Re: [alsa-devel] [PATCH 15/27] ALSA: hda - Use timecounter_initialize interface

[Sagar Arun Kamble]

On 1/5/2018 9:13 PM, Pierre-Louis Bossart wrote:

On 1/5/18 4:06 AM, Sagar Arun Kamble wrote:



On 1/3/2018 1:23 AM, Pierre-Louis Bossart wrote:

On 1/2/18 12:21 PM, Richard Cochran wrote:

On Tue, Jan 02, 2018 at 11:15:45AM -0600, Pierre-Louis Bossart wrote:
I wrote the code for HDaudio and I remember wasting time trying to 
figure
out the gory details of the cycle counter stuff when all I wanted 
was a
conversion from a 24MHz counter to ns values using a 125/3 
operation in the

right order - as explained in the comments


Would using clocks_calc_mult_shift() work for you?


In theory yes, but I'd need to re-check what the results would be.
I remember applying the 1/3 factor separately to avoid wrap-around 
after 4 hours [1], but I can't remember the details on the analysis. 
I can't figure out what the 'maxsec' argument should be either.


I am not sure if I understood the wrap-around correctly. Is 
AZX_REG_WALL_CLK 64bit or 32bit and in the comments which 20 bits are 
being referred.


it's a 32-bit counter.
off the top of my head, the idea was that the integer arithmetic 
should not degrade the precision (42ns) and that means you need to be 
careful with the fractional part, especially if the errors with the 
fractional part accumulate over time (I think this was the case when I 
looked several years ago). That's the main reason why I did the 
division by 3 last, after the read, so that the precision is not 
impacted by the interval between two reads.


timecounter interface ensures to add the fractional ns time so it has 
good precision. In the table below 3600 indicates interval of 1min 
between two counter reads
and 14400 indicates that interval of 4 minutes.  Shift factor controls 
the ns precision we want.
Considering duration of 195 years of interval between counter reads at 
24mhz we get mult=83 and shift=1 which will give cycle duration = 41.5.
What is the maximum duration of interval over which this read can be 
done in audio driver.
You also need to be careful with the multiplication factor otherwise 
you will exceed the 64-bit resolution. For example with the 14400 
factor, you cannot handle a counter larger than 2^64/14400, which 
gives you 14826 hours or 1.69 years. It's one of those 'nobody will 
ever need more than 640KB' value. The 125 factor gives you 195 years 
without saturating.




Keeping maxsec at lower value will ensure good precision but the mult 
factor derived then might lead to overflow if the interval of counter 
read is big.
Keeping maxsec high will reduce the mult factor and will marginally 
impact the precision (of the order of 6 decimal places of fraction 
nano second).


For 24mhz clock I am getting following scale factors at different 
maxsec values. I think these are as good as 125/3

125/3 gives scale factor of 41.67

maxsec, mult, shift, scale factor (mult/(2^shift))
0, 2796202667, 26, 41.6667163372039794921875
3600, 87381333, 21,   41.66507720947265625
14400, 21845333, 19, 41.660308837890625

I see sound driver uses only timecounter_read so conversions should 
be fine.
If there are usages of timecounter_cyc2time then we will have to take 
care of updating the timecounter often as
timecounter API internally counts time backwards if counter is spaced 
more than 1/2 the range.


Thanks
Sagar
[1] 
http://elixir.free-electrons.com/linux/latest/source/sound/hda/hdac_stream.c#L486 

Re: [alsa-devel] [PATCH 15/27] ALSA: hda - Use timecounter_initialize interface

[Sagar Arun Kamble]

On 1/5/2018 9:13 PM, Pierre-Louis Bossart wrote:

On 1/5/18 4:06 AM, Sagar Arun Kamble wrote:



On 1/3/2018 1:23 AM, Pierre-Louis Bossart wrote:

On 1/2/18 12:21 PM, Richard Cochran wrote:

On Tue, Jan 02, 2018 at 11:15:45AM -0600, Pierre-Louis Bossart wrote:
I wrote the code for HDaudio and I remember wasting time trying to 
figure
out the gory details of the cycle counter stuff when all I wanted 
was a
conversion from a 24MHz counter to ns values using a 125/3 
operation in the

right order - as explained in the comments


Would using clocks_calc_mult_shift() work for you?


In theory yes, but I'd need to re-check what the results would be.
I remember applying the 1/3 factor separately to avoid wrap-around 
after 4 hours [1], but I can't remember the details on the analysis. 
I can't figure out what the 'maxsec' argument should be either.


I am not sure if I understood the wrap-around correctly. Is 
AZX_REG_WALL_CLK 64bit or 32bit and in the comments which 20 bits are 
being referred.


it's a 32-bit counter.
off the top of my head, the idea was that the integer arithmetic 
should not degrade the precision (42ns) and that means you need to be 
careful with the fractional part, especially if the errors with the 
fractional part accumulate over time (I think this was the case when I 
looked several years ago). That's the main reason why I did the 
division by 3 last, after the read, so that the precision is not 
impacted by the interval between two reads.


timecounter interface ensures to add the fractional ns time so it has 
good precision. In the table below 3600 indicates interval of 1min 
between two counter reads
and 14400 indicates that interval of 4 minutes.  Shift factor controls 
the ns precision we want.
Considering duration of 195 years of interval between counter reads at 
24mhz we get mult=83 and shift=1 which will give cycle duration = 41.5.
What is the maximum duration of interval over which this read can be 
done in audio driver.
You also need to be careful with the multiplication factor otherwise 
you will exceed the 64-bit resolution. For example with the 14400 
factor, you cannot handle a counter larger than 2^64/14400, which 
gives you 14826 hours or 1.69 years. It's one of those 'nobody will 
ever need more than 640KB' value. The 125 factor gives you 195 years 
without saturating.




Keeping maxsec at lower value will ensure good precision but the mult 
factor derived then might lead to overflow if the interval of counter 
read is big.
Keeping maxsec high will reduce the mult factor and will marginally 
impact the precision (of the order of 6 decimal places of fraction 
nano second).


For 24mhz clock I am getting following scale factors at different 
maxsec values. I think these are as good as 125/3

125/3 gives scale factor of 41.67

maxsec, mult, shift, scale factor (mult/(2^shift))
0, 2796202667, 26, 41.6667163372039794921875
3600, 87381333, 21,   41.66507720947265625
14400, 21845333, 19, 41.660308837890625

I see sound driver uses only timecounter_read so conversions should 
be fine.
If there are usages of timecounter_cyc2time then we will have to take 
care of updating the timecounter often as
timecounter API internally counts time backwards if counter is spaced 
more than 1/2 the range.


Thanks
Sagar
[1] 
http://elixir.free-electrons.com/linux/latest/source/sound/hda/hdac_stream.c#L486 

Re: [Intel-wired-lan] [PATCH 01/27] timecounter: Make cyclecounter struct part of timecounter struct

[Sagar Arun Kamble]

Thanks Aaron.

Regards,
Sagar

On 1/9/2018 3:50 AM, Brown, Aaron F wrote:

From: Intel-wired-lan [mailto:intel-wired-lan-boun...@osuosl.org] On
Behalf Of Sagar Arun Kamble
Sent: Thursday, December 14, 2017 11:38 PM
To: linux-kernel@vger.kernel.org
Cc: alsa-de...@alsa-project.org; linux-r...@vger.kernel.org;
net...@vger.kernel.org; Richard Cochran <richardcoch...@gmail.com>;
Stephen Boyd <sb...@codeaurora.org>; Chris Wilson <chris@chris-
wilson.co.uk>; John Stultz <john.stu...@linaro.org>; intel-wired-
l...@lists.osuosl.org; Thomas Gleixner <t...@linutronix.de>; Kamble, Sagar A
<sagar.a.kam...@intel.com>; kvm...@lists.cs.columbia.edu; linux-arm-
ker...@lists.infradead.org
Subject: [Intel-wired-lan] [PATCH 01/27] timecounter: Make cyclecounter
struct part of timecounter struct

There is no real need for the users of timecounters to define cyclecounter
and timecounter variables separately. Since timecounter will always be
based on cyclecounter, have cyclecounter struct as member of timecounter
struct.

v2: Rebase.

Suggested-by: Chris Wilson <ch...@chris-wilson.co.uk>
Signed-off-by: Sagar Arun Kamble <sagar.a.kam...@intel.com>
Cc: Chris Wilson <ch...@chris-wilson.co.uk>
Cc: Richard Cochran <richardcoch...@gmail.com>
Cc: John Stultz <john.stu...@linaro.org>
Cc: Thomas Gleixner <t...@linutronix.de>
Cc: Stephen Boyd <sb...@codeaurora.org>
Cc: linux-kernel@vger.kernel.org
Cc: linux-arm-ker...@lists.infradead.org
Cc: net...@vger.kernel.org
Cc: intel-wired-...@lists.osuosl.org
Cc: linux-r...@vger.kernel.org
Cc: alsa-de...@alsa-project.org
Cc: kvm...@lists.cs.columbia.edu
Acked-by: Jeff Kirsher <jeffrey.t.kirs...@intel.com> (Intel drivers)
---
  arch/microblaze/kernel/timer.c | 20 ++--
  drivers/clocksource/arm_arch_timer.c   | 19 ++--
  drivers/net/ethernet/amd/xgbe/xgbe-dev.c   |  3 +-
  drivers/net/ethernet/amd/xgbe/xgbe-ptp.c   |  9 +++---
  drivers/net/ethernet/amd/xgbe/xgbe.h   |  1 -
  drivers/net/ethernet/broadcom/bnx2x/bnx2x.h|  1 -
  drivers/net/ethernet/broadcom/bnx2x/bnx2x_main.c   | 20 ++--
  drivers/net/ethernet/freescale/fec.h   |  1 -
  drivers/net/ethernet/freescale/fec_ptp.c   | 30 +-
  drivers/net/ethernet/intel/e1000e/e1000.h  |  1 -
  drivers/net/ethernet/intel/e1000e/netdev.c | 27 
  drivers/net/ethernet/intel/e1000e/ptp.c|  2 +-
  drivers/net/ethernet/intel/igb/igb.h   |  1 -
  drivers/net/ethernet/intel/igb/igb_ptp.c   | 25 ---
  drivers/net/ethernet/intel/ixgbe/ixgbe.h   |  1 -
  drivers/net/ethernet/intel/ixgbe/ixgbe_ptp.c   | 17 +-
  drivers/net/ethernet/mellanox/mlx4/en_clock.c  | 28 -
  drivers/net/ethernet/mellanox/mlx4/mlx4_en.h   |  1 -
  .../net/ethernet/mellanox/mlx5/core/lib/clock.c| 34 ++--
  drivers/net/ethernet/qlogic/qede/qede_ptp.c| 20 ++--
  drivers/net/ethernet/ti/cpts.c | 36 --
  drivers/net/ethernet/ti/cpts.h |  1 -
  include/linux/mlx5/driver.h|  1 -
  include/linux/timecounter.h|  4 +--
  include/sound/hdaudio.h|  1 -
  kernel/time/timecounter.c  | 28 -
  sound/hda/hdac_stream.c|  7 +++--
  virt/kvm/arm/arch_timer.c  |  6 ++--
  28 files changed, 163 insertions(+), 182 deletions(-)


For Intel e1000e and igb drivers:
Tested-by: Aaron Brown <aaron.f.br...@intel.com>

Re: [Intel-wired-lan] [PATCH 01/27] timecounter: Make cyclecounter struct part of timecounter struct

[Sagar Arun Kamble]

Thanks Aaron.

Regards,
Sagar

On 1/9/2018 3:50 AM, Brown, Aaron F wrote:

From: Intel-wired-lan [mailto:intel-wired-lan-boun...@osuosl.org] On
Behalf Of Sagar Arun Kamble
Sent: Thursday, December 14, 2017 11:38 PM
To: linux-kernel@vger.kernel.org
Cc: alsa-de...@alsa-project.org; linux-r...@vger.kernel.org;
net...@vger.kernel.org; Richard Cochran ;
Stephen Boyd ; Chris Wilson ; John Stultz ; intel-wired-
l...@lists.osuosl.org; Thomas Gleixner ; Kamble, Sagar A
; kvm...@lists.cs.columbia.edu; linux-arm-
ker...@lists.infradead.org
Subject: [Intel-wired-lan] [PATCH 01/27] timecounter: Make cyclecounter
struct part of timecounter struct

There is no real need for the users of timecounters to define cyclecounter
and timecounter variables separately. Since timecounter will always be
based on cyclecounter, have cyclecounter struct as member of timecounter
struct.

v2: Rebase.

Suggested-by: Chris Wilson 
Signed-off-by: Sagar Arun Kamble 
Cc: Chris Wilson 
Cc: Richard Cochran 
Cc: John Stultz 
Cc: Thomas Gleixner 
Cc: Stephen Boyd 
Cc: linux-kernel@vger.kernel.org
Cc: linux-arm-ker...@lists.infradead.org
Cc: net...@vger.kernel.org
Cc: intel-wired-...@lists.osuosl.org
Cc: linux-r...@vger.kernel.org
Cc: alsa-de...@alsa-project.org
Cc: kvm...@lists.cs.columbia.edu
Acked-by: Jeff Kirsher  (Intel drivers)
---
  arch/microblaze/kernel/timer.c | 20 ++--
  drivers/clocksource/arm_arch_timer.c   | 19 ++--
  drivers/net/ethernet/amd/xgbe/xgbe-dev.c   |  3 +-
  drivers/net/ethernet/amd/xgbe/xgbe-ptp.c   |  9 +++---
  drivers/net/ethernet/amd/xgbe/xgbe.h   |  1 -
  drivers/net/ethernet/broadcom/bnx2x/bnx2x.h|  1 -
  drivers/net/ethernet/broadcom/bnx2x/bnx2x_main.c   | 20 ++--
  drivers/net/ethernet/freescale/fec.h   |  1 -
  drivers/net/ethernet/freescale/fec_ptp.c   | 30 +-
  drivers/net/ethernet/intel/e1000e/e1000.h  |  1 -
  drivers/net/ethernet/intel/e1000e/netdev.c | 27 
  drivers/net/ethernet/intel/e1000e/ptp.c|  2 +-
  drivers/net/ethernet/intel/igb/igb.h   |  1 -
  drivers/net/ethernet/intel/igb/igb_ptp.c   | 25 ---
  drivers/net/ethernet/intel/ixgbe/ixgbe.h   |  1 -
  drivers/net/ethernet/intel/ixgbe/ixgbe_ptp.c   | 17 +-
  drivers/net/ethernet/mellanox/mlx4/en_clock.c  | 28 -
  drivers/net/ethernet/mellanox/mlx4/mlx4_en.h   |  1 -
  .../net/ethernet/mellanox/mlx5/core/lib/clock.c| 34 ++--
  drivers/net/ethernet/qlogic/qede/qede_ptp.c| 20 ++--
  drivers/net/ethernet/ti/cpts.c | 36 --
  drivers/net/ethernet/ti/cpts.h |  1 -
  include/linux/mlx5/driver.h|  1 -
  include/linux/timecounter.h|  4 +--
  include/sound/hdaudio.h|  1 -
  kernel/time/timecounter.c  | 28 -
  sound/hda/hdac_stream.c|  7 +++--
  virt/kvm/arm/arch_timer.c  |  6 ++--
  28 files changed, 163 insertions(+), 182 deletions(-)


For Intel e1000e and igb drivers:
Tested-by: Aaron Brown 

Re: [alsa-devel] [PATCH 15/27] ALSA: hda - Use timecounter_initialize interface

[Sagar Arun Kamble]

On 1/3/2018 1:23 AM, Pierre-Louis Bossart wrote:

On 1/2/18 12:21 PM, Richard Cochran wrote:

On Tue, Jan 02, 2018 at 11:15:45AM -0600, Pierre-Louis Bossart wrote:
I wrote the code for HDaudio and I remember wasting time trying to 
figure

out the gory details of the cycle counter stuff when all I wanted was a
conversion from a 24MHz counter to ns values using a 125/3 operation 
in the

right order - as explained in the comments


Would using clocks_calc_mult_shift() work for you?


In theory yes, but I'd need to re-check what the results would be.
I remember applying the 1/3 factor separately to avoid wrap-around 
after 4 hours [1], but I can't remember the details on the analysis. I 
can't figure out what the 'maxsec' argument should be either.


I am not sure if I understood the wrap-around correctly. Is 
AZX_REG_WALL_CLK 64bit or 32bit and in the comments which 20 bits are 
being referred.


Keeping maxsec at lower value will ensure good precision but the mult 
factor derived then might lead to overflow if the interval of counter 
read is big.
Keeping maxsec high will reduce the mult factor and will marginally 
impact the precision (of the order of 6 decimal places of fraction nano 
second).


For 24mhz clock I am getting following scale factors at different maxsec 
values. I think these are as good as 125/3

125/3 gives scale factor of 41.67

maxsec, mult, shift, scale factor (mult/(2^shift))
0, 2796202667, 26, 41.6667163372039794921875
3600, 87381333, 21,   41.66507720947265625
14400, 21845333, 19, 41.660308837890625

I see sound driver uses only timecounter_read so conversions should be fine.
If there are usages of timecounter_cyc2time then we will have to take 
care of updating the timecounter often as
timecounter API internally counts time backwards if counter is spaced 
more than 1/2 the range.


Thanks
Sagar
[1] 
http://elixir.free-electrons.com/linux/latest/source/sound/hda/hdac_stream.c#L486 

Re: [alsa-devel] [PATCH 15/27] ALSA: hda - Use timecounter_initialize interface

[Sagar Arun Kamble]

On 1/3/2018 1:23 AM, Pierre-Louis Bossart wrote:

On 1/2/18 12:21 PM, Richard Cochran wrote:

On Tue, Jan 02, 2018 at 11:15:45AM -0600, Pierre-Louis Bossart wrote:
I wrote the code for HDaudio and I remember wasting time trying to 
figure

out the gory details of the cycle counter stuff when all I wanted was a
conversion from a 24MHz counter to ns values using a 125/3 operation 
in the

right order - as explained in the comments


Would using clocks_calc_mult_shift() work for you?


In theory yes, but I'd need to re-check what the results would be.
I remember applying the 1/3 factor separately to avoid wrap-around 
after 4 hours [1], but I can't remember the details on the analysis. I 
can't figure out what the 'maxsec' argument should be either.


I am not sure if I understood the wrap-around correctly. Is 
AZX_REG_WALL_CLK 64bit or 32bit and in the comments which 20 bits are 
being referred.


Keeping maxsec at lower value will ensure good precision but the mult 
factor derived then might lead to overflow if the interval of counter 
read is big.
Keeping maxsec high will reduce the mult factor and will marginally 
impact the precision (of the order of 6 decimal places of fraction nano 
second).


For 24mhz clock I am getting following scale factors at different maxsec 
values. I think these are as good as 125/3

125/3 gives scale factor of 41.67

maxsec, mult, shift, scale factor (mult/(2^shift))
0, 2796202667, 26, 41.6667163372039794921875
3600, 87381333, 21,   41.66507720947265625
14400, 21845333, 19, 41.660308837890625

I see sound driver uses only timecounter_read so conversions should be fine.
If there are usages of timecounter_cyc2time then we will have to take 
care of updating the timecounter often as
timecounter API internally counts time backwards if counter is spaced 
more than 1/2 the range.


Thanks
Sagar
[1] 
http://elixir.free-electrons.com/linux/latest/source/sound/hda/hdac_stream.c#L486 

Re: [alsa-devel] [PATCH 15/27] ALSA: hda - Use timecounter_initialize interface

[Sagar Arun Kamble]

On 12/28/2017 10:19 PM, Richard Cochran wrote:

On Tue, Dec 26, 2017 at 01:07:35PM +0530, Sagar Arun Kamble wrote:

Or can we provide simpler versions for covering some defaults?  At
least reducing the number of arguments would make things easier.

Thought about specifying 1. cyclecounter read func 2. frequency 3. width of
counter as parameters here
which can get rid of mult, shift params. But this is not easy as most of the
drivers do not specify
cyclecounter frequency and instead hard-code the mult/shift factors.

You are talking about using clocks_calc_mult_shift() here, right? (See
the usage example in drivers/net/ethernet/ti/cpts.c).

Yes

This is a good idea, and it is worth getting the driver authors' input
to figure out the correct parameters.

I bet we can use that almost everywhere.  If there are any drivers
that cannot be converted, then we can leave some sort of low level
legacy initialization method.

Agree

Thanks,
Richard

Re: [alsa-devel] [PATCH 15/27] ALSA: hda - Use timecounter_initialize interface

[Sagar Arun Kamble]

On 12/28/2017 10:19 PM, Richard Cochran wrote:

On Tue, Dec 26, 2017 at 01:07:35PM +0530, Sagar Arun Kamble wrote:

Or can we provide simpler versions for covering some defaults?  At
least reducing the number of arguments would make things easier.

Thought about specifying 1. cyclecounter read func 2. frequency 3. width of
counter as parameters here
which can get rid of mult, shift params. But this is not easy as most of the
drivers do not specify
cyclecounter frequency and instead hard-code the mult/shift factors.

You are talking about using clocks_calc_mult_shift() here, right? (See
the usage example in drivers/net/ethernet/ti/cpts.c).

Yes

This is a good idea, and it is worth getting the driver authors' input
to figure out the correct parameters.

I bet we can use that almost everywhere.  If there are any drivers
that cannot be converted, then we can leave some sort of low level
legacy initialization method.

Agree

Thanks,
Richard

Re: [alsa-devel] [PATCH 15/27] ALSA: hda - Use timecounter_initialize interface

[Sagar Arun Kamble]

On 12/15/2017 10:40 PM, Takashi Iwai wrote:

On Fri, 15 Dec 2017 17:51:25 +0100,
Richard Cochran wrote:

On Fri, Dec 15, 2017 at 12:10:47PM +0100, Takashi Iwai wrote:


-   struct cyclecounter *cc = _dev->tc.cc;
-   cc->read = azx_cc_read;
-   cc->mask = CLOCKSOURCE_MASK(32);
-   cc->mult = 125; /* saturation after 195 years */
-   cc->shift = 0;

I want to get away from this mess of open coded structure
initialization and use a proper functional interface instead.

I agree that a proper functional interface would be better, too.
But not a form like foo(501, 21, 10, 499, 5678).
In C syntax, you may more easily pass a wrong value than open codes.


nsec = 0; /* audio time is elapsed time since trigger */
-   timecounter_init(tc, nsec);
+   timecounter_initialize(tc,
+  azx_cc_read,
+  CLOCKSOURCE_MASK(32),
+  125, /* saturation after 195 years */
+  0,
+  nsec);

Hmm, a function with so many arguments is difficult to remember and is
often error-prone.  By this transition, it becomes harder to read
through.

Please suggest a better way.

I have no good idea ATM, sorry.

Or can we provide simpler versions for covering some defaults?  At
least reducing the number of arguments would make things easier.
Thought about specifying 1. cyclecounter read func 2. frequency 3. width 
of counter as parameters here
which can get rid of mult, shift params. But this is not easy as most of 
the drivers do not specify

cyclecounter frequency and instead hard-code the mult/shift factors.
How about passing initialized cyclecounter struct?


Takashi

Re: [alsa-devel] [PATCH 15/27] ALSA: hda - Use timecounter_initialize interface

[Sagar Arun Kamble]

On 12/15/2017 10:40 PM, Takashi Iwai wrote:

On Fri, 15 Dec 2017 17:51:25 +0100,
Richard Cochran wrote:

On Fri, Dec 15, 2017 at 12:10:47PM +0100, Takashi Iwai wrote:


-   struct cyclecounter *cc = _dev->tc.cc;
-   cc->read = azx_cc_read;
-   cc->mask = CLOCKSOURCE_MASK(32);
-   cc->mult = 125; /* saturation after 195 years */
-   cc->shift = 0;

I want to get away from this mess of open coded structure
initialization and use a proper functional interface instead.

I agree that a proper functional interface would be better, too.
But not a form like foo(501, 21, 10, 499, 5678).
In C syntax, you may more easily pass a wrong value than open codes.


nsec = 0; /* audio time is elapsed time since trigger */
-   timecounter_init(tc, nsec);
+   timecounter_initialize(tc,
+  azx_cc_read,
+  CLOCKSOURCE_MASK(32),
+  125, /* saturation after 195 years */
+  0,
+  nsec);

Hmm, a function with so many arguments is difficult to remember and is
often error-prone.  By this transition, it becomes harder to read
through.

Please suggest a better way.

I have no good idea ATM, sorry.

Or can we provide simpler versions for covering some defaults?  At
least reducing the number of arguments would make things easier.
Thought about specifying 1. cyclecounter read func 2. frequency 3. width 
of counter as parameters here
which can get rid of mult, shift params. But this is not easy as most of 
the drivers do not specify

cyclecounter frequency and instead hard-code the mult/shift factors.
How about passing initialized cyclecounter struct?


Takashi

[PATCH 04/27] clocksource/arm_arch_timer: Use timecounter_initialize interface

[Sagar Arun Kamble]

With new interface timecounter_initialize we can initialize timecounter
fields and underlying cyclecounter together. Update arch_timer_kvm_info
timecounter init with this new function.

Signed-off-by: Sagar Arun Kamble <sagar.a.kam...@intel.com>
Cc: Richard Cochran <richardcoch...@gmail.com>
Cc: Mark Rutland <mark.rutl...@arm.com>
Cc: Marc Zyngier <marc.zyng...@arm.com>
Cc: Daniel Lezcano <daniel.lezc...@linaro.org>
Cc: Thomas Gleixner <t...@linutronix.de>
Cc: linux-arm-ker...@lists.infradead.org
Cc: linux-kernel@vger.kernel.org
---
 drivers/clocksource/arm_arch_timer.c | 15 +++
 1 file changed, 7 insertions(+), 8 deletions(-)

diff --git a/drivers/clocksource/arm_arch_timer.c 
b/drivers/clocksource/arm_arch_timer.c
index 31543e5..d96393f 100644
--- a/drivers/clocksource/arm_arch_timer.c
+++ b/drivers/clocksource/arm_arch_timer.c
@@ -910,10 +910,7 @@ static u64 arch_counter_get_cntvct_mem(void)
return ((u64) vct_hi << 32) | vct_lo;
 }
 
-static struct arch_timer_kvm_info arch_timer_kvm_info = {
-   .timecounter.cc.read = arch_counter_read_cc,
-   .timecounter.cc.mask = CLOCKSOURCE_MASK(56),
-};
+static struct arch_timer_kvm_info arch_timer_kvm_info;
 
 struct arch_timer_kvm_info *arch_timer_get_kvm_info(void)
 {
@@ -923,7 +920,6 @@ struct arch_timer_kvm_info *arch_timer_get_kvm_info(void)
 static void __init arch_counter_register(unsigned type)
 {
u64 start_count;
-   struct cyclecounter *cc = _timer_kvm_info.timecounter.cc;
 
/* Register the CP15 based counter if we have one */
if (type & ARCH_TIMER_TYPE_CP15) {
@@ -943,9 +939,12 @@ static void __init arch_counter_register(unsigned type)
start_count = arch_timer_read_counter();
clocksource_register_hz(_counter, arch_timer_rate);
 
-   cc->mult = clocksource_counter.mult;
-   cc->shift = clocksource_counter.shift;
-   timecounter_init(_timer_kvm_info.timecounter, start_count);
+   timecounter_initialize(_timer_kvm_info.timecounter,
+  arch_counter_read_cc,
+  CLOCKSOURCE_MASK(56),
+  clocksource_counter.mult,
+  clocksource_counter.shift,
+  start_count);
 
/* 56 bits minimum, so we assume worst case rollover */
sched_clock_register(arch_timer_read_counter, 56, arch_timer_rate);
-- 
1.9.1

[PATCH 04/27] clocksource/arm_arch_timer: Use timecounter_initialize interface

[Sagar Arun Kamble]

With new interface timecounter_initialize we can initialize timecounter
fields and underlying cyclecounter together. Update arch_timer_kvm_info
timecounter init with this new function.

Signed-off-by: Sagar Arun Kamble 
Cc: Richard Cochran 
Cc: Mark Rutland 
Cc: Marc Zyngier 
Cc: Daniel Lezcano 
Cc: Thomas Gleixner 
Cc: linux-arm-ker...@lists.infradead.org
Cc: linux-kernel@vger.kernel.org
---
 drivers/clocksource/arm_arch_timer.c | 15 +++
 1 file changed, 7 insertions(+), 8 deletions(-)

diff --git a/drivers/clocksource/arm_arch_timer.c 
b/drivers/clocksource/arm_arch_timer.c
index 31543e5..d96393f 100644
--- a/drivers/clocksource/arm_arch_timer.c
+++ b/drivers/clocksource/arm_arch_timer.c
@@ -910,10 +910,7 @@ static u64 arch_counter_get_cntvct_mem(void)
return ((u64) vct_hi << 32) | vct_lo;
 }
 
-static struct arch_timer_kvm_info arch_timer_kvm_info = {
-   .timecounter.cc.read = arch_counter_read_cc,
-   .timecounter.cc.mask = CLOCKSOURCE_MASK(56),
-};
+static struct arch_timer_kvm_info arch_timer_kvm_info;
 
 struct arch_timer_kvm_info *arch_timer_get_kvm_info(void)
 {
@@ -923,7 +920,6 @@ struct arch_timer_kvm_info *arch_timer_get_kvm_info(void)
 static void __init arch_counter_register(unsigned type)
 {
u64 start_count;
-   struct cyclecounter *cc = _timer_kvm_info.timecounter.cc;
 
/* Register the CP15 based counter if we have one */
if (type & ARCH_TIMER_TYPE_CP15) {
@@ -943,9 +939,12 @@ static void __init arch_counter_register(unsigned type)
start_count = arch_timer_read_counter();
clocksource_register_hz(_counter, arch_timer_rate);
 
-   cc->mult = clocksource_counter.mult;
-   cc->shift = clocksource_counter.shift;
-   timecounter_init(_timer_kvm_info.timecounter, start_count);
+   timecounter_initialize(_timer_kvm_info.timecounter,
+  arch_counter_read_cc,
+  CLOCKSOURCE_MASK(56),
+  clocksource_counter.mult,
+  clocksource_counter.shift,
+  start_count);
 
/* 56 bits minimum, so we assume worst case rollover */
sched_clock_register(arch_timer_read_counter, 56, arch_timer_rate);
-- 
1.9.1

[PATCH 05/27] amd-xgbe: Use timecounter_initialize interface

[Sagar Arun Kamble]

With new interface timecounter_initialize we can initialize timecounter
fields and underlying cyclecounter together. Update xgbe tstamp_tc
timecounter init with this new function.

Signed-off-by: Sagar Arun Kamble <sagar.a.kam...@intel.com>
Cc: Richard Cochran <richardcoch...@gmail.com>
Cc: Tom Lendacky <thomas.lenda...@amd.com>
Cc: net...@vger.kernel.org
Cc: linux-kernel@vger.kernel.org
---
 drivers/net/ethernet/amd/xgbe/xgbe-ptp.c | 13 ++---
 1 file changed, 6 insertions(+), 7 deletions(-)

diff --git a/drivers/net/ethernet/amd/xgbe/xgbe-ptp.c 
b/drivers/net/ethernet/amd/xgbe/xgbe-ptp.c
index 5ea4edf..486437b 100644
--- a/drivers/net/ethernet/amd/xgbe/xgbe-ptp.c
+++ b/drivers/net/ethernet/amd/xgbe/xgbe-ptp.c
@@ -228,7 +228,6 @@ void xgbe_ptp_register(struct xgbe_prv_data *pdata)
 {
struct ptp_clock_info *info = >ptp_clock_info;
struct ptp_clock *clock;
-   struct cyclecounter *cc = >tstamp_tc.cc;
u64 dividend;
 
snprintf(info->name, sizeof(info->name), "%s",
@@ -258,12 +257,12 @@ void xgbe_ptp_register(struct xgbe_prv_data *pdata)
pdata->tstamp_addend = div_u64(dividend, pdata->ptpclk_rate);
 
/* Setup the timecounter */
-   cc->read = xgbe_cc_read;
-   cc->mask = CLOCKSOURCE_MASK(64);
-   cc->mult = 1;
-   cc->shift = 0;
-
-   timecounter_init(>tstamp_tc, ktime_to_ns(ktime_get_real()));
+   timecounter_initialize(>tstamp_tc,
+  xgbe_cc_read,
+  CLOCKSOURCE_MASK(64),
+  1,
+  0,
+  ktime_to_ns(ktime_get_real()));
 
/* Disable all timestamping to start */
XGMAC_IOWRITE(pdata, MAC_TSCR, 0);
-- 
1.9.1

[PATCH 05/27] amd-xgbe: Use timecounter_initialize interface

[Sagar Arun Kamble]

With new interface timecounter_initialize we can initialize timecounter
fields and underlying cyclecounter together. Update xgbe tstamp_tc
timecounter init with this new function.

Signed-off-by: Sagar Arun Kamble 
Cc: Richard Cochran 
Cc: Tom Lendacky 
Cc: net...@vger.kernel.org
Cc: linux-kernel@vger.kernel.org
---
 drivers/net/ethernet/amd/xgbe/xgbe-ptp.c | 13 ++---
 1 file changed, 6 insertions(+), 7 deletions(-)

diff --git a/drivers/net/ethernet/amd/xgbe/xgbe-ptp.c 
b/drivers/net/ethernet/amd/xgbe/xgbe-ptp.c
index 5ea4edf..486437b 100644
--- a/drivers/net/ethernet/amd/xgbe/xgbe-ptp.c
+++ b/drivers/net/ethernet/amd/xgbe/xgbe-ptp.c
@@ -228,7 +228,6 @@ void xgbe_ptp_register(struct xgbe_prv_data *pdata)
 {
struct ptp_clock_info *info = >ptp_clock_info;
struct ptp_clock *clock;
-   struct cyclecounter *cc = >tstamp_tc.cc;
u64 dividend;
 
snprintf(info->name, sizeof(info->name), "%s",
@@ -258,12 +257,12 @@ void xgbe_ptp_register(struct xgbe_prv_data *pdata)
pdata->tstamp_addend = div_u64(dividend, pdata->ptpclk_rate);
 
/* Setup the timecounter */
-   cc->read = xgbe_cc_read;
-   cc->mask = CLOCKSOURCE_MASK(64);
-   cc->mult = 1;
-   cc->shift = 0;
-
-   timecounter_init(>tstamp_tc, ktime_to_ns(ktime_get_real()));
+   timecounter_initialize(>tstamp_tc,
+  xgbe_cc_read,
+  CLOCKSOURCE_MASK(64),
+  1,
+  0,
+  ktime_to_ns(ktime_get_real()));
 
/* Disable all timestamping to start */
XGMAC_IOWRITE(pdata, MAC_TSCR, 0);
-- 
1.9.1

[PATCH 07/27] fec: Use timecounter_initialize interface

[Sagar Arun Kamble]

With new interface timecounter_initialize we can initialize timecounter
fields and underlying cyclecounter together. Update fec ethernet
timecounter init with this new function.

Signed-off-by: Sagar Arun Kamble <sagar.a.kam...@intel.com>
Cc: Richard Cochran <richardcoch...@gmail.com>
Cc: Fugang Duan <fugang.d...@nxp.com>
Cc: net...@vger.kernel.org
Cc: linux-kernel@vger.kernel.org
---
 drivers/net/ethernet/freescale/fec_ptp.c | 15 +++
 1 file changed, 7 insertions(+), 8 deletions(-)

diff --git a/drivers/net/ethernet/freescale/fec_ptp.c 
b/drivers/net/ethernet/freescale/fec_ptp.c
index b1261d1..1ba7216 100644
--- a/drivers/net/ethernet/freescale/fec_ptp.c
+++ b/drivers/net/ethernet/freescale/fec_ptp.c
@@ -278,14 +278,13 @@ void fec_ptp_start_cyclecounter(struct net_device *ndev)
writel(FEC_T_CTRL_ENABLE | FEC_T_CTRL_PERIOD_RST,
fep->hwp + FEC_ATIME_CTRL);
 
-   memset(>tc.cc, 0, sizeof(fep->tc.cc));
-   fep->tc.cc.read = fec_ptp_read;
-   fep->tc.cc.mask = CLOCKSOURCE_MASK(31);
-   fep->tc.cc.shift = 31;
-   fep->tc.cc.mult = FEC_CC_MULT;
-
-   /* reset the ns time counter */
-   timecounter_init(>tc, ktime_to_ns(ktime_get_real()));
+   /* Initialize the ns time counter */
+   timecounter_initialize(>tc,
+  fec_ptp_read,
+  CLOCKSOURCE_MASK(31),
+  FEC_CC_MULT,
+  31,
+  ktime_to_ns(ktime_get_real()));
 
spin_unlock_irqrestore(>tmreg_lock, flags);
 }
-- 
1.9.1

[PATCH 07/27] fec: Use timecounter_initialize interface

[Sagar Arun Kamble]

With new interface timecounter_initialize we can initialize timecounter
fields and underlying cyclecounter together. Update fec ethernet
timecounter init with this new function.

Signed-off-by: Sagar Arun Kamble 
Cc: Richard Cochran 
Cc: Fugang Duan 
Cc: net...@vger.kernel.org
Cc: linux-kernel@vger.kernel.org
---
 drivers/net/ethernet/freescale/fec_ptp.c | 15 +++
 1 file changed, 7 insertions(+), 8 deletions(-)

diff --git a/drivers/net/ethernet/freescale/fec_ptp.c 
b/drivers/net/ethernet/freescale/fec_ptp.c
index b1261d1..1ba7216 100644
--- a/drivers/net/ethernet/freescale/fec_ptp.c
+++ b/drivers/net/ethernet/freescale/fec_ptp.c
@@ -278,14 +278,13 @@ void fec_ptp_start_cyclecounter(struct net_device *ndev)
writel(FEC_T_CTRL_ENABLE | FEC_T_CTRL_PERIOD_RST,
fep->hwp + FEC_ATIME_CTRL);
 
-   memset(>tc.cc, 0, sizeof(fep->tc.cc));
-   fep->tc.cc.read = fec_ptp_read;
-   fep->tc.cc.mask = CLOCKSOURCE_MASK(31);
-   fep->tc.cc.shift = 31;
-   fep->tc.cc.mult = FEC_CC_MULT;
-
-   /* reset the ns time counter */
-   timecounter_init(>tc, ktime_to_ns(ktime_get_real()));
+   /* Initialize the ns time counter */
+   timecounter_initialize(>tc,
+  fec_ptp_read,
+  CLOCKSOURCE_MASK(31),
+  FEC_CC_MULT,
+  31,
+  ktime_to_ns(ktime_get_real()));
 
spin_unlock_irqrestore(>tmreg_lock, flags);
 }
-- 
1.9.1

[PATCH 08/27] e1000e: Use timecounter_initialize interface

[Sagar Arun Kamble]

With new interface timecounter_initialize we can initialize timecounter
fields and underlying cyclecounter together. Update e1000e timecounter
init with this new function.

Signed-off-by: Sagar Arun Kamble <sagar.a.kam...@intel.com>
Cc: Richard Cochran <richardcoch...@gmail.com>
Cc: Jeff Kirsher <jeffrey.t.kirs...@intel.com>
Cc: intel-wired-...@lists.osuosl.org
Cc: net...@vger.kernel.org
Cc: linux-kernel@vger.kernel.org
---
 drivers/net/ethernet/intel/e1000e/e1000.h  |  4 
 drivers/net/ethernet/intel/e1000e/netdev.c | 31 +-
 2 files changed, 22 insertions(+), 13 deletions(-)

diff --git a/drivers/net/ethernet/intel/e1000e/e1000.h 
b/drivers/net/ethernet/intel/e1000e/e1000.h
index b59f82a..d6b0e59 100644
--- a/drivers/net/ethernet/intel/e1000e/e1000.h
+++ b/drivers/net/ethernet/intel/e1000e/e1000.h
@@ -341,6 +341,10 @@ struct e1000_adapter {
struct work_struct tx_hwtstamp_work;
spinlock_t systim_lock; /* protects SYSTIML/H regsters */
struct timecounter tc;
+   u64 (*cc_read)(const struct cyclecounter *cc);
+   u64 cc_mask;
+   u32 cc_mult;
+   u32 cc_shift;
struct ptp_clock *ptp_clock;
struct ptp_clock_info ptp_clock_info;
struct pm_qos_request pm_qos_req;
diff --git a/drivers/net/ethernet/intel/e1000e/netdev.c 
b/drivers/net/ethernet/intel/e1000e/netdev.c
index c9f7ba3..6be1327 100644
--- a/drivers/net/ethernet/intel/e1000e/netdev.c
+++ b/drivers/net/ethernet/intel/e1000e/netdev.c
@@ -3536,7 +3536,7 @@ s32 e1000e_get_base_timinca(struct e1000_adapter 
*adapter, u32 *timinca)
incperiod = INCPERIOD_96MHZ;
incvalue = INCVALUE_96MHZ;
shift = INCVALUE_SHIFT_96MHZ;
-   adapter->tc.cc.shift = shift + INCPERIOD_SHIFT_96MHZ;
+   adapter->cc_shift = shift + INCPERIOD_SHIFT_96MHZ;
break;
case e1000_pch_lpt:
if (er32(TSYNCRXCTL) & E1000_TSYNCRXCTL_SYSCFI) {
@@ -3544,13 +3544,13 @@ s32 e1000e_get_base_timinca(struct e1000_adapter 
*adapter, u32 *timinca)
incperiod = INCPERIOD_96MHZ;
incvalue = INCVALUE_96MHZ;
shift = INCVALUE_SHIFT_96MHZ;
-   adapter->tc.cc.shift = shift + INCPERIOD_SHIFT_96MHZ;
+   adapter->cc_shift = shift + INCPERIOD_SHIFT_96MHZ;
} else {
/* Stable 25MHz frequency */
incperiod = INCPERIOD_25MHZ;
incvalue = INCVALUE_25MHZ;
shift = INCVALUE_SHIFT_25MHZ;
-   adapter->tc.cc.shift = shift;
+   adapter->cc_shift = shift;
}
break;
case e1000_pch_spt:
@@ -3559,7 +3559,7 @@ s32 e1000e_get_base_timinca(struct e1000_adapter 
*adapter, u32 *timinca)
incperiod = INCPERIOD_24MHZ;
incvalue = INCVALUE_24MHZ;
shift = INCVALUE_SHIFT_24MHZ;
-   adapter->tc.cc.shift = shift;
+   adapter->cc_shift = shift;
break;
}
return -EINVAL;
@@ -3569,13 +3569,13 @@ s32 e1000e_get_base_timinca(struct e1000_adapter 
*adapter, u32 *timinca)
incperiod = INCPERIOD_24MHZ;
incvalue = INCVALUE_24MHZ;
shift = INCVALUE_SHIFT_24MHZ;
-   adapter->tc.cc.shift = shift;
+   adapter->cc_shift = shift;
} else {
/* Stable 38400KHz frequency */
incperiod = INCPERIOD_38400KHZ;
incvalue = INCVALUE_38400KHZ;
shift = INCVALUE_SHIFT_38400KHZ;
-   adapter->tc.cc.shift = shift;
+   adapter->cc_shift = shift;
}
break;
case e1000_82574:
@@ -3584,7 +3584,7 @@ s32 e1000e_get_base_timinca(struct e1000_adapter 
*adapter, u32 *timinca)
incperiod = INCPERIOD_25MHZ;
incvalue = INCVALUE_25MHZ;
shift = INCVALUE_SHIFT_25MHZ;
-   adapter->tc.cc.shift = shift;
+   adapter->cc_shift = shift;
break;
default:
return -EINVAL;
@@ -3953,9 +3953,14 @@ static void e1000e_systim_reset(struct e1000_adapter 
*adapter)
return;
}
 
-   /* reset the systim ns time counter */
+   /* reinitialize the systim ns time counter */
spin_lock_irqsave(>systim_lock, flags);
-   timecounter_init(>tc, ktime_to_ns(ktime_get_real()));
+   timecounter_initialize(>tc,
+  adapter->cc_read,
+  adapter->cc_mask,
+ 

[PATCH 08/27] e1000e: Use timecounter_initialize interface

[Sagar Arun Kamble]

With new interface timecounter_initialize we can initialize timecounter
fields and underlying cyclecounter together. Update e1000e timecounter
init with this new function.

Signed-off-by: Sagar Arun Kamble 
Cc: Richard Cochran 
Cc: Jeff Kirsher 
Cc: intel-wired-...@lists.osuosl.org
Cc: net...@vger.kernel.org
Cc: linux-kernel@vger.kernel.org
---
 drivers/net/ethernet/intel/e1000e/e1000.h  |  4 
 drivers/net/ethernet/intel/e1000e/netdev.c | 31 +-
 2 files changed, 22 insertions(+), 13 deletions(-)

diff --git a/drivers/net/ethernet/intel/e1000e/e1000.h 
b/drivers/net/ethernet/intel/e1000e/e1000.h
index b59f82a..d6b0e59 100644
--- a/drivers/net/ethernet/intel/e1000e/e1000.h
+++ b/drivers/net/ethernet/intel/e1000e/e1000.h
@@ -341,6 +341,10 @@ struct e1000_adapter {
struct work_struct tx_hwtstamp_work;
spinlock_t systim_lock; /* protects SYSTIML/H regsters */
struct timecounter tc;
+   u64 (*cc_read)(const struct cyclecounter *cc);
+   u64 cc_mask;
+   u32 cc_mult;
+   u32 cc_shift;
struct ptp_clock *ptp_clock;
struct ptp_clock_info ptp_clock_info;
struct pm_qos_request pm_qos_req;
diff --git a/drivers/net/ethernet/intel/e1000e/netdev.c 
b/drivers/net/ethernet/intel/e1000e/netdev.c
index c9f7ba3..6be1327 100644
--- a/drivers/net/ethernet/intel/e1000e/netdev.c
+++ b/drivers/net/ethernet/intel/e1000e/netdev.c
@@ -3536,7 +3536,7 @@ s32 e1000e_get_base_timinca(struct e1000_adapter 
*adapter, u32 *timinca)
incperiod = INCPERIOD_96MHZ;
incvalue = INCVALUE_96MHZ;
shift = INCVALUE_SHIFT_96MHZ;
-   adapter->tc.cc.shift = shift + INCPERIOD_SHIFT_96MHZ;
+   adapter->cc_shift = shift + INCPERIOD_SHIFT_96MHZ;
break;
case e1000_pch_lpt:
if (er32(TSYNCRXCTL) & E1000_TSYNCRXCTL_SYSCFI) {
@@ -3544,13 +3544,13 @@ s32 e1000e_get_base_timinca(struct e1000_adapter 
*adapter, u32 *timinca)
incperiod = INCPERIOD_96MHZ;
incvalue = INCVALUE_96MHZ;
shift = INCVALUE_SHIFT_96MHZ;
-   adapter->tc.cc.shift = shift + INCPERIOD_SHIFT_96MHZ;
+   adapter->cc_shift = shift + INCPERIOD_SHIFT_96MHZ;
} else {
/* Stable 25MHz frequency */
incperiod = INCPERIOD_25MHZ;
incvalue = INCVALUE_25MHZ;
shift = INCVALUE_SHIFT_25MHZ;
-   adapter->tc.cc.shift = shift;
+   adapter->cc_shift = shift;
}
break;
case e1000_pch_spt:
@@ -3559,7 +3559,7 @@ s32 e1000e_get_base_timinca(struct e1000_adapter 
*adapter, u32 *timinca)
incperiod = INCPERIOD_24MHZ;
incvalue = INCVALUE_24MHZ;
shift = INCVALUE_SHIFT_24MHZ;
-   adapter->tc.cc.shift = shift;
+   adapter->cc_shift = shift;
break;
}
return -EINVAL;
@@ -3569,13 +3569,13 @@ s32 e1000e_get_base_timinca(struct e1000_adapter 
*adapter, u32 *timinca)
incperiod = INCPERIOD_24MHZ;
incvalue = INCVALUE_24MHZ;
shift = INCVALUE_SHIFT_24MHZ;
-   adapter->tc.cc.shift = shift;
+   adapter->cc_shift = shift;
} else {
/* Stable 38400KHz frequency */
incperiod = INCPERIOD_38400KHZ;
incvalue = INCVALUE_38400KHZ;
shift = INCVALUE_SHIFT_38400KHZ;
-   adapter->tc.cc.shift = shift;
+   adapter->cc_shift = shift;
}
break;
case e1000_82574:
@@ -3584,7 +3584,7 @@ s32 e1000e_get_base_timinca(struct e1000_adapter 
*adapter, u32 *timinca)
incperiod = INCPERIOD_25MHZ;
incvalue = INCVALUE_25MHZ;
shift = INCVALUE_SHIFT_25MHZ;
-   adapter->tc.cc.shift = shift;
+   adapter->cc_shift = shift;
break;
default:
return -EINVAL;
@@ -3953,9 +3953,14 @@ static void e1000e_systim_reset(struct e1000_adapter 
*adapter)
return;
}
 
-   /* reset the systim ns time counter */
+   /* reinitialize the systim ns time counter */
spin_lock_irqsave(>systim_lock, flags);
-   timecounter_init(>tc, ktime_to_ns(ktime_get_real()));
+   timecounter_initialize(>tc,
+  adapter->cc_read,
+  adapter->cc_mask,
+  adapter->cc_mult,
+  adapter->cc_shift,
+ 

[PATCH 03/27] microblaze: Use timecounter_initialize interface

[Sagar Arun Kamble]

With new interface timecounter_initialize we can initialize timecounter
fields and underlying cyclecounter together. Update xilinx_tc init with
this new function.

Signed-off-by: Sagar Arun Kamble <sagar.a.kam...@intel.com>
Cc: Richard Cochran <richardcoch...@gmail.com>
Cc: Michal Simek <mon...@monstr.eu>
Cc: Daniel Lezcano <daniel.lezc...@linaro.org>
Cc: Neil Armstrong <narmstr...@baylibre.com>
Cc: Nicolai Stange <nicsta...@gmail.com>
Cc: Arnd Bergmann <a...@arndb.de>
Cc: Rob Herring <r...@kernel.org>
Cc: Thomas Gleixner <t...@linutronix.de>
Cc: linux-kernel@vger.kernel.org
---
 arch/microblaze/kernel/timer.c | 19 ++-
 1 file changed, 10 insertions(+), 9 deletions(-)

diff --git a/arch/microblaze/kernel/timer.c b/arch/microblaze/kernel/timer.c
index b7f89e9..b32a896 100644
--- a/arch/microblaze/kernel/timer.c
+++ b/arch/microblaze/kernel/timer.c
@@ -204,20 +204,21 @@ static u64 xilinx_cc_read(const struct cyclecounter *cc)
return xilinx_read(NULL);
 }
 
-static struct timecounter xilinx_tc = {
-   .cc.read = xilinx_cc_read,
-   .cc.mask = CLOCKSOURCE_MASK(32),
-   .cc.mult = 0,
-   .cc.shift = 8,
-};
+static struct timecounter xilinx_tc;
 
 static int __init init_xilinx_timecounter(void)
 {
-   struct cyclecounter *cc = _tc.cc;
+   u32 mult = 0;
+   u32 shift = 8;
 
-   cc->mult = div_sc(timer_clock_freq, NSEC_PER_SEC, cc->shift);
+   mult = div_sc(timer_clock_freq, NSEC_PER_SEC, shift);
 
-   timecounter_init(_tc, sched_clock());
+   timecounter_initialize(_tc,
+  xilinx_cc_read,
+  CLOCKSOURCE_MASK(32),
+  mult,
+  shift,
+  sched_clock());
 
return 0;
 }
-- 
1.9.1

[PATCH 03/27] microblaze: Use timecounter_initialize interface

[Sagar Arun Kamble]

With new interface timecounter_initialize we can initialize timecounter
fields and underlying cyclecounter together. Update xilinx_tc init with
this new function.

Signed-off-by: Sagar Arun Kamble 
Cc: Richard Cochran 
Cc: Michal Simek 
Cc: Daniel Lezcano 
Cc: Neil Armstrong 
Cc: Nicolai Stange 
Cc: Arnd Bergmann 
Cc: Rob Herring 
Cc: Thomas Gleixner 
Cc: linux-kernel@vger.kernel.org
---
 arch/microblaze/kernel/timer.c | 19 ++-
 1 file changed, 10 insertions(+), 9 deletions(-)

diff --git a/arch/microblaze/kernel/timer.c b/arch/microblaze/kernel/timer.c
index b7f89e9..b32a896 100644
--- a/arch/microblaze/kernel/timer.c
+++ b/arch/microblaze/kernel/timer.c
@@ -204,20 +204,21 @@ static u64 xilinx_cc_read(const struct cyclecounter *cc)
return xilinx_read(NULL);
 }
 
-static struct timecounter xilinx_tc = {
-   .cc.read = xilinx_cc_read,
-   .cc.mask = CLOCKSOURCE_MASK(32),
-   .cc.mult = 0,
-   .cc.shift = 8,
-};
+static struct timecounter xilinx_tc;
 
 static int __init init_xilinx_timecounter(void)
 {
-   struct cyclecounter *cc = _tc.cc;
+   u32 mult = 0;
+   u32 shift = 8;
 
-   cc->mult = div_sc(timer_clock_freq, NSEC_PER_SEC, cc->shift);
+   mult = div_sc(timer_clock_freq, NSEC_PER_SEC, shift);
 
-   timecounter_init(_tc, sched_clock());
+   timecounter_initialize(_tc,
+  xilinx_cc_read,
+  CLOCKSOURCE_MASK(32),
+  mult,
+  shift,
+  sched_clock());
 
return 0;
 }
-- 
1.9.1

[PATCH 11/27] net/mlx4: Use timecounter_initialize interface

[Sagar Arun Kamble]

With new interface timecounter_initialize we can initialize timecounter
fields and underlying cyclecounter together. Update mlx4 timecounter
init with this new function.

Signed-off-by: Sagar Arun Kamble <sagar.a.kam...@intel.com>
Cc: Richard Cochran <richardcoch...@gmail.com>
Cc: Tariq Toukan <tar...@mellanox.com>
Cc: net...@vger.kernel.org
Cc: linux-r...@vger.kernel.org
Cc: linux-kernel@vger.kernel.org
---
 drivers/net/ethernet/mellanox/mlx4/en_clock.c | 17 +
 1 file changed, 9 insertions(+), 8 deletions(-)

diff --git a/drivers/net/ethernet/mellanox/mlx4/en_clock.c 
b/drivers/net/ethernet/mellanox/mlx4/en_clock.c
index 35987b5..dd736cc 100644
--- a/drivers/net/ethernet/mellanox/mlx4/en_clock.c
+++ b/drivers/net/ethernet/mellanox/mlx4/en_clock.c
@@ -264,6 +264,7 @@ void mlx4_en_init_timestamp(struct mlx4_en_dev *mdev)
 {
struct mlx4_dev *dev = mdev->dev;
unsigned long flags;
+   u32 mult, shift;
 
/* mlx4_en_init_timestamp is called for each netdev.
 * mdev->ptp_clock is common for all ports, skip initialization if
@@ -274,17 +275,17 @@ void mlx4_en_init_timestamp(struct mlx4_en_dev *mdev)
 
seqlock_init(>clock_lock);
 
-   memset(>clock.cc, 0, sizeof(mdev->clock.cc));
-   mdev->clock.cc.read = mlx4_en_read_clock;
-   mdev->clock.cc.mask = CLOCKSOURCE_MASK(48);
-   mdev->clock.cc.shift = freq_to_shift(dev->caps.hca_core_clock);
-   mdev->clock.cc.mult =
-   clocksource_khz2mult(1000 * dev->caps.hca_core_clock,
-mdev->clock.cc.shift);
+   shift = freq_to_shift(dev->caps.hca_core_clock);
+   mult = clocksource_khz2mult(1000 * dev->caps.hca_core_clock, shift);
mdev->nominal_c_mult = mdev->clock.cc.mult;
 
write_seqlock_irqsave(>clock_lock, flags);
-   timecounter_init(>clock, ktime_to_ns(ktime_get_real()));
+   timecounter_initialize(>clock,
+  mlx4_en_read_clock,
+  CLOCKSOURCE_MASK(48),
+  mult,
+  shift,
+  ktime_to_ns(ktime_get_real()));
write_sequnlock_irqrestore(>clock_lock, flags);
 
/* Configure the PHC */
-- 
1.9.1

[PATCH 11/27] net/mlx4: Use timecounter_initialize interface

[Sagar Arun Kamble]

With new interface timecounter_initialize we can initialize timecounter
fields and underlying cyclecounter together. Update mlx4 timecounter
init with this new function.

Signed-off-by: Sagar Arun Kamble 
Cc: Richard Cochran 
Cc: Tariq Toukan 
Cc: net...@vger.kernel.org
Cc: linux-r...@vger.kernel.org
Cc: linux-kernel@vger.kernel.org
---
 drivers/net/ethernet/mellanox/mlx4/en_clock.c | 17 +
 1 file changed, 9 insertions(+), 8 deletions(-)

diff --git a/drivers/net/ethernet/mellanox/mlx4/en_clock.c 
b/drivers/net/ethernet/mellanox/mlx4/en_clock.c
index 35987b5..dd736cc 100644
--- a/drivers/net/ethernet/mellanox/mlx4/en_clock.c
+++ b/drivers/net/ethernet/mellanox/mlx4/en_clock.c
@@ -264,6 +264,7 @@ void mlx4_en_init_timestamp(struct mlx4_en_dev *mdev)
 {
struct mlx4_dev *dev = mdev->dev;
unsigned long flags;
+   u32 mult, shift;
 
/* mlx4_en_init_timestamp is called for each netdev.
 * mdev->ptp_clock is common for all ports, skip initialization if
@@ -274,17 +275,17 @@ void mlx4_en_init_timestamp(struct mlx4_en_dev *mdev)
 
seqlock_init(>clock_lock);
 
-   memset(>clock.cc, 0, sizeof(mdev->clock.cc));
-   mdev->clock.cc.read = mlx4_en_read_clock;
-   mdev->clock.cc.mask = CLOCKSOURCE_MASK(48);
-   mdev->clock.cc.shift = freq_to_shift(dev->caps.hca_core_clock);
-   mdev->clock.cc.mult =
-   clocksource_khz2mult(1000 * dev->caps.hca_core_clock,
-mdev->clock.cc.shift);
+   shift = freq_to_shift(dev->caps.hca_core_clock);
+   mult = clocksource_khz2mult(1000 * dev->caps.hca_core_clock, shift);
mdev->nominal_c_mult = mdev->clock.cc.mult;
 
write_seqlock_irqsave(>clock_lock, flags);
-   timecounter_init(>clock, ktime_to_ns(ktime_get_real()));
+   timecounter_initialize(>clock,
+  mlx4_en_read_clock,
+  CLOCKSOURCE_MASK(48),
+  mult,
+  shift,
+  ktime_to_ns(ktime_get_real()));
write_sequnlock_irqrestore(>clock_lock, flags);
 
/* Configure the PHC */
-- 
1.9.1

[PATCH 06/27] bnx2x: Use timecounter_initialize interface

[Sagar Arun Kamble]

With new interface timecounter_initialize we can initialize timecounter
fields and underlying cyclecounter together. Update bnx2x timecounter
init with this new function.

Signed-off-by: Sagar Arun Kamble <sagar.a.kam...@intel.com>
Cc: Richard Cochran <richardcoch...@gmail.com>
Cc: Ariel Elior <ariel.el...@cavium.com>
Cc: everest-linux...@cavium.com
Cc: net...@vger.kernel.org
Cc: linux-kernel@vger.kernel.org
---
 drivers/net/ethernet/broadcom/bnx2x/bnx2x_main.c | 18 ++
 1 file changed, 6 insertions(+), 12 deletions(-)

diff --git a/drivers/net/ethernet/broadcom/bnx2x/bnx2x_main.c 
b/drivers/net/ethernet/broadcom/bnx2x/bnx2x_main.c
index 83624ad..1e10a81 100644
--- a/drivers/net/ethernet/broadcom/bnx2x/bnx2x_main.c
+++ b/drivers/net/ethernet/broadcom/bnx2x/bnx2x_main.c
@@ -15269,15 +15269,6 @@ static u64 bnx2x_cyclecounter_read(const struct 
cyclecounter *cc)
return phc_cycles;
 }
 
-static void bnx2x_init_cyclecounter(struct cyclecounter *cc)
-{
-   memset(cc, 0, sizeof(*cc));
-   cc->read = bnx2x_cyclecounter_read;
-   cc->mask = CYCLECOUNTER_MASK(64);
-   cc->shift = 0;
-   cc->mult = 1;
-}
-
 static int bnx2x_send_reset_timesync_ramrod(struct bnx2x *bp)
 {
struct bnx2x_func_state_params func_params = {NULL};
@@ -15511,9 +15502,12 @@ void bnx2x_init_ptp(struct bnx2x *bp)
 * unload / load (e.g. MTU change) while it is running.
 */
if (!bp->timecounter_init_done) {
-   bnx2x_init_cyclecounter(>timecounter.cc);
-   timecounter_init(>timecounter,
-ktime_to_ns(ktime_get_real()));
+   timecounter_initialize(>timecounter,
+  bnx2x_cyclecounter_read,
+  CYCLECOUNTER_MASK(64),
+  1,
+  0,
+  ktime_to_ns(ktime_get_real()));
bp->timecounter_init_done = 1;
}
 
-- 
1.9.1

[PATCH 06/27] bnx2x: Use timecounter_initialize interface

[Sagar Arun Kamble]

With new interface timecounter_initialize we can initialize timecounter
fields and underlying cyclecounter together. Update bnx2x timecounter
init with this new function.

Signed-off-by: Sagar Arun Kamble 
Cc: Richard Cochran 
Cc: Ariel Elior 
Cc: everest-linux...@cavium.com
Cc: net...@vger.kernel.org
Cc: linux-kernel@vger.kernel.org
---
 drivers/net/ethernet/broadcom/bnx2x/bnx2x_main.c | 18 ++
 1 file changed, 6 insertions(+), 12 deletions(-)

diff --git a/drivers/net/ethernet/broadcom/bnx2x/bnx2x_main.c 
b/drivers/net/ethernet/broadcom/bnx2x/bnx2x_main.c
index 83624ad..1e10a81 100644
--- a/drivers/net/ethernet/broadcom/bnx2x/bnx2x_main.c
+++ b/drivers/net/ethernet/broadcom/bnx2x/bnx2x_main.c
@@ -15269,15 +15269,6 @@ static u64 bnx2x_cyclecounter_read(const struct 
cyclecounter *cc)
return phc_cycles;
 }
 
-static void bnx2x_init_cyclecounter(struct cyclecounter *cc)
-{
-   memset(cc, 0, sizeof(*cc));
-   cc->read = bnx2x_cyclecounter_read;
-   cc->mask = CYCLECOUNTER_MASK(64);
-   cc->shift = 0;
-   cc->mult = 1;
-}
-
 static int bnx2x_send_reset_timesync_ramrod(struct bnx2x *bp)
 {
struct bnx2x_func_state_params func_params = {NULL};
@@ -15511,9 +15502,12 @@ void bnx2x_init_ptp(struct bnx2x *bp)
 * unload / load (e.g. MTU change) while it is running.
 */
if (!bp->timecounter_init_done) {
-   bnx2x_init_cyclecounter(>timecounter.cc);
-   timecounter_init(>timecounter,
-ktime_to_ns(ktime_get_real()));
+   timecounter_initialize(>timecounter,
+  bnx2x_cyclecounter_read,
+  CYCLECOUNTER_MASK(64),
+  1,
+  0,
+  ktime_to_ns(ktime_get_real()));
bp->timecounter_init_done = 1;
}
 
-- 
1.9.1

[PATCH 12/27] net/mlx5: Use timecounter_initialize interface

[Sagar Arun Kamble]

With new interface timecounter_initialize we can initialize timecounter
fields and underlying cyclecounter together. Update mlx5 timecounter
init with this new function.

Signed-off-by: Sagar Arun Kamble <sagar.a.kam...@intel.com>
Cc: Richard Cochran <richardcoch...@gmail.com>
Cc: Saeed Mahameed <sae...@mellanox.com>
Cc: Matan Barak <mat...@mellanox.com>
Cc: Leon Romanovsky <leo...@mellanox.com>
Cc: Eugenia Emantayev <euge...@mellanox.com>
Cc: Eitan Rabin <ra...@mellanox.com>
Cc: Feras Daoud <fera...@mellanox.com>
Cc: Sagar Arun Kamble <sagar.a.kam...@intel.com>
Cc: Miroslav Lichvar <mlich...@redhat.com>
Cc: net...@vger.kernel.org
Cc: linux-r...@vger.kernel.org
Cc: linux-kernel@vger.kernel.org
---
 drivers/net/ethernet/mellanox/mlx5/core/lib/clock.c | 18 +++---
 1 file changed, 11 insertions(+), 7 deletions(-)

diff --git a/drivers/net/ethernet/mellanox/mlx5/core/lib/clock.c 
b/drivers/net/ethernet/mellanox/mlx5/core/lib/clock.c
index 8cb6838..071f78a 100644
--- a/drivers/net/ethernet/mellanox/mlx5/core/lib/clock.c
+++ b/drivers/net/ethernet/mellanox/mlx5/core/lib/clock.c
@@ -458,6 +458,7 @@ void mlx5_init_clock(struct mlx5_core_dev *mdev)
u64 ns;
u64 frac = 0;
u32 dev_freq;
+   u32 mult, shift;
 
dev_freq = MLX5_CAP_GEN(mdev, device_frequency_khz);
if (!dev_freq) {
@@ -465,13 +466,16 @@ void mlx5_init_clock(struct mlx5_core_dev *mdev)
return;
}
rwlock_init(>lock);
-   cc->read = read_internal_timer;
-   cc->shift = MLX5_CYCLES_SHIFT;
-   cc->mult = clocksource_khz2mult(dev_freq, cc->shift);
-   clock->nominal_c_mult = cc->mult;
-   cc->mask = CLOCKSOURCE_MASK(41);
-
-   timecounter_init(>tc, ktime_to_ns(ktime_get_real()));
+   shift = MLX5_CYCLES_SHIFT;
+   mult = clocksource_khz2mult(dev_freq, shift);
+   clock->nominal_c_mult = mult;
+
+   timecounter_initialize(>tc,
+  read_internal_timer,
+  CLOCKSOURCE_MASK(41),
+  mult,
+  shift,
+  ktime_to_ns(ktime_get_real()));
 
/* Calculate period in seconds to call the overflow watchdog - to make
 * sure counter is checked at least once every wrap around.
-- 
1.9.1

[PATCH 12/27] net/mlx5: Use timecounter_initialize interface

[Sagar Arun Kamble]

With new interface timecounter_initialize we can initialize timecounter
fields and underlying cyclecounter together. Update mlx5 timecounter
init with this new function.

Signed-off-by: Sagar Arun Kamble 
Cc: Richard Cochran 
Cc: Saeed Mahameed 
Cc: Matan Barak 
Cc: Leon Romanovsky 
Cc: Eugenia Emantayev 
Cc: Eitan Rabin 
Cc: Feras Daoud 
Cc: Sagar Arun Kamble 
Cc: Miroslav Lichvar 
Cc: net...@vger.kernel.org
Cc: linux-r...@vger.kernel.org
Cc: linux-kernel@vger.kernel.org
---
 drivers/net/ethernet/mellanox/mlx5/core/lib/clock.c | 18 +++---
 1 file changed, 11 insertions(+), 7 deletions(-)

diff --git a/drivers/net/ethernet/mellanox/mlx5/core/lib/clock.c 
b/drivers/net/ethernet/mellanox/mlx5/core/lib/clock.c
index 8cb6838..071f78a 100644
--- a/drivers/net/ethernet/mellanox/mlx5/core/lib/clock.c
+++ b/drivers/net/ethernet/mellanox/mlx5/core/lib/clock.c
@@ -458,6 +458,7 @@ void mlx5_init_clock(struct mlx5_core_dev *mdev)
u64 ns;
u64 frac = 0;
u32 dev_freq;
+   u32 mult, shift;
 
dev_freq = MLX5_CAP_GEN(mdev, device_frequency_khz);
if (!dev_freq) {
@@ -465,13 +466,16 @@ void mlx5_init_clock(struct mlx5_core_dev *mdev)
return;
}
rwlock_init(>lock);
-   cc->read = read_internal_timer;
-   cc->shift = MLX5_CYCLES_SHIFT;
-   cc->mult = clocksource_khz2mult(dev_freq, cc->shift);
-   clock->nominal_c_mult = cc->mult;
-   cc->mask = CLOCKSOURCE_MASK(41);
-
-   timecounter_init(>tc, ktime_to_ns(ktime_get_real()));
+   shift = MLX5_CYCLES_SHIFT;
+   mult = clocksource_khz2mult(dev_freq, shift);
+   clock->nominal_c_mult = mult;
+
+   timecounter_initialize(>tc,
+  read_internal_timer,
+  CLOCKSOURCE_MASK(41),
+  mult,
+  shift,
+  ktime_to_ns(ktime_get_real()));
 
/* Calculate period in seconds to call the overflow watchdog - to make
 * sure counter is checked at least once every wrap around.
-- 
1.9.1

[PATCH 15/27] ALSA: hda - Use timecounter_initialize interface

[Sagar Arun Kamble]

With new interface timecounter_initialize we can initialize timecounter
fields and underlying cyclecounter together. Update azx timecounter
init with this new function.

Signed-off-by: Sagar Arun Kamble <sagar.a.kam...@intel.com>
Cc: Richard Cochran <richardcoch...@gmail.com>
Cc: Jaroslav Kysela <pe...@perex.cz>
Cc: Takashi Iwai <ti...@suse.com>
Cc: Thomas Gleixner <t...@linutronix.de>
Cc: Vinod Koul <vinod.k...@intel.com>
Cc: alsa-de...@alsa-project.org
Cc: linux-kernel@vger.kernel.org
---
 sound/hda/hdac_stream.c | 14 ++
 1 file changed, 6 insertions(+), 8 deletions(-)

diff --git a/sound/hda/hdac_stream.c b/sound/hda/hdac_stream.c
index 9426c1a..ad91dde 100644
--- a/sound/hda/hdac_stream.c
+++ b/sound/hda/hdac_stream.c
@@ -477,12 +477,8 @@ static void azx_timecounter_init(struct hdac_stream 
*azx_dev,
 bool force, u64 last)
 {
struct timecounter *tc = _dev->tc;
-   struct cyclecounter *cc = _dev->tc.cc;
u64 nsec;
 
-   cc->read = azx_cc_read;
-   cc->mask = CLOCKSOURCE_MASK(32);
-
/*
 * Converting from 24 MHz to ns means applying a 125/3 factor.
 * To avoid any saturation issues in intermediate operations,
@@ -493,11 +489,13 @@ static void azx_timecounter_init(struct hdac_stream 
*azx_dev,
 * overflows occur after about 4 hours or less, not a option.
 */
 
-   cc->mult = 125; /* saturation after 195 years */
-   cc->shift = 0;
-
nsec = 0; /* audio time is elapsed time since trigger */
-   timecounter_init(tc, nsec);
+   timecounter_initialize(tc,
+  azx_cc_read,
+  CLOCKSOURCE_MASK(32),
+  125, /* saturation after 195 years */
+  0,
+  nsec);
if (force) {
/*
 * force timecounter to use predefined value,
-- 
1.9.1

[PATCH 15/27] ALSA: hda - Use timecounter_initialize interface

[Sagar Arun Kamble]

With new interface timecounter_initialize we can initialize timecounter
fields and underlying cyclecounter together. Update azx timecounter
init with this new function.

Signed-off-by: Sagar Arun Kamble 
Cc: Richard Cochran 
Cc: Jaroslav Kysela 
Cc: Takashi Iwai 
Cc: Thomas Gleixner 
Cc: Vinod Koul 
Cc: alsa-de...@alsa-project.org
Cc: linux-kernel@vger.kernel.org
---
 sound/hda/hdac_stream.c | 14 ++
 1 file changed, 6 insertions(+), 8 deletions(-)

diff --git a/sound/hda/hdac_stream.c b/sound/hda/hdac_stream.c
index 9426c1a..ad91dde 100644
--- a/sound/hda/hdac_stream.c
+++ b/sound/hda/hdac_stream.c
@@ -477,12 +477,8 @@ static void azx_timecounter_init(struct hdac_stream 
*azx_dev,
 bool force, u64 last)
 {
struct timecounter *tc = _dev->tc;
-   struct cyclecounter *cc = _dev->tc.cc;
u64 nsec;
 
-   cc->read = azx_cc_read;
-   cc->mask = CLOCKSOURCE_MASK(32);
-
/*
 * Converting from 24 MHz to ns means applying a 125/3 factor.
 * To avoid any saturation issues in intermediate operations,
@@ -493,11 +489,13 @@ static void azx_timecounter_init(struct hdac_stream 
*azx_dev,
 * overflows occur after about 4 hours or less, not a option.
 */
 
-   cc->mult = 125; /* saturation after 195 years */
-   cc->shift = 0;
-
nsec = 0; /* audio time is elapsed time since trigger */
-   timecounter_init(tc, nsec);
+   timecounter_initialize(tc,
+  azx_cc_read,
+  CLOCKSOURCE_MASK(32),
+  125, /* saturation after 195 years */
+  0,
+  nsec);
if (force) {
/*
 * force timecounter to use predefined value,
-- 
1.9.1

[PATCH 14/27] net: cpts: Use timecounter_initialize interface

[Sagar Arun Kamble]

With new interface timecounter_initialize we can initialize timecounter
fields and underlying cyclecounter together. Update cpts timecounter
init with this new function.

Signed-off-by: Sagar Arun Kamble <sagar.a.kam...@intel.com>
Cc: "David S. Miller" <da...@davemloft.net>
Cc: Grygorii Strashko <grygorii.stras...@ti.com>
Cc: Richard Cochran <richardcoch...@gmail.com>
Cc: Bhumika Goyal <bhumi...@gmail.com>
Cc: Thomas Gleixner <t...@linutronix.de>
Cc: Ivan Khoronzhuk <ivan.khoronz...@linaro.org>
Cc: net...@vger.kernel.org
Cc: linux-kernel@vger.kernel.org
---
 drivers/net/ethernet/ti/cpts.c | 29 +
 drivers/net/ethernet/ti/cpts.h |  3 +++
 2 files changed, 20 insertions(+), 12 deletions(-)

diff --git a/drivers/net/ethernet/ti/cpts.c b/drivers/net/ethernet/ti/cpts.c
index b8fe843..e6afc94 100644
--- a/drivers/net/ethernet/ti/cpts.c
+++ b/drivers/net/ethernet/ti/cpts.c
@@ -447,7 +447,12 @@ int cpts_register(struct cpts *cpts)
cpts_write32(cpts, CPTS_EN, control);
cpts_write32(cpts, TS_PEND_EN, int_enable);
 
-   timecounter_init(>tc, ktime_to_ns(ktime_get_real()));
+   timecounter_initialize(>tc,
+  cpts_systim_read,
+  cpts->mask,
+  cpts->mult,
+  cpts->shift,
+  ktime_to_ns(ktime_get_real()));
 
cpts->clock = ptp_clock_register(>info, cpts->dev);
if (IS_ERR(cpts->clock)) {
@@ -484,7 +489,8 @@ void cpts_unregister(struct cpts *cpts)
 }
 EXPORT_SYMBOL_GPL(cpts_unregister);
 
-static void cpts_calc_mult_shift(struct cpts *cpts)
+static void cpts_calc_mult_shift(struct cpts *cpts, u64 mask,
+u32 *mult, u32 *shift)
 {
struct cyclecounter *cc = >tc.cc;
u64 frac, maxsec, ns;
@@ -495,7 +501,7 @@ static void cpts_calc_mult_shift(struct cpts *cpts)
/* Calc the maximum number of seconds which we can run before
 * wrapping around.
 */
-   maxsec = cc->mask;
+   maxsec = mask;
do_div(maxsec, freq);
/* limit conversation rate to 10 sec as higher values will produce
 * too small mult factors and so reduce the conversion accuracy
@@ -508,18 +514,18 @@ static void cpts_calc_mult_shift(struct cpts *cpts)
dev_info(cpts->dev, "cpts: overflow check period %lu (jiffies)\n",
 cpts->ov_check_period);
 
-   if (cc->mult || cc->shift)
+   if (*mult || *shift)
return;
 
-   clocks_calc_mult_shift(>mult, >shift,
+   clocks_calc_mult_shift(mult, shift,
   freq, NSEC_PER_SEC, maxsec);
 
frac = 0;
-   ns = cyclecounter_cyc2ns(cc, freq, cc->mask, );
+   ns = cyclecounter_cyc2ns(cc, freq, mask, );
 
dev_info(cpts->dev,
 "CPTS: ref_clk_freq:%u calc_mult:%u calc_shift:%u error:%lld 
nsec/sec\n",
-freq, cc->mult, cc->shift,
+freq, *mult, *shift,
 (ns - NSEC_PER_SEC));
 }
 
@@ -571,15 +577,14 @@ struct cpts *cpts_create(struct device *dev, void __iomem 
*regs,
 
clk_prepare(cpts->refclk);
 
-   cpts->tc.cc.read = cpts_systim_read;
-   cpts->tc.cc.mask = CLOCKSOURCE_MASK(32);
+   cpts->mask = CLOCKSOURCE_MASK(32);
cpts->info = cpts_info;
 
-   cpts_calc_mult_shift(cpts);
-   /* save tc.cc.mult original value as it can be modified
+   cpts_calc_mult_shift(cpts, cpts->mask, >mult, >shift);
+   /* save mult original value as it can be modified
 * by cpts_ptp_adjfreq().
 */
-   cpts->cc_mult = cpts->tc.cc.mult;
+   cpts->cc_mult = cpts->mult;
 
return cpts;
 }
diff --git a/drivers/net/ethernet/ti/cpts.h b/drivers/net/ethernet/ti/cpts.h
index a7174eb..da50d34 100644
--- a/drivers/net/ethernet/ti/cpts.h
+++ b/drivers/net/ethernet/ti/cpts.h
@@ -118,6 +118,9 @@ struct cpts {
spinlock_t lock; /* protects time registers */
u32 cc_mult; /* for the nominal frequency */
struct timecounter tc;
+   u64 mask;
+   u32 mult;
+   u32 shift;
int phc_index;
struct clk *refclk;
struct list_head events;
-- 
1.9.1

[PATCH 14/27] net: cpts: Use timecounter_initialize interface

[Sagar Arun Kamble]

With new interface timecounter_initialize we can initialize timecounter
fields and underlying cyclecounter together. Update cpts timecounter
init with this new function.

Signed-off-by: Sagar Arun Kamble 
Cc: "David S. Miller" 
Cc: Grygorii Strashko 
Cc: Richard Cochran 
Cc: Bhumika Goyal 
Cc: Thomas Gleixner 
Cc: Ivan Khoronzhuk 
Cc: net...@vger.kernel.org
Cc: linux-kernel@vger.kernel.org
---
 drivers/net/ethernet/ti/cpts.c | 29 +
 drivers/net/ethernet/ti/cpts.h |  3 +++
 2 files changed, 20 insertions(+), 12 deletions(-)

diff --git a/drivers/net/ethernet/ti/cpts.c b/drivers/net/ethernet/ti/cpts.c
index b8fe843..e6afc94 100644
--- a/drivers/net/ethernet/ti/cpts.c
+++ b/drivers/net/ethernet/ti/cpts.c
@@ -447,7 +447,12 @@ int cpts_register(struct cpts *cpts)
cpts_write32(cpts, CPTS_EN, control);
cpts_write32(cpts, TS_PEND_EN, int_enable);
 
-   timecounter_init(>tc, ktime_to_ns(ktime_get_real()));
+   timecounter_initialize(>tc,
+  cpts_systim_read,
+  cpts->mask,
+  cpts->mult,
+  cpts->shift,
+  ktime_to_ns(ktime_get_real()));
 
cpts->clock = ptp_clock_register(>info, cpts->dev);
if (IS_ERR(cpts->clock)) {
@@ -484,7 +489,8 @@ void cpts_unregister(struct cpts *cpts)
 }
 EXPORT_SYMBOL_GPL(cpts_unregister);
 
-static void cpts_calc_mult_shift(struct cpts *cpts)
+static void cpts_calc_mult_shift(struct cpts *cpts, u64 mask,
+u32 *mult, u32 *shift)
 {
struct cyclecounter *cc = >tc.cc;
u64 frac, maxsec, ns;
@@ -495,7 +501,7 @@ static void cpts_calc_mult_shift(struct cpts *cpts)
/* Calc the maximum number of seconds which we can run before
 * wrapping around.
 */
-   maxsec = cc->mask;
+   maxsec = mask;
do_div(maxsec, freq);
/* limit conversation rate to 10 sec as higher values will produce
 * too small mult factors and so reduce the conversion accuracy
@@ -508,18 +514,18 @@ static void cpts_calc_mult_shift(struct cpts *cpts)
dev_info(cpts->dev, "cpts: overflow check period %lu (jiffies)\n",
 cpts->ov_check_period);
 
-   if (cc->mult || cc->shift)
+   if (*mult || *shift)
return;
 
-   clocks_calc_mult_shift(>mult, >shift,
+   clocks_calc_mult_shift(mult, shift,
   freq, NSEC_PER_SEC, maxsec);
 
frac = 0;
-   ns = cyclecounter_cyc2ns(cc, freq, cc->mask, );
+   ns = cyclecounter_cyc2ns(cc, freq, mask, );
 
dev_info(cpts->dev,
 "CPTS: ref_clk_freq:%u calc_mult:%u calc_shift:%u error:%lld 
nsec/sec\n",
-freq, cc->mult, cc->shift,
+freq, *mult, *shift,
 (ns - NSEC_PER_SEC));
 }
 
@@ -571,15 +577,14 @@ struct cpts *cpts_create(struct device *dev, void __iomem 
*regs,
 
clk_prepare(cpts->refclk);
 
-   cpts->tc.cc.read = cpts_systim_read;
-   cpts->tc.cc.mask = CLOCKSOURCE_MASK(32);
+   cpts->mask = CLOCKSOURCE_MASK(32);
cpts->info = cpts_info;
 
-   cpts_calc_mult_shift(cpts);
-   /* save tc.cc.mult original value as it can be modified
+   cpts_calc_mult_shift(cpts, cpts->mask, >mult, >shift);
+   /* save mult original value as it can be modified
 * by cpts_ptp_adjfreq().
 */
-   cpts->cc_mult = cpts->tc.cc.mult;
+   cpts->cc_mult = cpts->mult;
 
return cpts;
 }
diff --git a/drivers/net/ethernet/ti/cpts.h b/drivers/net/ethernet/ti/cpts.h
index a7174eb..da50d34 100644
--- a/drivers/net/ethernet/ti/cpts.h
+++ b/drivers/net/ethernet/ti/cpts.h
@@ -118,6 +118,9 @@ struct cpts {
spinlock_t lock; /* protects time registers */
u32 cc_mult; /* for the nominal frequency */
struct timecounter tc;
+   u64 mask;
+   u32 mult;
+   u32 shift;
int phc_index;
struct clk *refclk;
struct list_head events;
-- 
1.9.1

[PATCH 18/27] bnx2x: Use timecounter_reset interface

[Sagar Arun Kamble]

With new interface timecounter_reset we can update the start time for
timecounter. Update bnx2x_ptp_settime with this new function.

Signed-off-by: Sagar Arun Kamble <sagar.a.kam...@intel.com>
Cc: Richard Cochran <richardcoch...@gmail.com>
Cc: Ariel Elior <ariel.el...@cavium.com>
Cc: everest-linux...@cavium.com
Cc: net...@vger.kernel.org
Cc: linux-kernel@vger.kernel.org
---
 drivers/net/ethernet/broadcom/bnx2x/bnx2x_main.c | 4 ++--
 1 file changed, 2 insertions(+), 2 deletions(-)

diff --git a/drivers/net/ethernet/broadcom/bnx2x/bnx2x_main.c 
b/drivers/net/ethernet/broadcom/bnx2x/bnx2x_main.c
index 1e10a81..e30e736 100644
--- a/drivers/net/ethernet/broadcom/bnx2x/bnx2x_main.c
+++ b/drivers/net/ethernet/broadcom/bnx2x/bnx2x_main.c
@@ -13849,8 +13849,8 @@ static int bnx2x_ptp_settime(struct ptp_clock_info *ptp,
 
DP(BNX2X_MSG_PTP, "PTP settime called, ns = %llu\n", ns);
 
-   /* Re-init the timecounter */
-   timecounter_init(>timecounter, ns);
+   /* Reset the timecounter */
+   timecounter_reset(>timecounter, ns);
 
return 0;
 }
-- 
1.9.1

[PATCH 16/27] timecounter: Introduce timecounter_reset

[Sagar Arun Kamble]

timecounter_init initializes the cyclecounter and sets the start time.
Sometimes drivers using timecounter just want to update start time.
This patch creates new function timecounter_reset that just resets the
timecounter start time and associated state.

Suggested-by: Richard Cochran <richardcoch...@gmail.com>
Signed-off-by: Sagar Arun Kamble <sagar.a.kam...@intel.com>
Cc: Richard Cochran <richardcoch...@gmail.com>
Cc: Chris Wilson <ch...@chris-wilson.co.uk>
Cc: John Stultz <john.stu...@linaro.org>
Cc: Thomas Gleixner <t...@linutronix.de>
Cc: Stephen Boyd <sb...@codeaurora.org>
Cc: linux-kernel@vger.kernel.org
---
 include/linux/timecounter.h | 17 +++--
 kernel/time/timecounter.c   |  9 +
 2 files changed, 24 insertions(+), 2 deletions(-)

diff --git a/include/linux/timecounter.h b/include/linux/timecounter.h
index 59d3fd7..e0fd741 100644
--- a/include/linux/timecounter.h
+++ b/include/linux/timecounter.h
@@ -98,7 +98,7 @@ static inline void timecounter_adjtime(struct timecounter 
*tc, s64 delta)
 
 /**
  * timecounter_init - initialize a time counter
- * @tc:Pointer to time counter which is to be 
initialized/reset
+ * @tc:Pointer to time counter which is to be 
initialized
  * @start_tstamp:  Arbitrary initial time stamp.
  *
  * After this call the current cycle register (roughly) corresponds to
@@ -131,8 +131,21 @@ extern void timecounter_initialize(struct timecounter *tc,
   u64 start_tstamp);
 
 /**
+ * timecounter_reset - reset a time counter
+ * @tc:Pointer to time counter which is to be reset
+ * @start_tstamp:  Arbitrary initial time stamp.
+ *
+ * After this call the current cycle register (roughly) corresponds to
+ * the supplied time stamp. Every call to timecounter_read() increments
+ * the time stamp counter by the number of elapsed nanoseconds.
+ */
+extern void timecounter_reset(struct timecounter *tc,
+ u64 start_tstamp);
+
+/**
  * timecounter_read - return nanoseconds elapsed since timecounter_init() or
- *timecounter_initialize() plus the initial time stamp
+ *timecounter_initialize() or timecounter_reset() plus
+ *   the initial time stamp
  * @tc:  Pointer to time counter.
  *
  * In other words, keeps track of time since the same epoch as
diff --git a/kernel/time/timecounter.c b/kernel/time/timecounter.c
index 6d915752..3f62fe0 100644
--- a/kernel/time/timecounter.c
+++ b/kernel/time/timecounter.c
@@ -47,6 +47,15 @@ void timecounter_initialize(struct timecounter *tc,
 }
 EXPORT_SYMBOL_GPL(timecounter_initialize);
 
+void timecounter_reset(struct timecounter *tc,
+  u64 start_tstamp)
+{
+   tc->cycle_last = tc->cc.read(>cc);
+   tc->nsec = start_tstamp;
+   tc->frac = 0;
+}
+EXPORT_SYMBOL_GPL(timecounter_reset);
+
 /**
  * timecounter_read_delta - get nanoseconds since last call of this function
  * @tc: Pointer to time counter
-- 
1.9.1

[PATCH 18/27] bnx2x: Use timecounter_reset interface

[Sagar Arun Kamble]

With new interface timecounter_reset we can update the start time for
timecounter. Update bnx2x_ptp_settime with this new function.

Signed-off-by: Sagar Arun Kamble 
Cc: Richard Cochran 
Cc: Ariel Elior 
Cc: everest-linux...@cavium.com
Cc: net...@vger.kernel.org
Cc: linux-kernel@vger.kernel.org
---
 drivers/net/ethernet/broadcom/bnx2x/bnx2x_main.c | 4 ++--
 1 file changed, 2 insertions(+), 2 deletions(-)

diff --git a/drivers/net/ethernet/broadcom/bnx2x/bnx2x_main.c 
b/drivers/net/ethernet/broadcom/bnx2x/bnx2x_main.c
index 1e10a81..e30e736 100644
--- a/drivers/net/ethernet/broadcom/bnx2x/bnx2x_main.c
+++ b/drivers/net/ethernet/broadcom/bnx2x/bnx2x_main.c
@@ -13849,8 +13849,8 @@ static int bnx2x_ptp_settime(struct ptp_clock_info *ptp,
 
DP(BNX2X_MSG_PTP, "PTP settime called, ns = %llu\n", ns);
 
-   /* Re-init the timecounter */
-   timecounter_init(>timecounter, ns);
+   /* Reset the timecounter */
+   timecounter_reset(>timecounter, ns);
 
return 0;
 }
-- 
1.9.1

[PATCH 16/27] timecounter: Introduce timecounter_reset

[Sagar Arun Kamble]

timecounter_init initializes the cyclecounter and sets the start time.
Sometimes drivers using timecounter just want to update start time.
This patch creates new function timecounter_reset that just resets the
timecounter start time and associated state.

Suggested-by: Richard Cochran 
Signed-off-by: Sagar Arun Kamble 
Cc: Richard Cochran 
Cc: Chris Wilson 
Cc: John Stultz 
Cc: Thomas Gleixner 
Cc: Stephen Boyd 
Cc: linux-kernel@vger.kernel.org
---
 include/linux/timecounter.h | 17 +++--
 kernel/time/timecounter.c   |  9 +
 2 files changed, 24 insertions(+), 2 deletions(-)

diff --git a/include/linux/timecounter.h b/include/linux/timecounter.h
index 59d3fd7..e0fd741 100644
--- a/include/linux/timecounter.h
+++ b/include/linux/timecounter.h
@@ -98,7 +98,7 @@ static inline void timecounter_adjtime(struct timecounter 
*tc, s64 delta)
 
 /**
  * timecounter_init - initialize a time counter
- * @tc:Pointer to time counter which is to be 
initialized/reset
+ * @tc:Pointer to time counter which is to be 
initialized
  * @start_tstamp:  Arbitrary initial time stamp.
  *
  * After this call the current cycle register (roughly) corresponds to
@@ -131,8 +131,21 @@ extern void timecounter_initialize(struct timecounter *tc,
   u64 start_tstamp);
 
 /**
+ * timecounter_reset - reset a time counter
+ * @tc:Pointer to time counter which is to be reset
+ * @start_tstamp:  Arbitrary initial time stamp.
+ *
+ * After this call the current cycle register (roughly) corresponds to
+ * the supplied time stamp. Every call to timecounter_read() increments
+ * the time stamp counter by the number of elapsed nanoseconds.
+ */
+extern void timecounter_reset(struct timecounter *tc,
+ u64 start_tstamp);
+
+/**
  * timecounter_read - return nanoseconds elapsed since timecounter_init() or
- *timecounter_initialize() plus the initial time stamp
+ *timecounter_initialize() or timecounter_reset() plus
+ *   the initial time stamp
  * @tc:  Pointer to time counter.
  *
  * In other words, keeps track of time since the same epoch as
diff --git a/kernel/time/timecounter.c b/kernel/time/timecounter.c
index 6d915752..3f62fe0 100644
--- a/kernel/time/timecounter.c
+++ b/kernel/time/timecounter.c
@@ -47,6 +47,15 @@ void timecounter_initialize(struct timecounter *tc,
 }
 EXPORT_SYMBOL_GPL(timecounter_initialize);
 
+void timecounter_reset(struct timecounter *tc,
+  u64 start_tstamp)
+{
+   tc->cycle_last = tc->cc.read(>cc);
+   tc->nsec = start_tstamp;
+   tc->frac = 0;
+}
+EXPORT_SYMBOL_GPL(timecounter_reset);
+
 /**
  * timecounter_read_delta - get nanoseconds since last call of this function
  * @tc: Pointer to time counter
-- 
1.9.1

[PATCH 22/27] ixgbe: Use timecounter_reset interface

[Sagar Arun Kamble]

With new interface timecounter_reset we can update the start time for
timecounter. Update ixgbe_ptp_settime with this new function.

Signed-off-by: Sagar Arun Kamble <sagar.a.kam...@intel.com>
Cc: Richard Cochran <richardcoch...@gmail.com>
Cc: Jeff Kirsher <jeffrey.t.kirs...@intel.com>
Cc: intel-wired-...@lists.osuosl.org
Cc: net...@vger.kernel.org
Cc: linux-kernel@vger.kernel.org
---
 drivers/net/ethernet/intel/ixgbe/ixgbe_ptp.c | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/drivers/net/ethernet/intel/ixgbe/ixgbe_ptp.c 
b/drivers/net/ethernet/intel/ixgbe/ixgbe_ptp.c
index 073c1ef..35f85ef 100644
--- a/drivers/net/ethernet/intel/ixgbe/ixgbe_ptp.c
+++ b/drivers/net/ethernet/intel/ixgbe/ixgbe_ptp.c
@@ -508,7 +508,7 @@ static int ixgbe_ptp_settime(struct ptp_clock_info *ptp,
 
/* reset the timecounter */
spin_lock_irqsave(>tmreg_lock, flags);
-   timecounter_init(>hw_tc, ns);
+   timecounter_reset(>hw_tc, ns);
spin_unlock_irqrestore(>tmreg_lock, flags);
 
if (adapter->ptp_setup_sdp)
-- 
1.9.1

[PATCH 22/27] ixgbe: Use timecounter_reset interface

[Sagar Arun Kamble]

With new interface timecounter_reset we can update the start time for
timecounter. Update ixgbe_ptp_settime with this new function.

Signed-off-by: Sagar Arun Kamble 
Cc: Richard Cochran 
Cc: Jeff Kirsher 
Cc: intel-wired-...@lists.osuosl.org
Cc: net...@vger.kernel.org
Cc: linux-kernel@vger.kernel.org
---
 drivers/net/ethernet/intel/ixgbe/ixgbe_ptp.c | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/drivers/net/ethernet/intel/ixgbe/ixgbe_ptp.c 
b/drivers/net/ethernet/intel/ixgbe/ixgbe_ptp.c
index 073c1ef..35f85ef 100644
--- a/drivers/net/ethernet/intel/ixgbe/ixgbe_ptp.c
+++ b/drivers/net/ethernet/intel/ixgbe/ixgbe_ptp.c
@@ -508,7 +508,7 @@ static int ixgbe_ptp_settime(struct ptp_clock_info *ptp,
 
/* reset the timecounter */
spin_lock_irqsave(>tmreg_lock, flags);
-   timecounter_init(>hw_tc, ns);
+   timecounter_reset(>hw_tc, ns);
spin_unlock_irqrestore(>tmreg_lock, flags);
 
if (adapter->ptp_setup_sdp)
-- 
1.9.1

[PATCH 23/27] net/mlx4: Use timecounter_reset interface

[Sagar Arun Kamble]

With new interface timecounter_reset we can update the start time for
timecounter. Update mlx4_en_phc_settime with this new function.

Signed-off-by: Sagar Arun Kamble <sagar.a.kam...@intel.com>
Cc: Richard Cochran <richardcoch...@gmail.com>
Cc: Tariq Toukan <tar...@mellanox.com>
Cc: net...@vger.kernel.org
Cc: linux-r...@vger.kernel.org
Cc: linux-kernel@vger.kernel.org
---
 drivers/net/ethernet/mellanox/mlx4/en_clock.c | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/drivers/net/ethernet/mellanox/mlx4/en_clock.c 
b/drivers/net/ethernet/mellanox/mlx4/en_clock.c
index dd736cc..bfed4ac 100644
--- a/drivers/net/ethernet/mellanox/mlx4/en_clock.c
+++ b/drivers/net/ethernet/mellanox/mlx4/en_clock.c
@@ -207,7 +207,7 @@ static int mlx4_en_phc_settime(struct ptp_clock_info *ptp,
 
/* reset the timecounter */
write_seqlock_irqsave(>clock_lock, flags);
-   timecounter_init(>clock, ns);
+   timecounter_reset(>clock, ns);
write_sequnlock_irqrestore(>clock_lock, flags);
 
return 0;
-- 
1.9.1

[PATCH 21/27] igb: Use timecounter_reset interface

[Sagar Arun Kamble]

With new interface timecounter_reset we can update the start time for
timecounter. Update igb_ptp_settime_82576 with this new function.

Signed-off-by: Sagar Arun Kamble <sagar.a.kam...@intel.com>
Cc: Richard Cochran <richardcoch...@gmail.com>
Cc: Jeff Kirsher <jeffrey.t.kirs...@intel.com>
Cc: intel-wired-...@lists.osuosl.org
Cc: net...@vger.kernel.org
Cc: linux-kernel@vger.kernel.org
---
 drivers/net/ethernet/intel/igb/igb_ptp.c | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/drivers/net/ethernet/intel/igb/igb_ptp.c 
b/drivers/net/ethernet/intel/igb/igb_ptp.c
index f6c1a8d..90344b2 100644
--- a/drivers/net/ethernet/intel/igb/igb_ptp.c
+++ b/drivers/net/ethernet/intel/igb/igb_ptp.c
@@ -330,7 +330,7 @@ static int igb_ptp_settime_82576(struct ptp_clock_info *ptp,
 
spin_lock_irqsave(>tmreg_lock, flags);
 
-   timecounter_init(>tc, ns);
+   timecounter_reset(>tc, ns);
 
spin_unlock_irqrestore(>tmreg_lock, flags);
 
-- 
1.9.1

[PATCH 21/27] igb: Use timecounter_reset interface

[Sagar Arun Kamble]

With new interface timecounter_reset we can update the start time for
timecounter. Update igb_ptp_settime_82576 with this new function.

Signed-off-by: Sagar Arun Kamble 
Cc: Richard Cochran 
Cc: Jeff Kirsher 
Cc: intel-wired-...@lists.osuosl.org
Cc: net...@vger.kernel.org
Cc: linux-kernel@vger.kernel.org
---
 drivers/net/ethernet/intel/igb/igb_ptp.c | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/drivers/net/ethernet/intel/igb/igb_ptp.c 
b/drivers/net/ethernet/intel/igb/igb_ptp.c
index f6c1a8d..90344b2 100644
--- a/drivers/net/ethernet/intel/igb/igb_ptp.c
+++ b/drivers/net/ethernet/intel/igb/igb_ptp.c
@@ -330,7 +330,7 @@ static int igb_ptp_settime_82576(struct ptp_clock_info *ptp,
 
spin_lock_irqsave(>tmreg_lock, flags);
 
-   timecounter_init(>tc, ns);
+   timecounter_reset(>tc, ns);
 
spin_unlock_irqrestore(>tmreg_lock, flags);
 
-- 
1.9.1

[PATCH 23/27] net/mlx4: Use timecounter_reset interface

[Sagar Arun Kamble]

With new interface timecounter_reset we can update the start time for
timecounter. Update mlx4_en_phc_settime with this new function.

Signed-off-by: Sagar Arun Kamble 
Cc: Richard Cochran 
Cc: Tariq Toukan 
Cc: net...@vger.kernel.org
Cc: linux-r...@vger.kernel.org
Cc: linux-kernel@vger.kernel.org
---
 drivers/net/ethernet/mellanox/mlx4/en_clock.c | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/drivers/net/ethernet/mellanox/mlx4/en_clock.c 
b/drivers/net/ethernet/mellanox/mlx4/en_clock.c
index dd736cc..bfed4ac 100644
--- a/drivers/net/ethernet/mellanox/mlx4/en_clock.c
+++ b/drivers/net/ethernet/mellanox/mlx4/en_clock.c
@@ -207,7 +207,7 @@ static int mlx4_en_phc_settime(struct ptp_clock_info *ptp,
 
/* reset the timecounter */
write_seqlock_irqsave(>clock_lock, flags);
-   timecounter_init(>clock, ns);
+   timecounter_reset(>clock, ns);
write_sequnlock_irqrestore(>clock_lock, flags);
 
return 0;
-- 
1.9.1

[PATCH 24/27] net/mlx5: Use timecounter_reset interface

[Sagar Arun Kamble]

With new interface timecounter_reset we can update the start time for
timecounter. Update mlx5_ptp_settime with this new function.

Signed-off-by: Sagar Arun Kamble <sagar.a.kam...@intel.com>
Cc: Richard Cochran <richardcoch...@gmail.com>
Cc: Saeed Mahameed <sae...@mellanox.com>
Cc: Matan Barak <mat...@mellanox.com>
Cc: Leon Romanovsky <leo...@mellanox.com>
Cc: Eugenia Emantayev <euge...@mellanox.com>
Cc: Eitan Rabin <ra...@mellanox.com>
Cc: Feras Daoud <fera...@mellanox.com>
Cc: Tariq Toukan <tar...@mellanox.com>
Cc: Thomas Gleixner <t...@linutronix.de>
Cc: net...@vger.kernel.org
Cc: linux-r...@vger.kernel.org
Cc: linux-kernel@vger.kernel.org
---
 drivers/net/ethernet/mellanox/mlx5/core/lib/clock.c | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/drivers/net/ethernet/mellanox/mlx5/core/lib/clock.c 
b/drivers/net/ethernet/mellanox/mlx5/core/lib/clock.c
index 071f78a..e490522 100644
--- a/drivers/net/ethernet/mellanox/mlx5/core/lib/clock.c
+++ b/drivers/net/ethernet/mellanox/mlx5/core/lib/clock.c
@@ -122,7 +122,7 @@ static int mlx5_ptp_settime(struct ptp_clock_info *ptp,
unsigned long flags;
 
write_lock_irqsave(>lock, flags);
-   timecounter_init(>tc, ns);
+   timecounter_reset(>tc, ns);
write_unlock_irqrestore(>lock, flags);
 
return 0;
-- 
1.9.1

[PATCH 24/27] net/mlx5: Use timecounter_reset interface

[Sagar Arun Kamble]

With new interface timecounter_reset we can update the start time for
timecounter. Update mlx5_ptp_settime with this new function.

Signed-off-by: Sagar Arun Kamble 
Cc: Richard Cochran 
Cc: Saeed Mahameed 
Cc: Matan Barak 
Cc: Leon Romanovsky 
Cc: Eugenia Emantayev 
Cc: Eitan Rabin 
Cc: Feras Daoud 
Cc: Tariq Toukan 
Cc: Thomas Gleixner 
Cc: net...@vger.kernel.org
Cc: linux-r...@vger.kernel.org
Cc: linux-kernel@vger.kernel.org
---
 drivers/net/ethernet/mellanox/mlx5/core/lib/clock.c | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/drivers/net/ethernet/mellanox/mlx5/core/lib/clock.c 
b/drivers/net/ethernet/mellanox/mlx5/core/lib/clock.c
index 071f78a..e490522 100644
--- a/drivers/net/ethernet/mellanox/mlx5/core/lib/clock.c
+++ b/drivers/net/ethernet/mellanox/mlx5/core/lib/clock.c
@@ -122,7 +122,7 @@ static int mlx5_ptp_settime(struct ptp_clock_info *ptp,
unsigned long flags;
 
write_lock_irqsave(>lock, flags);
-   timecounter_init(>tc, ns);
+   timecounter_reset(>tc, ns);
write_unlock_irqrestore(>lock, flags);
 
return 0;
-- 
1.9.1

[PATCH 25/27] qede: Use timecounter_reset interface

[Sagar Arun Kamble]

With new interface timecounter_reset we can update the start time for
timecounter. Update qede_ptp_settime with this new function.

Signed-off-by: Sagar Arun Kamble <sagar.a.kam...@intel.com>
Cc: Richard Cochran <richardcoch...@gmail.com>
Cc: Ariel Elior <ariel.el...@cavium.com>
Cc: everest-linux...@cavium.com
Cc: net...@vger.kernel.org
Cc: linux-kernel@vger.kernel.org
---
 drivers/net/ethernet/qlogic/qede/qede_ptp.c | 4 ++--
 1 file changed, 2 insertions(+), 2 deletions(-)

diff --git a/drivers/net/ethernet/qlogic/qede/qede_ptp.c 
b/drivers/net/ethernet/qlogic/qede/qede_ptp.c
index 0d054dc..c2de311 100644
--- a/drivers/net/ethernet/qlogic/qede/qede_ptp.c
+++ b/drivers/net/ethernet/qlogic/qede/qede_ptp.c
@@ -129,9 +129,9 @@ static int qede_ptp_settime(struct ptp_clock_info *info,
 
DP_VERBOSE(edev, QED_MSG_DEBUG, "PTP settime called, ns = %llu\n", ns);
 
-   /* Re-init the timecounter */
+   /* Reset the timecounter */
spin_lock_bh(>lock);
-   timecounter_init(>tc, ns);
+   timecounter_reset(>tc, ns);
spin_unlock_bh(>lock);
 
return 0;
-- 
1.9.1

[PATCH 26/27] net: cpts: Use timecounter_reset interface

[Sagar Arun Kamble]

With new interface timecounter_reset we can update the start time for
timecounter. Update cpts_ptp_settime with this new function.

Signed-off-by: Sagar Arun Kamble <sagar.a.kam...@intel.com>
Cc: "David S. Miller" <da...@davemloft.net>
Cc: Grygorii Strashko <grygorii.stras...@ti.com>
Cc: Bhumika Goyal <bhumi...@gmail.com>
Cc: Richard Cochran <richardcoch...@gmail.com>
Cc: Thomas Gleixner <t...@linutronix.de>
Cc: net...@vger.kernel.org
Cc: linux-kernel@vger.kernel.org
---
 drivers/net/ethernet/ti/cpts.c | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/drivers/net/ethernet/ti/cpts.c b/drivers/net/ethernet/ti/cpts.c
index e6afc94..41bac4d 100644
--- a/drivers/net/ethernet/ti/cpts.c
+++ b/drivers/net/ethernet/ti/cpts.c
@@ -268,7 +268,7 @@ static int cpts_ptp_settime(struct ptp_clock_info *ptp,
ns = timespec64_to_ns(ts);
 
spin_lock_irqsave(>lock, flags);
-   timecounter_init(>tc, ns);
+   timecounter_reset(>tc, ns);
spin_unlock_irqrestore(>lock, flags);
 
return 0;
-- 
1.9.1

[PATCH 25/27] qede: Use timecounter_reset interface

[Sagar Arun Kamble]

With new interface timecounter_reset we can update the start time for
timecounter. Update qede_ptp_settime with this new function.

Signed-off-by: Sagar Arun Kamble 
Cc: Richard Cochran 
Cc: Ariel Elior 
Cc: everest-linux...@cavium.com
Cc: net...@vger.kernel.org
Cc: linux-kernel@vger.kernel.org
---
 drivers/net/ethernet/qlogic/qede/qede_ptp.c | 4 ++--
 1 file changed, 2 insertions(+), 2 deletions(-)

diff --git a/drivers/net/ethernet/qlogic/qede/qede_ptp.c 
b/drivers/net/ethernet/qlogic/qede/qede_ptp.c
index 0d054dc..c2de311 100644
--- a/drivers/net/ethernet/qlogic/qede/qede_ptp.c
+++ b/drivers/net/ethernet/qlogic/qede/qede_ptp.c
@@ -129,9 +129,9 @@ static int qede_ptp_settime(struct ptp_clock_info *info,
 
DP_VERBOSE(edev, QED_MSG_DEBUG, "PTP settime called, ns = %llu\n", ns);
 
-   /* Re-init the timecounter */
+   /* Reset the timecounter */
spin_lock_bh(>lock);
-   timecounter_init(>tc, ns);
+   timecounter_reset(>tc, ns);
spin_unlock_bh(>lock);
 
return 0;
-- 
1.9.1

[PATCH 26/27] net: cpts: Use timecounter_reset interface

[Sagar Arun Kamble]

With new interface timecounter_reset we can update the start time for
timecounter. Update cpts_ptp_settime with this new function.

Signed-off-by: Sagar Arun Kamble 
Cc: "David S. Miller" 
Cc: Grygorii Strashko 
Cc: Bhumika Goyal 
Cc: Richard Cochran 
Cc: Thomas Gleixner 
Cc: net...@vger.kernel.org
Cc: linux-kernel@vger.kernel.org
---
 drivers/net/ethernet/ti/cpts.c | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/drivers/net/ethernet/ti/cpts.c b/drivers/net/ethernet/ti/cpts.c
index e6afc94..41bac4d 100644
--- a/drivers/net/ethernet/ti/cpts.c
+++ b/drivers/net/ethernet/ti/cpts.c
@@ -268,7 +268,7 @@ static int cpts_ptp_settime(struct ptp_clock_info *ptp,
ns = timespec64_to_ns(ts);
 
spin_lock_irqsave(>lock, flags);
-   timecounter_init(>tc, ns);
+   timecounter_reset(>tc, ns);
spin_unlock_irqrestore(>lock, flags);
 
return 0;
-- 
1.9.1

[PATCH 20/27] e1000e: Use timecounter_reset interface

[Sagar Arun Kamble]

With new interface timecounter_reset we can update the start time for
timecounter. Update e1000e_phc_settime with this new function.

Signed-off-by: Sagar Arun Kamble <sagar.a.kam...@intel.com>
Cc: Richard Cochran <richardcoch...@gmail.com>
Cc: Jeff Kirsher <jeffrey.t.kirs...@intel.com>
Cc: intel-wired-...@lists.osuosl.org
Cc: net...@vger.kernel.org
Cc: linux-kernel@vger.kernel.org
---
 drivers/net/ethernet/intel/e1000e/ptp.c | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/drivers/net/ethernet/intel/e1000e/ptp.c 
b/drivers/net/ethernet/intel/e1000e/ptp.c
index 03d5f2a..5b368aa 100644
--- a/drivers/net/ethernet/intel/e1000e/ptp.c
+++ b/drivers/net/ethernet/intel/e1000e/ptp.c
@@ -222,7 +222,7 @@ static int e1000e_phc_settime(struct ptp_clock_info *ptp,
 
/* reset the timecounter */
spin_lock_irqsave(>systim_lock, flags);
-   timecounter_init(>tc, ns);
+   timecounter_reset(>tc, ns);
spin_unlock_irqrestore(>systim_lock, flags);
 
return 0;
-- 
1.9.1

[PATCH 20/27] e1000e: Use timecounter_reset interface

[Sagar Arun Kamble]

With new interface timecounter_reset we can update the start time for
timecounter. Update e1000e_phc_settime with this new function.

Signed-off-by: Sagar Arun Kamble 
Cc: Richard Cochran 
Cc: Jeff Kirsher 
Cc: intel-wired-...@lists.osuosl.org
Cc: net...@vger.kernel.org
Cc: linux-kernel@vger.kernel.org
---
 drivers/net/ethernet/intel/e1000e/ptp.c | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/drivers/net/ethernet/intel/e1000e/ptp.c 
b/drivers/net/ethernet/intel/e1000e/ptp.c
index 03d5f2a..5b368aa 100644
--- a/drivers/net/ethernet/intel/e1000e/ptp.c
+++ b/drivers/net/ethernet/intel/e1000e/ptp.c
@@ -222,7 +222,7 @@ static int e1000e_phc_settime(struct ptp_clock_info *ptp,
 
/* reset the timecounter */
spin_lock_irqsave(>systim_lock, flags);
-   timecounter_init(>tc, ns);
+   timecounter_reset(>tc, ns);
spin_unlock_irqrestore(>systim_lock, flags);
 
return 0;
-- 
1.9.1

[PATCH 17/27] amd-xgbe: Use timecounter_reset interface

[Sagar Arun Kamble]

With new interface timecounter_reset we can update the start time for
timecounter. Update xgbe_config_tstamp and xgbe_settime with this new
function.

Signed-off-by: Sagar Arun Kamble <sagar.a.kam...@intel.com>
Cc: Richard Cochran <richardcoch...@gmail.com>
Cc: Tom Lendacky <thomas.lenda...@amd.com>
Cc: net...@vger.kernel.org
Cc: linux-kernel@vger.kernel.org
---
 drivers/net/ethernet/amd/xgbe/xgbe-dev.c | 4 ++--
 drivers/net/ethernet/amd/xgbe/xgbe-ptp.c | 2 +-
 2 files changed, 3 insertions(+), 3 deletions(-)

diff --git a/drivers/net/ethernet/amd/xgbe/xgbe-dev.c 
b/drivers/net/ethernet/amd/xgbe/xgbe-dev.c
index 5005c87..aee99de 100644
--- a/drivers/net/ethernet/amd/xgbe/xgbe-dev.c
+++ b/drivers/net/ethernet/amd/xgbe/xgbe-dev.c
@@ -1621,8 +1621,8 @@ static int xgbe_config_tstamp(struct xgbe_prv_data *pdata,
xgbe_update_tstamp_addend(pdata, pdata->tstamp_addend);
xgbe_set_tstamp_time(pdata, 0, 0);
 
-   /* Initialize the timecounter */
-   timecounter_init(>tstamp_tc, ktime_to_ns(ktime_get_real()));
+   /* Reset the timecounter */
+   timecounter_reset(>tstamp_tc, ktime_to_ns(ktime_get_real()));
 
return 0;
 }
diff --git a/drivers/net/ethernet/amd/xgbe/xgbe-ptp.c 
b/drivers/net/ethernet/amd/xgbe/xgbe-ptp.c
index 486437b..5fcde50 100644
--- a/drivers/net/ethernet/amd/xgbe/xgbe-ptp.c
+++ b/drivers/net/ethernet/amd/xgbe/xgbe-ptp.c
@@ -211,7 +211,7 @@ static int xgbe_settime(struct ptp_clock_info *info,
 
spin_lock_irqsave(>tstamp_lock, flags);
 
-   timecounter_init(>tstamp_tc, nsec);
+   timecounter_reset(>tstamp_tc, nsec);
 
spin_unlock_irqrestore(>tstamp_lock, flags);
 
-- 
1.9.1

[PATCH 27/27] timecounter: Remove timecounter_init

[Sagar Arun Kamble]

With all timecounter users now initializing timecounter using
timecounter_initialize remove timecounter_init function.

Signed-off-by: Sagar Arun Kamble <sagar.a.kam...@intel.com>
Cc: Richard Cochran <richardcoch...@gmail.com>
Cc: John Stultz <john.stu...@linaro.org>
Cc: Thomas Gleixner <t...@linutronix.de>
Cc: Stephen Boyd <sb...@codeaurora.org>
Cc: linux-kernel@vger.kernel.org
---
 include/linux/timecounter.h | 20 +++-
 kernel/time/timecounter.c   | 16 
 2 files changed, 7 insertions(+), 29 deletions(-)

diff --git a/include/linux/timecounter.h b/include/linux/timecounter.h
index e0fd741..e3e1d1e 100644
--- a/include/linux/timecounter.h
+++ b/include/linux/timecounter.h
@@ -46,8 +46,7 @@ struct cyclecounter {
 /**
  * struct timecounter - layer above a %struct cyclecounter which counts 
nanoseconds
  * Contains the state needed by timecounter_read() to detect
- * cycle counter wrap around. Initialize with timecounter_init() when
- * underlying cyclecounter is initialized, with timecounter_initialize() to
+ * cycle counter wrap around. Initialize with timecounter_initialize() to
  * initialize cyclecounter and timecounter fields. Also used to convert
  * cycle counts into the corresponding nanosecond counts with
  * timecounter_cyc2time(). Users of this code are responsible for
@@ -97,18 +96,6 @@ static inline void timecounter_adjtime(struct timecounter 
*tc, s64 delta)
 }
 
 /**
- * timecounter_init - initialize a time counter
- * @tc:Pointer to time counter which is to be 
initialized
- * @start_tstamp:  Arbitrary initial time stamp.
- *
- * After this call the current cycle register (roughly) corresponds to
- * the initial time stamp. Every call to timecounter_read() increments
- * the time stamp counter by the number of elapsed nanoseconds.
- */
-extern void timecounter_init(struct timecounter *tc,
-u64 start_tstamp);
-
-/**
  * timecounter_initialize - initialize a time counter and underlying
cyclecounter
  * @tc:Pointer to time counter which is to be 
initialized
@@ -143,9 +130,8 @@ extern void timecounter_reset(struct timecounter *tc,
  u64 start_tstamp);
 
 /**
- * timecounter_read - return nanoseconds elapsed since timecounter_init() or
- *timecounter_initialize() or timecounter_reset() plus
- *   the initial time stamp
+ * timecounter_read - return nanoseconds elapsed since timecounter_initialize()
+ *   or timecounter_reset() plus the initial time stamp
  * @tc:  Pointer to time counter.
  *
  * In other words, keeps track of time since the same epoch as
diff --git a/kernel/time/timecounter.c b/kernel/time/timecounter.c
index 3f62fe0..0a02ef0 100644
--- a/kernel/time/timecounter.c
+++ b/kernel/time/timecounter.c
@@ -18,17 +18,6 @@
 #include 
 #include 
 
-void timecounter_init(struct timecounter *tc, u64 start_tstamp)
-{
-   struct cyclecounter *cc = >cc;
-
-   tc->cycle_last = cc->read(cc);
-   tc->nsec = start_tstamp;
-   tc->mask = (1ULL << cc->shift) - 1;
-   tc->frac = 0;
-}
-EXPORT_SYMBOL_GPL(timecounter_init);
-
 void timecounter_initialize(struct timecounter *tc,
u64 (*read)(const struct cyclecounter *cc),
u64 mask,
@@ -43,7 +32,10 @@ void timecounter_initialize(struct timecounter *tc,
cc->mult = mult;
cc->shift = shift;
 
-   timecounter_init(tc, start_tstamp);
+   tc->cycle_last = cc->read(cc);
+   tc->nsec = start_tstamp;
+   tc->mask = (1ULL << cc->shift) - 1;
+   tc->frac = 0;
 }
 EXPORT_SYMBOL_GPL(timecounter_initialize);
 
-- 
1.9.1

[PATCH 17/27] amd-xgbe: Use timecounter_reset interface

[Sagar Arun Kamble]

With new interface timecounter_reset we can update the start time for
timecounter. Update xgbe_config_tstamp and xgbe_settime with this new
function.

Signed-off-by: Sagar Arun Kamble 
Cc: Richard Cochran 
Cc: Tom Lendacky 
Cc: net...@vger.kernel.org
Cc: linux-kernel@vger.kernel.org
---
 drivers/net/ethernet/amd/xgbe/xgbe-dev.c | 4 ++--
 drivers/net/ethernet/amd/xgbe/xgbe-ptp.c | 2 +-
 2 files changed, 3 insertions(+), 3 deletions(-)

diff --git a/drivers/net/ethernet/amd/xgbe/xgbe-dev.c 
b/drivers/net/ethernet/amd/xgbe/xgbe-dev.c
index 5005c87..aee99de 100644
--- a/drivers/net/ethernet/amd/xgbe/xgbe-dev.c
+++ b/drivers/net/ethernet/amd/xgbe/xgbe-dev.c
@@ -1621,8 +1621,8 @@ static int xgbe_config_tstamp(struct xgbe_prv_data *pdata,
xgbe_update_tstamp_addend(pdata, pdata->tstamp_addend);
xgbe_set_tstamp_time(pdata, 0, 0);
 
-   /* Initialize the timecounter */
-   timecounter_init(>tstamp_tc, ktime_to_ns(ktime_get_real()));
+   /* Reset the timecounter */
+   timecounter_reset(>tstamp_tc, ktime_to_ns(ktime_get_real()));
 
return 0;
 }
diff --git a/drivers/net/ethernet/amd/xgbe/xgbe-ptp.c 
b/drivers/net/ethernet/amd/xgbe/xgbe-ptp.c
index 486437b..5fcde50 100644
--- a/drivers/net/ethernet/amd/xgbe/xgbe-ptp.c
+++ b/drivers/net/ethernet/amd/xgbe/xgbe-ptp.c
@@ -211,7 +211,7 @@ static int xgbe_settime(struct ptp_clock_info *info,
 
spin_lock_irqsave(>tstamp_lock, flags);
 
-   timecounter_init(>tstamp_tc, nsec);
+   timecounter_reset(>tstamp_tc, nsec);
 
spin_unlock_irqrestore(>tstamp_lock, flags);
 
-- 
1.9.1

[PATCH 27/27] timecounter: Remove timecounter_init

[Sagar Arun Kamble]

With all timecounter users now initializing timecounter using
timecounter_initialize remove timecounter_init function.

Signed-off-by: Sagar Arun Kamble 
Cc: Richard Cochran 
Cc: John Stultz 
Cc: Thomas Gleixner 
Cc: Stephen Boyd 
Cc: linux-kernel@vger.kernel.org
---
 include/linux/timecounter.h | 20 +++-
 kernel/time/timecounter.c   | 16 
 2 files changed, 7 insertions(+), 29 deletions(-)

diff --git a/include/linux/timecounter.h b/include/linux/timecounter.h
index e0fd741..e3e1d1e 100644
--- a/include/linux/timecounter.h
+++ b/include/linux/timecounter.h
@@ -46,8 +46,7 @@ struct cyclecounter {
 /**
  * struct timecounter - layer above a %struct cyclecounter which counts 
nanoseconds
  * Contains the state needed by timecounter_read() to detect
- * cycle counter wrap around. Initialize with timecounter_init() when
- * underlying cyclecounter is initialized, with timecounter_initialize() to
+ * cycle counter wrap around. Initialize with timecounter_initialize() to
  * initialize cyclecounter and timecounter fields. Also used to convert
  * cycle counts into the corresponding nanosecond counts with
  * timecounter_cyc2time(). Users of this code are responsible for
@@ -97,18 +96,6 @@ static inline void timecounter_adjtime(struct timecounter 
*tc, s64 delta)
 }
 
 /**
- * timecounter_init - initialize a time counter
- * @tc:Pointer to time counter which is to be 
initialized
- * @start_tstamp:  Arbitrary initial time stamp.
- *
- * After this call the current cycle register (roughly) corresponds to
- * the initial time stamp. Every call to timecounter_read() increments
- * the time stamp counter by the number of elapsed nanoseconds.
- */
-extern void timecounter_init(struct timecounter *tc,
-u64 start_tstamp);
-
-/**
  * timecounter_initialize - initialize a time counter and underlying
cyclecounter
  * @tc:Pointer to time counter which is to be 
initialized
@@ -143,9 +130,8 @@ extern void timecounter_reset(struct timecounter *tc,
  u64 start_tstamp);
 
 /**
- * timecounter_read - return nanoseconds elapsed since timecounter_init() or
- *timecounter_initialize() or timecounter_reset() plus
- *   the initial time stamp
+ * timecounter_read - return nanoseconds elapsed since timecounter_initialize()
+ *   or timecounter_reset() plus the initial time stamp
  * @tc:  Pointer to time counter.
  *
  * In other words, keeps track of time since the same epoch as
diff --git a/kernel/time/timecounter.c b/kernel/time/timecounter.c
index 3f62fe0..0a02ef0 100644
--- a/kernel/time/timecounter.c
+++ b/kernel/time/timecounter.c
@@ -18,17 +18,6 @@
 #include 
 #include 
 
-void timecounter_init(struct timecounter *tc, u64 start_tstamp)
-{
-   struct cyclecounter *cc = >cc;
-
-   tc->cycle_last = cc->read(cc);
-   tc->nsec = start_tstamp;
-   tc->mask = (1ULL << cc->shift) - 1;
-   tc->frac = 0;
-}
-EXPORT_SYMBOL_GPL(timecounter_init);
-
 void timecounter_initialize(struct timecounter *tc,
u64 (*read)(const struct cyclecounter *cc),
u64 mask,
@@ -43,7 +32,10 @@ void timecounter_initialize(struct timecounter *tc,
cc->mult = mult;
cc->shift = shift;
 
-   timecounter_init(tc, start_tstamp);
+   tc->cycle_last = cc->read(cc);
+   tc->nsec = start_tstamp;
+   tc->mask = (1ULL << cc->shift) - 1;
+   tc->frac = 0;
 }
 EXPORT_SYMBOL_GPL(timecounter_initialize);
 
-- 
1.9.1

[PATCH 19/27] net: fec: ptp: Use timecounter_reset interface

[Sagar Arun Kamble]

With new interface timecounter_reset we can update the start time for
timecounter. Update fec_ptp_settime with this new function.

Signed-off-by: Sagar Arun Kamble <sagar.a.kam...@intel.com>
Cc: Richard Cochran <richardcoch...@gmail.com>
Cc: Fugang Duan <fugang.d...@nxp.com>
Cc: net...@vger.kernel.org
Cc: linux-kernel@vger.kernel.org
---
 drivers/net/ethernet/freescale/fec_ptp.c | 3 +--
 1 file changed, 1 insertion(+), 2 deletions(-)

diff --git a/drivers/net/ethernet/freescale/fec_ptp.c 
b/drivers/net/ethernet/freescale/fec_ptp.c
index 1ba7216..d03ea0e 100644
--- a/drivers/net/ethernet/freescale/fec_ptp.c
+++ b/drivers/net/ethernet/freescale/fec_ptp.c
@@ -419,7 +419,6 @@ static int fec_ptp_settime(struct ptp_clock_info *ptp,
 {
struct fec_enet_private *fep =
container_of(ptp, struct fec_enet_private, ptp_caps);
-
u64 ns;
unsigned long flags;
u32 counter;
@@ -439,7 +438,7 @@ static int fec_ptp_settime(struct ptp_clock_info *ptp,
 
spin_lock_irqsave(>tmreg_lock, flags);
writel(counter, fep->hwp + FEC_ATIME);
-   timecounter_init(>tc, ns);
+   timecounter_reset(>tc, ns);
spin_unlock_irqrestore(>tmreg_lock, flags);
mutex_unlock(>ptp_clk_mutex);
return 0;
-- 
1.9.1

[PATCH 19/27] net: fec: ptp: Use timecounter_reset interface

[Sagar Arun Kamble]

With new interface timecounter_reset we can update the start time for
timecounter. Update fec_ptp_settime with this new function.

Signed-off-by: Sagar Arun Kamble 
Cc: Richard Cochran 
Cc: Fugang Duan 
Cc: net...@vger.kernel.org
Cc: linux-kernel@vger.kernel.org
---
 drivers/net/ethernet/freescale/fec_ptp.c | 3 +--
 1 file changed, 1 insertion(+), 2 deletions(-)

diff --git a/drivers/net/ethernet/freescale/fec_ptp.c 
b/drivers/net/ethernet/freescale/fec_ptp.c
index 1ba7216..d03ea0e 100644
--- a/drivers/net/ethernet/freescale/fec_ptp.c
+++ b/drivers/net/ethernet/freescale/fec_ptp.c
@@ -419,7 +419,6 @@ static int fec_ptp_settime(struct ptp_clock_info *ptp,
 {
struct fec_enet_private *fep =
container_of(ptp, struct fec_enet_private, ptp_caps);
-
u64 ns;
unsigned long flags;
u32 counter;
@@ -439,7 +438,7 @@ static int fec_ptp_settime(struct ptp_clock_info *ptp,
 
spin_lock_irqsave(>tmreg_lock, flags);
writel(counter, fep->hwp + FEC_ATIME);
-   timecounter_init(>tc, ns);
+   timecounter_reset(>tc, ns);
spin_unlock_irqrestore(>tmreg_lock, flags);
mutex_unlock(>ptp_clk_mutex);
return 0;
-- 
1.9.1

[PATCH 13/27] qede: Use timecounter_initialize interface

[Sagar Arun Kamble]

With new interface timecounter_initialize we can initialize timecounter
fields and underlying cyclecounter together. Update qede timecounter
init with this new function.

Signed-off-by: Sagar Arun Kamble <sagar.a.kam...@intel.com>
Cc: Richard Cochran <richardcoch...@gmail.com>
Cc: Ariel Elior <ariel.el...@cavium.com>
Cc: everest-linux...@cavium.com
Cc: net...@vger.kernel.org
Cc: linux-kernel@vger.kernel.org
---
 drivers/net/ethernet/qlogic/qede/qede_ptp.c | 13 ++---
 1 file changed, 6 insertions(+), 7 deletions(-)

diff --git a/drivers/net/ethernet/qlogic/qede/qede_ptp.c 
b/drivers/net/ethernet/qlogic/qede/qede_ptp.c
index 95bb8a8..0d054dc 100644
--- a/drivers/net/ethernet/qlogic/qede/qede_ptp.c
+++ b/drivers/net/ethernet/qlogic/qede/qede_ptp.c
@@ -427,13 +427,12 @@ static int qede_ptp_init(struct qede_dev *edev, bool 
init_tc)
 * unload / load (e.g. MTU change) while it is running.
 */
if (init_tc) {
-   memset(>tc.cc, 0, sizeof(ptp->tc.cc));
-   ptp->tc.cc.read = qede_ptp_read_cc;
-   ptp->tc.cc.mask = CYCLECOUNTER_MASK(64);
-   ptp->tc.cc.shift = 0;
-   ptp->tc.cc.mult = 1;
-
-   timecounter_init(>tc, ktime_to_ns(ktime_get_real()));
+   timecounter_initialize(>tc,
+  qede_ptp_read_cc,
+  CYCLECOUNTER_MASK(64),
+  1,
+  0,
+  ktime_to_ns(ktime_get_real()));
}
 
return rc;
-- 
1.9.1

[PATCH 13/27] qede: Use timecounter_initialize interface

[Sagar Arun Kamble]

With new interface timecounter_initialize we can initialize timecounter
fields and underlying cyclecounter together. Update qede timecounter
init with this new function.

Signed-off-by: Sagar Arun Kamble 
Cc: Richard Cochran 
Cc: Ariel Elior 
Cc: everest-linux...@cavium.com
Cc: net...@vger.kernel.org
Cc: linux-kernel@vger.kernel.org
---
 drivers/net/ethernet/qlogic/qede/qede_ptp.c | 13 ++---
 1 file changed, 6 insertions(+), 7 deletions(-)

diff --git a/drivers/net/ethernet/qlogic/qede/qede_ptp.c 
b/drivers/net/ethernet/qlogic/qede/qede_ptp.c
index 95bb8a8..0d054dc 100644
--- a/drivers/net/ethernet/qlogic/qede/qede_ptp.c
+++ b/drivers/net/ethernet/qlogic/qede/qede_ptp.c
@@ -427,13 +427,12 @@ static int qede_ptp_init(struct qede_dev *edev, bool 
init_tc)
 * unload / load (e.g. MTU change) while it is running.
 */
if (init_tc) {
-   memset(>tc.cc, 0, sizeof(ptp->tc.cc));
-   ptp->tc.cc.read = qede_ptp_read_cc;
-   ptp->tc.cc.mask = CYCLECOUNTER_MASK(64);
-   ptp->tc.cc.shift = 0;
-   ptp->tc.cc.mult = 1;
-
-   timecounter_init(>tc, ktime_to_ns(ktime_get_real()));
+   timecounter_initialize(>tc,
+  qede_ptp_read_cc,
+  CYCLECOUNTER_MASK(64),
+  1,
+  0,
+  ktime_to_ns(ktime_get_real()));
}
 
return rc;
-- 
1.9.1

[PATCH 10/27] ixgbe: Use timecounter_initialize interface

[Sagar Arun Kamble]

With new interface timecounter_initialize we can initialize timecounter
fields and underlying cyclecounter together. Update ixgbe ptp timecounter
init with this new function.

Signed-off-by: Sagar Arun Kamble <sagar.a.kam...@intel.com>
Cc: Richard Cochran <richardcoch...@gmail.com>
Cc: Jeff Kirsher <jeffrey.t.kirs...@intel.com>
Cc: intel-wired-...@lists.osuosl.org
Cc: net...@vger.kernel.org
Cc: linux-kernel@vger.kernel.org
---
 drivers/net/ethernet/intel/ixgbe/ixgbe.h  |  6 +++-
 drivers/net/ethernet/intel/ixgbe/ixgbe_main.c |  4 +--
 drivers/net/ethernet/intel/ixgbe/ixgbe_ptp.c  | 45 +++
 3 files changed, 32 insertions(+), 23 deletions(-)

diff --git a/drivers/net/ethernet/intel/ixgbe/ixgbe.h 
b/drivers/net/ethernet/intel/ixgbe/ixgbe.h
index 5c391a0..67e8b5c 100644
--- a/drivers/net/ethernet/intel/ixgbe/ixgbe.h
+++ b/drivers/net/ethernet/intel/ixgbe/ixgbe.h
@@ -739,6 +739,10 @@ struct ixgbe_adapter {
unsigned long last_rx_timestamp;
spinlock_t tmreg_lock;
struct timecounter hw_tc;
+   u64 (*cc_read)(const struct cyclecounter *cc);
+   u64 cc_mask;
+   u32 cc_mult;
+   u32 cc_shift;
u32 base_incval;
u32 tx_hwtstamp_timeouts;
u32 tx_hwtstamp_skipped;
@@ -994,7 +998,7 @@ static inline void ixgbe_ptp_rx_hwtstamp(struct ixgbe_ring 
*rx_ring,
 
 int ixgbe_ptp_set_ts_config(struct ixgbe_adapter *adapter, struct ifreq *ifr);
 int ixgbe_ptp_get_ts_config(struct ixgbe_adapter *adapter, struct ifreq *ifr);
-void ixgbe_ptp_start_cyclecounter(struct ixgbe_adapter *adapter);
+void ixgbe_ptp_start_timecounter(struct ixgbe_adapter *adapter);
 void ixgbe_ptp_reset(struct ixgbe_adapter *adapter);
 void ixgbe_ptp_check_pps_event(struct ixgbe_adapter *adapter);
 #ifdef CONFIG_PCI_IOV
diff --git a/drivers/net/ethernet/intel/ixgbe/ixgbe_main.c 
b/drivers/net/ethernet/intel/ixgbe/ixgbe_main.c
index 62a1891..86a337b 100644
--- a/drivers/net/ethernet/intel/ixgbe/ixgbe_main.c
+++ b/drivers/net/ethernet/intel/ixgbe/ixgbe_main.c
@@ -7332,7 +7332,7 @@ static void ixgbe_watchdog_link_is_up(struct 
ixgbe_adapter *adapter)
adapter->last_rx_ptp_check = jiffies;
 
if (test_bit(__IXGBE_PTP_RUNNING, >state))
-   ixgbe_ptp_start_cyclecounter(adapter);
+   ixgbe_ptp_start_timecounter(adapter);
 
switch (link_speed) {
case IXGBE_LINK_SPEED_10GB_FULL:
@@ -7400,7 +7400,7 @@ static void ixgbe_watchdog_link_is_down(struct 
ixgbe_adapter *adapter)
adapter->flags2 |= IXGBE_FLAG2_SEARCH_FOR_SFP;
 
if (test_bit(__IXGBE_PTP_RUNNING, >state))
-   ixgbe_ptp_start_cyclecounter(adapter);
+   ixgbe_ptp_start_timecounter(adapter);
 
e_info(drv, "NIC Link is Down\n");
netif_carrier_off(netdev);
diff --git a/drivers/net/ethernet/intel/ixgbe/ixgbe_ptp.c 
b/drivers/net/ethernet/intel/ixgbe/ixgbe_ptp.c
index 6e9f2c0..073c1ef 100644
--- a/drivers/net/ethernet/intel/ixgbe/ixgbe_ptp.c
+++ b/drivers/net/ethernet/intel/ixgbe/ixgbe_ptp.c
@@ -1075,7 +1075,8 @@ static void ixgbe_ptp_link_speed_adjust(struct 
ixgbe_adapter *adapter,
 }
 
 /**
- * ixgbe_ptp_start_cyclecounter - create the cycle counter from hw
+ * ixgbe_ptp_start_timecounter - create the cycle counter from hw and
+ *  initialize corresponding timecounter.
  * @adapter: pointer to the adapter structure
  *
  * This function should be called to set the proper values for the TIMINCA
@@ -1084,10 +1085,9 @@ static void ixgbe_ptp_link_speed_adjust(struct 
ixgbe_adapter *adapter,
  * structure. It should be called whenever a new TIMINCA value is necessary,
  * such as during initialization or when the link speed changes.
  */
-void ixgbe_ptp_start_cyclecounter(struct ixgbe_adapter *adapter)
+void ixgbe_ptp_start_timecounter(struct ixgbe_adapter *adapter)
 {
struct ixgbe_hw *hw = >hw;
-   struct cyclecounter cc;
unsigned long flags;
u32 incval = 0;
u32 tsauxc = 0;
@@ -1104,9 +1104,9 @@ void ixgbe_ptp_start_cyclecounter(struct ixgbe_adapter 
*adapter)
 * proper fix to this problem would require modification of the
 * timecounter delta calculations.
 */
-   cc.mask = CLOCKSOURCE_MASK(64);
-   cc.mult = 1;
-   cc.shift = 0;
+   adapter->cc_mask = CLOCKSOURCE_MASK(64);
+   adapter->cc_mult = 1;
+   adapter->cc_shift = 0;
 
switch (hw->mac.type) {
case ixgbe_mac_X550EM_x:
@@ -1118,13 +1118,13 @@ void ixgbe_ptp_start_cyclecounter(struct ixgbe_adapter 
*adapter)
 */
fuse0 = IXGBE_READ_REG(hw, IXGBE_FUSES0_GROUP(0));
if (!(fuse0 & IXGBE_FUSES0_300MHZ)) {
-   cc.mult = 3;
-   cc.shift = 2;
+   adapter->cc_mult = 3;
+   adapter->cc_shift = 2;
}
/* fallthroug

[PATCH 02/27] timecounter: Introduce timecounter_initialize to update timecounter and cyclecounter

[Sagar Arun Kamble]

With cyclecounter coupled with timecounter, we should move to use
interface that initializes entire timecounter structure. This patch
creates function timecounter_initialize that takes cyclecounter
parameters and start time and initializes the timecounter and underlying
cyclecounter.
Function timecounter_init which requires initialized cyclecounter can be
removed once all drivers are migrated to this new interface.

Suggested-by: Richard Cochran <richardcoch...@gmail.com>
Signed-off-by: Sagar Arun Kamble <sagar.a.kam...@intel.com>
Cc: Richard Cochran <richardcoch...@gmail.com>
Cc: Chris Wilson <ch...@chris-wilson.co.uk>
Cc: John Stultz <john.stu...@linaro.org>
Cc: Thomas Gleixner <t...@linutronix.de>
Cc: Stephen Boyd <sb...@codeaurora.org>
Cc: linux-kernel@vger.kernel.org
---
 include/linux/timecounter.h | 41 -
 kernel/time/timecounter.c   | 18 ++
 2 files changed, 50 insertions(+), 9 deletions(-)

diff --git a/include/linux/timecounter.h b/include/linux/timecounter.h
index 6daca06..59d3fd7 100644
--- a/include/linux/timecounter.h
+++ b/include/linux/timecounter.h
@@ -46,13 +46,14 @@ struct cyclecounter {
 /**
  * struct timecounter - layer above a %struct cyclecounter which counts 
nanoseconds
  * Contains the state needed by timecounter_read() to detect
- * cycle counter wrap around. Initialize with
- * timecounter_init(). Also used to convert cycle counts into the
- * corresponding nanosecond counts with timecounter_cyc2time(). Users
- * of this code are responsible for initializing the underlying
- * cycle counter hardware, locking issues and reading the time
- * more often than the cycle counter wraps around. The nanosecond
- * counter will only wrap around after ~585 years.
+ * cycle counter wrap around. Initialize with timecounter_init() when
+ * underlying cyclecounter is initialized, with timecounter_initialize() to
+ * initialize cyclecounter and timecounter fields. Also used to convert
+ * cycle counts into the corresponding nanosecond counts with
+ * timecounter_cyc2time(). Users of this code are responsible for
+ * locking issues and reading the time more often than the cycle counter
+ * wraps around. The nanosecond counter will only wrap around after ~585
+ * years.
  *
  * @cc:the cycle counter used by this instance
  * @cycle_last:most recent cycle counter value seen by
@@ -108,8 +109,30 @@ extern void timecounter_init(struct timecounter *tc,
 u64 start_tstamp);
 
 /**
- * timecounter_read - return nanoseconds elapsed since timecounter_init()
- *plus the initial time stamp
+ * timecounter_initialize - initialize a time counter and underlying
+   cyclecounter
+ * @tc:Pointer to time counter which is to be 
initialized
+ * @read:  Pointer to function that returns the current cycle value
+ * @mask:  bitmask for two's complement
+ * subtraction of non 64 bit counters,
+ * @mult:  cycle to nanosecond multiplier
+ * @shift: cycle to nanosecond divisor (power of two)
+ * @start_tstamp:  Arbitrary initial time stamp.
+ *
+ * After this call the current cycle register (roughly) corresponds to
+ * the initial time stamp. Every call to timecounter_read() increments
+ * the time stamp counter by the number of elapsed nanoseconds.
+ */
+extern void timecounter_initialize(struct timecounter *tc,
+  u64 (*read)(const struct cyclecounter *cc),
+  u64 mask,
+  u32 mult,
+  u32 shift,
+  u64 start_tstamp);
+
+/**
+ * timecounter_read - return nanoseconds elapsed since timecounter_init() or
+ *timecounter_initialize() plus the initial time stamp
  * @tc:  Pointer to time counter.
  *
  * In other words, keeps track of time since the same epoch as
diff --git a/kernel/time/timecounter.c b/kernel/time/timecounter.c
index 7919acb..6d915752 100644
--- a/kernel/time/timecounter.c
+++ b/kernel/time/timecounter.c
@@ -29,6 +29,24 @@ void timecounter_init(struct timecounter *tc, u64 
start_tstamp)
 }
 EXPORT_SYMBOL_GPL(timecounter_init);
 
+void timecounter_initialize(struct timecounter *tc,
+   u64 (*read)(const struct cyclecounter *cc),
+   u64 mask,
+   u32 mult,
+   u32 shift,
+   u64 start_tstamp)
+{
+   struct cyclecounter *cc = >cc;
+
+   cc->read = read;
+   cc->mask = mask;
+   cc->mult = mult;
+   cc->shift = shift;
+
+   timecounter_init(tc, start_tstamp);
+}
+EXPORT_SYMBOL_GPL(timecounter_initialize);

[PATCH 10/27] ixgbe: Use timecounter_initialize interface

[Sagar Arun Kamble]

With new interface timecounter_initialize we can initialize timecounter
fields and underlying cyclecounter together. Update ixgbe ptp timecounter
init with this new function.

Signed-off-by: Sagar Arun Kamble 
Cc: Richard Cochran 
Cc: Jeff Kirsher 
Cc: intel-wired-...@lists.osuosl.org
Cc: net...@vger.kernel.org
Cc: linux-kernel@vger.kernel.org
---
 drivers/net/ethernet/intel/ixgbe/ixgbe.h  |  6 +++-
 drivers/net/ethernet/intel/ixgbe/ixgbe_main.c |  4 +--
 drivers/net/ethernet/intel/ixgbe/ixgbe_ptp.c  | 45 +++
 3 files changed, 32 insertions(+), 23 deletions(-)

diff --git a/drivers/net/ethernet/intel/ixgbe/ixgbe.h 
b/drivers/net/ethernet/intel/ixgbe/ixgbe.h
index 5c391a0..67e8b5c 100644
--- a/drivers/net/ethernet/intel/ixgbe/ixgbe.h
+++ b/drivers/net/ethernet/intel/ixgbe/ixgbe.h
@@ -739,6 +739,10 @@ struct ixgbe_adapter {
unsigned long last_rx_timestamp;
spinlock_t tmreg_lock;
struct timecounter hw_tc;
+   u64 (*cc_read)(const struct cyclecounter *cc);
+   u64 cc_mask;
+   u32 cc_mult;
+   u32 cc_shift;
u32 base_incval;
u32 tx_hwtstamp_timeouts;
u32 tx_hwtstamp_skipped;
@@ -994,7 +998,7 @@ static inline void ixgbe_ptp_rx_hwtstamp(struct ixgbe_ring 
*rx_ring,
 
 int ixgbe_ptp_set_ts_config(struct ixgbe_adapter *adapter, struct ifreq *ifr);
 int ixgbe_ptp_get_ts_config(struct ixgbe_adapter *adapter, struct ifreq *ifr);
-void ixgbe_ptp_start_cyclecounter(struct ixgbe_adapter *adapter);
+void ixgbe_ptp_start_timecounter(struct ixgbe_adapter *adapter);
 void ixgbe_ptp_reset(struct ixgbe_adapter *adapter);
 void ixgbe_ptp_check_pps_event(struct ixgbe_adapter *adapter);
 #ifdef CONFIG_PCI_IOV
diff --git a/drivers/net/ethernet/intel/ixgbe/ixgbe_main.c 
b/drivers/net/ethernet/intel/ixgbe/ixgbe_main.c
index 62a1891..86a337b 100644
--- a/drivers/net/ethernet/intel/ixgbe/ixgbe_main.c
+++ b/drivers/net/ethernet/intel/ixgbe/ixgbe_main.c
@@ -7332,7 +7332,7 @@ static void ixgbe_watchdog_link_is_up(struct 
ixgbe_adapter *adapter)
adapter->last_rx_ptp_check = jiffies;
 
if (test_bit(__IXGBE_PTP_RUNNING, >state))
-   ixgbe_ptp_start_cyclecounter(adapter);
+   ixgbe_ptp_start_timecounter(adapter);
 
switch (link_speed) {
case IXGBE_LINK_SPEED_10GB_FULL:
@@ -7400,7 +7400,7 @@ static void ixgbe_watchdog_link_is_down(struct 
ixgbe_adapter *adapter)
adapter->flags2 |= IXGBE_FLAG2_SEARCH_FOR_SFP;
 
if (test_bit(__IXGBE_PTP_RUNNING, >state))
-   ixgbe_ptp_start_cyclecounter(adapter);
+   ixgbe_ptp_start_timecounter(adapter);
 
e_info(drv, "NIC Link is Down\n");
netif_carrier_off(netdev);
diff --git a/drivers/net/ethernet/intel/ixgbe/ixgbe_ptp.c 
b/drivers/net/ethernet/intel/ixgbe/ixgbe_ptp.c
index 6e9f2c0..073c1ef 100644
--- a/drivers/net/ethernet/intel/ixgbe/ixgbe_ptp.c
+++ b/drivers/net/ethernet/intel/ixgbe/ixgbe_ptp.c
@@ -1075,7 +1075,8 @@ static void ixgbe_ptp_link_speed_adjust(struct 
ixgbe_adapter *adapter,
 }
 
 /**
- * ixgbe_ptp_start_cyclecounter - create the cycle counter from hw
+ * ixgbe_ptp_start_timecounter - create the cycle counter from hw and
+ *  initialize corresponding timecounter.
  * @adapter: pointer to the adapter structure
  *
  * This function should be called to set the proper values for the TIMINCA
@@ -1084,10 +1085,9 @@ static void ixgbe_ptp_link_speed_adjust(struct 
ixgbe_adapter *adapter,
  * structure. It should be called whenever a new TIMINCA value is necessary,
  * such as during initialization or when the link speed changes.
  */
-void ixgbe_ptp_start_cyclecounter(struct ixgbe_adapter *adapter)
+void ixgbe_ptp_start_timecounter(struct ixgbe_adapter *adapter)
 {
struct ixgbe_hw *hw = >hw;
-   struct cyclecounter cc;
unsigned long flags;
u32 incval = 0;
u32 tsauxc = 0;
@@ -1104,9 +1104,9 @@ void ixgbe_ptp_start_cyclecounter(struct ixgbe_adapter 
*adapter)
 * proper fix to this problem would require modification of the
 * timecounter delta calculations.
 */
-   cc.mask = CLOCKSOURCE_MASK(64);
-   cc.mult = 1;
-   cc.shift = 0;
+   adapter->cc_mask = CLOCKSOURCE_MASK(64);
+   adapter->cc_mult = 1;
+   adapter->cc_shift = 0;
 
switch (hw->mac.type) {
case ixgbe_mac_X550EM_x:
@@ -1118,13 +1118,13 @@ void ixgbe_ptp_start_cyclecounter(struct ixgbe_adapter 
*adapter)
 */
fuse0 = IXGBE_READ_REG(hw, IXGBE_FUSES0_GROUP(0));
if (!(fuse0 & IXGBE_FUSES0_300MHZ)) {
-   cc.mult = 3;
-   cc.shift = 2;
+   adapter->cc_mult = 3;
+   adapter->cc_shift = 2;
}
/* fallthrough */
case ixgbe_mac_x550em_a:
case ixgbe_mac_X550:
-   cc.read = ixgbe_

[PATCH 02/27] timecounter: Introduce timecounter_initialize to update timecounter and cyclecounter

[Sagar Arun Kamble]

With cyclecounter coupled with timecounter, we should move to use
interface that initializes entire timecounter structure. This patch
creates function timecounter_initialize that takes cyclecounter
parameters and start time and initializes the timecounter and underlying
cyclecounter.
Function timecounter_init which requires initialized cyclecounter can be
removed once all drivers are migrated to this new interface.

Suggested-by: Richard Cochran 
Signed-off-by: Sagar Arun Kamble 
Cc: Richard Cochran 
Cc: Chris Wilson 
Cc: John Stultz 
Cc: Thomas Gleixner 
Cc: Stephen Boyd 
Cc: linux-kernel@vger.kernel.org
---
 include/linux/timecounter.h | 41 -
 kernel/time/timecounter.c   | 18 ++
 2 files changed, 50 insertions(+), 9 deletions(-)

diff --git a/include/linux/timecounter.h b/include/linux/timecounter.h
index 6daca06..59d3fd7 100644
--- a/include/linux/timecounter.h
+++ b/include/linux/timecounter.h
@@ -46,13 +46,14 @@ struct cyclecounter {
 /**
  * struct timecounter - layer above a %struct cyclecounter which counts 
nanoseconds
  * Contains the state needed by timecounter_read() to detect
- * cycle counter wrap around. Initialize with
- * timecounter_init(). Also used to convert cycle counts into the
- * corresponding nanosecond counts with timecounter_cyc2time(). Users
- * of this code are responsible for initializing the underlying
- * cycle counter hardware, locking issues and reading the time
- * more often than the cycle counter wraps around. The nanosecond
- * counter will only wrap around after ~585 years.
+ * cycle counter wrap around. Initialize with timecounter_init() when
+ * underlying cyclecounter is initialized, with timecounter_initialize() to
+ * initialize cyclecounter and timecounter fields. Also used to convert
+ * cycle counts into the corresponding nanosecond counts with
+ * timecounter_cyc2time(). Users of this code are responsible for
+ * locking issues and reading the time more often than the cycle counter
+ * wraps around. The nanosecond counter will only wrap around after ~585
+ * years.
  *
  * @cc:the cycle counter used by this instance
  * @cycle_last:most recent cycle counter value seen by
@@ -108,8 +109,30 @@ extern void timecounter_init(struct timecounter *tc,
 u64 start_tstamp);
 
 /**
- * timecounter_read - return nanoseconds elapsed since timecounter_init()
- *plus the initial time stamp
+ * timecounter_initialize - initialize a time counter and underlying
+   cyclecounter
+ * @tc:Pointer to time counter which is to be 
initialized
+ * @read:  Pointer to function that returns the current cycle value
+ * @mask:  bitmask for two's complement
+ * subtraction of non 64 bit counters,
+ * @mult:  cycle to nanosecond multiplier
+ * @shift: cycle to nanosecond divisor (power of two)
+ * @start_tstamp:  Arbitrary initial time stamp.
+ *
+ * After this call the current cycle register (roughly) corresponds to
+ * the initial time stamp. Every call to timecounter_read() increments
+ * the time stamp counter by the number of elapsed nanoseconds.
+ */
+extern void timecounter_initialize(struct timecounter *tc,
+  u64 (*read)(const struct cyclecounter *cc),
+  u64 mask,
+  u32 mult,
+  u32 shift,
+  u64 start_tstamp);
+
+/**
+ * timecounter_read - return nanoseconds elapsed since timecounter_init() or
+ *timecounter_initialize() plus the initial time stamp
  * @tc:  Pointer to time counter.
  *
  * In other words, keeps track of time since the same epoch as
diff --git a/kernel/time/timecounter.c b/kernel/time/timecounter.c
index 7919acb..6d915752 100644
--- a/kernel/time/timecounter.c
+++ b/kernel/time/timecounter.c
@@ -29,6 +29,24 @@ void timecounter_init(struct timecounter *tc, u64 
start_tstamp)
 }
 EXPORT_SYMBOL_GPL(timecounter_init);
 
+void timecounter_initialize(struct timecounter *tc,
+   u64 (*read)(const struct cyclecounter *cc),
+   u64 mask,
+   u32 mult,
+   u32 shift,
+   u64 start_tstamp)
+{
+   struct cyclecounter *cc = >cc;
+
+   cc->read = read;
+   cc->mask = mask;
+   cc->mult = mult;
+   cc->shift = shift;
+
+   timecounter_init(tc, start_tstamp);
+}
+EXPORT_SYMBOL_GPL(timecounter_initialize);
+
 /**
  * timecounter_read_delta - get nanoseconds since last call of this function
  * @tc: Pointer to time counter
-- 
1.9.1

[PATCH 09/27] igb: Use timecounter_initialize interface

[Sagar Arun Kamble]

With new interface timecounter_initialize we can initialize timecounter
fields and underlying cyclecounter together. Update igb ptp timecounter
init with this new function.

Signed-off-by: Sagar Arun Kamble <sagar.a.kam...@intel.com>
Cc: Richard Cochran <richardcoch...@gmail.com>
Cc: Jeff Kirsher <jeffrey.t.kirs...@intel.com>
Cc: intel-wired-...@lists.osuosl.org
Cc: net...@vger.kernel.org
Cc: linux-kernel@vger.kernel.org
---
 drivers/net/ethernet/intel/igb/igb.h |  4 
 drivers/net/ethernet/intel/igb/igb_ptp.c | 23 ++-
 2 files changed, 18 insertions(+), 9 deletions(-)

diff --git a/drivers/net/ethernet/intel/igb/igb.h 
b/drivers/net/ethernet/intel/igb/igb.h
index 4eac4f2..4ef5632 100644
--- a/drivers/net/ethernet/intel/igb/igb.h
+++ b/drivers/net/ethernet/intel/igb/igb.h
@@ -566,6 +566,10 @@ struct igb_adapter {
unsigned int ptp_flags;
spinlock_t tmreg_lock;
struct timecounter tc;
+   u64 (*cc_read)(const struct cyclecounter *cc);
+   u64 cc_mask;
+   u32 cc_mult;
+   u32 cc_shift;
u32 tx_hwtstamp_timeouts;
u32 tx_hwtstamp_skipped;
u32 rx_hwtstamp_cleared;
diff --git a/drivers/net/ethernet/intel/igb/igb_ptp.c 
b/drivers/net/ethernet/intel/igb/igb_ptp.c
index 0745eff..f6c1a8d 100644
--- a/drivers/net/ethernet/intel/igb/igb_ptp.c
+++ b/drivers/net/ethernet/intel/igb/igb_ptp.c
@@ -1126,10 +1126,10 @@ void igb_ptp_init(struct igb_adapter *adapter)
adapter->ptp_caps.gettime64 = igb_ptp_gettime_82576;
adapter->ptp_caps.settime64 = igb_ptp_settime_82576;
adapter->ptp_caps.enable = igb_ptp_feature_enable;
-   adapter->tc.cc.read = igb_ptp_read_82576;
-   adapter->tc.cc.mask = CYCLECOUNTER_MASK(64);
-   adapter->tc.cc.mult = 1;
-   adapter->tc.cc.shift = IGB_82576_TSYNC_SHIFT;
+   adapter->cc_read = igb_ptp_read_82576;
+   adapter->cc_mask = CYCLECOUNTER_MASK(64);
+   adapter->cc_mult = 1;
+   adapter->cc_shift = IGB_82576_TSYNC_SHIFT;
adapter->ptp_flags |= IGB_PTP_OVERFLOW_CHECK;
break;
case e1000_82580:
@@ -1145,10 +1145,10 @@ void igb_ptp_init(struct igb_adapter *adapter)
adapter->ptp_caps.gettime64 = igb_ptp_gettime_82576;
adapter->ptp_caps.settime64 = igb_ptp_settime_82576;
adapter->ptp_caps.enable = igb_ptp_feature_enable;
-   adapter->tc.cc.read = igb_ptp_read_82580;
-   adapter->tc.cc.mask = CYCLECOUNTER_MASK(IGB_NBITS_82580);
-   adapter->tc.cc.mult = 1;
-   adapter->tc.cc.shift = 0;
+   adapter->cc_read = igb_ptp_read_82580;
+   adapter->cc_mask = CYCLECOUNTER_MASK(IGB_NBITS_82580);
+   adapter->cc_mult = 1;
+   adapter->cc_shift = 0;
adapter->ptp_flags |= IGB_PTP_OVERFLOW_CHECK;
break;
case e1000_i210:
@@ -1289,7 +1289,12 @@ void igb_ptp_reset(struct igb_adapter *adapter)
 
igb_ptp_write_i210(adapter, );
} else {
-   timecounter_init(>tc, ktime_to_ns(ktime_get_real()));
+   timecounter_initialize(>tc,
+  adapter->cc_read,
+  adapter->cc_mask,
+  adapter->cc_mult,
+  adapter->cc_shift,
+  ktime_to_ns(ktime_get_real()));
}
 out:
spin_unlock_irqrestore(>tmreg_lock, flags);
-- 
1.9.1

[PATCH 01/27] timecounter: Make cyclecounter struct part of timecounter struct

[Sagar Arun Kamble]

There is no real need for the users of timecounters to define cyclecounter
and timecounter variables separately. Since timecounter will always be
based on cyclecounter, have cyclecounter struct as member of timecounter
struct.

v2: Rebase.

Suggested-by: Chris Wilson <ch...@chris-wilson.co.uk>
Signed-off-by: Sagar Arun Kamble <sagar.a.kam...@intel.com>
Cc: Chris Wilson <ch...@chris-wilson.co.uk>
Cc: Richard Cochran <richardcoch...@gmail.com>
Cc: John Stultz <john.stu...@linaro.org>
Cc: Thomas Gleixner <t...@linutronix.de>
Cc: Stephen Boyd <sb...@codeaurora.org>
Cc: linux-kernel@vger.kernel.org
Cc: linux-arm-ker...@lists.infradead.org
Cc: net...@vger.kernel.org
Cc: intel-wired-...@lists.osuosl.org
Cc: linux-r...@vger.kernel.org
Cc: alsa-de...@alsa-project.org
Cc: kvm...@lists.cs.columbia.edu
Acked-by: Jeff Kirsher <jeffrey.t.kirs...@intel.com> (Intel drivers)
---
 arch/microblaze/kernel/timer.c | 20 ++--
 drivers/clocksource/arm_arch_timer.c   | 19 ++--
 drivers/net/ethernet/amd/xgbe/xgbe-dev.c   |  3 +-
 drivers/net/ethernet/amd/xgbe/xgbe-ptp.c   |  9 +++---
 drivers/net/ethernet/amd/xgbe/xgbe.h   |  1 -
 drivers/net/ethernet/broadcom/bnx2x/bnx2x.h|  1 -
 drivers/net/ethernet/broadcom/bnx2x/bnx2x_main.c   | 20 ++--
 drivers/net/ethernet/freescale/fec.h   |  1 -
 drivers/net/ethernet/freescale/fec_ptp.c   | 30 +-
 drivers/net/ethernet/intel/e1000e/e1000.h  |  1 -
 drivers/net/ethernet/intel/e1000e/netdev.c | 27 
 drivers/net/ethernet/intel/e1000e/ptp.c|  2 +-
 drivers/net/ethernet/intel/igb/igb.h   |  1 -
 drivers/net/ethernet/intel/igb/igb_ptp.c   | 25 ---
 drivers/net/ethernet/intel/ixgbe/ixgbe.h   |  1 -
 drivers/net/ethernet/intel/ixgbe/ixgbe_ptp.c   | 17 +-
 drivers/net/ethernet/mellanox/mlx4/en_clock.c  | 28 -
 drivers/net/ethernet/mellanox/mlx4/mlx4_en.h   |  1 -
 .../net/ethernet/mellanox/mlx5/core/lib/clock.c| 34 ++--
 drivers/net/ethernet/qlogic/qede/qede_ptp.c| 20 ++--
 drivers/net/ethernet/ti/cpts.c | 36 --
 drivers/net/ethernet/ti/cpts.h |  1 -
 include/linux/mlx5/driver.h|  1 -
 include/linux/timecounter.h|  4 +--
 include/sound/hdaudio.h|  1 -
 kernel/time/timecounter.c  | 28 -
 sound/hda/hdac_stream.c|  7 +++--
 virt/kvm/arm/arch_timer.c  |  6 ++--
 28 files changed, 163 insertions(+), 182 deletions(-)

diff --git a/arch/microblaze/kernel/timer.c b/arch/microblaze/kernel/timer.c
index 7de941c..b7f89e9 100644
--- a/arch/microblaze/kernel/timer.c
+++ b/arch/microblaze/kernel/timer.c
@@ -199,27 +199,25 @@ static u64 xilinx_read(struct clocksource *cs)
return (u64)xilinx_clock_read();
 }
 
-static struct timecounter xilinx_tc = {
-   .cc = NULL,
-};
-
 static u64 xilinx_cc_read(const struct cyclecounter *cc)
 {
return xilinx_read(NULL);
 }
 
-static struct cyclecounter xilinx_cc = {
-   .read = xilinx_cc_read,
-   .mask = CLOCKSOURCE_MASK(32),
-   .shift = 8,
+static struct timecounter xilinx_tc = {
+   .cc.read = xilinx_cc_read,
+   .cc.mask = CLOCKSOURCE_MASK(32),
+   .cc.mult = 0,
+   .cc.shift = 8,
 };
 
 static int __init init_xilinx_timecounter(void)
 {
-   xilinx_cc.mult = div_sc(timer_clock_freq, NSEC_PER_SEC,
-   xilinx_cc.shift);
+   struct cyclecounter *cc = _tc.cc;
+
+   cc->mult = div_sc(timer_clock_freq, NSEC_PER_SEC, cc->shift);
 
-   timecounter_init(_tc, _cc, sched_clock());
+   timecounter_init(_tc, sched_clock());
 
return 0;
 }
diff --git a/drivers/clocksource/arm_arch_timer.c 
b/drivers/clocksource/arm_arch_timer.c
index 57cb2f0..31543e5 100644
--- a/drivers/clocksource/arm_arch_timer.c
+++ b/drivers/clocksource/arm_arch_timer.c
@@ -179,11 +179,6 @@ static u64 arch_counter_read_cc(const struct cyclecounter 
*cc)
.flags  = CLOCK_SOURCE_IS_CONTINUOUS,
 };
 
-static struct cyclecounter cyclecounter __ro_after_init = {
-   .read   = arch_counter_read_cc,
-   .mask   = CLOCKSOURCE_MASK(56),
-};
-
 struct ate_acpi_oem_info {
char oem_id[ACPI_OEM_ID_SIZE + 1];
char oem_table_id[ACPI_OEM_TABLE_ID_SIZE + 1];
@@ -915,7 +910,10 @@ static u64 arch_counter_get_cntvct_mem(void)
return ((u64) vct_hi << 32) | vct_lo;
 }
 
-static struct arch_timer_kvm_info arch_timer_kvm_info;
+static struct arch_timer_kvm_info arch_timer_kvm_info = {
+   .timecounter.cc.read = arch_counter_read_cc,
+   .timecounter.cc.mask = CLOCKSOURCE_MASK(56),
+};
 
 struct arch_timer_kvm_info 

[PATCH 01/27] timecounter: Make cyclecounter struct part of timecounter struct

[Sagar Arun Kamble]

There is no real need for the users of timecounters to define cyclecounter
and timecounter variables separately. Since timecounter will always be
based on cyclecounter, have cyclecounter struct as member of timecounter
struct.

v2: Rebase.

Suggested-by: Chris Wilson 
Signed-off-by: Sagar Arun Kamble 
Cc: Chris Wilson 
Cc: Richard Cochran 
Cc: John Stultz 
Cc: Thomas Gleixner 
Cc: Stephen Boyd 
Cc: linux-kernel@vger.kernel.org
Cc: linux-arm-ker...@lists.infradead.org
Cc: net...@vger.kernel.org
Cc: intel-wired-...@lists.osuosl.org
Cc: linux-r...@vger.kernel.org
Cc: alsa-de...@alsa-project.org
Cc: kvm...@lists.cs.columbia.edu
Acked-by: Jeff Kirsher  (Intel drivers)
---
 arch/microblaze/kernel/timer.c | 20 ++--
 drivers/clocksource/arm_arch_timer.c   | 19 ++--
 drivers/net/ethernet/amd/xgbe/xgbe-dev.c   |  3 +-
 drivers/net/ethernet/amd/xgbe/xgbe-ptp.c   |  9 +++---
 drivers/net/ethernet/amd/xgbe/xgbe.h   |  1 -
 drivers/net/ethernet/broadcom/bnx2x/bnx2x.h|  1 -
 drivers/net/ethernet/broadcom/bnx2x/bnx2x_main.c   | 20 ++--
 drivers/net/ethernet/freescale/fec.h   |  1 -
 drivers/net/ethernet/freescale/fec_ptp.c   | 30 +-
 drivers/net/ethernet/intel/e1000e/e1000.h  |  1 -
 drivers/net/ethernet/intel/e1000e/netdev.c | 27 
 drivers/net/ethernet/intel/e1000e/ptp.c|  2 +-
 drivers/net/ethernet/intel/igb/igb.h   |  1 -
 drivers/net/ethernet/intel/igb/igb_ptp.c   | 25 ---
 drivers/net/ethernet/intel/ixgbe/ixgbe.h   |  1 -
 drivers/net/ethernet/intel/ixgbe/ixgbe_ptp.c   | 17 +-
 drivers/net/ethernet/mellanox/mlx4/en_clock.c  | 28 -
 drivers/net/ethernet/mellanox/mlx4/mlx4_en.h   |  1 -
 .../net/ethernet/mellanox/mlx5/core/lib/clock.c| 34 ++--
 drivers/net/ethernet/qlogic/qede/qede_ptp.c| 20 ++--
 drivers/net/ethernet/ti/cpts.c | 36 --
 drivers/net/ethernet/ti/cpts.h |  1 -
 include/linux/mlx5/driver.h|  1 -
 include/linux/timecounter.h|  4 +--
 include/sound/hdaudio.h|  1 -
 kernel/time/timecounter.c  | 28 -
 sound/hda/hdac_stream.c|  7 +++--
 virt/kvm/arm/arch_timer.c  |  6 ++--
 28 files changed, 163 insertions(+), 182 deletions(-)

diff --git a/arch/microblaze/kernel/timer.c b/arch/microblaze/kernel/timer.c
index 7de941c..b7f89e9 100644
--- a/arch/microblaze/kernel/timer.c
+++ b/arch/microblaze/kernel/timer.c
@@ -199,27 +199,25 @@ static u64 xilinx_read(struct clocksource *cs)
return (u64)xilinx_clock_read();
 }
 
-static struct timecounter xilinx_tc = {
-   .cc = NULL,
-};
-
 static u64 xilinx_cc_read(const struct cyclecounter *cc)
 {
return xilinx_read(NULL);
 }
 
-static struct cyclecounter xilinx_cc = {
-   .read = xilinx_cc_read,
-   .mask = CLOCKSOURCE_MASK(32),
-   .shift = 8,
+static struct timecounter xilinx_tc = {
+   .cc.read = xilinx_cc_read,
+   .cc.mask = CLOCKSOURCE_MASK(32),
+   .cc.mult = 0,
+   .cc.shift = 8,
 };
 
 static int __init init_xilinx_timecounter(void)
 {
-   xilinx_cc.mult = div_sc(timer_clock_freq, NSEC_PER_SEC,
-   xilinx_cc.shift);
+   struct cyclecounter *cc = _tc.cc;
+
+   cc->mult = div_sc(timer_clock_freq, NSEC_PER_SEC, cc->shift);
 
-   timecounter_init(_tc, _cc, sched_clock());
+   timecounter_init(_tc, sched_clock());
 
return 0;
 }
diff --git a/drivers/clocksource/arm_arch_timer.c 
b/drivers/clocksource/arm_arch_timer.c
index 57cb2f0..31543e5 100644
--- a/drivers/clocksource/arm_arch_timer.c
+++ b/drivers/clocksource/arm_arch_timer.c
@@ -179,11 +179,6 @@ static u64 arch_counter_read_cc(const struct cyclecounter 
*cc)
.flags  = CLOCK_SOURCE_IS_CONTINUOUS,
 };
 
-static struct cyclecounter cyclecounter __ro_after_init = {
-   .read   = arch_counter_read_cc,
-   .mask   = CLOCKSOURCE_MASK(56),
-};
-
 struct ate_acpi_oem_info {
char oem_id[ACPI_OEM_ID_SIZE + 1];
char oem_table_id[ACPI_OEM_TABLE_ID_SIZE + 1];
@@ -915,7 +910,10 @@ static u64 arch_counter_get_cntvct_mem(void)
return ((u64) vct_hi << 32) | vct_lo;
 }
 
-static struct arch_timer_kvm_info arch_timer_kvm_info;
+static struct arch_timer_kvm_info arch_timer_kvm_info = {
+   .timecounter.cc.read = arch_counter_read_cc,
+   .timecounter.cc.mask = CLOCKSOURCE_MASK(56),
+};
 
 struct arch_timer_kvm_info *arch_timer_get_kvm_info(void)
 {
@@ -925,6 +923,7 @@ struct arch_timer_kvm_info *arch_timer_get_kvm_info(void)
 static void __init arch_counter_register(unsigned type)
 {
u64 start_count;
+   struct cyclecounter *cc = _timer_kvm_info.ti

[PATCH 09/27] igb: Use timecounter_initialize interface

[Sagar Arun Kamble]

With new interface timecounter_initialize we can initialize timecounter
fields and underlying cyclecounter together. Update igb ptp timecounter
init with this new function.

Signed-off-by: Sagar Arun Kamble 
Cc: Richard Cochran 
Cc: Jeff Kirsher 
Cc: intel-wired-...@lists.osuosl.org
Cc: net...@vger.kernel.org
Cc: linux-kernel@vger.kernel.org
---
 drivers/net/ethernet/intel/igb/igb.h |  4 
 drivers/net/ethernet/intel/igb/igb_ptp.c | 23 ++-
 2 files changed, 18 insertions(+), 9 deletions(-)

diff --git a/drivers/net/ethernet/intel/igb/igb.h 
b/drivers/net/ethernet/intel/igb/igb.h
index 4eac4f2..4ef5632 100644
--- a/drivers/net/ethernet/intel/igb/igb.h
+++ b/drivers/net/ethernet/intel/igb/igb.h
@@ -566,6 +566,10 @@ struct igb_adapter {
unsigned int ptp_flags;
spinlock_t tmreg_lock;
struct timecounter tc;
+   u64 (*cc_read)(const struct cyclecounter *cc);
+   u64 cc_mask;
+   u32 cc_mult;
+   u32 cc_shift;
u32 tx_hwtstamp_timeouts;
u32 tx_hwtstamp_skipped;
u32 rx_hwtstamp_cleared;
diff --git a/drivers/net/ethernet/intel/igb/igb_ptp.c 
b/drivers/net/ethernet/intel/igb/igb_ptp.c
index 0745eff..f6c1a8d 100644
--- a/drivers/net/ethernet/intel/igb/igb_ptp.c
+++ b/drivers/net/ethernet/intel/igb/igb_ptp.c
@@ -1126,10 +1126,10 @@ void igb_ptp_init(struct igb_adapter *adapter)
adapter->ptp_caps.gettime64 = igb_ptp_gettime_82576;
adapter->ptp_caps.settime64 = igb_ptp_settime_82576;
adapter->ptp_caps.enable = igb_ptp_feature_enable;
-   adapter->tc.cc.read = igb_ptp_read_82576;
-   adapter->tc.cc.mask = CYCLECOUNTER_MASK(64);
-   adapter->tc.cc.mult = 1;
-   adapter->tc.cc.shift = IGB_82576_TSYNC_SHIFT;
+   adapter->cc_read = igb_ptp_read_82576;
+   adapter->cc_mask = CYCLECOUNTER_MASK(64);
+   adapter->cc_mult = 1;
+   adapter->cc_shift = IGB_82576_TSYNC_SHIFT;
adapter->ptp_flags |= IGB_PTP_OVERFLOW_CHECK;
break;
case e1000_82580:
@@ -1145,10 +1145,10 @@ void igb_ptp_init(struct igb_adapter *adapter)
adapter->ptp_caps.gettime64 = igb_ptp_gettime_82576;
adapter->ptp_caps.settime64 = igb_ptp_settime_82576;
adapter->ptp_caps.enable = igb_ptp_feature_enable;
-   adapter->tc.cc.read = igb_ptp_read_82580;
-   adapter->tc.cc.mask = CYCLECOUNTER_MASK(IGB_NBITS_82580);
-   adapter->tc.cc.mult = 1;
-   adapter->tc.cc.shift = 0;
+   adapter->cc_read = igb_ptp_read_82580;
+   adapter->cc_mask = CYCLECOUNTER_MASK(IGB_NBITS_82580);
+   adapter->cc_mult = 1;
+   adapter->cc_shift = 0;
adapter->ptp_flags |= IGB_PTP_OVERFLOW_CHECK;
break;
case e1000_i210:
@@ -1289,7 +1289,12 @@ void igb_ptp_reset(struct igb_adapter *adapter)
 
igb_ptp_write_i210(adapter, );
} else {
-   timecounter_init(>tc, ktime_to_ns(ktime_get_real()));
+   timecounter_initialize(>tc,
+  adapter->cc_read,
+  adapter->cc_mask,
+  adapter->cc_mult,
+  adapter->cc_shift,
+  ktime_to_ns(ktime_get_real()));
}
 out:
spin_unlock_irqrestore(>tmreg_lock, flags);
-- 
1.9.1

[PATCH 00/27] timecounter/cyclecounter struct/interface update

[Sagar Arun Kamble]

This series makes cyclecounter part of timecounter and introduces two
new interface functions timecounter_initialize() for initializing both
timecounter and cyclecounter fields and timecounter_reset() for setting
start time of timecounter. Updates all drivers with new functions and
removes function timecounter_init().

Cc: Sagar Arun Kamble <sagar.a.kam...@intel.com>
Cc: Richard Cochran <richardcoch...@gmail.com>
Cc: John Stultz <john.stu...@linaro.org>
Cc: Thomas Gleixner <t...@linutronix.de>
Cc: Stephen Boyd <sb...@codeaurora.org>
Cc: linux-kernel@vger.kernel.org

Sagar Arun Kamble (27):
  timecounter: Make cyclecounter struct part of timecounter struct
  timecounter: Introduce timecounter_initialize to update timecounter
and cyclecounter
  microblaze: Use timecounter_initialize interface
  clocksource/arm_arch_timer: Use timecounter_initialize interface
  amd-xgbe: Use timecounter_initialize interface
  bnx2x: Use timecounter_initialize interface
  fec: Use timecounter_initialize interface
  e1000e: Use timecounter_initialize interface
  igb: Use timecounter_initialize interface
  ixgbe: Use timecounter_initialize interface
  net/mlx4: Use timecounter_initialize interface
  net/mlx5: Use timecounter_initialize interface
  qede: Use timecounter_initialize interface
  net: cpts: Use timecounter_initialize interface
  ALSA: hda - Use timecounter_initialize interface
  timecounter: Introduce timecounter_reset
  amd-xgbe: Use timecounter_reset interface
  bnx2x: Use timecounter_reset interface
  net: fec: ptp: Use timecounter_reset interface
  e1000e: Use timecounter_reset interface
  igb: Use timecounter_reset interface
  ixgbe: Use timecounter_reset interface
  net/mlx4: Use timecounter_reset interface
  net/mlx5: Use timecounter_reset interface
  qede: Use timecounter_reset interface
  net: cpts: Use timecounter_reset interface
  timecounter: Remove timecounter_init

 arch/microblaze/kernel/timer.c | 23 -
 drivers/clocksource/arm_arch_timer.c   | 16 +++---
 drivers/net/ethernet/amd/xgbe/xgbe-dev.c   |  5 +-
 drivers/net/ethernet/amd/xgbe/xgbe-ptp.c   | 18 +++
 drivers/net/ethernet/amd/xgbe/xgbe.h   |  1 -
 drivers/net/ethernet/broadcom/bnx2x/bnx2x.h|  1 -
 drivers/net/ethernet/broadcom/bnx2x/bnx2x_main.c   | 24 -
 drivers/net/ethernet/freescale/fec.h   |  1 -
 drivers/net/ethernet/freescale/fec_ptp.c   | 34 ++---
 drivers/net/ethernet/intel/e1000e/e1000.h  |  5 +-
 drivers/net/ethernet/intel/e1000e/netdev.c | 34 +++--
 drivers/net/ethernet/intel/e1000e/ptp.c|  2 +-
 drivers/net/ethernet/intel/igb/igb.h   |  5 +-
 drivers/net/ethernet/intel/igb/igb_ptp.c   | 30 ++-
 drivers/net/ethernet/intel/ixgbe/ixgbe.h   |  7 ++-
 drivers/net/ethernet/intel/ixgbe/ixgbe_main.c  |  4 +-
 drivers/net/ethernet/intel/ixgbe/ixgbe_ptp.c   | 58 --
 drivers/net/ethernet/mellanox/mlx4/en_clock.c  | 29 +--
 drivers/net/ethernet/mellanox/mlx4/mlx4_en.h   |  1 -
 .../net/ethernet/mellanox/mlx5/core/lib/clock.c| 38 +++---
 drivers/net/ethernet/qlogic/qede/qede_ptp.c| 21 
 drivers/net/ethernet/ti/cpts.c | 45 ++---
 drivers/net/ethernet/ti/cpts.h |  4 +-
 include/linux/mlx5/driver.h|  1 -
 include/linux/timecounter.h| 52 +--
 include/sound/hdaudio.h|  1 -
 kernel/time/timecounter.c  | 49 --
 sound/hda/hdac_stream.c| 17 +++
 virt/kvm/arm/arch_timer.c  |  6 +--
 29 files changed, 291 insertions(+), 241 deletions(-)

-- 
1.9.1

[PATCH 00/27] timecounter/cyclecounter struct/interface update

[Sagar Arun Kamble]

This series makes cyclecounter part of timecounter and introduces two
new interface functions timecounter_initialize() for initializing both
timecounter and cyclecounter fields and timecounter_reset() for setting
start time of timecounter. Updates all drivers with new functions and
removes function timecounter_init().

Cc: Sagar Arun Kamble 
Cc: Richard Cochran 
Cc: John Stultz 
Cc: Thomas Gleixner 
Cc: Stephen Boyd 
Cc: linux-kernel@vger.kernel.org

Sagar Arun Kamble (27):
  timecounter: Make cyclecounter struct part of timecounter struct
  timecounter: Introduce timecounter_initialize to update timecounter
and cyclecounter
  microblaze: Use timecounter_initialize interface
  clocksource/arm_arch_timer: Use timecounter_initialize interface
  amd-xgbe: Use timecounter_initialize interface
  bnx2x: Use timecounter_initialize interface
  fec: Use timecounter_initialize interface
  e1000e: Use timecounter_initialize interface
  igb: Use timecounter_initialize interface
  ixgbe: Use timecounter_initialize interface
  net/mlx4: Use timecounter_initialize interface
  net/mlx5: Use timecounter_initialize interface
  qede: Use timecounter_initialize interface
  net: cpts: Use timecounter_initialize interface
  ALSA: hda - Use timecounter_initialize interface
  timecounter: Introduce timecounter_reset
  amd-xgbe: Use timecounter_reset interface
  bnx2x: Use timecounter_reset interface
  net: fec: ptp: Use timecounter_reset interface
  e1000e: Use timecounter_reset interface
  igb: Use timecounter_reset interface
  ixgbe: Use timecounter_reset interface
  net/mlx4: Use timecounter_reset interface
  net/mlx5: Use timecounter_reset interface
  qede: Use timecounter_reset interface
  net: cpts: Use timecounter_reset interface
  timecounter: Remove timecounter_init

 arch/microblaze/kernel/timer.c | 23 -
 drivers/clocksource/arm_arch_timer.c   | 16 +++---
 drivers/net/ethernet/amd/xgbe/xgbe-dev.c   |  5 +-
 drivers/net/ethernet/amd/xgbe/xgbe-ptp.c   | 18 +++
 drivers/net/ethernet/amd/xgbe/xgbe.h   |  1 -
 drivers/net/ethernet/broadcom/bnx2x/bnx2x.h|  1 -
 drivers/net/ethernet/broadcom/bnx2x/bnx2x_main.c   | 24 -
 drivers/net/ethernet/freescale/fec.h   |  1 -
 drivers/net/ethernet/freescale/fec_ptp.c   | 34 ++---
 drivers/net/ethernet/intel/e1000e/e1000.h  |  5 +-
 drivers/net/ethernet/intel/e1000e/netdev.c | 34 +++--
 drivers/net/ethernet/intel/e1000e/ptp.c|  2 +-
 drivers/net/ethernet/intel/igb/igb.h   |  5 +-
 drivers/net/ethernet/intel/igb/igb_ptp.c   | 30 ++-
 drivers/net/ethernet/intel/ixgbe/ixgbe.h   |  7 ++-
 drivers/net/ethernet/intel/ixgbe/ixgbe_main.c  |  4 +-
 drivers/net/ethernet/intel/ixgbe/ixgbe_ptp.c   | 58 --
 drivers/net/ethernet/mellanox/mlx4/en_clock.c  | 29 +--
 drivers/net/ethernet/mellanox/mlx4/mlx4_en.h   |  1 -
 .../net/ethernet/mellanox/mlx5/core/lib/clock.c| 38 +++---
 drivers/net/ethernet/qlogic/qede/qede_ptp.c| 21 
 drivers/net/ethernet/ti/cpts.c | 45 ++---
 drivers/net/ethernet/ti/cpts.h |  4 +-
 include/linux/mlx5/driver.h|  1 -
 include/linux/timecounter.h| 52 +--
 include/sound/hdaudio.h|  1 -
 kernel/time/timecounter.c  | 49 --
 sound/hda/hdac_stream.c| 17 +++
 virt/kvm/arm/arch_timer.c  |  6 +--
 29 files changed, 291 insertions(+), 241 deletions(-)

-- 
1.9.1

Re: [PATCH 1/1] timecounter: Make cyclecounter struct part of timecounter struct

[Sagar Arun Kamble]

On 12/2/2017 11:34 PM, Richard Cochran wrote:

On Sat, Dec 02, 2017 at 10:01:35AM +0530, Sagar Arun Kamble wrote:

There is no real need for the users of timecounters to define cyclecounter
and timecounter variables separately. Since timecounter will always be
based on cyclecounter, have cyclecounter struct as member of timecounter
struct.

Overall, this is a welcome change.  However, it doesn't go far enough,
IMHO, and I'll explain that more below.


diff --git a/drivers/net/ethernet/mellanox/mlx4/en_clock.c 
b/drivers/net/ethernet/mellanox/mlx4/en_clock.c
index 0247885..35987b5 100644
--- a/drivers/net/ethernet/mellanox/mlx4/en_clock.c
+++ b/drivers/net/ethernet/mellanox/mlx4/en_clock.c

...

As it stands now, timecounter_init() is used for two totally different
reasons.  Some callers only want to set the time, ...


@@ -207,7 +207,7 @@ static int mlx4_en_phc_settime(struct ptp_clock_info *ptp,
  
  	/* reset the timecounter */

write_seqlock_irqsave(>clock_lock, flags);
-   timecounter_init(>clock, >cycles, ns);
+   timecounter_init(>clock, ns);
write_sequnlock_irqrestore(>clock_lock, flags);
  
  	return 0;

... while others initialize the data structure the first time:


@@ -274,17 +274,17 @@ void mlx4_en_init_timestamp(struct mlx4_en_dev *mdev)
  
  	seqlock_init(>clock_lock);
  
-	memset(>cycles, 0, sizeof(mdev->cycles));

-   mdev->cycles.read = mlx4_en_read_clock;
-   mdev->cycles.mask = CLOCKSOURCE_MASK(48);
-   mdev->cycles.shift = freq_to_shift(dev->caps.hca_core_clock);
-   mdev->cycles.mult =
-   clocksource_khz2mult(1000 * dev->caps.hca_core_clock, 
mdev->cycles.shift);
-   mdev->nominal_c_mult = mdev->cycles.mult;
+   memset(>clock.cc, 0, sizeof(mdev->clock.cc));
+   mdev->clock.cc.read = mlx4_en_read_clock;
+   mdev->clock.cc.mask = CLOCKSOURCE_MASK(48);
+   mdev->clock.cc.shift = freq_to_shift(dev->caps.hca_core_clock);
+   mdev->clock.cc.mult =
+   clocksource_khz2mult(1000 * dev->caps.hca_core_clock,
+mdev->clock.cc.shift);
+   mdev->nominal_c_mult = mdev->clock.cc.mult;
  
  	write_seqlock_irqsave(>clock_lock, flags);

-   timecounter_init(>clock, >cycles,
-ktime_to_ns(ktime_get_real()));
+   timecounter_init(>clock, ktime_to_ns(ktime_get_real()));

I'd like to see two followup patches to this one:

1. Convert timecounter_init() callers to a new timecounter_reset()
function where the intent is to reset the time.

2. Change timecounter_init() to take the cyclecounter fields as
arguments.

void timecounter_init(struct timecounter *tc,
  u64 (*read)(const struct cyclecounter *cc),
  u64 mask,
  u32 mult,
  u32 shift,
  u64 start_tstamp);

Then we can clean up all this stuff:

mdev->clock.cc.read = mlx4_en_read_clock;
mdev->clock.cc.mask = CLOCKSOURCE_MASK(48);
mdev->clock.cc.shift = freq_to_shift(dev->caps.hca_core_clock);
mdev->clock.cc.mult = clocksource_khz2mult(...);

This second step can be phased in by calling the new function
timecounter_initialize() and converting the drivers one by one.


Yes. Will make these changes and share new patchset.

Thank you for the review Richard.

Regards,
Sagar


diff --git a/include/linux/timecounter.h b/include/linux/timecounter.h
index 2496ad4..6daca06 100644
--- a/include/linux/timecounter.h
+++ b/include/linux/timecounter.h

...

@@ -98,7 +98,6 @@ static inline void timecounter_adjtime(struct timecounter 
*tc, s64 delta)
  /**
   * timecounter_init - initialize a time counter
   * @tc:   Pointer to time counter which is to be 
initialized/reset
- * @cc:A cycle counter, ready to be used.

This "ready to used" requirement should go.  The init() function
should make the instance ready to be used all at once.

Thanks,
Richard

Re: [PATCH 1/1] timecounter: Make cyclecounter struct part of timecounter struct

[Sagar Arun Kamble]

On 12/2/2017 11:34 PM, Richard Cochran wrote:

On Sat, Dec 02, 2017 at 10:01:35AM +0530, Sagar Arun Kamble wrote:

There is no real need for the users of timecounters to define cyclecounter
and timecounter variables separately. Since timecounter will always be
based on cyclecounter, have cyclecounter struct as member of timecounter
struct.

Overall, this is a welcome change.  However, it doesn't go far enough,
IMHO, and I'll explain that more below.


diff --git a/drivers/net/ethernet/mellanox/mlx4/en_clock.c 
b/drivers/net/ethernet/mellanox/mlx4/en_clock.c
index 0247885..35987b5 100644
--- a/drivers/net/ethernet/mellanox/mlx4/en_clock.c
+++ b/drivers/net/ethernet/mellanox/mlx4/en_clock.c

...

As it stands now, timecounter_init() is used for two totally different
reasons.  Some callers only want to set the time, ...


@@ -207,7 +207,7 @@ static int mlx4_en_phc_settime(struct ptp_clock_info *ptp,
  
  	/* reset the timecounter */

write_seqlock_irqsave(>clock_lock, flags);
-   timecounter_init(>clock, >cycles, ns);
+   timecounter_init(>clock, ns);
write_sequnlock_irqrestore(>clock_lock, flags);
  
  	return 0;

... while others initialize the data structure the first time:


@@ -274,17 +274,17 @@ void mlx4_en_init_timestamp(struct mlx4_en_dev *mdev)
  
  	seqlock_init(>clock_lock);
  
-	memset(>cycles, 0, sizeof(mdev->cycles));

-   mdev->cycles.read = mlx4_en_read_clock;
-   mdev->cycles.mask = CLOCKSOURCE_MASK(48);
-   mdev->cycles.shift = freq_to_shift(dev->caps.hca_core_clock);
-   mdev->cycles.mult =
-   clocksource_khz2mult(1000 * dev->caps.hca_core_clock, 
mdev->cycles.shift);
-   mdev->nominal_c_mult = mdev->cycles.mult;
+   memset(>clock.cc, 0, sizeof(mdev->clock.cc));
+   mdev->clock.cc.read = mlx4_en_read_clock;
+   mdev->clock.cc.mask = CLOCKSOURCE_MASK(48);
+   mdev->clock.cc.shift = freq_to_shift(dev->caps.hca_core_clock);
+   mdev->clock.cc.mult =
+   clocksource_khz2mult(1000 * dev->caps.hca_core_clock,
+mdev->clock.cc.shift);
+   mdev->nominal_c_mult = mdev->clock.cc.mult;
  
  	write_seqlock_irqsave(>clock_lock, flags);

-   timecounter_init(>clock, >cycles,
-ktime_to_ns(ktime_get_real()));
+   timecounter_init(>clock, ktime_to_ns(ktime_get_real()));

I'd like to see two followup patches to this one:

1. Convert timecounter_init() callers to a new timecounter_reset()
function where the intent is to reset the time.

2. Change timecounter_init() to take the cyclecounter fields as
arguments.

void timecounter_init(struct timecounter *tc,
  u64 (*read)(const struct cyclecounter *cc),
  u64 mask,
  u32 mult,
  u32 shift,
  u64 start_tstamp);

Then we can clean up all this stuff:

mdev->clock.cc.read = mlx4_en_read_clock;
mdev->clock.cc.mask = CLOCKSOURCE_MASK(48);
mdev->clock.cc.shift = freq_to_shift(dev->caps.hca_core_clock);
mdev->clock.cc.mult = clocksource_khz2mult(...);

This second step can be phased in by calling the new function
timecounter_initialize() and converting the drivers one by one.


Yes. Will make these changes and share new patchset.

Thank you for the review Richard.

Regards,
Sagar


diff --git a/include/linux/timecounter.h b/include/linux/timecounter.h
index 2496ad4..6daca06 100644
--- a/include/linux/timecounter.h
+++ b/include/linux/timecounter.h

...

@@ -98,7 +98,6 @@ static inline void timecounter_adjtime(struct timecounter 
*tc, s64 delta)
  /**
   * timecounter_init - initialize a time counter
   * @tc:   Pointer to time counter which is to be 
initialized/reset
- * @cc:A cycle counter, ready to be used.

This "ready to used" requirement should go.  The init() function
should make the instance ready to be used all at once.

Thanks,
Richard

[PATCH 1/1] timecounter: Make cyclecounter struct part of timecounter struct

[Sagar Arun Kamble]

There is no real need for the users of timecounters to define cyclecounter
and timecounter variables separately. Since timecounter will always be
based on cyclecounter, have cyclecounter struct as member of timecounter
struct.

Suggested-by: Chris Wilson <ch...@chris-wilson.co.uk>
Signed-off-by: Sagar Arun Kamble <sagar.a.kam...@intel.com>
Cc: Chris Wilson <ch...@chris-wilson.co.uk>
Cc: Richard Cochran <richardcoch...@gmail.com>
Cc: John Stultz <john.stu...@linaro.org>
Cc: Thomas Gleixner <t...@linutronix.de>
Cc: Stephen Boyd <sb...@codeaurora.org>
Cc: linux-kernel@vger.kernel.org
Cc: linux-arm-ker...@lists.infradead.org
Cc: net...@vger.kernel.org
Cc: intel-wired-...@lists.osuosl.org
Cc: linux-r...@vger.kernel.org
Cc: alsa-de...@alsa-project.org
Cc: kvm...@lists.cs.columbia.edu
---
 arch/microblaze/kernel/timer.c | 20 ++--
 drivers/clocksource/arm_arch_timer.c   | 19 ++--
 drivers/net/ethernet/amd/xgbe/xgbe-dev.c   |  3 +-
 drivers/net/ethernet/amd/xgbe/xgbe-ptp.c   |  9 +++---
 drivers/net/ethernet/amd/xgbe/xgbe.h   |  1 -
 drivers/net/ethernet/broadcom/bnx2x/bnx2x.h|  1 -
 drivers/net/ethernet/broadcom/bnx2x/bnx2x_main.c   | 20 ++--
 drivers/net/ethernet/freescale/fec.h   |  1 -
 drivers/net/ethernet/freescale/fec_ptp.c   | 30 +-
 drivers/net/ethernet/intel/e1000e/e1000.h  |  1 -
 drivers/net/ethernet/intel/e1000e/netdev.c | 27 
 drivers/net/ethernet/intel/e1000e/ptp.c|  2 +-
 drivers/net/ethernet/intel/igb/igb.h   |  1 -
 drivers/net/ethernet/intel/igb/igb_ptp.c   | 25 ---
 drivers/net/ethernet/intel/ixgbe/ixgbe.h   |  1 -
 drivers/net/ethernet/intel/ixgbe/ixgbe_ptp.c   | 17 +-
 drivers/net/ethernet/mellanox/mlx4/en_clock.c  | 28 -
 drivers/net/ethernet/mellanox/mlx4/mlx4_en.h   |  1 -
 .../net/ethernet/mellanox/mlx5/core/lib/clock.c| 34 ++--
 drivers/net/ethernet/qlogic/qede/qede_ptp.c| 20 ++--
 drivers/net/ethernet/ti/cpts.c | 36 --
 drivers/net/ethernet/ti/cpts.h |  1 -
 include/linux/mlx5/driver.h|  1 -
 include/linux/timecounter.h|  4 +--
 include/sound/hdaudio.h|  1 -
 kernel/time/timecounter.c  | 28 -
 sound/hda/hdac_stream.c|  7 +++--
 virt/kvm/arm/arch_timer.c  |  6 ++--
 28 files changed, 163 insertions(+), 182 deletions(-)

diff --git a/arch/microblaze/kernel/timer.c b/arch/microblaze/kernel/timer.c
index 7de941c..b7f89e9 100644
--- a/arch/microblaze/kernel/timer.c
+++ b/arch/microblaze/kernel/timer.c
@@ -199,27 +199,25 @@ static u64 xilinx_read(struct clocksource *cs)
return (u64)xilinx_clock_read();
 }
 
-static struct timecounter xilinx_tc = {
-   .cc = NULL,
-};
-
 static u64 xilinx_cc_read(const struct cyclecounter *cc)
 {
return xilinx_read(NULL);
 }
 
-static struct cyclecounter xilinx_cc = {
-   .read = xilinx_cc_read,
-   .mask = CLOCKSOURCE_MASK(32),
-   .shift = 8,
+static struct timecounter xilinx_tc = {
+   .cc.read = xilinx_cc_read,
+   .cc.mask = CLOCKSOURCE_MASK(32),
+   .cc.mult = 0,
+   .cc.shift = 8,
 };
 
 static int __init init_xilinx_timecounter(void)
 {
-   xilinx_cc.mult = div_sc(timer_clock_freq, NSEC_PER_SEC,
-   xilinx_cc.shift);
+   struct cyclecounter *cc = _tc.cc;
+
+   cc->mult = div_sc(timer_clock_freq, NSEC_PER_SEC, cc->shift);
 
-   timecounter_init(_tc, _cc, sched_clock());
+   timecounter_init(_tc, sched_clock());
 
return 0;
 }
diff --git a/drivers/clocksource/arm_arch_timer.c 
b/drivers/clocksource/arm_arch_timer.c
index 57cb2f0..31543e5 100644
--- a/drivers/clocksource/arm_arch_timer.c
+++ b/drivers/clocksource/arm_arch_timer.c
@@ -179,11 +179,6 @@ static u64 arch_counter_read_cc(const struct cyclecounter 
*cc)
.flags  = CLOCK_SOURCE_IS_CONTINUOUS,
 };
 
-static struct cyclecounter cyclecounter __ro_after_init = {
-   .read   = arch_counter_read_cc,
-   .mask   = CLOCKSOURCE_MASK(56),
-};
-
 struct ate_acpi_oem_info {
char oem_id[ACPI_OEM_ID_SIZE + 1];
char oem_table_id[ACPI_OEM_TABLE_ID_SIZE + 1];
@@ -915,7 +910,10 @@ static u64 arch_counter_get_cntvct_mem(void)
return ((u64) vct_hi << 32) | vct_lo;
 }
 
-static struct arch_timer_kvm_info arch_timer_kvm_info;
+static struct arch_timer_kvm_info arch_timer_kvm_info = {
+   .timecounter.cc.read = arch_counter_read_cc,
+   .timecounter.cc.mask = CLOCKSOURCE_MASK(56),
+};
 
 struct arch_timer_kvm_info *arch_timer_get_kvm_info(void)
 {
@@ -925,6 +923,7 @@ struct arch_timer_kvm_info *arch_timer_get_kv

[PATCH 1/1] timecounter: Make cyclecounter struct part of timecounter struct

[Sagar Arun Kamble]

There is no real need for the users of timecounters to define cyclecounter
and timecounter variables separately. Since timecounter will always be
based on cyclecounter, have cyclecounter struct as member of timecounter
struct.

Suggested-by: Chris Wilson 
Signed-off-by: Sagar Arun Kamble 
Cc: Chris Wilson 
Cc: Richard Cochran 
Cc: John Stultz 
Cc: Thomas Gleixner 
Cc: Stephen Boyd 
Cc: linux-kernel@vger.kernel.org
Cc: linux-arm-ker...@lists.infradead.org
Cc: net...@vger.kernel.org
Cc: intel-wired-...@lists.osuosl.org
Cc: linux-r...@vger.kernel.org
Cc: alsa-de...@alsa-project.org
Cc: kvm...@lists.cs.columbia.edu
---
 arch/microblaze/kernel/timer.c | 20 ++--
 drivers/clocksource/arm_arch_timer.c   | 19 ++--
 drivers/net/ethernet/amd/xgbe/xgbe-dev.c   |  3 +-
 drivers/net/ethernet/amd/xgbe/xgbe-ptp.c   |  9 +++---
 drivers/net/ethernet/amd/xgbe/xgbe.h   |  1 -
 drivers/net/ethernet/broadcom/bnx2x/bnx2x.h|  1 -
 drivers/net/ethernet/broadcom/bnx2x/bnx2x_main.c   | 20 ++--
 drivers/net/ethernet/freescale/fec.h   |  1 -
 drivers/net/ethernet/freescale/fec_ptp.c   | 30 +-
 drivers/net/ethernet/intel/e1000e/e1000.h  |  1 -
 drivers/net/ethernet/intel/e1000e/netdev.c | 27 
 drivers/net/ethernet/intel/e1000e/ptp.c|  2 +-
 drivers/net/ethernet/intel/igb/igb.h   |  1 -
 drivers/net/ethernet/intel/igb/igb_ptp.c   | 25 ---
 drivers/net/ethernet/intel/ixgbe/ixgbe.h   |  1 -
 drivers/net/ethernet/intel/ixgbe/ixgbe_ptp.c   | 17 +-
 drivers/net/ethernet/mellanox/mlx4/en_clock.c  | 28 -
 drivers/net/ethernet/mellanox/mlx4/mlx4_en.h   |  1 -
 .../net/ethernet/mellanox/mlx5/core/lib/clock.c| 34 ++--
 drivers/net/ethernet/qlogic/qede/qede_ptp.c| 20 ++--
 drivers/net/ethernet/ti/cpts.c | 36 --
 drivers/net/ethernet/ti/cpts.h |  1 -
 include/linux/mlx5/driver.h|  1 -
 include/linux/timecounter.h|  4 +--
 include/sound/hdaudio.h|  1 -
 kernel/time/timecounter.c  | 28 -
 sound/hda/hdac_stream.c|  7 +++--
 virt/kvm/arm/arch_timer.c  |  6 ++--
 28 files changed, 163 insertions(+), 182 deletions(-)

diff --git a/arch/microblaze/kernel/timer.c b/arch/microblaze/kernel/timer.c
index 7de941c..b7f89e9 100644
--- a/arch/microblaze/kernel/timer.c
+++ b/arch/microblaze/kernel/timer.c
@@ -199,27 +199,25 @@ static u64 xilinx_read(struct clocksource *cs)
return (u64)xilinx_clock_read();
 }
 
-static struct timecounter xilinx_tc = {
-   .cc = NULL,
-};
-
 static u64 xilinx_cc_read(const struct cyclecounter *cc)
 {
return xilinx_read(NULL);
 }
 
-static struct cyclecounter xilinx_cc = {
-   .read = xilinx_cc_read,
-   .mask = CLOCKSOURCE_MASK(32),
-   .shift = 8,
+static struct timecounter xilinx_tc = {
+   .cc.read = xilinx_cc_read,
+   .cc.mask = CLOCKSOURCE_MASK(32),
+   .cc.mult = 0,
+   .cc.shift = 8,
 };
 
 static int __init init_xilinx_timecounter(void)
 {
-   xilinx_cc.mult = div_sc(timer_clock_freq, NSEC_PER_SEC,
-   xilinx_cc.shift);
+   struct cyclecounter *cc = _tc.cc;
+
+   cc->mult = div_sc(timer_clock_freq, NSEC_PER_SEC, cc->shift);
 
-   timecounter_init(_tc, _cc, sched_clock());
+   timecounter_init(_tc, sched_clock());
 
return 0;
 }
diff --git a/drivers/clocksource/arm_arch_timer.c 
b/drivers/clocksource/arm_arch_timer.c
index 57cb2f0..31543e5 100644
--- a/drivers/clocksource/arm_arch_timer.c
+++ b/drivers/clocksource/arm_arch_timer.c
@@ -179,11 +179,6 @@ static u64 arch_counter_read_cc(const struct cyclecounter 
*cc)
.flags  = CLOCK_SOURCE_IS_CONTINUOUS,
 };
 
-static struct cyclecounter cyclecounter __ro_after_init = {
-   .read   = arch_counter_read_cc,
-   .mask   = CLOCKSOURCE_MASK(56),
-};
-
 struct ate_acpi_oem_info {
char oem_id[ACPI_OEM_ID_SIZE + 1];
char oem_table_id[ACPI_OEM_TABLE_ID_SIZE + 1];
@@ -915,7 +910,10 @@ static u64 arch_counter_get_cntvct_mem(void)
return ((u64) vct_hi << 32) | vct_lo;
 }
 
-static struct arch_timer_kvm_info arch_timer_kvm_info;
+static struct arch_timer_kvm_info arch_timer_kvm_info = {
+   .timecounter.cc.read = arch_counter_read_cc,
+   .timecounter.cc.mask = CLOCKSOURCE_MASK(56),
+};
 
 struct arch_timer_kvm_info *arch_timer_get_kvm_info(void)
 {
@@ -925,6 +923,7 @@ struct arch_timer_kvm_info *arch_timer_get_kvm_info(void)
 static void __init arch_counter_register(unsigned type)
 {
u64 start_count;
+   struct cyclecounter *cc = _timer_kvm_info.timecounter.cc;
 
/* Register the CP15 based counter if

[PATCH 1/1] timecounter: Make cyclecounter struct part of timecounter struct

[Sagar Arun Kamble]

There is no real need for the users of timecounters to define cyclecounter
and timecounter variables separately. Since timecounter will always be
based on cyclecounter, have cyclecounter struct as member of timecounter
struct.

Suggested-by: Chris Wilson <ch...@chris-wilson.co.uk>
Signed-off-by: Sagar Arun Kamble <sagar.a.kam...@intel.com>
Cc: Chris Wilson <ch...@chris-wilson.co.uk>
Cc: John Stultz <john.stu...@linaro.org>
Cc: Thomas Gleixner <t...@linutronix.de>
Cc: Stephen Boyd <sb...@codeaurora.org>
Cc: linux-kernel@vger.kernel.org
Cc: linux-arm-ker...@lists.infradead.org
Cc: net...@vger.kernel.org
Cc: intel-wired-...@lists.osuosl.org
Cc: linux-r...@vger.kernel.org
Cc: alsa-de...@alsa-project.org
Cc: kvm...@lists.cs.columbia.edu
---
 arch/microblaze/kernel/timer.c | 20 ++--
 drivers/clocksource/arm_arch_timer.c   | 19 ++--
 drivers/net/ethernet/amd/xgbe/xgbe-dev.c   |  3 +-
 drivers/net/ethernet/amd/xgbe/xgbe-ptp.c   |  9 +++---
 drivers/net/ethernet/amd/xgbe/xgbe.h   |  1 -
 drivers/net/ethernet/broadcom/bnx2x/bnx2x.h|  1 -
 drivers/net/ethernet/broadcom/bnx2x/bnx2x_main.c   | 20 ++--
 drivers/net/ethernet/freescale/fec.h   |  1 -
 drivers/net/ethernet/freescale/fec_ptp.c   | 30 +-
 drivers/net/ethernet/intel/e1000e/e1000.h  |  1 -
 drivers/net/ethernet/intel/e1000e/netdev.c | 27 
 drivers/net/ethernet/intel/e1000e/ptp.c|  2 +-
 drivers/net/ethernet/intel/igb/igb.h   |  1 -
 drivers/net/ethernet/intel/igb/igb_ptp.c   | 25 ---
 drivers/net/ethernet/intel/ixgbe/ixgbe.h   |  1 -
 drivers/net/ethernet/intel/ixgbe/ixgbe_ptp.c   | 17 +-
 drivers/net/ethernet/mellanox/mlx4/en_clock.c  | 28 -
 drivers/net/ethernet/mellanox/mlx4/mlx4_en.h   |  1 -
 .../net/ethernet/mellanox/mlx5/core/lib/clock.c| 34 ++--
 drivers/net/ethernet/qlogic/qede/qede_ptp.c| 20 ++--
 drivers/net/ethernet/ti/cpts.c | 36 --
 drivers/net/ethernet/ti/cpts.h |  1 -
 include/linux/mlx5/driver.h|  1 -
 include/linux/timecounter.h|  4 +--
 include/sound/hdaudio.h|  1 -
 kernel/time/timecounter.c  | 28 -
 sound/hda/hdac_stream.c|  7 +++--
 virt/kvm/arm/arch_timer.c  |  6 ++--
 28 files changed, 163 insertions(+), 182 deletions(-)

diff --git a/arch/microblaze/kernel/timer.c b/arch/microblaze/kernel/timer.c
index 7de941c..b7f89e9 100644
--- a/arch/microblaze/kernel/timer.c
+++ b/arch/microblaze/kernel/timer.c
@@ -199,27 +199,25 @@ static u64 xilinx_read(struct clocksource *cs)
return (u64)xilinx_clock_read();
 }
 
-static struct timecounter xilinx_tc = {
-   .cc = NULL,
-};
-
 static u64 xilinx_cc_read(const struct cyclecounter *cc)
 {
return xilinx_read(NULL);
 }
 
-static struct cyclecounter xilinx_cc = {
-   .read = xilinx_cc_read,
-   .mask = CLOCKSOURCE_MASK(32),
-   .shift = 8,
+static struct timecounter xilinx_tc = {
+   .cc.read = xilinx_cc_read,
+   .cc.mask = CLOCKSOURCE_MASK(32),
+   .cc.mult = 0,
+   .cc.shift = 8,
 };
 
 static int __init init_xilinx_timecounter(void)
 {
-   xilinx_cc.mult = div_sc(timer_clock_freq, NSEC_PER_SEC,
-   xilinx_cc.shift);
+   struct cyclecounter *cc = _tc.cc;
+
+   cc->mult = div_sc(timer_clock_freq, NSEC_PER_SEC, cc->shift);
 
-   timecounter_init(_tc, _cc, sched_clock());
+   timecounter_init(_tc, sched_clock());
 
return 0;
 }
diff --git a/drivers/clocksource/arm_arch_timer.c 
b/drivers/clocksource/arm_arch_timer.c
index 57cb2f0..31543e5 100644
--- a/drivers/clocksource/arm_arch_timer.c
+++ b/drivers/clocksource/arm_arch_timer.c
@@ -179,11 +179,6 @@ static u64 arch_counter_read_cc(const struct cyclecounter 
*cc)
.flags  = CLOCK_SOURCE_IS_CONTINUOUS,
 };
 
-static struct cyclecounter cyclecounter __ro_after_init = {
-   .read   = arch_counter_read_cc,
-   .mask   = CLOCKSOURCE_MASK(56),
-};
-
 struct ate_acpi_oem_info {
char oem_id[ACPI_OEM_ID_SIZE + 1];
char oem_table_id[ACPI_OEM_TABLE_ID_SIZE + 1];
@@ -915,7 +910,10 @@ static u64 arch_counter_get_cntvct_mem(void)
return ((u64) vct_hi << 32) | vct_lo;
 }
 
-static struct arch_timer_kvm_info arch_timer_kvm_info;
+static struct arch_timer_kvm_info arch_timer_kvm_info = {
+   .timecounter.cc.read = arch_counter_read_cc,
+   .timecounter.cc.mask = CLOCKSOURCE_MASK(56),
+};
 
 struct arch_timer_kvm_info *arch_timer_get_kvm_info(void)
 {
@@ -925,6 +923,7 @@ struct arch_timer_kvm_info *arch_timer_get_kvm_info(void)
 static void __init arch_counter_re

[PATCH 1/1] timecounter: Make cyclecounter struct part of timecounter struct

[Sagar Arun Kamble]

There is no real need for the users of timecounters to define cyclecounter
and timecounter variables separately. Since timecounter will always be
based on cyclecounter, have cyclecounter struct as member of timecounter
struct.

Suggested-by: Chris Wilson 
Signed-off-by: Sagar Arun Kamble 
Cc: Chris Wilson 
Cc: John Stultz 
Cc: Thomas Gleixner 
Cc: Stephen Boyd 
Cc: linux-kernel@vger.kernel.org
Cc: linux-arm-ker...@lists.infradead.org
Cc: net...@vger.kernel.org
Cc: intel-wired-...@lists.osuosl.org
Cc: linux-r...@vger.kernel.org
Cc: alsa-de...@alsa-project.org
Cc: kvm...@lists.cs.columbia.edu
---
 arch/microblaze/kernel/timer.c | 20 ++--
 drivers/clocksource/arm_arch_timer.c   | 19 ++--
 drivers/net/ethernet/amd/xgbe/xgbe-dev.c   |  3 +-
 drivers/net/ethernet/amd/xgbe/xgbe-ptp.c   |  9 +++---
 drivers/net/ethernet/amd/xgbe/xgbe.h   |  1 -
 drivers/net/ethernet/broadcom/bnx2x/bnx2x.h|  1 -
 drivers/net/ethernet/broadcom/bnx2x/bnx2x_main.c   | 20 ++--
 drivers/net/ethernet/freescale/fec.h   |  1 -
 drivers/net/ethernet/freescale/fec_ptp.c   | 30 +-
 drivers/net/ethernet/intel/e1000e/e1000.h  |  1 -
 drivers/net/ethernet/intel/e1000e/netdev.c | 27 
 drivers/net/ethernet/intel/e1000e/ptp.c|  2 +-
 drivers/net/ethernet/intel/igb/igb.h   |  1 -
 drivers/net/ethernet/intel/igb/igb_ptp.c   | 25 ---
 drivers/net/ethernet/intel/ixgbe/ixgbe.h   |  1 -
 drivers/net/ethernet/intel/ixgbe/ixgbe_ptp.c   | 17 +-
 drivers/net/ethernet/mellanox/mlx4/en_clock.c  | 28 -
 drivers/net/ethernet/mellanox/mlx4/mlx4_en.h   |  1 -
 .../net/ethernet/mellanox/mlx5/core/lib/clock.c| 34 ++--
 drivers/net/ethernet/qlogic/qede/qede_ptp.c| 20 ++--
 drivers/net/ethernet/ti/cpts.c | 36 --
 drivers/net/ethernet/ti/cpts.h |  1 -
 include/linux/mlx5/driver.h|  1 -
 include/linux/timecounter.h|  4 +--
 include/sound/hdaudio.h|  1 -
 kernel/time/timecounter.c  | 28 -
 sound/hda/hdac_stream.c|  7 +++--
 virt/kvm/arm/arch_timer.c  |  6 ++--
 28 files changed, 163 insertions(+), 182 deletions(-)

diff --git a/arch/microblaze/kernel/timer.c b/arch/microblaze/kernel/timer.c
index 7de941c..b7f89e9 100644
--- a/arch/microblaze/kernel/timer.c
+++ b/arch/microblaze/kernel/timer.c
@@ -199,27 +199,25 @@ static u64 xilinx_read(struct clocksource *cs)
return (u64)xilinx_clock_read();
 }
 
-static struct timecounter xilinx_tc = {
-   .cc = NULL,
-};
-
 static u64 xilinx_cc_read(const struct cyclecounter *cc)
 {
return xilinx_read(NULL);
 }
 
-static struct cyclecounter xilinx_cc = {
-   .read = xilinx_cc_read,
-   .mask = CLOCKSOURCE_MASK(32),
-   .shift = 8,
+static struct timecounter xilinx_tc = {
+   .cc.read = xilinx_cc_read,
+   .cc.mask = CLOCKSOURCE_MASK(32),
+   .cc.mult = 0,
+   .cc.shift = 8,
 };
 
 static int __init init_xilinx_timecounter(void)
 {
-   xilinx_cc.mult = div_sc(timer_clock_freq, NSEC_PER_SEC,
-   xilinx_cc.shift);
+   struct cyclecounter *cc = _tc.cc;
+
+   cc->mult = div_sc(timer_clock_freq, NSEC_PER_SEC, cc->shift);
 
-   timecounter_init(_tc, _cc, sched_clock());
+   timecounter_init(_tc, sched_clock());
 
return 0;
 }
diff --git a/drivers/clocksource/arm_arch_timer.c 
b/drivers/clocksource/arm_arch_timer.c
index 57cb2f0..31543e5 100644
--- a/drivers/clocksource/arm_arch_timer.c
+++ b/drivers/clocksource/arm_arch_timer.c
@@ -179,11 +179,6 @@ static u64 arch_counter_read_cc(const struct cyclecounter 
*cc)
.flags  = CLOCK_SOURCE_IS_CONTINUOUS,
 };
 
-static struct cyclecounter cyclecounter __ro_after_init = {
-   .read   = arch_counter_read_cc,
-   .mask   = CLOCKSOURCE_MASK(56),
-};
-
 struct ate_acpi_oem_info {
char oem_id[ACPI_OEM_ID_SIZE + 1];
char oem_table_id[ACPI_OEM_TABLE_ID_SIZE + 1];
@@ -915,7 +910,10 @@ static u64 arch_counter_get_cntvct_mem(void)
return ((u64) vct_hi << 32) | vct_lo;
 }
 
-static struct arch_timer_kvm_info arch_timer_kvm_info;
+static struct arch_timer_kvm_info arch_timer_kvm_info = {
+   .timecounter.cc.read = arch_counter_read_cc,
+   .timecounter.cc.mask = CLOCKSOURCE_MASK(56),
+};
 
 struct arch_timer_kvm_info *arch_timer_get_kvm_info(void)
 {
@@ -925,6 +923,7 @@ struct arch_timer_kvm_info *arch_timer_get_kvm_info(void)
 static void __init arch_counter_register(unsigned type)
 {
u64 start_count;
+   struct cyclecounter *cc = _timer_kvm_info.timecounter.cc;
 
/* Register the CP15 based counter if we have one */

Re: Creating cyclecounter and lock member in timecounter structure [ Was Re: [RFC 1/4] drm/i915/perf: Add support to correlate GPU timestamp with system time]

[Sagar Arun Kamble]

On 12/1/2017 2:33 AM, Saeed Mahameed wrote:

On Mon, Nov 27, 2017 at 2:05 AM, Sagar Arun Kamble
<sagar.a.kam...@intel.com> wrote:


On 11/24/2017 7:01 PM, Thomas Gleixner wrote:

On Fri, 24 Nov 2017, Sagar Arun Kamble wrote:

On 11/24/2017 12:29 AM, Thomas Gleixner wrote:

On Thu, 23 Nov 2017, Sagar Arun Kamble wrote:

We needed inputs on possible optimization that can be done to
timecounter/cyclecounter structures/usage.
This mail is in response to review of patch
https://patchwork.freedesktop.org/patch/188448/.

As Chris's observation below, about dozen of timecounter users in the
kernel
have below structures
defined individually:

spinlock_t lock;
struct cyclecounter cc;
struct timecounter tc;

Can we move lock and cc to tc? That way it will be convenient.
Also it will allow unifying the locking/overflow watchdog handling
across
all
drivers.

Looks like none of the timecounter usage sites has a real need to
separate
timecounter and cyclecounter.

Yes. Will share patch for this change.


The lock is a different question. The locking of the various drivers
differs and I have no idea how you want to handle that. Just sticking
the
lock into the datastructure and then not making use of it in the
timercounter code and leave it to the callsites does not make sense.

Most of the locks are held around timecounter_read. In some instances it
is held when cyclecounter is updated standalone or is updated along with
timecounter calls.  Was thinking if we move the lock in timecounter
functions, drivers just have to do locking around its operations on
cyclecounter. But then another problem I see is there are variation of
locking calls like lock_irqsave, lock_bh, write_lock_irqsave (some using
rwlock_t). Should this all locking be left to driver only then?

You could have the lock in the struct and protect the inner workings in
the
related core functions.

That might remove locking requirements from some of the callers and the
others still have their own thing around it.


For drivers having static/fixed cyclecounter, we can rely only on lock
inside timecounter.
Most of the network drivers update cyclecounter at runtime and they will
have to rely on two locks if
we add one to timecounter. This may not be efficient for them. Also the lock
in timecounter has to be less restrictive (may be seqlock) I guess.

Cc'd Mellanox list for inputs on this.

I have started feeling that the current approach of drivers managing the
locks is the right one so better leave the
lock out of timecounter.


I agree here,

In mlx5 we rely on our own read/write lock to serialize access to
mlx5_clock struct (mlx5 timecounter and cyclecounter).
the access is not as simple as
lock()
call time_counter_API
unlock()

Sometimes we also explicitly update/adjust timecycles counters with
mlx5 specific calculations after we read the timecounter all inside
our lock.
e.g.
@mlx5_ptp_adjfreq()

 write_lock_irqsave(>lock, flags);
 timecounter_read(>tc);
 clock->cycles.mult = neg_adj ? clock->nominal_c_mult - diff :
clock->nominal_c_mult + diff;
 write_unlock_irqrestore(>lock, flags);

So i don't think it will be a simple task to have a generic thread
safe timecounter API, without the need to specifically adjust it for
all driver use-cases.
Also as said above, in runtime it is not obvious in which context the
timecounter will be accessed irq/soft irq/user.

let's keep it as is, and let the driver decide which locking scheme is
most suitable for it.


Yes. Thanks for your inputs Saeed.

Regards
Sagar



Thanks,
Saeed.


Thanks,

 tglx


--
To unsubscribe from this list: send the line "unsubscribe linux-rdma" in
the body of a message to majord...@vger.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html

Re: Creating cyclecounter and lock member in timecounter structure [ Was Re: [RFC 1/4] drm/i915/perf: Add support to correlate GPU timestamp with system time]

[Sagar Arun Kamble]

On 12/1/2017 2:33 AM, Saeed Mahameed wrote:

On Mon, Nov 27, 2017 at 2:05 AM, Sagar Arun Kamble
 wrote:


On 11/24/2017 7:01 PM, Thomas Gleixner wrote:

On Fri, 24 Nov 2017, Sagar Arun Kamble wrote:

On 11/24/2017 12:29 AM, Thomas Gleixner wrote:

On Thu, 23 Nov 2017, Sagar Arun Kamble wrote:

We needed inputs on possible optimization that can be done to
timecounter/cyclecounter structures/usage.
This mail is in response to review of patch
https://patchwork.freedesktop.org/patch/188448/.

As Chris's observation below, about dozen of timecounter users in the
kernel
have below structures
defined individually:

spinlock_t lock;
struct cyclecounter cc;
struct timecounter tc;

Can we move lock and cc to tc? That way it will be convenient.
Also it will allow unifying the locking/overflow watchdog handling
across
all
drivers.

Looks like none of the timecounter usage sites has a real need to
separate
timecounter and cyclecounter.

Yes. Will share patch for this change.


The lock is a different question. The locking of the various drivers
differs and I have no idea how you want to handle that. Just sticking
the
lock into the datastructure and then not making use of it in the
timercounter code and leave it to the callsites does not make sense.

Most of the locks are held around timecounter_read. In some instances it
is held when cyclecounter is updated standalone or is updated along with
timecounter calls.  Was thinking if we move the lock in timecounter
functions, drivers just have to do locking around its operations on
cyclecounter. But then another problem I see is there are variation of
locking calls like lock_irqsave, lock_bh, write_lock_irqsave (some using
rwlock_t). Should this all locking be left to driver only then?

You could have the lock in the struct and protect the inner workings in
the
related core functions.

That might remove locking requirements from some of the callers and the
others still have their own thing around it.


For drivers having static/fixed cyclecounter, we can rely only on lock
inside timecounter.
Most of the network drivers update cyclecounter at runtime and they will
have to rely on two locks if
we add one to timecounter. This may not be efficient for them. Also the lock
in timecounter has to be less restrictive (may be seqlock) I guess.

Cc'd Mellanox list for inputs on this.

I have started feeling that the current approach of drivers managing the
locks is the right one so better leave the
lock out of timecounter.


I agree here,

In mlx5 we rely on our own read/write lock to serialize access to
mlx5_clock struct (mlx5 timecounter and cyclecounter).
the access is not as simple as
lock()
call time_counter_API
unlock()

Sometimes we also explicitly update/adjust timecycles counters with
mlx5 specific calculations after we read the timecounter all inside
our lock.
e.g.
@mlx5_ptp_adjfreq()

 write_lock_irqsave(>lock, flags);
 timecounter_read(>tc);
 clock->cycles.mult = neg_adj ? clock->nominal_c_mult - diff :
clock->nominal_c_mult + diff;
 write_unlock_irqrestore(>lock, flags);

So i don't think it will be a simple task to have a generic thread
safe timecounter API, without the need to specifically adjust it for
all driver use-cases.
Also as said above, in runtime it is not obvious in which context the
timecounter will be accessed irq/soft irq/user.

let's keep it as is, and let the driver decide which locking scheme is
most suitable for it.


Yes. Thanks for your inputs Saeed.

Regards
Sagar



Thanks,
Saeed.


Thanks,

 tglx


--
To unsubscribe from this list: send the line "unsubscribe linux-rdma" in
the body of a message to majord...@vger.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html

Re: Creating cyclecounter and lock member in timecounter structure [ Was Re: [RFC 1/4] drm/i915/perf: Add support to correlate GPU timestamp with system time]

[Sagar Arun Kamble]

On 11/24/2017 7:01 PM, Thomas Gleixner wrote:

On Fri, 24 Nov 2017, Sagar Arun Kamble wrote:

On 11/24/2017 12:29 AM, Thomas Gleixner wrote:

On Thu, 23 Nov 2017, Sagar Arun Kamble wrote:

We needed inputs on possible optimization that can be done to
timecounter/cyclecounter structures/usage.
This mail is in response to review of patch
https://patchwork.freedesktop.org/patch/188448/.

As Chris's observation below, about dozen of timecounter users in the
kernel
have below structures
defined individually:

spinlock_t lock;
struct cyclecounter cc;
struct timecounter tc;

Can we move lock and cc to tc? That way it will be convenient.
Also it will allow unifying the locking/overflow watchdog handling across
all
drivers.

Looks like none of the timecounter usage sites has a real need to separate
timecounter and cyclecounter.

Yes. Will share patch for this change.


The lock is a different question. The locking of the various drivers
differs and I have no idea how you want to handle that. Just sticking the
lock into the datastructure and then not making use of it in the
timercounter code and leave it to the callsites does not make sense.

Most of the locks are held around timecounter_read. In some instances it
is held when cyclecounter is updated standalone or is updated along with
timecounter calls.  Was thinking if we move the lock in timecounter
functions, drivers just have to do locking around its operations on
cyclecounter. But then another problem I see is there are variation of
locking calls like lock_irqsave, lock_bh, write_lock_irqsave (some using
rwlock_t). Should this all locking be left to driver only then?

You could have the lock in the struct and protect the inner workings in the
related core functions.

That might remove locking requirements from some of the callers and the
others still have their own thing around it.


For drivers having static/fixed cyclecounter, we can rely only on lock inside 
timecounter.
Most of the network drivers update cyclecounter at runtime and they will have 
to rely on two locks if
we add one to timecounter. This may not be efficient for them. Also the lock in 
timecounter has to be less restrictive (may be seqlock) I guess.

Cc'd Mellanox list for inputs on this.

I have started feeling that the current approach of drivers managing the locks 
is the right one so better leave the
lock out of timecounter.


Thanks,

tglx

Re: Creating cyclecounter and lock member in timecounter structure [ Was Re: [RFC 1/4] drm/i915/perf: Add support to correlate GPU timestamp with system time]

[Sagar Arun Kamble]

On 11/24/2017 7:01 PM, Thomas Gleixner wrote:

On Fri, 24 Nov 2017, Sagar Arun Kamble wrote:

On 11/24/2017 12:29 AM, Thomas Gleixner wrote:

On Thu, 23 Nov 2017, Sagar Arun Kamble wrote:

We needed inputs on possible optimization that can be done to
timecounter/cyclecounter structures/usage.
This mail is in response to review of patch
https://patchwork.freedesktop.org/patch/188448/.

As Chris's observation below, about dozen of timecounter users in the
kernel
have below structures
defined individually:

spinlock_t lock;
struct cyclecounter cc;
struct timecounter tc;

Can we move lock and cc to tc? That way it will be convenient.
Also it will allow unifying the locking/overflow watchdog handling across
all
drivers.

Looks like none of the timecounter usage sites has a real need to separate
timecounter and cyclecounter.

Yes. Will share patch for this change.


The lock is a different question. The locking of the various drivers
differs and I have no idea how you want to handle that. Just sticking the
lock into the datastructure and then not making use of it in the
timercounter code and leave it to the callsites does not make sense.

Most of the locks are held around timecounter_read. In some instances it
is held when cyclecounter is updated standalone or is updated along with
timecounter calls.  Was thinking if we move the lock in timecounter
functions, drivers just have to do locking around its operations on
cyclecounter. But then another problem I see is there are variation of
locking calls like lock_irqsave, lock_bh, write_lock_irqsave (some using
rwlock_t). Should this all locking be left to driver only then?

You could have the lock in the struct and protect the inner workings in the
related core functions.

That might remove locking requirements from some of the callers and the
others still have their own thing around it.


For drivers having static/fixed cyclecounter, we can rely only on lock inside 
timecounter.
Most of the network drivers update cyclecounter at runtime and they will have 
to rely on two locks if
we add one to timecounter. This may not be efficient for them. Also the lock in 
timecounter has to be less restrictive (may be seqlock) I guess.

Cc'd Mellanox list for inputs on this.

I have started feeling that the current approach of drivers managing the locks 
is the right one so better leave the
lock out of timecounter.


Thanks,

tglx

[PATCH 1/1] timecounter: Make cyclecounter struct part of timecounter struct

[Sagar Arun Kamble]

There is no real need for the users of timecounters to define cyclecounter
and timecounter variables separately. Since timecounter will always be
based on cyclecounter, have cyclecounter struct as member of timecounter
struct.

Suggested-by: Chris Wilson <ch...@chris-wilson.co.uk>
Signed-off-by: Sagar Arun Kamble <sagar.a.kam...@intel.com>
Cc: John Stultz <john.stu...@linaro.org>
Cc: Thomas Gleixner <t...@linutronix.de>
Cc: Stephen Boyd <sb...@codeaurora.org>
Cc: linux-kernel@vger.kernel.org
---
 arch/microblaze/kernel/timer.c | 20 ++---
 drivers/clocksource/arm_arch_timer.c   | 19 ++--
 drivers/net/ethernet/amd/xgbe/xgbe-dev.c   |  3 +-
 drivers/net/ethernet/amd/xgbe/xgbe-ptp.c   |  9 +++---
 drivers/net/ethernet/amd/xgbe/xgbe.h   |  1 -
 drivers/net/ethernet/broadcom/bnx2x/bnx2x.h|  1 -
 drivers/net/ethernet/broadcom/bnx2x/bnx2x_main.c   | 20 ++---
 drivers/net/ethernet/freescale/fec.h   |  1 -
 drivers/net/ethernet/freescale/fec_ptp.c   | 31 ++-
 drivers/net/ethernet/intel/e1000e/e1000.h  |  1 -
 drivers/net/ethernet/intel/e1000e/netdev.c | 27 -
 drivers/net/ethernet/intel/e1000e/ptp.c|  2 +-
 drivers/net/ethernet/intel/igb/igb.h   |  1 -
 drivers/net/ethernet/intel/igb/igb_ptp.c   | 25 
 drivers/net/ethernet/intel/ixgbe/ixgbe.h   |  1 -
 drivers/net/ethernet/intel/ixgbe/ixgbe_ptp.c   | 17 +--
 drivers/net/ethernet/mellanox/mlx4/en_clock.c  | 28 -
 drivers/net/ethernet/mellanox/mlx4/mlx4_en.h   |  1 -
 drivers/net/ethernet/mellanox/mlx5/core/en.h   |  1 -
 drivers/net/ethernet/mellanox/mlx5/core/en_clock.c | 33 ++--
 drivers/net/ethernet/qlogic/qede/qede_ptp.c| 20 ++---
 drivers/net/ethernet/ti/cpts.c | 35 +++---
 drivers/net/ethernet/ti/cpts.h |  1 -
 include/linux/timecounter.h|  4 +--
 include/sound/hdaudio.h|  1 -
 kernel/time/timecounter.c  | 28 -
 sound/hda/hdac_stream.c|  7 +++--
 27 files changed, 160 insertions(+), 178 deletions(-)

diff --git a/arch/microblaze/kernel/timer.c b/arch/microblaze/kernel/timer.c
index 7de941c..b7f89e9 100644
--- a/arch/microblaze/kernel/timer.c
+++ b/arch/microblaze/kernel/timer.c
@@ -199,27 +199,25 @@ static u64 xilinx_read(struct clocksource *cs)
return (u64)xilinx_clock_read();
 }
 
-static struct timecounter xilinx_tc = {
-   .cc = NULL,
-};
-
 static u64 xilinx_cc_read(const struct cyclecounter *cc)
 {
return xilinx_read(NULL);
 }
 
-static struct cyclecounter xilinx_cc = {
-   .read = xilinx_cc_read,
-   .mask = CLOCKSOURCE_MASK(32),
-   .shift = 8,
+static struct timecounter xilinx_tc = {
+   .cc.read = xilinx_cc_read,
+   .cc.mask = CLOCKSOURCE_MASK(32),
+   .cc.mult = 0,
+   .cc.shift = 8,
 };
 
 static int __init init_xilinx_timecounter(void)
 {
-   xilinx_cc.mult = div_sc(timer_clock_freq, NSEC_PER_SEC,
-   xilinx_cc.shift);
+   struct cyclecounter *cc = _tc.cc;
+
+   cc->mult = div_sc(timer_clock_freq, NSEC_PER_SEC, cc->shift);
 
-   timecounter_init(_tc, _cc, sched_clock());
+   timecounter_init(_tc, sched_clock());
 
return 0;
 }
diff --git a/drivers/clocksource/arm_arch_timer.c 
b/drivers/clocksource/arm_arch_timer.c
index fd4b7f6..1d60ba0 100644
--- a/drivers/clocksource/arm_arch_timer.c
+++ b/drivers/clocksource/arm_arch_timer.c
@@ -177,11 +177,6 @@ static u64 arch_counter_read_cc(const struct cyclecounter 
*cc)
.flags  = CLOCK_SOURCE_IS_CONTINUOUS,
 };
 
-static struct cyclecounter cyclecounter __ro_after_init = {
-   .read   = arch_counter_read_cc,
-   .mask   = CLOCKSOURCE_MASK(56),
-};
-
 struct ate_acpi_oem_info {
char oem_id[ACPI_OEM_ID_SIZE + 1];
char oem_table_id[ACPI_OEM_TABLE_ID_SIZE + 1];
@@ -877,7 +872,10 @@ static u64 arch_counter_get_cntvct_mem(void)
return ((u64) vct_hi << 32) | vct_lo;
 }
 
-static struct arch_timer_kvm_info arch_timer_kvm_info;
+static struct arch_timer_kvm_info arch_timer_kvm_info = {
+   .timecounter.cc.read = arch_counter_read_cc,
+   .timecounter.cc.mask = CLOCKSOURCE_MASK(56),
+};
 
 struct arch_timer_kvm_info *arch_timer_get_kvm_info(void)
 {
@@ -887,6 +885,7 @@ struct arch_timer_kvm_info *arch_timer_get_kvm_info(void)
 static void __init arch_counter_register(unsigned type)
 {
u64 start_count;
+   struct cyclecounter *cc = _timer_kvm_info.timecounter.cc;
 
/* Register the CP15 based counter if we have one */
if (type & ARCH_TIMER_TYPE_CP15) {
@@ -905,10 +904,10 @@ static void __init arch_counter_register(unsigned t

[PATCH 1/1] timecounter: Make cyclecounter struct part of timecounter struct

[Sagar Arun Kamble]

There is no real need for the users of timecounters to define cyclecounter
and timecounter variables separately. Since timecounter will always be
based on cyclecounter, have cyclecounter struct as member of timecounter
struct.

Suggested-by: Chris Wilson 
Signed-off-by: Sagar Arun Kamble 
Cc: John Stultz 
Cc: Thomas Gleixner 
Cc: Stephen Boyd 
Cc: linux-kernel@vger.kernel.org
---
 arch/microblaze/kernel/timer.c | 20 ++---
 drivers/clocksource/arm_arch_timer.c   | 19 ++--
 drivers/net/ethernet/amd/xgbe/xgbe-dev.c   |  3 +-
 drivers/net/ethernet/amd/xgbe/xgbe-ptp.c   |  9 +++---
 drivers/net/ethernet/amd/xgbe/xgbe.h   |  1 -
 drivers/net/ethernet/broadcom/bnx2x/bnx2x.h|  1 -
 drivers/net/ethernet/broadcom/bnx2x/bnx2x_main.c   | 20 ++---
 drivers/net/ethernet/freescale/fec.h   |  1 -
 drivers/net/ethernet/freescale/fec_ptp.c   | 31 ++-
 drivers/net/ethernet/intel/e1000e/e1000.h  |  1 -
 drivers/net/ethernet/intel/e1000e/netdev.c | 27 -
 drivers/net/ethernet/intel/e1000e/ptp.c|  2 +-
 drivers/net/ethernet/intel/igb/igb.h   |  1 -
 drivers/net/ethernet/intel/igb/igb_ptp.c   | 25 
 drivers/net/ethernet/intel/ixgbe/ixgbe.h   |  1 -
 drivers/net/ethernet/intel/ixgbe/ixgbe_ptp.c   | 17 +--
 drivers/net/ethernet/mellanox/mlx4/en_clock.c  | 28 -
 drivers/net/ethernet/mellanox/mlx4/mlx4_en.h   |  1 -
 drivers/net/ethernet/mellanox/mlx5/core/en.h   |  1 -
 drivers/net/ethernet/mellanox/mlx5/core/en_clock.c | 33 ++--
 drivers/net/ethernet/qlogic/qede/qede_ptp.c| 20 ++---
 drivers/net/ethernet/ti/cpts.c | 35 +++---
 drivers/net/ethernet/ti/cpts.h |  1 -
 include/linux/timecounter.h|  4 +--
 include/sound/hdaudio.h|  1 -
 kernel/time/timecounter.c  | 28 -
 sound/hda/hdac_stream.c|  7 +++--
 27 files changed, 160 insertions(+), 178 deletions(-)

diff --git a/arch/microblaze/kernel/timer.c b/arch/microblaze/kernel/timer.c
index 7de941c..b7f89e9 100644
--- a/arch/microblaze/kernel/timer.c
+++ b/arch/microblaze/kernel/timer.c
@@ -199,27 +199,25 @@ static u64 xilinx_read(struct clocksource *cs)
return (u64)xilinx_clock_read();
 }
 
-static struct timecounter xilinx_tc = {
-   .cc = NULL,
-};
-
 static u64 xilinx_cc_read(const struct cyclecounter *cc)
 {
return xilinx_read(NULL);
 }
 
-static struct cyclecounter xilinx_cc = {
-   .read = xilinx_cc_read,
-   .mask = CLOCKSOURCE_MASK(32),
-   .shift = 8,
+static struct timecounter xilinx_tc = {
+   .cc.read = xilinx_cc_read,
+   .cc.mask = CLOCKSOURCE_MASK(32),
+   .cc.mult = 0,
+   .cc.shift = 8,
 };
 
 static int __init init_xilinx_timecounter(void)
 {
-   xilinx_cc.mult = div_sc(timer_clock_freq, NSEC_PER_SEC,
-   xilinx_cc.shift);
+   struct cyclecounter *cc = _tc.cc;
+
+   cc->mult = div_sc(timer_clock_freq, NSEC_PER_SEC, cc->shift);
 
-   timecounter_init(_tc, _cc, sched_clock());
+   timecounter_init(_tc, sched_clock());
 
return 0;
 }
diff --git a/drivers/clocksource/arm_arch_timer.c 
b/drivers/clocksource/arm_arch_timer.c
index fd4b7f6..1d60ba0 100644
--- a/drivers/clocksource/arm_arch_timer.c
+++ b/drivers/clocksource/arm_arch_timer.c
@@ -177,11 +177,6 @@ static u64 arch_counter_read_cc(const struct cyclecounter 
*cc)
.flags  = CLOCK_SOURCE_IS_CONTINUOUS,
 };
 
-static struct cyclecounter cyclecounter __ro_after_init = {
-   .read   = arch_counter_read_cc,
-   .mask   = CLOCKSOURCE_MASK(56),
-};
-
 struct ate_acpi_oem_info {
char oem_id[ACPI_OEM_ID_SIZE + 1];
char oem_table_id[ACPI_OEM_TABLE_ID_SIZE + 1];
@@ -877,7 +872,10 @@ static u64 arch_counter_get_cntvct_mem(void)
return ((u64) vct_hi << 32) | vct_lo;
 }
 
-static struct arch_timer_kvm_info arch_timer_kvm_info;
+static struct arch_timer_kvm_info arch_timer_kvm_info = {
+   .timecounter.cc.read = arch_counter_read_cc,
+   .timecounter.cc.mask = CLOCKSOURCE_MASK(56),
+};
 
 struct arch_timer_kvm_info *arch_timer_get_kvm_info(void)
 {
@@ -887,6 +885,7 @@ struct arch_timer_kvm_info *arch_timer_get_kvm_info(void)
 static void __init arch_counter_register(unsigned type)
 {
u64 start_count;
+   struct cyclecounter *cc = _timer_kvm_info.timecounter.cc;
 
/* Register the CP15 based counter if we have one */
if (type & ARCH_TIMER_TYPE_CP15) {
@@ -905,10 +904,10 @@ static void __init arch_counter_register(unsigned type)
clocksource_counter.flags |= CLOCK_SOURCE_SUSPEND_NONSTOP;
start_count = arch_timer_read_counter();
clocksource_regi

Re: Creating cyclecounter and lock member in timecounter structure [ Was Re: [RFC 1/4] drm/i915/perf: Add support to correlate GPU timestamp with system time]

[Sagar Arun Kamble]

On 11/24/2017 12:29 AM, Thomas Gleixner wrote:

On Thu, 23 Nov 2017, Sagar Arun Kamble wrote:

We needed inputs on possible optimization that can be done to
timecounter/cyclecounter structures/usage.
This mail is in response to review of patch
https://patchwork.freedesktop.org/patch/188448/.

As Chris's observation below, about dozen of timecounter users in the kernel
have below structures
defined individually:

spinlock_t lock;
struct cyclecounter cc;
struct timecounter tc;

Can we move lock and cc to tc? That way it will be convenient.
Also it will allow unifying the locking/overflow watchdog handling across all
drivers.

Looks like none of the timecounter usage sites has a real need to separate
timecounter and cyclecounter.


Yes. Will share patch for this change.


The lock is a different question. The locking of the various drivers
differs and I have no idea how you want to handle that. Just sticking the
lock into the datastructure and then not making use of it in the
timercounter code and leave it to the callsites does not make sense.


Most of the locks are held around timecounter_read. In some instances it is 
held when cyclecounter is
updated standalone or is updated along with timecounter calls.
Was thinking if we move the lock in timecounter functions, drivers just have to 
do locking around its
operations on cyclecounter. But then another problem I see is there are 
variation of locking calls
like lock_irqsave, lock_bh, write_lock_irqsave (some using rwlock_t). Should 
this all locking be left
to driver only then?


Thanks,

tglx


Thanks
Sagar

Re: Creating cyclecounter and lock member in timecounter structure [ Was Re: [RFC 1/4] drm/i915/perf: Add support to correlate GPU timestamp with system time]

[Sagar Arun Kamble]

On 11/24/2017 12:29 AM, Thomas Gleixner wrote:

On Thu, 23 Nov 2017, Sagar Arun Kamble wrote:

We needed inputs on possible optimization that can be done to
timecounter/cyclecounter structures/usage.
This mail is in response to review of patch
https://patchwork.freedesktop.org/patch/188448/.

As Chris's observation below, about dozen of timecounter users in the kernel
have below structures
defined individually:

spinlock_t lock;
struct cyclecounter cc;
struct timecounter tc;

Can we move lock and cc to tc? That way it will be convenient.
Also it will allow unifying the locking/overflow watchdog handling across all
drivers.

Looks like none of the timecounter usage sites has a real need to separate
timecounter and cyclecounter.


Yes. Will share patch for this change.


The lock is a different question. The locking of the various drivers
differs and I have no idea how you want to handle that. Just sticking the
lock into the datastructure and then not making use of it in the
timercounter code and leave it to the callsites does not make sense.


Most of the locks are held around timecounter_read. In some instances it is 
held when cyclecounter is
updated standalone or is updated along with timecounter calls.
Was thinking if we move the lock in timecounter functions, drivers just have to 
do locking around its
operations on cyclecounter. But then another problem I see is there are 
variation of locking calls
like lock_irqsave, lock_bh, write_lock_irqsave (some using rwlock_t). Should 
this all locking be left
to driver only then?


Thanks,

tglx


Thanks
Sagar

Creating cyclecounter and lock member in timecounter structure [ Was Re: [RFC 1/4] drm/i915/perf: Add support to correlate GPU timestamp with system time]

[Sagar Arun Kamble]

Hi,

We needed inputs on possible optimization that can be done to 
timecounter/cyclecounter structures/usage.
This mail is in response to review of patch 
https://patchwork.freedesktop.org/patch/188448/.

As Chris's observation below, about dozen of timecounter users in the kernel 
have below structures
defined individually:

spinlock_t lock;
struct cyclecounter cc;
struct timecounter tc;

Can we move lock and cc to tc? That way it will be convenient.
Also it will allow unifying the locking/overflow watchdog handling across all 
drivers.

Please suggest.

Thanks
Sagar

On 11/15/2017 5:55 PM, Chris Wilson wrote:

Quoting Sagar Arun Kamble (2017-11-15 12:13:51)

  #include 
  #include 
@@ -2149,6 +2150,14 @@ struct i915_perf_stream {
  * @oa_config: The OA configuration used by the stream.
  */
 struct i915_oa_config *oa_config;
+
+   /**
+* System time correlation variables.
+*/
+   struct cyclecounter cc;
+   spinlock_t systime_lock;
+   struct timespec64 start_systime;
+   struct timecounter tc;

This pattern is repeated a lot by struct timecounter users. (I'm still
trying to understand why the common case is not catered for by a
convenience timecounter api.)


  };
  
  /**

diff --git a/drivers/gpu/drm/i915/i915_perf.c b/drivers/gpu/drm/i915/i915_perf.c
index 00be015..72ddc34 100644
--- a/drivers/gpu/drm/i915/i915_perf.c
+++ b/drivers/gpu/drm/i915/i915_perf.c
@@ -192,6 +192,7 @@
   */
  
  #include 

+#include 
  #include 
  #include 
  
@@ -2391,6 +2392,56 @@ static unsigned int i915_perf_poll(struct file *file, poll_table *wait)

  }
  
  /**

+ * i915_cyclecounter_read - read raw cycle/timestamp counter
+ * @cc: cyclecounter structure
+ */
+static u64 i915_cyclecounter_read(const struct cyclecounter *cc)
+{
+   struct i915_perf_stream *stream = container_of(cc, typeof(*stream), cc);
+   struct drm_i915_private *dev_priv = stream->dev_priv;
+   u64 ts_count;
+
+   intel_runtime_pm_get(dev_priv);
+   ts_count = I915_READ64_2x32(GEN4_TIMESTAMP,
+   GEN7_TIMESTAMP_UDW);
+   intel_runtime_pm_put(dev_priv);
+
+   return ts_count;
+}
+
+static void i915_perf_init_cyclecounter(struct i915_perf_stream *stream)
+{
+   struct drm_i915_private *dev_priv = stream->dev_priv;
+   int cs_ts_freq = dev_priv->perf.oa.timestamp_frequency;
+   struct cyclecounter *cc = >cc;
+   u32 maxsec;
+
+   cc->read = i915_cyclecounter_read;
+   cc->mask = CYCLECOUNTER_MASK(CS_TIMESTAMP_WIDTH(dev_priv));
+   maxsec = cc->mask / cs_ts_freq;
+
+   clocks_calc_mult_shift(>mult, >shift, cs_ts_freq,
+  NSEC_PER_SEC, maxsec);
+}
+
+static void i915_perf_init_timecounter(struct i915_perf_stream *stream)
+{
+#define SYSTIME_START_OFFSET   35 /* Counter read takes about 350us */
+   unsigned long flags;
+   u64 ns;
+
+   i915_perf_init_cyclecounter(stream);
+   spin_lock_init(>systime_lock);
+
+   getnstimeofday64(>start_systime);
+   ns = timespec64_to_ns(>start_systime) + SYSTIME_START_OFFSET;

Use ktime directly. Or else Arnd will be back with a patch to fix it.
(All non-ktime interfaces are effectively deprecated; obsolete for
drivers.)


+   spin_lock_irqsave(>systime_lock, flags);
+   timecounter_init(>tc, >cc, ns);
+   spin_unlock_irqrestore(>systime_lock, flags);
+}
+
+/**
   * i915_perf_enable_locked - handle `I915_PERF_IOCTL_ENABLE` ioctl
   * @stream: A disabled i915 perf stream
   *
@@ -2408,6 +2459,8 @@ static void i915_perf_enable_locked(struct 
i915_perf_stream *stream)
 /* Allow stream->ops->enable() to refer to this */
 stream->enabled = true;
  
+   i915_perf_init_timecounter(stream);

+
 if (stream->ops->enable)
 stream->ops->enable(stream);
  }
diff --git a/drivers/gpu/drm/i915/i915_reg.h b/drivers/gpu/drm/i915/i915_reg.h
index cfdf4f8..e7e6966 100644
--- a/drivers/gpu/drm/i915/i915_reg.h
+++ b/drivers/gpu/drm/i915/i915_reg.h
@@ -8882,6 +8882,12 @@ enum skl_power_gate {
  
  /* Gen4+ Timestamp and Pipe Frame time stamp registers */

  #define GEN4_TIMESTAMP _MMIO(0x2358)
+#define GEN7_TIMESTAMP_UDW _MMIO(0x235C)
+#define PRE_GEN7_TIMESTAMP_WIDTH   32
+#define GEN7_TIMESTAMP_WIDTH   36
+#define CS_TIMESTAMP_WIDTH(dev_priv) \
+   (INTEL_GEN(dev_priv) < 7 ? PRE_GEN7_TIMESTAMP_WIDTH : \
+  GEN7_TIMESTAMP_WIDTH)

s/PRE_GEN7/GEN4/ would be consistent.
If you really want to add support for earlier, I9XX_.

Ok. I can accept the justification, and we are not the only ones who do
the cyclecounter -> timecounter correction like this.
-Chris

Creating cyclecounter and lock member in timecounter structure [ Was Re: [RFC 1/4] drm/i915/perf: Add support to correlate GPU timestamp with system time]

[Sagar Arun Kamble]

Hi,

We needed inputs on possible optimization that can be done to 
timecounter/cyclecounter structures/usage.
This mail is in response to review of patch 
https://patchwork.freedesktop.org/patch/188448/.

As Chris's observation below, about dozen of timecounter users in the kernel 
have below structures
defined individually:

spinlock_t lock;
struct cyclecounter cc;
struct timecounter tc;

Can we move lock and cc to tc? That way it will be convenient.
Also it will allow unifying the locking/overflow watchdog handling across all 
drivers.

Please suggest.

Thanks
Sagar

On 11/15/2017 5:55 PM, Chris Wilson wrote:

Quoting Sagar Arun Kamble (2017-11-15 12:13:51)

  #include 
  #include 
@@ -2149,6 +2150,14 @@ struct i915_perf_stream {
  * @oa_config: The OA configuration used by the stream.
  */
 struct i915_oa_config *oa_config;
+
+   /**
+* System time correlation variables.
+*/
+   struct cyclecounter cc;
+   spinlock_t systime_lock;
+   struct timespec64 start_systime;
+   struct timecounter tc;

This pattern is repeated a lot by struct timecounter users. (I'm still
trying to understand why the common case is not catered for by a
convenience timecounter api.)


  };
  
  /**

diff --git a/drivers/gpu/drm/i915/i915_perf.c b/drivers/gpu/drm/i915/i915_perf.c
index 00be015..72ddc34 100644
--- a/drivers/gpu/drm/i915/i915_perf.c
+++ b/drivers/gpu/drm/i915/i915_perf.c
@@ -192,6 +192,7 @@
   */
  
  #include 

+#include 
  #include 
  #include 
  
@@ -2391,6 +2392,56 @@ static unsigned int i915_perf_poll(struct file *file, poll_table *wait)

  }
  
  /**

+ * i915_cyclecounter_read - read raw cycle/timestamp counter
+ * @cc: cyclecounter structure
+ */
+static u64 i915_cyclecounter_read(const struct cyclecounter *cc)
+{
+   struct i915_perf_stream *stream = container_of(cc, typeof(*stream), cc);
+   struct drm_i915_private *dev_priv = stream->dev_priv;
+   u64 ts_count;
+
+   intel_runtime_pm_get(dev_priv);
+   ts_count = I915_READ64_2x32(GEN4_TIMESTAMP,
+   GEN7_TIMESTAMP_UDW);
+   intel_runtime_pm_put(dev_priv);
+
+   return ts_count;
+}
+
+static void i915_perf_init_cyclecounter(struct i915_perf_stream *stream)
+{
+   struct drm_i915_private *dev_priv = stream->dev_priv;
+   int cs_ts_freq = dev_priv->perf.oa.timestamp_frequency;
+   struct cyclecounter *cc = >cc;
+   u32 maxsec;
+
+   cc->read = i915_cyclecounter_read;
+   cc->mask = CYCLECOUNTER_MASK(CS_TIMESTAMP_WIDTH(dev_priv));
+   maxsec = cc->mask / cs_ts_freq;
+
+   clocks_calc_mult_shift(>mult, >shift, cs_ts_freq,
+  NSEC_PER_SEC, maxsec);
+}
+
+static void i915_perf_init_timecounter(struct i915_perf_stream *stream)
+{
+#define SYSTIME_START_OFFSET   35 /* Counter read takes about 350us */
+   unsigned long flags;
+   u64 ns;
+
+   i915_perf_init_cyclecounter(stream);
+   spin_lock_init(>systime_lock);
+
+   getnstimeofday64(>start_systime);
+   ns = timespec64_to_ns(>start_systime) + SYSTIME_START_OFFSET;

Use ktime directly. Or else Arnd will be back with a patch to fix it.
(All non-ktime interfaces are effectively deprecated; obsolete for
drivers.)


+   spin_lock_irqsave(>systime_lock, flags);
+   timecounter_init(>tc, >cc, ns);
+   spin_unlock_irqrestore(>systime_lock, flags);
+}
+
+/**
   * i915_perf_enable_locked - handle `I915_PERF_IOCTL_ENABLE` ioctl
   * @stream: A disabled i915 perf stream
   *
@@ -2408,6 +2459,8 @@ static void i915_perf_enable_locked(struct 
i915_perf_stream *stream)
 /* Allow stream->ops->enable() to refer to this */
 stream->enabled = true;
  
+   i915_perf_init_timecounter(stream);

+
 if (stream->ops->enable)
 stream->ops->enable(stream);
  }
diff --git a/drivers/gpu/drm/i915/i915_reg.h b/drivers/gpu/drm/i915/i915_reg.h
index cfdf4f8..e7e6966 100644
--- a/drivers/gpu/drm/i915/i915_reg.h
+++ b/drivers/gpu/drm/i915/i915_reg.h
@@ -8882,6 +8882,12 @@ enum skl_power_gate {
  
  /* Gen4+ Timestamp and Pipe Frame time stamp registers */

  #define GEN4_TIMESTAMP _MMIO(0x2358)
+#define GEN7_TIMESTAMP_UDW _MMIO(0x235C)
+#define PRE_GEN7_TIMESTAMP_WIDTH   32
+#define GEN7_TIMESTAMP_WIDTH   36
+#define CS_TIMESTAMP_WIDTH(dev_priv) \
+   (INTEL_GEN(dev_priv) < 7 ? PRE_GEN7_TIMESTAMP_WIDTH : \
+  GEN7_TIMESTAMP_WIDTH)

s/PRE_GEN7/GEN4/ would be consistent.
If you really want to add support for earlier, I9XX_.

Ok. I can accept the justification, and we are not the only ones who do
the cyclecounter -> timecounter correction like this.
-Chris

[PATCH 1/2] iommu: Disable preemption around use of this_cpu_ptr()

[Sagar Arun Kamble]

From: Chris Wilson 

Between acquiring the this_cpu_ptr() and using it, ideally we don't want
to be preempted and work on another CPU's private data. this_cpu_ptr()
checks whether or not preemption is disable, and get_cpu_ptr() provides
a convenient wrapper for operating on the cpu ptr inside a preemption
disabled critical section (which currently is provided by the
spinlock). Indeed if we disable preemption around this_cpu_ptr,
we do not need the CPU local spinlock - so long as take care that no other
CPU is running that code as do perform the cross-CPU cache flushing and
teardown, but that is a subject for another patch.

[  167.997877] BUG: using smp_processor_id() in preemptible [] code: 
usb-storage/216
[  167.997940] caller is debug_smp_processor_id+0x17/0x20
[  167.997945] CPU: 7 PID: 216 Comm: usb-storage Tainted: G U  
4.7.0-rc1-gfxbench-RO_Patchwork_1057+ #1
[  167.997948] Hardware name: Hewlett-Packard HP Pro 3500 Series/2ABF, BIOS 
8.11 10/24/2012
[  167.997951]   880118b7f9c8 8140dca5 
0007
[  167.997958]  81a3a7e9 880118b7f9f8 8142a927 

[  167.997965]  8800d499ed58 0001 000f 
880118b7fa08
[  167.997971] Call Trace:
[  167.997977]  [] dump_stack+0x67/0x92
[  167.997981]  [] check_preemption_disabled+0xd7/0xe0
[  167.997985]  [] debug_smp_processor_id+0x17/0x20
[  167.997990]  [] alloc_iova_fast+0xb7/0x210
[  167.997994]  [] intel_alloc_iova+0x7f/0xd0
[  167.997998]  [] intel_map_sg+0xbd/0x240
[  167.998002]  [] ? debug_lockdep_rcu_enabled+0x1d/0x20
[  167.998009]  [] usb_hcd_map_urb_for_dma+0x4b9/0x5a0
[  167.998013]  [] usb_hcd_submit_urb+0xe9/0xaa0
[  167.998017]  [] ? mark_held_locks+0x6f/0xa0
[  167.998022]  [] ? __raw_spin_lock_init+0x1c/0x50
[  167.998025]  [] ? debug_lockdep_rcu_enabled+0x1d/0x20
[  167.998028]  [] usb_submit_urb+0x3f3/0x5a0
[  167.998032]  [] ? trace_hardirqs_on_caller+0x122/0x1b0
[  167.998035]  [] usb_sg_wait+0x67/0x150
[  167.998039]  [] 
usb_stor_bulk_transfer_sglist.part.3+0x82/0xd0
[  167.998042]  [] usb_stor_bulk_srb+0x4c/0x60
[  167.998045]  [] usb_stor_Bulk_transport+0x17e/0x420
[  167.998049]  [] usb_stor_invoke_transport+0x242/0x540
[  167.998052]  [] ? debug_lockdep_rcu_enabled+0x1d/0x20
[  167.998058]  [] usb_stor_transparent_scsi_command+0x9/0x10
[  167.998061]  [] usb_stor_control_thread+0x158/0x260
[  167.998064]  [] ? fill_inquiry_response+0x20/0x20
[  167.998067]  [] ? fill_inquiry_response+0x20/0x20
[  167.998071]  [] kthread+0xea/0x100
[  167.998078]  [] ret_from_fork+0x1f/0x40
[  167.998081]  [] ? kthread_create_on_node+0x1f0/0x1f0

v2: convert preempt_disable(); var = this_cpu_ptr() to var = get_cpu_ptr()
v3: Actually use get_cpu_ptr (not get_cpu_var). Drop the spinlock
removal, concentrate on the immediate bug fix.

Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=96293
Signed-off-by: Chris Wilson 
Cc: Joonas Lahtinen 
Cc: Joerg Roedel 
Cc: io...@lists.linux-foundation.org
Cc: linux-kernel@vger.kernel.org
Reviewed-by: Joonas Lahtinen 
---
 drivers/iommu/iova.c | 8 ++--
 1 file changed, 6 insertions(+), 2 deletions(-)

diff --git a/drivers/iommu/iova.c b/drivers/iommu/iova.c
index ba764a0..e23001b 100644
--- a/drivers/iommu/iova.c
+++ b/drivers/iommu/iova.c
@@ -420,8 +420,10 @@ retry:
 
/* Try replenishing IOVAs by flushing rcache. */
flushed_rcache = true;
+   preempt_disable();
for_each_online_cpu(cpu)
free_cpu_cached_iovas(cpu, iovad);
+   preempt_enable();
goto retry;
}
 
@@ -749,7 +751,7 @@ static bool __iova_rcache_insert(struct iova_domain *iovad,
bool can_insert = false;
unsigned long flags;
 
-   cpu_rcache = this_cpu_ptr(rcache->cpu_rcaches);
+   cpu_rcache = get_cpu_ptr(rcache->cpu_rcaches);
spin_lock_irqsave(_rcache->lock, flags);
 
if (!iova_magazine_full(cpu_rcache->loaded)) {
@@ -779,6 +781,7 @@ static bool __iova_rcache_insert(struct iova_domain *iovad,
iova_magazine_push(cpu_rcache->loaded, iova_pfn);
 
spin_unlock_irqrestore(_rcache->lock, flags);
+   put_cpu_ptr(rcache->cpu_rcaches);
 
if (mag_to_free) {
iova_magazine_free_pfns(mag_to_free, iovad);
@@ -812,7 +815,7 @@ static unsigned long __iova_rcache_get(struct iova_rcache 
*rcache,
bool has_pfn = false;
unsigned long flags;
 
-   cpu_rcache = this_cpu_ptr(rcache->cpu_rcaches);
+   cpu_rcache = get_cpu_ptr(rcache->cpu_rcaches);
spin_lock_irqsave(_rcache->lock, flags);
 
if (!iova_magazine_empty(cpu_rcache->loaded)) {
@@ -834,6 +837,7 @@ static unsigned long __iova_rcache_get(struct iova_rcache 
*rcache,
iova_pfn = 

[PATCH 1/2] iommu: Disable preemption around use of this_cpu_ptr()

[Sagar Arun Kamble]

From: Chris Wilson 

Between acquiring the this_cpu_ptr() and using it, ideally we don't want
to be preempted and work on another CPU's private data. this_cpu_ptr()
checks whether or not preemption is disable, and get_cpu_ptr() provides
a convenient wrapper for operating on the cpu ptr inside a preemption
disabled critical section (which currently is provided by the
spinlock). Indeed if we disable preemption around this_cpu_ptr,
we do not need the CPU local spinlock - so long as take care that no other
CPU is running that code as do perform the cross-CPU cache flushing and
teardown, but that is a subject for another patch.

[  167.997877] BUG: using smp_processor_id() in preemptible [] code: 
usb-storage/216
[  167.997940] caller is debug_smp_processor_id+0x17/0x20
[  167.997945] CPU: 7 PID: 216 Comm: usb-storage Tainted: G U  
4.7.0-rc1-gfxbench-RO_Patchwork_1057+ #1
[  167.997948] Hardware name: Hewlett-Packard HP Pro 3500 Series/2ABF, BIOS 
8.11 10/24/2012
[  167.997951]   880118b7f9c8 8140dca5 
0007
[  167.997958]  81a3a7e9 880118b7f9f8 8142a927 

[  167.997965]  8800d499ed58 0001 000f 
880118b7fa08
[  167.997971] Call Trace:
[  167.997977]  [] dump_stack+0x67/0x92
[  167.997981]  [] check_preemption_disabled+0xd7/0xe0
[  167.997985]  [] debug_smp_processor_id+0x17/0x20
[  167.997990]  [] alloc_iova_fast+0xb7/0x210
[  167.997994]  [] intel_alloc_iova+0x7f/0xd0
[  167.997998]  [] intel_map_sg+0xbd/0x240
[  167.998002]  [] ? debug_lockdep_rcu_enabled+0x1d/0x20
[  167.998009]  [] usb_hcd_map_urb_for_dma+0x4b9/0x5a0
[  167.998013]  [] usb_hcd_submit_urb+0xe9/0xaa0
[  167.998017]  [] ? mark_held_locks+0x6f/0xa0
[  167.998022]  [] ? __raw_spin_lock_init+0x1c/0x50
[  167.998025]  [] ? debug_lockdep_rcu_enabled+0x1d/0x20
[  167.998028]  [] usb_submit_urb+0x3f3/0x5a0
[  167.998032]  [] ? trace_hardirqs_on_caller+0x122/0x1b0
[  167.998035]  [] usb_sg_wait+0x67/0x150
[  167.998039]  [] 
usb_stor_bulk_transfer_sglist.part.3+0x82/0xd0
[  167.998042]  [] usb_stor_bulk_srb+0x4c/0x60
[  167.998045]  [] usb_stor_Bulk_transport+0x17e/0x420
[  167.998049]  [] usb_stor_invoke_transport+0x242/0x540
[  167.998052]  [] ? debug_lockdep_rcu_enabled+0x1d/0x20
[  167.998058]  [] usb_stor_transparent_scsi_command+0x9/0x10
[  167.998061]  [] usb_stor_control_thread+0x158/0x260
[  167.998064]  [] ? fill_inquiry_response+0x20/0x20
[  167.998067]  [] ? fill_inquiry_response+0x20/0x20
[  167.998071]  [] kthread+0xea/0x100
[  167.998078]  [] ret_from_fork+0x1f/0x40
[  167.998081]  [] ? kthread_create_on_node+0x1f0/0x1f0

v2: convert preempt_disable(); var = this_cpu_ptr() to var = get_cpu_ptr()
v3: Actually use get_cpu_ptr (not get_cpu_var). Drop the spinlock
removal, concentrate on the immediate bug fix.

Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=96293
Signed-off-by: Chris Wilson 
Cc: Joonas Lahtinen 
Cc: Joerg Roedel 
Cc: io...@lists.linux-foundation.org
Cc: linux-kernel@vger.kernel.org
Reviewed-by: Joonas Lahtinen 
---
 drivers/iommu/iova.c | 8 ++--
 1 file changed, 6 insertions(+), 2 deletions(-)

diff --git a/drivers/iommu/iova.c b/drivers/iommu/iova.c
index ba764a0..e23001b 100644
--- a/drivers/iommu/iova.c
+++ b/drivers/iommu/iova.c
@@ -420,8 +420,10 @@ retry:
 
/* Try replenishing IOVAs by flushing rcache. */
flushed_rcache = true;
+   preempt_disable();
for_each_online_cpu(cpu)
free_cpu_cached_iovas(cpu, iovad);
+   preempt_enable();
goto retry;
}
 
@@ -749,7 +751,7 @@ static bool __iova_rcache_insert(struct iova_domain *iovad,
bool can_insert = false;
unsigned long flags;
 
-   cpu_rcache = this_cpu_ptr(rcache->cpu_rcaches);
+   cpu_rcache = get_cpu_ptr(rcache->cpu_rcaches);
spin_lock_irqsave(_rcache->lock, flags);
 
if (!iova_magazine_full(cpu_rcache->loaded)) {
@@ -779,6 +781,7 @@ static bool __iova_rcache_insert(struct iova_domain *iovad,
iova_magazine_push(cpu_rcache->loaded, iova_pfn);
 
spin_unlock_irqrestore(_rcache->lock, flags);
+   put_cpu_ptr(rcache->cpu_rcaches);
 
if (mag_to_free) {
iova_magazine_free_pfns(mag_to_free, iovad);
@@ -812,7 +815,7 @@ static unsigned long __iova_rcache_get(struct iova_rcache 
*rcache,
bool has_pfn = false;
unsigned long flags;
 
-   cpu_rcache = this_cpu_ptr(rcache->cpu_rcaches);
+   cpu_rcache = get_cpu_ptr(rcache->cpu_rcaches);
spin_lock_irqsave(_rcache->lock, flags);
 
if (!iova_magazine_empty(cpu_rcache->loaded)) {
@@ -834,6 +837,7 @@ static unsigned long __iova_rcache_get(struct iova_rcache 
*rcache,
iova_pfn = iova_magazine_pop(cpu_rcache->loaded, limit_pfn);
 
spin_unlock_irqrestore(_rcache->lock, flags);
+   put_cpu_ptr(rcache->cpu_rcaches);
 
return 

[PATCH 2/2] iommu: Remove cpu-local spinlock

[Sagar Arun Kamble]

From: Chris Wilson 

By avoiding cross-CPU usage of the per-cpu iova cache, we can forgo
having a spinlock inside the per-cpu struct. The only place where we
actually may touch another CPU's data is when performing a cache flush
after running out of memory. Here, we can instead schedule a task to run
on the other CPU to do the flush before trying again.

Signed-off-by: Chris Wilson 
Cc: Joonas Lahtinen 
Cc: Joerg Roedel 
Cc: io...@lists.linux-foundation.org
Cc: linux-kernel@vger.kernel.org
Reviewed-by: Joonas Lahtinen 
---
 drivers/iommu/iova.c | 29 ++---
 1 file changed, 6 insertions(+), 23 deletions(-)

diff --git a/drivers/iommu/iova.c b/drivers/iommu/iova.c
index e23001b..36cdc8e 100644
--- a/drivers/iommu/iova.c
+++ b/drivers/iommu/iova.c
@@ -390,6 +390,11 @@ free_iova(struct iova_domain *iovad, unsigned long pfn)
 }
 EXPORT_SYMBOL_GPL(free_iova);
 
+static void free_this_cached_iovas(void *info)
+{
+   free_cpu_cached_iovas(smp_processor_id(), info);
+}
+
 /**
  * alloc_iova_fast - allocates an iova from rcache
  * @iovad: - iova domain in question
@@ -413,17 +418,12 @@ alloc_iova_fast(struct iova_domain *iovad, unsigned long 
size,
 retry:
new_iova = alloc_iova(iovad, size, limit_pfn, true);
if (!new_iova) {
-   unsigned int cpu;
-
if (flushed_rcache)
return 0;
 
/* Try replenishing IOVAs by flushing rcache. */
flushed_rcache = true;
-   preempt_disable();
-   for_each_online_cpu(cpu)
-   free_cpu_cached_iovas(cpu, iovad);
-   preempt_enable();
+   on_each_cpu(free_this_cached_iovas, iovad, true);
goto retry;
}
 
@@ -647,7 +647,6 @@ struct iova_magazine {
 };
 
 struct iova_cpu_rcache {
-   spinlock_t lock;
struct iova_magazine *loaded;
struct iova_magazine *prev;
 };
@@ -729,7 +728,6 @@ static void init_iova_rcaches(struct iova_domain *iovad)
continue;
for_each_possible_cpu(cpu) {
cpu_rcache = per_cpu_ptr(rcache->cpu_rcaches, cpu);
-   spin_lock_init(_rcache->lock);
cpu_rcache->loaded = iova_magazine_alloc(GFP_KERNEL);
cpu_rcache->prev = iova_magazine_alloc(GFP_KERNEL);
}
@@ -749,10 +747,8 @@ static bool __iova_rcache_insert(struct iova_domain *iovad,
struct iova_magazine *mag_to_free = NULL;
struct iova_cpu_rcache *cpu_rcache;
bool can_insert = false;
-   unsigned long flags;
 
cpu_rcache = get_cpu_ptr(rcache->cpu_rcaches);
-   spin_lock_irqsave(_rcache->lock, flags);
 
if (!iova_magazine_full(cpu_rcache->loaded)) {
can_insert = true;
@@ -780,7 +776,6 @@ static bool __iova_rcache_insert(struct iova_domain *iovad,
if (can_insert)
iova_magazine_push(cpu_rcache->loaded, iova_pfn);
 
-   spin_unlock_irqrestore(_rcache->lock, flags);
put_cpu_ptr(rcache->cpu_rcaches);
 
if (mag_to_free) {
@@ -813,10 +808,8 @@ static unsigned long __iova_rcache_get(struct iova_rcache 
*rcache,
struct iova_cpu_rcache *cpu_rcache;
unsigned long iova_pfn = 0;
bool has_pfn = false;
-   unsigned long flags;
 
cpu_rcache = get_cpu_ptr(rcache->cpu_rcaches);
-   spin_lock_irqsave(_rcache->lock, flags);
 
if (!iova_magazine_empty(cpu_rcache->loaded)) {
has_pfn = true;
@@ -836,7 +829,6 @@ static unsigned long __iova_rcache_get(struct iova_rcache 
*rcache,
if (has_pfn)
iova_pfn = iova_magazine_pop(cpu_rcache->loaded, limit_pfn);
 
-   spin_unlock_irqrestore(_rcache->lock, flags);
put_cpu_ptr(rcache->cpu_rcaches);
 
return iova_pfn;
@@ -866,17 +858,11 @@ static void free_cpu_iova_rcache(unsigned int cpu, struct 
iova_domain *iovad,
 struct iova_rcache *rcache)
 {
struct iova_cpu_rcache *cpu_rcache = per_cpu_ptr(rcache->cpu_rcaches, 
cpu);
-   unsigned long flags;
-
-   spin_lock_irqsave(_rcache->lock, flags);
-
iova_magazine_free_pfns(cpu_rcache->loaded, iovad);
iova_magazine_free(cpu_rcache->loaded);
 
iova_magazine_free_pfns(cpu_rcache->prev, iovad);
iova_magazine_free(cpu_rcache->prev);
-
-   spin_unlock_irqrestore(_rcache->lock, flags);
 }
 
 /*
@@ -910,16 +896,13 @@ void free_cpu_cached_iovas(unsigned int cpu, struct 
iova_domain *iovad)
 {
struct iova_cpu_rcache *cpu_rcache;
struct iova_rcache *rcache;
-   unsigned long flags;
int i;
 
for (i = 0; i < IOVA_RANGE_CACHE_MAX_SIZE; ++i) {
rcache = >rcaches[i];
cpu_rcache = per_cpu_ptr(rcache->cpu_rcaches, 

[PATCH 2/2] iommu: Remove cpu-local spinlock

[Sagar Arun Kamble]

From: Chris Wilson 

By avoiding cross-CPU usage of the per-cpu iova cache, we can forgo
having a spinlock inside the per-cpu struct. The only place where we
actually may touch another CPU's data is when performing a cache flush
after running out of memory. Here, we can instead schedule a task to run
on the other CPU to do the flush before trying again.

Signed-off-by: Chris Wilson 
Cc: Joonas Lahtinen 
Cc: Joerg Roedel 
Cc: io...@lists.linux-foundation.org
Cc: linux-kernel@vger.kernel.org
Reviewed-by: Joonas Lahtinen 
---
 drivers/iommu/iova.c | 29 ++---
 1 file changed, 6 insertions(+), 23 deletions(-)

diff --git a/drivers/iommu/iova.c b/drivers/iommu/iova.c
index e23001b..36cdc8e 100644
--- a/drivers/iommu/iova.c
+++ b/drivers/iommu/iova.c
@@ -390,6 +390,11 @@ free_iova(struct iova_domain *iovad, unsigned long pfn)
 }
 EXPORT_SYMBOL_GPL(free_iova);
 
+static void free_this_cached_iovas(void *info)
+{
+   free_cpu_cached_iovas(smp_processor_id(), info);
+}
+
 /**
  * alloc_iova_fast - allocates an iova from rcache
  * @iovad: - iova domain in question
@@ -413,17 +418,12 @@ alloc_iova_fast(struct iova_domain *iovad, unsigned long 
size,
 retry:
new_iova = alloc_iova(iovad, size, limit_pfn, true);
if (!new_iova) {
-   unsigned int cpu;
-
if (flushed_rcache)
return 0;
 
/* Try replenishing IOVAs by flushing rcache. */
flushed_rcache = true;
-   preempt_disable();
-   for_each_online_cpu(cpu)
-   free_cpu_cached_iovas(cpu, iovad);
-   preempt_enable();
+   on_each_cpu(free_this_cached_iovas, iovad, true);
goto retry;
}
 
@@ -647,7 +647,6 @@ struct iova_magazine {
 };
 
 struct iova_cpu_rcache {
-   spinlock_t lock;
struct iova_magazine *loaded;
struct iova_magazine *prev;
 };
@@ -729,7 +728,6 @@ static void init_iova_rcaches(struct iova_domain *iovad)
continue;
for_each_possible_cpu(cpu) {
cpu_rcache = per_cpu_ptr(rcache->cpu_rcaches, cpu);
-   spin_lock_init(_rcache->lock);
cpu_rcache->loaded = iova_magazine_alloc(GFP_KERNEL);
cpu_rcache->prev = iova_magazine_alloc(GFP_KERNEL);
}
@@ -749,10 +747,8 @@ static bool __iova_rcache_insert(struct iova_domain *iovad,
struct iova_magazine *mag_to_free = NULL;
struct iova_cpu_rcache *cpu_rcache;
bool can_insert = false;
-   unsigned long flags;
 
cpu_rcache = get_cpu_ptr(rcache->cpu_rcaches);
-   spin_lock_irqsave(_rcache->lock, flags);
 
if (!iova_magazine_full(cpu_rcache->loaded)) {
can_insert = true;
@@ -780,7 +776,6 @@ static bool __iova_rcache_insert(struct iova_domain *iovad,
if (can_insert)
iova_magazine_push(cpu_rcache->loaded, iova_pfn);
 
-   spin_unlock_irqrestore(_rcache->lock, flags);
put_cpu_ptr(rcache->cpu_rcaches);
 
if (mag_to_free) {
@@ -813,10 +808,8 @@ static unsigned long __iova_rcache_get(struct iova_rcache 
*rcache,
struct iova_cpu_rcache *cpu_rcache;
unsigned long iova_pfn = 0;
bool has_pfn = false;
-   unsigned long flags;
 
cpu_rcache = get_cpu_ptr(rcache->cpu_rcaches);
-   spin_lock_irqsave(_rcache->lock, flags);
 
if (!iova_magazine_empty(cpu_rcache->loaded)) {
has_pfn = true;
@@ -836,7 +829,6 @@ static unsigned long __iova_rcache_get(struct iova_rcache 
*rcache,
if (has_pfn)
iova_pfn = iova_magazine_pop(cpu_rcache->loaded, limit_pfn);
 
-   spin_unlock_irqrestore(_rcache->lock, flags);
put_cpu_ptr(rcache->cpu_rcaches);
 
return iova_pfn;
@@ -866,17 +858,11 @@ static void free_cpu_iova_rcache(unsigned int cpu, struct 
iova_domain *iovad,
 struct iova_rcache *rcache)
 {
struct iova_cpu_rcache *cpu_rcache = per_cpu_ptr(rcache->cpu_rcaches, 
cpu);
-   unsigned long flags;
-
-   spin_lock_irqsave(_rcache->lock, flags);
-
iova_magazine_free_pfns(cpu_rcache->loaded, iovad);
iova_magazine_free(cpu_rcache->loaded);
 
iova_magazine_free_pfns(cpu_rcache->prev, iovad);
iova_magazine_free(cpu_rcache->prev);
-
-   spin_unlock_irqrestore(_rcache->lock, flags);
 }
 
 /*
@@ -910,16 +896,13 @@ void free_cpu_cached_iovas(unsigned int cpu, struct 
iova_domain *iovad)
 {
struct iova_cpu_rcache *cpu_rcache;
struct iova_rcache *rcache;
-   unsigned long flags;
int i;
 
for (i = 0; i < IOVA_RANGE_CACHE_MAX_SIZE; ++i) {
rcache = >rcaches[i];
cpu_rcache = per_cpu_ptr(rcache->cpu_rcaches, cpu);
-   spin_lock_irqsave(_rcache->lock, flags);
iova_magazine_free_pfns(cpu_rcache->loaded, iovad);
 

Re: [PATCH v2 1/4] drm: Adding new flag to restrict bitmask drm properties as 32 bit type and 32 bit value pair

[Sagar Arun Kamble]

Adding Rob and Rusty in the review thread.

Kindly review these patches for interface being proposed to set
color/alpha property of planes modeled after glBlendFunc.

 http://lists.freedesktop.org/archives/intel-gfx/2014-March/042350.html
 http://lists.freedesktop.org/archives/intel-gfx/2014-March/042351.html
 http://lists.freedesktop.org/archives/intel-gfx/2014-March/042352.html
 http://lists.freedesktop.org/archives/intel-gfx/2014-March/042587.html
 http://lists.freedesktop.org/archives/intel-gfx/2014-March/042354.html

thanks,
Sagar


On Tue, 2014-03-25 at 20:02 +0530, sagar.a.kam...@intel.com wrote:
> From: Sagar Kamble 
> 
> With this patch new flag DRM_MODE_PROP_32BIT_PAIR is added that will help 
> make use
> of 64 bit value of bitmask property as two 32 bit values.
> 
> Cc: airl...@linux.ie
> Cc: dri-de...@lists.freedesktop.org
> Cc: linux-kernel@vger.kernel.org
> Signed-off-by: Sagar Kamble 
> ---
>  drivers/gpu/drm/drm_crtc.c  | 22 --
>  include/uapi/drm/drm_mode.h |  3 +++
>  2 files changed, 19 insertions(+), 6 deletions(-)
> 
> diff --git a/drivers/gpu/drm/drm_crtc.c b/drivers/gpu/drm/drm_crtc.c
> index 4e43fc2..d0d03ec 100644
> --- a/drivers/gpu/drm/drm_crtc.c
> +++ b/drivers/gpu/drm/drm_crtc.c
> @@ -2993,10 +2993,13 @@ int drm_property_add_enum(struct drm_property 
> *property, int index,
>  
>   /*
>* Bitmask enum properties have the additional constraint of values
> -  * from 0 to 63
> +  * from 0 to 63. For properties with 32BIT_PAIR Flag set this constraint
> +  * range is 0 to 31.
>*/
> - if ((property->flags & DRM_MODE_PROP_BITMASK) && (value > 63))
> - return -EINVAL;
> + if (property->flags & DRM_MODE_PROP_BITMASK)
> + if (((property->flags & DRM_MODE_PROP_32BIT_PAIR) && (value > 
> 31)) ||
> + (value > 63))
> + return -EINVAL;
>  
>   if (!list_empty(>enum_blob_list)) {
>   list_for_each_entry(prop_enum, >enum_blob_list, head) 
> {
> @@ -3305,9 +3308,16 @@ static bool drm_property_change_is_valid(struct 
> drm_property *property,
>   } else if (property->flags & DRM_MODE_PROP_BITMASK) {
>   int i;
>   uint64_t valid_mask = 0;
> - for (i = 0; i < property->num_values; i++)
> - valid_mask |= (1ULL << property->values[i]);
> - return !(value & ~valid_mask);
> + uint32_t valid_32bit_mask = 0;
> + if (property->flags & DRM_MODE_PROP_32BIT_PAIR) {
> + for (i = 0; i < property->num_values; i++)
> + valid_32bit_mask |= (1UL << 
> property->values[i]);
> + return !((value & 0x) & ~valid_32bit_mask);
> + } else {
> + for (i = 0; i < property->num_values; i++)
> + valid_mask |= (1ULL << property->values[i]);
> + return !(value & ~valid_mask);
> + }
>   } else if (property->flags & DRM_MODE_PROP_BLOB) {
>   /* Only the driver knows */
>   return true;
> diff --git a/include/uapi/drm/drm_mode.h b/include/uapi/drm/drm_mode.h
> index f104c26..5e3a7d9 100644
> --- a/include/uapi/drm/drm_mode.h
> +++ b/include/uapi/drm/drm_mode.h
> @@ -250,6 +250,9 @@ struct drm_mode_get_connector {
>  #define DRM_MODE_PROP_ENUM   (1<<3) /* enumerated type with text strings */
>  #define DRM_MODE_PROP_BLOB   (1<<4)
>  #define DRM_MODE_PROP_BITMASK(1<<5) /* bitmask of enumerated types */
> +#define DRM_MODE_PROP_32BIT_PAIR (1<<6) /* 32 bit bitmask of enumerated types
> +  * and 32 bit of value of the type */
> +
>  
>  struct drm_mode_property_enum {
>   __u64 value;


--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majord...@vger.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/

Re: [PATCH v2 1/4] drm: Adding new flag to restrict bitmask drm properties as 32 bit type and 32 bit value pair

[Sagar Arun Kamble]

Adding Rob and Rusty in the review thread.

Kindly review these patches for interface being proposed to set
color/alpha property of planes modeled after glBlendFunc.

 http://lists.freedesktop.org/archives/intel-gfx/2014-March/042350.html
 http://lists.freedesktop.org/archives/intel-gfx/2014-March/042351.html
 http://lists.freedesktop.org/archives/intel-gfx/2014-March/042352.html
 http://lists.freedesktop.org/archives/intel-gfx/2014-March/042587.html
 http://lists.freedesktop.org/archives/intel-gfx/2014-March/042354.html

thanks,
Sagar


On Tue, 2014-03-25 at 20:02 +0530, sagar.a.kam...@intel.com wrote:
 From: Sagar Kamble sagar.a.kam...@intel.com
 
 With this patch new flag DRM_MODE_PROP_32BIT_PAIR is added that will help 
 make use
 of 64 bit value of bitmask property as two 32 bit values.
 
 Cc: airl...@linux.ie
 Cc: dri-de...@lists.freedesktop.org
 Cc: linux-kernel@vger.kernel.org
 Signed-off-by: Sagar Kamble sagar.a.kam...@intel.com
 ---
  drivers/gpu/drm/drm_crtc.c  | 22 --
  include/uapi/drm/drm_mode.h |  3 +++
  2 files changed, 19 insertions(+), 6 deletions(-)
 
 diff --git a/drivers/gpu/drm/drm_crtc.c b/drivers/gpu/drm/drm_crtc.c
 index 4e43fc2..d0d03ec 100644
 --- a/drivers/gpu/drm/drm_crtc.c
 +++ b/drivers/gpu/drm/drm_crtc.c
 @@ -2993,10 +2993,13 @@ int drm_property_add_enum(struct drm_property 
 *property, int index,
  
   /*
* Bitmask enum properties have the additional constraint of values
 -  * from 0 to 63
 +  * from 0 to 63. For properties with 32BIT_PAIR Flag set this constraint
 +  * range is 0 to 31.
*/
 - if ((property-flags  DRM_MODE_PROP_BITMASK)  (value  63))
 - return -EINVAL;
 + if (property-flags  DRM_MODE_PROP_BITMASK)
 + if (((property-flags  DRM_MODE_PROP_32BIT_PAIR)  (value  
 31)) ||
 + (value  63))
 + return -EINVAL;
  
   if (!list_empty(property-enum_blob_list)) {
   list_for_each_entry(prop_enum, property-enum_blob_list, head) 
 {
 @@ -3305,9 +3308,16 @@ static bool drm_property_change_is_valid(struct 
 drm_property *property,
   } else if (property-flags  DRM_MODE_PROP_BITMASK) {
   int i;
   uint64_t valid_mask = 0;
 - for (i = 0; i  property-num_values; i++)
 - valid_mask |= (1ULL  property-values[i]);
 - return !(value  ~valid_mask);
 + uint32_t valid_32bit_mask = 0;
 + if (property-flags  DRM_MODE_PROP_32BIT_PAIR) {
 + for (i = 0; i  property-num_values; i++)
 + valid_32bit_mask |= (1UL  
 property-values[i]);
 + return !((value  0x)  ~valid_32bit_mask);
 + } else {
 + for (i = 0; i  property-num_values; i++)
 + valid_mask |= (1ULL  property-values[i]);
 + return !(value  ~valid_mask);
 + }
   } else if (property-flags  DRM_MODE_PROP_BLOB) {
   /* Only the driver knows */
   return true;
 diff --git a/include/uapi/drm/drm_mode.h b/include/uapi/drm/drm_mode.h
 index f104c26..5e3a7d9 100644
 --- a/include/uapi/drm/drm_mode.h
 +++ b/include/uapi/drm/drm_mode.h
 @@ -250,6 +250,9 @@ struct drm_mode_get_connector {
  #define DRM_MODE_PROP_ENUM   (13) /* enumerated type with text strings */
  #define DRM_MODE_PROP_BLOB   (14)
  #define DRM_MODE_PROP_BITMASK(15) /* bitmask of enumerated types */
 +#define DRM_MODE_PROP_32BIT_PAIR (16) /* 32 bit bitmask of enumerated types
 +  * and 32 bit of value of the type */
 +
  
  struct drm_mode_property_enum {
   __u64 value;


--
To unsubscribe from this list: send the line unsubscribe linux-kernel in
the body of a message to majord...@vger.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/

Re: [PATCH v4 1/1] drm/i915: Enabling 128x128 and 256x256 ARGB Cursor Support

[Sagar Arun Kamble]

Gentle reminder for reviewing this and i-g-t patch.

On Mon, 2014-03-10 at 17:06 +0530, sagar.a.kam...@intel.com wrote:
> From: Sagar Kamble 
> 
> With this patch we allow larger cursor planes of sizes 128x128
> and 256x256.
> 
> v2: Added more precise check on size while setting cursor plane.
> 
> v3: Changes related to restructuring cursor size restrictions
> and DRM_DEBUG usage.
> 
> v4: Indentation related changes for setting cursor control and
> implementing DRM_CAP_CURSOR_WIDTH and DRM_CAP_CURSOR_HEIGHT
> 
> Testcase: igt/kms_cursor_crc
> Cc: Daniel Vetter 
> Cc: Jani Nikula 
> Cc: David Airlie 
> Cc: dri-de...@lists.freedesktop.org
> Cc: linux-kernel@vger.kernel.org
> Signed-off-by: G, Pallavi 
> Signed-off-by: Sagar Kamble 
> ---
>  drivers/gpu/drm/i915/i915_reg.h  |  4 +++
>  drivers/gpu/drm/i915/intel_display.c | 53 
> 
>  drivers/gpu/drm/i915/intel_drv.h |  7 +
>  3 files changed, 59 insertions(+), 5 deletions(-)
> 
> diff --git a/drivers/gpu/drm/i915/i915_reg.h b/drivers/gpu/drm/i915/i915_reg.h
> index 146609a..aee8258 100644
> --- a/drivers/gpu/drm/i915/i915_reg.h
> +++ b/drivers/gpu/drm/i915/i915_reg.h
> @@ -3551,7 +3551,11 @@ enum punit_power_well {
>  /* New style CUR*CNTR flags */
>  #define   CURSOR_MODE0x27
>  #define   CURSOR_MODE_DISABLE   0x00
> +#define   CURSOR_MODE_128_32B_AX 0x02
> +#define   CURSOR_MODE_256_32B_AX 0x03
>  #define   CURSOR_MODE_64_32B_AX 0x07
> +#define   CURSOR_MODE_128_ARGB_AX ((1 << 5) | CURSOR_MODE_128_32B_AX)
> +#define   CURSOR_MODE_256_ARGB_AX ((1 << 5) | CURSOR_MODE_256_32B_AX)
>  #define   CURSOR_MODE_64_ARGB_AX ((1 << 5) | CURSOR_MODE_64_32B_AX)
>  #define   MCURSOR_PIPE_SELECT(1 << 28)
>  #define   MCURSOR_PIPE_A 0x00
> diff --git a/drivers/gpu/drm/i915/intel_display.c 
> b/drivers/gpu/drm/i915/intel_display.c
> index 0868afb..ec6a073 100644
> --- a/drivers/gpu/drm/i915/intel_display.c
> +++ b/drivers/gpu/drm/i915/intel_display.c
> @@ -7440,10 +7440,26 @@ static void i9xx_update_cursor(struct drm_crtc *crtc, 
> u32 base)
>   bool visible = base != 0;
>  
>   if (intel_crtc->cursor_visible != visible) {
> + int16_t width = intel_crtc->cursor_width;
>   uint32_t cntl = I915_READ(CURCNTR(pipe));
>   if (base) {
>   cntl &= ~(CURSOR_MODE | MCURSOR_PIPE_SELECT);
> - cntl |= CURSOR_MODE_64_ARGB_AX | MCURSOR_GAMMA_ENABLE;
> + cntl |= MCURSOR_GAMMA_ENABLE;
> +
> + switch (width) {
> + case 64:
> + cntl |= CURSOR_MODE_64_ARGB_AX;
> + break;
> + case 128:
> + cntl |= CURSOR_MODE_128_ARGB_AX;
> + break;
> + case 256:
> + cntl |= CURSOR_MODE_256_ARGB_AX;
> + break;
> + default:
> + WARN_ON(1);
> + return;
> + }
>   cntl |= pipe << 28; /* Connect to correct pipe */
>   } else {
>   cntl &= ~(CURSOR_MODE | MCURSOR_GAMMA_ENABLE);
> @@ -7468,10 +7484,25 @@ static void ivb_update_cursor(struct drm_crtc *crtc, 
> u32 base)
>   bool visible = base != 0;
>  
>   if (intel_crtc->cursor_visible != visible) {
> + int16_t width = intel_crtc->cursor_width;
>   uint32_t cntl = I915_READ(CURCNTR_IVB(pipe));
>   if (base) {
>   cntl &= ~CURSOR_MODE;
> - cntl |= CURSOR_MODE_64_ARGB_AX | MCURSOR_GAMMA_ENABLE;
> + cntl |= MCURSOR_GAMMA_ENABLE;
> + switch (width) {
> + case 64:
> + cntl |= CURSOR_MODE_64_ARGB_AX;
> + break;
> + case 128:
> + cntl |= CURSOR_MODE_128_ARGB_AX;
> + break;
> + case 256:
> + cntl |= CURSOR_MODE_256_ARGB_AX;
> + break;
> + default:
> + WARN_ON(1);
> + return;
> + }
>   } else {
>   cntl &= ~(CURSOR_MODE | MCURSOR_GAMMA_ENABLE);
>   cntl |= CURSOR_MODE_DISABLE;
> @@ -7567,9 +7598,11 @@ static int intel_crtc_cursor_set(struct drm_crtc *crtc,
>   goto finish;
>   }
>  
> - /* Currently we only support 64x64 cursors */
> - if (width != 64 || height != 64) {
> - DRM_ERROR("we currently only support 64x64 cursors\n");
> + /* Check for which cursor types we support */
> + if (!((width == 64 && height == 64) ||
> + (width == 128 && height == 128 

Re: [PATCH v4 1/1] drm/i915: Enabling 128x128 and 256x256 ARGB Cursor Support

[Sagar Arun Kamble]

Gentle reminder for reviewing this and i-g-t patch.

On Mon, 2014-03-10 at 17:06 +0530, sagar.a.kam...@intel.com wrote:
 From: Sagar Kamble sagar.a.kam...@intel.com
 
 With this patch we allow larger cursor planes of sizes 128x128
 and 256x256.
 
 v2: Added more precise check on size while setting cursor plane.
 
 v3: Changes related to restructuring cursor size restrictions
 and DRM_DEBUG usage.
 
 v4: Indentation related changes for setting cursor control and
 implementing DRM_CAP_CURSOR_WIDTH and DRM_CAP_CURSOR_HEIGHT
 
 Testcase: igt/kms_cursor_crc
 Cc: Daniel Vetter daniel.vet...@ffwll.ch
 Cc: Jani Nikula jani.nik...@linux.intel.com
 Cc: David Airlie airl...@linux.ie
 Cc: dri-de...@lists.freedesktop.org
 Cc: linux-kernel@vger.kernel.org
 Signed-off-by: G, Pallavi pallav...@intel.com
 Signed-off-by: Sagar Kamble sagar.a.kam...@intel.com
 ---
  drivers/gpu/drm/i915/i915_reg.h  |  4 +++
  drivers/gpu/drm/i915/intel_display.c | 53 
 
  drivers/gpu/drm/i915/intel_drv.h |  7 +
  3 files changed, 59 insertions(+), 5 deletions(-)
 
 diff --git a/drivers/gpu/drm/i915/i915_reg.h b/drivers/gpu/drm/i915/i915_reg.h
 index 146609a..aee8258 100644
 --- a/drivers/gpu/drm/i915/i915_reg.h
 +++ b/drivers/gpu/drm/i915/i915_reg.h
 @@ -3551,7 +3551,11 @@ enum punit_power_well {
  /* New style CUR*CNTR flags */
  #define   CURSOR_MODE0x27
  #define   CURSOR_MODE_DISABLE   0x00
 +#define   CURSOR_MODE_128_32B_AX 0x02
 +#define   CURSOR_MODE_256_32B_AX 0x03
  #define   CURSOR_MODE_64_32B_AX 0x07
 +#define   CURSOR_MODE_128_ARGB_AX ((1  5) | CURSOR_MODE_128_32B_AX)
 +#define   CURSOR_MODE_256_ARGB_AX ((1  5) | CURSOR_MODE_256_32B_AX)
  #define   CURSOR_MODE_64_ARGB_AX ((1  5) | CURSOR_MODE_64_32B_AX)
  #define   MCURSOR_PIPE_SELECT(1  28)
  #define   MCURSOR_PIPE_A 0x00
 diff --git a/drivers/gpu/drm/i915/intel_display.c 
 b/drivers/gpu/drm/i915/intel_display.c
 index 0868afb..ec6a073 100644
 --- a/drivers/gpu/drm/i915/intel_display.c
 +++ b/drivers/gpu/drm/i915/intel_display.c
 @@ -7440,10 +7440,26 @@ static void i9xx_update_cursor(struct drm_crtc *crtc, 
 u32 base)
   bool visible = base != 0;
  
   if (intel_crtc-cursor_visible != visible) {
 + int16_t width = intel_crtc-cursor_width;
   uint32_t cntl = I915_READ(CURCNTR(pipe));
   if (base) {
   cntl = ~(CURSOR_MODE | MCURSOR_PIPE_SELECT);
 - cntl |= CURSOR_MODE_64_ARGB_AX | MCURSOR_GAMMA_ENABLE;
 + cntl |= MCURSOR_GAMMA_ENABLE;
 +
 + switch (width) {
 + case 64:
 + cntl |= CURSOR_MODE_64_ARGB_AX;
 + break;
 + case 128:
 + cntl |= CURSOR_MODE_128_ARGB_AX;
 + break;
 + case 256:
 + cntl |= CURSOR_MODE_256_ARGB_AX;
 + break;
 + default:
 + WARN_ON(1);
 + return;
 + }
   cntl |= pipe  28; /* Connect to correct pipe */
   } else {
   cntl = ~(CURSOR_MODE | MCURSOR_GAMMA_ENABLE);
 @@ -7468,10 +7484,25 @@ static void ivb_update_cursor(struct drm_crtc *crtc, 
 u32 base)
   bool visible = base != 0;
  
   if (intel_crtc-cursor_visible != visible) {
 + int16_t width = intel_crtc-cursor_width;
   uint32_t cntl = I915_READ(CURCNTR_IVB(pipe));
   if (base) {
   cntl = ~CURSOR_MODE;
 - cntl |= CURSOR_MODE_64_ARGB_AX | MCURSOR_GAMMA_ENABLE;
 + cntl |= MCURSOR_GAMMA_ENABLE;
 + switch (width) {
 + case 64:
 + cntl |= CURSOR_MODE_64_ARGB_AX;
 + break;
 + case 128:
 + cntl |= CURSOR_MODE_128_ARGB_AX;
 + break;
 + case 256:
 + cntl |= CURSOR_MODE_256_ARGB_AX;
 + break;
 + default:
 + WARN_ON(1);
 + return;
 + }
   } else {
   cntl = ~(CURSOR_MODE | MCURSOR_GAMMA_ENABLE);
   cntl |= CURSOR_MODE_DISABLE;
 @@ -7567,9 +7598,11 @@ static int intel_crtc_cursor_set(struct drm_crtc *crtc,
   goto finish;
   }
  
 - /* Currently we only support 64x64 cursors */
 - if (width != 64 || height != 64) {
 - DRM_ERROR(we currently only support 64x64 cursors\n);
 + /* Check for which cursor types we support */
 + if (!((width == 64  height == 64) ||
 + (width == 128  height == 128  

Re: [PATCH v3 1/1] drm/i915: Enabling 128x128 and 256x256 ARGB Cursor Support

[Sagar Arun Kamble]

On Sun, 2014-03-09 at 00:34 +0530, Sagar Arun Kamble wrote:
> On Sat, 2014-03-08 at 13:51 -0500, Alex Deucher wrote:
> > On Sat, Mar 8, 2014 at 1:49 PM,   wrote:
> > > From: Sagar Kamble 
> > >
> > > With this patch we allow larger cursor planes of sizes 128x128
> > > and 256x256.
> > >
> > > v2: Added more precise check on size while setting cursor plane.
> > >
> > > v3: Changes related to restructuring cursor size restrictions
> > > and DRM_DEBUG usage.
> > >
> > 
> > I'm not sure how useful it is since you appear to be able to
> > selectively adjust the cursor size, but I recently added support for
> > exposing the cursor size as a drm cap:
> > http://git.kernel.org/cgit/linux/kernel/git/torvalds/linux.git/commit/?id=8716ed4e7bed4e4c7e3f37940e950ddc0362f450
> > 
> > Alex
> Thanks Alex. This is useful for cursor test to enumerate sub-tests based
> on these caps.
Realized after sending mail that we just get to know current cursor
width and height. Can we have capability that exposes all supported
cursor sizes?
> > 
> > > Testcase: igt/kms_cursor_crc
> > > Cc: Daniel Vetter 
> > > Cc: Jani Nikula 
> > > Cc: David Airlie 
> > > Cc: dri-de...@lists.freedesktop.org
> > > Cc: linux-kernel@vger.kernel.org
> > > Signed-off-by: G, Pallavi 
> > > Signed-off-by: Sagar Kamble 
> > > ---
> > >  drivers/gpu/drm/i915/i915_reg.h  |  4 
> > >  drivers/gpu/drm/i915/intel_display.c | 36 
> > > +++-
> > >  2 files changed, 35 insertions(+), 5 deletions(-)
> > >
> > > diff --git a/drivers/gpu/drm/i915/i915_reg.h 
> > > b/drivers/gpu/drm/i915/i915_reg.h
> > > index 146609a..aee8258 100644
> > > --- a/drivers/gpu/drm/i915/i915_reg.h
> > > +++ b/drivers/gpu/drm/i915/i915_reg.h
> > > @@ -3551,7 +3551,11 @@ enum punit_power_well {
> > >  /* New style CUR*CNTR flags */
> > >  #define   CURSOR_MODE  0x27
> > >  #define   CURSOR_MODE_DISABLE   0x00
> > > +#define   CURSOR_MODE_128_32B_AX 0x02
> > > +#define   CURSOR_MODE_256_32B_AX 0x03
> > >  #define   CURSOR_MODE_64_32B_AX 0x07
> > > +#define   CURSOR_MODE_128_ARGB_AX ((1 << 5) | CURSOR_MODE_128_32B_AX)
> > > +#define   CURSOR_MODE_256_ARGB_AX ((1 << 5) | CURSOR_MODE_256_32B_AX)
> > >  #define   CURSOR_MODE_64_ARGB_AX ((1 << 5) | CURSOR_MODE_64_32B_AX)
> > >  #define   MCURSOR_PIPE_SELECT  (1 << 28)
> > >  #define   MCURSOR_PIPE_A   0x00
> > > diff --git a/drivers/gpu/drm/i915/intel_display.c 
> > > b/drivers/gpu/drm/i915/intel_display.c
> > > index 0868afb..e59e8fd 100644
> > > --- a/drivers/gpu/drm/i915/intel_display.c
> > > +++ b/drivers/gpu/drm/i915/intel_display.c
> > > @@ -7440,10 +7440,22 @@ static void i9xx_update_cursor(struct drm_crtc 
> > > *crtc, u32 base)
> > > bool visible = base != 0;
> > >
> > > if (intel_crtc->cursor_visible != visible) {
> > > +   int16_t width = intel_crtc->cursor_width;
> > > uint32_t cntl = I915_READ(CURCNTR(pipe));
> > > if (base) {
> > > cntl &= ~(CURSOR_MODE | MCURSOR_PIPE_SELECT);
> > > -   cntl |= CURSOR_MODE_64_ARGB_AX | 
> > > MCURSOR_GAMMA_ENABLE;
> > > +
> > > +   switch (width) {
> > > +   case 64:
> > > +   cntl |= CURSOR_MODE_64_ARGB_AX | 
> > > MCURSOR_GAMMA_ENABLE;
> > > +   break;
> > > +   case 128:
> > > +   cntl |= CURSOR_MODE_128_ARGB_AX | 
> > > MCURSOR_GAMMA_ENABLE;
> > > +   break;
> > > +   case 256:
> > > +   cntl |= CURSOR_MODE_256_ARGB_AX | 
> > > MCURSOR_GAMMA_ENABLE;
> > > +   break;
> > > +   }
> > > cntl |= pipe << 28; /* Connect to correct pipe */
> > > } else {
> > > cntl &= ~(CURSOR_MODE | MCURSOR_GAMMA_ENABLE);
> > > @@ -7468,10 +7480,22 @@ static void ivb_update_cursor(struct drm_crtc 
> > > *crtc, u32 base)
> > > bool visib

Re: [PATCH v3 1/1] drm/i915: Enabling 128x128 and 256x256 ARGB Cursor Support

[Sagar Arun Kamble]

On Sat, 2014-03-08 at 13:51 -0500, Alex Deucher wrote:
> On Sat, Mar 8, 2014 at 1:49 PM,   wrote:
> > From: Sagar Kamble 
> >
> > With this patch we allow larger cursor planes of sizes 128x128
> > and 256x256.
> >
> > v2: Added more precise check on size while setting cursor plane.
> >
> > v3: Changes related to restructuring cursor size restrictions
> > and DRM_DEBUG usage.
> >
> 
> I'm not sure how useful it is since you appear to be able to
> selectively adjust the cursor size, but I recently added support for
> exposing the cursor size as a drm cap:
> http://git.kernel.org/cgit/linux/kernel/git/torvalds/linux.git/commit/?id=8716ed4e7bed4e4c7e3f37940e950ddc0362f450
> 
> Alex
Thanks Alex. This is useful for cursor test to enumerate sub-tests based
on these caps.
> 
> > Testcase: igt/kms_cursor_crc
> > Cc: Daniel Vetter 
> > Cc: Jani Nikula 
> > Cc: David Airlie 
> > Cc: dri-de...@lists.freedesktop.org
> > Cc: linux-kernel@vger.kernel.org
> > Signed-off-by: G, Pallavi 
> > Signed-off-by: Sagar Kamble 
> > ---
> >  drivers/gpu/drm/i915/i915_reg.h  |  4 
> >  drivers/gpu/drm/i915/intel_display.c | 36 
> > +++-
> >  2 files changed, 35 insertions(+), 5 deletions(-)
> >
> > diff --git a/drivers/gpu/drm/i915/i915_reg.h 
> > b/drivers/gpu/drm/i915/i915_reg.h
> > index 146609a..aee8258 100644
> > --- a/drivers/gpu/drm/i915/i915_reg.h
> > +++ b/drivers/gpu/drm/i915/i915_reg.h
> > @@ -3551,7 +3551,11 @@ enum punit_power_well {
> >  /* New style CUR*CNTR flags */
> >  #define   CURSOR_MODE  0x27
> >  #define   CURSOR_MODE_DISABLE   0x00
> > +#define   CURSOR_MODE_128_32B_AX 0x02
> > +#define   CURSOR_MODE_256_32B_AX 0x03
> >  #define   CURSOR_MODE_64_32B_AX 0x07
> > +#define   CURSOR_MODE_128_ARGB_AX ((1 << 5) | CURSOR_MODE_128_32B_AX)
> > +#define   CURSOR_MODE_256_ARGB_AX ((1 << 5) | CURSOR_MODE_256_32B_AX)
> >  #define   CURSOR_MODE_64_ARGB_AX ((1 << 5) | CURSOR_MODE_64_32B_AX)
> >  #define   MCURSOR_PIPE_SELECT  (1 << 28)
> >  #define   MCURSOR_PIPE_A   0x00
> > diff --git a/drivers/gpu/drm/i915/intel_display.c 
> > b/drivers/gpu/drm/i915/intel_display.c
> > index 0868afb..e59e8fd 100644
> > --- a/drivers/gpu/drm/i915/intel_display.c
> > +++ b/drivers/gpu/drm/i915/intel_display.c
> > @@ -7440,10 +7440,22 @@ static void i9xx_update_cursor(struct drm_crtc 
> > *crtc, u32 base)
> > bool visible = base != 0;
> >
> > if (intel_crtc->cursor_visible != visible) {
> > +   int16_t width = intel_crtc->cursor_width;
> > uint32_t cntl = I915_READ(CURCNTR(pipe));
> > if (base) {
> > cntl &= ~(CURSOR_MODE | MCURSOR_PIPE_SELECT);
> > -   cntl |= CURSOR_MODE_64_ARGB_AX | 
> > MCURSOR_GAMMA_ENABLE;
> > +
> > +   switch (width) {
> > +   case 64:
> > +   cntl |= CURSOR_MODE_64_ARGB_AX | 
> > MCURSOR_GAMMA_ENABLE;
> > +   break;
> > +   case 128:
> > +   cntl |= CURSOR_MODE_128_ARGB_AX | 
> > MCURSOR_GAMMA_ENABLE;
> > +   break;
> > +   case 256:
> > +   cntl |= CURSOR_MODE_256_ARGB_AX | 
> > MCURSOR_GAMMA_ENABLE;
> > +   break;
> > +   }
> > cntl |= pipe << 28; /* Connect to correct pipe */
> > } else {
> > cntl &= ~(CURSOR_MODE | MCURSOR_GAMMA_ENABLE);
> > @@ -7468,10 +7480,22 @@ static void ivb_update_cursor(struct drm_crtc 
> > *crtc, u32 base)
> > bool visible = base != 0;
> >
> > if (intel_crtc->cursor_visible != visible) {
> > +   int16_t width = intel_crtc->cursor_width;
> > uint32_t cntl = I915_READ(CURCNTR_IVB(pipe));
> > if (base) {
> > cntl &= ~CURSOR_MODE;
> > -   cntl |= CURSOR_MODE_64_ARGB_AX | 
> > MCURSOR_GAMMA_ENABLE;
> > +
> > +   switch (width) {
> > +   case 64:
> > +   cntl |= CURSOR_MODE_64_ARGB_AX | 
> > MCURSOR_GAMMA_ENABLE;
> > +   break;
> > +   case 128:
> > +   cntl |= CURSOR_MODE_128_ARGB_AX | 
> > MCURSOR_GAMMA_ENABLE;
> > +   break;
> > +   case 256:
> > +   cntl |= CURSOR_MODE_256_ARGB_AX | 
> > MCURSOR_GAMMA_ENABLE;
> > +   break;
> > +   }
> > } else {
> > cntl &= ~(CURSOR_MODE | MCURSOR_GAMMA_ENABLE);
> > cntl |= CURSOR_MODE_DISABLE;
> > @@ 

Re: [PATCH v3 1/1] drm/i915: Enabling 128x128 and 256x256 ARGB Cursor Support

[Sagar Arun Kamble]

On Sat, 2014-03-08 at 13:51 -0500, Alex Deucher wrote:
 On Sat, Mar 8, 2014 at 1:49 PM,  sagar.a.kam...@intel.com wrote:
  From: Sagar Kamble sagar.a.kam...@intel.com
 
  With this patch we allow larger cursor planes of sizes 128x128
  and 256x256.
 
  v2: Added more precise check on size while setting cursor plane.
 
  v3: Changes related to restructuring cursor size restrictions
  and DRM_DEBUG usage.
 
 
 I'm not sure how useful it is since you appear to be able to
 selectively adjust the cursor size, but I recently added support for
 exposing the cursor size as a drm cap:
 http://git.kernel.org/cgit/linux/kernel/git/torvalds/linux.git/commit/?id=8716ed4e7bed4e4c7e3f37940e950ddc0362f450
 
 Alex
Thanks Alex. This is useful for cursor test to enumerate sub-tests based
on these caps.
 
  Testcase: igt/kms_cursor_crc
  Cc: Daniel Vetter daniel.vet...@ffwll.ch
  Cc: Jani Nikula jani.nik...@linux.intel.com
  Cc: David Airlie airl...@linux.ie
  Cc: dri-de...@lists.freedesktop.org
  Cc: linux-kernel@vger.kernel.org
  Signed-off-by: G, Pallavi pallav...@intel.com
  Signed-off-by: Sagar Kamble sagar.a.kam...@intel.com
  ---
   drivers/gpu/drm/i915/i915_reg.h  |  4 
   drivers/gpu/drm/i915/intel_display.c | 36 
  +++-
   2 files changed, 35 insertions(+), 5 deletions(-)
 
  diff --git a/drivers/gpu/drm/i915/i915_reg.h 
  b/drivers/gpu/drm/i915/i915_reg.h
  index 146609a..aee8258 100644
  --- a/drivers/gpu/drm/i915/i915_reg.h
  +++ b/drivers/gpu/drm/i915/i915_reg.h
  @@ -3551,7 +3551,11 @@ enum punit_power_well {
   /* New style CUR*CNTR flags */
   #define   CURSOR_MODE  0x27
   #define   CURSOR_MODE_DISABLE   0x00
  +#define   CURSOR_MODE_128_32B_AX 0x02
  +#define   CURSOR_MODE_256_32B_AX 0x03
   #define   CURSOR_MODE_64_32B_AX 0x07
  +#define   CURSOR_MODE_128_ARGB_AX ((1  5) | CURSOR_MODE_128_32B_AX)
  +#define   CURSOR_MODE_256_ARGB_AX ((1  5) | CURSOR_MODE_256_32B_AX)
   #define   CURSOR_MODE_64_ARGB_AX ((1  5) | CURSOR_MODE_64_32B_AX)
   #define   MCURSOR_PIPE_SELECT  (1  28)
   #define   MCURSOR_PIPE_A   0x00
  diff --git a/drivers/gpu/drm/i915/intel_display.c 
  b/drivers/gpu/drm/i915/intel_display.c
  index 0868afb..e59e8fd 100644
  --- a/drivers/gpu/drm/i915/intel_display.c
  +++ b/drivers/gpu/drm/i915/intel_display.c
  @@ -7440,10 +7440,22 @@ static void i9xx_update_cursor(struct drm_crtc 
  *crtc, u32 base)
  bool visible = base != 0;
 
  if (intel_crtc-cursor_visible != visible) {
  +   int16_t width = intel_crtc-cursor_width;
  uint32_t cntl = I915_READ(CURCNTR(pipe));
  if (base) {
  cntl = ~(CURSOR_MODE | MCURSOR_PIPE_SELECT);
  -   cntl |= CURSOR_MODE_64_ARGB_AX | 
  MCURSOR_GAMMA_ENABLE;
  +
  +   switch (width) {
  +   case 64:
  +   cntl |= CURSOR_MODE_64_ARGB_AX | 
  MCURSOR_GAMMA_ENABLE;
  +   break;
  +   case 128:
  +   cntl |= CURSOR_MODE_128_ARGB_AX | 
  MCURSOR_GAMMA_ENABLE;
  +   break;
  +   case 256:
  +   cntl |= CURSOR_MODE_256_ARGB_AX | 
  MCURSOR_GAMMA_ENABLE;
  +   break;
  +   }
  cntl |= pipe  28; /* Connect to correct pipe */
  } else {
  cntl = ~(CURSOR_MODE | MCURSOR_GAMMA_ENABLE);
  @@ -7468,10 +7480,22 @@ static void ivb_update_cursor(struct drm_crtc 
  *crtc, u32 base)
  bool visible = base != 0;
 
  if (intel_crtc-cursor_visible != visible) {
  +   int16_t width = intel_crtc-cursor_width;
  uint32_t cntl = I915_READ(CURCNTR_IVB(pipe));
  if (base) {
  cntl = ~CURSOR_MODE;
  -   cntl |= CURSOR_MODE_64_ARGB_AX | 
  MCURSOR_GAMMA_ENABLE;
  +
  +   switch (width) {
  +   case 64:
  +   cntl |= CURSOR_MODE_64_ARGB_AX | 
  MCURSOR_GAMMA_ENABLE;
  +   break;
  +   case 128:
  +   cntl |= CURSOR_MODE_128_ARGB_AX | 
  MCURSOR_GAMMA_ENABLE;
  +   break;
  +   case 256:
  +   cntl |= CURSOR_MODE_256_ARGB_AX | 
  MCURSOR_GAMMA_ENABLE;
  +   break;
  +   }
  } else {
  cntl = ~(CURSOR_MODE | MCURSOR_GAMMA_ENABLE);
  cntl |= CURSOR_MODE_DISABLE;
  @@ -7567,9 +7591,11 @@ static int intel_crtc_cursor_set(struct drm_crtc 
  *crtc,

Re: [PATCH v3 1/1] drm/i915: Enabling 128x128 and 256x256 ARGB Cursor Support

[Sagar Arun Kamble]

On Sun, 2014-03-09 at 00:34 +0530, Sagar Arun Kamble wrote:
 On Sat, 2014-03-08 at 13:51 -0500, Alex Deucher wrote:
  On Sat, Mar 8, 2014 at 1:49 PM,  sagar.a.kam...@intel.com wrote:
   From: Sagar Kamble sagar.a.kam...@intel.com
  
   With this patch we allow larger cursor planes of sizes 128x128
   and 256x256.
  
   v2: Added more precise check on size while setting cursor plane.
  
   v3: Changes related to restructuring cursor size restrictions
   and DRM_DEBUG usage.
  
  
  I'm not sure how useful it is since you appear to be able to
  selectively adjust the cursor size, but I recently added support for
  exposing the cursor size as a drm cap:
  http://git.kernel.org/cgit/linux/kernel/git/torvalds/linux.git/commit/?id=8716ed4e7bed4e4c7e3f37940e950ddc0362f450
  
  Alex
 Thanks Alex. This is useful for cursor test to enumerate sub-tests based
 on these caps.
Realized after sending mail that we just get to know current cursor
width and height. Can we have capability that exposes all supported
cursor sizes?
  
   Testcase: igt/kms_cursor_crc
   Cc: Daniel Vetter daniel.vet...@ffwll.ch
   Cc: Jani Nikula jani.nik...@linux.intel.com
   Cc: David Airlie airl...@linux.ie
   Cc: dri-de...@lists.freedesktop.org
   Cc: linux-kernel@vger.kernel.org
   Signed-off-by: G, Pallavi pallav...@intel.com
   Signed-off-by: Sagar Kamble sagar.a.kam...@intel.com
   ---
drivers/gpu/drm/i915/i915_reg.h  |  4 
drivers/gpu/drm/i915/intel_display.c | 36 
   +++-
2 files changed, 35 insertions(+), 5 deletions(-)
  
   diff --git a/drivers/gpu/drm/i915/i915_reg.h 
   b/drivers/gpu/drm/i915/i915_reg.h
   index 146609a..aee8258 100644
   --- a/drivers/gpu/drm/i915/i915_reg.h
   +++ b/drivers/gpu/drm/i915/i915_reg.h
   @@ -3551,7 +3551,11 @@ enum punit_power_well {
/* New style CUR*CNTR flags */
#define   CURSOR_MODE  0x27
#define   CURSOR_MODE_DISABLE   0x00
   +#define   CURSOR_MODE_128_32B_AX 0x02
   +#define   CURSOR_MODE_256_32B_AX 0x03
#define   CURSOR_MODE_64_32B_AX 0x07
   +#define   CURSOR_MODE_128_ARGB_AX ((1  5) | CURSOR_MODE_128_32B_AX)
   +#define   CURSOR_MODE_256_ARGB_AX ((1  5) | CURSOR_MODE_256_32B_AX)
#define   CURSOR_MODE_64_ARGB_AX ((1  5) | CURSOR_MODE_64_32B_AX)
#define   MCURSOR_PIPE_SELECT  (1  28)
#define   MCURSOR_PIPE_A   0x00
   diff --git a/drivers/gpu/drm/i915/intel_display.c 
   b/drivers/gpu/drm/i915/intel_display.c
   index 0868afb..e59e8fd 100644
   --- a/drivers/gpu/drm/i915/intel_display.c
   +++ b/drivers/gpu/drm/i915/intel_display.c
   @@ -7440,10 +7440,22 @@ static void i9xx_update_cursor(struct drm_crtc 
   *crtc, u32 base)
   bool visible = base != 0;
  
   if (intel_crtc-cursor_visible != visible) {
   +   int16_t width = intel_crtc-cursor_width;
   uint32_t cntl = I915_READ(CURCNTR(pipe));
   if (base) {
   cntl = ~(CURSOR_MODE | MCURSOR_PIPE_SELECT);
   -   cntl |= CURSOR_MODE_64_ARGB_AX | 
   MCURSOR_GAMMA_ENABLE;
   +
   +   switch (width) {
   +   case 64:
   +   cntl |= CURSOR_MODE_64_ARGB_AX | 
   MCURSOR_GAMMA_ENABLE;
   +   break;
   +   case 128:
   +   cntl |= CURSOR_MODE_128_ARGB_AX | 
   MCURSOR_GAMMA_ENABLE;
   +   break;
   +   case 256:
   +   cntl |= CURSOR_MODE_256_ARGB_AX | 
   MCURSOR_GAMMA_ENABLE;
   +   break;
   +   }
   cntl |= pipe  28; /* Connect to correct pipe */
   } else {
   cntl = ~(CURSOR_MODE | MCURSOR_GAMMA_ENABLE);
   @@ -7468,10 +7480,22 @@ static void ivb_update_cursor(struct drm_crtc 
   *crtc, u32 base)
   bool visible = base != 0;
  
   if (intel_crtc-cursor_visible != visible) {
   +   int16_t width = intel_crtc-cursor_width;
   uint32_t cntl = I915_READ(CURCNTR_IVB(pipe));
   if (base) {
   cntl = ~CURSOR_MODE;
   -   cntl |= CURSOR_MODE_64_ARGB_AX | 
   MCURSOR_GAMMA_ENABLE;
   +
   +   switch (width) {
   +   case 64:
   +   cntl |= CURSOR_MODE_64_ARGB_AX | 
   MCURSOR_GAMMA_ENABLE;
   +   break;
   +   case 128:
   +   cntl |= CURSOR_MODE_128_ARGB_AX | 
   MCURSOR_GAMMA_ENABLE;
   +   break;
   +   case 256:
   +   cntl |= CURSOR_MODE_256_ARGB_AX | 
   MCURSOR_GAMMA_ENABLE

Re: [Intel-gfx] [PATCH v5 11/11] drm/i915: Calling rotate and inverse rotate transformations after clipping

[Sagar Arun Kamble]

On Tue, 2014-02-11 at 16:56 +0200, Ville Syrjälä wrote:
> On Tue, Feb 11, 2014 at 05:02:31PM +0530, Sagar Arun Kamble wrote:
> > On Mon, 2014-02-10 at 15:32 +0200, Ville Syrjälä wrote:
> > > On Mon, Feb 10, 2014 at 01:01:18PM +0530, sagar.a.kam...@intel.com wrote:
> > > > From: Sagar Kamble 
> > > > 
> > > > With clipped sprites these transformations are not working. these
> > > > functions transform complete sprite irrespective of clipping present.
> > > > This leads to invisible portion of sprite show up when rotate 180 if
> > > > it was out of visible area before.
> > > > 
> > > > v4: Moved rotate transform for source rectangle after clipping.
> > > > Added rotate and inverse rotate transform for destination rect.
> > > 
> > > Still NAK.
> > > 
> > > I just pushed rotation support to my glplane test app [1], and with
> > > with that my rotated clipping code works exactly as intended.
> > > 
> > > [1] git://gitorious.org/vsyrjala/glplane.git
> > I tried this app. I think I am considering 180 degree rotation of
> > clipped sprite plane differently. I have captured output with these
> > rotate transforms moved before(clip-rotated) and after(rotate-clipped)
> > clipping code.
> > 
> > Which is valid? Rotating entire sprite and then clipping or Rotating
> > clipped portion?
> > 
> > Reference and Rotated output is attached FYI.
> > 
> > If rotating entire sprite is correct then this patch 11/11 is not needed
> > and can be abandoned.
> 
> The way I think of these things is roughly this:
> 
> You have the user specified source rectangle, where the coordinates specify
> the viewport into the framebuffer. This coordinate space is oriented the
> same was as the framebuffer itself, ie. the first pixel of the
> framebuffer is at coordinates 0,0. So plane rotation doesn't affect this
> at all.
> 
> Then you have the user specified destination/crtc rectangle, where the
> coordinates specify the position of the plane within the crtc coordinate
> space. So the first visible pixel the pipe will push out is at
> coordinates 0,0. So again plane rotation doesn't affect this.
> 
> Then you have the rotation which simply specifies the transformation to
> be applied to the pixels when they "move" from the source rectangle to
> the destination rectangle. So w/ 0 degree rotation the pixel at
> src_x,src_y in the framebuffer will appear at position crtc_x,crtc_y
> on the crtc output. With 180 degree rotation the pixel at src_x,src_y
> will appear at crtc_x+crtc_w-1,crtc_y+crtc_h-1.
> 
> As clipping happens in the crtc coordinate space, we need to orient
> the source coordindates the same way to get the correct clipping result.
> So for example with 0 degrees rotation clipping the left side of the
> destination rectangle must result in clipping the left side of the source
> rectangle as well. And with 180 degree rotation clipping the destination
> rectangle on the left side must result in clipping the source rectangle
> on the right side. Left and right in each case referring to the original
> unrotate coordinates.
> 
> So let's say we have the following situation w/ 180 degree rotation.
> The letters inside the rects represented specific named pixels,
> the FB rectangle represents the FB as specified by addfb2 ioctl,
> the CRTC rectangle represents the pipe output (0,0 -> PIPESRC.w,h):
> 
> FB: CRTC:
> 0,0 ___ 0.0 __
> |  abcd   | | |
> |  efgh   | | |
> |_| |hgfe |
> |dcba_|
> unclipped coordinates specified by user:
> src_x=2,src_y=0 crtc_x=0,crtc_y=2
> src_w=4.src_h=2 crtc_w=4,crtc_h=2
> 
> clipped coordinates:
> src_x=2,src_y=0 crtc_x=0,crtc_y=2
> src_w=4.src_h=2 crtc_w=4,crtc_h=2
> 
> 
> Then the user moves the sprite one pixel to the left resulting on some
> clipping (the X pixels). Note that the unclipped source coordinates do
> not change here at all, in fact crtc_x is the only thing changed by the
> user:
> 
> FB: CRTC:
> 0,0 ___ 0.0 __
> |  abcX   | | |
> |  efgX   | | |
> |_| Xgfe  |
> Xcba__|
> unclipped coordinates specified by user:
> src_x=2,src_y=0 crtc_x=-1,crtc_y=2
> src_w=4.src_h=2 crtc_w= 4,crtc_h=2
> 
> clipped coordinates:
> src_x=2,src_y=0 crtc_x=0,crtc_y=2
> src_w=3.src_h=2 crtc_w=3,crtc_h=2
> 


Understood this now. Thank you Ville for providing this elaborate
graphical view of the rotation cases.

-Sagar

--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majord...@vger.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/

Re: [Intel-gfx] [PATCH v5 11/11] drm/i915: Calling rotate and inverse rotate transformations after clipping

[Sagar Arun Kamble]

On Tue, 2014-02-11 at 12:59 +0100, Daniel Vetter wrote:
> On Tue, Feb 11, 2014 at 05:02:31PM +0530, Sagar Arun Kamble wrote:
> > On Mon, 2014-02-10 at 15:32 +0200, Ville Syrjälä wrote:
> > > On Mon, Feb 10, 2014 at 01:01:18PM +0530, sagar.a.kam...@intel.com wrote:
> > > > From: Sagar Kamble 
> > > > 
> > > > With clipped sprites these transformations are not working. these
> > > > functions transform complete sprite irrespective of clipping present.
> > > > This leads to invisible portion of sprite show up when rotate 180 if
> > > > it was out of visible area before.
> > > > 
> > > > v4: Moved rotate transform for source rectangle after clipping.
> > > > Added rotate and inverse rotate transform for destination rect.
> > > 
> > > Still NAK.
> > > 
> > > I just pushed rotation support to my glplane test app [1], and with
> > > with that my rotated clipping code works exactly as intended.
> > > 
> > > [1] git://gitorious.org/vsyrjala/glplane.git
> > I tried this app. I think I am considering 180 degree rotation of
> > clipped sprite plane differently. I have captured output with these
> > rotate transforms moved before(clip-rotated) and after(rotate-clipped)
> > clipping code.
> > 
> > Which is valid? Rotating entire sprite and then clipping or Rotating
> > clipped portion?
> > 
> > Reference and Rotated output is attached FYI.
> > 
> > If rotating entire sprite is correct then this patch 11/11 is not needed
> > and can be abandoned.
> 
> I think doing the clipping first (as a viewport) and then rotating the
> viewport makes more sense to me. Dunno what other people think, but I
> guess we should document that somewhere.
> -Daniel

Agree. Thought of use-case where if media player(that will use sprite
for video) is moved such that it is off the screen partially, after
rotation as well we should see same portion of clip inside(rotated
though).

> 
> > 
> > thanks,
> > Sagar
> > > > 
> > > > Cc: Daniel Vetter 
> > > > Cc: Jani Nikula 
> > > > Cc: David Airlie 
> > > > Cc: dri-de...@lists.freedesktop.org
> > > > Cc: linux-kernel@vger.kernel.org
> > > > Cc: vijay.a.purushotha...@intel.com
> > > > Signed-off-by: Sagar Kamble 
> > > > ---
> > > >  drivers/gpu/drm/i915/intel_sprite.c | 16 
> > > >  1 file changed, 12 insertions(+), 4 deletions(-)
> > > > 
> > > > diff --git a/drivers/gpu/drm/i915/intel_sprite.c 
> > > > b/drivers/gpu/drm/i915/intel_sprite.c
> > > > index 62b9f84..799f6a9 100644
> > > > --- a/drivers/gpu/drm/i915/intel_sprite.c
> > > > +++ b/drivers/gpu/drm/i915/intel_sprite.c
> > > > @@ -769,9 +769,6 @@ intel_update_plane(struct drm_plane *plane, struct 
> > > > drm_crtc *crtc,
> > > > max_scale = intel_plane->max_downscale << 16;
> > > > min_scale = intel_plane->can_scale ? 1 : (1 << 16);
> > > >  
> > > > -   drm_rect_rotate(, fb->width << 16, fb->height << 16,
> > > > -   intel_plane->rotation);
> > > > -
> > > > hscale = drm_rect_calc_hscale_relaxed(, , min_scale, 
> > > > max_scale);
> > > > BUG_ON(hscale < 0);
> > > >  
> > > > @@ -785,6 +782,13 @@ intel_update_plane(struct drm_plane *plane, struct 
> > > > drm_crtc *crtc,
> > > > crtc_w = drm_rect_width();
> > > > crtc_h = drm_rect_height();
> > > >  
> > > > +   drm_rect_rotate(, fb->width << 16, fb->height << 16,
> > > > +   intel_plane->rotation);
> > > > +
> > > > +   drm_rect_rotate(, intel_crtc->config.pipe_src_w,
> > > > +   intel_crtc->config.pipe_src_h,
> > > > +   intel_plane->rotation);
> > > > +
> > > > if (visible) {
> > > > /* check again in case clipping clamped the results */
> > > > hscale = drm_rect_calc_hscale(, , min_scale, 
> > > > max_scale);
> > > > @@ -811,7 +815,11 @@ intel_update_plane(struct drm_plane *plane, struct 
> > > > drm_crtc *crtc,
> > > >  drm_rect_height() * vscale - 
> > > > drm_rect_height());
> > > &

Re: [PATCH v2 02/11] drm: Add support_bits parameter to drm_property_create_bitmask()

[Sagar Arun Kamble]

Reviewed-by: Sagar Kamble 
Tested-by: Sagar Kamble 

On Mon, 2014-02-10 at 16:05 +0200, ville.syrj...@linux.intel.com wrote:
> From: Ville Syrjälä 
> 
> Make drm_property_create_bitmask() a bit more generic by allowing the
> caller to specify which bits are in fact supported. This allows multiple
> callers to use the same enum list, but still create different versions
> of the same property with different list of supported bits.
> 
> v2: Populate values[] array as non-sparse
> Make supported_bits 64bit
> Fix up omapdrm call site (Rob)
> 
> Cc: Tomi Valkeinen 
> Cc: Rob Clark 
> Cc: Sagar Kamble 
> Cc: dri-de...@lists.freedesktop.org
> Signed-off-by: Ville Syrjälä 
> ---
> This should also make my original "drm: Add 
> drm_mode_create_rotation_property()"
> patch do the right thing. So patch 03/11 of the series needs to be replaced 
> with the
> original.
> 
>  drivers/gpu/drm/drm_crtc.c   | 18 ++
>  drivers/gpu/drm/omapdrm/omap_plane.c |  5 -
>  include/drm/drm_crtc.h   |  3 ++-
>  3 files changed, 20 insertions(+), 6 deletions(-)
> 
> diff --git a/drivers/gpu/drm/drm_crtc.c b/drivers/gpu/drm/drm_crtc.c
> index 3b7d32d..6e099069 100644
> --- a/drivers/gpu/drm/drm_crtc.c
> +++ b/drivers/gpu/drm/drm_crtc.c
> @@ -2906,10 +2906,12 @@ EXPORT_SYMBOL(drm_property_create_enum);
>  struct drm_property *drm_property_create_bitmask(struct drm_device *dev,
>int flags, const char *name,
>const struct drm_prop_enum_list *props,
> -  int num_values)
> +  int num_props,
> +  uint64_t supported_bits)
>  {
>   struct drm_property *property;
> - int i, ret;
> + int i, ret, index = 0;
> + int num_values = hweight64(supported_bits);
>  
>   flags |= DRM_MODE_PROP_BITMASK;
>  
> @@ -2917,8 +2919,16 @@ struct drm_property 
> *drm_property_create_bitmask(struct drm_device *dev,
>   if (!property)
>   return NULL;
>  
> - for (i = 0; i < num_values; i++) {
> - ret = drm_property_add_enum(property, i,
> + for (i = 0; i < num_props; i++) {
> + if (!(supported_bits & (1ULL << props[i].type)))
> + continue;
> +
> + if (WARN_ON(index >= num_values)) {
> + drm_property_destroy(dev, property);
> + return NULL;
> + }
> +
> + ret = drm_property_add_enum(property, index++,
> props[i].type,
> props[i].name);
>   if (ret) {
> diff --git a/drivers/gpu/drm/omapdrm/omap_plane.c 
> b/drivers/gpu/drm/omapdrm/omap_plane.c
> index 046d5e6..0d97650 100644
> --- a/drivers/gpu/drm/omapdrm/omap_plane.c
> +++ b/drivers/gpu/drm/omapdrm/omap_plane.c
> @@ -309,7 +309,10 @@ void omap_plane_install_properties(struct drm_plane 
> *plane,
>   { DRM_REFLECT_Y,  "reflect-y" },
>   };
>   prop = drm_property_create_bitmask(dev, 0, "rotation",
> - props, ARRAY_SIZE(props));
> + props, ARRAY_SIZE(props),
> + BIT(DRM_ROTATE_0) | BIT(DRM_ROTATE_90) |
> + BIT(DRM_ROTATE_180) | 
> BIT(DRM_ROTATE_270) |
> + BIT(DRM_REFLECT_X) | 
> BIT(DRM_REFLECT_Y));
>   if (prop == NULL)
>   return;
>   priv->rotation_prop = prop;
> diff --git a/include/drm/drm_crtc.h b/include/drm/drm_crtc.h
> index d5c46c1..4e6d2aa 100644
> --- a/include/drm/drm_crtc.h
> +++ b/include/drm/drm_crtc.h
> @@ -1070,7 +1070,8 @@ extern struct drm_property 
> *drm_property_create_enum(struct drm_device *dev, int
>  struct drm_property *drm_property_create_bitmask(struct drm_device *dev,
>int flags, const char *name,
>const struct drm_prop_enum_list *props,
> -  int num_values);
> +  int num_props,
> +  uint64_t supported_bits);
>  struct drm_property *drm_property_create_range(struct drm_device *dev, int 
> flags,
>const char *name,
>uint64_t min, uint64_t max);


--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majord...@vger.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/

Re: [PATCH v2 02/11] drm: Add support_bits parameter to drm_property_create_bitmask()

[Sagar Arun Kamble]

Reviewed-by: Sagar Kamble sagar.a.kam...@intel.com
Tested-by: Sagar Kamble sagar.a.kam...@intel.com

On Mon, 2014-02-10 at 16:05 +0200, ville.syrj...@linux.intel.com wrote:
 From: Ville Syrjälä ville.syrjala at linux.intel.com
 
 Make drm_property_create_bitmask() a bit more generic by allowing the
 caller to specify which bits are in fact supported. This allows multiple
 callers to use the same enum list, but still create different versions
 of the same property with different list of supported bits.
 
 v2: Populate values[] array as non-sparse
 Make supported_bits 64bit
 Fix up omapdrm call site (Rob)
 
 Cc: Tomi Valkeinen tomi.valkei...@ti.com
 Cc: Rob Clark robdcl...@gmail.com
 Cc: Sagar Kamble sagar.a.kam...@intel.com
 Cc: dri-de...@lists.freedesktop.org
 Signed-off-by: Ville Syrjälä ville.syrj...@linux.intel.com
 ---
 This should also make my original drm: Add 
 drm_mode_create_rotation_property()
 patch do the right thing. So patch 03/11 of the series needs to be replaced 
 with the
 original.
 
  drivers/gpu/drm/drm_crtc.c   | 18 ++
  drivers/gpu/drm/omapdrm/omap_plane.c |  5 -
  include/drm/drm_crtc.h   |  3 ++-
  3 files changed, 20 insertions(+), 6 deletions(-)
 
 diff --git a/drivers/gpu/drm/drm_crtc.c b/drivers/gpu/drm/drm_crtc.c
 index 3b7d32d..6e099069 100644
 --- a/drivers/gpu/drm/drm_crtc.c
 +++ b/drivers/gpu/drm/drm_crtc.c
 @@ -2906,10 +2906,12 @@ EXPORT_SYMBOL(drm_property_create_enum);
  struct drm_property *drm_property_create_bitmask(struct drm_device *dev,
int flags, const char *name,
const struct drm_prop_enum_list *props,
 -  int num_values)
 +  int num_props,
 +  uint64_t supported_bits)
  {
   struct drm_property *property;
 - int i, ret;
 + int i, ret, index = 0;
 + int num_values = hweight64(supported_bits);
  
   flags |= DRM_MODE_PROP_BITMASK;
  
 @@ -2917,8 +2919,16 @@ struct drm_property 
 *drm_property_create_bitmask(struct drm_device *dev,
   if (!property)
   return NULL;
  
 - for (i = 0; i  num_values; i++) {
 - ret = drm_property_add_enum(property, i,
 + for (i = 0; i  num_props; i++) {
 + if (!(supported_bits  (1ULL  props[i].type)))
 + continue;
 +
 + if (WARN_ON(index = num_values)) {
 + drm_property_destroy(dev, property);
 + return NULL;
 + }
 +
 + ret = drm_property_add_enum(property, index++,
 props[i].type,
 props[i].name);
   if (ret) {
 diff --git a/drivers/gpu/drm/omapdrm/omap_plane.c 
 b/drivers/gpu/drm/omapdrm/omap_plane.c
 index 046d5e6..0d97650 100644
 --- a/drivers/gpu/drm/omapdrm/omap_plane.c
 +++ b/drivers/gpu/drm/omapdrm/omap_plane.c
 @@ -309,7 +309,10 @@ void omap_plane_install_properties(struct drm_plane 
 *plane,
   { DRM_REFLECT_Y,  reflect-y },
   };
   prop = drm_property_create_bitmask(dev, 0, rotation,
 - props, ARRAY_SIZE(props));
 + props, ARRAY_SIZE(props),
 + BIT(DRM_ROTATE_0) | BIT(DRM_ROTATE_90) |
 + BIT(DRM_ROTATE_180) | 
 BIT(DRM_ROTATE_270) |
 + BIT(DRM_REFLECT_X) | 
 BIT(DRM_REFLECT_Y));
   if (prop == NULL)
   return;
   priv-rotation_prop = prop;
 diff --git a/include/drm/drm_crtc.h b/include/drm/drm_crtc.h
 index d5c46c1..4e6d2aa 100644
 --- a/include/drm/drm_crtc.h
 +++ b/include/drm/drm_crtc.h
 @@ -1070,7 +1070,8 @@ extern struct drm_property 
 *drm_property_create_enum(struct drm_device *dev, int
  struct drm_property *drm_property_create_bitmask(struct drm_device *dev,
int flags, const char *name,
const struct drm_prop_enum_list *props,
 -  int num_values);
 +  int num_props,
 +  uint64_t supported_bits);
  struct drm_property *drm_property_create_range(struct drm_device *dev, int 
 flags,
const char *name,
uint64_t min, uint64_t max);


--
To unsubscribe from this list: send the line unsubscribe linux-kernel in
the body of a message to majord...@vger.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/

Re: [Intel-gfx] [PATCH v5 11/11] drm/i915: Calling rotate and inverse rotate transformations after clipping

[Sagar Arun Kamble]

On Tue, 2014-02-11 at 12:59 +0100, Daniel Vetter wrote:
 On Tue, Feb 11, 2014 at 05:02:31PM +0530, Sagar Arun Kamble wrote:
  On Mon, 2014-02-10 at 15:32 +0200, Ville Syrjälä wrote:
   On Mon, Feb 10, 2014 at 01:01:18PM +0530, sagar.a.kam...@intel.com wrote:
From: Sagar Kamble sagar.a.kam...@intel.com

With clipped sprites these transformations are not working. these
functions transform complete sprite irrespective of clipping present.
This leads to invisible portion of sprite show up when rotate 180 if
it was out of visible area before.

v4: Moved rotate transform for source rectangle after clipping.
Added rotate and inverse rotate transform for destination rect.
   
   Still NAK.
   
   I just pushed rotation support to my glplane test app [1], and with
   with that my rotated clipping code works exactly as intended.
   
   [1] git://gitorious.org/vsyrjala/glplane.git
  I tried this app. I think I am considering 180 degree rotation of
  clipped sprite plane differently. I have captured output with these
  rotate transforms moved before(clip-rotated) and after(rotate-clipped)
  clipping code.
  
  Which is valid? Rotating entire sprite and then clipping or Rotating
  clipped portion?
  
  Reference and Rotated output is attached FYI.
  
  If rotating entire sprite is correct then this patch 11/11 is not needed
  and can be abandoned.
 
 I think doing the clipping first (as a viewport) and then rotating the
 viewport makes more sense to me. Dunno what other people think, but I
 guess we should document that somewhere.
 -Daniel

Agree. Thought of use-case where if media player(that will use sprite
for video) is moved such that it is off the screen partially, after
rotation as well we should see same portion of clip inside(rotated
though).

 
  
  thanks,
  Sagar

Cc: Daniel Vetter daniel.vet...@ffwll.ch
Cc: Jani Nikula jani.nik...@linux.intel.com
Cc: David Airlie airl...@linux.ie
Cc: dri-de...@lists.freedesktop.org
Cc: linux-kernel@vger.kernel.org
Cc: vijay.a.purushotha...@intel.com
Signed-off-by: Sagar Kamble sagar.a.kam...@intel.com
---
 drivers/gpu/drm/i915/intel_sprite.c | 16 
 1 file changed, 12 insertions(+), 4 deletions(-)

diff --git a/drivers/gpu/drm/i915/intel_sprite.c 
b/drivers/gpu/drm/i915/intel_sprite.c
index 62b9f84..799f6a9 100644
--- a/drivers/gpu/drm/i915/intel_sprite.c
+++ b/drivers/gpu/drm/i915/intel_sprite.c
@@ -769,9 +769,6 @@ intel_update_plane(struct drm_plane *plane, struct 
drm_crtc *crtc,
max_scale = intel_plane-max_downscale  16;
min_scale = intel_plane-can_scale ? 1 : (1  16);
 
-   drm_rect_rotate(src, fb-width  16, fb-height  16,
-   intel_plane-rotation);
-
hscale = drm_rect_calc_hscale_relaxed(src, dst, min_scale, 
max_scale);
BUG_ON(hscale  0);
 
@@ -785,6 +782,13 @@ intel_update_plane(struct drm_plane *plane, struct 
drm_crtc *crtc,
crtc_w = drm_rect_width(dst);
crtc_h = drm_rect_height(dst);
 
+   drm_rect_rotate(src, fb-width  16, fb-height  16,
+   intel_plane-rotation);
+
+   drm_rect_rotate(dst, intel_crtc-config.pipe_src_w,
+   intel_crtc-config.pipe_src_h,
+   intel_plane-rotation);
+
if (visible) {
/* check again in case clipping clamped the results */
hscale = drm_rect_calc_hscale(src, dst, min_scale, 
max_scale);
@@ -811,7 +815,11 @@ intel_update_plane(struct drm_plane *plane, struct 
drm_crtc *crtc,
 drm_rect_height(dst) * vscale - 
drm_rect_height(src));
 
drm_rect_rotate_inv(src, fb-width  16, fb-height 
 16,
-   intel_plane-rotation);
+   intel_plane-rotation);
+
+   drm_rect_rotate_inv(dst, intel_crtc-config.pipe_src_w,
+   intel_crtc-config.pipe_src_h,
+   intel_plane-rotation);
 
/* sanity check to make sure the src viewport wasn't 
enlarged */
WARN_ON(src.x1  (int) src_x ||
-- 
1.8.5

___
Intel-gfx mailing list
intel-...@lists.freedesktop.org
http://lists.freedesktop.org/mailman/listinfo/intel-gfx
   
  
 
 
 
 
 
 


--
To unsubscribe from this list: send the line unsubscribe linux-kernel in
the body of a message to majord...@vger.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/

Re: [Intel-gfx] [PATCH v5 11/11] drm/i915: Calling rotate and inverse rotate transformations after clipping

[Sagar Arun Kamble]

On Tue, 2014-02-11 at 16:56 +0200, Ville Syrjälä wrote:
 On Tue, Feb 11, 2014 at 05:02:31PM +0530, Sagar Arun Kamble wrote:
  On Mon, 2014-02-10 at 15:32 +0200, Ville Syrjälä wrote:
   On Mon, Feb 10, 2014 at 01:01:18PM +0530, sagar.a.kam...@intel.com wrote:
From: Sagar Kamble sagar.a.kam...@intel.com

With clipped sprites these transformations are not working. these
functions transform complete sprite irrespective of clipping present.
This leads to invisible portion of sprite show up when rotate 180 if
it was out of visible area before.

v4: Moved rotate transform for source rectangle after clipping.
Added rotate and inverse rotate transform for destination rect.
   
   Still NAK.
   
   I just pushed rotation support to my glplane test app [1], and with
   with that my rotated clipping code works exactly as intended.
   
   [1] git://gitorious.org/vsyrjala/glplane.git
  I tried this app. I think I am considering 180 degree rotation of
  clipped sprite plane differently. I have captured output with these
  rotate transforms moved before(clip-rotated) and after(rotate-clipped)
  clipping code.
  
  Which is valid? Rotating entire sprite and then clipping or Rotating
  clipped portion?
  
  Reference and Rotated output is attached FYI.
  
  If rotating entire sprite is correct then this patch 11/11 is not needed
  and can be abandoned.
 
 The way I think of these things is roughly this:
 
 You have the user specified source rectangle, where the coordinates specify
 the viewport into the framebuffer. This coordinate space is oriented the
 same was as the framebuffer itself, ie. the first pixel of the
 framebuffer is at coordinates 0,0. So plane rotation doesn't affect this
 at all.
 
 Then you have the user specified destination/crtc rectangle, where the
 coordinates specify the position of the plane within the crtc coordinate
 space. So the first visible pixel the pipe will push out is at
 coordinates 0,0. So again plane rotation doesn't affect this.
 
 Then you have the rotation which simply specifies the transformation to
 be applied to the pixels when they move from the source rectangle to
 the destination rectangle. So w/ 0 degree rotation the pixel at
 src_x,src_y in the framebuffer will appear at position crtc_x,crtc_y
 on the crtc output. With 180 degree rotation the pixel at src_x,src_y
 will appear at crtc_x+crtc_w-1,crtc_y+crtc_h-1.
 
 As clipping happens in the crtc coordinate space, we need to orient
 the source coordindates the same way to get the correct clipping result.
 So for example with 0 degrees rotation clipping the left side of the
 destination rectangle must result in clipping the left side of the source
 rectangle as well. And with 180 degree rotation clipping the destination
 rectangle on the left side must result in clipping the source rectangle
 on the right side. Left and right in each case referring to the original
 unrotate coordinates.
 
 So let's say we have the following situation w/ 180 degree rotation.
 The letters inside the rects represented specific named pixels,
 the FB rectangle represents the FB as specified by addfb2 ioctl,
 the CRTC rectangle represents the pipe output (0,0 - PIPESRC.w,h):
 
 FB: CRTC:
 0,0 ___ 0.0 __
 |  abcd   | | |
 |  efgh   | | |
 |_| |hgfe |
 |dcba_|
 unclipped coordinates specified by user:
 src_x=2,src_y=0 crtc_x=0,crtc_y=2
 src_w=4.src_h=2 crtc_w=4,crtc_h=2
 
 clipped coordinates:
 src_x=2,src_y=0 crtc_x=0,crtc_y=2
 src_w=4.src_h=2 crtc_w=4,crtc_h=2
 
 
 Then the user moves the sprite one pixel to the left resulting on some
 clipping (the X pixels). Note that the unclipped source coordinates do
 not change here at all, in fact crtc_x is the only thing changed by the
 user:
 
 FB: CRTC:
 0,0 ___ 0.0 __
 |  abcX   | | |
 |  efgX   | | |
 |_| Xgfe  |
 Xcba__|
 unclipped coordinates specified by user:
 src_x=2,src_y=0 crtc_x=-1,crtc_y=2
 src_w=4.src_h=2 crtc_w= 4,crtc_h=2
 
 clipped coordinates:
 src_x=2,src_y=0 crtc_x=0,crtc_y=2
 src_w=3.src_h=2 crtc_w=3,crtc_h=2
 


Understood this now. Thank you Ville for providing this elaborate
graphical view of the rotation cases.

-Sagar

--
To unsubscribe from this list: send the line unsubscribe linux-kernel in
the body of a message to majord...@vger.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/