Change-Id: I66007a7f7e4e27fb129121f36143dce3cfb43738
Signed-off-by: Monk Liu
---
drivers/gpu/drm/amd/amdgpu/amdgpu.h | 31 ++
drivers/gpu/drm/amd/amdgpu/gfx_v8_0.c | 61 +++
2 files changed, 92 insertions(+)
diff --git
*** BLURB HERE ***
to enable GFX SRIOV we need insert some new packag in gfx ring
which called META-DATA write command, and meta-data is composed
by two parts: CE and DE.
CE's meta-data write need prior to CNTX_CNTL package while
DE's should following that.
meta-data will notify CP about the CSA
META-DATA is used in GFX cmd submit, we have two
types of META-DATA now, one is legacy and the other
is for chained-ib-preemption, which is used in vulkan
UMD.
Change-Id: I17324905b4822a93fd3ffb3e8f86e544871b89ab
Signed-off-by: Monk Liu
---
ce's META-DATA write need prior to cntx_cntrl package
and de's should following.
Change-Id: I2bf436d8a16bae38ce484f846f0b399e16cfe80a
Signed-off-by: Monk Liu
---
drivers/gpu/drm/amd/amdgpu/amdgpu.h| 1 +
drivers/gpu/drm/amd/amdgpu/amdgpu_ib.c | 3 +++
> -Original Message-
> From: Michel Dänzer [mailto:mic...@daenzer.net]
> Sent: Thursday, January 12, 2017 1:45 PM
> To: Liu, Monk ; Christian König
> ; Yu, Xiangliang
> Cc: amd-gfx@lists.freedesktop.org
> Subject: Re: [V3
On 12/01/17 12:21 PM, Liu, Monk wrote:
>
>if (in_interrupt())
> BUG();
Current coding style is
BUG_ON(in_interrupt());
according to https://kernelnewbies.org/FAQ/BUG .
--
Earthling Michel Dänzer | http://www.amd.com
Libre software enthusiast
Xiangliang
please BUG() when register access occured in RUNTIME and IRQ context, e.g.:
for register read:
if (amdgpu_sriov_runtime(adev)) {
if (in_interrupt())
BUG();
else
return amdgpu_virt_kiq_rreg(adev, reg, v);
}
and also for register write, with above addressed,
Xiangliang
please BUG() when register access occured in RUNTIME and IRQ context, e.g.:
if (amdgpu_sriov_runtime(adev)) {
}
return amdgpu_virt_kiq_wreg(adev, reg, v);
with above addressed, Reviewed-by: Monk Liu
发件人: amd-gfx
Hi Xie,
Perhaps you want to use `fprintf(stderr, "...")` over `printf("..")` and
lose the space before the start parenthesis. Also, line wrap
your commit message.
Side note, use git send-email so that the patch is inline and not a HTML
email for easy review and application of the patch.
Kind
Hi,
Provide more specific error message if non-privileged user runs amdgpu_test
Before this change, the error message is:" WARNING - Suite initialization
failed..." People might think this is a driver problem.
Tested with non-privileged user. Now the error message is like.
...
Rex Zhu wrote:
in profiling mode, powerplay will fix power state
as stable as possible.and disable gfx cg and LBPW feature.
profile_standard: as a prerequisite, ensure power and thermal
sustainable, set clocks ratio as close to the highest clock
ratio as possible.
profile_min_sclk: fix mclk as
Hi,
v2: Use strerror instead of %m. %m is a GNU C Library extension.
Thanks,
Alex Bin Xie
From: Xie, AlexBin
Sent: Wednesday, January 11, 2017 4:14 PM
To: amd-gfx@lists.freedesktop.org
Subject: More specific libdrm error message
Hi,
Provide more specific
Difference families may have different numbers of rings. Use
the variable rather than a hardcoded number.
Signed-off-by: Alex Deucher
---
drivers/gpu/drm/amd/amdgpu/amdgpu_cs.c | 4 ++--
1 file changed, 2 insertions(+), 2 deletions(-)
diff --git
> -Original Message-
> From: amd-gfx [mailto:amd-gfx-boun...@lists.freedesktop.org] On Behalf
> Of Deucher, Alexander
> Sent: Wednesday, January 11, 2017 2:25 PM
> To: Wentland, Harry; amd-gfx@lists.freedesktop.org
> Cc: Aberback, Joshua
> Subject: RE: [PATCH 05/10] drm/amd/display:
> -Original Message-
> From: amd-gfx [mailto:amd-gfx-boun...@lists.freedesktop.org] On Behalf
> Of Harry Wentland
> Sent: Wednesday, January 11, 2017 1:11 PM
> To: amd-gfx@lists.freedesktop.org
> Cc: Aberback, Joshua
> Subject: [PATCH 05/10] drm/amd/display: Support 64-bit Polaris11 5k VSR
From: Joshua Aberback
Change-Id: If711bd494fcece078ef9c09eb542b510dcd68885
Signed-off-by: Joshua Aberback
Reviewed-by: Tony Cheng
Acked-by: Harry Wentland
---
From: Ding Wang
Use DVO as pixel clock for DP before enabling link
PHY. Otherwise, when switching from HDMI passive dongle
to DP on the same connector, the PHY PLL is used as
pixel clock, and CRTC would stop working.
Change-Id: Iad2e20eaac6675ce53f9fec6781e75151c43666f
From: Dave Airlie
These aren't needed, and aren't really used in too many places.
Change-Id: I1a52c694d6d8a84f5756fed4f60f295173cba6b5
Signed-off-by: Dave Airlie
Reviewed-by: Tony Cheng
Reviewed-by: Harry Wentland
From: Hersen Wu
1. remove the sleep mechanism while set_blank true from
the timing generator.
Since Hw sequencer is the one that manages the flow
he will be responsible for wait for blanck in a critical places.
Change-Id: Idd2136fced775b5c9012c4034a6ee1306fce0945
From: Yongqiang Sun
Change-Id: Ic6c140575b5ffb6ce32e75d3a3725312a8be6942
Signed-off-by: Yongqiang Sun
Reviewed-by: Tony Cheng
Acked-by: Harry Wentland
---
* bunch of fixes, including for Polaris 12
* some cleanup, in particular to use HDMI header for HDMI defines
Harry
Arindam Nath (1):
drm/amd/display: fix cursor disappearing after resume
Dave Airlie (2):
drm/amd/display: drop min/max wrappers
drm/amd/display: start using linux hdmi
From: Dave Airlie
DAL has defines for things, and it doesn't even use them itself.
Change-Id: Ic94412df212f8d66b4ad482fcd70de46b7b8c61a
Signed-off-by: Dave Airlie
Reviewed-by: Tony Cheng
Reviewed-by: Harry Wentland
From: Arindam Nath
Since during suspend, the cursor registers are cleared,
once the system resumes back, the cursor remains disabled.
cursor_set_attributes() only sets the cursor attributes
along with cursor size and surface address, but does not
enable the cursor back on.
From: Joshua Aberback
- pass full asic_id info into bw_calc_init instead of only version enum
- 64-bit Polaris11 needs an extra microsecond of dmif_urgent_latency
- add helper to convert from asic_id.family to bw_calc version enum
Change-Id:
From: Leon Elazar
blank/unblanck functionality apply_ctx_for_surface will update planes visibility
by manipulating Blender and CRTC HW modules.
Change-Id: I0fa0b46cb36de203b07af7a846a740e81fda6233
Signed-off-by: Leon Elazar
Reviewed-by: Dmytro
From: Joshua Aberback
Change-Id: I77c287c6d6fe2e46f232e707383347f11ade91c0
Signed-off-by: Joshua Aberback
Reviewed-by: Tony Cheng
Acked-by: Harry Wentland
---
> -Original Message-
> From: amd-gfx [mailto:amd-gfx-boun...@lists.freedesktop.org] On Behalf
> Of Michel Dänzer
> Sent: Wednesday, January 11, 2017 4:36 AM
> To: amd-gfx@lists.freedesktop.org
> Cc: Harald Judt; Johannes Hirte
> Subject: [PATCH] Revert "drm/amdgpu: Only update the CUR_SIZE
From: Nicolai Hähnle
Unlock the resv lock only if we were the ones to lock it in the first
place.
Signed-off-by: Nicolai Hähnle
Reviewed-by: Edward O'Callaghan
Reviewed-by: Christian König
From: Nicolai Hähnle
ttm_bo_init checks that the reservation object is locked. This is
the caller's responsibility when resv != NULL. Otherwise, the inline
reservation object of the newly allocated buffer is used and must
explicitly be locked.
Uninterruptible w/w locks
[ML] do you mean If some where want to use vmid0_gds_address I should just use
vmid0_csa_addr + 4096 ?
Yes.
Can I get your RB after above two place modified in patch ?
Yeah, with that fixed the patch is Reviewed-by: Christian König
.
Regards,
Christian.
Am
Reviewed-by: Monk Liu
-Original Message-
From: amd-gfx [mailto:amd-gfx-boun...@lists.freedesktop.org] On Behalf Of
Xiangliang Yu
Sent: Wednesday, January 11, 2017 9:18 PM
To: amd-gfx@lists.freedesktop.org
Cc: Yu, Xiangliang
Subject: [V3 09/11] drm/amdgpu/virt:
Sounds make sense!
BR Monk
-Original Message-
From: Yu, Xiangliang
Sent: Wednesday, January 11, 2017 11:04 PM
To: Yu, Xiangliang ; Liu, Monk ;
amd-gfx@lists.freedesktop.org
Cc: Liu, Shaoyun
Subject: RE: [V3 06/11]
> > We will discuss it tomorrow, I'm not sure of your approach right now
> >
> > We can stop running driver right after IDH_FLR_NOTIFICATION received
> > and continue driver right after IDH_FLR_NOTIFICATION_CMPL
If missing the interrupt, I think we can't easy to handle the error. That what
i
> We will discuss it tomorrow, I'm not sure of your approach right now
>
> We can stop running driver right after IDH_FLR_NOTIFICATION received and
> continue driver right after IDH_FLR_NOTIFICATION_CMPL
I have verified the method on my platform and work fine.
Let's discuss it tomorrow if you
> -Original Message-
> From: Liu, Monk
> Sent: Wednesday, January 11, 2017 10:33 PM
> To: Yu, Xiangliang ; amd-
> g...@lists.freedesktop.org
> Subject: RE: [V3 05/11] drm/amdgpu/virt: add high level interfaces for virt
>
> See in lines
>
>
> -Original
We should BUG and not access register at all if found during RUNTIME &&
IRQ_context
By far.
BR Monk
-Original Message-
From: Christian König [mailto:deathsim...@vodafone.de]
Sent: Wednesday, January 11, 2017 10:44 PM
To: Liu, Monk ; Yu, Xiangliang
We will discuss it tomorrow, I'm not sure of your approach right now
We can stop running driver right after IDH_FLR_NOTIFICATION received and
continue driver right after IDH_FLR_NOTIFICATION_CMPL
BR Monk
-Original Message-
From: Yu, Xiangliang
Sent: Wednesday, January 11, 2017 10:43
>Similar to the GDS address inside the VMs I think the code actually using it
>should calculate it from the BO address.
[ML] do you mean If some where want to use vmid0_gds_address I should just use
vmid0_csa_addr + 4096 ?
>The validate buffer is only temporarily used during mapping the CSA
> -Original Message-
> From: Liu, Monk
> Sent: Wednesday, January 11, 2017 10:28 PM
> To: Yu, Xiangliang ; amd-
> g...@lists.freedesktop.org
> Cc: Yu, Xiangliang ; Liu, Shaoyun
>
> Subject: RE: [V3 06/11] drm/amdgpu/virt:
See in lines
-Original Message-
From: Yu, Xiangliang
Sent: Wednesday, January 11, 2017 10:27 PM
To: Liu, Monk ; amd-gfx@lists.freedesktop.org
Subject: RE: [V3 05/11] drm/amdgpu/virt: add high level interfaces for virt
> This patch is not derived from my work
> This patch is not derived from my work (amd-sriov-4.3/4.6), please don't add
> my "signed-off-by: Monk Liu " line on it, Seems you add an upper layer
> interface to invoke on gpu_request/release routines, And I will also take role
> in the reviewing of it:
>
> + adev->virt.caps &=
Am 11.01.2017 um 15:09 schrieb Nicolai Hähnle:
On 11.01.2017 12:56, Christian König wrote:
Am 11.01.2017 um 08:31 schrieb Nicolai Hähnle:
From: Nicolai Hähnle
ttm_bo_init checks that the reservation object is locked. This is
the caller's responsibility when resv !=
Because if we are in interrupt , we are forbid to do schedule, and use kiq to
read register will invoke fence_wait() ...
That won't work. Locking a mutext like it is done in the write path can
cause scheduling as well.
If we need to push writes through the KIQ in interrupt context we need
to
Hi Xiangliang
This patch is not derived from my work (amd-sriov-4.3/4.6), please don't add my
"signed-off-by: Monk Liu " line on it,
Seems you add an upper layer interface to invoke on gpu_request/release
routines,
And I will also take role in the reviewing of it:
+
On 11.01.2017 12:56, Christian König wrote:
Am 11.01.2017 um 08:31 schrieb Nicolai Hähnle:
From: Nicolai Hähnle
ttm_bo_init checks that the reservation object is locked. This is
the caller's responsibility when resv != NULL. Otherwise, the inline
reservation object of
> Patch #7 is missing a comment that this applies during driver load/unload,
> cause that isn't obvious.
Not just only for loading/unloading, resume process will need to
request/release full gpu access after FLR (function level reset). Later I'll
upstream FLR patches.
>
> With that fixed the
On 11.01.2017 12:56, Christian König wrote:
Am 11.01.2017 um 08:31 schrieb Nicolai Hähnle:
From: Nicolai Hähnle
ttm_bo_init checks that the reservation object is locked. This is
the caller's responsibility when resv != NULL. Otherwise, the inline
reservation object of
Because if we are in interrupt , we are forbid to do schedule, and use kiq to
read register will invoke fence_wait() ...
If you think that's odds, we can use BUG_ON() to stop driver running if we need
read registers in SRIOV case.
And to fully support register reading, we need to implement a
Patches #1-#3, #5 and #8 are Reviewed-by: Christian König
.
Patch #6, #9 and #10 are Acked-by: Christian König
.
Patch #4: Just send out a question on this.
Patch #7 is missing a comment that this applies during driver
load/unload, cause
Am 11.01.2017 um 14:18 schrieb Xiangliang Yu:
For virtualization, it is must for driver to use KIQ to access
registers when it is out of GPU full access mode.
Signed-off-by: Xiangliang Yu
Signed-off-by: Monk Liu
---
Current amdgpu reset process only works on bare-metal and for
SRIOV many inside it need re-work to adapt to vf device.
This is a temporary workaround to skip gpu reset.
Signed-off-by: Monk Liu
Signed-off-by: Xiangliang Yu
---
Virtualization don't need the dc, disable it.
Signed-off-by: Xiangliang Yu
---
drivers/gpu/drm/amd/amdgpu/amdgpu_device.c | 3 +++
1 file changed, 3 insertions(+)
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c
b/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c
index
Virtual display is default setting for virtualization, enable it.
Signed-off-by: Xiangliang Yu
---
drivers/gpu/drm/amd/amdgpu/amdgpu_virt.c | 4
1 file changed, 4 insertions(+)
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_virt.c
For gpu vf device, first need to request full gpu access before
accessing gpu registers, and release full gpu access after the
access is done.
Signed-off-by: Xiangliang Yu
Signed-off-by: Monk Liu
---
drivers/gpu/drm/amd/amdgpu/amdgpu_device.c | 9
Call VI virtualization functions if device is Vf.
Signed-off-by: Xiangliang Yu
Signed-off-by: Monk Liu
---
drivers/gpu/drm/amd/amdgpu/vi.c | 34 --
1 file changed, 32 insertions(+), 2 deletions(-)
diff --git
This patch series will enable AMD GPU virtualization and wires it
into amdgpu modules.
Changes in V3:
1. rename field name;
2. remove extra fence_get()/fence_put();
3. fixed coding style;
4. update patch comment;
Changes in V2:
1. define new high level interface for non-asic specific;
2. add
For virtualization, it is must for driver to use KIQ to access
registers when it is out of GPU full access mode.
Signed-off-by: Xiangliang Yu
Signed-off-by: Monk Liu
---
drivers/gpu/drm/amd/amdgpu/Makefile| 2 +-
Add new flag to define gpu runtime that is out of full gpu access.
Signed-off-by: Xiangliang Yu
Signed-off-by: Monk Liu
---
drivers/gpu/drm/amd/amdgpu/amdgpu_virt.h | 4
1 file changed, 4 insertions(+)
diff --git
VI has asic specific virt support, which including mailbox and
golden registers init.
Signed-off-by: Xiangliang Yu
Signed-off-by: Monk Liu
Signed-off-by: shaoyunl
---
drivers/gpu/drm/amd/amdgpu/Makefile | 2 +-
During virtual runtime, need to send command to kiq ring to
read/write GPU registers. Add two interface to support the two
actions.
Signed-off-by: Xiangliang Yu
Signed-off-by: Monk Linu
---
drivers/gpu/drm/amd/amdgpu/amdgpu.h | 2 ++
Add high level interfaces that is not relate to specific asic. So
asic files just need to implement the interfaces to support
virtualization.
Signed-off-by: Xiangliang Yu
Signed-off-by: Monk Liu
---
drivers/gpu/drm/amd/amdgpu/amdgpu_virt.c | 57
Implement emit_rreg/wreg function for kiq ring.
Signed-off-by: Xiangliang Yu
Signed-off-by: Monk Liu
---
drivers/gpu/drm/amd/amdgpu/amdgpu_virt.h | 3 ++-
drivers/gpu/drm/amd/amdgpu/gfx_v8_0.c| 37
2 files changed,
Am 11.01.2017 um 11:43 schrieb Monk Liu:
Change-Id: I528c2f324830aaa21ab8d8250bc80a2a6bab33bd
Signed-off-by: Monk Liu
---
drivers/gpu/drm/amd/amdgpu/amdgpu_cs.c | 14 ++
drivers/gpu/drm/amd/amdgpu/amdgpu_device.c | 12
Am 11.01.2017 um 11:43 schrieb Monk Liu:
for SRIOV usage, CSA is only used per device and each
VM will map on it.
Change-Id: Ibc63d87be59d7d89b96117e4d1b7e970a872580d
Signed-off-by: Monk Liu
---
drivers/gpu/drm/amd/amdgpu/Makefile | 2 +-
Am 11.01.2017 um 11:43 schrieb Monk Liu:
and implement CSA functions in this file
Change-Id: Ife0eff7b13b8b5946f005a39f6ecb8db1cb72c38
Signed-off-by: Monk Liu
---
drivers/gpu/drm/amd/amdgpu/amdgpu_virt.h | 8
drivers/gpu/drm/amd/amdgpu/amdgpu_vm.h | 3 +++
2
Reviewed-by: Tom St Denis
From: amd-gfx on behalf of Huang Rui
Sent: Tuesday, January 10, 2017 21:30
To: amd-gfx@lists.freedesktop.org
Cc: StDenis, Tom; Huang, Ray
Subject: [PATCH]
Am 11.01.2017 um 08:31 schrieb Nicolai Hähnle:
From: Nicolai Hähnle
ttm_bo_init checks that the reservation object is locked. This is
the caller's responsibility when resv != NULL. Otherwise, the inline
reservation object of the newly allocated buffer is used and must
Am 11.01.2017 um 08:31 schrieb Nicolai Hähnle:
From: Nicolai Hähnle
Unlock the resv lock only if we were the ones to lock it in the first
place.
Signed-off-by: Nicolai Hähnle
Reviewed-by: Christian König .
---
Am 11.01.2017 um 04:48 schrieb Rex Zhu:
Change-Id: I1495894d01ca10e5d496e71f5d4b1a93a12336f3
Signed-off-by: Rex Zhu
Not deeply into VCE programming any more, but both patches look like
they make sense and are Acked-by: Christian König
.
Regards,
Ah, that issue again. Yeah that was fixed in the meantime. Please remove
the extra fance_get()/fence_put() for upstreaming.
If an old branch still has that issue we should backport the proper
fixes and not work around them like this.
Thanks for looking into it,
Christian.
Am 11.01.2017 um
Am 11.01.2017 um 07:31 schrieb Flora Cui:
Change-Id: I9fcdb4bc69e5f5c80eadb1e5fd31570802b0ce70
Signed-off-by: Flora Cui
Reviewed-by: Christian König .
---
drivers/gpu/drm/amd/amdgpu/gmc_v6_0.c | 23 +--
1 file changed, 9
v2:
drop invoking gem_va and update CSA address in CS routines,
v3:
reserve CSA bo and call vm_bo_rmv() to remove CSA mappings
prior to vm_fini()
Monk Liu (3):
drm/amdgpu:new field members for SRIOV
drm/amdgpu:add new file for SRIOV
drm/amdgpu:invoke CSA functions
for SRIOV usage, CSA is only used per device and each
VM will map on it.
Change-Id: Ibc63d87be59d7d89b96117e4d1b7e970a872580d
Signed-off-by: Monk Liu
---
drivers/gpu/drm/amd/amdgpu/Makefile | 2 +-
drivers/gpu/drm/amd/amdgpu/amdgpu_virt.c | 95
and implement CSA functions in this file
Change-Id: Ife0eff7b13b8b5946f005a39f6ecb8db1cb72c38
Signed-off-by: Monk Liu
---
drivers/gpu/drm/amd/amdgpu/amdgpu_virt.h | 8
drivers/gpu/drm/amd/amdgpu/amdgpu_vm.h | 3 +++
2 files changed, 11 insertions(+)
diff --git
Change-Id: I528c2f324830aaa21ab8d8250bc80a2a6bab33bd
Signed-off-by: Monk Liu
---
drivers/gpu/drm/amd/amdgpu/amdgpu_cs.c | 14 ++
drivers/gpu/drm/amd/amdgpu/amdgpu_device.c | 12
drivers/gpu/drm/amd/amdgpu/amdgpu_kms.c| 14 ++
From: Michel Dänzer
This reverts commits 7c83d7abc9997cf1efac2c0ce384b5e8453ee870 and
a1f49cc179ce6b7b7758ae3ff5cdb138d0ee0f56.
They caused the HW cursor to disappear under various circumstances in
the wild. I wasn't able to reproduce any of them, and I'm not sure
what's
Hi Christian
latest updates on the fence grab/release
I recall the impression of why I added those grab/release pare:
I checked the history, the first time I added those pare is because they are
first introduced/implemented in amd-sriov-4.3, and that branch's fence
implement has big gap
77 matches
Mail list logo