On Wed, Jun 24, 2015 at 10:06 PM, Alex Deucher wrote:
> On Fri, Jun 12, 2015 at 4:08 AM, Oded Gabbay wrote:
>> From: Ben Goz
>>
>> Signed-off-by: Ben Goz
>> Acked-by: Oded Gabbay
>
> This patch effectively reverts a fix for a hw bug IIRC.
>
> Alex
Hmm,
From: Ben Goz
v2: add missing MTYPE_NONCACHED enum
Signed-off-by: Ben Goz
Acked-by: Oded Gabbay
Reviewed-by: Alex Deucher
---
drivers/gpu/drm/amd/amdgpu/cikd.h | 6 +
drivers/gpu/drm/amd/amdgpu/gfx_v7_0.c | 42 ++
drivers/gpu/drm/amd/amdgpu/gfx_v8_0.c
On Wed, Jun 24, 2015 at 10:31 PM, Alex Deucher wrote:
> On Wed, Jun 24, 2015 at 3:28 PM, Oded Gabbay wrote:
>> On Wed, Jun 24, 2015 at 10:25 PM, Alex Deucher
>> wrote:
>>> On Fri, Jun 12, 2015 at 4:08 AM, Oded Gabbay
>>> wrote:
>>>> From: Ben Goz
On Wed, Jun 24, 2015 at 10:25 PM, Alex Deucher wrote:
> On Fri, Jun 12, 2015 at 4:08 AM, Oded Gabbay wrote:
>> From: Ben Goz
>>
>> Signed-off-by: Ben Goz
>> Acked-by: Oded Gabbay
>> Reviewed-by: Alex Deucher
>
> Does this depend on some other patch?
nges up to 7861c7a4ca5c637f3cad1e343602a15e857140fd:
drm/amdkfd: fix some range checks in address watch ioctl (2015-06-16 15:42:26
+0300)
Dan Carpenter (1):
drm/amdkfd: fix some range checks in address watch ioctl
Oded Gabbay (1):
drm/radeon: remov
On Thu, Jun 11, 2015 at 6:19 PM, Dan Carpenter
wrote:
>
> diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_chardev.c
> b/drivers/gpu/drm/amd/amdkfd/kfd_chardev.c
> index 96c904b..54a608a 100644
> --- a/drivers/gpu/drm/amd/amdkfd/kfd_chardev.c
> +++ b/drivers/gpu/drm/amd/amdkfd/kfd_chardev.c
> @@ -553
On Thu, Jun 11, 2015 at 6:19 PM, Dan Carpenter
wrote:
> #include "kfd_dbgmgr.h"
> +#include "../../radeon/cik_reg.h"
>
> static long kfd_ioctl(struct file *, unsigned int, unsigned long);
We can't include radeon files in kfd_chardev.c, because amdkfd works
with two device drivers - radeon and
From: Ben Goz
Signed-off-by: Ben Goz
Reviewed-by: Yair Shachar
Signed-off-by: Oded Gabbay
---
drivers/gpu/drm/amd/amdkfd/kfd_topology.c | 5 +
drivers/gpu/drm/amd/amdkfd/kfd_topology.h | 1 +
2 files changed, 6 insertions(+)
diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_topology.c
b
This patch makes use of the new amd headers (that are part of the new
amdgpu driver), instead of private defines.
Signed-off-by: Oded Gabbay
---
drivers/gpu/drm/amd/amdkfd/cik_regs.h| 11 ---
.../drm/amd/amdkfd/kfd_device_queue_manager_cik.c| 12
drivers
From: Ben Goz
Signed-off-by: Ben Goz
Signed-off-by: Oded Gabbay
---
drivers/gpu/drm/amd/amdkfd/kfd_packet_manager.c | 91 +++--
1 file changed, 87 insertions(+), 4 deletions(-)
diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_packet_manager.c
b/drivers/gpu/drm/amd/amdkfd
From: Ben Goz
The MAP_QUEUES packet length for Carrizo is different than for Kaveri.
Therefore, we now need to calculate the runlist length with regard to the
underlying H/W.
Signed-off-by: Ben Goz
Signed-off-by: Oded Gabbay
---
drivers/gpu/drm/amd/amdkfd/kfd_packet_manager.c | 8 +++-
1
From: Ben Goz
This patch adds support for the VI APU in the DQM module.
Most of the functionality of DQM is shared between CI and VI. Therefore,
only a handful of functions are required to be in the
H/W-specific part of DQM.
Signed-off-by: Ben Goz
Signed-off-by: Oded Gabbay
---
.../drm/amd
From: Ben Goz
This patch implements all the VI MQD manager functions.
This is done in a different file as the MQD format is different
between CI and VI
Signed-off-by: Ben Goz
Signed-off-by: Oded Gabbay
---
drivers/gpu/drm/amd/amdkfd/Makefile | 3 +-
drivers/gpu/drm/amd/amdkfd
From: Ben Goz
Signed-off-by: Ben Goz
Signed-off-by: Oded Gabbay
---
drivers/gpu/drm/amd/amdkfd/kfd_pm4_headers_vi.h | 398
1 file changed, 398 insertions(+)
create mode 100644 drivers/gpu/drm/amd/amdkfd/kfd_pm4_headers_vi.h
diff --git a/drivers/gpu/drm/amd/amdkfd
From: Ben Goz
This patch adds the PCI IDs of supported CZ devices to the
supported_devices structure in amdkfd. That structure is used during the
amdkfd probing stage, to check if the currently probed device is eligible
to be handled by amdkfd.
Signed-off-by: Ben Goz
Signed-off-by: Oded Gabbay
Signed-off-by: Oded Gabbay
---
drivers/gpu/drm/amd/amdkfd/Kconfig | 2 +-
1 file changed, 1 insertion(+), 1 deletion(-)
diff --git a/drivers/gpu/drm/amd/amdkfd/Kconfig
b/drivers/gpu/drm/amd/amdkfd/Kconfig
index 8dfac37..e13c67c 100644
--- a/drivers/gpu/drm/amd/amdkfd/Kconfig
+++ b/drivers/gpu
located in separate pci
bar pages, the danger of sharing registers between the drivers
is minimal.
Having said that, we are planning to move the doorbells as well
to amdgpu.
Signed-off-by: Ben Goz
Signed-off-by: Oded Gabbay
---
MAINTAINERS | 2 +
drivers
amdkfd on amdgpu and for
debugging purposes. For users who would like to run HSA on Kaveri, please
use the radeon graphic driver.
Note: CONFIG_DRM_AMDGPU_CIK must be selected for amdgpu to handle Kaveri.
Signed-off-by: Oded Gabbay
---
MAINTAINERS | 1
amdkfd
- amdgpu_amdkfd_interrupt - call the ISR of amdkfd
- amdgpu_amdkfd_suspend - suspend callback from amdgpu
- amdgpu_amdkfd_resume - resume callback from amdgpu
This patch also modifies the relevant amdgpu files, to use this new
interface.
Signed-off-by: Oded Gabbay
---
MAINT
This patch splits the KGD_ENGINE_SDMA to KGD_ENGINE_SDMA1 and
KGD_ENGINE_SDMA2 to match CZ definitions.
Signed-off-by: Oded Gabbay
---
drivers/gpu/drm/amd/include/kgd_kfd_interface.h | 3 ++-
drivers/gpu/drm/radeon/radeon_kfd.c | 3 ++-
2 files changed, 4 insertions(+), 2 deletions
bell packet type for Carrizo
Oded Gabbay (5):
drm/radeon: Modify kgd_engine_type enum to match CZ
drm/amdgpu: Add H/W agnostic amdgpu <--> amdkfd interface
drm/amdgpu: add amdgpu <--> amdkfd gfx7 interface
drm/amdkfd: Add dependency of DRM_AMDGPU to Kconfig
drm/amdkfd: Use gen
Signed-off-by: Oded Gabbay
---
drivers/gpu/drm/radeon/radeon.h | 1 -
1 file changed, 1 deletion(-)
diff --git a/drivers/gpu/drm/radeon/radeon.h b/drivers/gpu/drm/radeon/radeon.h
index 46eb0fa..1985a90 100644
--- a/drivers/gpu/drm/radeon/radeon.h
+++ b/drivers/gpu/drm/radeon/radeon.h
@@ -2456,7
From: Ben Goz
Signed-off-by: Ben Goz
Acked-by: Oded Gabbay
---
drivers/gpu/drm/amd/amdgpu/gfx_v8_0.c | 2 +-
1 file changed, 1 insertion(+), 1 deletion(-)
diff --git a/drivers/gpu/drm/amd/amdgpu/gfx_v8_0.c
b/drivers/gpu/drm/amd/amdgpu/gfx_v8_0.c
index e4aeb74..7683d7f 100644
--- a/drivers
From: Ben Goz
Signed-off-by: Ben Goz
Acked-by: Oded Gabbay
Reviewed-by: Alex Deucher
---
drivers/gpu/drm/amd/amdgpu/gfx_v7_0.c | 42 ++
drivers/gpu/drm/amd/amdgpu/gfx_v8_0.c | 47 ++
drivers/gpu/drm/amd/amdgpu/sdma_v3_0.c | 28
Hi Alex,
Here are two patches from Ben that:
- Add initializations to amdgpu that we need inside amdkfd
- Configure doorbell to maximum slots
These changes were part of the 1.0F.2 release and you even r-b the first patch.
I prefer to upstream these patches through your tree, as they do not touc
-
Ben Goz (1):
drm/amdkfd: make reset wavefronts per process per device
Oded Gabbay (3):
drm/amdkfd: add debug print to kfd_events.c
drm/amdkfd: Add missing properties to CZ device info
drm/amdkfd: remove not used defines from cik_regs.h
Valentin Rothberg (1):
drm/a
On Sat, Jun 6, 2015 at 10:09 PM, Christian König
wrote:
> On 06.06.2015 19:08, Oded Gabbay wrote:
>
> Hi Alex,
> I think you have a mistake in this patch.
> You renamed AMDGPU_GEM_CREATE_CPU_GTT_WC to AMDGPU_GEM_CREATE_CPU_GTT_USWC,
> however, AMDGPU_GEM_CREATE_CPU_GTT_WC was
Signed-off-by: Oded Gabbay
---
drivers/gpu/drm/amd/amdkfd/cik_regs.h | 177 +++---
1 file changed, 13 insertions(+), 164 deletions(-)
diff --git a/drivers/gpu/drm/amd/amdkfd/cik_regs.h
b/drivers/gpu/drm/amd/amdkfd/cik_regs.h
index 01ff332..183be5b 100644
--- a
This patch adds two missing properties initializations to the device
info structure of CZ.
As we don't have CZ support yet, it isn't critical, but its important to
fix this now instead of forgetting about it later.
Signed-off-by: Oded Gabbay
---
drivers/gpu/drm/amd/amdkfd/kfd_de
From: Ben Goz
This commit moves the reset wavefront flag to per process per device
data structure, so we can support multiple devices.
Signed-off-by: Ben Goz
Signed-off-by: Oded Gabbay
---
.../gpu/drm/amd/amdkfd/kfd_device_queue_manager.c | 7 ---
drivers/gpu/drm/amd/amdkfd/kfd_priv.h
On Fri, Jun 5, 2015 at 11:27 AM, Valentin Rothberg
wrote:
> The CONFIG_ prefix is reserved for Kconfig options in Make and CPP
> syntax. Various static analysis tools rely on this naming convention
> and check if CONFIG_ prefixed symbols are defined Kconfig. Hence add
> yet another prefix AMD_ t
Hi Alex,
I think you have a mistake in this patch.
You renamed AMDGPU_GEM_CREATE_CPU_GTT_WC to AMDGPU_GEM_CREATE_CPU_GTT_USWC,
however, AMDGPU_GEM_CREATE_CPU_GTT_WC was defined as (1 << 3) and
AMDGPU_GEM_CREATE_CPU_GTT_USWC is defined as (1 << 2)
Oded
On Wed, May 27, 2015 at 6:22 AM Alex Deucher
Hi Valentin,
Thanks for catching that.
I would be grateful if you could fix this yourself.
Oded
On Thu, Jun 4, 2015 at 4:45 PM Valentin Rothberg
wrote:
> Hi Yair,
>
> your commit fbeb661bfa89 ("drm/amdkfd: Add skeleton H/W debugger
> module support") has shown up in today's linux-next tree (i.e
Hi Dave,
One minor fix for last -rc of 4.1. The fix moves the update of an attribute in
topology to *after* it is actually initialized.
Thanks,
Oded
The following changes since commit c65b99f046843d2455aa231747b5a07a999a9f3d:
Linux 4.1-rc6 (2015-05-31 19:01:07 -0700)
are available i
Hi Dave,
Resending pull request of amdkfd for 4.2
I fixed the three things you asked me:
- rename get_dbgmgr_mutex to kfd_get_dbgmgr_mutex to namespace it
- change void* to uint64_t inside ioctl arguments
- use kmalloc instead of kzalloc because we use copy_from_user
immediately after it
Follo
Hi Dave,
Another pull request of amdkfd for 4.2
drm-amdkfd-next-2015-06-01:
- Add the H/W debugger support module, including new IOCTLs to:
- register/unregister a process as a debugged process
- Set address watch-point in the debugged process's GPU kernel
- Do a wave control operation in
orcing it to end the shader explicitly.
Signed-off-by: Ben Goz
Signed-off-by: Oded Gabbay
---
drivers/gpu/drm/amd/amdkfd/kfd_dbgdev.c| 65 ++
.../gpu/drm/amd/amdkfd/kfd_device_queue_manager.c | 8 ++-
drivers/gpu/drm/amd/amdkfd/kfd_priv.h | 7 +++
Skidanov
Signed-off-by: Oded Gabbay
---
drivers/gpu/drm/amd/include/kgd_kfd_interface.h | 8 ++
drivers/gpu/drm/radeon/cikd.h | 9 --
drivers/gpu/drm/radeon/radeon_kfd.c | 38 +++--
3 files changed, 50 insertions(+), 5 deletions
From: Yair Shachar
Signed-off-by: Yair Shachar
Signed-off-by: Oded Gabbay
---
drivers/gpu/drm/amd/amdkfd/kfd_chardev.c | 104 ++-
1 file changed, 103 insertions(+), 1 deletion(-)
diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_chardev.c
b/drivers/gpu/drm/amd/amdkfd
From: Yair Shachar
Signed-off-by: Yair Shachar
Signed-off-by: Oded Gabbay
---
drivers/gpu/drm/amd/amdkfd/kfd_chardev.c | 88 +++-
1 file changed, 87 insertions(+), 1 deletion(-)
diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_chardev.c
b/drivers/gpu/drm/amd/amdkfd
From: Yair Shachar
Signed-off-by: Yair Shachar
Signed-off-by: Oded Gabbay
---
drivers/gpu/drm/amd/amdkfd/kfd_chardev.c | 72 +++-
1 file changed, 70 insertions(+), 2 deletions(-)
diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_chardev.c
b/drivers/gpu/drm/amd/amdkfd
From: Yair Shachar
The address watch operation gives the ability to specify watch points
which will generate a shader breakpoint, based on a specified single
address or range of addresses.
There is support for read/write/any access modes.
Signed-off-by: Yair Shachar
Signed-off-by: Oded Gabbay
existing wave front(s)
Signed-off-by: Yair Shachar
Signed-off-by: Oded Gabbay
---
drivers/gpu/drm/amd/amdkfd/kfd_dbgdev.c| 405 +
drivers/gpu/drm/amd/amdkfd/kfd_dbgmgr.c| 16 +
drivers/gpu/drm/amd/amdkfd/kfd_dbgmgr.h| 5 +-
.../gpu/drm/amd
tions through the
kernel driver.
Signed-off-by: Yair Shachar
Signed-off-by: Oded Gabbay
---
drivers/gpu/drm/amd/amdkfd/Makefile| 3 +-
drivers/gpu/drm/amd/amdkfd/kfd_dbgdev.c| 142 ++
drivers/gpu/drm/amd/amdkfd/kfd_dbgdev.h| 193 ++
dr
CP to preempt queues which are currently being debugged.
The number of queues that can be designated as static is limited by the
number of HQDs (Hardware Queue Descriptors).
Signed-off-by: Yair Shachar
Signed-off-by: Oded Gabbay
---
drivers/gpu/drm/amd/amdkfd/kfd_device.c| 2 +
.../gp
e passes a
pointer to a structure (in userspace) that contains the value of the
arguments. The kernel driver is responsible to parse this structure and
validate its contents.
Signed-off-by: Yair Shachar
Signed-off-by: Oded Gabbay
---
drivers/gpu/drm/amd/amdkfd/kfd_char
From: Yair Shachar
This patch adds new interface functions to the kfd2kgd interface file. The
new functions allow to perform H/W debugger operations by writing to GPU
registers.
Signed-off-by: Yair Shachar
Signed-off-by: Oded Gabbay
---
drivers/gpu/drm/amd/include/kgd_kfd_interface.h | 13
This patch-set adds the H/W debugger module support to amdkfd.
The H/W debugger module support enables a userspace debugger, e.g gdb, to
debug GPU kernels running in HSA mode.
The available operations in this patch-set are setting watchpoints in the
GPU kernel and controlling wave-fronts.
It is
On Wed, May 20, 2015 at 4:37 AM, Joe Perches wrote:
> Use the generic mechanism to declare a bitmap instead of unsigned long.
>
> It seems that "struct kfd_process.allocated_queue_bitmap" is unused.
> Maybe it could be deleted instead.
>
> Signed-off-by: Joe Perches
> ---
> drivers/gpu/drm/amd/a
From: Alexey Skidanov
This patch fixes a bug where the number of watch points
was shown before it was actually calculated
Signed-off-by: Alexey Skidanov
Cc: stable at vger.kernel.org
Signed-off-by: Oded Gabbay
---
drivers/gpu/drm/amd/amdkfd/kfd_topology.c | 4 ++--
1 file changed, 2
t handling module
drm/amdkfd: add events IOCTL set definitions
drm/amdkfd: Add the events module
drm/amdkfd: Implement events IOCTLs
Firo Yang (1):
drm/amdkfd: Remove unessary void pointer cast
Oded Gabbay (7):
drm/amdkfd: reformat some debug prints
drm/amdk
On May 19, 2015 3:16 AM, "Dave Airlie" wrote:
>
>
>>>
>>> drm-amdkfd-next-2015-05-18:
>>>
>>> - Add the interrupts & events modules, including new IOCTLs to create
and wait
>>> on events. The HSA RT open source stack is mainly using events to
know when
>>> a dispatched work has been completed.
On Mon, May 18, 2015 at 9:29 AM, Oded Gabbay wrote:
> Hi Dave,
>
> Here is the pull request of amdkfd for 4.2
>
> drm-amdkfd-next-2015-05-18:
>
> - Add the interrupts & events modules, including new IOCTLs to create and
> wait
> on events. The HSA RT open source
events IOCTL set definitions
drm/amdkfd: Add the events module
drm/amdkfd: Implement events IOCTLs
Firo Yang (1):
drm/amdkfd: Remove unessary void pointer cast
Oded Gabbay (7):
drm/amdkfd: reformat some debug prints
drm/amdkfd: Use new struct for asic specific ops
Leaving AMD soon so need to update my email address to @gmail.com
Signed-off-by: Oded Gabbay
---
MAINTAINERS | 2 +-
1 file changed, 1 insertion(+), 1 deletion(-)
diff --git a/MAINTAINERS b/MAINTAINERS
index 781e099..d68df93 100644
--- a/MAINTAINERS
+++ b/MAINTAINERS
@@ -631,7 +631,7 @@ F
Signed-off-by: Oded Gabbay
---
drivers/gpu/drm/amd/amdkfd/kfd_module.c | 4 ++--
1 file changed, 2 insertions(+), 2 deletions(-)
diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_module.c
b/drivers/gpu/drm/amd/amdkfd/kfd_module.c
index e4fc96e..ca8410e 100644
--- a/drivers/gpu/drm/amd/amdkfd
From: Andrew Lewycky
Signed-off-by: Andrew Lewycky
Signed-off-by: Oded Gabbay
---
drivers/gpu/drm/amd/amdkfd/kfd_chardev.c | 36 +++-
1 file changed, 31 insertions(+), 5 deletions(-)
diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_chardev.c
b/drivers/gpu/drm/amd
. The HSA process isn't waiting on an event that handles
this exception.
The default behavior is not to send a SIGTERM and suffice
with a dmesg error print.
Reviewed-by: Ben Goz
Signed-off-by: Oded Gabbay
---
drivers/gpu/drm/amd/amdkfd/kfd_events.c | 12 +---
drivers/gpu/dr
From: Alexey Skidanov
Signed-off-by: Alexey Skidanov
Signed-off-by: Oded Gabbay
---
drivers/gpu/drm/amd/amdkfd/cik_event_interrupt.c | 5 -
drivers/gpu/drm/amd/amdkfd/cik_int.h | 1 +
drivers/gpu/drm/amd/amdkfd/kfd_events.c | 21 +
drivers/gpu
notifying
AMDKFD module on PPR failure.
The process triggering a PPR failure will be notified by
appropriate event or SIGTERM signal will be sent to it.
Signed-off-by: Alexey Skidanov
Signed-off-by: Oded Gabbay
---
drivers/gpu/drm/amd/amdkfd/kfd_device.c | 29 ++
drivers/gpu/drm/amd/amdkfd
IOCTL path
Reformatted debug prints in kfd_event_mmap
Map correct size (as received from mmap) in kfd_event_mmap
Signed-off-by: Andrew Lewycky
Signed-off-by: Oded Gabbay
---
drivers/gpu/drm/amd/amdkfd/Makefile | 2 +-
drivers/gpu/drm/amd/amdkfd/cik_event_interrupt.c | 63 ++
driver
- AMDKFD_IOC_WAIT_EVENTS:
Wait on event(s) until they are signaled
v2:
- Move the limit of the signal events to kfd_ioctl.h so it
can be used by userspace
- Increase limit of signal events to 4096
Signed-off-by: Andrew Lewycky
Signed-off-by: Oded Gabbay
---
drivers/gpu/drm/amd/amdkfd/kfd_chardev.c
interrupts because we have no back-pressure to the hardware.
However, only interrupts that are "wanted" by amdkfd, are copied into
the amdkfd s/w interrupt ring, in order to minimize the chances
for overflow of the ring.
Signed-off-by: Andrew Lewycky
Signed-off-by: Oded Gabbay
---
This patch adds a new interface function to the kfd->kgd interface.
The function is kgd_init_interrupts() and its function is to
initialize a pipe's interrupts.
The function currently enables the timestamp interrupt and the
bad opcode interrupt.
Signed-off-by: Oded Gabbay
---
drivers
/amdkfd: Add bad opcode exception handling
Andrew Lewycky (3):
drm/amdkfd: Add interrupt handling module
drm/amdkfd: add events IOCTL set definitions
drm/amdkfd: Implement events IOCTLs
Oded Gabbay (4):
drm/radeon: Add init interrupt kfd->kgd interface
drm/amdkfd: Add the events module
at:
git://people.freedesktop.org/~gabbayo/linux tags/drm-amdkfd-fixes-2015-05-07
for you to fetch changes up to 79b066bd76d501cfe8328142153da301f5ca11d1:
drm/amdkfd: Initialize sdma vm when creating sdma queue (2015-05-07 17:38:06
+0300)
---
On Tue, May 5, 2015 at 7:28 PM, Alex Deucher wrote:
> On Tue, May 5, 2015 at 5:03 AM, Oded Gabbay wrote:
>> Signed-off-by: Oded Gabbay
>> ---
>
> Maybe call it, init_vm() in cause you need to do something queue
> specific for other rings? Either way, this series is:
>
Signed-off-by: Oded Gabbay
---
drivers/gpu/drm/amd/amdkfd/kfd_device_queue_manager.c| 15 +--
drivers/gpu/drm/amd/amdkfd/kfd_device_queue_manager.h| 3 +++
.../gpu/drm/amd/amdkfd/kfd_device_queue_manager_cik.c| 16
drivers/gpu/drm/amd/amdkfd
This patch creates a new structure for asic specific operations, instead
of using the existing structure of operations.
This is done to make the code flow more logic, readable and maintainable.
The change is done only to the device queue manager module at this point.
Signed-off-by: Oded Gabbay
From: Xihan Zhang
This patch fixes a bug where sdma vm wasn't initialized when
an sdma queue was created in HWS mode.
This caused GPUVM faults to appear on dmesg and it is one of the
causes that SDMA queues are not working.
Signed-off-by: Xihan Zhang
Reviewed-by: Ben Goz
Signed-off-by:
This patch sets the local memory size that is reported to userspace to 0.
This is done to make sure that userspace won't try to allocate local memory
for HSA.
As long as amdkfd doesn't support allocating local memory for HSA,
we need this patch.
Signed-off-by: Oded Gabbay
---
drive
Sometimes we might unregister process that have queues, because we couldn't
preempt the queues. Until now we blocked it with BUG_ON but instead just
print it as debug.
Reviewed-by: Ben Goz
Signed-off-by: Oded Gabbay
---
drivers/gpu/drm/amd/amdkfd/kfd_device_queue_manager.c | 5 +++--
1
Signed-off-by: Oded Gabbay
---
drivers/gpu/drm/amd/amdkfd/kfd_doorbell.c | 15 +++
drivers/gpu/drm/amd/amdkfd/kfd_kernel_queue.c | 5 +++--
2 files changed, 10 insertions(+), 10 deletions(-)
diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_doorbell.c
b/drivers/gpu/drm/amd/amdkfd
On 04/23/2015 12:58 PM, Firo Yang wrote:
> kmalloc() returns a void pointer - no need to cast it in
> drivers/gpu/drm/amd/amdkfd/kfd_process.c::kfd_process_destroy_delayed()
>
> Signed-off-by: Firo Yang
> ---
> drivers/gpu/drm/amd/amdkfd/kfd_process.c | 3 +--
> 1 file changed, 1 insertion(+)
Signed-off-by: Oded Gabbay
---
drivers/gpu/drm/amd/amdkfd/kfd_module.c | 4 ++--
1 file changed, 2 insertions(+), 2 deletions(-)
diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_module.c
b/drivers/gpu/drm/amd/amdkfd/kfd_module.c
index e4fc96e..ca8410e 100644
--- a/drivers/gpu/drm/amd/amdkfd
From: Andrew Lewycky
Signed-off-by: Andrew Lewycky
Signed-off-by: Oded Gabbay
---
drivers/gpu/drm/amd/amdkfd/kfd_chardev.c | 36 +++-
1 file changed, 31 insertions(+), 5 deletions(-)
diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_chardev.c
b/drivers/gpu/drm/amd
. The HSA process isn't waiting on an event that handles this exception.
The default behavior is not to send a SIGTERM and suffice with a dmesg error
print.
Reviewed-by: Ben Goz
Signed-off-by: Oded Gabbay
---
drivers/gpu/drm/amd/amdkfd/kfd_events.c | 12 +---
drivers/gpu/drm/amd/a
From: Alexey Skidanov
Signed-off-by: Alexey Skidanov
Signed-off-by: Oded Gabbay
---
drivers/gpu/drm/amd/amdkfd/cik_event_interrupt.c | 5 -
drivers/gpu/drm/amd/amdkfd/cik_int.h | 1 +
drivers/gpu/drm/amd/amdkfd/kfd_events.c | 21 +
drivers/gpu
notifying
AMDKFD module on PPR failure.
The process triggering a PPR failure will be notified by appropriate event
or SIGTERM signal will be sent to it.
Signed-off-by: Alexey Skidanov
Signed-off-by: Oded Gabbay
---
drivers/gpu/drm/amd/amdkfd/kfd_device.c | 29 ++
drivers/gpu/drm/amd/amdkfd
on of 256 events per process, in order for the search
for SQ interrupts to be a direct lookup. In the future, this limitation will be
lifted.
Signed-off-by: Andrew Lewycky
Signed-off-by: Oded Gabbay
---
drivers/gpu/drm/amd/amdkfd/Makefile | 2 +-
drivers/gpu/drm/amd/a
- AMDKFD_IOC_WAIT_EVENTS:
Wait on event(s) until they are signaled
Signed-off-by: Andrew Lewycky
Signed-off-by: Oded Gabbay
---
drivers/gpu/drm/amd/amdkfd/kfd_chardev.c | 45 +++
include/uapi/linux/kfd_ioctl.h | 94 +++-
2 files changed, 137 insertions
interrupts because we have no back-pressure to the hardware.
However, only interrupts that are "wanted" by amdkfd, are copied into the
amdkfd s/w interrupt ring, in order to minimize the chances for overflow of the
ring.
Signed-off-by: Andrew Lewycky
Signed-off-by: Oded Gabbay
---
This patch adds a new interface function to the kfd->kgd interface.
The function is kgd_init_interrupts() and its function is to initialize a
pipe's interrupts.
The function currently enables the timestamp interrupt and the bad opcode
interrupt.
Signed-off-by: Oded Gabbay
---
drivers
--
John Stultz (1):
drm/amdkfd: Convert timestamping to use 64bit time accessors
Oded Gabbay (3):
drm/amdkfd: rename fence_wait_timeout
drm/amdkfd: Remove unused field from struct qcm_process_device
drm/amdkfd: add debug prints for process teardown
Xihan Zhang (1):
From: xihazhang
The current code can only support one kgd instance. We have to
support multiple kgd instances in one system. i.e two amdgpu or two
radeon or one amdgpu + one radeon or more than two kgd instances.
Signed-off-by: Xihan Zhang
Signed-off-by: Oded Gabbay
---
drivers/gpu/drm/amd
On 03/13/2015 01:24 PM, Arnd Bergmann wrote:
> On Thursday 12 March 2015 10:23:40 John Stultz wrote:
>> Convert the timestamping in the amdkfd driver to use
>> a timespec64 and 64bit time accessors.
>>
>> Cc: Arnd Bergmann
>> Cc: Oded Gabbay
>>
Hi Dave,
Please pull the following three bug fixes for -rc5:
- Fixing SDMA initialization when in non-HWS mode (debug mode)
- Memory leak fix when destroying kernel queue
- Fix number of available compute pipelines according to new firmware
Thanks,
Oded
The following changes since comm
On 03/12/2015 11:36 AM, Christian König wrote:
> On 12.03.2015 10:30, Oded Gabbay wrote:
>>
>> On 03/12/2015 11:23 AM, Christian König wrote:
>>> On 12.03.2015 10:02, Michel Dänzer wrote:
>>>> On 12.03.2015 06:14, Alex Deucher wrote:
>>>>
This patch updates the Kaveri MEC firmware to #396 (from #391).
The MEC firmware is mainly used for amdkfd - AMD's HSA Linux kernel driver.
Signed-off-by: Oded Gabbay
---
radeon/kaveri_mec.bin | Bin 17024 -> 17024 bytes
radeon/kaveri_mec2.bin | Bin 17024 -> 17024 bytes
2 files
s the ttm bo size to determine when to select top down
>>>>>> allocation but since the ttm bo is not initialized yet the
>>>>>> check is always false.
>>>>>>
>>>>>> Noticed-by: Oded Gabbay
>>>>>> Signed-off-by:
On 03/12/2015 11:04 AM, Michel Dänzer wrote:
> On 12.03.2015 17:48, Oded Gabbay wrote:
>> From: Ben Goz
>>
>> The current CP firmware can handle Usermode Queues only on MEC1.
>> To reflect this firmware change, this commit reduces number of compute
>> pipel
From: Ben Goz
The current CP firmware can handle Usermode Queues only on MEC1.
To reflect this firmware change, this commit reduces number of compute pipelines
to 4 - 1, from 8 - 1 (the first pipeline is allocated for kgd).
Signed-off-by: Ben Goz
Signed-off-by: Oded Gabbay
---
drivers/gpu
-by: Ben Goz
Signed-off-by: Oded Gabbay
---
drivers/gpu/drm/amd/amdkfd/kfd_device_queue_manager.c | 10 +-
1 file changed, 9 insertions(+), 1 deletion(-)
diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_device_queue_manager.c
b/drivers/gpu/drm/amd/amdkfd/kfd_device_queue_manager.c
index
From: Ben Goz
This patch adds a missing destruction of mqd, when destroying a kernel queue.
Without the destruction, there is a memory leakage when repeatedly creating and
destroying kernel queues.
Signed-off-by: Ben Goz
Signed-off-by: Oded Gabbay
---
drivers/gpu/drm/amd/amdkfd
Signed-off-by: Oded Gabbay
---
drivers/gpu/drm/amd/amdkfd/kfd_process.c | 6 ++
1 file changed, 6 insertions(+)
diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_process.c
b/drivers/gpu/drm/amd/amdkfd/kfd_process.c
index a369c14..945d622 100644
--- a/drivers/gpu/drm/amd/amdkfd/kfd_process.c
Signed-off-by: Oded Gabbay
Reviewed-by: Ben Goz
---
drivers/gpu/drm/amd/amdkfd/kfd_priv.h | 2 --
1 file changed, 2 deletions(-)
diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_priv.h
b/drivers/gpu/drm/amd/amdkfd/kfd_priv.h
index 5a44f2f..b7bd7af 100644
--- a/drivers/gpu/drm/amd/amdkfd/kfd_priv.h
fence_wait_timeout() is an exported kernel symbol, so we should rename our
local function to something different.
Signed-off-by: Oded Gabbay
---
drivers/gpu/drm/amd/amdkfd/kfd_device_queue_manager.c | 4 ++--
1 file changed, 2 insertions(+), 2 deletions(-)
diff --git a/drivers/gpu/drm/amd
On 02/24/2015 04:56 PM, Oded Gabbay wrote:
> In a nutshell:
>
> This RFC proposes a control mechanism for VRAM (GPU local memory) memory
> pinning that is initiated by HSA processes. This control mechanism is proposed
> in order to prevent starvation of graphic applications d
In a nutshell:
This RFC proposes a control mechanism for VRAM (GPU local memory) memory
pinning that is initiated by HSA processes. This control mechanism is proposed
in order to prevent starvation of graphic applications due to high VRAM usage
by HSA processes.
TOC:
--
+0200)
Oded Gabbay (2):
drm/amdkfd: Initialize only amdkfd's assigned pipelines
drm/amdkfd: don't set get_pipes_num() as inline
drivers/gpu/drm/amd/amdkfd/kfd_device_queue_manager.c | 10 --
drivers/gpu/drm/amd/amdkfd/kfd_device_queue_man
601 - 700 of 1121 matches
Mail list logo