On Fri, Apr 21, 2017 at 12:59 PM, Ville Syrjälä
wrote:
> On Fri, Apr 21, 2017 at 10:49:49AM -0400, Ilia Mirkin wrote:
>> On Fri, Apr 21, 2017 at 3:58 AM, Gerd Hoffmann wrote:
>> > While working on graphics support for virtual machines on ppc64
Depending on usage patterns, the current LRU policy may create a
non-injective mapping between userspace ring ids and kernel rings.
This behaviour is undesired as apps that attempt to fill all HW blocks
would be unable to reach some of them.
This change forces the LRU policy to create bijective
V2 updates:
* Rebased
* All patches now have r-b
Second part of the split of the series:
Add support for high priority scheduling in amdgpu v8
These patches should be close to being good enough to land.
The first two patches are simple fixes I've ported from the ROCm branch. These
still need
The MQD structure matches the reg layout. Take advantage of this to
simplify HQD programming.
Note that the ACTIVE field still needs to be programmed last.
Suggested-by: Felix Kuehling
Reviewed-by: Felix Kuehling
Signed-off-by: Andres Rodriguez
Add amdgpu_queue_mgr, a mechanism that allows disjointing usermode's
ring ids from the kernel's ring ids.
The queue manager maintains a per-file descriptor map of user ring ids
to amdgpu_ring pointers. Once a map is created it is permanent (this is
required to maintain FIFO execution guarantees
Use an LRU policy to map usermode rings to HW compute queues.
Most compute clients use one queue, and usually the first queue
available. This results in poor pipe/queue work distribution when
multiple compute apps are running. In most cases pipe 0 queue 0 is
the only queue that gets used.
In
On 17-04-21 05:21 PM, Andres Rodriguez wrote:
> Hey Felix,
>
> I wanted to confirm my understanding of your email.
>
> > Reading the updated shadow_wptr
>
> You are referring to the copy_from_user() call, correct? Just wanted
> to make sure I didn't misunderstand and you meant the write to
>
Hey Felix,
I wanted to confirm my understanding of your email.
> Reading the updated shadow_wptr
You are referring to the copy_from_user() call, correct? Just wanted to
make sure I didn't misunderstand and you meant the write to
mmCP_HQD_PQ_WPTR.
> after the doorbell has been enabled in
I spotted a potential race condition in this change: Reading the updated
shadow_wptr must happen after the doorbell has been enabled in the HQD.
Once the doorbell is enabled, the CP will capture any further
submissions to the queue. If you activate the doorbell later (in
mqd_commit), the CP would
Patches 1 and 2 are Reviewed-by: Felix Kuehling
Note that we're making another change to this programming sequence in
the upcoming release. Conceptually the MQD represents the state of an
inactive queue. So the CP_HQD_ACTIVE and DOORBELL_EN bits should be 0 in
the MQD.
Pipes provide better concurrency than queues, therefore we want to make
sure that apps use queues from different pipes whenever possible.
Optimize for the trivial case where an app will consume rings in order,
therefore we don't want adjacent rings to belong to the same pipe.
Reviewed-by: Edward
Instead of picking an arbitrary queue for KIQ, search for one according
to policy. The queue must be unused.
Also report the KIQ as an unavailable resource to KFD.
In testing I ran into KCQ initialization issues when using pipes 2/3 of
MEC2 for the KIQ. Therefore the policy disallows grabbing
Instead of taking the first pipe and giving the rest to kfd, take the
first 2 queues of each pipe.
Effectively, amdgpu and amdkfd own the same number of queues. But
because the queues are spread over multiple pipes the hardware will be
able to better handle concurrent compute workloads.
amdgpu
The assumption that we are only using the first pipe no longer holds.
Instead, calculate the queue_mask from the queue_bitmap.
Acked-by: Felix Kuehling
Signed-off-by: Andres Rodriguez
---
drivers/gpu/drm/amd/amdgpu/gfx_v8_0.c | 20
Take ownership of pipe initialization away from KFD.
Note that hpd_eop_gpu_addr was already large enough to accomodate all
pipes.
Reviewed-by: Edward O'Callaghan
Reviewed-by: Felix Kuehling
Acked-by: Christian König
The current implementation is hardcoded to enable ME1/PIPE0 interrupts
only.
This patch allows amdgpu to enable interrupts for any pipe of ME1.
v2: added gfx9 support
v3: use soc15_grbm_select for gfx9
Acked-by: Felix Kuehling
Signed-off-by: Andres Rodriguez
Update the KGD to KFD interface to allow sharing pipes with queue
granularity instead of pipe granularity.
This allows for more interesting pipe/queue splits.
v2: fix overflow check for res.queue_mask
v3: fix shift overflow when setting res.queue_mask
v4: fix comment in is_pipeline_enabled()
v5:
V4 updates:
* Rebased onto latest 4.12-wip (ba16d6c), since I got test reports that the
patch series failed to apply
V3 updates:
* Fixed kfd set resources grabbing MEC1 queues
* kfdtest now passes on Kaveri with the amdgpu or radeon driver
V2 updates:
* Fixed wrong HPD offset in
This information is already available in adev.
Reviewed-by: Edward O'Callaghan
Reviewed-by: Felix Kuehling
Acked-by: Christian König
Signed-off-by: Andres Rodriguez
---
Make amdgpu the owner of all per-pipe state of the HQDs.
This change will allow us to split the queues between kfd and amdgpu
with a queue granularity instead of pipe granularity.
This patch fixes kfd allocating an HDP_EOP region for its 3 pipes which
goes unused.
v2: support for gfx9
v3: fix
Previously the queue/pipe split with kfd operated with pipe
granularity. This patch allows amdgpu to take ownership of an arbitrary
set of queues.
It also consolidates the last few magic numbers in the compute
initialization process into mec_init.
v2: support for gfx9
v3: renamed
The return value from copy_form_user is 0 for the success case.
Reviewed-by: Felix Kuehling
Signed-off-by: Andres Rodriguez
---
drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gfx_v8.c | 2 +-
1 file changed, 1 insertion(+), 1 deletion(-)
diff --git
Rename straggler instances of r(adeon)dev to a(mdgpu)dev
Reviewed-by: Edward O'Callaghan
Reviewed-by: Felix Kuehling
Acked-by: Christian König
Signed-off-by: Andres Rodriguez
---
Handle HQD deactivation timeouts instead of ignoring them.
Reviewed-by: Edward O'Callaghan
Acked-by: Christian König
Acked-by: Felix Kuehling
Signed-off-by: Andres Rodriguez
---
The MQD programming sequence currently exists in 3 different places.
Refactor it to absorb all the duplicates.
The success path remains mostly identical except for a slightly
different order in the non-kiq case. This shouldn't matter if the HQD
is disabled.
The error handling paths have been
Rename MEC_HPD_SIZE to GFXN_MEC_HPD_SIZE to clarify it is specific to a
gfx generation.
Signed-off-by: Andres Rodriguez
---
drivers/gpu/drm/amd/amdgpu/gfx_v7_0.c | 11 +--
drivers/gpu/drm/amd/amdgpu/gfx_v8_0.c | 15 +++
The gfxv7 contains a slightly different version of cik_mqd called
bonaire_mqd. This can introduce subtle bugs if fixes are not applied in
both places.
Reviewed-by: Edward O'Callaghan
Acked-by: Christian König
Acked-by: Felix Kuehling
On 2017-04-21 12:23 PM, Mario Kleiner wrote:
Hi
These two patches were written and tested against Harry's
hwentland amd-staging-dc-drm-next branch from a few days ago.
Not sure if that is the right one to base dal/dc patches against,
but it looks recent and active enough?
Yes, I suggest you
On 17-04-21 01:01 PM, Christian König wrote:
> Am 21.04.2017 um 17:43 schrieb Felix Kuehling:
>> On 17-04-21 03:11 AM, Christian König wrote:
>>> Hi Alex,
>>>
No. For the current source code, I think the premap and no-op is not
working.
>>> Indeed, we don't set bo->mem.bus.addr in
Am 21.04.2017 um 12:05 schrieb Chunming Zhou:
Change-Id: Ie5958cf6dbdc1c8278e61d9158483472d6f5c6e3
Signed-off-by: Chunming Zhou
That patch should come earlier in the series.
---
drivers/gpu/drm/amd/amdgpu/amdgpu.h| 1 +
Am 21.04.2017 um 12:05 schrieb Chunming Zhou:
Change-Id: I64da2701c9fdcf986afb90ba1492a78d5bef1b6c
Signed-off-by: Chunming Zhou
---
drivers/gpu/drm/amd/amdgpu/amdgpu_vm.c | 60 ++
1 file changed, 60 insertions(+)
diff --git
Am 21.04.2017 um 12:05 schrieb Chunming Zhou:
ctx_alloc will check dedicated flag for its process by this flag.
Change-Id: I5f80dc39dc9d44660a96a2b710b0dbb4d3b9039d
Signed-off-by: Chunming Zhou
---
drivers/gpu/drm/amd/amdgpu/amdgpu_ctx.c | 9 --
Am 21.04.2017 um 12:05 schrieb Chunming Zhou:
Change-Id: I652d89200c38a238ff50092107fa1ee31317728a
Signed-off-by: Chunming Zhou
Reviewed-by: Christian König for this one.
---
drivers/gpu/drm/amd/amdgpu/amdgpu_ctx.c | 9 ++---
1 file
Am 21.04.2017 um 18:27 schrieb Deucher, Alexander:
-Original Message-
From: amd-gfx [mailto:amd-gfx-boun...@lists.freedesktop.org] On Behalf
Of Chunming Zhou
Sent: Friday, April 21, 2017 6:06 AM
To: amd-gfx@lists.freedesktop.org
Cc: Zhou, David(ChunMing)
Subject: [PATCH 0/5] ***
__setup doesn't work in modules.
s8250_options is only compiled if the driver is not a module.
I came up with a different way to parse another module's options. But I
can't get the commandline that I would parse in that way:
@@ -787,12 +793,45 @@ long amdgpu_drm_ioctl(struct file *filp,
On Fri, Apr 21, 2017 at 10:49:49AM -0400, Ilia Mirkin wrote:
> On Fri, Apr 21, 2017 at 3:58 AM, Gerd Hoffmann wrote:
> > While working on graphics support for virtual machines on ppc64 (which
> > exists in both little and big endian variants) I've figured the comments
> > for
> -Original Message-
> From: amd-gfx [mailto:amd-gfx-boun...@lists.freedesktop.org] On Behalf
> Of Chunming Zhou
> Sent: Friday, April 21, 2017 6:06 AM
> To: amd-gfx@lists.freedesktop.org
> Cc: Zhou, David(ChunMing)
> Subject: [PATCH 0/5] *** Dedicated vmid per process ***
>
> The current
Hi
These two patches were written and tested against Harry's
hwentland amd-staging-dc-drm-next branch from a few days ago.
Not sure if that is the right one to base dal/dc patches against,
but it looks recent and active enough?
They fix pageflip scheduling to always flip at the right target
Make sure we do not program a hw pageflip inside vblank 'n' iff the
atomic flip is comitted while inside the same vblank 'n'. We must
defer such a flip by one refresh cycle to vblank 'n+1'.
Without this, pageflips programmed via X11 GLX_OML_sync_control extensions
glXSwapBuffersMscOML(...,
Since DC now uses CRTC_VERTICAL_INTERRUPT0 as VBLANK irq trigger
and vblank interrupts actually happen earliest at start of vblank,
instead of a bit before vblank, we no longer need some of the
fudging logic to deal with too early vblank irq handling (grep for
lb_vblank_lead_lines). This itself
> -Original Message-
> From: amd-gfx [mailto:amd-gfx-boun...@lists.freedesktop.org] On Behalf
> Of Roger.He
> Sent: Friday, April 21, 2017 3:07 AM
> To: amd-gfx@lists.freedesktop.org
> Cc: Zhou, David(ChunMing); He, Hongbo
> Subject: [PATCH 1/3] drm/amdgpu: fix indent
>
> Change-Id:
On 17-04-21 03:11 AM, Christian König wrote:
> Hi Alex,
>
>> No. For the current source code, I think the premap and no-op is not
>> working.
>>
> Indeed, we don't set bo->mem.bus.addr in amdgpu_ttm_io_mem_reserve()
> any more. Felix will probably want to fix that for the KFD branch.
I vaguely
Thanks, Christian for adding me.
On 2017-04-21 09:27 AM, Christian König wrote:
Adding Harry to this mail thread as well, cause is one of the people
really affected by this.
Christian.
Am 21.04.2017 um 15:21 schrieb Christian König:
Am 21.04.2017 um 15:12 schrieb Gerd Hoffmann:
Hi,
On Fri, Apr 21, 2017 at 3:58 AM, Gerd Hoffmann wrote:
> While working on graphics support for virtual machines on ppc64 (which
> exists in both little and big endian variants) I've figured the comments
> for various drm fourcc formats in the header file don't match reality.
>
>
On Fri, Apr 21, 2017 at 4:43 AM, Chunming Zhou wrote:
> v2: address Alex's comment, add AMDGPU_DEFAULT_GTT_SIZE_MB.
>
> Change-Id: I0417651a8e6fd23b28f9b30a110a196c332e2d03
> Signed-off-by: Chunming Zhou
Reviewed-by: Alex Deucher
Adding Harry to this mail thread as well, cause is one of the people
really affected by this.
Christian.
Am 21.04.2017 um 15:21 schrieb Christian König:
Am 21.04.2017 um 15:12 schrieb Gerd Hoffmann:
Hi,
"native" to me feels more like "native to the GPU" since these things
really are
Am 21.04.2017 um 15:12 schrieb Gerd Hoffmann:
Hi,
"native" to me feels more like "native to the GPU" since these things
really are tied to the GPU not the CPU. That's also why I went with the
explicit endianness originally so that the driver could properly declare
what the GPU supports.
Hi,
> > "native" to me feels more like "native to the GPU" since these things
> > really are tied to the GPU not the CPU. That's also why I went with the
> > explicit endianness originally so that the driver could properly declare
> > what the GPU supports.
> And to be honest I would really
Am 21.04.2017 um 13:49 schrieb Ville Syrjälä:
On Fri, Apr 21, 2017 at 02:40:18PM +0300, Pekka Paalanen wrote:
On Fri, 21 Apr 2017 14:08:04 +0300
Ville Syrjälä wrote:
On Fri, Apr 21, 2017 at 11:50:18AM +0200, Gerd Hoffmann wrote:
On Fr, 2017-04-21 at 12:25
Thank you!
-Original Message-
From: Christian König [mailto:deathsim...@vodafone.de]
Sent: Friday, April 21, 2017 7:47 PM
To: He, Hongbo; amd-gfx@lists.freedesktop.org
Cc: Zhou, David(ChunMing)
Subject: Re: [PATCH 3/3] drm/amdgpu: fix gpu reset issue
See amdgpu_gpu_reset():
>
On Fri, Apr 21, 2017 at 02:40:18PM +0300, Pekka Paalanen wrote:
> On Fri, 21 Apr 2017 14:08:04 +0300
> Ville Syrjälä wrote:
>
> > On Fri, Apr 21, 2017 at 11:50:18AM +0200, Gerd Hoffmann wrote:
> > > On Fr, 2017-04-21 at 12:25 +0300, Ville Syrjälä wrote:
> > > >
Am 21.04.2017 um 12:00 schrieb Chunming Zhou:
[ 132.036658] amdgpu :22:00.0: VM IB without ID
[ 132.036709] [drm:amdgpu_job_run [amdgpu]] *ERROR* Error scheduling IBs (-22)
[ 132.036755] [drm:amd_sched_main [amdgpu]] *ERROR* Failed to run job!
root cause is fence is signaled during sync
See amdgpu_gpu_reset():
/* block scheduler */
for (i = 0; i < AMDGPU_MAX_RINGS; ++i) {
struct amdgpu_ring *ring = adev->rings[i];
if (!ring)
continue;
kthread_park(ring->sched.thread);
On Fri, 21 Apr 2017 14:08:04 +0300
Ville Syrjälä wrote:
> On Fri, Apr 21, 2017 at 11:50:18AM +0200, Gerd Hoffmann wrote:
> > On Fr, 2017-04-21 at 12:25 +0300, Ville Syrjälä wrote:
> > > On Fri, Apr 21, 2017 at 09:58:24AM +0200, Gerd Hoffmann wrote:
> > > >
Hi,
> > I personally find "native" more intuitive, but at the end of the day I
> > don't mind much. If people prefer "host" over "native" I'll change it.
>
> "native" to me feels more like "native to the GPU" since these things
> really are tied to the GPU not the CPU.
Ok, then maybe "host"
On Fri, Apr 21, 2017 at 11:50:18AM +0200, Gerd Hoffmann wrote:
> On Fr, 2017-04-21 at 12:25 +0300, Ville Syrjälä wrote:
> > On Fri, Apr 21, 2017 at 09:58:24AM +0200, Gerd Hoffmann wrote:
> > > While working on graphics support for virtual machines on ppc64 (which
> > > exists in both little and
On Fri, Apr 21, 2017 at 11:50:18AM +0200, Gerd Hoffmann wrote:
> On Fr, 2017-04-21 at 12:25 +0300, Ville Syrjälä wrote:
> > On Fri, Apr 21, 2017 at 09:58:24AM +0200, Gerd Hoffmann wrote:
> > > While working on graphics support for virtual machines on ppc64 (which
> > > exists in both little and
Change-Id: Ie5958cf6dbdc1c8278e61d9158483472d6f5c6e3
Signed-off-by: Chunming Zhou
---
drivers/gpu/drm/amd/amdgpu/amdgpu.h| 1 +
drivers/gpu/drm/amd/amdgpu/amdgpu_device.c | 1 +
drivers/gpu/drm/amd/amdgpu/amdgpu_vm.c | 10 ++
3 files changed, 12
Change-Id: Id728e20366c8a1ae90d4e901dc80e136e2a613bb
Signed-off-by: Chunming Zhou
---
drivers/gpu/drm/amd/amdgpu/amdgpu_vm.c | 17 -
drivers/gpu/drm/amd/amdgpu/amdgpu_vm.h | 2 ++
2 files changed, 18 insertions(+), 1 deletion(-)
diff --git
ctx_alloc will check dedicated flag for its process by this flag.
Change-Id: I5f80dc39dc9d44660a96a2b710b0dbb4d3b9039d
Signed-off-by: Chunming Zhou
---
drivers/gpu/drm/amd/amdgpu/amdgpu_ctx.c | 9 --
drivers/gpu/drm/amd/amdgpu/amdgpu_vm.c | 50
The current kernel implementation, which grabs the idle VMID from pool when
emitting the job may:
The back-to-back submission from one process could use different VMID.
The submission to different queues from single process could use different
VMID
It works well in most case but cannot
Change-Id: I64da2701c9fdcf986afb90ba1492a78d5bef1b6c
Signed-off-by: Chunming Zhou
---
drivers/gpu/drm/amd/amdgpu/amdgpu_vm.c | 60 ++
1 file changed, 60 insertions(+)
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_vm.c
Change-Id: I652d89200c38a238ff50092107fa1ee31317728a
Signed-off-by: Chunming Zhou
---
drivers/gpu/drm/amd/amdgpu/amdgpu_ctx.c | 9 ++---
1 file changed, 6 insertions(+), 3 deletions(-)
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_ctx.c
[ 132.036658] amdgpu :22:00.0: VM IB without ID
[ 132.036709] [drm:amdgpu_job_run [amdgpu]] *ERROR* Error scheduling IBs (-22)
[ 132.036755] [drm:amd_sched_main [amdgpu]] *ERROR* Failed to run job!
root cause is fence is signaled during sync transfer.
Change-Id:
> During GPU reset, IOCTL still submit command to GPU.
No it won't. We disable the GPU scheduler during GPU reset, so that isn't an
issue.
Could you point out the related code? Thanks.
Thanks
Roger(Hongbo.He)
-Original Message-
From: Christian König
On Fr, 2017-04-21 at 12:25 +0300, Ville Syrjälä wrote:
> On Fri, Apr 21, 2017 at 09:58:24AM +0200, Gerd Hoffmann wrote:
> > While working on graphics support for virtual machines on ppc64 (which
> > exists in both little and big endian variants) I've figured the comments
> > for various drm fourcc
On Fri, Apr 21, 2017 at 11:38:28AM +0200, Gerd Hoffmann wrote:
> Hi,
>
> > > Leaving the yuv formats as-is. I have no idea if and how those are used
> > > on bigendian machines.
>
> > just an idea - since we are not sure how the remaining formats are being
> > used, should those be marked
Hi,
> > Leaving the yuv formats as-is. I have no idea if and how those are used
> > on bigendian machines.
> just an idea - since we are not sure how the remaining formats are being
> used, should those be marked somehow uncertain whether they are little
> or native endian?
ATM the yuv don't
On Fri, Apr 21, 2017 at 09:58:24AM +0200, Gerd Hoffmann wrote:
> While working on graphics support for virtual machines on ppc64 (which
> exists in both little and big endian variants) I've figured the comments
> for various drm fourcc formats in the header file don't match reality.
>
> Comments
Am 21.04.2017 um 10:43 schrieb Chunming Zhou:
v2: address Alex's comment, add AMDGPU_DEFAULT_GTT_SIZE_MB.
Change-Id: I0417651a8e6fd23b28f9b30a110a196c332e2d03
Signed-off-by: Chunming Zhou
Reviewed-by: Christian König
---
Hi Christian:
During GPU reset, IOCTL still submit command to GPU.
That will result in lots of strange problems when we test this feature.
I have no better idea to handle this case. At that time, I have tried with
return error directly rather than waiting GPU reset completion.
But that
Hi Roger,
Could you try to move 'id->current_gpu_reset_count =
atomic_read(>gpu_reset_counter);' in amdgpu_vm_grab_id to
no_flush_needed label.
Regards,
David Zhou
On 2017年04月21日 16:27, Christian König wrote:
NAK, that is exactly what we wanted to avoid.
We used to have an exclusive lock
v2: address Alex's comment, add AMDGPU_DEFAULT_GTT_SIZE_MB.
Change-Id: I0417651a8e6fd23b28f9b30a110a196c332e2d03
Signed-off-by: Chunming Zhou
---
drivers/gpu/drm/amd/amdgpu/amdgpu.h | 1 +
drivers/gpu/drm/amd/amdgpu/gmc_v6_0.c | 3 ++-
On 2017年04月20日 20:58, Christian König wrote:
Am 20.04.2017 um 14:43 schrieb Alex Deucher:
On Wed, Apr 19, 2017 at 11:47 PM, Chunming Zhou
wrote:
Change-Id: Icdf2486a2d1116e71dc8958cda679a4a83838489
Signed-off-by: Chunming Zhou
---
NAK, that is exactly what we wanted to avoid.
We used to have an exclusive lock for this and it cause a whole bunch of
problems.
Please elaborate why that should be necessary.
Regards,
Christian.
Am 21.04.2017 um 09:08 schrieb Roger.He:
Change-Id: Ib77d33a09f348ebf2e3a9d7861411f4b951ebf7c
On 21.04.2017 10:07, Christian König wrote:
From: Christian König
We need to check if the VM is swapped out before trying to update it.
Signed-off-by: Christian König
Oops. That makes a lot of sense...
Fixes: 23e0563e48f7 ("drm/amdgpu:
From: Christian König
We need to check if the VM is swapped out before trying to update it.
Signed-off-by: Christian König
---
drivers/gpu/drm/amd/amdgpu/amdgpu_gem.c | 68 ++---
1 file changed, 37 insertions(+),
On Fri, 21 Apr 2017 09:58:24 +0200
Gerd Hoffmann wrote:
> While working on graphics support for virtual machines on ppc64 (which
> exists in both little and big endian variants) I've figured the comments
> for various drm fourcc formats in the header file don't match reality.
While working on graphics support for virtual machines on ppc64 (which
exists in both little and big endian variants) I've figured the comments
for various drm fourcc formats in the header file don't match reality.
Comments says the RGB formats are little endian, but in practice they
are native
Change-Id: Ib77d33a09f348ebf2e3a9d7861411f4b951ebf7c
Signed-off-by: Roger.He
---
drivers/gpu/drm/amd/amdgpu/amdgpu.h| 6 ++
drivers/gpu/drm/amd/amdgpu/amdgpu_device.c | 16 +++-
drivers/gpu/drm/amd/amdgpu/amdgpu_drv.c| 4
Change-Id: I3a81680648fbaca74662c85709c864fa3a4a75bd
Signed-off-by: Roger.He
---
drivers/gpu/drm/amd/amdgpu/amdgpu_device.c | 34 +++---
1 file changed, 17 insertions(+), 17 deletions(-)
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c
Change-Id: Id925f4e241c4192127880d2017fbf2979aa09fc7
Signed-off-by: Roger.He
---
drivers/gpu/drm/amd/amdgpu/amdgpu_device.c | 46 ++
1 file changed, 46 insertions(+)
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c
On 2017年04月21日 15:08, Roger.He wrote:
Change-Id: Ib77d33a09f348ebf2e3a9d7861411f4b951ebf7c
Signed-off-by: Roger.He
Please add more explaination in comment, like "prevent ioctl during gpu
reset, otherwise many un-expected behaviour happen."
Regards,
David Zhou
---
On 2017年04月21日 15:07, Roger.He wrote:
Change-Id: Id925f4e241c4192127880d2017fbf2979aa09fc7
Signed-off-by: Roger.He
---
drivers/gpu/drm/amd/amdgpu/amdgpu_device.c | 46 ++
1 file changed, 46 insertions(+)
diff --git
Hi Alex,
No. For the current source code, I think the premap and no-op is not
working.
Indeed, we don't set bo->mem.bus.addr in amdgpu_ttm_io_mem_reserve() any
more. Felix will probably want to fix that for the KFD branch.
Anyway, as I said not unmapping the page tables is harmless
85 matches
Mail list logo