Am 11.11.23 um 14:11 schrieb Rajneesh Bhardwaj:
To maximize per cpu execution context for the work items, use the
recommended settings i.e. WQ_DFL_ACTIVE(256). There is no apparent
reason to throttle to 16 while process tear down.
Well big NAK to this. During process tear down it can be that hu
Am 10.11.23 um 16:07 schrieb Lazar, Lijo:
On 11/10/2023 8:18 PM, Christian König wrote:
Am 09.11.23 um 08:38 schrieb Lijo Lazar:
cancel_work is not backported to all custom kernels.
Well this is pretty clear NAK to pushing this upstream. We absolutely
can't add workaround for
AGP
address setup into amdgpu_bo_gpu_offset_no_check().
v2: check mem_type before checking agp
Fixes: 67318cb84341 ("drm/amdgpu/gmc11: set gart placement GC11")
Reported-by: Jesse Zhang
Reported-by: Yifan Zhang
Signed-off-by: Alex Deucher
Cc: christian.koe...@amd.com
Reviewed-by: Ch
Am 10.11.23 um 15:47 schrieb Alex Deucher:
This worked by luck if the GART aperture ended up at 0. When
we ended up moving GART on some chips, the GART aperture ended
up offsetting the the AGP address since the resource->start is
a GART offset, not an MC address. Fix this by moving the AGP
addr
Am 09.11.23 um 08:38 schrieb Lijo Lazar:
cancel_work is not backported to all custom kernels.
Well this is pretty clear NAK to pushing this upstream. We absolutely
can't add workaround for older kernels.
You could keep this in the backported kernel, but why should cancel_work
not be availab
Just call amdgpu_gmc_agp_addr() and check the return value for !=
AMDGPU_BO_INVALID_OFFSET;
The problem is simply that we can't cache that result anywhere because
bo->resource->start is essentially the offset into the GART and not the
MC address.
That must have been sneaked in years ago when
No, that's broken as well.
The problem is in amdgpu_ttm_alloc_gart():
if (addr != AMDGPU_BO_INVALID_OFFSET) {
bo->resource->start = addr >> PAGE_SHIFT;
return 0;
}
bo->resource->start is relative to the GART address, so we can't assign
the AGP ad
Am 10.11.23 um 13:52 schrieb Yifan Zhang:
For BOs in AGP aperture, tbo.resource->start includes AGP aperture start.
Well big NAK to that. tbo.resource->start should never ever include the
AGP aperture start in the first place.
How did that happen?
Regards,
Christian.
Don't add it again i
Am 09.11.23 um 15:41 schrieb Alex Deucher:
We've had misc reports of random IOMMU page faults when
this is used. It's just a rarely used optimization anyway, so
let's just disable it.
Signed-off-by: Alex Deucher
Acked-by: Christian König for the series.
---
drivers/gpu
ated after scheduler init, so
change the ordering to fix this.
v2: Leave logic in UVD and VCE code
Fixes: 56e449603f0ac5 ("drm/sched: Convert the GPU scheduler to variable number of
run-queues")
Signed-off-by: Alex Deucher
Cc: ltuiko...@gmail.com
Reviewed-by: Christian König
---
We should not leak the pointer where we couldn't grab the reference
on to the caller because it can be that the error handling still
tries to put the reference then.
Signed-off-by: Christian König
---
drivers/gpu/drm/amd/amdgpu/amdgpu_bo_list.c | 1 +
1 file changed, 1 insertion(+)
diff
Otherwise userspace can spam the logs by using incorrect input values.
Signed-off-by: Christian König
---
drivers/gpu/drm/amd/amdgpu/amdgpu_cs.c | 2 +-
1 file changed, 1 insertion(+), 1 deletion(-)
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_cs.c
b/drivers/gpu/drm/amd/amdgpu/amdgpu_cs.c
-by: Christian König
---
drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd.h| 9 ++-
.../gpu/drm/amd/amdgpu/amdgpu_amdkfd_gpuvm.c | 64 +--
drivers/gpu/drm/amd/amdkfd/kfd_chardev.c | 15 ++---
3 files changed, 52 insertions(+), 36 deletions(-)
diff --git a/drivers/gpu/drm/
Am 08.11.23 um 22:23 schrieb Felix Kuehling:
On 2023-11-08 07:28, Christian König wrote:
Not necessary objections to this patch here, but rather how this new
state is used later on.
The fundamental problem is that re-validating things in
amdgpu_vm_handle_moved() won't work in all
such as GFXIP9.4.3.
Acked-by: Felix Kuehling
Signed-off-by: Rajneesh Bhardwaj
Reviewed-by: Christian König
Going to push this to drm-misc-next.
Thanks,
Christian.
---
Changes in v2:
- Absorbed the feedback provided by Christian in the commit message and
the comment.
drivers/gpu/drm
Not necessary objections to this patch here, but rather how this new
state is used later on.
The fundamental problem is that re-validating things in
amdgpu_vm_handle_moved() won't work in all cases.
The general approach for both classic CS IOCTL as well as user queues is
the following:
1.
Am 07.11.23 um 17:58 schrieb Felix Kuehling:
mem = bo->tbo.resource may be NULL in amdgpu_vm_bo_update.
Fixes: 180253782038 ("drm/ttm: stop allocating dummy resources during BO
creation")
Signed-off-by: Felix Kuehling
Reviewed-by: Christian König
---
drivers/gpu/d
Am 07.11.23 um 20:45 schrieb Rajneesh Bhardwaj:
When a TTM BO is getting freed, to optimize the clearing operation on
the workqueue, schedule it closer to a NUMA node where the memory was
allocated. This avoids the cases where the ttm_bo_delayed_delete gets
scheduled on the CPU cores that are acr
g/archives/amd-gfx/2023-November/100966.html
Signed-off-by: Alex Deucher
Cc: Mikhail Gavrilov
Reviewed-by: Christian König
---
drivers/gpu/drm/amd/amdgpu/amdgpu_object.c | 3 ---
drivers/gpu/drm/amd/amdgpu/gmc_v10_0.c | 1 +
drivers/gpu/drm/amd/amdgpu/gmc_v11_0.c | 1 +
drivers/g
Am 07.11.23 um 16:14 schrieb Alex Deucher:
On Tue, Nov 7, 2023 at 9:52 AM Christian König
wrote:
This is only needed for certain UVD/VCE hardware/firmware versions.
We currently enable it for all UVD and VCE versions.
Leo needs to decide that, but I would rather like to keep the call to
wrote:
On Tue, Nov 7, 2023 at 5:52 AM Christian König
wrote:
Am 03.11.23 um 23:10 schrieb Alex Deucher:
On Fri, Nov 3, 2023 at 4:17 PM Alex Deucher wrote:
On Thu, Oct 26, 2023 at 4:17 PM Luben Tuikov wrote:
Pushed to drm-misc-next.
BTW, I'm seeing the following on older GPUs with VC
0-26 15:52, Luben Tuikov wrote:
On 2023-10-26 15:32, Alex Deucher wrote:
On Thu, Oct 26, 2023 at 2:22 AM Christian König
wrote:
Am 25.10.23 um 19:19 schrieb Alex Deucher:
Rather than doing this in the IP code for the SDMA paging
engine, move it up to the core device level init level.
This
Am 06.11.23 um 16:47 schrieb Tatsuyuki Ishi:
On Nov 6, 2023, at 22:44, Christian König
wrote:
Am 02.11.23 um 15:04 schrieb Tatsuyuki Ishi:
In Vulkan, it is the application's responsibility to perform adequate
synchronization before a sparse unmap, replace or BO destroy operation.
Unti
Am 02.11.23 um 15:04 schrieb Tatsuyuki Ishi:
In Vulkan, it is the application's responsibility to perform adequate
synchronization before a sparse unmap, replace or BO destroy operation.
Until now, the kernel applied the same rule as implicitly-synchronized
APIs like OpenGL, which with per-VM BO
Am 06.11.23 um 08:56 schrieb Tatsuyuki Ishi:
On Oct 31, 2023, at 23:07, Christian König
wrote:
Am 31.10.23 um 14:59 schrieb Bas Nieuwenhuizen:
On Tue, Oct 31, 2023 at 2:57 PM Christian König
wrote:
Am 31.10.23 um 14:40 schrieb Tatsuyuki Ishi:
> The curr
Am 01.11.23 um 23:00 schrieb Felix Kuehling:
On 2023-10-31 11:18, Alex Deucher wrote:
On Tue, Oct 31, 2023 at 11:12 AM Christian König
wrote:
When clearing the root PD fails we need to properly release it again.
Signed-off-by: Christian König
Acked-by: Alex Deucher
Has this been submitted
Am 01.11.23 um 20:13 schrieb Arunpravin Paneer Selvam:
Hi Christian,
On 10/30/2023 9:34 PM, Christian König wrote:
Am 30.10.23 um 13:22 schrieb Arunpravin Paneer Selvam:
If the size returned by drm buddy allocator is higher than
the required size, we take the higher size to calculate
the
Am 01.11.23 um 17:26 schrieb Arunpravin Paneer Selvam:
Replace seq64 bo lock sequences with drm_exec.
Signed-off-by: Alex Deucher
Reviewed-by: Christian König
---
drivers/gpu/drm/amd/amdgpu/amdgpu_seq64.c | 73 ++-
1 file changed, 33 insertions(+), 40 deletions
s in VRAM")
Signed-off-by: Alex Deucher
Cc: alexey.kli...@linaro.org
Acked-by: Christian König
---
drivers/gpu/drm/amd/amdgpu/amdgpu_gfx.c | 2 ++
1 file changed, 2 insertions(+)
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_gfx.c
b/drivers/gpu/drm/amd/amdgpu/amdgpu_gfx.c
index c92
needs to be mapped read only in the user VM (Alex)
v2:
- Instead of just one define for TOP/BOTTOM
reserved space separate them into two (Christian)
- Fix the CPU and VA calculations and while at it
also cleanup error handling and kerneldoc (Christian)
Signed-off-by: Christian König
Am 02.11.23 um 03:36 schrieb Lang Yu:
On 10/31/ , Christian König wrote:
Am 31.10.23 um 14:59 schrieb Bas Nieuwenhuizen:
On Tue, Oct 31, 2023 at 2:57 PM Christian König
wrote:
Am 31.10.23 um 14:40 schrieb Tatsuyuki Ishi:
> The current amdgpu_gem_va_update_vm only tries to perf
When clearing the root PD fails we need to properly release it again.
Signed-off-by: Christian König
---
drivers/gpu/drm/amd/amdgpu/amdgpu_vm.c | 31 +-
1 file changed, 16 insertions(+), 15 deletions(-)
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_vm.c
b/drivers/gpu
Am 31.10.23 um 15:39 schrieb Tatsuyuki Ishi:
On Oct 31, 2023, at 22:55, Christian König wrote:
Am 31.10.23 um 14:40 schrieb Tatsuyuki Ishi:
In short, eviction never really belonged to the vm_status state machine.
I strongly disagree to that.
Even when evicted, the BO could belong to
Am 31.10.23 um 15:14 schrieb Michel Dänzer:
On 10/31/23 14:40, Tatsuyuki Ishi wrote:
In Vulkan, it is the application's responsibility to perform adequate
synchronization before a sparse unmap, replace or BO destroy operation.
Until now, the kernel applied the same rule as implicitly-synchroni
Am 31.10.23 um 14:59 schrieb Bas Nieuwenhuizen:
On Tue, Oct 31, 2023 at 2:57 PM Christian König
wrote:
Am 31.10.23 um 14:40 schrieb Tatsuyuki Ishi:
> The current amdgpu_gem_va_update_vm only tries to perform
updates for the
> BO specified in the GEM ioctl; however,
Am 31.10.23 um 14:40 schrieb Tatsuyuki Ishi:
All the state changes are handled in the TTM move callback; doing it again
here just leads to more confusion.
The state move here is because we need to track which PDs/PTs are
already validated and which have new locations reflected in the PDEs.
W
Am 31.10.23 um 14:40 schrieb Tatsuyuki Ishi:
The current amdgpu_gem_va_update_vm only tries to perform updates for the
BO specified in the GEM ioctl; however, when a binding is split, the
adjacent bindings also need to be updated. Such updates currently ends up
getting deferred until next submiss
Am 31.10.23 um 14:40 schrieb Tatsuyuki Ishi:
In short, eviction never really belonged to the vm_status state machine.
I strongly disagree to that.
Even when evicted, the BO could belong to either the moved or done state.
The "evicted" state needed to handle both cases, causing greater confusi
Hi Alexey,
trying to answer some of the questions since Alex is currently on vacation.
Am 30.10.23 um 17:01 schrieb Alexey Klimov:
Hi Alex,
On Thu, 26 Oct 2023 at 19:53, Alex Deucher wrote:
On Thu, Oct 26, 2023 at 1:33 PM Alexey Klimov wrote:
#regzbot introduced: 1cfb4d612127
#regzbot titl
Am 31.10.23 um 03:55 schrieb YuanShang:
The variable "chunk_ptr" should be a pointer pointing
to a struct drm_amdgpu_cs_chunk instead of to a pointer
of that.
Signed-off-by: YuanShang
Good catch, Reviewed-by: Christian König
---
drivers/gpu/drm/amd/amdgpu/amdgpu_cs.c | 2 +
Am 30.10.23 um 18:56 schrieb Felix Kuehling:
On 2023-10-30 13:48, Christian König wrote:
Am 30.10.23 um 18:38 schrieb Felix Kuehling:
On 2023-10-30 12:16, Christian König wrote:
@@ -1904,6 +1906,19 @@ kfd_process_gpuid_from_node(struct
kfd_process *p, struct kfd_node *node,
return
Am 30.10.23 um 18:38 schrieb Felix Kuehling:
On 2023-10-30 12:16, Christian König wrote:
@@ -1904,6 +1906,19 @@ kfd_process_gpuid_from_node(struct
kfd_process *p, struct kfd_node *node,
return -EINVAL;
}
+static void signal_eviction_fence(struct kfd_process *p)
+{
+ spin_lock
Am 30.10.23 um 16:16 schrieb Felix Kuehling:
On 2023-10-30 4:23, Christian König wrote:
Am 28.10.23 um 00:39 schrieb Felix Kuehling:
Make restore workers freezable so we don't have to explicitly flush
them
in suspend and GPU reset code paths, and we don't accidentally try to
restore
d-off-by: Arunpravin Paneer Selvam
Acked-by: Christian König
IIRC that hack with the start address is actually not needed any more,
but we need to double check this.
Christian.
---
drivers/gpu/drm/amd/amdgpu/amdgpu_vram_mgr.c | 15 +--
1 file changed, 13 insertions(+), 2 deleti
Am 30.10.23 um 14:38 schrieb Rob Clark:
On Mon, Oct 30, 2023 at 1:05 AM Christian König
wrote:
Am 27.10.23 um 18:58 schrieb Rob Clark:
From: Rob Clark
In cases where the # is known ahead of time, it is silly to do the table
resize dance.
Ah, yes that was my initial implementation as well
Am 28.10.23 um 02:48 schrieb Yuran Pereira:
Hello,
On Fri, Oct 27, 2023 at 11:57:45AM -0400, Hamza Mahfooz wrote:
On 10/27/23 11:55, Lakha, Bhawanpreet wrote:
[AMD Official Use Only - General]
There was a consensus to use memset instead of {0}. I remember making
changes related to that previ
Am 28.10.23 um 00:39 schrieb Felix Kuehling:
Make restore workers freezable so we don't have to explicitly flush them
in suspend and GPU reset code paths, and we don't accidentally try to
restore BOs while the GPU is suspended. Not having to flush restore_work
also helps avoid lock/fence dependen
+ exec->objects = kmalloc(sz, GFP_KERNEL);
Please use k*v*malloc() here since we can't predict how large that will be.
With that fixed the patch is Reviewed-by: Christian König
.
Regards,
Christian.
/* If allocation here fails, just delay that till the first use
Am 27.10.23 um 22:41 schrieb Alex Deucher:
For pptable structs that use flexible array sizes, use flexible arrays.
Link: https://bugs.launchpad.net/ubuntu/+source/linux/+bug/2039926
Signed-off-by: Alex Deucher
Acked-by: Christian König
---
.../drm/amd/pm/powerplay/hwmgr/pptable_v1_0.h
Am 26.10.23 um 20:56 schrieb Alex Deucher:
Since they were moved to VRAM, we need to use the IO
variants of memcpy.
Fixes: 1cfb4d612127 ("drm/amdgpu: put MQDs in VRAM")
Signed-off-by: Alex Deucher
Reviewed-by: Christian König
---
drivers/gpu/drm/amd/amdgpu/gfx_v1
Am 26.10.23 um 21:32 schrieb Alex Deucher:
On Thu, Oct 26, 2023 at 1:45 PM Luben Tuikov wrote:
Update the GPU Scheduler maintainer email.
Cc: Alex Deucher
Cc: Christian König
Cc: Daniel Vetter
Cc: Dave Airlie
Cc: AMD Graphics
Cc: Direct Rendering Infrastructure - Development
Signed-off
veloped-by: Umio Yasuno
Signed-off-by: Umio Yasuno
Reviewed-by: Christian König for the series.
---
This thread has been inactive for nearly 5 months, so I re-created the patch.
drivers/gpu/drm/amd/amdgpu/amdgpu_fdinfo.c | 3 ---
1 file changed, 3 deletions(-)
diff --git a/drivers/gpu/drm/
remove hardware queue, queue id = 3
Fixes: d0c860f33553 ("drm/amdgpu: rework lock handling for flush_tlb v2")
Reported-by: Li Ma
Signed-off-by: Yifan Zhang
Reviewed-by: Christian König
Thanks for looking into this,
Christian.
---
drivers/gpu/drm/amd/amdgpu/amdgpu_device.c |
Am 10.10.23 um 12:17 schrieb Shashank Sharma:
This patch adds support for userqueue eviction fences. In general, when
a process wants to map VRAM memory but TTM can't find enough space, it
attempts to evict BOs from its LRU list. This fence will prevent the TTM
manager from evicting the process's
Am 26.10.23 um 12:24 schrieb Arunpravin Paneer Selvam:
Temporarily remove the seq64 mapping sequence.
Signed-off-by: Arunpravin Paneer Selvam
Reviewed-by: Christian König
Please push to amd-staging-drm-next ASAP and ping Kenny when that's
merged (or if it doesn't merge aut
Am 25.10.23 um 20:45 schrieb Felix Kuehling:
On 2023-10-25 02:12, Christian König wrote:
Am 24.10.23 um 21:20 schrieb David Francis:
dmaunmap can call ttm_bo_validate, which expects the
ttm dma_resv to be held.
Well first of all the dma_resv object isn't related to TTM.
Acquire the
Deucher
I don't know of hand if the high level function really cover everything,
so only Acked-by: Christian König for now.
Christian.
---
drivers/gpu/drm/amd/amdgpu/amdgpu_device.c | 15 +++
drivers/gpu/drm/amd/amdgpu/amdgpu_sdma.c | 21 -
drivers/gp
Am 24.10.23 um 21:20 schrieb David Francis:
dmaunmap can call ttm_bo_validate, which expects the
ttm dma_resv to be held.
Well first of all the dma_resv object isn't related to TTM.
Acquire the locks in amdgpu_amdkfd_gpuvm_dmaunmap_mem.
Because the dmaunmap step can now fail, two new number
Am 25.10.23 um 06:24 schrieb Luben Tuikov:
From: Alex Deucher
Rather than doing this in the IP code for the SDMA paging
engine, move it up to the core device level init level.
This should fix the scheduler init ordering.
v2: Fix checkpatch parens complaint; long lines. (Luben)
Signed-off-by:
ned-off-by: Andrey Grodzovsky
Reviewed-by: Christian König
Link: https://www.spinics.net/lists/amd-gfx/msg74112.html
Andrey separated the scheduler initialization from the ring init because
we need some of the rings for XGMI initialization which in turn in
necessary to figure out the
Am 24.10.23 um 01:41 schrieb Felix Kuehling:
[sorry, I hit send too early]
On 2023-10-23 11:15, Christian König wrote:
Am 23.10.23 um 15:06 schrieb Daniel Tang:
That commit causes the screen to freeze a few moments after running
clinfo on v6.6-rc7 and ROCm 5.6. Sometimes the rest of the
Am 20.10.23 um 11:59 schrieb Emily Deng:
Issue: Dead heappen during gpu recover, the call sequence as below:
amdgpu_device_gpu_recover->amdgpu_amdkfd_pre_reset->flush_delayed_work->
amdgpu_amdkfd_gpuvm_restore_process_bos->amdgpu_sync_wait
Resolving a deadlock with a timeout is illegal in gene
Am 24.10.23 um 04:55 schrieb Luben Tuikov:
On 2023-10-23 01:49, Christian König wrote:
Am 23.10.23 um 05:23 schrieb Luben Tuikov:
Initialize ring schedulers before using them, very early in the amdgpu boot,
at PCI probe time, specifically at frame-buffer dumb-create at fill-buffer.
This was
Am 23.10.23 um 15:06 schrieb Daniel Tang:
That commit causes the screen to freeze a few moments after running
clinfo on v6.6-rc7 and ROCm 5.6. Sometimes the rest of the computer
including ssh also freezes. On v6.5-rc1, it only results in a NULL pointer
deference message in dmesg and the process t
Am 20.10.23 um 11:59 schrieb Emily Deng:
Add error handling for amdgpu_sync_wait.
Signed-off-by: Emily Deng
Reviewed-by: Christian König for this one.
Going to discuss with Felix later today what we do with the timeout.
Christian.
---
drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gpuvm.c
Am 20.10.23 um 21:47 schrieb Felix Kuehling:
On 2023-10-20 09:10, Christian König wrote:
No, the wait forever is what is expected and perfectly valid user
experience.
Waiting with a timeout on the other hand sounds like a really bad
idea to me.
Every wait with a timeout needs a
lized.
Cc: Christian König
Cc: Alex Deucher
Cc: Felix Kuehling
Cc: AMD Graphics
Signed-off-by: Luben Tuikov
---
drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.c | 14 ++
1 file changed, 14 insertions(+)
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.c
b/drivers/gpu/drm/amd/amdgpu/am
ssage-
From: Christian König
Sent: Friday, October 20, 2023 3:29 PM
To: Deng, Emily ; amd-gfx@lists.freedesktop.org
Subject: Re: [PATCH 1/2] drm/amdgpu: Add timeout for sync wait
Am 20.10.23 um 08:13 schrieb Emily Deng:
Issue: Dead heappen during gpu recover, the call se
Looks like the KFD still needs this.
Signed-off-by: Christian König
Fixes: 8abc1eb2987a ("drm/amdkfd: switch over to using drm_exec v3")
---
drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gpuvm.c | 2 +-
1 file changed, 1 insertion(+), 1 deletion(-)
diff --git a/drivers/gpu/drm/
Am 17.10.23 um 16:36 schrieb Stanley.Yang:
This is workaround, kiq ring test failed in suspend stage when do ras
recovery for gfx v9_4_3.
Any idea why that failed? Problems like this usually point to an
incorrect init or in this case re-init procedure and are actually what
the ring test shoul
Am 20.10.23 um 08:13 schrieb Emily Deng:
You need a patch description and this patch here needs to come first and
not second.
Christian.
Signed-off-by: Emily Deng
---
drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gpuvm.c | 9 ++---
drivers/gpu/drm/amd/amdgpu/amdgpu_mes.c | 6
Am 20.10.23 um 08:13 schrieb Emily Deng:
Issue: Dead heappen during gpu recover, the call sequence as below:
amdgpu_device_gpu_recover->amdgpu_amdkfd_pre_reset->flush_delayed_work->
amdgpu_amdkfd_gpuvm_restore_process_bos->amdgpu_sync_wait
It is because the amdgpu_sync_wait is waiting for the b
Am 17.10.23 um 15:04 schrieb Alex Deucher:
On Tue, Oct 17, 2023 at 8:22 AM Christian König
wrote:
Looks like RADV is actually hitting this.
Signed-off-by: Christian König
Fixes: ca6c1e210aa7 ("drm/amdgpu: use the new drm_exec object for CS v3")
Acked-by: Alex Deucher
Pushed t
Am 18.10.23 um 19:05 schrieb Shyam Sundar S K:
On 10/18/2023 9:37 PM, Christian König wrote:
Am 18.10.23 um 17:47 schrieb Mario Limonciello:
On 10/18/2023 08:40, Christian König wrote:
Am 18.10.23 um 11:28 schrieb Shyam Sundar S K:
On 10/18/2023 2:50 PM, Ilpo Järvinen wrote:
On Wed, 18
Am 18.10.23 um 17:47 schrieb Mario Limonciello:
On 10/18/2023 08:40, Christian König wrote:
Am 18.10.23 um 11:28 schrieb Shyam Sundar S K:
On 10/18/2023 2:50 PM, Ilpo Järvinen wrote:
On Wed, 18 Oct 2023, Shyam Sundar S K wrote:
In order to provide GPU inputs to TA for the Smart PC
Am 18.10.23 um 11:28 schrieb Shyam Sundar S K:
On 10/18/2023 2:50 PM, Ilpo Järvinen wrote:
On Wed, 18 Oct 2023, Shyam Sundar S K wrote:
In order to provide GPU inputs to TA for the Smart PC solution to work, we
need to have interface between the PMF driver and the AMDGPU driver.
Add the i
alled amdgpu_ctx_priority_is_valid() already first thing in the
function.
Cc: Alex Deucher
Cc: Christian König
Signed-off-by: Luben Tuikov
Acked-by: Christian König
---
drivers/gpu/drm/amd/amdgpu/amdgpu_ctx.c | 15 ---
1 file changed, 8 insertions(+), 7 deletions(-)
diff --git a/drivers/gpu/dr
Am 17.10.23 um 23:13 schrieb Felix Kuehling:
abo->tbo.resource may be NULL in amdgpu_vm_bo_update.
Fixes: 180253782038 ("drm/ttm: stop allocating dummy resources during BO
creation")
Signed-off-by: Felix Kuehling
Reviewed-by: Christian König
---
drivers/gpu/drm/amd/amdgp
-by: Christian König
---
drivers/gpu/drm/amd/amdgpu/gmc_v9_0.c | 7 ++-
1 file changed, 2 insertions(+), 5 deletions(-)
diff --git a/drivers/gpu/drm/amd/amdgpu/gmc_v9_0.c
b/drivers/gpu/drm/amd/amdgpu/gmc_v9_0.c
index 3a1050344b59..b1eb81ca64bc 100644
--- a/drivers/gpu/drm/amd/amdgpu
Am 17.10.23 um 14:42 schrieb Hamza Mahfooz:
On 10/17/23 08:10, Christian König wrote:
Looks like RADV is actually hitting this.
Signed-off-by: Christian König
Fixes: ca6c1e210aa7 ("drm/amdgpu: use the new drm_exec object for CS
v3")
Do you think this will fix the following iss
Looks like RADV is actually hitting this.
Signed-off-by: Christian König
Fixes: ca6c1e210aa7 ("drm/amdgpu: use the new drm_exec object for CS v3")
---
drivers/gpu/drm/amd/amdgpu/amdgpu_cs.c | 3 ++-
1 file changed, 2 insertions(+), 1 deletion(-)
diff --git a/drivers/gpu/drm/
Am 17.10.23 um 09:25 schrieb Shashank Sharma:
Hello Christian, Felix,
Thanks for your comments, mine inline.
On 17/10/2023 07:55, Christian König wrote:
Am 17.10.23 um 00:15 schrieb Felix Kuehling:
On 2023-10-16 13:08, Shashank Sharma wrote:
This patch does the following:
- moves vm
Am 17.10.23 um 00:15 schrieb Felix Kuehling:
On 2023-10-16 13:08, Shashank Sharma wrote:
This patch does the following:
- moves vm->task_info struct to fpriv->task_info.
- makes task_info allocation dynamic.
- adds reference counting support for task_info structure.
- adds some new helper functi
Am 16.10.23 um 18:52 schrieb Bokun Zhang:
- Update SRIOV header with RB decouple flag
Signed-off-by: Bokun Zhang
---
drivers/gpu/drm/amd/amdgpu/amdgv_sriovmsg.h | 5 +++--
1 file changed, 3 insertions(+), 2 deletions(-)
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgv_sriovmsg.h
b/drivers/gpu
Am 12.10.23 um 09:41 schrieb Yifan Zhang:
flush the correct vmid tlb for specific pasid on gmc 11.
Signed-off-by: Yifan Zhang
Good catch, Reviewed-by: Christian König
---
drivers/gpu/drm/amd/amdgpu/gmc_v11_0.c | 2 +-
1 file changed, 1 insertion(+), 1 deletion(-)
diff --git a/drivers
when map the Bos into same VM, otherwise
we may trigger kernel general protection when iterate mappings from bo_va.
Signed-off-by: Felix Kuehling
Acked-by: Christian König
Reviewed-by: Ramesh Errabolu
Reviewed-By: Xiaogang Chen
Tested-By: Xiaogang Chen
---
drivers/gpu/drm/amd/amdgpu
Hi Feifei,
yeah, that is correct behavior. The GMC callback should *not* get called
during resume in a reset, because the reset needs to take care of
invalidating the TLB anyway.
If the later doesn't work any more we need to re-iterate the reset
procedure and not mess with this here.
Regar
Am 07.10.23 um 11:50 schrieb Greg KH:
On Sun, Sep 10, 2023 at 03:43:01PM -0500, Bryan Jennings wrote:
This is also causing log spam on 5.15. It was included in 5.15.128 as
commit 4921792e04f2125b5eadef9dbe9417a8354c7eff. I encountered this and
found https://gitlab.freedesktop.org/drm/amd/-/iss
Am 06.10.23 um 16:41 schrieb Alex Deucher:
On Fri, Oct 6, 2023 at 9:07 AM Christian König
wrote:
bo->tbo.resource can easily be NULL here.
Signed-off-by: Christian König
Add a link to the bug report?
Ah, crap. Forgotten to add the link before pushing that. But I've added
a C
tps://gitlab.freedesktop.org/drm/amd/-/issues/2362
Signed-off-by: Mario Limonciello
Acked-by: Christian König
---
v4->v5:
* New patch
---
drivers/gpu/drm/amd/display/amdgpu_dm/amdgpu_dm.c | 2 ++
1 file changed, 2 insertions(+)
diff --git a/drivers/gpu/drm/amd/display/amdgpu_dm/amdgpu_dm.c
b/drivers/gpu/d
ff-by: Mario Limonciello
Acked-by: Christian König
---
drivers/gpu/drm/amd/amdgpu/amdgpu_device.c | 36 +-
1 file changed, 29 insertions(+), 7 deletions(-)
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c
b/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c
index a362152
Am 09.10.23 um 03:50 schrieb Xu, Feifei:
[AMD Official Use Only - General]
Hi,
Based on your description, the above code should use "||" instead of "&&",
&& is to add more restriction here. To avoid skipping necessary TLB flush by
return.
For Asics < GFX11, !adev->gfx.is_poweron is always t
to allocate memory before suspend
starts so that it can potentially be evicted into swap instead.
Signed-off-by: Mario Limonciello
Apart from the commit message Reviewed-by: Christian König
.
Regards,
Christian.
---
v4->v5:
* New patch
---
drivers/gpu/drm/amd/amdgpu/amdgpu_devic
may fail.
So move this step into prepare() to move evict majority of
resources and update all non-pmops callers to call the same callback.
Link: https://gitlab.freedesktop.org/drm/amd/-/issues/2362
Signed-off-by: Mario Limonciello
Reviewed-by: Christian König for this one.
---
v4->
Am 06.10.23 um 16:24 schrieb Alex Deucher:
On Fri, Oct 6, 2023 at 4:32 AM Christian König
wrote:
Am 06.10.23 um 07:21 schrieb Lijo Lazar:
Add sysfs attribute to read power management log. A snapshot is
captured to the buffer when the attribute is read.
Signed-off-by: Lijo Lazar
---
v2
bo->tbo.resource can easily be NULL here.
Signed-off-by: Christian König
---
drivers/gpu/drm/amd/amdgpu/amdgpu_object.h | 2 +-
1 file changed, 1 insertion(+), 1 deletion(-)
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_object.h
b/drivers/gpu/drm/amd/amdgpu/amdgpu_object.h
index f3ee83cdf
Am 06.10.23 um 07:21 schrieb Lijo Lazar:
Add sysfs attribute to read power management log. A snapshot is
captured to the buffer when the attribute is read.
Signed-off-by: Lijo Lazar
---
v2: Pass PAGE_SIZE as the max size of input buffer
drivers/gpu/drm/amd/pm/amdgpu_pm.c | 40 ++
Am 05.10.23 um 16:59 schrieb Mario Limonciello:
On 10/5/2023 09:39, Christian König wrote:
Am 04.10.23 um 19:18 schrieb Mario Limonciello:
Linux PM core has a prepare() callback run before suspend.
If the system is under high memory pressure, the resources may need
to be evicted into swap
Am 04.10.23 um 19:18 schrieb Mario Limonciello:
Linux PM core has a prepare() callback run before suspend.
If the system is under high memory pressure, the resources may need
to be evicted into swap instead. If the storage backing for swap
is offlined during the suspend() step then such a call
: 256 bytes
Encode IB size alignment: 4 bytes
Also bump amdgpu driver version for this update.
Signed-off-by: Boyuan Zhang
Reviewed-by: Christian König
---
drivers/gpu/drm/amd/amdgpu/amdgpu_drv.c | 3 ++-
drivers/gpu/drm/amd/amdgpu/amdgpu_kms.c | 22 +++---
2 files
801 - 900 of 1887 matches
Mail list logo