On 2022-08-30 02:00, Christian König wrote:
Am 29.08.22 um 21:30 schrieb Felix Kuehling:
Am 2022-08-29 um 14:59 schrieb Christian König:
Am 29.08.22 um 18:07 schrieb Felix Kuehling:
Am 2022-08-29 um 11:38 schrieb Christian König:
Am 27.08.22 um 01:16 schrieb Felix Kuehling:
BOs can be in a
Am 2022-08-29 um 14:59 schrieb Christian König:
Am 29.08.22 um 18:07 schrieb Felix Kuehling:
Am 2022-08-29 um 11:38 schrieb Christian König:
Am 27.08.22 um 01:16 schrieb Felix Kuehling:
BOs can be in a different location than was intended at allocation
time,
for example when restoring fails
Am 2022-08-29 um 11:38 schrieb Christian König:
Am 27.08.22 um 01:16 schrieb Felix Kuehling:
BOs can be in a different location than was intended at allocation time,
for example when restoring fails after an eviction or BOs get pinned in
system memory. On some GPUs the MTYPE for coherent
Am 2022-08-29 um 11:00 schrieb Alex Sierra:
[Why] Devices with CPU XGMI iolink do not support PCIe peer access.
Signed-off-by: Alex Sierra
Acked-by: Alex Deucher
Reviewed-by: Felix Kuehling
---
drivers/gpu/drm/amd/amdgpu/amdgpu_device.c | 3 ++-
1 file changed, 2 insertions(+), 1
Am 2022-08-28 um 11:28 schrieb Christian König:
Am 26.08.22 um 23:49 schrieb Felix Kuehling:
On 2022-08-26 11:47, Alex Sierra wrote:
[Why] Devices with CPU XGMI iolink do not support PCIe peer access.
Signed-off-by: Alex Sierra
---
drivers/gpu/drm/amd/amdgpu/amdgpu_device.c | 3 ++-
1
every time the page
tables are updated.
Signed-off-by: Felix Kuehling
---
.../gpu/drm/amd/amdgpu/amdgpu_amdkfd_gpuvm.c | 9 -
drivers/gpu/drm/amd/amdgpu/amdgpu_vm.c| 19 +++
drivers/gpu/drm/amd/amdgpu/amdgpu_vm.h| 1 +
3 files changed, 28 insertions(+), 1
On 2022-08-26 11:47, Alex Sierra wrote:
[Why] Devices with CPU XGMI iolink do not support PCIe peer access.
Signed-off-by: Alex Sierra
---
drivers/gpu/drm/amd/amdgpu/amdgpu_device.c | 3 ++-
1 file changed, 2 insertions(+), 1 deletion(-)
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_device
Am 2022-08-25 um 04:58 schrieb Lijo Lazar:
HDP flush is used early in the init sequence as part of memory controller
block initialization. Hence remapping of HDP registers needed for flush
needs to happen earlier.
This also fixes the AER error reported as Unsupported Request during
driver load.
Am 2022-08-25 um 09:31 schrieb Christian König:
Use DMA_RESV_USAGE_BOOKKEEP for VM page table updates and KFD preemption fence.
Signed-off-by: Christian König
---
drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gpuvm.c | 2 +-
drivers/gpu/drm/amd/amdgpu/amdgpu_vm_sdma.c | 3 ++-
2 files chang
Am 2022-08-24 um 01:10 schrieb Lazar, Lijo:
On 8/23/2022 10:34 PM, Tom Seewald wrote:
On Sat, Aug 20, 2022 at 2:53 AM Lazar, Lijo wrote:
Missed the remap part, the offset is here -
https://elixir.bootlin.com/linux/v6.0-rc1/source/drivers/gpu/drm/amd/amdgpu/nv.c#L680
The trace is com
Do you mind squashing the two patches. It would make them easier to
review because it makes it easier to see that the same functions are
using both.
Thanks,
Felix
On 2022-08-24 16:01, Daniel Phillips wrote:
Patch 1 adds kfd_lock_pdd_by_id and patch 2 adds kfd_unlock_pdd helpers,
broken out
er to extend UTCL2
reach"
Signed-off-by: Mukul Joshi
Reviewed-by: Felix Kuehling
---
drivers/gpu/drm/amd/amdgpu/mmhub_v9_4.c | 12 ++--
1 file changed, 10 insertions(+), 2 deletions(-)
diff --git a/drivers/gpu/drm/amd/amdgpu/mmhub_v9_4.c
b/drivers/gpu/drm/amd/amdgpu/mmhub_v9
Am 2022-08-22 um 11:30 schrieb Daniel Phillips:
Convert most of the "mutex_lock; kfd_process_device_data_by_id" occurrences
in kfd_chardev to use the kfd_lock_pdd_by_id. These will now consistently
log debug output if the lookup fails. Sites where kfd_process_device_data_by_id
is used without loc
n same core")
Signed-off-by: Randy Dunlap
Cc: Gautham R. Shenoy
Cc: Michael Ellerman
Cc: Nicholas Piggin
Cc: Christophe Leroy
Cc: linuxppc-...@lists.ozlabs.org
Cc: amd-gfx@lists.freedesktop.org
Cc: Felix Kuehling
Cc: Alex Deucher
Cc: Christian König
Cc: "Pan, Xinhui"
Acked-by: Fel
On 2022-08-18 15:34, Randy Dunlap wrote:
Hi--
On 8/18/22 12:15, Sudip Mukherjee wrote:
On Thu, Aug 18, 2022 at 4:10 PM Randy Dunlap wrote:
On 8/18/22 03:43, Sudip Mukherjee wrote:
On Thu, Aug 18, 2022 at 3:09 AM Randy Dunlap wrote:
On 8/17/22 19:01, Alex Deucher wrote:
On Wed, Aug 17, 2
Am 2022-08-18 um 14:19 schrieb shaoyunl:
The additional call is caused by merge conflict
Signed-off-by: shaoyunl
Reviewed-by: Felix Kuehling
---
drivers/gpu/drm/amd/amdgpu/amdgpu_device.c | 2 --
1 file changed, 2 deletions(-)
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c
Am 2022-08-12 um 21:27 schrieb Bas Nieuwenhuizen:
This way callsites can choose between READ/BOOKKEEP reservations.
Signed-off-by: Bas Nieuwenhuizen
---
drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gpuvm.c | 5 +
drivers/gpu/drm/amd/amdgpu/amdgpu_cs.c | 9 +++--
drivers/gpu/dr
Am 2022-08-17 um 11:04 schrieb Felix Kuehling:
Am 2022-08-16 um 23:09 schrieb Lang Yu:
Then we can remove the burden of maintaining codes to
parse family_id from gfx version in rocr,
i.e., remove DevIDToAddrLibFamily().
I'm OK with the change. But you won't be able
Am 2022-08-16 um 23:09 schrieb Lang Yu:
Then we can remove the burden of maintaining codes to
parse family_id from gfx version in rocr,
i.e., remove DevIDToAddrLibFamily().
I'm OK with the change. But you won't be able to remove
DevIDToAddrLibFamily as long as ROCr needs to support older kerne
Am 2022-08-15 um 15:25 schrieb Mukul Joshi:
There are no backing hardware registers for ih_soft ring.
As a result, don't try to access hardware registers for read
and write pointers when processing interrupts on the IH soft
ring.
Signed-off-by: Mukul Joshi
Reviewed-by: Felix Kue
Am 2022-08-14 um 13:27 schrieb Christian König:
Am 12.08.22 um 22:56 schrieb Mukul Joshi:
There are no backing hardware registers for ih_soft ring.
As a result, don't try to access hardware registers for read
and write pointers when processing interrupts on the IH soft
ring.
Mhm, the original
On 2022-08-12 18:05, Andrey Grodzovsky wrote:
On 2022-08-12 14:38, Kim, Jonathan wrote:
[Public]
Hi Andrey,
Here's the load/unload stack trace. This is a 2 GPU xGMI system. I
put dbg_xgmi_hive_get/put refcount print post kobj get/put.
It's stuck at 2 on unload. If it's an 8 GPU system,
On 2022-08-12 16:56, Mukul Joshi wrote:
There are no backing hardware registers for ih_soft ring.
As a result, don't try to access hardware registers for read
and write pointers when processing interrupts on the IH soft
ring.
Signed-off-by: Mukul Joshi
The patch looks good to me. But you pr
x for this:
https://lore.kernel.org/all/20220706183302.1719795-1-ramesh.errab...@amd.com/
You commented on a version of his patch:
https://lore.kernel.org/all/20220629161241.GM11460@kadam/
Did this get lost somehow? Anyway, your patch looks good to me and I'm
going to apply it to amd-stagin
On 2022-08-12 16:30, Tejun Heo wrote:
On Fri, Aug 12, 2022 at 04:26:47PM -0400, Felix Kuehling wrote:
Hi workqueue maintainers,
In the KFD (amdgpu) driver we found a need to schedule bottom half interrupt
handlers on CPU cores different from the one where the top-half interrupt
handler runs to
Hi workqueue maintainers,
In the KFD (amdgpu) driver we found a need to schedule bottom half
interrupt handlers on CPU cores different from the one where the
top-half interrupt handler runs to avoid the interrupt handler stalling
the bottom half in extreme scenarios. See my latest patch that t
On 2022-08-12 09:55, Philip Yang wrote:
On 2022-08-11 15:04, Felix Kuehling wrote:
On systems that support SMT (hyperthreading) schedule the bottom half of
the KFD interrupt handler on the same core. This makes it possible to
reserve a core for interrupt handling and have the bottom half run
before.
Use for_each_cpu_wrap instead of open-coding it.
Signed-off-by: Felix Kuehling
---
drivers/gpu/drm/amd/amdkfd/kfd_device.c | 20
1 file changed, 16 insertions(+), 4 deletions(-)
diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_device.c
b/drivers/gpu/drm/amd/amdkfd
Am 2022-08-11 um 09:42 schrieb Jonathan Kim:
When an xgmi node is added to the hive, it takes another hive
reference for its reset domain.
This extra reference was not dropped on device removal from the
hive so drop it.
Signed-off-by: Jonathan Kim
---
drivers/gpu/drm/amd/amdgpu/amdgpu_xgmi.c
Am 2022-08-10 um 19:41 schrieb Felix Kuehling:
On systems that support SMT (hyperthreading) schedule the bottom half of
the KFD interrupt handler on the same core. This makes it possible to
reserve a core for interrupt handling and have the bottom half run on
that same core.
On systems without
before.
Use for_each_cpu_wrap instead of open-coding it.
Signed-off-by: Felix Kuehling
---
drivers/gpu/drm/amd/amdkfd/kfd_device.c | 20
1 file changed, 16 insertions(+), 4 deletions(-)
diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_device.c
b/drivers/gpu/drm/amd/amdkfd
and drop the svm_bo
reference if it isn't.
Signed-off-by: Felix Kuehling
---
drivers/gpu/drm/amd/amdkfd/kfd_svm.c | 17 +++--
drivers/gpu/drm/amd/amdkfd/kfd_svm.h | 1 -
2 files changed, 7 insertions(+), 11 deletions(-)
diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_svm.c
b/dri
hangs when kfd_ioctl_wait_events is interrupted by a
signal.
Signed-off-by: Felix Kuehling
---
drivers/gpu/drm/amd/amdkfd/kfd_chardev.c | 2 +-
drivers/gpu/drm/amd/amdkfd/kfd_events.c | 24
drivers/gpu/drm/amd/amdkfd/kfd_priv.h| 2 +-
3 files changed, 14 insertions
On 2022-08-04 12:01, Joseph Greathouse wrote:
Enable translate_further on Arcturus and Aldebaran server chips
in order to increase the UTCL2 reach from 8 GiB to 64 GiB,
which is more in line with the amount of framebuffer DRAM in
the devices.
Signed-off-by: Joseph Greathouse
Acked-by: Felix
Only allocate doorbells when the first queue is created on a GPU or the
doorbells need to be mapped into CPU or GPU virtual address space. This
avoids allocating doorbells unnecessarily and can allow more processes
to use KFD on multi-GPU systems.
Signed-off-by: Felix Kuehling
---
drivers/gpu
Your patches are missing Signed-off-by lines. If you use "git commit
-s", git should add those automatically for your convenience. Other than
that, the patches look good to me. With Signed-off-by added, the series is
Reviewed-by: Felix Kuehling
Am 2022-07-28 um 23:16 schrieb Danie
Am 2022-07-27 um 23:30 schrieb Yu Zhe:
time_is_before_jiffies deals with timer wrapping correctly.
Signed-off-by: Yu Zhe
Thank you. This patch looks good to me. I'm applying it to
amd-staging-drm-next.
Reviewed-by: Felix Kuehling
---
drivers/gpu/drm/amd/amdkfd/kfd_interrupt.
Am 2022-07-27 um 23:52 schrieb Alex Deucher:
On Wed, Jul 27, 2022 at 7:08 PM Felix Kuehling wrote:
Don't drop the stub fence reference after installing it as a replacement
for the eviction fence. dma_resv_replace_fences doesn't take another
reference to the fence, so it takes owners
Am 2022-07-25 um 06:32 schrieb Lang Yu:
No need to reference the BO here, dmabuf framework will handle that.
OK. I guess I needed to do that manually for the userptr attachment, and
copy/pasted it unnecessarily for the dmabuf attachment.
Reviewed-by: Felix Kuehling
Signed-off-by
_fences v2")
CC: Christian König
Signed-off-by: Felix Kuehling
---
drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gpuvm.c | 1 -
1 file changed, 1 deletion(-)
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gpuvm.c
b/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gpuvm.c
index 87a3a3ae9448..a6c7dcd8c345
This patch introduces a build warning for me:
CC [M] drivers/gpu/drm/amd/amdgpu/../amdkfd/kfd_interrupt.o
In file included from /home/fkuehlin/compute/kernel/include/linux/spinlock.h:54,
from /home/fkuehlin/compute/kernel/include/linux/mmzone.h:8,
from /home/f
Am 2022-07-26 um 19:43 schrieb Rajneesh Bhardwaj:
For typesafety, use container_of() instead of implicit cast from struct
ttm_tt to struct amdgpu_ttm_tt.
Cc: Christian König
Signed-off-by: Rajneesh Bhardwaj
---
drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.c | 44 ++---
1 file c
Am 2022-07-25 um 17:17 schrieb Philip Yang:
This will be used to split giant svm range into smaller ranges, to
support VRAM overcommitment by giant range and improve GPU retry fault
recover on giant range.
Signed-off-by: Philip Yang
Reviewed-by: Felix Kuehling
---
drivers/gpu/drm/amd
Am 2022-07-26 um 07:15 schrieb Lang Yu:
The kgd_mem memory allocated in amdgpu_amdkfd_gpuvm_import_dmabuf()
is not freed properly.
Explicitly free it in amdgpu_amdkfd_gpuvm_free_memory_of_gpu()
under condition "mem->bo->kfd_bo != mem".
Suggested-by: Felix Kuehling
Signed
Am 2022-07-25 um 22:18 schrieb Lang Yu:
On 07/25/ , Felix Kuehling wrote:
Am 2022-07-25 um 20:40 schrieb Lang Yu:
On 07/25/ , Felix Kuehling wrote:
Am 2022-07-25 um 20:15 schrieb Lang Yu:
On 07/25/ , Felix Kuehling wrote:
Am 2022-07-25 um 06:32 schrieb Lang Yu:
We have memory leak issue
Am 2022-07-25 um 20:40 schrieb Lang Yu:
On 07/25/ , Felix Kuehling wrote:
Am 2022-07-25 um 20:15 schrieb Lang Yu:
On 07/25/ , Felix Kuehling wrote:
Am 2022-07-25 um 06:32 schrieb Lang Yu:
We have memory leak issue in current implenmention, i.e.,
the allocated struct kgd_mem memory is not
Am 2022-07-25 um 20:15 schrieb Lang Yu:
On 07/25/ , Felix Kuehling wrote:
Am 2022-07-25 um 06:32 schrieb Lang Yu:
We have memory leak issue in current implenmention, i.e.,
the allocated struct kgd_mem memory is not handled properly.
The idea is always creating a buffer object when importing
a
Am 2022-07-25 um 13:19 schrieb Philip Yang:
Giant svm range split to smaller ranges, align the range start address
to max svm range pages to improve MMU TLB usage.
Signed-off-by: Philip Yang
Reviewed-by: Felix Kuehling
---
drivers/gpu/drm/amd/amdkfd/kfd_svm.c | 49
Am 2022-07-25 um 13:19 schrieb Philip Yang:
This will be used to split giant svm range into smaller ranges, to
support VRAM overcommitment by giant range and improve GPU retry fault
recover on giant range.
Signed-off-by: Philip Yang
---
drivers/gpu/drm/amd/amdkfd/kfd_migrate.c | 2 ++
dri
Am 2022-07-25 um 13:14 schrieb Daniel Phillips:
Rounding up allocations in the allocation path caused test regressions,
so now just round in the availability path.
---
drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gpuvm.c | 11 +--
1 file changed, 5 insertions(+), 6 deletions(-)
diff --git
"Check whether
to prevent eviction of @f by @mm".
With that fixed, the patch is
Reviewed-by: Felix Kuehling
*
* @f: [IN] fence
* @mm: [IN] mm that needs to be verified
+ *
+ * Check if @mm is same as that of the fence @f, if same return TRUE else
+ * return FALSE.
+ *
Am 2022-07-25 um 08:23 schrieb Philip Yang:
Giant svm range split to smaller ranges, align the range start address
to max svm range pages to improve MMU TLB usage.
Signed-off-by: Philip Yang
---
drivers/gpu/drm/amd/amdkfd/kfd_svm.c | 52 +++-
1 file changed, 36 insert
Am 2022-07-25 um 08:23 schrieb Philip Yang:
This will be used to split giant svm range into smaller ranges, to
support VRAM overcommitment by giant range and improve GPU retry fault
recover on giant range.
Signed-off-by: Philip Yang
---
drivers/gpu/drm/amd/amdkfd/kfd_migrate.c | 2 ++
drive
Am 2022-07-25 um 06:32 schrieb Lang Yu:
We have memory leak issue in current implenmention, i.e.,
the allocated struct kgd_mem memory is not handled properly.
The idea is always creating a buffer object when importing
a gfx BO based dmabuf.
Signed-off-by: Lang Yu
---
.../gpu/drm/amd/amdgpu
ses system crash with
different backtrace.
The fix is to increase process refcount and then decrease the refcount
after mmu_notifier_get success.
Signed-off-by: Philip Yang
Reviewed-by: Felix Kuehling
---
drivers/gpu/drm/amd/amdkfd/kfd_process.c | 6 ++
1 file changed, 6 inserti
+ snoop = true;
+
+
With the two extra blank lines removed, this patch is
Reviewed-by: Felix Kuehling
Please check whether a similar cleanup can be made in
svm_range_get_pte_flags, or maybe even, whether common code can be
factored out of those two functions.
Re
prange creation and free.
Signed-off-by: Alex Sierra
Reviewed-by: Felix Kuehling
---
drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd.h| 4 ++
.../gpu/drm/amd/amdgpu/amdgpu_amdkfd_gpuvm.c | 23 ---
drivers/gpu/drm/amd/amdkfd/kfd_svm.c | 60 +--
3 files changed
On 2022-07-18 06:50, David Hildenbrand wrote:
On 15.07.22 17:05, Alex Sierra wrote:
[WHY]
It makes more sense to have these helpers in zone specific header
file, rather than the generic mm.h
Signed-off-by: Alex Sierra
Acked-by: David Hildenbrand
Thank you! I don't think I have the authorit
Xinhui submitted this patch instead, which should address the same
issue: "drm/amdgpu: Remove one duplicated ef removal"
Alex, can you pick up that patch for drm-fixes for 5.19, if it's not too
late?
Thanks,
Felix
On 2022-07-18 10:58, Mike Lothian wrote:
Is this likely to land before 5.1
On 2022-07-11 21:56, Alex Sierra wrote:
[WHY]
Unified memory with xnack off should be tracked, as userptr mappings
and legacy allocations do. To avoid oversuscribe system memory when
xnack off.
[How]
Exposing functions reserve_mem_limit and unreserve_mem_limit to SVM
API and call them on every pr
itted.
Signed-off-by: Felix Kuehling
---
drivers/gpu/drm/ttm/ttm_device.c | 3 +++
1 file changed, 3 insertions(+)
diff --git a/drivers/gpu/drm/ttm/ttm_device.c b/drivers/gpu/drm/ttm/ttm_device.c
index be24bb6cefd0..165a6cbb45d5 100644
--- a/drivers/gpu/drm/ttm/ttm_device.c
+++ b/drivers/gpu/dr
On 2022-07-08 19:17, Ramesh Errabolu wrote:
- Update defconfig for PCI_P2PDMA
- Update defconfig for DMABUF_MOVE_NOTIFY
- Update defconfig for HSA_AMD_P2P
---
The patch is missing a Signed-off-by. With that fixed
Reviewed-by: Felix Kuehling
Notes:
Following procedure
Am 2022-07-13 um 05:14 schrieb shikai guo:
From: Shikai Guo
While executing KFDMemoryTest.MMBench, test case will allocate 4KB size memory
1000 times.
Every time, user space will get 2M memory.APU VRAM is 512M, there is not enough
memory to be allocated.
So the 2M aligned feature is not sui
Eric sent out the corresponding user mode patches to the mailing list as
well. It looks a bit weird, because it looks like they're part of the
same patch series. But patch 2 and 3 are actually user mode patches. The
interesting one is patch 3.
Do we still need a link to a user mode patch in t
;
lock((work_completion)(&svms->deferred_list_work));
lock(&process->mutex);
Signed-off-by: Philip Yang
Reviewed-by: Felix Kuehling
---
drivers/gpu/drm/amd/amdkfd/kfd_process.c | 21 +
1 file changed, 9 insertions(+), 12 deletions(-)
On 2022-07-11 14:41, Eric Huang wrote:
To expose unified memory for ctx save/resotre area feature
availablity to libhsakmt.
Signed-off-by: Eric Huang
Reviewed-by: Felix Kuehling
---
include/uapi/linux/kfd_ioctl.h | 3 ++-
1 file changed, 2 insertions(+), 1 deletion(-)
diff --git a
On 2022-07-08 07:28, David Hildenbrand wrote:
On 07.07.22 21:03, Alex Sierra wrote:
[WHY]
Have a cleaner way to expose all page zone helpers in one header
What exactly is a "page zone"? Do you mean a buddy zone as in
include/linux/mmzone.h ?
Zone as in ZONE_DEVICE. Maybe we could extend mmzone
I think this could also be fixed by not taking the process_info lock in
svm_range_restore_work and svm_range_set_attr. I'm not even sure why
we're taking this lock in the SVM code. I think that was copied from the
restore workers in amdgpu_amdkfd_gpuvm.c because there it's used to
protect the B
On 2022-07-07 21:53, xinhui pan wrote:
That has been done in BO release notify.
Signed-off-by: xinhui pan
Reviewed-by: Felix Kuehling
---
drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gpuvm.c | 5 -
1 file changed, 5 deletions(-)
diff --git a/drivers/gpu/drm/amd/amdgpu
On 2022-07-07 12:45, Philip Yang wrote:
MMU notifier callback unmap the svm range update page table may free the
PTB BO, then amdgpu_fill_buffer zero BO memory could cause deadlock as
kmalloc may trigger MMU notifier.
amdgpu_vm_pt_clear setup PTB BO memory with initial value, and no
sensitive da
Am 2022-07-07 um 09:39 schrieb philip yang:
On 2022-07-07 06:28, xinhui pan wrote:
Queue would be freed when create_queue_cpsch fails
So lets do queue cleanup otherwise various list and memory issues
happen.
This bug was introduced when adding MES support, as we used to ignore
execute_queue
Am 2022-07-07 um 05:54 schrieb Christian König:
Am 07.07.22 um 11:50 schrieb xinhui pan:
Fence is accessed by dma_resv_add_fence() now.
Use amdgpu_amdkfd_remove_eviction_fence instead.
Signed-off-by: xinhui pan
---
drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gpuvm.c | 4 ++--
1 file changed, 2
"drm/amdkfd: Extend KFD device topology to surface peer-to-peer
links")
Signed-off-by: Ramesh Errabolu
Reported-by: Dan Carpenter
Reviewed-by: Felix Kuehling
---
drivers/gpu/drm/amd/amdkfd/kfd_topology.c | 34 +++
1 file changed, 17 insertions(+), 17 deletions(-)
On 2022-07-05 15:02, philip yang wrote:
On 2022-07-05 14:48, Felix Kuehling wrote:
On 2022-06-10 21:03, Philip Yang wrote:
If destroying/unmapping queue failed, application may destroy queue
again, cause below kernel warning backtrace.
For outstanding queues, either applications forget to
On 2022-06-10 21:03, Philip Yang wrote:
If destroying/unmapping queue failed, application may destroy queue
again, cause below kernel warning backtrace.
For outstanding queues, either applications forget to destroy or failed
to destroy, kfd_process_notifier_release will remove queue sysfs
object
On 2022-07-05 12:09, philip yang wrote:
On 2022-06-27 20:23, Alex Sierra wrote:
[WHY]
Unified memory with xnack off should be tracked, as userptr mappings
and legacy allocations do. To avoid oversuscribe system memory when
xnack off.
I think this also apply to XNACK ON (remove p->xnack_enabl
header->ErrorEventId = Event->EventId;
+ header->ErrorReason = ErrPayload;
+ header->DebugOffset = q->ctx_save_restore_size;
+ header->DebugSize = q->debug_memory_size;
+ }
Is there a way to refac
gets
mapped. So just map root PD after updating vm->update_funcs
in amdgpu_vm_make_compute whether the vm_update_mode changed
or not.
v3:
- Add some comments suggested by Christian.
v2:
- Don't rename vm_validate_pt_pd_bos and make it public.
Signed-off-by: Lang Yu
Reviewed-by: F
On 2022-06-30 11:19, Eric Huang wrote:
On 2022-06-29 19:29, Felix Kuehling wrote:
On 2022-06-29 18:53, Eric Huang wrote:
On 2022-06-29 18:20, Felix Kuehling wrote:
On 2022-06-28 17:43, Eric Huang wrote:
Two changes:
1. reducing unnecessary evict/unmap when range is not mapped to gpu.
2
: Philip Yang
Reviewed-by: Felix Kuehling
---
include/uapi/linux/kfd_ioctl.h | 37 ++
1 file changed, 37 insertions(+)
diff --git a/include/uapi/linux/kfd_ioctl.h b/include/uapi/linux/kfd_ioctl.h
index c648ed7c5ff1..f239e260796b 100644
--- a/include/uapi/linux
Am 2022-06-28 um 10:50 schrieb Philip Yang:
Indicate SMI profiling events available.
Signed-off-by: Philip Yang
Reviewed-by: Felix Kuehling
---
include/uapi/linux/kfd_ioctl.h | 3 ++-
1 file changed, 2 insertions(+), 1 deletion(-)
diff --git a/include/uapi/linux/kfd_ioctl.h b/include
ang
Reviewed-by: Felix Kuehling
---
drivers/gpu/drm/amd/amdkfd/kfd_smi_events.c | 14 ++
1 file changed, 10 insertions(+), 4 deletions(-)
diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_smi_events.c
b/drivers/gpu/drm/amd/amdkfd/kfd_smi_events.c
index e5896b7a16dd..0472b56de
Am 2022-06-28 um 10:50 schrieb Philip Yang:
SVM range unmapped from GPUs when range is unmapped from CPU, or with
xnack on from MMU notifier when range is evicted or migrated.
Signed-off-by: Philip Yang
Reviewed-by: Felix Kuehling
---
drivers/gpu/drm/amd/amdkfd/kfd_smi_events.c | 9
restore.
Signed-off-by: Philip Yang
Reviewed-by: Felix Kuehling
---
drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd.h| 2 +-
.../gpu/drm/amd/amdgpu/amdgpu_amdkfd_gpuvm.c | 12 ---
drivers/gpu/drm/amd/amdkfd/kfd_chardev.c | 4 +--
drivers/gpu/drm/amd/amdkfd/kfd_device.c | 4
eviction.
Signed-off-by: Philip Yang
Reviewed-by: Felix Kuehling
---
drivers/gpu/drm/amd/amdkfd/kfd_migrate.c| 53 -
drivers/gpu/drm/amd/amdkfd/kfd_migrate.h| 5 +-
drivers/gpu/drm/amd/amdkfd/kfd_smi_events.c | 22 +
drivers/gpu/drm/amd/amdkfd
tamp < AMDGPU_SVM_RANGE_RETRY_FAULT_PENDING) {
+ if (ktime_before(timestamp, ktime_add_ns(prange->validate_timestamp,
+ AMDGPU_SVM_RANGE_RETRY_FAULT_PENDING))) {
You changed the timestamp units from us to ns. I think you'll need to
update AMDGPU_SVM_RANGE_RETRY_FAULT_PENDING (multiply with 100
On 2022-06-29 18:53, Eric Huang wrote:
On 2022-06-29 18:20, Felix Kuehling wrote:
On 2022-06-28 17:43, Eric Huang wrote:
Two changes:
1. reducing unnecessary evict/unmap when range is not mapped to gpu.
2. adding always evict when flags is set to always_mapped.
Signed-off-by: Eric Huang
distribute CUs across the SEs in a pairwise manner to
assume WGP mode at all times.
Signed-off-by: Jonathan Kim
Looks good to me. Three nit-picks inline. With that fixed, the patch is
Reviewed-by: Felix Kuehling
---
drivers/gpu/drm/amd/amdkfd/kfd_mqd_manager.c | 12
On 2022-06-28 17:43, Eric Huang wrote:
Two changes:
1. reducing unnecessary evict/unmap when range is not mapped to gpu.
2. adding always evict when flags is set to always_mapped.
Signed-off-by: Eric Huang
---
drivers/gpu/drm/amd/amdkfd/kfd_svm.c | 10 --
1 file changed, 8 insertions
On 2022-06-29 03:33, David Hildenbrand wrote:
On 29.06.22 05:54, Alex Sierra wrote:
is_pinnable_page() and folio_is_pinnable() were renamed to
is_longterm_pinnable_page() and folio_is_longterm_pinnable()
respectively. These functions are used in the FOLL_LONGTERM flag
context.
Subject talks abo
Am 2022-06-28 um 20:03 schrieb Errabolu, Ramesh:
[AMD Official Use Only - General]
My responses are inline
-Original Message-
From: Kuehling, Felix
Sent: Tuesday, June 28, 2022 6:41 PM
To: Errabolu, Ramesh ; amd-gfx@lists.freedesktop.org;
dan.carpen...@oracle.com
Subject: Re: [PATCH]
Am 2022-06-28 um 19:25 schrieb Ramesh Errabolu:
The patch fixes couple of warnings, as reported by Smatch
a static analyzer
Signed-off-by: Ramesh Errabolu
Reported-by: Dan Carpenter
---
drivers/gpu/drm/amd/amdkfd/kfd_topology.c | 36 ---
1 file changed, 19 insertions(+
Am 2022-06-27 um 12:01 schrieb Eric Huang:
No. There is only internal link for now, because it is under review.
Once it is submitted, external link should be in gerritgit for libhsakmt.
Hi Eric,
For anything that requires ioctl API changes, the user mode and kernel
mode changes need to be rev
.
Signed-off-by: Alex Sierra
Acked-by: Felix Kuehling
Reviewed-by: Alistair Popple
[hch: rebased ontop of the refcount changes,
removed is_dev_private_or_coherent_page]
Signed-off-by: Christoph Hellwig
---
include/linux/memremap.h | 19 +++
mm/memcontrol.c
Am 2022-06-10 um 13:13 schrieb Graham Sider:
Store MES scheduler and MES KIQ version numbers in amdgpu_mes for GFX11.
Signed-off-by: Graham Sider
Acked-by: Felix Kuehling
---
drivers/gpu/drm/amd/amdgpu/amdgpu_mes.h | 3 +++
drivers/gpu/drm/amd/amdgpu/mes_v11_0.c | 12
Am 2022-06-10 um 13:13 schrieb Graham Sider:
Update MES API to support oversubscription without aggregated doorbell
for usermode queues.
Signed-off-by: Graham Sider
Acked-by: Felix Kuehling
---
drivers/gpu/drm/amd/amdgpu/amdgpu_mes.c | 1 +
drivers/gpu/drm/amd/amdgpu
Am 2022-06-10 um 13:13 schrieb Graham Sider:
Starting with GFX11, MES requires wptr BOs to be GTT allocated/mapped to
GART for usermode queues in order to support oversubscription. In the
case that work is submitted to an unmapped queue, MES must have a GART
wptr address to determine whether the
: David Yat Sin
Reviewed-by: Felix Kuehling
---
drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd.h| 1 +
.../gpu/drm/amd/amdgpu/amdgpu_amdkfd_gpuvm.c | 38 +--
drivers/gpu/drm/amd/amdkfd/kfd_chardev.c | 34 +
include/uapi/linux/kfd_ioctl.h
Am 2022-06-10 um 11:46 schrieb Alex Deucher:
drivers/gpu/drm/amd/amdgpu/../amdkfd/kfd_topology.c:1542:11:
warning: variable 'i' set but not used [-Wunused-but-set-variable]
Reported-by: kernel test robot
Signed-off-by: Alex Deucher
Thank you for taking care of this.
Reviewed
le to which BO belongs
+ *
+ * Return: void
I don't think you need to state a void return explicitly. [+David],
since you were looking into KFD documentation and kernel-doc comments
lately, do you have any feedback on the kernel-doc syntax?
Other than that, this patch is
Reviewed-by: Felix Kue
901 - 1000 of 2256 matches
Mail list logo
