Oh, I see your point, but that actually presents for a cleanup patch, and mine
is to add a condition to fix memory leak, I think they different purpose and
should be separated,
I can add one more patch to cleanup it with that "create_bo_kenel" to make code
more tight and clean
BR Monk
This condition is wrong, you are right, and I followed up a patch to correct it
Should be :
If (!amdgpu_sriov_vf()) || !adev->in_reset)
BR Monk
-Original Message-
From: Quan, Evan
Sent: 2017年9月20日 9:32
To: Koenig, Christian ; Liu, Monk ;
Hi Monk,
I think your change affects barematal case. Per my confirmation, vega10 cannot
boot with the change applied.
If the change is only intended to cover gpu reset case of sriov, maybe the
logic should be
If (!(amdgpu_sriov_vf(adev) && adev->in_sriov_reset))
Regards,
Evan
>-Original
>> +vma->vm_pgoff = amdgpu_bo_mmap_offset(bo) >> PAGE_SHIFT;
> Maybe better use "vma->vm_pgoff += amdgpu_bo_mmap_offset(bo) >> PAGE_SHIFT;",
> but I'm not sure.
> How other drivers handle this?
This is a good catch. Looks like pgoff is honored during prime mmap, not a fake
offset here.
>> +
On 19/09/17 02:33 PM, Christian König wrote:
[root@carrizo ~]# xxd -l 1024 -s 0xC
/sys/kernel/debug/dri/0/amdgpu_iova
Actually 0xC is a special address, e.g. video BIOS if I'm not
completely mistaken.
Not sure why that would be mapped by the driver but I can also read the
kernel's
Signed-off-by: Felix Kuehling
---
drivers/gpu/drm/amd/include/linux/chash.h | 22 +++---
1 file changed, 15 insertions(+), 7 deletions(-)
diff --git a/drivers/gpu/drm/amd/include/linux/chash.h
b/drivers/gpu/drm/amd/include/linux/chash.h
index
Signed-off-by: Felix Kuehling
---
drivers/gpu/drm/amd/lib/chash.c | 30 +++---
1 file changed, 23 insertions(+), 7 deletions(-)
diff --git a/drivers/gpu/drm/amd/lib/chash.c b/drivers/gpu/drm/amd/lib/chash.c
index 1bc4287..e07e6f3 100644
---
Fine from my end.
Acked-by: Dave Airlie
Dave.
On 20 September 2017 at 00:27, Christian König
wrote:
> Am 19.09.2017 um 16:22 schrieb Alex Deucher:
>>
>> It was not clear. The rest of the driver is MIT/X11.
>>
>> Signed-off-by: Alex Deucher
Signed-off-by: Eric Huang
---
drivers/gpu/drm/amd/powerplay/hwmgr/hwmgr.c | 75 -
drivers/gpu/drm/amd/powerplay/inc/hwmgr.h | 6 +++
2 files changed, 80 insertions(+), 1 deletion(-)
diff --git a/drivers/gpu/drm/amd/powerplay/hwmgr/hwmgr.c
Eric Huang (4):
drm/amdgpu: add cgs query info of pci bus devfn
drm/amd/powerplay: add register thermal interrupt in hwmgr_hw_init
drm/amd/powerplay: implement register thermal interrupt for Vega10
drm/amd/powerplay: change alert temperature range
drivers/gpu/drm/amd/amdgpu/amdgpu_cgs.c
Change to more meaningful range that triggers thermal
interrupts.
Signed-off-by: Eric Huang
---
drivers/gpu/drm/amd/powerplay/hwmgr/hardwaremanager.c | 4 ++--
1 file changed, 2 insertions(+), 2 deletions(-)
diff --git
Use it to replace the hard coded value in amdgpu_vm_bo_update_mapping().
Change-Id: I85d89d401b8dbcf01ca9c55c281e552db874fde5
Signed-off-by: Yong Zhao
---
drivers/gpu/drm/amd/amdgpu/amdgpu.h| 4
drivers/gpu/drm/amd/amdgpu/amdgpu_vm.c | 2 +-
Signed-off-by: Eric Huang
---
drivers/gpu/drm/amd/amdgpu/amdgpu_cgs.c | 3 +++
drivers/gpu/drm/amd/include/cgs_common.h | 1 +
2 files changed, 4 insertions(+)
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_cgs.c
b/drivers/gpu/drm/amd/amdgpu/amdgpu_cgs.c
index
[root@carrizo ~]# xxd -l 1024 -s 0xC
/sys/kernel/debug/dri/0/amdgpu_iova
Actually 0xC is a special address, e.g. video BIOS if I'm not
completely mistaken.
Take a look at /proc/iomem and try reading something which is marked as
"System RAM".
If that still works we are lost and
As I predicted ...
[root@carrizo ~]# xxd -l 1024 -s 0xC
/sys/kernel/debug/dri/0/amdgpu_iova
000c: 55aa 74e9 a902 U.t.
000c0010: 4c02 4942 L.IB
000c0020: 4daa 2e93 0004
Signed-off-by: Tom St Denis
(v2): Don't create iova file when iommu is disabled
---
drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.c | 4
1 file changed, 4 insertions(+)
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.c
b/drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.c
index
When max_bytes is not 8 bytes aligned and bo size is larger than
max_bytes, the last 8 bytes in a ttm node may be left unchanged.
For example, on pre SDMA 4.0, max_bytes = 0x1f, and the bo size
is 0x20, the problem will happen.
In order to fix the problem, we separately store the max nums
On 19/09/17 01:18 PM, Christian König wrote:
Am 19.09.2017 um 19:14 schrieb Tom St Denis:
On 19/09/17 01:10 PM, Christian König wrote:
As far as I know we don't need #ifdefs cause there are dummy
functions when IOMMU is not compiled in.
But this patch is a NAK since it circumvents the
Am 19.09.2017 um 19:14 schrieb Tom St Denis:
On 19/09/17 01:10 PM, Christian König wrote:
As far as I know we don't need #ifdefs cause there are dummy
functions when IOMMU is not compiled in.
But this patch is a NAK since it circumvents the /dev/mem restriction
when IOMMU is disabled and
On 19/09/17 01:10 PM, Christian König wrote:
As far as I know we don't need #ifdefs cause there are dummy functions
when IOMMU is not compiled in.
But this patch is a NAK since it circumvents the /dev/mem restriction
when IOMMU is disabled and that is not something we can easily allow.
I
As far as I know we don't need #ifdefs cause there are dummy functions
when IOMMU is not compiled in.
But this patch is a NAK since it circumvents the /dev/mem restriction
when IOMMU is disabled and that is not something we can easily allow.
Regards,
Christian.
Am 19.09.2017 um 17:36
This is a reference patch series for the kernel series:
"drm/amdgpu: introduce AMDGPU_GEM_CREATE_EXPLICIT_SYNC v2"
Andres Rodriguez (2):
radv: factor out radv_alloc_memory
radv: disable implicit sync for radv allocated bos v2
src/amd/vulkan/radv_device.c | 22
Implicit sync kicks in when a buffer is used by two different amdgpu
contexts simultaneously. Jobs that use explicit synchronization
mechanisms end up needlessly waiting to be scheduled for long periods
of time in order to achieve serialized execution.
This patch disables implicit synchronization
This allows us to pass extra parameters to the memory allocation
operation that are not defined in the vulkan spec. This is useful for
internal usage.
Signed-off-by: Andres Rodriguez
---
src/amd/vulkan/radv_device.c | 19 ++-
src/amd/vulkan/radv_private.h |
Hi Andres,
The explicit sync should not be used for DrI3 and DRI2 but for cross process
memory sharing, right?
We still have to rely on implicit sync to guarantee the. Correct order of
rendering and present.
Could you confirm?
Thanks.
Sent from my iPhone
On 19 Sep 2017, at 9:57 PM, Andres
Booting with IOMMU disabled in the bios breaks this.
I'll submit a fix in a minute. Should be easy.
Tom
On 19/09/17 11:25 AM, Tom St Denis wrote:
On 19/09/17 11:17 AM, Deucher, Alexander wrote:
-Original Message-
From: amd-gfx [mailto:amd-gfx-boun...@lists.freedesktop.org] On Behalf
On 19/09/17 11:17 AM, Deucher, Alexander wrote:
-Original Message-
From: amd-gfx [mailto:amd-gfx-boun...@lists.freedesktop.org] On Behalf
Of Tom St Denis
Sent: Tuesday, September 19, 2017 8:13 AM
To: amd-gfx@lists.freedesktop.org
Cc: StDenis, Tom
Subject: [PATCH] drm/amd/amdgpu: add
Am 19.09.2017 um 16:22 schrieb Alex Deucher:
It was not clear. The rest of the driver is MIT/X11.
Signed-off-by: Alex Deucher
Reviewed-by: Christian König
---
Dave, any objections?
drivers/gpu/drm/amd/amdgpu/amdgpu_trace_points.c |
On Tue, Sep 19, 2017 at 2:41 AM, Monk Liu wrote:
> previously a patch has typo error, correct it
Please squash this into the previous broken patch before committing it.
Alex
>
> Change-Id: I91bcefd7148b5db1c7d957c868e13a46ca40ef74
> Signed-off-by: Monk Liu
It was not clear. The rest of the driver is MIT/X11.
Signed-off-by: Alex Deucher
---
Dave, any objections?
drivers/gpu/drm/amd/amdgpu/amdgpu_trace_points.c | 19 +++
1 file changed, 19 insertions(+)
diff --git
Correct.
The idea is to only set AMDGPU_GEM_CREATE_EXPLICIT_SYNC for buffers that
are not associated with dri2/3 or PRIME.
Regards,
Andres
On 2017-09-19 10:10 AM, Mao, David wrote:
Hi Andres,
The explicit sync should not be used for DrI3 and DRI2 but for cross
process memory sharing,
On Tue, Sep 19, 2017 at 4:30 AM, Christian König
wrote:
> I don't know why, but the HDP is generally not part of the register spec.
>
> So you can neither find HDP_DEBUG0 nor HDP_READ_CACHE_INVALIDATE in it as
> far as I know.
>
> Point is that the HDP invalidates it's
Introduce a flag to signal that access to a BO will be synchronized
through an external mechanism.
Currently all buffers shared between contexts are subject to implicit
synchronization. However, this is only required for protocols that
currently don't support an explicit synchronization mechanism
Am 19.09.2017 um 14:13 schrieb Tom St Denis:
Signed-off-by: Tom St Denis
Reviewed-by: Christian König
(v2): Add domain to iova debugfs
(v3): Add true read/write methods to access system memory of pages
mapped to the device
(v4): Move
Signed-off-by: Tom St Denis
(v2): Add domain to iova debugfs
(v3): Add true read/write methods to access system memory of pages
mapped to the device
(v4): Move get_domain call out of loop and return on error
(v5): Just use kmap/kunmap
---
Am 19.09.2017 um 03:05 schrieb Felix Kuehling:
Thanks for the reviews. I rebased this on amd-staging-drm-next, retested
and submitted.
Christian, do you want to do some graphics PASID and VMFault work on top
of that? I think I'll be working on more KFD upstreaming this week and
maybe look at
Am 19.09.2017 um 08:41 schrieb Monk Liu:
previously a patch has typo error, correct it
Change-Id: I91bcefd7148b5db1c7d957c868e13a46ca40ef74
Signed-off-by: Monk Liu
Acked-by: Christian König
---
drivers/gpu/drm/amd/amdgpu/amdgpu_psp.c | 2 +-
Am 19.09.2017 um 08:41 schrieb Monk Liu:
otherwise a gpu hang will make application couldn't be killed
under timedout=0 mode
v2:
Fix memoryleak job/job->s_fence issue
unlock mn
remove the ERROR msg after waiting being interrupted
Change-Id: I6051b5b3ae1188983f49325a2438c84a6c12374a
Am 19.09.2017 um 13:20 schrieb Tom St Denis:
Signed-off-by: Tom St Denis
(v2): Add domain to iova debugfs
(v3): Add true read/write methods to access system memory of pages
mapped to the device
(v4): Move get_domain call out of loop and return on error
---
> Well the question is why does the CPC still needs the MQD commands and how
> can we prevent that?
You are right, I'll see what can do
-Original Message-
From: Christian König [mailto:ckoenig.leichtzumer...@gmail.com]
Sent: 2017年9月19日 16:27
To: Liu, Monk ; Koenig,
It actually looks like a valid cleanup to me.
Patch is Reviewed-by: Christian König .
Regards,
Christian.
Am 19.09.2017 um 05:20 schrieb Andres Rodriguez:
This is a small cleanup patch from my initial naive attempt at
extracting a TTM bo in amdgpu_sync_resv(). It
Am 18.09.2017 um 21:18 schrieb Yong Zhao:
When max_bytes is not 8 bytes aligned and bo size is larger than
max_bytes, the last 8 bytes in a ttm node may be left unchanged.
For example, on pre SDMA 4.0, max_bytes = 0x1f, and the bo size
is 0x20, the problem will happen.
In order to fix
Am 18.09.2017 um 19:33 schrieb Tom St Denis:
Signed-off-by: Tom St Denis
Mhm, I sometimes have good use for those. But just adding a printk at
the right place does the job as well.
So patch is Reviewed-by: Christian König .
Regards,
Am 18.09.2017 um 19:33 schrieb Tom St Denis:
Signed-off-by: Tom St Denis
Reviewed-by: Christian König
(v2): add domains and avoid strcmp
---
drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.c | 54 ++---
On 09/19/2017 04:30 PM, Christian König wrote:
I don't know why, but the HDP is generally not part of the register spec.
AFAIW, these regs may be used for HW guys to debug some special cases.
Usually they are not suggested to touch formally.
(e.g. GFX cannot access PRT unmap range, but with
drm/amdgpu: simplify trace point code
amdgpu_trace_points.c does not declare license and it can be replaced by a
simpler way.
Signed-off-by: Ken Xue
---
drivers/gpu/drm/amd/amdgpu/Makefile | 2 +-
drivers/gpu/drm/amd/amdgpu/amdgpu_device.c | 2 ++
I don't know why, but the HDP is generally not part of the register spec.
So you can neither find HDP_DEBUG0 nor HDP_READ_CACHE_INVALIDATE in it
as far as I know.
Point is that the HDP invalidates it's read cache on any register write
(the register itself doesn't matter). So far we used the
Am 19.09.2017 um 06:14 schrieb Liu, Monk:
Christian,
That sounds at least a bit better. But my question is why doesn't this work
like it does on Tonga, e.g. correctly clean things up?
Tonga also suffer with this issue, just that we fixed it in the branch for CSP
customer and staging code
>using this register to replace MM_HDP_DEBUG0 is suggested from a HDP
HW guys
I'm OK with this line.
Thanks for explain.
David Zhou
On 2017年09月19日 15:00, Liu, Monk wrote:
First, I didn't check if windows did this way or not, because I don't sure if
windows is always doing the right thing,
First, I didn't check if windows did this way or not, because I don't sure if
windows is always doing the right thing, e.g. for GFX preemption I didn't copy
windows scheme and we found couple bugs in windows but not in linux ...
So please don't assume we should copy from windows, unless it's
What question ? please reply here
-Original Message-
From: Zhou, David(ChunMing)
Sent: 2017年9月19日 12:25
To: Liu, Monk ; Koenig, Christian ;
amd-gfx@lists.freedesktop.org
Subject: RE: [PATCH 12/18] drm/amdgpu:use formal register to trigger hdp
previously a patch has typo error, correct it
Change-Id: I91bcefd7148b5db1c7d957c868e13a46ca40ef74
Signed-off-by: Monk Liu
---
drivers/gpu/drm/amd/amdgpu/amdgpu_psp.c | 2 +-
1 file changed, 1 insertion(+), 1 deletion(-)
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_psp.c
GPU reset will require all hw doing hw_init thus
ucode_init_bo will be invoked again, which lead to
memory leak
skip the fw_buf allocation during sriov gpu reset to avoid
memory leak.
Change-Id: I31131eda1bd45ea2f5bdc50c5da5fc5a9fe9027d
Signed-off-by: Monk Liu
---
otherwise a gpu hang will make application couldn't be killed
under timedout=0 mode
v2:
Fix memoryleak job/job->s_fence issue
unlock mn
remove the ERROR msg after waiting being interrupted
Change-Id: I6051b5b3ae1188983f49325a2438c84a6c12374a
Signed-off-by: Monk Liu
---
54 matches
Mail list logo