On 2017年11月14日 20:00, Christian König wrote:
Am 14.11.2017 um 12:52 schrieb Christian König:
Am 14.11.2017 um 10:07 schrieb Monk Liu:
this can fix the memory leak under the case that not all
BO are freed during "takedown" stage, because originally
it blocks following kfree on mgr.
NAK, all
Can you try this patch ? I noticed there is a new API to use to wait for
flips instead of what we are using now.
Thanks,
Andrey
On 2017-11-14 01:06 PM, Michel Dänzer wrote:
I sometimes get the attached dmesg splat when the GDM login screen times
out and fades to black (presumably specifically
Fixes distorted colors on some cards on resume from suspend.
This reverts commit b9729b17a414f99c61f4db9ac9f9ed987fa0cbfe.
Bug: https://bugs.freedesktop.org/show_bug.cgi?id=98832
Signed-off-by: Alex Deucher
Cc: sta...@vger.kernel.org
---
drivers/gpu/drm/radeon/radeon_fb.c | 1 -
1 file changed,
A second-level user mode trap handler can be installed. The CWSR trap
handler jumps to the secondary trap handler conditionally for any
conditions not handled by it. This can be used e.g. for debugging or
catching math exceptions.
When CWSR is disabled, the user mode trap handler is installed as
f
The trap handler is like an interrupt handler running on the GPU
compute unit. It is needed for supporting CWSR (compute wave
save/restore).
This file defines an array with the pre-compiled GFXv8 shader ISA.
The assembly code is included for reference in #if 0 ... #endif.
Signed-off-by: Shaoyun.l
This hardware feature allows the GPU to preempt shader execution in
the middle of a compute wave, save the state and restore it later
to resume execution.
Memory for saving the state is allocated per queue in user mode and
the address and size passed to the create_queue ioctl. The size
depends on
The PQM doesn't change after process creation. So initialize it in
kfd_create_process_device_data.
Signed-off-by: Felix Kuehling
---
drivers/gpu/drm/amd/amdkfd/kfd_process.c | 1 +
drivers/gpu/drm/amd/amdkfd/kfd_process_queue_manager.c | 4 +---
2 files changed, 2 insertions(+), 3
Hi,
might this be related to this bug I reportet some time ago?
https://bugs.freedesktop.org/show_bug.cgi?id=103489
I currently work around it be reverting the following commits:
8459f633b65a81979b9b233ed0a91a9fdc98b7bd
c196cbe065da1c9b5f7683af9cdd0889760151c1
e6f4f9cfde0f586c276190f541ecfc80d
On 11/14/2017 01:06 PM, Michel Dänzer wrote:
I sometimes get the attached dmesg splat when the GDM login screen times
out and fades to black (presumably specifically when it finally sets
DPMS off, which disables the vblank interrupt).
As you mention this and also that this happens intermitten
On 2017-11-14 12:57 PM, Michel Dänzer wrote:
> From: Michel Dänzer
>
> The main motivation was that it was at least theoretically possible for
> stream_for_cursor to be dereferenced when it was NULL.
>
> The code also becomes slightly smaller, and hopefully easier to
> understand.
>
Thanks. We
Michel,
Would you mind attaching the full dmesg? We'll have a look.
Thanks,
Jordan
-Original Message-
From: Michel Dänzer [mailto:mic...@daenzer.net]
Sent: Tuesday, November 14, 2017 1:06 PM
To: amd-gfx@lists.freedesktop.org
Cc: dc_upstream
Subject: [DC] [CRTC:44:crtc-0] vblank wait ti
DC always work off current context. If we don't swap in a context with plane
removed, HWSS will be doing the wrong programming when we enable streams and
planes.
-Original Message-
From: Grodzovsky, Andrey
Sent: Tuesday, November 14, 2017 10:08 AM
To: Li, Sun peng (Leo) ; Wentland, Har
I sometimes get the attached dmesg splat when the GDM login screen times
out and fades to black (presumably specifically when it finally sets
DPMS off, which disables the vblank interrupt).
I've seen this for the first time today, so it might have been
introduced by one of the DC changes which la
From: Michel Dänzer
The main motivation was that it was at least theoretically possible for
stream_for_cursor to be dereferenced when it was NULL.
The code also becomes slightly smaller, and hopefully easier to
understand.
Signed-off-by: Michel Dänzer
---
.../drm/amd/display/dc/dcn10/dcn10_hw
> -Original Message-
> From: amd-gfx [mailto:amd-gfx-boun...@lists.freedesktop.org] On Behalf
> Of Michel Dänzer
> Sent: Tuesday, November 14, 2017 12:52 PM
> To: amd-gfx@lists.freedesktop.org
> Subject: [PATCH] amdgpu: Don't use DRM_ERROR when failing to allocate a
> BO
>
> From: Michel D
From: Michel Dänzer
This can be triggered by userspace, e.g. trying to allocate too large a
BO, so it shouldn't log anything by default.
Callers need to handle failure anyway.
Signed-off-by: Michel Dänzer
---
drivers/gpu/drm/amd/amdgpu/amdgpu_gem.c| 2 +-
drivers/gpu/drm/amd/amdgpu/amdgpu
On 11/14/2017 11:00 AM, Leo Li wrote:
On 2017-11-10 01:40 PM, Andrey Grodzovsky wrote:
On 11/10/2017 01:38 PM, Andrey Grodzovsky wrote:
On 11/09/2017 03:05 PM, Harry Wentland wrote:
From: "Leo (Sunpeng) Li"
This is a followup to the following revert:
Rex ZhuRevert "drm/amd/displ
On 2017-11-10 01:40 PM, Andrey Grodzovsky wrote:
On 11/10/2017 01:38 PM, Andrey Grodzovsky wrote:
On 11/09/2017 03:05 PM, Harry Wentland wrote:
From: "Leo (Sunpeng) Li"
This is a followup to the following revert:
Rex Zhu Revert "drm/amd/display: Match actual state during S3
On Tue, Nov 14, 2017 at 7:13 AM, Christian König
wrote:
> Otherwise we can't correctly CPU map TTM buffers.
>
> Signed-off-by: Christian König
Acked-by: Alex Deucher
> ---
> drivers/gpu/drm/amd/amdgpu/amdgpu_prime.c | 6 +-
> 1 file changed, 5 insertions(+), 1 deletion(-)
>
> diff --git a
Thought so, in this case the patch is Reviewed-by: Christian König
.
Regards,
Christian.
Am 14.11.2017 um 16:02 schrieb Liu, Monk:
Yeah CSA is allocated right after in GMC hw init (and after WB_INIT)
So we can put the free of it right before WB_INIT, thus avoiding doing it in
each GFX IP's sw
Am 14.11.2017 um 16:22 schrieb Oak Zeng:
From: ozeng
v1: Properly allocate TLB invalidation engine to avoid conflict.
v2: Added comments to codes
Change-Id: I9a32cbcb133c07d7efad1a7bd07c6e2098c747e7
Signed-off-by: Oak Zeng
Reviewed-by: Christian König
---
drivers/gpu/drm/amd/amdgpu/gmc
Makes sense, Reviewed-by: Christian König fro
this one as well.
Regards,
Christian.
Am 14.11.2017 um 16:00 schrieb Liu, Monk:
Actually " drm_mm_takedown" below will give you the warning if some BO is still
in MM
That means below part is totally useless:
-spin_lock(&mgr->lock);
-
On 14/11/17 03:37 PM, Tom St Denis wrote:
> On 14/11/17 09:36 AM, Harry Wentland wrote:
>> On 2017-11-14 08:30 AM, Tom St Denis wrote:
>>> Hi all,
>>>
>>> Found when testing the tip of drm-next on my A12-9800 Carrizo.
>>>
>>> [root@carrizo linux]# git bisect good
>>> 138a3358c17918bb5cd5aa1ea9e8760
On 14/11/17 01:18 PM, Christian König wrote:
> Am 14.11.2017 um 11:02 schrieb Michel Dänzer:
>> On 13/11/17 10:54 AM, Christian König wrote:
>>>
>>> + } else {
>>> + locked = reservation_object_trylock(bo->resv);
>>> + if (!locked)
>>> + c
On Tue, Nov 14, 2017 at 10:22 AM, Oak Zeng wrote:
> From: ozeng
>
> v1: Properly allocate TLB invalidation engine to avoid conflict.
> v2: Added comments to codes
>
> Change-Id: I9a32cbcb133c07d7efad1a7bd07c6e2098c747e7
> Signed-off-by: Oak Zeng
Reviewed-by: Alex Deucher
> ---
> drivers/gpu/
From: ozeng
v1: Properly allocate TLB invalidation engine to avoid conflict.
v2: Added comments to codes
Change-Id: I9a32cbcb133c07d7efad1a7bd07c6e2098c747e7
Signed-off-by: Oak Zeng
---
drivers/gpu/drm/amd/amdgpu/gmc_v9_0.c | 15 ---
1 file changed, 12 insertions(+), 3 deletions(-)
On 2017-11-10 04:14 AM, S, Shirish wrote:
> From: Shirish S
>
> This patch fixes static checker warning of
> "warn: cast after binop" introduced by
> 56087b31 drm/amd/display: fix high part address in
> dm_plane_helper_prepare_fb()
>
> Signed-off-by: Shirish S
Reviewed-by: Harry Wentland
Fe
Ok, got it.
Thanks,
Andrey
On 11/14/2017 10:10 AM, Cheng, Tony wrote:
DC always work off current context. If we don't swap in a context with plane
removed, HWSS will be doing the wrong programming when we enable streams and
planes.
-Original Message-
From: Grodzovsky, Andrey
Sent: T
On 11/13/2017 04:53 PM, Leo Li wrote:
On 2017-11-10 02:00 PM, Andrey Grodzovsky wrote:
On 11/09/2017 03:05 PM, Harry Wentland wrote:
From: "Leo (Sunpeng) Li"
When disabling pipe splitting, we need to make sure we disable both
planes used.
This should be done for Linux as well.
Change-
Yeah CSA is allocated right after in GMC hw init (and after WB_INIT)
So we can put the free of it right before WB_INIT, thus avoiding doing it in
each GFX IP's sw_init
-Original Message-
From: Christian König [mailto:ckoenig.leichtzumer...@gmail.com]
Sent: 2017年11月14日 19:55
To: Liu, Mo
Actually " drm_mm_takedown" below will give you the warning if some BO is still
in MM
That means below part is totally useless:
>> -spin_lock(&mgr->lock);
>> -if (!drm_mm_clean(&mgr->mm)) {
>> -spin_unlock(&mgr->lock);
>> -return -EBUSY;
>> -}
>> -
And the worse thi
ok
-Original Message-
From: Christian König [mailto:ckoenig.leichtzumer...@gmail.com]
Sent: 2017年11月14日 19:49
To: Liu, Monk ; amd-gfx@lists.freedesktop.org
Subject: Re: [PATCH 03/10] drm/amdgpu:fix NULL pointer access during drv remove
Am 14.11.2017 um 10:07 schrieb Monk Liu:
> NULL poin
On 2017-11-14 08:30 AM, Tom St Denis wrote:
> Hi all,
>
> Found when testing the tip of drm-next on my A12-9800 Carrizo.
>
> [root@carrizo linux]# git bisect good
> 138a3358c17918bb5cd5aa1ea9e8760ac69c515d is the first bad commit
> commit 138a3358c17918bb5cd5aa1ea9e8760ac69c515d
> Author: Yongqia
Quoting Chris Wilson (2017-11-14 14:39:29)
> Quoting Christian König (2017-11-14 14:24:35)
> > The amdgpu issue to also need signaled fences in the reservation objects
> > should be fixed by now.
> >
> > Optimize the list by keeping only the not signaled yet fences around.
> >
> > v2: temporary p
Quoting Christian König (2017-11-14 14:24:36)
> The amdgpu issue to also need signaled fences in the reservation objects
> should
> be fixed by now.
>
> Optimize the handling by replacing a signaled fence when adding a new
> shared one.
>
> Signed-off-by: Christian König
Reviewed-by: Chris Wil
On 2017-11-14 09:33 AM, Michel Dänzer wrote:
> On 14/11/17 01:29 PM, Tom St Denis wrote:
>> Hi Harry,
>>
>> At display/dc/dcn10/dcn10_hw_sequencer.c:2140
>>
>> if (num_planes > 0) {
>> struct dc_stream_state *stream_for_cursor;
>>
>> program_all_pipe_in_tree(dc, top_pipe_to_prog
Quoting Christian König (2017-11-14 14:24:35)
> The amdgpu issue to also need signaled fences in the reservation objects
> should be fixed by now.
>
> Optimize the list by keeping only the not signaled yet fences around.
>
> v2: temporary put the signaled fences at the end of the new container
>
On 14/11/17 09:36 AM, Harry Wentland wrote:
On 2017-11-14 08:30 AM, Tom St Denis wrote:
Hi all,
Found when testing the tip of drm-next on my A12-9800 Carrizo.
[root@carrizo linux]# git bisect good
138a3358c17918bb5cd5aa1ea9e8760ac69c515d is the first bad commit
commit 138a3358c17918bb5cd5aa1ea
On 2017-11-14 07:29 AM, Tom St Denis wrote:
> Hi Harry,
>
> At display/dc/dcn10/dcn10_hw_sequencer.c:2140
>
> if (num_planes > 0) {
> struct dc_stream_state *stream_for_cursor;
>
> program_all_pipe_in_tree(dc, top_pipe_to_program, context);
>
> for (i = 0; i < dc->re
Quoting Christian König (2017-11-14 14:24:44)
> Am 06.11.2017 um 17:22 schrieb Chris Wilson:
> > Quoting Christian König (2017-10-30 14:59:04)
> >> From: Christian König
> >>
> >> The amdgpu issue to also need signaled fences in the reservation objects
> >> should
> >> be fixed by now.
> >>
> >>
On 14/11/17 01:29 PM, Tom St Denis wrote:
> Hi Harry,
>
> At display/dc/dcn10/dcn10_hw_sequencer.c:2140
>
> if (num_planes > 0) {
> struct dc_stream_state *stream_for_cursor;
>
> program_all_pipe_in_tree(dc, top_pipe_to_program, context);
>
> for (i = 0; i < dc->res_
Am 06.11.2017 um 17:22 schrieb Chris Wilson:
Quoting Christian König (2017-10-30 14:59:04)
From: Christian König
The amdgpu issue to also need signaled fences in the reservation objects should
be fixed by now.
Optimize the handling by replacing a signaled fence when adding a new
shared one.
The amdgpu issue to also need signaled fences in the reservation objects
should be fixed by now.
Optimize the list by keeping only the not signaled yet fences around.
v2: temporary put the signaled fences at the end of the new container
v3: put the old fence at the end of the new container as wel
The amdgpu issue to also need signaled fences in the reservation objects should
be fixed by now.
Optimize the handling by replacing a signaled fence when adding a new
shared one.
Signed-off-by: Christian König
---
drivers/dma-buf/reservation.c | 18 +++---
1 file changed, 15 inserti
The suite stalls the CP, until RCA is done the suite is
disabled to not disrupt regression testing.
Signed-off-by: Andrey Grodzovsky
---
tests/amdgpu/amdgpu_test.c| 2 +-
tests/amdgpu/amdgpu_test.h| 5 +
tests/amdgpu/deadlock_tests.c | 19 +++
3 files changed, 25 in
Hi all,
Found when testing the tip of drm-next on my A12-9800 Carrizo.
[root@carrizo linux]# git bisect good
138a3358c17918bb5cd5aa1ea9e8760ac69c515d is the first bad commit
commit 138a3358c17918bb5cd5aa1ea9e8760ac69c515d
Author: Yongqiang Sun
Date: Tue Nov 7 11:01:34 2017 -0500
drm/amd/
Am 14.11.2017 um 13:53 schrieb Andrey Grodzovsky:
On 11/14/2017 03:44 AM, Christian König wrote:
Am 13.11.2017 um 18:01 schrieb Andrey Grodzovsky:
Allocates 1 TB of memory. Test is disabled by default
since it's triggers OOM killer.
v2:
FIx the test to only alloc the BO and assert if return
On 11/14/2017 03:44 AM, Christian König wrote:
Am 13.11.2017 um 18:01 schrieb Andrey Grodzovsky:
Allocates 1 TB of memory. Test is disabled by default
since it's triggers OOM killer.
v2:
FIx the test to only alloc the BO and assert if return value
not equal to -ENOMEM and remove test disable
Hi Harry,
At display/dc/dcn10/dcn10_hw_sequencer.c:2140
if (num_planes > 0) {
struct dc_stream_state *stream_for_cursor;
program_all_pipe_in_tree(dc, top_pipe_to_program, context);
for (i = 0; i < dc->res_pool->pipe_count; i++) {
Am 14.11.2017 um 11:02 schrieb Michel Dänzer:
On 13/11/17 10:54 AM, Christian König wrote:
Deleted BOs with the same reservation object can be reaped even if they
can't be reserved.
v2: rebase and we still need to remove/add the BO from/to the LRU.
v3: fix remove/add one more time, cleanup the
Otherwise we can't correctly CPU map TTM buffers.
Signed-off-by: Christian König
---
drivers/gpu/drm/amd/amdgpu/amdgpu_prime.c | 6 +-
1 file changed, 5 insertions(+), 1 deletion(-)
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_prime.c
b/drivers/gpu/drm/amd/amdgpu/amdgpu_prime.c
index 90a
Am 14.11.2017 um 12:52 schrieb Christian König:
Am 14.11.2017 um 10:07 schrieb Monk Liu:
this can fix the memory leak under the case that not all
BO are freed during "takedown" stage, because originally
it blocks following kfree on mgr.
NAK, all BOs must be freed before we take down at least t
Am 14.11.2017 um 10:07 schrieb Monk Liu:
Change-Id: I984697748672421a05f179a7bf4eeb0255b28bdc
Signed-off-by: Monk Liu
Acked-by: Christian König
---
drivers/gpu/drm/amd/amdgpu/amdgpu_device.c | 3 ++-
1 file changed, 2 insertions(+), 1 deletion(-)
diff --git a/drivers/gpu/drm/amd/amdgpu/
Am 14.11.2017 um 10:07 schrieb Monk Liu:
those RLC used buffers are not cleared in GFX's sw_fini
Change-Id: I0c9204753caa327cfbabdfb1465d2c03012ee977
Signed-off-by: Monk Liu
Reviewed-by: Christian König
---
drivers/gpu/drm/amd/amdgpu/gfx_v7_0.c | 8
drivers/gpu/drm/amd/amdgpu/g
Am 14.11.2017 um 10:07 schrieb Monk Liu:
use bo_create/free_kernel instead of manually doing it
Change-Id: I128651d2b2207f6e16b484b050f3232c08b53de3
Signed-off-by: Monk Liu
Reviewed-by: Christian König
---
drivers/gpu/drm/amd/amdgpu/amdgpu_ucode.c | 43 +++
1
Am 14.11.2017 um 10:07 schrieb Monk Liu:
instead of doing it in each GFX ip's sw_fini
Mhm, is that allocated in the same way as well?
Change-Id: Idf0fd500d4fc385cf7a930cc56305070c937bf20
Signed-off-by: Monk Liu
---
drivers/gpu/drm/amd/amdgpu/amdgpu_device.c | 1 +
drivers/gpu/drm/amd/amd
Am 14.11.2017 um 10:07 schrieb Monk Liu:
Change-Id: I5d3ea5facc23bcc94806a098ae9455ce1564fa09
Signed-off-by: Monk Liu
Reviewed-by: Christian König
---
drivers/gpu/drm/amd/amdgpu/amdgpu_device.c | 3 +--
1 file changed, 1 insertion(+), 2 deletions(-)
diff --git a/drivers/gpu/drm/amd/amdg
Am 14.11.2017 um 10:07 schrieb Monk Liu:
this can fix the memory leak under the case that not all
BO are freed during "takedown" stage, because originally
it blocks following kfree on mgr.
NAK, all BOs must be freed before we take down at least the GTT manager.
That we run into a memory leak i
Am 14.11.2017 um 10:07 schrieb Monk Liu:
Move "reserve_fw_vram_fini" to the correct place
which should in amdgpu_ttm_fini() since reserve_fw_vram_init is
in amdgpu_ttm_init().
this can fix error report and memory leak in "vram_mgr_fini()".
Change-Id: I4b1dc7eecda62407f30a5a16f59555eee49f04a3
Si
Am 14.11.2017 um 10:07 schrieb Monk Liu:
NULL pointer is because original logic will step into
set_pde_pte() even after the gart.ptr is freed due to
there are twice gart_unbind() on all gart area.
also, there are other minor fixes:
1,since gart_init only create dummy page, the corresponding
gart
Am 14.11.2017 um 10:07 schrieb Monk Liu:
Change-Id: Id641e125f3dafc54223e49ee444ab64249e7e3a1
Signed-off-by: Monk Liu
---
drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.c | 11 +++
1 file changed, 3 insertions(+), 8 deletions(-)
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.c
b/drivers/
Am 14.11.2017 um 10:07 schrieb Monk Liu:
for gart_ram_alloc/free, they are never used in driver thus
ripe them out totally.
for gart_vram_pin/unpin, they are not needed becuase we can
use bo_creat_kernel/free to replace the original manual way
in the gart_vram_alloc/free, thus gart_vram_pin/unpi
Am 14.11.2017 um 03:30 schrieb Chunming Zhou:
On 2017年11月14日 01:22, Christian König wrote:
Am 13.11.2017 um 17:12 schrieb Oak Zeng:
From: ozeng
Properly allocate TLB invalidation engine to avoid conflict.
The latest engine allocation on gfx9 is:
Engine 0, 1: idle
Engine 2, 3: firmware
Engin
On 13/11/17 10:54 AM, Christian König wrote:
> Needed for the next patch.
>
> v2: actually predicate all unlocks
>
> Signed-off-by: Christian König
[...]
> @@ -531,24 +533,28 @@ static int ttm_bo_cleanup_refs_and_unlock(struct
> ttm_buffer_object *bo,
> return -EBUSY;
>
On 13/11/17 10:54 AM, Christian König wrote:
> Deleted BOs with the same reservation object can be reaped even if they
> can't be reserved.
>
> v2: rebase and we still need to remove/add the BO from/to the LRU.
> v3: fix remove/add one more time, cleanup the logic a bit
> v4: we should still check
those RLC used buffers are not cleared in GFX's sw_fini
Change-Id: I0c9204753caa327cfbabdfb1465d2c03012ee977
Signed-off-by: Monk Liu
---
drivers/gpu/drm/amd/amdgpu/gfx_v7_0.c | 8
drivers/gpu/drm/amd/amdgpu/gfx_v8_0.c | 9 +
drivers/gpu/drm/amd/amdgpu/gfx_v9_0.c | 8
3
Change-Id: I984697748672421a05f179a7bf4eeb0255b28bdc
Signed-off-by: Monk Liu
---
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.c
b/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c
index ccb33ef.
use bo_create/free_kernel instead of manually doing it
Change-Id: I128651d2b2207f6e16b484b050f3232c08b53de3
Signed-off-by: Monk Liu
---
drivers/gpu/drm/amd/amdgpu/amdgpu_ucode.c | 43 +++
1 file changed, 9 insertions(+), 34 deletions(-)
diff --git a/drivers/gpu/drm/a
Change-Id: I5d3ea5facc23bcc94806a098ae9455ce1564fa09
Signed-off-by: Monk Liu
---
drivers/gpu/drm/amd/amdgpu/amdgpu_device.c | 3 +--
1 file changed, 1 insertion(+), 2 deletions(-)
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c
b/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c
index 9471f47.
instead of doing it in each GFX ip's sw_fini
Change-Id: Idf0fd500d4fc385cf7a930cc56305070c937bf20
Signed-off-by: Monk Liu
---
drivers/gpu/drm/amd/amdgpu/amdgpu_device.c | 1 +
drivers/gpu/drm/amd/amdgpu/amdgpu_virt.c | 6 ++
drivers/gpu/drm/amd/amdgpu/amdgpu_virt.h | 1 +
drivers/gpu/drm
this can fix the memory leak under the case that not all
BO are freed during "takedown" stage, because originally
it blocks following kfree on mgr.
Change-Id: I56e832ac317215a4d9f58edb8e75ea722158715c
Signed-off-by: Monk Liu
---
drivers/gpu/drm/amd/amdgpu/amdgpu_gtt_mgr.c | 6 --
drivers/gp
NULL pointer is because original logic will step into
set_pde_pte() even after the gart.ptr is freed due to
there are twice gart_unbind() on all gart area.
also, there are other minor fixes:
1,since gart_init only create dummy page, the corresponding
gart_fini shouldn't do more like unbinding all
Move "reserve_fw_vram_fini" to the correct place
which should in amdgpu_ttm_fini() since reserve_fw_vram_init is
in amdgpu_ttm_init().
this can fix error report and memory leak in "vram_mgr_fini()".
Change-Id: I4b1dc7eecda62407f30a5a16f59555eee49f04a3
Signed-off-by: Monk Liu
---
drivers/gpu/drm
patch 3 is to fix kernel driver cannot removed bug (new regression)
patch 1,2,6,7,8,10 are the cleanups to make code more readable and simple
patch 4,5,9 are the fixes for memory leak bug after driver unloaded
Monk Liu (10):
drm/amdgpu:cleanup stolen vga memory finish
drm/amdgpu:cleanup GMC
Change-Id: Id641e125f3dafc54223e49ee444ab64249e7e3a1
Signed-off-by: Monk Liu
---
drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.c | 11 +++
1 file changed, 3 insertions(+), 8 deletions(-)
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.c
b/drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.c
index 99e812d
for gart_ram_alloc/free, they are never used in driver thus
ripe them out totally.
for gart_vram_pin/unpin, they are not needed becuase we can
use bo_creat_kernel/free to replace the original manual way
in the gart_vram_alloc/free, thus gart_vram_pin/unpin can
also be riped out.
Change-Id: Ic6094
Am 14.11.2017 um 03:40 schrieb Chunming Zhou:
On 2017年11月13日 17:54, Christian König wrote:
Deleted BOs with the same reservation object can be reaped even if they
can't be reserved.
v2: rebase and we still need to remove/add the BO from/to the LRU.
v3: fix remove/add one more time, cleanup th
Am 13.11.2017 um 23:34 schrieb Felix Kuehling:
On 2017-11-10 07:02 AM, Christian König wrote:
Am 10.11.2017 um 12:33 schrieb Roger He:
since sometimes VRAM size is bigger than system memory
Change-Id: I5b14d18ed7a9f79810cc50c023ac9e240bddf101
Signed-off-by: Roger He
---
drivers/gpu/drm/amd
Am 13.11.2017 um 22:09 schrieb Alex Deucher:
We did this for gfx6 and 8, but somehow missed gfx7.
Signed-off-by: Alex Deucher
This one is Reviewed-by: Christian König , the
rest of the series is Acked-by: Christian König .
Regards,
Christian.
---
drivers/gpu/drm/amd/amdgpu/gfx_v7_0.c |
Am 13.11.2017 um 18:01 schrieb Andrey Grodzovsky:
Allocates 1 TB of memory. Test is disabled by default
since it's triggers OOM killer.
v2:
FIx the test to only alloc the BO and assert if return value
not equal to -ENOMEM and remove test disable on start.
Signed-off-by: Andrey Grodzovsky
I w
80 matches
Mail list logo