Skip to contentSkip to conversation

drm/amd/display: use GFP_ATOMIC for bounding box

PATCH In review Alex Deucher 4 participants 5 messages Jun 4, 2024 – Jun 5, 2024

Subsystems (3)
lore
Share & tools
lore

Alex Deucher

author maintainer
#0 lore raw

PATCH 1/1 drm/amd/display: use GFP_ATOMIC for bounding box

This can be called in atomic context. Should fix:

BUG: sleeping function called from invalid context at include/linux/sched/mm.h:306

in_atomic(): 1, irqs_disabled(): 0, non_block: 0, pid: 449, name: kworker/u64:8
preempt_count: 2, expected: 0
RCU nest depth: 0, expected: 0
Preemption disabled at:
ffffffffc0ce1580>] dc_fpu_begin+0x30/0xd0 [amdgpu]

CPU: 5 PID: 449 Comm: kworker/u64:8 Tainted: G        W          6.8.0+ #35
Hardware name: System manufacturer System Product Name/ROG STRIX X570-E GAMING WIFI II, BIOS 4204 02/24/2022
Workqueue: events_unbound async_run_entry_fn
Fixes 88c61827cedc ("drm/amd/display: dynamically allocate dml2_configuration_options structures")
Cc George Zhang <[email protected]>
Cc Arnd Bergmann <[email protected]>
---
 drivers/gpu/drm/amd/display/dc/resource/dcn32/dcn32_resource.c  | 2 +-
 .../gpu/drm/amd/display/dc/resource/dcn321/dcn321_resource.c    | 2 +-
 2 files changed, 2 insertions(+), 2 deletions(-)
drivers/gpu/drm/amd/display/dc/resource/dcn32/dcn32_resource.c +1 -1
index 0f11d7c8791c..3fe0d5334145 100644
@@ -2009,7 +2009,7 @@ static void dcn32_update_bw_bounding_box(struct dc *dc, struct clk_bw_params *bw
 {
 	struct dml2_configuration_options *dml2_opt;
 
-	dml2_opt = kmemdup(&dc->dml2_options, sizeof(dc->dml2_options), GFP_KERNEL);
+	dml2_opt = kmemdup(&dc->dml2_options, sizeof(dc->dml2_options), GFP_ATOMIC);
 	if (!dml2_opt)
 		return;
 
drivers/gpu/drm/amd/display/dc/resource/dcn321/dcn321_resource.c +1 -1
index 07ca6f58447d..a61cf5741275 100644
@@ -1583,7 +1583,7 @@ static void dcn321_update_bw_bounding_box(struct dc *dc, struct clk_bw_params *b
 {
 	struct dml2_configuration_options *dml2_opt;
 
-	dml2_opt = kmemdup(&dc->dml2_options, sizeof(dc->dml2_options), GFP_KERNEL);
+	dml2_opt = kmemdup(&dc->dml2_options, sizeof(dc->dml2_options), GFP_ATOMIC);
 	if (!dml2_opt)
 		return;
 
Signature · 1 line
2.45.1
#1 lore raw

Am 04.06.24 um 15:50 schrieb Alex Deucher:

Quoted 11 lines

This can be called in atomic context. Should fix:

BUG: sleeping function called from invalid context at include/linux/sched/mm.h:306

in_atomic(): 1, irqs_disabled(): 0, non_block: 0, pid: 449, name: kworker/u64:8
preempt_count: 2, expected: 0
RCU nest depth: 0, expected: 0
Preemption disabled at:
ffffffffc0ce1580>] dc_fpu_begin+0x30/0xd0 [amdgpu]

CPU: 5 PID: 449 Comm: kworker/u64:8 Tainted: G        W          6.8.0+ #35
Hardware name: System manufacturer System Product Name/ROG STRIX X570-E GAMING WIFI II, BIOS 4204 02/24/2022
Workqueue: events_unbound async_run_entry_fn

That most likely only papers over the real problem and is not a valid fix.

The question is why is that an atomic context?  If the function is used under a spinlock then this might indeed be the right fix.

If it's because of floating point operation then that here won't work either.

In that case the only real fix is to avoid the allocation altogether.

Regards,
Christian.

Quoted 39 lines
Fixes 88c61827cedc ("drm/amd/display: dynamically allocate dml2_configuration_options structures")
Cc George Zhang <[email protected]>
Cc Arnd Bergmann <[email protected]>
---
  drivers/gpu/drm/amd/display/dc/resource/dcn32/dcn32_resource.c  | 2 +-
  .../gpu/drm/amd/display/dc/resource/dcn321/dcn321_resource.c    | 2 +-
  2 files changed, 2 insertions(+), 2 deletions(-)
drivers/gpu/drm/amd/display/dc/resource/dcn32/dcn32_resource.c +1 -1
index 0f11d7c8791c..3fe0d5334145 100644
@@ -2009,7 +2009,7 @@ static void dcn32_update_bw_bounding_box(struct dc *dc, struct clk_bw_params *bw
  {
  	struct dml2_configuration_options *dml2_opt;
 
-	dml2_opt = kmemdup(&dc->dml2_options, sizeof(dc->dml2_options), GFP_KERNEL);
+	dml2_opt = kmemdup(&dc->dml2_options, sizeof(dc->dml2_options), GFP_ATOMIC);
  	if (!dml2_opt)
  		return;
 
drivers/gpu/drm/amd/display/dc/resource/dcn321/dcn321_resource.c +1 -1
index 07ca6f58447d..a61cf5741275 100644
@@ -1583,7 +1583,7 @@ static void dcn321_update_bw_bounding_box(struct dc *dc, struct clk_bw_params *b
  {
  	struct dml2_configuration_options *dml2_opt;
 
-	dml2_opt = kmemdup(&dc->dml2_options, sizeof(dc->dml2_options), GFP_KERNEL);
+	dml2_opt = kmemdup(&dc->dml2_options, sizeof(dc->dml2_options), GFP_ATOMIC);
  	if (!dml2_opt)
  		return;
 
#2 lore raw
Quoted 16 lines · Christian König

Am 04.06.24 um 15:50 schrieb Alex Deucher:

Quoted 11 lines

This can be called in atomic context. Should fix:

BUG: sleeping function called from invalid context at include/linux/sched/mm.h:306

in_atomic(): 1, irqs_disabled(): 0, non_block: 0, pid: 449, name: kworker/u64:8
preempt_count: 2, expected: 0
RCU nest depth: 0, expected: 0
Preemption disabled at:
ffffffffc0ce1580>] dc_fpu_begin+0x30/0xd0 [amdgpu]

CPU: 5 PID: 449 Comm: kworker/u64:8 Tainted: G        W          6.8.0+ #35
Hardware name: System manufacturer System Product Name/ROG STRIX X570-E GAMING WIFI II, BIOS 4204 02/24/2022
Workqueue: events_unbound async_run_entry_fn

That most likely only papers over the real problem and is not a valid fix.

Quoted 7 lines

The question is why is that an atomic context? If the function is used under a spinlock then this might indeed be the right fix.

If it's because of floating point operation then that here won't work either.

In that case the only real fix is to avoid the allocation altogether.

Reverting the patch will bring back the stack size warning. I guess the display folks will need to look more closely at this.

Alex

Quoted 44 lines

Regards,
Christian.

Quoted 39 lines
Fixes 88c61827cedc ("drm/amd/display: dynamically allocate dml2_configuration_options structures")
Cc George Zhang <[email protected]>
Cc Arnd Bergmann <[email protected]>
---
  drivers/gpu/drm/amd/display/dc/resource/dcn32/dcn32_resource.c  | 2 +-
  .../gpu/drm/amd/display/dc/resource/dcn321/dcn321_resource.c    | 2 +-
  2 files changed, 2 insertions(+), 2 deletions(-)
drivers/gpu/drm/amd/display/dc/resource/dcn32/dcn32_resource.c +1 -1
index 0f11d7c8791c..3fe0d5334145 100644
@@ -2009,7 +2009,7 @@ static void dcn32_update_bw_bounding_box(struct dc *dc, struct clk_bw_params *bw
  {
      struct dml2_configuration_options *dml2_opt;
 
-     dml2_opt = kmemdup(&dc->dml2_options, sizeof(dc->dml2_options), GFP_KERNEL);
+     dml2_opt = kmemdup(&dc->dml2_options, sizeof(dc->dml2_options), GFP_ATOMIC);
      if (!dml2_opt)
              return;
 
drivers/gpu/drm/amd/display/dc/resource/dcn321/dcn321_resource.c +1 -1
index 07ca6f58447d..a61cf5741275 100644
@@ -1583,7 +1583,7 @@ static void dcn321_update_bw_bounding_box(struct dc *dc, struct clk_bw_params *b
  {
      struct dml2_configuration_options *dml2_opt;
 
-     dml2_opt = kmemdup(&dc->dml2_options, sizeof(dc->dml2_options), GFP_KERNEL);
+     dml2_opt = kmemdup(&dc->dml2_options, sizeof(dc->dml2_options), GFP_ATOMIC);
      if (!dml2_opt)
              return;
 
#3 lore raw
In reply to Christian König
Quoted 13 lines · Christian König

Am 04.06.24 um 15:50 schrieb Alex Deucher:

Quoted 11 lines

This can be called in atomic context. Should fix:

BUG: sleeping function called from invalid context at include/linux/sched/mm.h:306

in_atomic(): 1, irqs_disabled(): 0, non_block: 0, pid: 449, name: kworker/u64:8
preempt_count: 2, expected: 0
RCU nest depth: 0, expected: 0
Preemption disabled at:
ffffffffc0ce1580>] dc_fpu_begin+0x30/0xd0 [amdgpu]

CPU: 5 PID: 449 Comm: kworker/u64:8 Tainted: G        W          6.8.0+ #35
Hardware name: System manufacturer System Product Name/ROG STRIX X570-E GAMING WIFI II, BIOS 4204 02/24/2022
Workqueue: events_unbound async_run_entry_fn
Quoted 7 lines

That most likely only papers over the real problem and is not a valid fix.

The question is why is that an atomic context?  If the function is used under a spinlock then this might indeed be the right fix.

If it's because of floating point operation then that here won't work either.

It looks like there is only one caller, and this turns on floating point instructions just for the call:

        if (dc->res_pool->funcs->update_bw_bounding_box) {
                DC_FP_START();
                dc->res_pool->funcs->update_bw_bounding_box(dc, dc->clk_mgr->bw_params);
                DC_FP_END();
        }

but then they are enabled again inside of the function.

If we can drop the outer DC_FP_START(), that means the GFP_KERNEL allocation works. On the other hand if we actually have to enabled it before calling into the function (e.g. because there is an architecture that has incompatible function calling conventions when FP is enabled), the inner one is redundant, but we can potentially move the kmemdup() into the caller and pass the copy by refernence.

Arnd

#4 lore raw

Am 04.06.24 um 16:57 schrieb Arnd Bergmann:

Quoted 31 lines · Christian König
Quoted 12 lines

Am 04.06.24 um 15:50 schrieb Alex Deucher:

Quoted 11 lines

This can be called in atomic context. Should fix:

BUG: sleeping function called from invalid context at include/linux/sched/mm.h:306

in_atomic(): 1, irqs_disabled(): 0, non_block: 0, pid: 449, name: kworker/u64:8
preempt_count: 2, expected: 0
RCU nest depth: 0, expected: 0
Preemption disabled at:
ffffffffc0ce1580>] dc_fpu_begin+0x30/0xd0 [amdgpu]

CPU: 5 PID: 449 Comm: kworker/u64:8 Tainted: G        W          6.8.0+ #35
Hardware name: System manufacturer System Product Name/ROG STRIX X570-E GAMING WIFI II, BIOS 4204 02/24/2022
Workqueue: events_unbound async_run_entry_fn
Quoted 7 lines

That most likely only papers over the real problem and is not a valid fix.

The question is why is that an atomic context?  If the function is used under a spinlock then this might indeed be the right fix.

If it's because of floating point operation then that here won't work either.

It looks like there is only one caller, and this turns on floating point instructions just for the call:

         if (dc->res_pool->funcs->update_bw_bounding_box) {
                 DC_FP_START();
                 dc->res_pool->funcs->update_bw_bounding_box(dc, dc->clk_mgr->bw_params);
                 DC_FP_END();
         }

but then they are enabled again inside of the function.

Quoted 8 lines

If we can drop the outer DC_FP_START(), that means the GFP_KERNEL allocation works. On the other hand if we actually have to enabled it before calling into the function (e.g. because there is an architecture that has incompatible function calling conventions when FP is enabled), the inner one is redundant, but we can potentially move the kmemdup() into the caller and pass the copy by refernence.

Yeah exactly that's the case.

The DC_FP_START() and DC_FP_END() calls need to be outside of the function  because the compiler has no idea that it can't move any flouting point instructions outside of the critical section between the two functions.

So yes the calls to DC_FP_START() and DC_FP_END() from within floating point enabled code should be forbidden somehow.

And when that is done the caller should allocate any parameters needed and pass them by reference to avoid the GFP_ATOMIC.

Regards,
Christian.

Quoted 2 lines

Arnd