Skip to contentSkip to conversation

drm/amd/dc/dce120: Fix audio register mapping, stop triggering KASAN

PATCH In review Lyude Paul 3 participants 4 messages Nov 14, 2022 – Nov 18, 2022

Subsystems (4)
lore
Share & tools
lore

Lyude Paul

author
#0 lore raw

PATCH 1/1 drm/amd/dc/dce120: Fix audio register mapping, stop triggering KASAN

There's been a very long running bug that seems to have been neglected for a while, where amdgpu consistently triggers a KASAN error at start:

  BUG: KASAN: global-out-of-bounds in read_indirect_azalia_reg+0x1d4/0x2a0 [amdgpu]
  Read of size 4 at addr ffffffffc2274b28 by task modprobe/1889

After digging through amd's rather creative method for accessing registers, I eventually discovered the problem likely has to do with the fact that on my dce120 GPU there are supposedly 7 sets of audio registers. But we only define a register mapping for 6 sets.

So, fix this and fix the KASAN warning finally.

---
Sending this one separately from the rest of my fixes since:

* It's definitely completely unrelated to the Gitlab 2171 issue
* I'm not sure if this is the correct fix since it's in DC

 drivers/gpu/drm/amd/display/dc/dce120/dce120_resource.c | 3 ++-
 1 file changed, 2 insertions(+), 1 deletion(-)
drivers/gpu/drm/amd/display/dc/dce120/dce120_resource.c +2 -1
index 1b70b78e2fa15..af631085e88c5 100644
@@ -359,7 +359,8 @@ static const struct dce_audio_registers audio_regs[] = {
 	audio_regs(2),
 	audio_regs(3),
 	audio_regs(4),
-	audio_regs(5)
+	audio_regs(5),
+	audio_regs(6),
 };
 
 #define DCE120_AUD_COMMON_MASK_SH_LIST(mask_sh)\
Signature · 1 line
2.37.3
Reviewed-by on v1 1/1
#1 lore raw
Quoted 12 lines · Lyude Paul

There's been a very long running bug that seems to have been neglected for a while, where amdgpu consistently triggers a KASAN error at start:

  BUG: KASAN: global-out-of-bounds in read_indirect_azalia_reg+0x1d4/0x2a0 [amdgpu]
  Read of size 4 at addr ffffffffc2274b28 by task modprobe/1889

After digging through amd's rather creative method for accessing registers, I eventually discovered the problem likely has to do with the fact that on my dce120 GPU there are supposedly 7 sets of audio registers. But we only define a register mapping for 6 sets.

Quoted 4 lines

So, fix this and fix the KASAN warning finally.

This is the correct fix for asics having 7 audio instances. It looks correct to me, assuming DCE12 actually has 7 audio instances. @Wentland, Harry Do you know off hand? If you can confirm that, the patch is:

Quoted 26 lines

---
Sending this one separately from the rest of my fixes since:

* It's definitely completely unrelated to the Gitlab 2171 issue
* I'm not sure if this is the correct fix since it's in DC

 drivers/gpu/drm/amd/display/dc/dce120/dce120_resource.c | 3 ++-
 1 file changed, 2 insertions(+), 1 deletion(-)
drivers/gpu/drm/amd/display/dc/dce120/dce120_resource.c +2 -2
index 1b70b78e2fa15..af631085e88c5 100644
@@ -359,7 +359,8 @@ static const struct dce_audio_registers audio_regs[] = {
        audio_regs(2),
        audio_regs(3),
        audio_regs(4),
-       audio_regs(5)
+       audio_regs(5),
+       audio_regs(6),
 };
 
 #define DCE120_AUD_COMMON_MASK_SH_LIST(mask_sh)\
--

2.37.3

#2 lore raw
Quoted 16 lines · Alex Deucher
Quoted 14 lines · Lyude Paul

There's been a very long running bug that seems to have been neglected for a while, where amdgpu consistently triggers a KASAN error at start:

  BUG: KASAN: global-out-of-bounds in read_indirect_azalia_reg+0x1d4/0x2a0 [amdgpu]
  Read of size 4 at addr ffffffffc2274b28 by task modprobe/1889

After digging through amd's rather creative method for accessing registers, I eventually discovered the problem likely has to do with the fact that on my dce120 GPU there are supposedly 7 sets of audio registers. But we only define a register mapping for 6 sets.

So, fix this and fix the KASAN warning finally.

Quoted 8 lines
Quoted 2 lines

This is the correct fix for asics having 7 audio instances. It looks correct to me, assuming DCE12 actually has 7 audio instances. @Wentland, Harry Do you know off hand? If you can confirm that, the patch is:

The driver currently defines 7 audio instances, whether or not it actually has 7 remains to be seen but the code as is is broken, so I'll apply this. If it turns out there are only 6 instances we can fix the count later. Applied. Thanks!

Alex

Quoted 28 lines
Quoted 28 lines

---
Sending this one separately from the rest of my fixes since:

* It's definitely completely unrelated to the Gitlab 2171 issue
* I'm not sure if this is the correct fix since it's in DC

 drivers/gpu/drm/amd/display/dc/dce120/dce120_resource.c | 3 ++-
 1 file changed, 2 insertions(+), 1 deletion(-)
drivers/gpu/drm/amd/display/dc/dce120/dce120_resource.c +2 -2
index 1b70b78e2fa15..af631085e88c5 100644
@@ -359,7 +359,8 @@ static const struct dce_audio_registers audio_regs[] = {
        audio_regs(2),
        audio_regs(3),
        audio_regs(4),
-       audio_regs(5)
+       audio_regs(5),
+       audio_regs(6),
 };
 
 #define DCE120_AUD_COMMON_MASK_SH_LIST(mask_sh)\
--

2.37.3

maintainer Reviewed-by
#3 lore raw
Quoted 26 lines · Alex Deucher
Quoted 16 lines · Alex Deucher
Quoted 14 lines · Lyude Paul

There's been a very long running bug that seems to have been neglected for a while, where amdgpu consistently triggers a KASAN error at start:

  BUG: KASAN: global-out-of-bounds in read_indirect_azalia_reg+0x1d4/0x2a0 [amdgpu]
  Read of size 4 at addr ffffffffc2274b28 by task modprobe/1889

After digging through amd's rather creative method for accessing registers, I eventually discovered the problem likely has to do with the fact that on my dce120 GPU there are supposedly 7 sets of audio registers. But we only define a register mapping for 6 sets.

So, fix this and fix the KASAN warning finally.

Quoted 8 lines
Quoted 2 lines

This is the correct fix for asics having 7 audio instances. It looks correct to me, assuming DCE12 actually has 7 audio instances. @Wentland, Harry Do you know off hand? If you can confirm that, the patch is:

Quoted 5 lines

The driver currently defines 7 audio instances, whether or not it actually has 7 remains to be seen but the code as is is broken, so I'll apply this. If it turns out there are only 6 instances we can fix the count later. Applied. Thanks!

Good catch. I seem to recall some DCE generations had an extra audio device and if DCE12 defines 7 audio instances then this is the correct fix.

Harry

Quoted 30 lines

Alex

Quoted 28 lines
Quoted 28 lines

---
Sending this one separately from the rest of my fixes since:

* It's definitely completely unrelated to the Gitlab 2171 issue
* I'm not sure if this is the correct fix since it's in DC

 drivers/gpu/drm/amd/display/dc/dce120/dce120_resource.c | 3 ++-
 1 file changed, 2 insertions(+), 1 deletion(-)
drivers/gpu/drm/amd/display/dc/dce120/dce120_resource.c +2 -2
index 1b70b78e2fa15..af631085e88c5 100644
@@ -359,7 +359,8 @@ static const struct dce_audio_registers audio_regs[] = {
        audio_regs(2),
        audio_regs(3),
        audio_regs(4),
-       audio_regs(5)
+       audio_regs(5),
+       audio_regs(6),
 };
 
 #define DCE120_AUD_COMMON_MASK_SH_LIST(mask_sh)\
--

2.37.3