Skip to contentSkip to conversation

nouveau: explicitly wait on the fence in nouveau_bo_move_m2mf

PATCH In review Karol Herbst 5 participants 17 messages Aug 19, 2022 – Feb 1, 2023

Subsystems (4)
lore
Share & tools
lore
#0 lore raw

PATCH 1/1 nouveau: explicitly wait on the fence in nouveau_bo_move_m2mf

It is a bit unlcear to us why that's helping, but it does and unbreaks suspend/resume on a lot of GPUs without any known drawbacks.

---
 drivers/gpu/drm/nouveau/nouveau_bo.c | 9 +++++++++
 1 file changed, 9 insertions(+)
drivers/gpu/drm/nouveau/nouveau_bo.c +9 -0
index 35bb0bb3fe61..126b3c6e12f9 100644
@@ -822,6 +822,15 @@ nouveau_bo_move_m2mf(struct ttm_buffer_object *bo, int evict,
 		if (ret == 0) {
 			ret = nouveau_fence_new(chan, false, &fence);
 			if (ret == 0) {
+				/* TODO: figure out a better solution here
+				 *
+				 * wait on the fence here explicitly as going through
+				 * ttm_bo_move_accel_cleanup somehow doesn't seem to do it.
+				 *
+				 * Without this the operation can timeout and we'll fallback to a
+				 * software copy, which might take several minutes to finish.
+				 */
+				nouveau_fence_wait(fence, false, false);
 				ret = ttm_bo_move_accel_cleanup(bo,
 								&fence->base,
 								evict, false,
Signature · 1 line
2.37.1
maintainer Reviewed-by on v1 1/1
#1 lore raw
Quoted 30 lines · Karol Herbst

It is a bit unlcear to us why that's helping, but it does and unbreaks suspend/resume on a lot of GPUs without any known drawbacks.

---
 drivers/gpu/drm/nouveau/nouveau_bo.c | 9 +++++++++
 1 file changed, 9 insertions(+)
drivers/gpu/drm/nouveau/nouveau_bo.c +9 -0
index 35bb0bb3fe61..126b3c6e12f9 100644
@@ -822,6 +822,15 @@ nouveau_bo_move_m2mf(struct ttm_buffer_object *bo, int evict,
 		if (ret == 0) {
 			ret = nouveau_fence_new(chan, false, &fence);
 			if (ret == 0) {
+				/* TODO: figure out a better solution here
+				 *
+				 * wait on the fence here explicitly as going through
+				 * ttm_bo_move_accel_cleanup somehow doesn't seem to do it.
+				 *
+				 * Without this the operation can timeout and we'll fallback to a
+				 * software copy, which might take several minutes to finish.
+				 */
+				nouveau_fence_wait(fence, false, false);
 				ret = ttm_bo_move_accel_cleanup(bo,
 								&fence->base,
 								evict, false,
Signature · 3 lines
Cheers,
 Lyude Paul (she/her)
 Software Engineer at Red Hat
#2 lore raw
In reply to Karol Herbst

Hi,

Quoted 34 lines · Karol Herbst

It is a bit unlcear to us why that's helping, but it does and unbreaks suspend/resume on a lot of GPUs without any known drawbacks.

---
 drivers/gpu/drm/nouveau/nouveau_bo.c | 9 +++++++++
 1 file changed, 9 insertions(+)
drivers/gpu/drm/nouveau/nouveau_bo.c +9 -0
index 35bb0bb3fe61..126b3c6e12f9 100644
@@ -822,6 +822,15 @@ nouveau_bo_move_m2mf(struct ttm_buffer_object *bo, int evict,
 		if (ret == 0) {
 			ret = nouveau_fence_new(chan, false, &fence);
 			if (ret == 0) {
+				/* TODO: figure out a better solution here
+				 *
+				 * wait on the fence here explicitly as going through
+				 * ttm_bo_move_accel_cleanup somehow doesn't seem to do it.
+				 *
+				 * Without this the operation can timeout and we'll fallback to a
+				 * software copy, which might take several minutes to finish.
+				 */
+				nouveau_fence_wait(fence, false, false);
 				ret = ttm_bo_move_accel_cleanup(bo,
 								&fence->base,
 								evict, false,
Signature · 1 line
2.37.1

While this is marked for 5.15+ only, a user in Debian was seeing the suspend issue as well on 5.10.y and did confirm the commit fixes the issue as well in the 5.10.y series:

https://bugs.debian.org/cgi-bin/bugreport.cgi?bug=989705#69

Karol, Lyude, should that as well be picked for 5.10.y?

Regards,
Salvatore

#3 lore raw
Quoted 43 lines · Salvatore Bonaccorso

Hi,

Quoted 31 lines · Karol Herbst

It is a bit unlcear to us why that's helping, but it does and unbreaks suspend/resume on a lot of GPUs without any known drawbacks.

---
 drivers/gpu/drm/nouveau/nouveau_bo.c | 9 +++++++++
 1 file changed, 9 insertions(+)
drivers/gpu/drm/nouveau/nouveau_bo.c +9 -1
index 35bb0bb3fe61..126b3c6e12f9 100644
@@ -822,6 +822,15 @@ nouveau_bo_move_m2mf(struct ttm_buffer_object *bo, int evict,
              if (ret == 0) {
                      ret = nouveau_fence_new(chan, false, &fence);
                      if (ret == 0) {
+                             /* TODO: figure out a better solution here
+                              *
+                              * wait on the fence here explicitly as going through
+                              * ttm_bo_move_accel_cleanup somehow doesn't seem to do it.
+                              *
+                              * Without this the operation can timeout and we'll fallback to a
+                              * software copy, which might take several minutes to finish.
+                              */
+                             nouveau_fence_wait(fence, false, false);
                              ret = ttm_bo_move_accel_cleanup(bo,
                                                              &fence->base,
                                                              evict, false,
--
Quoted 3 lines

2.37.1

While this is marked for 5.15+ only, a user in Debian was seeing the suspend issue as well on 5.10.y and did confirm the commit fixes the issue as well in the 5.10.y series:

Quoted 4 lines

https://bugs.debian.org/cgi-bin/bugreport.cgi?bug=989705#69

Karol, Lyude, should that as well be picked for 5.10.y?

mhh from the original report 5.10 was fine, but maybe something got backported and it broke it? I'll try to do some testing on my machine and see what I can figure out, but it could also be a debian only issue at this point.

Quoted 3 lines

Regards,
Salvatore

#4 lore raw

Hi,

Quoted 46 lines · Karol Herbst
Quoted 45 lines · Salvatore Bonaccorso

Hi,

Quoted 31 lines · Karol Herbst

It is a bit unlcear to us why that's helping, but it does and unbreaks suspend/resume on a lot of GPUs without any known drawbacks.

---
 drivers/gpu/drm/nouveau/nouveau_bo.c | 9 +++++++++
 1 file changed, 9 insertions(+)
drivers/gpu/drm/nouveau/nouveau_bo.c +9 -1
index 35bb0bb3fe61..126b3c6e12f9 100644
@@ -822,6 +822,15 @@ nouveau_bo_move_m2mf(struct ttm_buffer_object *bo, int evict,
              if (ret == 0) {
                      ret = nouveau_fence_new(chan, false, &fence);
                      if (ret == 0) {
+                             /* TODO: figure out a better solution here
+                              *
+                              * wait on the fence here explicitly as going through
+                              * ttm_bo_move_accel_cleanup somehow doesn't seem to do it.
+                              *
+                              * Without this the operation can timeout and we'll fallback to a
+                              * software copy, which might take several minutes to finish.
+                              */
+                             nouveau_fence_wait(fence, false, false);
                              ret = ttm_bo_move_accel_cleanup(bo,
                                                              &fence->base,
                                                              evict, false,
--
Quoted 3 lines

2.37.1

While this is marked for 5.15+ only, a user in Debian was seeing the suspend issue as well on 5.10.y and did confirm the commit fixes the issue as well in the 5.10.y series:

https://bugs.debian.org/cgi-bin/bugreport.cgi?bug=989705#69

Quoted 7 lines
Quoted 2 lines

Karol, Lyude, should that as well be picked for 5.10.y?

mhh from the original report 5.10 was fine, but maybe something got backported and it broke it? I'll try to do some testing on my machine and see what I can figure out, but it could also be a debian only issue at this point.

Right, this is a possiblity, thanks for looking into it!

Computer Enthusiastic, can you verify the problem as well in a non-Debian patched upstream kernel directly from the 5.10.y series (latest 5.10.144) and verify the fix there?

Regards,
Salvatore

#5 lore raw

Hello,

Il giorno mar 20 set 2022 alle ore 13:59 Salvatore Bonaccorso <[email protected]> ha scritto:
[..]

Quoted 6 lines

Computer Enthusiastic, can you verify the problem as well in a non-Debian patched upstream kernel directly from the 5.10.y series (latest 5.10.144) and verify the fix there?

Regards,
Salvatore

I've tested the vanilla kernel 5.10.145 (it was the latest one week ago) without Debian kernel patches, but using the kernel config file from the latest kernel for Debian Stable:
- without the Karol's patch: it always fails both suspend to ram and hibernate to disk with the usual behavior (a very long time to suspend or hibernate, then it fails on resume with a garbled screen)
- with the Karol's patch: it succeeds both suspend and hibernate and it correctly resumes afterwards.

The kernel was tested using the following graphic adapter:
Graphics: Device-1: NVIDIA G96CM [GeForce 9600M GT] driver: nouveau v: kernel
Device-2: Suyin Acer HD Crystal Eye webcam type: USB driver:
uvcvideo

          Display: x11 server: X.Org 1.20.11 driver: loaded:
modesetting unloaded: fbdev,vesa
          resolution: 1280x800~60Hz
          OpenGL: renderer: NV96 v: 3.3 Mesa 20.3.5

Therefore, 5.10.y series of the kernel need to be patched to work correctly at least with the aforementioned graphic card.

The script I used to compile the kernel are attached for further reference and verification.

Hope that helps.

#6 lore raw
In reply to Karol Herbst

Hello,

Il giorno ven 19 ago 2022 alle ore 22:09 Karol Herbst <[email protected]> ha scritto:

Quoted 32 lines

It is a bit unlcear to us why that's helping, but it does and unbreaks suspend/resume on a lot of GPUs without any known drawbacks.

---
 drivers/gpu/drm/nouveau/nouveau_bo.c | 9 +++++++++
 1 file changed, 9 insertions(+)
drivers/gpu/drm/nouveau/nouveau_bo.c +9 -1
index 35bb0bb3fe61..126b3c6e12f9 100644
@@ -822,6 +822,15 @@ nouveau_bo_move_m2mf(struct ttm_buffer_object *bo, int evict,
                if (ret == 0) {
                        ret = nouveau_fence_new(chan, false, &fence);
                        if (ret == 0) {
+                               /* TODO: figure out a better solution here
+                                *
+                                * wait on the fence here explicitly as going through
+                                * ttm_bo_move_accel_cleanup somehow doesn't seem to do it.
+                                *
+                                * Without this the operation can timeout and we'll fallback to a
+                                * software copy, which might take several minutes to finish.
+                                */
+                               nouveau_fence_wait(fence, false, false);
                                ret = ttm_bo_move_accel_cleanup(bo,
                                                                &fence->base,
                                                                evict, false,
--
Quoted 2 lines

2.37.1

Do you think it could be possible to make the patch land in kernel 5.10.x in the near future ?

Is there something I can do to help it to happen ?

Thanks.

#7 lore raw
Quoted 43 lines · Computer Enthusiastic

Hello,

Il giorno ven 19 ago 2022 alle ore 22:09 Karol Herbst <[email protected]> ha scritto:

Quoted 32 lines

It is a bit unlcear to us why that's helping, but it does and unbreaks suspend/resume on a lot of GPUs without any known drawbacks.

---
 drivers/gpu/drm/nouveau/nouveau_bo.c | 9 +++++++++
 1 file changed, 9 insertions(+)
drivers/gpu/drm/nouveau/nouveau_bo.c +9 -1
index 35bb0bb3fe61..126b3c6e12f9 100644
@@ -822,6 +822,15 @@ nouveau_bo_move_m2mf(struct ttm_buffer_object *bo, int evict,
                if (ret == 0) {
                        ret = nouveau_fence_new(chan, false, &fence);
                        if (ret == 0) {
+                               /* TODO: figure out a better solution here
+                                *
+                                * wait on the fence here explicitly as going through
+                                * ttm_bo_move_accel_cleanup somehow doesn't seem to do it.
+                                *
+                                * Without this the operation can timeout and we'll fallback to a
+                                * software copy, which might take several minutes to finish.
+                                */
+                               nouveau_fence_wait(fence, false, false);
                                ret = ttm_bo_move_accel_cleanup(bo,
                                                                &fence->base,
                                                                evict, false,
--
Quoted 2 lines

2.37.1

Do you think it could be possible to make the patch land in kernel 5.10.x in the near future ?

Quoted 2 lines

Is there something I can do to help it to happen ?

you can send it out to the stable mailing list and I ack it there. I just totally forgot about it, sorry.

Quoted 2 lines

Thanks.

#8 lore raw

Hello Karol,

Il giorno mar 1 nov 2022 alle ore 11:46 Karol Herbst <[email protected]> ha scritto:

Quoted 46 lines
Quoted 43 lines · Computer Enthusiastic

Hello,

Il giorno ven 19 ago 2022 alle ore 22:09 Karol Herbst <[email protected]> ha scritto:

Quoted 32 lines

It is a bit unlcear to us why that's helping, but it does and unbreaks suspend/resume on a lot of GPUs without any known drawbacks.

---
 drivers/gpu/drm/nouveau/nouveau_bo.c | 9 +++++++++
 1 file changed, 9 insertions(+)
drivers/gpu/drm/nouveau/nouveau_bo.c +9 -1
index 35bb0bb3fe61..126b3c6e12f9 100644
@@ -822,6 +822,15 @@ nouveau_bo_move_m2mf(struct ttm_buffer_object *bo, int evict,
                if (ret == 0) {
                        ret = nouveau_fence_new(chan, false, &fence);
                        if (ret == 0) {
+                               /* TODO: figure out a better solution here
+                                *
+                                * wait on the fence here explicitly as going through
+                                * ttm_bo_move_accel_cleanup somehow doesn't seem to do it.
+                                *
+                                * Without this the operation can timeout and we'll fallback to a
+                                * software copy, which might take several minutes to finish.
+                                */
+                               nouveau_fence_wait(fence, false, false);
                                ret = ttm_bo_move_accel_cleanup(bo,
                                                                &fence->base,
                                                                evict, false,
--
Quoted 2 lines

2.37.1

Do you think it could be possible to make the patch land in kernel 5.10.x in the near future ?

Quoted 8 lines
Quoted 2 lines

Is there something I can do to help it to happen ?

you can send it out to the stable mailing list and I ack it there. I just totally forgot about it, sorry.

Quoted 2 lines

Thanks.

As you suggested, I sent a request to a request to linux-kernel [1], but it was not acknowledged.

What else can I do to help the patch land in the 5.10.x kernel in the future ?

Thanks and happy Christmas.

[1] https://marc.info/?l=linux-kernel&m=166883482714930&w=2

#9 lore raw

Hello,

The patch "[Nouveau] [PATCH] nouveau: explicitly wait on the fence in nouveau_bo_move_m2mf" [1] was marked for kernels v5.15+ and it was merged upstream.

The same patch [1] works with kernel 5.10.y, but it is not been merged upstream so far.

According to Karol Herbst suggestion [2], I'm sending this message to ask for merging it into 5.10 kernel.

Thanks in advance.

---
[1] https://lore.kernel.org/nouveau/[email protected]/

[2] https://lore.kernel.org/nouveau/CACO55tv0jO2TmuWcwFiAUQB-__DZVwhv7WNN9MfgMXV053gknw@mail.gmail.com/

maintainer
#10 lore raw
Quoted 11 lines · Computer Enthusiastic

Hello,

The patch "[Nouveau] [PATCH] nouveau: explicitly wait on the fence in nouveau_bo_move_m2mf" [1] was marked for kernels v5.15+ and it was merged upstream.

The same patch [1] works with kernel 5.10.y, but it is not been merged upstream so far.

According to Karol Herbst suggestion [2], I'm sending this message to ask for merging it into 5.10 kernel.

We need to know the git commit id. And have you tested it on 5.10.y? And why are you stuck on 5.10.y for this type of hardware? Why not move to 5.15.y or 6.1.y?

And as my bot says:

<formletter>

This is not the correct way to submit patches for inclusion in the stable kernel tree. Please read:
https://www.kernel.org/doc/html/latest/process/stable-kernel-rules.html
for how to do this properly.

</formletter>

thanks,

greg k-h

#11 lore raw

Hi Greg,

I'm not the reporter, so would like to confirm him explicitly, but I believe I can give some context:

Quoted 10 lines · Greg KH
Quoted 9 lines · Computer Enthusiastic

Hello,

The patch "[Nouveau] [PATCH] nouveau: explicitly wait on the fence in nouveau_bo_move_m2mf" [1] was marked for kernels v5.15+ and it was merged upstream.

The same patch [1] works with kernel 5.10.y, but it is not been merged upstream so far.

Quoted 6 lines
Quoted 2 lines

According to Karol Herbst suggestion [2], I'm sending this message to ask for merging it into 5.10 kernel.

We need to know the git commit id. And have you tested it on 5.10.y? And why are you stuck on 5.10.y for this type of hardware? Why not move to 5.15.y or 6.1.y?

This would be commit 6b04ce966a73 ("nouveau: explicitly wait on the fence in nouveau_bo_move_m2mf") in mainline, applied in 6.0-rc3 and backported to 5.19.6 and 5.15.64.

Computer Enthusiastic, tested it on 5.10.y:
https://lore.kernel.org/nouveau/CAHSpYy1mcTns0JS6eivjK82CZ9_ajSwH-H7gtDwCkNyfvihaAw@mail.gmail.com/

It was reported in Debian by the user originally as
https://bugs.debian.org/989705#69 after updating to the 5.10.y series in Debian bullseye.

I guess the user could move to the next stable release Debian bookworm, once it's released (it's currently in the last milestones to finalize, cf.
https://release.debian.org/ but we are not yet there). In the next release this will be automatically be fixed indeed.

Computer Enthusiastic, can you confirm please to Greg in particular the first questions, in particular to confirm the commit fixes the suspend issue?

Regards,
Salvatore

#12 lore raw

Hello Greg,
Hello Salvatore,

Quoted 39 lines · Salvatore Bonaccorso

Hi Greg,

I'm not the reporter, so would like to confirm him explicitly, but I believe I can give some context:

Quoted 10 lines · Greg KH
Quoted 9 lines · Computer Enthusiastic

Hello,

The patch "[Nouveau] [PATCH] nouveau: explicitly wait on the fence in nouveau_bo_move_m2mf" [1] was marked for kernels v5.15+ and it was merged upstream.

The same patch [1] works with kernel 5.10.y, but it is not been merged upstream so far.

Quoted 6 lines
Quoted 2 lines

According to Karol Herbst suggestion [2], I'm sending this message to ask for merging it into 5.10 kernel.

We need to know the git commit id. And have you tested it on 5.10.y? And why are you stuck on 5.10.y for this type of hardware? Why not move to 5.15.y or 6.1.y?

This would be commit 6b04ce966a73 ("nouveau: explicitly wait on the fence in nouveau_bo_move_m2mf") in mainline, applied in 6.0-rc3 and backported to 5.19.6 and 5.15.64.

Computer Enthusiastic, tested it on 5.10.y:
https://lore.kernel.org/nouveau/CAHSpYy1mcTns0JS6eivjK82CZ9_ajSwH-H7gtDwCkNyfvihaAw@mail.gmail.com/

It was reported in Debian by the user originally as
https://bugs.debian.org/989705#69 after updating to the 5.10.y series in Debian bullseye.

I guess the user could move to the next stable release Debian bookworm, once it's released (it's currently in the last milestones to finalize, cf.
https://release.debian.org/ but we are not yet there). In the next release this will be automatically be fixed indeed.

Quoted 5 lines

Computer Enthusiastic, can you confirm please to Greg in particular the first questions, in particular to confirm the commit fixes the suspend issue?

Regards,
Salvatore

Thanks for replaying to my request: I really appreciate.

I apologize if my request was not formally correct.

The upstream kernel 5.10.y hangs on suspend or fails to resume if it is suspended to ram or suspended to disk (if nouveau kernel module is used with some nvidia graphic cards).

I confirm the commit ID 6b04ce966a73 (by Karol Herbst) fixes the aforementioned suspend to ram and suspend to disk issues with kernel 5.10.y . It tested it with my own computer.

The last kernel version I tested is 5.10.165, that I patched and installed in Debian Stable (11.6) that I'm currently running and that I tested again today.

It would be nice if the next point release of Debian Stable could ship a kernel that includes patch commit ID 6b04ce966a73 for the benefit of nouveau module users.

Thanks again.

maintainer
#13 lore raw
Quoted 61 lines · Computer Enthusiastic

Hello Greg,
Hello Salvatore,

Quoted 39 lines · Salvatore Bonaccorso

Hi Greg,

I'm not the reporter, so would like to confirm him explicitly, but I believe I can give some context:

Quoted 10 lines · Greg KH
Quoted 9 lines · Computer Enthusiastic

Hello,

The patch "[Nouveau] [PATCH] nouveau: explicitly wait on the fence in nouveau_bo_move_m2mf" [1] was marked for kernels v5.15+ and it was merged upstream.

The same patch [1] works with kernel 5.10.y, but it is not been merged upstream so far.

Quoted 6 lines
Quoted 2 lines

According to Karol Herbst suggestion [2], I'm sending this message to ask for merging it into 5.10 kernel.

We need to know the git commit id. And have you tested it on 5.10.y? And why are you stuck on 5.10.y for this type of hardware? Why not move to 5.15.y or 6.1.y?

This would be commit 6b04ce966a73 ("nouveau: explicitly wait on the fence in nouveau_bo_move_m2mf") in mainline, applied in 6.0-rc3 and backported to 5.19.6 and 5.15.64.

Computer Enthusiastic, tested it on 5.10.y:
https://lore.kernel.org/nouveau/CAHSpYy1mcTns0JS6eivjK82CZ9_ajSwH-H7gtDwCkNyfvihaAw@mail.gmail.com/

It was reported in Debian by the user originally as
https://bugs.debian.org/989705#69 after updating to the 5.10.y series in Debian bullseye.

I guess the user could move to the next stable release Debian bookworm, once it's released (it's currently in the last milestones to finalize, cf.
https://release.debian.org/ but we are not yet there). In the next release this will be automatically be fixed indeed.

Quoted 5 lines

Computer Enthusiastic, can you confirm please to Greg in particular the first questions, in particular to confirm the commit fixes the suspend issue?

Regards,
Salvatore

Thanks for replaying to my request: I really appreciate.

I apologize if my request was not formally correct.

The upstream kernel 5.10.y hangs on suspend or fails to resume if it is suspended to ram or suspended to disk (if nouveau kernel module is used with some nvidia graphic cards).

I confirm the commit ID 6b04ce966a73 (by Karol Herbst) fixes the aforementioned suspend to ram and suspend to disk issues with kernel 5.10.y . It tested it with my own computer.

Quoted 7 lines

The last kernel version I tested is 5.10.165, that I patched and installed in Debian Stable (11.6) that I'm currently running and that I tested again today.

It would be nice if the next point release of Debian Stable could ship a kernel that includes patch commit ID 6b04ce966a73 for the benefit of nouveau module users.

Ok, I've queued it up for 5.10.y now, thanks.

greg k-h

maintainer
#14 lore raw

Thanks a ton for the help Greg!

Quoted 74 lines · Greg KH
Quoted 61 lines · Computer Enthusiastic

Hello Greg,
Hello Salvatore,

Quoted 39 lines · Salvatore Bonaccorso

Hi Greg,

I'm not the reporter, so would like to confirm him explicitly, but I believe I can give some context:

Quoted 10 lines · Greg KH
Quoted 9 lines · Computer Enthusiastic

Hello,

The patch "[Nouveau] [PATCH] nouveau: explicitly wait on the fence in nouveau_bo_move_m2mf" [1] was marked for kernels v5.15+ and it was merged upstream.

The same patch [1] works with kernel 5.10.y, but it is not been merged upstream so far.

Quoted 6 lines
Quoted 2 lines

According to Karol Herbst suggestion [2], I'm sending this message to ask for merging it into 5.10 kernel.

We need to know the git commit id. And have you tested it on 5.10.y? And why are you stuck on 5.10.y for this type of hardware? Why not move to 5.15.y or 6.1.y?

This would be commit 6b04ce966a73 ("nouveau: explicitly wait on the fence in nouveau_bo_move_m2mf") in mainline, applied in 6.0-rc3 and backported to 5.19.6 and 5.15.64.

Computer Enthusiastic, tested it on 5.10.y:
https://lore.kernel.org/nouveau/CAHSpYy1mcTns0JS6eivjK82CZ9_ajSwH-H7gtDwCkNyfvihaAw@mail.gmail.com/

It was reported in Debian by the user originally as
https://bugs.debian.org/989705#69 after updating to the 5.10.y series in Debian bullseye.

I guess the user could move to the next stable release Debian bookworm, once it's released (it's currently in the last milestones to finalize, cf.
https://release.debian.org/ but we are not yet there). In the next release this will be automatically be fixed indeed.

Quoted 5 lines

Computer Enthusiastic, can you confirm please to Greg in particular the first questions, in particular to confirm the commit fixes the suspend issue?

Regards,
Salvatore

Thanks for replaying to my request: I really appreciate.

I apologize if my request was not formally correct.

The upstream kernel 5.10.y hangs on suspend or fails to resume if it is suspended to ram or suspended to disk (if nouveau kernel module is used with some nvidia graphic cards).

I confirm the commit ID 6b04ce966a73 (by Karol Herbst) fixes the aforementioned suspend to ram and suspend to disk issues with kernel 5.10.y . It tested it with my own computer.

Quoted 7 lines

The last kernel version I tested is 5.10.165, that I patched and installed in Debian Stable (11.6) that I'm currently running and that I tested again today.

It would be nice if the next point release of Debian Stable could ship a kernel that includes patch commit ID 6b04ce966a73 for the benefit of nouveau module users.

Ok, I've queued it up for 5.10.y now, thanks.

greg k-h

Signature · 3 lines
Cheers,
 Lyude Paul (she/her)
 Software Engineer at Red Hat
#15 lore raw

Hello Greg,

Quoted 77 lines · Lyude Paul

Thanks a ton for the help Greg!

Quoted 74 lines · Greg KH
Quoted 61 lines · Computer Enthusiastic

Hello Greg,
Hello Salvatore,

Quoted 39 lines · Salvatore Bonaccorso

Hi Greg,

I'm not the reporter, so would like to confirm him explicitly, but I believe I can give some context:

Quoted 10 lines · Greg KH
Quoted 9 lines · Computer Enthusiastic

Hello,

The patch "[Nouveau] [PATCH] nouveau: explicitly wait on the fence in nouveau_bo_move_m2mf" [1] was marked for kernels v5.15+ and it was merged upstream.

The same patch [1] works with kernel 5.10.y, but it is not been merged upstream so far.

Quoted 6 lines
Quoted 2 lines

According to Karol Herbst suggestion [2], I'm sending this message to ask for merging it into 5.10 kernel.

We need to know the git commit id. And have you tested it on 5.10.y? And why are you stuck on 5.10.y for this type of hardware? Why not move to 5.15.y or 6.1.y?

This would be commit 6b04ce966a73 ("nouveau: explicitly wait on the fence in nouveau_bo_move_m2mf") in mainline, applied in 6.0-rc3 and backported to 5.19.6 and 5.15.64.

Computer Enthusiastic, tested it on 5.10.y:
https://lore.kernel.org/nouveau/CAHSpYy1mcTns0JS6eivjK82CZ9_ajSwH-H7gtDwCkNyfvihaAw@mail.gmail.com/

It was reported in Debian by the user originally as
https://bugs.debian.org/989705#69 after updating to the 5.10.y series in Debian bullseye.

I guess the user could move to the next stable release Debian bookworm, once it's released (it's currently in the last milestones to finalize, cf.
https://release.debian.org/ but we are not yet there). In the next release this will be automatically be fixed indeed.

Quoted 5 lines

Computer Enthusiastic, can you confirm please to Greg in particular the first questions, in particular to confirm the commit fixes the suspend issue?

Regards,
Salvatore

Thanks for replaying to my request: I really appreciate.

I apologize if my request was not formally correct.

The upstream kernel 5.10.y hangs on suspend or fails to resume if it is suspended to ram or suspended to disk (if nouveau kernel module is used with some nvidia graphic cards).

I confirm the commit ID 6b04ce966a73 (by Karol Herbst) fixes the aforementioned suspend to ram and suspend to disk issues with kernel 5.10.y . It tested it with my own computer.

Quoted 7 lines

The last kernel version I tested is 5.10.165, that I patched and installed in Debian Stable (11.6) that I'm currently running and that I tested again today.

It would be nice if the next point release of Debian Stable could ship a kernel that includes patch commit ID 6b04ce966a73 for the benefit of nouveau module users.

Ok, I've queued it up for 5.10.y now, thanks.

greg k-h

Thank you so much.

Many thanks to Salvatore for the help and, of course, to Karl for the patch and to all of you who made it possible.

#16 lore raw
In reply to Karol Herbst

Hello,

Il giorno ven 19 ago 2022 alle ore 22:09 Karol Herbst <[email protected]> ha scritto:

Quoted 32 lines

It is a bit unlcear to us why that's helping, but it does and unbreaks suspend/resume on a lot of GPUs without any known drawbacks.

---
 drivers/gpu/drm/nouveau/nouveau_bo.c | 9 +++++++++
 1 file changed, 9 insertions(+)
drivers/gpu/drm/nouveau/nouveau_bo.c +9 -1
index 35bb0bb3fe61..126b3c6e12f9 100644
@@ -822,6 +822,15 @@ nouveau_bo_move_m2mf(struct ttm_buffer_object *bo, int evict,
                if (ret == 0) {
                        ret = nouveau_fence_new(chan, false, &fence);
                        if (ret == 0) {
+                               /* TODO: figure out a better solution here
+                                *
+                                * wait on the fence here explicitly as going through
+                                * ttm_bo_move_accel_cleanup somehow doesn't seem to do it.
+                                *
+                                * Without this the operation can timeout and we'll fallback to a
+                                * software copy, which might take several minutes to finish.
+                                */
+                               nouveau_fence_wait(fence, false, false);
                                ret = ttm_bo_move_accel_cleanup(bo,
                                                                &fence->base,
                                                                evict, false,
--
Quoted 2 lines

2.37.1

Could it be possible to make land the aforementioned patch to the 5.10.x kernel version ? It is currently for >= 5.15.x kernel version only.

Thanks.