Skip to contentSkip to conversation

Various minor fixes to Nouveau

PATCH In review Timur Tabi 3 participants 10 messages Aug 4, 2025 – Aug 23, 2025

Subsystems (3)
lore
Share & tools
lore

Timur Tabi

author
#0 lore raw

cover Various minor fixes to Nouveau

Three minor fixes to Nouveau, discovered while working on Nova.

Timur Tabi (3):
drm/nouveau: fix error path in nvkm_gsp_fwsec_v2 drm/nouveau: remove unused increment in gm200_flcn_pio_imem_wr drm/nouveau: remove unused memory target test

 drivers/gpu/drm/nouveau/nvkm/falcon/gm200.c     | 15 ++++-----------
 drivers/gpu/drm/nouveau/nvkm/subdev/gsp/fwsec.c |  5 +++--
 2 files changed, 7 insertions(+), 13 deletions(-)

base-commit: 6531a2cf07ef156956840853692755cc7e1621b7

Signature · 1 line
2.43.0
author
#1 lore raw

PATCH v2 1/3 drm/nouveau: fix error path in nvkm_gsp_fwsec_v2

Function nvkm_gsp_fwsec_v2() sets 'ret' if the kmemdup() call fails, but it never uses or returns 'ret' after that point. We always need to release the firmware regardless, so do that and then check for error.

Fixes 176fdcbddfd2 ("drm/nouveau/gsp/r535: add support for booting GSP-RM")
---
 drivers/gpu/drm/nouveau/nvkm/subdev/gsp/fwsec.c | 5 +++--
 1 file changed, 3 insertions(+), 2 deletions(-)
drivers/gpu/drm/nouveau/nvkm/subdev/gsp/fwsec.c +3 -2
index 52412965fac1..5b721bd9d799 100644
@@ -209,11 +209,12 @@ nvkm_gsp_fwsec_v2(struct nvkm_gsp *gsp, const char *name,
 	fw->boot_addr = bld->start_tag << 8;
 	fw->boot_size = bld->code_size;
 	fw->boot = kmemdup(bl->data + hdr->data_offset + bld->code_off, fw->boot_size, GFP_KERNEL);
-	if (!fw->boot)
-		ret = -ENOMEM;
 
 	nvkm_firmware_put(bl);
 
+	if (!fw->boot)
+		return -ENOMEM;
+
 	/* Patch in interface data. */
 	return nvkm_gsp_fwsec_patch(gsp, fw, desc->InterfaceOffset, init_cmd);
 }
Signature · 1 line
2.43.0
#2 lore raw

Re: [PATCH v2 1/3] drm/nouveau: fix error path in nvkm_gsp_fwsec_v2

Quoted 9 lines · Timur Tabi

Function nvkm_gsp_fwsec_v2() sets 'ret' if the kmemdup() call fails, but
it never uses or returns 'ret' after that point.  We always need to release
the firmware regardless, so do that and then check for error.

Fixes 176fdcbddfd2 ("drm/nouveau/gsp/r535: add support for booting

GSP-RM")

That looks like a bug (leak?) to me.

+Cc stable?

P.

Quoted 27 lines

---
 drivers/gpu/drm/nouveau/nvkm/subdev/gsp/fwsec.c | 5 +++--

 1 file changed, 3 insertions(+), 2 deletions(-)
drivers/gpu/drm/nouveau/nvkm/subdev/gsp/fwsec.c +0 -0
index 52412965fac1..5b721bd9d799 100644
@@ -209,11 +209,12 @@ nvkm_gsp_fwsec_v2(struct nvkm_gsp *gsp, const
 char *name,
  	fw->boot_addr = bld->start_tag << 8;
  	fw->boot_size = bld->code_size;

  fw->boot = kmemdup(bl->data + hdr->data_offset + bld-

Quoted 1 line

code_off, fw->boot_size, GFP_KERNEL);

drivers/gpu/drm/nouveau/nvkm/subdev/gsp/fwsec.c +0 -2
-	if (!fw->boot)
-		ret = -ENOMEM;


  nvkm_firmware_put(bl);

drivers/gpu/drm/nouveau/nvkm/subdev/gsp/fwsec.c +3 -0
+	if (!fw->boot)
+		return -ENOMEM;
+

  /* Patch in interface data. */
  return nvkm_gsp_fwsec_patch(gsp, fw, desc->InterfaceOffset, init_cmd);
 }

maintainer
#3 lore raw

Re: [PATCH v2 1/3] drm/nouveau: fix error path in nvkm_gsp_fwsec_v2

In reply to Timur Tabi
Quoted 5 lines · Timur Tabi

Function nvkm_gsp_fwsec_v2() sets 'ret' if the kmemdup() call fails, but it never uses or returns 'ret' after that point. We always need to release the firmware regardless, so do that and then check for error.

Fixes 176fdcbddfd2 ("drm/nouveau/gsp/r535: add support for booting GSP-RM")

As mentioned by Philipp, we should Cc stable for this one.

Quoted 10 lines
---
  drivers/gpu/drm/nouveau/nvkm/subdev/gsp/fwsec.c | 5 +++--
  1 file changed, 3 insertions(+), 2 deletions(-)
drivers/gpu/drm/nouveau/nvkm/subdev/gsp/fwsec.c +0 -0
index 52412965fac1..5b721bd9d799 100644

drivers/gpu/drm/nouveau/nvkm/subdev/gsp/fwsec.c nvkm_gsp_fwsec_v2(struct nvkm_gsp *gsp, const char *name, · line 216

Quoted 10 lines
drivers/gpu/drm/nouveau/nvkm/subdev/gsp/fwsec.c +2 -2
@@ -209,11 +209,12 @@ nvkm_gsp_fwsec_v2(struct nvkm_gsp *gsp, const char *name,
  	fw->boot_addr = bld->start_tag << 8;
  	fw->boot_size = bld->code_size;
  	fw->boot = kmemdup(bl->data + hdr->data_offset + bld->code_off, fw->boot_size, GFP_KERNEL);
-	if (!fw->boot)
-		ret = -ENOMEM;
 
  	nvkm_firmware_put(bl);
 
+	if (!fw->boot)
+		return -ENOMEM;

Good catch! It's also good that you moved the return below the nvkm_firmware_put() call.

But don't we also need to revert the preceding call to nvkm_falcon_fw_ctor()?

Quoted 4 lines
drivers/gpu/drm/nouveau/nvkm/subdev/gsp/fwsec.c +1 -0
+
  	/* Patch in interface data. */
  	return nvkm_gsp_fwsec_patch(gsp, fw, desc->InterfaceOffset, init_cmd);
  }
author
#4 lore raw

Re: [PATCH v2 1/3] drm/nouveau: fix error path in nvkm_gsp_fwsec_v2

Quoted 7 lines · Danilo Krummrich

drivers/gpu/drm/nouveau/nvkm/subdev/gsp/fwsec.c nvkm_gsp_fwsec_v2(struct nvkm_gsp *gsp, const char *name, · line 216

Quoted 2 lines
drivers/gpu/drm/nouveau/nvkm/subdev/gsp/fwsec.c +2 -0
+	if (!fw->boot)
+		return -ENOMEM;

Good catch! It's also good that you moved the return below the nvkm_firmware_put() call.

But don't we also need to revert the preceding call to nvkm_falcon_fw_ctor()?

I don't know. I cannot decipher Nouveau's cleanup code.

If it does need to be reverted, then we have that problem in a lot of places, in both nvkm_gsp_fwsec_v2() and nvkm_gsp_fwsec_v3().

At one point Ben told me that Nouveau cleans itself up on failure, eliminating the need to clean up on every "return ret" exit.  You can see this through fwsec.c -- none of the return-error calls do any cleanup.

I tried to test this, but I didn't get far. I modified nvkm_gsp_fwsec_v2 to return failure and did a trace of nvkm_falcon_fw_ctor() and nvkm_falcon_fw_ctor(), and this is what I found:

[ 1281.361048] nvkm_falcon_fw_ctor::204 fw=ffff8e898022f540 (null)                                 
[ 1281.367036] nvkm_falcon_fw_ctor::204 fw=ffff8e898022f640 (null)                                 
[ 1281.379502] nvkm_falcon_fw_ctor::204 fw=ffffcfc3408e7908 (null)                                 
[ 1281.422846] nvkm_falcon_fw_dtor::157 fw=ffff8e89812480d0 (null)                                 
[ 1281.429326] nvkm_falcon_fw_dtor::157 fw=ffff8e898022f640 booter-unload                          
[ 1281.435884] nvkm_falcon_fw_dtor::157 fw=ffff8e898022f540 booter-load  

What's throwing me off is:

[ 1281.379502] nvkm_falcon_fw_ctor::204 fw=ffffcfc3408e7908 (null)                                 
[ 1281.422846] nvkm_falcon_fw_dtor::157 fw=ffff8e89812480d0 (null)                                 

These two should match. For some reason, there is no dtor call for 7908, but more disturbingly, there is no ctor call for 80d0. This doesn't happen if nvkm_gsp_fwsec_v2 succeeds.

author
#5 lore raw

PATCH v2 2/3 drm/nouveau: remove unused increment in gm200_flcn_pio_imem_wr

In reply to Timur Tabi

The 'tag' parameter is passed by value and is not actually used after being incremented, so remove the increment. It's the function that calls gm200_flcn_pio_imem_wr that is supposed to (and does) increment 'tag'.

---
 drivers/gpu/drm/nouveau/nvkm/falcon/gm200.c | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)
drivers/gpu/drm/nouveau/nvkm/falcon/gm200.c +1 -1
index b7da3ab44c27..6a004c6e6742 100644
@@ -103,7 +103,7 @@ gm200_flcn_pio_imem_wr_init(struct nvkm_falcon *falcon, u8 port, bool sec, u32 i
 static void
 gm200_flcn_pio_imem_wr(struct nvkm_falcon *falcon, u8 port, const u8 *img, int len, u16 tag)
 {
-	nvkm_falcon_wr32(falcon, 0x188 + (port * 0x10), tag++);
+	nvkm_falcon_wr32(falcon, 0x188 + (port * 0x10), tag);
 	while (len >= 4) {
 		nvkm_falcon_wr32(falcon, 0x184 + (port * 0x10), *(u32 *)img);
 		img += 4;
Signature · 1 line
2.43.0
Reviewed-by on v2 2/3
#6 lore raw

Re: [PATCH v2 2/3] drm/nouveau: remove unused increment in gm200_flcn_pio_imem_wr

Quoted 5 lines · Timur Tabi

The 'tag' parameter is passed by value and is not actually used after being incremented, so remove the increment.  It's the function that calls gm200_flcn_pio_imem_wr that is supposed to (and does) increment 'tag'.

Quoted 17 lines

---
 drivers/gpu/drm/nouveau/nvkm/falcon/gm200.c | 2 +-

 1 file changed, 1 insertion(+), 1 deletion(-)
drivers/gpu/drm/nouveau/nvkm/falcon/gm200.c +0 -0
index b7da3ab44c27..6a004c6e6742 100644
@@ -103,7 +103,7 @@ gm200_flcn_pio_imem_wr_init(struct nvkm_falcon *falcon, u8 port, bool sec, u32 i

 static void
 gm200_flcn_pio_imem_wr(struct nvkm_falcon *falcon, u8 port, const u8 *img, int len, u16 tag) {

drivers/gpu/drm/nouveau/nvkm/falcon/gm200.c +1 -1
-	nvkm_falcon_wr32(falcon, 0x188 + (port * 0x10), tag++);
+	nvkm_falcon_wr32(falcon, 0x188 + (port * 0x10), tag);

  while (len >= 4) {
  nvkm_falcon_wr32(falcon, 0x184 + (port * 0x10), *(u32 *)img); img += 4;

author
#7 lore raw

PATCH v2 3/3 drm/nouveau: remove unused memory target test

In reply to Timur Tabi

The memory target check is a hold-over from a refactor. It's harmless but distracting, so just remove it.

---
 drivers/gpu/drm/nouveau/nvkm/falcon/gm200.c | 13 +++----------
 1 file changed, 3 insertions(+), 10 deletions(-)
drivers/gpu/drm/nouveau/nvkm/falcon/gm200.c +3 -10
index 6a004c6e6742..7c43397c19e6 100644
@@ -249,9 +249,11 @@ int
 gm200_flcn_fw_load(struct nvkm_falcon_fw *fw)
 {
 	struct nvkm_falcon *falcon = fw->falcon;
-	int target, ret;
+	int ret;
 
 	if (fw->inst) {
+		int target;
+
 		nvkm_falcon_mask(falcon, 0x048, 0x00000001, 0x00000001);
 
 		switch (nvkm_memory_target(fw->inst)) {
@@ -285,15 +287,6 @@ gm200_flcn_fw_load(struct nvkm_falcon_fw *fw)
 	}
 
 	if (fw->boot) {
-		switch (nvkm_memory_target(&fw->fw.mem.memory)) {
-		case NVKM_MEM_TARGET_VRAM: target = 4; break;
-		case NVKM_MEM_TARGET_HOST: target = 5; break;
-		case NVKM_MEM_TARGET_NCOH: target = 6; break;
-		default:
-			WARN_ON(1);
-			return -EINVAL;
-		}
-
 		ret = nvkm_falcon_pio_wr(falcon, fw->boot, 0, 0,
 					 IMEM, falcon->code.limit - fw->boot_size, fw->boot_size,
 					 fw->boot_addr >> 8, false);
Signature · 1 line
2.43.0
maintainer
#8 lore raw
In reply to Timur Tabi
Quoted 1 line · Timur Tabi

Three minor fixes to Nouveau, discovered while working on Nova.

Thanks for this series!

Quoted 4 lines

Timur Tabi (3):
drm/nouveau: fix error path in nvkm_gsp_fwsec_v2 drm/nouveau: remove unused increment in gm200_flcn_pio_imem_wr drm/nouveau: remove unused memory target test

Please also add a Fixes tag for the last two patches.

author
#9 lore raw
Quoted 11 lines · Danilo Krummrich
Quoted 1 line · Timur Tabi

Three minor fixes to Nouveau, discovered while working on Nova.

Thanks for this series!

Quoted 4 lines

Timur Tabi (3):
   drm/nouveau: fix error path in nvkm_gsp_fwsec_v2 drm/nouveau: remove unused increment in gm200_flcn_pio_imem_wr drm/nouveau: remove unused memory target test

Please also add a Fixes tag for the last two patches.

Done. Please see v3, posted on the 12th.