Skip to contentSkip to conversation

drm/amdkfd: Check for null pointer after calling kmemdup

PATCH In review Dragos-Marian Panait 4 participants 9 messages Jan 3, 2023 – Jan 5, 2023 Stable branch

lore
Share & tools
lore
#0 lore raw

cover drm/amdkfd: Check for null pointer after calling kmemdup

The following commit is needed to fix CVE-2022-3108:
https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git/commit/?id=abfaf0eee97925905e742aa3b0b72e04a918fa9e

Jiasheng Jiang (1):
drm/amdkfd: Check for null pointer after calling kmemdup

 drivers/gpu/drm/amd/amdkfd/kfd_crat.c | 3 +++
 1 file changed, 3 insertions(+)

base-commit: c652c812211c7a427d16be1d3f904eb02eb4265f

Signature · 1 line
2.38.1
#1 lore raw

PATCH 1/1 drm/amdkfd: Check for null pointer after calling kmemdup

From: Jiasheng Jiang <[email protected]>

[ Upstream commit abfaf0eee97925905e742aa3b0b72e04a918fa9e ]

As the possible failure of the allocation, kmemdup() may return NULL pointer.
Therefore, it should be better to check the 'props2' in order to prevent the dereference of NULL pointer.

Fixes 3a87177eb141 ("drm/amdkfd: Add topology support for dGPUs")
---
 drivers/gpu/drm/amd/amdkfd/kfd_crat.c | 3 +++
 1 file changed, 3 insertions(+)
drivers/gpu/drm/amd/amdkfd/kfd_crat.c +3 -0
index e2780643f4c3..b05ca3e639b1 100644
@@ -397,6 +397,9 @@ static int kfd_parse_subtype_iolink(struct crat_subtype_iolink *iolink,
 			return -ENODEV;
 		/* same everything but the other direction */
 		props2 = kmemdup(props, sizeof(*props2), GFP_KERNEL);
+		if (!props2)
+			return -ENOMEM;
+
 		props2->node_from = id_to;
 		props2->node_to = id_from;
 		props2->kobj = NULL;
Signature · 1 line
2.38.1
maintainer
#2 lore raw
Quoted 15 lines · Dragos-Marian Panait

From: Jiasheng Jiang <[email protected]>

[ Upstream commit abfaf0eee97925905e742aa3b0b72e04a918fa9e ]

As the possible failure of the allocation, kmemdup() may return NULL pointer.
Therefore, it should be better to check the 'props2' in order to prevent the dereference of NULL pointer.

Fixes 3a87177eb141 ("drm/amdkfd: Add topology support for dGPUs")
Quoted 3 lines
---
 drivers/gpu/drm/amd/amdkfd/kfd_crat.c | 3 +++
 1 file changed, 3 insertions(+)

For obvious reasons, I can't take a patch for 4.19.y and not newer kernel releases, right?

Please provide backports for all kernels if you really need to see this merged. And note, it's not a real bug at all, and given that a CVE was allocated for it that makes me want to even more reject it to show the whole folly of that mess.

thanks,

greg k-h

#3 lore raw

Am 04.01.23 um 13:41 schrieb Greg KH:

Quoted 19 lines · Dragos-Marian Panait
Quoted 15 lines

From: Jiasheng Jiang <[email protected]>

[ Upstream commit abfaf0eee97925905e742aa3b0b72e04a918fa9e ]

As the possible failure of the allocation, kmemdup() may return NULL pointer.
Therefore, it should be better to check the 'props2' in order to prevent the dereference of NULL pointer.

Fixes 3a87177eb141 ("drm/amdkfd: Add topology support for dGPUs")
Quoted 3 lines
---
  drivers/gpu/drm/amd/amdkfd/kfd_crat.c | 3 +++
  1 file changed, 3 insertions(+)
Quoted 7 lines

For obvious reasons, I can't take a patch for 4.19.y and not newer kernel releases, right?

Please provide backports for all kernels if you really need to see this merged. And note, it's not a real bug at all, and given that a CVE was allocated for it that makes me want to even more reject it to show the whole folly of that mess.

Well as far as I can see this is nonsense to back port.

The code in question is only used only once during driver load and then never again, that exactly this allocation fails while tons of other are made before and after is extremely unlikely.

It's nice to have it fixed in newer kernels, but not worth a backport and certainly not stuff for a CVE.

Regards,
Christian.

Quoted 4 lines

thanks,

greg k-h

#4 lore raw
Quoted 29 lines · Christian König

Am 04.01.23 um 13:41 schrieb Greg KH:

Quoted 19 lines · Dragos-Marian Panait
Quoted 15 lines

From: Jiasheng Jiang <[email protected]>

[ Upstream commit abfaf0eee97925905e742aa3b0b72e04a918fa9e ]

As the possible failure of the allocation, kmemdup() may return NULL pointer.
Therefore, it should be better to check the 'props2' in order to prevent the dereference of NULL pointer.

Fixes 3a87177eb141 ("drm/amdkfd: Add topology support for dGPUs")
Quoted 3 lines
---
  drivers/gpu/drm/amd/amdkfd/kfd_crat.c | 3 +++
  1 file changed, 3 insertions(+)
Quoted 7 lines

For obvious reasons, I can't take a patch for 4.19.y and not newer kernel releases, right?

Please provide backports for all kernels if you really need to see this merged. And note, it's not a real bug at all, and given that a CVE was allocated for it that makes me want to even more reject it to show the whole folly of that mess.

Quoted 8 lines

Well as far as I can see this is nonsense to back port.

The code in question is only used only once during driver load and then never again, that exactly this allocation fails while tons of other are made before and after is extremely unlikely.

It's nice to have it fixed in newer kernels, but not worth a backport and certainly not stuff for a CVE.

It's already fixed in Linus' tree:
https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git/commit/?id=abfaf0eee97925905e742aa3b0b72e04a918fa9e

Alex

Quoted 10 lines

Regards,
Christian.

Quoted 4 lines

thanks,

greg k-h

maintainer
#5 lore raw
Quoted 36 lines · Alex Deucher
Quoted 35 lines · Christian König

Am 04.01.23 um 13:41 schrieb Greg KH:

Quoted 19 lines · Dragos-Marian Panait
Quoted 15 lines

From: Jiasheng Jiang <[email protected]>

[ Upstream commit abfaf0eee97925905e742aa3b0b72e04a918fa9e ]

As the possible failure of the allocation, kmemdup() may return NULL pointer.
Therefore, it should be better to check the 'props2' in order to prevent the dereference of NULL pointer.

Fixes 3a87177eb141 ("drm/amdkfd: Add topology support for dGPUs")
Quoted 3 lines
---
  drivers/gpu/drm/amd/amdkfd/kfd_crat.c | 3 +++
  1 file changed, 3 insertions(+)
Quoted 7 lines

For obvious reasons, I can't take a patch for 4.19.y and not newer kernel releases, right?

Please provide backports for all kernels if you really need to see this merged. And note, it's not a real bug at all, and given that a CVE was allocated for it that makes me want to even more reject it to show the whole folly of that mess.

Well as far as I can see this is nonsense to back port.

The code in question is only used only once during driver load and then never again, that exactly this allocation fails while tons of other are made before and after is extremely unlikely.

Quoted 5 lines
Quoted 2 lines

It's nice to have it fixed in newer kernels, but not worth a backport and certainly not stuff for a CVE.

It's already fixed in Linus' tree:
https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git/commit/?id=abfaf0eee97925905e742aa3b0b72e04a918fa9e

Yes, that's what the above commit shows...

confused,

greg k-h

#6 lore raw
Quoted 37 lines · Greg KH
Quoted 36 lines · Alex Deucher
Quoted 35 lines · Christian König

Am 04.01.23 um 13:41 schrieb Greg KH:

Quoted 19 lines · Dragos-Marian Panait
Quoted 15 lines

From: Jiasheng Jiang <[email protected]>

[ Upstream commit abfaf0eee97925905e742aa3b0b72e04a918fa9e ]

As the possible failure of the allocation, kmemdup() may return NULL pointer.
Therefore, it should be better to check the 'props2' in order to prevent the dereference of NULL pointer.

Fixes 3a87177eb141 ("drm/amdkfd: Add topology support for dGPUs")
Quoted 3 lines
---
   drivers/gpu/drm/amd/amdkfd/kfd_crat.c | 3 +++
   1 file changed, 3 insertions(+)
Quoted 7 lines

For obvious reasons, I can't take a patch for 4.19.y and not newer kernel releases, right?

Please provide backports for all kernels if you really need to see this merged. And note, it's not a real bug at all, and given that a CVE was allocated for it that makes me want to even more reject it to show the whole folly of that mess.

Well as far as I can see this is nonsense to back port.

The code in question is only used only once during driver load and then never again, that exactly this allocation fails while tons of other are made before and after is extremely unlikely.

It's nice to have it fixed in newer kernels, but not worth a backport and certainly not stuff for a CVE.

Quoted 7 lines
Quoted 2 lines

It's already fixed in Linus' tree:
https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git/commit/?id=abfaf0eee97925905e742aa3b0b72e04a918fa9e

Yes, that's what the above commit shows...

confused,

greg k-h

Just for completeness, I also sent out patches for 5.4 and 5.10 stable branches.
5.15 stable branch already has this change: https://git.kernel.org/pub/scm/linux/kernel/git/stable/linux.git/commit/?h=linux-5.15.y&id=5609b7803947eea1711516dd8659c7ed39f5a868

Dragos

maintainer
#7 lore raw
Quoted 41 lines · Dragos-Marian Panait
Quoted 39 lines · Greg KH
Quoted 38 lines · Alex Deucher
Quoted 27 lines · Christian König

Am 04.01.23 um 13:41 schrieb Greg KH:

Quoted 19 lines · Dragos-Marian Panait
Quoted 15 lines

From: Jiasheng Jiang <[email protected]>

[ Upstream commit abfaf0eee97925905e742aa3b0b72e04a918fa9e ]

As the possible failure of the allocation, kmemdup() may return NULL pointer.
Therefore, it should be better to check the 'props2' in order to prevent the dereference of NULL pointer.

Fixes 3a87177eb141 ("drm/amdkfd: Add topology support for dGPUs")
Quoted 3 lines
---
   drivers/gpu/drm/amd/amdkfd/kfd_crat.c | 3 +++
   1 file changed, 3 insertions(+)
Quoted 7 lines

For obvious reasons, I can't take a patch for 4.19.y and not newer kernel releases, right?

Please provide backports for all kernels if you really need to see this merged. And note, it's not a real bug at all, and given that a CVE was allocated for it that makes me want to even more reject it to show the whole folly of that mess.

Quoted 8 lines

Well as far as I can see this is nonsense to back port.

The code in question is only used only once during driver load and then never again, that exactly this allocation fails while tons of other are made before and after is extremely unlikely.

It's nice to have it fixed in newer kernels, but not worth a backport and certainly not stuff for a CVE.

It's already fixed in Linus' tree:
https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git/commit/?id=abfaf0eee97925905e742aa3b0b72e04a918fa9e

Quoted 8 lines
Quoted 5 lines

Yes, that's what the above commit shows...

confused,

greg k-h

Just for completeness, I also sent out patches for 5.4 and 5.10 stable
branches.
5.15 stable branch already has this change: https://git.kernel.org/pub/scm/linux/kernel/git/stable/linux.git/commit/?h=linux-5.15.y&id=5609b7803947eea1711516dd8659c7ed39f5a868

Again, this is not a real bug and someone needs to go and invalidate that CVE so you don't have to worry about it anymore. I suggest that you do that if your company cares about tracking CVEs.

thanks,

greg k-h

maintainer
#8 lore raw
In reply to Dragos-Marian Panait
Quoted 1 line · Dragos-Marian Panait

The following commit is needed to fix CVE-2022-3108:

That's a funny cve, given that you can not ever trigger it in a system, right? Why was a CVE allocated for that?

{sigh}