Hello arm-soc maintainers,
Please pull this patch removing the refcounting of kernel pages within the TEE shared memory helper.
Thanks, Jens
The following changes since commit 6de23f81a5e08be8fbf5e8d7e9febc72a5b5f27f:
Linux 7.0-rc1 (2026-02-22 13:18:59 -0800)
are available in the Git repository at:
git://git.kernel.org/pub/scm/linux/kernel/git/jenswi/linux-tee.git tags/tee-fix-for-v7.0
for you to fetch changes up to 08d9a4580f71120be3c5b221af32dca00a48ceb0:
tee: shm: Remove refcounting of kernel pages (2026-03-03 09:03:04 +0100)
---------------------------------------------------------------- TEE shared memory update for 7.0
Remove refcounting of kernel pages in register_shm_helper() to support slab allocations.
---------------------------------------------------------------- Matthew Wilcox (1): tee: shm: Remove refcounting of kernel pages
drivers/tee/tee_shm.c | 27 --------------------------- 1 file changed, 27 deletions(-)
On Mon, Mar 16, 2026 at 08:41:50AM +0100, Jens Wiklander wrote:
Hello arm-soc maintainers,
Please pull this patch removing the refcounting of kernel pages within the TEE shared memory helper.
Thanks, Jens
The following changes since commit 6de23f81a5e08be8fbf5e8d7e9febc72a5b5f27f:
Linux 7.0-rc1 (2026-02-22 13:18:59 -0800)
are available in the Git repository at:
git://git.kernel.org/pub/scm/linux/kernel/git/jenswi/linux-tee.git tags/tee-fix-for-v7.0
for you to fetch changes up to 08d9a4580f71120be3c5b221af32dca00a48ceb0:
tee: shm: Remove refcounting of kernel pages (2026-03-03 09:03:04 +0100)
TEE shared memory update for 7.0
Remove refcounting of kernel pages in register_shm_helper() to support slab allocations.
That's a supposed fix for a year old commit, not current RC, and the commit msg does not explain what is the bug being fixed here. Looks like it is doing improvement, not a fix. Can you provide here more explanation of the actual bug being fixed to justify this going to current RC fixes?
Best regards, Krzysztof
Hi,
On Tue, Mar 17, 2026 at 9:51 AM Krzysztof Kozlowski krzk@kernel.org wrote:
On Mon, Mar 16, 2026 at 08:41:50AM +0100, Jens Wiklander wrote:
Hello arm-soc maintainers,
Please pull this patch removing the refcounting of kernel pages within the TEE shared memory helper.
Thanks, Jens
The following changes since commit 6de23f81a5e08be8fbf5e8d7e9febc72a5b5f27f:
Linux 7.0-rc1 (2026-02-22 13:18:59 -0800)
are available in the Git repository at:
git://git.kernel.org/pub/scm/linux/kernel/git/jenswi/linux-tee.git tags/tee-fix-for-v7.0
for you to fetch changes up to 08d9a4580f71120be3c5b221af32dca00a48ceb0:
tee: shm: Remove refcounting of kernel pages (2026-03-03 09:03:04 +0100)
TEE shared memory update for 7.0
Remove refcounting of kernel pages in register_shm_helper() to support slab allocations.
That's a supposed fix for a year old commit, not current RC, and the commit msg does not explain what is the bug being fixed here. Looks like it is doing improvement, not a fix. Can you provide here more explanation of the actual bug being fixed to justify this going to current RC fixes?
I'm under the impression that it is a fix rather than an improvement. However, since it has been stewing for a while, it's clearly not urgent. I don't have anything specific on the bug fix, other than that we're supposed to stop taking those refcounts. Perhaps Sumit, Marco, or Matthew can add something?
Cheers, Jens
On 26-03-17, Jens Wiklander wrote:
Hi,
On Tue, Mar 17, 2026 at 9:51 AM Krzysztof Kozlowski krzk@kernel.org wrote:
On Mon, Mar 16, 2026 at 08:41:50AM +0100, Jens Wiklander wrote:
Hello arm-soc maintainers,
Please pull this patch removing the refcounting of kernel pages within the TEE shared memory helper.
Thanks, Jens
The following changes since commit 6de23f81a5e08be8fbf5e8d7e9febc72a5b5f27f:
Linux 7.0-rc1 (2026-02-22 13:18:59 -0800)
are available in the Git repository at:
git://git.kernel.org/pub/scm/linux/kernel/git/jenswi/linux-tee.git tags/tee-fix-for-v7.0
for you to fetch changes up to 08d9a4580f71120be3c5b221af32dca00a48ceb0:
tee: shm: Remove refcounting of kernel pages (2026-03-03 09:03:04 +0100)
TEE shared memory update for 7.0
Remove refcounting of kernel pages in register_shm_helper() to support slab allocations.
That's a supposed fix for a year old commit, not current RC, and the commit msg does not explain what is the bug being fixed here. Looks like it is doing improvement, not a fix. Can you provide here more explanation of the actual bug being fixed to justify this going to current RC fixes?
I'm under the impression that it is a fix rather than an improvement. However, since it has been stewing for a while, it's clearly not urgent. I don't have anything specific on the bug fix, other than that we're supposed to stop taking those refcounts. Perhaps Sumit, Marco, or Matthew can add something?
It's a long outstanding fix regarding memory page management. The userspace gets informed about this memory API misuse via a warning print incl. a backtrace. This gets triggered only if userspace uses OP-TEE trusted-keys or the OP-TEE EFI variables. Therefore I would like to get it into 7.0 but 7.1 fine too.
Regards, Marco
Cheers, Jens
On Tue, Mar 17, 2026, at 13:53, Marco Felsch wrote:
On 26-03-17, Jens Wiklander wrote:
I'm under the impression that it is a fix rather than an improvement. However, since it has been stewing for a while, it's clearly not urgent. I don't have anything specific on the bug fix, other than that we're supposed to stop taking those refcounts. Perhaps Sumit, Marco, or Matthew can add something?
It's a long outstanding fix regarding memory page management. The userspace gets informed about this memory API misuse via a warning print incl. a backtrace. This gets triggered only if userspace uses OP-TEE trusted-keys or the OP-TEE EFI variables. Therefore I would like to get it into 7.0 but 7.1 fine too.
Yes, that is clearly a bugfix. I think we'd also want this to be marked 'Cc: stable@vger.kernel.org' as it should also apply to 6.18-LTS and 6.19. Greg tends to also pick up bugfixes that don't have this, but it's generally better to be explicit here, especially when it's not obvious what the impact of the fix is.
I've merge the pull request now. If you think this should be in 6.18-lts and Greg doesn't pick it up by himself, maybe let him know about it.
Arnd
On Wed, Mar 18, 2026 at 01:39:41PM +0100, Arnd Bergmann wrote:
On Tue, Mar 17, 2026, at 13:53, Marco Felsch wrote:
On 26-03-17, Jens Wiklander wrote:
I'm under the impression that it is a fix rather than an improvement. However, since it has been stewing for a while, it's clearly not urgent. I don't have anything specific on the bug fix, other than that we're supposed to stop taking those refcounts. Perhaps Sumit, Marco, or Matthew can add something?
It's a long outstanding fix regarding memory page management. The userspace gets informed about this memory API misuse via a warning print incl. a backtrace. This gets triggered only if userspace uses OP-TEE trusted-keys or the OP-TEE EFI variables. Therefore I would like to get it into 7.0 but 7.1 fine too.
Yes, that is clearly a bugfix. I think we'd also want this to be marked 'Cc: stable@vger.kernel.org' as it should also apply to 6.18-LTS and 6.19. Greg tends to also pick up bugfixes that don't have this, but it's generally better to be explicit here, especially when it's not obvious what the impact of the fix is.
I've merge the pull request now. If you think this should be in 6.18-lts and Greg doesn't pick it up by himself, maybe let him know about it.
Thanks Arnd for picking up this fix. Last week I was on business travel hence the delay in my response. I see Sasha already picked it up for backports.
-Sumit
op-tee@lists.trustedfirmware.org
-
Arnd Bergmann -
Jens Wiklander -
Krzysztof Kozlowski -
Marco Felsch -
Sumit Garg