Re: [PATCH] arm64/gcs: Don't call gcs_free() during flush_gcs()

[Catalin Marinas]

On Thu, Jun 12, 2025 at 12:40:42PM +0100, Mark Brown wrote:
> On Wed, Jun 11, 2025 at 06:34:15PM +0100, Catalin Marinas wrote:
> > However, I thought there was another slightly misplaced call to
> > gcs_free() via arch_release_task_struct(). I wouldn't touch the user
> > memory with vm_munmap() when releasing a task structure. Is this needed
> > because the shadow stack is allocated automatically on thread creation,
> > so we need something to free it when the thread died?
> 
> Yeah, I've got another patch written but not sent for that (since it
> doesn't actually overlap) but like you say I need to check that things
> are joined up for threads that had a GCS created automatically for
> compatibility before I send it out.

OK. For this patch:

Reviewed-by: Catalin Marinas <catalin.marinas@xxxxxxx>

> > Another caller of gcs_free() is deactivate_mm(). It's not clear to me
> > when we need to free the shadow stack on this path. On the exit_mm()
> > path for example we have mmput() -> exit_mmap() that takes care of
> > unmapping everything. Similarly on the exec_mmap() path.
> 
> We need that one to clean up the GCS for threads that had it allocated
> for compatibility, you can see the leak that results without it easily
> with the glibc testsuite (or anything else that does threads, the glibc
> tests just spot it).  Most of the checking for arch_release_task_struct()
> is verifying that deactivate_mm() is guaranteed to be called eveywhere
> it's relevant, I need to page that back in.

Makes sense. I think we should only keep gcs_free() in one place,
ideally deactivate_mm() as that's more related to mm rather than the
task_struct.

-- 
Catalin