Use credential guards for scoped credential override with automatic
restoration on scope exit.

Signed-off-by: Christian Brauner <brauner@kernel.org>
---
 kernel/acct.c | 6 +-----
 1 file changed, 1 insertion(+), 5 deletions(-)

diff --git a/kernel/acct.c b/kernel/acct.c
index 61630110e29d..c1028f992529 100644
--- a/kernel/acct.c
+++ b/kernel/acct.c
@@ -520,12 +520,10 @@ static void fill_ac(struct bsd_acct_struct *acct)
 static void acct_write_process(struct bsd_acct_struct *acct)
 {
 	struct file *file = acct->file;
-	const struct cred *cred;
 	acct_t *ac = &acct->ac;
 
 	/* Perform file operations on behalf of whoever enabled accounting */
-	cred = override_creds(file->f_cred);
-
+	with_creds(file->f_cred);
 	/*
 	 * First check to see if there is enough free_space to continue
 	 * the process accounting system. Then get freeze protection. If
@@ -538,8 +536,6 @@ static void acct_write_process(struct bsd_acct_struct *acct)
 		__kernel_write(file, ac, sizeof(acct_t), &pos);
 		file_end_write(file);
 	}
-
-	revert_creds(cred);
 }
 
 static void do_acct_process(struct bsd_acct_struct *acct)

-- 
2.47.3

On Mon, 3 Nov 2025 at 20:27, Christian Brauner <brauner@kernel.org> wrote:
>
>         /* Perform file operations on behalf of whoever enabled accounting */
> -       cred = override_creds(file->f_cred);
> -
> +       with_creds(file->f_cred);

I'd almost prefer if we *only* did "scoped_with_creds()" and didn't
have this version at all.

Most of the cases want that anyway, and the couple of plain
"with_creds()" cases look like they would only be cleaned up by making
the cred scoping more explicit.

What do you think?

Anyway, I approve of the whole series, obviously, I just suspect we
could narrow down the new interface a bit more.

                Linus

On Tue, Nov 04, 2025 at 08:04:28AM +0900, Linus Torvalds wrote:
> On Mon, 3 Nov 2025 at 20:27, Christian Brauner <brauner@kernel.org> wrote:
> >
> >         /* Perform file operations on behalf of whoever enabled accounting */
> > -       cred = override_creds(file->f_cred);
> > -
> > +       with_creds(file->f_cred);
> 
> I'd almost prefer if we *only* did "scoped_with_creds()" and didn't
> have this version at all.
> 
> Most of the cases want that anyway, and the couple of plain
> "with_creds()" cases look like they would only be cleaned up by making
> the cred scoping more explicit.
> 
> What do you think?

Yeah, good idea. I reworked it all so now we're only left with:

scoped_with_creds()
scoped_with_kernel_creds()

It increases the indentation for about 3 cases but otherwise is safer.
It's all in:

https://git.kernel.org/pub/scm/linux/kernel/git/vfs/vfs.git/log/?h=kernel-6.19.cred

On Tue, Nov 4, 2025 at 12:04 AM Linus Torvalds
<torvalds@linux-foundation.org> wrote:
>
> On Mon, 3 Nov 2025 at 20:27, Christian Brauner <brauner@kernel.org> wrote:
> >
> >         /* Perform file operations on behalf of whoever enabled accounting */
> > -       cred = override_creds(file->f_cred);
> > -
> > +       with_creds(file->f_cred);
>
> I'd almost prefer if we *only* did "scoped_with_creds()" and didn't
> have this version at all.
>
> Most of the cases want that anyway, and the couple of plain
> "with_creds()" cases look like they would only be cleaned up by making
> the cred scoping more explicit.
>
> What do you think?

I had a similar reaction but for another reason.

The 'with' lingo reminds me of python with statement (e.g.
with open_file('example.txt', 'w') as file:), which implies a scope.
So in my head I am reading "with_creds" as with_creds_do.

Add to that the dubious practice (IMO) of scoped statements
without an explicit {} scope and this can become a source of
human brainos, but maybe the only problematic brain is mine..

Thanks,
Amir.