2020-12-07 00:05:40

by Jann Horn

Subject: [PATCH pidfd] signal: Add missing __user annotation to copy_siginfo_from_user_any

copy_siginfo_from_user_any() takes a userspace pointer as second
argument; annotate the parameter type accordingly.

Signed-off-by: Jann Horn <[email protected]>
---
I'm messing around with clang's version of __user annotation checking
and it spotted this issue:

kernel/signal.c:3759:44: warning: casting to dereferenceable pointer removes 'noderef' attribute [-Wnoderef]
ret = copy_siginfo_from_user_any(&kinfo, info);
^~~~
Untracked cast to function pointer at kernel/signal.c:4294:26


Christian, since this is pidfd code, can you take this through your tree?
Or should I send this to akpm (or someone else)?

kernel/signal.c | 3 ++-
1 file changed, 2 insertions(+), 1 deletion(-)

diff --git a/kernel/signal.c b/kernel/signal.c
index ef8f2a28d37c..4693191dc17c 100644
--- a/kernel/signal.c
+++ b/kernel/signal.c
@@ -3685,7 +3685,8 @@ static bool access_pidfd_pidns(struct pid *pid)
return true;
}

-static int copy_siginfo_from_user_any(kernel_siginfo_t *kinfo, siginfo_t *info)
+static int copy_siginfo_from_user_any(kernel_siginfo_t *kinfo,
+ siginfo_t __user *info)
{
#ifdef CONFIG_COMPAT
/*

base-commit: 0477e92881850d44910a7e94fc2c46f96faa131f
--
2.29.2.576.ga3fc446d84-goog

2020-12-07 10:36:00

by Christian Brauner

Subject: Re: [PATCH pidfd] signal: Add missing __user annotation to copy_siginfo_from_user_any

On Mon, Dec 07, 2020 at 01:02:52AM +0100, Jann Horn wrote:
> copy_siginfo_from_user_any() takes a userspace pointer as second
> argument; annotate the parameter type accordingly.
>
> Signed-off-by: Jann Horn <[email protected]>
> ---
> I'm messing around with clang's version of __user annotation checking
> and it spotted this issue:
>
> kernel/signal.c:3759:44: warning: casting to dereferenceable pointer removes 'noderef' attribute [-Wnoderef]
> ret = copy_siginfo_from_user_any(&kinfo, info);
> ^~~~
> Untracked cast to function pointer at kernel/signal.c:4294:26
>
>
> Christian, since this is pidfd code, can you take this through your tree?
> Or should I send this to akpm (or someone else)?

I'm picking it up now. :) Thanks!

Christian