Re: [PATCH] rseq: fix using an uninitialized stack variable in rseq_exit_user_update

[Peter Zijlstra <peterz@infradead.org>]

On Mon, Jun 01, 2026 at 09:59:35AM -0400, Mathieu Desnoyers wrote:
> On 2026-05-31 22:13, Qing Wang wrote:
> > There is an bug which is an uninitialized stack variable use in
> > `rseq_exit_user_update()` reported by syzbot:
> > 
> > BUG: KMSAN: kernel-infoleak in rseq_set_ids_get_csaddr include/linux/rseq_entry.h:502 [inline]
> > 
> > The local variable:
> > ```c
> > 	struct rseq_ids ids = {
> > 		.cpu_id	 = task_cpu(t),
> > 		.mm_cid	 = task_mm_cid(t),
> > 		.node_id = cpu_to_node(ids.cpu_id),
> > 	};
> > ```
> > According to the C standard, the evaluation order of expressions in an
> > initializer list is indeterminately sequenced. The compiler (Clang, in this
> > KMSAN build) evaluates `cpu_to_node(ids.cpu_id)` *before* `ids.cpu_id` is
> > initialized with `task_cpu(t)`.

W.T.F. would they (C committee) do that :-(, and since this is
apparently a known thing, why aren't the compilers issuing WARNs about
this pattern?

Clearly any non address-of self reference in a struct initializer is
buggered.

And ideally it would actually observe the data-dependency and respect
it, irrespective of the lunacy of the C standard.

> > This is fixed by moving the assignment of ids.node_id outside the structure
> > initialization.
> > 
> > Reported-by: syzbot+185a631927096f9da2fc@syzkaller.appspotmail.com
> > Closes: https://syzkaller.appspot.com/bug?extid=185a631927096f9da2fc
> > Fixes: 82f572449cfe ("rseq: Implement read only ABI enforcement for optimized RSEQ V2 mode")
> > Signed-off-by: Qing Wang <wangqing7171@gmail.com>
> > ---
> >   include/linux/rseq_entry.h | 2 +-
> >   1 file changed, 1 insertion(+), 1 deletion(-)
> > 
> > diff --git a/include/linux/rseq_entry.h b/include/linux/rseq_entry.h
> > index 63bc72086e75..e05f4c18f39e 100644
> > --- a/include/linux/rseq_entry.h
> > +++ b/include/linux/rseq_entry.h
> > @@ -638,8 +638,8 @@ static __always_inline bool rseq_exit_user_update(struct pt_regs *regs, struct t
> >   	struct rseq_ids ids = {
> >   		.cpu_id	 = task_cpu(t),
> >   		.mm_cid	 = task_mm_cid(t),
> > -		.node_id = cpu_to_node(ids.cpu_id),
> >   	};
> > +	ids.node_id = cpu_to_node(ids.cpu_id);
> 
> I find the fix odd. How about:
> 
> int cpu = task_cpu(t);
> struct rseq_ids ids = {
> 	.cpu_id = cpu,
> 	.mm_cid = task_mm_cid(t),
> 	.node_id = cpu_to_node(cpu),
> };
> 
> instead ?

Yes, much saner.