Re: [syzbot] [fs?] [mm?] WARNING in path_noexec (2)

[Christian Brauner <brauner@kernel.org>]

On Tue, Sep 30, 2025 at 03:04:22PM -0700, Kees Cook wrote:
> On Tue, Sep 30, 2025 at 01:17:34PM -0700, syzbot wrote:
> > Reported-by: syzbot+a9391462075ffb9f77c6@syzkaller.appspotmail.com
> > 
> > ------------[ cut here ]------------
> > WARNING: CPU: 1 PID: 6000 at fs/exec.c:119 path_noexec+0x1af/0x200 fs/exec.c:118
> 
> Christian, this is:
> 
> bool path_noexec(const struct path *path)
> {
>         /* If it's an anonymous inode make sure that we catch any shenanigans. */
>         VFS_WARN_ON_ONCE(IS_ANON_FILE(d_inode(path->dentry)) &&
>                          !(path->mnt->mnt_sb->s_iflags & SB_I_NOEXEC));
>         return (path->mnt->mnt_flags & MNT_NOEXEC) ||
>                (path->mnt->mnt_sb->s_iflags & SB_I_NOEXEC);
> }
> 
> > C reproducer:   https://syzkaller.appspot.com/x/repro.c?x=13e5fd6f980000
> 
> I think is from the created fd_dma_buf. I expect this would fix it:
> 
> 
> diff --git a/drivers/dma-buf/dma-buf.c b/drivers/dma-buf/dma-buf.c
> index 2bcf9ceca997..6e2ab1a4560d 100644
> --- a/drivers/dma-buf/dma-buf.c
> +++ b/drivers/dma-buf/dma-buf.c
> @@ -189,6 +189,8 @@ static int dma_buf_fs_init_context(struct fs_context *fc)
>  {
>  	struct pseudo_fs_context *ctx;
>  
> +	fc->s_iflags |= SB_I_NOEXEC;
> +	fc->s_iflags |= SB_I_NODEV;

Yeah, that seems like a good thing to do.
I'm quite happy that the VFS_WARN_ON_ONCE() in there is catching all
this!

Do you want to send a real patch I can pick up?

>  	ctx = init_pseudo(fc, DMA_BUF_MAGIC);
>  	if (!ctx)
>  		return -ENOMEM;
> 
> 
> Which reminds me, this still isn't landed either for secretmem:
> https://lore.kernel.org/all/20250707171735.GE1880847@ZenIV/

It should be in mainline as:
commit 98f99394a104cc80296da34a62d4e1ad04127013 ("secretmem: use SB_I_NOEXEC")