linux-kernel - Re: [PATCH] fs: use debug-only asserts around fd allocation and install

lists.openwall.net		lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC
Open Source and information security mailing list archives

Hash Suite: Windows password security audit tool. GUI, reports in PDF.

[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]

Message-ID: <CAGudoHFH70YpLYXnhJq4MDtjJ6FiY59Xn-D_kTB9xsE2UTJD_g@mail.gmail.com>
Date: Wed, 12 Mar 2025 18:21:01 +0100
From: Mateusz Guzik <mjguzik@...il.com>
To: brauner@...nel.org
Cc: viro@...iv.linux.org.uk, jack@...e.cz, linux-kernel@...r.kernel.org, 
	linux-fsdevel@...r.kernel.org
Subject: Re: [PATCH] fs: use debug-only asserts around fd allocation and install

On Wed, Mar 12, 2025 at 5:19 PM Mateusz Guzik <mjguzik@...il.com> wrote:
>
> This also restores the check which got removed in 52732bb9abc9ee5b
> ("fs/file.c: remove sanity_check and add likely/unlikely in alloc_fd()")
> for performance reasons -- they no longer apply with a debug-only
> variant.
>
> Signed-off-by: Mateusz Guzik <mjguzik@...il.com>
> ---
>
> I have about 0 opinion whether this should be BUG or WARN, the code was
> already inconsistent on this front. If you want the latter, I'll have 0
> complaints if you just sed it and commit as yours.
>
> This reminded me to sort out that litmus test for smp_rmb, hopefully
> soon(tm) as it is now nagging me.
>
>  fs/file.c | 5 +++--
>  1 file changed, 3 insertions(+), 2 deletions(-)
>
> diff --git a/fs/file.c b/fs/file.c
> index 6c159ede55f1..09460ec74ef8 100644
> --- a/fs/file.c
> +++ b/fs/file.c
> @@ -582,6 +582,7 @@ static int alloc_fd(unsigned start, unsigned end, unsigned flags)
>
>         __set_open_fd(fd, fdt, flags & O_CLOEXEC);
>         error = fd;
> +       VFS_BUG_ON(rcu_access_pointer(fdt->fd[fd]) != NULL);
>

when restoring this check i dutifully copy-pasted the original. I only
now mentally registered it uses a rcu primitive to do the load, while
the others do a plain load. arguably the former is closer to being
correct and it definitely does not hurt

so this line should replace the other 2 lines below. i can send a v2
to that effect, but given the triviality of the edit, perhaps you will
be happy to sort it out

>  out:
>         spin_unlock(&files->file_lock);
> @@ -647,7 +648,7 @@ void fd_install(unsigned int fd, struct file *file)
>                 rcu_read_unlock_sched();
>                 spin_lock(&files->file_lock);
>                 fdt = files_fdtable(files);
> -               WARN_ON(fdt->fd[fd] != NULL);
> +               VFS_BUG_ON(fdt->fd[fd] != NULL);
>                 rcu_assign_pointer(fdt->fd[fd], file);
>                 spin_unlock(&files->file_lock);
>                 return;
> @@ -655,7 +656,7 @@ void fd_install(unsigned int fd, struct file *file)
>         /* coupled with smp_wmb() in expand_fdtable() */
>         smp_rmb();
>         fdt = rcu_dereference_sched(files->fdt);
> -       BUG_ON(fdt->fd[fd] != NULL);
> +       VFS_BUG_ON(fdt->fd[fd] != NULL);
>         rcu_assign_pointer(fdt->fd[fd], file);
>         rcu_read_unlock_sched();
>  }
> --
> 2.43.0
>


-- 
Mateusz Guzik <mjguzik gmail.com>