On Sat, 2016-10-08 at 10:12 +0200, Peter Zijlstra wrote:
> On Fri, Oct 07, 2016 at 07:26:36PM -0400, Jeff Layton wrote:
>
> >
> > Well spotted. Yeah, I think you're right. The assertion is this:
> >
> > percpu_rwsem_assert_held(&file_rwsem);
> >
> > I'm guessing this is probably fallout from the lglock to rwsem
> > conversion (commit aba376607383).
> >
> > From a quick glance, I think we probably just need to down_read the
> > file_rwsem in locks_remove_lease, prior to taking the flc_lock, and
> > release it just afterward.
>
> Correct on all that.
>
> >
> > I do want to go over the code a little more
> > closely though to make sure other codepaths aren't missing that lock
> > though.
>
> Urg, sorry for missing these, I went through it again and found the
> below to be missing.
>
> ---
> Subject: fs/locks: Add missing file_sem locks
>
> I overlooked a few code-paths that can lead to
> locks_delete_global_locks().
>
> Reported-by: Dmitry Vyukov <dvyukov@xxxxxxxxxx>
> Signed-off-by: Peter Zijlstra (Intel) <peterz@xxxxxxxxxxxxx>
> ---
>
> diff --git a/fs/locks.c b/fs/locks.c
> index 133fb2543d21..c623490863a9 100644
> --- a/fs/locks.c
> +++ b/fs/locks.c
> @@ -1604,6 +1604,7 @@ int fcntl_getlease(struct file *filp)
>
> ctx = smp_load_acquire(&inode->i_flctx);
> if (ctx && !list_empty_careful(&ctx->flc_lease)) {
> + percpu_down_read_preempt_disable(&file_rwsem);
> spin_lock(&ctx->flc_lock);
> time_out_leases(file_inode(filp), &dispose);
> list_for_each_entry(fl, &ctx->flc_lease, fl_list) {
> @@ -1613,6 +1614,8 @@ int fcntl_getlease(struct file *filp)
> break;
> }
> spin_unlock(&ctx->flc_lock);
> + percpu_up_read_preempt_enable(&file_rwsem);
> +
> locks_dispose_list(&dispose);
> }
> return type;
> @@ -2522,11 +2525,14 @@ locks_remove_lease(struct file *filp, struct file_lock_context *ctx)
> if (list_empty(&ctx->flc_lease))
> return;
>
> + percpu_down_read_preempt_disable(&file_rwsem);
> spin_lock(&ctx->flc_lock);
> list_for_each_entry_safe(fl, tmp, &ctx->flc_lease, fl_list)
> if (filp == fl->fl_file)
> lease_modify(fl, F_UNLCK, &dispose);
> spin_unlock(&ctx->flc_lock);
> + percpu_up_read_preempt_enable(&file_rwsem);
> +
> locks_dispose_list(&dispose);
> }
>
Looks correct to me.
Peter, do you need me to pick this patch up or do you plan to get it to
Linus another way?
Thanks, and:
Reviewed-by: Jeff Layton <jlayton@xxxxxxxxxxxxxxx>
--
To unsubscribe from this list: send the line "unsubscribe linux-fsdevel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at http://vger.kernel.org/majordomo-info.html
