Re: [PATCH] xfs: Use WARN_ON rather than BUG() for bailout mount-operation

[Austin Kim]

2019ë 8ì 29ì (ë) ìí 4:56, Christoph Hellwig <hch@xxxxxxxxxxxxx>ëì ìì:
>
> On Wed, Aug 28, 2019 at 03:47:49PM +0900, Austin Kim wrote:
> > If the CONFIG_BUG is enabled, BUG() is executed and then system is crashed.
> > However, the bailout for mount is no longer proceeding.
> >
> > For this reason, using WARN_ON rather than BUG() could prevent this situation.
> > ---
> >  fs/xfs/xfs_mount.c | 3 +--
> >  1 file changed, 1 insertion(+), 2 deletions(-)
> >
> > diff --git a/fs/xfs/xfs_mount.c b/fs/xfs/xfs_mount.c
> > index 322da69..10fe000 100644
> > --- a/fs/xfs/xfs_mount.c
> > +++ b/fs/xfs/xfs_mount.c
> > @@ -213,8 +213,7 @@ xfs_initialize_perag(
> >                       goto out_hash_destroy;
> >
> >               spin_lock(&mp->m_perag_lock);
> > -             if (radix_tree_insert(&mp->m_perag_tree, index, pag)) {
> > -                     BUG();
> > +             if (WARN_ON(radix_tree_insert(&mp->m_perag_tree, index, pag))){
>
> Please make this a WARN_ON_ONCE so that we don't see a flodding of
> messages in case of this error.
>
Hello, Mr. Christoph
Thanks for good feedback.
If the kernel log is flooded with error message, as you pointed out,
it may cause other side-effect.(e.g: system non-responsive or lockup)

To. Mr. Darrick J. Wong
If you or other kernel developers do not disagree with the
idea(WARN_ON_ONCE instead of WARN_ON),
do I have to resend the patch with new revision?

The title, the commit message and patch might be changed as followings;
======
xfs: Use WARN_ON_ONCE rather than BUG() for bailout mount-operation

If the CONFIG_BUG is enabled, BUG() is executed and then system is crashed.
However, the bailout for mount is no longer proceeding.

For this reason, using WARN_ON_ONCE rather than BUG() could prevent
this situation.

diff --git a/fs/xfs/xfs_mount.c b/fs/xfs/xfs_mount.c
index 322da69..d831c13 100644
--- a/fs/xfs/xfs_mount.c
+++ b/fs/xfs/xfs_mount.c
@@ -213,8 +213,7 @@ xfs_initialize_perag(
                        goto out_hash_destroy;

                spin_lock(&mp->m_perag_lock);
-               if (radix_tree_insert(&mp->m_perag_tree, index, pag)) {
-                       BUG();
+               if (WARN_ON_ONCE(radix_tree_insert(&mp->m_perag_tree,
index, pag))) {
                        spin_unlock(&mp->m_perag_lock);
                        radix_tree_preload_end();
                        error = -EEXIST;
======

BR,
Guillermo Austin Kim