linux-kernel - Re: [PATCH 1/2] ocfs2: Add ocfs2_emergency

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] [day] [month] [year] [list]

Message-ID: <ehstl65oxyl2nmckl5axluz2arhz7dm6k63mlddtqb4qf2o7m2@ler4qv6wnbj2>
Date: Mon, 1 Dec 2025 15:43:59 +0800
From: Heming Zhao <heming.zhao@...e.com>
To: Ahmet Eray Karadag <eraykrdg1@...il.com>
Cc: mark@...heh.com, jlbec@...lplan.org, joseph.qi@...ux.alibaba.com, 
	ocfs2-devel@...ts.linux.dev, linux-kernel@...r.kernel.org, david.hunter.linux@...il.com, 
	skhan@...uxfoundation.org, Albin Babu Varghese <albinbabuvarghese20@...il.com>
Subject: Re: [PATCH 1/2] ocfs2: Add ocfs2_emergency_state helper and apply to
 setattr

On Fri, Nov 28, 2025 at 12:06:28PM +0300, Ahmet Eray Karadag wrote:
> On Tue, Nov 25, 2025 at 10:58:32AM +0800, Heming Zhao wrote:
> > On Tue, Nov 18, 2025 at 03:26:44AM +0300, Ahmet Eray Karadag wrote:
> > > To centralize error checking, follow the pattern of other filesystems
> > > like ext4 (which uses `ext4_emergency_state()`), and prepare for
> > > future enhancements, this patch introduces a new helper function:
> > > `ocfs2_emergency_state()`.
> > > 
> > > The purpose of this helper is to provide a single, unified location
> > > for checking all filesystem-level emergency conditions. In this
> > > initial implementation, the function only checks for the existing
> > > hard and soft read-only modes, returning -EROFS if either is set.
> > > 
> > > This provides a foundation where future checks (e.g., for fatal error
> > > states returning -EIO, or shutdown states) can be easily added in
> > > one place.
> > > 
> > > This patch also adds this new check to the beginning of
> > > `ocfs2_setattr()`. This ensures that operations like `ftruncate`
> > > (which triggered the original BUG) fail-fast with -EROFS when the
> > > filesystem is already in a read-only state.
> > > 
> > > Suggested-by: Heming Zhao <heming.zhao@...e.com>
> > > Co-developed-by: Albin Babu Varghese <albinbabuvarghese20@...il.com>
> > > Signed-off-by: Albin Babu Varghese <albinbabuvarghese20@...il.com>
> > > Signed-off-by: Ahmet Eray Karadag <eraykrdg1@...il.com>
> > > ---
> > >  fs/ocfs2/file.c  | 6 ++++++
> > >  fs/ocfs2/ocfs2.h | 8 ++++++++
> > >  2 files changed, 14 insertions(+)
> > > 
> > > diff --git a/fs/ocfs2/file.c b/fs/ocfs2/file.c
> > > index 21d797ccccd0..03a98985ac92 100644
> > > --- a/fs/ocfs2/file.c
> > > +++ b/fs/ocfs2/file.c
> > > @@ -1137,6 +1137,12 @@ int ocfs2_setattr(struct mnt_idmap *idmap, struct dentry *dentry,
> > >  					from_kgid(&init_user_ns, attr->ia_gid) : 0);
> > >  
> > >  	/* ensuring we don't even attempt to truncate a symlink */
> > > +	status = ocfs2_emergency_state(osb);
> > > +	if (status < 0) {
> > > +		mlog_errno(status);
> > > +		goto bail;
> > > +	}
> > > +
> > 
> > I prefer the ext4_emerency_state() style, using "if (unlikely(status))" here.
> > 
> > >  	if (S_ISLNK(inode->i_mode))
> > >  		attr->ia_valid &= ~ATTR_SIZE;
> > >  
> > > diff --git a/fs/ocfs2/ocfs2.h b/fs/ocfs2/ocfs2.h
> > > index 6aaa94c554c1..2ac16632dd19 100644
> > > --- a/fs/ocfs2/ocfs2.h
> > > +++ b/fs/ocfs2/ocfs2.h
> > > @@ -680,6 +680,14 @@ static inline int ocfs2_is_soft_readonly(struct ocfs2_super *osb)
> > >  	return ret;
> > >  }
> > >  
> > > +static inline int ocfs2_emergency_state(struct ocfs2_super *osb)
> > > +{
> > > +	if (ocfs2_is_soft_readonly(osb) || ocfs2_is_hard_readonly(osb)) {
> > > +		return -EROFS;
> > > +	}
> > > +	return 0;
> > > +}
> > > +
> > 
> > calling ocfs2_is_[soft|hard]_readonly() is a little expensive because it
> > involves acquiring the spinlock osb->osb_lock twice. In my view, we could
> > introduce a new function, such as ocfs2_is_readonly():
> In ext4, the helper functions check read-only flags without acquiring a lock.
> Would it be problematic to do the same in ocfs2?

Comparing with the ext4 code, in my view, OCFS2 incorrectly mixes the normal
read-only state with the emergency read-only state. For the minimal code change,
we must currently acquire the spinlock to access osb->osb_flags.

Thanks,
Heming
> > 
> > ```
> > static inline int ocfs2_is_readonly(struct ocfs2_super *osb)
> > {
> >         int ret;
> > 
> >         spin_lock(&osb->osb_lock);
> >         ret = osb->osb_flags & (OCFS2_OSB_SOFT_RO | OCFS2_OSB_HARD_RO);
> >         spin_unlock(&osb->osb_lock);
> > 
> >         return ret
> > }
> > ```
> > 
> > Thanks,
> > Heming
> > 
> > >  static inline int ocfs2_clusterinfo_valid(struct ocfs2_super *osb)
> > >  {
> > >  	return (osb->s_feature_incompat &
> > > -- 
> > > 2.43.0
> > > 
> I agree with the other comments. We will send v2.