linux-kernel - Re: [PATCH] fs: add missing fences to I

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: <CAGudoHFi-9qYPnb9xPtGsUq+Mf_8h+uk4iQDo1TggZYPgTv6fA@mail.gmail.com>
Date: Mon, 6 Oct 2025 03:16:43 +0200
From: Mateusz Guzik <mjguzik@...il.com>
To: Hillf Danton <hdanton@...a.com>
Cc: brauner@...nel.org, viro@...iv.linux.org.uk, jack@...e.cz, 
	linux-kernel@...r.kernel.org, linux-fsdevel@...r.kernel.org
Subject: Re: [PATCH] fs: add missing fences to I_NEW handling

On Mon, Oct 6, 2025 at 2:57 AM Hillf Danton <hdanton@...a.com> wrote:
>
> On Mon,  6 Oct 2025 01:15:26 +0200 Mateusz Guzik wrote:
> > Suppose there are 2 CPUs racing inode hash lookup func (say ilookup5())
> > and unlock_new_inode().
> >
> > In principle the latter can clear the I_NEW flag before prior stores
> > into the inode were made visible.
> >
> Given difficulty following up here, could you specify why the current
> mem barrier [1] in unlock_new_inode() is not enough?
>

That fence synchronizes against threads which went to sleep.

In the example I'm providing this did not happen.

  193 static inline void wait_on_inode(struct inode *inode)
  194 {
  195         wait_var_event(inode_state_wait_address(inode, __I_NEW),
  196                       !(READ_ONCE(inode->i_state) & I_NEW));

  303 #define wait_var_event(var, condition)                                  \
  304 do {                                                                    \
  305         might_sleep();                                                  \
  306         if (condition)                                                  \
  307                 break;                                                  \

I_NEW is tested here without any locks or fences.

  308         __wait_var_event(var, condition);                               \
  309 } while (0)

> [1] https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git/tree/fs/inode.c#n1190
>
> > The former can in turn observe I_NEW is cleared and proceed to use the
> > inode, while possibly reading from not-yet-published areas.
> >
> > Signed-off-by: Mateusz Guzik <mjguzik@...il.com>
> > ---
> >
> > I don't think this is a serious bug in the sense I doubt anyone ever ran
> > into it, but this is an issue on paper.
> >
> > I'm doing some changes in the area and I figured I'll get this bit out
> > of the way.
> >
> >  fs/dcache.c               | 4 ++++
> >  fs/inode.c                | 8 ++++++++
> >  include/linux/writeback.h | 4 ++++
> >  3 files changed, 16 insertions(+)
> >
> > diff --git a/fs/dcache.c b/fs/dcache.c
> > index a067fa0a965a..806d6a665124 100644
> > --- a/fs/dcache.c
> > +++ b/fs/dcache.c
> > @@ -1981,6 +1981,10 @@ void d_instantiate_new(struct dentry *entry, struct inode *inode)
> >       spin_lock(&inode->i_lock);
> >       __d_instantiate(entry, inode);
> >       WARN_ON(!(inode->i_state & I_NEW));
> > +     /*
> > +      * Pairs with smp_rmb in wait_on_inode().
> > +      */
> > +     smp_wmb();
> >       inode->i_state &= ~I_NEW & ~I_CREATING;
> >       /*
> >        * Pairs with the barrier in prepare_to_wait_event() to make sure
> > diff --git a/fs/inode.c b/fs/inode.c
> > index ec9339024ac3..842ee973c8b6 100644
> > --- a/fs/inode.c
> > +++ b/fs/inode.c
> > @@ -1181,6 +1181,10 @@ void unlock_new_inode(struct inode *inode)
> >       lockdep_annotate_inode_mutex_key(inode);
> >       spin_lock(&inode->i_lock);
> >       WARN_ON(!(inode->i_state & I_NEW));
> > +     /*
> > +      * Pairs with smp_rmb in wait_on_inode().
> > +      */
> > +     smp_wmb();
> >       inode->i_state &= ~I_NEW & ~I_CREATING;
> >       /*
> >        * Pairs with the barrier in prepare_to_wait_event() to make sure
> > @@ -1198,6 +1202,10 @@ void discard_new_inode(struct inode *inode)
> >       lockdep_annotate_inode_mutex_key(inode);
> >       spin_lock(&inode->i_lock);
> >       WARN_ON(!(inode->i_state & I_NEW));
> > +     /*
> > +      * Pairs with smp_rmb in wait_on_inode().
> > +      */
> > +     smp_wmb();
> >       inode->i_state &= ~I_NEW;
> >       /*
> >        * Pairs with the barrier in prepare_to_wait_event() to make sure
> > diff --git a/include/linux/writeback.h b/include/linux/writeback.h
> > index 22dd4adc5667..e1e1231a6830 100644
> > --- a/include/linux/writeback.h
> > +++ b/include/linux/writeback.h
> > @@ -194,6 +194,10 @@ static inline void wait_on_inode(struct inode *inode)
> >  {
> >       wait_var_event(inode_state_wait_address(inode, __I_NEW),
> >                      !(READ_ONCE(inode->i_state) & I_NEW));
> > +     /*
> > +      * Pairs with routines clearing I_NEW.
> > +      */
> > +     smp_rmb();
> >  }
>
> Why is this needed as nobody cares I_NEW after wait?
>
> >
> >  #ifdef CONFIG_CGROUP_WRITEBACK
> > --
> > 2.34.1