Re: [PATCH] vfs: check dentry is still valid in get_link()

From: Al Viro <viro@zeniv.linux.org.uk>
To: Dave Chinner <david@fromorbit.com>
Cc: Brian Foster <bfoster@redhat.com>, Ian Kent <raven@themaw.net>,
	"Darrick J. Wong" <djwong@kernel.org>,
	Christoph Hellwig <hch@lst.de>,
	Miklos Szeredi <miklos@szeredi.hu>,
	David Howells <dhowells@redhat.com>,
	Kernel Mailing List <linux-kernel@vger.kernel.org>,
	linux-fsdevel <linux-fsdevel@vger.kernel.org>,
	xfs <linux-xfs@vger.kernel.org>
Subject: Re: [PATCH] vfs: check dentry is still valid in get_link()
Date: Tue, 18 Jan 2022 03:17:13 +0000	[thread overview]
Message-ID: <YeYxOadA0HgYfBjt@zeniv-ca.linux.org.uk> (raw)
In-Reply-To: <20220118030041.GB59729@dread.disaster.area>

On Tue, Jan 18, 2022 at 02:00:41PM +1100, Dave Chinner wrote:
> > IOW, how far is xfs_inode_mark_reclaimable() from being callable in RCU
> > callback context?
> 
> AIUI, not very close at all,
> 
> I'm pretty sure we can't put it under RCU callback context at all
> because xfs_fs_destroy_inode() can take sleeping locks, perform
> transactions, do IO, run rcu_read_lock() critical sections, etc.
> This means that needs to run an a full task context and so can't run
> from RCU callback context at all.

Umm...  AFAICS, this
	pag = xfs_perag_get(mp, XFS_INO_TO_AGNO(mp, ip->i_ino));
	spin_lock(&pag->pag_ici_lock);
	spin_lock(&ip->i_flags_lock);

	trace_xfs_inode_set_reclaimable(ip);
	ip->i_flags &= ~(XFS_NEED_INACTIVE | XFS_INACTIVATING);
	ip->i_flags |= XFS_IRECLAIMABLE;
	xfs_perag_set_inode_tag(pag, XFS_INO_TO_AGINO(mp, ip->i_ino),
	XFS_ICI_RECLAIM_TAG);

	spin_unlock(&ip->i_flags_lock);
	spin_unlock(&pag->pag_ici_lock);
	xfs_perag_put(pag);
in the end of xfs_inodegc_set_reclaimable() could go into ->free_inode()
just fine.  It's xfs_inodegc_queue() I'm not sure about - the part
about flush_work() in there...

I'm not familiar with that code; could you point me towards some
docs/old postings/braindump/whatnot?