[PATCH v5 12/12] gfs2: Fix mmap + page fault deadlocks for direct I/O

From: Andreas Gruenbacher <agruenba@redhat.com>
To: Linus Torvalds <torvalds@linux-foundation.org>,
	Alexander Viro <viro@zeniv.linux.org.uk>,
	Christoph Hellwig <hch@infradead.org>,
	"Darrick J. Wong" <djwong@kernel.org>
Cc: Jan Kara <jack@suse.cz>, Matthew Wilcox <willy@infradead.org>,
	cluster-devel@redhat.com, linux-fsdevel@vger.kernel.org,
	linux-kernel@vger.kernel.org, ocfs2-devel@oss.oracle.com,
	Andreas Gruenbacher <agruenba@redhat.com>
Subject: [PATCH v5 12/12] gfs2: Fix mmap + page fault deadlocks for direct I/O
Date: Tue,  3 Aug 2021 21:18:18 +0200	[thread overview]
Message-ID: <20210803191818.993968-13-agruenba@redhat.com> (raw)
In-Reply-To: <20210803191818.993968-1-agruenba@redhat.com>

Also disable page faults during direct I/O requests and implement the same
kind of retry logic as in the buffered I/O case.

Direct I/O requests differ from buffered I/O requests in that they use
bio_iov_iter_get_pages for grabbing page references and faulting in pages
instead of triggering physical page faults.  Those manual page faults can be
disabled with the new iocb->noio flag.

Signed-off-by: Andreas Gruenbacher <agruenba@redhat.com>
---
 fs/gfs2/file.c | 47 +++++++++++++++++++++++++++++++++++++++++++++--
 1 file changed, 45 insertions(+), 2 deletions(-)

diff --git a/fs/gfs2/file.c b/fs/gfs2/file.c
index d98f690097e2..ed42b7675551 100644
--- a/fs/gfs2/file.c
+++ b/fs/gfs2/file.c
@@ -782,21 +782,47 @@ static ssize_t gfs2_file_direct_read(struct kiocb *iocb, struct iov_iter *to,
 	struct file *file = iocb->ki_filp;
 	struct gfs2_inode *ip = GFS2_I(file->f_mapping->host);
 	size_t count = iov_iter_count(to);
+	size_t written = 0;
 	ssize_t ret;
 
+	/*
+	 * In this function, we disable page faults when we're holding the
+	 * inode glock while doing I/O.  If a page fault occurs, we drop the
+	 * inode glock, fault in the pages manually, and retry.
+	 *
+	 * Unlike generic_file_read_iter, for reads, iomap_dio_rw can trigger
+	 * physical as well as manual page faults, and we need to disable both
+	 * kinds.
+	 */
+
 	if (!count)
 		return 0; /* skip atime */
 
 	gfs2_holder_init(ip->i_gl, LM_ST_DEFERRED, 0, gh);
+retry:
 	ret = gfs2_glock_nq(gh);
 	if (ret)
 		goto out_uninit;
 
-	ret = iomap_dio_rw(iocb, to, &gfs2_iomap_ops, NULL, 0, 0);
+	pagefault_disable();
+	to->noio = true;
+	ret = iomap_dio_rw(iocb, to, &gfs2_iomap_ops, NULL,
+			   IOMAP_DIO_PARTIAL, written);
+	to->noio = false;
+	pagefault_enable();
+
 	gfs2_glock_dq(gh);
+	if (ret > 0)
+		written = ret;
+	if (unlikely(iov_iter_count(to) && (ret > 0 || ret == -EFAULT)) &&
+	    iter_is_iovec(to) &&
+	    fault_in_iov_iter_writeable(to, SIZE_MAX) != 0)
+		goto retry;
 out_uninit:
 	gfs2_holder_uninit(gh);
-	return ret;
+	if (ret < 0)
+		return ret;
+	return written;
 }
 
 static ssize_t gfs2_file_direct_write(struct kiocb *iocb, struct iov_iter *from,
@@ -809,6 +835,15 @@ static ssize_t gfs2_file_direct_write(struct kiocb *iocb, struct iov_iter *from,
 	loff_t offset = iocb->ki_pos;
 	ssize_t ret;
 
+	/*
+	 * In this function, we disable page faults when we're holding the
+	 * inode glock while doing I/O.  If a page fault occurs, we drop the
+	 * inode glock, fault in the pages manually, and retry.
+	 *
+	 * For writes, iomap_dio_rw only triggers manual page faults, so we
+	 * don't need to disable physical ones.
+	 */
+
 	/*
 	 * Deferred lock, even if its a write, since we do no allocation on
 	 * this path. All we need to change is the atime, and this lock mode
@@ -818,6 +853,7 @@ static ssize_t gfs2_file_direct_write(struct kiocb *iocb, struct iov_iter *from,
 	 * VFS does.
 	 */
 	gfs2_holder_init(ip->i_gl, LM_ST_DEFERRED, 0, gh);
+retry:
 	ret = gfs2_glock_nq(gh);
 	if (ret)
 		goto out_uninit;
@@ -826,11 +862,18 @@ static ssize_t gfs2_file_direct_write(struct kiocb *iocb, struct iov_iter *from,
 	if (offset + len > i_size_read(&ip->i_inode))
 		goto out;
 
+	from->noio = true;
 	ret = iomap_dio_rw(iocb, from, &gfs2_iomap_ops, NULL, 0, 0);
+	from->noio = false;
+
 	if (ret == -ENOTBLK)
 		ret = 0;
 out:
 	gfs2_glock_dq(gh);
+	if (unlikely(ret == -EFAULT) &&
+	    iter_is_iovec(from) &&
+	    fault_in_iov_iter_readable(from, len) == len)
+		goto retry;
 out_uninit:
 	gfs2_holder_uninit(gh);
 	return ret;
-- 
2.26.3

