From: Dave Chinner <david@fromorbit.com>
To: linux-fsdevel@vger.kernel.org
Cc: linux-kernel@vger.kernel.org, xfs@oss.sgi.com
Subject: [PATCH 1/2] vfs: split generic splice code from i_mutex locking
Date: Mon, 18 Jul 2011 14:04:43 +1000 [thread overview]
Message-ID: <1310961884-11634-2-git-send-email-david@fromorbit.com> (raw)
In-Reply-To: <1310961884-11634-1-git-send-email-david@fromorbit.com>
From: Dave Chinner <dchinner@redhat.com>
XFS holds locks that should be nested inside the inode->i_mutex when
generic_file_splice_write is called. This function takes the
i_mutex, and so we get a lock inversion that triggers lockdep
warnings and has been found to cause real deadlocks.
XFS does not need the splice code to take the i_mutex to do the page
cache manipulation, so add a new function
generic_file_splice_write_unlocked() that avoids the locking of the
i_mutex for XFS to call.
Signed-off-by: Dave Chinner <dchinner@redhat.com>
---
fs/splice.c | 39 +++++++++++++++++++++++++++++++++++----
include/linux/fs.h | 2 ++
2 files changed, 37 insertions(+), 4 deletions(-)
diff --git a/fs/splice.c b/fs/splice.c
index aa866d3..c15137d 100644
--- a/fs/splice.c
+++ b/fs/splice.c
@@ -980,8 +980,9 @@ ssize_t splice_from_pipe(struct pipe_inode_info *pipe, struct file *out,
*
*/
ssize_t
-generic_file_splice_write(struct pipe_inode_info *pipe, struct file *out,
- loff_t *ppos, size_t len, unsigned int flags)
+__generic_file_splice_write(struct pipe_inode_info *pipe, struct file *out,
+ loff_t *ppos, size_t len, unsigned int flags,
+ int need_imutex)
{
struct address_space *mapping = out->f_mapping;
struct inode *inode = mapping->host;
@@ -1001,13 +1002,15 @@ generic_file_splice_write(struct pipe_inode_info *pipe, struct file *out,
if (ret <= 0)
break;
- mutex_lock_nested(&inode->i_mutex, I_MUTEX_CHILD);
+ if (need_imutex)
+ mutex_lock_nested(&inode->i_mutex, I_MUTEX_CHILD);
ret = file_remove_suid(out);
if (!ret) {
file_update_time(out);
ret = splice_from_pipe_feed(pipe, &sd, pipe_to_file);
}
- mutex_unlock(&inode->i_mutex);
+ if (need_imutex)
+ mutex_unlock(&inode->i_mutex);
} while (ret > 0);
splice_from_pipe_end(pipe, &sd);
@@ -1033,8 +1036,36 @@ generic_file_splice_write(struct pipe_inode_info *pipe, struct file *out,
return ret;
}
+/**
+ * generic_file_splice_write - splice data from a pipe to a file
+ * @pipe: pipe info
+ * @out: file to write to
+ * @ppos: position in @out
+ * @len: number of bytes to splice
+ * @flags: splice modifier flags
+ *
+ * Description:
+ * Will either move or copy pages (determined by @flags options) from
+ * the given pipe inode to the given file.
+ *
+ */
+ssize_t
+generic_file_splice_write(struct pipe_inode_info *pipe, struct file *out,
+ loff_t *ppos, size_t len, unsigned int flags)
+{
+ return __generic_file_splice_write(pipe, out, ppos, len, flags, 1);
+}
EXPORT_SYMBOL(generic_file_splice_write);
+ssize_t
+generic_file_splice_write_unlocked(struct pipe_inode_info *pipe,
+ struct file *out, loff_t *ppos,
+ size_t len, unsigned int flags)
+{
+ return __generic_file_splice_write(pipe, out, ppos, len, flags, 0);
+}
+EXPORT_SYMBOL(generic_file_splice_write_unlocked);
+
static int write_pipe_buf(struct pipe_inode_info *pipe, struct pipe_buffer *buf,
struct splice_desc *sd)
{
diff --git a/include/linux/fs.h b/include/linux/fs.h
index 54c49e5..3a8b984 100644
--- a/include/linux/fs.h
+++ b/include/linux/fs.h
@@ -2368,6 +2368,8 @@ extern ssize_t default_file_splice_read(struct file *, loff_t *,
struct pipe_inode_info *, size_t, unsigned int);
extern ssize_t generic_file_splice_write(struct pipe_inode_info *,
struct file *, loff_t *, size_t, unsigned int);
+extern ssize_t generic_file_splice_write_unlocked(struct pipe_inode_info *,
+ struct file *, loff_t *, size_t, unsigned int);
extern ssize_t generic_splice_sendpage(struct pipe_inode_info *pipe,
struct file *out, loff_t *, size_t len, unsigned int flags);
extern long do_splice_direct(struct file *in, loff_t *ppos, struct file *out,
--
1.7.5.1
next prev parent reply other threads:[~2011-07-18 4:04 UTC|newest]
Thread overview: 9+ messages / expand[flat|nested] mbox.gz Atom feed top
2011-07-18 4:04 [PATCH 0/2] splice: i_mutex vs splice write deadlock Dave Chinner
2011-07-18 4:04 ` Dave Chinner [this message]
2011-07-18 4:04 ` [PATCH 2/2] xfs: fix splice/direct-IO deadlock Dave Chinner
2011-07-19 3:10 ` [PATCH 0/2] splice: i_mutex vs splice write deadlock Christoph Hellwig
2011-07-19 4:07 ` Dave Chinner
-- strict thread matches above, loose matches on Subject: below --
2011-08-08 6:45 [PATCH 0/2] splice: i_mutex vs splice write deadlock V2 Dave Chinner
2011-08-08 6:45 ` [PATCH 1/2] vfs: split generic splice code from i_mutex locking Dave Chinner
2011-08-09 11:36 ` Jan Kara
2011-08-10 10:12 ` Christoph Hellwig
2012-11-28 2:12 [PATCH 0/2] splice: fix direct IO/splice deadlock Dave Chinner
2012-11-28 2:12 ` [PATCH 1/2] vfs: split generic splice code from i_mutex locking Dave Chinner
Reply instructions:
You may reply publicly to this message via plain-text email
using any one of the following methods:
* Save the following mbox file, import it into your mail client,
and reply-to-all from there: mbox
Avoid top-posting and favor interleaved quoting:
https://en.wikipedia.org/wiki/Posting_style#Interleaved_style
* Reply using the --to, --cc, and --in-reply-to
switches of git-send-email(1):
git send-email \
--in-reply-to=1310961884-11634-2-git-send-email-david@fromorbit.com \
--to=david@fromorbit.com \
--cc=linux-fsdevel@vger.kernel.org \
--cc=linux-kernel@vger.kernel.org \
--cc=xfs@oss.sgi.com \
/path/to/YOUR_REPLY
https://kernel.org/pub/software/scm/git/docs/git-send-email.html
* If your mail client supports setting the In-Reply-To header
via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line
before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for NNTP newsgroup(s).