diff mbox

[v5,08/17] dax: set errors in mapping when writeback fails

Message ID 20170531124540.8782-9-jlayton@redhat.com (mailing list archive)
State New, archived
Headers show

Commit Message

Jeff Layton May 31, 2017, 12:45 p.m. UTC
Jan's description for this patch is much better than mine, so I'm
quoting it verbatim here:

-----------------8<-----------------
DAX currently doesn't set errors in the mapping when cache flushing
fails in dax_writeback_mapping_range(). Since this function can get
called only from fsync(2) or sync(2), this is actually as good as it can
currently get since we correctly propagate the error up from
dax_writeback_mapping_range() to filemap_fdatawrite()

However, in the future better writeback error handling will enable us to
properly report these errors on fsync(2) even if there are multiple file
descriptors open against the file or if sync(2) gets called before
fsync(2). So convert DAX to using standard error reporting through the
mapping.
-----------------8<-----------------

For now, only do this when the FS_WB_ERRSEQ flag is set. The
AS_EIO/AS_ENOSPC flags are not currently cleared in the older code when
writeback initiation fails, only when we discover an error after waiting
on writeback to complete, so we only want to do this with errseq_t based
error handling to prevent seeing duplicate errors on fsync.

Signed-off-by: Jeff Layton <jlayton@redhat.com>
Reviewed-by: Jan Kara <jack@suse.cz>
Reviewed-by: Christoph Hellwig <hch@lst.de>
Reviewed-and-Tested-by: Ross Zwisler <ross.zwisler@linux.intel.com>
---
 fs/dax.c | 18 +++++++++++++++++-
 1 file changed, 17 insertions(+), 1 deletion(-)

Comments

Ross Zwisler June 6, 2017, 1:01 a.m. UTC | #1
On Wed, May 31, 2017 at 08:45:31AM -0400, Jeff Layton wrote:
> Jan's description for this patch is much better than mine, so I'm
> quoting it verbatim here:
> 
> -----------------8<-----------------
> DAX currently doesn't set errors in the mapping when cache flushing
> fails in dax_writeback_mapping_range(). Since this function can get
> called only from fsync(2) or sync(2), this is actually as good as it can
> currently get since we correctly propagate the error up from
> dax_writeback_mapping_range() to filemap_fdatawrite()
> 
> However, in the future better writeback error handling will enable us to
> properly report these errors on fsync(2) even if there are multiple file
> descriptors open against the file or if sync(2) gets called before
> fsync(2). So convert DAX to using standard error reporting through the
> mapping.
> -----------------8<-----------------
> 
> For now, only do this when the FS_WB_ERRSEQ flag is set. The
> AS_EIO/AS_ENOSPC flags are not currently cleared in the older code when
> writeback initiation fails, only when we discover an error after waiting
> on writeback to complete, so we only want to do this with errseq_t based
> error handling to prevent seeing duplicate errors on fsync.
> 
> Signed-off-by: Jeff Layton <jlayton@redhat.com>
> Reviewed-by: Jan Kara <jack@suse.cz>
> Reviewed-by: Christoph Hellwig <hch@lst.de>
> Reviewed-and-Tested-by: Ross Zwisler <ross.zwisler@linux.intel.com>

Re-tested this version of the series with some injected DAX errors, and it
looks good.
Jeff Layton June 6, 2017, 1:08 a.m. UTC | #2
On Mon, 2017-06-05 at 19:01 -0600, Ross Zwisler wrote:
> On Wed, May 31, 2017 at 08:45:31AM -0400, Jeff Layton wrote:
> > Jan's description for this patch is much better than mine, so I'm
> > quoting it verbatim here:
> > 
> > -----------------8<-----------------
> > DAX currently doesn't set errors in the mapping when cache flushing
> > fails in dax_writeback_mapping_range(). Since this function can get
> > called only from fsync(2) or sync(2), this is actually as good as it can
> > currently get since we correctly propagate the error up from
> > dax_writeback_mapping_range() to filemap_fdatawrite()
> > 
> > However, in the future better writeback error handling will enable us to
> > properly report these errors on fsync(2) even if there are multiple file
> > descriptors open against the file or if sync(2) gets called before
> > fsync(2). So convert DAX to using standard error reporting through the
> > mapping.
> > -----------------8<-----------------
> > 
> > For now, only do this when the FS_WB_ERRSEQ flag is set. The
> > AS_EIO/AS_ENOSPC flags are not currently cleared in the older code when
> > writeback initiation fails, only when we discover an error after waiting
> > on writeback to complete, so we only want to do this with errseq_t based
> > error handling to prevent seeing duplicate errors on fsync.
> > 
> > Signed-off-by: Jeff Layton <jlayton@redhat.com>
> > Reviewed-by: Jan Kara <jack@suse.cz>
> > Reviewed-by: Christoph Hellwig <hch@lst.de>
> > Reviewed-and-Tested-by: Ross Zwisler <ross.zwisler@linux.intel.com>
> 
> Re-tested this version of the series with some injected DAX errors, and it
> looks good.

Excellent! Thanks very much for helping test it.
diff mbox

Patch

diff --git a/fs/dax.c b/fs/dax.c
index c22eaf162f95..42788d8505c7 100644
--- a/fs/dax.c
+++ b/fs/dax.c
@@ -856,8 +856,24 @@  int dax_writeback_mapping_range(struct address_space *mapping,
 
 			ret = dax_writeback_one(bdev, dax_dev, mapping,
 					indices[i], pvec.pages[i]);
-			if (ret < 0)
+			if (ret < 0) {
+				/*
+				 * For fs' that use errseq_t based error
+				 * tracking, we must call mapping_set_error
+				 * here to ensure that fsync on all open fds
+				 * get back an error. Doing this with the old
+				 * wb error tracking infrastructure is
+				 * problematic though, as DAX writeback is
+				 * synchronous, and the error flags are not
+				 * cleared when initiation fails, only when
+				 * it fails after the write has been submitted
+				 * to the backing store.
+				 */
+				if (mapping->host->i_sb->s_type->fs_flags &
+						FS_WB_ERRSEQ)
+					mapping_set_error(mapping, ret);
 				goto out;
+			}
 		}
 	}
 out: