[PATCH] VFS mpage.c:mpage_end_io_read wrong behavior when I/O failure occurs

["fs@ercist.iscas.ac.cn" <fs@ercist.iscas.ac.cn>]

Related FS:
     JFS

Related files:
     fs/mpage.c

Bug description:
     Make a partition in USB storage HDD, create a test file with
any contents.
     Write a program, do: open(O_RDWR) - write - fsync - close.
After each operation, pause for a while, such as 3 seconds. Between
open and fsync, unplug USB wire to force an I/O error. fsync will
return OK instead of EIO.

Bug analysis:
sys_fsync does 3 jobs:
   1)submit all pages taged with PAGECACHE_TAG_DIRTY to disk I/O,
   2)do file-system related fsync operations
   3)wait all pages taged with PAGECACHE_TAG_WRITEBACK to complete.
Here, all disk I/O are asynchronous, when finished, mpage_end_io_write
will remove page from mapping's PAGECACHE_TAG_WRITEBACK tree. If I/O
fails, it will also set PG_error flag, but it FORGET to set the
mapping->flags AS_EIO flag. So in step 3, sys_fsync won't notice these
pages, then no error returns.
static int mpage_end_io_write(...)
{
	...
		if (!uptodate)
			SetPageError(page);<-- Not set mapping->flags
	...
}

static int wait_on_page_writeback_range(struct address_space *mapping,
				pgoff_t start, pgoff_t end)
{
         ...
	int ret = 0;
	...	

	while ((index <= end) && <-- For I/O error page, no loop here
		(nr_pages = pagevec_lookup_tag(&pvec, mapping,  &index,
		PAGECACHE_TAG_WRITEBACK,
		min(end - index, (pgoff_t)PAGEVEC_SIZE-1) + 1)) != 0) {
			...
	}

	if (test_and_clear_bit(AS_ENOSPC, &mapping->flags))
		ret = -ENOSPC;
	if (test_and_clear_bit(AS_EIO, &mapping->flags)) <-- Not set
		ret = -EIO;

	return ret; <-- return 0
}

Way around:
Set AS_EIO to mapping->flags if I/O error happens.

Patch:
diff -uNp linux-2.6.11.11-orig/fs/mpage.c linux-2.6.11.11-new/fs/mpage.c
--- linux-2.6.11.11-orig/fs/mpage.c	2005-05-27 01:06:46.000000000 -0400
+++ linux-2.6.11.11-new/fs/mpage.c	2005-06-02 14:29:42.264334920 -0400
@@ -79,8 +79,11 @@ static int mpage_end_io_write(struct bio
  		if (--bvec >= bio->bi_io_vec)
  			prefetchw(&bvec->bv_page->flags);

-		if (!uptodate)
+		if (!uptodate){
  			SetPageError(page);
+			if(page->mapping)
+				set_bit(AS_EIO, &page->mapping->flags);
+		}
  		end_page_writeback(page);
  	} while (bvec >= bio->bi_io_vec);
  	bio_put(bio);


Signed-off-by: Qu Fuping<fs@ercist.iscas.ac.cn>

-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/