With delayed allocation we lock the page in write_cache_pages and try to build
an in memory extent of contiguous blocks. This is needed so that we can get
large contiguous blocks request. Now with range_cyclic mode in write_cache_pages
if we have not done an I/O we loop back to 0 index and try to write the page.
That would imply we will attempt to take page lock of lower index page holding
the page lock of higher index page. This can cause a dead lock with other writeback
thread.
Signed-off-by: Aneesh Kumar K.V <[email protected]>
---
fs/ext4/inode.c | 23 +++++++++++++++++++++--
1 files changed, 21 insertions(+), 2 deletions(-)
diff --git a/fs/ext4/inode.c b/fs/ext4/inode.c
index 61e8fc0..c80e038 100644
--- a/fs/ext4/inode.c
+++ b/fs/ext4/inode.c
@@ -2437,6 +2437,7 @@ static int ext4_da_writepages(struct address_space *mapping,
int no_nrwrite_index_update;
int pages_written = 0;
long pages_skipped;
+ int range_cyclic = 0, cycled = 1, io_done = 0;
int needed_blocks, ret = 0, nr_to_writebump = 0;
struct ext4_sb_info *sbi = EXT4_SB(mapping->host->i_sb);
@@ -2488,9 +2489,17 @@ static int ext4_da_writepages(struct address_space *mapping,
if (wbc->range_start == 0 && wbc->range_end == LLONG_MAX)
range_whole = 1;
- if (wbc->range_cyclic)
+ if (wbc->range_cyclic) {
index = mapping->writeback_index;
- else
+ wbc->range_start = index << PAGE_CACHE_SHIFT;
+ wbc->range_end = LLONG_MAX;
+ wbc->range_cyclic = 0;
+ range_cyclic = 1;
+ if (index == 0)
+ cycled = 1;
+ else
+ cycled = 0;
+ } else
index = wbc->range_start >> PAGE_CACHE_SHIFT;
mpd.wbc = wbc;
@@ -2504,6 +2513,7 @@ static int ext4_da_writepages(struct address_space *mapping,
wbc->no_nrwrite_index_update = 1;
pages_skipped = wbc->pages_skipped;
+retry:
while (!ret && wbc->nr_to_write > 0) {
/*
@@ -2546,6 +2556,7 @@ static int ext4_da_writepages(struct address_space *mapping,
pages_written += mpd.pages_written;
wbc->pages_skipped = pages_skipped;
ret = 0;
+ io_done = 1;
} else if (wbc->nr_to_write)
/*
* There is no more writeout needed
@@ -2554,6 +2565,13 @@ static int ext4_da_writepages(struct address_space *mapping,
*/
break;
}
+ if (!io_done && !cycled) {
+ cycled = 1;
+ index = 0;
+ wbc->range_start = index << PAGE_CACHE_SHIFT;
+ wbc->range_end = mapping->writeback_index - 1;
+ goto retry;
+ }
if (pages_skipped != wbc->pages_skipped)
printk(KERN_EMERG "This should not happen leaving %s "
"with nr_to_write = %ld ret = %d\n",
@@ -2561,6 +2579,7 @@ static int ext4_da_writepages(struct address_space *mapping,
/* Update index */
index += pages_written;
+ wbc->range_cyclic = range_cyclic;
if (wbc->range_cyclic || (range_whole && wbc->nr_to_write > 0))
/*
* set the writeback_index so that range_cyclic
--
tg: (6ebb071..) range_cyclic_fix (depends on: fix_list_corruption)
I've rewritten the patch description for clarity's sake; can you
confirm I didn't mess anything up?
- Ted
ext4: Implement range_cyclic in ext4_da_writepages instead of write_cache_pages
From: "Aneesh Kumar K.V" <[email protected]>
With delayed allocation we lock the page in write_cache_pages() and
try to build an in memory extent of contiguous blocks. This is needed
so that we can get large contiguous blocks request. If range_cyclic
mode is enabled, write_cache_pages() will loop back to the 0 index if
no I/O has been done yet, and try to start writing from the beginning
of the range. That causes an attempt to take the page lock of lower
index page while holding the page lock of higher index page, which can
cause a dead lock with another writeback thread.
The solution is to implement the range_cyclic behavior in
ext4_da_writepages() instead.
http://bugzilla.kernel.org/show_bug.cgi?id=12579
Signed-off-by: Aneesh Kumar K.V <[email protected]>
Signed-off-by: "Theodore Ts'o" <[email protected]>