[PATCH v2] iomap: avoid deadlock if memory reclaim is triggered in writepage path

[Yafang Shao <laoar.shao@gmail.com>]

Recently there is a XFS deadlock on our server with an old kernel.
This deadlock is caused by allocating memory in xfs_map_blocks() while
doing writeback on behalf of memroy reclaim. Although this deadlock happens
on an old kernel, I think it could happen on the upstream as well. This
issue only happens once and can't be reproduced, so I haven't tried to
reproduce it on upsteam kernel.

Bellow is the call trace of this deadlock.
[480594.790087] INFO: task redis-server:16212 blocked for more than 120 seconds.
[480594.790087] "echo 0 > /proc/sys/kernel/hung_task_timeout_secs" disables this message.
[480594.790088] redis-server    D ffffffff8168bd60     0 16212  14347 0x00000004
[480594.790090]  ffff880da128f070 0000000000000082 ffff880f94a2eeb0 ffff880da128ffd8
[480594.790092]  ffff880da128ffd8 ffff880da128ffd8 ffff880f94a2eeb0 ffff88103f9d6c40
[480594.790094]  0000000000000000 7fffffffffffffff ffff88207ffc0ee8 ffffffff8168bd60
[480594.790096] Call Trace:
[480594.790101]  [<ffffffff8168dce9>] schedule+0x29/0x70
[480594.790103]  [<ffffffff8168b749>] schedule_timeout+0x239/0x2c0
[480594.790111]  [<ffffffff8168d28e>] io_schedule_timeout+0xae/0x130
[480594.790114]  [<ffffffff8168d328>] io_schedule+0x18/0x20
[480594.790116]  [<ffffffff8168bd71>] bit_wait_io+0x11/0x50
[480594.790118]  [<ffffffff8168b895>] __wait_on_bit+0x65/0x90
[480594.790121]  [<ffffffff811814e1>] wait_on_page_bit+0x81/0xa0
[480594.790125]  [<ffffffff81196ad2>] shrink_page_list+0x6d2/0xaf0
[480594.790130]  [<ffffffff811975a3>] shrink_inactive_list+0x223/0x710
[480594.790135]  [<ffffffff81198225>] shrink_lruvec+0x3b5/0x810
[480594.790139]  [<ffffffff8119873a>] shrink_zone+0xba/0x1e0
[480594.790141]  [<ffffffff81198c20>] do_try_to_free_pages+0x100/0x510
[480594.790143]  [<ffffffff8119928d>] try_to_free_mem_cgroup_pages+0xdd/0x170
[480594.790145]  [<ffffffff811f32de>] mem_cgroup_reclaim+0x4e/0x120
[480594.790147]  [<ffffffff811f37cc>] __mem_cgroup_try_charge+0x41c/0x670
[480594.790153]  [<ffffffff811f5cb6>] __memcg_kmem_newpage_charge+0xf6/0x180
[480594.790157]  [<ffffffff8118c72d>] __alloc_pages_nodemask+0x22d/0x420
[480594.790162]  [<ffffffff811d0c7a>] alloc_pages_current+0xaa/0x170
[480594.790165]  [<ffffffff811db8fc>] new_slab+0x30c/0x320
[480594.790168]  [<ffffffff811dd17c>] ___slab_alloc+0x3ac/0x4f0
[480594.790204]  [<ffffffff81685656>] __slab_alloc+0x40/0x5c
[480594.790206]  [<ffffffff811dfc43>] kmem_cache_alloc+0x193/0x1e0
[480594.790233]  [<ffffffffa04fab67>] kmem_zone_alloc+0x97/0x130 [xfs]
[480594.790247]  [<ffffffffa04f90ba>] _xfs_trans_alloc+0x3a/0xa0 [xfs]
[480594.790261]  [<ffffffffa04f915c>] xfs_trans_alloc+0x3c/0x50 [xfs]
[480594.790276]  [<ffffffffa04e958b>] xfs_iomap_write_allocate+0x1cb/0x390 [xfs]
[480594.790299]  [<ffffffffa04d3616>] xfs_map_blocks+0x1a6/0x210 [xfs]
[480594.790312]  [<ffffffffa04d416b>] xfs_do_writepage+0x17b/0x550 [xfs]
[480594.790314]  [<ffffffff8118d881>] write_cache_pages+0x251/0x4d0 [xfs]
[480594.790338]  [<ffffffffa04d3e05>] xfs_vm_writepages+0xc5/0xe0 [xfs]
[480594.790341]  [<ffffffff8118ebfe>] do_writepages+0x1e/0x40
[480594.790343]  [<ffffffff811837b5>] __filemap_fdatawrite_range+0x65/0x80
[480594.790346]  [<ffffffff81183901>] filemap_write_and_wait_range+0x41/0x90
[480594.790360]  [<ffffffffa04df2c6>] xfs_file_fsync+0x66/0x1e0 [xfs]
[480594.790363]  [<ffffffff81231cf5>] do_fsync+0x65/0xa0
[480594.790365]  [<ffffffff81231fe3>] SyS_fdatasync+0x13/0x20
[480594.790367]  [<ffffffff81698d09>] system_call_fastpath+0x16/0x1b

Note that xfs_iomap_write_allocate() is replaced by xfs_convert_blocks() in
commit 4ad765edb02a ("xfs: move xfs_iomap_write_allocate to xfs_aops.c")
and write_cache_pages() is replaced by iomap_writepages() in
commit 598ecfbaa742 ("iomap: lift the xfs writeback code to iomap").
So for upsteam, the call trace should be,
xfs_vm_writepages
  -> iomap_writepages
     -> write_cache_pages
        -> iomap_do_writepage
           -> xfs_map_blocks
              -> xfs_convert_blocks
                 -> xfs_bmapi_convert_delalloc
                    -> xfs_trans_alloc //It should alloc page with GFP_NOFS

Signed-off-by: Yafang Shao <laoar.shao@gmail.com>

---
v1 - >v2:
- retile the subject from "xfs: avoid deadlock when tigger memory reclam in xfs_map_blocks()"
- set GFP_NOFS in iomap_do_writepage(), per Dave.
---
 fs/iomap/buffered-io.c | 14 +++++++++++++-
 1 file changed, 13 insertions(+), 1 deletion(-)

diff --git a/fs/iomap/buffered-io.c b/fs/iomap/buffered-io.c
index a1ed762..f5176e3 100644
--- a/fs/iomap/buffered-io.c
+++ b/fs/iomap/buffered-io.c
@@ -17,6 +17,7 @@
 #include <linux/bio.h>
 #include <linux/sched/signal.h>
 #include <linux/migrate.h>
+#include <linux/sched/mm.h>
 #include "trace.h"
 
 #include "../internal.h"
@@ -1478,9 +1479,11 @@ static void iomap_writepage_end_bio(struct bio *bio)
 {
 	struct iomap_writepage_ctx *wpc = data;
 	struct inode *inode = page->mapping->host;
+	unsigned int nofs_flag;
 	pgoff_t end_index;
 	u64 end_offset;
 	loff_t offset;
+	int ret;
 
 	trace_iomap_writepage(inode, page_offset(page), PAGE_SIZE);
 
@@ -1571,7 +1574,16 @@ static void iomap_writepage_end_bio(struct bio *bio)
 		end_offset = offset;
 	}
 
-	return iomap_writepage_map(wpc, wbc, inode, page, end_offset);
+	/*
+	 * We can allocate memory here while doing writeback on behalf of
+	 * memory reclaim.  To avoid memory allocation deadlocks set the
+	 * task-wide nofs context for the following operations.
+	 */
+	nofs_flag = memalloc_nofs_save();
+	ret = iomap_writepage_map(wpc, wbc, inode, page, end_offset);
+	memalloc_nofs_restore(nofs_flag);
+
+	return ret;
 
 redirty:
 	redirty_page_for_writepage(wbc, page);
-- 
1.8.3.1