UBIFS: handle allocation failures in UBIFS write path
Running kernel 2.6.37, my PPC-based device occasionally gets an
order-2 allocation failure in UBIFS, which causes the root FS to
become unwritable:
kswapd0: page allocation failure. order:2, mode:0x4050
Call Trace:
[c787dc30] [c00085b8] show_stack+0x7c/0x194 (unreliable)
[c787dc70] [c0061aec] __alloc_pages_nodemask+0x4f0/0x57c
[c787dd00] [c0061b98] __get_free_pages+0x20/0x50
[c787dd10] [c00e4f88] ubifs_jnl_write_data+0x54/0x200
[c787dd50] [c00e82d4] do_writepage+0x94/0x198
[c787dd90] [c00675e4] shrink_page_list+0x40c/0x77c
[c787de40] [c0067de0] shrink_inactive_list+0x1e0/0x370
[c787de90] [c0068224] shrink_zone+0x2b4/0x2b8
[c787df00] [c0068854] kswapd+0x408/0x5d4
[c787dfb0] [c0037bcc] kthread+0x80/0x84
[c787dff0] [c000ef44] kernel_thread+0x4c/0x68
Similar problems were encountered last April by Tomasz Stanislawski:
http://patchwork.ozlabs.org/patch/50965/
This patch implements Artem's suggested fix: fall back to a
mutex-protected static buffer, allocated at mount time. I tested it
by forcing execution down the failure path, and didn't see any ill
effects.
Artem: massaged the patch a little, improved it so that we'd not
allocate the write reserve buffer when we are in R/O mode.
Signed-off-by: Matthew L. Creech <mlcreech@gmail.com>
Signed-off-by: Artem Bityutskiy <Artem.Bityutskiy@nokia.com>
diff --git a/fs/ubifs/journal.c b/fs/ubifs/journal.c
index 914f1bd..aed25e8 100644
--- a/fs/ubifs/journal.c
+++ b/fs/ubifs/journal.c
@@ -690,7 +690,7 @@
{
struct ubifs_data_node *data;
int err, lnum, offs, compr_type, out_len;
- int dlen = UBIFS_DATA_NODE_SZ + UBIFS_BLOCK_SIZE * WORST_COMPR_FACTOR;
+ int dlen = COMPRESSED_DATA_NODE_BUF_SZ, allocated = 1;
struct ubifs_inode *ui = ubifs_inode(inode);
dbg_jnl("ino %lu, blk %u, len %d, key %s",
@@ -698,9 +698,19 @@
DBGKEY(key));
ubifs_assert(len <= UBIFS_BLOCK_SIZE);
- data = kmalloc(dlen, GFP_NOFS);
- if (!data)
- return -ENOMEM;
+ data = kmalloc(dlen, GFP_NOFS | __GFP_NOWARN);
+ if (!data) {
+ /*
+ * Fall-back to the write reserve buffer. Note, we might be
+ * currently on the memory reclaim path, when the kernel is
+ * trying to free some memory by writing out dirty pages. The
+ * write reserve buffer helps us to guarantee that we are
+ * always able to write the data.
+ */
+ allocated = 0;
+ mutex_lock(&c->write_reserve_mutex);
+ data = c->write_reserve_buf;
+ }
data->ch.node_type = UBIFS_DATA_NODE;
key_write(c, key, &data->key);
@@ -736,7 +746,10 @@
goto out_ro;
finish_reservation(c);
- kfree(data);
+ if (!allocated)
+ mutex_unlock(&c->write_reserve_mutex);
+ else
+ kfree(data);
return 0;
out_release:
@@ -745,7 +758,10 @@
ubifs_ro_mode(c, err);
finish_reservation(c);
out_free:
- kfree(data);
+ if (!allocated)
+ mutex_unlock(&c->write_reserve_mutex);
+ else
+ kfree(data);
return err;
}
diff --git a/fs/ubifs/super.c b/fs/ubifs/super.c
index d4b4cb4..e360c7a 100644
--- a/fs/ubifs/super.c
+++ b/fs/ubifs/super.c
@@ -1213,6 +1213,13 @@
if (c->bulk_read == 1)
bu_init(c);
+ if (!c->ro_mount) {
+ c->write_reserve_buf = kmalloc(COMPRESSED_DATA_NODE_BUF_SZ,
+ GFP_KERNEL);
+ if (!c->write_reserve_buf)
+ goto out_free;
+ }
+
c->mounting = 1;
err = ubifs_read_superblock(c);
@@ -1482,6 +1489,7 @@
out_cbuf:
kfree(c->cbuf);
out_free:
+ kfree(c->write_reserve_buf);
kfree(c->bu.buf);
vfree(c->ileb_buf);
vfree(c->sbuf);
@@ -1520,6 +1528,7 @@
kfree(c->cbuf);
kfree(c->rcvrd_mst_node);
kfree(c->mst_node);
+ kfree(c->write_reserve_buf);
kfree(c->bu.buf);
vfree(c->ileb_buf);
vfree(c->sbuf);
@@ -1605,6 +1614,10 @@
goto out;
}
+ c->write_reserve_buf = kmalloc(COMPRESSED_DATA_NODE_BUF_SZ, GFP_KERNEL);
+ if (!c->write_reserve_buf)
+ goto out;
+
err = ubifs_lpt_init(c, 0, 1);
if (err)
goto out;
@@ -1669,6 +1682,8 @@
c->bgt = NULL;
}
free_wbufs(c);
+ kfree(c->write_reserve_buf);
+ c->write_reserve_buf = NULL;
vfree(c->ileb_buf);
c->ileb_buf = NULL;
ubifs_lpt_free(c, 1);
@@ -1712,6 +1727,8 @@
free_wbufs(c);
vfree(c->orph_buf);
c->orph_buf = NULL;
+ kfree(c->write_reserve_buf);
+ c->write_reserve_buf = NULL;
vfree(c->ileb_buf);
c->ileb_buf = NULL;
ubifs_lpt_free(c, 1);
@@ -1942,6 +1959,7 @@
mutex_init(&c->mst_mutex);
mutex_init(&c->umount_mutex);
mutex_init(&c->bu_mutex);
+ mutex_init(&c->write_reserve_mutex);
init_waitqueue_head(&c->cmt_wq);
c->buds = RB_ROOT;
c->old_idx = RB_ROOT;
diff --git a/fs/ubifs/ubifs.h b/fs/ubifs/ubifs.h
index 3624950..8c40ad3 100644
--- a/fs/ubifs/ubifs.h
+++ b/fs/ubifs/ubifs.h
@@ -151,6 +151,12 @@
*/
#define WORST_COMPR_FACTOR 2
+/*
+ * How much memory is needed for a buffer where we comress a data node.
+ */
+#define COMPRESSED_DATA_NODE_BUF_SZ \
+ (UBIFS_DATA_NODE_SZ + UBIFS_BLOCK_SIZE * WORST_COMPR_FACTOR)
+
/* Maximum expected tree height for use by bottom_up_buf */
#define BOTTOM_UP_HEIGHT 64
@@ -1005,6 +1011,11 @@
* @bu_mutex: protects the pre-allocated bulk-read buffer and @c->bu
* @bu: pre-allocated bulk-read information
*
+ * @write_reserve_mutex: protects @write_reserve_buf
+ * @write_reserve_buf: on the write path we allocate memory, which might
+ * sometimes be unavailable, in which case we use this
+ * write reserve buffer
+ *
* @log_lebs: number of logical eraseblocks in the log
* @log_bytes: log size in bytes
* @log_last: last LEB of the log
@@ -1256,6 +1267,9 @@
struct mutex bu_mutex;
struct bu_info bu;
+ struct mutex write_reserve_mutex;
+ void *write_reserve_buf;
+
int log_lebs;
long long log_bytes;
int log_last;
