hi,
another bug triggered with a corrupted image, this time running fsx on
it.
[ 8898.458022] BUG: unable to handle kernel paging request at fffffff8
[ 8898.458022] IP: [<c023028c>] hfs_bnode_split+0x216/0x280
[ 8898.458022] *pde = 00008067 *pte = 00000000
[ 8898.458022] Oops: 0002 [#1] PREEMPT DEBUG_PAGEALLOC
[ 8898.458022] Modules linked in:
[ 8898.458022]
[ 8898.458022] Pid: 30558, comm: fsx Not tainted (2.6.27-rc4-00131-g83097ac-dirty #33)
[ 8898.458022] EIP: 0060:[<c023028c>] EFLAGS: 00010246 CPU: 0
[ 8898.458022] EIP is at hfs_bnode_split+0x216/0x280
[ 8898.458022] EAX: 00000002 EBX: fffffff4 ECX: 00000000 EDX: c6b48be6
[ 8898.458022] ESI: c6ac25a0 EDI: c6ac26c0 EBP: c6b48c00 ESP: c6b48bc4
[ 8898.458022] DS: 007b ES: 007b FS: 0000 GS: 0033 SS: 0068
[ 8898.458022] Process fsx (pid: 30558, ti=c6b48000 task=c6a06700 task.ti=c6b48000)
[ 8898.458022] Stack: c6a93950 00000010 00000fd8 00000fb8 00000830 000007ac 00000000 00000822
[ 8898.458022] 00000fb8 00000200 01ff0000 00001000 00000fb8 c6ac26c0 00000fb6 c6b48c4c
[ 8898.458022] c02303ca 000000f8 c6b48c68 c6b48e70 0000000f c6ac26c0 c6a93950 00000000
[ 8898.458022] Call Trace:
[ 8898.458022] [<c02303ca>] ? hfsplus_brec_insert+0xd4/0x2bc
[ 8898.458022] [<c022d7be>] ? hfsplus_create_cat+0x30c/0x38f
[ 8898.458022] [<c013dc3a>] ? trace_hardirqs_on+0xb/0xd
[ 8898.458022] [<c017cd20>] ? check_bytes_and_report+0x21/0x8f
[ 8898.458022] [<c017d8a3>] ? __slab_alloc+0xb1/0x503
[ 8898.458022] [<c017dfc5>] ? kmem_cache_alloc+0x4e/0xba
[ 8898.458022] [<c06ade04>] ? sub_preempt_count+0x9d/0xab
[ 8898.458022] [<c013dc3a>] ? trace_hardirqs_on+0xb/0xd
[ 8898.458022] [<c06abe19>] ? _spin_unlock+0x27/0x3c
[ 8898.458022] [<c019ad34>] ? __mark_inode_dirty+0x12f/0x137
[ 8898.458022] [<c06ade04>] ? sub_preempt_count+0x9d/0xab
[ 8898.458022] [<c06abe19>] ? _spin_unlock+0x27/0x3c
[ 8898.458022] [<c022e3db>] ? hfsplus_create+0x35/0x7a
[ 8898.458022] [<c01889ee>] ? vfs_create+0xa6/0x114
[ 8898.458022] [<c018ac61>] ? do_filp_open+0x1ad/0x62f
[ 8898.458022] [<c06abe19>] ? _spin_unlock+0x27/0x3c
[ 8898.458022] [<c0193632>] ? alloc_fd+0xbf/0xc9
[ 8898.458022] [<c06ade04>] ? sub_preempt_count+0x9d/0xab
[ 8898.458022] [<c0193632>] ? alloc_fd+0xbf/0xc9
[ 8898.458022] [<c0180249>] ? do_sys_open+0x42/0xb8
[ 8898.458022] [<c0180301>] ? sys_open+0x1e/0x26
[ 8898.458022] [<c01038bd>] ? sysenter_do_call+0x12/0x31
[ 8898.458022] =======================
[ 8898.458022] Code: c2 89 f8 66 89 55 f0 8d 55 e6 e8 63 ed ff ff 8b 56 0c 83 c4 14 85 d2 74 3f 8b 45 c4 e8 f5 f1 ff ff 31 c9 8d 55 e6 89 c3 8b 46 08 <89> 43 04 89 d8 6a 0e e8 1d ee ff ff 8b 43 04 8d 55 e6 0f c8 31
[ 8898.458022] EIP: [<c023028c>] hfs_bnode_split+0x216/0x280 SS:ESP 0068:c6b48bc4
[ 8898.458022] ---[ end trace 552e26de6e943a04 ]---
Problem is caused by not checking and propagating the return
valie of hfs_bnode_find()
Signed-off-by: Eric Sesterhenn <[email protected]>
--- linux/fs/hfsplus/brec.c.orig 2008-08-26 19:18:56.000000000 +0200
+++ linux/fs/hfsplus/brec.c 2008-08-26 19:19:27.000000000 +0200
@@ -304,6 +304,8 @@ static struct hfs_bnode *hfs_bnode_split
/* update next bnode header */
if (new_node->next) {
struct hfs_bnode *next_node = hfs_bnode_find(tree, new_node->next);
+ if (IS_ERR(next_node))
+ return next_node;
next_node->prev = new_node->this;
hfs_bnode_read(next_node, &node_desc, 0, sizeof(node_desc));
node_desc.prev = cpu_to_be32(next_node->prev);
Hi,
On Tue, 26 Aug 2008, Eric Sesterhenn wrote:
> --- linux/fs/hfsplus/brec.c.orig 2008-08-26 19:18:56.000000000 +0200
> +++ linux/fs/hfsplus/brec.c 2008-08-26 19:19:27.000000000 +0200
> @@ -304,6 +304,8 @@ static struct hfs_bnode *hfs_bnode_split
> /* update next bnode header */
> if (new_node->next) {
> struct hfs_bnode *next_node = hfs_bnode_find(tree, new_node->next);
> + if (IS_ERR(next_node))
> + return next_node;
> next_node->prev = new_node->this;
> hfs_bnode_read(next_node, &node_desc, 0, sizeof(node_desc));
> node_desc.prev = cpu_to_be32(next_node->prev);
Al Viro already fixed this for HFS in a better way, could you please adopt
his solution?
bye, Roman