While testing more corrupted images with hfsplus, i came across
one which triggered the following bug:
[15840.675016] BUG: unable to handle kernel paging request at
fffffffb
[15840.675016] IP: [<
c0116a4f>] kmap+0x15/0x56
[15840.675016] *pde =
00008067 *pte =
00000000
[15840.675016] Oops: 0000 [#1] PREEMPT DEBUG_PAGEALLOC
[15840.675016] Modules linked in:
[15840.675016]
[15840.675016] Pid: 11575, comm: ln Not tainted (
2.6.27-rc4-00123-gd3ee1b4-dirty #29)
[15840.675016] EIP: 0060:[<
c0116a4f>] EFLAGS:
00010202 CPU: 0
[15840.675016] EIP is at kmap+0x15/0x56
[15840.675016] EAX:
00000246 EBX:
fffffffb ECX:
00000000 EDX:
cab919c0
[15840.675016] ESI:
000007dd EDI:
cab0bcf4 EBP:
cab0bc98 ESP:
cab0bc94
[15840.675016] DS: 007b ES: 007b FS: 0000 GS: 0033 SS: 0068
[15840.675016] Process ln (pid: 11575, ti=
cab0b000 task=
cab919c0 task.ti=
cab0b000)
[15840.675016] Stack:
00000000 cab0bcdc c0231cfb 00000000 cab0bce0 00000800 ca9290c0 fffffffb
[15840.675016]
cab145d0 cab919c0 cab15998 22222222 22222222 22222222 00000001 cab15960
[15840.675016]
000007dd cab0bcf4 cab0bd04 c022cb3a cab0bcf4 cab15a6c ca9290c0 00000000
[15840.675016] Call Trace:
[15840.675016] [<
c0231cfb>] ? hfsplus_block_allocate+0x6f/0x2d3
[15840.675016] [<
c022cb3a>] ? hfsplus_file_extend+0xc4/0x1db
[15840.675016] [<
c022ce41>] ? hfsplus_get_block+0x8c/0x19d
[15840.675016] [<
c06adde4>] ? sub_preempt_count+0x9d/0xab
[15840.675016] [<
c019ece6>] ? __block_prepare_write+0x147/0x311
[15840.675016] [<
c0161934>] ? __grab_cache_page+0x52/0x73
[15840.675016] [<
c019ef4f>] ? block_write_begin+0x79/0xd5
[15840.675016] [<
c022cdb5>] ? hfsplus_get_block+0x0/0x19d
[15840.675016] [<
c019f22a>] ? cont_write_begin+0x27f/0x2af
[15840.675016] [<
c022cdb5>] ? hfsplus_get_block+0x0/0x19d
[15840.675016] [<
c0139ebe>] ? tick_program_event+0x28/0x4c
[15840.675016] [<
c013bd35>] ? trace_hardirqs_off+0xb/0xd
[15840.675016] [<
c022b723>] ? hfsplus_write_begin+0x2d/0x32
[15840.675016] [<
c022cdb5>] ? hfsplus_get_block+0x0/0x19d
[15840.675016] [<
c0161988>] ? pagecache_write_begin+0x33/0x107
[15840.675016] [<
c01879e5>] ? __page_symlink+0x3c/0xae
[15840.675016] [<
c019ad34>] ? __mark_inode_dirty+0x12f/0x137
[15840.675016] [<
c0187a70>] ? page_symlink+0x19/0x1e
[15840.675016] [<
c022e6eb>] ? hfsplus_symlink+0x41/0xa6
[15840.675016] [<
c01886a9>] ? vfs_symlink+0x99/0x101
[15840.675016] [<
c018a2f6>] ? sys_symlinkat+0x6b/0xad
[15840.675016] [<
c018a348>] ? sys_symlink+0x10/0x12
[15840.675016] [<
c01038bd>] ? sysenter_do_call+0x12/0x31
[15840.675016] =======================
[15840.675016] Code: 00 00 75 10 83 3d 88 2f ec c0 02 75 07 89 d0 e8 12 56 05 00 5d c3 55 ba 06 00 00 00 89 e5 53 89 c3 b8 3d eb 7e c0 e8 16 74 00 00 <8b> 03 c1 e8 1e 69 c0 d8 02 00 00 05 b8 69 8e c0 2b 80 c4 02 00
[15840.675016] EIP: [<
c0116a4f>] kmap+0x15/0x56 SS:ESP 0068:
cab0bc94
[15840.675016] ---[ end trace
4fea40dad6b70e5f ]---
This happens because the return value of read_mapping_page() is passed on
to kmap unchecked. The bug is triggered after the first
read_mapping_page() in hfsplus_block_allocate(), this patch fixes all
three usages in this functions but leaves the ones further down in the
file unchanged.
Signed-off-by: Eric Sesterhenn <snakebyte@gmx.de>
Cc: Roman Zippel <zippel@linux-m68k.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
mutex_lock(&HFSPLUS_SB(sb).alloc_file->i_mutex);
mapping = HFSPLUS_SB(sb).alloc_file->i_mapping;
page = read_mapping_page(mapping, offset / PAGE_CACHE_BITS, NULL);
+ if (IS_ERR(page)) {
+ start = size;
+ goto out;
+ }
pptr = kmap(page);
curr = pptr + (offset & (PAGE_CACHE_BITS - 1)) / 32;
i = offset % 32;
break;
page = read_mapping_page(mapping, offset / PAGE_CACHE_BITS,
NULL);
+ if (IS_ERR(page)) {
+ start = size;
+ goto out;
+ }
curr = pptr = kmap(page);
if ((size ^ offset) / PAGE_CACHE_BITS)
end = pptr + PAGE_CACHE_BITS / 32;
offset += PAGE_CACHE_BITS;
page = read_mapping_page(mapping, offset / PAGE_CACHE_BITS,
NULL);
+ if (IS_ERR(page)) {
+ start = size;
+ goto out;
+ }
pptr = kmap(page);
curr = pptr;
end = pptr + PAGE_CACHE_BITS / 32;