ocfs2: fix recursive semaphore deadlock in fiemap call

syzbot detected a OCFS2 hang due to a recursive semaphore on a
FS_IOC_FIEMAP of the extent list on a specially crafted mmap file.

context_switch kernel/sched/core.c:5357 [inline]
__schedule+0x1798/0x4cc0 kernel/sched/core.c:6961
__schedule_loop kernel/sched/core.c:7043 [inline]
schedule+0x165/0x360 kernel/sched/core.c:7058
schedule_preempt_disabled+0x13/0x30 kernel/sched/core.c:7115
rwsem_down_write_slowpath+0x872/0xfe0 kernel/locking/rwsem.c:1185
__down_write_common kernel/locking/rwsem.c:1317 [inline]
__down_write kernel/locking/rwsem.c:1326 [inline]
down_write+0x1ab/0x1f0 kernel/locking/rwsem.c:1591
ocfs2_page_mkwrite+0x2ff/0xc40 fs/ocfs2/mmap.c:142
do_page_mkwrite+0x14d/0x310 mm/memory.c:3361
wp_page_shared mm/memory.c:3762 [inline]
do_wp_page+0x268d/0x5800 mm/memory.c:3981
handle_pte_fault mm/memory.c:6068 [inline]
__handle_mm_fault+0x1033/0x5440 mm/memory.c:6195
handle_mm_fault+0x40a/0x8e0 mm/memory.c:6364
do_user_addr_fault+0x764/0x1390 arch/x86/mm/fault.c:1387
handle_page_fault arch/x86/mm/fault.c:1476 [inline]
exc_page_fault+0x76/0xf0 arch/x86/mm/fault.c:1532
asm_exc_page_fault+0x26/0x30 arch/x86/include/asm/idtentry.h:623
RIP: 0010:copy_user_generic arch/x86/include/asm/uaccess_64.h:126 [inline]
RIP: 0010:raw_copy_to_user arch/x86/include/asm/uaccess_64.h:147 [inline]
RIP: 0010:_inline_copy_to_user include/linux/uaccess.h:197 [inline]
RIP: 0010:_copy_to_user+0x85/0xb0 lib/usercopy.c:26
Code: e8 00 bc f7 fc 4d 39 fc 72 3d 4d 39 ec 77 38 e8 91 b9 f7 fc 4c 89
f7 89 de e8 47 25 5b fd 0f 01 cb 4c 89 ff 48 89 d9 4c 89 f6 <f3> a4 0f
1f 00 48 89 cb 0f 01 ca 48 89 d8 5b 41 5c 41 5d 41 5e 41
RSP: 0018:ffffc9000403f950 EFLAGS: 00050256
RAX: ffffffff84c7f101 RBX: 0000000000000038 RCX: 0000000000000038
RDX: 0000000000000000 RSI: ffffc9000403f9e0 RDI: 0000200000000060
RBP: ffffc9000403fa90 R08: ffffc9000403fa17 R09: 1ffff92000807f42
R10: dffffc0000000000 R11: fffff52000807f43 R12: 0000200000000098
R13: 00007ffffffff000 R14: ffffc9000403f9e0 R15: 0000200000000060
copy_to_user include/linux/uaccess.h:225 [inline]
fiemap_fill_next_extent+0x1c0/0x390 fs/ioctl.c:145
ocfs2_fiemap+0x888/0xc90 fs/ocfs2/extent_map.c:806
ioctl_fiemap fs/ioctl.c:220 [inline]
do_vfs_ioctl+0x1173/0x1430 fs/ioctl.c:532
__do_sys_ioctl fs/ioctl.c:596 [inline]
__se_sys_ioctl+0x82/0x170 fs/ioctl.c:584
do_syscall_x64 arch/x86/entry/syscall_64.c:63 [inline]
do_syscall_64+0xfa/0x3b0 arch/x86/entry/syscall_64.c:94
entry_SYSCALL_64_after_hwframe+0x77/0x7f
RIP: 0033:0x7f5f13850fd9
RSP: 002b:00007ffe3b3518b8 EFLAGS: 00000246 ORIG_RAX: 0000000000000010
RAX: ffffffffffffffda RBX: 0000200000000000 RCX: 00007f5f13850fd9
RDX: 0000200000000040 RSI: 00000000c020660b RDI: 0000000000000004
RBP: 6165627472616568 R08: 0000000000000000 R09: 0000000000000000
R10: 0000000000000000 R11: 0000000000000246 R12: 00007ffe3b3518f0
R13: 00007ffe3b351b18 R14: 431bde82d7b634db R15: 00007f5f1389a03b

ocfs2_fiemap() takes a read lock of the ip_alloc_sem semaphore (since
v2.6.22-527-g7307de80510a) and calls fiemap_fill_next_extent() to read the
extent list of this running mmap executable. The user supplied buffer to
hold the fiemap information page faults calling ocfs2_page_mkwrite() which
will take a write lock (since v2.6.27-38-g00dc417fa3e7) of the same
semaphore. This recursive semaphore will hold filesystem locks and causes
a hang of the fileystem.

The ip_alloc_sem protects the inode extent list and size. Release the
read semphore before calling fiemap_fill_next_extent() in ocfs2_fiemap()
and ocfs2_fiemap_inline(). This does an unnecessary semaphore lock/unlock
on the last extent but simplifies the error path.

Link: https://lkml.kernel.org/r/[e-mail]
Fixes: 00dc417fa3e7 ("ocfs2: fiemap support")
Signed-off-by: Mark Tinguely <[e-mail]>
Reported-by: [e-mail]
Closes: https://syzkaller.appspot.com/bug?extid=541dcc6ee768f77103e7
Reviewed-by: Joseph Qi <[e-mail]>
Cc: Mark Fasheh <[e-mail]>
Cc: Joel Becker <[e-mail]>
Cc: Junxiao Bi <[e-mail]>
Cc: Changwei Ge <[e-mail]>
Cc: Jun Piao <[e-mail]>
Cc: <[e-mail]>
Signed-off-by: Andrew Morton <[e-mail]>

fs/ocfs2/extent_map.c

···
  * it not only handles the fiemap for inlined files, but also deals
  * with the fast symlink, cause they have no difference for extent
  * mapping per se.
+ *
+ * Must be called with ip_alloc_sem semaphore held.
  */
 static int ocfs2_fiemap_inline(struct inode *inode, struct buffer_head *di_bh,
 			       struct fiemap_extent_info *fieinfo,
···
 	u64 phys;
 	u32 flags = FIEMAP_EXTENT_DATA_INLINE|FIEMAP_EXTENT_LAST;
 	struct ocfs2_inode_info *oi = OCFS2_I(inode);
+	lockdep_assert_held_read(&oi->ip_alloc_sem);
 
 	di = (struct ocfs2_dinode *)di_bh->b_data;
 	if (ocfs2_inode_is_fast_symlink(inode))
···
 			phys += offsetof(struct ocfs2_dinode,
 					 id2.i_data.id_data);
 
+		/* Release the ip_alloc_sem to prevent deadlock on page fault */
+		up_read(&OCFS2_I(inode)->ip_alloc_sem);
 		ret = fiemap_fill_next_extent(fieinfo, 0, phys, id_count,
 					      flags);
+		down_read(&OCFS2_I(inode)->ip_alloc_sem);
 		if (ret < 0)
 			return ret;
 	}
···
 		len_bytes = (u64)le16_to_cpu(rec.e_leaf_clusters) << osb->s_clustersize_bits;
 		phys_bytes = le64_to_cpu(rec.e_blkno) << osb->sb->s_blocksize_bits;
 		virt_bytes = (u64)le32_to_cpu(rec.e_cpos) << osb->s_clustersize_bits;
-
+		/* Release the ip_alloc_sem to prevent deadlock on page fault */
+		up_read(&OCFS2_I(inode)->ip_alloc_sem);
 		ret = fiemap_fill_next_extent(fieinfo, virt_bytes, phys_bytes,
 					      len_bytes, fe_flags);
+		down_read(&OCFS2_I(inode)->ip_alloc_sem);
 		if (ret)
 			break;