[patch 05/17] ocfs2: protect extent tree in ocfs2_prepare_inode_for_write()

[patch 05/17] ocfs2: protect extent tree in ocfs2_prepare_inode_for_write()

[akpm]

From: Shuning Zhang <sunny.s.zhang@oracle.com>
Subject: ocfs2: protect extent tree in ocfs2_prepare_inode_for_write()

When the extent tree is modified, it should be protected by inode cluster
lock and ip_alloc_sem.

The extent tree is accessed and modified in the
ocfs2_prepare_inode_for_write, but isn't protected by ip_alloc_sem.

The following is a case.  The function ocfs2_fiemap is accessing the
extent tree, which is modified at the same time.

[47145.974472] kernel BUG at fs/ocfs2/extent_map.c:475!
[47145.974480] invalid opcode: 0000 [#1] SMP
[47145.974489] Modules linked in: tun ocfs2 ocfs2_nodemanager configfs
ocfs2_stackglue xen_netback xen_blkback xen_gntalloc xen_gntdev xen_evtchn
xenfs xen_privcmd vfat fat bnx2fc fcoe libfcoe libfc scsi_transport_fc sunrpc
bridge 8021q mrp garp stp llc ib_iser rdma_cm ib_cm iw_cm ib_sa ib_mad
ib_core ib_addr dm_round_robin dm_multipath sg pcspkr raid1 shpchp
ipmi_devintf ipmi_msghandler ext4 jbd2 mbcache2 sd_mod nvme nvme_core bnxt_en
xhci_pci xhci_hcd crc32c_intel be2iscsi bnx2i cnic uio cxgb4i cxgb4 cxgb3i
libcxgbi ipv6 cxgb3 mdio qla4xxx wmi dm_mirror dm_region_hash dm_log dm_mod
iscsi_tcp libiscsi_tcp libiscsi scsi_transport_iscsi iscsi_ibft
iscsi_boot_sysfs
[47145.974636] CPU: 16 PID: 14047 Comm: o2info Not tainted
4.1.12-124.23.1.el6uek.x86_64 #2
[47145.974646] Hardware name: Oracle Corporation ORACLE SERVER X7-2L/ASM, MB
MECH, X7-2L, BIOS 42040600 10/19/2018
[47145.974658] task: ffff88019487e200 ti: ffff88003daa4000 task.ti:
ffff88003daa4000
[47145.974667] RIP: e030:[<ffffffffa05e4840>]  [<ffffffffa05e4840>]
ocfs2_get_clusters_nocache.isra.11+0x390/0x550 [ocfs2]
[47145.974708] RSP: e02b:ffff88003daa7d88  EFLAGS: 00010287
[47145.974713] RAX: 00000000000000de RBX: ffff8801d1104030 RCX:
ffff8801d1104e10
[47145.974719] RDX: 00000000000000de RSI: 000000000009ec40 RDI:
ffff8801d1104e24
[47145.974725] RBP: ffff88003daa7df8 R08: ffff88003daa7e38 R09:
0000000000000000
[47145.974732] R10: 000000000009ec3f R11: 0000000000000246 R12:
000000000009ec3f
[47145.974739] R13: ffff88004c419000 R14: 0000000000000002 R15:
ffff88003daa7e28
[47145.974754] FS:  00007fdbccc92720(0000) GS:ffff880358800000(0000)
knlGS:ffff880358800000
[47145.974764] CS:  e033 DS: 0000 ES: 0000 CR0: 0000000080050033
[47145.974772] CR2: 00007fd5dfcd8350 CR3: 0000000208677000 CR4:
0000000000042660
[47145.974785] Stack:
[47145.974790]  ffff88003daa7df8 00002000a05e249b ffff8801d1104000
ffff88003daa7e2c
[47145.974802]  ffff88003daa7e38 ffff88000cc484c0 ffff880145f5b478
0000000000000000
[47145.974811]  0000000000002000 ffff88000cc484c0 ffff88003daa7ea0
0000000000000000
[47145.974820] Call Trace:
[47145.974837]  [<ffffffffa05e53e3>] ocfs2_fiemap+0x1e3/0x430 [ocfs2]
[47145.974848]  [<ffffffff816f644f>] ? xen_hypervisor_callback+0x7f/0x120
[47145.974855]  [<ffffffff816f6448>] ? xen_hypervisor_callback+0x78/0x120
[47145.974861]  [<ffffffff816f64a3>] ? xen_hypervisor_callback+0xd3/0x120
[47145.974872]  [<ffffffff81220905>] do_vfs_ioctl+0x155/0x510
[47145.974878]  [<ffffffff81220d41>] SyS_ioctl+0x81/0xa0
[47145.974885]  [<ffffffff816f1b9f>] ? system_call_after_swapgs+0xe9/0x190
[47145.974891]  [<ffffffff816f1b98>] ? system_call_after_swapgs+0xe2/0x190
[47145.974899]  [<ffffffff816f1b91>] ? system_call_after_swapgs+0xdb/0x190
[47145.974905]  [<ffffffff816f1c5e>] system_call_fastpath+0x18/0xd8
[47145.974910] Code: 18 48 c7 c6 60 7f 65 a0 31 c0 bb e2 ff ff ff 48 8b 4a 40
48 8b 7a 28 48 c7 c2 78 2d 66 a0 e8 38 4f 05 00 e9 28 fe ff ff 0f 1f 00 <0f>
0b 66 0f 1f 44 00 00 bb 86 ff ff ff e9 13 fe ff ff 66 0f 1f
[47145.975000] RIP  [<ffffffffa05e4840>]
ocfs2_get_clusters_nocache.isra.11+0x390/0x550 [ocfs2]
[47145.975018]  RSP <ffff88003daa7d88>
[47145.989999] ---[ end trace c8aa0c8180e869dc ]---
[47146.087579] Kernel panic - not syncing: Fatal exception
[47146.087691] Kernel Offset: disabled

This issue can be reproduced every week in a production environment.

This issue is related to the usage mode.  If others use ocfs2 in this
mode, the kernel will panic frequently.

[akpm@linux-foundation.org: coding style fixes]
Link: http://lkml.kernel.org/r/1568772175-2906-2-git-send-email-sunny.s.zhang@oracle.com
Signed-off-by: Shuning Zhang <sunny.s.zhang@oracle.com>
Reviewed-by: Junxiao Bi <junxiao.bi@oracle.com>
Reviewed-by: Gang He <ghe@suse.com>
Cc: Mark Fasheh <mark@fasheh.com>
Cc: Joel Becker <jlbec@evilplan.org>
Cc: Joseph Qi <jiangqi903@gmail.com>
Cc: Changwei Ge <gechangwei@live.cn>
Cc: Jun Piao <piaojun@huawei.com>
Cc: <stable@vger.kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
---

 fs/ocfs2/file.c |  125 ++++++++++++++++++++++++++++++++++++----------
 1 file changed, 99 insertions(+), 26 deletions(-)

--- a/fs/ocfs2/file.c~ocfs2-protect-extent-tree-in-the-ocfs2_prepare_inode_for_write
+++ a/fs/ocfs2/file.c
@@ -2125,26 +2125,89 @@ out:
 	return ret;
 }
 
+static int ocfs2_inode_lock_for_extent_tree(struct inode *inode,
+					    struct buffer_head **di_bh,
+					    int meta_level,
+					    int overwrite_io,
+					    int write_sem,
+					    int wait)
+{
+	int ret = 0;
+
+	if (wait)
+		ret = ocfs2_inode_lock(inode, NULL, meta_level);
+	else
+		ret = ocfs2_try_inode_lock(inode,
+			overwrite_io ? NULL : di_bh, meta_level);
+	if (ret < 0)
+		goto out;
+
+	if (wait) {
+		if (write_sem)
+			down_write(&OCFS2_I(inode)->ip_alloc_sem);
+		else
+			down_read(&OCFS2_I(inode)->ip_alloc_sem);
+	} else {
+		if (write_sem)
+			ret = down_write_trylock(&OCFS2_I(inode)->ip_alloc_sem);
+		else
+			ret = down_read_trylock(&OCFS2_I(inode)->ip_alloc_sem);
+
+		if (!ret) {
+			ret = -EAGAIN;
+			goto out_unlock;
+		}
+	}
+
+	return ret;
+
+out_unlock:
+	brelse(*di_bh);
+	ocfs2_inode_unlock(inode, meta_level);
+out:
+	return ret;
+}
+
+static void ocfs2_inode_unlock_for_extent_tree(struct inode *inode,
+					       struct buffer_head **di_bh,
+					       int meta_level,
+					       int write_sem)
+{
+	if (write_sem)
+		up_write(&OCFS2_I(inode)->ip_alloc_sem);
+	else
+		up_read(&OCFS2_I(inode)->ip_alloc_sem);
+
+	brelse(*di_bh);
+	*di_bh = NULL;
+
+	if (meta_level >= 0)
+		ocfs2_inode_unlock(inode, meta_level);
+}
+
 static int ocfs2_prepare_inode_for_write(struct file *file,
 					 loff_t pos, size_t count, int wait)
 {
 	int ret = 0, meta_level = 0, overwrite_io = 0;
+	int write_sem = 0;
 	struct dentry *dentry = file->f_path.dentry;
 	struct inode *inode = d_inode(dentry);
 	struct buffer_head *di_bh = NULL;
+	u32 cpos;
+	u32 clusters;
 
 	/*
 	 * We start with a read level meta lock and only jump to an ex
 	 * if we need to make modifications here.
 	 */
 	for(;;) {
-		if (wait)
-			ret = ocfs2_inode_lock(inode, NULL, meta_level);
-		else
-			ret = ocfs2_try_inode_lock(inode,
-				overwrite_io ? NULL : &di_bh, meta_level);
+		ret = ocfs2_inode_lock_for_extent_tree(inode,
+						       &di_bh,
+						       meta_level,
+						       overwrite_io,
+						       write_sem,
+						       wait);
 		if (ret < 0) {
-			meta_level = -1;
 			if (ret != -EAGAIN)
 				mlog_errno(ret);
 			goto out;
@@ -2156,15 +2219,8 @@ static int ocfs2_prepare_inode_for_write
 		 */
 		if (!wait && !overwrite_io) {
 			overwrite_io = 1;
-			if (!down_read_trylock(&OCFS2_I(inode)->ip_alloc_sem)) {
-				ret = -EAGAIN;
-				goto out_unlock;
-			}
 
 			ret = ocfs2_overwrite_io(inode, di_bh, pos, count);
-			brelse(di_bh);
-			di_bh = NULL;
-			up_read(&OCFS2_I(inode)->ip_alloc_sem);
 			if (ret < 0) {
 				if (ret != -EAGAIN)
 					mlog_errno(ret);
@@ -2183,7 +2239,10 @@ static int ocfs2_prepare_inode_for_write
 		 * set inode->i_size at the end of a write. */
 		if (should_remove_suid(dentry)) {
 			if (meta_level == 0) {
-				ocfs2_inode_unlock(inode, meta_level);
+				ocfs2_inode_unlock_for_extent_tree(inode,
+								   &di_bh,
+								   meta_level,
+								   write_sem);
 				meta_level = 1;
 				continue;
 			}
@@ -2197,18 +2256,32 @@ static int ocfs2_prepare_inode_for_write
 
 		ret = ocfs2_check_range_for_refcount(inode, pos, count);
 		if (ret == 1) {
-			ocfs2_inode_unlock(inode, meta_level);
-			meta_level = -1;
+			ocfs2_inode_unlock_for_extent_tree(inode,
+							   &di_bh,
+							   meta_level,
+							   write_sem);
+			ret = ocfs2_inode_lock_for_extent_tree(inode,
+							       &di_bh,
+							       meta_level,
+							       overwrite_io,
+							       1,
+							       wait);
+			write_sem = 1;
+			if (ret < 0) {
+				if (ret != -EAGAIN)
+					mlog_errno(ret);
+				goto out;
+			}
 
-			ret = ocfs2_prepare_inode_for_refcount(inode,
-							       file,
-							       pos,
-							       count,
-							       &meta_level);
+			cpos = pos >> OCFS2_SB(inode->i_sb)->s_clustersize_bits;
+			clusters =
+				ocfs2_clusters_for_bytes(inode->i_sb, pos + count) - cpos;
+			ret = ocfs2_refcount_cow(inode, di_bh, cpos, clusters, UINT_MAX);
 		}
 
 		if (ret < 0) {
-			mlog_errno(ret);
+			if (ret != -EAGAIN)
+				mlog_errno(ret);
 			goto out_unlock;
 		}
 
@@ -2219,10 +2292,10 @@ out_unlock:
 	trace_ocfs2_prepare_inode_for_write(OCFS2_I(inode)->ip_blkno,
 					    pos, count, wait);
 
-	brelse(di_bh);
-
-	if (meta_level >= 0)
-		ocfs2_inode_unlock(inode, meta_level);
+	ocfs2_inode_unlock_for_extent_tree(inode,
+					   &di_bh,
+					   meta_level,
+					   write_sem);
 
 out:
 	return ret;
_

Re: [patch 05/17] ocfs2: protect extent tree in ocfs2_prepare_inode_for_write()

[Linus Torvalds]

On Tue, Nov 5, 2019 at 9:16 PM <akpm@linux-foundation.org> wrote:
>
> From: Shuning Zhang <sunny.s.zhang@oracle.com>
> Subject: ocfs2: protect extent tree in ocfs2_prepare_inode_for_write()
>
> When the extent tree is modified, it should be protected by inode cluster
> lock and ip_alloc_sem.
>
> The extent tree is accessed and modified in the
> ocfs2_prepare_inode_for_write, but isn't protected by ip_alloc_sem.

This patch results in a new warning for me:

  fs/ocfs2/file.c:2101:12: warning: ‘ocfs2_prepare_inode_for_refcount’
defined but not used [-Wunused-function]
   2101 | static int ocfs2_prepare_inode_for_refcount(struct inode *inode,
        |            ^~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~

and I'm wondering why nobody seems  to have noticed that or fixed
things? Because it does look like this removed the only use of that
function, and everybody who compiled this should have seen this
warning?

Was this not tested at all?

              Linus

Re: [patch 05/17] ocfs2: protect extent tree in ocfs2_prepare_inode_for_write()

[Linus Torvalds]

On Tue, Nov 5, 2019 at 9:16 PM <akpm@linux-foundation.org> wrote:
>
> From: Shuning Zhang <sunny.s.zhang@oracle.com>
> Subject: ocfs2: protect extent tree in ocfs2_prepare_inode_for_write()
>
> When the extent tree is modified, it should be protected by inode cluster
> lock and ip_alloc_sem.
>
> The extent tree is accessed and modified in the
> ocfs2_prepare_inode_for_write, but isn't protected by ip_alloc_sem.

This patch results in a new warning for me:

  fs/ocfs2/file.c:2101:12: warning: ‘ocfs2_prepare_inode_for_refcount’
defined but not used [-Wunused-function]
   2101 | static int ocfs2_prepare_inode_for_refcount(struct inode *inode,
        |            ^~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~

and I'm wondering why nobody seems  to have noticed that or fixed
things? Because it does look like this removed the only use of that
function, and everybody who compiled this should have seen this
warning?

Was this not tested at all?

              Linus

Re: [patch 05/17] ocfs2: protect extent tree in ocfs2_prepare_inode_for_write()

[Andrew Morton]

On Wed, 6 Nov 2019 08:41:44 -0800 Linus Torvalds <torvalds@linux-foundation.org> wrote:

> On Tue, Nov 5, 2019 at 9:16 PM <akpm@linux-foundation.org> wrote:
> >
> > From: Shuning Zhang <sunny.s.zhang@oracle.com>
> > Subject: ocfs2: protect extent tree in ocfs2_prepare_inode_for_write()
> >
> > When the extent tree is modified, it should be protected by inode cluster
> > lock and ip_alloc_sem.
> >
> > The extent tree is accessed and modified in the
> > ocfs2_prepare_inode_for_write, but isn't protected by ip_alloc_sem.
> 
> This patch results in a new warning for me:
> 
>   fs/ocfs2/file.c:2101:12: warning: ‘ocfs2_prepare_inode_for_refcount’
> defined but not used [-Wunused-function]
>    2101 | static int ocfs2_prepare_inode_for_refcount(struct inode *inode,
>         |            ^~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
> 
> and I'm wondering why nobody seems  to have noticed that or fixed
> things? Because it does look like this removed the only use of that
> function, and everybody who compiled this should have seen this
> warning?
> 

Oop, sorry, that's in a separate patch which I failed to squish.  Shall
send now.