Skip to contentSkip to conversation

fs/ntfs3: fix null pointer dereference in d_flags_for_inode

PATCH In review Liangbin Lian 2 participants 5 messages May 6, 2022 – Jun 11, 2022 NTFS3 filesystem

lore
Share & tools
lore
#0 lore raw

PATCH 1/1 fs/ntfs3: fix null pointer dereference in d_flags_for_inode

ntfs_read_mft may return inode with null i_op, cause null pointer dereference in d_flags_for_inode (inode->i_op->get_link). Reproduce:
- sudo mount -t ntfs3 -o loop ntfs.img ntfs
- ls ntfs/'$Extend/$Quota'

The call trace is shown below (striped):

 BUG: kernel NULL pointer dereference, address: 0000000000000008
 CPU: 0 PID: 577 Comm: ls Tainted: G           OE     5.16.0-0.bpo.4-amd64 #1  Debian 5.16.12-1~bpo11+1
 RIP: 0010:d_flags_for_inode+0x65/0x90
 Call Trace:
 ntfs_lookup
 +--- dir_search_u
 |    +--- ntfs_iget5
 |         +--- ntfs_read_mft
 +--- d_splice_alias
      +--- __d_add
           +--- d_flags_for_inode
---
 fs/ntfs3/inode.c | 1 -
 1 file changed, 1 deletion(-)
fs/ntfs3/inode.c +0 -1
index 9eab11e3b..b68d26fa8 100644
@@ -45,7 +45,6 @@ static struct inode *ntfs_read_mft(struct inode *inode,
 	struct MFT_REC *rec;
 	struct runs_tree *run;
 
-	inode->i_op = NULL;
 	/* Setup 'uid' and 'gid' */
 	inode->i_uid = sbi->options->fs_uid;
 	inode->i_gid = sbi->options->fs_gid;
Signature · 1 line
2.32.0 (Apple Git-132)
#1 lore raw

Hello.

Thank you for reporting this bug.
The bug happens because we don't initialize i_op for records in $Extend. We tested patch on our side, let me know if patch helps you too.

fs/ntfs3: Fix missing i_op in ntfs_read_mft

There is null pointer dereference because i_op == NULL.
The bug happens because we don't initialize i_op for records in $Extend.

Fixes 82cae269cfa9 ("fs/ntfs3: Add initialization of super block")
fs/ntfs3/inode.c +1 -0
index 879952254071..b2cc1191be69 100644
@@ -430,6 +430,7 @@ static struct inode *ntfs_read_mft(struct inode *inode,
        } else if (fname && fname->home.low == cpu_to_le32(MFT_REC_EXTEND) &&
                   fname->home.seq == cpu_to_le16(MFT_REC_EXTEND)) {
                /* Records in $Extend are not a files or general directories. */
+               inode->i_op = &ntfs_file_inode_operations;
        } else {
                err = -EINVAL;
                goto out;
Quoted 35 lines · Liangbin Lian

ntfs_read_mft may return inode with null i_op, cause null pointer dereference in d_flags_for_inode (inode->i_op->get_link). Reproduce:
- sudo mount -t ntfs3 -o loop ntfs.img ntfs
- ls ntfs/'$Extend/$Quota'

The call trace is shown below (striped):

  BUG: kernel NULL pointer dereference, address: 0000000000000008
  CPU: 0 PID: 577 Comm: ls Tainted: G           OE     5.16.0-0.bpo.4-amd64 #1  Debian 5.16.12-1~bpo11+1
  RIP: 0010:d_flags_for_inode+0x65/0x90
  Call Trace:
  ntfs_lookup
  +--- dir_search_u
  |    +--- ntfs_iget5
  |         +--- ntfs_read_mft
  +--- d_splice_alias
       +--- __d_add
            +--- d_flags_for_inode
---
  fs/ntfs3/inode.c | 1 -
  1 file changed, 1 deletion(-)
fs/ntfs3/inode.c +0 -1
index 9eab11e3b..b68d26fa8 100644
@@ -45,7 +45,6 @@ static struct inode *ntfs_read_mft(struct inode *inode,
  	struct MFT_REC *rec;
  	struct runs_tree *run;
 
-	inode->i_op = NULL;
  	/* Setup 'uid' and 'gid' */
  	inode->i_uid = sbi->options->fs_uid;
  	inode->i_gid = sbi->options->fs_gid;
author
#2 lore raw

Hello.
inode->i_op already initialized when inode alloc, this bug was introduced by inode->i_op = NULL;, just delete this line.
Please check my patch, maybe it's a better one, I have tested it on my project.

Quoted 66 lines · Almaz Alexandrovich

Hello.

Thank you for reporting this bug.
The bug happens because we don't initialize i_op for records in $Extend. We tested patch on our side, let me know if patch helps you too.

fs/ntfs3: Fix missing i_op in ntfs_read_mft

There is null pointer dereference because i_op == NULL.
The bug happens because we don't initialize i_op for records in $Extend.

Fixes 82cae269cfa9 ("fs/ntfs3: Add initialization of super block")
fs/ntfs3/inode.c +1 -0
index 879952254071..b2cc1191be69 100644
@@ -430,6 +430,7 @@ static struct inode *ntfs_read_mft(struct inode *inode,
         } else if (fname && fname->home.low == cpu_to_le32(MFT_REC_EXTEND) &&
                    fname->home.seq == cpu_to_le16(MFT_REC_EXTEND)) {
                 /* Records in $Extend are not a files or general directories. */
+               inode->i_op = &ntfs_file_inode_operations;
         } else {
                 err = -EINVAL;
                 goto out;
Quoted 35 lines · Liangbin Lian

ntfs_read_mft may return inode with null i_op, cause null pointer dereference in d_flags_for_inode (inode->i_op->get_link). Reproduce:
- sudo mount -t ntfs3 -o loop ntfs.img ntfs
- ls ntfs/'$Extend/$Quota'

The call trace is shown below (striped):

  BUG: kernel NULL pointer dereference, address: 0000000000000008
  CPU: 0 PID: 577 Comm: ls Tainted: G           OE     5.16.0-0.bpo.4-amd64 #1  Debian 5.16.12-1~bpo11+1
  RIP: 0010:d_flags_for_inode+0x65/0x90
  Call Trace:
  ntfs_lookup
  +--- dir_search_u
  |    +--- ntfs_iget5
  |         +--- ntfs_read_mft
  +--- d_splice_alias
       +--- __d_add
            +--- d_flags_for_inode
---
  fs/ntfs3/inode.c | 1 -
  1 file changed, 1 deletion(-)
fs/ntfs3/inode.c +0 -1
index 9eab11e3b..b68d26fa8 100644
@@ -45,7 +45,6 @@ static struct inode *ntfs_read_mft(struct inode *inode,
      struct MFT_REC *rec;
      struct runs_tree *run;
 
-     inode->i_op = NULL;
      /* Setup 'uid' and 'gid' */
      inode->i_uid = sbi->options->fs_uid;
      inode->i_gid = sbi->options->fs_gid;
#3 lore raw

Hello.

In the end of ntfs_read_mft function we must assign correct i_op:
inode->i_op = &ntfs_dir_inode_operations;
<...>
inode->i_op = &ntfs_link_inode_operations;
<...>
inode->i_op = &ntfs_file_inode_operations;
<...>
inode->i_op = &ntfs_special_inode_operations;

In this if .. else if .. else is an error:
records in $Extend doesn't get correct i_op.
This was fixed in my patch.

Line in beginning of ntfs_read_mft function
inode->i_op = NULL;
triggered null pointer dereference because
inode->i_op = &ntfs_file_inode_operations;
is missing for records in $Extend.

If I just remove inode->i_op = NULL,
then in i_op will be some previous value. Sometimes this value is correct, sometimes it's not.

I'm thankful, that you've spent time to find and debug this issue. This was reflected in line Reported-by: Liangbin Lian <[email protected]> I hope I've made more clear my previous message.

Quoted 72 lines · 练亮斌

Hello.
inode->i_op already initialized when inode alloc, this bug was introduced by inode->i_op = NULL;, just delete this line.
Please check my patch, maybe it's a better one, I have tested it on my project.

Quoted 66 lines · Almaz Alexandrovich

Hello.

Thank you for reporting this bug.
The bug happens because we don't initialize i_op for records in $Extend. We tested patch on our side, let me know if patch helps you too.

fs/ntfs3: Fix missing i_op in ntfs_read_mft

There is null pointer dereference because i_op == NULL.
The bug happens because we don't initialize i_op for records in $Extend.

Fixes 82cae269cfa9 ("fs/ntfs3: Add initialization of super block")
fs/ntfs3/inode.c +1 -0
index 879952254071..b2cc1191be69 100644
@@ -430,6 +430,7 @@ static struct inode *ntfs_read_mft(struct inode *inode,
          } else if (fname && fname->home.low == cpu_to_le32(MFT_REC_EXTEND) &&
                     fname->home.seq == cpu_to_le16(MFT_REC_EXTEND)) {
                  /* Records in $Extend are not a files or general directories. */
+               inode->i_op = &ntfs_file_inode_operations;
          } else {
                  err = -EINVAL;
                  goto out;
Quoted 35 lines · Liangbin Lian

ntfs_read_mft may return inode with null i_op, cause null pointer dereference in d_flags_for_inode (inode->i_op->get_link). Reproduce:
- sudo mount -t ntfs3 -o loop ntfs.img ntfs
- ls ntfs/'$Extend/$Quota'

The call trace is shown below (striped):

   BUG: kernel NULL pointer dereference, address: 0000000000000008
   CPU: 0 PID: 577 Comm: ls Tainted: G           OE     5.16.0-0.bpo.4-amd64 #1  Debian 5.16.12-1~bpo11+1
   RIP: 0010:d_flags_for_inode+0x65/0x90
   Call Trace:
   ntfs_lookup
   +--- dir_search_u
   |    +--- ntfs_iget5
   |         +--- ntfs_read_mft
   +--- d_splice_alias
        +--- __d_add
             +--- d_flags_for_inode
---
   fs/ntfs3/inode.c | 1 -
   1 file changed, 1 deletion(-)
fs/ntfs3/inode.c +0 -1
index 9eab11e3b..b68d26fa8 100644
@@ -45,7 +45,6 @@ static struct inode *ntfs_read_mft(struct inode *inode,
       struct MFT_REC *rec;
       struct runs_tree *run;
 
-     inode->i_op = NULL;
       /* Setup 'uid' and 'gid' */
       inode->i_uid = sbi->options->fs_uid;
       inode->i_gid = sbi->options->fs_gid;
author
#4 lore raw

Hello.

If ntfs_file_inode_operations is ok for $Extend, that's fine with me.

Konstantin Komarov <[email protected]>

Quoted 104 lines · 于2022年6月1日周三 00:34写道:

Hello.

In the end of ntfs_read_mft function we must assign correct i_op:
inode->i_op = &ntfs_dir_inode_operations;
<...>
inode->i_op = &ntfs_link_inode_operations;
<...>
inode->i_op = &ntfs_file_inode_operations;
<...>
inode->i_op = &ntfs_special_inode_operations;

In this if .. else if .. else is an error:
records in $Extend doesn't get correct i_op.
This was fixed in my patch.

Line in beginning of ntfs_read_mft function
inode->i_op = NULL;
triggered null pointer dereference because
inode->i_op = &ntfs_file_inode_operations;
is missing for records in $Extend.

If I just remove inode->i_op = NULL,
then in i_op will be some previous value. Sometimes this value is correct, sometimes it's not.

I'm thankful, that you've spent time to find and debug this issue. This was reflected in line Reported-by: Liangbin Lian <[email protected]> I hope I've made more clear my previous message.

Quoted 72 lines · 练亮斌

Hello.
inode->i_op already initialized when inode alloc, this bug was introduced by inode->i_op = NULL;, just delete this line.
Please check my patch, maybe it's a better one, I have tested it on my project.

Quoted 66 lines · Almaz Alexandrovich

Hello.

Thank you for reporting this bug.
The bug happens because we don't initialize i_op for records in $Extend. We tested patch on our side, let me know if patch helps you too.

fs/ntfs3: Fix missing i_op in ntfs_read_mft

There is null pointer dereference because i_op == NULL.
The bug happens because we don't initialize i_op for records in $Extend.

Fixes 82cae269cfa9 ("fs/ntfs3: Add initialization of super block")
fs/ntfs3/inode.c +1 -0
index 879952254071..b2cc1191be69 100644
@@ -430,6 +430,7 @@ static struct inode *ntfs_read_mft(struct inode *inode,
          } else if (fname && fname->home.low == cpu_to_le32(MFT_REC_EXTEND) &&
                     fname->home.seq == cpu_to_le16(MFT_REC_EXTEND)) {
                  /* Records in $Extend are not a files or general directories. */
+               inode->i_op = &ntfs_file_inode_operations;
          } else {
                  err = -EINVAL;
                  goto out;
Quoted 35 lines · Liangbin Lian

ntfs_read_mft may return inode with null i_op, cause null pointer dereference in d_flags_for_inode (inode->i_op->get_link). Reproduce:
- sudo mount -t ntfs3 -o loop ntfs.img ntfs
- ls ntfs/'$Extend/$Quota'

The call trace is shown below (striped):

   BUG: kernel NULL pointer dereference, address: 0000000000000008
   CPU: 0 PID: 577 Comm: ls Tainted: G           OE     5.16.0-0.bpo.4-amd64 #1  Debian 5.16.12-1~bpo11+1
   RIP: 0010:d_flags_for_inode+0x65/0x90
   Call Trace:
   ntfs_lookup
   +--- dir_search_u
   |    +--- ntfs_iget5
   |         +--- ntfs_read_mft
   +--- d_splice_alias
        +--- __d_add
             +--- d_flags_for_inode
---
   fs/ntfs3/inode.c | 1 -
   1 file changed, 1 deletion(-)
fs/ntfs3/inode.c +0 -1
index 9eab11e3b..b68d26fa8 100644
@@ -45,7 +45,6 @@ static struct inode *ntfs_read_mft(struct inode *inode,
       struct MFT_REC *rec;
       struct runs_tree *run;
 
-     inode->i_op = NULL;
       /* Setup 'uid' and 'gid' */
       inode->i_uid = sbi->options->fs_uid;
       inode->i_gid = sbi->options->fs_gid;