[PATCH] ext4: fix bug in extents parsing when number of entries in header is zero

[PATCH] ext4: fix bug in extents parsing when number of entries in header is zero

[Luís Henriques]

When walking through an inode extents, the ext4_ext_binsearch_idx() function
assumes that the extent header has been previously validated.  However,
there are no checks that verify that the number of entries (eh->eh_entries)
is non-zero.  And this will lead to problems because the EXT_FIRST_INDEX()
and EXT_LAST_INDEX() will return garbage and result in this:

[  135.245946] ------------[ cut here ]------------
[  135.247579] kernel BUG at fs/ext4/extents.c:2258!
[  135.249045] invalid opcode: 0000 [#1] PREEMPT SMP
[  135.250320] CPU: 2 PID: 238 Comm: tmp118 Not tainted 5.19.0-rc8+ #4
[  135.252067] Hardware name: QEMU Standard PC (i440FX + PIIX, 1996), BIOS rel-1.15.0-0-g2dd4b9b-rebuilt.opensuse.org 04/01/2014
[  135.255065] RIP: 0010:ext4_ext_map_blocks+0xc20/0xcb0
[  135.256475] Code:
[  135.261433] RSP: 0018:ffffc900005939f8 EFLAGS: 00010246
[  135.262847] RAX: 0000000000000024 RBX: ffffc90000593b70 RCX: 0000000000000023
[  135.264765] RDX: ffff8880038e5f10 RSI: 0000000000000003 RDI: ffff8880046e922c
[  135.266670] RBP: ffff8880046e9348 R08: 0000000000000001 R09: ffff888002ca580c
[  135.268576] R10: 0000000000002602 R11: 0000000000000000 R12: 0000000000000024
[  135.270477] R13: 0000000000000000 R14: 0000000000000024 R15: 0000000000000000
[  135.272394] FS:  00007fdabdc56740(0000) GS:ffff88807dd00000(0000) knlGS:0000000000000000
[  135.274510] CS:  0010 DS: 0000 ES: 0000 CR0: 0000000080050033
[  135.276075] CR2: 00007ffc26bd4f00 CR3: 0000000006261004 CR4: 0000000000170ea0
[  135.277952] Call Trace:
[  135.278635]  <TASK>
[  135.279247]  ? preempt_count_add+0x6d/0xa0
[  135.280358]  ? percpu_counter_add_batch+0x55/0xb0
[  135.281612]  ? _raw_read_unlock+0x18/0x30
[  135.282704]  ext4_map_blocks+0x294/0x5a0
[  135.283745]  ? xa_load+0x6f/0xa0
[  135.284562]  ext4_mpage_readpages+0x3d6/0x770
[  135.285646]  read_pages+0x67/0x1d0
[  135.286492]  ? folio_add_lru+0x51/0x80
[  135.287441]  page_cache_ra_unbounded+0x124/0x170
[  135.288510]  filemap_get_pages+0x23d/0x5a0
[  135.289457]  ? path_openat+0xa72/0xdd0
[  135.290332]  filemap_read+0xbf/0x300
[  135.291158]  ? _raw_spin_lock_irqsave+0x17/0x40
[  135.292192]  new_sync_read+0x103/0x170
[  135.293014]  vfs_read+0x15d/0x180
[  135.293745]  ksys_read+0xa1/0xe0
[  135.294461]  do_syscall_64+0x3c/0x80
[  135.295284]  entry_SYSCALL_64_after_hwframe+0x46/0xb0

Unfortunately, __ext4_ext_check() only verifies that eh->eh_entries doesn't
exceed eh->eh_max.  And since an empty leaf seems to be a valid value in
same cases, adding this extra check there isn't an option.

This patch simply adds the check directly in ext4_ext_binsearch_idx() and
propagates this error so that the kernel doesn't hit this BUG_ON() in
ext4_ext_determine_hole().

Link: https://bugzilla.kernel.org/show_bug.cgi?id=215941
Signed-off-by: Luís Henriques <lhenriques@suse.de>
---
 fs/ext4/extents.c | 13 ++++++++++---
 1 file changed, 10 insertions(+), 3 deletions(-)

Hi!

This bug is easily reproducible using the filesystem image provided --
it's just a matter of mounting it and run:

    $ cat /mnt/foo/bar/xattr

Anyway, I hope my analysis of the bug is correct -- the root cause seems
to be an extent header with an invalid value for in eh_entries, which will
later cause the BUG_ON().

Cheers,
--
Luís

diff --git a/fs/ext4/extents.c b/fs/ext4/extents.c
index c148bb97b527..53cfe2c681c4 100644
--- a/fs/ext4/extents.c
+++ b/fs/ext4/extents.c
@@ -738,7 +738,7 @@ void ext4_ext_drop_refs(struct ext4_ext_path *path)
  * binary search for the closest index of the given block
  * the header must be checked before calling this
  */
-static void
+static int
 ext4_ext_binsearch_idx(struct inode *inode,
 			struct ext4_ext_path *path, ext4_lblk_t block)
 {
@@ -748,6 +748,11 @@ ext4_ext_binsearch_idx(struct inode *inode,
 
 	ext_debug(inode, "binsearch for %u(idx):  ", block);
 
+	if (eh->eh_entries == 0) {
+		EXT4_ERROR_INODE(inode, "No entries in extent header!");
+		return -EFSCORRUPTED;
+	}
+
 	l = EXT_FIRST_INDEX(eh) + 1;
 	r = EXT_LAST_INDEX(eh);
 	while (l <= r) {
@@ -791,7 +796,7 @@ ext4_ext_binsearch_idx(struct inode *inode,
 		BUG_ON(chix != path->p_idx);
 	}
 #endif
-
+	return 0;
 }
 
 /*
@@ -919,7 +924,9 @@ ext4_find_extent(struct inode *inode, ext4_lblk_t block,
 		ext_debug(inode, "depth %d: num %d, max %d\n",
 			  ppos, le16_to_cpu(eh->eh_entries), le16_to_cpu(eh->eh_max));
 
-		ext4_ext_binsearch_idx(inode, path + ppos, block);
+		ret = ext4_ext_binsearch_idx(inode, path + ppos, block);
+		if (ret < 0)
+			goto err;
 		path[ppos].p_block = ext4_idx_pblock(path[ppos].p_idx);
 		path[ppos].p_depth = i;
 		path[ppos].p_ext = NULL;

Re: [PATCH] ext4: fix bug in extents parsing when number of entries in header is zero

[Luís Henriques]

On Fri, Aug 05, 2022 at 03:00:25PM +0100, Luís Henriques wrote:
> When walking through an inode extents, the ext4_ext_binsearch_idx() function
> assumes that the extent header has been previously validated.  However,
> there are no checks that verify that the number of entries (eh->eh_entries)
> is non-zero.  And this will lead to problems because the EXT_FIRST_INDEX()
> and EXT_LAST_INDEX() will return garbage and result in this:
> 
> [  135.245946] ------------[ cut here ]------------
> [  135.247579] kernel BUG at fs/ext4/extents.c:2258!
> [  135.249045] invalid opcode: 0000 [#1] PREEMPT SMP
> [  135.250320] CPU: 2 PID: 238 Comm: tmp118 Not tainted 5.19.0-rc8+ #4
> [  135.252067] Hardware name: QEMU Standard PC (i440FX + PIIX, 1996), BIOS rel-1.15.0-0-g2dd4b9b-rebuilt.opensuse.org 04/01/2014
> [  135.255065] RIP: 0010:ext4_ext_map_blocks+0xc20/0xcb0
> [  135.256475] Code:
> [  135.261433] RSP: 0018:ffffc900005939f8 EFLAGS: 00010246
> [  135.262847] RAX: 0000000000000024 RBX: ffffc90000593b70 RCX: 0000000000000023
> [  135.264765] RDX: ffff8880038e5f10 RSI: 0000000000000003 RDI: ffff8880046e922c
> [  135.266670] RBP: ffff8880046e9348 R08: 0000000000000001 R09: ffff888002ca580c
> [  135.268576] R10: 0000000000002602 R11: 0000000000000000 R12: 0000000000000024
> [  135.270477] R13: 0000000000000000 R14: 0000000000000024 R15: 0000000000000000
> [  135.272394] FS:  00007fdabdc56740(0000) GS:ffff88807dd00000(0000) knlGS:0000000000000000
> [  135.274510] CS:  0010 DS: 0000 ES: 0000 CR0: 0000000080050033
> [  135.276075] CR2: 00007ffc26bd4f00 CR3: 0000000006261004 CR4: 0000000000170ea0
> [  135.277952] Call Trace:
> [  135.278635]  <TASK>
> [  135.279247]  ? preempt_count_add+0x6d/0xa0
> [  135.280358]  ? percpu_counter_add_batch+0x55/0xb0
> [  135.281612]  ? _raw_read_unlock+0x18/0x30
> [  135.282704]  ext4_map_blocks+0x294/0x5a0
> [  135.283745]  ? xa_load+0x6f/0xa0
> [  135.284562]  ext4_mpage_readpages+0x3d6/0x770
> [  135.285646]  read_pages+0x67/0x1d0
> [  135.286492]  ? folio_add_lru+0x51/0x80
> [  135.287441]  page_cache_ra_unbounded+0x124/0x170
> [  135.288510]  filemap_get_pages+0x23d/0x5a0
> [  135.289457]  ? path_openat+0xa72/0xdd0
> [  135.290332]  filemap_read+0xbf/0x300
> [  135.291158]  ? _raw_spin_lock_irqsave+0x17/0x40
> [  135.292192]  new_sync_read+0x103/0x170
> [  135.293014]  vfs_read+0x15d/0x180
> [  135.293745]  ksys_read+0xa1/0xe0
> [  135.294461]  do_syscall_64+0x3c/0x80
> [  135.295284]  entry_SYSCALL_64_after_hwframe+0x46/0xb0
> 
> Unfortunately, __ext4_ext_check() only verifies that eh->eh_entries doesn't
> exceed eh->eh_max.  And since an empty leaf seems to be a valid value in
> same cases, adding this extra check there isn't an option.
> 
> This patch simply adds the check directly in ext4_ext_binsearch_idx() and
> propagates this error so that the kernel doesn't hit this BUG_ON() in
> ext4_ext_determine_hole().
> 
> Link: https://bugzilla.kernel.org/show_bug.cgi?id=215941
> Signed-off-by: Luís Henriques <lhenriques@suse.de>
> ---
>  fs/ext4/extents.c | 13 ++++++++++---
>  1 file changed, 10 insertions(+), 3 deletions(-)
> 
> Hi!
> 
> This bug is easily reproducible using the filesystem image provided --
> it's just a matter of mounting it and run:
> 
>     $ cat /mnt/foo/bar/xattr
> 
> Anyway, I hope my analysis of the bug is correct -- the root cause seems
> to be an extent header with an invalid value for in eh_entries, which will
> later cause the BUG_ON().

Although I did got any feedback yet, it looks like this patch also fixes
bugzilla #216283.  This issue is quite similar, but the BUG_ON() (a
different one) is hit on the write path.  Doing something like:

  $ echo 123 > /mnt/foo/bar/acl ; sync

is enough to crash the kernel with that image.  Also, in the bug my patch
initially refers to, the eh_entries field is '0' right on the root inode
(i.e., in the extent header in the inode.i_block).  For this other bug,
this happens in a non-root node.

Cheers,
--
Luís

> 
> Cheers,
> --
> Luís
> 
> diff --git a/fs/ext4/extents.c b/fs/ext4/extents.c
> index c148bb97b527..53cfe2c681c4 100644
> --- a/fs/ext4/extents.c
> +++ b/fs/ext4/extents.c
> @@ -738,7 +738,7 @@ void ext4_ext_drop_refs(struct ext4_ext_path *path)
>   * binary search for the closest index of the given block
>   * the header must be checked before calling this
>   */
> -static void
> +static int
>  ext4_ext_binsearch_idx(struct inode *inode,
>  			struct ext4_ext_path *path, ext4_lblk_t block)
>  {
> @@ -748,6 +748,11 @@ ext4_ext_binsearch_idx(struct inode *inode,
>  
>  	ext_debug(inode, "binsearch for %u(idx):  ", block);
>  
> +	if (eh->eh_entries == 0) {
> +		EXT4_ERROR_INODE(inode, "No entries in extent header!");
> +		return -EFSCORRUPTED;
> +	}
> +
>  	l = EXT_FIRST_INDEX(eh) + 1;
>  	r = EXT_LAST_INDEX(eh);
>  	while (l <= r) {
> @@ -791,7 +796,7 @@ ext4_ext_binsearch_idx(struct inode *inode,
>  		BUG_ON(chix != path->p_idx);
>  	}
>  #endif
> -
> +	return 0;
>  }
>  
>  /*
> @@ -919,7 +924,9 @@ ext4_find_extent(struct inode *inode, ext4_lblk_t block,
>  		ext_debug(inode, "depth %d: num %d, max %d\n",
>  			  ppos, le16_to_cpu(eh->eh_entries), le16_to_cpu(eh->eh_max));
>  
> -		ext4_ext_binsearch_idx(inode, path + ppos, block);
> +		ret = ext4_ext_binsearch_idx(inode, path + ppos, block);
> +		if (ret < 0)
> +			goto err;
>  		path[ppos].p_block = ext4_idx_pblock(path[ppos].p_idx);
>  		path[ppos].p_depth = i;
>  		path[ppos].p_ext = NULL;

Re: [PATCH] ext4: fix bug in extents parsing when number of entries in header is zero

[Baokun Li]

在 2022/8/5 22:00, Luís Henriques 写道:
> When walking through an inode extents, the ext4_ext_binsearch_idx() function
> assumes that the extent header has been previously validated.  However,
> there are no checks that verify that the number of entries (eh->eh_entries)
> is non-zero.  And this will lead to problems because the EXT_FIRST_INDEX()
> and EXT_LAST_INDEX() will return garbage and result in this:
>
> [  135.245946] ------------[ cut here ]------------
> [  135.247579] kernel BUG at fs/ext4/extents.c:2258!
> [  135.249045] invalid opcode: 0000 [#1] PREEMPT SMP
> [  135.250320] CPU: 2 PID: 238 Comm: tmp118 Not tainted 5.19.0-rc8+ #4
> [  135.252067] Hardware name: QEMU Standard PC (i440FX + PIIX, 1996), BIOS rel-1.15.0-0-g2dd4b9b-rebuilt.opensuse.org 04/01/2014
> [  135.255065] RIP: 0010:ext4_ext_map_blocks+0xc20/0xcb0
> [  135.256475] Code:
> [  135.261433] RSP: 0018:ffffc900005939f8 EFLAGS: 00010246
> [  135.262847] RAX: 0000000000000024 RBX: ffffc90000593b70 RCX: 0000000000000023
> [  135.264765] RDX: ffff8880038e5f10 RSI: 0000000000000003 RDI: ffff8880046e922c
> [  135.266670] RBP: ffff8880046e9348 R08: 0000000000000001 R09: ffff888002ca580c
> [  135.268576] R10: 0000000000002602 R11: 0000000000000000 R12: 0000000000000024
> [  135.270477] R13: 0000000000000000 R14: 0000000000000024 R15: 0000000000000000
> [  135.272394] FS:  00007fdabdc56740(0000) GS:ffff88807dd00000(0000) knlGS:0000000000000000
> [  135.274510] CS:  0010 DS: 0000 ES: 0000 CR0: 0000000080050033
> [  135.276075] CR2: 00007ffc26bd4f00 CR3: 0000000006261004 CR4: 0000000000170ea0
> [  135.277952] Call Trace:
> [  135.278635]  <TASK>
> [  135.279247]  ? preempt_count_add+0x6d/0xa0
> [  135.280358]  ? percpu_counter_add_batch+0x55/0xb0
> [  135.281612]  ? _raw_read_unlock+0x18/0x30
> [  135.282704]  ext4_map_blocks+0x294/0x5a0
> [  135.283745]  ? xa_load+0x6f/0xa0
> [  135.284562]  ext4_mpage_readpages+0x3d6/0x770
> [  135.285646]  read_pages+0x67/0x1d0
> [  135.286492]  ? folio_add_lru+0x51/0x80
> [  135.287441]  page_cache_ra_unbounded+0x124/0x170
> [  135.288510]  filemap_get_pages+0x23d/0x5a0
> [  135.289457]  ? path_openat+0xa72/0xdd0
> [  135.290332]  filemap_read+0xbf/0x300
> [  135.291158]  ? _raw_spin_lock_irqsave+0x17/0x40
> [  135.292192]  new_sync_read+0x103/0x170
> [  135.293014]  vfs_read+0x15d/0x180
> [  135.293745]  ksys_read+0xa1/0xe0
> [  135.294461]  do_syscall_64+0x3c/0x80
> [  135.295284]  entry_SYSCALL_64_after_hwframe+0x46/0xb0
>
> Unfortunately, __ext4_ext_check() only verifies that eh->eh_entries doesn't
> exceed eh->eh_max.  And since an empty leaf seems to be a valid value in
> same cases, adding this extra check there isn't an option.
>
> This patch simply adds the check directly in ext4_ext_binsearch_idx() and
> propagates this error so that the kernel doesn't hit this BUG_ON() in
> ext4_ext_determine_hole().
>
> Link:https://bugzilla.kernel.org/show_bug.cgi?id=215941
> Signed-off-by: Luís Henriques<lhenriques@suse.de>
> ---
>   fs/ext4/extents.c | 13 ++++++++++---
>   1 file changed, 10 insertions(+), 3 deletions(-)
>
> Hi!
>
> This bug is easily reproducible using the filesystem image provided --
> it's just a matter of mounting it and run:
>
>      $ cat /mnt/foo/bar/xattr

Hi Luís,
yeah, that's a good catch!
> Anyway, I hope my analysis of the bug is correct -- the root cause seems
> to be an extent header with an invalid value for in eh_entries, which will
> later cause the BUG_ON().
>
> Cheers,
> --
> Luís
But there's a little bit of a deviation in your understanding of the 
problem,
so the patch doesn't look good.
The issue is caused by the contradiction between eh_entries and eh_depth.
Therefore, we need to check the contradiction instead of adding a 
judgment to ext4_ext_binsearch_idx.
So the right fix is to add a check to __ext4_ext_check like:

diff --git a/fs/ext4/extents.c b/fs/ext4/extents.c
index c148bb97b527..2dfd35f727cb 100644
--- a/fs/ext4/extents.c
+++ b/fs/ext4/extents.c
@@ -460,6 +460,10 @@ static int __ext4_ext_check(const char *function, 
unsigned int line,
                 error_msg = "invalid eh_entries";
                 goto corrupted;
         }
+       if (unlikely((eh->eh_entries == 0) && (depth > 0))) {
+               error_msg = "contradictory eh_entries and eh_depth";
+               goto corrupted;
+       }
         if (!ext4_valid_extent_entries(inode, eh, lblk, &pblk, depth)) {
                 error_msg = "invalid extent entries";
                 goto corrupted;

In this way, we can fix this issue and check for header exceptions 
before calling ext4_ext_binsearch_idx.

Thanks!
-- 
With Best Regards,
Baokun Li

Re: [PATCH] ext4: fix bug in extents parsing when number of entries in header is zero

[Luís Henriques]

On Fri, Aug 12, 2022 at 10:33:20AM +0800, Baokun Li wrote:
> 在 2022/8/5 22:00, Luís Henriques 写道:
...
> > This bug is easily reproducible using the filesystem image provided --
> > it's just a matter of mounting it and run:
> > 
> >      $ cat /mnt/foo/bar/xattr
> 
> Hi Luís,
> yeah, that's a good catch!
> > Anyway, I hope my analysis of the bug is correct -- the root cause seems
> > to be an extent header with an invalid value for in eh_entries, which will
> > later cause the BUG_ON().
> > 
> > Cheers,
> > --
> > Luís
> But there's a little bit of a deviation in your understanding of the
> problem,
> so the patch doesn't look good.
> The issue is caused by the contradiction between eh_entries and eh_depth.

Ah! This makes a lot of sense and I can confirm this is exactly what
happens in both bugzilla images.  Thanks a lot for your feedback!

> Therefore, we need to check the contradiction instead of adding a judgment
> to ext4_ext_binsearch_idx.
> So the right fix is to add a check to __ext4_ext_check like:
> 
> diff --git a/fs/ext4/extents.c b/fs/ext4/extents.c
> index c148bb97b527..2dfd35f727cb 100644
> --- a/fs/ext4/extents.c
> +++ b/fs/ext4/extents.c
> @@ -460,6 +460,10 @@ static int __ext4_ext_check(const char *function,
> unsigned int line,
>                 error_msg = "invalid eh_entries";
>                 goto corrupted;
>         }
> +       if (unlikely((eh->eh_entries == 0) && (depth > 0))) {
> +               error_msg = "contradictory eh_entries and eh_depth";
> +               goto corrupted;
> +       }
>         if (!ext4_valid_extent_entries(inode, eh, lblk, &pblk, depth)) {
>                 error_msg = "invalid extent entries";
>                 goto corrupted;
> 
> In this way, we can fix this issue and check for header exceptions before
> calling ext4_ext_binsearch_idx.

Awesome, I'll send out v2 with the suggested change.  It makes sense to
have this check and it should fix both bugs.

On the other hand, I still wonder wether the extra check in my original
patch is correct or not.  I spent a good amount of time trying to find out
if eh_entries can be 0 at that point (in ext4_ext_binsearch_idx()) and
couldn't find a situation where it could.  And running the fstests with
that check didn't show any problem.  But yeah, my understanding of the
whole code is far from perfect.

Cheers,
--
Luís