Skip to contentSkip to conversation

tools/kvm_stat: fix attack vector with user controlled FUSE mounts

PATCH In review Matthias Gerstner 2 participants 7 messages Nov 3, 2022 – Nov 7, 2022 KVM

lore
Share & tools
lore
#0 lore raw

PATCH 1/1 tools/kvm_stat: fix attack vector with user controlled FUSE mounts

The first field in /proc/mounts can be influenced by unprivileged users through the widespread fusermount setuid-root program. Example:

```
user$ mkdir ~/mydebugfs
user$ export _FUSE_COMMFD=0
user$ fusermount ~/mydebugfs -ononempty,fsname=debugfs
user$ grep debugfs /proc/mounts
debugfs /home/user/mydebugfs fuse rw,nosuid,nodev,relatime,user_id=1000,group_id=100 0 0
```

If there is no debugfs already mounted in the system then this can be used by unprivileged users to trick kvm_stat into using a user controlled file system location for obtaining KVM statistics.

To exploit this also a race condition has to be won, since the code checks for the existence of the 'kvm' subdirectory of the resulting path. This doesn't work on a FUSE mount, because the root user is not allowed to access non-root FUSE mounts for security reasons. If an attacker manages to unmount the FUSE mount in time again then kvm_stat would be using the resulting path, though.

The impact if winning the race condition is mostly a denial-of-service or damaged information integrity. The files in debugfs are only opened for reading. So the attacker can cause very large data to be read in by kvm_stat or fake data to be processed by kvm_stat. I don't see any viable way to turn this into a privilege escalation.

The fix is simply to use the file system type field instead. Whitespace in the mount path is escaped in /proc/mounts thus no further safety measures in the parsing should be necessary to make this correct.

---
 tools/kvm/kvm_stat/kvm_stat | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)
tools/kvm/kvm_stat/kvm_stat +1 -1
index 9c366b3a676d..88a73999aa58 100755
@@ -1756,7 +1756,7 @@ def assign_globals():
 
     debugfs = ''
     for line in open('/proc/mounts'):
-        if line.split(' ')[0] == 'debugfs':
+        if line.split(' ')[2] == 'debugfs':
             debugfs = line.split(' ')[1]
             break
     if debugfs == '':
Signature · 1 line
2.37.3
maintainer
#1 lore raw
Quoted 22 lines · Matthias Gerstner

The first field in /proc/mounts can be influenced by unprivileged users through the widespread fusermount setuid-root program. Example:

```
user$ mkdir ~/mydebugfs
user$ export _FUSE_COMMFD=0
user$ fusermount ~/mydebugfs -ononempty,fsname=debugfs
user$ grep debugfs /proc/mounts
debugfs /home/user/mydebugfs fuse rw,nosuid,nodev,relatime,user_id=1000,group_id=100 0 0
```

If there is no debugfs already mounted in the system then this can be used by unprivileged users to trick kvm_stat into using a user controlled file system location for obtaining KVM statistics.

To exploit this also a race condition has to be won, since the code checks for the existence of the 'kvm' subdirectory of the resulting path. This doesn't work on a FUSE mount, because the root user is not allowed to access non-root FUSE mounts for security reasons. If an attacker manages to unmount the FUSE mount in time again then kvm_stat would be using the resulting path, though.

Quoted 5 lines

The impact if winning the race condition is mostly a denial-of-service or damaged information integrity. The files in debugfs are only opened for reading. So the attacker can cause very large data to be read in by kvm_stat or fake data to be processed by kvm_stat. I don't see any viable way to turn this into a privilege escalation.

Ok, thanks for confirming. I will tweak the commit message.

Paolo

Quoted 20 lines

The fix is simply to use the file system type field instead. Whitespace in the mount path is escaped in /proc/mounts thus no further safety measures in the parsing should be necessary to make this correct.

---
  tools/kvm/kvm_stat/kvm_stat | 2 +-
  1 file changed, 1 insertion(+), 1 deletion(-)
tools/kvm/kvm_stat/kvm_stat +1 -1
index 9c366b3a676d..88a73999aa58 100755
@@ -1756,7 +1756,7 @@ def assign_globals():
 
      debugfs = ''
      for line in open('/proc/mounts'):
-        if line.split(' ')[0] == 'debugfs':
+        if line.split(' ')[2] == 'debugfs':
              debugfs = line.split(' ')[1]
              break
      if debugfs == '':
maintainer
#2 lore raw
In reply to Matthias Gerstner
Quoted 3 lines · Matthias Gerstner

The fix is simply to use the file system type field instead. Whitespace in the mount path is escaped in /proc/mounts thus no further safety measures in the parsing should be necessary to make this correct.

Can you please send a patch to replace seq_printf with seq_escape in afs_show_devname though? I couldn't find anything that prevents cell->name and volume->name from containing a space, so better safe than sorry.

Paolo

#3 lore raw
Quoted 9 lines · Paolo Bonzini
Quoted 3 lines · Matthias Gerstner

The fix is simply to use the file system type field instead. Whitespace in the mount path is escaped in /proc/mounts thus no further safety measures in the parsing should be necessary to make this correct.

Can you please send a patch to replace seq_printf with seq_escape in afs_show_devname though? I couldn't find anything that prevents cell->name and volume->name from containing a space, so better safe than sorry.

I only checked this during runtime using a tmpfs and assumed this would be true for all file systems.

Sure I can come up with a patch. Should I send a new single patch containing both changes, a new patch series with two patches or do I need to send the afs change to a different mailing list? Sorry - I'm new to kernel development.

Cheers

Matthias

maintainer
#4 lore raw
Quoted 5 lines · Matthias Gerstner

Sure I can come up with a patch. Should I send a new single patch containing both changes, a new patch series with two patches or do I need to send the afs change to a different mailing list? Sorry - I'm new to kernel development.

Yes, please send the AFS patch to [email protected]. In the meanwhile I will queue this patch as it is already an improvement.

Paolo

maintainer
#5 lore raw
In reply to Matthias Gerstner
Quoted 7 lines · Matthias Gerstner

The fix is simply to use the file system type field instead. Whitespace in the mount path is escaped in /proc/mounts thus no further safety measures in the parsing should be necessary to make this correct.

---
  tools/kvm/kvm_stat/kvm_stat | 2 +-
  1 file changed, 1 insertion(+), 1 deletion(-)

Matthias, both this patch and the one you sent to linux-afs need to include a "Signed-off-by" line, for example:

###
### Signed-off-by: Matthias Gerstner <[email protected]>
###

The meaning of this is visible at https://developercertificate.org/.

For this patch you can just reply to the message with the above line (without the "###" in front) and I'll accept it. However, for linux-afs I suggest that you just resend it. Just committing your patch with the "-s" command line argument will include the line for you.

Thanks,

Paolo

Quoted 13 lines
tools/kvm/kvm_stat/kvm_stat +1 -1
index 9c366b3a676d..88a73999aa58 100755
@@ -1756,7 +1756,7 @@ def assign_globals():
 
      debugfs = ''
      for line in open('/proc/mounts'):
-        if line.split(' ')[0] == 'debugfs':
+        if line.split(' ')[2] == 'debugfs':
              debugfs = line.split(' ')[1]
              break
      if debugfs == '':
#6 lore raw
Quoted 41 lines · Paolo Bonzini
Quoted 7 lines · Matthias Gerstner

The fix is simply to use the file system type field instead. Whitespace in the mount path is escaped in /proc/mounts thus no further safety measures in the parsing should be necessary to make this correct.

---
  tools/kvm/kvm_stat/kvm_stat | 2 +-
  1 file changed, 1 insertion(+), 1 deletion(-)

Matthias, both this patch and the one you sent to linux-afs need to include a "Signed-off-by" line, for example:

###
### Signed-off-by: Matthias Gerstner <[email protected]>
###

The meaning of this is visible at https://developercertificate.org/.

For this patch you can just reply to the message with the above line (without the "###" in front) and I'll accept it. However, for linux-afs I suggest that you just resend it. Just committing your patch with the "-s" command line argument will include the line for you.

Thanks,

Paolo

Quoted 13 lines
tools/kvm/kvm_stat/kvm_stat +1 -1
index 9c366b3a676d..88a73999aa58 100755
@@ -1756,7 +1756,7 @@ def assign_globals():
 
      debugfs = ''
      for line in open('/proc/mounts'):
-        if line.split(' ')[0] == 'debugfs':
+        if line.split(' ')[2] == 'debugfs':
              debugfs = line.split(' ')[1]
              break
      if debugfs == '':