Bug 889878 (CVE-2022-3715) – app-shells/bash: heap buffer overflow yielding "memory problems"

• Product: Gentoo Security
• Status: CONFIRMED ---

[Description John Helmert III 2023-01-05 18:55:43 UTC]

CVE-2022-3715:

A flaw was found in the bash package, where a heap-buffer overflow can occur in valid parameter_transform. This issue may lead to memory problems.

No idea about a patch. Reported upstream at: https://lists.gnu.org/archive/html/bug-bash/2022-08/msg00147.html

[Comment 1 Allen Webb 2023-03-07 14:59:02 UTC]

Red hat is tracking this here:
https://bugzilla.redhat.com/show_bug.cgi?id=2126720

And links to the mailing list thread with the fix:
https://lists.gnu.org/archive/html/bug-bash/2022-08/msg00147.html

```
   Repeat-By:
   1. Build bash with asan 2. Run with AFL++ crafted input (in attachment)
   Fix:
diff -uprN bash-5.1-orig/subst.c bash-5.1/subst.c
--- bash-5.1-orig/subst.c       2022-08-29 18:56:11.071802865 +0000
+++ bash-5.1/subst.c    2022-08-29 18:56:48.187435415 +0000
@@ -7916,7 +7916,7 @@ static int
 valid_parameter_transform (xform)
      char *xform;
 {
-  if (xform[1])
+  if (xform[0] && xform[1])
     return 0;

   /* check for valid values of xform[0] */

   Kind regards, Ivan Kapranov.
```

[Comment 2 John Helmert III 2023-03-11 04:37:32 UTC]

Thanks! Any idea about impact?

[Comment 3 Allen Webb 2023-03-16 18:22:41 UTC]

It is hard to say without seeing what the repro case looks like. Since it is in a shell, it probably already requires significant access in order to reach the bug. In other words an attacker who can already run shell doesn't have much to gain by exploiting a memory bug in the shell interpreter.

Some notable exceptions are
* the potential for a noexec bypasses
* the potential a privileged process already uses shell in a way that could be tricked into matching a repro case

I would consider this worth fixing, but not urgent. It is surprising to me that upstream didn't take the fix (yet).

[Comment 4 Mike Gilbert 2023-03-16 20:52:16 UTC]

After reading the message thread, it sounds like the reporter has failed to sufficiently demonstrate that a problem even exists. I suspect they just submitted some fuzzer output without reviewing it for context.

[Comment 5 Allen Webb 2023-03-20 20:21:36 UTC]

I took a look to see if I could find a way to reach it and the only way seems to be here:
https://github.com/bminor/bash/blob/ec8113b9861375e4e17b3307372569d429dec814/subst.c#L8679


```
if (xform[0] == 0 || valid_parameter_transform (xform) == 0)
```

So the check in the patch is already performed before the only call site for valid_parameter_transform. It could easily become an issue in the future, but I think it is safe to close this.

[Comment 6 Mike Gilbert 2023-03-20 21:28:29 UTC]

The `xform[0] == 0 ||` check was added in bash-5.2, which was released a few weeks after the issue was reported on the bash mailing list.

[Comment 7 Allen Webb 2023-03-20 21:30:44 UTC]

In that case we should either pick that commit or upgrade past the fix.