pmcd (and probably others) in pcp-3.6.3 crash when decoding crafted PDUs where the numcreds field exceeds the number of elements actually contained in the PDU.
On 32-bit architectures, the size passed to malloc can be too small, leading to a heap-based buffer overflow. On 64-bit architectures, the multiplication is performed with 64 bits, so no overflow occurs, and the crash happens because __pmDecodeCreds reads beyond the end of an allocated buffer.
pmcd runs with root privileges and starts automatically when the pcp package is installed. It listens on 0.0.0.0 and does not require authentication.
Created attachment 598767 [details]
preliminary patch to __pmDecodeCreds to fix this
Attached preliminary patch from Nathan Scott (upstream). Has been tested
against the reproducer script.
(In reply to comment #8)
> Created attachment 598767 [details]
> preliminary patch to __pmDecodeCreds to fix this
> Attached preliminary patch from Nathan Scott (upstream). Has been tested
> against the reproducer script.
This calculation can overflow:
need = sizeof(creds_t) + ((numcred-1) * sizeof(__pmCred));
There should be a check *before* the calculation that this cannot happen, something like numcred < (INT_MAX - sizeof(creds_t))/sizeof(__pmCred).
Created attachment 599108 [details]
Updated patch to address pcp credentials pdu buffer overflow exploit
Incorporates an additional check to guard against malloc size overflow. Uses a tighter restriction than the MAXINT-based suggestion, based on current uses and likely future uses of this PDU.
Ressign to Nathan since he is following thru and this seems to be the way
to go for these bugs.
I have a pcpqa test (513) which uses Florian's initial test case to exercise this change. I'll hang onto it privately until we get a/ confirmation on permission to make use of this, and b/ public knowledge of the issue.
(In reply to comment #10)
> Created attachment 599108 [details]
> Updated patch to address pcp credentials pdu buffer overflow exploit
> Incorporates an additional check to guard against malloc size overflow.
> Uses a tighter restriction than the MAXINT-based suggestion, based on
> current uses and likely future uses of this PDU.
This looks good, thanks.
This issue has been addressed in pcp-3.6.5
This issue was addressed in Fedora and EPEL via the following security updates: