[CVE-2008-2363] Heap overflow in PartsBatch class via .nzb files

Description of problem:

Version-Release number of selected component (if applicable):
pan-0.132-2.fc8.x86_64

How reproducible:
Always

Steps to Reproduce:
  Start pan from the command line

Actual results:

pan: parts.cc:244: void pan::Parts::set_parts(const pan::PartBatch&): Assertion
`pch == part_mid_buf + part_mid_buf_len' failed.
Aborted

Expected results:

PAN starts

Additional info:

Created attachment 306290
Tasks.nzb causing corrupted double-linked list abort from glibc

Debugged the problem further. It happens when trying to load a huge tasks.nzb
(around 95000 lines). When the file is split into two parts by copying and
removing extra <file>...</file> sections, each part loads successfully.
This most likely means state corruption in nzb parser by certain input.

Attaching the bad nzb files that cause: assertion, segmentation fault, or abort
inside glibc (double-linked list corruption detected).

Created attachment 306292
Glibc-generated stack trace from abort

Created attachment 306293
Tasks.nzb causing pan to segfault

Created attachment 306294
Stack trace from segfault

Created attachment 306296
Tasks.nzb causing assertion in pan.

Created attachment 306297
Stack trace from assertion

Found the root cause of the problem.

PartsBatch class has inconsistent storage for Parts. For each new batch, it does
not clear() the 'parts' variable, but only resize()s it if the new batch is
bigger (pan/data/parts.cc, lines 307 and 316). If the new batch is smaller than
the previous, extra entries are untouched.

The problem happens when pan attempts to sort the full vector
(pan/tasks/nzb.cc:128), thus possibly mixing the parts from the current batch
with the parts from the previous batches.

The solution is to clear() the 'parts' vector on each init(), then push_back()
the new parts in add_part, so the vector size always corresponds to the real
number of found parts.

Created attachment 306880
Fix the assertion in PAN due to sorting of an array containing redundant elements

Added proper handling of PartBatch::parts vector.
Fixed allocation policy of Part::packed_mid element to re-allocate it every
time rather than trying to reuse a previous instance.

Signed-off by: <email address hidden>

Note that when compiled with -O0 for ease of debugging, abort or segfault never
happen, assertion always happens instead.

I'm assigning this issue CVE-2008-2363. From the look of this, it looks like a
pretty generic heap overflow, which means we can't rule out arbitrary code
execution.

Added dependency on bug 433970.
Building pan rpm with this assertion fix for F9 is impossible without
compilation fixes here: https://bugzilla.redhat.com/attachment.cgi?id=306299

Moving to Security Response product, creating tracking bugs.

Current Pan in Fedora is 0.135+ and this was fixed in 0.133 (http://pan.rebelbase.com/download/releases/0.133/).