Hello there, I have been fuzzing objdump with American Fuzzy Lop and AddressSanitizer. Please find attached the minimized file causing the issue ("Input") and the ASAN report log ("Output"). Below is the reduced stacktrace with links to the corresponding source lines on a GitHub mirror. The command I used was `objdump -D <file>`. Let me know if there is any additional information I can provide. -- Input: ef51bcdcaae667058b002f94b5dafd05.12926af7cc4fab77f87a3ec70a329100.min Output: ef51bcdcaae667058b002f94b5dafd05.12926af7cc4fab77f87a3ec70a329100.txt Error in "ieee_object_p": stack-buffer-overflow in ieee_object_p at bfd/ieee.c:1985 (see https://github.com/bminor/binutils-gdb/blob/561bf3e950e410fbcac06523d43039f1f58150ca/bfd/ieee.c#L1985) in bfd_check_format_matches at bfd/format.c:311 (see https://github.com/bminor/binutils-gdb/blob/561bf3e950e410fbcac06523d43039f1f58150ca/bfd/format.c#L311) in display_object_bfd at binutils/objdump.c:3602 (see https://github.com/bminor/binutils-gdb/blob/561bf3e950e410fbcac06523d43039f1f58150ca/binutils/objdump.c#L3602) in display_any_bfd at binutils/objdump.c:3693 (see https://github.com/bminor/binutils-gdb/blob/561bf3e950e410fbcac06523d43039f1f58150ca/binutils/objdump.c#L3693) in display_file at binutils/objdump.c:3714 (see https://github.com/bminor/binutils-gdb/blob/561bf3e950e410fbcac06523d43039f1f58150ca/binutils/objdump.c#L3714) in main at binutils/objdump.c:4016 (see https://github.com/bminor/binutils-gdb/blob/561bf3e950e410fbcac06523d43039f1f58150ca/binutils/objdump.c#L4016)
Created attachment 10114 [details] testcase
Created attachment 10115 [details] report
Additional Information: The command used was `objdump -D <file>`. The compilation flags used were `-g -O2 -fno-omit-frame-pointer -fsanitize=address -fno-sanitize-recover=undefined`. The configuration settings used were `--enable-targets=all --disable-shared`.
Created attachment 10139 [details] Proposed patch Hi Aadamski, This is a difficult one. Although I can reproduce the problem, it goes away whenever I make changes to the sources, including just adding comments! I suspect a compiler bug, but I cannot prove it at the moment. Are you able to test out patches and see if they make a difference ? If so please could you try out the uploaded patch which might make a difference. For me, the bug goes away, but I cannot tell if I have really fixed the problem, or if there is still some kind of stack corruption going on. Cheers Nick
(In reply to Nick Clifton from comment #4) > Are you able to test out patches and see if they make a difference ? If so > please could you try out the uploaded patch which might make a difference. > For me, the bug goes away, but I cannot tell if I have really fixed the > problem, or if there is still some kind of stack corruption going on. I tried your patch with both the raw and the minimized testcases; it seems not to crash anymore. I have started an instance of AFL in "crash exploration" mode in order to generate more diverse testcases. I will report back in a few hours. Thanks a lot for all your bug fixes, you're doing god's work! :-)
Sounds like it's fixed for me!
The master branch has been updated by Nick Clifton <nickc@sourceware.org>: https://sourceware.org/git/gitweb.cgi?p=binutils-gdb.git;h=63634bb4a107877dd08b6282e28e11cfd1a1649e commit 63634bb4a107877dd08b6282e28e11cfd1a1649e Author: Nick Clifton <nickc@redhat.com> Date: Thu Jun 15 12:44:23 2017 +0100 Avoid a possible compiler bug by using a static buffer instead of a stack local buffer. PR binutils/21582 * ieee.c (ieee_object_p): Use a static buffer to avoid compiler bugs.
Hi Aadamski, Thanks for the confirmation - patch applied. Cheers Nick