Debugging Random Program Behavior

I was looking at a weird coredump the other day. From the core, the program was trying to write to virtual address 0x6 and crashed on memcpy.

There's a piece of code looks like

if (a == 1) {
} else {

And from the coredump, a is indeed 1. However the execution took the else branch and crashed on memcpy in do_bar.

  1. Disassembled the code in gdb using disassemble /s. It's so much better than just disassemble or even disassemble /m. With link time optimization, more functions are being inlined, which makes reading plain assembly harder. disassemble /s would annotate each block of instructions with reference back to the source file. It helps understand the assembly much easier.
  2. Read the assembly and find where the bogus address came from. It boiled down to a single instruction lea 0x2(%r14) %r15, where %r15 was supposed to be set to %r14+2 but it's set to 0x6 instead. gdb is able to provide register values for each frame by unwinding the stack. So it looks like some kind of CPU/firmware bug.
  3. It's running E5-2680 v4, at microcode version 0xb000014 according to /proc/cpuinfo.
  4. It's Broadwell according to
  5. Got BIOS version Version: F06_3B06 from dmidecode.
  6. Once you have enough keywords, Google is your friend. And... there you go! Looks like I should try update the firmware.
Show Comments

Get the latest posts delivered right to your inbox.