1

I came across a question the other day on reddit: How can we use PAGE_GUARD-based memory breakpoints in GDB (not hardware breakpoints)?

Ollydbg, x64dbg and IDA PRO all support these types of breakpoints, but I couldnt find a way in GDB.

If this is not available, is it possible to set the PAGE_GUARD bit manually in GDB?

1 Answer 1

2

According to man mmap

  PROT_NONE  The memory cannot be accessed at all.
  PROT_READ  The memory can be read.
  PROT_WRITE The memory can be modified.
  PROT_EXEC  The memory can be executed.

PROT_NONE will act like a guard page by hitting a SIGSEGV when accessed.

The page with PROT_NONE looks like this in the map during runtime

    0x7ffff7ff7000     0x7ffff7ff8000 ---p     1000 0

gdb allows you to call arbitrary functions in the process space. A simple solution would be to run this under gdb

print mprotect($address,0x1000,0)

This would set PROT_NONE = 0 permissions on the page and it will act as a guard page. If after hitting SIGSEGV you want to remap the page as rw (PROT_READ|PROT_WRITE)

print mprotect($address,0x1000,3)

If you want to add an extra page mapped as guard page like the page heaps in windows, you can call mmap.

print /a mmap($address+0x1000,0x1000,0,0x22,-1,0)

Here 0x22 is MAP_PRIVATE|MAP_ANONYMOUS

3
  • Thanks! I am surprised though gdb doesnt offer it directly. Mar 31, 2019 at 8:02
  • Hm, another thing: is there really no way to get the current protection of the page? Otherwise it would be hard to undo the memory-breakpoint and expose me to anti-debugging actions. Mar 31, 2019 at 8:31
  • There's no way to read the protections from any api/syscall. Worst case you can parse /proc/<pid>/maps. Some gdb extensions like pwndbg give a vmmap command which does this.
    – sudhackar
    Mar 31, 2019 at 9:18

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge that you have read and understand our privacy policy and code of conduct.

Not the answer you're looking for? Browse other questions tagged or ask your own question.