[Bug server/30453] gdbserver cannot set breakpoints when /proc/pid/mem is not writable

["adrian.ratiu at collabora dot com" <sourceware-bugzilla@sourceware.org>]

https://sourceware.org/bugzilla/show_bug.cgi?id=30453

Adrian Ratiu <adrian.ratiu at collabora dot com> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
                 CC|                            |adrian.ratiu at collabora dot com

--- Comment #8 from Adrian Ratiu <adrian.ratiu at collabora dot com> ---
Hello, sorry for necro-bumping this, but I'm working on Chromebooks on behalf
of Collabora and encountered this problem - of not having a functional remote
debugger past GDB 9.2 (yes CrOS based systems are still using v9.2...).

I got GDB working again by creating a kernel patch as you suggested above, to
relax the /proc/pid/mem hardening restriction for ptrace-scoped processes, but
it got rejected by ChromiumOS upstream as too risky.

Please see the attached kernel patch and this reply [1] if you want more
details on that.

I also wrote a patch for gdbserver which adds back the the ptrace r/w as a
fallback which works nicely for my test cases. I've attached the patch and
tried to explain as best I can in the commit message.

Please consider the perspective of ChromeOS/ChromiumOS users, for which users
GDB has a regression because basic functionality used to work and is broken in
more recent releases.

Would the gdbserver patch be acceptable to fix this case? I can modify it if
necessary to make it acceptable, for example after writing it I noticed that
gdb/linux-nat.c contains proc_mem_file_is_writable () which does something
similar, maybe I could reuse that logic?

Thanks,
Adrian 

[1] https://issuetracker.corp.google.com/issues/268356054#comment61

-- 
You are receiving this mail because:
You are on the CC list for the bug.