This is the mail archive of the
gdb-patches@sourceware.org
mailing list for the GDB project.
Re: [PATCH] Fix buffer overflow in ada-lang.c:move_bits
- From: Pedro Alves <palves at redhat dot com>
- To: Joel Brobecker <brobecker at adacore dot com>, Tom Tromey <tom at tromey dot com>
- Cc: gdb-patches at sourceware dot org
- Date: Thu, 8 Nov 2018 19:11:44 +0000
- Subject: Re: [PATCH] Fix buffer overflow in ada-lang.c:move_bits
- References: <20181024162037.21024-1-tom@tromey.com> <20181101153517.GA2705@adacore.com>
On 11/01/2018 03:35 PM, Joel Brobecker wrote:
> Hi Tom,
>
>> -fsanitize=address showed that ada-lang.c:move_bits can run off the
>> end of the source buffer. I believe this patch fixes the problem, by
>> arranging not to read from the source buffer once there are sufficient
>> bits in the accumulator.
>>
>> gdb/ChangeLog
>> 2018-10-23 Tom Tromey <tom@tromey.com>
>>
>> * ada-lang.c (move_bits): Don't run off the end of the source
>> buffer.
>
> Thanks for the patch!
>
> This is a part of the code that always forces me to think twice
> (or ten times), each time I try to touch it. I should really start
> adding comments to this code that detail what we are trying to do
> as we do it.
>
> I tested your change through our testsuite on the various baremetal
> targets we have, and noticed that it causes regressions on ppc and arm
> targets. It's hopefully something small, but just being back from
> a holiday, I'm a bit tied up at work; I'll put that issue on my TODO
> list to look at further.
I was going to suggest that this would benefit from unit tests in
the style of dwarf2read.c:copy_bitwise's, but, actually, isn't this
exactly the same as copy_bitwise? Can we get rid of ada-lang.c:move_bits?
(And maybe move copy_bitwise elsewhere?)
Thanks,
Pedro Alves