[Ltrace-devel] [PATCH v3] Tracing PLT-less MIPS binaries

Faraz Shahbazker faraz.shahbazker at imgtec.com
Wed Feb 11 00:52:55 UTC 2015


On 02/10/2015 03:56 PM, Petr Machata wrote:
> Eugene Rudoy <gene at freetz.org> writes:
>> # test command: ./ltrace /usr/bin/openssl speed 2>&1 | tee openssl_speed.log
>> # result: a lot of "unexpected breakpoint at"-lines
>> #           output doesn't contain any call to libssl-functions
> 
> This could be anything.  Somehow the internal logic is thrown off and
> sees events that it can't make sense of.  After it rejects all other
> cases, it assumes it's a breakpoint, but doesn't see one with a given
> address in its breakpoint tables.  It could be a misinterpreted
> singlestep, unbiased or double-biased calls, or something else still.
> Faraz will have to figure this out.
As far as I can make out, the unexpected breakpoint is related to signal handling; will dig further. 

Compiler built with-mips-plt is not a direct indicator of whether binaries will have PLT. On my test system, wget has PLT while openssl does not. With this patch, openssl does print function traces in the output, in spite of the many "unexpected breakpoint" errors.

Regards,
Faraz Shahbazker



More information about the Ltrace-devel mailing list