public inbox for gdb@sourceware.org
 help / color / mirror / Atom feed
* How to get software recording to work?
@ 2022-01-21 15:15 Simon Sobisch
       [not found] ` <DM8PR11MB5749520182B9D686E8254730DE5E9@DM8PR11MB5749.namprd11.prod.outlook.com>
  0 siblings, 1 reply; 2+ messages in thread
From: Simon Sobisch @ 2022-01-21 15:15 UTC (permalink / raw)
  To: gdb

(gdb) record full
(gdb) continue
Continuing.
Process record does not support instruction 0xfae64 at address 
0x7ffff7df0f48.
Process record: failed to record execution log.

Program stopped.
0x00007ffff7df0f48 in _dl_runtime_resolve_xsave () from 
/lib64/ld-linux-x86-64.so.2

This is produced in an older environment with GDB 9.2; using a different 
environment with GDB 11.1:

Process record does not support instruction 0xc5 at address 0x7ffff6c79024.
Process record: failed to record execution log.

Program stopped.
0x00007ffff6c79024 in __memset_avx2_unaligned_erms () from 
/usr/lib64/libc.so.6


Can this recording be done reliably? Which places would need to be fixed 
/ adjusted to get this working?

Thanks for any pointers,
Simon


^ permalink raw reply	[flat|nested] 2+ messages in thread

* RE: How to get software recording to work?
       [not found]   ` <932be60f-b2d3-11fe-929c-4ad9d702be8c@gnu.org>
@ 2022-01-24  9:27     ` Metzger, Markus T
  0 siblings, 0 replies; 2+ messages in thread
From: Metzger, Markus T @ 2022-01-24  9:27 UTC (permalink / raw)
  To: Simon Sobisch; +Cc: gdb

Hello Simon,

>Thanks Markus for your answers (just in case: they weren't CC'ed to the
>mailing list, not sure if this was by intent).

It was not.  My email client sometimes removes mailing lists when I 'reply all' and I forgot to re-add it.  Thanks for pointing it out.


>> The issue is lack of AVX support in the record full instruction decoder.
>>
>>> Can this recording be done reliably? Which places would need to be fixed
>>> / adjusted to get this working?
>>
>> AFAIK record full also has issues with multi-threaded processes.  To fix the
>instruction decode issue, the best approach IMHO would be to extend the
>instruction decoder that is used for disassembly and use that for record full.
>
>Is this something Intel may have a look at?
>
>I guess Instruction/Processor specific instructions are possibly best
>implemented by the vendors, aren't them?

Intel is maintaining the binutils disassembler.  It is a bit unfortunate that record full uses a separate instruction decoder.

Given the issues with multi-threaded debuggee processes, I'm not sure how usable record full would be.  Or how much it would be used if we were to enable AVX.  I remember it adding significant recording overhead.

Regards,
Markus.
Intel Deutschland GmbH
Registered Address: Am Campeon 10, 85579 Neubiberg, Germany
Tel: +49 89 99 8853-0, www.intel.de <http://www.intel.de>
Managing Directors: Christin Eisenschmid, Sharon Heck, Tiffany Doon Silva  
Chairperson of the Supervisory Board: Nicole Lau
Registered Office: Munich
Commercial Register: Amtsgericht Muenchen HRB 186928

^ permalink raw reply	[flat|nested] 2+ messages in thread

end of thread, other threads:[~2022-01-24  9:27 UTC | newest]

Thread overview: 2+ messages (download: mbox.gz / follow: Atom feed)
-- links below jump to the message on this page --
2022-01-21 15:15 How to get software recording to work? Simon Sobisch
     [not found] ` <DM8PR11MB5749520182B9D686E8254730DE5E9@DM8PR11MB5749.namprd11.prod.outlook.com>
     [not found]   ` <932be60f-b2d3-11fe-929c-4ad9d702be8c@gnu.org>
2022-01-24  9:27     ` Metzger, Markus T

This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for read-only IMAP folder(s) and NNTP newsgroup(s).