[Bug gdb/10743] New: Process record doesn't support instruction 0xf6e

[Bug gdb/10743] New: Process record doesn't support instruction 0xf6e

[jan dot kratochvil at redhat dot com]

Process record doesn't support instruction 0xf6e at address 0x7ffff6c42cf1.
Process record: failed to record execution log.

Program received signal SIGTRAP, Trace/breakpoint trap.
strrchr () at ../sysdeps/x86_64/strrchr.S:26
26		movd	%esi, %xmm1
Current language:  auto
The current source language is "auto; currently asm".
(gdb) bt
#0  strrchr () at ../sysdeps/x86_64/strrchr.S:26
#1  0x000000000044ba4a in main (argc=2, argv=0x7fffffffd538) at gdb.c:38
(gdb) q

Fedora 12 x86_64
gdb-7.0
glibc-2.10.90-24.x86_64

-- 
           Summary: Process record doesn't support instruction 0xf6e
           Product: gdb
           Version: unknown
            Status: UNCONFIRMED
          Severity: normal
          Priority: P2
         Component: gdb
        AssignedTo: unassigned at sourceware dot org
        ReportedBy: jan dot kratochvil at redhat dot com
                CC: gdb-prs at sourceware dot org
  GCC host triplet: x86_64-redhat-linux-gnu
GCC target triplet: x86_64-redhat-linux-gnu


http://sourceware.org/bugzilla/show_bug.cgi?id=10743

------- You are receiving this mail because: -------
You are on the CC list for the bug, or are watching someone who is.

[Bug gdb/10743] Process record doesn't support instruction 0xf6e

[jan dot kratochvil at redhat dot com]

------- Additional Comments From jan dot kratochvil at redhat dot com  2009-10-07 15:12 -------
Checked it affects also Fedora 11 - expecting it is since glibc-2.10 (May 2009).

When the record mode singlesteps it may simulate an unknown CPU to not to use
the optimized functions according to the glibc __init_cpu_features() detection.
Still the full CPU support sure would be better, this is now a bit blocker.


-- 


http://sourceware.org/bugzilla/show_bug.cgi?id=10743

------- You are receiving this mail because: -------
You are on the CC list for the bug, or are watching someone who is.

[Bug gdb/10743] Process record doesn't support instruction 0xf6e

[teawater at gmail dot com]

------- Additional Comments From teawater at gmail dot com  2009-10-07 15:51 -------
Thanks Jan, this insn is still not support by prec.

-- 


http://sourceware.org/bugzilla/show_bug.cgi?id=10743

------- You are receiving this mail because: -------
You are on the CC list for the bug, or are watching someone who is.

[Bug gdb/10743] Process record doesn't support instruction 0xf6e

[pierre dot poissinger at gmail dot com]

------- Additional Comments From pierre dot poissinger at gmail dot com  2009-10-11 10:50 -------
I have exactly the same problem on ubuntu 9.04 x86_64 with glibc 2.9-4ubuntu6.1
but with memset.

To overcome this, I do the following in gdb before starting target record, to
work around:

set __x86_64_preferred_memory_instruction=0

Following memset.S (and cacheinfo.c), this will force memset to use 'regular
instruction' instead of SSE - and it allows me to reverse debug :-)

Not sure this is honored in glibc 2.10 for strrchr but you could give a shoot.

-- 
           What    |Removed                     |Added
----------------------------------------------------------------------------
                 CC|                            |pierre dot poissinger at
                   |                            |gmail dot com


http://sourceware.org/bugzilla/show_bug.cgi?id=10743

------- You are receiving this mail because: -------
You are on the CC list for the bug, or are watching someone who is.

[Bug gdb/10743] Process record doesn't support instruction 0xf6e

[teawater at gmail dot com]

------- Additional Comments From teawater at gmail dot com  2009-10-12 07:17 -------
Thanks  Pierre, this idea is very cool.

-- 


http://sourceware.org/bugzilla/show_bug.cgi?id=10743

------- You are receiving this mail because: -------
You are on the CC list for the bug, or are watching someone who is.

[Bug gdb/10743] Process record doesn't support instruction 0xf6e

[jan dot kratochvil at redhat dot com]

------- Additional Comments From jan dot kratochvil at redhat dot com  2009-10-12 08:19 -------
It does not work, __x86_64_preferred_memory_instruction is not used in strrchr,
please try:

echo -e "#include <string.h>\nint main (int argc, char **argv) { return strrchr
(argv[0], '/') - argv[0]; }" | gcc -o 1 -Wall -g -x c -; ./gdb -nx -ex start -ex
'set __x86_64_preferred_memory_instruction=0' -ex record -ex c ./1
GNU gdb (GDB) 7.0.50.20091012-cvs
Copyright (C) 2009 Free Software Foundation, Inc.
License GPLv3+: GNU GPL version 3 or later <http://gnu.org/licenses/gpl.html>
This is free software: you are free to change and redistribute it.
There is NO WARRANTY, to the extent permitted by law.  Type "show copying"
and "show warranty" for details.
This GDB was configured as "x86_64-unknown-linux-gnu".
For bug reporting instructions, please see:
<http://www.gnu.org/software/gdb/bugs/>...
Reading symbols from /home/jkratoch/redhat/gdb-clean/1...done.
Temporary breakpoint 1 at 0x4004d3: file <stdin>, line 2.
Starting program: /home/jkratoch/redhat/gdb-clean/1 

Temporary breakpoint 1, main (argc=1, argv=0x7fffffffd4c8) at <stdin>:2
2	<stdin>: No such file or directory.
	in <stdin>
Continuing.
Process record doesn't support instruction 0xf6e at address 0x36b5a7f961.
Process record: failed to record execution log.

Program received signal SIGTRAP, Trace/breakpoint trap.
strrchr () at ../sysdeps/x86_64/strrchr.S:26
26		movd	%esi, %xmm1
Current language:  auto
The current source language is "auto; currently asm".
(gdb) p __x86_64_preferred_memory_instruction
$1 = 0
(gdb) up
#1  0x00000000004004e7 in main (argc=1, argv=0x7fffffffd4c8) at <stdin>:2
2	<stdin>: No such file or directory.
	in <stdin>
(gdb) x/i $pc-5
0x4004e2 <main+30>:	callq  0x4003c8 <strrchr@plt>
(gdb) x/i 0x4003c8
0x4003c8 <strrchr@plt>:	jmpq   *0x2004b2(%rip)        # 0x600880
<_GLOBAL_OFFSET_TABLE_+32>
(gdb) x/gx 0x600880
0x600880 <_GLOBAL_OFFSET_TABLE_+32>:	0x00000036b5a7f960
(gdb) x/i 0x00000036b5a7f960
0x36b5a7f960 <strrchr>:	movd   %esi,%xmm1

__x86_64_preferred_memory_instruction is just not in use there, neither if you
set it to any value or if you watch it.


-- 


http://sourceware.org/bugzilla/show_bug.cgi?id=10743

------- You are receiving this mail because: -------
You are on the CC list for the bug, or are watching someone who is.

[Bug gdb/10743] Process record doesn't support instruction 0xf6e

[pierre dot poissinger at gmail dot com]

------- Additional Comments From pierre dot poissinger at gmail dot com  2009-10-12 09:56 -------
Too bad.
I checked the glibc 2.10 src and the new string functions use SSE no matter what
:-( - so no 'easy' workaround.

So, unless you hack the glibc to support this workaround, I guess you could
still create your own lib with strlen, strchr, strchrnul, strrchr, memchr, and
rawmemchr and use preload to force ld to link with them instead of glibc ones.


-- 


http://sourceware.org/bugzilla/show_bug.cgi?id=10743

------- You are receiving this mail because: -------
You are on the CC list for the bug, or are watching someone who is.

[Bug gdb/10743] Process record doesn't support instruction 0xf6e

[jan dot kratochvil at redhat dot com]

-- 
           What    |Removed                     |Added
----------------------------------------------------------------------------
  BugsThisDependsOn|                            |10760


http://sourceware.org/bugzilla/show_bug.cgi?id=10743

------- You are receiving this mail because: -------
You are on the CC list for the bug, or are watching someone who is.

[Bug gdb/10743] Process record doesn't support instruction 0xf6e

[jan dot kratochvil at redhat dot com]

------- Additional Comments From jan dot kratochvil at redhat dot com  2010-08-01 09:07 -------
At least the Comment 5 reproducer seems to be fixed on:
GNU gdb (GDB) 7.2.50.20100731-cvs


-- 
           What    |Removed                     |Added
----------------------------------------------------------------------------
             Status|UNCONFIRMED                 |RESOLVED
         Resolution|                            |FIXED


http://sourceware.org/bugzilla/show_bug.cgi?id=10743

------- You are receiving this mail because: -------
You are on the CC list for the bug, or are watching someone who is.