pthread segfault

pthread segfault

[Scott R. Parish]

I compiled 2.0.1 from source.

dom0 is running debian--working fine

dom1 is running arch--boots fine but applications linked to pthread
segfault when pthread goes to access gs. I would think that this is the
tls problem, but i renamed tls.

Any ideas?

Thanks
sRp

-- 
Scott R. Parish



-------------------------------------------------------
SF email is sponsored by - The IT Product Guide
Read honest & candid reviews on hundreds of IT Products from real users.
Discover which products truly live up to the hype. Start reading now. 
http://productguide.itmanagersjournal.com/

Re: pthread segfault

[Ian Pratt]

> I compiled 2.0.1 from source.
> 
> dom0 is running debian--working fine
> 
> dom1 is running arch--boots fine but applications linked to pthread
> segfault when pthread goes to access gs. I would think that this is the
> tls problem, but i renamed tls.

This is pretty surprising. 

The -ve segment emulation stuff should be able to handle all the
different forms of tcb and tls access we know about.

What happens if you don't move tls out the way? It'll be slower,
but should work. (is it falling back to 686? -- "ldd /bin/sleep"
will tell you).

What glibc version is it using?

Can you get a debugger to look at the operands to the faulting
instuction.

Thanks,
Ian


-------------------------------------------------------
SF email is sponsored by - The IT Product Guide
Read honest & candid reviews on hundreds of IT Products from real users.
Discover which products truly live up to the hype. Start reading now. 
http://productguide.itmanagersjournal.com/

Re: pthread segfault

[Scott R. Parish]

On Wed, Nov 24, 2004 at 02:11:49AM +0000, Ian Pratt wrote:

> > dom1 is running arch--boots fine but applications linked to pthread
> > segfault when pthread goes to access gs. I would think that this is the
> > tls problem, but i renamed tls.
> 
> What happens if you don't move tls out the way? It'll be slower,
> but should work. (is it falling back to 686? -- "ldd /bin/sleep"
> will tell you).

ah, that works!

> What glibc version is it using?

2.3.3

> Can you get a debugger to look at the operands to the faulting
> instuction.

# gdb ls
GNU gdb 6.2.1
Copyright 2004 Free Software Foundation, Inc.
GDB is free software, covered by the GNU General Public License, and you are
welcome to change it and/or distribute copies of it under certain conditions.
Type "show copying" to see the conditions.
There is absolutely no warranty for GDB.  Type "show warranty" for details.
This GDB was configured as "i686-pc-linux-gnu"...(no debugging symbols found)...Using host libthread_db library "/lib/libthread_db.so.1".

(gdb) run
Starting program: /bin/ls 
(no debugging symbols found)...(no debugging symbols found)...(no debugging symbols found)...[Thread debugging using libthread_db enabled]
[New Thread 16384 (LWP 260)]
(no debugging symbols found)...
Program received signal SIGSEGV, Segmentation fault.
[Switching to Thread 16384 (LWP 260)]
0xb7e56d43 in _pthread_cleanup_push_defer () from /lib/libpthread.so.0
(gdb) bt
#0  0xb7e56d43 in _pthread_cleanup_push_defer () from /lib/libpthread.so.0
#1  0xb7f0d35b in _IO_link_in_internal () from /lib/libc.so.6
#2  0xb7f0abd6 in _IO_new_file_init () from /lib/libc.so.6
#3  0xb7f010ec in __fopen_internal () from /lib/libc.so.6
#4  0xb7f0115d in fopen@@GLIBC_2.1 () from /lib/libc.so.6
#5  0xb7f7303b in setmntent () from /lib/libc.so.6
#6  0xb7f77e29 in get_proc_path () from /lib/libc.so.6
#7  0xb7f7819e in phys_pages_info () from /lib/libc.so.6
#8  0xb7f4a394 in posix_sysconf () from /lib/libc.so.6
#9  0xb7f4a718 in sysconf () from /lib/libc.so.6
#10 0xb7ecd7ec in qsort () from /lib/libc.so.6
#11 0x0804a7bc in ?? ()
...
(gdb) disassemble 0xb7e56d43
Dump of assembler code for function _pthread_cleanup_push_defer:
0xb7e56d30 <_pthread_cleanup_push_defer+0>:     push   %ebp
0xb7e56d31 <_pthread_cleanup_push_defer+1>:     mov    %esp,%ebp
0xb7e56d33 <_pthread_cleanup_push_defer+3>:     mov    0x8(%ebp),%edx
0xb7e56d36 <_pthread_cleanup_push_defer+6>:     mov    0xc(%ebp),%eax
0xb7e56d39 <_pthread_cleanup_push_defer+9>:     mov    %eax,(%edx)
0xb7e56d3b <_pthread_cleanup_push_defer+11>:    mov    0x10(%ebp),%eax
0xb7e56d3e <_pthread_cleanup_push_defer+14>:    mov    %eax,0x4(%edx)
0xb7e56d41 <_pthread_cleanup_push_defer+17>:    xor    %eax,%eax
0xb7e56d43 <_pthread_cleanup_push_defer+19>:    mov    %gs:0x81,%al
0xb7e56d49 <_pthread_cleanup_push_defer+25>:    movsbl %al,%eax
0xb7e56d4c <_pthread_cleanup_push_defer+28>:    mov    %eax,0x8(%edx)
0xb7e56d4f <_pthread_cleanup_push_defer+31>:    mov    %gs:0x7c,%eax
...
(gdb) info reg
eax            0x0      0
ecx            0x0      0
edx            0xbfffd648       -1073752504
ebx            0xb7fd1044       -1208151996
esp            0xbfffd634       0xbfffd634
ebp            0xbfffd634       0xbfffd634
esi            0x805e720        134604576
edi            0x1      1
eip            0xb7e56d43       0xb7e56d43
eflags         0x10246  66118
cs             0x73     115
ss             0x7b     123
ds             0x7b     123
es             0x7b     123
fs             0x0      0
gs             0x33     51


-- 
Scott R. Parish



-------------------------------------------------------
SF email is sponsored by - The IT Product Guide
Read honest & candid reviews on hundreds of IT Products from real users.
Discover which products truly live up to the hype. Start reading now. 
http://productguide.itmanagersjournal.com/

Re: pthread segfault

[Scott R. Parish]

On Wed, Nov 24, 2004 at 02:11:49AM +0000, Ian Pratt wrote:

> (is it falling back to 686? -- "ldd /bin/sleep" will tell you).

not quite sure what i was suppose to look for, but..:

# ldd /usr/bin/sleep
        linux-gate.so.1 =>  (0xfbffd000)
        libm.so.6 => /lib/libm.so.6 (0xb7fc4000)
        librt.so.1 => /lib/librt.so.1 (0xb7fb1000)
        libc.so.6 => /lib/libc.so.6 (0xb7e80000)
        /lib/ld-linux.so.2 => /lib/ld-linux.so.2 (0xb7fea000)
        libpthread.so.0 => /lib/libpthread.so.0 (0xb7e2f000)


-- 
Scott R. Parish



-------------------------------------------------------
SF email is sponsored by - The IT Product Guide
Read honest & candid reviews on hundreds of IT Products from real users.
Discover which products truly live up to the hype. Start reading now. 
http://productguide.itmanagersjournal.com/

Re: pthread segfault

[Keir Fraser]

Sounds like LDT support has got broken. Are you using Linux 2.4 or
2.6? You might also want to try the xen-2.0-testing.bk tree.

 -- Keir

> I compiled 2.0.1 from source.
> 
> dom0 is running debian--working fine
> 
> dom1 is running arch--boots fine but applications linked to pthread
> segfault when pthread goes to access gs. I would think that this is the
> tls problem, but i renamed tls.
> 
> Any ideas?
> 
> Thanks
> sRp
> 
> -- 
> Scott R. Parish
> 
> 
> 
> -------------------------------------------------------
> SF email is sponsored by - The IT Product Guide
> Read honest & candid reviews on hundreds of IT Products from real users.
> Discover which products truly live up to the hype. Start reading now. 
> http://productguide.itmanagersjournal.com/
> _______________________________________________
> Xen-devel mailing list
> Xen-devel@lists.sourceforge.net
> https://lists.sourceforge.net/lists/listinfo/xen-devel



-------------------------------------------------------
SF email is sponsored by - The IT Product Guide
Read honest & candid reviews on hundreds of IT Products from real users.
Discover which products truly live up to the hype. Start reading now. 
http://productguide.itmanagersjournal.com/

Re: pthread segfault

[Scott R. Parish]

On Wed, Nov 24, 2004 at 08:50:25AM +0000, Keir Fraser wrote:

> Sounds like LDT support has got broken. Are you using Linux 2.4 or
> 2.6?

I ran "make world" and "make install" in the source directory with no
changes; all domains are 2.6.9.

> You might also want to try the xen-2.0-testing.bk tree.

I don't have bitkeeper access (have developed on gnu arch), but i
grabbed xen-2.0-testing-src.tgz and same story with it as with 2.0.1.

sRp

-- 
Scott R. Parish



-------------------------------------------------------
SF email is sponsored by - The IT Product Guide
Read honest & candid reviews on hundreds of IT Products from real users.
Discover which products truly live up to the hype. Start reading now. 
http://productguide.itmanagersjournal.com/