Re: 5.0 linux emul problem?

[Jukka Marin <jmarin%embedtronics.fi@localhost>]

On Tue, Feb 22, 2011 at 11:16:22PM +0000, Christos Zoulas wrote:

> > 21345      1 mgcld    NAMI  "/emul/linux/usr/tmp/.flexlm"
> > 21345      1 mgcld    NAMI  "/usr/tmp/.flexlm"
> > 21345      1 mgcld    RET   lstat64 0
> > 21345      1 mgcld    CALL  open(0x8191720,2,0)
> > 21345      1 mgcld    NAMI  "/emul/linux/var/tmp/lockmgcld"
> > 21345      1 mgcld    NAMI  "/var/tmp/lockmgcld"
> > 21345      1 mgcld    RET   open 4
> > 21345      1 mgcld    CALL  fcntl64(4,6,0xbb3f968c)
> > 21345      1 mgcld    RET   fcntl64 -1 unknown errno 35
> > 21345      1 mgcld    CALL  time(0)
> > 21345      1 mgcld    RET   time 1298384794/0x4d63c79a
> > 21345      1 mgcld    CALL  write(1,0xbb66e000,0x30)
> > 21345      1 mgcld    GIO   fd 1 wrote 48 bytes
> >       "16:26:34 (mgcld) lost lock, exiting (errno: 11)\n"
> > 21345      1 mgcld    RET   write 48/0x30
> > 21345      1 mgcld    CALL  time(0)
> > 21345      1 mgcld    RET   time 1298384794/0x4d63c79a
> > 21345      1 mgcld    CALL  time(0)
> > 21345      1 mgcld    RET   time 1298384794/0x4d63c79a
> > 21345      1 mgcld    CALL  write(1,0xbb66e000,0x38)
> > 21345      1 mgcld    GIO   fd 1 wrote 56 bytes
> >       "16:26:34 (mgcld) EXITING DUE TO SIGNAL 41 Exit reason 9\n"
> 
> Sorry for asking the obvious, but when you remove it, does the process
> re-create it? You could add some debugging, but it seems to me that as
> you said the SETLK call thinks it is already locked.

Yes, I have tried removing all files and the process creates it again.

Hmm.. maybe I'm misreading the dump but it seems that mgcld first creates
(opens) the file, calls fcntl64(4,6,xxxx), then forks, and the new process
also opens and attempts to fcntl64(4,6,xxxx) the file (the first process
never closes file 4, afaict).  But if this was so, how could the binary
work on linux, either?

I'll put the complete trace log at http://www.jmp.fi/~jmarin/mgcld.log if
someone whose brain as awake wants to take a look..

  -jm