Hello Trond,
sorry for my dumb question, but I don't know where to search. Is it possibl=
e=20
to somewhere find the NFS client patches that go into the kernel versions?
We are mounting /var from an initrd giving the 'nolock' option. I havn't=20
analyzed it in detail yet, but the filelocking code from ypbind gets some=20
strange results causing ypbind not starting up at all. It worked fine with=
=20
2.6.9, but not with 2.6.11-rc4.=20
Afer mounting /var again, when the system is already started up and=20
explicitely giving the 'lock' option, ypbind starts up fine.
I'm now interested in the kernel changes that cause the problems.=20
The nfs-server for /var is unfs3, but I really think thats a client problem=
=2E I=20
also so far couldn't find something of interest in an ethereal log between=
=20
the client and server on the unfs3 port.
Thanks,
Bernd
=2D-=20
Bernd Schubert
Physikalisch Chemisches Institut / Theoretische Chemie
Universit=E4t Heidelberg
INF 229
69120 Heidelberg
e-mail: [email protected]
-------------------------------------------------------
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://ads.osdn.com/?ad_id=6595&alloc_id=14396&op=click
_______________________________________________
NFS maillist - [email protected]
https://lists.sourceforge.net/lists/listinfo/nfs
m=E5 den 21.02.2005 Klokka 23:35 (+0100) skreiv Bernd Schubert:
> Hello Trond,
The best place to search for NFS client patches that have actually been
merged in is to use the web-based Bitkeeper browser:
http://linux.bkbits.net:8080/linux-2.6/src?nav=3Dindex.html
That allows you to go to each file, and see how recently it has changed,
and which changesets (Bitkeeper-speak for "patches") affect it.
> The nfs-server for /var is unfs3, but I really think thats a client probl=
em. I=20
> also so far couldn't find something of interest in an ethereal log betwee=
n=20
> the client and server on the unfs3 port.
You should really always be mounting using "nolock" when you are running
against unfs3. Is that what you have normally been using?
Cheers,
Trond
--=20
Trond Myklebust <[email protected]>
-------------------------------------------------------
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://ads.osdn.com/?ad_id=6595&alloc_id=14396&op=click
_______________________________________________
NFS maillist - [email protected]
https://lists.sourceforge.net/lists/listinfo/nfs
m=E5 den 21.02.2005 Klokka 19:03 (-0500) skreiv Trond Myklebust:
> You should really always be mounting using "nolock" when you are running
> against unfs3. Is that what you have normally been using?
Duh... Sorry, you've already answered that question in your mail....
Do you have a "strace" output of what ypbind is seeing?
Cheers,
Trond
--=20
Trond Myklebust <[email protected]>
-------------------------------------------------------
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://ads.osdn.com/?ad_id=6595&alloc_id=14396&op=click
_______________________________________________
NFS maillist - [email protected]
https://lists.sourceforge.net/lists/listinfo/nfs
Hello Trond,
thanks for the bitkeeper information, it seems I need to become familiar wi=
th=20
it :)
On Tuesday 22 February 2005 01:05, you wrote:
> m=E5 den 21.02.2005 Klokka 19:03 (-0500) skreiv Trond Myklebust:
> > You should really always be mounting using "nolock" when you are running
> > against unfs3. Is that what you have normally been using?
>
> Duh... Sorry, you've already answered that question in your mail....
>
> Do you have a "strace" output of what ypbind is seeing?
To make it easier to understand I isolated the code and testing it as own=20
program.
=2E..
access("/etc/ld.so.nohwcap", F_OK) =3D -1 ENOENT (No such file or=20
directory)
open("/lib/tls/i686/cmov/libc.so.6", O_RDONLY) =3D 3
read(3, "\177ELF\1\1\1\0\0\0\0\0\0\0\0\0\3\0\3\0\1\0\0\0\300\350"..., 512) =
=3D=20
512
fstat64(3, {st_mode=3DS_IFREG|0644, st_size=3D1281552, ...}) =3D 0
old_mmap(0x41019000, 1284940, PROT_READ|PROT_EXEC, MAP_PRIVATE, 3, 0) =3D=20
0x41019000
old_mmap(0x41148000, 32768, PROT_READ|PROT_WRITE, MAP_PRIVATE|MAP_FIXED, 3,=
=20
0x12f000) =3D 0x41148000
old_mmap(0x41150000, 11084, PROT_READ|PROT_WRITE, MAP_PRIVATE|MAP_FIXED|
MAP_ANONYMOUS, -1, 0) =3D 0x41150000
close(3) =3D 0
old_mmap(NULL, 4096, PROT_READ|PROT_WRITE, MAP_PRIVATE|MAP_ANONYMOUS, -1, 0=
) =3D=20
0x4001e000
set_thread_area({entry_number:-1 -> 6, base_addr:0x4001e460, limit:1048575,=
=20
seg_32bit:1, contents:0, read_exec_only:0, limit_in_pages:1,=20
seg_not_present:0, useable:1}) =3D 0
munmap(0x40001000, 115498) =3D 0
open("/var/run/test_lock.pid", O_RDWR|O_CREAT, 0644) =3D 3
fcntl64(3, F_GETFD) =3D 0
fcntl64(3, F_SETFD, FD_CLOEXEC) =3D 0
fcntl64(3, F_GETLK, {type=3DF_WRLCK, whence=3DSEEK_SET, start=3D0, len=3D0,=
pid=3D3595})=20
=3D 0
fstat64(1, {st_mode=3DS_IFCHR|0600, st_rdev=3Dmakedev(136, 0), ...}) =3D 0
mmap2(NULL, 4096, PROT_READ|PROT_WRITE, MAP_PRIVATE|MAP_ANONYMOUS, -1, 0) =
=3D=20
0x40001000
write(1, " lock.l_type =3D 1\n", 17 lock.l_type =3D 1
) =3D 17
write(1, " F_UNLCK =3D 2\n", 13 F_UNLCK =3D 2
) =3D 13
write(1, "errno: 0: Success\n", 18errno: 0: Success
) =3D 18
write(1, "ypbind-mt already running (pid 3"..., 47ypbind-mt already running=
=20
(pid 3595) - exiting
) =3D 47
munmap(0x40001000, 4096) =3D 0
exit_group(1) =3D ?
The PID file is of course empty, so I don't know whats causing the problem=
=20
there, but I'm also not familiar with file locking.
#include <sys/param.h>
#include <sys/types.h>
#include <sys/socket.h>
#include <sys/stat.h>
#include <dirent.h>
#include <stdio.h>
#include <ctype.h>
#include <fcntl.h>
#include <errno.h>
#include <stdlib.h>
#include <unistd.h>
#include <syslog.h>
#include <string.h>
#include <signal.h>
#include <locale.h>
#include <libintl.h>
#include <rpc/rpc.h>
#if defined(HAVE_RPC_SVC_SOC_H)
#include <rpc/svc_soc.h>
#endif /* HAVE_RPC_SVC_SOC_H */
#include <rpcsvc/ypclnt.h>
#include <rpc/pmap_clnt.h>
#include <pthread.h>
#if defined(HAVE_NSS_H)
#include <nss.h>
#endif
#define _(String) gettext (String)
#ifdef HAVE_PATHS_H
#include <paths.h>
#endif
#define _YPBIND_PIDFILE "/var/run/test_lock.pid"
void main ( void )
{
int lock_fd;
struct flock lock ;
int left, written;
pid_t pid;
char pbuf[ 10 ], *ptr;
int flags;
lock_fd =3D open ( _YPBIND_PIDFILE, O_CREAT | O_RDWR,
S_IRUSR | S_IWUSR | S_IRGRP | S_IROTH );
if ( lock_fd < 0 )
printf ( "cannot create pidfile %s" , _YPBIND_PIDFILE );
/* Make sure file gets correctly closed when process finished. */
flags =3D fcntl ( lock_fd, F_GETFD, 0 );
if ( flags =3D=3D -1 ) {
/* Cannot get file flags. */
close ( lock_fd );
return ;
}
flags |=3D FD_CLOEXEC; /* Close on exit. */
if ( fcntl ( lock_fd, F_SETFD, flags ) < 0 ) {
/* Cannot set new flags. */
close ( lock_fd );
return ;
}
lock.l_type =3D F_WRLCK;
lock.l_start =3D 0;
lock.l_whence =3D SEEK_SET;
lock.l_len =3D 0;
/* Is the pidfile locked by another ypserv ? */
if ( fcntl ( lock_fd, F_GETLK, &lock ) < 0 ) {
if ( errno !=3D ENOLCK ) {
printf ( "fcntl error: %s" , strerror ( errno ) );
/* XXX look, which pid is in pidfile */
}
pid =3D 0;
}
else if ( lock.l_type =3D=3D F_UNLCK )
pid =3D 0; /* false, region is not locked by=
=20
another proc */
else {
printf(" lock.l_type =3D %i\n", lock.l_type);
printf(" F_UNLCK =3D %i\n", F_UNLCK);
pid =3D lock.l_pid; /* true, return pid of lock owner =
*/
printf("errno: %i: %s\n", errno, strerror(errno));
}
if ( pid !=3D 0 ) {
printf ( "ypbind-mt already running (pid %d) - exiting\n" ,
pid );
exit ( 1 );
}
printf("locking seems to work fine, pid=3D%i\n", pid);
}
Thanks a lot for your help,
Bernd
-------------------------------------------------------
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://ads.osdn.com/?ad_id=6595&alloc_id=14396&op=click
_______________________________________________
NFS maillist - [email protected]
https://lists.sourceforge.net/lists/listinfo/nfs
NFS: Further fixes for the -onolock case.
Duh... GETLK returns F_UNLCK if and only if the lock could be placed.
Signed-off-by: Trond Myklebust <[email protected]>
---
file.c | 6 +++---
1 files changed, 3 insertions(+), 3 deletions(-)
Index: linux-2.6.11-rc4/fs/nfs/file.c
===================================================================
--- linux-2.6.11-rc4.orig/fs/nfs/file.c
+++ linux-2.6.11-rc4/fs/nfs/file.c
@@ -320,10 +320,10 @@ static int do_getlk(struct file *filp, i
status = NFS_PROTO(inode)->lock(filp, cmd, fl);
else {
struct file_lock *cfl = posix_test_lock(filp, fl);
- if (cfl != NULL) {
+
+ fl->fl_type = F_UNLCK;
+ if (cfl != NULL)
memcpy(fl, cfl, sizeof(*fl));
- fl->fl_type = F_UNLCK;
- }
}
unlock_kernel();
return status;
On Tuesday 22 February 2005 03:16, Trond Myklebust wrote:
> ty den 22.02.2005 Klokka 02:00 (+0100) skreiv Bernd Schubert:
> > To make it easier to understand I isolated the code and testing it as o=
wn
> > program.
>
> Thanks. Does the following patch fix things for you?
>
Thank you very very much, the patch fixes the problem!
Thanks again,
Bernd
=2D-=20
Bernd Schubert
Physikalisch Chemisches Institut / Theoretische Chemie
Universit=E4t Heidelberg
INF 229
69120 Heidelberg
e-mail: [email protected]
-------------------------------------------------------
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://ads.osdn.com/?ad_id=6595&alloc_id=14396&op=click
_______________________________________________
NFS maillist - [email protected]
https://lists.sourceforge.net/lists/listinfo/nfs