flock(2) cannot block other process which try to get same flock?

[Minoru Usui]

Hi, 

I faced strange behaviour of flock(2) in upstream kernel(3.1.0-rc8+).
It seems flock cannot block other processes which try to get same flock.
If someone know the cause of this behaviour, please let me know.

My test is very very simple.
It tries to get flock against same file by 5 processes.
The result is following.

-----------------------------------------------------------------
# ./flock.sh
lock(pid:2384)
# unlock(pid:2384)
lock(pid:2385)   --- lock(a)
lock(pid:2384)   --- lock(b)
unlock(pid:2385) --- unlock(a)
unlock(pid:2384) --- unlock(b)
   :
-----------------------------------------------------------------

I setup auditd and check audit.log, it indicated same behaviour.

-----------------------------------------------------------------
type=SYSCALL msg=audit(1317375911.187:105): arch=c000003e syscall=73 success=yes exit=0 a0=0 a1=2 a2=0 a3=7fffe664f920 items=0 ppid=2382 pid=2384 auid=0 uid=0 gid=0 euid=0 suid=0 fsuid=0 egid=0 sgid=0 fsgid=0 tty=pts1 ses=1 comm="flock" exe="/home/usui/devel/tp/flock" key=(null)

type=SYSCALL msg=audit(1317375912.187:106): arch=c000003e syscall=73 success=yes exit=0 a0=0 a1=8 a2=0 a3=8 items=0 ppid=1 pid=2384 auid=0 uid=0 gid=0 euid=0 suid=0 fsuid=0 egid=0 sgid=0 fsgid=0 tty=pts1 ses=1 comm="flock" exe="/home/usui/devel/tp/flock" key=(null)

type=SYSCALL msg=audit(1317375911.187:107): arch=c000003e syscall=73 success=yes exit=0 a0=0 a1=2 a2=0 a3=7fffdc2e2c70 items=0 ppid=1 pid=2385 auid=0 uid=0 gid=0 euid=0 suid=0 fsuid=0 egid=0 sgid=0 fsgid=0 tty=pts1 ses=1 comm="flock" exe="/home/usui/devel/tp/flock" key=(null)
  -> lock(a) (pid=2385, a1=2)

type=SYSCALL msg=audit(1317375912.187:108): arch=c000003e syscall=73 success=yes exit=0 a0=0 a1=2 a2=0 a3=7fffe664f920 items=0 ppid=1 pid=2384 auid=0 uid=0 gid=0 euid=0 suid=0 fsuid=0 egid=0 sgid=0 fsgid=0 tty=pts1 ses=1 comm="flock" exe="/home/usui/devel/tp/flock" key=(null)
  -> lock(b) (pid=2384, a1=2)

type=SYSCALL msg=audit(1317375913.187:109): arch=c000003e syscall=73 success=yes exit=0 a0=0 a1=8 a2=0 a3=8 items=0 ppid=1 pid=2384 auid=0 uid=0 gid=0 euid=0 suid=0 fsuid=0 egid=0 sgid=0 fsgid=0 tty=pts1 ses=1 comm="flock" exe="/home/usui/devel/tp/flock" key=(null)
  -> unlock(a) (pid=2385, a1=8)

type=SYSCALL msg=audit(1317375913.187:110): arch=c000003e syscall=73 success=yes exit=0 a0=0 a1=8 a2=0 a3=8 items=0 ppid=1 pid=2385 auid=0 uid=0 gid=0 euid=0 suid=0 fsuid=0 egid=0 sgid=0 fsgid=0 tty=pts1 ses=1 comm="flock" exe="/home/usui/devel/tp/flock" key=(null)
  -> unlock(b) (pid=2384, a1=8)
-----------------------------------------------------------------



Test programs are following.
-----------------------------------------------------------------
# cat -n flock.sh
     1  #! /bin/sh
     2  
     3  CNT=5
     4  
     5  if [ $# -ne 0 ]
     6  then
     7    CNT=$1
     8  fi
     9  
    10  for i in `seq 1 $CNT`
    11  do
    12          ./flock &
    13  done

# cat -n flock.c
     1  #include <stdio.h>
     2  #include <sys/file.h>
     3  #include <sys/types.h>
     4  #include <sys/stat.h>
     5  #include <fcntl.h>
     6  #include <errno.h>
     7  #include <unistd.h>
     8  #include <stdlib.h>
     9  
    10  #define LOCKFILE "/tmp/flocktest"
    11  int main(void)
    12  {
    13          int fd;
    14          int ret;
    15          int pid = getpid();
    16  
    17          while(1) {
    18                  if ((fd = open(LOCKFILE, O_RDONLY) < 0)) {
    19                                  perror("open()");
    20                                  exit(1);
    21                  }
    22  
    23                  if (ret = flock(fd, LOCK_EX) < 0) {
    24                          perror("flock(LOCK_EX)");
    25                          exit(1);
    26                  }
    27                  printf("lock(pid:%d)\n", pid);
    28                  fflush(stdout);
    29  
    30                  sleep(1);
    31  
    32                  printf("unlock(pid:%d)\n", pid);
    33                  fflush(stdout);
    34                  if (ret = flock(fd, LOCK_UN) < 0) {
    35                          perror("flock(LOCK_UN)");
    36                          exit(1);
    37                  }
    38  
    39                  if (close(fd) < 0) {
    40                          perror("flock(LOCK_UN)");
    41                          exit(1);
    42                  }
    43          }
    44  
    45          return 0;
    46  }
-----------------------------------------------------------------
-- 
Minoru Usui <usui@xxxxxxxxxxxxxxxxx>
--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/