I'm trying to solve a problem using mandatory locks but am having some
difficulty in doing so. (if there's a more appropriate place for
discussing this please ignore the rest of this post. pointers to that
place would be appreciated ;) )
my problem:
when I lock a file with a mandatory write lock (ie. fcntl, +s-x bits and
mand mount option. for code see below) it is still possible:
- for me to rm the file in question
- for the file to be read by an other process
It's not possible to cat to the file or cp another file over it (as
expected). For my application I want a locked file to be completely
locked/protected from other processes.
If I retry the same without setting the +s-x bits the cp & cat succeed,
so something special is being done in the first case.
According to all docs (i can find) mandatory locks should block both the
read and write system calls (i cant find anything in it regarding unlink
though...). In my case however it seems as if only write calls are
blocked but not read calls?
If anyone could shed any light on this it would be much appreciated.
Esger
system details:
linux 2.4.17 partially suse & xfs patched kernel
filesystem is reiserfs
i386 dual pIII450
test app code:
int main ()
{
int fd ;
fd = open ("image.jpg", O_RDWR) ;
if (fd == -1)
{
printf ("error %d while opening file\n", errno) ;
exit (1) ;
}
struct flock lock ;
lock.l_type = F_WRLCK ;
lock.l_whence = SEEK_SET ;
lock.l_start = 0 ;
lock.l_len = 0 ;
lock.l_pid = 0 ; // ignored
int err = fcntl (fd, F_SETLK, &lock) ;
if (err == -1)
{
printf ("error %d while locking file\n", errno) ;
exit (1) ;
}
else
printf ("file locked\n") ;
while (1)
sleep (1) ;
}
-- NeoMail - Webmail that doesn't suck... as much. http://neomail.sourceforge.net - To unsubscribe from this list: send the line "unsubscribe linux-kernel" in the body of a message to majordomo@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html Please read the FAQ at http://www.tux.org/lkml/