[lkml]   [2009]   [Jan]   [18]   [last100]   RSS Feed
Views: [wrap][no wrap]   [headers]  [forward] 
Messages in this thread
SubjectRe: splice_to_pipe() hung (blocked for more than 120 seconds)
On Sun, Jan 18, 2009 at 2:44 PM, Vegard Nossum <> wrote:
> So in short: Is it possible that inode_double_lock() in
> splice_from_pipe() first locks the pipe mutex, THEN locks the
> file/socket mutex? In that case, there should be a lock imbalance,
> because pipe_wait() would unlock the pipe while the file/socket mutex
> is held.
> That would possibly explain the sporadicity of the lockup; it depends
> on the actual order of the double lock.
> Why doesn't lockdep report that? Hm. I guess it is because these are
> both inode mutexes and lockdep can't detect a locking imbalance within
> the same lock class?
> Anyway, that's just a theory. :-) Will try to confirm by simplifying
> the test-case.

Hm, I do believe this _is_ evidence in favour of the theory:

top - 09:03:57 up 2:16, 2 users, load average: 129.27, 49.28, 21.57
Tasks: 161 total, 1 running, 95 sleeping, 1 stopped, 64 zombie


#define _GNU_SOURCE

#include <sys/socket.h>
#include <sys/types.h>

#include <fcntl.h>
#include <errno.h>
#include <pthread.h>
#include <signal.h>
#include <stdio.h>
#include <stdlib.h>
#include <string.h>
#include <unistd.h>

static int sock_fd[2];
static int pipe_fd[2];

#define N 16384

static void *do_write(void *unused)
unsigned int i;
for (i = 0; i < N; ++i)
write(pipe_fd[1], "x", 1);

return NULL;
static void *do_read(void *unused)
unsigned int i;
char c;
for (i = 0; i < N; ++i)
read(sock_fd[0], &c, 1);

return NULL;
static void *do_splice(void *unused)
unsigned int i;
for (i = 0; i < N; ++i)
splice(pipe_fd[0], NULL, sock_fd[1], NULL, 1, 0);

return NULL;
int main(int argc, char *argv[])
pthread_t writer;
pthread_t reader;
pthread_t splicer[2];
while (1) {
if (socketpair(AF_UNIX, SOCK_STREAM, 0, sock_fd) == -1)
if (pipe(pipe_fd) == -1)
pthread_create(&writer, NULL, &do_write, NULL);
pthread_create(&reader, NULL, &do_read, NULL);
pthread_create(&splicer[0], NULL, &do_splice, NULL);
pthread_create(&splicer[1], NULL, &do_splice, NULL);

pthread_join(writer, NULL);
pthread_join(reader, NULL);
pthread_join(splicer[0], NULL);
pthread_join(splicer[1], NULL);

printf("failed to deadlock, retrying...\n");
$ gcc splice.c -lpthread
$ ./a.out &
$ ./a.out &
$ ./a.out &
(as many as you want; then wait for a bit -- ten seconds works for me)
$ killall -9 a.out
(not all will die -- those are now zombies)


"The animistic metaphor of the bug that maliciously sneaked in while
the programmer was not looking is intellectually dishonest as it
disguises that the error is the programmer's own creation."
-- E. W. Dijkstra, EWD1036

 \ /
  Last update: 2009-01-18 15:13    [from the cache]
©2003-2014 Jasper Spaans. hosted at Digital Ocean