Hi,
I've hit some deadlock in kernel xenstore client exposed via
/proc/xen/xenbus. Steps to reproduce are simple:
int main() {
struct xs_handle *xs;
xs = xs_open(0);
xs_watch(xs, "domid", "token");
xs_read(xs, 0, "name", NULL);
return 0;
}
xs_watch internally creates new thread, which uses read to wait for the
watch. And in the same time, the program tries to read some value,
but actually it hangs at sending the command (before even sending a path to be
read). Strace gives this (simplified for readability):
[pid 2494] write(3, "\4\0\0\0\0\0\0\0\0\0\0\0\f\0\0\0", 160 = 16
[pid 2494] write(3, "domid\0", 6) = 6
[pid 2494] write(3, "token\0", 6) = 6
[pid 2495] read(3, <unfinished ...>
[pid 2494] futex(0x71c0d4, FUTEX_WAIT_PRIVATE, 1, NULL <unfinished ...>
[pid 2495] <... read resumed>
"\17\0\0\0\377\377\377\377\220~\255\27\f\0\0\0", 16) = 16
[pid 2495] read(3, "domid\0token\0", 12) = 12
[pid 2495] read(3, "\4\0\0\0\0\0\0\0\0\0\0\0\3\0\0\0", 16) = 16
[pid 2495] read(3, "OK\0", 3) = 3
[pid 2495] futex(0x71c0d4, FUTEX_WAKE_OP_PRIVATE, 1, 1, 0x71c0d0,
{FUTEX_OP_SET, 0, FUTEX_OP_CMP_GT, 1} <unfinished ...>
[pid 2494] <... futex resumed> ) = 0
[pid 2495] <... futex resumed> ) = 1
[pid 2494] futex(0x71c0a8, FUTEX_WAIT_PRIVATE, 2, NULL <unfinished ...>
[pid 2495] futex(0x71c0a8, FUTEX_WAKE_PRIVATE, 1 <unfinished ...>
[pid 2494] <... futex resumed> ) = -1 EAGAIN (Resource
temporarily unavailable)
[pid 2495] <... futex resumed> ) = 0
[pid 2494] futex(0x71c0a8, FUTEX_WAKE_PRIVATE, 1 <unfinished ...>
[pid 2495] read(3, <unfinished ...>
[pid 2494] <... futex resumed> ) = 0
[pid 2494] rt_sigaction(SIGPIPE, {SIG_DFL, [], SA_RESTORER,
0x7fc78c1488f0}, NULL, 8) = 0
[pid 2494] rt_sigaction(SIGPIPE, {SIG_IGN, [], SA_RESTORER,
0x7fc78c1488f0}, {SIG_DFL, [], SA_RESTORER, 0x7fc78c1488f0}, 8) = 0
[pid 2494] write(3, "\2\0\0\0\0\0\0\0\0\0\0\0\5\0\0\0", 16
And thats all - 2494 is waiting on write, 2495 is waiting on read.
On 3.12.x it is working. On 3.17.0 and 3.18.7 it is broken. I haven't
checked versions in the middle.
Any ideas?
--
Best Regards,
Marek Marczykowski-Górecki
Invisible Things Lab
A: Because it messes up the order in which people normally read text.
Q: Why is top-posting such a bad thing?
pgpTtYH7GsIrQ.pgp
Description: PGP signature
_______________________________________________ Xen-devel mailing list Xen-devel@lists.xen.org http://lists.xen.org/xen-devel
