d/freebsd-skq
1
0
Fork 0
Code Issues Pull Requests Packages Projects Releases Wiki Activity

Fix possible deadlock where worker process sends an event to the main process

Browse Source 
while the main process sends control message to the worker process, but worker
process hasn't started control thread yet, because it waits for reply from the
main process.

The fix is to start the control thread before sending any events.

Reported and fix suggested by:	Mikolaj Golub <to.my.trociny@gmail.com>
MFC after:	3 days
This commit is contained in:
pjd 2010-09-22 19:03:11 +00:00
parent 3657e3ff87
commit 9433a082e8
2 changed files with 26 additions and 8 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

16
sbin/hastd/primary.c
View File

@ -807,10 +807,20 @@ hastd_primary(struct hast_resource *res)
proto_send(res->hr_event, NULL, 0);
init_local(res);
if (real_remote(res) && init_remote(res, NULL, NULL))
sync_start();
init_ggate(res);
init_environment(res);
/*
* Create the control thread before sending any event to the parent,
* as we can deadlock when parent sends control request to worker,
* but worker has no control thread started yet, so parent waits.
* In the meantime worker sends an event to the parent, but parent
* is unable to handle the event, because it waits for control
* request response.
*/
error = pthread_create(&td, NULL, ctrl_thread, res);
assert(error == 0);
if (real_remote(res) && init_remote(res, NULL, NULL))
sync_start();
error = pthread_create(&td, NULL, ggate_recv_thread, res);
assert(error == 0);
error = pthread_create(&td, NULL, local_send_thread, res);
@ -823,8 +833,6 @@ hastd_primary(struct hast_resource *res)
assert(error == 0);
error = pthread_create(&td, NULL, sync_thread, res);
assert(error == 0);
error = pthread_create(&td, NULL, ctrl_thread, res);
assert(error == 0);
(void)guard_thread(res);
}

18
sbin/hastd/secondary.c
View File

@ -393,17 +393,27 @@ hastd_secondary(struct hast_resource *res, struct nv *nvin)
pjdlog_errno(LOG_WARNING, "Unable to set connection timeout");
init_local(res);
init_remote(res, nvin);
init_environment();
/*
* Create the control thread before sending any event to the parent,
* as we can deadlock when parent sends control request to worker,
* but worker has no control thread started yet, so parent waits.
* In the meantime worker sends an event to the parent, but parent
* is unable to handle the event, because it waits for control
* request response.
*/
error = pthread_create(&td, NULL, ctrl_thread, res);
assert(error == 0);
init_remote(res, nvin);
event_send(res, EVENT_CONNECT);
error = pthread_create(&td, NULL, recv_thread, res);
assert(error == 0);
error = pthread_create(&td, NULL, disk_thread, res);
assert(error == 0);
error = pthread_create(&td, NULL, send_thread, res);
assert(error == 0);
(void)ctrl_thread(res);
(void)send_thread(res);
}
static void