aboutsummaryrefslogtreecommitdiffstats
path: root/erts
diff options
context:
space:
mode:
authorMaxim Fedorov <[email protected]>2018-11-19 14:15:43 -0800
committerMaxim Fedorov <[email protected]>2018-11-26 11:27:12 -0800
commit63077f5bb3011ec9a14aa9b9e39ed31e4525078b (patch)
treed8399c688de7e1cfda7655f0f2dfc6b2bbb75755 /erts
parentd743f51960384dd17ae3cb905ec57e8012c3abd7 (diff)
downloadotp-63077f5bb3011ec9a14aa9b9e39ed31e4525078b.tar.gz
otp-63077f5bb3011ec9a14aa9b9e39ed31e4525078b.tar.bz2
otp-63077f5bb3011ec9a14aa9b9e39ed31e4525078b.zip
erts: fix attempt to start timer when executing on dirty scheduler
Since OTP R20, there is a possibility for MAJOR garbage collection to run on dirty scheduler. So DistEntry destructor is being called on dirty scheduler as well. This, in turn, leads to an attempt to schedule timer on a dirty scheduler too, which is impossible (and will assert on debug build, but does succeed for release build, creating an infinite busy loop, since aux work wakes scheduler up, but dirty scheduler cannot execute aus work). There is a similar method in erl_hl_timer, see erts_start_timer_callback.
Diffstat (limited to 'erts')
-rw-r--r--erts/emulator/beam/erl_node_tables.c21
1 files changed, 19 insertions, 2 deletions
diff --git a/erts/emulator/beam/erl_node_tables.c b/erts/emulator/beam/erl_node_tables.c
index f4dc60941a..18ed782ae3 100644
--- a/erts/emulator/beam/erl_node_tables.c
+++ b/erts/emulator/beam/erl_node_tables.c
@@ -421,8 +421,25 @@ static void schedule_delete_dist_entry(DistEntry* dep)
*
* Note that timeouts do not guarantee thread progress.
*/
- erts_schedule_thr_prgr_later_op(start_timer_delete_dist_entry,
- dep, &dep->later_op);
+ ErtsSchedulerData *esdp = erts_get_scheduler_data();
+ if (esdp && !ERTS_SCHEDULER_IS_DIRTY(esdp)) {
+ erts_schedule_thr_prgr_later_op(start_timer_delete_dist_entry,
+ dep, &dep->later_op);
+ } else {
+ /*
+ * Since OTP 20, it's possible that destructor is executed on
+ * a dirty scheduler. Aux work cannot be done on a dirty
+ * scheduler, and scheduling any aux work on a dirty scheduler
+ * makes the scheduler to loop infinitely.
+ * To avoid this, make a spot jump: schedule this function again
+ * on a first normal scheduler. It is guaranteed to be always
+ * online. Since it's a rare event, this shall not pose a big
+ * utilisation hit.
+ */
+ erts_schedule_misc_aux_work(1,
+ (void (*)(void *))schedule_delete_dist_entry,
+ (void *) dep);
+ }
}
static void