| Message ID | 147981906281.19181.14128282832689294852.stgit@Solace.fritz.box (mailing list archive) |
|---|---|
| State | New, archived |
| Headers | show |
On Tue, Nov 22, 2016 at 01:51:02PM +0100, Dario Faggioli wrote: > Since b047f888d489 ("xen: sched: leave CPUs doing tasklet > work alone") a cpu executing a tasklet, is not marked as > idle. > > Therefore: > - avoid asserting that we can't find the idle vcpu running > on one of them, which is not true, > - avoid triggering a preemption on them (and add an assert > checking that). > > This fixes a bug identified by OSSTest, in flight 102372 > (on ARM, but it's not at all ARM specific), where the > ASSERT() was triggering like this: > > (XEN) Xen call trace: > (XEN) [<0022af78>] sched_credit2.c#runq_tickle+0x3e8/0x61c (PC) > (XEN) [<0022aedc>] sched_credit2.c#runq_tickle+0x34c/0x61c (LR) > (XEN) [<0022b644>] sched_credit2.c#csched2_context_saved+0x128/0x1a4 > (XEN) [<0023303c>] context_saved+0x7c/0xa4 > (XEN) [<0024f660>] domain.c#schedule_tail+0x2b4/0x308 > (XEN) [<0024faac>] context_switch+0x80/0x94 > (XEN) [<0022ff48>] schedule.c#schedule+0x76c/0x7ec > (XEN) [<002338d4>] softirq.c#__do_softirq+0xcc/0xec > (XEN) [<00233968>] do_softirq+0x18/0x28 > (XEN) [<00261084>] leave_hypervisor_tail+0x58/0x88 > (XEN) [<002649d0>] entry.o#return_to_guest+0xc/0xb8 > (XEN) > (XEN) **************************************** > (XEN) Panic on CPU 1: > (XEN) Assertion '!is_idle_vcpu(cur->vcpu)' failed at sched_credit2.c:1009 > (XEN) **************************************** > > Signed-off-by: Dario Faggioli <dario.faggioli@citrix.com> > Reviewed-by: Jan Beulich <jbeulich@suse.com> Release-acked-by: Wei Liu <wei.liu2@citrix.com>
diff --git a/xen/common/sched_credit2.c b/xen/common/sched_credit2.c index 1f26553..c2c563d 100644 --- a/xen/common/sched_credit2.c +++ b/xen/common/sched_credit2.c @@ -1006,7 +1006,12 @@ runq_tickle(const struct scheduler *ops, struct csched2_vcpu *new, s_time_t now) cur = CSCHED2_VCPU(curr_on_cpu(i)); - ASSERT(!is_idle_vcpu(cur->vcpu)); + /* + * Even if the cpu is not in rqd->idle, it may be running the + * idle vcpu, if it's doing tasklet work. Just skip it. + */ + if ( is_idle_vcpu(cur->vcpu) ) + continue; /* Update credits for current to see if we want to preempt. */ burn_credits(rqd, cur, now); @@ -1042,6 +1047,7 @@ runq_tickle(const struct scheduler *ops, struct csched2_vcpu *new, s_time_t now) return; } + ASSERT(!is_idle_vcpu(curr_on_cpu(ipid))); SCHED_STAT_CRANK(tickled_busy_cpu); tickle: BUG_ON(ipid == -1);
Since b047f888d489 ("xen: sched: leave CPUs doing tasklet work alone") a cpu executing a tasklet, is not marked as idle. Therefore: - avoid asserting that we can't find the idle vcpu running on one of them, which is not true, - avoid triggering a preemption on them (and add an assert checking that). This fixes a bug identified by OSSTest, in flight 102372 (on ARM, but it's not at all ARM specific), where the ASSERT() was triggering like this: (XEN) Xen call trace: (XEN) [<0022af78>] sched_credit2.c#runq_tickle+0x3e8/0x61c (PC) (XEN) [<0022aedc>] sched_credit2.c#runq_tickle+0x34c/0x61c (LR) (XEN) [<0022b644>] sched_credit2.c#csched2_context_saved+0x128/0x1a4 (XEN) [<0023303c>] context_saved+0x7c/0xa4 (XEN) [<0024f660>] domain.c#schedule_tail+0x2b4/0x308 (XEN) [<0024faac>] context_switch+0x80/0x94 (XEN) [<0022ff48>] schedule.c#schedule+0x76c/0x7ec (XEN) [<002338d4>] softirq.c#__do_softirq+0xcc/0xec (XEN) [<00233968>] do_softirq+0x18/0x28 (XEN) [<00261084>] leave_hypervisor_tail+0x58/0x88 (XEN) [<002649d0>] entry.o#return_to_guest+0xc/0xb8 (XEN) (XEN) **************************************** (XEN) Panic on CPU 1: (XEN) Assertion '!is_idle_vcpu(cur->vcpu)' failed at sched_credit2.c:1009 (XEN) **************************************** Signed-off-by: Dario Faggioli <dario.faggioli@citrix.com> Reviewed-by: Jan Beulich <jbeulich@suse.com> --- Cc: George Dunlap <george.dunlap@citrix.com> Cc: Wei Liu <wei.liu2@citrix.com> --- Changes from v1: * ASSERT() moved down, as suggested during review. --- xen/common/sched_credit2.c | 8 +++++++- 1 file changed, 7 insertions(+), 1 deletion(-)