| Message ID | 20240418072340.2090877-1-linan666@huaweicloud.com (mailing list archive) |
|---|---|
| State | New, archived |
| Headers | show |
| Series | blk-iocost: do not WARNING if iocg has already offlined | expand |
在 2024/04/18 15:23, linan666@huaweicloud.com 写道: > From: Li Nan <linan122@huawei.com> > > In iocg_pay_debt(), warn is triggered if 'active_list' is empty, which > is intended to confirm iocg is avitve when it has debt. However, warn > can be triggered during removing cgroup controller, as > iocg_waitq_timer_fn() is awakened at that time. > > WARNING: CPU: 0 PID: 2344971 at block/blk-iocost.c:1402 iocg_pay_debt+0x14c/0x190 This line doesn't match the code from mainline, please mention that which kernel release you're testing. Other than that, ioc_pd_free() indeed clear 'active_list' before canceling the timer, this patch looks good to me. Reviewed-by: Yu Kuai <yukuai3@huawei.com> > Call trace: > iocg_pay_debt+0x14c/0x190 > iocg_kick_waitq+0x438/0x4c0 > iocg_waitq_timer_fn+0xd8/0x130 > __run_hrtimer+0x144/0x45c > __hrtimer_run_queues+0x16c/0x244 > hrtimer_interrupt+0x2cc/0x7b0 > > The warn in this situation is meaningless. Since this iocg is being > removed, the state of the 'active_list' is irrelevant, and 'waitq_timer' > is canceled after removing 'active_list' in ioc_pd_free(), which ensure > iocg is freed after iocg_waitq_timer_fn() returns. > > Therefore, add the check if iocg has already offlined to avoid warn > when removing cgroup controller. > > Signed-off-by: Li Nan <linan122@huawei.com> > --- > block/blk-iocost.c | 2 +- > 1 file changed, 1 insertion(+), 1 deletion(-) > > diff --git a/block/blk-iocost.c b/block/blk-iocost.c > index baa20c85799d..2e109c016a39 100644 > --- a/block/blk-iocost.c > +++ b/block/blk-iocost.c > @@ -1440,7 +1440,7 @@ static void iocg_pay_debt(struct ioc_gq *iocg, u64 abs_vpay, > lockdep_assert_held(&iocg->waitq.lock); > > /* make sure that nobody messed with @iocg */ > - WARN_ON_ONCE(list_empty(&iocg->active_list)); > + WARN_ON_ONCE(list_empty(&iocg->active_list) && iocg->pd.online); > WARN_ON_ONCE(iocg->inuse > 1); > > iocg->abs_vdebt -= min(abs_vpay, iocg->abs_vdebt); >
在 2024/4/18 16:52, Yu Kuai 写道: > > > 在 2024/04/18 15:23, linan666@huaweicloud.com 写道: >> From: Li Nan <linan122@huawei.com> >> >> In iocg_pay_debt(), warn is triggered if 'active_list' is empty, which >> is intended to confirm iocg is avitve when it has debt. However, warn >> can be triggered during removing cgroup controller, as >> iocg_waitq_timer_fn() is awakened at that time. >> >> WARNING: CPU: 0 PID: 2344971 at block/blk-iocost.c:1402 >> iocg_pay_debt+0x14c/0x190 > > This line doesn't match the code from mainline, please mention that > which kernel release you're testing. > Thanks for your review. I tested at 5.10, but mainline has the same issue. > Other than that, ioc_pd_free() indeed clear 'active_list' before > canceling the timer, this patch looks good to me. > > Reviewed-by: Yu Kuai <yukuai3@huawei.com>
Hello, On Thu, Apr 18, 2024 at 03:23:40PM +0800, linan666@huaweicloud.com wrote: > From: Li Nan <linan122@huawei.com> > > In iocg_pay_debt(), warn is triggered if 'active_list' is empty, which > is intended to confirm iocg is avitve when it has debt. However, warn > can be triggered during removing cgroup controller, as Maybe saying "a blkcg is being removed" is clearer? > iocg_waitq_timer_fn() is awakened at that time. > > WARNING: CPU: 0 PID: 2344971 at block/blk-iocost.c:1402 iocg_pay_debt+0x14c/0x190 > Call trace: > iocg_pay_debt+0x14c/0x190 > iocg_kick_waitq+0x438/0x4c0 > iocg_waitq_timer_fn+0xd8/0x130 > __run_hrtimer+0x144/0x45c > __hrtimer_run_queues+0x16c/0x244 > hrtimer_interrupt+0x2cc/0x7b0 > > The warn in this situation is meaningless. Since this iocg is being > removed, the state of the 'active_list' is irrelevant, and 'waitq_timer' > is canceled after removing 'active_list' in ioc_pd_free(), which ensure > iocg is freed after iocg_waitq_timer_fn() returns. > > Therefore, add the check if iocg has already offlined to avoid warn > when removing cgroup controller. > > Signed-off-by: Li Nan <linan122@huawei.com> > --- > block/blk-iocost.c | 2 +- > 1 file changed, 1 insertion(+), 1 deletion(-) > > diff --git a/block/blk-iocost.c b/block/blk-iocost.c > index baa20c85799d..2e109c016a39 100644 > --- a/block/blk-iocost.c > +++ b/block/blk-iocost.c > @@ -1440,7 +1440,7 @@ static void iocg_pay_debt(struct ioc_gq *iocg, u64 abs_vpay, > lockdep_assert_held(&iocg->waitq.lock); > > /* make sure that nobody messed with @iocg */ > - WARN_ON_ONCE(list_empty(&iocg->active_list)); > + WARN_ON_ONCE(list_empty(&iocg->active_list) && iocg->pd.online); Can you add a comment explaining why we need the pd.online test? Other than the above nits, looks great to me. Please feel free to add Acked-by: Tejun Heo <tj@kernel.org> Thanks.
在 2024/4/19 0:14, Tejun Heo 写道: > Hello, > > On Thu, Apr 18, 2024 at 03:23:40PM +0800, linan666@huaweicloud.com wrote: >> From: Li Nan <linan122@huawei.com> >> >> In iocg_pay_debt(), warn is triggered if 'active_list' is empty, which >> is intended to confirm iocg is avitve when it has debt. However, warn >> can be triggered during removing cgroup controller, as > > Maybe saying "a blkcg is being removed" is clearer? Thanks for your suggestion. I will correct my expression in next version. > >> iocg_waitq_timer_fn() is awakened at that time. >> >> WARNING: CPU: 0 PID: 2344971 at block/blk-iocost.c:1402 iocg_pay_debt+0x14c/0x190 >> Call trace: >> iocg_pay_debt+0x14c/0x190 >> iocg_kick_waitq+0x438/0x4c0 >> iocg_waitq_timer_fn+0xd8/0x130 >> __run_hrtimer+0x144/0x45c >> __hrtimer_run_queues+0x16c/0x244 >> hrtimer_interrupt+0x2cc/0x7b0 >> >> The warn in this situation is meaningless. Since this iocg is being >> removed, the state of the 'active_list' is irrelevant, and 'waitq_timer' >> is canceled after removing 'active_list' in ioc_pd_free(), which ensure >> iocg is freed after iocg_waitq_timer_fn() returns. >> >> Therefore, add the check if iocg has already offlined to avoid warn >> when removing cgroup controller. >> >> Signed-off-by: Li Nan <linan122@huawei.com> >> --- >> block/blk-iocost.c | 2 +- >> 1 file changed, 1 insertion(+), 1 deletion(-) >> >> diff --git a/block/blk-iocost.c b/block/blk-iocost.c >> index baa20c85799d..2e109c016a39 100644 >> --- a/block/blk-iocost.c >> +++ b/block/blk-iocost.c >> @@ -1440,7 +1440,7 @@ static void iocg_pay_debt(struct ioc_gq *iocg, u64 abs_vpay, >> lockdep_assert_held(&iocg->waitq.lock); >> >> /* make sure that nobody messed with @iocg */ >> - WARN_ON_ONCE(list_empty(&iocg->active_list)); >> + WARN_ON_ONCE(list_empty(&iocg->active_list) && iocg->pd.online); > > Can you add a comment explaining why we need the pd.online test? Yeah, I will add comment in next version. > > Other than the above nits, looks great to me. Please feel free to add > > Acked-by: Tejun Heo <tj@kernel.org> > > Thanks. >
diff --git a/block/blk-iocost.c b/block/blk-iocost.c index baa20c85799d..2e109c016a39 100644 --- a/block/blk-iocost.c +++ b/block/blk-iocost.c @@ -1440,7 +1440,7 @@ static void iocg_pay_debt(struct ioc_gq *iocg, u64 abs_vpay, lockdep_assert_held(&iocg->waitq.lock); /* make sure that nobody messed with @iocg */ - WARN_ON_ONCE(list_empty(&iocg->active_list)); + WARN_ON_ONCE(list_empty(&iocg->active_list) && iocg->pd.online); WARN_ON_ONCE(iocg->inuse > 1); iocg->abs_vdebt -= min(abs_vpay, iocg->abs_vdebt);