Messages in this thread | | | From | Marco Elver <> | Date | Wed, 13 Mar 2024 15:41:18 +0100 | Subject | Re: [PATCH v2 2/4] perf: Enqueue SIGTRAP always via task_work. |
| |
On Tue, 12 Mar 2024 at 19:08, Sebastian Andrzej Siewior <bigeasy@linutronix.de> wrote: > > A signal is delivered by raising irq_work() which works from any context > including NMI. irq_work() can be delayed if the architecture does not > provide an interrupt vector. In order not to lose a signal, the signal > is injected via task_work during event_sched_out(). > > Instead going via irq_work, the signal could be added directly via > task_work. The signal is sent to current and can be enqueued on its > return path to userland instead of triggering irq_work. A dummy IRQ is > required in the NMI case to ensure the task_work is handled before > returning to user land. For this irq_work is used. An alternative would > be just raising an interrupt like arch_send_call_function_single_ipi(). > > During testing with `remove_on_exec' it become visible that the event > can be enqueued via NMI during execve(). The task_work must not be kept > because free_event() will complain later. Also the new task will not > have a sighandler installed. > > Queue signal via task_work. Remove perf_event::pending_sigtrap and > and use perf_event::pending_work instead. Raise irq_work in the NMI case > for a dummy interrupt. Remove the task_work if the event is freed. > > Signed-off-by: Sebastian Andrzej Siewior <bigeasy@linutronix.de> > --- > include/linux/perf_event.h | 3 +-- > kernel/events/core.c | 45 +++++++++++++++++--------------------- > 2 files changed, 21 insertions(+), 27 deletions(-) > > diff --git a/include/linux/perf_event.h b/include/linux/perf_event.h > index d2a15c0c6f8a9..24ac6765146c7 100644 > --- a/include/linux/perf_event.h > +++ b/include/linux/perf_event.h > @@ -781,7 +781,6 @@ struct perf_event { > unsigned int pending_wakeup; > unsigned int pending_kill; > unsigned int pending_disable; > - unsigned int pending_sigtrap; > unsigned long pending_addr; /* SIGTRAP */ > struct irq_work pending_irq; > struct callback_head pending_task; > @@ -959,7 +958,7 @@ struct perf_event_context { > struct rcu_head rcu_head; > > /* > - * Sum (event->pending_sigtrap + event->pending_work) > + * Sum (event->pending_work + event->pending_work) > * > * The SIGTRAP is targeted at ctx->task, as such it won't do changing > * that until the signal is delivered. > diff --git a/kernel/events/core.c b/kernel/events/core.c > index c7a0274c662c8..e9926baaa1587 100644 > --- a/kernel/events/core.c > +++ b/kernel/events/core.c > @@ -2283,21 +2283,6 @@ event_sched_out(struct perf_event *event, struct perf_event_context *ctx) > state = PERF_EVENT_STATE_OFF; > } > > - if (event->pending_sigtrap) { > - bool dec = true; > - > - event->pending_sigtrap = 0; > - if (state != PERF_EVENT_STATE_OFF && > - !event->pending_work) { > - event->pending_work = 1; > - dec = false; > - WARN_ON_ONCE(!atomic_long_inc_not_zero(&event->refcount)); > - task_work_add(current, &event->pending_task, TWA_RESUME); > - } > - if (dec) > - local_dec(&event->ctx->nr_pending); > - } > - > perf_event_set_state(event, state); > > if (!is_software_event(event)) > @@ -6741,11 +6726,6 @@ static void __perf_pending_irq(struct perf_event *event) > * Yay, we hit home and are in the context of the event. > */ > if (cpu == smp_processor_id()) { > - if (event->pending_sigtrap) { > - event->pending_sigtrap = 0; > - perf_sigtrap(event); > - local_dec(&event->ctx->nr_pending); > - } > if (event->pending_disable) { > event->pending_disable = 0; > perf_event_disable_local(event); > @@ -9592,14 +9572,17 @@ static int __perf_event_overflow(struct perf_event *event, > > if (regs) > pending_id = hash32_ptr((void *)instruction_pointer(regs)) ?: 1; > - if (!event->pending_sigtrap) { > - event->pending_sigtrap = pending_id; > + if (!event->pending_work) { > + event->pending_work = pending_id; > local_inc(&event->ctx->nr_pending); > - irq_work_queue(&event->pending_irq); > + WARN_ON_ONCE(!atomic_long_inc_not_zero(&event->refcount)); > + task_work_add(current, &event->pending_task, TWA_RESUME); > + if (in_nmi()) > + irq_work_queue(&event->pending_irq);
Some brief code comments here would help having to dig through git history to understand this.
> } else if (event->attr.exclude_kernel && valid_sample) { > /* > * Should not be able to return to user space without > - * consuming pending_sigtrap; with exceptions: > + * consuming pending_work; with exceptions: > * > * 1. Where !exclude_kernel, events can overflow again > * in the kernel without returning to user space. > @@ -9609,7 +9592,7 @@ static int __perf_event_overflow(struct perf_event *event, > * To approximate progress (with false negatives), > * check 32-bit hash of the current IP. > */ > - WARN_ON_ONCE(event->pending_sigtrap != pending_id); > + WARN_ON_ONCE(event->pending_work != pending_id); > } > > event->pending_addr = 0; > @@ -13049,6 +13032,13 @@ static void sync_child_event(struct perf_event *child_event) > &parent_event->child_total_time_running); > } > > +static bool task_work_cb_match(struct callback_head *cb, void *data) > +{ > + struct perf_event *event = container_of(cb, struct perf_event, pending_task); > + > + return event == data; > +} > + > static void > perf_event_exit_event(struct perf_event *event, struct perf_event_context *ctx) > { > @@ -13088,6 +13078,11 @@ perf_event_exit_event(struct perf_event *event, struct perf_event_context *ctx) > * Kick perf_poll() for is_event_hup(); > */ > perf_event_wakeup(parent_event); > + if (event->pending_work && > + task_work_cancel_match(current, task_work_cb_match, event)) {
Brief comment which case this covers would be good.
> + put_event(event); > + local_dec(&event->ctx->nr_pending); > + } > free_event(event); > put_event(parent_event); > return; > -- > 2.43.0 >
| |