napi_schedule() is expected to be called either:
* From an interrupt, where raised softirqs are handled on IRQ exit
* Fom a softirq disabled section, where raised softirqs are handled on
the next call to local_bh_enable().
* From a softirq handler, where raised softirqs are handled on the next
round in do_softirq(), or further deferred to a dedicated kthread.
r8152 may call napi_schedule() on device resume time from a bare task
context without disabling softirqs as the following trace shows:
__raise_softirq_irqoff
__napi_schedule
rtl8152_runtime_resume.isra.0
rtl8152_resume
usb_resume_interface.isra.0
usb_resume_both
__rpm_callback
rpm_callback
rpm_resume
__pm_runtime_resume
usb_autoresume_device
usb_remote_wakeup
hub_event
process_one_work
worker_thread
kthread
ret_from_fork
ret_from_fork_asm
This may result in the NET_RX softirq vector to be ignored until the
next interrupt or softirq handling. The delay can be long if the
above kthread leaves the CPU idle and the tick is stopped for a while,
as reported with the following message:
NOHZ tick-stop error: local softirq work is pending, handler #08!!!
Fix this with disabling softirqs while calling napi_schedule(). The
call to local_bh_enable() will take care of the NET_RX raised vector.
Reported-by: Paul Menzel <pmenzel@xxxxxxxxxxxxx>
Closes: 354a2690-9bbf-4ccb-8769-fa94707a9340@xxxxxxxxxxxxx
Signed-off-by: Frederic Weisbecker <frederic@xxxxxxxxxx>
---
drivers/net/usb/r8152.c | 5 ++++-
1 file changed, 4 insertions(+), 1 deletion(-)
diff --git a/drivers/net/usb/r8152.c b/drivers/net/usb/r8152.c
index 468c73974046..1325460ae457 100644
--- a/drivers/net/usb/r8152.c
+++ b/drivers/net/usb/r8152.c
@@ -8537,8 +8537,11 @@ static int rtl8152_runtime_resume(struct r8152 *tp)
clear_bit(SELECTIVE_SUSPEND, &tp->flags);
smp_mb__after_atomic();
- if (!list_empty(&tp->rx_done))
+ if (!list_empty(&tp->rx_done)) {
+ local_bh_disable();
napi_schedule(&tp->napi);
+ local_bh_enable();
+ }
usb_submit_urb(tp->intr_urb, GFP_NOIO);
} else {
--
2.46.0
