net: Handle napi_schedule() calls from non-interrupt

napi_schedule() is expected to be called either:

* From an interrupt, where raised softirqs are handled on IRQ exit

* From 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.

Other bare tasks context may end up ignoring the raised NET_RX vector
until the next random softirq handling opportunity, which may not
happen before a while if the CPU goes idle afterwards with the tick
stopped.

Such "misuses" have been detected on several places thanks to messages
of the kind:

	"NOHZ tick-stop error: local softirq work is pending, handler #8!!!"

For example:

       __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

And also:

* drivers/net/usb/r8152.c::rtl_work_func_t
* drivers/net/netdevsim/netdev.c::nsim_start_xmit

There is a long history of issues of this kind:

	019edd0 ("ath10k: sdio: Add missing BH locking around napi_schdule()")
	3300685 ("idpf: disable local BH when scheduling napi for marker packets")
	e3d5d70 ("net: lan78xx: fix "softirq work is pending" error")
	e55c27e ("mt76: mt7615: add missing bh-disable around rx napi schedule")
	c0182aa ("mt76: mt7915: add missing bh-disable around tx napi enable/schedule")
	970be1d ("mt76: disable BH around napi_schedule() calls")
	019edd0 ("ath10k: sdio: Add missing BH locking around napi_schdule()")
	30bfec4 ("can: rx-offload: can_rx_offload_threaded_irq_finish(): add new  function to be called from threaded interrupt")
	e63052a ("mlx5e: add add missing BH locking around napi_schdule()")
	83a0c6e ("i40e: Invoke softirqs after napi_reschedule")
	bd4ce94 ("mlx4: Invoke softirqs after napi_reschedule")
	8cf699e ("mlx4: do not call napi_schedule() without care")
	ec13ee8 ("virtio_net: invoke softirqs after __napi_schedule")

This shows that relying on the caller to arrange a proper context for
the softirqs to be handled while calling napi_schedule() is very fragile
and error prone. Also fixing them can also prove challenging if the
caller may be called from different kinds of contexts.

Therefore fix this from napi_schedule() itself with waking up ksoftirqd
when softirqs are raised from task contexts.

Reported-by: Paul Menzel <[email protected]>
Reported-by: Jakub Kicinski <[email protected]>
Reported-by: Francois Romieu <[email protected]>
Closes: https://lore.kernel.org/lkml/[email protected]/
Cc: Breno Leitao <[email protected]>
Signed-off-by: Frederic Weisbecker <[email protected]>
Reviewed-by: Eric Dumazet <[email protected]>
Link: https://patch.msgid.link/[email protected]
Signed-off-by: Jakub Kicinski <[email protected]>

Author: Frederic Weisbecker

net/core/dev.c

@@ -4757,7 +4757,7 @@ static inline void ____napi_schedule(struct softnet_data *sd,
 	 * we have to raise NET_RX_SOFTIRQ.
 	 */
 	if (!sd->in_net_rx_action)
-		__raise_softirq_irqoff(NET_RX_SOFTIRQ);
+		raise_softirq_irqoff(NET_RX_SOFTIRQ);
 }
 
 #ifdef CONFIG_RPS