CVS log for src/sys/net/pktqueue.c

Up to [cvs.NetBSD.org] / src / sys / net

Request diff between arbitrary revisions

Keyword substitution: kv
Default branch: MAIN

s/explcit/explicit/ in comment.

In pktq_flush():
- Run a dummy softint at IPL_SOFTNET on all CPUs to ensure that the
  ISR for this pktqueue is not running (addresses a pre-existing XXX).
- Hold the barrier lock around the critical section to ensure that
  implicit pktq_barrier() calls via pktq_ifdetach() are held off during
  the critical section.
- Ensure the critical section completes in minimal time by not freeing
  memory during the critical section; instead, just build a list of the
  packets pulled out of the per-CPU queues and free them after the critical
  section is over.

Re-factor how pktq_barrier() is issued by if_detach().

Rather than excplicitly referencing ip_pktq and ip6_pktq in if_detach(),
instead add all pktqueues to a global list.  This list is then used in
the new pktq_ifdetach() function to issue a barrier on all pktqueues.

Note that the performance of this list is not critical; it will seldom
be accessed (then pktqueues are created/destroyed and when network
interfaces are detached), and so a simple synchronization strategy using
a rwlock is sufficient.

pktqueue: Re-factor sysctl handling.

Provide a new pktq_sysctl_setup() function that attaches standard
pktq sysctl nodes below a specified parent node, with either a
fixed node ID or CTL_CREATE to dynamically assign node IDs.  Make
all of the sysctl handlers private to pktqueue.c, and remove the
INET- and INET6-specific pktqueue sysctl code from net/if.c.

pktq_rps_hash(): Make the "funcp" argument const.

pktq_dequeue(): Prevent packets from getting stuck beind barrier markers.

pktq_barrier() ensures that all packets enqueued before the barrier have
been dequeued before the barrier returns.  However, previously, pktq_dequeue()
would return NULL when a barrier marker was encountered.  If there were
packets queued up behind the marker and no additional softint were scheduled
for the pktqueue, those packets would end up stranded.  pktq_dequeue() now
continues to the next slot after the marker, ensuring that processing can
continue after the barrier has been signaled.

Fix net.*.rps_hash=toeplitz-othercpus on one CPU systems.

Fix typo in comment.

Make pktq_rps_hash() pluggable for each interface type.  Reviewed by gdt@n.o, thorpej@n.o, and riastradh@n.o, thanks.

Sync with HEAD.

Sync with HEAD.

Remove strange padding #define and replace with anonymous struct/union

pktqueue(9): Use percpu_create to allow early initialization.

Otherwise pktqueues can't be created before all CPUs are detected --
they will have a queue only for the primary CPU, not for others.

This will also be necessary if we want to add CPU hotplug (still need
some way to block hotplug during pktq_set_maxlen but it's a start).

Merge changes from current as of 20200406

Sync with head.

Use percpu_foreach_xcall() to gather volatile per-cpu counters.  These
must be serialized against the interrupts / soft-interrupts in which
they're manipulated, as well as protected from non-atomic 64-bit memory
loads on 32-bit platforms.

Sync with HEAD

Sync with HEAD

Resolve a couple of conflicts (result of the uimin/uimax changes)

- Fix a bug that drop counter shows incorrect vaule like
  "net.inet.ip.ifq.drops = 72059810241052672"
- Change pktq's length sysctl to uint64_t.

update from HEAD

Sync with HEAD

remove checks for failure after memory allocation calls that cannot fail:

  kmem_alloc() with KM_SLEEP
  kmem_zalloc() with KM_SLEEP
  percpu_alloc()
  pserialize_create()
  psref_class_create()

all of these paths include an assertion that the allocation has not failed,
so callers should not assert that again.

Rebase to HEAD as of a few days ago.

Rebase.

file pktqueue.c was added on branch tls-maxphys on 2014-08-20 00:04:34 +0000

file pktqueue.c was added on branch tls-earlyentropy on 2014-08-10 06:56:15 +0000

Fix pktq_enqueue for rump

Add _RUMP_NATIVE_ABI to the macro condition for i386 and x86_64 because
_RUMPKERNEL is not defined for them. See sys/rump/Makefile.rump.

Found by ATF

Restore RPS of pktq_enqueue unless _RUMPKERNEL

It's a workaround and would be fixed in rump soon.

ok pooka@

Move sysctl_pktq_{maxlen,count} to pktqueue.c and make them global

They will be used by bridge.

ok rmind@

Add 3rd argument to pktq_create to pass sc

It will be used to pass bridge sc for bridge_forward softint.

ok rmind@

pktqueue: add or fix some comments, remove some header inclusions.

Restore the assert in RUMP's softint_schedule_cpu() and just ensure
curcpu() in the caller.

Implement pktq_set_maxlen() and let sysctl net.inet.{ip,ip6}.ifq.maxlen be
changed on the fly again.

- Implement pktqueue interface for lockless IP input queue.
- Replace ipintrq and ip6intrq with the pktqueue mechanism.
- Eliminate kernel-lock from ipintr() and ip6intr().
- Some preparation work to push softnet_lock out of ipintr().

Discussed on tech-net.