Commit Graph

126 Commits

Author SHA1 Message Date
Gleb Smirnoff
cf3254aac8 Do not leak hooks in ng_bypass().
Submitted by:	Alexander Motin <mav alkar.net>
2006-10-11 14:33:08 +00:00
Gleb Smirnoff
b96baf0a65 When counting nodes second time, use the same criteria as for
the first time.

PR:		kern/98529
Submitted by:	Michael Heyman
2006-06-07 12:42:15 +00:00
Gleb Smirnoff
27e216594b Use NET_CALLOUT_MPSAFE for netgraph callout initializer. 2006-06-06 08:05:27 +00:00
John Baldwin
b1e30c4c4e Conditionally acquire Giant in netgraph callouts to honor mpsafenet=0.
Reported by:	sekes <gexlie at gmail dot com>
MFC after:	1 week
2006-06-02 20:35:39 +00:00
Gleb Smirnoff
2955ee1802 - Print also node ID in ktr(9) messages. [1]
- Use fixed length for function name, making ktrdump(8) output
  easier to read.

Suggested by:	julian [1]
2006-01-12 22:41:32 +00:00
Gleb Smirnoff
22b286280c Remove old debugging leftover.
Reviewed by:	julian
2006-01-12 21:03:09 +00:00
Gleb Smirnoff
1be0418cbc Fix wording in last commit.
Submitted by:	julian
2006-01-12 10:15:51 +00:00
Gleb Smirnoff
3b33fbe7d4 Add ktr(9) hooks to easier tracing of the netgraph item flow through
netgraph.
2006-01-11 15:29:48 +00:00
Gleb Smirnoff
4c9b591060 Some whitespace and style cleanup. 2005-11-15 10:54:20 +00:00
Gleb Smirnoff
ac5dd14182 Fix two races which happen when netgraph is restructuring:
- Introduce ng_topo_mtx, a mutex to protect topology changes.
  - In ng_destroy_node() protect with ng_topo_mtx the process
    of checking and pointing at ng_deadnode. [1]
  - In ng_con_part2() check that our peer is not a ng_deadnode,
    and protect the check with ng_topo_mtx.
  - Add KASSERTs to ng_acquire_read/write, to make more
    understandible synopsis in case if called on ng_deadnode.

Reported by:	Roselyn Lee [1]
2005-11-02 15:23:47 +00:00
Gleb Smirnoff
4be5933577 Rework the ng_item queueing on nodes:
- Introduce a new flags NGQF_QREADER and NGQF_QWRITER,
    which tell how the item should be actually applied,
    overriding NGQF_READER/NGQF_WRITER flags.
  - Do not differ between pending reader or writer. Use only
    one flag that is raised, when there are pending items.
  - Schedule netgraph ISR in ng_queue_rw(), so that callers
    do not need to do this job.
  - Fix several comments.

Submitted by:	julian
2005-11-02 14:27:24 +00:00
Gleb Smirnoff
eb2405dde8 - When flushing node input queue, check whether item has a callback. If
it does, then call it suppling ENOENT as error value.
- Add assert, that helped to catch the above error.
2005-10-13 11:55:50 +00:00
Gleb Smirnoff
32b33288f7 After rev. 1.103 the oitem and ierror are no longer needed, remove them. 2005-10-12 10:18:44 +00:00
Gleb Smirnoff
714fb86548 Fix a regression introduced in rev. 1.107. If an item once had a writer
semantics, and then was reused for next node, it still would be applied
as writer again.
  To fix the regression the decision is made never to alter item->el_flags
after the item has been allocated. This requires checking for overrides
both in ng_dequeue() and in ng_snd_item().

  Details:
  - Caller of the ng_apply_item() knows what is the current access to
    node and specifies it to ng_apply_item(). The latter drops the
    given access after item has beem applied.
  - ng_dequeue() needs to be supplied with int pointer, where it stores
    the obtained access on node.
  - Check for node/hook access overrides in ng_dequeue().
2005-10-11 13:48:38 +00:00
Gleb Smirnoff
1bf8e0faed Fix an item leak, that happens when some node calls ng_callout() two
times consequently, without checking whether callout has been serviced
or not. (ng_pptpgre and ng_ppp were catched in this behavior).

  - In ng_callout() save old item before calling callout_reset(). If the
    latter has returned 1, then free this item.
  - In ng_uncallout() clear c->c_arg.

Problem reported by:	Alexandre Kardanev
2005-09-08 14:32:19 +00:00
Gleb Smirnoff
ba5b359aef Fix build. 2005-09-06 20:36:38 +00:00
Gleb Smirnoff
b32cfb3228 In INVARIANTS case also check that nodes do not pass queues of mbufs
each other.
2005-09-06 17:02:13 +00:00
Gleb Smirnoff
c6118fcc1c Raise one more bit in READER_MASK. I believe that before this change
it was possible to have 1 reader and 1 writer thread working on
a node simultaneously.

Reviewed by:	julian
2005-09-06 16:58:25 +00:00
Gleb Smirnoff
6064e568ec Use non-debug macros inside debugging functions, to prevent
important information from being rewritten.
2005-09-02 19:52:54 +00:00
Gleb Smirnoff
6f683eeed2 Cleanup the reader/writer policy in netgraph(4). Assign
either reader or writer flag on item in the function, that
allocates the item. Do not modify these flags when item is
applied or queued.
  The only exceptions are node and hook overrides - they can
change item flags to writer.
2005-08-26 15:14:33 +00:00
Gleb Smirnoff
03b25f5ddc In ng_callout() assert that supplied arguments are non-NULL. 2005-08-21 19:48:51 +00:00
Gleb Smirnoff
5bc15201f4 Fix cut-n-paste error, introduced in rev. 1.103. 2005-07-21 22:15:37 +00:00
Gleb Smirnoff
e58d779daa Catch up with netgraph.h rev. 1.57 and fix build. 2005-07-21 20:34:40 +00:00
Gleb Smirnoff
f912c0f77c Problem description:
At the end of ng_snd_item(), node queue is processed. In certain
netgraph setups deep recursive calls can occur.
  For example this happens, when two nodes are connected and can send
items to each other in both directions. If, for some reason, both nodes
have a lot of items in their queues, then the processing thread will
recurse between these two nodes, delivering items left and right, going
deeper in the stack. Other setups can suffer from deep recursion, too.
The following factors can influence risk of deep netgraph call:
 - periodical write-access events on node
 - combination of slow link and fast one in one graph
 - net.inet.ip.fastforwarding

Changes made:

 - In ng_acquire_{read,write}() do not dequeue another item. Instead,
   call ng_setisr() for this node.
 - At the end of ng_snd_item(), do not process queue. Call ng_setisr(),
   if there are any dequeueable items on node queue.
 - In ng_setisr() narrow worklist mutex holding.
 - In ng_setisr() assert queue mutex.

Theoretically, the first two changes should negatively affect performance.
To check this, some profiling was made:

1) In general real tasks, no noticable performance difference was found.

2) The following test was made: two multithreaded nodes and one
single-threaded were connected into a ring. A large queues of packets
were sent around this ring. Time to pass the ring N times was measured.
This is a very vacuous test: no items/mbufs are allocated, no upcalls or
downcalls outside of netgraph. It doesn't represent a real load, it is
a stress test for ng_acquire_{read,write}() and item queueing functions.
Surprisingly, the performance impact was positive! New code is 13% faster
on UP and 17% faster on SMP, in this particular test.

The problem was originally found, described, analyzed and original patch
was written by Roselyn Lee from Vernier Networks. Thanks!

Submitted by:		Roselyn Lee <rosel verniernetworks com>
2005-07-21 12:08:37 +00:00
Gleb Smirnoff
8afe16d57b In the splnet times, netgraph was functional and synchronous. Nowadays,
an item may be queued and processed later. While this is OK for mbufs,
this is a problem for control messages.

In the framework:
- Add optional callback function pointer to an item. When item gets
  applied the callback is executed from ng_apply_item().
- Add new flag NG_PROGRESS. If this flag is supplied, then return
  EINPROGRESS instead of 0 in case if item failed to deliver
  synchronously and was queued.
- Honor NG_PROGRESS in ng_snd_item().

In ng_socket:
- When userland sends control message add callback to the item.
- If ng_snd_item() returns EINPROGRESS, then sleep.

This change fixes possible races in ngctl(8) scripts.

Reviewed by:	julian
Approved by:	re (scottl)
2005-07-05 17:35:20 +00:00
Gleb Smirnoff
42282202ee - Extend interface of ng_getqblk(), so that malloc wait flags are
specified by caller.
- Change ng_send_item() interface - use 'flags' argument instead of
  boolean 'queue'.
- Extend ng_send_fn(), ng_package_data() and ng_package_msg()
  interface - add possibility to pass flags. Rename ng_send_fn() to
  ng_send_fn1(). Create macro for ng_send_fn().
- Update all macros, that use ng_package_data() and ng_package_msg().

Reviewed by:	julian
2005-05-16 17:07:03 +00:00
Gleb Smirnoff
8253c060cf Move queue element routines upper, so that all related
declarations are in one place, to improve readability.
No functional changes.
2005-05-14 10:07:17 +00:00
Gleb Smirnoff
1489164f6b Use uma(9) for allocating netgraph items:
- ng_getqblk() simply runs uma_zalloc().
  - ng_free_item() simply frees.
  - ngq_mtx is pushed down under NETGRAPH_DEBUG.
  - NGQF_FREE is removed.

Increase default maxalloc to 512.

Reviewed by:	julian
2005-05-14 09:25:18 +00:00
Gleb Smirnoff
aacdb11479 Since there is no way to queue a function call to node, create
ng_queue_fn() - a queue version of ng_send_fn().
2005-05-13 11:35:02 +00:00
Gleb Smirnoff
8bb55179ad Plug item leak in case when NGI_FN is applied to invalid node.
Submitted by:	Roselyn Lee
MFC after:	3 days
2005-03-10 19:27:08 +00:00
Gleb Smirnoff
687809752d Make netgraph ISR and callout MPSAFE.
Reviewed by:	rwatson, ru
2005-02-12 09:52:36 +00:00
Gleb Smirnoff
1fbb36ff80 Rename ng_callout_trapoline to ng_callout_trampoline.
Requested by:	ru
2005-01-26 09:01:50 +00:00
Gleb Smirnoff
7b26345646 With recent changes to _callout_stop_safe() we can remove a hack
in ng_uncallout().
2005-01-25 22:08:19 +00:00
Gleb Smirnoff
19724144d6 Fix an evil typo.
Submitted by:	Roselyn Lee
MFC after:	3 days
2005-01-24 13:32:19 +00:00
Warner Losh
c398230b64 /* -> /*- for license, minor formatting changes 2005-01-07 01:45:51 +00:00
Maksim Yevmenkin
aa38f8f98c Introduce new startup level SI_SUB_NETGRAPH that is after
SI_SUB_INIT_IF but before SI_SUB_DRIVERS. Make Netgraph(4)
framework initialize at SI_SUB_NETGRAPH level.

This does not address the bigger problem: MODULE_DEPEND
does not seem to work when modules are compiled in the
kernel, but it fixes the problem with Netgraph Bluetooth
device drivers reported by a few folks.

PR:		i386/69876
Reviewed by:	julian, rik, scottl
MFC after:	3 days
2005-01-06 17:45:03 +00:00
Gleb Smirnoff
8e853a2c74 Move systm.h up, since it is required by mbuf.h.
Requested by:	ru
2004-12-23 13:09:37 +00:00
Gleb Smirnoff
77a58296d8 - sort includes
- remove duplicate include sys/sysctl.h
2004-12-23 10:48:10 +00:00
Gleb Smirnoff
d58e678f22 Assert queue mutex in ng_dequeue() and ng_queue_rw(). 2004-12-19 14:58:13 +00:00
Gleb Smirnoff
f9d9e1b4ec Mechanically rename s/ng_timeout/ng_callout/g, s/ng_untimeout/ng_uncallout/g.
This is done to keep both versions in RELENG_5 and support both APIs.

Reviewed by:	scottl
Approved by:	julian (mentor), implicitly
2004-12-01 11:56:32 +00:00
Gleb Smirnoff
3eadb26df8 Partically backout previous commit. Since _callout_stop_safe() clears
out c->c_func, we can't take it after callout_stop(). To take it before
we need to acquire callout_lock, to avoid race. This commit narrows
down area where lock is held, but hack is still present.

This should be redesigned.

Approved by:	julian (mentor)
2004-11-04 21:30:18 +00:00
Gleb Smirnoff
30bef41b8a - Make ng_timeout() to use callout() interface instead of timeout().
- Remove callout-hacking from ng_untimeout().

Approved by:	julian (mentor)
MFC after:	1 month
2004-11-02 21:24:30 +00:00
Gleb Smirnoff
7610f57454 When making a peer of unknown type framework tries to load module
using linker_load_module(). This works OK if NGM_MKPEER message came
from userland and we have process associated with thread. But when
NGM_MKPEER was queued because target node was busy, linker_load_module()
is called from netisr thread leading to panic.
To workaround that we do not load modules by framework, instead ng_socket
loads module (if this is required) before sending NGM_MKPEER.
However, the race condition between return from NgSendMsg() and actual
creation of node still exist and needs to be solved.

PR:		kern/62789
Approved by:	julian
2004-07-27 20:30:56 +00:00
Julian Elischer
be4252b367 Slight cosmetic changes.
Also introduce a macro to be called by persistent nodes to signal their
persistence during shutdown to hide this mechanism from the node author.

Make node flags have a consistent style in naming.

Document the change.
2004-07-20 17:15:38 +00:00
Julian Elischer
505fad52f7 Reverse a lock/unlock pair that were the wrong way around in some code that
is obviously not run a lot. (but is in some test cases).
This code is not usually run because it covers a case that doesn't
happen a lot (removing a node that has data traversing it).
2004-07-18 22:57:46 +00:00
Poul-Henning Kamp
3e019deaed Do a pass over all modules in the kernel and make them return EOPNOTSUPP
for unknown events.

A number of modules return EINVAL in this instance, and I have left
those alone for now and instead taught MOD_QUIESCE to accept this
as "didn't do anything".
2004-07-15 08:26:07 +00:00
Marcel Moolenaar
f33ca0c961 Update for the KDB framework:
o  Call kdb_enter() instead of Debugger().
o  Change comments accordingly.
2004-07-10 21:45:58 +00:00
Julian Elischer
8ed370fd90 Remove the home-grown metadata facility in favour of the now generic
mbuf tags facility. Netgraph modules will all need a recompile.

Submitted by:	Gleb Smirnoff <glebius@cell.sick.ru>
2004-06-30 22:51:29 +00:00
Julian Elischer
1020444971 Not quite sure how that one got past me.. 2004-06-26 01:22:29 +00:00
Julian Elischer
1cf3fa7934 Make the frameworkl responsible for not passing the nodes a NULL mbuf pointer.
this allows the nodes to not test for this..

Submitted by:	Gleb Smirnoff <glebius@cell.sick.ru>
2004-06-25 21:11:14 +00:00