Commit Graph

59 Commits

Author SHA1 Message Date
jh
9c043d5908 Use g_eventlock to protect against losing wakeups in the g_event process
and replace tsleep(9) with msleep(9) which doesn't use a timeout. The
previously used timeout caused the event process to wake up ten times
per second on an idle system.

one_event() is now called with the topology lock held and it returns
with both the topology and event locks held when there are no more
events in the queue.

Reported by:	mav, Marius Nünnerich
Reviewed by:	freebsd-geom
2010-11-22 16:47:53 +00:00
jh
6604019e26 Extend the g_eventlock mutex coverage in one_event() to include setting
of the EV_DONE flag and use the mutex to protect against losing wakeups
in g_waitfor_event().

Reported by:	davidxu
Tested by:	davidxu
Discussed on:	freebsd-current
2010-11-03 16:19:35 +00:00
mjacob
d16c7840fb Try and narrow the gap in which you act on an event that has been canceled.
Obtained from:	Jaako Heinonen
MFC after:	1 month
2010-06-08 22:40:02 +00:00
pjd
e429b3be95 When orphaning a provider, cancel events related to it.
Without this change the following situation was possible:

1. Provider is orphaned from within class' access() method on last write
   close - orphan provider event is send.
2. GEOM detects last write close on a provider and sends new provider event.
3. g_orphan_register() is called, and calls all orphan methods of attached
   consumers.
4. New provider event is executed on orphaned provider, all classes can
   taste already orphaned provider, and some may attach consumers to it.
   Those consumers will never go away, because the g_orphan_register()
   was already called.

We end up with a zombie provider.

With this change, at step 3, we will cancel new provider event.

How to repeat this problem:

	# mdconfig -a -t malloc -s 10m
	# geli init -i 0 md0
	# geli attach md0
	# newfs -L test /dev/md0.eli
	# mount /dev/ufs/test /mnt/tmp
	# geli detach -l md0.eli
	# umount /mnt/tmp
	# glabel status
            Name  Status  Components
        ufs/test  N/A     N/A

Reviewed by:	phk
Approved by:	re (kensmith)
2007-09-27 20:18:34 +00:00
pjd
d06bfaa1a9 Delay an orphan event if provider has still in-flight I/O requests.
This way GEOM classes can safely detach from provider when an orphan
event is received. This fixes 'detach with active requests' panic for
gstripe/gconcat under load.

PR:		kern/102766
Submitted by:	mjacob
OK'ed by:	phk
MFC after:	1 week
2006-09-10 09:11:54 +00:00
phk
f24fdd9686 Remove a race condition that could result in processes being stuck
waiting for geom events to happen:

Instead of maintaining a count of outstanding events, simply look if
the queue is empty.  Make sure to not remove events from the queue
until they are executed in order to not open a new race.

Much work by:	pjd
Tested by:	kris
MT6:		yes, should be.
2005-09-04 19:14:19 +00:00
pjd
0435460c63 Correct typo. 2005-05-18 21:53:08 +00:00
pjd
412bff1f39 Introduce g_waitidlelock() function which is simlar to g_waitidle(),
but should be called with the topology lock held and returns with the
topology lock held and empty event queue.

Approved by:	phk (sometime ago)
2004-11-09 23:20:50 +00:00
phk
93429f2778 Add a new per-thread private flag: TDP_GEOM.
This flag gets set whenever the thread posts an event on the GEOM
event queue, and if the flag is set when the thread is prepared
to return to userland from the kernel, g_waitidle() will be called
to make sure that the posted events have completed.

This can replace an insufficient number of g_waitidle() calls in
various other places, and has the advantage of being failsafe:  Any
system call which does a VOP_OPEN()/VOP_CLOSE will now correctly
wait for any geom events it posted as part of spoils or tastes.

Assert that topology and Giant is not held in g_waitidle().
2004-10-23 20:49:17 +00:00
phk
042d10f8ef Make withering water tight.
When we orphan/wither a provider, an attached geom+consumer could
end up being withered as a result and it may be in front of us in
the normal object scanning order so we need to do multi-pass.  On
the other hand, there may be withering stuff we can't get rid off
(yet), so we need to keep track of both the existence of withering
stuff and if there is more we can do at this time.
2004-07-08 16:17:14 +00:00
phk
9a7074f4cb Rearrange some of the GEOM debugging tools to be more structured.
Retire g_sanity() and corresponding debugflag (0x8)

  Retire g_{stall,release}_events().

  Under #ifdef DIAGNOSTIC:

    Make g_valid_obj() an official function and have it return an an
    non-zero integer which indicates the kind of object when found.

    Implement G_VALID_{CLASS,GEOM,CONSUMER,PROVIDER}() macros based
    on g_valid_obj().

    Sprinkle calls to these macros liberally over the infrastructure.

    Always check that we do not free a live object.
2004-03-10 08:49:08 +00:00
pjd
1af94a8033 Now we have g_topology_assert_not(), so use it to detect deadlocks.
Approved by:	phk, scottl (mentor)
2004-02-10 15:55:17 +00:00
phk
98a9a50f22 Don't mingle malloc/g_event flags.
Spotted by:	pjd@
2004-02-02 10:58:07 +00:00
phk
0592926b47 Make sure to keep track of canceled events.
Submitted by:	Pawel Jakub Dawidek <nick@garage.freebsd.pl>
2004-01-23 21:09:38 +00:00
phk
fe692735bf Prevent withering of the provider we're orphaning from happening until
we do it ourselves.

Nailed by:	Simon Heath <heath@cng.fr>
2003-12-23 11:37:05 +00:00
phk
888f0e51c6 KASSERT against multiple orphanings of providers. 2003-12-07 10:04:43 +00:00
phk
aeb3231da5 Fix a harmless bug and add a ')' in a debugging printf.
Submitted by: "Bjoern A. Zeeb" <bzeeb-lists@lists.zabbadoz.net>
2003-11-18 07:54:12 +00:00
phk
ed350390ed Destroy providers maked with G_PF_WITHER when the last consumer has detached. 2003-10-12 11:34:35 +00:00
obrien
40ee2bd5c8 Use __FBSDID().
Approved by:	phk
2003-06-11 06:49:16 +00:00
phk
69b4e9ce01 Add missing va_end() calls.
Noticed by:	tmm
2003-06-07 10:16:53 +00:00
phk
a49f8acd23 Get rid of trivial function g_destroy_event(). 2003-05-02 05:26:19 +00:00
phk
26e81abf93 Add a new flag, EV_CANCELED, and use it to make g_waitfor_event() return
EAGAIN if an event got canceled.
2003-05-01 19:43:52 +00:00
phk
2ae9241429 When events on a reference is cancelled, check our doorstep first,
it might be an orphan.
2003-05-01 19:24:00 +00:00
phk
ffd04bfbdd Introduce a g_waitfor_event() function which posts an event and waits for
it to be run (or cancelled) and use this instead of home-rolled versions.
2003-04-23 21:28:27 +00:00
phk
7a7d406136 More of the event stuff can now be private to geom_event.c 2003-04-23 20:54:42 +00:00
phk
3bbfe9f43b Rename g_call_me() to g_post_event(), and give it a flag
argument to determine if we can M_WAITOK in malloc.
2003-04-23 20:46:12 +00:00
phk
b4c26d2e1a Remove the now unused hardcoded g_post_event() event support. 2003-04-23 20:25:33 +00:00
phk
7c462ceb2b Turn EV_NEW_PROVIDER into a g_call_me() event. 2003-04-23 20:16:13 +00:00
phk
67a31a6107 Convert EV_SPOILED event to use g_call_me(). 2003-04-23 20:06:38 +00:00
phk
16bcc257a6 Turn the hardwired NEW_CLASS event into a g_call_me() event. 2003-04-23 19:34:38 +00:00
phk
51e68a1f27 Move the shutdown eventhandler stuff to a more logical place. 2003-04-23 19:15:27 +00:00
phk
03b1ca36ba Time has run from the "run GEOM in userland" harness, and the new regression
test is built to test GEOM as running in the kernel.

This commit is basically "unifdef -D_KERNEL" to remove the mainly #include
related code to support the userland-harness.
2003-04-13 09:02:06 +00:00
phk
bef7fcb806 Fix a bug which resulted in orphanization getting confused every now
and then.
2003-04-12 16:36:19 +00:00
phk
39cbb43bab Add handling for cancelled events in the g_call_me() methods. 2003-04-02 21:10:04 +00:00
phk
cedf04c475 Change events to have an array of "void *" references, and give the
event posting functions varargs to fill these.

Attribute g_call_me() to appropriate g_geom's where necessary.

Add a flag argument to g_call_me() methods which will be used to signal
cancellation of events in the future.

This commit should be a no-op.
2003-04-02 20:41:18 +00:00
phk
dc9d8429c2 Save a lock: Grab the stall_events SX lock exclusively so it also serialize
OAM reqests.
2003-03-24 13:35:06 +00:00
phk
8e76f1cefa Introduce g_cancel_events() and use it a couple of places where it makes
sense.
2003-03-23 23:01:40 +00:00
phk
3e8d794231 Introduce an SX lock which allows us to stall event processing
during OAM operations.
2003-03-23 21:58:09 +00:00
phk
127bfb5486 #ifdef notyet a bit of code which needs not yet committed refcounting to
work correctly.
2003-03-16 16:29:04 +00:00
phk
29e1315705 If we run out of consumers while orphaning them, and the provider's geom
is withering, destroy the provider when done.

This was exposed by the recent change to geom_dev's orphaning logic.
2003-03-10 23:41:41 +00:00
phk
8da8aa2622 Stamp out Danglish. 2003-03-09 10:02:31 +00:00
imp
cf874b345d Back out M_* changes, per decision of the TRB.
Approved by: trb
2003-02-19 05:47:46 +00:00
alfred
bf8e8a6e8f Remove M_TRYWAIT/M_WAITOK/M_WAIT. Callers should use 0.
Merge M_NOWAIT/M_DONTWAIT into a single flag M_NOWAIT.
2003-01-21 08:56:16 +00:00
phk
6ef829cd7d Remove g_silence(). It does not do anything anymore. 2003-01-13 08:46:32 +00:00
phk
a500538d00 Polish a bit here and there.
Reenable the geom.ctl device so people can play with gbde.

Sponsored by:	DARPA & NAI Labs
2002-11-04 09:31:02 +00:00
phk
23e3bcb405 Make sure we don't loose our topology lock in a call_me() handler.
Sponsored by:	DARPA & NAI Labs.
2002-09-30 08:27:29 +00:00
phk
83aec680cb Implement g_call_me() as a way for geom methods to schedule operations
to be performed in the event-thread.

To do this, we need to lock the eventlist with g_eventlock (nee g_doorlock),
since g_call_me() being called from the UP/DOWN paths will not be able to
aquire g_topology_lock.

This also means that for now these events are not referenced on any
particular consumer/provider/geom.

For UP/DOWN path use, this will not become a problem since the access()
function will make sure we drain any bio's before we dismantle.

Sponsored by:   DARPA & NAI Labs.
2002-09-27 20:38:36 +00:00
phk
a09a7e0aa7 Ok, include also the two tests which actually does effect the claims
of the last commit message.

Sponsored by:	DARPA & NAI Labs.
2002-09-27 20:30:27 +00:00
phk
85fb4b0590 Hook into the shutdown EVENTHANDLER and stop tasting things after we
get notified to make things settle a bit faster.

Sponsored by:	DARPA & NAI Labs.
2002-09-27 20:23:35 +00:00
phk
582654c687 Rename the doorlock to eventlock, it gets to protect a bit more in the future.
Sponsored by:	DARPA & NAI Labs.
2002-09-27 20:18:16 +00:00