ac306aa704
Sort.
2018-08-27 15:11:04 +00:00
88b7f7da4f
Fix argument types of atomic64_add/sub to be sensible.
...
Technically Linux does use long long rather than int64_t, but that's
silly.
2018-08-27 15:10:52 +00:00
101f4868d0
Implement ATOMIC64_INIT.
2018-08-27 15:10:41 +00:00
e733144d22
Need <uvm/uvm.h> for struct uvm_object.
2018-08-27 15:10:25 +00:00
8def7a33b3
Sprinkle AGP conditionals where needed.
2018-08-27 15:10:12 +00:00
290bbba29a
Work around a namespacing clash between dtrace and linux symbol.
2018-08-27 15:09:58 +00:00
11e2f4e40e
Use dtrace probes to implement i915_trace.h.
2018-08-27 15:09:35 +00:00
1d9ce7071c
Hash-locked atomic64 to give ports like powerpc32 a chance at radeon.
2018-08-27 15:08:54 +00:00
fdce7bed9d
Delete spurious file accidentally committed.
2018-08-27 15:08:25 +00:00
936018793f
dtrace sdt probes for linux rcu.
2018-08-27 15:07:59 +00:00
70435d9314
Don't skimp on the last batch of work when worker is dying.
...
This is capitalism. We can't afford to waste time the worker could
spend on labour even if we work 'em to death!
2018-08-27 15:07:44 +00:00
077afc9a15
Work around broken empty expansion of SDT_PROBE* if !KDTRACE_HOOKS.
2018-08-27 15:07:29 +00:00
e794a29ba6
sdt dtrace probes for linux kmap.
2018-08-27 15:07:11 +00:00
5248a24ccc
sdt dtrace probes for linux idr.
2018-08-27 15:06:54 +00:00
ccfee5f718
Sprinkle dtrace probes all over the Linux workqueue stuff.
2018-08-27 15:06:37 +00:00
0466378a17
Fix post-cancellation case of queue_delayed_work.
...
Request the callout queue immediately, as if it hadn't been cancelled
at all, rather than reschedule for the next tick.
2018-08-27 15:06:20 +00:00
e27b343568
Rework Linux workqueue synchronization yet again.
...
- Use a low bit in the pointer to the queue, rather than whether the
pointer is null or not, to determine whether the work item is
queued/scheduled.
- _Preserve_ the pointer to the queue after we release the work.
- Release the work _before_ executing the function, not after.
This simplifies some things: we no longer have to distinguish whether
the work is queued or running in the functions where we are trying to
modify it. The pointer has to be preserved even after the work is
released so that we can flush the workqueue after the work has been
released.
2018-08-27 15:06:02 +00:00
a1c732f301
Fix flush_delayed_work semantics: run it now if it was scheduled.
2018-08-27 15:05:44 +00:00
168b5894a5
Reorder fields to get better space savings than bit fields.
2018-08-27 15:05:30 +00:00
31bca41163
Document workqueue semantics as I understand it.
2018-08-27 15:05:16 +00:00
1fca7189a5
Avoid the prospect of callout calls piling up.
...
Don't ever callout_schedule the callout while an existing call may be
in progress.
Echo some cases from cancel_delayed_work in flush_delayed_work.
2018-08-27 15:05:01 +00:00
296411f744
Don't let more than one thread claim credit for cancellation.
2018-08-27 15:04:45 +00:00
50219d6a99
Factor out waiting for current work.
2018-08-27 15:04:32 +00:00
6bba4bb671
Merge two similar branches.
2018-08-27 15:04:19 +00:00
d2984e1fab
Deduplicate callout destruction.
2018-08-27 15:03:59 +00:00
450ff8bbdc
Deduplicate callout initialization.
2018-08-27 15:03:45 +00:00
a1f195dae7
Handle a bazillion additional cases I hadn't considered.
...
There's a lot of duplication here. To be deduplicated in a
subsequent commit.
2018-08-27 15:03:32 +00:00
6810725695
Fix semantics of flush_work and flush_delayed_work.
...
- Change return type to void.
=> Upstream it is bool, but exactly one of hundreds of callers
actually use it, and I don't think the semantics is clear.
- Make sure to wait for whichever of the current work _and_ the next
batch queued is currently there in the workqueue.
- Don't retry a cancelled callout. Cancellation in the state
DELAYED_WORK_CANCELLED is guaranteed.
2018-08-27 15:03:20 +00:00
9cb8c351bd
Reverse the sense of some branches for simplicity.
2018-08-27 15:03:07 +00:00
5f1b90f39a
Fix bugs in workqueue destruction.
2018-08-27 15:02:52 +00:00
7a09cf8a76
Fully initialize workqueue struct. Zalloc for paranoia.
2018-08-27 15:02:38 +00:00
551d54d11f
No need to retry in cancel_delayed_work_sync after all.
2018-08-27 15:02:19 +00:00
535de6e400
Deduplicate post-cancellation logic.
2018-08-27 15:02:08 +00:00
5891db5c6e
Don't touch a work after it is released.
2018-08-27 15:01:47 +00:00
01334e6900
Non-waiting cancellation may need to release work too.
2018-08-27 15:01:13 +00:00
61b3d58635
After cancel and callout_halt, may need to reload workqueue.
2018-08-27 15:00:57 +00:00
5638bd9cfa
Don't wait forever to flush if the worker thread has no work.
2018-08-27 15:00:27 +00:00
6c0c863ee4
Sprinkle some assertions.
2018-08-27 14:59:58 +00:00
7d2f7a0702
Exercise greater care in acquiring and releasing work.
2018-08-27 14:59:20 +00:00
2e56c8b44a
Check ifp->if_capenable, not ifp->if_capabilities, when enable some features.
...
Not a real problem since ifp->if_capabilities = 0 at the moment.
2018-08-27 14:59:04 +00:00
b8dd8a9b3f
Add missing break.
2018-08-27 14:58:57 +00:00
ddd95367de
Add missing namespace for flush_scheduled_work.
2018-08-27 14:58:40 +00:00
dd8f93d733
Add KASSERT about state after callout fires.
2018-08-27 14:58:24 +00:00
ac596c7fca
Work is not necessarily queued in flush_work either.
...
Regularize style: use initializer only if the wq is guaranteed, and
confirmed by KASSERT, to be set
2018-08-27 14:58:09 +00:00
43c369bcf4
Work is not necessarily queued in cancel_work.
2018-08-27 14:57:55 +00:00
9ee469e1db
Draft rewrite of Linux workqueue reimplementation.
...
Just use an explicit thread; don't attempt to fudge it with
workqueue(9). No doubt there are various mistakes in here, but they
should be easier to get right than the mega-kludgerific nonsense that
preceded this draft.
2018-08-27 14:57:21 +00:00
ae1f9ce6c8
Add missing -z flag to usage()
2018-08-27 14:55:46 +00:00
f3083eb260
Mark some NetBSDisms as such.
2018-08-27 14:54:32 +00:00
45e24dc2e9
Read from the data register, not the index we just wrote.
2018-08-27 14:54:08 +00:00
126550ec16
Scale offset by size of unit, gen6_pte_t.
2018-08-27 14:53:54 +00:00
riastradh