Commit Graph

79 Commits

Author SHA1 Message Date
simonb 21d2b8b53d We don't need to include <uvm/uvm_extern.h> before <sys/sysctl.h> anymore. 2002-03-17 11:10:43 +00:00
lukem adc783d537 add RCSIDs 2001-11-12 15:25:01 +00:00
enami 163c9dd7c1 Defopt CALLWHEEL_STATS. 2001-09-13 05:22:16 +00:00
thorpej 16c229ea7c Optimization suggested by Bill Sommerfeld: Keep a hint as to the
"earliest" firing callout in a bucket.  This allows us to skip
the scan up the bucket if no callouts are due in the bucket.

A cheap O(1) hint update is done at callout insertion (if new callout
is earlier than hint) and removal (is bucket empty).  A thorough
refresh of the hint is done when the bucket is traversed.

This doesn't matter much on machines with small values of hz
(e.g. i386), but on systems with large values of hz (e.g. Alpha),
it has a definite positive effect.

Also, keep the callwheel stats in evcnts, so that you can view them
with "vmstat -e".
2001-09-11 04:32:19 +00:00
simonb cbbd901bdc Declare schedhz. 2001-05-06 13:46:34 +00:00
thorpej 2f89e3d744 Explicitly include <machine/intr.h> if __HAVE_GENERIC_SOFT_INTERRUPTS. 2001-01-17 18:21:41 +00:00
thorpej d74e432ed3 Make softclock a generic soft interrupt of the API is available,
adding the requisite void * argument to softclock().
2001-01-15 20:19:50 +00:00
mycroft 66610a4779 Introduce PROC_PC(), which is used to get a process's user PC. If this is
defined, call addupc_intr() directly from statclock() in the system time case,
using the same P_OWEUPC path if the copyin/copyout fails.
Use this in i386 to remove profiling code from the normal userret() path.
2000-12-10 19:29:30 +00:00
sommerfeld 340951f9d1 On second thought.. pass cpu_info * to roundrobin() explicitly. 2000-08-26 04:01:16 +00:00
sommerfeld ec08310fab More MP clock/scheduler changes:
- Periodically invoke roundrobin() from hardclock() on all cpu's rather
than from a timer callout; this allows time-slicing on non-primary cpu's.
 - Make pscnt per-cpu.
 - Notice psdiv changes on each cpu, and adjust pscnt at that point.
Also, invoke setstatclockrate() from the clock interrupt when each cpu
notices the divisor change, rather than when starting/stopping the
profiling clock.
2000-08-26 03:34:36 +00:00
thorpej f759220f40 Define the MI parts of the "big kernel lock" perimeter. From
Bill Sommerfeld.
2000-08-22 17:28:28 +00:00
eeh 3787c3f7fd Should use an `intrptr_t' for address calculations rather than `int'. 2000-08-22 16:44:51 +00:00
thorpej 25fe521af4 Fix a locking glitch in callwheel_slock handling. Noted by Bill Sommerfeld. 2000-08-22 15:30:59 +00:00
thorpej 14c0be9cd4 Protect hardclock_ticks and softclock_ticks with the callwheel
lock to prevent a race between hardclock() and callout_reset().
2000-08-21 23:51:33 +00:00
thorpej b7e86fa7a8 spllowersoftclock() is already void; no need to cast it. 2000-08-21 23:43:30 +00:00
thorpej 012500bf1f Add a lock for the callwheel (callout facility), and only go to
splclock() while holding it.
2000-08-21 23:40:56 +00:00
thorpej cd32ace8bb ANSI'ify. 2000-08-01 04:57:28 +00:00
thorpej c0c8481a2a New hzto() function from FreeBSD and Artur Grabowski <art@stacken.kth.se>.
Stops sleeps from returning early (by up to a clock tick), and return 0
ticks for timeouts that should happen now or in the past.

Returning 0 is different from the legacy hzto() interface, and callers
need to check for it.
2000-07-13 17:06:15 +00:00
mrg 32aa199ccf remove include of <vm/vm.h> 2000-06-27 17:41:07 +00:00
thorpej 5b281c5932 Move schedticks and cp_time into schedstate_percpu. Also, allow
non-primary CPUs to call hardclock(), but make them bail about
before updating global timekeeping state (that's the job of the
primary CPU).
2000-06-03 20:42:42 +00:00
simonb 38cc1b3975 Add new sysctl node "KERN_SYSVIPC_INFO" with "KERN_SYSVIPC_MSG_INFO",
"KERN_SYSVIPC_SEM_INFO" and "KERN_SYSVIPC_SHM_INFO" to return the
info and data structures for the relevent SysV IPC types.  The return
structures use fixed-size types and should be compat32 safe.  All
user-visible changes are protected with
	#if !defined(_POSIX_C_SOURCE) && !defined(_XOPEN_SOURCE)

Make all variable declarations extern in msg.h, sem.h and shm.h and
add relevent variable declarations to sysv_*.c and remove unneeded
header files from those .c files.

Make compat14 SysV IPC conversion functions and sysctl_file() static.

Change the data pointer to "void *" in sysctl_clockrate(),
sysctl_ntptime(), sysctl_file() and sysctl_doeproc().
2000-06-02 15:53:03 +00:00
mycroft da42c608fe Use a better multiplier for the 60Hz case. 2000-05-29 23:48:33 +00:00
mycroft 7513b8e18d Update an outdated comment.
Allow all powers of 2 from 2^0 to 2^16 for hz.
Enable hz==1200.
2000-05-29 15:05:10 +00:00
mycroft 8dcf08ff77 Improve the time_adj multiplier for the 100Hz and 1000Hz cases, and add a
1200Hz case.
2000-05-29 14:58:59 +00:00
augustss 264f1d27c6 Get rid of register declarations. 2000-03-30 09:27:11 +00:00
enami f9c7a69ff5 Call the routine to calculate callwheelsize from allocsys() instead of
main() since some port like alpha and mips calls allocsys() before main()
is called.  While I'm here, I renamed some function.
2000-03-24 11:57:14 +00:00
thorpej 2b58edac40 Remove the CALLWHEEL_SORT code. It was implemented just for experimenting,
and I had no plans to ever enable it.  A record of the code is now in the
CVS history of the file, so we can unclutter now.
2000-03-23 20:51:09 +00:00
thorpej b667a5a357 New callout mechanism with two major improvements over the old
timeout()/untimeout() API:
- Clients supply callout handle storage, thus eliminating problems of
  resource allocation.
- Insertion and removal of callouts is constant time, important as
  this facility is used quite a lot in the kernel.

The old timeout()/untimeout() API has been removed from the kernel.
2000-03-23 06:30:07 +00:00
thorpej a0397a2573 Move callout initialization to a single location; no need to duplicate
that code all over the place.
2000-01-19 20:05:30 +00:00
sommerfeld c450ebbbe7 If using kernel PLL (for NTP), initialize "fixtick" to a reasonable
approximation of reality if the MD code doesn't.  This variable is the
equivalent of "tickfix" for the non-NTP path.

This allows an alpha kernel (where hz=1024) with "options NTP" to
synch up quite nicely (as opposed to having an frequency error of
~560ppm, which is outside the capture range of the PLL).
1999-09-06 20:44:02 +00:00
thorpej eb20bbc780 Change the semantics of splsoftclock() to be like other spl*() functions,
that is priority is rasied.  Add a new spllowersoftclock() to provide the
atomic drop-to-softclock semantics that the old splsoftclock() provided,
and update calls accordingly.

This fixes a problem with using the "rnd" pseudo-device from within
interrupt context to extract random data (e.g. from within the softnet
interrupt) where doing so would incorrectly unblock interrupts (causing
all sorts of lossage).

XXX 4 platforms do not have priority-raising capability: newsmips, sparc,
XXX sparc64, and VAX.  This platforms still have this bug until their
XXX spl*() functions are fixed.
1999-08-05 18:08:08 +00:00
christos a32f7169fc Align struct timeval time to the same alignment requirements of a quad.
This broke the sparc elf kernel which in microtime uses ldd to load both
words at the same time. The a.out kernel, just got lucky.
1999-05-04 16:16:54 +00:00
ross e47e3c9f45 schedclk() -> schedclock(), for consistency with hardclock(), statclock(), ...
update comments for recent scheduler mods
1999-02-28 18:14:57 +00:00
mycroft d4955ba8a9 While I'm on a fixed point kick, improve the NTP clock factor correction to
give <.1% error in all (supported) cases.  It doesn't cost us much.
1999-02-23 17:41:48 +00:00
ross b4a33c4e60 Scheduler bug fixes and reorganization
* fix the ancient nice(1) bug, where nice +20 processes incorrectly
  steal 10 - 20% of the CPU, (or even more depending on load average)
* provide a new schedclk() mechanism at a new clock at schedhz, so high
  platform hz values don't cause nice +0 processes to look like they are
  niced
* change the algorithm slightly, and reorganize the code a lot
* fix percent-CPU calculation bugs, and eliminate some no-op code

=== nice bug === Correctly divide the scheduler queues between niced and
compute-bound processes. The current nice weight of two (sort of, see
`algorithm change' below) neatly divides the USRPRI queues in half; this
should have been used to clip p_estcpu, instead of UCHAR_MAX.  Besides
being the wrong amount, clipping an unsigned char to UCHAR_MAX is a no-op,
and it was done after decay_cpu() which can only _reduce_ the value.  It
has to be kept <= NICE_WEIGHT * PRIO_MAX - PPQ or processes can
scheduler-penalize themselves onto the same queue as nice +20 processes.
(Or even a higher one.)

=== New schedclk() mechansism === Some platforms should be cutting down
stathz before hitting the scheduler, since the scheduler algorithm only
works right in the vicinity of 64 Hz. Rather than prescale hz, then scale
back and forth by 4 every time p_estcpu is touched (each occurance an
abstraction violation), use p_estcpu without scaling and require schedhz
to be generated directly at the right frequency. Use a default stathz (well,
actually, profhz) / 4, so nothing changes unless a platform defines schedhz
and a new clock.  Define these for alpha, where hz==1024, and nice was
totally broke.

=== Algorithm change === The nice value used to be added to the
exponentially-decayed scheduler history value p_estcpu, in _addition_ to
be incorporated directly (with greater wieght) into the priority calculation.
At first glance, it appears to be a pointless increase of 1/8 the nice
effect (pri = p_estcpu/4 + nice*2), but it's actually at least 3x that
because it will ramp up linearly but be decayed only exponentially, thus
converging to an additional .75 nice for a loadaverage of one. I killed
this, it makes the behavior hard to control, almost impossible to analyze,
and the effect (~~nothing at for the first second, then somewhat increased
niceness after three seconds or more, depending on load average) pointless.

=== Other bugs === hz -> profhz in the p_pctcpu = f(p_cpticks) calcuation.
Collect scheduler functionality. Try to put each abstraction in just one
place.
1999-02-23 02:56:03 +00:00
jonathan 6c0abe64fc defopt NTP and PPS_SYNC, in preparation for adding PPS support. 1998-04-22 07:08:11 +00:00
ross 7516424fe6 Teach the NTP PLL how to lock when hz == 1000. 1998-01-31 10:42:11 +00:00
gwr 55f621803a Moved db_show_callout() to ddb/db_xxx.c 1997-05-21 19:55:45 +00:00
tls 02697a5d47 add case for 512Hz in NTP code 1997-05-05 19:25:26 +00:00
mycroft 85b2440284 Use splclock() to block time updates, not splhigh(). 1997-02-28 04:45:35 +00:00
cgd 008f5aedf9 apply patch from PR 2788 (from Dennis Ferguson <dennis@jnx.com>) to
more smoothly apply "tickfix"es microsecond deltas (when compensating
for clocks running at > 1000Hz).
1997-01-15 04:59:39 +00:00
cgd 99fb250b5c fix from PR 2787 (from Dennis Ferguson <dennis@jnx.com>): when adjtime
is running (and NTP is not enabled), the adjtime()-handling code clobbers
any tickfix that may be necessary for systems with clocks with frequency
greater than 1000Hz.
1997-01-15 04:27:35 +00:00
cgd 30ca7eaa8d clean up a few spaces vs. tabs bogons 1996-11-15 23:51:23 +00:00
abrown 062b2cdd6b Fix unused variable warning for an "int i" left over from the old disk
statistics code.
1996-07-24 16:13:04 +00:00
thorpej 0462d3063c Remove old-style disk instrumentation code. 1996-07-12 22:00:42 +00:00
briggs 0a55dbede5 Do not declare tickfixcnt ifdef NTP. 1996-06-09 04:51:03 +00:00
christos 4ef330b934 remove include of <sys/cpu.h> 1996-04-22 01:38:12 +00:00
christos 0babd67cbc Fix db_printf formats. 1996-03-30 22:23:18 +00:00
mycroft 6b1f642521 Calculate the equivalent of `SHIFT_HZ' at run time, in initclocks(). 1996-03-15 07:56:00 +00:00
mycroft 79bb0bd2c7 Move an assignment inside #ifdef NTP. 1996-03-08 06:27:30 +00:00