38d3366f0f
Fix type of `code' arg to sendsig().
1994-11-08 04:22:39 +00:00
f875a9efe6
Allocate pv_pages differently.
1994-11-08 01:17:19 +00:00
d254e95680
Garbage collect some variables which are no longer used.
1994-11-07 10:19:28 +00:00
21e911d911
Update for new autoconfig.
1994-11-07 09:03:48 +00:00
cfae52f542
It's legitimate, if foolish, to mask SIGCONT.
1994-11-07 05:57:28 +00:00
297c06de67
Comment previous change.
1994-11-07 05:32:50 +00:00
86a2079aef
When restarting syscalls, make sure we adjust the PC by the correct amount
...
depending on what instruction we used to enter the kernel. Push the
instruction length into tf_err on entry.
1994-11-07 05:26:08 +00:00
ed993b3930
Two trivial optimizations.
1994-11-07 05:09:04 +00:00
d5f99262a9
Since npxdna() and npxsave() are only ever called at spl0 now, we can get rid
...
of all the overhead of frobbing the IDT, and just use a simple flag to tell
npxintr() to ignore interrupts. This allows lots of code reduction.
Also, update some comments and add some DIAGNOSTIC code to detect is these
assumptions ever become wrong.
1994-11-07 03:39:37 +00:00
5133a6cfd7
Only save the FPU state if somebody else needs the FPU, or when forking.
1994-11-06 23:43:17 +00:00
f3ef31e4aa
Oops; forgot to adjust stack offset.
1994-11-06 22:05:08 +00:00
37863af7bb
Forgot two underscores.
1994-11-06 21:51:47 +00:00
d4510670f1
fix for bug 559, as reported by Jason Thorpe. (fix by me)
1994-11-06 21:49:08 +00:00
58b3010e87
Use %eax slightly more effectively in cpu_switch().
1994-11-06 20:51:46 +00:00
2f38797073
Trivial changes to make the code more consistent.
1994-11-06 20:39:50 +00:00
915336b97c
Approximate the divide by multiplying by 3433/4096. This is faster than
...
the previous code, and more accurate. Also avoid saving and restoring
registers.
(The multiplication algorithm was generated by GCC.)
1994-11-06 20:33:35 +00:00
ddb6e42460
Actually, don't bother checking p == 0 at all.
1994-11-06 20:30:09 +00:00
5cf38ebdb4
The case of curproc == 0 `can't happen' unless we're in the idle loop, so
...
allow it to panic by way of a reference to 0 in other cases. Explicitly test
for p == 0 in the specific cases where it's needed, and only calculate pcb if
needed.
1994-11-06 20:28:57 +00:00
9a84611b06
Move the spl0() from the idle loop to the top of cpu_switch(). (This also
...
avoids needing to call spllower() below.) Set curproc immediately after
removing the new process for the run queue, before allowing interrupts again,
for the benefit of schedcpu().
1994-11-06 20:21:29 +00:00
205d425f2f
As pointed out by Bruce, splsoftclock() needs to process pending interrupts
...
when lowering the priority. Do this by simply calling splx(). For the
benefit of NFS (for now), modify splx() to return a value so that
splsoftclock() does. This work is optimized out for every use of splx()
that ignores the value, so it only affects relevant code.
Also, clean up this file a little.
1994-11-06 01:37:47 +00:00
4cd7c0f34e
reset scsi bus at attach time
1994-11-05 09:35:54 +00:00
880176b15b
proc0.p_addr must be initialized before we can take a probe trap, so
...
that the pcb_onfault will be work
1994-11-05 09:31:05 +00:00
0d4b376377
only print the vec if needed
1994-11-05 09:28:06 +00:00
c2191a2ad0
Fully prototype ssdtosd() and sdtossd().
1994-11-05 03:20:36 +00:00
34c4570d61
Only use ssdtosd() when creating, expanding, or copying the LDT. Keep a copy
...
of the result in the PCB so it can be loaded quickly on context switches.
1994-11-05 03:17:33 +00:00
867ac5e15f
Add sdtossd() by HOSOKAWA Tatsumi, and a C version of ssdtosd() by me.
1994-11-05 02:47:02 +00:00
600aa2c5f8
Trivial optimizations. Also, make sure we get the overflow correct if
...
called at spl >= clock.
1994-11-05 02:16:26 +00:00
2a9b6f7959
Fix stupid bug in last change.
1994-11-04 23:30:15 +00:00
4ec6774e12
Add a simplified version of wdcommand() that can be used in some cases.
...
Avoid needing a wd_softc inside wdprobe().
1994-11-04 23:18:06 +00:00
38420d7db1
Don't need cd_indirect.
1994-11-04 22:51:59 +00:00
3df6bf0c0e
Trivial stylistic changes.
1994-11-04 21:32:23 +00:00
098c6e36e6
Various cleanup. In particular, be much more careful with buffer flags.
1994-11-04 21:19:48 +00:00
e11ffbc64b
Fix a call to math_abort().
1994-11-04 20:47:07 +00:00
854d8d96dc
Rearrange slightly.
1994-11-04 19:25:34 +00:00
cfc4633919
Remove outdated comment.
1994-11-04 19:17:47 +00:00
1ec8f8d19c
Make IRQ_SLAVE a plain number, too, and remove IRQ{0-15}.
1994-11-04 19:13:49 +00:00
95f0ee6cd5
Remove icu.h.
1994-11-04 19:11:13 +00:00
2252d31c99
Remove use of IRQ0.
1994-11-04 19:10:46 +00:00
8d01374523
Don't need icu.h any more.
1994-11-04 19:08:19 +00:00
b7acf321cd
No longer needs icu.h.
1994-11-04 18:59:34 +00:00
195574a7de
Change all of the uses of IRQ{0-15} and bit masks to use plain numbers
...
rather than bit shifting.
1994-11-04 18:57:15 +00:00
d66f8d7cdb
Clean up deleted files.
1994-11-04 18:54:43 +00:00
29414c5a0a
Change all of the uses of IRQ{0-15} and bit masks to use plain numbers
...
rather than bit shifting.
1994-11-04 18:34:50 +00:00
6cd3ad1d11
ulimit builtin (PR #388 )
...
This public domain code, originally by Doug Gwyn, Doug Kingston, Eric
Gisin, and Michael Rendell was ripped from pdksh 5.0.8 and hacked for
use with ash.
1994-11-04 17:13:27 +00:00
bf98208622
Actually pass all the arguments to amd that we meant to. Reported by
...
Dave Leonard (D.Leonard@pleiades.newcastle.edu.au ).
1994-11-04 13:27:32 +00:00
6d9009aacb
bugs don't get any dumber than this.
1994-11-04 09:59:07 +00:00
0ba30791b5
Remove references to isa_devtab.
1994-11-04 09:57:52 +00:00
f8183b2d0e
Make a wrapper match function to check the bus and device numbers, rather
...
than insisting that every driver do it.
1994-11-04 09:42:18 +00:00
9af4fee97d
Use kvtop() rather than address arithmetic in a couple of places.
1994-11-04 08:55:53 +00:00
a980084e46
EISA devices have no DRQ.
1994-11-04 08:24:25 +00:00
mycroft