| Age | Commit message (Collapse) | Author |
|---|
|
|
|
ok hugh@
|
|
<machine/param.h>, which was not the case for these arches.
|
|
|
|
boolean_t pmap_extract(struct pmap *, vaddr_t, paddr_t *).
Matches NetBSD. Tested by various people on various platforms.
|
|
> date: 2000/05/08 18:51:17; author: ragge; state: Exp; lines: +6 -2
> branches: 1.4.2;
> Get correct MAC address out of rom on VAX 4000/200. From Michael Kukat.
|
|
okay art@ miod@
|
|
Discussed with pjanzen@
|
|
Also contemplate the curious clr device, wherefor art thine code?
|
|
supported framebuffer (simple monochrome) can finish this off.
If anyone has run across docs for LCG or SPGFX, please get in touch.
|
|
changes. Mostly from NetBSD.
|
|
kern_ipc_10.c for other compat modules.
|
|
|
|
|
|
|
|
|
|
- Change pmap_change_wiring to pmap_unwire because it's only called that way.
- Remove pmap_pageable because it's seldom implemented and when it is, it's
either almost useless or incorrect. The same information is already passed
to the pmap anyway by pmap_enter and pmap_unwire.
|
|
And remove that memory price comment from 1981. It is amusing, but also
confusing because the math in there is only correct on vax.
|
|
Move it from cpu_startup() to main().
|
|
We might want to use them on types that are bigger than vaddr_t.
Fix all callers that pass pointers without casts.
|
|
CLSIZE -> 1
CLBYTES -> PAGE_SIZE
OLOFSET -> PAGE_MASK
etc.
At the same time some archs needed some cleaning in vmparam.h so that
goes in at the same time.
|
|
|
|
MI RealTek network adaptor. Will resurrect or rename rl at such time as
we actually support these drives.
|
|
difference between the BGTR and foo: is too large for a byte displacement.
as should give an error or at least a warning here, but it doesn't;
instead it merrily outputs a completely bogus displacement.
This fixes problems with EMODD on numbers with negative exponents.
|
|
boot to halt on boot attempts after a failed network boot attempt.
Also, prototype netmountroot() and return 0 in netclose().
|
|
1GB i386 machines needs this. The fix is heavily based on Jason Thorpe's
found in NetBSD. Here is his original commit message:
Instead of checking vm_physmem[<physseg>].pgs to determine if
uvm_page_init() has completed, add a boolean uvm.page_init_done,
and test against that. Use this same boolean (rather than
pmap_initialized) in pmap_growkernel() to determine if we are
being called via uvm_page_init() to grow the kernel address space.
This fixes a problem on some i386 configurations where pmap_init()
itself was needing to have the kernel page table grown, and since
pmap_initialized was not yet set to TRUE, pmap_growkernel() was
choosing the wrong code path.
|
|
|
|
|
|
|
|
right thing if booting off something other than sd0a.
RB_ASKNAME will be reimplemented later.
|
|
Also a large quantity of whitespace fixes.
|
|
|
|
could test this in the next snapshot.
|
|
: date: 2000/04/28 00:08:51; author: matt; state: Exp; lines: +2 -2
: Dump NRCV to 8 from 5. Note that a 8K NFS reply takes 6 full size
: Ethernet packets so the chances are that if the server was fast enough
: this driver would continuously lose one of the fragments. Eventually
: causing the kernel load to fail. Add 2 extra for random ARP or other
: packets that might be picked up.
|
|
Ambiguity is because DEC's docs differ from observed values.
Only cosmetic, at any rate.
|
|
Entries for KA681 and KA691, and a conservative default case to keep
things sane if an unknown cpu is encountered by me. Also some KNF.
Anyone with a 4000/{4,5,6}00 able to test?
|
|
- Using CVTDL and subtracting to seperate integer from fraction does
not work if the integer is >32 bits long; instead, rearrange the bits
into a quadword, use ASHQ to truncate, and then subtract.
Also:
- Set the condition codes properly; this fixes the other problem with
>32-bit-integer parts in libm by letting modf subtract the fractional
part (which *is* a double) to get the integral part in a double.
- move the zero checks earlier in the routine
- instead of 'ret' in zeroexit, use brw goback
|
|
Won't try to match VAX_STYP_692 until we can find a 4000 70x and learn
more about this model.
|
|
This fixes bcopy, copyin/copyout, and kcopy. Much testing was done by hugh@
and me to find the fastest replacement out of half a dozen candidates.
Also, correct register masks in numerous places.
|
|
|
|
|
|
|
|
VS4000/vlc or MV3100/{3,4}0. This seems to produce more consistent
detection than checking the configuration and test register.
|
|
It's not yet clear which is better used to differentiate these models.
|
|
|
|
|
|
|
|
|
|
sigcode to call a handler, SP was set in the middle of the struct
trapframe/struct sigcontext pair placed on the stack. This was adequate
when only one signal is being delivered at a time, but when two signals are
delivered in one syscall, the second invocation of sendsig() overwrote
some of the first sendsig() invocation's stack data.
The result was that under certain circumstances, (e.g., sending signals to
backgrounded processes from ksh) the argument passed to signal handlers
when the second signal (the one corresponding to the first sendsig()
invocation) was delivered was bogus data from the stack. Therefore,
programs that depended on the "sigraised" argument to be accurate
blew up.
This fix ensures that sendsig() always leaves a process's SP out of the way
of structures that will be used. AP is not moved, though, so sigreturn
can find the args correctly. Also, sigcode is changed to compensate for the
SP change.
|
|
High kludge factor, but the payoff is vsbus scsi for free.
Most of the work is by Michael Kukat and ragge, with a few other
synchronisations thrown in.
|
