Mostly from Hubert Feyrer <hubertf@netbsd.org>, with tweaks from me.
+ Tweak "NetBSD developers" section
- move non developers into separate "other contributers" section
- list everyone, even if they're a member of another group
such as core or portmasters
+ Update "releng members" for reality.
0xc0200000. Tidy up to remove dead comments and code.
Allow more than one L1 entry for the kernel space and use the 'spare'
memory below the kernel code for the initial page tables in the same
way that the iq80310 does.
inetd) started from rc can now be run under systrace by setting
${name}_systrace to a set of flags to be passed to systrace. Note
that policies need to generated before you do this.
inetd) started from rc can now be run under systrace by setting
${name}_systrace to a set of flags to be passed to systrace. Note
that policies need to generated before you do this.
now be displayed correctly (fixes part 1 of pr 17480)
* Prevent field manipulations on null fields, the fields have nothing in
them so they should not be manipulated.
in PR kern/17506):
If we reset the mouse and discover a different protocol after the reset
than before, reset it again after one second. This makes protocol detection
after a KVA switched to another machine more reliable and should not affect
the standard situation with one mouse connected to one machine.
BIOS calls can be used for device probing etc.
And now it's getting nasty:
The kvm86 code needs a TSS, and it is most convenient to use proc0's
instead of doing some static allocation. (We might reconsider this if
we want to use vm86 for console initialization, ie much earlier.)
For the TSS slot to be allocated, we have to move the call to
i386_proc0_tss_ldt_init() up.
Since the npx code twiddles CR0 the corresponding pcb field must
be synchronized later. It would probably be cleaner to do this in
the npx driver.
8086 machine. Ifff it works, it is much easier and more elegant than
going to real real mode:
-simpler code
-no need for "identity" memory mappings
-easy passing of buffers for bulk data to functions
-some more control
There is no interrupt support ATM, and it lacks a function to access
random virtual memory of the VM. MP issues to consider.
lukem