guts were actually functionally equivalent to the current guts, but were
much larger, filled with bugs, and indeed poked around at the disklabel
when some of those bugs prevented them from ever using the disklabel!
had a few bugs fixed that let the problem slip in, and since bootxx's
Makefile now goes out of its way to satisfy installboot's undocumented
and totally unreasonable assumptions about the bootxx file it's operating
on. No point in fixing the assumptions, because sooner rather than later
this incarnation of installboot is going to die.
assuming that there's always going to be space for the whole boot
block info struct. (the assumption would cause a malloc'd region
to be overrun, if it proved false.)
call to __main(), and therefore saves the size of the call and the
size of a stub implementation of __main().
in the primary boot block, don't bother saving/restoring the argument
passed in from the caller. There is no such argument (that we care
about, at least) to the primary. (for secondary, it's the firmware
FD being used.)
Clean up the "Region 1" related definitions, and define load addresses,
max load size, and max total size for as many boot block types as we can.
(types = unified, primary, secondary). We can't always define all
values for all boot blocks, though.
Make CPP flags selection less gross.
Use objcopy rather than headersize (yay, evil gets a stake to the heart!).
Use a little shell script to verify that the sizes of the boot blocks are OK.
Do not compile too much more of libsa than we actually have to.
* Map the message buffer with access_type = VM_PROT_READ|VM_PROT_WRITE `just
because'.
* Map the file system buffers with access_type = VM_PROT_READ|VM_PROT_WRITE to
avoid possible problems with pagemove().
* Do not use VM_PROT_EXEC with either of the above.
* Map pages for /dev/mem with access_type = prot. Also, DO NOT use
pmap_kenter() for this, as we DO NOT want to lose modification information.
* Map pages in dumpsys() with VM_PROT_READ.
* Map pages in m68k mappedcopyin()/mappedcopyout() and writeback() with
access_type = prot.
* For now, bus_dma*(), pmap_map(), vmapbuf(), and similar functions still use
access_type = 0. This should probably be revisited.
allocated from a pool, and the MIPS and Alpha use KSEG to map pool
pages. So, mb_map wasn't actually being used. Saves around 4MB of
kernel virtual address space in a typical configuration.
Garbage-collect the related VM_MBUF_SIZE constant.
"BUS_SPACE_ALIGNED_POINTER()".
Equal to the param.h "ALIGNED_POINTER()" normally, but obeys additional
requirements of the bus_space_xxx_n() macros. (BUS_SPACE_DEBUG)
by two processors concurrently. This means that we cannot modify the
lev1map in use by the processor which wishes to use the PROM.
Fix this by creating a separate lev1map for PROM users. This lev1map
is a copy of the kernel_lev1map, with the exception of the necessary
PROM mapping. When a processor wishes to use the PROM, it switches
its PTBR to point at the prom_lev1map, performs the PROM operation,
and switches back to its previous lev1map.
Note that kernels without multiprocessor support use the old method
of modifying the current lev1map.
Also, serialize access to the PROM via a spin lock.
additional processors are spun up on multiprocessor Alpha systems.
Now, each processor gets its own idle thread (the primary processor
uses proc0). This idle thread is used in switch_exit(), rather than
explicitly referencing proc0.
Also, make `curproc', `fpcurproc', and `curpcb' per-cpu values. This
required some data structure rearrangement; cpu info is now statically
allocated in the BSS, rather than via malloc(), and cpu_softc is gone.
(Modeled somewhat after NetBSD/sparc's multiprocessor info structures.)
* 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.
* PCI VGA (tested with Imagine-128)
* ISA SoundBlaster (tested with SB1!)
* AudioPCI (tested with ... AudioPCI!)
* PnP SoundBlaster (currently loses in 16-bit mode due to extent lossage)
* SonicVibes (currently loses; see sv.c commit log)
cgd