negative-logic flags (flags 0x01 and 0x02 for npx0, defaulting to unset = on).
This changes the default from off to on. The options have been in current
for several months with no problems reported.
Added a boot-time negative-logic flag for the old I5886_FAST_BCOPY option
which went away too soon (flag 0x04 for npx0, defaulting to unset = on).
Added a boot-time way to set the memory size (iosiz in config, iosize in
userconfig for npx0).
LINT:
Removed old options. Documented npx0's flags and iosiz.
options.i386:
Removed old options.
identcpu.c:
Don't set the function pointers here. Setting them has to be delayed
until after userconfig has had a chance to disable them and until after
a good npx0 has been detected.
machdep.c:
Use npx0's iosize instead of MAXMEM if it is nonzero.
support.s:
Added vectors and glue code for copyin() and copyout().
Fixed ifdefs for i586_bzero().
Added ifdefs for i586_bcopy().
npx.c:
Set the function pointers here.
Clear hw_float when an npx exists but is too broken to use.
Restored style from a year or three ago in npxattach().
if I586_CPU is defined. Note there is a runtime check so the code
won't be run for non-Pentium CPUs anyway.
2.2 candidate, this code has been tested for almost half year in -current.
handlers if interrupts are nested more than a few (3) deep. This
only reduces the maximum nesting level by 1 with the standard
drivers unless there is a related bug somewhere, but can't hurt
much (the worst case is returning to hoggish interrupt handler like
wdintr(), but such interrupt handlers hurt anyway).
Fixed a previously harmless race incrementing the interrupt nesting
level.
This should be in 2.1.6 and 2.2.
hardware interrupt counts add up to the total. Previously, software
interrupts generated by splz() were counted in the total. These
software interrupts seem to be very rare - there have apparently been
0 of them on freefall among the last 352448857 interrupts.
mappings with mlock. This problem only occurred because of the
quick unmap code not respecting the wired-ness of pages in the
process. In the future, we need to eliminate the dependency
intrinsic to the design of the code that wired pages actually
be mapped. It is kind-of bogus not to have wired pages mapped,
but it is also a weakness for the code to fall flat because
of a missing page.
This show fix a problem that Tor Egge has been having, and also
should be included into 2.2-RELEASE.
(1) deleted #if 0
pc98/pc98/mse.c
(2) hold per-unit I/O ports in ed_softc
pc98/pc98/if_ed.c
pc98/pc98/if_ed98.h
(3) merge more files by segregating changes into headers.
new file (moved from pc98/pc98):
i386/isa/aic_98.h
deleted:
well, it's already in the commit message so I won't repeat the
long list here ;)
Submitted by: The FreeBSD(98) Development Team
AHC_FORCE_PIO - This forces the driver to use PIO even on systems that
say they have memory mapped the controller's registers. This
seems to fix Ken Lam's problems. I've also placed this option
in the GENERIC kernel file so that we are guaranteed to install
even on these flakey machines.
AHC_SHARE_SCBS - This option attempts to share the external SCB SRAM on
the 398X controllers allowing a totoll of 255 non-paged SCBs.
This doesn't work quite yet, so this option is mostly here to
help 398X owners to experiment and give me feedback until this
works properly.
I decided to do this for every hardclock() call instead of lazily
in microtime(). The lazy method is simpler but has more overhead
if microtime() is called a lot.
CPU_THISTICKLEN() is now a no-op and should probably go away.
Previously it did nothing directly but had the side effect of
setting i586_last_tick for CPU_CLOCKUPDATE() and i586_avg_tick for
debugging. CPU_CLOCKUPDATE() now uses a better method and
i586_avg_tick is too much trouble to maintain.
Reduced nesting of #includes in the usual case.
Increased nesting of #includes when CLOCK_HAIR is defined. This
is a kludge to get typedefs for inline functions only when the
inline functions are used. Normally only kern_clock.c defines
this. kern_clock.c can't include the i386 headers directly.
Removed unused LOCORE support.
- use a more accurate and more efficient method of compensating for
overheads. The old method counted too much time against leaf
functions.
- normally use the Pentium timestamp counter if available.
On Pentiums, the times are now accurate to within a couple of cpu
clock cycles per function call in the (unlikely) event that there
are no cache misses in or caused by the profiling code.
- optionally use an arbitrary Pentium event counter if available.
- optionally regress to using the i8254 counter.
- scaled the i8254 counter by a factor of 128. Now the i8254 counters
overflow slightly faster than the TSC counters for a 150MHz Pentium :-)
(after about 16 seconds). This is to avoid fractional overheads.
files.i386:
permon.c temporarily has to be classified as a profiling-routine
because a couple of functions in it may be called from profiling code.
options.i386:
- I586_CTR_GUPROF is currently unused (oops).
- I586_PMC_GUPROF should be something like 0x70000 to enable (but not
use unless prof_machdep.c is changed) support for Pentium event
counters. 7 is a control mode and the counter number 0 is somewhere
in the 0000 bits (see perfmon.h for the encoding).
profile.h:
- added declarations.
- cleaned up separation of user mode declarations.
prof_machdep.c:
Mostly clock-select changes. The default clock can be changed by
editing kmem. There should be a sysctl for this.
subr_prof.c:
- added copyright.
- calibrate overheads for the new method.
- documented new method.
- fixed races and and machine dependencies in start/stop code.
mcount.c:
Use the new overhead compensation method.
gmon.h:
- changed GPROF4 counter type from unsigned to int. Oops, this should
be machine-dependent and/or int32_t.
- reorganized overhead counters.
Submitted by: Pentium event counter changes mostly by wollman
- use CROSSJUMP() and CROSSJUMP_LABEL() for conditional jumps from idle()
into cpu_switch() and vice versa.
- moved badsw code to after cpu_switch().
Cosmetic changes:
- moved sw0 string to be immediately after its caller (badsw).
- removed unused #include.
previous snap. Specifically, kern_exit and kern_exec now makes a
call into the pmap module to do a very fast removal of pages from the
address space. Additionally, the pmap module now updates the PG_MAPPED
and PG_WRITABLE flags. This is an optional optimization, but helpful
on the X86.
- fixed a sloppy common-style declaration.
- removed an unused macro.
- moved once-used macros to the one file where they are used.
- removed unused forward struct declarations.
- removed __pure.
- declared inline functions as inline in their prototype as well
as in theire definition (gcc unfortunately allows the prototype
to be inconsistent).
- staticized.
<net/if_arp.h> and fixed the things that depended on it. The nested
include just allowed unportable programs to compile and made my
simple #include checking program report that networking code doesn't
need to include <sys/socket.h>.
I586_OPTIMIZED_BCOPY is configured.
Similarly for bzero/I586_OPTIMIZED_BZERO.
Fake 586's had better have a hardware FPU with non-broken exception
handling (we mask exceptions, but broken exception handling may trap
on the instructions that do the masking). I guess this means that
the routines won't work on most 386's or FPUless 486's even when they
have a h/w FPU.
These are based on using the FPU to do 64-bit stores. They also
use i586-optimized instruction ordering, i586-optimized cache
management and a couple of other tricks. They should work on any
i*86 with a h/w FPU, but are slower on at least i386's and i486's.
They come close to saturating the memory bus on i586's. bzero()
can maintain a 3-3-3-3 burst cycle to 66 MHz non-EDO main memory
on a P133 (but is too slow to keep up with a 2-2-2-2 burst cycle
for EDO - someone with EDO should fix this). bcopy() is several
cycles short of keeping up with a 3-3-3-3 cycle for writing. For
a P133 writing to 66 MHz main memory, it just manages an N-3-3-3,
3-3-3-3 pair of burst cycles, where N is typically 6.
The new routines are not used by default. They are always configured
and can be enabled at runtime using a debugger or an lkm to change
their function pointer, or at compile time using new options (see
another log message).
Removed old, dead i586_bzero() and i686_bzero(). Read-before-write is
usually bad for i586's. It doubles the memory traffic unless the data
is already cached, and data is (or should be) very rarely cached for
large bzero()s (the system should prefer uncached pages for cleaning),
and the amount of data handled by small bzero()s is relatively small
in the kernel.
Improved comments about overlapping copies.
Removed unused #include.
First, change sysinstall and the Makefile rules to not build the kernel
nlist directly into sysinstall now. Instead, spit it out as an ascii
file in /stand and parse it from sysinstall later. This solves the chicken-n-
egg problem of building sysinstall into the fsimage before BOOTMFS is built
and can have its symbols extracted. Now we generate the symbol file in
release.8.
Second, add Poul-Henning's USERCONFIG_BOOT changes. These have two
effects:
1. Userconfig is always entered, rather than only after a -c
(don't scream yet, it's not as bad as it sounds).
2. Userconfig reads a message string which can optionally be
written just past the boot blocks. This string "preloads"
the userconfig input buffer and is parsed as user input.
If the first command is not "USERCONFIG", userconfig will
treat this as an implied "quit" (which is why you don't need
to scream - you never even know you went through userconfig
and back out again if you don't specifically ask for it),
otherwise it will read and execute the following commands
until a "quit" is seen or the end is reached, in which case
the normal userconfig command prompt will then be presented.
How to create your own startup sequences, using any boot.flp image
from the next snap forward (not yet, but soon):
% dd of=/dev/rfd0 seek=1 bs=512 count=1 conv=sync <<WAKKA_WAKKA_DOO
USERCONFIG
irq ed0 10
iomem ed0 0xcc000
disable ed1
quit
WAKKA_WAKKA_DOO
Third, add an intro screen to UserConfig so that users aren't just thrown
into this strange screen if userconfig is auto-launched. The default
boot.flp startup sequence is now, in fact, this:
USERCONFIG
intro
visual
(Since visual never returns, we don't need a following "quit").
Submitted-By: phk & jkh
dependent operation, and not really a correct name. invltlb and invlpg
are more descriptive, and in the case of invlpg, a real opcode.
Additionally, fix the tlb management code for 386 machines.
lcall 7,0 (ie: ldt slot 0) and lcall 0x87,0 (ldt slot 16, it's shifted
three bits to the left). I was fiddling with this so long ago, I don't
recall the specifics.
with this quite a while ago when somebody reported a BSD/OS 2.1 binary
that wouldn't run. I'm pretty sure they tried it and I'm pretty sure
they mentioned to me that the patch worked.
comparisons in the inb() and outb() macros. I decided that int args
are OK here. Any type that can hold a u_int16_t without overflow
is correct, and 32-bit types are optimal.
Introduced a few tens of warnings (100 in LINT) for use of pessimized
(short) types for the port arg. Only a few drivers are affected by
this. u_short pessimizations aren't detected.
Added `__extension__' before the statement-expression in inb() so
that it can be compiled without warnings by gcc -pedantic.
the prototype.
Put the jump table for i486_bzero() in the data section. This
speeds up i486_bzero() a little on Pentiums without significantly
affecting its speed on 486's.
Don't waste time falling through 14 nop's to return from do1 in
i486_bzero().
Use fastmove() for counts >= 1024 (was > 1024). Cosmetic.
Fixed profiling of fastmove().
Restored meaningful labels from the pre-1.1 version in fastmove().
Local labels are evil.
Fixed (high resolution non-) profiling of __bb_init_func().
I maintain that it saves more power to simply "hlt" the CPU than to
spend tons of time trying to tell the APM bios to do the same.
In particular if you do it 100 times a second...
instead of 0 if there is no input.
syscons.c:
Added missing spl locking in sccncheckc(). Return the same value as
sccngetc() would. It is wrong for sccngetc() to return non-ASCII, but
stripping the non-ASCII bits doesn't help.
(1) Add PC98 support to apm_bios.h and ns16550.h, remove pc98/pc98/ic
(2) Move PC98 specific code out of cpufunc.h (to pc98.h)
(3) Let the boot subtrees look more alike
Submitted by: The FreeBSD(98) Development Team
<freebsd98-hackers@jp.freebsd.org>
modified. Pages that are removed by the pageout daemon were
the worst affected. Additionally, numerous minor cleanups,
including better handling of busy page table pages. This
commit fixes the worst of the pmap problems recently introduced.
biosextmem > 65536, but biosextmem is a 16-bit quantity so it is
guaranteed to be < 65536. Related cruft for biosbasemem was
mostly cleaned up in rev.1.26.
The default level works with minimal overhead, but one can also enable
full, efficient use of a 512K cache. (Parameters can be generated
to support arbitrary cache sizes also.)
returned by the RTC, use the bootblock supplied value. Also, map the
'stolen by BIOS' memory in the same manner as the ISA-hole memory, since
it is really an extenstion of the BIOS. This is necessary for 32-bit
BIOS functions such as APM support on laptops, and the loss of memory
for non-necessary functions seems to be at most 4k.
Reviewed by: phk
Obtained from: email conversation with jtk@atria.com
nearest .01 Mhz rather than simply truncating it downwards.
This hack makes this 89.999928 Mhz clock correctly round to the closer
90.00-MHz rather than 89.99-MHz:
> i586 clock: 89999928 Hz, i8254 clock: 1193152 Hz
> CPU: Pentium (90.00-MHz 586-class CPU)
Changed i586_ctr_bias from long long to u_int. Only the low 32 bits
are used now that microtime uses a multiplication to do the scaling.
Previously the high 32 bits had to match those of rdtsc() to prevent
overflow traps and invalid timeval adjustments.
early enough when the dump device is specified in the config file.
Removed stale comment about configuration root and swap devices.
Don't bother clearing dumplo when dumpdev is set to NODEV. Everything
is controlled by dumpdev.
Fixed the kern.dumpdev sysctl. Writes were handle bogusly.
problem with the 'shell scripts' was found, but there was a 'strange'
problem found with a 486 laptop that we could not find. This commit
backs the code back to 25-jul, and will be re-entered after the snapshot
in smaller (more easily tested) chunks.
performance issues.
1) The pmap module has had too many inlines, and so the
object file is simply bigger than it needs to be.
Some common code is also merged into subroutines.
2) Removal of some *evil* PHYS_TO_VM_PAGE macro calls.
Unfortunately, a few have needed to be added also.
The removal caused the need for more vm_page_lookups.
I added lookup hints to minimize the need for the
page table lookup operations.
3) Removal of some bogus performance improvements, that
mostly made the code more complex (tracking individual
page table page updates unnecessarily). Those improvements
actually hurt 386 processors perf (not that people who
worry about perf use 386 processors anymore :-)).
4) Changed pv queue manipulations/structures to be TAILQ's.
5) The pv queue code has had some performance problems since
day one. Some significant scalability issues are resolved
by threading the pv entries from the pmap AND the physical
address instead of just the physical address. This makes
certain pmap operations run much faster. This does
not affect most micro-benchmarks, but should help loaded system
performance *significantly*. DG helped and came up with most
of the solution for this one.
6) Most if not all pmap bit operations follow the pattern:
pmap_test_bit();
pmap_clear_bit();
That made for twice the necessary pv list traversal. The
pmap interface now supports only pmap_tc_bit type operations:
pmap_[test/clear]_modified, pmap_[test/clear]_referenced.
Additionally, the modified routine now takes a vm_page_t arg
instead of a phys address. This eliminates a PHYS_TO_VM_PAGE
operation.
7) Several rewrites of routines that contain redundant code to
use common routines, so that there is a greater likelihood of
keeping the cache footprint smaller.
make it more intelligible, improve the partially bogus locking, and
allow for a ``quick re-acquiration'' from a pending release of timer 0
that happened ``recently'', so it was not processed yet by clkintr().
This latter modification now finally allows to play XBoing over
pcaudio without losing sounds or getting complaints. ;-) (XBoing
opens/writes/closes the sound device all over the day.)
Correct locking for sysbeep().
Extensively (:-) reviewed by: bde
Finished the constant poisoning that was begun in rev.1.14. Consts
aren't very poisonous (or useful) unless -Wcast-qual is in CFLAGS,
and it isn't in the default CFLAGS.
type identification code out of machdep.c and into a new file of its
own. Hopefully other grot can be moved out of machdep.c as well
(by other people) into more descriptively-named files.
Testing with the high frequency of 20000 Hz (to find problems) only found
the problem that this frequency is too high for slow i386's.
Disable interrupts while setting the timer frequency. This was unnecessary
before rev.1.57 and forgotten in rev.1.57. The critical (i8254) interrupts
are disabled in another way at boot time but not in the sysctl to change
the frequency.
Fixed profiling of system times. It was pre-4.4Lite and didn't support
statclocks. System times were too small by a factor of 8.
Handle deferred profiling ticks the 4.4Lite way: use addupc_task() instead
of addupc(). Call addupc_task() directly instead of using the ADDUPC()
macro.
Removed vestigial support for PROFTIMER.
switch.s:
Removed addupc().
resourcevar.h:
Removed ADDUPC() and declarations of addupc().
cpu.h:
Updated a comment. i386's never were tahoe's, and the deferred profiling
tick became (possibly) multiple ticks in 4.4Lite.
Obtained from: mostly from NetBSD
multiplication in addupc() overflowed for addresses >= 256K, assuming
the usual profil(2) scale parameter of 0x8000. addupc() will go away
soon.
Submitted by: John Polstra <jdp@polstra.com>
is only used by the icu support modules and by a few drivers that know
too much about the icu (most only use it to convert `n' to `IRQn'). isa.h
is only used by ioconf.c and by a few drivers that know too much about
isa addresses (a few have to, because config is deficient).
enable flag instead of enable_intr() to restore it to its usual state.
getit() is only called from DELAY() so there is no point in optimising
its speed (this wasn't so clear when it was extern), and using
enable_intr() made it inconvenient to call DELAY() from probes that need
to run with interrupts disabled.
1) Make it much less likely to miss a wakeup in vm_page_free_wakeup
2) Create a new entry point into pmap: pmap_ts_referenced, eliminates
the need to scan the pv lists twice in many cases. Perhaps there
is alot more to do here to work on minimizing pv list manipulation
3) Minor improvements to vm_pageout including the use of pmap_ts_ref.
4) Major changes and code improvement to pmap. This code has had
several serious bugs in page table page manipulation. In order
to simplify the problem, and hopefully solve it for once and all,
page table pages are no longer "managed" with the pv list stuff.
Page table pages are only (mapped and held/wired) or
(free and unused) now. Page table pages are never inactive,
active or cached. These changes have probably fixed the
hold count problems, but if they haven't, then the code is
simpler anyway for future bugfixing.
5) The pmap code has been sorely in need of re-organization, and I
have taken a first (of probably many) steps. Please tell me
if you have any ideas.
All new code is "#ifdef PC98"ed so this should make no difference to
PC/AT (and its clones) users.
Ok'd by: core
Submitted by: FreeBSD(98) development team
It is called from copyin and copyout.
The new routine is conditioned on I586_CPU and I586_FAST_BCOPY, so you
need
options "I586_FAST_BCOPY"
(quotes essenstial) in your kernel config file.
Also, if you have other kernel types configured in your kernel, an
additional check to make sure it is running on a Pentium is inserted.
(It is not clear why it doesn't help on P6s, it may be just that the
Orion chipset doesn't prefetch as efficiently as Tritons and friends.)
Bruce can now hack this away. :)
alias `pcb_cr3' instead. That is still one alias too many, but is convenient
for me since I've replaced the tss in the pcb by a few scalar variables in
the pcb.
altfmt was abused to sometimes screw up the disassembly of the bytes
following unconditional jump instructions. Gas doesn't pad to a longword
boundary like the comment said - that is the programmer's responsibility.
queue in vm_fault.
Move the PG_BUSY in vm_fault to the correct place.
Remove redundant/unnecessary code in pmap.c.
Properly block on rundown of page table pages, if they are busy.
I think that the VM system is in pretty good shape now, and the following
individuals (among others, in no particular order) have helped with this
recent bunch of bugs, thanks! If I left anyone out, I apologize!
Stephen McKay, Stephen Hocking, Eric J. Chet, Dan O'Brien, James Raynard,
Marc Fournier.
some problems with the page-table page management code, since it can't
deal with the notion of page-table pages being paged out or in transit.
Also, clean up some stylistic issues per some suggestions from
Stephen McKay.
process during pmap_copy. This minimizes unnecessary swapping or creation of
swap space. If there is a hold_count flaw for page-table
pages, clear the page before freeing it to lessen the chance of a system
crash -- this is a robustness thing only, NOT a fix.
close some windows that are opened up by page table allocations. The
prefaulting code no longer uses hold counts, but now uses the busy
flag for synchronization.
ansi and traditional cpp.
The nesting rules of macros are different, which required some changes.
Use __CONCAT(x,y) instead of /**/.
Redo some comments to use /* */ rather than "# comment" because the ansi
cpp cares about those, and also cares about quote matching.
queue corruption problems, and to apply Gary Palmer's code cleanups.
David Greenman helped with these problems also. There is still
a hang problem using X in small memory machines.
comparing the PTD pointers, they needed to be masked by PG_FRAME, and
they weren't. Also, the "improved" non-386 code wasn't really an
improvement, so I simplified and fixed the code. This might have
caused some of the panics caused by the VM megacommit.
With this fix from Stephen, we are getting the target fork performance
that I have been trying to attain: P5-166, before the mega-commit: 700-800usecs,
after: 600usecs, with Stephen's fix: 500usecs!!! Also, this could be the
solution of some strange panic problems...
Reviewed by: dyson@freebsd.org
Submitted by: Stephen McKay <syssgm@devetir.qld.gov.au>
contributions or ideas from Stephen McKay <syssgm@devetir.qld.gov.au>,
Alan Cox <alc@cs.rice.edu>, David Greenman <davidg@freebsd.org> and me:
More usage of the TAILQ macros. Additional minor fix to queue.h.
Performance enhancements to the pageout daemon.
Addition of a wait in the case that the pageout daemon
has to run immediately.
Slightly modify the pageout algorithm.
Significant revamp of the pmap/fork code:
1) PTE's and UPAGES's are NO LONGER in the process's map.
2) PTE's and UPAGES's reside in their own objects.
3) TOTAL elimination of recursive page table pagefaults.
4) The page directory now resides in the PTE object.
5) Implemented pmap_copy, thereby speeding up fork time.
6) Changed the pv entries so that the head is a pointer
and not an entire entry.
7) Significant cleanup of pmap_protect, and pmap_remove.
8) Removed significant amounts of machine dependent
fork code from vm_glue. Pushed much of that code into
the machine dependent pmap module.
9) Support more completely the reuse of already zeroed
pages (Page table pages and page directories) as being
already zeroed.
Performance and code cleanups in vm_map:
1) Improved and simplified allocation of map entries.
2) Improved vm_map_copy code.
3) Corrected some minor problems in the simplify code.
Implemented splvm (combo of splbio and splimp.) The VM code now
seldom uses splhigh.
Improved the speed of and simplified kmem_malloc.
Minor mod to vm_fault to avoid using pre-zeroed pages in the case
of objects with backing objects along with the already
existant condition of having a vnode. (If there is a backing
object, there will likely be a COW... With a COW, it isn't
necessary to start with a pre-zeroed page.)
Minor reorg of source to perhaps improve locality of ref.
boundary, which means that it doesn't mark the start of the data
section (which is then inaccessible to the programmer ??).
Hopefully fixes recent locore reboot problems.
