1) Start using TSM.
Struct procs continue to point to upages structure, after being freed.
Struct vmspace continues to point to pte object and kva space for kstack.
u_map is now superfluous.
2) vm_map's don't need to be reference counted. They always exist either
in the kernel or in a vmspace. The vmspaces are managed by reference
counts.
3) Remove the "wired" vm_map nonsense.
4) No need to keep a cache of kernel stack kva's.
5) Get rid of strange looking ++var, and change to var++.
6) Change more data structures to use our "zone" allocator. Added
struct proc, struct vmspace and struct vnode. This saves a significant
amount of kva space and physical memory. Additionally, this enables
TSM for the zone managed memory.
7) Keep ioopt disabled for now.
8) Remove the now bogus "single use" map concept.
9) Use generation counts or id's for data structures residing in TSM, where
it allows us to avoid unneeded restart overhead during traversals, where
blocking might occur.
10) Account better for memory deficits, so the pageout daemon will be able
to make enough memory available (experimental.)
11) Fix some vnode locking problems. (From Tor, I think.)
12) Add a check in ufs_lookup, to avoid lots of unneeded calls to bcmp.
(experimental.)
13) Significantly shrink, cleanup, and make slightly faster the vm_fault.c
code. Use generation counts, get rid of unneded collpase operations,
and clean up the cluster code.
14) Make vm_zone more suitable for TSM.
This commit is partially as a result of discussions and contributions from
other people, including DG, Tor Egge, PHK, and probably others that I
have forgotten to attribute (so let me know, if I forgot.)
This is not the infamous, final cleanup of the vnode stuff, but a necessary
step. Vnode mgmt should be correct, but things might still change, and
there is still some missing stuff (like ioopt, and physical backing of
non-merged cache files, debugging of layering concepts.)
Move sigjmp_buf and jmp_buf structure definitions to machine/setjmp.h
so that i386 can continue to use int as the basic register type and
alpha can use long. Bruce was concerned about possible differing
alignment. I've left the definition of _JBLEN in machine/setjmp.h
even though Bruce's example used the number directly. I don't know if
any other code relies on _JBLEN, so I left it to avoid potential
breakage.
config option in pmap. Fix a problem with faulting in pages. Clean-up
some loose ends in swap pager memory management.
The system should be much more stable, but all subtile bugs aren't fixed yet.
This is Junichi's v1.0 driver.
NOTE: Major device numbers have been changed to avoid conflict with other
FreeBSD 3.0 devices. The new numbers should be considered "official."
This driver is still considered "beta" quality, although we have been
playing with it. Please submit bugs to junichi and myself.
Submitted by: junichi@astec.co.jp
address range. They may have been trashed earlier in the boot
process, or the directory header may simply be bogus.
PR: 5140
Submitted by: Joel Faedi <Joel.Faedi@esial.u-nancy.fr>
Brought-to-attention-by: Derek Inksetter <derek@saidev.com>, bde
used, and caused a reference to an uninitialised variable (state).
I think I've fixed it now, but since nothing in the tree seems to use it,
I'm not sure.
It failed to recognize the PCI bus in a system that had only an
old chip-set (class code 000000) and a Cyclom multiport serial
card on PCI bus 0, but no VGA card or disk or network controller.
PR: i386/5300
Submitted by: Nickolay N. Dudorov <nnd@itfs.nsk.su>
- A nonprofiling version of s_lock (called s_lock_np) is used
by mcount.
- When profiling is active, more registers are clobbered in
seemingly simple assembly routines. This means that some
callers needed to save/restore extra registers.
- The stack pointer must have space for a 'fake' return address
in idle, to avoid stack underflow.
noticed some major enhancements available for UP situations. The number
of UP TLB flushes is decreased much more than significantly with these
changes. Since a TLB flush appears to cost minimally approx 80 cycles,
this is a "nice" enhancement, equiv to eliminating between 40 and 160
instructions per TLB flush.
Changes include making sure that kernel threads all use the same PTD,
and eliminate unneeded PTD switches at context switch time.
Wrappered and enabled by the define BETTER_CLOCK (on by default in smpyests.h)
apic_vector.s also contains a small change I (smp) made to eliminate
the double level INT problem. It seems stable, but I haven't the tools
in place to prove it fixes the problem.
Reviewed by: smp@csn.net
Submitted by: Tor Egge <Tor.Egge@idi.ntnu.no>
workaround. Note that this currently eats up two pages extra in the system;
this could be alleviated by aligning idt correctly, and then only dealing with
that (as opposed to the current method of allocated two pages and copying the
IDT table to that, and then setting that to be the IDT table).
make isa_dmacascade, isa_dmastart, isa_dmadone, and find_isadev MUCH
easier to be found by starting them at the beginging of the line...
remove braces inside of ifdef RESOURCE_CHECK... found by % in vi...
"high resolution" profiling. The available clocks are:
- the i8254 clock
- on non-SMP i586's and i686's: the TSC
- on systems with I586_PMC_GUPROF configured, and PERFMON configured
and available: all the performance counters.
This is unfinshed (there are problems with locking out the PERFMON
device driver, and with losing calibration after switching the clock),
but better than static configuration or writing to kmem.
Changed ifdefs to avoid generating code for non-working option
combinations.
time, but was left at 0. This caused the "can't happen" case in
splz_swi to happen for panics when tsleep() calls splx(safepri)
and there is a SWI_AST pending. This was harmless because the
the error handling happens to be right. Debugging this was tricky
because debugger traps force SWI_AST_MASK on in `cpl'.
there is a natural place to initialize `safepri' in a future commit.
Spinoffs:
- spl0() gets called in the unlikely event that isa is not configured.
- configure() has better control over enabling interrupts.
- it is now less unclear that interrupts aren't actually enabled early.
Rev.1.48 of autoconf.c seems to have done the opposite of what was
intended - moving the isa_configure() call delayed the spl0() side
effect.
Added some comments about the bogons. Removed the splhigh() call since
it is a no-op.
interval [VM_MIN_KERNEL_ADDRESS, etext] was used instead of
[btext, etext). Added a comment about this being completely
wrong for LKMs. This only affects interpreting the instructions
after the return to attempt decide the number of args. The
attempt usually fails anyway.
checking was mostly wrong at the boundaries. For the lower limit,
VM_MIN_KERNEL_ADDRESS was used instead of btext and there was an
off-by-(`size' - 1) error. For the upper limit, &etext was used
instead of etext and there was an off-by-1 error. The bugs were
harmless because `size' is not too large and some memory is mapped
just beyond the ends. We still depend on the former to avoid
having to handle the case where the memory range covers the whole
text section, and on the latter to prevent problems when we map
just beyond an end to allow writing an address range that overlaps
the end.
Fixed placement of a nearby comment.
one traditionally reserved for swap devices. The restrictions
should now be the same as the ones for dumpsys(). The restriction
on the partition should be removed someday, and dumpsys() shouldn't
repeat all the checks.
place that depended on it. The "bazillion warnings" mentioned in the
log for rev.1.45 apparently aren't a problem any more. It is hard
to be sure because the SIMPLELOCK_DEBUG option turns off (and breaks)
things in the SMP case.
it in struct proc instead.
This fixes a boatload of compiler warning, and removes a lot of cruft
from the sources.
I have not removed the /*ARGSUSED*/, they will require some looking at.
libkvm, ps and other userland struct proc frobbing programs will need
recompiled.
in <machine/cpu.h>. Moved the declarations to <machine/cputypes.h>.
Fixed style bugs in the moved code. Fixed everything that depended on
the nested include. Don't include <machine/cpu.h> (in the changed files)
unless something in it is used directly.
and fixed everything that dependended on it being declared in the old
place. It is used in "machine-independent" code in subr_prof.c.
Moved declaration of btext from subr_prof.c to <machine/cpu.h>. It
is machine-dependent.
follow.
* Rename/reorder all of the pccard structures, change many of the member
names to be descriptive, and follow more closely other 'bus' drivers
naming schemes.
* Rename a bunch of parameter and local variable names to be more
consistant in the code.
* Renamed the PCCARD 'crd' device to be the 'card' device
* KNF and make the code consistant where it was obvious.
* ifdef'd out some unused code
i was at it, do no longer insist on `PCVT_FREEBSD' being declared in
the config file, but default it to a reasonable value.
More cleanup to follow, but this part is safe for RELENG_2_2, too.
Distribute all but the most fundamental malloc types. This time I also
remembered the trick to making things static: Put "static" in front of
them.
A couple of finer points by: bde
has a PS/2 port, this is a good thing. Note, older 386/486 boxes may
lockup the keyboard controller with this enabled, but most of these kinds
of machines don't run -current, so the benefits outweigh the downsides.
Discussed with: Kazutaka YOKOTA <yokota@zodiac.mech.utsunomiya-u.ac.jp>
- CPU_CYRIX_NO_LOCK enables weak locking. If this option is not set and
FAILESAFE is defined, NO_LOCK bit of CCR1 is cleared.
- CPU_WT_ALLOC enables write-through allocation.
in a P6 SMP system. Some MB bios'es don't set the registers up correctly
for the AP's. Additionally, set the memory between 0xa0000 and 0xbffff
as write combining.
It seems I didn't count my 0's properly when adding the new masks into
icu_vector.s pushing SWI_AST_MASK off the end of the array and screwing
up the indexing for SWI_CLOCK_MASK.
Fix the bug icu_vector.s and also reformat the code in both icu_vector.s and
apic_vector.s so that it will be much harder to make the same mistake in
the future.
Submitted by: Bruce Evans <bde@zeta.org.au>
machine generates an NMI for each floating point error, just like an old XT.
Since it is ISA only, reading the EISA status port yields 0xff, which would
give a spurious EISA panic. The simplest thing to do is to ignore the 0xff.
report slightly more than 64MB of total memory. This can happen due to the
total being the sum of both base and extended memory.
Submitted by: Alan Cox <alc@cs.rice.edu>
Add cpu_rootconf and cpu_dumpconf so that configuring these
two devices can be better controlled by the MI configuration
code.
machdep.c:
MD initialization code for the new callout interface.
trap.c:
Add support for printing out whether cam interrupts are masked
during a panic.
possible. (This is not really a typographical improvement in the
case of the K6 it seems, but AMD appearantly want it too look
that way). Also if bootverbose, dump some more info about the
chip.
these structs for conflics...
it still exist that two PnP cards can colide, but this is up to the user
to make sure it doesn't happen...
other modifications to pnp.c to format output properly, and hide more
output behind bootverbose flag...
fix some bugons in pnp.h that would of made it difficult for inclusion
in external programs (for import of pnpinfo)
PR: 4486
Submitted by: tegge@idi.ntnu.no (Tor Egge)
Implement a function is_adapter_memory() in order to determine what
should nto be dumped at all. Currently, only populated with the ``ISA
memory hole''. Adapter regions of other busses should be added.
mode, the slash is a comment leader, while under non-elf it is a divide
symbol (what a concept! :-). Theoretically, #APP/#NO_APP can change this
but that doesn't seem to mesh too well with macros and line continuation.
while waiting for an interrupt (rather than spinning on the runqueue status
bits), since the other cpu can put stuff in there and the sleeping cpu may
not get an interrupt for a while. When we have a reschedule IPI, this can
come back.
Pointed out by: fsmp
are met:
1) The BIOS indicates that there is exactly 64MB of RAM, and
2) The memory size isn't specified with the MAXMEM option or
the npx0 msize hack,
...then do a speculative memory probe beyond the 64MB's until the
first bad page is encountered. This is an admitted hack, but should
nonetheless deal with detecting the correct amount of memory in nearly
all of the modern systems with >64MB of RAM.
Also made a change that will cause the list of detected memory chunks
to be printed if bootverbose is set.
holding CPU along with the lock. When a CPU fails to get the lock
it compares its own id to the holder id. If they are the same it
panic()s, as simple locks are binary, and this would cause a deadlock.
Controlled by smptests.h: SL_DEBUG, ON by default.
Some minor cleanup.
Add a simplelock to deal with disable_intr()/enable_intr() as used in UP kernel.
UP kernel expects that this is enough to guarantee exclusive access to
regions of code bracketed by these 2 functions.
Add a simplelock to bracket clock accesses in clock.c: clock_lock.
Help from: Bruce Evans <bde@zeta.org.au>
and the sound driver which uses auto dma.
The dma interface functionality remains however it now checks
to see if a dma is operating in auto dma mode and if so it bypasses
the busy flag check . I have modified the sound driver 3.5 to
adjust for this new behavior and tested it under FreeBSD 3.0 -current
This patch also includes the new function isa_dmastop.
Submitted by: Amancio Hasty <hasty@rah.star-gate.com>
smp_active = 1 used to indicate that the system had frozen previously
started AP's, while smp_active = 0 was "AP's not yet started". I have split
this into smp_started (which is set when the AP's come online), and
smp_active is left for turning on/off AP scheduling.
- We now have enough per-cpu idle context, the real idle loop has been
revived (cpu's halt now with nothing to do).
- Some preliminary support for running some operations outside the
global lock (eg: zeroing "free but not yet zeroed pages") is present
but appears to cause problems. Off by default.
- the smp_active sysctl now behaves differently. It's merely a 'true/false'
option. Setting smp_active to zero causes the AP's to halt in the idle
loop and stop scheduling processes.
- bootstrap is a lot safer. Instead of sharing a statically compiled in
stack a number of times (which has caused lots of problems) and then
abandoning it, we use the idle context to boot the AP's directly. This
should help >2 cpu support since the bootlock stuff was in doubt.
- print physical apic id in traps.. helps identify private pages getting
out of sync. (You don't want to know how much hair I tore out with this!)
More cleanup to follow, this is more of a checkpoint than a
'finished' thing.
irqs can't work (at best, the first one attached wins). It used to
be necessary to skip this check because of bogus irqs in the sound
drivers, but the sound drivers have been fixed, except possibly the
OSS ones.
wasted.
Fixed type mismatches for functions with vm_prot_t's as args. vm_prot_t
is u_char, so the prototypes should have used promoteof(u_char) to match
the old-style function definitions. They use just vm_prot_t. This depends
on gcc features to work. I fixed the definitions since this is easiest.
The correct fix may be to change vm_prot_t to u_int, to optimize for time
instead of space.
Removed a stale comment.
Added a new variable, 'bsp_apic_ready', which is set as soon as the bootstrap
CPU has initialized its local APIC. Conditionalize the GENSPLR functions
to call ss_lock ONLY after bsp_apic_ready is TRUE; This should prevent
any problems with races between the time the 1st AP becomes ready and the
time smp_active is set.
region protected by the simplelock 'cpl_lock'.
Notes:
- this code is currently controlled on a section by section basis with
defines in machine/param.h. All sections are currently enabled.
- this code is not as clean as I would like, but that can wait till later.
- the "giant lock" still surrounds most instances of this "cpl region".
I still have to do the code that arbitrates setting cpl between the
top and bottom halves of the kernel.
- the possibility of deadlock exists, I am committing the code at this
point so as to exercise it and detect any such cases B4 the "giant lock"
is removed.
Made NEW_STRATEGY default.
Removed misc. old cruft.
Centralized simple locks into mp_machdep.c
Centralized simple lock macros into param.h
More cleanup in the direction of making splxx()/cpl MP-safe.
Several new fine-grained locks.
New FAST_INTR() methods:
- separate simplelock for FAST_INTR, no more giant lock.
- FAST_INTR()s no longer checks ipending on way out of ISR.
sio made MP-safe (I hope).
We now tsleep() in kthread_init() between start_init()
and prepare_usermode() while waiting for ALL the idle_loop()
processes to come online.
Debugged & tested by: "Thomas D. Dean" <tomdean@ix.netcom.com>
Reviewed by: David Greenman <dg@root.com>
as chargeable CPU usage. This should mitigate the problem of processes
doing disk I/O hogging the CPU. Various users have reported the
problem, and test code shows that the problem should now be gone.
Work done by BSDI, Jonathan Lemon <jlemon@americantv.com>,
Mike Smith <msmith@gsoft.com.au>, Sean Eric Fagan <sef@kithrup.com>,
and probably alot of others.
Submitted by: Jnathan Lemon <jlemon@americantv.com>
This code was eliminated when the PEND_INTS algorithm was added. But it was
discovered that PEND_INTS only worsen latency for FAST_INTR() routines,
which can't be marked pending.
Noticed & debugged by: dave adkins <adkin003@gold.tc.umn.edu>
Initially functionality is confined to 32-bit BIOS functions, however
it is envisioned that BIOS support may be enlisted for other
activities in the future.
Mask the read value from the count register in order to return zero correctly
after TC, as per intel datasheet : "If it is not autoinitialised, this
register will have a count of FFFFH after TC"
comments. Remove reduntant extra addition that was unncessary, and
unneeded mask (asuming inb works correctly).
Submitted by: Stephen McKay <syssgm@dtir.qld.gov.au>
handlers don't skew the results of isa_dmastatus. The function can be
safely called with interrupts disabled.
Submitted by: Stephen McKay <syssgm@dtir.qld.gov.au>
- removed TEST_ALTTIMER.
- removed APIC_PIN0_TIMER.
- removed TIMER_ALL.
mplock.s:
- minor update of try_mplock for new algorithm where a CPU uses try_mplock
instead of get_mplock in the ISRs.
- removed TEST_ALTTIMER.
- removed APIC_PIN0_TIMER.
- removed TIMER_ALL.
apic_vector.s:
- new algorithm where a CPU uses try_mplock instead of get_mplock:
if successful continue as before.
if fail set ipending bit, mask INT (to avoid recursion), cleanup & iret.
This allows the CPU to return to successful work, while the ISR will be run
by the CPU holding the lock as part of the doreti dance.
Macros to convert the Lite2 lock manager primitives to the names used
in the kernel proper. This allows us to hide them from the lock
manager till they can be turned on.
smp.h:
declarations for the new simplelock functions.
- s_lock_init()
- s_lock()
- s_lock_try()
- s_unlock()
Created lock for IO APIC and apic_imen (SMP version of imen)
- imen_lock
Code to use imen_lock for access from apic_ipl.s and apic_vector.s.
Moved this code *outside* of mp_lock.
It seems to work!!!
and he says he's happy to see forward movement in aligning our defaults
with a 16 bit world, the 8 bit folk already being veterans by this
point who know how to use userconfig.
In any case, perhaps Warner will soon come to save us all with his Dynamic
Probing(tm) feature and this will all become totally moot in any case,
so it's probably not worth arguing about either way.
