CVS log for src/sys/arch/hppa/hppa/pmap.c

Up to [cvs.NetBSD.org] / src / sys / arch / hppa / hppa

Request diff between arbitrary revisions

Keyword substitution: kv
Default branch: MAIN

check for empty ranges in pmap_page_physload() itself

avoid calling pmap_page_physload() with an empty range
this avoids a panic in uvm_pmap_physload()
now my C360 boots again

Pull up following revision(s) (requested by skrll in ticket #1474):

	sys/arch/hppa/hppa/pmap.c: revision 1.120

Re-reading the PA2.0 Cache Move-In rules tells me we do indeed need to
purge the translations from the TLBs in pmap_procwr.

PR/56867: hppa: intermittent SIGSEGV reports in t_ptrace_wait's stepN
and setstepN test cases

Re-reading the PA2.0 Cache Move-In rules tells me we do indeed need to
purge the translations from the TLBs in pmap_procwr.

PR/56867: hppa: intermittent SIGSEGV reports in t_ptrace_wait's stepN and setstepN test cases

Pull up following revision(s) (requested by skrll in ticket #1467):

	sys/arch/hppa/include/pmap.h: revision 1.41
	sys/arch/hppa/hppa/pmap.c: revision 1.117

port-hppa/56849: Wacko kernel memory accounting in current/hppa

Two fixes:
- Don't include direct mapped memory in pmap statistics
- Decrement pmap statistics counts in pmap_kremove

Oops... revert commit mistake

Handle 'NA' (non-access) traps for the lpa and probe instructions.  The
change is inspired by OpenBSD with a bunch of my own, mainly stylistic,
changes.

Thanks to Tom Lane for the analysis.

PR/56118: sporadic app crashes in HPPA -current

port-hppa/56849: Wacko kernel memory accounting in current/hppa

Two fixes:
- Don't include direct mapped memory in pmap statistics
- Decrement pmap statistics counts in pmap_kremove

sys: Use membar_release/acquire around reference drop.

This just goes through my recent reference count membar audit and
changes membar_exit to membar_release and membar_enter to
membar_acquire -- this should make everything cheaper on most CPUs
without hurting correctness, because membar_acquire is generally
cheaper than membar_enter.

sys: Membar audit around reference count releases.

If two threads are using an object that is freed when the reference
count goes to zero, we need to ensure that all memory operations
related to the object happen before freeing the object.

Using an atomic_dec_uint_nv(&refcnt) == 0 ensures that only one
thread takes responsibility for freeing, but it's not enough to
ensure that the other thread's memory operations happen before the
freeing.

Consider:

	  Thread A			  Thread B
	obj->foo = 42;			obj->baz = 73;
	mumble(&obj->bar);		grumble(&obj->quux);
	/* membar_exit(); */		/* membar_exit(); */
	atomic_dec -- not last		atomic_dec -- last
					/* membar_enter(); */
					KASSERT(invariant(obj->foo,
					    obj->bar));
					free_stuff(obj);

The memory barriers ensure that

	obj->foo = 42;
	mumble(&obj->bar);

in thread A happens before

	KASSERT(invariant(obj->foo, obj->bar));
	free_stuff(obj);

in thread B.  Without them, this ordering is not guaranteed.

So in general it is necessary to do

	membar_exit();
	if (atomic_dec_uint_nv(&obj->refcnt) != 0)
		return;
	membar_enter();

to release a reference, for the `last one out hit the lights' style
of reference counting.  (This is in contrast to the style where one
thread blocks new references and then waits under a lock for existing
ones to drain with a condvar -- no membar needed thanks to mutex(9).)

I searched for atomic_dec to find all these.  Obviously we ought to
have a better abstraction for this because there's so much copypasta.
This is a stop-gap measure to fix actual bugs until we have that.  It
would be nice if an abstraction could gracefully handle the different
styles of reference counting in use -- some years ago I drafted an
API for this, but making it cover everything got a little out of hand
(particularly with struct vnode::v_usecount) and I ended up setting
it aside to work on psref/localcount instead for better scalability.

I got bored of adding #ifdef __HAVE_ATOMIC_AS_MEMBAR everywhere, so I
only put it on things that look performance-critical on 5sec review.
We should really adopt membar_enter_preatomic/membar_exit_postatomic
or something (except they are applicable only to atomic r/m/w, not to
atomic_load/store_*, making the naming annoying) and get rid of all
the ifdefs.

Remove trailing \n from UVMHIST_LOG() format strings.

- Alter the convention for uvm_page_array slightly, so the basic search
  parameters can't change part way through a search: move the "uobj" and
  "flags" arguments over to uvm_page_array_init() and store those with the
  array.

- With that, detect when it's not possible to find any more pages in the
  tree with the given search parameters, and avoid repeated tree lookups if
  the caller loops over uvm_page_array_fill_and_peek().

Convert pmap_check_alias into a function that resolves any aliases issues
to deal with the locking around entering a mapping of the same page with
and alias issue in the same pmap (pmap lock already held in pmap_enter)

This has the added benefit of only flushing the troublesome mappings.

The locking could be simplified further here and an PV list iterator
function with callback is probably useful.

Sync with HEAD

Sync with HEAD

To allow concurrent page faults on the same set of pages lock the PV
lists.  From ad@ and fixed up by me.

Remove __HAVE_UNLOCKED_PMAP

More KNF

Convert PMAPDEBUG to UVMHIST

Spaces to tabs

KNF

G/C a #include

Merge changes from current as of 20200406

Give pmap uvm_objects an empty pagerops to avoid special casing in UVM.
(This use of uvm_object causes a disproportionate amount of work.)

Sync with head.

Adjust for UVM locking changes.

UVM locking changes, proposed on tech-kern:

- Change the lock on uvm_object, vm_amap and vm_anon to be a RW lock.
- Break v_interlock and vmobjlock apart.  v_interlock remains a mutex.
- Do partial PV list locking in the x86 pmap.  Others to follow later.

Rename uvm_page_locked_p() -> uvm_page_owner_locked_p()

Merge from yamt-pagecache:

- do gang lookup of pages using radixtree.
- remove now unused uvm_object::uo_memq and vm_page::listq.queue.

update from HEAD

Sync with HEAD

Sync with HEAD.  (Note that most of these changes are simply $NetBSD$
tag issues.)

switch all ports to use uvm_init.c:uvm_md_init()

uvm_setpagesize() is now subsumed within this funciton.

Rebase to HEAD as of a few days ago.

Rebase.

Trailing whitespace.

sync with head

sync to -current

sync to -current

sync to latest -current.

Pull up following revision(s) (requested by skrll in ticket #62):
	sys/arch/hppa/hppa/pmap.c: revision 1.98
Fix PMAPDEBUG build

Fix PMAPDEBUG build

merge to -current.

Track if a page has an executable mapping and flush the icache (and
dcache) appropriately.

Fixes the lang/python26 build on my C3700 (PA8700) and chuq's J6700 with
PA8500.

Thanks to chuq for ideas and help with this.

Wrap a long line.

Add a KASSERT.

More whitespace.

Whitespace. hi perry.

Group the prototypes.

Simplify code a little.

Style.

Make pmap_check_alias static

G/C pmap_page_remove_locked prototype.

Cope with userland mappings of unmanaged pages (i.e. device memory being
mmap'ed), by properly making these mappings uncached, and correctly accounting
their removal to not get stuck with leftover pte pages.

From OpenBSD.

No need for sync_caches - f[di]cache already issue the sync instruction.

Whitespace.

Define PMAP_NEED_PROCWR and provide pmap_procwr so that the i-cache is
synchronised with the d-cache appropriately.

G/C pvh_aliases

Whitespace.

adapt hp700.  compile tested.

Catchup with rmind-uvmplock merge.

Welcome to 5.99.53!  Merge rmind-uvmplock branch:

- Reorganize locking in UVM and provide extra serialisation for pmap(9).
  New lock order: [vmpage-owner-lock] -> pmap-lock.

- Simplify locking in some pmap(9) modules by removing P->V locking.

- Use lock object on vmobjlock (and thus vnode_t::v_interlock) to share
  the locks amongst UVM objects where necessary (tmpfs, layerfs, unionfs).

- Rewrite and optimise x86 TLB shootdown code, make it simpler and cleaner.
  Add TLBSTATS option for x86 to collect statistics about TLB shootdowns.

- Unify /dev/mem et al in MI code and provide required locking (removes
  kernel-lock on some ports).  Also, avoid cache-aliasing issues.

Thanks to Andrew Doran and Joerg Sonnenberger, as their initial patches
formed the core changes of this branch.

Implement sharing of vnode_t::v_interlock amongst vnodes:
- Lock is shared amongst UVM objects using uvm_obj_setlock() or getnewvnode().
- Adjust vnode cache to handle unsharing, add VI_LOCKSHARE flag for that.
- Use sharing in tmpfs and layerfs for underlying object.
- Simplify locking in ubc_fault().
- Sprinkle some asserts.

Discussed with ad@.

Drop per-page locking i.e. pvh_lock and rely on locking provided by
upper layer, UVM.  Sprinkle asserts.

sync with head

Retire struct user, remove sys/user.h inclusions.  Note sys/user.h header
as obsolete.  Remove USER_TO_UAREA/UAREA_TO_USER macros.

Various #include fixes and review by matt@.

Don't rely on global APIs from internal.

Put VM_PAGE_TO_MD() definition in one place.  No functional changes.

Sync with HEAD.

We already have a flag PMAP_NOCACHE.  s/PMAP_UNMANAGED/PMAN_NOCACHE/.
Pointed out by Chuck Silvers, thanks.

Use VM_PAGE_TO_MD() to locate struct vm_page_md.  No functional
changes.

Implement pmap_physload_device(9) to replace xmd(4) MD backend.
Implement pmap_mmap(9) and use it from mem(4) and xmd(4).

Sync with HEAD.

sync with head.

Clean up; merge options DIRECT_PAGE into options XIP.

sync with head

Allow PVF_UNCACHEABLE in set/clear of pmap_changebit call (for now).

Re-define the definition of "device page"; device pages are pages of
device memory.  Pages which don't have vm_page (== can't be used for
generic use), but whose PV are tracked, are called "direct pages" from
now.

sync with head

Fix a typo; pointed out by Chuck Silvers.

Sync with HEAD.

Support PMAP_UNMANAGED in some pmaps.

(Others should be converted eventually, but no problem while managed
device page is not used.)

Pull across pmap_remove fix from OpenBSD.

Fix user-after-free bug in pmap_remove().  Page table pages are freed as soon
as the last page table entry that was actually used is cleared.  So make sure
we check the page table page is still there for every page we remove.
Otherwise we will bring back the tlb entry and cache lines when we touch the
freed page, and we will create an illegal alias (non-equivalent mapping)
as soon as the page gets re-used.

Fix PMAPDEBUG build for previous.

Not all PA CPUs have the U-bit (uncacheable) for non-IO memory.  In fact
most don't.  Deal with non-equivalent aliases by removing and flushing the
managed mappings, and flushing the unmanaged mappings.

When flushing caches/TLB flush the cache before purging the TLB just in
case the flush enters the mapping into the TLB.

The FP regs are accessed in both physical and virtual modes.  Make sure
they're allocated in directly mapped memory to avoid aliasing problems
and unnecessary cache flushing.

Fix various names and references. If we ever directly map the uarea and
pcb we can change things at that point.

Significantly helps my J210 (PA7200).

Mostly from OpenBSD.

Change struct uvm_object::vmobjlock to be dynamically allocated with
mutex_obj_alloc().  It allows us to share the locks among UVM objects.

Add a comment.

sync with head

Add #include "opt_cputype.h" where necessary.

Use VM_PAGE_TO_MD().  Only compile tested.

s/printf/db_printf/ in a few places.

Deal with PA2.0 cache move-in rules by purging the TLB as well as
flushing/purging the cache.

Fix typo in previous. It's not a good idea to let uvm mess with kernel
text pages.

Semi-colons help.

Add a prototype for pmap_page_physload.

Remove the dma24_ex method of dealing with BUS_DMA_24BIT and replace
with a uvm managed freelist of pages in the right range.

Remove unnecessary part of conditional.

DPRINTF format fixes.

Fix pmap_dump_table so that it doesn't loop indefinitely for a space that
doesn't map anything in the last PDE.

Also, don't loop unnecessarily through the spaces.

Flush only as much as is required. Pointed out by rmind.

- Use uvm_lwp_setuarea() instead of directly setting address to lwp_t::l_addr.
- Replace most remaining uses of l_addr with uvm_lwp_getuarea() or lwp_getpcb().
- Amend assembly in ports where it accesses PCB via struct user.
- Rename L_ADDR to L_PCB in few places.  Reduce sys/user.h inclusions.

Kill proc0paddr.  Use lwp0.l_addr instead.

Use lwp_getpcb() on hppa and ia64, clean from struct user usage.

Update a couple of comments.

Use the new flags argument to pmap_kenter_pa for PMAP_NOCACHE.

Add a flags argument to pmap_kenter_pa(9).
Patch showed on tech-kern@ http://mail-index.netbsd.org/tech-kern/2009/11/04/msg006434.html
No objections.

Remove uarea swap-out functionality:

- Addresses the issue described in PR/38828.
- Some simplification in threading and sleepq subsystems.
- Eliminates pmap_collect() and, as a side note, allows pmap optimisations.
- Eliminates XS_CTL_DATA_ONSTACK in scsipi code.
- Avoids few scans on LWP list and thus potentially long holds of proc_lock.
- Cuts ~1.5k lines of code.  Reduces amd64 kernel size by ~4k.
- Removes __SWAP_BROKEN cases.

Tested on x86, mips, acorn32 (thanks <mpumford>) and partly tested on
acorn26 (thanks to <bjh21>).

Discussed on <tech-kern>, reviewed by <ad>.

sync with head.

Fix non-DEBUG builds.

Remove NOP flushes. pvh_list == NULL.

Pull across a fix from OpenBSD, but use PDE_SIZE instead of ~PDE_MASK + 1.

Make sure that if a PDE isn't there, we skip to the start of the address
range covered by the next PDE and not somewhere in the middle.  The old could
have skipped over some valid PTE's causing them to stay behind in the pmap.
Since we would not flush the cache for those pages either this could also
cause memory corruption when dirty cache lines would be written back to
memory at a later stage.

Sync with HEAD.

Commit is split, to avoid a "too many arguments" protocol error.

sync with head.

Merge nick-hppapmap.

This is a port of the OpenBSD pmap and trap handling code to get us

        - Performance boost on some/all machines.
        - Well on the way to PA2.0 (in 32bit mode) support. Several
          machines probe hardware, but fail sometime after interrupts
          are enabled.

Other things changed / fixed on the branch are

        - update autoconf to use the OpenBSD code.
        - com @ dino is very close to being supported.
        - HPPA_REDZONE has been replaced with a working redzone which
          is enabled with DIAGNOSTIC.
        - UPAGES has been halved in size.
        - power(4) from OpenBSD to fix a few bugs.
        - updated list of modules from OpenBSD.
	- initial ports of uturn(4), astro(4) and elroy(4).
	- update some copyrights (remove advertising clause, etc.)

Thanks to mjf for some of the above, testing and listening.

Track unmanaged mappings to deal with loaning and non-equivalient aliases.

Sync with HEAD.

For correctness test pvh_attrs with lock held.

Enable a DPRINT

Oops, typo in previous.

Another comment update.

Update some comments.

Add a reference.

Whitespace.

Add a KASSERT to pmap_activate

Wrap long lines. KNF.

Put KASSERTs after DPRINTFs.

change pmap flags argument from int to u_int.
discussed with christos@ on source-changes-d@

Remove incorrect KASSERT.

XXX PVF_KENTER needs work.

Remove PMAPDEBUG define that crept in.

Track referrenced and modified attributes of pages correct, i.e. as
described in pmap(9).

Inherit uncacheable attribute if the page already has non-equiv aliases.

Tidy-up a comment while I'm here.

Use __func__ everywhere.

Fix-up a message while here.

Whitespace.

Simplify locking while UP only.

Make pmap_hptdump compile.

No need to apply PTE_PAGE to pa. It's already page aligned.

Add a comment.

Remove redundant part of a KASSERT

Remove some debug code.

Wrap a line. KNF.

Add a couple of KASSERTs.

bcopy -> memcpy

Fix previous - I forgot a ;

More DIAGNOSTIC changed to KASSERT.

s/u_int32_t/uint32_t/

Re-enable hp700_pagezero_{,un}map

Fix non-DIAGNOSTIC compiles.

Remove debug code so that a kernel compiles with clean sources from the
branch.

Fixup redzone stuff and enable it if DIAGNOSTIC.

Handle non-equivalent aliases.

Thanks to mjf for help with this.

Whitespace.

Make LOCKDEBUG kernels boot.

Rename a variable to better reflect what it's used for.

Convert some #ifdef PMAPDEBUG checks to KASSERTs.

Sync with HEAD.

Update a comment to make sense to >1 person.

It's VM_PAGE_TO_PHYS you muppet.

Before returning a page to UVM flush the cache to avoid aliasing problems.

Convert pmap_flush_page's 2nd arg from int to bool.

G/C #if 0 code

Typo.

Sync with HEAD.

kill MALLOC and FREE macros.

replace bitmask_snprintf(9) with snprintb(3)

No need for pmap_bootstrap_pv_page* with PMAP_{,UN}MAP_POOLPAGE.

Update haad-dm branch to haad-dm-base2.

Make kernel_pmap_ptr a const.  Requested by steve_martin.

Make pmap_kernel() a MI macro for struct pmap *kernel_pmap_ptr,
which is now the "API" provided by the pmap module.  pmap_kernel()
remains as the syntactic sugar.

Bonus cosmetics round: move all the pmap_t pointer typedefs into
uvm_pmap.h.

Thanks to Greg Oster for providing cpu muscle for doing test builds.

Improve debug output.

Improve locking.

Whitespace.

Bah! Fix my previous commit that broke the build. I should have converted
the cast I just removed to uint32_t.

Don't cast 'pve->pv_va' to int, seeing as its type is vaddr_t (uint32_t).

Restore PMAP_NC

Typo in comment.

Improve BTLB handing.

G/C some old stuff.

Pre-allocate pdes for all of KVA.

Correctly pmap_kenter_pa the various memory areas.

Fix up memory counts and reporting.

Commit work in progress.

Sync with HEAD.

sync with head.

sync with head.

Remove clause 3 and 4 from TNF licenses

Sync with HEAD.

sync with head

sync with HEAD

Sync with HEAD

Fix includes

Sync with HEAD

Sync with HEAD.

Sync with head.

typo in comment.

__FUNCTION__ -> __func__

Sync with HEAD.

sync with head

Sync with HEAD.

Sync with HEAD. amd64 Xen support needs testing.

pmap_create: M_VMMAP -> M_VMPMAP for pmap.

sync with HEAD

Merge the ppcoea-renovation branch to HEAD.

This branch was a major cleanup and rototill of many of the various OEA
cpu based PPC ports that focused on sharing as much code as possible
between the various ports to eliminate near-identical copies of files in
every tree.  Additionally there is a new PIC system that unifies the
interface to interrupt code for all different OEA ppc arches.  The work
for this branch was done by a variety of people, too long to list here.

TODO:
bebox still needs work to complete the transition to -renovation.
ofppc still needs a bunch of work, which I will be looking at.
ev64260 still needs to be renovated
amigappc was not attempted.

NOTES:
pmppc was removed as an arch, and moved to a evbppc target.

Sync with HEAD

sync with head.

Sync with head.

Like I said - don't futz with the trapframe in pmap_activate. Instead
update the trapframe space and protection register in cpu_lwp_fork and
setregs.

Sync with head.

Sync with head.

Update to HEAD.

pmap_activate shouldn't play with the trap frame - especially not now it's
called by the MI code.

Implement a way of tracking the vmspace allocated to a process and its
LWPs so that if it changes underneath us we can correct the space id in
the trapframe in pmap_activate for now.

An example of when this happens is vfork/exec.

An exec hook is probably the way to go.

Sync with HEAD.

Kill caddr_t; there will be some MI fallout, but it will be fixed shortly.

Remove hppa_{round,trunc}_page.

- sync with head.
- move sched_changepri back to kern_synch.c as it doesn't know PPQ anymore.

sync with head.

TRUE -> true, FALSE -> false

Replace the Mach-derived boolean_t type with the C99 bool type.  A
future commit will replace use of TRUE and FALSE with true and false.

Whitespace.

Sync with head.

sync with head.

sync with head.

s/independant/independent/, from Zafer.

Sync with head.

pmap_kenter_pa the kernel and it's data structures including the kernel
stack.

Fixes the "booting from disk memory corruption bug" which was a result
of pmap_extract silently failing against a scsipi_xfer data area allocat-
ed on kernel stack in _bus_dmamap_load_buffer

Add some debug code.

Whitespace fixes.

G/C pages_per_vm_page

KNF and comment update.

sync with head.

sync with head

Pull up following revision(s) (requested by skrll in ticket #116):
	sys/arch/hppa/hppa/pmap.c: revision 1.19
	sys/arch/hppa/hppa/pmap.c: revision 1.20
	sys/arch/hppa/hppa/pmap.c: revision 1.21
	sys/arch/hppa/hppa/pmap.c: revision 1.22
	sys/arch/hppa/hppa/pmap.c: revision 1.23
	sys/arch/hppa/hppa/pmap.c: revision 1.24
	sys/arch/hp700/conf/ld.script: revision 1.9
Fix PMAPDEBUG compile by sorting the printf format problems.
KNF a comment.
Use PMAP_PRINTF(PDB_INIT, ...) instead of various #ifdefs and add some
more debug code.
KNF a comment
Fix PR 33702.
Modernise the linker script and make sure we have a symbol after the
link_sets.
Make sure that pmap_bootstrap doesn't tell uvm that some of the kernel
phys pages are free. Previously uvm would very likely allocate the phys
pages used for the link_sets to a MALLOC and they'd get overwritten.
Thanks to David H. Gutteridge for testing various things.
Don't lose a page at the end of the 3 segments we uvm_page_physload.
(end/avail_end are excluive page numbers - as dicussed on tech-kern)

sync with head.

Don't lose a page at the end of the 3 segments we uvm_page_physload.
(end/avail_end are excluive page numbers - as dicussed on tech-kern)

Fix PR 33702.

Modernise the linker script and make sure we have a symbol after the
link_sets.

Make sure that pmap_bootstrap doesn't tell uvm that some of the kernel
phys pages are free. Previously uvm would very likely allocate the phys
pages used for the link_sets to a MALLOC and they'd get overwritten.

Thanks to David H. Gutteridge for testing various things.

KNF a comment

Use PMAP_PRINTF(PDB_INIT, ...) instead of various #ifdefs and add some
more debug code.

KNF a comment.

Fix PMAPDEBUG compile by sorting the printf format problems.

sync with head.

Remove leading __ from __(const|inline|signed|volatile) -- it is obsolete.

merge ktrace-lwp.

Sync with HEAD. Here we go again...

Pull up revision 1.16 (requested by chs in ticket #1980):
in pmap_enter(), preset the mod/ref bits based on the flags argument.
fixes 25640.

Pull up revision 1.16 (requested by chs in ticket #1980):
in pmap_enter(), preset the mod/ref bits based on the flags argument.
fixes 25640.

Pull up revision 1.16 (requested by chs in ticket #424):
in pmap_enter(), preset the mod/ref bits based on the flags argument.
fixes 25640.

in pmap_enter(), preset the mod/ref bits based on the flags argument.
fixes 25640.

in pmap_protect(), remove the assertion that any entries modified must be
managed.  after the yamt-km changes, the sti driver needs to add execute
permission to the (now unmanaged) mapping for its copy of the card firmware,
and this appears to work fine already.

Sync with HEAD

add hppa support for RAS, context stuff, siginfo and scheduler activations:
 - add a RAS hook in cpu_switch().
 - fill in the definition of struct mcontext.
 - implement cpu_upcall(), cpu_getmcontext(), cpu_getmcontext() and
   cpu_switchto().
 - for now, force the right priviledge bits and space regs in setcontext().
 - use correct values for __SIMPLELOCK_*.
 - move the user stack to start at a multiple of the pthread stack size
   so that libpthread can use the sp-masking trick.

in pmap_protect(), don't just return if the desired permissions include write.
we could be removing execute permission.

add a PMAP_NC flag for pmap_kenter_pa() to specific a non-cached mapping.
use this in mbus_dmamem_map() to fix corruption of DMA memory.
note that this TLB bit is ignored on some CPUs (PA7100 and probably
others of that era), so this doesn't fix the problem in general,
but it does work on newer models and will make things easier later.

de-__P, remove register, ansify.

update for LWPs, and some lite cleanup.

__KERNEL_RCSID()

Back out the following chagne:
    http://mail-index.netbsd.org/source-changes/2003/05/08/0068.html

There were some side-effects that I didn't anticipate, and fixing them
is proving to be more difficult than I thought, do just eject for now.
Maybe one day we can look at this again.

Fixes PR kern/21517.

Simplify the way the bounds of the managed kernel virtual address
space is advertised to UVM by making virtual_avail and virtual_end
first-class exported variables by UVM.  Machine-dependent code is
responsible for initializing them before main() is called.  Anything
that steals KVA must adjust these variables accordingly.

This reduces the number of instances of this info from 3 to 1, and
simplifies the pmap(9) interface by removing the pmap_virtual_space()
function call, and removing two arguments from pmap_steal_memory().

This also eliminates some kludges such as having to burn kernel_map
entries on space used by the kernel and stolen KVA.

This also eliminates use of VM_{MIN,MAX}_KERNEL_ADDRESS from MI code,
this giving MD code greater flexibility over the bounds of the managed
kernel virtual address space if a given port's specific platforms can
vary in this regard (this is especially true of the evb* ports).

Use PAGE_SIZE rather than NBPG.

sync kqueue branch with HEAD

catch up with -current.

No longer use BTLB entries to map the entire address spaces of I/O subsystems,
since BTLB entries can be scarce and very little of an I/O subsystem normally
needs to be mapped.

Instead, the pmap now allows mappings of I/O space to be entered with
pmap_kenter_pa.  bus_space mappings for small amounts of I/O space (as for
virtually all devices) are made this way, with BTLB entries still used for
large mappings for things like framebuffers.

This has led to more and cleaned-up uses of bus_space(9) and has caused
some autoconf cleanup.  Also, kgdb is now attached and connected before
autoconfiguration, which is much earlier than before.

Cleaned up BTLB support.  There are no longer BTLB function pointers
in struct hppa_cpu_info or anywhere else, now there are just hppa_btlb_*
functions.  Added support for machines with split I/D and variable-range
BTLBs.  Added support for purging BTLB entries.

Significant pmap changes to no longer rely on the "U-bit" (TLB_UNCACHEABLE)
to deal with aliasing of regular memory pages, because many processors don't
support it.

Now, the pmap marks all mappings of a page that has any non-equivalent
aliasing and any writable mapping, and the fault handlers watch for this
and flush other mappings out of the TLB and cache before (re)entering a
conflicting mapping.

When a page has non-equivalent aliasing, only one writable mapping at
a time may be in the TLB and cache.  If no writable mapping is in the
TLB and cache, any number of read-only mappings may be.

The PA7100LC/PA7300LC fault handlers have not been converted yet.

Made changes in where/how the kernel is linked, and how the pmap
maps it with BTLB entries, to minimize the number of BTLB entries
needed.

Because the CPU type was often guessed incorrectly, the mapping of
HP board number to system name now includes information about the
expected CPU type.

catch up with -current.

catch up with -current on kqueue branch

file pmap.c was added on branch gehenna-devsw on 2002-07-14 17:47:12 +0000

file pmap.c was added on branch kqueue on 2002-06-23 17:37:05 +0000

Added files to support generic HP PA-RISC based machines.  hp700-specific
files to follow.