Commit Graph

1252 Commits

Author SHA1 Message Date
thorpej
c0cc1ed476 Implement support for Address Space Numbers, greatly reducing the number
of TLB and I-cache flushes, significantly speeding up context switches.

Once again, many thanks to Chris Demetriou and Ross Harvey for code
review and debugging assistance!
1998-03-22 05:41:37 +00:00
thorpej
a8d86e5a7c Replace PMAP_ASNGEN_INVALID with PMAP_ASN_RESERVED. 1998-03-22 05:39:50 +00:00
mjacob
86b6520e41 more TS_WOPEN to tp->t_wopen changes 1998-03-21 23:36:19 +00:00
mycroft
0dae91d9af Eliminate uses of TS_WOPEN in hard-wired devices. 1998-03-21 22:52:59 +00:00
mjacob
34f87569b9 add some error defintions 1998-03-21 22:02:42 +00:00
thorpej
66d8f5b544 sync systypes w/ <machine/rpb.h> 1998-03-20 21:48:21 +00:00
thorpej
63c73f94e3 Add a few more systypes. 1998-03-20 21:48:03 +00:00
thorpej
0f95ffdc1d Nuke swpctxt(); it's only used by the Mach pmap, which we will only ever
use for reference.
1998-03-19 06:44:25 +00:00
thorpej
003c50d1d5 Add a macro to invalidate the TLB for a given pmap/va pair. TLB
invalidation algorithm:

	if (old mapping had PG_ASM set || pmap is active) {
		TIBS(va);
		if (also sync I-stream)
			imb();
	}

The check for "old mapping had PG_ASM" will get all kernel mappings (since
kernel mappings always have PG_ASM set).

This allows us to remove the bogus check for the kernel pmap in
active_pmap() - do so.

Use the new TLB invalidation macro whenever such action is needed.
1998-03-18 23:55:25 +00:00
thorpej
15adb17803 Eliminate the last argument from pmap_remove_mapping(); it makes its own
decisions about TLB invalidation.
1998-03-18 23:11:44 +00:00
thorpej
7ee4af11a7 Change active_pmap() to use the CPU mask (XXX and check for kernel pmap
as well, until some other changes are made).  Nuke active_user_pmap(),
and change the places that used it to use active_pmap() instead (as well
as make some DIAGNOSTIC consistency checks).
1998-03-18 22:50:50 +00:00
thorpej
605472f676 Optimize out a TLB invalidation in a common case of pmap_enter(): if
the PTE was previously invalid, no TLB invalidation is necessary because:

	(1) when a PTE is invalidated, its entry is flushed from the
	    TLB

	(2) the PALcode won't install an invalid PTE into the TLB.
1998-03-18 22:13:58 +00:00
thorpej
cfdf9a95ad Keep track of which CPUs are using a pmap by setting/clearing bits
in the pmap's CPU mask in pmap_activate()/pmap_deactivate().
1998-03-18 21:57:03 +00:00
thorpej
43614761e3 In cpu_exit() deactivate the address space before freeing the vmspace
structure.  We will continue to run on this context (which is the
global Lev1map at this point) right up until we switch to proc0's
context in switch_exit().
1998-03-18 20:38:07 +00:00
thorpej
87eb2cfded Don't call pmap_deactivate() if we jumped into the middle of cpu_switch()
from switch_exit(), since by this time, the vmspace will have already
been deactivated and freed.
1998-03-18 20:36:13 +00:00
thorpej
b637a998f4 Add ASN housekeeping and a CPU mask to the pmap. 1998-03-18 19:39:23 +00:00
thorpej
961a955498 Move the "are we active" macros out of the header file. 1998-03-18 19:27:46 +00:00
thorpej
d37acae24c Add a DIAGNOSTIC checks for the kernel pmap in pmap_create_lev1map()
and pmap_destroy_lev1map().  Correct a comment in another DIAGNOSTIC
panic.
1998-03-18 19:21:50 +00:00
thorpej
06b49b8f3e Change a couple of assert()s to DIAGNOSTIC panics. 1998-03-18 19:12:57 +00:00
thorpej
438599b408 Correct a comment in pmap_bootstrap(). 1998-03-18 19:04:42 +00:00
thorpej
56e004c995 Pass the max ASN from the HWRPB to pmap_boostrap(). 1998-03-18 19:02:49 +00:00
thorpej
426d2953f5 Add a macro to test if PG_ASM (Address Space Match) is set in a PTE. 1998-03-18 19:00:15 +00:00
bouyer
9f50fca1fd Add commented out "options FFS_EI" 1998-03-18 16:34:41 +00:00
thorpej
9c1e8fc2ed Implement the PMAP_NEW interface for UVM. 1998-03-17 05:15:24 +00:00
thorpej
1477f77353 Properly depend on the PMAP_NEW option. 1998-03-17 05:00:18 +00:00
thorpej
6bbfd3e9ff Use pmap_kenter_pa() in _bus_dmamem_map() if PMAP_NEW. 1998-03-17 04:59:36 +00:00
thorpej
f8cff5ab23 Add a software PTE bit that indicates that a va -> pa mapping was entered
in the physical->virtual list.
1998-03-17 04:53:43 +00:00
thorpej
00452b441f Move PTE-related constants here, and make them not depend on a hard-coded
page size (i.e. use the one initialized from the HWRPB at boot time).

Do a bit of cleanup while here, rendering old inherited constants obsolete.
1998-03-12 07:29:21 +00:00
thorpej
4d8723232d Garbage-collect a bunch of constants that were inherited, but are no
longer necessary or make sense.
1998-03-12 07:28:07 +00:00
thorpej
bd3c0e36cf Garbage-collect this a bit. 1998-03-12 06:47:11 +00:00
thorpej
00a597fe92 Use vm_page_alloc1() and vm_page_free1() as appropriate. 1998-03-12 06:27:36 +00:00
thorpej
1f8d640c4b Bump maxusers to 64. 1998-03-12 06:04:47 +00:00
thorpej
6e6e2d7ebf Dump maxusers to 32. 1998-03-12 06:04:31 +00:00
thorpej
d9a1f8ba36 Adjust the default and low-bound maxusers, now that the pmap can deal. 1998-03-12 06:04:14 +00:00
thorpej
dfe0937a7e If not DEBUG, use the Virtual Page Table to get the PTE for kernel mappings
in pmap_enter() and pmap_emulate_reference().
1998-03-12 02:59:22 +00:00
thorpej
30766180a3 Nuke these; they are long-since obsolete. 1998-03-12 01:28:01 +00:00
thorpej
152a4bfa60 Increase the maximum userspace address to 4TB. Leave the stack at 8G
for now, but make a note that we might want to move it down to 4G later.
1998-03-12 01:25:52 +00:00
thorpej
e046925c3a Massive cleanup and partial rewrite of the NetBSD/alpha pmap module.
Major change is that page table page management has been completely
rewritten.  Page tables are now accessed via K0SEG (no more KVA space
wasted on user page tables), and a much larger user address space is
supported.

Many thanks to Chris Demetriou and Ross Harvey for helpful insight and
debugging assistance.
1998-03-12 01:24:52 +00:00
thorpej
900e1c90bd Nuke ALPHA_STSIZE and ALPHA_MAX_PTSIZE. Add macros to compute and operate
on segments mapped by L1 and L2 PTEs.
1998-03-12 01:21:21 +00:00
thorpej
7225aae835 Move check for user-pmap-still-using-Lev1map from pmap_enter_ptpage()
to pmap_enter().
1998-03-09 22:31:23 +00:00
thorpej
daa9cfae50 Don't do the Segtabzero-for-dev-zero hack. 1998-03-09 20:43:28 +00:00
thorpej
e456fc0538 Simplify/speed up pagemove() somewhat by using the Virtual Page Table. 1998-03-09 20:17:03 +00:00
thorpej
78a173bada Define VPT_INDEX(), which computes the index into the Virtual Page Table
of the PTE that maps the specified virtual address.

Thanks to Chris Demetriou and Ross Harvey for clarifying the VPT.
1998-03-09 19:57:57 +00:00
thorpej
b3d7fd8f3f Just use vtophys() to get the PCB phys addr. 1998-03-07 04:20:45 +00:00
thorpej
9c236919e9 Rewrite pmap_extract(), and use it as appropriate in vtophys() rather
than (almost) duplicating the code.
1998-03-07 03:37:02 +00:00
thorpej
cd7d081d02 Export a pointer to the Virtual Page Table. 1998-03-07 03:15:43 +00:00
thorpej
f7d64ef778 Clean up a comment. 1998-03-07 03:15:06 +00:00
thorpej
f55b280ffc Clean this up a bit. 1998-03-07 01:44:08 +00:00
thorpej
dcbc04e7af Rename Sysptmap and Sysmap to Lev2map and Lev3map, respectively, and
add some comments describing in better detail what they are and how they
are used.
1998-03-07 01:43:54 +00:00
thorpej
60e6d2e48f Update a comment in pagemove(). 1998-03-07 01:26:04 +00:00