17e7b7e1bd
Make vnode_pager_putpage return an int, like other *_putpage routines.
1994-01-07 18:12:12 +00:00
4343c468c5
Trivial stylistic change to match other code.
1994-01-07 18:11:38 +00:00
f8258a5b44
Add missing return value.
1994-01-07 17:12:51 +00:00
c26286a5c1
bring changes down from magnum
1993-12-20 12:39:55 +00:00
b4d3382694
Canonicalize all #includes.
1993-12-17 07:56:32 +00:00
6c0eb81987
Fix typo in DIAGNOSTIC code.
1993-12-06 13:00:49 +00:00
053e467648
Use proc0's credentials in buffer used for IO to swap device.
...
Especially useful when swapping over NFS.
1993-11-29 16:05:20 +00:00
76dbc1192b
new specfs.h and fifo.h locations
1993-11-12 05:54:12 +00:00
9d0e9ed1c1
fix setup of b_dirtyoff and b_dirtyend when writing out a page.
...
from Mike Hibler.
1993-11-10 22:43:34 +00:00
56adbe979f
make absent absent
1993-11-10 08:22:05 +00:00
d580099dde
fix those last changes, and quiet a compiler warning.
1993-11-10 08:11:47 +00:00
554c91f74a
change some flags back to bitfields, so it compiles.
1993-11-03 14:49:22 +00:00
64c1e3b2f5
new device pager from mike hibler, hibler@lamp. this is in its 'virgin' state
1993-11-03 14:48:42 +00:00
215649adeb
from Mark Tinguely (tinguely@plains.NoDak.edu): record when processes are
...
swapped...
1993-10-19 02:54:09 +00:00
e98e3183fc
get rid of a few ref's to buffer_map
1993-10-11 02:11:28 +00:00
13a840d756
generalize kernacc() to any protections in the function "kerncheckprot()",
...
and prototype it.
1993-10-06 23:38:19 +00:00
f09334ef6c
we need the buffer hack again, now that we have a real vfs_bio...
1993-10-06 02:06:49 +00:00
95e5868ee2
allow an offset to *really* be passed in for mmap's on devices.
...
From Julian Elischer.
1993-10-02 00:00:21 +00:00
00e8bf23cd
Add `#ifdef DIAGNOSTIC' around check for dirty pages in persistent objects.
1993-09-23 22:20:29 +00:00
9c32bc786a
make allproc be volatile, and cast things accordingly.
...
suggested by torek, because CSRG had problems with reordering
of assignments to allproc leading to strange panics from kernels
compiled with gcc2...
1993-09-15 22:30:32 +00:00
b4940ad7a0
Wrap iprintf, vm_object_print, vm_map_print to allow passing a printf function.
1993-09-13 14:10:35 +00:00
37e8d5291e
get rid of bogosities of the form "vm_maxsaddr + MAXSSIZ", and replace
...
them with USRSTACK. the former was a hack by wfj to accomodate his screwy
execve() implementation. the latter is a (minor) reversion to net/2.
1993-09-04 01:29:22 +00:00
6a474f29e2
get rid of maxdmap, and seperate MAXDSIZ and MAXSSIZ in rlimit checking.
1993-09-04 00:37:54 +00:00
2cb87de98b
fixs long broken 'dequeue' macro which would expand to 'dequeue_head' which
...
is never defined. added fix from my private sources:
#define dequeue_head(queue) remqueue(queue, queue_first(queue))
no one noticed, because everyone used the 'remqueue' interface except me.
1993-09-01 15:55:26 +00:00
ae8b49f87b
#ifdef pc532 in same places as #ifdef i386
1993-09-01 08:49:45 +00:00
a87b4a30bd
pagers are now options, not pseudo-devices
1993-08-30 07:09:20 +00:00
7f550be7fd
Remove prototype for pmap_bootstrap()
1993-08-29 12:12:20 +00:00
405b084fe6
return some errors other than EIO from swstrategy()
1993-08-29 01:50:20 +00:00
15c3400ffe
Machine independant changes to VM for handling non-contiguous memory.
1993-08-27 23:45:55 +00:00
4e65827b6c
Moved vm_map_create()'s out-of-maps panic to a more appropriate place.
1993-08-19 03:13:24 +00:00
fbd4257135
Put `#ifdef notyet' around some code which is currently defunct.
1993-08-14 10:02:30 +00:00
22dff5862a
Removed the user area protection gunk I painstakingly removed in revision
...
1.9. It has not been necessary since then due to a segment fixup made to
machdep.c that prevents access to anything above MAX_VMUSER_ADDRESS.
1993-08-13 03:50:02 +00:00
acf8d8b584
deal with ddb bogosities
1993-08-07 06:03:02 +00:00
26817f18ab
fix prettyprinting changes that caused the kernel to not link if you
...
defined DIAGNOSTIC but not DDB...
1993-08-07 05:08:11 +00:00
a0aed26669
Remove a spurious `#' added two revisions ago.
1993-08-03 01:56:52 +00:00
4862b84c92
Add RCS identifiers (this time on the correct side of the branch), and
...
incorporate recent changes in netbsd-0-9 branch.
1993-08-01 19:22:24 +00:00
3643c134f1
Change "#endif FOO" to "#endif /* FOO */", to keep gcc -Wall, lint, etc.
...
from complaining about a trivial issue.
1993-07-29 21:42:19 +00:00
7b2afa7fe6
incorporate changes from 0-9-base to 0-9-ALPHA
1993-07-28 02:20:34 +00:00
36a6e635b3
Committing in src/sys/vm
...
Modified Files:
vm_object.c
call vm_object_remove() to remove an object from the hashtable when
vm_object_collapse() starts fiddling with pagers.
1993-07-28 00:00:01 +00:00
9d7e246a03
Cleanup pretty printers for maps and objects.
1993-07-22 13:03:36 +00:00
24e3cbad44
according to andrew, the bit with the comment:
...
> /* protect from the user area from user accesses. :-)
> addr -> addr + UPAGES*NBPG don't seem to be protected without
> this; the rest seems to be OK, and doesn't like being protected
> - andrew@werple.apana.org.au */
*is* necessary, so put it back in... it's in an #ifdef i386 part anyway.
1993-07-19 13:47:09 +00:00
aa381d0e73
vm_protect() of the user area is no longer required for i386 machines; the
...
protection is implemented by adjusting the bounds of the user segment
descriptors.
1993-07-18 08:19:34 +00:00
ad087b06a0
Finish moving struct definitions outside of function declarations.
1993-07-17 15:56:59 +00:00
97cae397ee
Finish moving struct definitions outside of function declarations.
1993-07-17 15:24:33 +00:00
fd3c8fa35a
fix a whole slew of off-by-one errors
1993-07-15 15:42:17 +00:00
4b90229cd2
do the "right thing" with mmap protections, once and for all.
...
in particular, if an object is mapped private, DO NOT grant
VM_PROT_ALL protections to it; it could be sensitive data.
1993-07-15 14:48:37 +00:00
b8b3b4471d
get rid of any support for vm_fault_wire() returning a result.
...
if it's going to be ifdef'd out until it's tested, it shouldn't
be put in to begin with.
also, minor ansifications and cleanups (mostly so tags would be happier).
1993-07-15 14:25:18 +00:00
999eb6ce5b
get rid of extraneous thread_wakeup()s, added at sef's prompting,
...
which would never be called anyway. leave his vm_page_activate()
change in though, as it is correct.
1993-07-15 14:07:51 +00:00
5287574e0a
get rid of bogus retrymalloc code (and it was bogus)...
1993-07-15 13:33:23 +00:00
b99e3b6666
get rid of bogus protections chacking which conflicts w/guido's
...
patch...
1993-07-07 11:25:32 +00:00
09844fbcfe
make the rlist code all go away. replace it with resource map code,
...
as written by Wolfgang Solfrank.
1993-07-07 06:04:12 +00:00
8ebc845441
got rid of (obviously) bogus assignment which followed free().
1993-07-02 10:26:49 +00:00
7a0e26cd13
Paul Kranenburg's VM deadlock patches (from patchkit 00147, parts 3 & 4).
...
Prevent dirty objects from being cached, and prevent vm_page_alloc() from
allocating too much memory to non-kernel objects.
1993-06-30 03:48:25 +00:00
5f8b3f4ff4
Oops. Get rid of the `#ifndef MYCROFT_IS_A_DORK'. B-)
1993-06-29 19:38:50 +00:00
d94e5eecbb
Fix some off-by-one errors.
1993-06-29 13:47:05 +00:00
5705dde7e9
Disallow access to the user pages, where the kernel stack is stored. This
...
is effected in useracc(), and also by a vm_protect() in vm_fork() for i386
CPUs. Without the latter a write to the user area, say USRSTACK+1000,
would hang a 386-based system.
1993-06-29 07:16:31 +00:00
918a434644
ANSIfications. Added support for vm_fault_wire() returning a result; code
...
to actually fail is currently disabled, as this would enable some new code
in vm_map_pageable() (disabled in this commit) that hasn't been used to
date. I'm fairly confident it is all OK, but shall test it some more once
the rest of the kernel is more stable, before enabling it.
1993-06-27 06:38:48 +00:00
51ac6a0997
ANSIfications.
1993-06-27 06:27:29 +00:00
0e761b8d7e
fix behavior when VM_MAXUSER_ADDRESS == (-UPAGES * NBPG)
1993-06-21 10:18:20 +00:00
800afb6fec
Remove annonying printf when adding swap.
1993-06-17 16:34:25 +00:00
1b44fd4c31
fix syntax errors caused by *not* having any one of the pagers...
1993-05-29 17:03:06 +00:00
ada4945e6b
patches to make kernel malloc() retryable. from <MDICKSON@CSI.compuserve.com>
1993-05-27 14:34:29 +00:00
cc5e8e00d5
add rcs ids and clean up headers
1993-05-20 03:59:08 +00:00
3cec79bb72
add counters for vmstat
1993-05-07 07:12:43 +00:00
db3da84702
add bruce evans's useracc fixes, from patchkit patch 137.
1993-05-07 07:03:50 +00:00
b235aec4f6
fix free page count determination
1993-05-07 06:59:45 +00:00
fd7c50fd61
patch from Pete Chown <pc123@cus.cam.ac.uk> to fix mmap
...
modes vs. read/write permissions problem.
1993-05-07 01:58:11 +00:00
645b67012b
make it panic, like i thought it did, for mb_map too small.
1993-04-29 08:57:02 +00:00
0b48d59355
Fix utter inanity I committed last night.
1993-04-29 00:46:21 +00:00
a203e5704b
Implement previous change more cleanly.
1993-04-28 04:00:49 +00:00
87dc98f60b
Introduce a new panic() condition in case I missed something.
1993-04-28 03:04:57 +00:00
95090f0a13
Don't use first block of swap area.
1993-04-28 03:04:25 +00:00
235bd1db44
Add consistent multiple-inclusion protection.
1993-04-19 03:45:34 +00:00
33d009a09c
Prevents you from creating double faults and such mmaping stuff on top
...
of the kernel, etc. Tested. Derived from BSDI public patches
1993-04-10 15:06:01 +00:00
4493ba8058
fixed stupid typo (missing comma) in hand-applied diff
1993-04-09 17:19:02 +00:00
a128c34f9c
From: Guido van Rooij <guido@gvr.win.tue.nl>
...
when mmapping a file, permissions are checked as it should be. When
mprotect()-ing the address range afterwards, no protection was checked
regarding the protection of the file originally opened. So
when you open /usr/bin/su RDONLY and SHARED you could afterwards change
the mmapped region to READ|WRITE. This gave the possibility to obtain
root privs obviously.
1993-04-09 17:16:30 +00:00
202313150e
changes, running on sef's machine to good effect, likely contributed
...
by torek. (sef told me to get them, and their origin...)
1993-04-09 17:13:17 +00:00
827c75471f
From: andrew@werple.apana.org.au (Andrew Herbert)
...
Here are some fixes I derived from the mach 3.0 VM system a couple of months
ago. At the time, I was giving the memory object routines a good looking
at, trying to fix the long-standing problem where vm_object_collapse()
sometimes fails to collapse objects left over from the exit of a forked
child. As bde has noted, the problem seems to occur when portions of the
parent are paged out. These "lost" memory objects, which can eat up a huge
amount of swap space, are reclaimed when the parent responsible for the
fork()s is killed.
1993-04-09 16:53:46 +00:00
3cac22992d
from sean eric fagan:
...
it seems to keep the vm system from deadlocking the system when it runs
out of swap + physical memory.
prevents the system from giving the last page(s) to anything but the
referenced "processes" (especially important is the pager process,
which should never have to wait for a free page).
1993-04-09 15:54:56 +00:00
64354bb7c6
broke out panics for *_map too small, and tried to coalesce the maps more
...
often.
1993-03-31 21:53:56 +00:00
e541169ce2
after 0.2.2 "stable" patches applied
1993-03-21 18:04:42 +00:00
61f282557f
initial import of 386bsd-0.1 sources
1993-03-21 09:45:37 +00:00
mycroft