Commit Graph

374 Commits

Author SHA1 Message Date
thorpej 1e3e1bfe09 Include the NFS option header. 1998-02-19 00:55:04 +00:00
cgd 3bbb7f7d45 Move pmap_map() function definition to MD headers, as appropriate. It's
an internal function, and the VM system shouldn't try to prototype it.
(Note that some ports _don't_ prototype it.)
1998-02-18 02:05:32 +00:00
mrg d90485202c - add defopt's for UVM, UVMHIST and PMAP_NEW.
- remove unnecessary UVMHIST_DECL's.
1998-02-10 14:08:44 +00:00
thorpej 223caea788 pmap_page_index() is not used in the MACHINE_NEW_NONCONTIG case. 1998-02-08 18:47:06 +00:00
thorpej d319b28938 Round allocations to page size in vm_bootstrap_steal_memory(). 1998-02-08 18:24:52 +00:00
mrg 8f7ee94e13 initial import of the new virtual memory system, UVM, into -current.
UVM was written by chuck cranor <chuck@maria.wustl.edu>, with some
minor portions derived from the old Mach code.  i provided some help
getting swap and paging working, and other bug fixes/ideas.  chuck
silvers <chuq@chuq.com> also provided some other fixes.

these are the changes to the old Mach VM system to allow both to be
available in the kernel.
1998-02-06 00:14:43 +00:00
mrg 5ef401eaaf oops, forgot to add this file earlier. 1998-02-05 14:58:45 +00:00
ross c297f000b7 NFS swap fix. Most files here are modified purely to add a tsleep() wmesg
string. The actual functional changes are:

	1. in vm_page_alloc(), restore some long lost code by Paul Kranenburg
	that reserves the last few pages for kernel objects, adding an
	additional escape for the pageout daemon.

	2. In vm_wait() (replaces VM_WAIT), recognize the amusing fatal
	deadlock where the pagedemon sleeps on a channel that only it ever
	wakes up, and add a timeout and printf. Mod 1 should generally
	prevent this from happening in any case.

	3. Fix a livelock in vm_pageout_page() caused by a pre-wakeup of
	page consumers prior to the pageout, which can easily fail over
	memory issues in NFS and IP code. Also, ++ cnt.v_pageouts only
	if the pageout succeeded.
1998-01-31 04:02:39 +00:00
drochner c0d5e99849 fix overflow of the vm_physmem[] array 1998-01-20 12:34:35 +00:00
thorpej e76d5a2fe1 Oops, don't need vm_page_physget() if PMAP_STEAL_MEMORY. 1998-01-10 03:02:58 +00:00
thorpej 51902c1094 In vm_page_physload(), check page_shift not cnt.v_page_size for
having been initialized.
1998-01-09 08:50:00 +00:00
thorpej 4001cf8069 In vm_mem_init(), don't test for cnt.v_page_size being == 0. It is perfectly
legitimate for this to be set if vm_set_page_size() hasn't been called (it's
how the page size is communicated to vm_set_page_size()).  Instead, test
for page_shift == 0; this is one of the values that vm_set_page_size()
initializes.
1998-01-09 06:00:50 +00:00
thorpej 54f8b4bf57 In the new non-contig case, redefine pmap_steal_memory() as an alternative
to vm_bootstrap_steal_memory(), since we may need to steal memory from
a direct-mapped segment at this early stage of the game (on e.g. MIPS
and Alpha).

This interface is activated by defining PMAP_STEAL_MEMORY in <machine/pmap.h>.
Otherwise, the pmap_virtual_space() prototype is in-scope for use by
vm_bootstrap_steal_memory().
1998-01-08 23:28:04 +00:00
thorpej 15efce41a3 Zorch pmap_startup() prototype. It was never implemented in machine-dependent
code, and is inlined in the MACHINE_NONCONTIG case of vm_page_bootstrap() now.
1998-01-08 23:13:05 +00:00
thorpej 589a1e440e Clean up and consolidate the old non-contig bootstrap functions, and
rename pmap_steal_memory() to vm_bootstrap_steal_memory().
1998-01-08 23:03:24 +00:00
mrg a20d56e92e add new version of non contiguous memory code, written by chuck cranor,
called "MACHINE_NEW_NONCONGIG".  this is required for UVM, the new VM
system (also written by chuck) that is coming soon.  adds new functions:
	vm_page_physload() -- tell the VM system about an area of memory.
	vm_physseg_find() -- returns index in vm_physmem array that this
		address is in.
and several new versions of old functions/macros defined in vm_page.h.

this is the MI portion.  sparc, and then later i386 portions to come.
all other ports need to change to this ASAP!  (alpha is already being
worked on)
1998-01-08 11:36:16 +00:00
thorpej d8ee45898e Garbage-collect __VM_PMAP_HACK. 1998-01-06 08:36:23 +00:00
perry 24920eefb1 RCSID Police. 1998-01-05 19:40:40 +00:00
thorpej 5739c88a75 Implement address space sharing (by keeping ref counts on the vmspace
structure).  Many thanks to Chuck Cranor for debugging assistence.
1998-01-03 02:53:00 +00:00
thorpej b9f1b716f3 Now that all ports have pmap_activate(), and it has an identical interface,
prototype it in <vm/pmap.h>
1998-01-03 01:12:59 +00:00
thorpej a322314f51 Split out the code that prepares a VM space for exec into a new
vmspace_exec() function.
1997-12-31 07:47:41 +00:00
pk 443a2a02f4 "pool" routines moved to kern/subr_pool.c 1997-12-15 11:18:41 +00:00
pk 5d1352c448 Guard against synchronous I/O completetion per swap-device. 1997-12-02 13:47:37 +00:00
pk 1736bb9de8 Make the code a bit more readable. 1997-12-01 17:10:48 +00:00
pk d4a6f93147 Make sure the damned buffer heads are initialized... 1997-12-01 16:33:15 +00:00
pk 46f604677d Avoid several race conditions, especially when a transaction ends in error.
Guard sw_reg_start() against synchronous completions of the device
strategy routines, avoiding possibly boundless recursion.
1997-11-29 00:35:43 +00:00
thorpej 97d2a58201 Fix slight argument bogosity with getgroups(), setgroups(), select(),
and swapctl().  For the former three, they use an 'int' in their user-land
prototype which was a 'u_int' in the kernel, which screwed up automatic
generation/checking of lint syscall stubs.  For the latter, the user-land
prototype uses a "const char *", but the syscall just used "char *".

From Chris Demetriou <cgd@pa.dec.com>.
1997-11-04 21:24:14 +00:00
thorpej 14cd99880c Fix the shared library versioning snafu caused by the recent changes
to the stat(2) family and msync(2).  This uses a primitive function
versioning scheme.

This reverts the libc shared library major version from 13 to 12, and
adds a few new interfaces to bring us to libc version 12.20.

From Frank van der Linden <fvdl@NetBSD.ORG>.
1997-10-20 22:05:06 +00:00
pk d24dbbbfc4 A `simplelock' suffices for most swap device list traversals, as noted
by Chuck Cranor.

Use an exclusive (long-term) lock before applying changes to the swap
lists. This lock is also used by the stats system call because of possible
blocking in copyout().
1997-10-17 19:06:05 +00:00
christos b82d33e84c PR/4269: Chris Demetriou: msync is missing the MS_* flags.
Also change caddr_t syscall arguments to void *
1997-10-16 23:29:26 +00:00
pk dfbda96da0 Use `long' to cast a pointer to an integer and back. 1997-10-14 08:50:18 +00:00
pk e95c79e586 Keep a private pool of auxiliary resources (vndxfer & vndbuf structures)
used to setup I/O to regular files.

Implemented in a somewhat generic way, for what it's worth.
1997-10-13 19:15:19 +00:00
mrg 55c881fbd2 when creating the per-device extent, use from 0 to nblks, and then
extent_alloc_region() any skipped space.  this is (basically) what
the previous swap code did (with rmaps rather than extents).
1997-10-10 13:16:24 +00:00
mrg 74f573d85d remove advertising clause from all my licenses. 1997-10-10 05:39:47 +00:00
chuck 081b0ea365 correct a comment 1997-09-22 15:22:12 +00:00
enami 8a45521059 Convert RTC_OFFSET, DEVPAGER, SWAPPAGER and VNODEPAGER to be declared
by defopt.
1997-09-20 00:44:05 +00:00
thorpej 9a31fc02fe Don't forget to increase the free page count in vm_page_free_memory(),
pointed out by Mark Brinicombe <mark@causality.com>.
1997-09-16 00:08:09 +00:00
pk 9f1374779b Remove a comment that's incorrect since last commit. 1997-09-10 18:26:33 +00:00
pk 2d8fd853ae Unlocking the target object within vm_object_overlay() is safe, since
the backing object is protected the OBJ_FADING flag.

In vm_object_deallocate(), gain a reference to the object we try to
collapse (per Manuel Bouyer).
1997-09-08 21:23:22 +00:00
chuck 1821ecc346 import sys_minherit() from openbsd. 1997-09-08 18:19:45 +00:00
pk c5d6596b7c vm_object_deallocate():
Unlock the object cache before looking for collapse candidates.
	Use vm_object_lock_try() to obtain a lock on a collapse candidate.
	This avoids a deadlock due to recursively entering
	vm_object_deallocate() in case a backing object goes away.

vm_object_collapse() and helper functions:
	The object passed in is locked; make sure it is locked when
	we leave vm_object_collapse().

   XXX: Currently, the object is not unlocked during while we possibly
	wait for pager data. We need to consider the effects of
	simultaneous calls to vm_object_collapse() on the same object
	in case unlocking is necessary.
1997-09-07 21:19:41 +00:00
pk 9bec4e7918 Don't hold on to object and queue locks while sleeping. 1997-09-07 20:41:59 +00:00
pk 0c386233d4 Adapt a diagnostic to big-endian machines.
(Not that it actually works on little-endians, but we can easily make it not
work on big-endians all the same).
1997-09-07 20:39:23 +00:00
pk fff424c798 Add missing `unlock(&vm_object_list_lock);' 1997-09-07 19:31:52 +00:00
mrg 66d1e506df remove last vestigates of __FORK_BRAINDAMANGE 1997-08-27 02:35:38 +00:00
drochner 448446ec38 Don't make coredump of areas which are mmap()ed from a device.
This can have bad side effects.
Closes PR kern/3252.
1997-07-22 10:06:43 +00:00
fvdl 933d4afe84 Make vm_map_clear_recursive behave exactly like the old lock_clear_recursive.
This doesn't make a difference in the way that it's currently being called,
but it's more intuitive.
1997-07-20 23:23:45 +00:00
fvdl 5a4803bc55 XXX - some userland code includes this, needing <sys/time.h>, so
include it, for the time being.
1997-07-07 10:57:34 +00:00
fvdl 206e9582ef Move lock.h and kern_lock.c to their rightful places, as in Lite2. Make
the VM code use the new interface.
1997-07-06 12:38:25 +00:00
drochner 8c0a9bab00 Don't cast 64bit (off_t) file sizes to vm_offset_t (32bit on many
architectures), truncate them intelligently instead.
The truncation is done centralized in vnode_pager.c.
This prevents from wrap-over effects when parts of large (>2^32 byte) files
are mmapped.
Don't allow to mmap above the numerical range of vm_offset_t.
This is considered a temporary solution until the vm system handles the
object sizes/offsets more cleanly.
1997-07-04 20:22:09 +00:00
thorpej 657f118b7f Update the status of a couple of items here... specifically, more aggressive
object collapse (DONE), machine-dependent hook for virtual memory allocation
(DONE - PMAP_PREFER()), and better coherency between page and buffer
caches (A LITTLE BETTER - we sync up the vnode pager in the sync(2)
system call now).
1997-06-26 01:02:07 +00:00
mrg 781dfea84f back out previous. look in unistd.h. 1997-06-23 01:15:03 +00:00
christos 796badce6b PR/3772: Matthias Scheler: Missing swapctl prototype. 1997-06-22 19:35:55 +00:00
pk 98ce8ab879 Remove __FORK_BRAINDAMAGEd code; it's no longer needed. 1997-06-19 20:54:15 +00:00
thorpej 3bd11220b6 Don't print a message if the user attempts SWAP_OFF. Return an error
instead.
1997-06-17 03:37:07 +00:00
thorpej c61704d37c Remove the code that sets dumpdev, for two reasons:
- It was incorrect.  A dev_t should _never_ be compared against
  NULL, as 0 is a valid value (major 0, minor 0).
- The operator mave have _intentionally_ set dumpdev to "none"
  (i.e. NODEV), and setting it the first time a VBLK swap device
  is added would break the semantics.
setroot() deals with all dump device selection.  Dumps and swap are now
in no way related.
1997-06-16 20:21:55 +00:00
mrg c78d0720ff - set dumpdev if not already set
- don't use nocoalease (from thorpej).
- fix priority list management (from pk).
- update comment for swapoff.
- free swap device extent name safely (from thorpej).
1997-06-16 13:35:17 +00:00
pk e5b0a8820b As long as we actually use `swapdev_vp', initialize it. 1997-06-13 15:37:17 +00:00
mrg 21cc3ddfd5 bring mrg-vm-swap2 onto mainilne. 1997-06-12 14:51:20 +00:00
thorpej 433c3cae0f Pull thorpej-bus-dma branch into mainline. 1997-06-06 23:10:23 +00:00
gwr 09da7e273e Eliminate vmspace.vm_pmap and all references to it unless
__VM_PMAP_HACK is defined (for temporary compatibility).
The __VM_PMAP_HACK code should be removed after all the
ports that define it have removed all vm_pmap references.
1997-05-16 21:39:50 +00:00
thorpej e73580920d Change to not require including <sys/dmap.h> 1997-04-17 00:12:50 +00:00
thorpej 27b49a0ede XXX Kludge for PowerPC; need to use kernel vfprintf(). 1997-04-17 00:12:08 +00:00
thorpej 90ece8ebcb Declare dmmin, dmmax, swapmap, and nswapmap here. 1997-04-17 00:08:50 +00:00
mycroft 448427b7ba Document a requirement of vm_object_bypass(). 1997-04-08 22:35:49 +00:00
mycroft 69e0aacce3 Remove the `backing_object still referenced' panic, since there's an expected
race with vnode_pager_sync().
1997-04-07 01:57:14 +00:00
mycroft 3a474fa002 Remove a printf() that's no longer needed. 1997-03-03 20:56:12 +00:00
mycroft 700dee447e DTRT if vm_object_lookup() returns NULL. This may happen while an object is
being created or destroyed.
1997-02-27 17:49:05 +00:00
tls fad978b853 a mutex on vnode_pager_sync avoids another race condition, only possible if sync were called again while we were sleeping. This should be the last of this nest of bugs. From mycroft. 1997-02-26 02:26:19 +00:00
thorpej cbb4319fb9 vm_object_shadow(): Add a DIAGNOSTIC check for shadowing a null object.
From Charles Hannum <mycroft@netbsd.org>
1997-02-25 23:28:09 +00:00
thorpej a49641e340 vm_map_copy_entry(): Don't set "needs_copy" on a zero-fill entry.
From Charles Hannum <mycroft@netbsd.org>
1997-02-25 23:27:08 +00:00
thorpej 7327c8cf62 From Charles Hannum:
During a collapse, abort if the object is paging; the pagedaemon can
trigger a collapse, causing deadlock.  (XXX)
1997-02-24 22:19:26 +00:00
mrg 82f19ac36c lite2 1997-02-23 09:58:53 +00:00
mrg 3b9c3fff2e restructure vm_object_overlay() to collapse objects in more cases.
change one diag panic() to a diag printf().  from mycroft.
1997-02-23 09:01:37 +00:00
mrg 8eba4aeacb don't clean fading objects. from mycroft. 1997-02-23 08:56:57 +00:00
thorpej a49beaa16a Possesive, Jason... possesive.... 1997-02-22 04:36:48 +00:00
thorpej 75be8504f4 More good stuff from Charles:
Eliminate a nasty race condition in vnode_pager_sync() by making
two passes.  First, find all of the objects to be cleaned and reference
them.  Second, traverse this list objects, cleaning them if there are
references other than us.
1997-02-22 04:11:42 +00:00
thorpej 255b9d74d8 Rewrite the object chain collapse code. Bugs fixed are too numerous
to describe here.  This should fix the problems with "hanging processes"
people have seen since the original object collapse code was committed.

From Charles Hannum <mycroft@netbsd.org>
1997-02-21 20:30:49 +00:00
thorpej 235a2d71f7 Use vm_object_paging() to test if an object has pages in transit.
From Charles Hannum <mycroft@netbsd.org>
1997-02-21 20:27:21 +00:00
thorpej dd8d0f1f00 Define a macro to test if an object has pages in transit. Clarify
some arithmetic, and nuke an unneeded prototype.

From Charles Hannum <mycroft@netbsd.org>.
1997-02-21 20:26:03 +00:00
thorpej 17b60fc6e9 In vnode_pager_sync(), we need to vm_object_deallocate() after cleaning
pages.  Noticed by Thor Simon <tls@netbsd.org>, confirmed by
Charles Hannum <mycroft@netbsd.org>
1997-02-21 20:22:39 +00:00
mrg 42ae012452 This is all from mycroft:
Create macros (with names borrowed from Mach 3) to manipulate
object->paging_in_progress, and use them.  When decreasing the
paging count, always make sure to wake up anyone waiting.
Fixes `vospgc', `vosca1', and `vosca2' hangs.
1997-02-18 13:39:31 +00:00
tls 127c6b54e7 Sigh. I initially had this exactly backwards: the object needs to be locked, the vnode does *not*. Avoids panic: locking against myself quickly reported by Charles. 1997-02-17 13:10:28 +00:00
tls a507fda378 vnode_pager_sync bugs: A) must lock the object, not just the vnode. B) Don't remove the object from paging queues, because we'd have to put it back on one. The pagedaemon will eventually take care of this for us anyway. This was causing us to *lose track of pages completely* every time vnode_pager_sync() was called... sigh. Credit goes to mycroft for untangling this mess. 1997-02-17 06:22:52 +00:00
tls 4b1213d27c sync needs to clean VM objects backed by vnode pagers 1997-02-13 02:54:06 +00:00
mrg 491e8af174 minor KNF from jason. 1997-02-05 08:26:25 +00:00
mrg eadd7792d9 add thread_sleep_msg() that takes a wait message, and change thread_sleep()
into a macro for backwards compat.  alter vm_object_sleep() to take advantage
of this.
1997-02-05 08:09:45 +00:00
mrg 85328f7b35 KNF. 1997-02-05 07:48:42 +00:00
thorpej 593fdd462d NFSCLIENT -> NFS 1997-01-31 03:06:06 +00:00
thorpej b7a5b84750 Treat the nswdev == 0 case like the nswap == 0 case; don't panic, just
report no swap space.  Suggested by Leo Weppleman <leo@netbsd.org>.
1997-01-22 07:28:21 +00:00
mrg 4f2a217c2b pull up vm_object_collapse() fix from mrg_vm_object_collapse branch.
many thanks to niklas hallqvist (niklas@appli.se) for this.
1997-01-03 18:03:14 +00:00
mrg f3d9c468e1 minor KNF changes. 1997-01-03 18:00:25 +00:00
cgd 8a3333b2a9 Fix an inconsistency that came in with Lite: setrq() was renamed to
setrunqueue(), but remrq() was never renamed.  Rename remrq() to
remrunqueue().
1996-11-06 20:19:19 +00:00
cgd ee08818e7a Fix an inconsistency that came in with Lite: setrq() was renamed to
setrunqueue(), but remrq() was never renamed.  Rename remrq() to
remrunqueue().  Also, move remrunqueue() prototype from vm/vm_extern.h
to sys/proc.h, so that it's in the same place as the setrunqueue() prototype
and other related prototypes.
1996-11-06 19:33:39 +00:00
gwr f5927ca260 Minor nit in my change regarding const and non-ansi code... 1996-10-27 21:50:33 +00:00
gwr 2ada6af51b Oops, part of that last commit was not meant to go in.
The definitions of va_alist, va_dcl belong elsewhere.
1996-10-27 20:50:00 +00:00
gwr a0d9bd5bb2 The hack-ish tricks to make the ANSI-style va_start macro work in
non-ANSI mode were not quite complete.  This makes non-ANSI work.
Perhaps it would be better to eliminate this hack altogether...
1996-10-27 20:30:52 +00:00
matthias 3ff0c0553e * The kernel-stack no longer needs to be wired-down on the pc532. 1996-10-23 07:20:09 +00:00
perry 9ab047883b Fix unused variable -Wxxx failure when MINIROOTSIZE defined. 1996-10-21 01:32:38 +00:00
christos 3dc7953eec back out previous kprintf changes 1996-10-12 21:50:05 +00:00