the same pages over and over.
* Increased the priority of the page writer a bit, so that it is higher than
the one of the page daemon.
* Added a sModifiedTemporaryPages counter to let the page_writer decide how
many pages are there to write back (temporary pages would go to the swap file
and are only written back when memory is low).
* In case there are more than 1024 modified (non-temporary) pages around, the
page writer will constantly write out pages, not only when being pushed.
* The page writer now temporarily always leave out temporary pages (as long as
we don't have a swap file).
* Shuffled functions around a bit.
git-svn-id: file:///srv/svn/repos/haiku/haiku/trunk@22852 a95241bf-73f2-0310-859d-f6bbb57e9c96
on pages are recorded in a history and can be printed via the
"page_allocations" command.
* Fixed a problem in the page scrubber. It temporarily removed pages
from the free list, which could have been reserved by someone else.
When actually allocating the reserved pages, that someone could find
free and clear lists empty and would therefore rightfully panic
("Had reserved page, but there is none!").
git-svn-id: file:///srv/svn/repos/haiku/haiku/trunk@22841 a95241bf-73f2-0310-859d-f6bbb57e9c96
non-reserve case.
* vm_page_allocate_page() could also lose a stolen page in case more free pages were available
after steal_pages() was called.
git-svn-id: file:///srv/svn/repos/haiku/haiku/trunk@22834 a95241bf-73f2-0310-859d-f6bbb57e9c96
vm_page_reserve_pages() and vm_page_allocate_page() will now steal pages from
the inactive queue as needed.
* We currently never steal active pages anymore, but this might need to be
revised later (therefore, the page scanner never waits anymore, but uses
mutex_trylock() to lock a cache).
* The page scanner and writer now both run at normal priority - let's see how
that will work out.
* Introduced an inactive queue.
* Instead of shuffling pages around in the queue (and therefore destroying LRU)
the page stealing mechanism now uses a marker page to be able to release the
page lock without losing its position in the queue.
* The page writer now always grabs the whole release count of the semaphore, so
that there won't be a huge backlog to catch up with.
* vm_page_num_free_pages() now also includes the inactive queue as well as the
reserved pages (they are no longer regarded as free pages).
* Added a insert_page_after() function that inserts a page after another one,
needed by the marker code.
* clear_page() now gets a vm_page instead of a physical address which simplified
some code.
* Removed superfluous initialization of the queues (if those aren't zeroed on
start, we would have serious problems, anyway).
* Removed old and unimplemented dump_free_page_table() ("free_pages") KDL
command.
git-svn-id: file:///srv/svn/repos/haiku/haiku/trunk@22506 a95241bf-73f2-0310-859d-f6bbb57e9c96
vm_remove_all_page_mappings() all computed the wrong virtual address for the
page in question, and would therefore return incorrect data!
Introduced a virtual_page_address() function that is now used by all of them.
So that's why we were still "losing" modified flags - it took me some hours
to find those (sometimes processes died after stealing pages)...
git-svn-id: file:///srv/svn/repos/haiku/haiku/trunk@22498 a95241bf-73f2-0310-859d-f6bbb57e9c96
it intended to. That resulted in more writable pages where you wouldn't want
them (ie. allowing the area to change pages in lower caches).
* We were losing modified pages: vm_unmap_pages() sometimes has to preserve
the modified flag (eg. when called from page fault).
* Both of these were responsible that stealing active pages would crash
applications - even if less likely, this could also have happened when
stealing inactive pages. Therefore, I've activated stealing active pages
again.
* The page writer now pushes the pages of busy vnodes to the end of the queue,
so that it won't pick them up again too soon (the vnode destruction would
be in the process of writing those pages back, anyway).
* The page thief now triggers the page writer to run once it has to steal
active pages. This might be a bit too aggressive, though.
git-svn-id: file:///srv/svn/repos/haiku/haiku/trunk@22495 a95241bf-73f2-0310-859d-f6bbb57e9c96
* The page_thief() will no longer steal any pages in B_LOW_MEMORY_NOTE state,
only in the more serious cases.
* I've disabled stealing active pages for now again; there seem to be some
problems with it (either with how we do it, or with other stuff).
* vm_page_schedule_write_page() now always releases the page writer semaphore,
resulting in many more written pages - this isn't optimal as long as there is
no I/O scheduler, but before it was much too rare when there are many dirty
pages.
* Customized the thread priorities a bit to make the page scanner/thief/writer
experience a bit nicer with our current scheduler.
* vm_page_reserve_pages() would return too early, it did not test if really
enough pages are free.
* Under certain circumstances, the wakeup from vm_page_reserve_pages() did not
work - we now always notify in vm_page_unreserve_pages() to work around that
problem.
* Checked if the page reservations are done when needed and are always balanced
for the whole kernel.
* vm_page_allocate_page() now panics if it can't deliver a reserved page.
* vm_page_allocate_page_run() can no longer steal reserved pages.
* With all of those changes, I could finally copy a file in emulation, ie. the
Luposian bug should finally be fixed, even though the system might still not
be perfectly stable under low memory. Will do some more testing.
* write_page() no longer prints something on failure.
* Dumping a whole page queue will now also write the cache type of each page.
git-svn-id: file:///srv/svn/repos/haiku/haiku/trunk@22486 a95241bf-73f2-0310-859d-f6bbb57e9c96
* vm_low_memory_state() now periodically recomputes the state in case the
low memory thread is waiting for something.
git-svn-id: file:///srv/svn/repos/haiku/haiku/trunk@22483 a95241bf-73f2-0310-859d-f6bbb57e9c96
allocate with the vm_cache locked - this is necessary to be able to steal pages
from itself (large files...).
The system doesn't actually lock up anymore, but it still renders itself unusable;
obviously the page thief does not work correctly, yet. The rest of the experience
is created by our current scheduler (the page thief runs and runs, but it doesn't
free any pages anymore).
git-svn-id: file:///srv/svn/repos/haiku/haiku/trunk@22466 a95241bf-73f2-0310-859d-f6bbb57e9c96
* his has to be used by the page writer to make sure the vnode is still valid.
* This should have been the final nail on the Luposian bug - I haven't tested
it yet, but we'll certainly see :-)
git-svn-id: file:///srv/svn/repos/haiku/haiku/trunk@22462 a95241bf-73f2-0310-859d-f6bbb57e9c96
interfere with the page thief, we always need to have reserved a page for
this upfront. I introduced a function to the vm_translation_map layer that
estimates how much pages a mapping might need at maximum. All functions that
map a page now call this and reserve the needed pages upfront.
It might not be a nice solution, but it works.
* The page thief could run into a panic when trying to call vm_cache_release_ref()
on a non-existing (NULL) cache.
* Also, it will now ignore wired active pages.
* There is still a race condition between the page writer and the vnode
destruction - writing a page back needs a valid vnode, but that might just
have been deleted.
git-svn-id: file:///srv/svn/repos/haiku/haiku/trunk@22455 a95241bf-73f2-0310-859d-f6bbb57e9c96
(currently, wired pages don't always have the PAGE_STATE_WIRED, but a
wired_count).
git-svn-id: file:///srv/svn/repos/haiku/haiku/trunk@22449 a95241bf-73f2-0310-859d-f6bbb57e9c96
the unmapped page.
* This is needed by everyone who calls this to make sure modifications to a
page aren't ignored. Namely, the page scanner and the page thief were
affected.
* Cleaned up locking the page's cache a bit in page_thief(); there is now
a helper class that takes care of everything.
git-svn-id: file:///srv/svn/repos/haiku/haiku/trunk@22438 a95241bf-73f2-0310-859d-f6bbb57e9c96
* the page writer don't allow to block, while all other writers do. This fixes
bug #1509. The reason the page writer needs this is because it marks several
pages from different caches as busy.
* Fixed a warning about ASSERT being defined already in BFS, since
util/DoublyLinkedList.h now includes debug.h.
git-svn-id: file:///srv/svn/repos/haiku/haiku/trunk@22434 a95241bf-73f2-0310-859d-f6bbb57e9c96
could be written back. This should stop the page thief from stealing active
pages that cannot be recreated easily :-)
git-svn-id: file:///srv/svn/repos/haiku/haiku/trunk@22404 a95241bf-73f2-0310-859d-f6bbb57e9c96
necessary because low memory handlers might need to wait for other services
while we need to make sure the page thief can always steal pages.
* Made the page thief more aggressive when memory is critical: added a third
run where it will also steal a small amount of active pages to keep things
going.
* The page thief did not maintain the cache's reference correctly in all
situations.
git-svn-id: file:///srv/svn/repos/haiku/haiku/trunk@22382 a95241bf-73f2-0310-859d-f6bbb57e9c96
components - now, we divide the initialization of that service into two parts
which allows this (before, most handlers were gone after boot).
* Added debugger command that dumps the low memory handlers.
* The slab allocator now registers its low memory handler with a higher
priority, so that it'll run before the potentially heavier ones.
git-svn-id: file:///srv/svn/repos/haiku/haiku/trunk@22374 a95241bf-73f2-0310-859d-f6bbb57e9c96
inserted into the list according to their priority; higher priority
handlers are now called first.
git-svn-id: file:///srv/svn/repos/haiku/haiku/trunk@22359 a95241bf-73f2-0310-859d-f6bbb57e9c96
Apparently not everyone checks the page state for PAGE_STATE_BUSY.
vm_page_write_modified_pages() does now at least.
* Fixed typos in comments.
git-svn-id: file:///srv/svn/repos/haiku/haiku/trunk@22349 a95241bf-73f2-0310-859d-f6bbb57e9c96
* We now have a page writer that takes some pages from the modified queue
and writes it back every few seconds. It can be triggered by the page
scanner to do that more often, though. That mechanism can be greatly
improved once we have our I/O scheduler working.
* Removed vm_page_write_modified_page() again - it was all "eaten up" by
the page writer.
* Reworked vm_page_write_modified_pages() a bit: it now uses
vm_test_map_modification() and vm_clear_map_flags() instead of the
iterating over all areas which wouldn't even work correctly.
The code is much simpler now, too.
* You usually put something to the tail of a queue, and remove the contents
from the head, not vice versa - changed queue implementation to reflect this.
* Additionally, there is now a enqueue_page_to_head() if you actually want the
opposite.
* vm_page_requeue() allows you to move a page in a queue to the head or tail.
* Replaced vm_clear_map_activation() with vm_clear_map_flags() which allows
you to clear other flags than PAGE_ACCESSED.
* The page scanner dumps now some arguments with each run.
* Removed the old disabled pageout_daemon() from NewOS.
git-svn-id: file:///srv/svn/repos/haiku/haiku/trunk@22348 a95241bf-73f2-0310-859d-f6bbb57e9c96
to the private VM types are including vm_types.h now.
* Removed vm_page, vm_area, vm_cache, and vm_address_space typedefs; it's
cleaner this way, and the actual types are only used in C++ files now,
anyway.
* And that caused changes in many files...
* Made commpage.h self-containing.
* Minor cleanup.
git-svn-id: file:///srv/svn/repos/haiku/haiku/trunk@22329 a95241bf-73f2-0310-859d-f6bbb57e9c96
in enqueue_page() - it doesn't really belong there.
* Minor cleanup.
git-svn-id: file:///srv/svn/repos/haiku/haiku/trunk@22328 a95241bf-73f2-0310-859d-f6bbb57e9c96
For some reason, shell wildcards do not include deleted files... :)
git-svn-id: file:///srv/svn/repos/haiku/haiku/trunk@22327 a95241bf-73f2-0310-859d-f6bbb57e9c96
opaque types for C.
* As a result, I've renamed some more source files to .cpp, and fixed
all warnings caused by that.
git-svn-id: file:///srv/svn/repos/haiku/haiku/trunk@22326 a95241bf-73f2-0310-859d-f6bbb57e9c96
introduced a new vm_page_write_modified_page().
* Resolved a TODO: vm_page_write_modified_pages() did not mark a to be
written page busy but unlocked its cache which could let someone else
steal that page in the mean time.
* Moved the logic when to move a page to the active or inactive queue to
a new function move_page_to_active_or_inactive_queue().
* Moved page_state_to_string() to vm_page(); it's now also used by the
"page" and "page_queue" KDL commands.
* Made the output of the "page_queue list" command more useful.
git-svn-id: file:///srv/svn/repos/haiku/haiku/trunk@22323 a95241bf-73f2-0310-859d-f6bbb57e9c96
are locked, there is now a vm_page_reserve_pages() call to ensure upfront that
there is a page for me when I need it, and may have locked some caches.
* The vm_soft_fault() routine now makes use of that feature.
* vm_page_allocate_page() now resets the vm_page::usage_count, so that the file
cache does not need to do this in read_chunk_into_cache() and
write_chunk_to_cache().
* In cache_io() however, it need to update the usage_count - and it does that
now. Since non-mapped caches don't have mappings, the page scanner will punish
the cache pages stronger than other pages which is accidently just what we
want.
git-svn-id: file:///srv/svn/repos/haiku/haiku/trunk@22319 a95241bf-73f2-0310-859d-f6bbb57e9c96
* Removed the vm_cache/vm_store ref_count duality that besides being a bit ugly
also created the page dameon cache retrieval problem: now, only areas (and
cache consumers) retrieve a reference to the store (and therefore, the vnode).
The page daemon doesn't need to care about this at all anymore, and the pseudo
references of the vm_cache could be removed again.
* Rearranged deletion of vnodes such that its ID can be reused directly after
fs_remove_vnode() has been called.
* vm_page_allocate_page() no longer panics when it runs out of pages, but just
waits for new pages to become available using the new sFreeCondition condition
variable - to make sure this happens in an acceptable time frame, it'll
trigger a run of the low memory handlers.
* Implemented a page_thief() that steals inactive pages from caches and puts
them into the free queue. It runs as a low memory handler.
* The file cache now sets the usage count on the pages it inserts into the
cache (needs some rework though, cache_io() doesn't do it yet).
* Instead of panicking, the kernel will currently dead lock in low memory
situations, since BFS does a bit too much in bfs_release_vnode().
* Some minor cleanup.
git-svn-id: file:///srv/svn/repos/haiku/haiku/trunk@22315 a95241bf-73f2-0310-859d-f6bbb57e9c96
* The page daemon now runs independently from the low memory state.
* It now also increases its frequency and scanned pages, as well as decreases
the interval between runs with increasing memory pressure.
git-svn-id: file:///srv/svn/repos/haiku/haiku/trunk@22314 a95241bf-73f2-0310-859d-f6bbb57e9c96
* vm_low_memory() no longer calls call_handlers() directly, but just triggers
a run immediately.
git-svn-id: file:///srv/svn/repos/haiku/haiku/trunk@22313 a95241bf-73f2-0310-859d-f6bbb57e9c96
* Fixed vm_page_allocate_page_run(): it did not take the pageState into account,
and would therefore return uninitialized memory (ie. B_CONTIGUOUS areas would
contain garbage).
Now, it stores if a page is cleared in a new vm_page::is_cleared field.
* Some cleanup.
git-svn-id: file:///srv/svn/repos/haiku/haiku/trunk@22306 a95241bf-73f2-0310-859d-f6bbb57e9c96
freshly booted, it would already contain > 20000 pages. The size is
now initialized to half of the available pages. Ideally it would
grow/shrink dynamically, though.
* Changed the hash function to yield a better distribution.
git-svn-id: file:///srv/svn/repos/haiku/haiku/trunk@22211 a95241bf-73f2-0310-859d-f6bbb57e9c96
a reference to a by them not yet referenced cache was not correct.
They only incremented the reference count, but a vnode cache reference
includes also a vnode reference. In case of the page daemon this would
cause vnode references to be lost (causing bug #1465).
* The page daemon used an unsafe method to access a yet unreferenced
page cache. There was nothing that prevented the cache from being
deleted while the page daemon tried to get a reference. The
vm_page::cache field is now protected by the page cache table
spinlock, too, which the new function
vm_cache_acquire_page_cache_ref(), used by the page daemon, also
acquires while trying to get the reference.
git-svn-id: file:///srv/svn/repos/haiku/haiku/trunk@22208 a95241bf-73f2-0310-859d-f6bbb57e9c96
* The vm_translation_map is now correctly held in all of the vm_ mapping
functions.
* Removed the old vm_daemons.c file - there is now a new vm_daemons.cpp
which contains the beginnings of our new page daemon.
So far, it's pretty static and not much tested. What it currently does
is to rescan all pages in the system with a two-handed clock algorithm
and push pages into the modified and inactive lists.
* These inactive pages aren't really stolen yet, even though their mappings
are removed (ie. their next access will cause a page fault). This should
slow down Haiku a bit more, great, huh? :-)
* The page daemon currently only runs on low memory situations, though.
git-svn-id: file:///srv/svn/repos/haiku/haiku/trunk@22156 a95241bf-73f2-0310-859d-f6bbb57e9c96
DEBUG_CACHE_LIST) that prints an unspectacular list of pointers to all
existing caches. Feel free to extend.
* Enhanced MultiAddressSpaceLocker:
- It supports choosing between read and write lock per address space,
now.
- Added AddAreaCacheAndLock(), which adds the address spaces of all
areas that are attached to a given area's cache, locks them, and
locks the cache. It makes sure that the area list didn't change in
the meantime and optionally also that all areas have their
no_cache_change flags cleared.
* Changed vm_copy_on_write_area() to take a cache instead of an area,
requiring it to be locked and all address spaces of affected areas to
be read-locked, plus all areas' no_cache_change flags to be cleared.
Callers simply use MultiAddressSpaceLocker:: AddAreaCacheAndLock() to
do that. This resolves an open TODO, that the areas' base, size, and
protection fields were accessed without their address spaces being
locked.
* vm_copy_area() does now always insert a cache for the target area. Not
doing that would cause source and target area being attached to
the same cache in case the target protection was read-only. This
would make them behave like cloned areas, which would lead to trouble
when one of the areas would be changed to writable later.
* Fixed the !writable -> writable case in vm_set_area_protection(). It
would simply change the protection of all mapped pages for this area,
including ones from lower caches, thus causing later writes to the
area to be seen by areas that shouldn't see them. This fixes a problem
with software breakpoints in gdb. They could cause other programs to
be dropped into the debugger.
* resize_area() uses MultiAddressSpaceLocker::AddAreaCacheAndLock() now,
too, and could be compacted quite a bit.
git-svn-id: file:///srv/svn/repos/haiku/haiku/trunk@22152 a95241bf-73f2-0310-859d-f6bbb57e9c96
space - but they were released upon deletion. It's probably not really
needed, but now all reserved areas also grab a reference to their address
space.
* Rearranged team tear down to be a bit more sane: the I/O context is removed
first (where semaphores/areas/ports/whatever might still be used), and the
address space is deleted last.
* delete_area() can now remove its address space reference again (due to the
two changes above), and therefore fixes bug #1374.
* cleaned up vm_address_space.c a bit (no functional change there, though).
git-svn-id: file:///srv/svn/repos/haiku/haiku/trunk@22130 a95241bf-73f2-0310-859d-f6bbb57e9c96
* vm_soft_fault() no longer touches the page source cache in case fault_get_page() failed.
* fault_find_page() now unlocks and releases the cache if reading in a page failed.
git-svn-id: file:///srv/svn/repos/haiku/haiku/trunk@22016 a95241bf-73f2-0310-859d-f6bbb57e9c96
if the page was already in the "modified" list before. Also, the source page (which is
either mapped directly or copied to the target page) is no longer marked busy before its
final destiny is decided (it didn't have any effect, anyway, since we had its cache
locked for the whole time, but it now preserves the modified state). This fixes bug #1369.
* vm_cache_write_modified() now filters out temporary caches (it's currently called on area
deletion).
git-svn-id: file:///srv/svn/repos/haiku/haiku/trunk@21971 a95241bf-73f2-0310-859d-f6bbb57e9c96
* Removed a few instances where the page state was set busy directly after
allocating it. This is a no-op, since a page is always busy after
allocation.
git-svn-id: file:///srv/svn/repos/haiku/haiku/trunk@21875 a95241bf-73f2-0310-859d-f6bbb57e9c96
it's reference to the vm_address_space - luckily, it doesn't even need a
reference, since it always runs in the current address space, which cannot
go away for obvious reasons.
git-svn-id: file:///srv/svn/repos/haiku/haiku/trunk@21865 a95241bf-73f2-0310-859d-f6bbb57e9c96
* fixes a dead lock in vm_soft_fault() - the locking scheme enforces you to
lock the address space before a vm_cache, not the other way, around. Since
we need to lock the cache that has our page in fault_get_page(), we violated
that scheme by relocking the address space in order to get access to the
vm_area. Now, we read lock the address space during the whole page fault;
added a TODO that explains why this might not really be desirable, if
we can avoid it (the only way would be to reverse that locking scheme
which would potentially cause the more busy vm_cache locks to be held
longer).
* vm_copy_area() uses the MultiAddressSpaceLocker, but actually forget to
call Lock() on it...
* delete_area() leaks vm_address_space references - but fixing this currently
causes other problems to be investigated; I'll open a bug for that.
git-svn-id: file:///srv/svn/repos/haiku/haiku/trunk@21862 a95241bf-73f2-0310-859d-f6bbb57e9c96
