data. This is available when BLOCK_CACHE_BLOCK_TRACING is 2 or greater.
* Completely untested as of now, though. Will do so soon.
git-svn-id: file:///srv/svn/repos/haiku/haiku/trunk@30104 a95241bf-73f2-0310-859d-f6bbb57e9c96
* Renamed DoublyLinkedList::Size() to Count(), since it actually counts the
items (ie. O(n)).
git-svn-id: file:///srv/svn/repos/haiku/haiku/trunk@29979 a95241bf-73f2-0310-859d-f6bbb57e9c96
IORequest.{h,cpp}.
* Introduced public <io_requests.h> header. Currently it only declares the
single function BFS uses.
git-svn-id: file:///srv/svn/repos/haiku/haiku/trunk@29446 a95241bf-73f2-0310-859d-f6bbb57e9c96
* Replaced the use of offsetof() for structs that aren't PODs. Add a
offset_of_member() macro to util/khash.h because that's what it's used for
in our cases.
* Change the signature of add_debugger_command()/remove_debugger_command() on
GCC > 2 to avoid the depricated conversion from string constants to char *.
* Adding some "suggested" parenthesis. I know that not everyone likes that, but
it pointed out at least one bug that is fixed here as well.
git-svn-id: file:///srv/svn/repos/haiku/haiku/trunk@29113 a95241bf-73f2-0310-859d-f6bbb57e9c96
will now accepted and regarded as sparse file data. FileMap::_Add() also
correctly joins multiple vecs with that offset together, FileMap::Translate()
will always report offset -1 even for offsets into that extent.
* read_file_io_vec_pages() (or rather, its backend common_file_io_vec_pages())
now supports sparse files, and will just clear the memory it should read from
offset -1 instead of passing a request to the vnode.
* ext2 now correctly reports sparse files. This should close bug #2889, as well
as #975.
git-svn-id: file:///srv/svn/repos/haiku/haiku/trunk@28648 a95241bf-73f2-0310-859d-f6bbb57e9c96
the fixes applied to the kernel version of it (most notably a correctly
working block_cache_discard(), and cache_detach_sub_transaction()).
Also switched to the new notification functions, even though it still works
synchronously in the fs_shell.
* Minor cleanup.
git-svn-id: file:///srv/svn/repos/haiku/haiku/trunk@28617 a95241bf-73f2-0310-859d-f6bbb57e9c96
in the cache is the same as in the condition variable. It now uses the
low-contention sCachesLock to accomplish this.
* Also added an ASSERT to make sure the notification is no longer part of the
list at this point.
* Improved KDL command output.
git-svn-id: file:///srv/svn/repos/haiku/haiku/trunk@28516 a95241bf-73f2-0310-859d-f6bbb57e9c96
discarded blocks correctly as well.
* cache_detach_sub_transaction() left cached_block::original_data unchanged even
if the parent data was to become current (in case the sub transaction didn't
change the block yet). This could cause outdated blocks to be written back.
* cache_detach_sub_transaction() also set cached_block::previous_transaction
for all blocks, not just the ones with a previous transaction. This could
cause blocks to be written twice for no reason.
* cache_start_sub_transaction() did not change the num_blocks count for
discarded blocks.
* block_cache_discard() now panics if the block was already changed in the
current transaction.
* Improved test application, added more tests, revealing the above bugs in
cache_detach_sub_transaction().
* Minor cleanup.
git-svn-id: file:///srv/svn/repos/haiku/haiku/trunk@28514 a95241bf-73f2-0310-859d-f6bbb57e9c96
from the block notifier, the cache could be deleted before we have the chance
to lock it. We now lock the sCachesLock, and see if this cache is still valid.
git-svn-id: file:///srv/svn/repos/haiku/haiku/trunk@28512 a95241bf-73f2-0310-859d-f6bbb57e9c96
cache_end_transaction(), and cache_start_sub_transaction()).
* Further work on the test application, it's now actually usable, first test
passes.
* dump_block() did erroneously print 'B' for the dirty flag; now both dirty and
discard have the 'D' (3rd and 5th column).
* block_cache::LowMemoryHandler() is now private (and got an underscore prefix).
* Minor cleanup, shuffled some methods around.
git-svn-id: file:///srv/svn/repos/haiku/haiku/trunk@28508 a95241bf-73f2-0310-859d-f6bbb57e9c96
discarding their changes. This functionality currently only works correctly
when no transactions are used.
* Started test application for the block cache, doesn't do anything yet.
git-svn-id: file:///srv/svn/repos/haiku/haiku/trunk@28496 a95241bf-73f2-0310-859d-f6bbb57e9c96
* Turned the checks for all those macros to "#if"s instead of "#ifdef"s.
* Introduced macro KDEBUG_LEVEL which serves as a master setting.
git-svn-id: file:///srv/svn/repos/haiku/haiku/trunk@28248 a95241bf-73f2-0310-859d-f6bbb57e9c96
added vm_memcpy_from_physical() and vm_memcpy_physical_page(), and
added respective functions to the vm_translation_map operations. The
architecture specific implementation can now decide how to implement
them most efficiently. Added generic implementations that can be used,
though.
* Changed vm_{get,put}_physical_page(). The former no longer accepts
flags (the only flag PHYSICAL_PAGE_DONT_WAIT wasn't needed anymore).
Instead it returns an implementation-specific handle that has to be
passed to the latter. Added vm_{get,put}_physical_page_current_cpu()
and *_debug() variants, that work only for the current CPU,
respectively when in the kernel debugger. Also adjusted the
vm_translation_map operations accordingly.
* Made consequent use of the physical memory operations in the source
tree.
* Also adjusted the m68k and ppc implementations with respect to the
vm_translation_map operation changes, but they are probably broken,
nevertheless.
* For x86 the generic physical page mapper isn't used anymore. It is
suboptimal in any case. For systems with small memory it is too much
overhead, since one can just map the complete physical memory (that's
not done yet, though). For systems with large memory it counteracts
the VM strategy to reuse the least recently used pages. Since those
pages will most likely not be mapped by the page mapper anymore, it
will keep remapping chunks. This was also the reason why building
Haiku in Haiku was significantly faster with only 256 MB RAM (since
that much could be kept mapped all the time).
Now we're using a different strategy: We have small pools of virtual
page slots per CPU that are used for the physical page operations
(memset_physical(), memcpy_*_physical()) with CPU-pinned thread.
Furthermore we have four slots per translation map, which are used to
map page tables.
These changes speed up the Haiku image build in Haiku significantly. On
my Core2 Duo 2.2 GHz 2 GB machine about 40% to 20 min 40 s (KDEBUG
disabled, block cache debug disabled). Still more than factor 3 slower
than FreeBSD and Linux, though.
git-svn-id: file:///srv/svn/repos/haiku/haiku/trunk@28244 a95241bf-73f2-0310-859d-f6bbb57e9c96
{memset,memcpy_to}_physical() functions.
Mapping lots of physical pages at once as done before was an actual
problem on systems with enough RAM, as the physical page mapper can map
only 64 chunks at a time. So multiple threads could play dining
philosophers, each getting only one of two chopsticks, waiting for
another one to be freed.
git-svn-id: file:///srv/svn/repos/haiku/haiku/trunk@28220 a95241bf-73f2-0310-859d-f6bbb57e9c96
PHYSICAL_PAGE_{NO,CAN}_WAIT into an actual flag
PHYSICAL_PAGE_DONT_WAIT.
* Pass the flags through to the chunk mapper callback.
git-svn-id: file:///srv/svn/repos/haiku/haiku/trunk@27979 a95241bf-73f2-0310-859d-f6bbb57e9c96
* Use the new VMCache::Read() flags parameter to directly read into the
physical page in the page fault handler instead of mapping it first.
git-svn-id: file:///srv/svn/repos/haiku/haiku/trunk@27966 a95241bf-73f2-0310-859d-f6bbb57e9c96
ignore every other cache (starting from the first).
* The consequence of this was that no blocks were written back automatically
for those caches, and their transactions were never idle, causing bug #2781.
git-svn-id: file:///srv/svn/repos/haiku/haiku/trunk@27831 a95241bf-73f2-0310-859d-f6bbb57e9c96
Currently it only contains KDEBUG and the block cache debugging macros.
git-svn-id: file:///srv/svn/repos/haiku/haiku/trunk@27816 a95241bf-73f2-0310-859d-f6bbb57e9c96
vector at the end of the file will be aligned to the given value.
* BFS uses an alignment of 512 bytes (should be block size of the
underlying device or BFS block size, whatever is less), which should
be fine, since file data are only stored in BFS blocks. This totally
avoids any partial operations at the I/O scheduler level, thus saving
disk operations. Not that I could measure any performance difference.
Theoretically it should help a lot though, particularly when dealing
with lots of small files, since we avoid using bounce buffers, which
are (a) limited in number and (b) require copying of the data.
git-svn-id: file:///srv/svn/repos/haiku/haiku/trunk@27246 a95241bf-73f2-0310-859d-f6bbb57e9c96
global cache list before locking it. This allows to get rid of the
block_cache::deleting field, as well as simplifies some code.
* This also fixes a possible deadlock I recently introduced (on destruction,
the locking order was wrong).
* Now uses an anonymous condition variable instead.
* Moved the block_cache initialization code into a dedicated method that will
now also fail in case the low resource handler couldn't be registered (as
pointed out by Salvatore).
git-svn-id: file:///srv/svn/repos/haiku/haiku/trunk@27074 a95241bf-73f2-0310-859d-f6bbb57e9c96
be passed on to the IORequest. Most relevantly physical pages can now
be written directly by passing B_PHYSICAL_IO_REQUEST.
* Added VMCache::WriteAsync() which is supposed to write pages
asynchronously. The base class version version falls back to the
synchronous Write(). Only VMVnodeCache implements WriteAsync() ATM,
VMAnonymousCache (swap support) still has to be adjusted accordingly.
* write_page() doesn't need to map the page anymore as it can write the
physical page directly.
* Modified the page writer to write pages asynchronously. This shouldn't
have any noticeable effect yet. It will though as soon as the I/O
scheduler reorders I/O operations.
git-svn-id: file:///srv/svn/repos/haiku/haiku/trunk@27056 a95241bf-73f2-0310-859d-f6bbb57e9c96
* Added file_map_set_mode() function that you can use to keep a whole file
cached. This is needed for the swap file support: FILE_MAP_CACHE_ALL will
not only precache all file_io_vecs when called, but it will also cause all
file_map_translate() calls to fail that would require further caching (ie.
if the file size had changed).
* Updated the fs_shell file map code to the latest one (with several bug fixes).
git-svn-id: file:///srv/svn/repos/haiku/haiku/trunk@26785 a95241bf-73f2-0310-859d-f6bbb57e9c96
actual caching in the file cache, i.e. all reads and writes go directly
to the underlying device. The implementation is not quite complete,
since the VM can still add pages to the cache when the file is mmap()ed,
which can lead to inconsistencies.
git-svn-id: file:///srv/svn/repos/haiku/haiku/trunk@26779 a95241bf-73f2-0310-859d-f6bbb57e9c96
user_{memcpy,memset}(), since that can cause a page fault, which needs
pages and might try to steal some from our cache.
git-svn-id: file:///srv/svn/repos/haiku/haiku/trunk@26700 a95241bf-73f2-0310-859d-f6bbb57e9c96
partially (e.g. due to hitting the end of file). The respective
classes have grown new methods and attributes to deal with that. The
"finished" callbacks have got additional parameters to indicate
whether the transfer was only partial and how much has been
transferred. Other callbacks and functions have a size_t* in/out
parameter instead of a simple size_t, now.
* vfs_{read,write}_pages() do now use the I/O request framework instead
of the underlying FS's {read,write}_pages() hooks (those should be
unused now). Furthermore they've got an additional "flags" parameter,
which is passed to IORequest::Init(), i.e. it allows to specify that
the given vecs refer to physical addresses.
* The file cache's read_into_cache() reads directly into physical
pages, now.
* Fixed bug in DoIO::IO(): The offset was not adjusted, so that all
pages were incorrectly transferred from/to the same location.
* Fixed broken subrequest scheduling loop head in
do_iterative_fd_io_iterate().
* Adjusted the test driver and implemented its io() hook. Using this
driver I/O requests are passed all the way from the VFS/VM to the
driver and through the I/O scheduler. It even seems to work. :-)
* Added missing const to the iovec* parameter of the IORequest::Init()
methods.
* Disabled some debug output by default. Added new optional debug
output.
git-svn-id: file:///srv/svn/repos/haiku/haiku/trunk@26692 a95241bf-73f2-0310-859d-f6bbb57e9c96
fs_vnode_ops::write_pages() to be called with fsReenter = true. Since
this is no longer the case, the argument has become superfluous. For
read_pages() it always was. Removed the argument from the functions
and all functions that propagated it.
* Some whitespace at the end of lines was removed.
git-svn-id: file:///srv/svn/repos/haiku/haiku/trunk@26579 a95241bf-73f2-0310-859d-f6bbb57e9c96
introduces the following relevant changes:
* VMCache:
- Renamed vm_cache to VMCache, merged it with vm_store and made it a
C++ class with virtual methods (replacing the store operations).
Turned the different store implementations into subclasses.
- Introduced MergeStore() callback, changed semantics of Commit().
- Changed locking and referencing semantics. A reference can only be
acquired/released with the cache locked. An unreferenced cache is
deleted and a mergeable cache merged when it is unlocked. This
removes the "busy" state of a cache and simplifies the page fault
code.
* Added VMAnonymousCache, which will implement swap support (work by
Zhao Shuai). It is not integrated and used yet, though.
* Enabled the mutex/recursive lock holder asserts.
* Fixed DoublyLinkedList::Swap().
* Generalized the low memory handler to a low resource handler. And made
semaphores and reserved memory handled resources. Made
vm_try_resource_memory() optionally wait (with timeout), and used that
feature to reserve memory for areas.
...
git-svn-id: file:///srv/svn/repos/haiku/haiku/trunk@26572 a95241bf-73f2-0310-859d-f6bbb57e9c96
* FileMap::_InvalidateAfter():
- Off-by-one error. The last extend (which normally should be kept)
was thrown away, too, but still accessed afterwards. Worst case
could be a write to free()d memory.
- Drop the last extend when it would be truncated to zero size.
* FileMap::Translate():
- Incorrect handling of B_BUFFER_OVERFLOW case in the
vfs_get_file_map() loop. After the loop the function would return
incorrectly, making the caller think all vectors in the provided
array had been initialized correctly. This could cause a file system
implementation using the file map to read from or write to random
disk locations, in the latter case possibly corrupting the file
system.
- Some readability improvements in the final loop. Removed incorrect
check.
git-svn-id: file:///srv/svn/repos/haiku/haiku/trunk@26523 a95241bf-73f2-0310-859d-f6bbb57e9c96
using it.
* IOW cache_prefetch_vnode() should work again now.
git-svn-id: file:///srv/svn/repos/haiku/haiku/trunk@26468 a95241bf-73f2-0310-859d-f6bbb57e9c96
systems like BFS only hold a read lock when reading the map, accessing the
file map wasn't safe.
* Made FileMap::Free() private, and renamed it to _Free().
git-svn-id: file:///srv/svn/repos/haiku/haiku/trunk@26295 a95241bf-73f2-0310-859d-f6bbb57e9c96
* adjacent vecs are now joined.
* partial invalidation no longer frees all cached extents.
* the array can now be larger than the needed number of entries, allowing
for a saner array allocation policy.
* it does no longer read the whole file map when the first translation is
requested, but only as much as required (it will still ask the file system
for the maximum file size, but it won't traverse further as long as the
initial request is fulfilled).
* This should help a lot with the ext2 file system that doesn't support real
file extents (but keeps a list of blocks).
git-svn-id: file:///srv/svn/repos/haiku/haiku/trunk@26260 a95241bf-73f2-0310-859d-f6bbb57e9c96
* Moved the file_extent array into its own structure (unnamed union problem
with GCC4), and added max_count field (currently unused).
* Minor cleanup.
git-svn-id: file:///srv/svn/repos/haiku/haiku/trunk@26259 a95241bf-73f2-0310-859d-f6bbb57e9c96
* Fixed most of the warnings resulting from that by removing actually not used
variables or moving declaration into the #IF. Left unused functions there
though, as I wouldn't know if they are supposed to be used again.
* Fixed two statements with no effect (clamping to MAX_ANCILLARY_DATA_LEN in
socket.cpp and unsetting fCounterSem in MessagingService.cpp).
* Some style cleanups.
git-svn-id: file:///srv/svn/repos/haiku/haiku/trunk@26253 a95241bf-73f2-0310-859d-f6bbb57e9c96
per cache.
* Changed the strategy vm_cache_acquire_page_cache_ref() uses to ensure
that the cache isn't deleted while trying to get a reference. Instead
of the global cache pages hash table lock, it holds the global cache
list lock now. We acquire + release this lock in delete_cache() after
removing all pages and just before deleting the object.
* Some small optimizations using the property that the cache's pages are
ordered, now (vm_cache_resize(), vm_page_write_modified_page_range(),
vm_page_schedule_write_page_range()).
* Replaced some code counting a cache's pages by simply using
vm_cache::page_count.
git-svn-id: file:///srv/svn/repos/haiku/haiku/trunk@26160 a95241bf-73f2-0310-859d-f6bbb57e9c96
