* Replace count_low/count_high with bigtime_t fields plus an int32.
sizeof(spinlock) is now 32 bytes with the debug option enabled.
* Adjust and clean up all spinlock code to use the new fields.
* Fold DEBUG_SPINLOCK_LATENCIES into the new code. Remove the bootloader
option and other flags for it (these were not compiled in by default.)
The new code should be much easier to understand and also more powerful.
However, the information transmitted to userland isn't as useful now;
the KDL command output will have the interesting information.
(Things could be reworked to transmit more interesting information to
userland again if desired, but as this code clearly hadn't been compiled
for many years, as it referred to global spinlocks that have been gone
for a very long time.)
Change-Id: I2cb34078bfdc7604f288a297b6cd1aa7ff9cc512
Reviewed-on: https://review.haiku-os.org/c/haiku/+/6943
Reviewed-by: waddlesplash <waddlesplash@gmail.com>
Tested-by: Commit checker robot <no-reply+buildbot@haiku-os.org>
team.cpp address_space change is needed in
arch_thread_init_kthread_stack to set initial thread page
translation map. It also allows to simplify some debugger code.
DEBUG_PAGE_ACCESS check is currently incorrectly implemented in
RISCV64VMTranslationMap and disabled to avoid panic.
gHtifRegs = 0 change is needed to avoid using HTIF when it is not
available. gHtifRegs != NULL check is used to detect that HTIF is
present. 0x40008000 is TinyEMU HTIF address. HTIF is emulator specific
device that provide serial IO and shutdown capability (and maybe
something else depending on virtual machine).
Change-Id: Ic4d567b28c49799ae0f55223dd983a752823bab4
Reviewed-on: https://review.haiku-os.org/c/haiku/+/4328
Reviewed-by: Alex von Gluck IV <kallisti5@unixzen.com>
Tested-by: Commit checker robot <no-reply+buildbot@haiku-os.org>
Each TCPEndpoint has two BufferQueue members, one for the send queue
and one for the receive queue.
If DEBUG_BUFFER_QUEUE is enabled, then most methods of BufferQueue
call BufferQueue::Verify(), sometimes twice. This member function
performs some sanity checking which requires iterating through every
net_buffer in the queue.
Disabling this in a debug build improved throughput by a factor of 5x
over the loopback interface on my laptop. Using iperf the measured
throughput went from 900Mbps to around 4.8Gbps.
This patch turns this sanity checking off for release builds.
* Rename DEBUG_BUFFER_QUEUE to DEBUG_TCP_BUFFER_QUEUE
* Change the default in BufferQueue.h to disabled
* Set DEBUG_TCP_BUFFER_QUEUE to KDEBUG_LEVEL_2 in
kernel_debug_config.h
Change-Id: I262dac5d7e2889d2942bbdcf6b667cc0cbafa4c8
Reviewed-on: https://review.haiku-os.org/c/haiku/+/2780
Reviewed-by: waddlesplash <waddlesplash@gmail.com>
It does have a performance impact (as it serializes all VMCache
creations), and it is not that useful, so move it to KDEBUG_LEVEL_2
so that the beta releases (on KDEBUG_LEVEL_1) are not affected
by it.
(The nightlies are on KDEBUG_LEVEL_2, so this will not affect
them.)
Change-Id: Iab35d58f9d210651e3fd9027b75ea27a7a6928e5
Reviewed-on: https://review.haiku-os.org/c/haiku/+/2612
Reviewed-by: waddlesplash <waddlesplash@gmail.com>
* This enables a mechanism to profile almost the complete boot process
(starting with main2()), if SYSTEM_PROFILER is defined to 1.
* You can access the profiling data using "profile -r".
This allows to use the debug features of the guarded heap also on
allocations made through the object cache API. This is obivously
horrible for performance and uses up huge amounts of memory, so the
initial and grow sizes are adjusted accordingly.
Note that this is a rather simple hack, using the object_cache pointer
to transport the allocation size. The alignment is neglected completely.
This is a very simple heap implementation that allocates memory so that
the end of each allocation always coincides with a page end and is
followed by a guard page which is marked non-present. Out of bounds
access (both read and write) therefore cause a crash (unhandled page
fault).
Note that this allocator is neither speed nor space efficient, indeed it
wastes huge amounts of pages and address space so it is quite easy to
hit limits. It is intended as a pure debug feature.
* Add optional stack trace capturing for slab memory manager tracing.
* Add allocation tracking for the slab allocator (enabled via
SLAB_ALLOCATION_TRACKING). The allocation tracking requires tracing
with stack traces to be enabled for object caches and/or the memory
manager.
- Add class AllocationTrackingInfo that associates an allocation with
its respective tracing entry. The structure is added to the end of
an allocation done by the memory manager. For the object caches
there's a separate array for each slab.
- Add code range markers to the slab code, so that the first caller
into the slab code can be retrieved from the stack traces.
- Add KDL command "allocations_per_caller" that lists all allocations
summarized by caller.
* Move debug definitions from slab_private.h to slab_debug.h.
git-svn-id: file:///srv/svn/repos/haiku/haiku/trunk@43072 a95241bf-73f2-0310-859d-f6bbb57e9c96
after each chunk allocation/deallocation.
* The commands that dump chunks also verify, whether chunks that look free
are in the free list.
git-svn-id: file:///srv/svn/repos/haiku/haiku/trunk@37360 a95241bf-73f2-0310-859d-f6bbb57e9c96
* The previous kernel heap has only an advantage for debugging, anyway, but the
actual reason for this change is bug #6232, and Haiku should crash for most
people since a week now.
git-svn-id: file:///srv/svn/repos/haiku/haiku/trunk@37327 a95241bf-73f2-0310-859d-f6bbb57e9c96
the default setting for the "debug syslog" feature (can still be overridden
in the boot loader). Per default enabled for kdebug level >= 1.
git-svn-id: file:///srv/svn/repos/haiku/haiku/trunk@35884 a95241bf-73f2-0310-859d-f6bbb57e9c96
* Added support to do larger raw allocations (up to one large chunk (128 pages))
in the slab areas. For an even larger allocation an area is created (haven't
seen that happen yet, though).
* Added kernel tracing (SLAB_MEMORY_MANAGER_TRACING).
* _FreeArea(): Copy and paste bug: The meta chunks of the to be freed area
would be added to the free lists instead of being removed from them. This
would corrupt the lists and also lead to all kinds of misuse of meta chunks.
object caches:
* Implemented CACHE_ALIGN_ON_SIZE. It is no longer set for all small object
caches, but the block allocator sets it on all power of two size caches.
* object_cache_reserve_internal(): Detect recursion and don't wait in such a
case. The function could deadlock itself, since
HashedObjectCache::CreateSlab() does allocate memory, thus potentially
reentering.
* object_cache_low_memory():
- I missed some returns when reworking that one in r35254, so the function
might stop early and also leave the cache in maintenance mode, which would
cause it to be ignored by object cache resizer and low memory handler from
that point on.
- Since ReturnSlab() potentially unlocks, the conditions weren't quite correct
and too many slabs could be freed.
- Simplified things a bit.
* object_cache_alloc(): Since object_cache_reserve_internal() does potentially
unlock the cache, the situation might have changed and their might not be an
empty slab available, but a partial one. The function would crash.
* Renamed the object cache tracing variable to SLAB_OBJECT_CACHE_TRACING.
* Renamed debugger command "cache_info" to "slab_cache" to avoid confusion with
the VMCache commands.
* ObjectCache::usage was not maintained anymore since I introduced the
MemoryManager. object_cache_get_usage() would thus always return 0 and the
block cache would not be considered cached memory. This was only of
informational relevance, though.
slab allocator misc.:
* Disable the object depots of block allocator caches for object sizes > 2 KB.
Allocations of those sizes aren't so common that the object depots yield any
benefit.
* The slab allocator is now fully self-sufficient. It allocates its bootstrap
memory from the MemoryManager, and the hash tables for HashedObjectCaches use
the block allocator instead of the heap, now.
* Added option to use the slab allocator for malloc() and friends
(USE_SLAB_ALLOCATOR_FOR_MALLOC). Currently disabled. Works in principle and
has virtually no lock contention. Handling for low memory situations is yet
missing, though.
* Improved the output of some debugger commands.
git-svn-id: file:///srv/svn/repos/haiku/haiku/trunk@35283 a95241bf-73f2-0310-859d-f6bbb57e9c96
access to a vm_page. It is basically an atomically accessed thread ID field
in the vm_page structure, which is explicitly set by macros marking the
critical sections. As a first positive effect I had to review quite a bit of
code and found several issues.
* Added several TODOs and comments. Some harmless ones, but also a few
troublesome ones in vm.cpp regarding page unmapping.
* file_cache: PrecacheIO::Prepare()/read_into_cache: Removed superfluous
vm_page_allocate_page() return value checks. It cannot fail anymore.
* Removed the heavily contended "pages" lock. We use different policies now:
- sModifiedTemporaryPages is accessed atomically.
- sPageDeficitLock and sFreePageCondition are protected by a new mutex.
- The page queues have individual locks (mutexes).
- Renamed set_page_state_nolock() to set_page_state(). Unless the caller says
otherwise, it does now lock the affected pages queues itself. Also changed
the return value to void -- we panic() anyway.
* set_page_state(): Add free/clear pages to the beginning of their respective
queues as this is more cache-friendly.
* Pages with the states PAGE_STATE_WIRED or PAGE_STATE_UNUSED are no longer
in any queue. They were in the "active" queue, but there's no good reason
to have them there. In case we decide to let the page daemon work the queues
(like FreeBSD) they would just be in the way.
* Pulled the common part of vm_page_allocate_page_run[_no_base]() into a helper
function. Also fixed a bug I introduced previously: The functions must not
vm_page_unreserve_pages() on success, since they remove the pages from the
free/clear queue without decrementing sUnreservedFreePages.
* vm_page_set_state(): Changed return type to void. The function cannot really
fail and no-one was checking it anyway.
* vm_page_free(), vm_page_set_state(): Added assertion: The page must not be
free/clear before. This is implied by the policy that no-one is allowed to
access free/clear pages without holding the respective queue's lock, which is
not the case at this point. This found the bug fixed in r34912.
* vm_page_requeue(): Added general assertions. panic() when requeuing of
free/clear pages is requested. Same reason as above.
* vm_clone_area(), B_FULL_LOCK case: Don't map busy pages. The implementation is
still not correct, though.
My usual -j8 Haiku build test runs another 10% faster, now. The total kernel
time drops about 18%. As hoped the new locks have only a fraction of the old
"pages" lock contention. Other locks lead the "most wanted list" now.
git-svn-id: file:///srv/svn/repos/haiku/haiku/trunk@34933 a95241bf-73f2-0310-859d-f6bbb57e9c96
* Added VMCache::MovePage() and MoveAllPages() to move pages between caches.
* VMAnonymousCache:
- _MergeSwapPages(): Avoid doing anything, if neither cache has swapped out
pages.
- _MergeSwapPages() does now also remove source cache pages that are
shadowed by consumer swap pages. This allows us to call _MergeSwapPages()
before _MergePagesSmallerSource(), save the swap page shadowing check
there and get rid of the vm_page::merge_swap flag. This is an
optimization based on the assumption that usually none or only few pages
are swapped out, so we save a lot of checks.
- Implemented _MergePagesSmallerConsumer() as an alternative to
_MergePagesSmallerSource(). The former is used when the source cache has
more pages than the consumer cache. It iterates over the consumer cache's
pages, moves them to the source and finally moves all pages back to the
consumer. The final move is relatively cheap (though unfortunately we
still have to update all pages' vm_page::cache field), so that overall we
save iterations of the main loop with the more expensive checks.
The optimizations particularly improve the common fork()+exec*() situations.
fork() uses CoW, which is implemented by putting two new empty caches between
the to be copied area and its cache. exec*() destroys one copy of the area,
its cache and thus causes merging of the other new cache with the old cache.
Since this usually happens in a very short time, the old cache does still
contain many pages and the new cache only few. Previously the many pages were
all checked and moved individually. Now we do that for the few pages instead.
A very extreme example of this situation is the Haiku image build. jam has a
huge heap (> 200 MB) and it fork()s+exec*()s for every action to be executed.
Since during the cache merging the cache is locked, any write access to a
heap page causes jam to block until the cache merging is done. Formerly that
took so long that it killed a lot of parallelism in multi-job builds. That
could be observed particularly well when lots of small actions where executed
(like the Link, XRes, Mimeset, SetType, SetVersion combos when building
executables/libraries/add-ons). Those look dramatically better now.
The overall speed improvement for a -j8 image build on my machine is only
about 15%, though.
git-svn-id: file:///srv/svn/repos/haiku/haiku/trunk@34784 a95241bf-73f2-0310-859d-f6bbb57e9c96
* When DEBUG_SPINLOCK_LATENCIES is 1, the system will panic if any spinlock is
held longer than DEBUG_LATENCY micro seconds (currently 200). If your system
doesn't boot anymore, a new safemode setting can disable the panic.
* Besides some problems during boot when the MTRRs are set up, 200 usecs work
fine here if all debug output is turned off (the output stuff is definitely
problematic, though I don't have a good idea on how to improve upon it a lot).
* Renamed the formerly BeOS compatible safemode settings to look better; there
is no need to be compatible there.
git-svn-id: file:///srv/svn/repos/haiku/haiku/trunk@33953 a95241bf-73f2-0310-859d-f6bbb57e9c96
have a simple dedicated heap for the kernel debugger with stacked allocation
pools (deleting a pool frees all memory allocated in it). The heap should
eventually be used for all commands that need temporary storage too large for
the stack instead of each using its own static buffer.
git-svn-id: file:///srv/svn/repos/haiku/haiku/trunk@30949 a95241bf-73f2-0310-859d-f6bbb57e9c96
enter/exit code. There's no real reason not to keep kernel breakpoints
enabled when in userland (unless there are breakpoints installed for the
team, of course).
* Enabled kernel breakpoints by default (check your kernel_debug_config.h,
if you have overridden it!), since they don't really add any overhead
anymore.
git-svn-id: file:///srv/svn/repos/haiku/haiku/trunk@30206 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
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