* The threads beside the main thread are killed earlier now (in the new
team_shutdown_team()), before removing the team from the team hash and from
its process group. This fixes#5296.
* Use a condition variable instead of a semaphore to wait for the non-main
threads to die. We notify the condition right after a thread has left the
team. The semaphore was released by the undertaker.
git-svn-id: file:///srv/svn/repos/haiku/haiku/trunk@35196 a95241bf-73f2-0310-859d-f6bbb57e9c96
- enqueue_in_run_queue() no longer returns whether rescheduling is supposed
to happen. Instead is sets cpu_ent::invoke_scheduler on the current CPU.
- reschedule() does now handle cpu_ent::invoke_scheduler_if_idle(). No need
to let all callers do that.
* thread_unblock[_locked]() no longer return whether rescheduling is supposed
to happen.
* Got rid of the B_INVOKE_SCHEDULER handling. The interrupt hooks really
can't know, when it makes sense to reschedule or not.
* Introduced scheduler_reschedule_if_necessary[_locked]() functions for
checking+invoking the scheduler.
* Some semaphore functions (e.g. delete_sem()) invoke the scheduler now, if
they wake up anything with greater priority.
I've also tried to add scheduler invocations in the condition variable and
mutex/rw_lock code, but that actually has a negative impact on performance,
probably because it causes too much ping-ponging between threads when
multiple locking primitives are involved.
git-svn-id: file:///srv/svn/repos/haiku/haiku/trunk@34657 a95241bf-73f2-0310-859d-f6bbb57e9c96
* simple_smp scheduler: Rewrote the interesting part of
enqueue_in_run_queue(). It always selects a target CPU for the inserted
thread, now. If no CPU is idle, the CPU running the thread with the lowest
priority is chosen. If the thread running on the target CPU has a lower
priority than the inserted one, it will be asked to reschedule. If that's
the current CPU, we'll return the correct value (wasn't done before at
all).
These changes help reducing latencies. On my machine in an idle system
playing music DebugAnalyzer shows maximum latencies of about 1 us. I still
find that a bit much, but it's several orders of magnitude better than
before. The -j8 Haiku image build time dropped about 10%.
git-svn-id: file:///srv/svn/repos/haiku/haiku/trunk@34635 a95241bf-73f2-0310-859d-f6bbb57e9c96
implemented by means of an additional member in cpu_ent.
* Removed thread::keep_scheduled and the related functions. The feature
wasn't used yet and wouldn't have worked as implemented anyway.
* Resurrected an older, SMP aware version of our simple scheduler and made it
the default instead of the affine scheduler. The latter is in no state to
be used yet. It causes enormous latencies (I've seen up to 0.1s) even when
six or seven CPUs were idle at the same time, totally killing parallelism.
That's also the reason why a -j8 build was slower than a -j2. This is no
longer the case. On my machine the -j2 build takes about 10% less time now
and the -j8 build saves another 20%. The latter is not particularly
impressive (compared with Linux), but that seems to be due to lock
contention.
git-svn-id: file:///srv/svn/repos/haiku/haiku/trunk@34615 a95241bf-73f2-0310-859d-f6bbb57e9c96
* "Classified" VMAddressSpace, i.e. turned the vm_address_space_*() functions
into methods, made all attributes (but "areas") private, and added
accessors.
* Also turned the vm.cpp functions vm_area_lookup() and
remove_area_from_address_space() into VMAddressSpace methods. The rest of
the area management functionality will follow soon.
git-svn-id: file:///srv/svn/repos/haiku/haiku/trunk@34447 a95241bf-73f2-0310-859d-f6bbb57e9c96
accesses the scheduler data inside the thread structure, freeing it first lead
to a crash when a thread couldn't be created.
git-svn-id: file:///srv/svn/repos/haiku/haiku/trunk@33940 a95241bf-73f2-0310-859d-f6bbb57e9c96
iterated over all known thread *IDs* with interrupts disabled.
Now it iterates over the team's thread list (going from back to front, since
new threads are added at the front of the singly linked queue).
* Alexandre restarted Tracker quite a lot, and let the shell script run to
reproduce a certain bug - and then wondered why ProcessController would
take several seconds to open its windows until it passed through more than
8 million IDs... :-)
git-svn-id: file:///srv/svn/repos/haiku/haiku/trunk@33737 a95241bf-73f2-0310-859d-f6bbb57e9c96
* scheduler_enqueue_in_runqueue() now allows the scheduler to return a hint as to whether a reschedule is desirable or not. This is used in a few other places in order to relegate scheduling decisions entirely to the scheduler rather than the priority hacks previously used. There are probably other places in the kernel that could now make use of that information to more intelligently call reschedule() though.
* Switch over the default scheduler to scheduler_affine().
git-svn-id: file:///srv/svn/repos/haiku/haiku/trunk@32554 a95241bf-73f2-0310-859d-f6bbb57e9c96
installed team debugger and adjusted the code accordingly. It's not needed yet,
but I intend to add support for software breakpoints and those require a bit of
uninitialization that needs to be synchronized with debugger changes and can't
be done with interrupts disabled.
git-svn-id: file:///srv/svn/repos/haiku/haiku/trunk@31194 a95241bf-73f2-0310-859d-f6bbb57e9c96
methods that used an "event mask" field. There was no need to introduce
a "flags" field for the same purpose.
* Renamed protected DefaultNotificationService methods (removed "_" prefix).
* Adjusted the code providing a notification service accordingly.
* Changed the event message several notification services generated by renaming
the "opcode" field to "event".
* Implemented the TEAM_ADDED event and also added a TEAM_EXEC event.
* Added notifications for threads and images.
* Added visitor-like iteration functions for teams, threads, and images.
git-svn-id: file:///srv/svn/repos/haiku/haiku/trunk@30126 a95241bf-73f2-0310-859d-f6bbb57e9c96
Don't really resume the undertaker thread after creating it, since at that
point the scheduler has not been started yet. Didn't do any harm, since it was
only the second thread anyway and the idle thread was rescheduled almost
immediately.
git-svn-id: file:///srv/svn/repos/haiku/haiku/trunk@29908 a95241bf-73f2-0310-859d-f6bbb57e9c96
makes them more convenient to use, since the caller doesn't need to know the
target threads' user thread structure.
* Adjusted the pthread rwlock implementation accordingly.
git-svn-id: file:///srv/svn/repos/haiku/haiku/trunk@29578 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
as public defines. They are now called THREAD_{MIN|MAX}_SET_PRIORITY to
better reflect what they are for. Minimum priority is now 1, ie. you no
longer can set another thread to the idle priority. This fixes part of
ticket #2959.
* set_thread_priority() will no longer allow to change the priority of the
idle thread to something else. This fixes the rest of ticket #2959.
* Automatic whitespace cleanup in OS.h.
git-svn-id: file:///srv/svn/repos/haiku/haiku/trunk@28521 a95241bf-73f2-0310-859d-f6bbb57e9c96
scheduler_set_thread_priority(). Setting the thread priority was the
only situation in which it was used.
* Renamed scheduler.cpp to scheduler_simple.cpp.
* The scheduler functions are no longer called directly. Instead there's
an operation vector now, which is initialized at kernel init time.
This allows for picking the most suitable scheduler for the machine
(e.g. a non-SMP scheduler on a non-SMP machine).
git-svn-id: file:///srv/svn/repos/haiku/haiku/trunk@28262 a95241bf-73f2-0310-859d-f6bbb57e9c96
the free queue.
* "thread" also prints the thread's I/O priority.
git-svn-id: file:///srv/svn/repos/haiku/haiku/trunk@28218 a95241bf-73f2-0310-859d-f6bbb57e9c96
This fixes a deadlock in SoundPlay that made it freeze when trying to play the 2nd sound.
git-svn-id: file:///srv/svn/repos/haiku/haiku/trunk@28179 a95241bf-73f2-0310-859d-f6bbb57e9c96
arch_vm_aspace_swap().
* The x86 implementation does now maintain a bit mask per
vm_translation_map_arch_info indicating on which CPUs the address
space is active. This allows flush_tmap() to avoid ICI for user
address spaces when the team isn't currently running on any other CPU.
In this context ICI is relatively expensive, particularly since we map
most pages via vm_map_page() and therefore invoke flush_tmap() pretty
much for every single page.
This optimization speeds up a "hello world" compilation about 20% on
my machine (KDEBUG turned off, freshly booted), but interestingly it
has virtually no effect on the "-j2" haiku build time.
git-svn-id: file:///srv/svn/repos/haiku/haiku/trunk@27912 a95241bf-73f2-0310-859d-f6bbb57e9c96
timed out thread has a higher priority than the currently running one.
Maybe we should even restrict this behavior to realtime threads.
git-svn-id: file:///srv/svn/repos/haiku/haiku/trunk@27897 a95241bf-73f2-0310-859d-f6bbb57e9c96
interrupt handler and will be executed right before returning from the
interrupt.
git-svn-id: file:///srv/svn/repos/haiku/haiku/trunk@27648 a95241bf-73f2-0310-859d-f6bbb57e9c96
still pretty much work in progress.
* Introduced init_thread_debug_info() which is used instead of
clear_thread_debug_info() when the thread is created. The latter
requires former initialization.
* user_debug_thread_deleted() is now already invoked in thread_exit(),
not in the undertaker.
git-svn-id: file:///srv/svn/repos/haiku/haiku/trunk@27531 a95241bf-73f2-0310-859d-f6bbb57e9c96
before being NULL checked on line 1234. I moved the null check to be after name
is initialized and removed it from line 1234.
git-svn-id: file:///srv/svn/repos/haiku/haiku/trunk@27428 a95241bf-73f2-0310-859d-f6bbb57e9c96
state_to_text().
* state_to_text() can now deal with a NULL thread argument.
git-svn-id: file:///srv/svn/repos/haiku/haiku/trunk@27163 a95241bf-73f2-0310-859d-f6bbb57e9c96
short form (like "threads").
* "thread" allows to specify more than one thread now.
git-svn-id: file:///srv/svn/repos/haiku/haiku/trunk@27152 a95241bf-73f2-0310-859d-f6bbb57e9c96
create_area_etc().
* When the new flag CREATE_AREA_DONT_WAIT is specified, the functions
don't wait for memory or pages to become available. They fail
immediately instead.
git-svn-id: file:///srv/svn/repos/haiku/haiku/trunk@27117 a95241bf-73f2-0310-859d-f6bbb57e9c96
include the guard pages. Adjusted the kernel and boot loader code
accordingly -- the guard pages size is added/not removed respectively.
The stack size passed to _kern_spawn_thread() is now the actually usable
size, and it is no longer possible to specify a size smaller than or
equal to the guard pages size.
* vm_create_anonymous_area(): Precommit two pages maximum -- a stack with
only one page usable size obviously doesn't need two pages.
git-svn-id: file:///srv/svn/repos/haiku/haiku/trunk@26819 a95241bf-73f2-0310-859d-f6bbb57e9c96
gTeamSpinlock.
* Renamed the static global variables in smp.c to match our style guide.
* Minor other cleanup.
* Removed superfluous white space.
git-svn-id: file:///srv/svn/repos/haiku/haiku/trunk@26730 a95241bf-73f2-0310-859d-f6bbb57e9c96
This allows drivers to lock the memory outside of the original team context.
* create_area_etc() got a struct team as first argument, but that should have
been a team_id.
* Removed delete_area_etc() - there is already vm_delete_area() doing the same
thing.
* Renamed vm_get_address_space_by_id() to vm_get_address_space(), as there is
no other method of getting an address space.
* Removed erroneous white space.
git-svn-id: file:///srv/svn/repos/haiku/haiku/trunk@26455 a95241bf-73f2-0310-859d-f6bbb57e9c96
* Missing backslash in UPDATE_THREAD_KERNEL_TIME() macro. Caused the
thread's kernel time never to be updated when the thread left the
kernel. The time would be added to the user time instead.
* Changed semantics of thread_at_kernel_exit(). It leaves interrupts
disabled when returning, now. All code using it would disable
interrupts afterwards anyway, and inbetween the thread could be
interrupted causing two more time updates (or actually not causing
them erroneously).
* The code in x86's interrupt handler entry (int_bottom) also checks
thread::in_kernel now, to determine whether the thread was already
considered to be in userland. This is necessary since the sysexit
instruction doesn't reenable interrupts, so that we have to do that
with the preceeding instruction. Hence an interrupt can occur
inbetween, which would incorrectly add the interrupt's kernel time
to the thread's user time.
* For userland threads team_create_thread_start() didn't invoke
thread_at_kernel_exit() or any equivalent. We do that in
arch_thread_enter_userspace() now.
git-svn-id: file:///srv/svn/repos/haiku/haiku/trunk@26240 a95241bf-73f2-0310-859d-f6bbb57e9c96
already a dead thread available. Fixes a race condition.
git-svn-id: file:///srv/svn/repos/haiku/haiku/trunk@25754 a95241bf-73f2-0310-859d-f6bbb57e9c96
B_DEBUG_SPINLOCK_CONTENTION is defined to 1. It typedefs spinlock to a
structure (thus breaking BeOS binary compatibility), containing a
counter which is incremented whenever a thread has to wait for the
spinlock.
* Added macros for spinlock initialization and access and changed
code using spinlocks accordingly. This breaks compilation for BeOS --
the macros should be defined in the respective compatibility wrappers.
* Added generic syscall to get the spinlock counters for the thread and
the team spinlocks.
git-svn-id: file:///srv/svn/repos/haiku/haiku/trunk@25752 a95241bf-73f2-0310-859d-f6bbb57e9c96
remains. This replaces the previous mechanism of switching the thread to
a dedicated death stack. We might consider moving more cleanup work to
the undertaker, but that seems a little more involved.
git-svn-id: file:///srv/svn/repos/haiku/haiku/trunk@25704 a95241bf-73f2-0310-859d-f6bbb57e9c96
address space that is fully locked and marked B_KERNEL_AREA. It can
thus be accessed by the kernel without additional checks.
* For each userland thread we do create a user_thread structure in that
area. The structure is accessible from userland via TLS, using the
private get_user_thread() function.
* Introduced private userland functions [un]defer_signals(). They can be
used to cheaply disable/re-enable signal delivery. They use the
user_thread::defer_signals/pending_signals fields which are
checked/updated by the kernel.
git-svn-id: file:///srv/svn/repos/haiku/haiku/trunk@25451 a95241bf-73f2-0310-859d-f6bbb57e9c96
Ingo Weinhold