This simplifies signalfd code, by avoiding it to remain attached to the
sighand during its lifetime.
In this way, the signalfd remain attached to the sighand only during
poll(2) (and select and epoll) and read(2). This also allows to remove
all the custom "tsk == current" checks in kernel/signal.c, since
dequeue_signal() will only be called by "current".
I think this is also what Ben was suggesting time ago.
The external effect of this, is that a thread can extract only its own
private signals and the group ones. I think this is an acceptable
behaviour, in that those are the signals the thread would be able to
fetch w/out signalfd.
Signed-off-by: Davide Libenzi <davidel@xmailserver.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
When using rt_mutex, a NULL pointer dereference is occurred at
enqueue_task_rt. Here is a scenario;
1) there are two threads, the thread A is fair_sched_class and
thread B is rt_sched_class.
2) Thread A is boosted up to rt_sched_class, because the thread A
has a rt_mutex lock and the thread B is waiting the lock.
3) At this time, when thread A create a new thread C, the thread
C has a rt_sched_class.
4) When doing wake_up_new_task() for the thread C, the priority
of the thread C is out of the RT priority range, because the
normal priority of thread A is not the RT priority. It makes
data corruption by overflowing the rt_prio_array.
The new thread C should be fair_sched_class.
The new thread should be valid scheduler class before queuing.
This patch fixes to set the suitable scheduler class.
Signed-off-by: Hiroshi Shimamoto <h-shimamoto@ct.jp.nec.com>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
Signed-off-by: Peter Zijlstra <a.p.zijlstra@chello.nl>
add /proc/sys/kernel/sched_compat_yield to make sys_sched_yield()
more agressive, by moving the yielding task to the last position
in the rbtree.
with sched_compat_yield=0:
PID USER PR NI VIRT RES SHR S %CPU %MEM TIME+ COMMAND
2539 mingo 20 0 1576 252 204 R 50 0.0 0:02.03 loop_yield
2541 mingo 20 0 1576 244 196 R 50 0.0 0:02.05 loop
with sched_compat_yield=1:
PID USER PR NI VIRT RES SHR S %CPU %MEM TIME+ COMMAND
2584 mingo 20 0 1576 248 196 R 99 0.0 0:52.45 loop
2582 mingo 20 0 1576 256 204 R 0 0.0 0:00.00 loop_yield
Signed-off-by: Ingo Molnar <mingo@elte.hu>
Signed-off-by: Peter Zijlstra <a.p.zijlstra@chello.nl>
It turned out, that the user namespace is released during the do_exit() in
exit_task_namespaces(), but the struct user_struct is released only during the
put_task_struct(), i.e. MUCH later.
On debug kernels with poisoned slabs this will cause the oops in
uid_hash_remove() because the head of the chain, which resides inside the
struct user_namespace, will be already freed and poisoned.
Since the uid hash itself is required only when someone can search it, i.e.
when the namespace is alive, we can safely unhash all the user_struct-s from
it during the namespace exiting. The subsequent free_uid() will complete the
user_struct destruction.
For example simple program
#include <sched.h>
char stack[2 * 1024 * 1024];
int f(void *foo)
{
return 0;
}
int main(void)
{
clone(f, stack + 1 * 1024 * 1024, 0x10000000, 0);
return 0;
}
run on kernel with CONFIG_USER_NS turned on will oops the
kernel immediately.
This was spotted during OpenVZ kernel testing.
Signed-off-by: Pavel Emelyanov <xemul@openvz.org>
Signed-off-by: Alexey Dobriyan <adobriyan@openvz.org>
Acked-by: "Serge E. Hallyn" <serue@us.ibm.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Surprisingly, but (spotted by Alexey Dobriyan) the uid hash still uses
list_heads, thus occupying twice as much place as it could. Convert it to
hlist_heads.
Signed-off-by: Pavel Emelyanov <xemul@openvz.org>
Signed-off-by: Alexey Dobriyan <adobriyan@openvz.org>
Acked-by: Serge Hallyn <serue@us.ibm.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
struct utsname is copied from master one without any exclusion.
Here is sample output from one proggie doing
sethostname("aaaaaaaaaaaaaaaaaaaaaaaaaaaaaa");
sethostname("bbbbbbbbbbbbbbbbbbbbbbbbbbbbbb");
and another
clone(,, CLONE_NEWUTS, ...)
uname()
hostname = 'aaaaaaaaaaaaaaaaaaaaaaaaabbbbb'
hostname = 'bbbaaaaaaaaaaaaaaaaaaaaaaaaaaa'
hostname = 'aaaaaaaabbbbbbbbbbbbbbbbbbbbbb'
hostname = 'aaaaaaaaaaaaaaaaaaaaaaaaaabbbb'
hostname = 'aaaaaaaaaaaaaaaaaaaaaaaaaaaabb'
hostname = 'aaabbbbbbbbbbbbbbbbbbbbbbbbbbb'
hostname = 'bbbbbbbbbbbbbbbbaaaaaaaaaaaaaa'
Hostname is sometimes corrupted.
Yes, even _the_ simplest namespace activity had bug in it. :-(
Signed-off-by: Alexey Dobriyan <adobriyan@sw.ru>
Acked-by: Serge Hallyn <serue@us.ibm.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Taking a cpu offline removes the cpu from the online mask before the
CPU_DEAD notification is done. The clock events layer does the cleanup
of the dead CPU from the CPU_DEAD notifier chain. tick_do_timer_cpu is
used to avoid xtime lock contention by assigning the task of jiffies
xtime updates to one CPU. If a CPU is taken offline, then this
assignment becomes stale. This went unnoticed because most of the time
the offline CPU went dead before the online CPU reached __cpu_die(),
where the CPU_DEAD state is checked. In the case that the offline CPU did
not reach the DEAD state before we reach __cpu_die(), the code in there
goes to sleep for 100ms. Due to the stale time update assignment, the
system is stuck forever.
Take the assignment away when a cpu is not longer in the cpu_online_mask.
We do this in the last call to tick_nohz_stop_sched_tick() when the offline
CPU is on the way to the final play_dead() idle entry.
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
When a cpu goes offline it is removed from the broadcast masks. If the
mask becomes empty the code shuts down the broadcast device. This is
wrong, because the broadcast device needs to be ready for the online
cpu going idle (into a c-state, which stops the local apic timer).
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
The jinxed VAIO refuses to resume without hitting keys on the keyboard
when this is not enforced. It is unclear why the cpu ends up in a lower
C State without notifying the clock events layer, but enforcing the
oneshot broadcast here is safe.
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
Timekeeping resume adjusts xtime by adding the slept time in seconds and
resets the reference value of the clock source (clock->cycle_last).
clock->cycle last is used to calculate the delta between the last xtime
update and the readout of the clock source in __get_nsec_offset(). xtime
plus the offset is the current time. The resume code ignores the delta
which had already elapsed between the last xtime update and the actual
time of suspend. If the suspend time is short, then we can see time
going backwards on resume.
Suspend:
offs_s = clock->read() - clock->cycle_last;
now = xtime + offs_s;
timekeeping_suspend_time = read_rtc();
Resume:
sleep_time = read_rtc() - timekeeping_suspend_time;
xtime.tv_sec += sleep_time;
clock->cycle_last = clock->read();
offs_r = clock->read() - clock->cycle_last;
now = xtime + offs_r;
if sleep_time_seconds == 0 and offs_r < offs_s, then time goes
backwards.
Fix this by storing the offset from the last xtime update and add it to
xtime during resume, when we reset clock->cycle_last:
sleep_time = read_rtc() - timekeeping_suspend_time;
xtime.tv_sec += sleep_time;
xtime += offs_s; /* Fixup xtime offset at suspend time */
clock->cycle_last = clock->read();
offs_r = clock->read() - clock->cycle_last;
now = xtime + offs_r;
Thanks to Marcelo for tracking this down on the OLPC and providing the
necessary details to analyze the root cause.
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
Cc: John Stultz <johnstul@us.ibm.com>
Cc: Tosatti <marcelo@kvack.org>
Lockdep complains about the access of rtc in timekeeping_suspend
inside the interrupt disabled region of the write locked xtime lock.
Move the access outside.
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
Cc: John Stultz <johnstul@us.ibm.com>
Seems to me that this timer will only get started on platforms that say
they don't want it?
Signed-off-by: Tony Breeds <tony@bakeyournoodle.com>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Gabriel Paubert <paubert@iram.es>
Cc: Zachary Amsden <zach@vmware.com>
Acked-by: Thomas Gleixner <tglx@linutronix.de>
Cc: John Stultz <johnstul@us.ibm.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
The semantics of call_usermodehelper_pipe() used to be that it would fork
the helper, and wait for the kernel thread to be started. This was
implemented by setting sub_info.wait to 0 (implicitly), and doing a
wait_for_completion().
As part of the cleanup done in 0ab4dc9227,
call_usermodehelper_pipe() was changed to pass 1 as the value for wait to
call_usermodehelper_exec().
This is equivalent to setting sub_info.wait to 1, which is a change from
the previous behaviour. Using 1 instead of 0 causes
__call_usermodehelper() to start the kernel thread running
wait_for_helper(), rather than directly calling ____call_usermodehelper().
The end result is that the calling kernel code blocks until the user mode
helper finishes. As the helper is expecting input on stdin, and now no one
is writing anything, everything locks up (observed in do_coredump).
The fix is to change the 1 to UMH_WAIT_EXEC (aka 0), indicating that we
want to wait for the kernel thread to be started, but not for the helper to
finish.
Signed-off-by: Michael Ellerman <michael@ellerman.id.au>
Acked-by: Andi Kleen <ak@suse.de>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
The futex list traversal on the compat side appears to have
a bug.
It's loop termination condition compares:
while (compat_ptr(uentry) != &head->list)
But that can't be right because "uentry" has the special
"pi" indicator bit still potentially set at bit 0. This
is cleared by fetch_robust_entry() into the "entry"
return value.
What this seems to mean is that the list won't terminate
when list iteration gets back to the the head. And we'll
also process the list head like a normal entry, which could
cause all kinds of problems.
So we should check for equality with "entry". That pointer
is of the non-compat type so we have to do a little casting
to keep the compiler and sparse happy.
The same problem can in theory occur with the 'pending'
variable, although that has not been reported from users
so far.
Based on the original patch from David Miller.
Acked-by: Ingo Molnar <mingo@elte.hu>
Cc: Thomas Gleixner <tglx@linutronix.de>
Cc: David Miller <davem@davemloft.net>
Signed-off-by: Arnd Bergmann <arnd@arndb.de>
Cc: <stable@kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
When PTRACE_SYSCALL was used and then PTRACE_DETACH is used, the
TIF_SYSCALL_TRACE flag is left set on the formerly-traced task. This
means that when a new tracer comes along and does PTRACE_ATTACH, it's
possible he gets a syscall tracing stop even though he's never used
PTRACE_SYSCALL. This happens if the task was in the middle of a system
call when the second PTRACE_ATTACH was done. The symptom is an
unexpected SIGTRAP when the tracer thinks that only SIGSTOP should have
been provoked by his ptrace calls so far.
A few machines already fixed this in ptrace_disable (i386, ia64, m68k).
But all other machines do not, and still have this bug. On x86_64, this
constitutes a regression in IA32 compatibility support.
Since all machines now use TIF_SYSCALL_TRACE for this, I put the
clearing of TIF_SYSCALL_TRACE in the generic ptrace_detach code rather
than adding it to every other machine's ptrace_disable.
Signed-off-by: Roland McGrath <roland@redhat.com>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
fix ideal_runtime:
- do not scale it using niced_granularity()
it is against sum_exec_delta, so its wall-time, not fair-time.
- move the whole check into __check_preempt_curr_fair()
so that wakeup preemption can also benefit from the new logic.
this also results in code size reduction:
text data bss dec hex filename
13391 228 1204 14823 39e7 sched.o.before
13369 228 1204 14801 39d1 sched.o.after
Signed-off-by: Peter Zijlstra <a.p.zijlstra@chello.nl>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
Second preparatory patch for fix-ideal runtime:
Mark prev_sum_exec_runtime at the beginning of our run, the same spot
that adds our wait period to wait_runtime. This seems a more natural
location to do this, and it also reduces the code a bit:
text data bss dec hex filename
13397 228 1204 14829 39ed sched.o.before
13391 228 1204 14823 39e7 sched.o.after
Signed-off-by: Peter Zijlstra <a.p.zijlstra@chello.nl>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
Preparatory patch for fix-ideal-runtime:
simplify __check_preempt_curr_fair(): get rid of the integer return.
text data bss dec hex filename
13404 228 1204 14836 39f4 sched.o.before
13393 228 1204 14825 39e9 sched.o.after
functionality is unchanged.
Signed-off-by: Peter Zijlstra <a.p.zijlstra@chello.nl>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
the cfs_rq->wait_runtime debug/statistics counter was not maintained
properly - fix this.
this also removes some code:
text data bss dec hex filename
13420 228 1204 14852 3a04 sched.o.before
13404 228 1204 14836 39f4 sched.o.after
Signed-off-by: Ingo Molnar <mingo@elte.hu>
Signed-off-by: Peter Zijlstra <a.p.zijlstra@chello.nl>
fix niced_granularity(). This resulted in under-scheduling for
CPU-bound negative nice level tasks (and this in turn caused
higher than necessary latencies in nice-0 tasks).
Signed-off-by: Ingo Molnar <mingo@elte.hu>
First fix the check
if (*imbalance + SCHED_LOAD_SCALE_FUZZ < busiest_load_per_task)
with this
if (*imbalance < busiest_load_per_task)
As the current check is always false for nice 0 tasks (as
SCHED_LOAD_SCALE_FUZZ is same as busiest_load_per_task for nice 0
tasks).
With the above change, imbalance was getting reset to 0 in the corner
case condition, making the FUZZ logic fail. Fix it by not corrupting the
imbalance and change the imbalance, only when it finds that the HT/MC
optimization is needed.
Signed-off-by: Suresh Siddha <suresh.b.siddha@intel.com>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
* git://git.kernel.org/pub/scm/linux/kernel/git/mingo/linux-2.6-sched:
sched: clean up task_new_fair()
sched: small schedstat fix
sched: fix wait_start_fair condition in update_stats_wait_end()
sched: call update_curr() in task_tick_fair()
sched: make the scheduler converge to the ideal latency
sched: fix sleeper bonus limit
Spotted by taoyue <yue.tao@windriver.com> and Jeremy Katz <jeremy.katz@windriver.com>.
collect_signal: sigqueue_free:
list_del_init(&first->list);
if (!list_empty(&q->list)) {
// not taken
}
q->flags &= ~SIGQUEUE_PREALLOC;
__sigqueue_free(first); __sigqueue_free(q);
Now, __sigqueue_free() is called twice on the same "struct sigqueue" with the
obviously bad implications.
In particular, this double free breaks the array_cache->avail logic, so the
same sigqueue could be "allocated" twice, and the bug can manifest itself via
the "impossible" BUG_ON(!SIGQUEUE_PREALLOC) in sigqueue_free/send_sigqueue.
Hopefully this can explain these mysterious bug-reports, see
http://marc.info/?t=118766926500003http://marc.info/?t=118466273000005
Alexey Dobriyan reports this patch makes the difference for the testcase, but
nobody has an access to the application which opened the problems originally.
Also, this patch removes tasklist lock/unlock, ->siglock is enough.
Signed-off-by: Oleg Nesterov <oleg@tv-sign.ru>
Cc: taoyue <yue.tao@windriver.com>
Cc: Jeremy Katz <jeremy.katz@windriver.com>
Cc: Sukadev Bhattiprolu <sukadev@us.ibm.com>
Cc: Alexey Dobriyan <adobriyan@sw.ru>
Cc: Ingo Molnar <mingo@elte.hu>
Cc: Thomas Gleixner <tglx@linutronix.de>
Cc: Roland McGrath <roland@redhat.com>
Cc: <stable@kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Mariusz Kozlowski reported lockdep's warning:
> =================================
> [ INFO: inconsistent lock state ]
> 2.6.23-rc2-mm1 #7
> ---------------------------------
> inconsistent {in-hardirq-W} -> {hardirq-on-W} usage.
> ifconfig/5492 [HC0[0]:SC0[0]:HE1:SE1] takes:
> (&tp->lock){+...}, at: [<de8706e0>] rtl8139_interrupt+0x27/0x46b [8139too]
> {in-hardirq-W} state was registered at:
> [<c0138eeb>] __lock_acquire+0x949/0x11ac
> [<c01397e7>] lock_acquire+0x99/0xb2
> [<c0452ff3>] _spin_lock+0x35/0x42
> [<de8706e0>] rtl8139_interrupt+0x27/0x46b [8139too]
> [<c0147a5d>] handle_IRQ_event+0x28/0x59
> [<c01493ca>] handle_level_irq+0xad/0x10b
> [<c0105a13>] do_IRQ+0x93/0xd0
> [<c010441e>] common_interrupt+0x2e/0x34
...
> other info that might help us debug this:
> 1 lock held by ifconfig/5492:
> #0: (rtnl_mutex){--..}, at: [<c0451778>] mutex_lock+0x1c/0x1f
>
> stack backtrace:
...
> [<c0452ff3>] _spin_lock+0x35/0x42
> [<de8706e0>] rtl8139_interrupt+0x27/0x46b [8139too]
> [<c01480fd>] free_irq+0x11b/0x146
> [<de871d59>] rtl8139_close+0x8a/0x14a [8139too]
> [<c03bde63>] dev_close+0x57/0x74
...
This shows that a driver's irq handler was running both in hard interrupt
and process contexts with irqs enabled. The latter was done during
free_irq() call and was possible only with CONFIG_DEBUG_SHIRQ enabled.
This was fixed by another patch.
But similar problem is possible with request_irq(): any locks taken from
irq handler could be vulnerable - especially with soft interrupts. This
patch fixes it by disabling local interrupts during handler's run. (It
seems, disabling softirqs should be enough, but it needs more checking
on possible races or other special cases).
Reported-by: Mariusz Kozlowski <m.kozlowski@tuxland.pl>
Signed-off-by: Jarek Poplawski <jarkao2@o2.pl>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Dependencies of CONFIG_SUSPEND and CONFIG_HIBERNATION introduced by commit
296699de6b "Introduce CONFIG_SUSPEND for
suspend-to-Ram and standby" are incorrect, as they don't cover the facts that
(1) not all architectures support suspend and (2) SMP hibernation is only
possible on X86 and PPC64 (if CONFIG_PPC64_SWSUSP is set).
Signed-off-by: Rafael J. Wysocki <rjw@sisk.pl>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Spotted by Marcin Kowalczyk <qrczak@knm.org.pl>.
sys_setpgid(child) fails if the child was forked by sub-thread.
Fix the "is it our child" check. The previous commit
ee0acf90d3 was not complete.
(this patch asks for the new same_thread_group() helper, but mainline doesn't
have it yet).
Signed-off-by: Oleg Nesterov <oleg@tv-sign.ru>
Acked-by: Roland McGrath <roland@redhat.com>
Cc: <stable@kernel.org>
Tested-by: "Marcin 'Qrczak' Kowalczyk" <qrczak@knm.org.pl>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
taskstats.ac_exitcode is assigned to task_struct.exit_code in bacct_add_tsk()
through the following kernel function calls:
do_exit()
taskstats_exit()
fill_pid()
bacct_add_tsk()
The problem is that in do_exit(), task_struct.exit_code is set to 'code' only
after taskstats_exit() has been called. So we need to move the assignment
before taskstats_exit().
Signed-off-by: Jonathan Lim <jlim@sgi.com>
Cc: Balbir Singh <balbir@in.ibm.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
cleanup: we have the 'se' and 'curr' entity-pointers already,
no need to use p->se and current->se.
Signed-off-by: Ingo Molnar <mingo@elte.hu>
Signed-off-by: Peter Zijlstra <a.p.zijlstra@chello.nl>
Signed-off-by: Mike Galbraith <efault@gmx.de>
small schedstat fix: the cfs_rq->wait_runtime 'sum of all runtimes'
statistics counters missed newly forked tasks and thus had a constant
negative skew. Fix this.
Signed-off-by: Ingo Molnar <mingo@elte.hu>
Signed-off-by: Peter Zijlstra <a.p.zijlstra@chello.nl>
Signed-off-by: Mike Galbraith <efault@gmx.de>
Peter Zijlstra noticed the following bug in SCHED_FEAT_SKIP_INITIAL (which
is disabled by default at the moment): it relies on se.wait_start_fair
being 0 while update_stats_wait_end() did not recognize a 0 value,
so instead of 'skipping' the initial interval we gave the new child
a maximum boost of +runtime-limit ...
(No impact on the default kernel, but nice to fix for completeness.)
Signed-off-by: Ingo Molnar <mingo@elte.hu>
Signed-off-by: Peter Zijlstra <a.p.zijlstra@chello.nl>
Signed-off-by: Mike Galbraith <efault@gmx.de>
update the fair-clock before using it for the key value.
[ mingo@elte.hu: small cleanups. ]
Signed-off-by: Ting Yang <tingy@cs.umass.edu>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
Signed-off-by: Mike Galbraith <efault@gmx.de>
Signed-off-by: Peter Zijlstra <a.p.zijlstra@chello.nl>
de-HZ-ification of the granularity defaults unearthed a pre-existing
property of CFS: while it correctly converges to the granularity goal,
it does not prevent run-time fluctuations in the range of
[-gran ... 0 ... +gran].
With the increase of the granularity due to the removal of HZ
dependencies, this becomes visible in chew-max output (with 5 tasks
running):
out: 28 . 27. 32 | flu: 0 . 0 | ran: 9 . 13 | per: 37 . 40
out: 27 . 27. 32 | flu: 0 . 0 | ran: 17 . 13 | per: 44 . 40
out: 27 . 27. 32 | flu: 0 . 0 | ran: 9 . 13 | per: 36 . 40
out: 29 . 27. 32 | flu: 2 . 0 | ran: 17 . 13 | per: 46 . 40
out: 28 . 27. 32 | flu: 0 . 0 | ran: 9 . 13 | per: 37 . 40
out: 29 . 27. 32 | flu: 0 . 0 | ran: 18 . 13 | per: 47 . 40
out: 28 . 27. 32 | flu: 0 . 0 | ran: 9 . 13 | per: 37 . 40
average slice is the ideal 13 msecs and the period is picture-perfect 40
msecs. But the 'ran' field fluctuates around 13.33 msecs and there's no
mechanism in CFS to keep that from happening: it's a perfectly valid
solution that CFS finds.
to fix this we add a granularity/preemption rule that knows about
the "target latency", which makes tasks that run longer than the ideal
latency run a bit less. The simplest approach is to simply decrease the
preemption granularity when a task overruns its ideal latency. For this
we have to track how much the task executed since its last preemption.
( this adds a new field to task_struct, but we can eliminate that
overhead in 2.6.24 by putting all the scheduler timestamps into an
anonymous union. )
with this change in place, chew-max output is fluctuation-less all
around:
out: 28 . 27. 39 | flu: 0 . 2 | ran: 13 . 13 | per: 41 . 40
out: 28 . 27. 39 | flu: 0 . 2 | ran: 13 . 13 | per: 41 . 40
out: 28 . 27. 39 | flu: 0 . 2 | ran: 13 . 13 | per: 41 . 40
out: 28 . 27. 39 | flu: 0 . 2 | ran: 13 . 13 | per: 41 . 40
out: 28 . 27. 39 | flu: 0 . 1 | ran: 13 . 13 | per: 41 . 40
out: 28 . 27. 39 | flu: 0 . 1 | ran: 13 . 13 | per: 41 . 40
this patch has no impact on any fastpath or on any globally observable
scheduling property. (unless you have sharp enough eyes to see
millisecond-level ruckles in glxgears smoothness :-)
Signed-off-by: Ingo Molnar <mingo@elte.hu>
Signed-off-by: Peter Zijlstra <a.p.zijlstra@chello.nl>
Signed-off-by: Mike Galbraith <efault@gmx.de>
There is an Amarok song switch time increase (regression) under
hefty load.
What is happening is that sleeper_bonus is never consumed, and only
rarely goes below runtime_limit, so for the most part, Amarok isn't
getting any bonus at all. We're keeping sleeper_bonus right at
runtime_limit (sched_latency == sched_runtime_limit == 40ms) forever, ie
we don't consume if we're lower that that, and don't add if we're above
it. One Amarok thread waking (or anybody else) will push us past the
threshold, so the next thread waking gets nada, but will reap pain from
the previous thread waking until we drop back to runtime_limit. It
looks to me like under load, some random task gets a bonus, and
everybody else pays, whether deserving or not.
This diff fixed the regression for me at any load rate.
Signed-off-by: Mike Galbraith <efault@gmx.de>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
Signed-off-by: Peter Zijlstra <a.p.zijlstra@chello.nl>
Fix bogus DEBUG_PREEMPT warning on x86_64, when cpu brought online after
bootup: current_is_keventd is right to note its use of smp_processor_id
is preempt-safe, but should use raw_smp_processor_id to avoid the warning.
Signed-off-by: Hugh Dickins <hugh@veritas.com>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
runtime limit and wakeup granularity used to be a function of
granularity and that was incorrect changed to sched_latency.
Fix this to make wakeup granularity a function of min-granularity,
and the runtime limit equal to latency.
Signed-off-by: Ingo Molnar <mingo@elte.hu>
due to adaptive granularity scheduling the role of sched_granularity
has changed to "minimum granularity", so rename the variable (and the
tunable) accordingly.
Signed-off-by: Ingo Molnar <mingo@elte.hu>
Signed-off-by: Peter Zijlstra <a.p.zijlstra@chello.nl>
Instead of specifying the preemption granularity, specify the wanted
latency. By fixing the granlarity to a constany the wakeup latency
it a function of the number of running tasks on the rq.
Invert this relation.
sysctl_sched_granularity becomes a minimum for the dynamic granularity
computed from the new sysctl_sched_latency.
Then use this latency to do more intelligent granularity decisions: if
there are fewer tasks running then we can schedule coarser. This helps
performance while still always keeping the latency target.
Signed-off-by: Peter Zijlstra <a.p.zijlstra@chello.nl>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
Make the lockdep sysctls not depend on CONFIG_SCHED_DEBUG.
Signed-off-by: Peter Zijlstra <a.p.zijlstra@chello.nl>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
fix task startup penalty miscalculation: sysctl_sched_granularity is
unsigned int and wait_runtime is long so we first have to convert it
to long before turning it negative ...
Signed-off-by: Ingo Molnar <mingo@elte.hu>
current code:
delta = calc_delta_mine(delta_exec, curr->load.weight, lw);
delta = min((u64)delta, cfs_rq->sleeper_bonus);
Notice that this calc_delta_mine() line is exactly delta_mine, which
gives:
delta = min((u64)delta_mine, cfs_rq->sleeper_bonus);
Signed-off-by: Peter Zijlstra <a.p.zijlstra@chello.nl>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
current code:
delta = min(cfs_rq->sleeper_bonus, (u64)delta_exec);
delta = calc_delta_mine(delta, curr->load.weight, lw);
delta = min((u64)delta, cfs_rq->sleeper_bonus);
drop the first min(), because we clip against sleeper_bonus in the 3rd line
again. That gives:
delta = calc_delta_mine(delta_exec, curr->load.weight, lw);
delta = min((u64)delta, cfs_rq->sleeper_bonus);
Signed-off-by: Peter Zijlstra <a.p.zijlstra@chello.nl>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
make the bonus balance more consistent: do not hand out a bonus if
there's too much in flight already, and only deduct as much from a
runner as it has the capacity. This makes the bonus engine a zero-sum
game (as intended).
this also simplifies the code:
text data bss dec hex filename
34770 2998 24 37792 93a0 sched.o.before
34749 2998 24 37771 938b sched.o.after
and it also avoids overscheduling in sleep-happy workloads like
hackbench.c.
Signed-off-by: Ingo Molnar <mingo@elte.hu>
Mitchell Erblich suggested a quality-of-implementation change to
not requeue SCHED_RR tasks if there's only a single task on the
runqueue, by checking for rq->nr_running == 1.
provide a more efficient implementation of that, to check that
particular RT priority-queue only.
[ From: mingo@elte.hu ]
Also first requeue the task then set need_resched - results in slightly
better machine-instruction ordering. Also clean up the code a bit.
Signed-off-by: Dmitry Adamushko <dmitry.adamushko@gmail.com>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
remove HZ dependency from the granularity default. Use 10 msec for
the base granularity, 1 msec for wakeup granularity and 25 msec for
batch wakeup granularity. (These defaults are close to the values
that the default HZ=250 setting got previously, and thus it's the
most common setting.)
Signed-off-by: Ingo Molnar <mingo@elte.hu>
when I built with CONFIG_FAIR_GROUP_SCHED=y, I need the following change
to make things right.
[ From: mingo@elte.hu ]
this config option is not upstream-configurable right now but lets fix
this for completeness.
Signed-off-by: Bruce Ashfield <bruce.ashfield@windriver.com>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
