| Age | Commit message (Collapse) | Author | Files | Lines |
|---|
|
Delegate the random insertion to the forced threaded interrupt
handler. Store the return IP of the hard interrupt handler in the irq
descriptor and feed it into the random generator as a source of
entropy.
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
|
|
CPU bringup calls into the random pool to initialize the stack
canary. During boot that works nicely even on RT as the might sleep
checks are disabled. During CPU hotplug the might sleep checks
trigger. Making the locks in random raw is a major PITA, so avoid the
call on RT is the only sensible solution. This is basically the same
randomness which we get during boot where the random pool has no
entropy and we rely on the TSC randomnness.
Reported-by: Carsten Emde <carsten.emde@osadl.org>
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
|
|
Disable on -RT. If this is invoked from irq-context we will have problems
to acquire the sleeping lock.
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
|
|
The root-lock is dropped before dev_hard_start_xmit() is invoked and after
setting the __QDISC___STATE_RUNNING bit. If this task is now pushed away
by a task with a higher priority then the task with the higher priority
won't be able to submit packets to the NIC directly instead they will be
enqueued into the Qdisc. The NIC will remain idle until the task(s) with
higher priority leave the CPU and the task with lower priority gets back
and finishes the job.
If we take always the busylock we ensure that the RT task can boost the
low-prio task and submit the packet.
Signed-off-by: Sebastian Andrzej Siewior <bigeasy@linutronix.de>
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
|
|
Upstream uses skb_dequeue() to acquire lock of `input_pkt_queue'. The reason is
to synchronize against a remote CPU which still thinks that the CPU is online
enqueues packets to this CPU.
There are no guarantees that the packet is enqueued before the callback is run,
it just hope.
RT however complains about an not initialized lock because it uses another lock
for `input_pkt_queue' due to the IRQ-off nature of the context.
Use the unlocked dequeue version for `input_pkt_queue'.
Signed-off-by: Sebastian Andrzej Siewior <bigeasy@linutronix.de>
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
|
|
Use the rps lock as rawlock so we can keep irq-off regions. It looks low
latency. However we can't kfree() from this context therefore we defer this
to the softirq and use the tofree_queue list for it (similar to process_queue).
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
|
|
In 2004 netif_rx_ni() gained a preempt_disable() section around
netif_rx() and its do_softirq() + testing for it. The do_softirq() part
is required because netif_rx() raises the softirq but does not invoke
it. The preempt_disable() is required to remain on the same CPU which added the
skb to the per-CPU list.
All this can be avoided be putting this into a local_bh_disable()ed
section. The local_bh_enable() part will invoke do_softirq() if
required.
Signed-off-by: Sebastian Andrzej Siewior <bigeasy@linutronix.de>
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
|
|
Delay RCU-selftests until ksoftirqd is up and running.
Signed-off-by: Sebastian Andrzej Siewior <bigeasy@linutronix.de>
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
|
|
i_dir_seq is a sequence counter with a lock which is represented by the lowest
bit. The writer atomically updates the counter which ensures that it can be
modified by only one writer at a time.
The commit introducing this change claims that the lock has been integrated
into the counter for space reasons within the inode struct. The i_dir_seq
member is within a union which shares also a pointer. That means by using
seqlock_t we would have a sequence counter and a lock without increasing the
size of the data structure on 64bit and 32bit would grow by 4 bytes. With
lockdep enabled the size would grow and on PREEMPT_RT the spinlock_t is also
larger.
In order to keep this construct working on PREEMPT_RT, the writer needs to
disable preemption while obtaining the lock on the sequence counter / starting
the write critical section. The writer acquires an otherwise unrelated
spinlock_t which serves the same purpose on !PREEMPT_RT. With enabled
preemption a high priority reader could preempt the writer and live lock the
system while waiting for the locked bit to disappear.
Another solution would be to have global spinlock_t which is always acquired
by the writer. The reader would then acquire the lock if the sequence count is
odd and by doing so force the writer out of the critical section. The global
spinlock_t could be replaced by a hashed lock based on the address of the inode
to lower the lock contention.
For now, manually disable preemption on PREEMPT_RT to avoid live locks.
Reported-by: Oleg.Karfich@wago.com
Signed-off-by: Sebastian Andrzej Siewior <bigeasy@linutronix.de>
|
|
__d_lookup_done() invokes wake_up_all() while holding a hlist_bl_lock()
which disables preemption. As a workaround convert it to swait.
Signed-off-by: Sebastian Andrzej Siewior <bigeasy@linutronix.de>
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
|
|
As explained by Alexander Fyodorov <halcy@yandex.ru>:
|read_lock(&tasklist_lock) in ptrace_stop() is converted to mutex on RT kernel,
|and it can remove __TASK_TRACED from task->state (by moving it to
|task->saved_state). If parent does wait() on child followed by a sys_ptrace
|call, the following race can happen:
|
|- child sets __TASK_TRACED in ptrace_stop()
|- parent does wait() which eventually calls wait_task_stopped() and returns
| child's pid
|- child blocks on read_lock(&tasklist_lock) in ptrace_stop() and moves
| __TASK_TRACED flag to saved_state
|- parent calls sys_ptrace, which calls ptrace_check_attach() and wait_task_inactive()
The patch is based on his initial patch where an additional check is
added in case the __TASK_TRACED moved to ->saved_state. The pi_lock is
taken in case the caller is interrupted between looking into ->state and
->saved_state.
[ Fix for ptrace_unfreeze_traced() by Oleg Nesterov ]
Signed-off-by: Sebastian Andrzej Siewior <bigeasy@linutronix.de>
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
|
|
Upstream commit '53da1d9456fe7f8 fix ptrace slowness' is nothing more
than a bandaid around the ptrace design trainwreck. It's not a
correctness issue, it's merily a cosmetic bandaid.
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
|
|
559271146efc ("mm/memcg: optimize user context object stock access") is a
classic example of optimizing for the cpu local BKL serialization without a
clear protection scope.
Disable MEMCG on RT for now.
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
|
|
On RT the seqcount_t is required even on UP because the softirq can be
preempted. The IRQ handler is threaded so it is also preemptible.
Disable preemption on 32bit-RT during value updates. There is no need to
disable interrupts on RT because the handler is run threaded. Therefore
disabling preemption is enough to guarantee that the update is not
interruped.
Signed-off-by: Sebastian Andrzej Siewior <bigeasy@linutronix.de>
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
|
|
raise_softirq_irqoff() disables interrupts and wakes the softirq
daemon, but after reenabling interrupts there is no preemption check,
so the execution of the softirq thread might be delayed arbitrarily.
In principle we could add that check to local_irq_enable/restore, but
that's overkill as the rasie_softirq_irqoff() sections are the only
ones which show this behaviour.
Reported-by: Carsten Emde <cbe@osadl.org>
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
|
|
cryptd has a per-CPU lock which protected with local_bh_disable() and
preempt_disable().
Add an explicit spin_lock to make the locking context more obvious and
visible to lockdep. Since it is a per-CPU lock, there should be no lock
contention on the actual spinlock.
There is a small race-window where we could be migrated to another CPU
after the cpu_queue has been obtain. This is not a problem because the
actual ressource is protected by the spinlock.
Signed-off-by: Sebastian Andrzej Siewior <bigeasy@linutronix.de>
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
|
|
|BUG: sleeping function called from invalid context at kernel/locking/rtmutex.c:915
|in_atomic(): 1, irqs_disabled(): 0, pid: 3194, name: rpc.nfsd
|Preemption disabled at:[<ffffffffa06bf0bb>] svc_xprt_received+0x4b/0xc0 [sunrpc]
|CPU: 6 PID: 3194 Comm: rpc.nfsd Not tainted 3.18.7-rt1 #9
|Hardware name: MEDION MS-7848/MS-7848, BIOS M7848W08.404 11/06/2014
| ffff880409630000 ffff8800d9a33c78 ffffffff815bdeb5 0000000000000002
| 0000000000000000 ffff8800d9a33c98 ffffffff81073c86 ffff880408dd6008
| ffff880408dd6000 ffff8800d9a33cb8 ffffffff815c3d84 ffff88040b3ac000
|Call Trace:
| [<ffffffff815bdeb5>] dump_stack+0x4f/0x9e
| [<ffffffff81073c86>] __might_sleep+0xe6/0x150
| [<ffffffff815c3d84>] rt_spin_lock+0x24/0x50
| [<ffffffffa06beec0>] svc_xprt_do_enqueue+0x80/0x230 [sunrpc]
| [<ffffffffa06bf0bb>] svc_xprt_received+0x4b/0xc0 [sunrpc]
| [<ffffffffa06c03ed>] svc_add_new_perm_xprt+0x6d/0x80 [sunrpc]
| [<ffffffffa06b2693>] svc_addsock+0x143/0x200 [sunrpc]
| [<ffffffffa072e69c>] write_ports+0x28c/0x340 [nfsd]
| [<ffffffffa072d2ac>] nfsctl_transaction_write+0x4c/0x80 [nfsd]
| [<ffffffff8117ee83>] vfs_write+0xb3/0x1d0
| [<ffffffff8117f889>] SyS_write+0x49/0xb0
| [<ffffffff815c4556>] system_call_fastpath+0x16/0x1b
Signed-off-by: Mike Galbraith <umgwanakikbuti@gmail.com>
Signed-off-by: Sebastian Andrzej Siewior <bigeasy@linutronix.de>
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
|
|
1)enqueue_to_backlog() (called from netif_rx) should be
bind to a particluar CPU. This can be achieved by
disabling migration. No need to disable preemption
2)Fixes crash "BUG: scheduling while atomic: ksoftirqd"
in case of RT.
If preemption is disabled, enqueue_to_backog() is called
in atomic context. And if backlog exceeds its count,
kfree_skb() is called. But in RT, kfree_skb() might
gets scheduled out, so it expects non atomic context.
-Replace preempt_enable(), preempt_disable() with
migrate_enable(), migrate_disable() respectively
-Replace get_cpu(), put_cpu() with get_cpu_light(),
put_cpu_light() respectively
Signed-off-by: Priyanka Jain <Priyanka.Jain@freescale.com>
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
Acked-by: Rajan Srivastava <Rajan.Srivastava@freescale.com>
Cc: <rostedt@goodmis.orgn>
Link: http://lkml.kernel.org/r/1337227511-2271-1-git-send-email-Priyanka.Jain@freescale.com
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
[bigeasy: Remove assumption about migrate_disable() from the description.]
Signed-off-by: Sebastian Andrzej Siewior <bigeasy@linutronix.de>
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
|
|
Avoid the preempt disable version of get_cpu_var(). The inner-lock should
provide enough serialisation.
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
|
|
Do not disable preemption while taking sleeping locks. All user look safe
for migrate_diable() only.
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
|
|
__raid_run_ops() disables preemption with get_cpu() around the access
to the raid5_percpu variables. That causes scheduling while atomic
spews on RT.
Serialize the access to the percpu data with a lock and keep the code
preemptible.
Reported-by: Udo van den Heuvel <udovdh@xs4all.nl>
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
Tested-by: Udo van den Heuvel <udovdh@xs4all.nl>
|
|
preempt_disable() and get_cpu() don't play well together with the sleeping
locks it tries to allocate later.
It seems to be enough to replace it with get_cpu_light() and migrate_disable().
Signed-off-by: Sebastian Andrzej Siewior <bigeasy@linutronix.de>
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
|
|
Signed-off-by: Sebastian Andrzej Siewior <bigeasy@linutronix.de>
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
|
|
PREEMPT_RT
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
Signed-off-by: Sebastian Andrzej Siewior <bigeasy@linutronix.de>
|
|
The printk header file includes ratelimit_types.h for its __ratelimit()
based usage. It requires it for the static initializer used in
printk_ratelimited(). It uses a raw_spinlock_t and includes the
spinlock_types.h. It makes no difference on non PREEMPT-RT builds but
PREEMPT-RT replaces the inner part of some locks and therefore includes
rtmutex.h and atomic.h which leads to recursive includes where defines
are missing.
By including only the raw_spinlock_t defines it avoids the atomic.h
related includes at this stage.
An example on powerpc:
| CALL scripts/atomic/check-atomics.sh
|In file included from include/linux/bug.h:5,
| from include/linux/page-flags.h:10,
| from kernel/bounds.c:10:
|arch/powerpc/include/asm/page_32.h: In function ‘clear_page’:
|arch/powerpc/include/asm/bug.h:87:4: error: implicit declaration of function ‘__WARN’ [-Werror=implicit-function-declaration]
| 87 | __WARN(); \
| | ^~~~~~
|arch/powerpc/include/asm/page_32.h:48:2: note: in expansion of macro ‘WARN_ON’
| 48 | WARN_ON((unsigned long)addr & (L1_CACHE_BYTES - 1));
| | ^~~~~~~
|arch/powerpc/include/asm/bug.h:58:17: error: invalid application of ‘sizeof’ to incomplete type ‘struct bug_entry’
| 58 | "i" (sizeof(struct bug_entry)), \
| | ^~~~~~
|arch/powerpc/include/asm/bug.h:89:3: note: in expansion of macro ‘BUG_ENTRY’
| 89 | BUG_ENTRY(PPC_TLNEI " %4, 0", \
| | ^~~~~~~~~
|arch/powerpc/include/asm/page_32.h:48:2: note: in expansion of macro ‘WARN_ON’
| 48 | WARN_ON((unsigned long)addr & (L1_CACHE_BYTES - 1));
| | ^~~~~~~
|In file included from arch/powerpc/include/asm/ptrace.h:298,
| from arch/powerpc/include/asm/hw_irq.h:12,
| from arch/powerpc/include/asm/irqflags.h:12,
| from include/linux/irqflags.h:16,
| from include/asm-generic/cmpxchg-local.h:6,
| from arch/powerpc/include/asm/cmpxchg.h:526,
| from arch/powerpc/include/asm/atomic.h:11,
| from include/linux/atomic.h:7,
| from include/linux/rwbase_rt.h:6,
| from include/linux/rwlock_types.h:55,
| from include/linux/spinlock_types.h:74,
| from include/linux/ratelimit_types.h:7,
| from include/linux/printk.h:10,
| from include/asm-generic/bug.h:22,
| from arch/powerpc/include/asm/bug.h:109,
| from include/linux/bug.h:5,
| from include/linux/page-flags.h:10,
| from kernel/bounds.c:10:
|include/linux/thread_info.h: In function ‘copy_overflow’:
|include/linux/thread_info.h:210:2: error: implicit declaration of function ‘WARN’ [-Werror=implicit-function-declaration]
| 210 | WARN(1, "Buffer overflow detected (%d < %lu)!\n", size, count);
| | ^~~~
The WARN / BUG include pulls in printk.h and then ptrace.h expects WARN
(from bug.h) which is not yet complete. Even hw_irq.h has WARN_ON()
statements.
On POWERPC64 there are missing atomic64 defines while building 32bit
VDSO:
| VDSO32C arch/powerpc/kernel/vdso32/vgettimeofday.o
|In file included from include/linux/atomic.h:80,
| from include/linux/rwbase_rt.h:6,
| from include/linux/rwlock_types.h:55,
| from include/linux/spinlock_types.h:74,
| from include/linux/ratelimit_types.h:7,
| from include/linux/printk.h:10,
| from include/linux/kernel.h:19,
| from arch/powerpc/include/asm/page.h:11,
| from arch/powerpc/include/asm/vdso/gettimeofday.h:5,
| from include/vdso/datapage.h:137,
| from lib/vdso/gettimeofday.c:5,
| from <command-line>:
|include/linux/atomic-arch-fallback.h: In function ‘arch_atomic64_inc’:
|include/linux/atomic-arch-fallback.h:1447:2: error: implicit declaration of function ‘arch_atomic64_add’; did you mean ‘arch_atomic_add’? [-Werror=impl
|icit-function-declaration]
| 1447 | arch_atomic64_add(1, v);
| | ^~~~~~~~~~~~~~~~~
| | arch_atomic_add
The generic fallback is not included, atomics itself are not used. If
kernel.h does not include printk.h then it comes later from the bug.h
include.
Signed-off-by: Sebastian Andrzej Siewior <bigeasy@linutronix.de>
|
|
The ww-mutex selftest operates directly on ww_mutex::base and assumes
its type is struct mutex. This isn't true on PREEMPT_RT which turns the
mutex into a rtmutex.
Add a ww_mutex_base_ abstraction which maps to the relevant mutex_ or
rt_mutex_ function.
Change the CONFIG_DEBUG_MUTEXES ifdef to DEBUG_WW_MUTEXES. The latter is
true for the MUTEX and RTMUTEX implementation of WW-MUTEX. The
assignment is required in order to pass the tests.
Signed-off-by: Sebastian Andrzej Siewior <bigeasy@linutronix.de>
|
|
The softirq context on PREEMPT_RT is different compared to !PREEMPT_RT.
As such lockdep_softirq_enter() is a nop and the all the "softirq safe"
tests fail on PREEMPT_RT because there is no difference.
Skip the softirq context tests on PREEMPT_RT.
Signed-off-by: Sebastian Andrzej Siewior <bigeasy@linutronix.de>
|
|
The tests with unbalanced lock() + unlock() operation leave a modified
preemption counter behind which is then reset to its original value
after the test.
The spin_lock() function on PREEMPT_RT does not include a
preempt_disable() statement but migrate_disable() and read_rcu_lock().
As a consequence both counter never get back to their original value and
system explodes later after the selftest.
In the double-unlock case on PREEMPT_RT, the migrate_disable() and RCU
code will trigger which should be avoided. These counter should not be
decremented below their initial value.
Save both counters and bring them back to their original value after the
test.
In the double-unlock case, increment both counter in advance to they
become balanced after the double unlock.
Signed-off-by: Sebastian Andrzej Siewior <bigeasy@linutronix.de>
|
|
The last column contains the results for the rtmutex tests.
Add it.
Signed-off-by: Sebastian Andrzej Siewior <bigeasy@linutronix.de>
|
|
There is not really a softirq context on PREEMPT_RT.
Softirqs on PREEMPT_RT are always invoked within the context of a threaded
interrupt handler or within ksoftirqd. The "in-softirq" context is preemptible
and is protected by a per-CPU lock to ensure mutual exclusion.
There is no difference on PREEMPT_RT between spin_lock_irq() and spin_lock()
because the former does not disable interrupts. Therefore if lock is used
in_softirq() and locked once with spin_lock_irq() then lockdep will report this
with "inconsistent {SOFTIRQ-ON-W} -> {IN-SOFTIRQ-W} usage".
Teach lockdep that we don't really do softirqs on -RT.
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
|
|
The locking selftest for ww-mutex expects to operate directly on the
base-mutex which becomes a rtmutex on PREEMPT_RT.
Add rt_mutex_lock_nest_lock(), follows mutex_lock_nest_lock() for
rtmutex.
Add rt_mutex_lock_killable(), follows mutex_lock_killable() for rtmutex.
Signed-off-by: Sebastian Andrzej Siewior <bigeasy@linutronix.de>
|
|
The lockdep selftest for ww-mutex assumes in a few cases the
ww_ctx->contending_lock assignment via __ww_mutex_check_kill() which
does not happen if the rtmutex detects the deadlock early.
The testcase passes if the deadlock handling here is removed. This means
that it will work if multiple threads/tasks are involved and not just a
single one.
Signed-off-by: Sebastian Andrzej Siewior <bigeasy@linutronix.de>
|
|
If migrate_enable() is used more often than its counter part then it
remains undetected and rq::nr_pinned will underflow, too.
Add a warning if migrate_enable() is attempted if without a matching a
migrate_disable().
Signed-off-by: Sebastian Andrzej Siewior <bigeasy@linutronix.de>
|
|
The functions local_lock related functions
local_lock_acquire()
local_lock_release()
are part of the internal implementation and should be avoided.
Define the lock as DEFINE_PER_CPU so the normal local_lock() function
can be used.
Signed-off-by: Sebastian Andrzej Siewior <bigeasy@linutronix.de>
|
|
rt_rwlock_is_contended() has not users. It makes no sense to use it as
rwlock_is_contended() because it is a sleeping lock on RT and preemption
is possible. It reports always != 0 if used by a writer and even if
there is a waiter then the lock might not be handed over if the
current owner has the highest priority.
Remove rt_rwlock_is_contended().
Reported-by: kernel test robot <lkp@intel.com>
Signed-off-by: Sebastian Andrzej Siewior <bigeasy@linutronix.de>
|
|
Some architectures are using stop_machine() while switching the opcode which
leads to latency spikes.
The architectures which use stop_machine() atm:
- ARM stop machine
- s390 stop machine
The architecures which use other sorcery:
- MIPS
- X86
- powerpc
- sparc
- arm64
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
[bigeasy: only ARM for now]
Signed-off-by: Sebastian Andrzej Siewior <bigeasy@linutronix.de>
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
|
|
Commit
9652dc2eb9e40 ("tcp: relax listening_hash operations")
removed the need to disable bottom half while acquiring
listening_hash.lock. There are still two callers left which disable
bottom half before the lock is acquired.
Drop local_bh_disable() around __inet_hash() which acquires
listening_hash->lock, invoke inet_ehash_nolisten() with disabled BH.
inet_unhash() conditionally acquires listening_hash->lock.
Reported-by: Mike Galbraith <efault@gmx.de>
Signed-off-by: Sebastian Andrzej Siewior <bigeasy@linutronix.de>
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
Link: https://lore.kernel.org/linux-rt-users/12d6f9879a97cd56c09fb53dee343cbb14f7f1f7.camel@gmx.de/
Link: https://lkml.kernel.org/r/X9CheYjuXWc75Spa@hirez.programming.kicks-ass.net
|
|
Commit
68d48e6a2df57 ("mm: workingset: add vmstat counter for shadow nodes")
introduced an IRQ-off check to ensure that a lock is held which also
disabled interrupts. This does not work the same way on -RT because none
of the locks, that are held, disable interrupts.
Replace this check with a lockdep assert which ensures that the lock is
held.
Cc: Peter Zijlstra <peterz@infradead.org>
Signed-off-by: Sebastian Andrzej Siewior <bigeasy@linutronix.de>
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
Link: https://lkml.kernel.org/r/20190211113829.sqf6bdi4c4cdd3rp@linutronix.de
|
|
All callers of cgroup_rstat_flush_locked() acquire cgroup_rstat_lock
either with spin_lock_irq() or spin_lock_irqsave().
cgroup_rstat_flush_locked() itself acquires cgroup_rstat_cpu_lock which
is a raw_spin_lock. This lock is also acquired in cgroup_rstat_updated()
in IRQ context and therefore requires _irqsave() locking suffix in
cgroup_rstat_flush_locked().
Since there is no difference between spin_lock_t and raw_spin_lock_t
on !RT lockdep does not complain here. On RT lockdep complains because
the interrupts were not disabled here and a deadlock is possible.
Acquire the raw_spin_lock_t with disabled interrupts.
Signed-off-by: Sebastian Andrzej Siewior <bigeasy@linutronix.de>
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
Link: https://www.spinics.net/lists/cgroups/msg23051.html
|
|
mmdrop() is invoked from finish_task_switch() by the incoming task to drop
the mm which was handed over by the previous task. mmdrop() can be quite
expensive which prevents an incoming real-time task from getting useful
work done.
Provide mmdrop_sched() which maps to mmdrop() on !RT kernels. On RT kernels
it delagates the eventually required invocation of __mmdrop() to RCU.
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
Signed-off-by: Sebastian Andrzej Siewior <bigeasy@linutronix.de>
Link: https://lore.kernel.org/r/20210928122411.648582026@linutronix.de
|
|
Anything which is done on behalf of a dead task at the end of
finish_task_switch() is preventing the incoming task from doing useful
work. While it is benefitial for fork heavy workloads to recycle the task
stack quickly, this is a latency source for real-time tasks.
Therefore delay the stack cleanup on RT enabled kernels.
Signed-off-by: Sebastian Andrzej Siewior <bigeasy@linutronix.de>
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
Link: https://lore.kernel.org/r/20210928122411.593486363@linutronix.de
|
|
Doing cleanups in the tail of schedule() is a latency punishment for the
incoming task. The point of invoking kprobes_task_flush() for a dead task
is that the instances are returned and cannot leak when __schedule() is
kprobed.
Move it into the delayed cleanup.
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
Cc: Masami Hiramatsu <mhiramat@kernel.org>
Signed-off-by: Sebastian Andrzej Siewior <bigeasy@linutronix.de>
Link: https://lore.kernel.org/r/20210928122411.537994026@linutronix.de
|
|
The queued remote wakeup mechanism has turned out to be suboptimal for RT
enabled kernels. The maximum latencies go up by a factor of > 5x in certain
scenarious.
This is caused by either long wake lists or by a large number of TTWU IPIs
which are processed back to back.
Disable it for RT.
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
Signed-off-by: Sebastian Andrzej Siewior <bigeasy@linutronix.de>
Link: https://lore.kernel.org/r/20210928122411.482262764@linutronix.de
|
|
Batched task migrations are a source for large latencies as they keep the
scheduler from running while processing the migrations.
Limit the batch size to 8 instead of 32 when running on a RT enabled
kernel.
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
Signed-off-by: Sebastian Andrzej Siewior <bigeasy@linutronix.de>
Link: https://lore.kernel.org/r/20210928122411.425097596@linutronix.de
|
|
The general rule that rcu_read_lock() held sections cannot voluntary sleep
does apply even on RT kernels. Though the substitution of spin/rw locks on
RT enabled kernels has to be exempt from that rule. On !RT a spin_lock()
can obviously nest inside a RCU read side critical section as the lock
acquisition is not going to block, but on RT this is not longer the case
due to the 'sleeping' spinlock substitution.
The RT patches contained a cheap hack to ignore the RCU nesting depth in
might_sleep() checks, which was a pragmatic but incorrect workaround.
Instead of generally ignoring the RCU nesting depth in __might_sleep() and
__might_resched() checks, pass the rcu_preempt_depth() via the offsets
argument to __might_resched() from spin/read/write_lock() which makes the
checks work correctly even in RCU read side critical sections.
The actual blocking on such a substituted lock within a RCU read side
critical section is already handled correctly in __schedule() by treating
it as a "preemption" of the RCU read side critical section.
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
Signed-off-by: Sebastian Andrzej Siewior <bigeasy@linutronix.de>
Link: https://lore.kernel.org/r/20210923165358.368305497@linutronix.de
|
|
The __might_resched() checks in the cond_resched_lock() variants use
PREEMPT_LOCK_OFFSET for preempt count offset checking which takes the
preemption disable by the spin_lock() which is still held at that point
into account.
On PREEMPT_RT enabled kernels spin/rw_lock held sections stay preemptible
which means PREEMPT_LOCK_OFFSET is 0, but that still triggers the
__might_resched() check because that takes RCU read side nesting into
account.
On RT enabled kernels spin/read/write_lock() issue rcu_read_lock() to
resemble the !RT semantics, which means in cond_resched_lock() the might
resched check will see preempt_count() == 0 and rcu_preempt_depth() == 1.
Introduce PREEMPT_LOCK_SCHED_OFFSET for those might resched checks and map
them depending on CONFIG_PREEMPT_RT.
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
Signed-off-by: Sebastian Andrzej Siewior <bigeasy@linutronix.de>
Link: https://lore.kernel.org/r/20210923165358.305969211@linutronix.de
|
|
For !RT kernels RCU nest depth in __might_resched() is always expected to
be 0, but on RT kernels it can be non zero while the preempt count is
expected to be always 0.
Instead of playing magic games in interpreting the 'preempt_offset'
argument, rename it to 'offsets' and use the lower 8 bits for the expected
preempt count, allow to hand in the expected RCU nest depth in the upper
bits and adopt the __might_resched() code and related checks and printks.
The affected call sites are updated in subsequent steps.
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
Signed-off-by: Sebastian Andrzej Siewior <bigeasy@linutronix.de>
Link: https://lore.kernel.org/r/20210923165358.243232823@linutronix.de
|
|
might_sleep() output is pretty informative, but can be confusing at times
especially with PREEMPT_RCU when the check triggers due to a voluntary
sleep inside a RCU read side critical section:
BUG: sleeping function called from invalid context at kernel/test.c:110
in_atomic(): 0, irqs_disabled(): 0, non_block: 0, pid: 415, name: kworker/u112:52
Preemption disabled at: migrate_disable+0x33/0xa0
in_atomic() is 0, but it still tells that preemption was disabled at
migrate_disable(), which is completely useless because preemption is not
disabled. But the interesting information to decode the above, i.e. the RCU
nesting depth, is not printed.
That becomes even more confusing when might_sleep() is invoked from
cond_resched_lock() within a RCU read side critical section. Here the
expected preemption count is 1 and not 0.
BUG: sleeping function called from invalid context at kernel/test.c:131
in_atomic(): 1, irqs_disabled(): 0, non_block: 0, pid: 415, name: kworker/u112:52
Preemption disabled at: test_cond_lock+0xf3/0x1c0
So in_atomic() is set, which is expected as the caller holds a spinlock,
but it's unclear why this is broken and the preempt disable IP is just
pointing at the correct place, i.e. spin_lock(), which is obviously not
helpful either.
Make that more useful in general:
- Print preempt_count() and the expected value
and for the CONFIG_PREEMPT_RCU case:
- Print the RCU read side critical section nesting depth
- Print the preempt disable IP only when preempt count
does not have the expected value.
So the might_sleep() dump from a within a preemptible RCU read side
critical section becomes:
BUG: sleeping function called from invalid context at kernel/test.c:110
in_atomic(): 0, irqs_disabled(): 0, non_block: 0, pid: 415, name: kworker/u112:52
preempt_count: 0, expected: 0
RCU nest depth: 1, expected: 0
and the cond_resched_lock() case becomes:
BUG: sleeping function called from invalid context at kernel/test.c:141
in_atomic(): 1, irqs_disabled(): 0, non_block: 0, pid: 415, name: kworker/u112:52
preempt_count: 1, expected: 1
RCU nest depth: 1, expected: 0
which makes is pretty obvious what's going on. For all other cases the
preempt disable IP is still printed as before:
BUG: sleeping function called from invalid context at kernel/test.c: 156
in_atomic(): 1, irqs_disabled(): 0, non_block: 0, pid: 1, name: swapper/0
preempt_count: 1, expected: 0
RCU nest depth: 0, expected: 0
Preemption disabled at:
[<ffffffff82b48326>] test_might_sleep+0xbe/0xf8
BUG: sleeping function called from invalid context at kernel/test.c: 163
in_atomic(): 1, irqs_disabled(): 0, non_block: 0, pid: 1, name: swapper/0
preempt_count: 1, expected: 0
RCU nest depth: 1, expected: 0
Preemption disabled at:
[<ffffffff82b48326>] test_might_sleep+0x1e4/0x280
This also prepares to provide a better debugging output for RT enabled
kernels and their spinlock substitutions.
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
Signed-off-by: Sebastian Andrzej Siewior <bigeasy@linutronix.de>
Link: https://lore.kernel.org/r/20210923165358.181022656@linutronix.de
|
|
Convert them to pr_*(). No functional change.
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
Signed-off-by: Sebastian Andrzej Siewior <bigeasy@linutronix.de>
Link: https://lore.kernel.org/r/20210923165358.117496067@linutronix.de
|
|
All callers hand in 0 and never will hand in anything else.
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
Signed-off-by: Sebastian Andrzej Siewior <bigeasy@linutronix.de>
Link: https://lore.kernel.org/r/20210923165358.054321586@linutronix.de
|
