Currently, RCU Tasks Trace (as well as the other two flavors of RCU Tasks)
use a single global callback list. This works well and is simple, but
expected changes in workload will cause this list to become a bottleneck.
This commit therefore creates per-CPU callback lists for the various
flavors of RCU Tasks, but continues queueing on a single list, namely
that of CPU 0. Later commits will dynamically vary the number of lists
in use to accommodate dynamic changes in workload.
Reported-by: Martin Lau <kafai@fb.com>
Cc: Neeraj Upadhyay <neeraj.iitr10@gmail.com>
Tested-by: kernel test robot <beibei.si@intel.com>
Signed-off-by: Paul E. McKenney <paulmck@kernel.org>
rcu_core() tries to ensure that its self-invocation in case of callbacks
overload only happen in softirq/rcuc mode. Indeed it doesn't make sense
to trigger local RCU core from nocb_cb kthread since it can execute
on a CPU different from the target rdp. Also in case of overload, the
nocb_cb kthread simply iterates a new loop of callbacks processing.
However the "offloaded" check that aims at preventing misplaced
rcu_core() invocations is wrong. First of all that state is volatile
and second: softirq/rcuc can execute while the target rdp is offloaded.
As a result rcu_core() can be invoked on the wrong CPU while in the
process of (de-)offloading.
Fix that with moving the rcu_core() self-invocation to rcu_core() itself,
irrespective of the rdp offloaded state.
Tested-by: Valentin Schneider <valentin.schneider@arm.com>
Tested-by: Sebastian Andrzej Siewior <bigeasy@linutronix.de>
Signed-off-by: Frederic Weisbecker <frederic@kernel.org>
Cc: Valentin Schneider <valentin.schneider@arm.com>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Sebastian Andrzej Siewior <bigeasy@linutronix.de>
Cc: Josh Triplett <josh@joshtriplett.org>
Cc: Joel Fernandes <joel@joelfernandes.org>
Cc: Boqun Feng <boqun.feng@gmail.com>
Cc: Neeraj Upadhyay <neeraju@codeaurora.org>
Cc: Uladzislau Rezki <urezki@gmail.com>
Cc: Thomas Gleixner <tglx@linutronix.de>
Signed-off-by: Paul E. McKenney <paulmck@kernel.org>
The current condition to limit the number of callbacks executed in a
row checks the offloaded state of the rdp. Not only is it volatile
but it is also misleading: the rcu_core() may well be executing
callbacks concurrently with NOCB kthreads, and the offloaded state
would then be verified on both cases. As a result the limit would
spuriously not apply anymore on softirq while in the middle of
(de-)offloading process.
Fix and clarify the condition with those constraints in mind:
_ If callbacks are processed either by rcuc or NOCB kthread, the call
to cond_resched_tasks_rcu_qs() is enough to take care of the overload.
_ If instead callbacks are processed by softirqs:
* If need_resched(), exit the callbacks processing
* Otherwise if CPU is idle we can continue
* Otherwise exit because a softirq shouldn't interrupt a task for too
long nor deprive other pending softirq vectors of the CPU.
Tested-by: Valentin Schneider <valentin.schneider@arm.com>
Tested-by: Sebastian Andrzej Siewior <bigeasy@linutronix.de>
Signed-off-by: Frederic Weisbecker <frederic@kernel.org>
Cc: Valentin Schneider <valentin.schneider@arm.com>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Sebastian Andrzej Siewior <bigeasy@linutronix.de>
Cc: Josh Triplett <josh@joshtriplett.org>
Cc: Joel Fernandes <joel@joelfernandes.org>
Cc: Boqun Feng <boqun.feng@gmail.com>
Cc: Neeraj Upadhyay <neeraju@codeaurora.org>
Cc: Uladzislau Rezki <urezki@gmail.com>
Cc: Thomas Gleixner <tglx@linutronix.de>
Signed-off-by: Paul E. McKenney <paulmck@kernel.org>
It's not entirely obvious why rdp->qlen_last_fqs_check is updated before
processing the queue only on offloaded rdp. There can be different
effect to that, either in favour of triggering the force quiescent state
path or not. For example:
1) If the number of callbacks has decreased since the last
rdp->qlen_last_fqs_check update (because we recently called
rcu_do_batch() and we executed below qhimark callbacks) and the number
of processed callbacks on a subsequent do_batch() arranges for
exceeding qhimark on non-offloaded but not on offloaded setup, then we
may spare a later run to the force quiescent state
slow path on __call_rcu_nocb_wake(), as compared to the non-offloaded
counterpart scenario.
Here is such an offloaded scenario instance:
qhimark = 1000
rdp->last_qlen_last_fqs_check = 3000
rcu_segcblist_n_cbs(rdp) = 2000
rcu_do_batch() {
if (offloaded)
rdp->last_qlen_fqs_check = rcu_segcblist_n_cbs(rdp) // 2000
// run 1000 callback
rcu_segcblist_n_cbs(rdp) = 1000
// Not updating rdp->qlen_last_fqs_check
if (count < rdp->qlen_last_fqs_check - qhimark)
rdp->qlen_last_fqs_check = count;
}
call_rcu() * 1001 {
__call_rcu_nocb_wake() {
// not taking the fqs slowpath:
// rcu_segcblist_n_cbs(rdp) == 2001
// rdp->qlen_last_fqs_check == 2000
// qhimark == 1000
if (len > rdp->qlen_last_fqs_check + qhimark)
...
}
In the case of a non-offloaded scenario, rdp->qlen_last_fqs_check
would be 1000 and the fqs slowpath would have executed.
2) If the number of callbacks has increased since the last
rdp->qlen_last_fqs_check update (because we recently queued below
qhimark callbacks) and the number of callbacks executed in rcu_do_batch()
doesn't exceed qhimark for either offloaded or non-offloaded setup,
then it's possible that the offloaded scenario later run the force
quiescent state slow path on __call_rcu_nocb_wake() while the
non-offloaded doesn't.
qhimark = 1000
rdp->last_qlen_last_fqs_check = 3000
rcu_segcblist_n_cbs(rdp) = 2000
rcu_do_batch() {
if (offloaded)
rdp->last_qlen_last_fqs_check = rcu_segcblist_n_cbs(rdp) // 2000
// run 100 callbacks
// concurrent queued 100
rcu_segcblist_n_cbs(rdp) = 2000
// Not updating rdp->qlen_last_fqs_check
if (count < rdp->qlen_last_fqs_check - qhimark)
rdp->qlen_last_fqs_check = count;
}
call_rcu() * 1001 {
__call_rcu_nocb_wake() {
// Taking the fqs slowpath:
// rcu_segcblist_n_cbs(rdp) == 3001
// rdp->qlen_last_fqs_check == 2000
// qhimark == 1000
if (len > rdp->qlen_last_fqs_check + qhimark)
...
}
In the case of a non-offloaded scenario, rdp->qlen_last_fqs_check
would be 3000 and the fqs slowpath would have executed.
The reason for updating rdp->qlen_last_fqs_check when invoking callbacks
for offloaded CPUs is that there is usually no point in waking up either
the rcuog or rcuoc kthreads while in this state. After all, both threads
are prohibited from indefinite sleeps.
The exception is when some huge number of callbacks are enqueued while
rcu_do_batch() is in the midst of invoking, in which case interrupting
the rcuog kthread's timed sleep might get more callbacks set up for the
next grace period.
Reported-and-tested-by: Valentin Schneider <valentin.schneider@arm.com>
Tested-by: Sebastian Andrzej Siewior <bigeasy@linutronix.de>
Original-patch-by: Thomas Gleixner <tglx@linutronix.de>
Signed-off-by: Frederic Weisbecker <frederic@kernel.org>
Cc: Valentin Schneider <valentin.schneider@arm.com>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Sebastian Andrzej Siewior <bigeasy@linutronix.de>
Cc: Josh Triplett <josh@joshtriplett.org>
Cc: Joel Fernandes <joel@joelfernandes.org>
Cc: Boqun Feng <boqun.feng@gmail.com>
Cc: Neeraj Upadhyay <neeraju@codeaurora.org>
Cc: Uladzislau Rezki <urezki@gmail.com>
Cc: Thomas Gleixner <tglx@linutronix.de>
Signed-off-by: Paul E. McKenney <paulmck@kernel.org>
When callbacks are offloaded, the NOCB kthreads handle the callbacks
progression on behalf of rcu_core().
However during the (de-)offloading process, the kthread may not be
entirely up to the task. As a result some callbacks grace period
sequence number may remain stale for a while because rcu_core() won't
take care of them either.
Fix this with forcing callbacks acceleration from rcu_core() as long
as the offloading process isn't complete.
Reported-and-tested-by: Valentin Schneider <valentin.schneider@arm.com>
Tested-by: Sebastian Andrzej Siewior <bigeasy@linutronix.de>
Signed-off-by: Frederic Weisbecker <frederic@kernel.org>
Cc: Valentin Schneider <valentin.schneider@arm.com>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Sebastian Andrzej Siewior <bigeasy@linutronix.de>
Cc: Josh Triplett <josh@joshtriplett.org>
Cc: Joel Fernandes <joel@joelfernandes.org>
Cc: Boqun Feng <boqun.feng@gmail.com>
Cc: Neeraj Upadhyay <neeraju@codeaurora.org>
Cc: Uladzislau Rezki <urezki@gmail.com>
Cc: Thomas Gleixner <tglx@linutronix.de>
Signed-off-by: Paul E. McKenney <paulmck@kernel.org>
While reporting a quiescent state for a given CPU, rcu_core() takes
advantage of the freshly loaded grace period sequence number and the
locked rnp to accelerate the callbacks whose sequence number have been
assigned a stale value.
This action is only necessary when the rdp isn't offloaded, otherwise
the NOCB kthreads already take care of the callbacks progression.
However the check for the offloaded state is volatile because it is
performed outside the IRQs disabled section. It's possible for the
offloading process to preempt rcu_core() at that point on PREEMPT_RT.
This is dangerous because rcu_core() may end up accelerating callbacks
concurrently with NOCB kthreads without appropriate locking.
Fix this with moving the offloaded check inside the rnp locking section.
Reported-and-tested-by: Valentin Schneider <valentin.schneider@arm.com>
Reviewed-by: Valentin Schneider <valentin.schneider@arm.com>
Tested-by: Sebastian Andrzej Siewior <bigeasy@linutronix.de>
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Sebastian Andrzej Siewior <bigeasy@linutronix.de>
Cc: Josh Triplett <josh@joshtriplett.org>
Cc: Joel Fernandes <joel@joelfernandes.org>
Cc: Boqun Feng <boqun.feng@gmail.com>
Cc: Neeraj Upadhyay <neeraju@codeaurora.org>
Cc: Uladzislau Rezki <urezki@gmail.com>
Cc: Thomas Gleixner <tglx@linutronix.de>
Signed-off-by: Frederic Weisbecker <frederic@kernel.org>
Signed-off-by: Paul E. McKenney <paulmck@kernel.org>
On PREEMPT_RT, if rcu_core() is preempted by the de-offloading process,
some work, such as callbacks acceleration and invocation, may be left
unattended due to the volatile checks on the offloaded state.
In the worst case this work is postponed until the next rcu_pending()
check that can take a jiffy to reach, which can be a problem in case
of callbacks flooding.
Solve that with invoking rcu_core() early in the de-offloading process.
This way any work dismissed by an ongoing rcu_core() call fooled by
a preempting deoffloading process will be caught up by a nearby future
recall to rcu_core(), this time fully aware of the de-offloading state.
Tested-by: Valentin Schneider <valentin.schneider@arm.com>
Tested-by: Sebastian Andrzej Siewior <bigeasy@linutronix.de>
Signed-off-by: Frederic Weisbecker <frederic@kernel.org>
Cc: Valentin Schneider <valentin.schneider@arm.com>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Sebastian Andrzej Siewior <bigeasy@linutronix.de>
Cc: Josh Triplett <josh@joshtriplett.org>
Cc: Joel Fernandes <joel@joelfernandes.org>
Cc: Boqun Feng <boqun.feng@gmail.com>
Cc: Neeraj Upadhyay <neeraju@codeaurora.org>
Cc: Uladzislau Rezki <urezki@gmail.com>
Cc: Thomas Gleixner <tglx@linutronix.de>
Signed-off-by: Paul E. McKenney <paulmck@kernel.org>
Currently SEGCBLIST_SOFTIRQ_ONLY is a bit of an exception among the
segcblist flags because it is an exclusive state that doesn't mix up
with the other flags. Remove it in favour of:
_ A flag specifying that rcu_core() needs to perform callbacks execution
and acceleration
and
_ A flag specifying we want the nocb lock to be held in any needed
circumstances
This clarifies the code and is more flexible: It allows to have a state
where rcu_core() runs with locking while offloading hasn't started yet.
This is a necessary step to prepare for triggering rcu_core() at the
very beginning of the de-offloading process so that rcu_core() won't
dismiss work while being preempted by the de-offloading process, at
least not without a pending subsequent rcu_core() that will quickly
catch up.
Reviewed-by: Valentin Schneider <Valentin.Schneider@arm.com>
Tested-by: Valentin Schneider <valentin.schneider@arm.com>
Tested-by: Sebastian Andrzej Siewior <bigeasy@linutronix.de>
Signed-off-by: Frederic Weisbecker <frederic@kernel.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Sebastian Andrzej Siewior <bigeasy@linutronix.de>
Cc: Josh Triplett <josh@joshtriplett.org>
Cc: Joel Fernandes <joel@joelfernandes.org>
Cc: Boqun Feng <boqun.feng@gmail.com>
Cc: Neeraj Upadhyay <neeraju@codeaurora.org>
Cc: Uladzislau Rezki <urezki@gmail.com>
Cc: Thomas Gleixner <tglx@linutronix.de>
Signed-off-by: Paul E. McKenney <paulmck@kernel.org>
rcu_nocb_lock_irqsave() can be preempted between the call to
rcu_segcblist_is_offloaded() and the actual locking. This matters now
that rcu_core() is preemptible on PREEMPT_RT and the (de-)offloading
process can interrupt the softirq or the rcuc kthread.
As a result we may locklessly call into code that requires nocb locking.
In practice this is a problem while we accelerate callbacks on rcu_core().
Simply disabling interrupts before (instead of after) checking the NOCB
offload state fixes the issue.
Reported-and-tested-by: Valentin Schneider <valentin.schneider@arm.com>
Tested-by: Sebastian Andrzej Siewior <bigeasy@linutronix.de>
Signed-off-by: Frederic Weisbecker <frederic@kernel.org>
Cc: Valentin Schneider <valentin.schneider@arm.com>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Sebastian Andrzej Siewior <bigeasy@linutronix.de>
Cc: Josh Triplett <josh@joshtriplett.org>
Cc: Joel Fernandes <joel@joelfernandes.org>
Cc: Boqun Feng <boqun.feng@gmail.com>
Cc: Neeraj Upadhyay <neeraju@codeaurora.org>
Cc: Uladzislau Rezki <urezki@gmail.com>
Cc: Thomas Gleixner <tglx@linutronix.de>
Signed-off-by: Paul E. McKenney <paulmck@kernel.org>
Currently, rcu_advance_cbs_nowake() checks that a grace period is in
progress, however, that grace period could end just after the check.
This commit rechecks that a grace period is still in progress while
holding the rcu_node structure's lock. The grace period cannot end while
the current CPU's rcu_node structure's ->lock is held, thus avoiding
false positives from the WARN_ON_ONCE().
As Daniel Vacek noted, it is not necessary for the rcu_node structure
to have a CPU that has not yet passed through its quiescent state.
Tested-by: Guillaume Morin <guillaume@morinfr.org>
Signed-off-by: Paul E. McKenney <paulmck@kernel.org>
Expedited RCU grace periods invoke sync_rcu_exp_select_node_cpus(), which
takes two passes over the leaf rcu_node structure's CPUs. The first
pass gathers up the current CPU and CPUs that are in dynticks idle mode.
The workqueue will report a quiescent state on their behalf later.
The second pass sends IPIs to the rest of the CPUs, but excludes the
current CPU, incorrectly assuming it has been included in the first
pass's list of CPUs.
Unfortunately the current CPU may have changed between the first and
second pass, due to the fact that the various rcu_node structures'
->lock fields have been dropped, thus momentarily enabling preemption.
This means that if the second pass's CPU was not on the first pass's
list, it will be ignored completely. There will be no IPI sent to
it, and there will be no reporting of quiescent states on its behalf.
Unfortunately, the expedited grace period will nevertheless be waiting
for that CPU to report a quiescent state, but with that CPU having no
reason to believe that such a report is needed.
The result will be an expedited grace period stall.
Fix this by no longer excluding the current CPU from consideration during
the second pass.
Fixes: b9ad4d6ed1 ("rcu: Avoid self-IPI in sync_rcu_exp_select_node_cpus()")
Reviewed-by: Neeraj Upadhyay <quic_neeraju@quicinc.com>
Signed-off-by: Frederic Weisbecker <frederic@kernel.org>
Cc: Uladzislau Rezki <urezki@gmail.com>
Cc: Neeraj Upadhyay <quic_neeraju@quicinc.com>
Cc: Boqun Feng <boqun.feng@gmail.com>
Cc: Josh Triplett <josh@joshtriplett.org>
Cc: Joel Fernandes <joel@joelfernandes.org>
Signed-off-by: Paul E. McKenney <paulmck@kernel.org>
In non-preemptible kernels, an unfortunately timed expedited grace period
can result in the rcu_exp_handler() IPI handler setting the rcu_data
structure's cpu_no_qs.b.exp field just as the target CPU enters idle.
There are situations in which this field will not be checked until after
that CPU exits idle. The resulting grace-period latency does not qualify
as "expedited".
This commit therefore checks this field upon non-preemptible idle entry in
the rcu_preempt_deferred_qs() function. It also qualifies the rcu_core()
preempt_count() check with IS_ENABLED(CONFIG_PREEMPT_COUNT) to prevent
false-positive quiescent states from count-free kernels.
Reported-by: Neeraj Upadhyay <neeraju@codeaurora.org>
Signed-off-by: Paul E. McKenney <paulmck@kernel.org>
If an RCU expedited grace period starts just when a CPU is in the process
of going offline, so that the outgoing CPU has completed its pass through
stop-machine but has not yet completed its final dive into the idle loop,
RCU will attempt to enable that CPU's scheduling-clock tick via a call
to tick_dep_set_cpu(). For this to happen, that CPU has to have been
online when the expedited grace period completed its CPU-selection phase.
This is pointless: The outgoing CPU has interrupts disabled, so it cannot
take a scheduling-clock tick anyway. In addition, the tick_dep_set_cpu()
function's eventual call to irq_work_queue_on() will splat as follows:
smpboot: CPU 1 is now offline
WARNING: CPU: 6 PID: 124 at kernel/irq_work.c:95
+irq_work_queue_on+0x57/0x60
Modules linked in:
CPU: 6 PID: 124 Comm: kworker/6:2 Not tainted 5.15.0-rc1+ #3
Hardware name: QEMU Standard PC (Q35 + ICH9, 2009), BIOS
+rel-1.14.0-0-g155821a-rebuilt.opensuse.org 04/01/2014
Workqueue: rcu_gp wait_rcu_exp_gp
RIP: 0010:irq_work_queue_on+0x57/0x60
Code: 8b 05 1d c7 ea 62 a9 00 00 f0 00 75 21 4c 89 ce 44 89 c7 e8
+9b 37 fa ff ba 01 00 00 00 89 d0 c3 4c 89 cf e8 3b ff ff ff eb ee <0f> 0b eb b7
+0f 0b eb db 90 48 c7 c0 98 2a 02 00 65 48 03 05 91
6f
RSP: 0000:ffffb12cc038fe48 EFLAGS: 00010282
RAX: 0000000000000001 RBX: 0000000000005208 RCX: 0000000000000020
RDX: 0000000000000001 RSI: 0000000000000001 RDI: ffff9ad01f45a680
RBP: 000000000004c990 R08: 0000000000000001 R09: ffff9ad01f45a680
R10: ffffb12cc0317db0 R11: 0000000000000001 R12: 00000000fffecee8
R13: 0000000000000001 R14: 0000000000026980 R15: ffffffff9e53ae00
FS: 0000000000000000(0000) GS:ffff9ad01f580000(0000)
+knlGS:0000000000000000
CS: 0010 DS: 0000 ES: 0000 CR0: 0000000080050033
CR2: 0000000000000000 CR3: 000000000de0c000 CR4: 00000000000006e0
DR0: 0000000000000000 DR1: 0000000000000000 DR2: 0000000000000000
DR3: 0000000000000000 DR6: 00000000fffe0ff0 DR7: 0000000000000400
Call Trace:
tick_nohz_dep_set_cpu+0x59/0x70
rcu_exp_wait_wake+0x54e/0x870
? sync_rcu_exp_select_cpus+0x1fc/0x390
process_one_work+0x1ef/0x3c0
? process_one_work+0x3c0/0x3c0
worker_thread+0x28/0x3c0
? process_one_work+0x3c0/0x3c0
kthread+0x115/0x140
? set_kthread_struct+0x40/0x40
ret_from_fork+0x22/0x30
---[ end trace c5bf75eb6aa80bc6 ]---
This commit therefore avoids invoking tick_dep_set_cpu() on offlined
CPUs to limit both futility and false-positive splats.
Signed-off-by: Paul E. McKenney <paulmck@kernel.org>
The sync_sched_exp_online_cleanup() is called from rcutree_online_cpu(),
which can be invoked with interrupts enabled. This means that
the ->cpu_no_qs.b.exp field is subject to data races from the
rcu_exp_handler() IPI handler, so this commit marks the load from
that field.
Signed-off-by: Paul E. McKenney <paulmck@kernel.org>
Having two fields for the same purpose with subtle differences on
different RCU flavours is confusing, especially when both fields always
exist on both RCU flavours.
Fortunately, it is now safe for preemptible RCU to rely on the rcu_data
structure's ->cpu_no_qs.b.exp field, just like non-preemptible RCU.
This commit therefore removes the ad-hoc ->exp_deferred_qs field.
Signed-off-by: Frederic Weisbecker <frederic@kernel.org>
Signed-off-by: Paul E. McKenney <paulmck@kernel.org>
On non-preemptible RCU, move clearing of the rcu_data structure's
->cpu_no_qs.b.exp filed to the actual expedited quiescent state report
function, matching hw preemptible RCU handles the ->exp_deferred_qs field.
This prepares for removing ->exp_deferred_qs in favor of ->cpu_no_qs.b.exp
for both preemptible and non-preemptible RCU.
Signed-off-by: Frederic Weisbecker <frederic@kernel.org>
Signed-off-by: Paul E. McKenney <paulmck@kernel.org>
Preemptible RCU does not use the rcu_data structure's ->cpu_no_qs.b.exp,
instead using a separate ->exp_deferred_qs field to record the need for
an expedited quiescent state.
In fact ->cpu_no_qs.b.exp should never be set in preemptible RCU because
preemptible RCU's expedited grace periods use other mechanisms to record
quiescent states.
This commit therefore removes the implicit rcu_qs() reference to
->cpu_no_qs.b.exp in favor of a direct reference to ->cpu_no_qs.b.norm.
Signed-off-by: Frederic Weisbecker <frederic@kernel.org>
Signed-off-by: Paul E. McKenney <paulmck@kernel.org>
When rt_runtime is modified from -1 to a valid control value, it may
cause the task to be throttled all the time. Operations like the following
will trigger the bug. E.g:
1. echo -1 > /proc/sys/kernel/sched_rt_runtime_us
2. Run a FIFO task named A that executes while(1)
3. echo 950000 > /proc/sys/kernel/sched_rt_runtime_us
When rt_runtime is -1, The rt period timer will not be activated when task
A enqueued. And then the task will be throttled after setting rt_runtime to
950,000. The task will always be throttled because the rt period timer is
not activated.
Fixes: d0b27fa778 ("sched: rt-group: synchonised bandwidth period")
Reported-by: Hulk Robot <hulkci@huawei.com>
Signed-off-by: Li Hua <hucool.lihua@huawei.com>
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Link: https://lkml.kernel.org/r/20211203033618.11895-1-hucool.lihua@huawei.com
All People I know including myself took a long time to figure out that
typical wakeup will always go to fast path and never go to slow path
except WF_FORK and WF_EXEC.
Vincent reminded me once in a linaro meeting and made me understand
slow path won't happen for WF_TTWU. But my other friends repeatedly
wasted a lot of time on testing this path like me before I reminded
them.
So obviously the code needs some document.
Signed-off-by: Barry Song <song.bao.hua@hisilicon.com>
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Link: https://lkml.kernel.org/r/20211016111109.5559-1-21cnbao@gmail.com
Add a new helper to deal with the swiotlb case. This keeps the code
nicely boundled and removes the not required call to
dma_direct_optimal_gfp_mask for the swiotlb case.
Signed-off-by: Christoph Hellwig <hch@lst.de>
Reviewed-by: Robin Murphy <robin.murphy@arm.com>
swiotlb_alloc and swiotlb_free are properly stubbed out if
CONFIG_DMA_RESTRICTED_POOL is not set, so skip the extra checks.
Signed-off-by: Christoph Hellwig <hch@lst.de>
Reviewed-by: Robin Murphy <robin.murphy@arm.com>
Instead of blindly running into a blocking operation for a non-blocking gfp,
return NULL and spew an error. Note that Kconfig prevents this for all
currently relevant platforms, and this is just a debug check.
Signed-off-by: Christoph Hellwig <hch@lst.de>
Reviewed-by: Robin Murphy <robin.murphy@arm.com>
If the architecture can't remap or set an address uncached there is no way
to fullfill a request for a coherent allocation. Return NULL in that case.
Note that this case currently does not happen, so this is a theoretical
fixup and/or a preparation for eventually supporting platforms that
can't support coherent allocations with the generic code.
Signed-off-by: Christoph Hellwig <hch@lst.de>
Reviewed-by: Robin Murphy <robin.murphy@arm.com>
Add a big central !dev_is_dma_coherent(dev) block to deal with as much
as of the uncached allocation schemes and document the schemes a bit
better.
Signed-off-by: Christoph Hellwig <hch@lst.de>
Reviewed-by: Robin Murphy <robin.murphy@arm.com>
Split the code for DMA_ATTR_NO_KERNEL_MAPPING allocations into a separate
helper to make dma_direct_alloc a little more readable.
Signed-off-by: Christoph Hellwig <hch@lst.de>
Reviewed-by: Robin Murphy <robin.murphy@arm.com>
Acked-by: David Rientjes <rientjes@google.com>
Add two local variables to track if we want to remap the returned
address using vmap or call dma_set_uncached and use that to simplify
the code flow.
Signed-off-by: Christoph Hellwig <hch@lst.de>
Reviewed-by: Robin Murphy <robin.murphy@arm.com>
We must never let unencrypted memory go back into the general page pool.
So if we fail to set it back to encrypted when freeing DMA memory, leak
the memory instead and warn the user.
Signed-off-by: Christoph Hellwig <hch@lst.de>
Reviewed-by: Robin Murphy <robin.murphy@arm.com>
Remapped allocations handle the encrypted bit through the pgprot passed
to vmap, so there is no call dma_set_decrypted. Note that this case is
currently entirely theoretical as no valid kernel configuration supports
remapped allocations and memory encryption currently.
Signed-off-by: Christoph Hellwig <hch@lst.de>
Factor out helpers the make dealing with memory encryption a little less
cumbersome.
Signed-off-by: Christoph Hellwig <hch@lst.de>
Reviewed-by: Robin Murphy <robin.murphy@arm.com>
The eprobes open code the reserving of the event on the ring buffer for
ftrace instead of using the ftrace event wrappers, which means that it
doesn't get affected by the filters, breaking the filtering logic on user
space.
Link: https://lkml.kernel.org/r/20211130024319.068451680@goodmis.org
Signed-off-by: Steven Rostedt (VMware) <rostedt@goodmis.org>
In case trace_event_buffer_lock_reserve() is called with preemption
enabled, the algorithm that defines the usage of the per cpu filter buffer
may fail if the task schedules to another CPU after determining which
buffer it will use.
Disable preemption when using the filter buffer. And because that same
buffer must be used throughout the call, keep preemption disabled until
the filter buffer is released.
This will also keep the semantics between the use case of when the filter
buffer is used, and when the ring buffer itself is used, as that case also
disables preemption until the ring buffer is released.
Link: https://lkml.kernel.org/r/20211130024318.880190623@goodmis.org
[ Fixed warning of assignment in if statement
Reported-by: kernel test robot <lkp@intel.com> ]
Signed-off-by: Steven Rostedt (VMware) <rostedt@goodmis.org>
The value read by this_cpu_read() is used later and its use is expected to
stay on the same CPU as being read. But this_cpu_read() does not warn if
it is called without preemption disabled, where as __this_cpu_read() will
check if preemption is disabled on CONFIG_DEBUG_PREEMPT
Currently all callers have preemption disabled, but there may be new
callers in the future that may not.
Link: https://lkml.kernel.org/r/20211130024318.698165354@goodmis.org
Signed-off-by: Steven Rostedt (VMware) <rostedt@goodmis.org>
Add '__rel_loc' new dynamic data location attribute which encodes
the data location from the next to the field itself.
The '__data_loc' is used for encoding the dynamic data location on
the trace event record. But '__data_loc' is not useful if the writer
doesn't know the event header (e.g. user event), because it records
the dynamic data offset from the entry of the record, not the field
itself.
This new '__rel_loc' attribute encodes the data location relatively
from the next of the field. For example, when there is a record like
below (the number in the parentheses is the size of fields)
|header(N)|common(M)|fields(K)|__data_loc(4)|fields(L)|data(G)|
In this case, '__data_loc' field will be
__data_loc = (G << 16) | (N+M+K+4+L)
If '__rel_loc' is used, this will be
|header(N)|common(M)|fields(K)|__rel_loc(4)|fields(L)|data(G)|
where
__rel_loc = (G << 16) | (L)
This case shows L bytes after the '__rel_loc' attribute field,
if there is no fields after the __rel_loc field, L must be 0.
This is relatively easy (and no need to consider the kernel header
change) when the event data fields are composed by user who doesn't
know header and common fields.
Link: https://lkml.kernel.org/r/163757341258.510314.4214431827833229956.stgit@devnote2
Cc: Beau Belgrave <beaub@linux.microsoft.com>
Cc: Namhyung Kim <namhyung@kernel.org>
Cc: Tom Zanussi <zanussi@kernel.org>
Signed-off-by: Masami Hiramatsu <mhiramat@kernel.org>
Signed-off-by: Steven Rostedt (VMware) <rostedt@goodmis.org>
Branch data available to BPF programs can be very useful to get stack traces
out of userspace application.
Commit fff7b64355 ("bpf: Add bpf_read_branch_records() helper") added BPF
support to capture branch records in x86. Enable this feature also for other
architectures as well by removing checks specific to x86.
If an architecture doesn't support branch records, bpf_read_branch_records()
still has appropriate checks and it will return an -EINVAL in that scenario.
Based on UAPI helper doc in include/uapi/linux/bpf.h, unsupported architectures
should return -ENOENT in such case. Hence, update the appropriate check to
return -ENOENT instead.
Selftest 'perf_branches' result on power9 machine which has the branch stacks
support:
- Before this patch:
[command]# ./test_progs -t perf_branches
#88/1 perf_branches/perf_branches_hw:FAIL
#88/2 perf_branches/perf_branches_no_hw:OK
#88 perf_branches:FAIL
Summary: 0/1 PASSED, 0 SKIPPED, 1 FAILED
- After this patch:
[command]# ./test_progs -t perf_branches
#88/1 perf_branches/perf_branches_hw:OK
#88/2 perf_branches/perf_branches_no_hw:OK
#88 perf_branches:OK
Summary: 1/2 PASSED, 0 SKIPPED, 0 FAILED
Selftest 'perf_branches' result on power9 machine which doesn't have branch
stack report:
- After this patch:
[command]# ./test_progs -t perf_branches
#88/1 perf_branches/perf_branches_hw:SKIP
#88/2 perf_branches/perf_branches_no_hw:OK
#88 perf_branches:OK
Summary: 1/1 PASSED, 1 SKIPPED, 0 FAILED
Fixes: fff7b64355 ("bpf: Add bpf_read_branch_records() helper")
Suggested-by: Peter Zijlstra <peterz@infradead.org>
Signed-off-by: Kajol Jain <kjain@linux.ibm.com>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
Link: https://lore.kernel.org/bpf/20211206073315.77432-1-kjain@linux.ibm.com
The variable @bcon has two meanings. It is used several times for iterating
the list of registered consoles. In the meantime, it holds the information
whether a boot console is first in @console_drivers list.
The information about the 1st console driver used to be important for
the decision whether to install the new console by default or not.
It allowed to re-evaluate the variable @need_default_console when
a real console with tty binding has been unregistered in the meantime.
The decision about the default console is not longer affected by @bcon
variable. The current code checks whether the first driver is real
and has tty binding directly.
The information about the first console is still used for two more
decisions:
1. It prevents duplicate output on non-boot consoles with
CON_CONSDEV flag set.
2. Early/boot consoles are unregistered when a real console with
CON_CONSDEV is registered and @keep_bootcon is not set.
The behavior in the real life is far from obvious. @bcon is set according
to the first console @console_drivers list. But the first position in
the list is special:
1. Consoles with CON_CONSDEV flag are put at the beginning of
the list. It is either the preferred console or any console
with tty binding registered by default.
2. Another console might become the first in the list when
the first console in the list is unregistered. It might
happen either explicitly or automatically when boot
consoles are unregistered.
There is one more important rule:
+ Boot consoles can't be registered when any real console
is already registered.
It is a puzzle. The main complication is the dependency on the first
position is the list and the complicated rules around it.
Let's try to make it easier:
1. Add variable @bootcon_enabled and set it by iterating all registered
consoles. The variable has obvious meaning and more predictable
behavior. Any speed optimization and other tricks are not worth it.
2. Use a generic name for the variable that is used to iterate
the list on registered console drivers.
Behavior change:
No, maybe surprisingly, there is _no_ behavior change!
Let's provide the proof by contradiction. Both operations, duplicate
output prevention and boot consoles removal, are done only when
the newly added console has CON_CONSDEV flag set. The behavior
would change when the new @bootcon_enabled has different value
than the original @bcon.
By other words, the behavior would change when the following conditions
are true:
+ a console with CON_CONSDEV flag is added
+ a real (non-boot) console is the first in the list
+ a boot console is later in the list
Now, a real console might be first in the list only when:
+ It was the first registered console. In this case, there can't be
any boot console because any later ones were rejected.
+ It was put at the first position because it had CON_CONSDEV flag
set. It was either the preferred console or it was a console with
tty binding registered by default. We are interested only in
a real consoles here. And real console with tty binding fulfills
conditions of the default console.
Now, there is always only one console that is either preferred
or fulfills conditions of the default console. It can't be already
in the list and being registered at the same time.
As a result, the above three conditions could newer be "true" at
the same time. Therefore the behavior can't change.
Final dilemma:
OK, the new code has the same behavior. But is the change in the right
direction? What if the handling of @console_drivers is updated in
the future?
OK, let's look at it from another angle:
1. The ordering of @console_drivers list is important only in
console_device() function. The first console driver with tty
binding gets associated with /dev/console.
2. CON_CONSDEV flag is shown in /proc/consoles. And it should be set
for the driver that is returned by console_device().
3. A boot console is removed and the duplicated output is prevented
when the real console with CON_CONSDEV flag is registered.
Now, in the ideal world:
+ The driver associated with /dev/console should be either a console
preferred via the command line, device tree, or SPCR. Or it should
be the first real console with tty binding registered by default.
+ The code should match the related boot and real console drivers.
It should unregister only the obsolete boot driver. And the duplicated
output should be prevented only on the related real driver.
It is clear that it is not guaranteed by the current code. Instead,
the current code looks like a maze of heuristics that try to achieve
the above.
It is result of adding several features over last few decades. For example,
a possibility to register more consoles, unregister consoles, boot
consoles, consoles without tty binding, device tree, SPCR, braille
consoles.
Anyway, there is no reason why the decision, about removing boot consoles
and preventing duplicated output, should depend on the first console
in the list. The current code does the decisions primary by CON_CONSDEV
flag that is used for the preferred console. It looks like a
good compromise. And the change seems to be in the right direction.
Signed-off-by: Petr Mladek <pmladek@suse.com>
Link: https://lore.kernel.org/r/20211122132649.12737-6-pmladek@suse.com
The variable @need_default_console is used to decide whether a newly
registered console should get enabled by default.
The logic is complicated. It can be modified in a register_console()
call. But it is always re-evaluated in the next call by the following
condition:
if (need_default_console || bcon || !console_drivers)
need_default_console = preferred_console < 0;
In short, the value is updated when either of the condition is valid:
+ the value is still, or again, "true"
+ boot/early console is still the first in @console_driver list
+ @console_driver list is empty
The value is updated according to @preferred_console. In particular,
it is set to "false" when a @preferred_console was set by
__add_preferred_console(). This happens when a non-braille console
was added via the command line, device tree, or SPCR.
It far from clear what this all means together. Let's look at
@need_default_console from another angle:
1. The value is "true" by default. It means that it is always set
according to @preferred_console during the first register_console()
call.
By other words, the first register_console() call will register
the console by default only when none non-braille console was defined
via the command line, device tree, or SPCR.
2. The value will always stay "false" when @preferred_console is set.
By other words, try_enable_default_console() will never get called
when a non-braille console is explicitly required.
4. The value might be set to "false" in try_enable_default_console()
when a console with tty binding (driver) gets enabled.
In this case CON_CONSDEV is set as well. It causes that the console
will be inserted as first into the list @console_driver. It might
be either real or boot/early console.
5. The value will be set _back_ to "true" in the next register_console()
call when:
+ The console added by the previous register_console() had been
a boot/early one.
+ The last console has been unregistered in the meantime and
a boot/early console became first in @console_drivers list
again. Or the list became empty.
By other words, the value will stay "false" only when the last
registered console was real, had tty binding, and was not removed
in the mean time.
The main logic looks clear:
+ Consoles are enabled by default only when no one is preferred
via the command line, device tree, or SPCR.
+ By default, any console is enabled until a real console
with tty binding gets registered.
The behavior when the real console with tty binding is later removed
is a bit unclear:
+ By default, any new console is registered again only when there
is no console or the first console in the list is a boot one.
The question is why the code is suddenly happy when a real console
without tty binding is the first in the list. It looks like an overlook
and bug.
Conclusion:
The state of @preferred_console and the first console in @console_driver
list should be enough to decide whether we need to enable the given console
by default.
The rules are simple. New consoles are _not_ enabled by default
when either of the following conditions is true:
+ @preferred_console is set. It means that a non-braille console
is explicitly configured via the command line, device tree, or SPCR.
+ A real console with tty binding is registered. Such a console will
have CON_CONSDEV flag set and will always be the first in
@console_drivers list.
Note:
The new code does not use @bcon variable. The meaning of the variable
is far from clear. The direct check of the first console in the list
makes it more clear that only real console fulfills requirements
of the default console.
Behavior change:
As already discussed above. There was one situation where the original
code worked a strange way. Let's have:
+ console A: real console without tty binding
+ console B: real console with tty binding
and do:
register_console(A); /* 1st step */
register_console(B); /* 2nd step */
unregister_console(B); /* 3rd step */
register_console(B); /* 4th step */
The original code will not register the console B in the 4th step.
@need_default_console is set to "false" in 2nd step. The real console
with tty binding (driver) is then removed in the 3rd step.
But @need_default_console will stay "false" in the 4th step because
there is no boot/early console and @registered_consoles list is not
empty.
The new code will register the console B in the 4th step because
it checks whether the first console has tty binding (->driver)
This behavior change should acceptable:
1. The scenario requires manual intervention (console removal).
The system should boot with the same consoles as before.
2. Console B is registered again probably because the user wants
to use it. The most likely scenario is that the related
module is reloaded.
3. It makes the behavior more consistent and predictable.
Signed-off-by: Petr Mladek <pmladek@suse.com>
Link: https://lore.kernel.org/r/20211122132649.12737-5-pmladek@suse.com
There is no need to clear @need_default_console when a console
preferred by the command line, device tree, or SPCR, gets enabled.
The code is called only when some non-braille console matched a console
in @console_cmdline array. It means that a non-braille console was added
in __add_preferred_console() and the variable preferred_console is set
to a number >= 0. As a result, @need_default_console is always set to
"false" in the magic condition:
if (need_default_console || bcon || !console_drivers)
need_default_console = preferred_console < 0;
This is one small step in removing the above magic condition
that is hard to follow.
The patch removes one superfluous assignment and should not change
the functionality.
Signed-off-by: Petr Mladek <pmladek@suse.com>
Link: https://lore.kernel.org/r/20211122132649.12737-4-pmladek@suse.com
The logic around the variable @has_preferred_console made my head
spin many times. Part of the problem is the ambiguous name.
There is the variable @preferred_console. It points to the last
non-braille console in @console_cmdline array. This array contains
consoles preferred via the command line, device tree, or SPCR.
Then there is the variable @has_preferred_console. It is set to
"true" when @preferred_console is enabled or when a console with
tty binding gets enabled by default.
It might get reset back by the magic condition:
if (!has_preferred_console || bcon || !console_drivers)
has_preferred_console = preferred_console >= 0;
It is a puzzle. Dumb explanation is that it gets re-evaluated
when:
+ it was not set before (see above when it gets set)
+ there is still an early console enabled (bcon)
+ there is no console enabled (!console_drivers)
This is still a puzzle.
It gets more clear when we see where the value is checked. The only
meaning of the variable is to decide whether we should try to enable
the new console by default.
Rename the variable according to the single situation where
the value is checked.
The rename requires an inverted logic. Otherwise, it is a simple
search & replace. It does not change the functionality.
Signed-off-by: Petr Mladek <pmladek@suse.com>
Reviewed-by: Sergey Senozhatsky <senozhatsky@chromium.org>
Link: https://lore.kernel.org/r/20211122132649.12737-3-pmladek@suse.com
Put the code enabling a console by default into a separate function
called try_enable_default_console().
Rename try_enable_new_console() to try_enable_preferred_console() to
make the purpose of the different variants more clear.
It is a code refactoring without any functional change.
Signed-off-by: Petr Mladek <pmladek@suse.com>
Reviewed-by: Sergey Senozhatsky <senozhatsky@chromium.org>
Link: https://lore.kernel.org/r/20211122132649.12737-2-pmladek@suse.com
Pull timer fix from Borislav Petkov:
- Prevent a tick storm when a dedicated timekeeper CPU in nohz_full
mode runs for prolonged periods with interrupts disabled and ends up
programming the next tick in the past, leading to that storm
* tag 'timers_urgent_for_v5.16_rc4' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip:
timers/nohz: Last resort update jiffies on nohz_full IRQ entry
