commit f90fff1e15 upstream.
If an exiting non-autoreaping task has already passed exit_notify() and
calls handle_posix_cpu_timers() from IRQ, it can be reaped by its parent
or debugger right after unlock_task_sighand().
If a concurrent posix_cpu_timer_del() runs at that moment, it won't be
able to detect timer->it.cpu.firing != 0: cpu_timer_task_rcu() and/or
lock_task_sighand() will fail.
Add the tsk->exit_state check into run_posix_cpu_timers() to fix this.
This fix is not needed if CONFIG_POSIX_CPU_TIMERS_TASK_WORK=y, because
exit_task_work() is called before exit_notify(). But the check still
makes sense, task_work_add(&tsk->posix_cputimers_work.work) will fail
anyway in this case.
Cc: stable@vger.kernel.org
Reported-by: Benoît Sevens <bsevens@google.com>
Fixes: 0bdd2ed413 ("sched: run_posix_cpu_timers: Don't check ->exit_state, use lock_task_sighand()")
Signed-off-by: Oleg Nesterov <oleg@redhat.com>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
commit 53dac34539 upstream.
hrtimers are migrated away from the dying CPU to any online target at
the CPUHP_AP_HRTIMERS_DYING stage in order not to delay bandwidth timers
handling tasks involved in the CPU hotplug forward progress.
However wakeups can still be performed by the outgoing CPU after
CPUHP_AP_HRTIMERS_DYING. Those can result again in bandwidth timers being
armed. Depending on several considerations (crystal ball power management
based election, earliest timer already enqueued, timer migration enabled or
not), the target may eventually be the current CPU even if offline. If that
happens, the timer is eventually ignored.
The most notable example is RCU which had to deal with each and every of
those wake-ups by deferring them to an online CPU, along with related
workarounds:
_ e787644caf (rcu: Defer RCU kthreads wakeup when CPU is dying)
_ 9139f93209 (rcu/nocb: Fix RT throttling hrtimer armed from offline CPU)
_ f7345ccc62 (rcu/nocb: Fix rcuog wake-up from offline softirq)
The problem isn't confined to RCU though as the stop machine kthread
(which runs CPUHP_AP_HRTIMERS_DYING) reports its completion at the end
of its work through cpu_stop_signal_done() and performs a wake up that
eventually arms the deadline server timer:
WARNING: CPU: 94 PID: 588 at kernel/time/hrtimer.c:1086 hrtimer_start_range_ns+0x289/0x2d0
CPU: 94 UID: 0 PID: 588 Comm: migration/94 Not tainted
Stopper: multi_cpu_stop+0x0/0x120 <- stop_machine_cpuslocked+0x66/0xc0
RIP: 0010:hrtimer_start_range_ns+0x289/0x2d0
Call Trace:
<TASK>
start_dl_timer
enqueue_dl_entity
dl_server_start
enqueue_task_fair
enqueue_task
ttwu_do_activate
try_to_wake_up
complete
cpu_stopper_thread
Instead of providing yet another bandaid to work around the situation, fix
it in the hrtimers infrastructure instead: always migrate away a timer to
an online target whenever it is enqueued from an offline CPU.
This will also allow to revert all the above RCU disgraceful hacks.
Fixes: 5c0930ccaa ("hrtimers: Push pending hrtimers away from outgoing CPU earlier")
Reported-by: Vlad Poenaru <vlad.wing@gmail.com>
Reported-by: Usama Arif <usamaarif642@gmail.com>
Signed-off-by: Frederic Weisbecker <frederic@kernel.org>
Signed-off-by: Paul E. McKenney <paulmck@kernel.org>
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
Cc: stable@vger.kernel.org
Tested-by: Paul E. McKenney <paulmck@kernel.org>
Link: https://lore.kernel.org/all/20250117232433.24027-1-frederic@kernel.org
Closes: 20241213203739.1519801-1-usamaarif642@gmail.com
Signed-off-by: Zhaoyang Li <lizy04@hust.edu.cn>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
[ Upstream commit a52067c24c ]
This reverts commit f590308536 ("timer debug: Hide kernel addresses via
%pK in /proc/timer_list")
The timer list helper SEQ_printf() uses either the real seq_printf() for
procfs output or vprintk() to print to the kernel log, when invoked from
SysRq-q. It uses %pK for printing pointers.
In the past %pK was prefered over %p as it would not leak raw pointer
values into the kernel log. Since commit ad67b74d24 ("printk: hash
addresses printed with %p") the regular %p has been improved to avoid this
issue.
Furthermore, restricted pointers ("%pK") were never meant to be used
through printk(). They can still unintentionally leak raw pointers or
acquire sleeping looks in atomic contexts.
Switch to the regular pointer formatting which is safer, easier to reason
about and sufficient here.
Signed-off-by: Thomas Weißschuh <thomas.weissschuh@linutronix.de>
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
Link: https://lore.kernel.org/lkml/20250113171731-dc10e3c1-da64-4af0-b767-7c7070468023@linutronix.de/
Link: https://lore.kernel.org/all/20250311-restricted-pointers-timer-v1-1-6626b91e54ab@linutronix.de
Signed-off-by: Sasha Levin <sashal@kernel.org>
[ Upstream commit 27af31e449 ]
When is_migration_base() is unused, it prevents kernel builds
with clang, `make W=1` and CONFIG_WERROR=y:
kernel/time/hrtimer.c:156:20: error: unused function 'is_migration_base' [-Werror,-Wunused-function]
156 | static inline bool is_migration_base(struct hrtimer_clock_base *base)
| ^~~~~~~~~~~~~~~~~
Fix this by marking it with __always_inline.
[ tglx: Use __always_inline instead of __maybe_unused and move it into the
usage sites conditional ]
Signed-off-by: Andy Shevchenko <andriy.shevchenko@linux.intel.com>
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
Link: https://lore.kernel.org/all/20250116160745.243358-1-andriy.shevchenko@linux.intel.com
Signed-off-by: Sasha Levin <sashal@kernel.org>
commit ed4fb6d7ef upstream.
The timerslack_ns setting is used to specify how much the hardware
timers should be delayed, to potentially dispatch multiple timers in a
single interrupt. This is a performance optimization. Timers of
realtime tasks (having a realtime scheduling policy) should not be
delayed.
This logic was inconsitently applied to the hrtimers, leading to delays
of realtime tasks which used timed waits for events (e.g. condition
variables). Due to the downstream override of the slack for rt tasks,
the procfs reported incorrect (non-zero) timerslack_ns values.
This is changed by setting the timer_slack_ns task attribute to 0 for
all tasks with a rt policy. By that, downstream users do not need to
specially handle rt tasks (w.r.t. the slack), and the procfs entry
shows the correct value of "0". Setting non-zero slack values (either
via procfs or PR_SET_TIMERSLACK) on tasks with a rt policy is ignored,
as stated in "man 2 PR_SET_TIMERSLACK":
Timer slack is not applied to threads that are scheduled under a
real-time scheduling policy (see sched_setscheduler(2)).
The special handling of timerslack on rt tasks in downstream users
is removed as well.
Signed-off-by: Felix Moessbauer <felix.moessbauer@siemens.com>
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
Link: https://lore.kernel.org/all/20240814121032.368444-2-felix.moessbauer@siemens.com
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
[ Upstream commit 6bb05a3333 ]
The following bug report happened with a PREEMPT_RT kernel:
BUG: sleeping function called from invalid context at kernel/locking/spinlock_rt.c:48
in_atomic(): 1, irqs_disabled(): 0, non_block: 0, pid: 2012, name: kwatchdog
preempt_count: 1, expected: 0
RCU nest depth: 0, expected: 0
get_random_u32+0x4f/0x110
clocksource_verify_choose_cpus+0xab/0x1a0
clocksource_verify_percpu.part.0+0x6b/0x330
clocksource_watchdog_kthread+0x193/0x1a0
It is due to the fact that clocksource_verify_choose_cpus() is invoked with
preemption disabled. This function invokes get_random_u32() to obtain
random numbers for choosing CPUs. The batched_entropy_32 local lock and/or
the base_crng.lock spinlock in driver/char/random.c will be acquired during
the call. In PREEMPT_RT kernel, they are both sleeping locks and so cannot
be acquired in atomic context.
Fix this problem by using migrate_disable() to allow smp_processor_id() to
be reliably used without introducing atomic context. preempt_disable() is
then called after clocksource_verify_choose_cpus() but before the
clocksource measurement is being run to avoid introducing unexpected
latency.
Fixes: 7560c02bdf ("clocksource: Check per-CPU clock synchronization when marked unstable")
Suggested-by: Sebastian Andrzej Siewior <bigeasy@linutronix.de>
Signed-off-by: Waiman Long <longman@redhat.com>
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
Reviewed-by: Paul E. McKenney <paulmck@kernel.org>
Reviewed-by: Sebastian Andrzej Siewior <bigeasy@linutronix.de>
Link: https://lore.kernel.org/all/20250131173323.891943-2-longman@redhat.com
Signed-off-by: Sasha Levin <sashal@kernel.org>
[ Upstream commit 1f566840a8 ]
The "Checking clocksource synchronization" message is normally printed
when clocksource_verify_percpu() is called for a given clocksource if
both the CLOCK_SOURCE_UNSTABLE and CLOCK_SOURCE_VERIFY_PERCPU flags
are set.
It is an informational message and so pr_info() is the correct choice.
Signed-off-by: Waiman Long <longman@redhat.com>
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
Reviewed-by: Paul E. McKenney <paulmck@kernel.org>
Acked-by: John Stultz <jstultz@google.com>
Link: https://lore.kernel.org/all/20250125015442.3740588-1-longman@redhat.com
Stable-dep-of: 6bb05a3333 ("clocksource: Use migrate_disable() to avoid calling get_random_u32() in atomic context")
Signed-off-by: Sasha Levin <sashal@kernel.org>
commit 2f8dea1692 upstream.
Consider a scenario where a CPU transitions from CPUHP_ONLINE to halfway
through a CPU hotunplug down to CPUHP_HRTIMERS_PREPARE, and then back to
CPUHP_ONLINE:
Since hrtimers_prepare_cpu() does not run, cpu_base.hres_active remains set
to 1 throughout. However, during a CPU unplug operation, the tick and the
clockevents are shut down at CPUHP_AP_TICK_DYING. On return to the online
state, for instance CFS incorrectly assumes that the hrtick is already
active, and the chance of the clockevent device to transition to oneshot
mode is also lost forever for the CPU, unless it goes back to a lower state
than CPUHP_HRTIMERS_PREPARE once.
This round-trip reveals another issue; cpu_base.online is not set to 1
after the transition, which appears as a WARN_ON_ONCE in enqueue_hrtimer().
Aside of that, the bulk of the per CPU state is not reset either, which
means there are dangling pointers in the worst case.
Address this by adding a corresponding startup() callback, which resets the
stale per CPU state and sets the online flag.
[ tglx: Make the new callback unconditionally available, remove the online
modification in the prepare() callback and clear the remaining
state in the starting callback instead of the prepare callback ]
Fixes: 5c0930ccaa ("hrtimers: Push pending hrtimers away from outgoing CPU earlier")
Signed-off-by: Koichiro Den <koichiro.den@canonical.com>
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
Cc: stable@vger.kernel.org
Link: https://lore.kernel.org/all/20241220134421.3809834-1-koichiro.den@canonical.com
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
commit f5807b0606 upstream.
Due to an unsigned cast, adjtimex() returns the wrong offest when using
ADJ_MICRO and the offset is negative. In this case a small negative offset
returns approximately 4.29 seconds (~ 2^32/1000 milliseconds) due to the
unsigned cast of the negative offset.
This cast was added when the kernel internal struct timex was changed to
use type long long for the time offset value to address the problem of a
64bit/32bit division on 32bit systems.
The correct cast would have been (s32), which is correct as time_offset can
only be in the range of [INT_MIN..INT_MAX] because the shift constant used
for calculating it is 32. But that's non-obvious.
Remove the cast and use div_s64() to cure the issue.
[ tglx: Fix white space damage, use div_s64() and amend the change log ]
Fixes: ead25417f8 ("timex: use __kernel_timex internally")
Signed-off-by: Marcelo Dalmas <marcelo.dalmas@ge.com>
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
Cc: stable@vger.kernel.org
Link: https://lore.kernel.org/all/SJ0P101MB03687BF7D5A10FD3C49C51E5F42E2@SJ0P101MB0368.NAMP101.PROD.OUTLOOK.COM
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
[ Upstream commit 92b043fd99 ]
The details about the handling of the "normal" values were moved
to the _msecs_to_jiffies() helpers in commit ca42aaf0c8 ("time:
Refactor msecs_to_jiffies"). However, the same commit still mentioned
__msecs_to_jiffies() in the added documentation.
Thus point to _msecs_to_jiffies() instead.
Fixes: ca42aaf0c8 ("time: Refactor msecs_to_jiffies")
Signed-off-by: Miguel Ojeda <ojeda@kernel.org>
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
Link: https://lore.kernel.org/all/20241025110141.157205-2-ojeda@kernel.org
Signed-off-by: Sasha Levin <sashal@kernel.org>
[ Upstream commit b05aefc1f5 ]
The documentation's intention is to compare msecs_to_jiffies() (first
sentence) with __msecs_to_jiffies() (second sentence), which is what the
original documentation did. One of the cleanups in commit f3cb80804b
("time: Fix various kernel-doc problems") may have thought the paragraph
was talking about the latter since that is what it is being documented.
Thus revert that part of the change.
Fixes: f3cb80804b ("time: Fix various kernel-doc problems")
Signed-off-by: Miguel Ojeda <ojeda@kernel.org>
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
Link: https://lore.kernel.org/all/20241025110141.157205-1-ojeda@kernel.org
Signed-off-by: Sasha Levin <sashal@kernel.org>
[ Upstream commit 6e62807c7f ]
If get_clock_desc() succeeds, it calls fget() for the clockid's fd,
and get the clk->rwsem read lock, so the error path should release
the lock to make the lock balance and fput the clockid's fd to make
the refcount balance and release the fd related resource.
However the below commit left the error path locked behind resulting in
unbalanced locking. Check timespec64_valid_strict() before
get_clock_desc() to fix it, because the "ts" is not changed
after that.
Fixes: d8794ac20a ("posix-clock: Fix missing timespec64 check in pc_clock_settime()")
Acked-by: Richard Cochran <richardcochran@gmail.com>
Signed-off-by: Jinjie Ruan <ruanjinjie@huawei.com>
Acked-by: Anna-Maria Behnsen <anna-maria@linutronix.de>
[pabeni@redhat.com: fixed commit message typo]
Signed-off-by: Paolo Abeni <pabeni@redhat.com>
Signed-off-by: Sasha Levin <sashal@kernel.org>
commit d8794ac20a upstream.
As Andrew pointed out, it will make sense that the PTP core
checked timespec64 struct's tv_sec and tv_nsec range before calling
ptp->info->settime64().
As the man manual of clock_settime() said, if tp.tv_sec is negative or
tp.tv_nsec is outside the range [0..999,999,999], it should return EINVAL,
which include dynamic clocks which handles PTP clock, and the condition is
consistent with timespec64_valid(). As Thomas suggested, timespec64_valid()
only check the timespec is valid, but not ensure that the time is
in a valid range, so check it ahead using timespec64_valid_strict()
in pc_clock_settime() and return -EINVAL if not valid.
There are some drivers that use tp->tv_sec and tp->tv_nsec directly to
write registers without validity checks and assume that the higher layer
has checked it, which is dangerous and will benefit from this, such as
hclge_ptp_settime(), igb_ptp_settime_i210(), _rcar_gen4_ptp_settime(),
and some drivers can remove the checks of itself.
Cc: stable@vger.kernel.org
Fixes: 0606f422b4 ("posix clocks: Introduce dynamic clocks")
Acked-by: Richard Cochran <richardcochran@gmail.com>
Suggested-by: Andrew Lunn <andrew@lunn.ch>
Suggested-by: Thomas Gleixner <tglx@linutronix.de>
Signed-off-by: Jinjie Ruan <ruanjinjie@huawei.com>
Link: https://patch.msgid.link/20241009072302.1754567-2-ruanjinjie@huawei.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
[ Upstream commit 5a830bbce3 ]
The hrtimer function callback must not be NULL. It has to be specified by
the call side but it is not validated by the hrtimer code. When a hrtimer
is queued without a function callback, the kernel crashes with a null
pointer dereference when trying to execute the callback in __run_hrtimer().
Introduce a validation before queuing the hrtimer in
hrtimer_start_range_ns().
[anna-maria: Rephrase commit message]
Signed-off-by: Phil Chang <phil.chang@mediatek.com>
Signed-off-by: Anna-Maria Behnsen <anna-maria@linutronix.de>
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
Reviewed-by: Anna-Maria Behnsen <anna-maria@linutronix.de>
Signed-off-by: Sasha Levin <sashal@kernel.org>
[ Upstream commit d0304569fb ]
Kernel timekeeping is designed to keep the change in cycles (since the last
timer interrupt) below max_cycles, which prevents multiplication overflow
when converting cycles to nanoseconds. However, if timer interrupts stop,
the clocksource_cyc2ns() calculation will eventually overflow.
Add protection against that. Simplify by folding together
clocksource_delta() and clocksource_cyc2ns() into cycles_to_nsec_safe().
Check against max_cycles, falling back to a slower higher precision
calculation.
Suggested-by: Thomas Gleixner <tglx@linutronix.de>
Signed-off-by: Adrian Hunter <adrian.hunter@intel.com>
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
Link: https://lore.kernel.org/r/20240325064023.2997-20-adrian.hunter@intel.com
Signed-off-by: Sasha Levin <sashal@kernel.org>
[ Upstream commit 3ce74f1a85 ]
tick_nohz_idle_got_tick() is called by cpuidle_reflect() within the idle
loop with interrupts enabled. This function modifies the struct
tick_sched's bitfield "got_idle_tick". However this bitfield is stored
within the same mask as other bitfields that can be modified from
interrupts.
Fortunately so far it looks like the only race that can happen is while
writing ->got_idle_tick to 0, an interrupt fires and writes the
->idle_active field to 0. It's then possible that the interrupted write
to ->got_idle_tick writes back the old value of ->idle_active back to 1.
However if that happens, the worst possible outcome is that the time
spent between that interrupt and the upcoming call to
tick_nohz_idle_exit() is accounted as idle, which is negligible quantity.
Still all the bitfield writes within this struct tick_sched's shadow
mask should be IRQ-safe. Therefore move this bitfield out to its own
storage to avoid further suprises.
Signed-off-by: Frederic Weisbecker <frederic@kernel.org>
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
Reviewed-by: Thomas Gleixner <tglx@linutronix.de>
Link: https://lore.kernel.org/r/20240225225508.11587-12-frederic@kernel.org
Signed-off-by: Sasha Levin <sashal@kernel.org>
commit 5916be8a53 upstream.
The addition of the bases argument to clock_was_set() fixed up all call
sites correctly except for do_adjtimex(). This uses CLOCK_REALTIME
instead of CLOCK_SET_WALL as argument. CLOCK_REALTIME is 0.
As a result the effect of that clock_was_set() notification is incomplete
and might result in timers expiring late because the hrtimer code does
not re-evaluate the affected clock bases.
Use CLOCK_SET_WALL instead of CLOCK_REALTIME to tell the hrtimers code
which clock bases need to be re-evaluated.
Fixes: 17a1b8826b ("hrtimer: Add bases argument to clock_was_set()")
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
Cc: stable@vger.kernel.org
Link: https://lore.kernel.org/all/877ccx7igo.ffs@tglx
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
commit 06c03c8edc upstream.
Using syzkaller with the recently reintroduced signed integer overflow
sanitizer produces this UBSAN report:
UBSAN: signed-integer-overflow in ../kernel/time/ntp.c:738:18
9223372036854775806 + 4 cannot be represented in type 'long'
Call Trace:
handle_overflow+0x171/0x1b0
__do_adjtimex+0x1236/0x1440
do_adjtimex+0x2be/0x740
The user supplied time_constant value is incremented by four and then
clamped to the operating range.
Before commit eea83d896e ("ntp: NTP4 user space bits update") the user
supplied value was sanity checked to be in the operating range. That change
removed the sanity check and relied on clamping after incrementing which
does not work correctly when the user supplied value is in the overflow
zone of the '+ 4' operation.
The operation requires CAP_SYS_TIME and the side effect of the overflow is
NTP getting out of sync.
Similar to the fixups for time_maxerror and time_esterror, clamp the user
space supplied value to the operating range.
[ tglx: Switch to clamping ]
Fixes: eea83d896e ("ntp: NTP4 user space bits update")
Signed-off-by: Justin Stitt <justinstitt@google.com>
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
Cc: Miroslav Lichvar <mlichvar@redhat.com>
Cc: stable@vger.kernel.org
Link: https://lore.kernel.org/all/20240517-b4-sio-ntp-c-v2-1-f3a80096f36f@google.com
Closes: https://github.com/KSPP/linux/issues/352
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
[ Upstream commit f2655ac2c0 ]
The current "nretries > 1 || nretries >= max_retries" check in
cs_watchdog_read() will always evaluate to true, and thus pr_warn(), if
nretries is greater than 1. The intent is instead to never warn on the
first try, but otherwise warn if the successful retry was the last retry.
Therefore, change that "||" to "&&".
Fixes: db3a34e174 ("clocksource: Retry clock read if long delays detected")
Reported-by: Borislav Petkov <bp@alien8.de>
Signed-off-by: Paul E. McKenney <paulmck@kernel.org>
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
Cc: stable@vger.kernel.org
Link: https://lore.kernel.org/all/20240802154618.4149953-2-paulmck@kernel.org
Signed-off-by: Sasha Levin <sashal@kernel.org>
[ Upstream commit 2ed08e4bc5 ]
On a 8-socket server the TSC is wrongly marked as 'unstable' and disabled
during boot time on about one out of 120 boot attempts:
clocksource: timekeeping watchdog on CPU227: wd-tsc-wd excessive read-back delay of 153560ns vs. limit of 125000ns,
wd-wd read-back delay only 11440ns, attempt 3, marking tsc unstable
tsc: Marking TSC unstable due to clocksource watchdog
TSC found unstable after boot, most likely due to broken BIOS. Use 'tsc=unstable'.
sched_clock: Marking unstable (119294969739, 159204297)<-(125446229205, -5992055152)
clocksource: Checking clocksource tsc synchronization from CPU 319 to CPUs 0,99,136,180,210,542,601,896.
clocksource: Switched to clocksource hpet
The reason is that for platform with a large number of CPUs, there are
sporadic big or huge read latencies while reading the watchog/clocksource
during boot or when system is under stress work load, and the frequency and
maximum value of the latency goes up with the number of online CPUs.
The cCurrent code already has logic to detect and filter such high latency
case by reading the watchdog twice and checking the two deltas. Due to the
randomness of the latency, there is a low probabilty that the first delta
(latency) is big, but the second delta is small and looks valid. The
watchdog code retries the readouts by default twice, which is not
necessarily sufficient for systems with a large number of CPUs.
There is a command line parameter 'max_cswd_read_retries' which allows to
increase the number of retries, but that's not user friendly as it needs to
be tweaked per system. As the number of required retries is proportional to
the number of online CPUs, this parameter can be calculated at runtime.
Scale and enlarge the number of retries according to the number of online
CPUs and remove the command line parameter completely.
[ tglx: Massaged change log and comments ]
Signed-off-by: Feng Tang <feng.tang@intel.com>
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
Tested-by: Jin Wang <jin1.wang@intel.com>
Tested-by: Paul E. McKenney <paulmck@kernel.org>
Reviewed-by: Waiman Long <longman@redhat.com>
Reviewed-by: Paul E. McKenney <paulmck@kernel.org>
Link: https://lore.kernel.org/r/20240221060859.1027450-1-feng.tang@intel.com
Stable-dep-of: f2655ac2c0 ("clocksource: Fix brown-bag boolean thinko in cs_watchdog_read()")
Signed-off-by: Sasha Levin <sashal@kernel.org>
[ Upstream commit 87d571d6fb ]
Using syzkaller alongside the newly reintroduced signed integer overflow
sanitizer spits out this report:
UBSAN: signed-integer-overflow in ../kernel/time/ntp.c:461:16
9223372036854775807 + 500 cannot be represented in type 'long'
Call Trace:
handle_overflow+0x171/0x1b0
second_overflow+0x2d6/0x500
accumulate_nsecs_to_secs+0x60/0x160
timekeeping_advance+0x1fe/0x890
update_wall_time+0x10/0x30
time_maxerror is unconditionally incremented and the result is checked
against NTP_PHASE_LIMIT, but the increment itself can overflow, resulting
in wrap-around to negative space.
Before commit eea83d896e ("ntp: NTP4 user space bits update") the user
supplied value was sanity checked to be in the operating range. That change
removed the sanity check and relied on clamping in handle_overflow() which
does not work correctly when the user supplied value is in the overflow
zone of the '+ 500' operation.
The operation requires CAP_SYS_TIME and the side effect of the overflow is
NTP getting out of sync.
Miroslav confirmed that the input value should be clamped to the operating
range and the same applies to time_esterror. The latter is not used by the
kernel, but the value still should be in the operating range as it was
before the sanity check got removed.
Clamp them to the operating range.
[ tglx: Changed it to clamping and included time_esterror ]
Fixes: eea83d896e ("ntp: NTP4 user space bits update")
Signed-off-by: Justin Stitt <justinstitt@google.com>
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
Cc: Miroslav Lichvar <mlichvar@redhat.com>
Link: https://lore.kernel.org/all/20240517-b4-sio-ntp-usec-v2-1-d539180f2b79@google.com
Closes: https://github.com/KSPP/linux/issues/354
Signed-off-by: Sasha Levin <sashal@kernel.org>
commit 6881e75237 upstream.
The recent fix for making the take over of the broadcast timer more
reliable retrieves a per CPU pointer in preemptible context.
This went unnoticed as compilers hoist the access into the non-preemptible
region where the pointer is actually used. But of course it's valid that
the compiler keeps it at the place where the code puts it which rightfully
triggers:
BUG: using smp_processor_id() in preemptible [00000000] code:
caller is hotplug_cpu__broadcast_tick_pull+0x1c/0xc0
Move it to the actual usage site which is in a non-preemptible region.
Fixes: f7d43dd206 ("tick/broadcast: Make takeover of broadcast hrtimer reliable")
Reported-by: David Wang <00107082@163.com>
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
Tested-by: Yu Liao <liaoyu15@huawei.com>
Cc: stable@vger.kernel.org
Link: https://lore.kernel.org/all/87ttg56ers.ffs@tglx
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
commit f7d43dd206 upstream.
Running the LTP hotplug stress test on a aarch64 machine results in
rcu_sched stall warnings when the broadcast hrtimer was owned by the
un-plugged CPU. The issue is the following:
CPU1 (owns the broadcast hrtimer) CPU2
tick_broadcast_enter()
// shutdown local timer device
broadcast_shutdown_local()
...
tick_broadcast_exit()
clockevents_switch_state(dev, CLOCK_EVT_STATE_ONESHOT)
// timer device is not programmed
cpumask_set_cpu(cpu, tick_broadcast_force_mask)
initiates offlining of CPU1
take_cpu_down()
/*
* CPU1 shuts down and does not
* send broadcast IPI anymore
*/
takedown_cpu()
hotplug_cpu__broadcast_tick_pull()
// move broadcast hrtimer to this CPU
clockevents_program_event()
bc_set_next()
hrtimer_start()
/*
* timer device is not programmed
* because only the first expiring
* timer will trigger clockevent
* device reprogramming
*/
What happens is that CPU2 exits broadcast mode with force bit set, then the
local timer device is not reprogrammed and CPU2 expects to receive the
expired event by the broadcast IPI. But this does not happen because CPU1
is offlined by CPU2. CPU switches the clockevent device to ONESHOT state,
but does not reprogram the device.
The subsequent reprogramming of the hrtimer broadcast device does not
program the clockevent device of CPU2 either because the pending expiry
time is already in the past and the CPU expects the event to be delivered.
As a consequence all CPUs which wait for a broadcast event to be delivered
are stuck forever.
Fix this issue by reprogramming the local timer device if the broadcast
force bit of the CPU is set so that the broadcast hrtimer is delivered.
[ tglx: Massage comment and change log. Add Fixes tag ]
Fixes: 989dcb645c ("tick: Handle broadcast wakeup of multiple cpus")
Signed-off-by: Yu Liao <liaoyu15@huawei.com>
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
Cc: stable@vger.kernel.org
Link: https://lore.kernel.org/r/20240711124843.64167-1-liaoyu15@huawei.com
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
[ Upstream commit 14274d0bd3 ]
So far, get_device_system_crosststamp() unconditionally passes
system_counterval.cycles to timekeeping_cycles_to_ns(). But when
interpolating system time (do_interp == true), system_counterval.cycles is
before tkr_mono.cycle_last, contrary to the timekeeping_cycles_to_ns()
expectations.
On x86, CONFIG_CLOCKSOURCE_VALIDATE_LAST_CYCLE will mitigate on
interpolating, setting delta to 0. With delta == 0, xtstamp->sys_monoraw
and xtstamp->sys_realtime are then set to the last update time, as
implicitly expected by adjust_historical_crosststamp(). On other
architectures, the resulting nonsense xtstamp->sys_monoraw and
xtstamp->sys_realtime corrupt the xtstamp (ts) adjustment in
adjust_historical_crosststamp().
Fix this by deriving xtstamp->sys_monoraw and xtstamp->sys_realtime from
the last update time when interpolating, by using the local variable
"cycles". The local variable already has the right value when
interpolating, unlike system_counterval.cycles.
Fixes: 2c756feb18 ("time: Add history to cross timestamp interface supporting slower devices")
Signed-off-by: Peter Hilber <peter.hilber@opensynergy.com>
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
Acked-by: John Stultz <jstultz@google.com>
Link: https://lore.kernel.org/r/20231218073849.35294-4-peter.hilber@opensynergy.com
Signed-off-by: Sasha Levin <sashal@kernel.org>
[ Upstream commit 87a4113088 ]
The cycle_between() helper checks if parameter test is in the open interval
(before, after). Colloquially speaking, this also applies to the counter
wrap-around special case before > after. get_device_system_crosststamp()
currently uses cycle_between() at the first call site to decide whether to
interpolate for older counter readings.
get_device_system_crosststamp() has the following problem with
cycle_between() testing against an open interval: Assume that, by chance,
cycles == tk->tkr_mono.cycle_last (in the following, "cycle_last" for
brevity). Then, cycle_between() at the first call site, with effective
argument values cycle_between(cycle_last, cycles, now), returns false,
enabling interpolation. During interpolation,
get_device_system_crosststamp() will then call cycle_between() at the
second call site (if a history_begin was supplied). The effective argument
values are cycle_between(history_begin->cycles, cycles, cycles), since
system_counterval.cycles == interval_start == cycles, per the assumption.
Due to the test against the open interval, cycle_between() returns false
again. This causes get_device_system_crosststamp() to return -EINVAL.
This failure should be avoided, since get_device_system_crosststamp() works
both when cycles follows cycle_last (no interpolation), and when cycles
precedes cycle_last (interpolation). For the case cycles == cycle_last,
interpolation is actually unneeded.
Fix this by changing cycle_between() into timestamp_in_interval(), which
now checks against the closed interval, rather than the open interval.
This changes the get_device_system_crosststamp() behavior for three corner
cases:
1. Bypass interpolation in the case cycles == tk->tkr_mono.cycle_last,
fixing the problem described above.
2. At the first timestamp_in_interval() call site, cycles == now no longer
causes failure.
3. At the second timestamp_in_interval() call site, history_begin->cycles
== system_counterval.cycles no longer causes failure.
adjust_historical_crosststamp() also works for this corner case,
where partial_history_cycles == total_history_cycles.
These behavioral changes should not cause any problems.
Fixes: 2c756feb18 ("time: Add history to cross timestamp interface supporting slower devices")
Signed-off-by: Peter Hilber <peter.hilber@opensynergy.com>
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
Link: https://lore.kernel.org/r/20231218073849.35294-3-peter.hilber@opensynergy.com
Signed-off-by: Sasha Levin <sashal@kernel.org>
[ Upstream commit 133e267ef4 ]
'days' is a s64 (from div_s64), and so should use a %lld specifier.
This was found by extending KUnit's assertion macros to use gcc's
__printf attribute.
Fixes: 2760105516 ("time: Improve performance of time64_to_tm()")
Signed-off-by: David Gow <davidgow@google.com>
Tested-by: Guenter Roeck <linux@roeck-us.net>
Reviewed-by: Justin Stitt <justinstitt@google.com>
Signed-off-by: Shuah Khan <skhan@linuxfoundation.org>
Signed-off-by: Sasha Levin <sashal@kernel.org>
commit dad6a09f31 upstream.
The hrtimers migration on CPU-down hotplug process has been moved
earlier, before the CPU actually goes to die. This leaves a small window
of opportunity to queue an hrtimer in a blind spot, leaving it ignored.
For example a practical case has been reported with RCU waking up a
SCHED_FIFO task right before the CPUHP_AP_IDLE_DEAD stage, queuing that
way a sched/rt timer to the local offline CPU.
Make sure such situations never go unnoticed and warn when that happens.
Fixes: 5c0930ccaa ("hrtimers: Push pending hrtimers away from outgoing CPU earlier")
Reported-by: Paul E. McKenney <paulmck@kernel.org>
Signed-off-by: Frederic Weisbecker <frederic@kernel.org>
Signed-off-by: Paul E. McKenney <paulmck@kernel.org>
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
Cc: stable@vger.kernel.org
Link: https://lore.kernel.org/r/20240129235646.3171983-4-boqun.feng@gmail.com
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
commit 6446495535 upstream.
There have been reports of the watchdog marking clocksources unstable on
machines with 8 NUMA nodes:
clocksource: timekeeping watchdog on CPU373:
Marking clocksource 'tsc' as unstable because the skew is too large:
clocksource: 'hpet' wd_nsec: 14523447520
clocksource: 'tsc' cs_nsec: 14524115132
The measured clocksource skew - the absolute difference between cs_nsec
and wd_nsec - was 668 microseconds:
cs_nsec - wd_nsec = 14524115132 - 14523447520 = 667612
The kernel used 200 microseconds for the uncertainty_margin of both the
clocksource and watchdog, resulting in a threshold of 400 microseconds (the
md variable). Both the cs_nsec and the wd_nsec value indicate that the
readout interval was circa 14.5 seconds. The observed behaviour is that
watchdog checks failed for large readout intervals on 8 NUMA node
machines. This indicates that the size of the skew was directly proportinal
to the length of the readout interval on those machines. The measured
clocksource skew, 668 microseconds, was evaluated against a threshold (the
md variable) that is suited for readout intervals of roughly
WATCHDOG_INTERVAL, i.e. HZ >> 1, which is 0.5 second.
The intention of 2e27e793e2 ("clocksource: Reduce clocksource-skew
threshold") was to tighten the threshold for evaluating skew and set the
lower bound for the uncertainty_margin of clocksources to twice
WATCHDOG_MAX_SKEW. Later in c37e85c135 ("clocksource: Loosen clocksource
watchdog constraints"), the WATCHDOG_MAX_SKEW constant was increased to
125 microseconds to fit the limit of NTP, which is able to use a
clocksource that suffers from up to 500 microseconds of skew per second.
Both the TSC and the HPET use default uncertainty_margin. When the
readout interval gets stretched the default uncertainty_margin is no
longer a suitable lower bound for evaluating skew - it imposes a limit
that is far stricter than the skew with which NTP can deal.
The root causes of the skew being directly proportinal to the length of
the readout interval are:
* the inaccuracy of the shift/mult pairs of clocksources and the watchdog
* the conversion to nanoseconds is imprecise for large readout intervals
Prevent this by skipping the current watchdog check if the readout
interval exceeds 2 * WATCHDOG_INTERVAL. Considering the maximum readout
interval of 2 * WATCHDOG_INTERVAL, the current default uncertainty margin
(of the TSC and HPET) corresponds to a limit on clocksource skew of 250
ppm (microseconds of skew per second). To keep the limit imposed by NTP
(500 microseconds of skew per second) for all possible readout intervals,
the margins would have to be scaled so that the threshold value is
proportional to the length of the actual readout interval.
As for why the readout interval may get stretched: Since the watchdog is
executed in softirq context the expiration of the watchdog timer can get
severely delayed on account of a ksoftirqd thread not getting to run in a
timely manner. Surely, a system with such belated softirq execution is not
working well and the scheduling issue should be looked into but the
clocksource watchdog should be able to deal with it accordingly.
Fixes: 2e27e793e2 ("clocksource: Reduce clocksource-skew threshold")
Suggested-by: Feng Tang <feng.tang@intel.com>
Signed-off-by: Jiri Wiesner <jwiesner@suse.de>
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
Tested-by: Paul E. McKenney <paulmck@kernel.org>
Reviewed-by: Feng Tang <feng.tang@intel.com>
Cc: stable@vger.kernel.org
Link: https://lore.kernel.org/r/20240122172350.GA740@incl
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Clark Williams