Running posix CPU timers in hard interrupt context has a few downsides:
- For PREEMPT_RT it cannot work as the expiry code needs to take
sighand lock, which is a 'sleeping spinlock' in RT. The original RT
approach of offloading the posix CPU timer handling into a high
priority thread was clumsy and provided no real benefit in general.
- For fine grained accounting it's just wrong to run this in context of
the timer interrupt because that way a process specific CPU time is
accounted to the timer interrupt.
- Long running timer interrupts caused by a large amount of expiring
timers which can be created and armed by unpriviledged user space.
There is no hard requirement to expire them in interrupt context.
If the signal is targeted at the task itself then it won't be delivered
before the task returns to user space anyway. If the signal is targeted at
a supervisor process then it might be slightly delayed, but posix CPU
timers are inaccurate anyway due to the fact that they are tied to the
tick.
Provide infrastructure to schedule task work which allows splitting the
posix CPU timer code into a quick check in interrupt context and a thread
context expiry and signal delivery function. This has to be enabled by
architectures as it requires that the architecture specific KVM
implementation handles pending task work before exiting to guest mode.
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
Signed-off-by: Ingo Molnar <mingo@kernel.org>
Reviewed-by: Oleg Nesterov <oleg@redhat.com>
Acked-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Link: https://lore.kernel.org/r/20200730102337.783470146@linutronix.de
Pull timer updates from Thomas Gleixner:
"Time, timers and related driver updates:
- Prevent unnecessary timer softirq invocations by extending the
tracking of the next expiring timer in the timer wheel beyond the
existing NOHZ functionality.
The tracking overhead at enqueue time is within the noise, but on
sensitive workloads the avoidance of the soft interrupt invocation
is a measurable improvement.
- The obligatory new clocksource driver for Ingenic X100 OST
- The usual fixes, improvements, cleanups and extensions for newer
chip variants all over the driver space"
* tag 'timers-core-2020-08-04' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip: (28 commits)
timers: Recalculate next timer interrupt only when necessary
clocksource/drivers/ingenic: Add support for the Ingenic X1000 OST.
dt-bindings: timer: Add Ingenic X1000 OST bindings.
clocksource/drivers: Replace HTTP links with HTTPS ones
clocksource/drivers/nomadik-mtu: Handle 32kHz clock
clocksource/drivers/sh_cmt: Use "kHz" for kilohertz
clocksource/drivers/imx: Add support for i.MX TPM driver with ARM64
clocksource/drivers/ingenic: Add high resolution timer support for SMP/SMT.
timers: Lower base clock forwarding threshold
timers: Remove must_forward_clk
timers: Spare timer softirq until next expiry
timers: Expand clk forward logic beyond nohz
timers: Reuse next expiry cache after nohz exit
timers: Always keep track of next expiry
timers: Optimize _next_timer_interrupt() level iteration
timers: Add comments about calc_index() ceiling work
timers: Move trigger_dyntick_cpu() to enqueue_timer()
timers: Use only bucket expiry for base->next_expiry value
timers: Preserve higher bits of expiration on index calculation
clocksource/drivers/timer-atmel-tcb: Add sama5d2 support
...
This modifies the first 32 bits out of the 128 bits of a random CPU's
net_rand_state on interrupt or CPU activity to complicate remote
observations that could lead to guessing the network RNG's internal
state.
Note that depending on some network devices' interrupt rate moderation
or binding, this re-seeding might happen on every packet or even almost
never.
In addition, with NOHZ some CPUs might not even get timer interrupts,
leaving their local state rarely updated, while they are running
networked processes making use of the random state. For this reason, we
also perform this update in update_process_times() in order to at least
update the state when there is user or system activity, since it's the
only case we care about.
Reported-by: Amit Klein <aksecurity@gmail.com>
Suggested-by: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Eric Dumazet <edumazet@google.com>
Cc: "Jason A. Donenfeld" <Jason@zx2c4.com>
Cc: Andy Lutomirski <luto@kernel.org>
Cc: Kees Cook <keescook@chromium.org>
Cc: Thomas Gleixner <tglx@linutronix.de>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: <stable@vger.kernel.org>
Signed-off-by: Willy Tarreau <w@1wt.eu>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
The nohz tick code recalculates the timer wheel's next expiry on each idle
loop iteration.
On the other hand, the base next expiry is now always cached and updated
upon timer enqueue and execution. Only timer dequeue may leave
base->next_expiry out of date (but then its stale value won't ever go past
the actual next expiry to be recalculated).
Since recalculating the next_expiry isn't a free operation, especially when
the last wheel level is reached to find out that no timer has been enqueued
at all, reuse the next expiry cache when it is known to be reliable, which
it is most of the time.
Signed-off-by: Frederic Weisbecker <frederic@kernel.org>
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
Link: https://lkml.kernel.org/r/20200723151641.12236-1-frederic@kernel.org
Now that the core timer infrastructure doesn't depend anymore on
periodic base->clk increments, even when the CPU is not in NO_HZ mode,
timer softirqs can be skipped until there are timers to expire.
Some spurious softirqs can still remain since base->next_expiry doesn't
keep track of canceled timers but this still reduces the number of softirqs
significantly: ~15 times less for HZ=1000 and ~5 times less for HZ=100.
Signed-off-by: Frederic Weisbecker <frederic@kernel.org>
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
Tested-by: Juri Lelli <juri.lelli@redhat.com>
Link: https://lkml.kernel.org/r/20200717140551.29076-11-frederic@kernel.org
As for next_expiry, the base->clk catch up logic will be expanded beyond
NOHZ in order to avoid triggering useless softirqs.
If softirqs should only fire to expire pending timers, periodic base->clk
increments must be skippable for random amounts of time. Therefore prepare
to catch-up with missing updates whenever an up-to-date base clock is
needed.
Signed-off-by: Frederic Weisbecker <frederic@kernel.org>
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
Tested-by: Juri Lelli <juri.lelli@redhat.com>
Link: https://lkml.kernel.org/r/20200717140551.29076-10-frederic@kernel.org
If a level has a timer that expires before reaching the next level, there
is no need to iterate further.
The next level is reached when the 3 lower bits of the current level are
cleared. If the next event happens before/during that, the next levels
won't provide an earlier expiration.
Signed-off-by: Frederic Weisbecker <frederic@kernel.org>
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
Tested-by: Juri Lelli <juri.lelli@redhat.com>
Link: https://lkml.kernel.org/r/20200717140551.29076-7-frederic@kernel.org
The bucket expiry time is the effective expriy time of timers and is
greater than or equal to the requested timer expiry time. This is due
to the guarantee that timers never expire early and the reduced expiry
granularity in the secondary wheel levels.
When a timer is enqueued, trigger_dyntick_cpu() checks whether the
timer is the new first timer. This check compares next_expiry with
the requested timer expiry value and not with the effective expiry
value of the bucket into which the timer was queued.
Storing the requested timer expiry value in base->next_expiry can lead
to base->clk going backwards if the requested timer expiry value is
smaller than base->clk. Commit 30c66fc30e ("timer: Prevent base->clk
from moving backward") worked around this by preventing the store when
timer->expiry is before base->clk, but did not fix the underlying
problem.
Use the expiry value of the bucket into which the timer is queued to
do the new first timer check. This fixes the base->clk going backward
problem.
The workaround of commit 30c66fc30e ("timer: Prevent base->clk from
moving backward") in trigger_dyntick_cpu() is not longer necessary as the
timers bucket expiry is guaranteed to be greater than or equal base->clk.
Signed-off-by: Anna-Maria Behnsen <anna-maria@linutronix.de>
Signed-off-by: Frederic Weisbecker <frederic@kernel.org>
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
Link: https://lkml.kernel.org/r/20200717140551.29076-4-frederic@kernel.org
The higher bits of the timer expiration are cropped while calling
calc_index() due to the implicit cast from unsigned long to unsigned int.
This loss shouldn't have consequences on the current code since all the
computation to calculate the index is done on the lower 32 bits.
However to prepare for returning the actual bucket expiration from
calc_index() in order to properly fix base->next_expiry updates, the higher
bits need to be preserved.
Signed-off-by: Frederic Weisbecker <frederic@kernel.org>
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
Link: https://lkml.kernel.org/r/20200717140551.29076-3-frederic@kernel.org
When an expiration delta falls into the last level of the wheel, that delta
has be compared against the maximum possible delay and reduced to fit in if
necessary.
However instead of comparing the delta against the maximum, the code
compares the actual expiry against the maximum. Then instead of fixing the
delta to fit in, it sets the maximum delta as the expiry value.
This can result in various undesired outcomes, the worst possible one
being a timer expiring 15 days ahead to fire immediately.
Fixes: 500462a9de ("timers: Switch to a non-cascading wheel")
Signed-off-by: Frederic Weisbecker <frederic@kernel.org>
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
Cc: stable@vger.kernel.org
Link: https://lkml.kernel.org/r/20200717140551.29076-2-frederic@kernel.org
When a timer is enqueued with a negative delta (ie: expiry is below
base->clk), it gets added to the wheel as expiring now (base->clk).
Yet the value that gets stored in base->next_expiry, while calling
trigger_dyntick_cpu(), is the initial timer->expires value. The
resulting state becomes:
base->next_expiry < base->clk
On the next timer enqueue, forward_timer_base() may accidentally
rewind base->clk. As a possible outcome, timers may expire way too
early, the worst case being that the highest wheel levels get spuriously
processed again.
To prevent from that, make sure that base->next_expiry doesn't get below
base->clk.
Fixes: a683f390b9 ("timers: Forward the wheel clock whenever possible")
Signed-off-by: Frederic Weisbecker <frederic@kernel.org>
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
Reviewed-by: Anna-Maria Behnsen <anna-maria@linutronix.de>
Tested-by: Juri Lelli <juri.lelli@redhat.com>
Cc: stable@vger.kernel.org
Link: https://lkml.kernel.org/r/20200703010657.2302-1-frederic@kernel.org
Instead of having all the sysctl handlers deal with user pointers, which
is rather hairy in terms of the BPF interaction, copy the input to and
from userspace in common code. This also means that the strings are
always NUL-terminated by the common code, making the API a little bit
safer.
As most handler just pass through the data to one of the common handlers
a lot of the changes are mechnical.
Signed-off-by: Christoph Hellwig <hch@lst.de>
Acked-by: Andrey Ignatov <rdna@fb.com>
Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
Pull timekeeping and timer updates from Thomas Gleixner:
"Core:
- Consolidation of the vDSO build infrastructure to address the
difficulties of cross-builds for ARM64 compat vDSO libraries by
restricting the exposure of header content to the vDSO build.
This is achieved by splitting out header content into separate
headers. which contain only the minimaly required information which
is necessary to build the vDSO. These new headers are included from
the kernel headers and the vDSO specific files.
- Enhancements to the generic vDSO library allowing more fine grained
control over the compiled in code, further reducing architecture
specific storage and preparing for adopting the generic library by
PPC.
- Cleanup and consolidation of the exit related code in posix CPU
timers.
- Small cleanups and enhancements here and there
Drivers:
- The obligatory new drivers: Ingenic JZ47xx and X1000 TCU support
- Correct the clock rate of PIT64b global clock
- setup_irq() cleanup
- Preparation for PWM and suspend support for the TI DM timer
- Expand the fttmr010 driver to support ast2600 systems
- The usual small fixes, enhancements and cleanups all over the
place"
* tag 'timers-core-2020-03-30' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip: (80 commits)
Revert "clocksource/drivers/timer-probe: Avoid creating dead devices"
vdso: Fix clocksource.h macro detection
um: Fix header inclusion
arm64: vdso32: Enable Clang Compilation
lib/vdso: Enable common headers
arm: vdso: Enable arm to use common headers
x86/vdso: Enable x86 to use common headers
mips: vdso: Enable mips to use common headers
arm64: vdso32: Include common headers in the vdso library
arm64: vdso: Include common headers in the vdso library
arm64: Introduce asm/vdso/processor.h
arm64: vdso32: Code clean up
linux/elfnote.h: Replace elf.h with UAPI equivalent
scripts: Fix the inclusion order in modpost
common: Introduce processor.h
linux/ktime.h: Extract common header for vDSO
linux/jiffies.h: Extract common header for vDSO
linux/time64.h: Extract common header for vDSO
linux/time32.h: Extract common header for vDSO
linux/time.h: Extract common header for vDSO
...
The timer_pending() function is mostly used in lockless contexts, so
Without proper annotations, KCSAN might detect a data-race [1].
Using hlist_unhashed_lockless() instead of hand-coding it seems
appropriate (as suggested by Paul E. McKenney).
[1]
BUG: KCSAN: data-race in del_timer / detach_if_pending
write to 0xffff88808697d870 of 8 bytes by task 10 on cpu 0:
__hlist_del include/linux/list.h:764 [inline]
detach_timer kernel/time/timer.c:815 [inline]
detach_if_pending+0xcd/0x2d0 kernel/time/timer.c:832
try_to_del_timer_sync+0x60/0xb0 kernel/time/timer.c:1226
del_timer_sync+0x6b/0xa0 kernel/time/timer.c:1365
schedule_timeout+0x2d2/0x6e0 kernel/time/timer.c:1896
rcu_gp_fqs_loop+0x37c/0x580 kernel/rcu/tree.c:1639
rcu_gp_kthread+0x143/0x230 kernel/rcu/tree.c:1799
kthread+0x1d4/0x200 drivers/block/aoe/aoecmd.c:1253
ret_from_fork+0x1f/0x30 arch/x86/entry/entry_64.S:352
read to 0xffff88808697d870 of 8 bytes by task 12060 on cpu 1:
del_timer+0x3b/0xb0 kernel/time/timer.c:1198
sk_stop_timer+0x25/0x60 net/core/sock.c:2845
inet_csk_clear_xmit_timers+0x69/0xa0 net/ipv4/inet_connection_sock.c:523
tcp_clear_xmit_timers include/net/tcp.h:606 [inline]
tcp_v4_destroy_sock+0xa3/0x3f0 net/ipv4/tcp_ipv4.c:2096
inet_csk_destroy_sock+0xf4/0x250 net/ipv4/inet_connection_sock.c:836
tcp_close+0x6f3/0x970 net/ipv4/tcp.c:2497
inet_release+0x86/0x100 net/ipv4/af_inet.c:427
__sock_release+0x85/0x160 net/socket.c:590
sock_close+0x24/0x30 net/socket.c:1268
__fput+0x1e1/0x520 fs/file_table.c:280
____fput+0x1f/0x30 fs/file_table.c:313
task_work_run+0xf6/0x130 kernel/task_work.c:113
tracehook_notify_resume include/linux/tracehook.h:188 [inline]
exit_to_usermode_loop+0x2b4/0x2c0 arch/x86/entry/common.c:163
Reported by Kernel Concurrency Sanitizer on:
CPU: 1 PID: 12060 Comm: syz-executor.5 Not tainted 5.4.0-rc3+ #0
Hardware name: Google Google Compute Engine/Google Compute Engine,
Signed-off-by: Eric Dumazet <edumazet@google.com>
Cc: Thomas Gleixner <tglx@linutronix.de>
[ paulmck: Pulled in Eric's later amendments. ]
Signed-off-by: Paul E. McKenney <paulmck@kernel.org>
When working commit 6dcd5d7a7a, a mistake was noticed by Linus:
schedule_timeout() was called without setting the task state to anything
particular.
It calls the scheduler, but doesn't delay anything, because the task stays
runnable. That happens because sched_submit_work() does nothing for tasks
in TASK_RUNNING state.
That turned out to be the intended behavior. Adding a WARN() is not useful
as the task could be woken up right after setting the state and before
reaching schedule_timeout().
Improve the comment about schedule_timeout() and describe that more
explicitly.
Signed-off-by: Alexander Popov <alex.popov@linux.com>
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
Link: https://lkml.kernel.org/r/20200117225900.16340-1-alex.popov@linux.com
The timer delayed for more than 3 seconds warning was triggered during
testing.
Workqueue: events_unbound sched_tick_remote
RIP: 0010:sched_tick_remote+0xee/0x100
...
Call Trace:
process_one_work+0x18c/0x3a0
worker_thread+0x30/0x380
kthread+0x113/0x130
ret_from_fork+0x22/0x40
The reason is that the code in collect_expired_timers() uses jiffies
unprotected:
if (next_event > jiffies)
base->clk = jiffies;
As the compiler is allowed to reload the value base->clk can advance
between the check and the store and in the worst case advance farther than
next event. That causes the timer expiry to be delayed until the wheel
pointer wraps around.
Convert the code to use READ_ONCE()
Fixes: 236968383c ("timers: Optimize collect_expired_timers() for NOHZ")
Signed-off-by: Li RongQing <lirongqing@baidu.com>
Signed-off-by: Liang ZhiCheng <liangzhicheng@baidu.com>
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
Cc: stable@vger.kernel.org
Link: https://lkml.kernel.org/r/1568894687-14499-1-git-send-email-lirongqing@baidu.com
When PREEMPT_RT is enabled, the soft interrupt thread can be preempted. If
the soft interrupt thread is preempted in the middle of a timer callback,
then calling del_timer_sync() can lead to two issues:
- If the caller is on a remote CPU then it has to spin wait for the timer
handler to complete. This can result in unbound priority inversion.
- If the caller originates from the task which preempted the timer
handler on the same CPU, then spin waiting for the timer handler to
complete is never going to end.
To avoid these issues, add a new lock to the timer base which is held
around the execution of the timer callbacks. If del_timer_sync() detects
that the timer callback is currently running, it blocks on the expiry
lock. When the callback is finished, the expiry lock is dropped by the
softirq thread which wakes up the waiter and the system makes progress.
This addresses both the priority inversion and the life lock issues.
This mechanism is not used for timers which are marked IRQSAFE as for those
preemption is disabled accross the callback and therefore this situation
cannot happen. The callbacks for such timers need to be individually
audited for RT compliance.
The same issue can happen in virtual machines when the vCPU which runs a
timer callback is scheduled out. If a second vCPU of the same guest calls
del_timer_sync() it will spin wait for the other vCPU to be scheduled back
in. The expiry lock mechanism would avoid that. It'd be trivial to enable
this when paravirt spinlocks are enabled in a guest, but it's not clear
whether this is an actual problem in the wild, so for now it's an RT only
mechanism.
As the softirq thread can be preempted with PREEMPT_RT=y, the SMP variant
of del_timer_sync() needs to be used on UP as well.
[ tglx: Refactored it for mainline ]
Signed-off-by: Anna-Maria Gleixner <anna-maria@linutronix.de>
Signed-off-by: Sebastian Andrzej Siewior <bigeasy@linutronix.de>
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
Acked-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Link: https://lkml.kernel.org/r/20190726185753.832418500@linutronix.de
Pull printk updates from Petr Mladek:
- Allow state reset of printk_once() calls.
- Prevent crashes when dereferencing invalid pointers in vsprintf().
Only the first byte is checked for simplicity.
- Make vsprintf warnings consistent and inlined.
- Treewide conversion of obsolete %pf, %pF to %ps, %pF printf
modifiers.
- Some clean up of vsprintf and test_printf code.
* tag 'printk-for-5.2' of git://git.kernel.org/pub/scm/linux/kernel/git/pmladek/printk:
lib/vsprintf: Make function pointer_string static
vsprintf: Limit the length of inlined error messages
vsprintf: Avoid confusion between invalid address and value
vsprintf: Prevent crash when dereferencing invalid pointers
vsprintf: Consolidate handling of unknown pointer specifiers
vsprintf: Factor out %pO handler as kobject_string()
vsprintf: Factor out %pV handler as va_format()
vsprintf: Factor out %p[iI] handler as ip_addr_string()
vsprintf: Do not check address of well-known strings
vsprintf: Consistent %pK handling for kptr_restrict == 0
vsprintf: Shuffle restricted_pointer()
printk: Tie printk_once / printk_deferred_once into .data.once for reset
treewide: Switch printk users from %pf and %pF to %ps and %pS, respectively
lib/test_printf: Switch to bitmap_zalloc()
