This commit marks the accesses in tree_stall.h so as to both avoid
undesirable compiler optimizations and to keep KCSAN focused on the
accesses of the core algorithm.
Signed-off-by: Paul E. McKenney <paulmck@kernel.org>
The kbuild test project found an oversized stack frame in rcu_gp_kthread()
for some kernel configurations. This oversizing was due to a very large
amount of inlining, which is unnecessary due to the fact that this code
executes infrequently. This commit therefore marks rcu_gp_init() and
rcu_gp_fqs_loop noinline_for_stack to conserve stack space.
Reported-by: kernel test robot <lkp@intel.com>
Tested-by: Rong Chen <rong.a.chen@intel.com>
[ paulmck: noinline_for_stack per Nathan Chancellor. ]
Reviewed-by: Nathan Chancellor <nathan@kernel.org>
Signed-off-by: Paul E. McKenney <paulmck@kernel.org>
Accesses to ->qsmask are normally protected by ->lock, but there is an
exception in the diagnostic code in rcu_check_boost_fail(). This commit
therefore applies data_race() to this access to avoid KCSAN complaining
about the C-language writes protected by ->lock.
Signed-off-by: Paul E. McKenney <paulmck@kernel.org>
This commit marks some interrupt-induced read-side data races in
__srcu_read_lock(), __srcu_read_unlock(), and srcu_torture_stats_print().
Signed-off-by: Paul E. McKenney <paulmck@kernel.org>
Systems with low-bandwidth consoles can have very large printk()
latencies, and on such systems it makes no sense to have the next RCU CPU
stall warning message start output before the prior message completed.
This commit therefore sets the time of the next stall only after the
prints have completed. While printing, the time of the next stall
message is set to ULONG_MAX/2 jiffies into the future.
Reviewed-by: Sergey Senozhatsky <senozhatsky@chromium.org>
Signed-off-by: Paul E. McKenney <paulmck@kernel.org>
rcu_cpu_stall_reset() is one of the functions virtual CPUs
execute during VM resume in order to handle jiffies skew
that can trigger false positive stall warnings. Paul has
pointed out that this approach is problematic because
rcu_cpu_stall_reset() disables RCU grace period stall-detection
virtually forever, while in fact it can just restart the
stall-detection timeout.
Suggested-by: "Paul E. McKenney" <paulmck@kernel.org>
Signed-off-by: Sergey Senozhatsky <senozhatsky@chromium.org>
Signed-off-by: Signed-off-by: Paul E. McKenney <paulmck@kernel.org>
The soft watchdog timer function checks if a virtual machine
was suspended and hence what looks like a lockup in fact
is a false positive.
This is what kvm_check_and_clear_guest_paused() does: it
tests guest PVCLOCK_GUEST_STOPPED (which is set by the host)
and if it's set then we need to touch all watchdogs and bail
out.
Watchdog timer function runs from IRQ, so PVCLOCK_GUEST_STOPPED
check works fine.
There is, however, one more watchdog that runs from IRQ, so
watchdog timer fn races with it, and that watchdog is not aware
of PVCLOCK_GUEST_STOPPED - RCU stall detector.
apic_timer_interrupt()
smp_apic_timer_interrupt()
hrtimer_interrupt()
__hrtimer_run_queues()
tick_sched_timer()
tick_sched_handle()
update_process_times()
rcu_sched_clock_irq()
This triggers RCU stalls on our devices during VM resume.
If tick_sched_handle()->rcu_sched_clock_irq() runs on a VCPU
before watchdog_timer_fn()->kvm_check_and_clear_guest_paused()
then there is nothing on this VCPU that touches watchdogs and
RCU reads stale gp stall timestamp and new jiffies value, which
makes it think that RCU has stalled.
Make RCU stall watchdog aware of PVCLOCK_GUEST_STOPPED and
don't report RCU stalls when we resume the VM.
Signed-off-by: Sergey Senozhatsky <senozhatsky@chromium.org>
Signed-off-by: Signed-off-by: Paul E. McKenney <paulmck@kernel.org>
KCSAN flags accesses to ->rcu_read_lock_nesting as data races, but
in the past, the overhead of marked accesses was excessive. However,
that was long ago, and much has changed since then, both in terms of
hardware and of compilers. Here is data taken on an eight-core laptop
using Intel(R) Core(TM) i9-10885H CPU @ 2.40GHz with a kernel built
using gcc version 9.3.0, with all data in nanoseconds.
Unmarked accesses (status quo), measured by three refscale runs:
Minimum reader duration: 3.286 2.851 3.395
Median reader duration: 3.698 3.531 3.4695
Maximum reader duration: 4.481 5.215 5.157
Marked accesses, also measured by three refscale runs:
Minimum reader duration: 3.501 3.677 3.580
Median reader duration: 4.053 3.723 3.895
Maximum reader duration: 7.307 4.999 5.511
This focused microbenhmark shows only sub-nanosecond differences which
are unlikely to be visible at the system level. This commit therefore
marks data-racing accesses to ->rcu_read_lock_nesting.
Signed-off-by: Paul E. McKenney <paulmck@kernel.org>
Accesses to the rcu_data structure's ->dynticks field have always been
fully ordered because it was not possible to prove that weaker ordering
was safe. However, with the removal of the rcu_eqs_special_set() function
and the advent of the Linux-kernel memory model, it is now easy to show
that two of the four original full memory barriers can be weakened to
acquire and release operations. The remaining pair must remain full
memory barriers. This change makes the memory ordering requirements
more evident, and it might well also speed up the to-idle and from-idle
fastpaths on some architectures.
The following litmus test, adapted from one supplied off-list by Frederic
Weisbecker, models the RCU grace-period kthread detecting an idle CPU
that is concurrently transitioning to non-idle:
C dynticks-from-idle
{
DYNTICKS=0; (* Initially idle. *)
}
P0(int *X, int *DYNTICKS)
{
int dynticks;
int x;
// Idle.
dynticks = READ_ONCE(*DYNTICKS);
smp_store_release(DYNTICKS, dynticks + 1);
smp_mb();
// Now non-idle
x = READ_ONCE(*X);
}
P1(int *X, int *DYNTICKS)
{
int dynticks;
WRITE_ONCE(*X, 1);
smp_mb();
dynticks = smp_load_acquire(DYNTICKS);
}
exists (1:dynticks=0 /\ 0:x=1)
Running "herd7 -conf linux-kernel.cfg dynticks-from-idle.litmus" verifies
this transition, namely, showing that if the RCU grace-period kthread (P1)
sees another CPU as idle (P0), then any memory access prior to the start
of the grace period (P1's write to X) will be seen by any RCU read-side
critical section following the to-non-idle transition (P0's read from X).
This is a straightforward use of full memory barriers to force ordering
in a store-buffering (SB) litmus test.
The following litmus test, also adapted from the one supplied off-list
by Frederic Weisbecker, models the RCU grace-period kthread detecting
a non-idle CPU that is concurrently transitioning to idle:
C dynticks-into-idle
{
DYNTICKS=1; (* Initially non-idle. *)
}
P0(int *X, int *DYNTICKS)
{
int dynticks;
// Non-idle.
WRITE_ONCE(*X, 1);
dynticks = READ_ONCE(*DYNTICKS);
smp_store_release(DYNTICKS, dynticks + 1);
smp_mb();
// Now idle.
}
P1(int *X, int *DYNTICKS)
{
int x;
int dynticks;
smp_mb();
dynticks = smp_load_acquire(DYNTICKS);
x = READ_ONCE(*X);
}
exists (1:dynticks=2 /\ 1:x=0)
Running "herd7 -conf linux-kernel.cfg dynticks-into-idle.litmus" verifies
this transition, namely, showing that if the RCU grace-period kthread
(P1) sees another CPU as newly idle (P0), then any pre-idle memory access
(P0's write to X) will be seen by any code following the grace period
(P1's read from X). This is a simple release-acquire pair forcing
ordering in a message-passing (MP) litmus test.
Of course, if the grace-period kthread detects the CPU as non-idle,
it will refrain from reporting a quiescent state on behalf of that CPU,
so there are no ordering requirements from the grace-period kthread in
that case. However, other subsystems call rcu_is_idle_cpu() to check
for CPUs being non-idle from an RCU perspective. That case is also
verified by the above litmus tests with the proviso that the sense of
the low-order bit of the DYNTICKS counter be inverted.
Unfortunately, on x86 smp_mb() is as expensive as a cache-local atomic
increment. This commit therefore weakens only the read from ->dynticks.
However, the updates are abstracted into a rcu_dynticks_inc() function
to ease any future changes that might be needed.
[ paulmck: Apply Linus Torvalds feedback. ]
Link: https://lore.kernel.org/lkml/20210721202127.2129660-4-paulmck@kernel.org/
Suggested-by: Linus Torvalds <torvalds@linux-foundation.org>
Acked-by: Mathieu Desnoyers <mathieu.desnoyers@efficios.com>
Signed-off-by: Paul E. McKenney <paulmck@kernel.org>
If rcu_print_task_stall() is invoked on an rcu_node structure that does
not contain any tasks blocking the current grace period, it takes an
early exit that fails to release that rcu_node structure's lock. This
results in a self-deadlock, which is detected by lockdep.
To reproduce this bug:
tools/testing/selftests/rcutorture/bin/kvm.sh --allcpus --duration 3 --trust-make --configs "TREE03" --kconfig "CONFIG_PROVE_LOCKING=y" --bootargs "rcutorture.stall_cpu=30 rcutorture.stall_cpu_block=1 rcutorture.fwd_progress=0 rcutorture.test_boost=0"
This will also result in other complaints, including RCU's scheduler
hook complaining about blocking rather than preemption and an rcutorture
writer stall.
Only a partial RCU CPU stall warning message will be printed because of
the self-deadlock.
This commit therefore releases the lock on the rcu_print_task_stall()
function's early exit path.
Fixes: c583bcb8f5 ("rcu: Don't invoke try_invoke_on_locked_down_task() with irqs disabled")
Tested-by: Qais Yousef <qais.yousef@arm.com>
Signed-off-by: Yanfei Xu <yanfei.xu@windriver.com>
Signed-off-by: Paul E. McKenney <paulmck@kernel.org>
The for loop in rcu_print_task_stall() always omits ts[0], which points
to the first task blocking the stalled grace period. This in turn fails
to count this first task, which means that ndetected will be equal to
zero when all CPUs have passed through their quiescent states and only
one task is blocking the stalled grace period. This zero value for
ndetected will in turn result in an incorrect "All QSes seen" message:
rcu: INFO: rcu_preempt detected stalls on CPUs/tasks:
rcu: Tasks blocked on level-1 rcu_node (CPUs 12-23):
(detected by 15, t=6504 jiffies, g=164777, q=9011209)
rcu: All QSes seen, last rcu_preempt kthread activity 1 (4295252379-4295252378), jiffies_till_next_fqs=1, root ->qsmask 0x2
BUG: sleeping function called from invalid context at include/linux/uaccess.h:156
in_atomic(): 1, irqs_disabled(): 0, non_block: 0, pid: 70613, name: msgstress04
INFO: lockdep is turned off.
Preemption disabled at:
[<ffff8000104031a4>] create_object.isra.0+0x204/0x4b0
CPU: 15 PID: 70613 Comm: msgstress04 Kdump: loaded Not tainted
5.12.2-yoctodev-standard #1
Hardware name: Marvell OcteonTX CN96XX board (DT)
Call trace:
dump_backtrace+0x0/0x2cc
show_stack+0x24/0x30
dump_stack+0x110/0x188
___might_sleep+0x214/0x2d0
__might_sleep+0x7c/0xe0
This commit therefore fixes the loop to include ts[0].
Fixes: c583bcb8f5 ("rcu: Don't invoke try_invoke_on_locked_down_task() with irqs disabled")
Tested-by: Qais Yousef <qais.yousef@arm.com>
Signed-off-by: Yanfei Xu <yanfei.xu@windriver.com>
Signed-off-by: Paul E. McKenney <paulmck@kernel.org>
Pull tracing fixes from Steven Rostedt:
"Fix tracepoint race between static_call and callback data
As callbacks to a tracepoint are paired with the data that is passed
in when the callback is registered to the tracepoint, it must have
that data passed to the callback when the tracepoint is triggered,
else bad things will happen. To keep the two together, they are both
assigned to a tracepoint structure and added to an array. The
tracepoint call site will dereference the structure (via RCU) and call
the callback in that structure along with the data in that structure.
This keeps the callback and data tightly coupled.
Because of the overhead that retpolines have on tracepoint callbacks,
if there's only one callback attached to a tracepoint (a common case),
then it is called via a static call (code modified to do a direct call
instead of an indirect call). But to implement this, the data had to
be decoupled from the callback, as now the callback is implemented via
a direct call from the static call and not an indirect call from the
dereferenced structure.
Note, the static call only calls a callback used when there's a single
callback attached to the tracepoint. If more than one callback is
attached to the same tracepoint, then the static call will call an
iterator function that goes back to dereferencing the structure
keeping the callback and its data tightly coupled again.
Issues can arise when going from 0 callbacks to one, as the static
call is assigned to the callback, and it must take care that the data
passed to it is loaded before the static call calls the callback.
Going from 1 to 2 callbacks is not an issue, as long as the static
call is updated to the iterator before the tracepoint structure array
is updated via RCU. Going from 2 to more or back down to 2 is not an
issue as the iterator can handle all theses cases. But going from 2 to
1, care must be taken as the static call is now calling a callback and
the data that is loaded must be the data for that callback.
Care was taken to ensure the callback and data would be in-sync, but
after a bug was reported, it became clear that not enough was done to
make sure that was the case. These changes address this.
The first change is to compare the old and new data instead of the old
and new callback, as it's the data that can corrupt the callback, even
if the callback is the same (something getting freed).
The next change is to convert these transitions into states, to make
it easier to know when a synchronization is needed, and to perform
those synchronizations. The problem with this patch is that it slows
down disabling all events from under a second, to making it take over
10 seconds to do the same work. But that is addressed in the final
patch.
The final patch uses the RCU state functions to keep track of the RCU
state between the transitions, and only needs to perform the
synchronization if an RCU synchronization hasn't been done already.
This brings the performance of disabling all events back to its
original value. That's because no synchronization is required between
disabling tracepoints but is required when enabling a tracepoint after
its been disabled. If an RCU synchronization happens after the
tracepoint is disabled, and before it is re-enabled, there's no need
to do the synchronization again.
Both the second and third patch have subtle complexities that they are
separated into two patches. But because the second patch causes such a
regression in performance, the third patch adds a "Fixes" tag to the
second patch, such that the two must be backported together and not
just the second patch"
* tag 'trace-v5.14-rc4-2' of git://git.kernel.org/pub/scm/linux/kernel/git/rostedt/linux-trace:
tracepoint: Use rcu get state and cond sync for static call updates
tracepoint: Fix static call function vs data state mismatch
tracepoint: static call: Compare data on transition from 2->1 callees
State transitions from 1->0->1 and N->2->1 callbacks require RCU
synchronization. Rather than performing the RCU synchronization every
time the state change occurs, which is quite slow when many tracepoints
are registered in batch, instead keep a snapshot of the RCU state on the
most recent transitions which belong to a chain, and conditionally wait
for a grace period on the last transition of the chain if one g.p. has
not elapsed since the last snapshot.
This applies to both RCU and SRCU.
This brings the performance regression caused by commit 231264d692
("Fix: tracepoint: static call function vs data state mismatch") back to
what it was originally.
Before this commit:
# trace-cmd start -e all
# time trace-cmd start -p nop
real 0m10.593s
user 0m0.017s
sys 0m0.259s
After this commit:
# trace-cmd start -e all
# time trace-cmd start -p nop
real 0m0.878s
user 0m0.000s
sys 0m0.103s
Link: https://lkml.kernel.org/r/20210805192954.30688-1-mathieu.desnoyers@efficios.com
Link: https://lore.kernel.org/io-uring/4ebea8f0-58c9-e571-fd30-0ce4f6f09c70@samba.org/
Cc: stable@vger.kernel.org
Cc: Ingo Molnar <mingo@redhat.com>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Andrew Morton <akpm@linux-foundation.org>
Cc: "Paul E. McKenney" <paulmck@kernel.org>
Cc: Stefan Metzmacher <metze@samba.org>
Fixes: 231264d692 ("Fix: tracepoint: static call function vs data state mismatch")
Signed-off-by: Mathieu Desnoyers <mathieu.desnoyers@efficios.com>
Reviewed-by: Paul E. McKenney <paulmck@kernel.org>
Signed-off-by: Steven Rostedt (VMware) <rostedt@goodmis.org>
The Energy Model (EM) provides useful information about device power in
each performance state to other subsystems like: Energy Aware Scheduler
(EAS). The energy calculation in EAS does arithmetic operation based on
the EM em_cpu_energy(). Current implementation of that function uses
em_perf_state::cost as a pre-computed cost coefficient equal to:
cost = power * max_frequency / frequency.
The 'power' is expressed in milli-Watts (or in abstract scale).
There are corner cases when the EAS energy calculation for two Performance
Domains (PDs) return the same value. The EAS compares these values to
choose smaller one. It might happen that this values are equal due to
rounding error. In such scenario, we need better resolution, e.g. 1000
times better. To provide this possibility increase the resolution in the
em_perf_state::cost for 64-bit architectures. The cost of increasing
resolution on 32-bit is pretty high (64-bit division) and is not justified
since there are no new 32bit big.LITTLE EAS systems expected which would
benefit from this higher resolution.
This patch allows to avoid the rounding to milli-Watt errors, which might
occur in EAS energy estimation for each PD. The rounding error is common
for small tasks which have small utilization value.
There are two places in the code where it makes a difference:
1. In the find_energy_efficient_cpu() where we are searching for
best_delta. We might suffer there when two PDs return the same result,
like in the example below.
Scenario:
Low utilized system e.g. ~200 sum_util for PD0 and ~220 for PD1. There
are quite a few small tasks ~10-15 util. These tasks would suffer for
the rounding error. These utilization values are typical when running games
on Android. One of our partners has reported 5..10mA less battery drain
when running with increased resolution.
Some details:
We have two PDs: PD0 (big) and PD1 (little)
Let's compare w/o patch set ('old') and w/ patch set ('new')
We are comparing energy w/ task and w/o task placed in the PDs
a) 'old' w/o patch set, PD0
task_util = 13
cost = 480
sum_util_w/o_task = 215
sum_util_w_task = 228
scale_cpu = 1024
energy_w/o_task = 480 * 215 / 1024 = 100.78 => 100
energy_w_task = 480 * 228 / 1024 = 106.87 => 106
energy_diff = 106 - 100 = 6
(this is equal to 'old' PD1's energy_diff in 'c)')
b) 'new' w/ patch set, PD0
task_util = 13
cost = 480 * 1000 = 480000
sum_util_w/o_task = 215
sum_util_w_task = 228
energy_w/o_task = 480000 * 215 / 1024 = 100781
energy_w_task = 480000 * 228 / 1024 = 106875
energy_diff = 106875 - 100781 = 6094
(this is not equal to 'new' PD1's energy_diff in 'd)')
c) 'old' w/o patch set, PD1
task_util = 13
cost = 160
sum_util_w/o_task = 283
sum_util_w_task = 293
scale_cpu = 355
energy_w/o_task = 160 * 283 / 355 = 127.55 => 127
energy_w_task = 160 * 296 / 355 = 133.41 => 133
energy_diff = 133 - 127 = 6
(this is equal to 'old' PD0's energy_diff in 'a)')
d) 'new' w/ patch set, PD1
task_util = 13
cost = 160 * 1000 = 160000
sum_util_w/o_task = 283
sum_util_w_task = 293
scale_cpu = 355
energy_w/o_task = 160000 * 283 / 355 = 127549
energy_w_task = 160000 * 296 / 355 = 133408
energy_diff = 133408 - 127549 = 5859
(this is not equal to 'new' PD0's energy_diff in 'b)')
2. Difference in the 6% energy margin filter at the end of
find_energy_efficient_cpu(). With this patch the margin comparison also
has better resolution, so it's possible to have better task placement
thanks to that.
Fixes: 27871f7a8a ("PM: Introduce an Energy Model management framework")
Reported-by: CCJ Yeh <CCj.Yeh@mediatek.com>
Reviewed-by: Dietmar Eggemann <dietmar.eggemann@arm.com>
Signed-off-by: Lukasz Luba <lukasz.luba@arm.com>
Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
SCHED_FLAG_KEEP_PARAMS can be passed to sched_setattr to specify that
the call must not touch scheduling parameters (nice or priority). This
is particularly handy for uclamp when used in conjunction with
SCHED_FLAG_KEEP_POLICY as that allows to issue a syscall that only
impacts uclamp values.
However, sched_setattr always checks whether the priorities and nice
values passed in sched_attr are valid first, even if those never get
used down the line. This is useless at best since userspace can
trivially bypass this check to set the uclamp values by specifying low
priorities. However, it is cumbersome to do so as there is no single
expression of this that skips both RT and CFS checks at once. As such,
userspace needs to query the task policy first with e.g. sched_getattr
and then set sched_attr.sched_priority accordingly. This is racy and
slower than a single call.
As the priority and nice checks are useless when SCHED_FLAG_KEEP_PARAMS
is specified, simply inherit them in this case to match the policy
inheritance of SCHED_FLAG_KEEP_POLICY.
Reported-by: Wei Wang <wvw@google.com>
Signed-off-by: Quentin Perret <qperret@google.com>
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Reviewed-by: Dietmar Eggemann <dietmar.eggemann@arm.com>
Reviewed-by: Qais Yousef <qais.yousef@arm.com>
Link: https://lore.kernel.org/r/20210805102154.590709-3-qperret@google.com
The UCLAMP_FLAG_IDLE flag is set on a runqueue when dequeueing the last
uclamp active task (that is, when buckets.tasks reaches 0 for all
buckets) to maintain the last uclamp.max and prevent blocked util from
suddenly becoming visible.
However, there is an asymmetry in how the flag is set and cleared which
can lead to having the flag set whilst there are active tasks on the rq.
Specifically, the flag is cleared in the uclamp_rq_inc() path, which is
called at enqueue time, but set in uclamp_rq_dec_id() which is called
both when dequeueing a task _and_ in the update_uclamp_active() path. As
a result, when both uclamp_rq_{dec,ind}_id() are called from
update_uclamp_active(), the flag ends up being set but not cleared,
hence leaving the runqueue in a broken state.
Fix this by clearing the flag in update_uclamp_active() as well.
Fixes: e496187da7 ("sched/uclamp: Enforce last task's UCLAMP_MAX")
Reported-by: Rick Yiu <rickyiu@google.com>
Signed-off-by: Quentin Perret <qperret@google.com>
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Reviewed-by: Qais Yousef <qais.yousef@arm.com>
Tested-by: Dietmar Eggemann <dietmar.eggemann@arm.com>
Link: https://lore.kernel.org/r/20210805102154.590709-2-qperret@google.com
A missing clock update is causing the following warning:
rq->clock_update_flags < RQCF_ACT_SKIP
WARNING: CPU: 112 PID: 2041 at kernel/sched/sched.h:1453
sub_running_bw.isra.0+0x190/0x1a0
...
CPU: 112 PID: 2041 Comm: sugov:112 Tainted: G W 5.14.0-rc1 #1
Hardware name: WIWYNN Mt.Jade Server System
B81.030Z1.0007/Mt.Jade Motherboard, BIOS 1.6.20210526 (SCP:
1.06.20210526) 2021/05/26
...
Call trace:
sub_running_bw.isra.0+0x190/0x1a0
migrate_task_rq_dl+0xf8/0x1e0
set_task_cpu+0xa8/0x1f0
try_to_wake_up+0x150/0x3d4
wake_up_q+0x64/0xc0
__up_write+0xd0/0x1c0
up_write+0x4c/0x2b0
cppc_set_perf+0x120/0x2d0
cppc_cpufreq_set_target+0xe0/0x1a4 [cppc_cpufreq]
__cpufreq_driver_target+0x74/0x140
sugov_work+0x64/0x80
kthread_worker_fn+0xe0/0x230
kthread+0x138/0x140
ret_from_fork+0x10/0x18
The task causing this is the `cppc_fie` DL task introduced by
commit 1eb5dde674 ("cpufreq: CPPC: Add support for frequency
invariance").
With CONFIG_ACPI_CPPC_CPUFREQ_FIE=y and schedutil cpufreq governor on
slow-switching system (like on this Ampere Altra WIWYNN Mt. Jade Arm
Server):
DL task `curr=sugov:112` lets `p=cppc_fie` migrate and since the latter
is in `non_contending` state, migrate_task_rq_dl() calls
sub_running_bw()->__sub_running_bw()->cpufreq_update_util()->
rq_clock()->assert_clock_updated()
on p.
Fix this by updating the clock for a non_contending task in
migrate_task_rq_dl() before calling sub_running_bw().
Reported-by: Bruno Goncalves <bgoncalv@redhat.com>
Signed-off-by: Dietmar Eggemann <dietmar.eggemann@arm.com>
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Reviewed-by: Daniel Bristot de Oliveira <bristot@kernel.org>
Acked-by: Juri Lelli <juri.lelli@redhat.com>
Link: https://lore.kernel.org/r/20210804135925.3734605-1-dietmar.eggemann@arm.com
Build failure in drivers/net/wwan/mhi_wwan_mbim.c:
add missing parameter (0, assuming we don't want buffer pre-alloc).
Conflict in drivers/net/dsa/sja1105/sja1105_main.c between:
589918df93 ("net: dsa: sja1105: be stateless with FDB entries on SJA1105P/Q/R/S/SJA1110 too")
0fac6aa098 ("net: dsa: sja1105: delete the best_effort_vlan_filtering mode")
Follow the instructions from the commit message of the former commit
- removed the if conditions. When looking at commit 589918df93 ("net:
dsa: sja1105: be stateless with FDB entries on SJA1105P/Q/R/S/SJA1110 too")
note that the mask_iotag fields get removed by the following patch.
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
On a 1->0->1 callbacks transition, there is an issue with the new
callback using the old callback's data.
Considering __DO_TRACE_CALL:
do { \
struct tracepoint_func *it_func_ptr; \
void *__data; \
it_func_ptr = \
rcu_dereference_raw((&__tracepoint_##name)->funcs); \
if (it_func_ptr) { \
__data = (it_func_ptr)->data; \
----> [ delayed here on one CPU (e.g. vcpu preempted by the host) ]
static_call(tp_func_##name)(__data, args); \
} \
} while (0)
It has loaded the tp->funcs of the old callback, so it will try to use the old
data. This can be fixed by adding a RCU sync anywhere in the 1->0->1
transition chain.
On a N->2->1 transition, we need an rcu-sync because you may have a
sequence of 3->2->1 (or 1->2->1) where the element 0 data is unchanged
between 2->1, but was changed from 3->2 (or from 1->2), which may be
observed by the static call. This can be fixed by adding an
unconditional RCU sync in transition 2->1.
Note, this fixes a correctness issue at the cost of adding a tremendous
performance regression to the disabling of tracepoints.
Before this commit:
# trace-cmd start -e all
# time trace-cmd start -p nop
real 0m0.778s
user 0m0.000s
sys 0m0.061s
After this commit:
# trace-cmd start -e all
# time trace-cmd start -p nop
real 0m10.593s
user 0m0.017s
sys 0m0.259s
A follow up fix will introduce a more lightweight scheme based on RCU
get_state and cond_sync, that will return the performance back to what it
was. As both this change and the lightweight versions are complex on their
own, for bisecting any issues that this may cause, they are kept as two
separate changes.
Link: https://lkml.kernel.org/r/20210805132717.23813-3-mathieu.desnoyers@efficios.com
Link: https://lore.kernel.org/io-uring/4ebea8f0-58c9-e571-fd30-0ce4f6f09c70@samba.org/
Cc: stable@vger.kernel.org
Cc: Ingo Molnar <mingo@redhat.com>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Andrew Morton <akpm@linux-foundation.org>
Cc: "Paul E. McKenney" <paulmck@kernel.org>
Cc: Stefan Metzmacher <metze@samba.org>
Fixes: d25e37d89d ("tracepoint: Optimize using static_call()")
Signed-off-by: Mathieu Desnoyers <mathieu.desnoyers@efficios.com>
Signed-off-by: Steven Rostedt (VMware) <rostedt@goodmis.org>
Pull ucounts fix from Eric Biederman:
"Fix a subtle locking versus reference counting bug in the ucount
changes, found by syzbot"
* 'for-v5.14' of git://git.kernel.org/pub/scm/linux/kernel/git/ebiederm/user-namespace:
ucounts: Fix race condition between alloc_ucounts and put_ucounts
Pull tracing fixes from Steven Rostedt:
"Various tracing fixes:
- Fix NULL pointer dereference caused by an error path
- Give histogram calculation fields a size, otherwise it breaks
synthetic creation based on them.
- Reject strings being used for number calculations.
- Fix recordmcount.pl warning on llvm building RISC-V allmodconfig
- Fix the draw_functrace.py script to handle the new trace output
- Fix warning of smp_processor_id() in preemptible code"
* tag 'trace-v5.14-rc4' of git://git.kernel.org/pub/scm/linux/kernel/git/rostedt/linux-trace:
tracing: Quiet smp_processor_id() use in preemptable warning in hwlat
scripts/tracing: fix the bug that can't parse raw_trace_func
scripts/recordmcount.pl: Remove check_objcopy() and $can_use_local
tracing: Reject string operand in the histogram expression
tracing / histogram: Give calculation hist_fields a size
tracing: Fix NULL pointer dereference in start_creating
The hardware latency detector (hwlat) has a mode that it runs one thread
across CPUs. The logic to move from the currently running CPU to the next
one in the list does a smp_processor_id() to find where it currently is.
Unfortunately, it's done with preemption enabled, and this triggers a
warning for using smp_processor_id() in a preempt enabled section.
As it is only using smp_processor_id() to get information on where it
currently is in order to simply move it to the next CPU, it doesn't really
care if it got moved in the mean time. It will simply balance out later if
such a case arises.
Switch smp_processor_id() to raw_smp_processor_id() to quiet that warning.
Link: https://lkml.kernel.org/r/20210804141848.79edadc0@oasis.local.home
Acked-by: Daniel Bristot de Oliveira <bristot@redhat.com>
Fixes: 8fa826b734 ("trace/hwlat: Implement the mode config option")
Signed-off-by: Steven Rostedt (VMware) <rostedt@goodmis.org>
When working on my user space applications, I found a bug in the synthetic
event code where the automated synthetic event field was not matching the
event field calculation it was attached to. Looking deeper into it, it was
because the calculation hist_field was not given a size.
The synthetic event fields are matched to their hist_fields either by
having the field have an identical string type, or if that does not match,
then the size and signed values are used to match the fields.
The problem arose when I tried to match a calculation where the fields
were "unsigned int". My tool created a synthetic event of type "u32". But
it failed to match. The string was:
diff=field1-field2:onmatch(event).trace(synth,$diff)
Adding debugging into the kernel, I found that the size of "diff" was 0.
And since it was given "unsigned int" as a type, the histogram fallback
code used size and signed. The signed matched, but the size of u32 (4) did
not match zero, and the event failed to be created.
This can be worse if the field you want to match is not one of the
acceptable fields for a synthetic event. As event fields can have any type
that is supported in Linux, this can cause an issue. For example, if a
type is an enum. Then there's no way to use that with any calculations.
Have the calculation field simply take on the size of what it is
calculating.
Link: https://lkml.kernel.org/r/20210730171951.59c7743f@oasis.local.home
Cc: Tom Zanussi <zanussi@kernel.org>
Cc: Masami Hiramatsu <mhiramat@kernel.org>
Cc: Namhyung Kim <namhyung@kernel.org>
Cc: Ingo Molnar <mingo@kernel.org>
Cc: Andrew Morton <akpm@linux-foundation.org>
Cc: stable@vger.kernel.org
Fixes: 100719dcef ("tracing: Add simple expression support to hist triggers")
Signed-off-by: Steven Rostedt (VMware) <rostedt@goodmis.org>
Test RTC_FEATURE_ALARM instead of relying on ops->set_alarm to know whether
alarms are available.
Signed-off-by: Alexandre Belloni <alexandre.belloni@bootlin.com>
Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
The functions get_online_cpus() and put_online_cpus() have been
deprecated during the CPU hotplug rework. They map directly to
cpus_read_lock() and cpus_read_unlock().
Replace deprecated CPU-hotplug functions with the official version.
The behavior remains unchanged.
Signed-off-by: Sebastian Andrzej Siewior <bigeasy@linutronix.de>
Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
After select_idle_sibling, p->recent_used_cpu is set to the
new target. However on the next wakeup, prev will be the same as
recent_used_cpu unless the load balancer has moved the task since the
last wakeup. It still works, but is less efficient than it could be.
This patch preserves recent_used_cpu for longer.
The impact on SIS efficiency is tiny so the SIS statistic patches were
used to track the hit rate for using recent_used_cpu. With perf bench
pipe on a 2-socket Cascadelake machine, the hit rate went from 57.14%
to 85.32%. For more intensive wakeup loads like hackbench, the hit rate
is almost negligible but rose from 0.21% to 6.64%. For scaling loads
like tbench, the hit rate goes from almost 0% to 25.42% overall. Broadly
speaking, on tbench, the success rate is much higher for lower thread
counts and drops to almost 0 as the workload scales to towards saturation.
Signed-off-by: Mel Gorman <mgorman@techsingularity.net>
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Link: https://lkml.kernel.org/r/20210804115857.6253-2-mgorman@techsingularity.net
SCHED_FLAG_SUGOV is supposed to be a kernel-only flag that userspace
cannot interact with. However, sched_getattr() currently reports it
in sched_flags if called on a sugov worker even though it is not
actually defined in a UAPI header. To avoid this, make sure to
clean-up the sched_flags field in sched_getattr() before returning to
userspace.
Signed-off-by: Quentin Perret <qperret@google.com>
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Link: https://lore.kernel.org/r/20210727101103.2729607-3-qperret@google.com
It is possible for sched_getattr() to incorrectly report the state of
the reset_on_fork flag when called on a deadline task.
Indeed, if the flag was set on a deadline task using sched_setattr()
with flags (SCHED_FLAG_RESET_ON_FORK | SCHED_FLAG_KEEP_PARAMS), then
p->sched_reset_on_fork will be set, but __setscheduler() will bail out
early, which means that the dl_se->flags will not get updated by
__setscheduler_params()->__setparam_dl(). Consequently, if
sched_getattr() is then called on the task, __getparam_dl() will
override kattr.sched_flags with the now out-of-date copy in dl_se->flags
and report the stale value to userspace.
To fix this, make sure to only copy the flags that are relevant to
sched_deadline to and from the dl_se->flags field.
Signed-off-by: Quentin Perret <qperret@google.com>
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Link: https://lore.kernel.org/r/20210727101103.2729607-2-qperret@google.com
Double enqueues in rt runqueues (list) have been reported while running
a simple test that spawns a number of threads doing a short sleep/run
pattern while being concurrently setscheduled between rt and fair class.
WARNING: CPU: 3 PID: 2825 at kernel/sched/rt.c:1294 enqueue_task_rt+0x355/0x360
CPU: 3 PID: 2825 Comm: setsched__13
RIP: 0010:enqueue_task_rt+0x355/0x360
Call Trace:
__sched_setscheduler+0x581/0x9d0
_sched_setscheduler+0x63/0xa0
do_sched_setscheduler+0xa0/0x150
__x64_sys_sched_setscheduler+0x1a/0x30
do_syscall_64+0x33/0x40
entry_SYSCALL_64_after_hwframe+0x44/0xae
list_add double add: new=ffff9867cb629b40, prev=ffff9867cb629b40,
next=ffff98679fc67ca0.
kernel BUG at lib/list_debug.c:31!
invalid opcode: 0000 [#1] PREEMPT_RT SMP PTI
CPU: 3 PID: 2825 Comm: setsched__13
RIP: 0010:__list_add_valid+0x41/0x50
Call Trace:
enqueue_task_rt+0x291/0x360
__sched_setscheduler+0x581/0x9d0
_sched_setscheduler+0x63/0xa0
do_sched_setscheduler+0xa0/0x150
__x64_sys_sched_setscheduler+0x1a/0x30
do_syscall_64+0x33/0x40
entry_SYSCALL_64_after_hwframe+0x44/0xae
__sched_setscheduler() uses rt_effective_prio() to handle proper queuing
of priority boosted tasks that are setscheduled while being boosted.
rt_effective_prio() is however called twice per each
__sched_setscheduler() call: first directly by __sched_setscheduler()
before dequeuing the task and then by __setscheduler() to actually do
the priority change. If the priority of the pi_top_task is concurrently
being changed however, it might happen that the two calls return
different results. If, for example, the first call returned the same rt
priority the task was running at and the second one a fair priority, the
task won't be removed by the rt list (on_list still set) and then
enqueued in the fair runqueue. When eventually setscheduled back to rt
it will be seen as enqueued already and the WARNING/BUG be issued.
Fix this by calling rt_effective_prio() only once and then reusing the
return value. While at it refactor code as well for clarity. Concurrent
priority inheritance handling is still safe and will eventually converge
to a new state by following the inheritance chain(s).
Fixes: 0782e63bc6 ("sched: Handle priority boosted tasks proper in setscheduler()")
[squashed Peterz changes; added changelog]
Reported-by: Mark Simmons <msimmons@redhat.com>
Signed-off-by: Juri Lelli <juri.lelli@redhat.com>
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Link: https://lkml.kernel.org/r/20210803104501.38333-1-juri.lelli@redhat.com
Implementing live patching on s390 requires each function's prologue to
contain a very special kind of nop, which gcc and clang don't generate.
However, the current code assumes that if CC_USING_NOP_MCOUNT is
defined, then whatever the compiler generates is good enough.
Move the CC_USING_NOP_MCOUNT check into the new ftrace_need_init_nop()
macro, that the architectures can override.
An alternative solution is to disable using -mnop-mcount in the
Makefile, however, this makes the build logic (even) more complicated
and forces the arch-specific code to deal with the useless __fentry__
symbol.
Signed-off-by: Ilya Leoshkevich <iii@linux.ibm.com>
Reviewed-by: Steven Rostedt (VMware) <rostedt@goodmis.org>
Link: https://lore.kernel.org/r/20210728212546.128248-2-iii@linux.ibm.com
Signed-off-by: Heiko Carstens <hca@linux.ibm.com>
Andrii Nakryiko says:
====================
bpf-next 2021-07-30
We've added 64 non-merge commits during the last 15 day(s) which contain
a total of 83 files changed, 5027 insertions(+), 1808 deletions(-).
The main changes are:
1) BTF-guided binary data dumping libbpf API, from Alan.
2) Internal factoring out of libbpf CO-RE relocation logic, from Alexei.
3) Ambient BPF run context and cgroup storage cleanup, from Andrii.
4) Few small API additions for libbpf 1.0 effort, from Evgeniy and Hengqi.
5) bpf_program__attach_kprobe_opts() fixes in libbpf, from Jiri.
6) bpf_{get,set}sockopt() support in BPF iterators, from Martin.
7) BPF map pinning improvements in libbpf, from Martynas.
8) Improved module BTF support in libbpf and bpftool, from Quentin.
9) Bpftool cleanups and documentation improvements, from Quentin.
10) Libbpf improvements for supporting CO-RE on old kernels, from Shuyi.
11) Increased maximum cgroup storage size, from Stanislav.
12) Small fixes and improvements to BPF tests and samples, from various folks.
* https://git.kernel.org/pub/scm/linux/kernel/git/bpf/bpf-next: (64 commits)
tools: bpftool: Complete metrics list in "bpftool prog profile" doc
tools: bpftool: Document and add bash completion for -L, -B options
selftests/bpf: Update bpftool's consistency script for checking options
tools: bpftool: Update and synchronise option list in doc and help msg
tools: bpftool: Complete and synchronise attach or map types
selftests/bpf: Check consistency between bpftool source, doc, completion
tools: bpftool: Slightly ease bash completion updates
unix_bpf: Fix a potential deadlock in unix_dgram_bpf_recvmsg()
libbpf: Add btf__load_vmlinux_btf/btf__load_module_btf
tools: bpftool: Support dumping split BTF by id
libbpf: Add split BTF support for btf__load_from_kernel_by_id()
tools: Replace btf__get_from_id() with btf__load_from_kernel_by_id()
tools: Free BTF objects at various locations
libbpf: Rename btf__get_from_id() as btf__load_from_kernel_by_id()
libbpf: Rename btf__load() as btf__load_into_kernel()
libbpf: Return non-null error on failures in libbpf_find_prog_btf_id()
bpf: Emit better log message if bpf_iter ctx arg btf_id == 0
tools/resolve_btfids: Emit warnings and patch zero id for missing symbols
bpf: Increase supported cgroup storage value size
libbpf: Fix race when pinning maps in parallel
...
====================
Link: https://lore.kernel.org/r/20210730225606.1897330-1-andrii@kernel.org
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Pull networking fixes from Jakub Kicinski:
"Networking fixes for 5.14-rc4, including fixes from bpf, can, WiFi
(mac80211) and netfilter trees.
Current release - regressions:
- mac80211: fix starting aggregation sessions on mesh interfaces
Current release - new code bugs:
- sctp: send pmtu probe only if packet loss in Search Complete state
- bnxt_en: add missing periodic PHC overflow check
- devlink: fix phys_port_name of virtual port and merge error
- hns3: change the method of obtaining default ptp cycle
- can: mcba_usb_start(): add missing urb->transfer_dma initialization
Previous releases - regressions:
- set true network header for ECN decapsulation
- mlx5e: RX, avoid possible data corruption w/ relaxed ordering and
LRO
- phy: re-add check for PHY_BRCM_DIS_TXCRXC_NOENRGY on the BCM54811
PHY
- sctp: fix return value check in __sctp_rcv_asconf_lookup
Previous releases - always broken:
- bpf:
- more spectre corner case fixes, introduce a BPF nospec
instruction for mitigating Spectre v4
- fix OOB read when printing XDP link fdinfo
- sockmap: fix cleanup related races
- mac80211: fix enabling 4-address mode on a sta vif after assoc
- can:
- raw: raw_setsockopt(): fix raw_rcv panic for sock UAF
- j1939: j1939_session_deactivate(): clarify lifetime of session
object, avoid UAF
- fix number of identical memory leaks in USB drivers
- tipc:
- do not blindly write skb_shinfo frags when doing decryption
- fix sleeping in tipc accept routine"
* tag 'net-5.14-rc4' of git://git.kernel.org/pub/scm/linux/kernel/git/netdev/net: (91 commits)
gve: Update MAINTAINERS list
can: esd_usb2: fix memory leak
can: ems_usb: fix memory leak
can: usb_8dev: fix memory leak
can: mcba_usb_start(): add missing urb->transfer_dma initialization
can: hi311x: fix a signedness bug in hi3110_cmd()
MAINTAINERS: add Yasushi SHOJI as reviewer for the Microchip CAN BUS Analyzer Tool driver
bpf: Fix leakage due to insufficient speculative store bypass mitigation
bpf: Introduce BPF nospec instruction for mitigating Spectre v4
sis900: Fix missing pci_disable_device() in probe and remove
net: let flow have same hash in two directions
nfc: nfcsim: fix use after free during module unload
tulip: windbond-840: Fix missing pci_disable_device() in probe and remove
sctp: fix return value check in __sctp_rcv_asconf_lookup
nfc: s3fwrn5: fix undefined parameter values in dev_err()
net/mlx5: Fix mlx5_vport_tbl_attr chain from u16 to u32
net/mlx5e: Fix nullptr in mlx5e_hairpin_get_mdev()
net/mlx5: Unload device upon firmware fatal error
net/mlx5e: Fix page allocation failure for ptp-RQ over SF
net/mlx5e: Fix page allocation failure for trap-RQ over SF
...
The event_trace_add_tracer() can fail. In this case, it leads to a crash
in start_creating with below call stack. Handle the error scenario
properly in trace_array_create_dir.
Call trace:
down_write+0x7c/0x204
start_creating.25017+0x6c/0x194
tracefs_create_file+0xc4/0x2b4
init_tracer_tracefs+0x5c/0x940
trace_array_create_dir+0x58/0xb4
trace_array_create+0x1bc/0x2b8
trace_array_get_by_name+0xdc/0x18c
Link: https://lkml.kernel.org/r/1627651386-21315-1-git-send-email-kamaagra@codeaurora.org
Cc: stable@vger.kernel.org
Fixes: 4114fbfd02 ("tracing: Enable creating new instance early boot")
Signed-off-by: Kamal Agrawal <kamaagra@codeaurora.org>
Signed-off-by: Steven Rostedt (VMware) <rostedt@goodmis.org>
The HW IRQ numbers generated by the PCI MSI layer can be quite large
on a pSeries machine when running under the IBM Hypervisor and they
appear as negative. Use '%lu' instead to show them correctly.
Signed-off-by: Cédric Le Goater <clg@kaod.org>
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
cycles_t has a different type across architectures: unsigned int,
unsinged long, or unsigned long long. Depending on architecture this
will generate this warning:
kernel/kcsan/debugfs.c: In function ‘microbenchmark’:
./include/linux/kern_levels.h:5:25: warning: format ‘%llu’ expects argument of type ‘long long unsigned int’, but argument 3 has type ‘cycles_t’ {aka ‘long unsigned int’} [-Wformat=]
To avoid this simply change the type of cycle to u64 in microbenchmark(),
since u64 is of type unsigned long long for all architectures.
Acked-by: Marco Elver <elver@google.com>
Link: https://lore.kernel.org/r/20210729142811.1309391-1-hca@linux.ibm.com
Signed-off-by: Heiko Carstens <hca@linux.ibm.com>
To avoid kernel build failure due to some missing .BTF-ids referenced
functions/types, the patch ([1]) tries to fill btf_id 0 for
these types.
In bpf verifier, for percpu variable and helper returning btf_id cases,
verifier already emitted proper warning with something like
verbose(env, "Helper has invalid btf_id in R%d\n", regno);
verbose(env, "invalid return type %d of func %s#%d\n",
fn->ret_type, func_id_name(func_id), func_id);
But this is not the case for bpf_iter context arguments.
I hacked resolve_btfids to encode btf_id 0 for struct task_struct.
With `./test_progs -n 7/5`, I got,
0: (79) r2 = *(u64 *)(r1 +0)
func 'bpf_iter_task' arg0 has btf_id 29739 type STRUCT 'bpf_iter_meta'
; struct seq_file *seq = ctx->meta->seq;
1: (79) r6 = *(u64 *)(r2 +0)
; struct task_struct *task = ctx->task;
2: (79) r7 = *(u64 *)(r1 +8)
; if (task == (void *)0) {
3: (55) if r7 != 0x0 goto pc+11
...
; BPF_SEQ_PRINTF(seq, "%8d %8d\n", task->tgid, task->pid);
26: (61) r1 = *(u32 *)(r7 +1372)
Type '(anon)' is not a struct
Basically, verifier will return btf_id 0 for task_struct.
Later on, when the code tries to access task->tgid, the
verifier correctly complains the type is '(anon)' and it is
not a struct. Users still need to backtrace to find out
what is going on.
Let us catch the invalid btf_id 0 earlier
and provide better message indicating btf_id is wrong.
The new error message looks like below:
R1 type=ctx expected=fp
; struct seq_file *seq = ctx->meta->seq;
0: (79) r2 = *(u64 *)(r1 +0)
func 'bpf_iter_task' arg0 has btf_id 29739 type STRUCT 'bpf_iter_meta'
; struct seq_file *seq = ctx->meta->seq;
1: (79) r6 = *(u64 *)(r2 +0)
; struct task_struct *task = ctx->task;
2: (79) r7 = *(u64 *)(r1 +8)
invalid btf_id for context argument offset 8
invalid bpf_context access off=8 size=8
[1] https://lore.kernel.org/bpf/20210727132532.2473636-1-hengqi.chen@gmail.com/
Signed-off-by: Yonghong Song <yhs@fb.com>
Signed-off-by: Andrii Nakryiko <andrii@kernel.org>
Link: https://lore.kernel.org/bpf/20210728183025.1461750-1-yhs@fb.com
In error handling branch "if (WARN_ON(node == NUMA_NO_NODE))", the
previously allocated memories are not released. Doing this before
allocating memory eliminates memory leaks.
tj: Note that the condition only occurs when the arch code is pretty broken
and the WARN_ON might as well be BUG_ON().
Signed-off-by: Zhen Lei <thunder.leizhen@huawei.com>
Reviewed-by: Lai Jiangshan <jiangshanlai@gmail.com>
Signed-off-by: Tejun Heo <tj@kernel.org>
console_verbose() increases console loglevel to
CONSOLE_LOGLEVEL_MOTORMOUTH, which provides more information
to debug a panic/oops.
Unfortunately, in Arista we maintain some DUTs (Device Under Test) that
are configured to have 9600 baud rate. While verbose console messages
have their value to post-analyze crashes, on such setup they:
- may prevent panic/oops messages being printed
- take too long to flush on console resulting in watchdog reboot
In all our setups we use kdump which saves dmesg buffer after panic,
so in reality those extra messages on console provide no additional value,
but rather add risk of not getting to __crash_kexec().
Provide printk.console_no_auto_verbose boot parameter, which allows
to switch off printk being verbose on oops/panic/lockdep.
Cc: Andrew Morton <akpm@linux-foundation.org>
Cc: John Ogness <john.ogness@linutronix.de>
Cc: Petr Mladek <pmladek@suse.com>
Cc: Sergey Senozhatsky <senozhatsky@chromium.org>
Cc: Steven Rostedt <rostedt@goodmis.org>
Signed-off-by: Dmitry Safonov <dima@arista.com>
Suggested-by: Petr Mladek <pmladek@suse.com>
Reviewed-by: Sergey Senozhatsky <senozhatsky@chromium.org>
Reviewed-by: Petr Mladek <pmladek@suse.com>
Tested-by: Petr Mladek <pmladek@suse.com>
Signed-off-by: Petr Mladek <pmladek@suse.com>
Link: https://lore.kernel.org/r/20210727130635.675184-3-dima@arista.com
Daniel Borkmann says:
====================
pull-request: bpf 2021-07-29
The following pull-request contains BPF updates for your *net* tree.
We've added 9 non-merge commits during the last 14 day(s) which contain
a total of 20 files changed, 446 insertions(+), 138 deletions(-).
The main changes are:
1) Fix UBSAN out-of-bounds splat for showing XDP link fdinfo, from Lorenz Bauer.
2) Fix insufficient Spectre v4 mitigation in BPF runtime, from Daniel Borkmann,
Piotr Krysiuk and Benedict Schlueter.
3) Batch of fixes for BPF sockmap found under stress testing, from John Fastabend.
====================
Signed-off-by: David S. Miller <davem@davemloft.net>
Spectre v4 gadgets make use of memory disambiguation, which is a set of
techniques that execute memory access instructions, that is, loads and
stores, out of program order; Intel's optimization manual, section 2.4.4.5:
A load instruction micro-op may depend on a preceding store. Many
microarchitectures block loads until all preceding store addresses are
known. The memory disambiguator predicts which loads will not depend on
any previous stores. When the disambiguator predicts that a load does
not have such a dependency, the load takes its data from the L1 data
cache. Eventually, the prediction is verified. If an actual conflict is
detected, the load and all succeeding instructions are re-executed.
af86ca4e30 ("bpf: Prevent memory disambiguation attack") tried to mitigate
this attack by sanitizing the memory locations through preemptive "fast"
(low latency) stores of zero prior to the actual "slow" (high latency) store
of a pointer value such that upon dependency misprediction the CPU then
speculatively executes the load of the pointer value and retrieves the zero
value instead of the attacker controlled scalar value previously stored at
that location, meaning, subsequent access in the speculative domain is then
redirected to the "zero page".
The sanitized preemptive store of zero prior to the actual "slow" store is
done through a simple ST instruction based on r10 (frame pointer) with
relative offset to the stack location that the verifier has been tracking
on the original used register for STX, which does not have to be r10. Thus,
there are no memory dependencies for this store, since it's only using r10
and immediate constant of zero; hence af86ca4e30 /assumed/ a low latency
operation.
However, a recent attack demonstrated that this mitigation is not sufficient
since the preemptive store of zero could also be turned into a "slow" store
and is thus bypassed as well:
[...]
// r2 = oob address (e.g. scalar)
// r7 = pointer to map value
31: (7b) *(u64 *)(r10 -16) = r2
// r9 will remain "fast" register, r10 will become "slow" register below
32: (bf) r9 = r10
// JIT maps BPF reg to x86 reg:
// r9 -> r15 (callee saved)
// r10 -> rbp
// train store forward prediction to break dependency link between both r9
// and r10 by evicting them from the predictor's LRU table.
33: (61) r0 = *(u32 *)(r7 +24576)
34: (63) *(u32 *)(r7 +29696) = r0
35: (61) r0 = *(u32 *)(r7 +24580)
36: (63) *(u32 *)(r7 +29700) = r0
37: (61) r0 = *(u32 *)(r7 +24584)
38: (63) *(u32 *)(r7 +29704) = r0
39: (61) r0 = *(u32 *)(r7 +24588)
40: (63) *(u32 *)(r7 +29708) = r0
[...]
543: (61) r0 = *(u32 *)(r7 +25596)
544: (63) *(u32 *)(r7 +30716) = r0
// prepare call to bpf_ringbuf_output() helper. the latter will cause rbp
// to spill to stack memory while r13/r14/r15 (all callee saved regs) remain
// in hardware registers. rbp becomes slow due to push/pop latency. below is
// disasm of bpf_ringbuf_output() helper for better visual context:
//
// ffffffff8117ee20: 41 54 push r12
// ffffffff8117ee22: 55 push rbp
// ffffffff8117ee23: 53 push rbx
// ffffffff8117ee24: 48 f7 c1 fc ff ff ff test rcx,0xfffffffffffffffc
// ffffffff8117ee2b: 0f 85 af 00 00 00 jne ffffffff8117eee0 <-- jump taken
// [...]
// ffffffff8117eee0: 49 c7 c4 ea ff ff ff mov r12,0xffffffffffffffea
// ffffffff8117eee7: 5b pop rbx
// ffffffff8117eee8: 5d pop rbp
// ffffffff8117eee9: 4c 89 e0 mov rax,r12
// ffffffff8117eeec: 41 5c pop r12
// ffffffff8117eeee: c3 ret
545: (18) r1 = map[id:4]
547: (bf) r2 = r7
548: (b7) r3 = 0
549: (b7) r4 = 4
550: (85) call bpf_ringbuf_output#194288
// instruction 551 inserted by verifier \
551: (7a) *(u64 *)(r10 -16) = 0 | /both/ are now slow stores here
// storing map value pointer r7 at fp-16 | since value of r10 is "slow".
552: (7b) *(u64 *)(r10 -16) = r7 /
// following "fast" read to the same memory location, but due to dependency
// misprediction it will speculatively execute before insn 551/552 completes.
553: (79) r2 = *(u64 *)(r9 -16)
// in speculative domain contains attacker controlled r2. in non-speculative
// domain this contains r7, and thus accesses r7 +0 below.
554: (71) r3 = *(u8 *)(r2 +0)
// leak r3
As can be seen, the current speculative store bypass mitigation which the
verifier inserts at line 551 is insufficient since /both/, the write of
the zero sanitation as well as the map value pointer are a high latency
instruction due to prior memory access via push/pop of r10 (rbp) in contrast
to the low latency read in line 553 as r9 (r15) which stays in hardware
registers. Thus, architecturally, fp-16 is r7, however, microarchitecturally,
fp-16 can still be r2.
Initial thoughts to address this issue was to track spilled pointer loads
from stack and enforce their load via LDX through r10 as well so that /both/
the preemptive store of zero /as well as/ the load use the /same/ register
such that a dependency is created between the store and load. However, this
option is not sufficient either since it can be bypassed as well under
speculation. An updated attack with pointer spill/fills now _all_ based on
r10 would look as follows:
[...]
// r2 = oob address (e.g. scalar)
// r7 = pointer to map value
[...]
// longer store forward prediction training sequence than before.
2062: (61) r0 = *(u32 *)(r7 +25588)
2063: (63) *(u32 *)(r7 +30708) = r0
2064: (61) r0 = *(u32 *)(r7 +25592)
2065: (63) *(u32 *)(r7 +30712) = r0
2066: (61) r0 = *(u32 *)(r7 +25596)
2067: (63) *(u32 *)(r7 +30716) = r0
// store the speculative load address (scalar) this time after the store
// forward prediction training.
2068: (7b) *(u64 *)(r10 -16) = r2
// preoccupy the CPU store port by running sequence of dummy stores.
2069: (63) *(u32 *)(r7 +29696) = r0
2070: (63) *(u32 *)(r7 +29700) = r0
2071: (63) *(u32 *)(r7 +29704) = r0
2072: (63) *(u32 *)(r7 +29708) = r0
2073: (63) *(u32 *)(r7 +29712) = r0
2074: (63) *(u32 *)(r7 +29716) = r0
2075: (63) *(u32 *)(r7 +29720) = r0
2076: (63) *(u32 *)(r7 +29724) = r0
2077: (63) *(u32 *)(r7 +29728) = r0
2078: (63) *(u32 *)(r7 +29732) = r0
2079: (63) *(u32 *)(r7 +29736) = r0
2080: (63) *(u32 *)(r7 +29740) = r0
2081: (63) *(u32 *)(r7 +29744) = r0
2082: (63) *(u32 *)(r7 +29748) = r0
2083: (63) *(u32 *)(r7 +29752) = r0
2084: (63) *(u32 *)(r7 +29756) = r0
2085: (63) *(u32 *)(r7 +29760) = r0
2086: (63) *(u32 *)(r7 +29764) = r0
2087: (63) *(u32 *)(r7 +29768) = r0
2088: (63) *(u32 *)(r7 +29772) = r0
2089: (63) *(u32 *)(r7 +29776) = r0
2090: (63) *(u32 *)(r7 +29780) = r0
2091: (63) *(u32 *)(r7 +29784) = r0
2092: (63) *(u32 *)(r7 +29788) = r0
2093: (63) *(u32 *)(r7 +29792) = r0
2094: (63) *(u32 *)(r7 +29796) = r0
2095: (63) *(u32 *)(r7 +29800) = r0
2096: (63) *(u32 *)(r7 +29804) = r0
2097: (63) *(u32 *)(r7 +29808) = r0
2098: (63) *(u32 *)(r7 +29812) = r0
// overwrite scalar with dummy pointer; same as before, also including the
// sanitation store with 0 from the current mitigation by the verifier.
2099: (7a) *(u64 *)(r10 -16) = 0 | /both/ are now slow stores here
2100: (7b) *(u64 *)(r10 -16) = r7 | since store unit is still busy.
// load from stack intended to bypass stores.
2101: (79) r2 = *(u64 *)(r10 -16)
2102: (71) r3 = *(u8 *)(r2 +0)
// leak r3
[...]
Looking at the CPU microarchitecture, the scheduler might issue loads (such
as seen in line 2101) before stores (line 2099,2100) because the load execution
units become available while the store execution unit is still busy with the
sequence of dummy stores (line 2069-2098). And so the load may use the prior
stored scalar from r2 at address r10 -16 for speculation. The updated attack
may work less reliable on CPU microarchitectures where loads and stores share
execution resources.
This concludes that the sanitizing with zero stores from af86ca4e30 ("bpf:
Prevent memory disambiguation attack") is insufficient. Moreover, the detection
of stack reuse from af86ca4e30 where previously data (STACK_MISC) has been
written to a given stack slot where a pointer value is now to be stored does
not have sufficient coverage as precondition for the mitigation either; for
several reasons outlined as follows:
1) Stack content from prior program runs could still be preserved and is
therefore not "random", best example is to split a speculative store
bypass attack between tail calls, program A would prepare and store the
oob address at a given stack slot and then tail call into program B which
does the "slow" store of a pointer to the stack with subsequent "fast"
read. From program B PoV such stack slot type is STACK_INVALID, and
therefore also must be subject to mitigation.
2) The STACK_SPILL must not be coupled to register_is_const(&stack->spilled_ptr)
condition, for example, the previous content of that memory location could
also be a pointer to map or map value. Without the fix, a speculative
store bypass is not mitigated in such precondition and can then lead to
a type confusion in the speculative domain leaking kernel memory near
these pointer types.
While brainstorming on various alternative mitigation possibilities, we also
stumbled upon a retrospective from Chrome developers [0]:
[...] For variant 4, we implemented a mitigation to zero the unused memory
of the heap prior to allocation, which cost about 1% when done concurrently
and 4% for scavenging. Variant 4 defeats everything we could think of. We
explored more mitigations for variant 4 but the threat proved to be more
pervasive and dangerous than we anticipated. For example, stack slots used
by the register allocator in the optimizing compiler could be subject to
type confusion, leading to pointer crafting. Mitigating type confusion for
stack slots alone would have required a complete redesign of the backend of
the optimizing compiler, perhaps man years of work, without a guarantee of
completeness. [...]
From BPF side, the problem space is reduced, however, options are rather
limited. One idea that has been explored was to xor-obfuscate pointer spills
to the BPF stack:
[...]
// preoccupy the CPU store port by running sequence of dummy stores.
[...]
2106: (63) *(u32 *)(r7 +29796) = r0
2107: (63) *(u32 *)(r7 +29800) = r0
2108: (63) *(u32 *)(r7 +29804) = r0
2109: (63) *(u32 *)(r7 +29808) = r0
2110: (63) *(u32 *)(r7 +29812) = r0
// overwrite scalar with dummy pointer; xored with random 'secret' value
// of 943576462 before store ...
2111: (b4) w11 = 943576462
2112: (af) r11 ^= r7
2113: (7b) *(u64 *)(r10 -16) = r11
2114: (79) r11 = *(u64 *)(r10 -16)
2115: (b4) w2 = 943576462
2116: (af) r2 ^= r11
// ... and restored with the same 'secret' value with the help of AX reg.
2117: (71) r3 = *(u8 *)(r2 +0)
[...]
While the above would not prevent speculation, it would make data leakage
infeasible by directing it to random locations. In order to be effective
and prevent type confusion under speculation, such random secret would have
to be regenerated for each store. The additional complexity involved for a
tracking mechanism that prevents jumps such that restoring spilled pointers
would not get corrupted is not worth the gain for unprivileged. Hence, the
fix in here eventually opted for emitting a non-public BPF_ST | BPF_NOSPEC
instruction which the x86 JIT translates into a lfence opcode. Inserting the
latter in between the store and load instruction is one of the mitigations
options [1]. The x86 instruction manual notes:
[...] An LFENCE that follows an instruction that stores to memory might
complete before the data being stored have become globally visible. [...]
The latter meaning that the preceding store instruction finished execution
and the store is at minimum guaranteed to be in the CPU's store queue, but
it's not guaranteed to be in that CPU's L1 cache at that point (globally
visible). The latter would only be guaranteed via sfence. So the load which
is guaranteed to execute after the lfence for that local CPU would have to
rely on store-to-load forwarding. [2], in section 2.3 on store buffers says:
[...] For every store operation that is added to the ROB, an entry is
allocated in the store buffer. This entry requires both the virtual and
physical address of the target. Only if there is no free entry in the store
buffer, the frontend stalls until there is an empty slot available in the
store buffer again. Otherwise, the CPU can immediately continue adding
subsequent instructions to the ROB and execute them out of order. On Intel
CPUs, the store buffer has up to 56 entries. [...]
One small upside on the fix is that it lifts constraints from af86ca4e30
where the sanitize_stack_off relative to r10 must be the same when coming
from different paths. The BPF_ST | BPF_NOSPEC gets emitted after a BPF_STX
or BPF_ST instruction. This happens either when we store a pointer or data
value to the BPF stack for the first time, or upon later pointer spills.
The former needs to be enforced since otherwise stale stack data could be
leaked under speculation as outlined earlier. For non-x86 JITs the BPF_ST |
BPF_NOSPEC mapping is currently optimized away, but others could emit a
speculation barrier as well if necessary. For real-world unprivileged
programs e.g. generated by LLVM, pointer spill/fill is only generated upon
register pressure and LLVM only tries to do that for pointers which are not
used often. The program main impact will be the initial BPF_ST | BPF_NOSPEC
sanitation for the STACK_INVALID case when the first write to a stack slot
occurs e.g. upon map lookup. In future we might refine ways to mitigate
the latter cost.
[0] https://arxiv.org/pdf/1902.05178.pdf
[1] https://msrc-blog.microsoft.com/2018/05/21/analysis-and-mitigation-of-speculative-store-bypass-cve-2018-3639/
[2] https://arxiv.org/pdf/1905.05725.pdf
Fixes: af86ca4e30 ("bpf: Prevent memory disambiguation attack")
Fixes: f7cf25b202 ("bpf: track spill/fill of constants")
Co-developed-by: Piotr Krysiuk <piotras@gmail.com>
Co-developed-by: Benedict Schlueter <benedict.schlueter@rub.de>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
Signed-off-by: Piotr Krysiuk <piotras@gmail.com>
Signed-off-by: Benedict Schlueter <benedict.schlueter@rub.de>
Acked-by: Alexei Starovoitov <ast@kernel.org>
