Except for waiting_for_gp list, there are no concurrent operations on
free_by_rcu, free_llist and free_llist_extra lists, so use
__llist_del_all() instead of llist_del_all(). waiting_for_gp list can be
deleted by RCU callback concurrently, so still use llist_del_all().
Acked-by: Stanislav Fomichev <sdf@google.com>
Signed-off-by: Hou Tao <houtao1@huawei.com>
Link: https://lore.kernel.org/r/20221021114913.60508-3-houtao@huaweicloud.com
Signed-off-by: Alexei Starovoitov <ast@kernel.org>
A busy irq work is an unfinished irq work and it can be either in the
pending state or in the running state. When destroying bpf memory
allocator, refill_work may be busy for PREEMPT_RT kernel in which irq
work is invoked in a per-CPU RT-kthread. It is also possible for kernel
with arch_irq_work_has_interrupt() being false (e.g. 1-cpu arm32 host or
mips) and irq work is inovked in timer interrupt.
The busy refill_work leads to various issues. The obvious one is that
there will be concurrent operations on free_by_rcu and free_list between
irq work and memory draining. Another one is call_rcu_in_progress will
not be reliable for the checking of pending RCU callback because
do_call_rcu() may have not been invoked by irq work yet. The other is
there will be use-after-free if irq work is freed before the callback
of irq work is invoked as shown below:
BUG: kernel NULL pointer dereference, address: 0000000000000000
#PF: supervisor instruction fetch in kernel mode
#PF: error_code(0x0010) - not-present page
PGD 12ab94067 P4D 12ab94067 PUD 1796b4067 PMD 0
Oops: 0010 [#1] PREEMPT_RT SMP
CPU: 5 PID: 64 Comm: irq_work/5 Not tainted 6.0.0-rt11+ #1
Hardware name: QEMU Standard PC (i440FX + PIIX, 1996)
RIP: 0010:0x0
Code: Unable to access opcode bytes at 0xffffffffffffffd6.
RSP: 0018:ffffadc080293e78 EFLAGS: 00010286
RAX: 0000000000000000 RBX: ffffcdc07fb6a388 RCX: ffffa05000a2e000
RDX: ffffa05000a2e000 RSI: ffffffff96cc9827 RDI: ffffcdc07fb6a388
......
Call Trace:
<TASK>
irq_work_single+0x24/0x60
irq_work_run_list+0x24/0x30
run_irq_workd+0x23/0x30
smpboot_thread_fn+0x203/0x300
kthread+0x126/0x150
ret_from_fork+0x1f/0x30
</TASK>
Considering the ease of concurrency handling, no overhead for
irq_work_sync() under non-PREEMPT_RT kernel and has-irq-work-interrupt
kernel and the short wait time used for irq_work_sync() under PREEMPT_RT
(When running two test_maps on PREEMPT_RT kernel and 72-cpus host, the
max wait time is about 8ms and the 99th percentile is 10us), just using
irq_work_sync() to wait for busy refill_work to complete before memory
draining and memory freeing.
Fixes: 7c8199e24f ("bpf: Introduce any context BPF specific memory allocator.")
Acked-by: Stanislav Fomichev <sdf@google.com>
Signed-off-by: Hou Tao <houtao1@huawei.com>
Link: https://lore.kernel.org/r/20221021114913.60508-2-houtao@huaweicloud.com
Signed-off-by: Alexei Starovoitov <ast@kernel.org>
post-6.0 issues.
-----BEGIN PGP SIGNATURE-----
iHUEABYKAB0WIQTTMBEPP41GrTpTJgfdBJ7gKXxAjgUCY1IgYgAKCRDdBJ7gKXxA
jpyRAQDkfa1LDkfbA4dQBZShkUhBX1k3AyRO1NWMjwwTxP3H8wD9HUz1BB3ynoKc
ipzQs7q5jbBvndczEksHiG2AC7SvQAI=
=wD9I
-----END PGP SIGNATURE-----
Merge tag 'mm-hotfixes-stable-2022-10-20' of git://git.kernel.org/pub/scm/linux/kernel/git/akpm/mm
Pull misc fixes from Andrew Morron:
"Seventeen hotfixes, mainly for MM.
Five are cc:stable and the remainder address post-6.0 issues"
* tag 'mm-hotfixes-stable-2022-10-20' of git://git.kernel.org/pub/scm/linux/kernel/git/akpm/mm:
nouveau: fix migrate_to_ram() for faulting page
mm/huge_memory: do not clobber swp_entry_t during THP split
hugetlb: fix memory leak associated with vma_lock structure
mm/page_alloc: reduce potential fragmentation in make_alloc_exact()
mm: /proc/pid/smaps_rollup: fix maple tree search
mm,hugetlb: take hugetlb_lock before decrementing h->resv_huge_pages
mm/mmap: fix MAP_FIXED address return on VMA merge
mm/mmap.c: __vma_adjust(): suppress uninitialized var warning
mm/mmap: undo ->mmap() when mas_preallocate() fails
init: Kconfig: fix spelling mistake "satify" -> "satisfy"
ocfs2: clear dinode links count in case of error
ocfs2: fix BUG when iput after ocfs2_mknod fails
gcov: support GCC 12.1 and newer compilers
zsmalloc: zs_destroy_pool: add size_class NULL check
mm/mempolicy: fix mbind_range() arguments to vma_merge()
mailmap: update email for Qais Yousef
mailmap: update Dan Carpenter's email address
Currently the NMI safety debugging is only performed on architectures
that don't support NMI-safe this_cpu_inc().
Reorder the code so that other architectures like x86 also detect bad
uses.
[ paulmck: Apply kernel test robot, Stephen Rothwell, and Zqiang feedback. ]
Signed-off-by: Frederic Weisbecker <frederic@kernel.org>
Signed-off-by: Paul E. McKenney <paulmck@kernel.org>
Tell about the need to protect against concurrent updaters who may
overflow the GP counter behind the current update.
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>
Using the NMI-unsafe reader API from within an NMI handler is very likely
to be buggy for three reasons:
1) NMIs aren't strictly re-entrant (a pending nested NMI will execute at
the end of the current one) so it should be fine to use a non-atomic
increment here. However, breakpoints can still interrupt NMIs and if
a breakpoint callback has a reader on that same ssp, a racy increment
can happen.
2) If the only reader site for a given srcu_struct structure is in an
NMI handler, then RCU should be used instead of SRCU.
3) Because of the previous reason (2), an srcu_struct structure having
an SRCU read side critical section in an NMI handler is likely to
have another one from a task context.
For all these reasons, warn if an NMI-unsafe reader API is used from an
NMI handler.
Signed-off-by: Frederic Weisbecker <frederic@kernel.org>
Reviewed-by: Joel Fernandes (Google) <joel@joelfernandes.org>
Signed-off-by: Paul E. McKenney <paulmck@kernel.org>
Running rcutorture with non-zero fqs_duration module parameter in a
kernel built with CONFIG_PREEMPTION=y results in the following splat:
BUG: using __this_cpu_read() in preemptible [00000000]
code: rcu_torture_fqs/398
caller is __this_cpu_preempt_check+0x13/0x20
CPU: 3 PID: 398 Comm: rcu_torture_fqs Not tainted 6.0.0-rc1-yoctodev-standard+
Call Trace:
<TASK>
dump_stack_lvl+0x5b/0x86
dump_stack+0x10/0x16
check_preemption_disabled+0xe5/0xf0
__this_cpu_preempt_check+0x13/0x20
rcu_force_quiescent_state.part.0+0x1c/0x170
rcu_force_quiescent_state+0x1e/0x30
rcu_torture_fqs+0xca/0x160
? rcu_torture_boost+0x430/0x430
kthread+0x192/0x1d0
? kthread_complete_and_exit+0x30/0x30
ret_from_fork+0x22/0x30
</TASK>
The problem is that rcu_force_quiescent_state() uses __this_cpu_read()
in preemptible code instead of the proper raw_cpu_read(). This commit
therefore changes __this_cpu_read() to raw_cpu_read().
Signed-off-by: Zqiang <qiang1.zhang@intel.com>
Reviewed-by: Joel Fernandes (Google) <joel@joelfernandes.org>
Signed-off-by: Paul E. McKenney <paulmck@kernel.org>
This commit adds a few words to the informative message that appears
every ten seconds in RCU Tasks and RCU Tasks Trace grace periods.
This message currently reads as follows:
rcu_tasks_wait_gp: rcu_tasks grace period 1046 is 10088 jiffies old.
After this change, it provides additional context, instead reading
as follows:
rcu_tasks_wait_gp: rcu_tasks grace period number 1046 (since boot) is 10088 jiffies old.
Reported-by: Alexei Starovoitov <ast@kernel.org>
Signed-off-by: Paul E. McKenney <paulmck@kernel.org>
The commit 3fcd6a230f ("x86/cpu: Avoid cpuinfo-induced IPIing of
idle CPUs") introduced rcu_is_idle_cpu() in order to identify the
current CPU idle state. But commit f3eca381bd ("x86/aperfmperf:
Replace arch_freq_get_on_cpu()") switched to using MAX_SAMPLE_AGE,
so rcu_is_idle_cpu() is no longer used. This commit therefore removes it.
Fixes: f3eca381bd ("x86/aperfmperf: Replace arch_freq_get_on_cpu()")
Signed-off-by: Yipeng Zou <zouyipeng@huawei.com>
Reviewed-by: Frederic Weisbecker <frederic@kernel.org>
Signed-off-by: Paul E. McKenney <paulmck@kernel.org>
Starting with GCC 12.1, the created .gcda format can't be read by gcov
tool. There are 2 significant changes to the .gcda file format that
need to be supported:
a) [gcov: Use system IO buffering]
(23eb66d1d46a34cb28c4acbdf8a1deb80a7c5a05) changed that all sizes in
the format are in bytes and not in words (4B)
b) [gcov: make profile merging smarter]
(72e0c742bd01f8e7e6dcca64042b9ad7e75979de) add a new checksum to the
file header.
Tested with GCC 7.5, 10.4, 12.2 and the current master.
Link: https://lkml.kernel.org/r/624bda92-f307-30e9-9aaa-8cc678b2dfb2@suse.cz
Signed-off-by: Martin Liska <mliska@suse.cz>
Tested-by: Peter Oberparleiter <oberpar@linux.ibm.com>
Reviewed-by: Peter Oberparleiter <oberpar@linux.ibm.com>
Cc: <stable@vger.kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
The patchable_function_entry(5) might output 5 single nop
instructions (depends on toolchain), which will clash with
bpf_arch_text_poke check for 5 bytes nop instruction.
Adding early init call for dispatcher that checks and change
the patchable entry into expected 5 nop instruction if needed.
There's no need to take text_mutex, because we are using it
in early init call which is called at pre-smp time.
Fixes: ceea991a01 ("bpf: Move bpf_dispatcher function out of ftrace locations")
Signed-off-by: Jiri Olsa <jolsa@kernel.org>
Acked-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Link: https://lore.kernel.org/r/20221018075934.574415-1-jolsa@kernel.org
Signed-off-by: Alexei Starovoitov <ast@kernel.org>
Making polled RCU grace periods account for expedited grace periods
required acquiring the leaf rcu_node structure's lock during early boot,
but after rcu_init() was called. This lock is irq-disabled, but the
code incorrectly assumes that irqs are always disabled when invoking
synchronize_rcu(). The exception is early boot before the scheduler has
started, which means that upon return from synchronize_rcu(), irqs will
be incorrectly enabled.
This commit fixes this bug by using irqsave/irqrestore locking primitives.
Fixes: bf95b2bc3e ("rcu: Switch polled grace-period APIs to ->gp_seq_polled")
Reported-by: Steven Rostedt <rostedt@goodmis.org>
Signed-off-by: Paul E. McKenney <paulmck@kernel.org>
This commit adds runtime checks to verify that a given srcu_struct uses
consistent NMI-safe (or not) read-side primitives globally, but based
on the per-CPU data. These global checks are made by the grace-period
code that must scan the srcu_data structures anyway, and are done only
in kernels built with CONFIG_PROVE_RCU=y.
Link: https://lore.kernel.org/all/20220910221947.171557773@linutronix.de/
Signed-off-by: Paul E. McKenney <paulmck@kernel.org>
Reviewed-by: Frederic Weisbecker <frederic@kernel.org>
Cc: Thomas Gleixner <tglx@linutronix.de>
Cc: John Ogness <john.ogness@linutronix.de>
Cc: Petr Mladek <pmladek@suse.com>
This commit adds runtime checks to verify that a given srcu_struct uses
consistent NMI-safe (or not) read-side primitives on a per-CPU basis.
Link: https://lore.kernel.org/all/20220910221947.171557773@linutronix.de/
Signed-off-by: Paul E. McKenney <paulmck@kernel.org>
Reviewed-by: Frederic Weisbecker <frederic@kernel.org>
Cc: Thomas Gleixner <tglx@linutronix.de>
Cc: John Ogness <john.ogness@linutronix.de>
Cc: Petr Mladek <pmladek@suse.com>
On strict load-store architectures, the use of this_cpu_inc() by
srcu_read_lock() and srcu_read_unlock() is not NMI-safe in TREE SRCU.
To see this suppose that an NMI arrives in the middle of srcu_read_lock(),
just after it has read ->srcu_lock_count, but before it has written
the incremented value back to memory. If that NMI handler also does
srcu_read_lock() and srcu_read_lock() on that same srcu_struct structure,
then upon return from that NMI handler, the interrupted srcu_read_lock()
will overwrite the NMI handler's update to ->srcu_lock_count, but
leave unchanged the NMI handler's update by srcu_read_unlock() to
->srcu_unlock_count.
This can result in a too-short SRCU grace period, which can in turn
result in arbitrary memory corruption.
If the NMI handler instead interrupts the srcu_read_unlock(), this
can result in eternal SRCU grace periods, which is not much better.
This commit therefore creates a pair of new srcu_read_lock_nmisafe()
and srcu_read_unlock_nmisafe() functions, which allow SRCU readers in
both NMI handlers and in process and IRQ context. It is bad practice
to mix the existing and the new _nmisafe() primitives on the same
srcu_struct structure. Use one set or the other, not both.
Just to underline that "bad practice" point, using srcu_read_lock() at
process level and srcu_read_lock_nmisafe() in your NMI handler will not,
repeat NOT, work. If you do not immediately understand why this is the
case, please review the earlier paragraphs in this commit log.
[ paulmck: Apply kernel test robot feedback. ]
[ paulmck: Apply feedback from Randy Dunlap. ]
[ paulmck: Apply feedback from John Ogness. ]
[ paulmck: Apply feedback from Frederic Weisbecker. ]
Link: https://lore.kernel.org/all/20220910221947.171557773@linutronix.de/
Signed-off-by: Paul E. McKenney <paulmck@kernel.org>
Acked-by: Randy Dunlap <rdunlap@infradead.org> # build-tested
Reviewed-by: Frederic Weisbecker <frederic@kernel.org>
Cc: Thomas Gleixner <tglx@linutronix.de>
Cc: John Ogness <john.ogness@linutronix.de>
Cc: Petr Mladek <pmladek@suse.com>
As previous commit, 'blk_trace_cleanup' will stop block trace if
block trace's state is 'Blktrace_running'.
So remove unnessary stop block trace in 'blk_trace_shutdown'.
Signed-off-by: Ye Bin <yebin10@huawei.com>
Reviewed-by: Christoph Hellwig <hch@lst.de>
Link: https://lore.kernel.org/r/20221019033602.752383-4-yebin@huaweicloud.com
Signed-off-by: Jens Axboe <axboe@kernel.dk>
When test as follows:
step1: ioctl(sda, BLKTRACESETUP, &arg)
step2: ioctl(sda, BLKTRACESTART, NULL)
step3: ioctl(sda, BLKTRACETEARDOWN, NULL)
step4: ioctl(sda, BLKTRACESETUP, &arg)
Got issue as follows:
debugfs: File 'dropped' in directory 'sda' already present!
debugfs: File 'msg' in directory 'sda' already present!
debugfs: File 'trace0' in directory 'sda' already present!
And also find syzkaller report issue like "KASAN: use-after-free Read in relay_switch_subbuf"
"https://syzkaller.appspot.com/bug?id=13849f0d9b1b818b087341691be6cc3ac6a6bfb7"
If remove block trace without stop(BLKTRACESTOP) block trace, '__blk_trace_remove'
will just set 'q->blk_trace' with NULL. However, debugfs file isn't removed, so
will report file already present when call BLKTRACESETUP.
static int __blk_trace_remove(struct request_queue *q)
{
struct blk_trace *bt;
bt = rcu_replace_pointer(q->blk_trace, NULL,
lockdep_is_held(&q->debugfs_mutex));
if (!bt)
return -EINVAL;
if (bt->trace_state != Blktrace_running)
blk_trace_cleanup(q, bt);
return 0;
}
If do test as follows:
step1: ioctl(sda, BLKTRACESETUP, &arg)
step2: ioctl(sda, BLKTRACESTART, NULL)
step3: ioctl(sda, BLKTRACETEARDOWN, NULL)
step4: remove sda
There will remove debugfs directory which will remove recursively all file
under directory.
>> blk_release_queue
>> debugfs_remove_recursive(q->debugfs_dir)
So all files which created in 'do_blk_trace_setup' are removed, and
'dentry->d_inode' is NULL. But 'q->blk_trace' is still in 'running_trace_lock',
'trace_note_tsk' will traverse 'running_trace_lock' all nodes.
>>trace_note_tsk
>> trace_note
>> relay_reserve
>> relay_switch_subbuf
>> d_inode(buf->dentry)->i_size
To solve above issues, reference commit '5afedf670caf', call 'blk_trace_cleanup'
unconditionally in '__blk_trace_remove' and first stop block trace in
'blk_trace_cleanup'.
Signed-off-by: Ye Bin <yebin10@huawei.com>
Reviewed-by: Christoph Hellwig <hch@lst.de>
Link: https://lore.kernel.org/r/20221019033602.752383-3-yebin@huaweicloud.com
Signed-off-by: Jens Axboe <axboe@kernel.dk>
-----BEGIN PGP SIGNATURE-----
iHUEABYIAB0WIQTFp0I1jqZrAX+hPRXbK58LschIgwUCY08JQQAKCRDbK58LschI
g0M0AQCWGrJcnQFut1qwR9efZUadwxtKGAgpaA/8Smd8+v7c8AD/SeHQuGfkFiD6
rx18hv1mTfG0HuPnFQy6YZQ98vmznwE=
=DaeS
-----END PGP SIGNATURE-----
Merge tag 'for-netdev' of git://git.kernel.org/pub/scm/linux/kernel/git/bpf/bpf-next
Daniel Borkmann says:
====================
pull-request: bpf-next 2022-10-18
We've added 33 non-merge commits during the last 14 day(s) which contain
a total of 31 files changed, 874 insertions(+), 538 deletions(-).
The main changes are:
1) Add RCU grace period chaining to BPF to wait for the completion
of access from both sleepable and non-sleepable BPF programs,
from Hou Tao & Paul E. McKenney.
2) Improve helper UAPI by explicitly defining BPF_FUNC_xxx integer
values. In the wild we have seen OS vendors doing buggy backports
where helper call numbers mismatched. This is an attempt to make
backports more foolproof, from Andrii Nakryiko.
3) Add libbpf *_opts API-variants for bpf_*_get_fd_by_id() functions,
from Roberto Sassu.
4) Fix libbpf's BTF dumper for structs with padding-only fields,
from Eduard Zingerman.
5) Fix various libbpf bugs which have been found from fuzzing with
malformed BPF object files, from Shung-Hsi Yu.
6) Clean up an unneeded check on existence of SSE2 in BPF x86-64 JIT,
from Jie Meng.
7) Fix various ASAN bugs in both libbpf and selftests when running
the BPF selftest suite on arm64, from Xu Kuohai.
8) Fix missing bpf_iter_vma_offset__destroy() call in BPF iter selftest
and use in-skeleton link pointer to remove an explicit bpf_link__destroy(),
from Jiri Olsa.
9) Fix BPF CI breakage by pointing to iptables-legacy instead of relying
on symlinked iptables which got upgraded to iptables-nft,
from Martin KaFai Lau.
10) Minor BPF selftest improvements all over the place, from various others.
* tag 'for-netdev' of git://git.kernel.org/pub/scm/linux/kernel/git/bpf/bpf-next: (33 commits)
bpf/docs: Update README for most recent vmtest.sh
bpf: Use rcu_trace_implies_rcu_gp() for program array freeing
bpf: Use rcu_trace_implies_rcu_gp() in local storage map
bpf: Use rcu_trace_implies_rcu_gp() in bpf memory allocator
rcu-tasks: Provide rcu_trace_implies_rcu_gp()
selftests/bpf: Use sys_pidfd_open() helper when possible
libbpf: Fix null-pointer dereference in find_prog_by_sec_insn()
libbpf: Deal with section with no data gracefully
libbpf: Use elf_getshdrnum() instead of e_shnum
selftest/bpf: Fix error usage of ASSERT_OK in xdp_adjust_tail.c
selftests/bpf: Fix error failure of case test_xdp_adjust_tail_grow
selftest/bpf: Fix memory leak in kprobe_multi_test
selftests/bpf: Fix memory leak caused by not destroying skeleton
libbpf: Fix memory leak in parse_usdt_arg()
libbpf: Fix use-after-free in btf_dump_name_dups
selftests/bpf: S/iptables/iptables-legacy/ in the bpf_nf and xdp_synproxy test
selftests/bpf: Alphabetize DENYLISTs
selftests/bpf: Add tests for _opts variants of bpf_*_get_fd_by_id()
libbpf: Introduce bpf_link_get_fd_by_id_opts()
libbpf: Introduce bpf_btf_get_fd_by_id_opts()
...
====================
Link: https://lore.kernel.org/r/20221018210631.11211-1-daniel@iogearbox.net
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
With Clang version 16+, -fsanitize=thread will turn
memcpy/memset/memmove calls in instrumented functions into
__tsan_memcpy/__tsan_memset/__tsan_memmove calls respectively.
Add these functions to the core KCSAN runtime, so that we (a) catch data
races with mem* functions, and (b) won't run into linker errors with
such newer compilers.
Cc: stable@vger.kernel.org # v5.10+
Signed-off-by: Marco Elver <elver@google.com>
Signed-off-by: Paul E. McKenney <paulmck@kernel.org>
This commit adds code to the RTWS_POLL_GET case of rcu_torture_writer()
to verify that the value of NUM_ACTIVE_RCU_POLL_OLDSTATE is sufficiently
large
Signed-off-by: Paul E. McKenney <paulmck@kernel.org>
This commit adds code to the RTWS_POLL_GET_FULL case
of rcu_torture_writer() to verify that the value of
NUM_ACTIVE_RCU_POLL_FULL_OLDSTATE is sufficiently large.
[ paulmck: Fix whitespace issue located by checkpatch.pl. ]
Signed-off-by: Paul E. McKenney <paulmck@kernel.org>
In preparation for RCU lazy changes, wake up the RCU nocb gp thread if
needed after an entrain. This change prevents the RCU barrier callback
from waiting in the queue for several seconds before the lazy callbacks
in front of it are serviced.
Reported-by: Joel Fernandes (Google) <joel@joelfernandes.org>
Signed-off-by: Frederic Weisbecker <frederic@kernel.org>
Signed-off-by: Joel Fernandes (Google) <joel@joelfernandes.org>
Signed-off-by: Paul E. McKenney <paulmck@kernel.org>
When the bypass cblist gets too big or its timeout has occurred, it is
flushed into the main cblist. However, the bypass timer is still running
and the behavior is that it would eventually expire and wake the GP
thread.
Since we are going to use the bypass cblist for lazy CBs, do the wakeup
soon as the flush for "too big or too long" bypass list happens.
Otherwise, long delays can happen for callbacks which get promoted from
lazy to non-lazy.
This is a good thing to do anyway (regardless of future lazy patches),
since it makes the behavior consistent with behavior of other code paths
where flushing into the ->cblist makes the GP kthread into a
non-sleeping state quickly.
[ Frederic Weisbecker: Changes to avoid unnecessary GP-thread wakeups plus
comment changes. ]
Reviewed-by: Frederic Weisbecker <frederic@kernel.org>
Signed-off-by: Joel Fernandes (Google) <joel@joelfernandes.org>
Signed-off-by: Paul E. McKenney <paulmck@kernel.org>
In kernels built with either CONFIG_RCU_NOCB_CPU_DEFAULT_ALL=y or
CONFIG_NO_HZ_FULL=y, additional CPUs must be added to rcu_nocb_mask.
Except that kernels booted without the rcu_nocbs= will not have
allocated rcu_nocb_mask. And the current rcu_init_nohz() function uses
its need_rcu_nocb_mask and offload_all local variables to track the
rcu_nocb and nohz_full state.
But there is a much simpler approach, namely creating a cpumask pointer
to track the default and then using cpumask_available() to check the
rcu_nocb_mask state. This commit takes this approach, thereby simplifying
and shortening the rcu_init_nohz() function.
Signed-off-by: Zhen Lei <thunder.leizhen@huawei.com>
Reviewed-by: Joel Fernandes (Google) <joel@joelfernandes.org>
Acked-by: Frederic Weisbecker <frederic@kernel.org>
Signed-off-by: Paul E. McKenney <paulmck@kernel.org>
The rnp->qsmask is locklessly accessed from rcutree_dying_cpu(). This
may help avoid load tearing due to concurrent access, KCSAN
issues, and preserve sanity of people reading the mask in tracing.
Reviewed-by: Frederic Weisbecker <frederic@kernel.org>
Signed-off-by: Joel Fernandes (Google) <joel@joelfernandes.org>
Signed-off-by: Paul E. McKenney <paulmck@kernel.org>
Once either rcutree_online_cpu() or rcutree_dead_cpu() is invoked
concurrently, the following rcu_boost_kthread_setaffinity() race can
occur:
CPU 1 CPU2
mask = rcu_rnp_online_cpus(rnp);
...
mask = rcu_rnp_online_cpus(rnp);
...
set_cpus_allowed_ptr(t, cm);
set_cpus_allowed_ptr(t, cm);
This results in CPU2's update being overwritten by that of CPU1, and
thus the possibility of ->boost_kthread_task continuing to run on a
to-be-offlined CPU.
This commit therefore eliminates this race by relying on the pre-existing
acquisition of ->boost_kthread_mutex to serialize the full process of
changing the affinity of ->boost_kthread_task.
Signed-off-by: Pingfan Liu <kernelfans@gmail.com>
Cc: David Woodhouse <dwmw@amazon.co.uk>
Cc: Frederic Weisbecker <frederic@kernel.org>
Cc: Neeraj Upadhyay <quic_neeraju@quicinc.com>
Cc: Josh Triplett <josh@joshtriplett.org>
Cc: Steven Rostedt <rostedt@goodmis.org>
Cc: Mathieu Desnoyers <mathieu.desnoyers@efficios.com>
Cc: Lai Jiangshan <jiangshanlai@gmail.com>
Cc: Joel Fernandes <joel@joelfernandes.org>
Cc: "Jason A. Donenfeld" <Jason@zx2c4.com>
Signed-off-by: Paul E. McKenney <paulmck@kernel.org>
The rcu_report_dead() function invokes rcu_report_exp_rdp() in order
to force an immediate expedited quiescent state on the outgoing
CPU, and then it invokes rcu_preempt_deferred_qs() to provide any
required deferred quiescent state of either sort. Because the call to
rcu_preempt_deferred_qs() provides the expedited RCU quiescent state if
requested, the call to rcu_report_exp_rdp() is potentially redundant.
One possible issue is a concurrent start of a new expedited RCU
grace period, but this situation is already handled correctly
by __sync_rcu_exp_select_node_cpus(). This function will detect
that the CPU is going offline via the error return from its call
to smp_call_function_single(). In that case, it will retry, and
eventually stop retrying due to rcu_report_exp_rdp() clearing the
->qsmaskinitnext bit corresponding to the target CPU. As a result,
__sync_rcu_exp_select_node_cpus() will report the necessary quiescent
state after dealing with any remaining CPU.
This change assumes that control does not enter rcu_report_dead() within
an RCU read-side critical section, but then again, the surviving call
to rcu_preempt_deferred_qs() has always made this assumption.
This commit therefore removes the call to rcu_report_exp_rdp(), thus
relying on rcu_preempt_deferred_qs() to handle both normal and expedited
quiescent states.
Signed-off-by: Zqiang <qiang1.zhang@intel.com>
Signed-off-by: Paul E. McKenney <paulmck@kernel.org>
NMI-safe variants of srcu_read_lock() and srcu_read_unlock() are needed
by printk(), which on many architectures entails read-modify-write
atomic operations. This commit prepares Tree SRCU for this change by
making both ->srcu_lock_count and ->srcu_unlock_count by atomic_long_t.
[ paulmck: Apply feedback from John Ogness. ]
Link: https://lore.kernel.org/all/20220910221947.171557773@linutronix.de/
Signed-off-by: Paul E. McKenney <paulmck@kernel.org>
Reviewed-by: Frederic Weisbecker <frederic@kernel.org>
Cc: Thomas Gleixner <tglx@linutronix.de>
Cc: John Ogness <john.ogness@linutronix.de>
Cc: Petr Mladek <pmladek@suse.com>
To support both sleepable and normal uprobe bpf program, the freeing of
trace program array chains a RCU-tasks-trace grace period and a normal
RCU grace period one after the other.
With the introduction of rcu_trace_implies_rcu_gp(),
__bpf_prog_array_free_sleepable_cb() can check whether or not a normal
RCU grace period has also passed after a RCU-tasks-trace grace period
has passed. If it is true, it is safe to invoke kfree() directly.
Signed-off-by: Hou Tao <houtao1@huawei.com>
Link: https://lore.kernel.org/r/20221014113946.965131-5-houtao@huaweicloud.com
Signed-off-by: Alexei Starovoitov <ast@kernel.org>
Local storage map is accessible for both sleepable and non-sleepable bpf
program, and its memory is freed by using both call_rcu_tasks_trace() and
kfree_rcu() to wait for both RCU-tasks-trace grace period and RCU grace
period to pass.
With the introduction of rcu_trace_implies_rcu_gp(), both
bpf_selem_free_rcu() and bpf_local_storage_free_rcu() can check whether
or not a normal RCU grace period has also passed after a RCU-tasks-trace
grace period has passed. If it is true, it is safe to call kfree()
directly.
Signed-off-by: Hou Tao <houtao1@huawei.com>
Link: https://lore.kernel.org/r/20221014113946.965131-4-houtao@huaweicloud.com
Signed-off-by: Alexei Starovoitov <ast@kernel.org>
The memory free logic in bpf memory allocator chains a RCU Tasks Trace
grace period and a normal RCU grace period one after the other, so it
can ensure that both sleepable and non-sleepable programs have finished.
With the introduction of rcu_trace_implies_rcu_gp(),
__free_rcu_tasks_trace() can check whether or not a normal RCU grace
period has also passed after a RCU Tasks Trace grace period has passed.
If it is true, freeing these elements directly, else freeing through
call_rcu().
Signed-off-by: Hou Tao <houtao1@huawei.com>
Link: https://lore.kernel.org/r/20221014113946.965131-3-houtao@huaweicloud.com
Signed-off-by: Alexei Starovoitov <ast@kernel.org>
As an accident of implementation, an RCU Tasks Trace grace period also
acts as an RCU grace period. However, this could change at any time.
This commit therefore creates an rcu_trace_implies_rcu_gp() that currently
returns true to codify this accident. Code relying on this accident
must call this function to verify that this accident is still happening.
Reported-by: Hou Tao <houtao@huaweicloud.com>
Signed-off-by: Paul E. McKenney <paulmck@kernel.org>
Cc: Alexei Starovoitov <ast@kernel.org>
Cc: Martin KaFai Lau <martin.lau@linux.dev>
Link: https://lore.kernel.org/r/20221014113946.965131-2-houtao@huaweicloud.com
Signed-off-by: Alexei Starovoitov <ast@kernel.org>
The calculation results of the variables 'func_addr' and 'func_size' are
not affected by the for loop and do not change due to the changes of
entries[i]. The performance can be improved by moving it outside the loop.
No functional change.
Signed-off-by: Zhen Lei <thunder.leizhen@huawei.com>
Reviewed-by: Petr Mladek <pmladek@suse.com>
Signed-off-by: Jiri Kosina <jkosina@suse.cz>
* Fix a recent regression where a sleeping kernfs function is called with
css_set_lock (spinlock) held.
* Revert the commit to enable cgroup1 support for cgroup_get_from_fd/file().
Multiple users assume that the lookup only works for cgroup2 and breaks
when fed a cgroup1 file. Instead, introduce a separate set of functions to
lookup both v1 and v2 and use them where the user explicitly wants to
support both versions.
* Compat update for tools/perf/util/bpf_skel/bperf_cgroup.bpf.c.
* Add Josef Bacik as a blkcg maintainer.
-----BEGIN PGP SIGNATURE-----
iIQEABYIACwWIQTfIjM1kS57o3GsC/uxYfJx3gVYGQUCY03MlA4cdGpAa2VybmVs
Lm9yZwAKCRCxYfJx3gVYGTkUAQD7fNcSPuc2m/BvW+gySKQkp9PZMA6E6yOIqirc
QKmIgwEAwWECW7RR1alhOGD50RtYkuYVsLD1+6Ka4eMHe+EhwA4=
=kGLI
-----END PGP SIGNATURE-----
Merge tag 'cgroup-for-6.1-rc1-fixes' of git://git.kernel.org/pub/scm/linux/kernel/git/tj/cgroup
Pull cgroup fixes from Tejun Heo:
- Fix a recent regression where a sleeping kernfs function is called
with css_set_lock (spinlock) held
- Revert the commit to enable cgroup1 support for cgroup_get_from_fd/file()
Multiple users assume that the lookup only works for cgroup2 and
breaks when fed a cgroup1 file. Instead, introduce a separate set of
functions to lookup both v1 and v2 and use them where the user
explicitly wants to support both versions.
- Compat update for tools/perf/util/bpf_skel/bperf_cgroup.bpf.c.
- Add Josef Bacik as a blkcg maintainer.
* tag 'cgroup-for-6.1-rc1-fixes' of git://git.kernel.org/pub/scm/linux/kernel/git/tj/cgroup:
blkcg: Update MAINTAINERS entry
mm: cgroup: fix comments for get from fd/file helpers
perf stat: Support old kernels for bperf cgroup counting
bpf: cgroup_iter: support cgroup1 using cgroup fd
cgroup: add cgroup_v1v2_get_from_[fd/file]()
Revert "cgroup: enable cgroup_get_from_file() on cgroup1"
cgroup: Reorganize css_set_lock and kernfs path processing
audit_filter_uring(), audit_filter_inode_name() are substantially
similar to audit_filter_syscall(). Move the core logic to
__audit_filter_op() which can be parametrized for all three.
On a Skylakex system, getpid() latency (all results aggregated
across 12 boot cycles):
Min Mean Median Max pstdev
(ns) (ns) (ns) (ns)
- 196.63 207.86 206.60 230.98 (+- 3.92%)
+ 183.73 196.95 192.31 232.49 (+- 6.04%)
Performance counter stats for 'bin/getpid' (3 runs) go from:
cycles 805.58 ( +- 4.11% )
instructions 1654.11 ( +- .05% )
IPC 2.06 ( +- 3.39% )
branches 430.02 ( +- .05% )
branch-misses 1.55 ( +- 7.09% )
L1-dcache-loads 440.01 ( +- .09% )
L1-dcache-load-misses 9.05 ( +- 74.03% )
to:
cycles 765.37 ( +- 6.66% )
instructions 1677.07 ( +- 0.04% )
IPC 2.20 ( +- 5.90% )
branches 431.10 ( +- 0.04% )
branch-misses 1.60 ( +- 11.25% )
L1-dcache-loads 521.04 ( +- 0.05% )
L1-dcache-load-misses 6.92 ( +- 77.60% )
(Both aggregated over 12 boot cycles.)
The increased L1-dcache-loads are due to some intermediate values now
coming from the stack.
The improvement in cycles is due to a slightly denser loop (the list
parameter in the list_for_each_entry_rcu() exit check now comes from
a register rather than a constant as before.)
Signed-off-by: Ankur Arora <ankur.a.arora@oracle.com>
Signed-off-by: Paul Moore <paul@paul-moore.com>
ctx->major contains the current syscall number. This is, of course, a
constant for the duration of the syscall. Unfortunately, GCC's alias
analysis cannot prove that it is not modified via a pointer in the
audit_filter_syscall() loop, and so always loads it from memory.
In and of itself the load isn't very expensive (ops dependent on the
ctx->major load are only used to determine the direction of control flow
and have short dependence chains and, in any case the related branches
get predicted perfectly in the fastpath) but still cache ctx->major
in a local for two reasons:
* ctx->major is in the first cacheline of struct audit_context and has
similar alignment as audit_entry::list audit_entry. For cases
with a lot of audit rules, doing this reduces one source of contention
from a potentially busy cache-set.
* audit_in_mask() (called in the hot loop in audit_filter_syscall())
does cast manipulation and error checking on ctx->major:
audit_in_mask(const struct audit_krule *rule, unsigned long val):
if (val > 0xffffffff)
return false;
word = AUDIT_WORD(val);
if (word >= AUDIT_BITMASK_SIZE)
return false;
bit = AUDIT_BIT(val);
return rule->mask[word] & bit;
The clauses related to the rule need to be evaluated in the loop, but
the rest is unnecessarily re-evaluated for every loop iteration.
(Note, however, that most of these are cheap ALU ops and the branches
are perfectly predicted. However, see discussion on cycles
improvement below for more on why it is still worth hoisting.)
On a Skylakex system change in getpid() latency (aggregated over
12 boot cycles):
Min Mean Median Max pstdev
(ns) (ns) (ns) (ns)
- 201.30 216.14 216.22 228.46 (+- 1.45%)
+ 196.63 207.86 206.60 230.98 (+- 3.92%)
Performance counter stats for 'bin/getpid' (3 runs) go from:
cycles 836.89 ( +- .80% )
instructions 2000.19 ( +- .03% )
IPC 2.39 ( +- .83% )
branches 430.14 ( +- .03% )
branch-misses 1.48 ( +- 3.37% )
L1-dcache-loads 471.11 ( +- .05% )
L1-dcache-load-misses 7.62 ( +- 46.98% )
to:
cycles 805.58 ( +- 4.11% )
instructions 1654.11 ( +- .05% )
IPC 2.06 ( +- 3.39% )
branches 430.02 ( +- .05% )
branch-misses 1.55 ( +- 7.09% )
L1-dcache-loads 440.01 ( +- .09% )
L1-dcache-load-misses 9.05 ( +- 74.03% )
(Both aggregated over 12 boot cycles.)
instructions: we reduce around 8 instructions/iteration because some of
the computation is now hoisted out of the loop (branch count does not
change because GCC, for reasons unclear, only hoists the computations
while keeping the basic-blocks.)
cycles: improve by about 5% (in aggregate and looking at individual run
numbers.) This is likely because we now waste fewer pipeline resources
on unnecessary instructions which allows the control flow to
speculatively execute further ahead shortening the execution of the loop
a little. The final gating factor on the performance of this loop
remains the long dependence chain due to the linked-list load.
Signed-off-by: Ankur Arora <ankur.a.arora@oracle.com>
Signed-off-by: Paul Moore <paul@paul-moore.com>
Introduce distinct struct balance_callback instead of performing function
pointer casting which will trip CFI. Avoids warnings as found by Clang's
future -Wcast-function-type-strict option:
In file included from kernel/sched/core.c:84:
kernel/sched/sched.h:1755:15: warning: cast from 'void (*)(struct rq *)' to 'void (*)(struct callback_head *)' converts to incompatible function type [-Wcast-function-type-strict]
head->func = (void (*)(struct callback_head *))func;
^~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
No binary differences result from this change.
This patch is a cleanup based on Brad Spengler/PaX Team's modifications
to sched code in their last public patch of grsecurity/PaX based on my
understanding of the code. Changes or omissions from the original code
are mine and don't reflect the original grsecurity/PaX code.
Reported-by: Sami Tolvanen <samitolvanen@google.com>
Signed-off-by: Kees Cook <keescook@chromium.org>
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Reviewed-by: Nathan Chancellor <nathan@kernel.org>
Link: https://github.com/ClangBuiltLinux/linux/issues/1724
Link: https://lkml.kernel.org/r/20221008000758.2957718-1-keescook@chromium.org
In commit 97886d9dcd ("sched: Migration changes for core scheduling"),
sched_group_cookie_match() was added to help determine if a cookie
matches the core state.
However, while it iterates the SMT group, it fails to actually use the
RQ for each of the CPUs iterated, use cpu_rq(cpu) instead of rq to fix
things.
Fixes: 97886d9dcd ("sched: Migration changes for core scheduling")
Signed-off-by: Lin Shengwang <linshengwang1@huawei.com>
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Link: https://lkml.kernel.org/r/20221008022709.642-1-linshengwang1@huawei.com
Since ftrace has trampolines, don't use thunks for the __fentry__ site
but instead require that every function called from there includes
accounting. This very much includes all the direct-call functions.
Additionally, ftrace uses ROP tricks in two places:
- return_to_handler(), and
- ftrace_regs_caller() when pt_regs->orig_ax is set by a direct-call.
return_to_handler() already uses a retpoline to replace an
indirect-jump to defeat IBT, since this is a jump-type retpoline, make
sure there is no accounting done and ALTERNATIVE the RET into a ret.
ftrace_regs_caller() does much the same and gets the same treatment.
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Link: https://lore.kernel.org/r/20220915111148.927545073@infradead.org
Since the pre-symbol function padding is an integral part of the
symbol make kallsyms report it as part of the symbol by reporting it
as sym-x instead of prev_sym+y.
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Link: https://lore.kernel.org/r/20220915111148.409656012@infradead.org
When indirect calls are switched to direct calls then it has to be ensured
that the call target is not the function, but the call thunk when call
depth tracking is enabled. But static calls are available before call
thunks have been set up.
Ensure a second run through the static call patching code after call thunks
have been created. When call thunks are not enabled this has no side
effects.
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Link: https://lore.kernel.org/r/20220915111148.306100465@infradead.org
* Raw data is also filled by bpf_perf_event_output.
* Add sample_flags to indicate raw data.
* This eliminates the segfaults as shown below:
Run ./samples/bpf/trace_output
BUG pid 9 cookie 1001000000004 sized 4
BUG pid 9 cookie 1001000000004 sized 4
BUG pid 9 cookie 1001000000004 sized 4
Segmentation fault (core dumped)
Fixes: 838d9bb62d ("perf: Use sample_flags for raw_data")
Signed-off-by: Sumanth Korikkar <sumanthk@linux.ibm.com>
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Acked-by: Namhyung Kim <namhyung@kernel.org>
Link: https://lkml.kernel.org/r/20221007081327.1047552-1-sumanthk@linux.ibm.com
Marco reported:
Due to the implementation of how SIGTRAP are delivered if
perf_event_attr::sigtrap is set, we've noticed 3 issues:
1. Missing SIGTRAP due to a race with event_sched_out() (more
details below).
2. Hardware PMU events being disabled due to returning 1 from
perf_event_overflow(). The only way to re-enable the event is
for user space to first "properly" disable the event and then
re-enable it.
3. The inability to automatically disable an event after a
specified number of overflows via PERF_EVENT_IOC_REFRESH.
The worst of the 3 issues is problem (1), which occurs when a
pending_disable is "consumed" by a racing event_sched_out(), observed
as follows:
CPU0 | CPU1
--------------------------------+---------------------------
__perf_event_overflow() |
perf_event_disable_inatomic() |
pending_disable = CPU0 | ...
| _perf_event_enable()
| event_function_call()
| task_function_call()
| /* sends IPI to CPU0 */
<IPI> | ...
__perf_event_enable() +---------------------------
ctx_resched()
task_ctx_sched_out()
ctx_sched_out()
group_sched_out()
event_sched_out()
pending_disable = -1
</IPI>
<IRQ-work>
perf_pending_event()
perf_pending_event_disable()
/* Fails to send SIGTRAP because no pending_disable! */
</IRQ-work>
In the above case, not only is that particular SIGTRAP missed, but also
all future SIGTRAPs because 'event_limit' is not reset back to 1.
To fix, rework pending delivery of SIGTRAP via IRQ-work by introduction
of a separate 'pending_sigtrap', no longer using 'event_limit' and
'pending_disable' for its delivery.
Additionally; and different to Marco's proposed patch:
- recognise that pending_disable effectively duplicates oncpu for
the case where it is set. As such, change the irq_work handler to
use ->oncpu to target the event and use pending_* as boolean toggles.
- observe that SIGTRAP targets the ctx->task, so the context switch
optimization that carries contexts between tasks is invalid. If
the irq_work were delayed enough to hit after a context switch the
SIGTRAP would be delivered to the wrong task.
- observe that if the event gets scheduled out
(rotation/migration/context-switch/...) the irq-work would be
insufficient to deliver the SIGTRAP when the event gets scheduled
back in (the irq-work might still be pending on the old CPU).
Therefore have event_sched_out() convert the pending sigtrap into a
task_work which will deliver the signal at return_to_user.
Fixes: 97ba62b278 ("perf: Add support for SIGTRAP on perf events")
Reported-by: Dmitry Vyukov <dvyukov@google.com>
Debugged-by: Dmitry Vyukov <dvyukov@google.com>
Reported-by: Marco Elver <elver@google.com>
Debugged-by: Marco Elver <elver@google.com>
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Reviewed-by: Marco Elver <elver@google.com>
Tested-by: Marco Elver <elver@google.com>
Replace the obsolete and ambiguous macro in_irq() with new
macro in_hardirq().
Signed-off-by: ye xingchen <ye.xingchen@zte.com.cn>
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
Acked-by: John Stultz <jstultz@google.com>
Link: https://lore.kernel.org/r/20221012012629.334966-1-ye.xingchen@zte.com.cn
-----BEGIN PGP SIGNATURE-----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=M+mV
-----END PGP SIGNATURE-----
Merge tag 'random-6.1-rc1-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/crng/random
Pull more random number generator updates from Jason Donenfeld:
"This time with some large scale treewide cleanups.
The intent of this pull is to clean up the way callers fetch random
integers. The current rules for doing this right are:
- If you want a secure or an insecure random u64, use get_random_u64()
- If you want a secure or an insecure random u32, use get_random_u32()
The old function prandom_u32() has been deprecated for a while
now and is just a wrapper around get_random_u32(). Same for
get_random_int().
- If you want a secure or an insecure random u16, use get_random_u16()
- If you want a secure or an insecure random u8, use get_random_u8()
- If you want secure or insecure random bytes, use get_random_bytes().
The old function prandom_bytes() has been deprecated for a while
now and has long been a wrapper around get_random_bytes()
- If you want a non-uniform random u32, u16, or u8 bounded by a
certain open interval maximum, use prandom_u32_max()
I say "non-uniform", because it doesn't do any rejection sampling
or divisions. Hence, it stays within the prandom_*() namespace, not
the get_random_*() namespace.
I'm currently investigating a "uniform" function for 6.2. We'll see
what comes of that.
By applying these rules uniformly, we get several benefits:
- By using prandom_u32_max() with an upper-bound that the compiler
can prove at compile-time is ≤65536 or ≤256, internally
get_random_u16() or get_random_u8() is used, which wastes fewer
batched random bytes, and hence has higher throughput.
- By using prandom_u32_max() instead of %, when the upper-bound is
not a constant, division is still avoided, because
prandom_u32_max() uses a faster multiplication-based trick instead.
- By using get_random_u16() or get_random_u8() in cases where the
return value is intended to indeed be a u16 or a u8, we waste fewer
batched random bytes, and hence have higher throughput.
This series was originally done by hand while I was on an airplane
without Internet. Later, Kees and I worked on retroactively figuring
out what could be done with Coccinelle and what had to be done
manually, and then we split things up based on that.
So while this touches a lot of files, the actual amount of code that's
hand fiddled is comfortably small"
* tag 'random-6.1-rc1-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/crng/random:
prandom: remove unused functions
treewide: use get_random_bytes() when possible
treewide: use get_random_u32() when possible
treewide: use get_random_{u8,u16}() when possible, part 2
treewide: use get_random_{u8,u16}() when possible, part 1
treewide: use prandom_u32_max() when possible, part 2
treewide: use prandom_u32_max() when possible, part 1
- Various performance optimizations, resulting in a 4%-9% speedup
in the mmtests/config-scheduler-perfpipe micro-benchmark.
- New interface to turn PSI on/off on a per cgroup level.
Signed-off-by: Ingo Molnar <mingo@kernel.org>
-----BEGIN PGP SIGNATURE-----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=91tB
-----END PGP SIGNATURE-----
Merge tag 'sched-psi-2022-10-14' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip
Pull PSI updates from Ingo Molnar:
- Various performance optimizations, resulting in a 4%-9% speedup in
the mmtests/config-scheduler-perfpipe micro-benchmark.
- New interface to turn PSI on/off on a per cgroup level.
* tag 'sched-psi-2022-10-14' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip:
sched/psi: Per-cgroup PSI accounting disable/re-enable interface
sched/psi: Cache parent psi_group to speed up group iteration
sched/psi: Consolidate cgroup_psi()
sched/psi: Add PSI_IRQ to track IRQ/SOFTIRQ pressure
sched/psi: Remove NR_ONCPU task accounting
sched/psi: Optimize task switch inside shared cgroups again
sched/psi: Move private helpers to sched/stats.h
sched/psi: Save percpu memory when !psi_cgroups_enabled
sched/psi: Don't create cgroup PSI files when psi_disabled
sched/psi: Fix periodic aggregation shut off
