As the base switch can be avoided completely when the base stays the same
the remove/enqueue handling can be more streamlined.
Split it out into a separate function which handles both in one go which is
way more efficient and makes the code simpler to follow.
Signed-off-by: Thomas Gleixner <tglx@kernel.org>
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Link: https://patch.msgid.link/20260224163430.737600486@kernel.org
The decision to keep a timer which is associated to the local CPU on that
CPU does not take NOHZ information into account. As a result there are a
lot of hrtimer base switch invocations which end up not switching the base
and stay on the local CPU. That's just work for nothing and can be further
improved.
If the local CPU is part of the NOISE housekeeping mask, then check:
1) Whether the local CPU has the tick running, which means it is
either not idle or already expecting a timer soon.
2) Whether the tick is stopped and need_resched() is set, which
means the CPU is about to exit idle.
This reduces the amount of hrtimer base switch attempts, which end up on
the local CPU anyway, significantly and prepares for further optimizations.
Signed-off-by: Thomas Gleixner <tglx@kernel.org>
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Link: https://patch.msgid.link/20260224163430.673473029@kernel.org
The decision whether to keep timers on the local CPU or on the CPU they are
associated to is suboptimal and causes the expensive switch_hrtimer_base()
mechanism to be invoked more than necessary. This is especially true for
pinned timers.
Rewrite the decision logic so that the current base is kept if:
1) The callback is running on the base
2) The timer is associated to the local CPU and the first expiring timer as
that allows to optimize for reprogramming avoidance
3) The timer is associated to the local CPU and pinned
4) The timer is associated to the local CPU and timer migration is
disabled.
Only #2 was covered by the original code, but especially #3 makes a
difference for high frequency rearming timers like the scheduler hrtick
timer. If timer migration is disabled, then #4 avoids most of the base
switches.
Signed-off-by: Thomas Gleixner <tglx@kernel.org>
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Link: https://patch.msgid.link/20260224163430.607935269@kernel.org
All 'u8' flags are true booleans, so make it entirely clear that these can
only contain true or false.
This is especially true for hrtimer::state, which has a historical leftover
of using the state with bitwise operations. That was used in the early
hrtimer implementation with several bits, but then converted to a boolean
state. But that conversion missed to replace the bit OR and bit check
operations all over the place, which creates suboptimal code. As of today
'state' is a misnomer because it's only purpose is to reflect whether the
timer is enqueued into the RB-tree or not. Rename it to 'is_queued' and
make all operations on it boolean.
This reduces text size from 8926 to 8732 bytes.
Signed-off-by: Thomas Gleixner <tglx@kernel.org>
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Link: https://patch.msgid.link/20260224163430.542427240@kernel.org
hrtimer_start() when invoked with an already armed timer traces like:
<comm>-.. [032] d.h2. 5.002263: hrtimer_cancel: hrtimer= ....
<comm>-.. [032] d.h1. 5.002263: hrtimer_start: hrtimer= ....
Which is incorrect as the timer doesn't get canceled. Just the expiry time
changes. The internal dequeue operation which is required for that is not
really interesting for trace analysis. But it makes it tedious to keep real
cancellations and the above case apart.
Remove the cancel tracing in hrtimer_start() and add a 'was_armed'
indicator to the hrtimer start tracepoint, which clearly indicates what the
state of the hrtimer is when hrtimer_start() is invoked:
<comm>-.. [032] d.h1. 6.200103: hrtimer_start: hrtimer= .... was_armed=0
<comm>-.. [032] d.h1. 6.200558: hrtimer_start: hrtimer= .... was_armed=1
Fixes: c6a2a17702 ("hrtimer: Add tracepoint for hrtimers")
Signed-off-by: Thomas Gleixner <tglx@kernel.org>
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Link: https://patch.msgid.link/20260224163430.208491877@kernel.org
Some clockevent devices are coupled to the system clocksource by
implementing a less than or equal comparator which compares the programmed
absolute expiry time against the underlying time counter.
The timekeeping core provides a function to convert and absolute
CLOCK_MONOTONIC based expiry time to a absolute clock cycles time which can
be directly fed into the comparator. That spares two time reads in the next
event progamming path, one to convert the absolute nanoseconds time to a
delta value and the other to convert the delta value back to a absolute
time value suitable for the comparator.
Provide a new clocksource callback which takes the absolute cycle value and
wire it up in clockevents_program_event(). Similar to clocksources allow
architectures to inline the rearm operation.
Signed-off-by: Thomas Gleixner <tglx@kernel.org>
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Link: https://patch.msgid.link/20260224163430.010425428@kernel.org
Some architectures have clockevent devices which are coupled to the system
clocksource by implementing a less than or equal comparator which compares
the programmed absolute expiry time against the underlying time
counter. Well known examples are TSC/TSC deadline timer and the S390 TOD
clocksource/comparator.
While the concept is nice it has some downsides:
1) The clockevents core code is strictly based on relative expiry times
as that's the most common case for clockevent device hardware. That
requires to convert the absolute expiry time provided by the caller
(hrtimers, NOHZ code) to a relative expiry time by reading and
substracting the current time.
The clockevent::set_next_event() callback must then read the counter
again to convert the relative expiry back into a absolute one.
2) The conversion factors from nanoseconds to counter clock cycles are
set up when the clockevent is registered. When NTP applies corrections
then the clockevent conversion factors can deviate from the
clocksource conversion substantially which either results in timers
firing late or in the worst case early. The early expiry then needs to
do a reprogam with a short delta.
In most cases this is papered over by the fact that the read in the
set_next_event() callback happens after the read which is used to
calculate the delta. So the tendency is that timers expire mostly
late.
All of this can be avoided by providing support for these devices in the
core code:
1) The timekeeping core keeps track of the last update to the clocksource
by storing the base nanoseconds and the corresponding clocksource
counter value. That's used to keep the conversion math for reading the
time within 64-bit in the common case.
This information can be used to avoid both reads of the underlying
clocksource in the clockevents reprogramming path:
delta = expiry - base_ns;
cycles = base_cycles + ((delta * clockevent::mult) >> clockevent::shift);
The resulting cycles value can be directly used to program the
comparator.
2) As #1 does not longer provide the "compensation" through the second
read the deviation of the clocksource and clockevent conversions
caused by NTP become more prominent.
This can be cured by letting the timekeeping core compute and store
the reverse conversion factors when the clocksource cycles to
nanoseconds factors are modified by NTP:
CS::MULT (1 << NS_TO_CYC_SHIFT)
--------------- = ----------------------
(1 << CS:SHIFT) NS_TO_CYC_MULT
Ergo: NS_TO_CYC_MULT = (1 << (CS::SHIFT + NS_TO_CYC_SHIFT)) / CS::MULT
The NS_TO_CYC_SHIFT value is calculated when the clocksource is
installed so that it aims for a one hour maximum sleep time.
Signed-off-by: Thomas Gleixner <tglx@kernel.org>
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Link: https://patch.msgid.link/20260224163429.944763521@kernel.org
On some architectures clocksource::read() boils down to a single
instruction, so the indirect function call is just a massive overhead
especially with speculative execution mitigations in effect.
Allow architectures to enable conditional inlining of that read to avoid
that by:
- providing a static branch to switch to the inlined variant
- disabling the branch before clocksource changes
- enabling the branch after a clocksource change, when the clocksource
indicates in a feature flag that it is the one which provides the
inlined variant
This is intentionally not a static call as that would only remove the
indirect call, but not the rest of the overhead.
Signed-off-by: Thomas Gleixner <tglx@kernel.org>
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Link: https://patch.msgid.link/20260224163429.675151545@kernel.org
The sequence of cancel and start is inefficient. It has to do the timer
lock/unlock twice and in the worst case has to reprogram the underlying
clock event device twice.
The reason why it is done this way is the usage of hrtimer_forward_now(),
which requires the timer to be inactive.
But that can be completely avoided as the forward can be done on a variable
and does not need any of the overrun accounting provided by
hrtimer_forward_now().
Implement a trivial forwarding mechanism and replace the cancel/reprogram
sequence with hrtimer_start(..., new_expiry).
For the non high resolution case the timer is not actually armed, but used
for storage so that code checking for expiry times can unconditially look
it up in the timer. So it is safe for that case to set the new expiry time
directly.
Signed-off-by: Thomas Gleixner <tglx@kernel.org>
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Link: https://patch.msgid.link/20260224163429.542178086@kernel.org
The hrtick timer is frequently rearmed before expiry and most of the time
the new expiry is past the armed one. As this happens on every context
switch it becomes expensive with scheduling heavy work loads especially in
virtual machines as the "hardware" reprogamming implies a VM exit.
hrtimer now provide a lazy rearm mode flag which skips the reprogamming if:
1) The timer was the first expiring timer before the rearm
2) The new expiry time is farther out than the armed time
This avoids a massive amount of reprogramming operations of the hrtick
timer for the price of eventually taking the alredy armed interrupt for
nothing.
Mark the hrtick timer accordingly.
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Signed-off-by: Thomas Gleixner <tglx@kernel.org>
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Link: https://patch.msgid.link/20260224163429.475409346@kernel.org
The hrtick timer is frequently rearmed before expiry and most of the time
the new expiry is past the armed one. As this happens on every context
switch it becomes expensive with scheduling heavy work loads especially in
virtual machines as the "hardware" reprogamming implies a VM exit.
Add a lazy rearm mode flag which skips the reprogamming if:
1) The timer was the first expiring timer before the rearm
2) The new expiry time is farther out than the armed time
This avoids a massive amount of reprogramming operations of the hrtick
timer for the price of eventually taking the alredy armed interrupt for
nothing.
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Signed-off-by: Thomas Gleixner <tglx@kernel.org>
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Link: https://patch.msgid.link/20260224163429.408524456@kernel.org
schedule() provides several mechanisms to update the hrtick timer:
1) When the next task is picked
2) When the balance callbacks are invoked before rq::lock is released
Each of them can result in a first expiring timer and cause a reprogram of
the clock event device.
Solve this by deferring the rearm to the end of schedule() right before
releasing rq::lock by setting a flag on entry which tells hrtick_start() to
cache the runtime constraint in rq::hrtick_delay without touching the timer
itself.
Right before releasing rq::lock evaluate the flags and either rearm or
cancel the hrtick timer.
Signed-off-by: Thomas Gleixner <tglx@kernel.org>
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Link: https://patch.msgid.link/20260224163429.273068659@kernel.org
The scheduler evaluates this via hrtimer_is_hres_active() every time it has
to update HRTICK. This needs to follow three pointers, which is expensive.
Provide a static branch based mechanism to avoid that.
Signed-off-by: Thomas Gleixner <tglx@kernel.org>
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Link: https://patch.msgid.link/20260224163429.136503358@kernel.org
The nominal duration for an EEVDF task to run is until its deadline. At
which point the deadline is moved ahead and a new task selection is done.
Try and predict the time 'lost' to higher scheduling classes. Since this is
an estimate, the timer can be both early or late. In case it is early
task_tick_fair() will take the !need_resched() path and restarts the timer.
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Signed-off-by: Thomas Gleixner <tglx@kernel.org>
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Reviewed-by: Juri Lelli <juri.lelli@redhat.com>
Link: https://patch.msgid.link/20260224163428.798198874@kernel.org
This converts some of the visually simpler cases that have been split
over multiple lines. I only did the ones that are easy to verify the
resulting diff by having just that final GFP_KERNEL argument on the next
line.
Somebody should probably do a proper coccinelle script for this, but for
me the trivial script actually resulted in an assertion failure in the
middle of the script. I probably had made it a bit _too_ trivial.
So after fighting that far a while I decided to just do some of the
syntactically simpler cases with variations of the previous 'sed'
scripts.
The more syntactically complex multi-line cases would mostly really want
whitespace cleanup anyway.
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
This is the exact same thing as the 'alloc_obj()' version, only much
smaller because there are a lot fewer users of the *alloc_flex()
interface.
As with alloc_obj() version, this was done entirely with mindless brute
force, using the same script, except using 'flex' in the pattern rather
than 'objs*'.
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
This was done entirely with mindless brute force, using
git grep -l '\<k[vmz]*alloc_objs*(.*, GFP_KERNEL)' |
xargs sed -i 's/\(alloc_objs*(.*\), GFP_KERNEL)/\1)/'
to convert the new alloc_obj() users that had a simple GFP_KERNEL
argument to just drop that argument.
Note that due to the extreme simplicity of the scripting, any slightly
more complex cases spread over multiple lines would not be triggered:
they definitely exist, but this covers the vast bulk of the cases, and
the resulting diff is also then easier to check automatically.
For the same reason the 'flex' versions will be done as a separate
conversion.
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
This is the result of running the Coccinelle script from
scripts/coccinelle/api/kmalloc_objs.cocci. The script is designed to
avoid scalar types (which need careful case-by-case checking), and
instead replace kmalloc-family calls that allocate struct or union
object instances:
Single allocations: kmalloc(sizeof(TYPE), ...)
are replaced with: kmalloc_obj(TYPE, ...)
Array allocations: kmalloc_array(COUNT, sizeof(TYPE), ...)
are replaced with: kmalloc_objs(TYPE, COUNT, ...)
Flex array allocations: kmalloc(struct_size(PTR, FAM, COUNT), ...)
are replaced with: kmalloc_flex(*PTR, FAM, COUNT, ...)
(where TYPE may also be *VAR)
The resulting allocations no longer return "void *", instead returning
"TYPE *".
Signed-off-by: Kees Cook <kees@kernel.org>
Pull tracing fixes from Steven Rostedt:
- Fix possible dereference of uninitialized pointer
When validating the persistent ring buffer on boot up, if the first
validation fails, a reference to "head_page" is performed in the
error path, but it skips over the initialization of that variable.
Move the initialization before the first validation check.
- Fix use of event length in validation of persistent ring buffer
On boot up, the persistent ring buffer is checked to see if it is
valid by several methods. One being to walk all the events in the
memory location to make sure they are all valid. The length of the
event is used to move to the next event. This length is determined by
the data in the buffer. If that length is corrupted, it could
possibly make the next event to check located at a bad memory
location.
Validate the length field of the event when doing the event walk.
- Fix function graph on archs that do not support use of ftrace_ops
When an architecture defines HAVE_DYNAMIC_FTRACE_WITH_ARGS, it means
that its function graph tracer uses the ftrace_ops of the function
tracer to call its callbacks. This allows a single registered
callback to be called directly instead of checking the callback's
meta data's hash entries against the function being traced.
For architectures that do not support this feature, it must always
call the loop function that tests each registered callback (even if
there's only one). The loop function tests each callback's meta data
against its hash of functions and will call its callback if the
function being traced is in its hash map.
The issue was that there was no check against this and the direct
function was being called even if the architecture didn't support it.
This meant that if function tracing was enabled at the same time as a
callback was registered with the function graph tracer, its callback
would be called for every function that the function tracer also
traced, even if the callback's meta data only wanted to be called
back for a small subset of functions.
Prevent the direct calling for those architectures that do not
support it.
- Fix references to trace_event_file for hist files
The hist files used event_file_data() to get a reference to the
associated trace_event_file the histogram was attached to. This would
return a pointer even if the trace_event_file is about to be freed
(via RCU). Instead it should use the event_file_file() helper that
returns NULL if the trace_event_file is marked to be freed so that no
new references are added to it.
- Wake up hist poll readers when an event is being freed
When polling on a hist file, the task is only awoken when a hist
trigger is triggered. This means that if an event is being freed
while there's a task waiting on its hist file, it will need to wait
until the hist trigger occurs to wake it up and allow the freeing to
happen. Note, the event will not be completely freed until all
references are removed, and a hist poller keeps a reference. But it
should still be woken when the event is being freed.
* tag 'trace-v7.0-2' of git://git.kernel.org/pub/scm/linux/kernel/git/trace/linux-trace:
tracing: Wake up poll waiters for hist files when removing an event
tracing: Fix checking of freed trace_event_file for hist files
fgraph: Do not call handlers direct when not using ftrace_ops
tracing: ring-buffer: Fix to check event length before using
ring-buffer: Fix possible dereference of uninitialized pointer
The event_hist_poll() function attempts to verify whether an event file is
being removed, but this check may not occur or could be unnecessarily
delayed. This happens because hist_poll_wakeup() is currently invoked only
from event_hist_trigger() when a hist command is triggered. If the event
file is being removed, no associated hist command will be triggered and a
waiter will be woken up only after an unrelated hist command is triggered.
Fix the issue by adding a call to hist_poll_wakeup() in
remove_event_file_dir() after setting the EVENT_FILE_FL_FREED flag. This
ensures that a task polling on a hist file is woken up and receives
EPOLLERR.
Cc: stable@vger.kernel.org
Cc: Mathieu Desnoyers <mathieu.desnoyers@efficios.com>
Cc: Tom Zanussi <zanussi@kernel.org>
Acked-by: Masami Hiramatsu (Google) <mhiramat@kernel.org>
Link: https://patch.msgid.link/20260219162737.314231-3-petr.pavlu@suse.com
Fixes: 1bd13edbbe ("tracing/hist: Add poll(POLLIN) support on hist file")
Signed-off-by: Petr Pavlu <petr.pavlu@suse.com>
Signed-off-by: Steven Rostedt (Google) <rostedt@goodmis.org>
The event_hist_open() and event_hist_poll() functions currently retrieve
a trace_event_file pointer from a file struct by invoking
event_file_data(), which simply returns file->f_inode->i_private. The
functions then check if the pointer is NULL to determine whether the event
is still valid. This approach is flawed because i_private is assigned when
an eventfs inode is allocated and remains set throughout its lifetime.
Instead, the code should call event_file_file(), which checks for
EVENT_FILE_FL_FREED. Using the incorrect access function may result in the
code potentially opening a hist file for an event that is being removed or
becoming stuck while polling on this file.
Correct the access method to event_file_file() in both functions.
Cc: stable@vger.kernel.org
Cc: Mathieu Desnoyers <mathieu.desnoyers@efficios.com>
Cc: Tom Zanussi <zanussi@kernel.org>
Link: https://patch.msgid.link/20260219162737.314231-2-petr.pavlu@suse.com
Fixes: 1bd13edbbe ("tracing/hist: Add poll(POLLIN) support on hist file")
Signed-off-by: Petr Pavlu <petr.pavlu@suse.com>
Acked-by: Masami Hiramatsu (Google) <mhiramat@kernel.org>
Signed-off-by: Steven Rostedt (Google) <rostedt@goodmis.org>
The function graph tracer was modified to us the ftrace_ops of the
function tracer. This simplified the code as well as allowed more features
of the function graph tracer.
Not all architectures were converted over as it required the
implementation of HAVE_DYNAMIC_FTRACE_WITH_ARGS to implement. For those
architectures, it still did it the old way where the function graph tracer
handle was called by the function tracer trampoline. The handler then had
to check the hash to see if the registered handlers wanted to be called by
that function or not.
In order to speed up the function graph tracer that used ftrace_ops, if
only one callback was registered with function graph, it would call its
function directly via a static call.
Now, if the architecture does not support the use of using ftrace_ops and
still has the ftrace function trampoline calling the function graph
handler, then by doing a direct call it removes the check against the
handler's hash (list of functions it wants callbacks to), and it may call
that handler for functions that the handler did not request calls for.
On 32bit x86, which does not support the ftrace_ops use with function
graph tracer, it shows the issue:
~# trace-cmd start -p function -l schedule
~# trace-cmd show
# tracer: function_graph
#
# CPU DURATION FUNCTION CALLS
# | | | | | | |
2) * 11898.94 us | schedule();
3) # 1783.041 us | schedule();
1) | schedule() {
------------------------------------------
1) bash-8369 => kworker-7669
------------------------------------------
1) | schedule() {
------------------------------------------
1) kworker-7669 => bash-8369
------------------------------------------
1) + 97.004 us | }
1) | schedule() {
[..]
Now by starting the function tracer is another instance:
~# trace-cmd start -B foo -p function
This causes the function graph tracer to trace all functions (because the
function trace calls the function graph tracer for each on, and the
function graph trace is doing a direct call):
~# trace-cmd show
# tracer: function_graph
#
# CPU DURATION FUNCTION CALLS
# | | | | | | |
1) 1.669 us | } /* preempt_count_sub */
1) + 10.443 us | } /* _raw_spin_unlock_irqrestore */
1) | tick_program_event() {
1) | clockevents_program_event() {
1) 1.044 us | ktime_get();
1) 6.481 us | lapic_next_event();
1) + 10.114 us | }
1) + 11.790 us | }
1) ! 181.223 us | } /* hrtimer_interrupt */
1) ! 184.624 us | } /* __sysvec_apic_timer_interrupt */
1) | irq_exit_rcu() {
1) 0.678 us | preempt_count_sub();
When it should still only be tracing the schedule() function.
To fix this, add a macro FGRAPH_NO_DIRECT to be set to 0 when the
architecture does not support function graph use of ftrace_ops, and set to
1 otherwise. Then use this macro to know to allow function graph tracer to
call the handlers directly or not.
Cc: stable@vger.kernel.org
Cc: Masami Hiramatsu <mhiramat@kernel.org>
Cc: Mathieu Desnoyers <mathieu.desnoyers@efficios.com>
Cc: Mark Rutland <mark.rutland@arm.com>
Link: https://patch.msgid.link/20260218104244.5f14dade@gandalf.local.home
Fixes: cc60ee813b ("function_graph: Use static_call and branch to optimize entry function")
Signed-off-by: Steven Rostedt (Google) <rostedt@goodmis.org>
Check the event length before adding it for accessing next index in
rb_read_data_buffer(). Since this function is used for validating
possibly broken ring buffers, the length of the event could be broken.
In that case, the new event (e + len) can point a wrong address.
To avoid invalid memory access at boot, check whether the length of
each event is in the possible range before using it.
Cc: stable@vger.kernel.org
Cc: Mathieu Desnoyers <mathieu.desnoyers@efficios.com>
Fixes: 5f3b6e839f ("ring-buffer: Validate boot range memory events")
Link: https://patch.msgid.link/177123421541.142205.9414352170164678966.stgit@devnote2
Signed-off-by: Masami Hiramatsu (Google) <mhiramat@kernel.org>
Signed-off-by: Steven Rostedt (Google) <rostedt@goodmis.org>
Pull bpf fixes from Alexei Starovoitov:
- Fix invalid write loop logic in libbpf's bpf_linker__add_buf() (Amery
Hung)
- Fix a potential use-after-free of BTF object (Anton Protopopov)
- Add feature detection to libbpf and avoid moving arena global
variables on older kernels (Emil Tsalapatis)
- Remove extern declaration of bpf_stream_vprintk() from libbpf headers
(Ihor Solodrai)
- Fix truncated netlink dumps in bpftool (Jakub Kicinski)
- Fix map_kptr grace period wait in bpf selftests (Kumar Kartikeya
Dwivedi)
- Remove hexdump dependency while building bpf selftests (Matthieu
Baerts)
- Complete fsession support in BPF trampolines on riscv (Menglong Dong)
* tag 'bpf-fixes' of git://git.kernel.org/pub/scm/linux/kernel/git/bpf/bpf:
selftests/bpf: Remove hexdump dependency
libbpf: Remove extern declaration of bpf_stream_vprintk()
selftests/bpf: Use vmlinux.h in test_xdp_meta
bpftool: Fix truncated netlink dumps
libbpf: Delay feature gate check until object prepare time
libbpf: Do not use PROG_TYPE_TRACEPOINT program for feature gating
bpf: Add a map/btf from a fd array more consistently
selftests/bpf: Fix map_kptr grace period wait
selftests/bpf: enable fsession_test on riscv64
selftests/bpf: Adjust selftest due to function rename
bpf, riscv: add fsession support for trampolines
bpf: Fix a potential use-after-free of BTF object
bpf, riscv: introduce emit_store_stack_imm64() for trampoline
libbpf: Fix invalid write loop logic in bpf_linker__add_buf()
libbpf: Add gating for arena globals relocation feature
Pull more non-MM updates from Andrew Morton:
- "two fixes in kho_populate()" fixes a couple of not-major issues in
the kexec handover code (Ran Xiaokai)
- misc singletons
* tag 'mm-nonmm-stable-2026-02-18-19-56' of git://git.kernel.org/pub/scm/linux/kernel/git/akpm/mm:
lib/group_cpus: handle const qualifier from clusters allocation type
kho: remove unnecessary WARN_ON(err) in kho_populate()
kho: fix missing early_memunmap() call in kho_populate()
scripts/gdb: implement x86_page_ops in mm.py
objpool: fix the overestimation of object pooling metadata size
selftests/memfd: use IPC semaphore instead of SIGSTOP/SIGCONT
delayacct: fix build regression on accounting tool
Pull more MM updates from Andrew Morton:
- "mm/vmscan: fix demotion targets checks in reclaim/demotion" fixes a
couple of issues in the demotion code - pages were failed demotion
and were finding themselves demoted into disallowed nodes (Bing Jiao)
- "Remove XA_ZERO from error recovery of dup_mmap()" fixes a rare
mapledtree race and performs a number of cleanups (Liam Howlett)
- "mm: add bitmap VMA flag helpers and convert all mmap_prepare to use
them" implements a lot of cleanups following on from the conversion
of the VMA flags into a bitmap (Lorenzo Stoakes)
- "support batch checking of references and unmapping for large folios"
implements batching to greatly improve the performance of reclaiming
clean file-backed large folios (Baolin Wang)
- "selftests/mm: add memory failure selftests" does as claimed (Miaohe
Lin)
* tag 'mm-stable-2026-02-18-19-48' of git://git.kernel.org/pub/scm/linux/kernel/git/akpm/mm: (36 commits)
mm/page_alloc: clear page->private in free_pages_prepare()
selftests/mm: add memory failure dirty pagecache test
selftests/mm: add memory failure clean pagecache test
selftests/mm: add memory failure anonymous page test
mm: rmap: support batched unmapping for file large folios
arm64: mm: implement the architecture-specific clear_flush_young_ptes()
arm64: mm: support batch clearing of the young flag for large folios
arm64: mm: factor out the address and ptep alignment into a new helper
mm: rmap: support batched checks of the references for large folios
tools/testing/vma: add VMA userland tests for VMA flag functions
tools/testing/vma: separate out vma_internal.h into logical headers
tools/testing/vma: separate VMA userland tests into separate files
mm: make vm_area_desc utilise vma_flags_t only
mm: update all remaining mmap_prepare users to use vma_flags_t
mm: update shmem_[kernel]_file_*() functions to use vma_flags_t
mm: update secretmem to use VMA flags on mmap_prepare
mm: update hugetlbfs to use VMA flags on mmap_prepare
mm: add basic VMA flag operation helper functions
tools: bitmap: add missing bitmap_[subset(), andnot()]
mm: add mk_vma_flags() bitmap flag macro helper
...
Pull sysctl updates from Joel Granados:
- Remove macros from proc handler converters
Replace the proc converter macros with "regular" functions. Though it
is more verbose than the macro version, it helps when debugging and
better aligns with coding-style.rst.
- General cleanup
Remove superfluous ctl_table forward declarations. Const qualify the
memory_allocation_profiling_sysctl and loadpin_sysctl_table arrays.
Add missing kernel doc to proc_dointvec_conv.
- Testing
This series was run through sysctl selftests/kunit test suite in
x86_64. And went into linux-next after rc4, giving it a good 3 weeks
of testing
* tag 'sysctl-7.00-rc1' of git://git.kernel.org/pub/scm/linux/kernel/git/sysctl/sysctl:
sysctl: replace SYSCTL_INT_CONV_CUSTOM macro with functions
sysctl: Replace unidirectional INT converter macros with functions
sysctl: Add kernel doc to proc_douintvec_conv
sysctl: Replace UINT converter macros with functions
sysctl: Add CONFIG_PROC_SYSCTL guards for converter macros
sysctl: clarify proc_douintvec_minmax doc
sysctl: Return -ENOSYS from proc_douintvec_conv when CONFIG_PROC_SYSCTL=n
sysctl: Remove unused ctl_table forward declarations
loadpin: Implement custom proc_handler for enforce
alloc_tag: move memory_allocation_profiling_sysctls into .rodata
sysctl: Add missing kernel-doc for proc_dointvec_conv
Pull SPDX updates from Greg KH:
"Here are two small changes that add some missing SPDX license lines to
some core kernel files. These are:
- adding SPDX license lines to kdb files
- adding SPDX license lines to the remaining kernel/ files
Both of these have been in linux-next for a while with no reported
issues"
* tag 'spdx-7.0-rc1' of git://git.kernel.org/pub/scm/linux/kernel/git/gregkh/spdx:
kernel: debug: Add SPDX license ids to kdb files
kernel: add SPDX-License-Identifier lines
Pull more block updates from Jens Axboe:
- Fix partial IOVA mapping cleanup in error handling
- Minor prep series ignoring discard return value, as
the inline value is always known
- Ensure BLK_FEAT_STABLE_WRITES is set for drbd
- Fix leak of folio in bio_iov_iter_bounce_read()
- Allow IOC_PR_READ_* for read-only open
- Another debugfs deadlock fix
- A few doc updates
* tag 'block-7.0-20260216' of git://git.kernel.org/pub/scm/linux/kernel/git/axboe/linux:
blk-mq: use NOIO context to prevent deadlock during debugfs creation
blk-stat: convert struct blk_stat_callback to kernel-doc
block: fix enum descriptions kernel-doc
block: update docs for bio and bvec_iter
block: change return type to void
nvmet: ignore discard return value
md: ignore discard return value
block: fix partial IOVA mapping cleanup in blk_rq_dma_map_iova
block: fix folio leak in bio_iov_iter_bounce_read()
block: allow IOC_PR_READ_* ioctls with BLK_OPEN_READ
drbd: always set BLK_FEAT_STABLE_WRITES
Pull pidfs updates from Christian Brauner:
- pid: introduce task_ppid_vnr() helper
- pidfs: convert rb-tree to rhashtable
Mateusz reported performance penalties during task creation because
pidfs uses pidmap_lock to add elements into the rbtree. Switch to an
rhashtable to have separate fine-grained locking and to decouple from
pidmap_lock moving all heavy manipulations outside of it
Also move inode allocation outside of pidmap_lock. With this there's
nothing happening for pidfs under pidmap_lock
- pid: reorder fields in pid_namespace to reduce false sharing
- Revert "pid: make __task_pid_nr_ns(ns => NULL) safe for zombie
callers"
- ipc: Add SPDX license id to mqueue.c
* tag 'kernel-7.0-rc1.misc' of git://git.kernel.org/pub/scm/linux/kernel/git/vfs/vfs:
pid: introduce task_ppid_vnr() helper
pidfs: implement ino allocation without the pidmap lock
Revert "pid: make __task_pid_nr_ns(ns => NULL) safe for zombie callers"
pid: reorder fields in pid_namespace to reduce false sharing
pidfs: convert rb-tree to rhashtable
ipc: Add SPDX license id to mqueue.c
Creating debugfs entries can trigger fs reclaim, which can enter back
into the block layer request_queue. This can cause deadlock if the
queue is frozen.
Previously, a WARN_ON_ONCE check was used in debugfs_create_files()
to detect this condition, but it was racy since the queue can be frozen
from another context at any time.
Introduce blk_debugfs_lock()/blk_debugfs_unlock() helpers that combine
the debugfs_mutex with memalloc_noio_save()/restore() to prevent fs
reclaim from triggering block I/O. Also add blk_debugfs_lock_nomemsave()
and blk_debugfs_unlock_nomemrestore() variants for callers that don't
need NOIO protection (e.g., debugfs removal or read-only operations).
Replace all raw debugfs_mutex lock/unlock pairs with these helpers,
using the _nomemsave/_nomemrestore variants where appropriate.
Reported-by: Yi Zhang <yi.zhang@redhat.com>
Closes: https://lore.kernel.org/all/CAHj4cs9gNKEYAPagD9JADfO5UH+OiCr4P7OO2wjpfOYeM-RV=A@mail.gmail.com/
Reported-by: Shinichiro Kawasaki <shinichiro.kawasaki@wdc.com>
Closes: https://lore.kernel.org/all/aYWQR7CtYdk3K39g@shinmob/
Suggested-by: Christoph Hellwig <hch@lst.de>
Signed-off-by: Yu Kuai <yukuai@fnnas.com>
Reviewed-by: Nilay Shroff <nilay@linux.ibm.com>
Signed-off-by: Jens Axboe <axboe@kernel.dk>
Pull kprobes updates from Masami Hiramatsu:
- Use a dedicated kernel thread to optimize the kprobes instead of
using workqueue thread. Since the kprobe optimizer waits a long time
for synchronize_rcu_task(), it can block other workers in the same
queue if it uses a workqueue.
- kprobe-events: return immediately if no new probe events are
specified on the kernel command line at boot time. This shortens
the kernel boot time.
- When a kprobe is fully removed from the kernel code, retry optimizing
another kprobe which is blocked by that kprobe.
* tag 'probes-v7.0' of git://git.kernel.org/pub/scm/linux/kernel/git/trace/linux-trace:
kprobes: Use dedicated kthread for kprobe optimizer
tracing: kprobe-event: Return directly when trace kprobes is empty
kprobes: retry blocked optprobe in do_free_cleaned_kprobes
do_con_write(), fbcon_redraw.*() invoke console_conditional_schedule()
which is a conditional scheduling point based on printk's internal
variables console_may_schedule. It may only be used if the console lock
is acquired for instance via console_lock() or console_trylock().
Prinkt sets the internal variable to 1 (and allows to schedule)
if the console lock has been acquired via console_lock(). The trylock
does not allow it.
The console_conditional_schedule() invocation in do_con_write() is
invoked shortly before console_unlock().
The console_conditional_schedule() invocation in fbcon_redraw.*()
original from fbcon_scroll() / vt's con_scroll() which originate from a
line feed.
In console_unlock() the variable is set to 0 (forbids to schedule) and
it tries to schedule while making progress printing. This is brand new
compared to when console_conditional_schedule() was added in v2.4.9.11.
In v2.6.38-rc3, console_unlock() (started its existence) iterated over
all consoles and flushed them with disabled interrupts. A scheduling
attempt here was not possible, it relied that a long print scheduled
before console_unlock().
Since commit 8d91f8b153 ("printk: do cond_resched() between lines
while outputting to consoles"), which appeared in v4.5-rc1,
console_unlock() attempts to schedule if it was allowed to schedule
while during console_lock(). Each record is idealy one line so after
every line feed.
This console_conditional_schedule() is also only relevant on
PREEMPT_NONE and PREEMPT_VOLUNTARY builds. In other configurations
cond_resched() becomes a nop and has no impact.
I'm bringing this all up just proof that it is not required anymore. It
becomes a problem on a PREEMPT_RT build with debug code enabled because
that might_sleep() in cond_resched() remains and triggers a warnings.
This is due to
legacy_kthread_func-> console_flush_one_record -> vt_console_print-> lf
-> con_scroll -> fbcon_scroll
and vt_console_print() acquires a spinlock_t which does not allow a
voluntary schedule. There is no need to fb_scroll() to schedule since
console_flush_one_record() attempts to schedule after each line.
!PREEMPT_RT is not affected because the legacy printing thread is only
enabled on PREEMPT_RT builds.
Therefore I suggest to remove console_conditional_schedule().
Cc: Simona Vetter <simona@ffwll.ch>
Cc: Helge Deller <deller@gmx.de>
Cc: linux-fbdev@vger.kernel.org
Cc: dri-devel@lists.freedesktop.org
Fixes: 5f53ca3ff8 ("printk: Implement legacy printer kthread for PREEMPT_RT")
Signed-off-by: Sebastian Andrzej Siewior <bigeasy@linutronix.de>
Acked-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Acked-by: Petr Mladek <pmladek@suse.com> # from printk() POV
Signed-off-by: Helge Deller <deller@gmx.de>
Pull tracing updates from Steven Rostedt:
"User visible changes:
- Add an entry into MAINTAINERS file for RUST versions of code
There's now RUST code for tracing and static branches. To
differentiate that code from the C code, add entries in for the
RUST version (with "[RUST]" around it) so that the right
maintainers get notified on changes.
- New bitmask-list option added to tracefs
When this is set, bitmasks in trace event are not displayed as hex
numbers, but instead as lists: e.g. 0-5,7,9 instead of 0000015f
- New show_event_filters file in tracefs
Instead of having to search all events/*/*/filter for any active
filters enabled in the trace instance, the file show_event_filters
will list them so that there's only one file that needs to be
examined to see if any filters are active.
- New show_event_triggers file in tracefs
Instead of having to search all events/*/*/trigger for any active
triggers enabled in the trace instance, the file
show_event_triggers will list them so that there's only one file
that needs to be examined to see if any triggers are active.
- Have traceoff_on_warning disable trace pintk buffer too
Recently recording of trace_printk() could go to other trace
instances instead of the top level instance. But if
traceoff_on_warning triggers, it doesn't stop the buffer with
trace_printk() and that data can easily be lost by being
overwritten. Have traceoff_on_warning also disable the instance
that has trace_printk() being written to it.
- Update the hist_debug file to show what function the field uses
When CONFIG_HIST_TRIGGERS_DEBUG is enabled, a hist_debug file
exists for every event. This displays the internal data of any
histogram enabled for that event. But it is lacking the function
that is called to process one of its fields. This is very useful
information that was missing when debugging histograms.
- Up the histogram stack size from 16 to 31
Stack traces can be used as keys for event histograms. Currently
the size of the stack that is stored is limited to just 16 entries.
But the storage space in the histogram is 256 bytes, meaning that
it can store up to 31 entries (plus one for the count of entries).
Instead of letting that space go to waste, up the limit from 16 to
31. This makes the keys much more useful.
- Fix permissions of per CPU file buffer_size_kb
The per CPU file of buffer_size_kb was incorrectly set to read only
in a previous cleanup. It should be writable.
- Reset "last_boot_info" if the persistent buffer is cleared
The last_boot_info shows address information of a persistent ring
buffer if it contains data from a previous boot. It is cleared when
recording starts again, but it is not cleared when the buffer is
reset. The data is useless after a reset so clear it on reset too.
Internal changes:
- A change was made to allow tracepoint callbacks to have preemption
enabled, and instead be protected by SRCU. This required some
updates to the callbacks for perf and BPF.
perf needed to disable preemption directly in its callback because
it expects preemption disabled in the later code.
BPF needed to disable migration, as its code expects to run
completely on the same CPU.
- Have irq_work wake up other CPU if current CPU is "isolated"
When there's a waiter waiting on ring buffer data and a new event
happens, an irq work is triggered to wake up that waiter. This is
noisy on isolated CPUs (running NO_HZ_FULL). Trigger an IPI to a
house keeping CPU instead.
- Use proper free of trigger_data instead of open coding it in.
- Remove redundant call of event_trigger_reset_filter()
It was called immediately in a function that was called right after
it.
- Workqueue cleanups
- Report errors if tracing_update_buffers() were to fail.
- Make the enum update workqueue generic for other parts of tracing
On boot up, a work queue is created to convert enum names into
their numbers in the trace event format files. This work queue can
also be used for other aspects of tracing that takes some time and
shouldn't be called by the init call code.
The blk_trace initialization takes a bit of time. Have the
initialization code moved to the new tracing generic work queue
function.
- Skip kprobe boot event creation call if there's no kprobes defined
on cmdline
The kprobe initialization to set up kprobes if they are defined on
the cmdline requires taking the event_mutex lock. This can be held
by other tracing code doing initialization for a long time. Since
kprobes added to the kernel command line need to be setup
immediately, as they may be tracing early initialization code, they
cannot be postponed in a work queue and must be setup in the
initcall code.
If there's no kprobe on the kernel cmdline, there's no reason to
take the mutex and slow down the boot up code waiting to get the
lock only to find out there's nothing to do. Simply exit out early
if there's no kprobes on the kernel cmdline.
If there are kprobes on the cmdline, then someone cares more about
tracing over the speed of boot up.
- Clean up the trigger code a bit
- Move code out of trace.c and into their own files
trace.c is now over 11,000 lines of code and has become more
difficult to maintain. Start splitting it up so that related code
is in their own files.
Move all the trace_printk() related code into trace_printk.c.
Move the __always_inline stack functions into trace.h.
Move the pid filtering code into a new trace_pid.c file.
- Better define the max latency and snapshot code
The latency tracers have a "max latency" buffer that is a copy of
the main buffer and gets swapped with it when a new high latency is
detected. This keeps the trace up to the highest latency around
where this max_latency buffer is never written to. It is only used
to save the last max latency trace.
A while ago a snapshot feature was added to tracefs to allow user
space to perform the same logic. It could also enable events to
trigger a "snapshot" if one of their fields hit a new high. This
was built on top of the latency max_latency buffer logic.
Because snapshots came later, they were dependent on the latency
tracers to be enabled. In reality, the latency tracers depend on
the snapshot code and not the other way around. It was just that
they came first.
Restructure the code and the kconfigs to have the latency tracers
depend on snapshot code instead. This actually simplifies the logic
a bit and allows to disable more when the latency tracers are not
defined and the snapshot code is.
- Fix a "false sharing" in the hwlat tracer code
The loop to search for latency in hardware was using a variable
that could be changed by user space for each sample. If the user
change this variable, it could cause a bus contention, and reading
that variable can show up as a large latency in the trace causing a
false positive. Read this variable at the start of the sample with
a READ_ONCE() into a local variable and keep the code from sharing
cache lines with readers.
- Fix function graph tracer static branch optimization code
When only one tracer is defined for function graph tracing, it uses
a static branch to call that tracer directly. When another tracer
is added, it goes into loop logic to call all the registered
callbacks.
The code was incorrect when going back to one tracer and never
re-enabled the static branch again to do the optimization code.
- And other small fixes and cleanups"
* tag 'trace-v7.0' of git://git.kernel.org/pub/scm/linux/kernel/git/trace/linux-trace: (46 commits)
function_graph: Restore direct mode when callbacks drop to one
tracing: Fix indentation of return statement in print_trace_fmt()
tracing: Reset last_boot_info if ring buffer is reset
tracing: Fix to set write permission to per-cpu buffer_size_kb
tracing: Fix false sharing in hwlat get_sample()
tracing: Move d_max_latency out of CONFIG_FSNOTIFY protection
tracing: Better separate SNAPSHOT and MAX_TRACE options
tracing: Add tracer_uses_snapshot() helper to remove #ifdefs
tracing: Rename trace_array field max_buffer to snapshot_buffer
tracing: Move pid filtering into trace_pid.c
tracing: Move trace_printk functions out of trace.c and into trace_printk.c
tracing: Use system_state in trace_printk_init_buffers()
tracing: Have trace_printk functions use flags instead of using global_trace
tracing: Make tracing_update_buffers() take NULL for global_trace
tracing: Make printk_trace global for tracing system
tracing: Move ftrace_trace_stack() out of trace.c and into trace.h
tracing: Move __trace_buffer_{un}lock_*() functions to trace.h
tracing: Make tracing_selftest_running global to the tracing subsystem
tracing: Make tracing_disabled global for tracing system
tracing: Clean up use of trace_create_maxlat_file()
...
Pull dma-mapping update from Marek Szyprowski:
"A small code cleanup for the DMA-mapping subsystem: removal of unused
hooks (Robin Murphy)"
* tag 'dma-mapping-7.0-2026-02-13' of git://git.kernel.org/pub/scm/linux/kernel/git/mszyprowski/linux:
dma-mapping: Remove dma_mark_clean (again)
The add_fd_from_fd_array() function takes a file descriptor as a
parameter and tries to add either map or btf to the corresponding
list of used objects. As was reported by Dan Carpenter, since the
commit c81e4322acf0 ("bpf: Fix a potential use-after-free of BTF
object"), the fdget() is called twice on the file descriptor, and
thus userspace, potentially, can replace the file pointed to by the
file descriptor in between the two calls. On practice, this shouldn't
break anything on the kernel side, but for consistency fix the code
such that only one fdget() is executed.
Reported-by: Dan Carpenter <dan.carpenter@linaro.org>
Closes: https://lore.kernel.org/r/aY689z7gHNv8rgVO@stanley.mountain/
Fixes: ccd2d799ed ("bpf: Fix a potential use-after-free of BTF object")
Signed-off-by: Anton Protopopov <a.s.protopopov@gmail.com>
Link: https://lore.kernel.org/r/20260213212949.759321-1-a.s.protopopov@gmail.com
Signed-off-by: Alexei Starovoitov <ast@kernel.org>
