At the moment running osnoise on a nohz_full CPU or uncontested FIFO
priority and a PREEMPT_RCU kernel might have the side effect of
extending grace periods too much. This will entice RCU to force a
context switch on the wayward CPU to end the grace period, all while
introducing unwarranted noise into the tracer. This behaviour is
unavoidable as overly extending grace periods might exhaust the system's
memory.
This same exact problem is what extended quiescent states (EQS) were
created for, conversely, rcu_momentary_dyntick_idle() emulates them by
performing a zero duration EQS. So let's make use of it.
In the common case rcu_momentary_dyntick_idle() is fairly inexpensive:
atomically incrementing a local per-CPU counter and doing a store. So it
shouldn't affect osnoise's measurements (which has a 1us granularity),
so we'll call it unanimously.
The uncommon case involve calling rcu_momentary_dyntick_idle() after
having the osnoise process:
- Receive an expedited quiescent state IPI with preemption disabled or
during an RCU critical section. (activates rdp->cpu_no_qs.b.exp
code-path).
- Being preempted within in an RCU critical section and having the
subsequent outermost rcu_read_unlock() called with interrupts
disabled. (t->rcu_read_unlock_special.b.blocked code-path).
Neither of those are possible at the moment, and are unlikely to be in
the future given the osnoise's loop design. On top of this, the noise
generated by the situations described above is unavoidable, and if not
exposed by rcu_momentary_dyntick_idle() will be eventually seen in
subsequent rcu_read_unlock() calls or schedule operations.
Link: https://lkml.kernel.org/r/20220307180740.577607-1-nsaenzju@redhat.com
Cc: stable@vger.kernel.org
Fixes: bce29ac9ce ("trace: Add osnoise tracer")
Signed-off-by: Nicolas Saenz Julienne <nsaenzju@redhat.com>
Acked-by: Paul E. McKenney <paulmck@kernel.org>
Acked-by: Daniel Bristot de Oliveira <bristot@kernel.org>
Signed-off-by: Steven Rostedt (Google) <rostedt@goodmis.org>
Nicolas reported that using:
# trace-cmd record -e all -M 10 -p osnoise --poll
Resulted in the following kernel warning:
------------[ cut here ]------------
WARNING: CPU: 0 PID: 1217 at kernel/tracepoint.c:404 tracepoint_probe_unregister+0x280/0x370
[...]
CPU: 0 PID: 1217 Comm: trace-cmd Not tainted 5.17.0-rc6-next-20220307-nico+ #19
RIP: 0010:tracepoint_probe_unregister+0x280/0x370
[...]
CR2: 00007ff919b29497 CR3: 0000000109da4005 CR4: 0000000000170ef0
Call Trace:
<TASK>
osnoise_workload_stop+0x36/0x90
tracing_set_tracer+0x108/0x260
tracing_set_trace_write+0x94/0xd0
? __check_object_size.part.0+0x10a/0x150
? selinux_file_permission+0x104/0x150
vfs_write+0xb5/0x290
ksys_write+0x5f/0xe0
do_syscall_64+0x3b/0x90
entry_SYSCALL_64_after_hwframe+0x44/0xae
RIP: 0033:0x7ff919a18127
[...]
---[ end trace 0000000000000000 ]---
The warning complains about an attempt to unregister an
unregistered tracepoint.
This happens on trace-cmd because it first stops tracing, and
then switches the tracer to nop. Which is equivalent to:
# cd /sys/kernel/tracing/
# echo osnoise > current_tracer
# echo 0 > tracing_on
# echo nop > current_tracer
The osnoise tracer stops the workload when no trace instance
is actually collecting data. This can be caused both by
disabling tracing or disabling the tracer itself.
To avoid unregistering events twice, use the existing
trace_osnoise_callback_enabled variable to check if the events
(and the workload) are actually active before trying to
deactivate them.
Link: https://lore.kernel.org/all/c898d1911f7f9303b7e14726e7cc9678fbfb4a0e.camel@redhat.com/
Link: https://lkml.kernel.org/r/938765e17d5a781c2df429a98f0b2e7cc317b022.1646823913.git.bristot@kernel.org
Cc: stable@vger.kernel.org
Cc: Marcelo Tosatti <mtosatti@redhat.com>
Fixes: 2fac8d6486 ("tracing/osnoise: Allow multiple instances of the same tracer")
Reported-by: Nicolas Saenz Julienne <nsaenzju@redhat.com>
Signed-off-by: Daniel Bristot de Oliveira <bristot@kernel.org>
Signed-off-by: Steven Rostedt (Google) <rostedt@goodmis.org>
__setup() handlers should generally return 1 to indicate that the
boot options have been handled.
Using invalid option values causes the entire kernel boot option
string to be reported as Unknown and added to init's environment
strings, polluting it.
Unknown kernel command line parameters "BOOT_IMAGE=/boot/bzImage-517rc6
kprobe_event=p,syscall_any,$arg1 trace_options=quiet
trace_clock=jiffies", will be passed to user space.
Run /sbin/init as init process
with arguments:
/sbin/init
with environment:
HOME=/
TERM=linux
BOOT_IMAGE=/boot/bzImage-517rc6
kprobe_event=p,syscall_any,$arg1
trace_options=quiet
trace_clock=jiffies
Return 1 from the __setup() handlers so that init's environment is not
polluted with kernel boot options.
Link: lore.kernel.org/r/64644a2f-4a20-bab3-1e15-3b2cdd0defe3@omprussia.ru
Link: https://lkml.kernel.org/r/20220303031744.32356-1-rdunlap@infradead.org
Cc: stable@vger.kernel.org
Fixes: 7bcfaf54f5 ("tracing: Add trace_options kernel command line parameter")
Fixes: e1e232ca6b ("tracing: Add trace_clock=<clock> kernel parameter")
Fixes: 970988e19e ("tracing/kprobe: Add kprobe_event= boot parameter")
Signed-off-by: Randy Dunlap <rdunlap@infradead.org>
Reported-by: Igor Zhbanov <i.zhbanov@omprussia.ru>
Acked-by: Masami Hiramatsu <mhiramat@kernel.org>
Signed-off-by: Steven Rostedt (Google) <rostedt@goodmis.org>
When trying to add a histogram against an event with the "cpu" field, it
was impossible due to "cpu" being a keyword to key off of the running CPU.
So to fix this, it was changed to "common_cpu" to match the other generic
fields (like "common_pid"). But since some scripts used "cpu" for keying
off of the CPU (for events that did not have "cpu" as a field, which is
most of them), a backward compatibility trick was added such that if "cpu"
was used as a key, and the event did not have "cpu" as a field name, then
it would fallback and switch over to "common_cpu".
This fix has a couple of subtle bugs. One was that when switching over to
"common_cpu", it did not change the field name, it just set a flag. But
the code still found a "cpu" field. The "cpu" field is used for filtering
and is returned when the event does not have a "cpu" field.
This was found by:
# cd /sys/kernel/tracing
# echo hist:key=cpu,pid:sort=cpu > events/sched/sched_wakeup/trigger
# cat events/sched/sched_wakeup/hist
Which showed the histogram unsorted:
{ cpu: 19, pid: 1175 } hitcount: 1
{ cpu: 6, pid: 239 } hitcount: 2
{ cpu: 23, pid: 1186 } hitcount: 14
{ cpu: 12, pid: 249 } hitcount: 2
{ cpu: 3, pid: 994 } hitcount: 5
Instead of hard coding the "cpu" checks, take advantage of the fact that
trace_event_field_field() returns a special field for "cpu" and "CPU" if
the event does not have "cpu" as a field. This special field has the
"filter_type" of "FILTER_CPU". Check that to test if the returned field is
of the CPU type instead of doing the string compare.
Also, fix the sorting bug by testing for the hist_field flag of
HIST_FIELD_FL_CPU when setting up the sort routine. Otherwise it will use
the special CPU field to know what compare routine to use, and since that
special field does not have a size, it returns tracing_map_cmp_none.
Cc: stable@vger.kernel.org
Fixes: 1e3bac71c5 ("tracing/histogram: Rename "cpu" to "common_cpu"")
Reported-by: Daniel Bristot de Oliveira <bristot@kernel.org>
Signed-off-by: Steven Rostedt (Google) <rostedt@goodmis.org>
When building with clang + CONFIG_DYNAMIC_FTRACE=n + W=1, there is a
warning:
kernel/trace/ftrace.c:7194:20: error: unused function 'ftrace_startup_enable' [-Werror,-Wunused-function]
static inline void ftrace_startup_enable(int command) { }
^
1 error generated.
Clang warns on instances of static inline functions in .c files with W=1
after commit 6863f5643d ("kbuild: allow Clang to find unused static
inline functions for W=1 build").
The ftrace_startup_enable() stub has been unused since
commit e1effa0144 ("ftrace: Annotate the ops operation on update"),
where its use outside of the CONFIG_DYNAMIC_TRACE section was replaced
by ftrace_startup_all(). Remove it to resolve the warning.
Link: https://lkml.kernel.org/r/20220214192847.488166-1-nathan@kernel.org
Reported-by: kernel test robot <lkp@intel.com>
Signed-off-by: Nathan Chancellor <nathan@kernel.org>
Signed-off-by: Steven Rostedt (Google) <rostedt@goodmis.org>
On a powerpc32 build with CONFIG_CC_OPTIMISE_FOR_SIZE, the inline
keyword is not honored and trace_trigger_soft_disabled() appears
approx 50 times in vmlinux.
Adding -Winline to the build, the following message appears:
./include/linux/trace_events.h:712:1: error: inlining failed in call to 'trace_trigger_soft_disabled': call is unlikely and code size would grow [-Werror=inline]
That function is rather big for an inlined function:
c003df60 <trace_trigger_soft_disabled>:
c003df60: 94 21 ff f0 stwu r1,-16(r1)
c003df64: 7c 08 02 a6 mflr r0
c003df68: 90 01 00 14 stw r0,20(r1)
c003df6c: bf c1 00 08 stmw r30,8(r1)
c003df70: 83 e3 00 24 lwz r31,36(r3)
c003df74: 73 e9 01 00 andi. r9,r31,256
c003df78: 41 82 00 10 beq c003df88 <trace_trigger_soft_disabled+0x28>
c003df7c: 38 60 00 00 li r3,0
c003df80: 39 61 00 10 addi r11,r1,16
c003df84: 4b fd 60 ac b c0014030 <_rest32gpr_30_x>
c003df88: 73 e9 00 80 andi. r9,r31,128
c003df8c: 7c 7e 1b 78 mr r30,r3
c003df90: 41 a2 00 14 beq c003dfa4 <trace_trigger_soft_disabled+0x44>
c003df94: 38 c0 00 00 li r6,0
c003df98: 38 a0 00 00 li r5,0
c003df9c: 38 80 00 00 li r4,0
c003dfa0: 48 05 c5 f1 bl c009a590 <event_triggers_call>
c003dfa4: 73 e9 00 40 andi. r9,r31,64
c003dfa8: 40 82 00 28 bne c003dfd0 <trace_trigger_soft_disabled+0x70>
c003dfac: 73 ff 02 00 andi. r31,r31,512
c003dfb0: 41 82 ff cc beq c003df7c <trace_trigger_soft_disabled+0x1c>
c003dfb4: 80 01 00 14 lwz r0,20(r1)
c003dfb8: 83 e1 00 0c lwz r31,12(r1)
c003dfbc: 7f c3 f3 78 mr r3,r30
c003dfc0: 83 c1 00 08 lwz r30,8(r1)
c003dfc4: 7c 08 03 a6 mtlr r0
c003dfc8: 38 21 00 10 addi r1,r1,16
c003dfcc: 48 05 6f 6c b c0094f38 <trace_event_ignore_this_pid>
c003dfd0: 38 60 00 01 li r3,1
c003dfd4: 4b ff ff ac b c003df80 <trace_trigger_soft_disabled+0x20>
However it is located in a hot path so inlining it is important.
But forcing inlining of the entire function by using __always_inline
leads to increasing the text size by approx 20 kbytes.
Instead, split the fonction in two parts, one part with the likely
fast path, flagged __always_inline, and a second part out of line.
With this change, on a powerpc32 with CONFIG_CC_OPTIMISE_FOR_SIZE
vmlinux text increases by only 1,4 kbytes, which is partly
compensated by a decrease of vmlinux data by 7 kbytes.
On ppc64_defconfig which has CONFIG_CC_OPTIMISE_FOR_SPEED, this
change reduces vmlinux text by more than 30 kbytes.
Link: https://lkml.kernel.org/r/69ce0986a52d026d381d612801d978aa4f977460.1644563295.git.christophe.leroy@csgroup.eu
Signed-off-by: Christophe Leroy <christophe.leroy@csgroup.eu>
Signed-off-by: Steven Rostedt (Google) <rostedt@goodmis.org>
Currently, the event probes save the type of the event they are attached
to when recording the event. For example:
# echo 'e:switch sched/sched_switch prev_state=$prev_state prev_prio=$prev_prio next_pid=$next_pid next_prio=$next_prio' > dynamic_events
# cat events/eprobes/switch/format
name: switch
ID: 1717
format:
field:unsigned short common_type; offset:0; size:2; signed:0;
field:unsigned char common_flags; offset:2; size:1; signed:0;
field:unsigned char common_preempt_count; offset:3; size:1; signed:0;
field:int common_pid; offset:4; size:4; signed:1;
field:unsigned int __probe_type; offset:8; size:4; signed:0;
field:u64 prev_state; offset:12; size:8; signed:0;
field:u64 prev_prio; offset:20; size:8; signed:0;
field:u64 next_pid; offset:28; size:8; signed:0;
field:u64 next_prio; offset:36; size:8; signed:0;
print fmt: "(%u) prev_state=0x%Lx prev_prio=0x%Lx next_pid=0x%Lx next_prio=0x%Lx", REC->__probe_type, REC->prev_state, REC->prev_prio, REC->next_pid, REC->next_prio
The __probe_type adds 4 bytes to every event.
One of the reasons for creating eprobes is to limit what is traced in an
event to be able to limit what is written into the ring buffer. Having
this redundant 4 bytes to every event takes away from this.
The event that is recorded can be retrieved from the event probe itself,
that is available when the trace is happening. For user space tools, it
could simply read the dynamic_event file to find the event they are for.
So there is really no reason to write this information into the ring
buffer for every event.
Link: https://lkml.kernel.org/r/20220218190057.2f5a19a8@gandalf.local.home
Acked-by: Masami Hiramatsu <mhiramat@kernel.org>
Reviewed-by: Joel Fernandes <joel@joelfernandes.org>
Signed-off-by: Steven Rostedt (Google) <rostedt@goodmis.org>
The kernel parameter "tp_printk_stop_on_boot" starts with "tp_printk" which is
the same as another kernel parameter "tp_printk". If "tp_printk" setup is
called before the "tp_printk_stop_on_boot", it will override the latter
and keep it from being set.
This is similar to other kernel parameter issues, such as:
Commit 745a600cf1 ("um: console: Ignore console= option")
or init/do_mounts.c:45 (setup function of "ro" kernel param)
Fix it by checking for a "_" right after the "tp_printk" and if that
exists do not process the parameter.
Link: https://lkml.kernel.org/r/20220208195421.969326-1-jsyoo5b@gmail.com
Signed-off-by: JaeSang Yoo <jsyoo5b@gmail.com>
[ Fixed up change log and added space after if condition ]
Signed-off-by: Steven Rostedt (Google) <rostedt@goodmis.org>
The patch ec5ce09875: "tracing: Allow whitespace to surround hist
trigger filter" from Jan 15, 2018, leads to the following Smatch
static checker warning:
kernel/trace/trace_events_hist.c:6199 event_hist_trigger_parse()
warn: 'p' can't be NULL.
Since p is always checked for a NULL value at the top of loop and
nothing in the rest of the loop will set it to NULL, the warning
is correct and might as well be 1 to silence the warning.
Link: https://lkml.kernel.org/r/a1d4c79766c0cf61e20438dc35244d216633fef6.1643319703.git.zanussi@kernel.org
Fixes: ec5ce09875 ("tracing: Allow whitespace to surround hist trigger filter")
Reported-by: kernel test robot <lkp@intel.com>
Reported-by: Dan Carpenter <dan.carpenter@oracle.com>
Signed-off-by: Tom Zanussi <zanussi@kernel.org>
Signed-off-by: Steven Rostedt (Google) <rostedt@goodmis.org>
The recent rename of event_hist_trigger_parse() caused smatch
re-evaluation of trace_events_hist.c and as a result an old warning
was found:
kernel/trace/trace_events_hist.c:6174 event_hist_trigger_parse()
error: we previously assumed 'glob' could be null (see line 6166)
glob should never be null (and apparently smatch can also figure that
out and skip the warning when using the cross-function DB (but which
can't be used with a 0day build as it takes too much time to
generate)).
Nonetheless for clarity, remove the test but add a WARN_ON() in case
the code ever changes.
Link: https://lkml.kernel.org/r/96925e5c1f116654ada7ea0613d930b1266b5e1c.1643319703.git.zanussi@kernel.org
Fixes: f404da6e1d ("tracing: Add 'last error' error facility for hist triggers")
Reported-by: kernel test robot <lkp@intel.com>
Reported-by: Dan Carpenter <dan.carpenter@oracle.com>
Signed-off-by: Tom Zanussi <zanussi@kernel.org>
Signed-off-by: Steven Rostedt (Google) <rostedt@goodmis.org>
First S390 complained that the sorting of the mcount sections at build
time caused the kernel to crash on their architecture. Now PowerPC is
complaining about it too. And also ARM64 appears to be having issues.
It may be necessary to also update the relocation table for the values
in the mcount table. Not only do we have to sort the table, but also
update the relocations that may be applied to the items in the table.
If the system is not relocatable, then it is fine to sort, but if it is,
some architectures may have issues (although x86 does not as it shifts all
addresses the same).
Add a HAVE_BUILDTIME_MCOUNT_SORT that an architecture can set to say it is
safe to do the sorting at build time.
Also update the config to compile in build time sorting in the sorttable
code in scripts/ to depend on CONFIG_BUILDTIME_MCOUNT_SORT.
Link: https://lore.kernel.org/all/944D10DA-8200-4BA9-8D0A-3BED9AA99F82@linux.ibm.com/
Link: https://lkml.kernel.org/r/20220127153821.3bc1ac6e@gandalf.local.home
Cc: Ingo Molnar <mingo@kernel.org>
Cc: Andrew Morton <akpm@linux-foundation.org>
Cc: Russell King <linux@armlinux.org.uk>
Cc: Yinan Liu <yinan@linux.alibaba.com>
Cc: Ard Biesheuvel <ardb@kernel.org>
Cc: Kees Cook <keescook@chromium.org>
Reported-by: Sachin Sant <sachinp@linux.ibm.com>
Reviewed-by: Mark Rutland <mark.rutland@arm.com>
Tested-by: Mark Rutland <mark.rutland@arm.com> [arm64]
Tested-by: Sachin Sant <sachinp@linux.ibm.com>
Fixes: 72b3942a17 ("scripts: ftrace - move the sort-processing in ftrace_init")
Signed-off-by: Steven Rostedt (Google) <rostedt@goodmis.org>
Pull ftrace fix from Steven Rostedt:
"Fix s390 breakage from sorting mcount tables.
The latest merge of the tracing tree sorts the mcount table at build
time. But s390 appears to do things differently (like always) and
replaces the sorted table back to the original unsorted one. As the
ftrace algorithm depends on it being sorted, bad things happen when it
is not, and s390 experienced those bad things.
Add a new config to tell the boot if the mcount table is sorted or
not, and allow s390 to opt out of it"
* tag 'trace-v5.17-3' of git://git.kernel.org/pub/scm/linux/kernel/git/rostedt/linux-trace:
ftrace: Fix assuming build time sort works for s390
To speed up the boot process, as mcount_loc needs to be sorted for ftrace
to work properly, sorting it at build time is more efficient than boot up
and can save milliseconds of time. Unfortunately, this change broke s390
as it will modify the mcount_loc location after the sorting takes place
and will put back the unsorted locations. Since the sorting is skipped at
boot up if it is believed that it was sorted at run time, ftrace can crash
as its algorithms are dependent on the list being sorted.
Add a new config BUILDTIME_MCOUNT_SORT that is set when
BUILDTIME_TABLE_SORT but not if S390 is set. Use this config to determine
if sorting should take place at boot up.
Link: https://lore.kernel.org/all/yt9dee51ctfn.fsf@linux.ibm.com/
Fixes: 72b3942a17 ("scripts: ftrace - move the sort-processing in ftrace_init")
Reported-by: Sven Schnelle <svens@linux.ibm.com>
Tested-by: Heiko Carstens <hca@linux.ibm.com>
Signed-off-by: Steven Rostedt (Google) <rostedt@goodmis.org>
Pull bitmap updates from Yury Norov:
- introduce for_each_set_bitrange()
- use find_first_*_bit() instead of find_next_*_bit() where possible
- unify for_each_bit() macros
* tag 'bitmap-5.17-rc1' of git://github.com/norov/linux:
vsprintf: rework bitmap_list_string
lib: bitmap: add performance test for bitmap_print_to_pagebuf
bitmap: unify find_bit operations
mm/percpu: micro-optimize pcpu_is_populated()
Replace for_each_*_bit_from() with for_each_*_bit() where appropriate
find: micro-optimize for_each_{set,clear}_bit()
include/linux: move for_each_bit() macros from bitops.h to find.h
cpumask: replace cpumask_next_* with cpumask_first_* where appropriate
tools: sync tools/bitmap with mother linux
all: replace find_next{,_zero}_bit with find_first{,_zero}_bit where appropriate
cpumask: use find_first_and_bit()
lib: add find_first_and_bit()
arch: remove GENERIC_FIND_FIRST_BIT entirely
include: move find.h from asm_generic to linux
bitops: move find_bit_*_le functions from le.h to find.h
bitops: protect find_first_{,zero}_bit properly
When we pass a negative value to the proc_doulongvec_minmax() function,
the function returns 0, but the corresponding interface value does not
change.
we can easily reproduce this problem with the following commands:
cd /proc/sys/fs/epoll
echo -1 > max_user_watches; echo $?; cat max_user_watches
This function requires a non-negative number to be passed in, so when a
negative number is passed in, -EINVAL is returned.
Link: https://lkml.kernel.org/r/20211220092627.3744624-1-libaokun1@huawei.com
Signed-off-by: Baokun Li <libaokun1@huawei.com>
Reported-by: Hulk Robot <hulkci@huawei.com>
Acked-by: Luis Chamberlain <mcgrof@kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
