linux-2.6-microblaze.git
5 months agotracing/boot: Add per-event histogram action options
Masami Hiramatsu [Tue, 10 Aug 2021 02:07:21 +0000 (11:07 +0900)]
tracing/boot: Add per-event histogram action options

Add a hist-trigger action syntax support to boot-time tracing.
Currently, boot-time tracing supports per-event actions as option
strings. However, for the histogram action, it has a special syntax
and usually needs a long action definition.
To make it readable and fit to the bootconfig syntax, this introduces
a new options for histogram.

Here are the histogram action options for boot-time tracing.

ftrace.[instance.INSTANCE.]event.GROUP.EVENT.hist {
     keys = <KEY>[,...]
     values = <VAL>[,...]
     sort = <SORT-KEY>[,...]
     size = <ENTRIES>
     name = <HISTNAME>
     var { <VAR> = <EXPR> ... }
     pause|continue|clear
     onmax|onchange { var = <VAR>; <ACTION> [= <PARAM>] }
     onmatch { event = <EVENT>; <ACTION> [= <PARAM>] }
     filter = <FILTER>
}

Where <ACTION> is one of below;

     trace = <EVENT>, <ARG1>[, ...]
     save = <ARG1>[, ...]
     snapshot

Link: https://lkml.kernel.org/r/162856124106.203126.10501871028479029087.stgit@devnote2
Signed-off-by: Masami Hiramatsu <mhiramat@kernel.org>
Signed-off-by: Steven Rostedt (VMware) <rostedt@goodmis.org>
5 months agotracing: Fix a typo in tracepoint.h
Huang Shijie [Mon, 2 Aug 2021 14:02:34 +0000 (14:02 +0000)]
tracing: Fix a typo in tracepoint.h

It should be @prev_pid, not @prev_prid.

Link: https://lkml.kernel.org/r/20210802140234.5383-1-shijie@os.amperecomputing.com
Signed-off-by: Huang Shijie <shijie@os.amperecomputing.com>
Signed-off-by: Steven Rostedt (VMware) <rostedt@goodmis.org>
5 months agotracing: Refactor TRACE_IRQFLAGS_SUPPORT in Kconfig
Masahiro Yamada [Sat, 31 Jul 2021 05:22:32 +0000 (14:22 +0900)]
tracing: Refactor TRACE_IRQFLAGS_SUPPORT in Kconfig

Make architectures select TRACE_IRQFLAGS_SUPPORT instead of
having many defines.

Link: https://lkml.kernel.org/r/20210731052233.4703-2-masahiroy@kernel.org
Acked-by: Heiko Carstens <hca@linux.ibm.com>
Acked-by: Vineet Gupta <vgupta@synopsys.com>   #arch/arc
Acked-by: Michael Ellerman <mpe@ellerman.id.au> (powerpc)
Acked-by: Catalin Marinas <catalin.marinas@arm.com>
Acked-by: Max Filippov <jcmvbkbc@gmail.com>
Signed-off-by: Masahiro Yamada <masahiroy@kernel.org>
Signed-off-by: Steven Rostedt (VMware) <rostedt@goodmis.org>
5 months agotracing: Simplify the Kconfig dependency of FTRACE
Masahiro Yamada [Sat, 31 Jul 2021 05:22:31 +0000 (14:22 +0900)]
tracing: Simplify the Kconfig dependency of FTRACE

The entire FTRACE block is surrounded by 'if TRACING_SUPPORT' ...
'endif'.

Using 'depends on' is a simpler way to guard FTRACE.

Link: https://lkml.kernel.org/r/20210731052233.4703-1-masahiroy@kernel.org
Signed-off-by: Masahiro Yamada <masahiroy@kernel.org>
Signed-off-by: Steven Rostedt (VMware) <rostedt@goodmis.org>
5 months agotracing: Allow execnames to be passed as args for synthetic events
Steven Rostedt (VMware) [Thu, 22 Jul 2021 14:27:07 +0000 (10:27 -0400)]
tracing: Allow execnames to be passed as args for synthetic events

Allow common_pid.execname to be saved in a variable in one histogram to be
passed to another histogram that can pass it as a parameter to a synthetic
event.

 ># echo 'hist:keys=pid:__arg__1=common_timestamp.usecs:arg2=common_pid.execname' \
       > events/sched/sched_waking/trigger
 ># echo 'wakeup_lat s32 pid; u64 delta; char wake_comm[]' > synthetic_events
 ># echo 'hist:keys=next_pid:pid=next_pid,delta=common_timestamp.usecs-$__arg__1,exec=$arg2'\
':onmatch(sched.sched_waking).trace(wakeup_lat,$pid,$delta,$exec)' \
 > events/sched/sched_switch/trigger

The above is a wake up latency synthetic event setup that passes the execname
of the common_pid that woke the task to the scheduling of that task, which
triggers a synthetic event that passes the original execname as a
parameter to display it.

 ># echo 1 > events/synthetic/enable
 ># cat trace
    <idle>-0       [006] d..4   186.863801: wakeup_lat: pid=1306 delta=65 wake_comm=kworker/u16:3
    <idle>-0       [000] d..4   186.863858: wakeup_lat: pid=163 delta=27 wake_comm=<idle>
    <idle>-0       [001] d..4   186.863903: wakeup_lat: pid=1307 delta=36 wake_comm=kworker/u16:4
    <idle>-0       [000] d..4   186.863927: wakeup_lat: pid=163 delta=5 wake_comm=<idle>
    <idle>-0       [006] d..4   186.863957: wakeup_lat: pid=1306 delta=24 wake_comm=kworker/u16:3
      sshd-1306    [006] d..4   186.864051: wakeup_lat: pid=61 delta=62 wake_comm=<idle>
    <idle>-0       [000] d..4   186.965030: wakeup_lat: pid=609 delta=18 wake_comm=<idle>
    <idle>-0       [006] d..4   186.987582: wakeup_lat: pid=1306 delta=65 wake_comm=kworker/u16:3
    <idle>-0       [000] d..4   186.987639: wakeup_lat: pid=163 delta=27 wake_comm=<idle>

Link: https://lkml.kernel.org/r/20210722142837.458596338@goodmis.org
Reviewed-by: Tom Zanussi <zanussi@kernel.org>
Reviewed-by: Masami Hiramatsu <mhiramat@kernel.org>
Signed-off-by: Steven Rostedt (VMware) <rostedt@goodmis.org>
5 months agotracing: Have histogram types be constant when possible
Steven Rostedt (VMware) [Thu, 22 Jul 2021 14:27:06 +0000 (10:27 -0400)]
tracing: Have histogram types be constant when possible

Instead of kstrdup("const", GFP_KERNEL), have the hist_field type simply
assign the constant hist_field->type = "const"; And when the value passed
to it is a variable, use "kstrdup_const(var, GFP_KERNEL);" which will just
copy the value if the variable is already a constant. This saves on having
to allocate when not needed.

All frees of the hist_field->type will need to use kfree_const().

Link: https://lkml.kernel.org/r/20210722142837.280718447@goodmis.org
Suggested-by: Masami Hiramatsu <mhiramat@kernel.org>
Reviewed-by: Masami Hiramatsu <mhiramat@kernel.org>
Signed-off-by: Steven Rostedt (VMware) <rostedt@goodmis.org>
5 months agotracing/histogram: Update the documentation for the buckets modifier
Steven Rostedt (VMware) [Wed, 7 Jul 2021 21:36:25 +0000 (17:36 -0400)]
tracing/histogram: Update the documentation for the buckets modifier

Update both the tracefs README file as well as the histogram.rst to
include an explanation of what the buckets modifier is and how to use it.
Include an example with the wakeup_latency example for both log2 and the
buckets modifiers as there was no existing log2 example.

Link: https://lkml.kernel.org/r/20210707213922.167218794@goodmis.org
Acked-by: Namhyung Kim <namhyung@kernel.org>
Reviewed-by: Masami Hiramatsu <mhiramat@kernel.org>
Signed-off-by: Steven Rostedt (VMware) <rostedt@goodmis.org>
5 months agotracing: Add linear buckets to histogram logic
Steven Rostedt (VMware) [Wed, 7 Jul 2021 21:36:24 +0000 (17:36 -0400)]
tracing: Add linear buckets to histogram logic

There's been several times I wished the histogram logic had a "grouping"
feature for the buckets. Currently, each bucket has a size of one. That
is, if you trace the amount of requested allocations, each allocation is
its own bucket, even if you are interested in what allocates 100 bytes or
less, 100 to 200, 200 to 300, etc.

Also, without grouping, it fills up the allocated histogram buckets
quickly. If you are tracking latency, and don't care if something is 200
microseconds off, or 201 microseconds off, but want to track them by say
10 microseconds each. This can not currently be done.

There is a log2 but that grouping get's too big too fast for a lot of
cases.

Introduce a "buckets=SIZE" command to each field where it will record in a
rounded number. For example:

 ># echo 'hist:keys=bytes_req.buckets=100:sort=bytes_req' > events/kmem/kmalloc/trigger
 ># cat events/kmem/kmalloc/hist
 # event histogram
 #
 # trigger info:
 hist:keys=bytes_req.buckets=100:vals=hitcount:sort=bytes_req.buckets=100:size=2048
 [active]
 #

 { bytes_req: ~ 0-99 } hitcount:       3149
 { bytes_req: ~ 100-199 } hitcount:       1468
 { bytes_req: ~ 200-299 } hitcount:         39
 { bytes_req: ~ 300-399 } hitcount:        306
 { bytes_req: ~ 400-499 } hitcount:        364
 { bytes_req: ~ 500-599 } hitcount:         32
 { bytes_req: ~ 600-699 } hitcount:         69
 { bytes_req: ~ 700-799 } hitcount:         37
 { bytes_req: ~ 1200-1299 } hitcount:         16
 { bytes_req: ~ 1400-1499 } hitcount:         30
 { bytes_req: ~ 2000-2099 } hitcount:          6
 { bytes_req: ~ 4000-4099 } hitcount:       2168
 { bytes_req: ~ 5000-5099 } hitcount:          6

 Totals:
     Hits: 7690
     Entries: 13
     Dropped: 0

Link: https://lkml.kernel.org/r/20210707213921.980359719@goodmis.org
Acked-by: Namhyung Kim <namhyung@kernel.org>
Reviewed-by: Tom Zanussi <zanussi@kernel.org>
Reviewed-by: Masami Hiramatsu <mhiramat@kernel.org>
Tested-by: Daniel Bristot de Oliveira <bristot@kernel.org>
Signed-off-by: Steven Rostedt (VMware) <rostedt@goodmis.org>
5 months agotracing/boot: Fix a hist trigger dependency for boot time tracing
Masami Hiramatsu [Tue, 10 Aug 2021 02:07:14 +0000 (11:07 +0900)]
tracing/boot: Fix a hist trigger dependency for boot time tracing

Fixes a build error when CONFIG_HIST_TRIGGERS=n with boot-time
tracing. Since the trigger_process_regex() is defined only
when CONFIG_HIST_TRIGGERS=y, if it is disabled, the 'actions'
event option also must be disabled.

Link: https://lkml.kernel.org/r/162856123376.203126.582144262622247352.stgit@devnote2
Fixes: 81a59555ff15 ("tracing/boot: Add per-event settings")
Signed-off-by: Masami Hiramatsu <mhiramat@kernel.org>
Signed-off-by: Steven Rostedt (VMware) <rostedt@goodmis.org>
5 months agotracing: Apply trace filters on all output channels
Pingfan Liu [Sat, 14 Aug 2021 03:45:38 +0000 (11:45 +0800)]
tracing: Apply trace filters on all output channels

The event filters are not applied on all of the output, which results in
the flood of printk when using tp_printk. Unfolding
event_trigger_unlock_commit_regs() into trace_event_buffer_commit(), so
the filters can be applied on every output.

Link: https://lkml.kernel.org/r/20210814034538.8428-1-kernelfans@gmail.com
Cc: stable@vger.kernel.org
Fixes: 0daa2302968c1 ("tracing: Add tp_printk cmdline to have tracepoints go to printk()")
Signed-off-by: Pingfan Liu <kernelfans@gmail.com>
Signed-off-by: Steven Rostedt (VMware) <rostedt@goodmis.org>
5 months agotracing / histogram: Fix NULL pointer dereference on strcmp() on NULL event name
Steven Rostedt (VMware) [Sun, 8 Aug 2021 04:30:11 +0000 (00:30 -0400)]
tracing / histogram: Fix NULL pointer dereference on strcmp() on NULL event name

The following commands:

 # echo 'read_max u64 size;' > synthetic_events
 # echo 'hist:keys=common_pid:count=count:onmax($count).trace(read_max,count)' > events/syscalls/sys_enter_read/trigger

Causes:

 BUG: kernel NULL pointer dereference, address: 0000000000000000
 #PF: supervisor read access in kernel mode
 #PF: error_code(0x0000) - not-present page
 PGD 0 P4D 0
 Oops: 0000 [#1] PREEMPT SMP
 CPU: 4 PID: 1763 Comm: bash Not tainted 5.14.0-rc2-test+ #155
 Hardware name: Hewlett-Packard HP Compaq Pro 6300 SFF/339A, BIOS K01
v03.03 07/14/2016
 RIP: 0010:strcmp+0xc/0x20
 Code: 75 f7 31 c0 0f b6 0c 06 88 0c 02 48 83 c0 01 84 c9 75 f1 4c 89 c0
c3 0f 1f 80 00 00 00 00 31 c0 eb 08 48 83 c0 01 84 d2 74 0f <0f> b6 14 07
3a 14 06 74 ef 19 c0 83 c8 01 c3 31 c0 c3 66 90 48 89
 RSP: 0018:ffffb5fdc0963ca8 EFLAGS: 00010246
 RAX: 0000000000000000 RBX: ffffffffb3a4e040 RCX: 0000000000000000
 RDX: 0000000000000000 RSI: ffff9714c0d0b640 RDI: 0000000000000000
 RBP: 0000000000000000 R08: 00000022986b7cde R09: ffffffffb3a4dff8
 R10: 0000000000000000 R11: 0000000000000000 R12: ffff9714c50603c8
 R13: 0000000000000000 R14: ffff97143fdf9e48 R15: ffff9714c01a2210
 FS:  00007f1fa6785740(0000) GS:ffff9714da400000(0000)
knlGS:0000000000000000
 CS:  0010 DS: 0000 ES: 0000 CR0: 0000000080050033
 CR2: 0000000000000000 CR3: 000000002d863004 CR4: 00000000001706e0
 Call Trace:
  __find_event_file+0x4e/0x80
  action_create+0x6b7/0xeb0
  ? kstrdup+0x44/0x60
  event_hist_trigger_func+0x1a07/0x2130
  trigger_process_regex+0xbd/0x110
  event_trigger_write+0x71/0xd0
  vfs_write+0xe9/0x310
  ksys_write+0x68/0xe0
  do_syscall_64+0x3b/0x90
  entry_SYSCALL_64_after_hwframe+0x44/0xae
 RIP: 0033:0x7f1fa6879e87

The problem was the "trace(read_max,count)" where the "count" should be
"$count" as "onmax()" only handles variables (although it really should be
able to figure out that "count" is a field of sys_enter_read). But there's
a path that does not find the variable and ends up passing a NULL for the
event, which ends up getting passed to "strcmp()".

Add a check for NULL to return and error on the command with:

 # cat error_log
  hist:syscalls:sys_enter_read: error: Couldn't create or find variable
  Command: hist:keys=common_pid:count=count:onmax($count).trace(read_max,count)
                                ^
Link: https://lkml.kernel.org/r/20210808003011.4037f8d0@oasis.local.home
Cc: Masami Hiramatsu <mhiramat@kernel.org>
Cc: stable@vger.kernel.org
Fixes: 50450603ec9cb tracing: Add 'onmax' hist trigger action support
Reviewed-by: Tom Zanussi <zanussi@kernel.org>
Signed-off-by: Steven Rostedt (VMware) <rostedt@goodmis.org>
5 months agoinit: Suppress wrong warning for bootconfig cmdline parameter
Masami Hiramatsu [Thu, 5 Aug 2021 02:10:51 +0000 (11:10 +0900)]
init: Suppress wrong warning for bootconfig cmdline parameter

Since the 'bootconfig' command line parameter is handled before
parsing the command line, it doesn't use early_param(). But in
this case, kernel shows a wrong warning message about it.

[    0.013714] Kernel command line: ro console=ttyS0  bootconfig console=tty0
[    0.013741] Unknown command line parameters: bootconfig

To suppress this message, add a dummy handler for 'bootconfig'.

Link: https://lkml.kernel.org/r/162812945097.77369.1849780946468010448.stgit@devnote2
Fixes: 86d1919a4fb0 ("init: print out unknown kernel parameters")
Reviewed-by: Andrew Halaney <ahalaney@redhat.com>
Signed-off-by: Masami Hiramatsu <mhiramat@kernel.org>
Signed-off-by: Steven Rostedt (VMware) <rostedt@goodmis.org>
5 months agotracing: define needed config DYNAMIC_FTRACE_WITH_ARGS
Lukas Bulwahn [Fri, 6 Aug 2021 19:50:27 +0000 (21:50 +0200)]
tracing: define needed config DYNAMIC_FTRACE_WITH_ARGS

Commit 2860cd8a2353 ("livepatch: Use the default ftrace_ops instead of
REGS when ARGS is available") intends to enable config LIVEPATCH when
ftrace with ARGS is available. However, the chain of configs to enable
LIVEPATCH is incomplete, as HAVE_DYNAMIC_FTRACE_WITH_ARGS is available,
but the definition of DYNAMIC_FTRACE_WITH_ARGS, combining DYNAMIC_FTRACE
and HAVE_DYNAMIC_FTRACE_WITH_ARGS, needed to enable LIVEPATCH, is missing
in the commit.

Fortunately, ./scripts/checkkconfigsymbols.py detects this and warns:

DYNAMIC_FTRACE_WITH_ARGS
Referencing files: kernel/livepatch/Kconfig

So, define the config DYNAMIC_FTRACE_WITH_ARGS analogously to the already
existing similar configs, DYNAMIC_FTRACE_WITH_REGS and
DYNAMIC_FTRACE_WITH_DIRECT_CALLS, in ./kernel/trace/Kconfig to connect the
chain of configs.

Link: https://lore.kernel.org/kernel-janitors/CAKXUXMwT2zS9fgyQHKUUiqo8ynZBdx2UEUu1WnV_q0OCmknqhw@mail.gmail.com/
Link: https://lkml.kernel.org/r/20210806195027.16808-1-lukas.bulwahn@gmail.com
Cc: Josh Poimboeuf <jpoimboe@redhat.com>
Cc: Jiri Kosina <jikos@kernel.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Miroslav Benes <mbenes@suse.cz>
Cc: stable@vger.kernel.org
Fixes: 2860cd8a2353 ("livepatch: Use the default ftrace_ops instead of REGS when ARGS is available")
Signed-off-by: Lukas Bulwahn <lukas.bulwahn@gmail.com>
Signed-off-by: Steven Rostedt (VMware) <rostedt@goodmis.org>
5 months agotrace/osnoise: Print a stop tracing message
Daniel Bristot de Oliveira [Sun, 18 Jul 2021 09:07:55 +0000 (11:07 +0200)]
trace/osnoise: Print a stop tracing message

When using osnoise/timerlat with stop tracing, sometimes it is
not clear in which CPU the stop condition was hit, mainly
when using some extra events.

Print a message informing in which CPU the trace stopped, like
in the example below:

          <idle>-0       [006] d.h.  2932.676616: #1672599 context    irq timer_latency     34689 ns
          <idle>-0       [006] dNh.  2932.676618: irq_noise: local_timer:236 start 2932.676615639 duration 2391 ns
          <idle>-0       [006] dNh.  2932.676620: irq_noise: virtio0-output.0:47 start 2932.676620180 duration 86 ns
          <idle>-0       [003] d.h.  2932.676621: #1673374 context    irq timer_latency      1200 ns
          <idle>-0       [006] d...  2932.676623: thread_noise: swapper/6:0 start 2932.676615964 duration 4339 ns
          <idle>-0       [003] dNh.  2932.676623: irq_noise: local_timer:236 start 2932.676620597 duration 1881 ns
          <idle>-0       [006] d...  2932.676623: sched_switch: prev_comm=swapper/6 prev_pid=0 prev_prio=120 prev_state=R ==> next_comm=timerlat/6 next_pid=852 next_prio=4
      timerlat/6-852     [006] ....  2932.676623: #1672599 context thread timer_latency     41931 ns
          <idle>-0       [003] d...  2932.676623: thread_noise: swapper/3:0 start 2932.676620854 duration 880 ns
          <idle>-0       [003] d...  2932.676624: sched_switch: prev_comm=swapper/3 prev_pid=0 prev_prio=120 prev_state=R ==> next_comm=timerlat/3 next_pid=849 next_prio=4
      timerlat/6-852     [006] ....  2932.676624: timerlat_main: stop tracing hit on cpu 6
      timerlat/3-849     [003] ....  2932.676624: #1673374 context thread timer_latency      4310 ns

Link: https://lkml.kernel.org/r/b30a0d7542adba019185f44ee648e60e14923b11.1626598844.git.bristot@kernel.org
Cc: Tom Zanussi <zanussi@kernel.org>
Cc: Namhyung Kim <namhyung@kernel.org>
Cc: Masami Hiramatsu <mhiramat@kernel.org>
Signed-off-by: Daniel Bristot de Oliveira <bristot@kernel.org>
Signed-off-by: Steven Rostedt (VMware) <rostedt@goodmis.org>
5 months agotrace/timerlat: Add a header with PREEMPT_RT additional fields
Daniel Bristot de Oliveira [Sun, 18 Jul 2021 09:07:54 +0000 (11:07 +0200)]
trace/timerlat: Add a header with PREEMPT_RT additional fields

Some extra flags are printed to the trace header when using the
PREEMPT_RT config. The extra flags are: need-resched-lazy,
preempt-lazy-depth, and migrate-disable.

Without printing these fields, the timerlat specific fields are
shifted by three positions, for example:

 # tracer: timerlat
 #
 #                                _-----=> irqs-off
 #                               / _----=> need-resched
 #                              | / _---=> hardirq/softirq
 #                              || / _--=> preempt-depth
 #                              || /
 #                              ||||             ACTIVATION
 #           TASK-PID      CPU# ||||   TIMESTAMP    ID            CONTEXT                LATENCY
 #              | |         |   ||||      |         |                  |                       |
           <idle>-0       [000] d..h...  3279.798871: #1     context    irq timer_latency       830 ns
            <...>-807     [000] .......  3279.798881: #1     context thread timer_latency     11301 ns

Add a new header for timerlat with the missing fields, to be used
when the PREEMPT_RT is enabled.

Link: https://lkml.kernel.org/r/babb83529a3211bd0805be0b8c21608230202c55.1626598844.git.bristot@kernel.org
Cc: Tom Zanussi <zanussi@kernel.org>
Cc: Namhyung Kim <namhyung@kernel.org>
Cc: Masami Hiramatsu <mhiramat@kernel.org>
Signed-off-by: Daniel Bristot de Oliveira <bristot@kernel.org>
Signed-off-by: Steven Rostedt (VMware) <rostedt@goodmis.org>
5 months agotrace/osnoise: Add a header with PREEMPT_RT additional fields
Daniel Bristot de Oliveira [Sun, 18 Jul 2021 09:07:53 +0000 (11:07 +0200)]
trace/osnoise: Add a header with PREEMPT_RT additional fields

Some extra flags are printed to the trace header when using the
PREEMPT_RT config. The extra flags are: need-resched-lazy,
preempt-lazy-depth, and migrate-disable.

Without printing these fields, the osnoise specific fields are
shifted by three positions, for example:

 # tracer: osnoise
 #
 #                                _-----=> irqs-off
 #                               / _----=> need-resched
 #                              | / _---=> hardirq/softirq
 #                              || / _--=> preempt-depth                            MAX
 #                              || /                                             SINGLE      Interference counters:
 #                              ||||               RUNTIME      NOISE  %% OF CPU  NOISE    +-----------------------------+
 #           TASK-PID      CPU# ||||   TIMESTAMP    IN US       IN US  AVAILABLE  IN US     HW    NMI    IRQ   SIRQ THREAD
 #              | |         |   ||||      |           |             |    |            |      |      |      |      |      |
            <...>-741     [000] .......  1105.690909: 1000000        234  99.97660      36     21      0   1001     22      3
            <...>-742     [001] .......  1105.691923: 1000000        281  99.97190     197      7      0   1012     35     14
            <...>-743     [002] .......  1105.691958: 1000000       1324  99.86760     118     11      0   1016    155    143
            <...>-744     [003] .......  1105.691998: 1000000        109  99.98910      21      4      0   1004     33      7
            <...>-745     [004] .......  1105.692015: 1000000       2023  99.79770      97     37      0   1023     52     18

Add a new header for osnoise with the missing fields, to be used
when the PREEMPT_RT is enabled.

Link: https://lkml.kernel.org/r/1f03289d2a51fde5a58c2e7def063dc630820ad1.1626598844.git.bristot@kernel.org
Cc: Tom Zanussi <zanussi@kernel.org>
Cc: Namhyung Kim <namhyung@kernel.org>
Cc: Masami Hiramatsu <mhiramat@kernel.org>
Signed-off-by: Daniel Bristot de Oliveira <bristot@kernel.org>
Signed-off-by: Steven Rostedt (VMware) <rostedt@goodmis.org>
5 months agotracepoint: Use rcu get state and cond sync for static call updates
Mathieu Desnoyers [Thu, 5 Aug 2021 19:29:54 +0000 (15:29 -0400)]
tracepoint: Use rcu get state and cond sync for static call updates

State transitions from 1->0->1 and N->2->1 callbacks require RCU
synchronization. Rather than performing the RCU synchronization every
time the state change occurs, which is quite slow when many tracepoints
are registered in batch, instead keep a snapshot of the RCU state on the
most recent transitions which belong to a chain, and conditionally wait
for a grace period on the last transition of the chain if one g.p. has
not elapsed since the last snapshot.

This applies to both RCU and SRCU.

This brings the performance regression caused by commit 231264d6927f
("Fix: tracepoint: static call function vs data state mismatch") back to
what it was originally.

Before this commit:

  # trace-cmd start -e all
  # time trace-cmd start -p nop

  real 0m10.593s
  user 0m0.017s
  sys 0m0.259s

After this commit:

  # trace-cmd start -e all
  # time trace-cmd start -p nop

  real 0m0.878s
  user 0m0.000s
  sys 0m0.103s

Link: https://lkml.kernel.org/r/20210805192954.30688-1-mathieu.desnoyers@efficios.com
Link: https://lore.kernel.org/io-uring/4ebea8f0-58c9-e571-fd30-0ce4f6f09c70@samba.org/
Cc: stable@vger.kernel.org
Cc: Ingo Molnar <mingo@redhat.com>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Andrew Morton <akpm@linux-foundation.org>
Cc: "Paul E. McKenney" <paulmck@kernel.org>
Cc: Stefan Metzmacher <metze@samba.org>
Fixes: 231264d6927f ("Fix: tracepoint: static call function vs data state mismatch")
Signed-off-by: Mathieu Desnoyers <mathieu.desnoyers@efficios.com>
Reviewed-by: Paul E. McKenney <paulmck@kernel.org>
Signed-off-by: Steven Rostedt (VMware) <rostedt@goodmis.org>
5 months agotracepoint: Fix static call function vs data state mismatch
Mathieu Desnoyers [Thu, 5 Aug 2021 13:27:16 +0000 (09:27 -0400)]
tracepoint: Fix static call function vs data state mismatch

On a 1->0->1 callbacks transition, there is an issue with the new
callback using the old callback's data.

Considering __DO_TRACE_CALL:

        do {                                                            \
                struct tracepoint_func *it_func_ptr;                    \
                void *__data;                                           \
                it_func_ptr =                                           \
                        rcu_dereference_raw((&__tracepoint_##name)->funcs); \
                if (it_func_ptr) {                                      \
                        __data = (it_func_ptr)->data;                   \

----> [ delayed here on one CPU (e.g. vcpu preempted by the host) ]

                        static_call(tp_func_##name)(__data, args);      \
                }                                                       \
        } while (0)

It has loaded the tp->funcs of the old callback, so it will try to use the old
data. This can be fixed by adding a RCU sync anywhere in the 1->0->1
transition chain.

On a N->2->1 transition, we need an rcu-sync because you may have a
sequence of 3->2->1 (or 1->2->1) where the element 0 data is unchanged
between 2->1, but was changed from 3->2 (or from 1->2), which may be
observed by the static call. This can be fixed by adding an
unconditional RCU sync in transition 2->1.

Note, this fixes a correctness issue at the cost of adding a tremendous
performance regression to the disabling of tracepoints.

Before this commit:

  # trace-cmd start -e all
  # time trace-cmd start -p nop

  real 0m0.778s
  user 0m0.000s
  sys 0m0.061s

After this commit:

  # trace-cmd start -e all
  # time trace-cmd start -p nop

  real 0m10.593s
  user 0m0.017s
  sys 0m0.259s

A follow up fix will introduce a more lightweight scheme based on RCU
get_state and cond_sync, that will return the performance back to what it
was. As both this change and the lightweight versions are complex on their
own, for bisecting any issues that this may cause, they are kept as two
separate changes.

Link: https://lkml.kernel.org/r/20210805132717.23813-3-mathieu.desnoyers@efficios.com
Link: https://lore.kernel.org/io-uring/4ebea8f0-58c9-e571-fd30-0ce4f6f09c70@samba.org/
Cc: stable@vger.kernel.org
Cc: Ingo Molnar <mingo@redhat.com>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Andrew Morton <akpm@linux-foundation.org>
Cc: "Paul E. McKenney" <paulmck@kernel.org>
Cc: Stefan Metzmacher <metze@samba.org>
Fixes: d25e37d89dd2 ("tracepoint: Optimize using static_call()")
Signed-off-by: Mathieu Desnoyers <mathieu.desnoyers@efficios.com>
Signed-off-by: Steven Rostedt (VMware) <rostedt@goodmis.org>
5 months agotracepoint: static call: Compare data on transition from 2->1 callees
Mathieu Desnoyers [Thu, 5 Aug 2021 13:27:15 +0000 (09:27 -0400)]
tracepoint: static call: Compare data on transition from 2->1 callees

On transition from 2->1 callees, we should be comparing .data rather
than .func, because the same callback can be registered twice with
different data, and what we care about here is that the data of array
element 0 is unchanged to skip rcu sync.

Link: https://lkml.kernel.org/r/20210805132717.23813-2-mathieu.desnoyers@efficios.com
Link: https://lore.kernel.org/io-uring/4ebea8f0-58c9-e571-fd30-0ce4f6f09c70@samba.org/
Cc: stable@vger.kernel.org
Cc: Ingo Molnar <mingo@redhat.com>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Andrew Morton <akpm@linux-foundation.org>
Cc: "Paul E. McKenney" <paulmck@kernel.org>
Cc: Stefan Metzmacher <metze@samba.org>
Fixes: 547305a64632 ("tracepoint: Fix out of sync data passing by static caller")
Signed-off-by: Mathieu Desnoyers <mathieu.desnoyers@efficios.com>
Signed-off-by: Steven Rostedt (VMware) <rostedt@goodmis.org>
5 months agotracing: Quiet smp_processor_id() use in preemptable warning in hwlat
Steven Rostedt (VMware) [Wed, 4 Aug 2021 18:18:48 +0000 (14:18 -0400)]
tracing: Quiet smp_processor_id() use in preemptable warning in hwlat

The hardware latency detector (hwlat) has a mode that it runs one thread
across CPUs. The logic to move from the currently running CPU to the next
one in the list does a smp_processor_id() to find where it currently is.
Unfortunately, it's done with preemption enabled, and this triggers a
warning for using smp_processor_id() in a preempt enabled section.

As it is only using smp_processor_id() to get information on where it
currently is in order to simply move it to the next CPU, it doesn't really
care if it got moved in the mean time. It will simply balance out later if
such a case arises.

Switch smp_processor_id() to raw_smp_processor_id() to quiet that warning.

Link: https://lkml.kernel.org/r/20210804141848.79edadc0@oasis.local.home
Acked-by: Daniel Bristot de Oliveira <bristot@redhat.com>
Fixes: 8fa826b7344d ("trace/hwlat: Implement the mode config option")
Signed-off-by: Steven Rostedt (VMware) <rostedt@goodmis.org>
5 months agoscripts/tracing: fix the bug that can't parse raw_trace_func
Hui Su [Fri, 11 Jun 2021 02:21:07 +0000 (10:21 +0800)]
scripts/tracing: fix the bug that can't parse raw_trace_func

Since commit 77271ce4b2c0 ("tracing: Add irq, preempt-count and need resched info
to default trace output"), the default trace output format has been changed to:
          <idle>-0       [009] d.h. 22420.068695: _raw_spin_lock_irqsave <-hrtimer_interrupt
          <idle>-0       [000] ..s. 22420.068695: _nohz_idle_balance <-run_rebalance_domains
          <idle>-0       [011] d.h. 22420.068695: account_process_tick <-update_process_times

origin trace output format:(before v3.2.0)
     # tracer: nop
     #
     #           TASK-PID    CPU#    TIMESTAMP  FUNCTION
     #              | |       |          |         |
          migration/0-6     [000]    50.025810: rcu_note_context_switch <-__schedule
          migration/0-6     [000]    50.025812: trace_rcu_utilization <-rcu_note_context_switch
          migration/0-6     [000]    50.025813: rcu_sched_qs <-rcu_note_context_switch
          migration/0-6     [000]    50.025815: rcu_preempt_qs <-rcu_note_context_switch
          migration/0-6     [000]    50.025817: trace_rcu_utilization <-rcu_note_context_switch
          migration/0-6     [000]    50.025818: debug_lockdep_rcu_enabled <-__schedule
          migration/0-6     [000]    50.025820: debug_lockdep_rcu_enabled <-__schedule

The draw_functrace.py(introduced in v2.6.28) can't parse the new version format trace_func,
So we need modify draw_functrace.py to adapt the new version trace output format.

Link: https://lkml.kernel.org/r/20210611022107.608787-1-suhui@zeku.com
Cc: stable@vger.kernel.org
Fixes: 77271ce4b2c0 tracing: Add irq, preempt-count and need resched info to default trace output
Signed-off-by: Hui Su <suhui@zeku.com>
Signed-off-by: Steven Rostedt (VMware) <rostedt@goodmis.org>
5 months agoscripts/recordmcount.pl: Remove check_objcopy() and $can_use_local
Nathan Chancellor [Mon, 2 Aug 2021 21:03:07 +0000 (14:03 -0700)]
scripts/recordmcount.pl: Remove check_objcopy() and $can_use_local

When building ARCH=riscv allmodconfig with llvm-objcopy, the objcopy
version warning from this script appears:

WARNING: could not find objcopy version or version is less than 2.17.
        Local function references are disabled.

The check_objcopy() function in scripts/recordmcount.pl is set up to
parse GNU objcopy's version string, not llvm-objcopy's, which triggers
the warning.

Commit 799c43415442 ("kbuild: thin archives make default for all archs")
made binutils 2.20 mandatory and commit ba64beb17493 ("kbuild: check the
minimum assembler version in Kconfig") enforces this at configuration
time so just remove check_objcopy() and $can_use_local instead, assuming
--globalize-symbol is always available.

llvm-objcopy has supported --globalize-symbol since LLVM 7.0.0 in 2018
and the minimum version for building the kernel with LLVM is 10.0.1 so
there is no issue introduced:

Link: https://github.com/llvm/llvm-project/commit/ee5be798dae30d5f9414b01f76ff807edbc881aa
Link: https://lkml.kernel.org/r/20210802210307.3202472-1-nathan@kernel.org
Reviewed-by: Nick Desaulniers <ndesaulniers@google.com>
Signed-off-by: Nathan Chancellor <nathan@kernel.org>
Signed-off-by: Steven Rostedt (VMware) <rostedt@goodmis.org>
5 months agotracing: Reject string operand in the histogram expression
Masami Hiramatsu [Tue, 27 Jul 2021 22:55:43 +0000 (07:55 +0900)]
tracing: Reject string operand in the histogram expression

Since the string type can not be the target of the addition / subtraction
operation, it must be rejected. Without this fix, the string type silently
converted to digits.

Link: https://lkml.kernel.org/r/162742654278.290973.1523000673366456634.stgit@devnote2
Cc: stable@vger.kernel.org
Fixes: 100719dcef447 ("tracing: Add simple expression support to hist triggers")
Signed-off-by: Masami Hiramatsu <mhiramat@kernel.org>
Signed-off-by: Steven Rostedt (VMware) <rostedt@goodmis.org>
5 months agotracing / histogram: Give calculation hist_fields a size
Steven Rostedt (VMware) [Fri, 30 Jul 2021 21:19:51 +0000 (17:19 -0400)]
tracing / histogram: Give calculation hist_fields a size

When working on my user space applications, I found a bug in the synthetic
event code where the automated synthetic event field was not matching the
event field calculation it was attached to. Looking deeper into it, it was
because the calculation hist_field was not given a size.

The synthetic event fields are matched to their hist_fields either by
having the field have an identical string type, or if that does not match,
then the size and signed values are used to match the fields.

The problem arose when I tried to match a calculation where the fields
were "unsigned int". My tool created a synthetic event of type "u32". But
it failed to match. The string was:

  diff=field1-field2:onmatch(event).trace(synth,$diff)

Adding debugging into the kernel, I found that the size of "diff" was 0.
And since it was given "unsigned int" as a type, the histogram fallback
code used size and signed. The signed matched, but the size of u32 (4) did
not match zero, and the event failed to be created.

This can be worse if the field you want to match is not one of the
acceptable fields for a synthetic event. As event fields can have any type
that is supported in Linux, this can cause an issue. For example, if a
type is an enum. Then there's no way to use that with any calculations.

Have the calculation field simply take on the size of what it is
calculating.

Link: https://lkml.kernel.org/r/20210730171951.59c7743f@oasis.local.home
Cc: Tom Zanussi <zanussi@kernel.org>
Cc: Masami Hiramatsu <mhiramat@kernel.org>
Cc: Namhyung Kim <namhyung@kernel.org>
Cc: Ingo Molnar <mingo@kernel.org>
Cc: Andrew Morton <akpm@linux-foundation.org>
Cc: stable@vger.kernel.org
Fixes: 100719dcef447 ("tracing: Add simple expression support to hist triggers")
Signed-off-by: Steven Rostedt (VMware) <rostedt@goodmis.org>
5 months agotracing: Fix NULL pointer dereference in start_creating
Kamal Agrawal [Fri, 30 Jul 2021 13:23:06 +0000 (18:53 +0530)]
tracing: Fix NULL pointer dereference in start_creating

The event_trace_add_tracer() can fail. In this case, it leads to a crash
in start_creating with below call stack. Handle the error scenario
properly in trace_array_create_dir.

Call trace:
down_write+0x7c/0x204
start_creating.25017+0x6c/0x194
tracefs_create_file+0xc4/0x2b4
init_tracer_tracefs+0x5c/0x940
trace_array_create_dir+0x58/0xb4
trace_array_create+0x1bc/0x2b8
trace_array_get_by_name+0xdc/0x18c

Link: https://lkml.kernel.org/r/1627651386-21315-1-git-send-email-kamaagra@codeaurora.org
Cc: stable@vger.kernel.org
Fixes: 4114fbfd02f1 ("tracing: Enable creating new instance early boot")
Signed-off-by: Kamal Agrawal <kamaagra@codeaurora.org>
Signed-off-by: Steven Rostedt (VMware) <rostedt@goodmis.org>
6 months agotracepoints: Update static_call before tp_funcs when adding a tracepoint
Steven Rostedt (VMware) [Fri, 23 Jul 2021 01:52:18 +0000 (21:52 -0400)]
tracepoints: Update static_call before tp_funcs when adding a tracepoint

Because of the significant overhead that retpolines pose on indirect
calls, the tracepoint code was updated to use the new "static_calls" that
can modify the running code to directly call a function instead of using
an indirect caller, and this function can be changed at runtime.

In the tracepoint code that calls all the registered callbacks that are
attached to a tracepoint, the following is done:

it_func_ptr = rcu_dereference_raw((&__tracepoint_##name)->funcs);
if (it_func_ptr) {
__data = (it_func_ptr)->data;
static_call(tp_func_##name)(__data, args);
}

If there's just a single callback, the static_call is updated to just call
that callback directly. Once another handler is added, then the static
caller is updated to call the iterator, that simply loops over all the
funcs in the array and calls each of the callbacks like the old method
using indirect calling.

The issue was discovered with a race between updating the funcs array and
updating the static_call. The funcs array was updated first and then the
static_call was updated. This is not an issue as long as the first element
in the old array is the same as the first element in the new array. But
that assumption is incorrect, because callbacks also have a priority
field, and if there's a callback added that has a higher priority than the
callback on the old array, then it will become the first callback in the
new array. This means that it is possible to call the old callback with
the new callback data element, which can cause a kernel panic.

static_call = callback1()
funcs[] = {callback1,data1};
callback2 has higher priority than callback1

CPU 1 CPU 2
----- -----

   new_funcs = {callback2,data2},
               {callback1,data1}

   rcu_assign_pointer(tp->funcs, new_funcs);

  /*
   * Now tp->funcs has the new array
   * but the static_call still calls callback1
   */

it_func_ptr = tp->funcs [ new_funcs ]
data = it_func_ptr->data [ data2 ]
static_call(callback1, data);

/* Now callback1 is called with
 * callback2's data */

[ KERNEL PANIC ]

   update_static_call(iterator);

To prevent this from happening, always switch the static_call to the
iterator before assigning the tp->funcs to the new array. The iterator will
always properly match the callback with its data.

To trigger this bug:

  In one terminal:

    while :; do hackbench 50; done

  In another terminal

    echo 1 > /sys/kernel/tracing/events/sched/sched_waking/enable
    while :; do
        echo 1 > /sys/kernel/tracing/set_event_pid;
        sleep 0.5
        echo 0 > /sys/kernel/tracing/set_event_pid;
        sleep 0.5
   done

And it doesn't take long to crash. This is because the set_event_pid adds
a callback to the sched_waking tracepoint with a high priority, which will
be called before the sched_waking trace event callback is called.

Note, the removal to a single callback updates the array first, before
changing the static_call to single callback, which is the proper order as
the first element in the array is the same as what the static_call is
being changed to.

Link: https://lore.kernel.org/io-uring/4ebea8f0-58c9-e571-fd30-0ce4f6f09c70@samba.org/
Cc: stable@vger.kernel.org
Fixes: d25e37d89dd2f ("tracepoint: Optimize using static_call()")
Reported-by: Stefan Metzmacher <metze@samba.org>
tested-by: Stefan Metzmacher <metze@samba.org>
Signed-off-by: Steven Rostedt (VMware) <rostedt@goodmis.org>
6 months agoftrace: Remove redundant initialization of variable ret
Colin Ian King [Wed, 21 Jul 2021 12:09:15 +0000 (13:09 +0100)]
ftrace: Remove redundant initialization of variable ret

The variable ret is being initialized with a value that is never
read, it is being updated later on. The assignment is redundant and
can be removed.

Link: https://lkml.kernel.org/r/20210721120915.122278-1-colin.king@canonical.com
Addresses-Coverity: ("Unused value")
Signed-off-by: Colin Ian King <colin.king@canonical.com>
Signed-off-by: Steven Rostedt (VMware) <rostedt@goodmis.org>
6 months agoftrace: Avoid synchronize_rcu_tasks_rude() call when not necessary
Nicolas Saenz Julienne [Wed, 21 Jul 2021 11:47:26 +0000 (13:47 +0200)]
ftrace: Avoid synchronize_rcu_tasks_rude() call when not necessary

synchronize_rcu_tasks_rude() triggers IPIs and forces rescheduling on
all CPUs. It is a costly operation and, when targeting nohz_full CPUs,
very disrupting (hence the name). So avoid calling it when 'old_hash'
doesn't need to be freed.

Link: https://lkml.kernel.org/r/20210721114726.1545103-1-nsaenzju@redhat.com
Signed-off-by: Nicolas Saenz Julienne <nsaenzju@redhat.com>
Signed-off-by: Steven Rostedt (VMware) <rostedt@goodmis.org>
6 months agotracing: Clean up alloc_synth_event()
Steven Rostedt (VMware) [Wed, 21 Jul 2021 23:53:41 +0000 (19:53 -0400)]
tracing: Clean up alloc_synth_event()

alloc_synth_event() currently has the following code to initialize the
event fields and dynamic_fields:

for (i = 0, j = 0; i < n_fields; i++) {
event->fields[i] = fields[i];

if (fields[i]->is_dynamic) {
event->dynamic_fields[j] = fields[i];
event->dynamic_fields[j]->field_pos = i;
event->dynamic_fields[j++] = fields[i];
event->n_dynamic_fields++;
}
}

1) It would make more sense to have all fields keep track of their
   field_pos.

2) event->dynmaic_fields[j] is assigned twice for no reason.

3) We can move updating event->n_dynamic_fields outside the loop, and just
   assign it to j.

This combination makes the code much cleaner.

Link: https://lkml.kernel.org/r/20210721195341.29bb0f77@oasis.local.home
Signed-off-by: Steven Rostedt (VMware) <rostedt@goodmis.org>
6 months agotracing/histogram: Rename "cpu" to "common_cpu"
Steven Rostedt (VMware) [Wed, 21 Jul 2021 15:00:53 +0000 (11:00 -0400)]
tracing/histogram: Rename "cpu" to "common_cpu"

Currently the histogram logic allows the user to write "cpu" in as an
event field, and it will record the CPU that the event happened on.

The problem with this is that there's a lot of events that have "cpu"
as a real field, and using "cpu" as the CPU it ran on, makes it
impossible to run histograms on the "cpu" field of events.

For example, if I want to have a histogram on the count of the
workqueue_queue_work event on its cpu field, running:

 ># echo 'hist:keys=cpu' > events/workqueue/workqueue_queue_work/trigger

Gives a misleading and wrong result.

Change the command to "common_cpu" as no event should have "common_*"
fields as that's a reserved name for fields used by all events. And
this makes sense here as common_cpu would be a field used by all events.

Now we can even do:

 ># echo 'hist:keys=common_cpu,cpu if cpu < 100' > events/workqueue/workqueue_queue_work/trigger
 ># cat events/workqueue/workqueue_queue_work/hist
 # event histogram
 #
 # trigger info: hist:keys=common_cpu,cpu:vals=hitcount:sort=hitcount:size=2048 if cpu < 100 [active]
 #

 { common_cpu:          0, cpu:          2 } hitcount:          1
 { common_cpu:          0, cpu:          4 } hitcount:          1
 { common_cpu:          7, cpu:          7 } hitcount:          1
 { common_cpu:          0, cpu:          7 } hitcount:          1
 { common_cpu:          0, cpu:          1 } hitcount:          1
 { common_cpu:          0, cpu:          6 } hitcount:          2
 { common_cpu:          0, cpu:          5 } hitcount:          2
 { common_cpu:          1, cpu:          1 } hitcount:          4
 { common_cpu:          6, cpu:          6 } hitcount:          4
 { common_cpu:          5, cpu:          5 } hitcount:         14
 { common_cpu:          4, cpu:          4 } hitcount:         26
 { common_cpu:          0, cpu:          0 } hitcount:         39
 { common_cpu:          2, cpu:          2 } hitcount:        184

Now for backward compatibility, I added a trick. If "cpu" is used, and
the field is not found, it will fall back to "common_cpu" and work as
it did before. This way, it will still work for old programs that use
"cpu" to get the actual CPU, but if the event has a "cpu" as a field, it
will get that event's "cpu" field, which is probably what it wants
anyway.

I updated the tracefs/README to include documentation about both the
common_timestamp and the common_cpu. This way, if that text is present in
the README, then an application can know that common_cpu is supported over
just plain "cpu".

Link: https://lkml.kernel.org/r/20210721110053.26b4f641@oasis.local.home
Cc: Namhyung Kim <namhyung@kernel.org>
Cc: Ingo Molnar <mingo@kernel.org>
Cc: Andrew Morton <akpm@linux-foundation.org>
Cc: stable@vger.kernel.org
Fixes: 8b7622bf94a44 ("tracing: Add cpu field for hist triggers")
Reviewed-by: Tom Zanussi <zanussi@kernel.org>
Reviewed-by: Masami Hiramatsu <mhiramat@kernel.org>
Signed-off-by: Steven Rostedt (VMware) <rostedt@goodmis.org>
6 months agotracing: Synthetic event field_pos is an index not a boolean
Steven Rostedt (VMware) [Wed, 21 Jul 2021 23:10:08 +0000 (19:10 -0400)]
tracing: Synthetic event field_pos is an index not a boolean

Performing the following:

 ># echo 'wakeup_lat s32 pid; u64 delta; char wake_comm[]' > synthetic_events
 ># echo 'hist:keys=pid:__arg__1=common_timestamp.usecs' > events/sched/sched_waking/trigger
 ># echo 'hist:keys=next_pid:pid=next_pid,delta=common_timestamp.usecs-$__arg__1:onmatch(sched.sched_waking).trace(wakeup_lat,$pid,$delta,prev_comm)'\
      > events/sched/sched_switch/trigger
 ># echo 1 > events/synthetic/enable

Crashed the kernel:

 BUG: kernel NULL pointer dereference, address: 000000000000001b
 #PF: supervisor read access in kernel mode
 #PF: error_code(0x0000) - not-present page
 PGD 0 P4D 0
 Oops: 0000 [#1] PREEMPT SMP
 CPU: 7 PID: 0 Comm: swapper/7 Not tainted 5.13.0-rc5-test+ #104
 Hardware name: Hewlett-Packard HP Compaq Pro 6300 SFF/339A, BIOS K01 v03.03 07/14/2016
 RIP: 0010:strlen+0x0/0x20
 Code: f6 82 80 2b 0b bc 20 74 11 0f b6 50 01 48 83 c0 01 f6 82 80 2b 0b bc
  20 75 ef c3 66 66 2e 0f 1f 84 00 00 00 00 00 0f 1f 40 00 <80> 3f 00 74 10
  48 89 f8 48 83 c0 01 80 38 9 f8 c3 31
 RSP: 0018:ffffaa75000d79d0 EFLAGS: 00010046
 RAX: 0000000000000002 RBX: ffff9cdb55575270 RCX: 0000000000000000
 RDX: ffff9cdb58c7a320 RSI: ffffaa75000d7b40 RDI: 000000000000001b
 RBP: ffffaa75000d7b40 R08: ffff9cdb40a4f010 R09: ffffaa75000d7ab8
 R10: ffff9cdb4398c700 R11: 0000000000000008 R12: ffff9cdb58c7a320
 R13: ffff9cdb55575270 R14: ffff9cdb58c7a000 R15: 0000000000000018
 FS:  0000000000000000(0000) GS:ffff9cdb5aa00000(0000) knlGS:0000000000000000
 CS:  0010 DS: 0000 ES: 0000 CR0: 0000000080050033
 CR2: 000000000000001b CR3: 00000000c0612006 CR4: 00000000001706e0
 Call Trace:
  trace_event_raw_event_synth+0x90/0x1d0
  action_trace+0x5b/0x70
  event_hist_trigger+0x4bd/0x4e0
  ? cpumask_next_and+0x20/0x30
  ? update_sd_lb_stats.constprop.0+0xf6/0x840
  ? __lock_acquire.constprop.0+0x125/0x550
  ? find_held_lock+0x32/0x90
  ? sched_clock_cpu+0xe/0xd0
  ? lock_release+0x155/0x440
  ? update_load_avg+0x8c/0x6f0
  ? enqueue_entity+0x18a/0x920
  ? __rb_reserve_next+0xe5/0x460
  ? ring_buffer_lock_reserve+0x12a/0x3f0
  event_triggers_call+0x52/0xe0
  trace_event_buffer_commit+0x1ae/0x240
  trace_event_raw_event_sched_switch+0x114/0x170
  __traceiter_sched_switch+0x39/0x50
  __schedule+0x431/0xb00
  schedule_idle+0x28/0x40
  do_idle+0x198/0x2e0
  cpu_startup_entry+0x19/0x20
  secondary_startup_64_no_verify+0xc2/0xcb

The reason is that the dynamic events array keeps track of the field
position of the fields array, via the field_pos variable in the
synth_field structure. Unfortunately, that field is a boolean for some
reason, which means any field_pos greater than 1 will be a bug (in this
case it was 2).

Link: https://lkml.kernel.org/r/20210721191008.638bce34@oasis.local.home
Cc: Masami Hiramatsu <mhiramat@kernel.org>
Cc: Namhyung Kim <namhyung@kernel.org>
Cc: Ingo Molnar <mingo@kernel.org>
Cc: Andrew Morton <akpm@linux-foundation.org>
Cc: stable@vger.kernel.org
Fixes: bd82631d7ccdc ("tracing: Add support for dynamic strings to synthetic events")
Reviewed-by: Tom Zanussi <zanussi@kernel.org>
Signed-off-by: Steven Rostedt (VMware) <rostedt@goodmis.org>
6 months agotracing: Fix bug in rb_per_cpu_empty() that might cause deadloop.
Haoran Luo [Wed, 21 Jul 2021 14:12:07 +0000 (14:12 +0000)]
tracing: Fix bug in rb_per_cpu_empty() that might cause deadloop.

The "rb_per_cpu_empty()" misinterpret the condition (as not-empty) when
"head_page" and "commit_page" of "struct ring_buffer_per_cpu" points to
the same buffer page, whose "buffer_data_page" is empty and "read" field
is non-zero.

An error scenario could be constructed as followed (kernel perspective):

1. All pages in the buffer has been accessed by reader(s) so that all of
them will have non-zero "read" field.

2. Read and clear all buffer pages so that "rb_num_of_entries()" will
return 0 rendering there's no more data to read. It is also required
that the "read_page", "commit_page" and "tail_page" points to the same
page, while "head_page" is the next page of them.

3. Invoke "ring_buffer_lock_reserve()" with large enough "length"
so that it shot pass the end of current tail buffer page. Now the
"head_page", "commit_page" and "tail_page" points to the same page.

4. Discard current event with "ring_buffer_discard_commit()", so that
"head_page", "commit_page" and "tail_page" points to a page whose buffer
data page is now empty.

When the error scenario has been constructed, "tracing_read_pipe" will
be trapped inside a deadloop: "trace_empty()" returns 0 since
"rb_per_cpu_empty()" returns 0 when it hits the CPU containing such
constructed ring buffer. Then "trace_find_next_entry_inc()" always
return NULL since "rb_num_of_entries()" reports there's no more entry
to read. Finally "trace_seq_to_user()" returns "-EBUSY" spanking
"tracing_read_pipe" back to the start of the "waitagain" loop.

I've also written a proof-of-concept script to construct the scenario
and trigger the bug automatically, you can use it to trace and validate
my reasoning above:

  https://github.com/aegistudio/RingBufferDetonator.git

Tests has been carried out on linux kernel 5.14-rc2
(2734d6c1b1a089fb593ef6a23d4b70903526fe0c), my fixed version
of kernel (for testing whether my update fixes the bug) and
some older kernels (for range of affected kernels). Test result is
also attached to the proof-of-concept repository.

Link: https://lore.kernel.org/linux-trace-devel/YPaNxsIlb2yjSi5Y@aegistudio/
Link: https://lore.kernel.org/linux-trace-devel/YPgrN85WL9VyrZ55@aegistudio
Cc: stable@vger.kernel.org
Fixes: bf41a158cacba ("ring-buffer: make reentrant")
Suggested-by: Linus Torvalds <torvalds@linuxfoundation.org>
Signed-off-by: Haoran Luo <www@aegistudio.net>
Signed-off-by: Steven Rostedt (VMware) <rostedt@goodmis.org>
6 months agoLinux 5.14-rc2
Linus Torvalds [Sun, 18 Jul 2021 21:13:49 +0000 (14:13 -0700)]
Linux 5.14-rc2

6 months agoMerge tag 'perf-tools-fixes-for-v5.14-2021-07-18' of git://git.kernel.org/pub/scm...
Linus Torvalds [Sun, 18 Jul 2021 19:20:27 +0000 (12:20 -0700)]
Merge tag 'perf-tools-fixes-for-v5.14-2021-07-18' of git://git./linux/kernel/git/acme/linux

Pull perf tools fixes from Arnaldo Carvalho de Melo:

 - Skip invalid hybrid PMU on hybrid systems when the atom (little) CPUs
   are offlined.

 - Fix 'perf test' problems related to the recently added hybrid
   (BIG/little) code.

 - Split ARM's coresight (hw tracing) decode by aux records to avoid
   fatal decoding errors.

 - Fix add event failure in 'perf probe' when running 32-bit perf in a
   64-bit kernel.

 - Fix 'perf sched record' failure when CONFIG_SCHEDSTATS is not set.

 - Fix memory and refcount leaks detected by ASAn when running 'perf
   test', should be clean of warnings now.

 - Remove broken definition of __LITTLE_ENDIAN from tools'
   linux/kconfig.h, which was breaking the build in some systems.

 - Cast PTHREAD_STACK_MIN to int as it may turn into 'long
   sysconf(__SC_THREAD_STACK_MIN_VALUE), breaking the build in some
   systems.

 - Fix libperf build error with LIBPFM4=1.

 - Sync UAPI files changed by the memfd_secret new syscall.

* tag 'perf-tools-fixes-for-v5.14-2021-07-18' of git://git.kernel.org/pub/scm/linux/kernel/git/acme/linux: (35 commits)
  perf sched: Fix record failure when CONFIG_SCHEDSTATS is not set
  perf probe: Fix add event failure when running 32-bit perf in a 64-bit kernel
  perf data: Close all files in close_dir()
  perf probe-file: Delete namelist in del_events() on the error path
  perf test bpf: Free obj_buf
  perf trace: Free strings in trace__parse_events_option()
  perf trace: Free syscall tp fields in evsel->priv
  perf trace: Free syscall->arg_fmt
  perf trace: Free malloc'd trace fields on exit
  perf lzma: Close lzma stream on exit
  perf script: Fix memory 'threads' and 'cpus' leaks on exit
  perf script: Release zstd data
  perf session: Cleanup trace_event
  perf inject: Close inject.output on exit
  perf report: Free generated help strings for sort option
  perf env: Fix memory leak of cpu_pmu_caps
  perf test maps__merge_in: Fix memory leak of maps
  perf dso: Fix memory leak in dso__new_map()
  perf test event_update: Fix memory leak of unit
  perf test event_update: Fix memory leak of evlist
  ...

6 months agoMerge tag 'xfs-5.14-fixes-1' of git://git.kernel.org/pub/scm/fs/xfs/xfs-linux
Linus Torvalds [Sun, 18 Jul 2021 18:27:25 +0000 (11:27 -0700)]
Merge tag 'xfs-5.14-fixes-1' of git://git./fs/xfs/xfs-linux

Pull xfs fixes from Darrick Wong:
 "A few fixes for issues in the new online shrink code, additional
  corrections for my recent bug-hunt w.r.t. extent size hints on
  realtime, and improved input checking of the GROWFSRT ioctl.

  IOW, the usual 'I somehow got bored during the merge window and
  resumed auditing the farther reaches of xfs':

   - Fix shrink eligibility checking when sparse inode clusters enabled

   - Reset '..' directory entries when unlinking directories to prevent
     verifier errors if fs is shrinked later

   - Don't report unusable extent size hints to FSGETXATTR

   - Don't warn when extent size hints are unusable because the sysadmin
     configured them that way

   - Fix insufficient parameter validation in GROWFSRT ioctl

   - Fix integer overflow when adding rt volumes to filesystem"

* tag 'xfs-5.14-fixes-1' of git://git.kernel.org/pub/scm/fs/xfs/xfs-linux:
  xfs: detect misaligned rtinherit directory extent size hints
  xfs: fix an integer overflow error in xfs_growfs_rt
  xfs: improve FSGROWFSRT precondition checking
  xfs: don't expose misaligned extszinherit hints to userspace
  xfs: correct the narrative around misaligned rtinherit/extszinherit dirs
  xfs: reset child dir '..' entry when unlinking child
  xfs: check for sparse inode clusters that cross new EOAG when shrinking

6 months agoMerge tag 'iomap-5.14-fixes-1' of git://git.kernel.org/pub/scm/fs/xfs/xfs-linux
Linus Torvalds [Sun, 18 Jul 2021 18:17:06 +0000 (11:17 -0700)]
Merge tag 'iomap-5.14-fixes-1' of git://git./fs/xfs/xfs-linux

Pull iomap fixes from Darrick Wong:
 "A handful of bugfixes for the iomap code.

  There's nothing especially exciting here, just fixes for UBSAN (not
  KASAN as I erroneously wrote in the tag message) warnings about
  undefined behavior in the SEEK_DATA/SEEK_HOLE code, and some
  reshuffling of per-page block state info to fix some problems with
  gfs2.

   - Fix KASAN warnings due to integer overflow in SEEK_DATA/SEEK_HOLE

   - Fix assertion errors when using inlinedata files on gfs2"

* tag 'iomap-5.14-fixes-1' of git://git.kernel.org/pub/scm/fs/xfs/xfs-linux:
  iomap: Don't create iomap_page objects in iomap_page_mkwrite_actor
  iomap: Don't create iomap_page objects for inline files
  iomap: Permit pages without an iop to enter writeback
  iomap: remove the length variable in iomap_seek_hole
  iomap: remove the length variable in iomap_seek_data

6 months agoMerge tag 'kbuild-fixes-v5.14' of git://git.kernel.org/pub/scm/linux/kernel/git/masah...
Linus Torvalds [Sun, 18 Jul 2021 18:10:30 +0000 (11:10 -0700)]
Merge tag 'kbuild-fixes-v5.14' of git://git./linux/kernel/git/masahiroy/linux-kbuild

Pull Kbuild fixes from Masahiro Yamada:

 - Restore the original behavior of scripts/setlocalversion when
   LOCALVERSION is set to empty.

 - Show Kconfig prompts even for 'make -s'

 - Fix the combination of COFNIG_LTO_CLANG=y and CONFIG_MODVERSIONS=y
   for older GNU Make versions

* tag 'kbuild-fixes-v5.14' of git://git.kernel.org/pub/scm/linux/kernel/git/masahiroy/linux-kbuild:
  Documentation: Fix intiramfs script name
  Kbuild: lto: fix module versionings mismatch in GNU make 3.X
  kbuild: do not suppress Kconfig prompts for silent build
  scripts/setlocalversion: fix a bug when LOCALVERSION is empty

6 months agoDocumentation: Fix intiramfs script name
Robert Richter [Thu, 15 Jul 2021 09:26:02 +0000 (11:26 +0200)]
Documentation: Fix intiramfs script name

Documentation was not changed when renaming the script in commit
80e715a06c2d ("initramfs: rename gen_initramfs_list.sh to
gen_initramfs.sh"). Fixing this.

Basically does:

 $ sed -i -e s/gen_initramfs_list.sh/gen_initramfs.sh/g $(git grep -l gen_initramfs_list.sh)

Fixes: 80e715a06c2d ("initramfs: rename gen_initramfs_list.sh to gen_initramfs.sh")
Signed-off-by: Robert Richter <rrichter@amd.com>
Signed-off-by: Masahiro Yamada <masahiroy@kernel.org>
6 months agoKbuild: lto: fix module versionings mismatch in GNU make 3.X
Lecopzer Chen [Thu, 15 Jul 2021 07:37:16 +0000 (15:37 +0800)]
Kbuild: lto: fix module versionings mismatch in GNU make 3.X

When building modules(CONFIG_...=m), I found some of module versions
are incorrect and set to 0.
This can be found in build log for first clean build which shows

WARNING: EXPORT symbol "XXXX" [drivers/XXX/XXX.ko] version generation failed,
symbol will not be versioned.

But in second build(incremental build), the WARNING disappeared and the
module version becomes valid CRC and make someone who want to change
modules without updating kernel image can't insert their modules.

The problematic code is
+ $(foreach n, $(filter-out FORCE,$^), \
+ $(if $(wildcard $(n).symversions), \
+ ; cat $(n).symversions >> $@.symversions))

For example:
  rm -f fs/notify/built-in.a.symversions    ; rm -f fs/notify/built-in.a; \
llvm-ar cDPrST fs/notify/built-in.a fs/notify/fsnotify.o \
fs/notify/notification.o fs/notify/group.o ...

`foreach n` shows nothing to `cat` into $(n).symversions because
`if $(wildcard $(n).symversions)` return nothing, but actually
they do exist during this line was executed.

-rw-r--r-- 1 root root 168580 Jun 13 19:10 fs/notify/fsnotify.o
-rw-r--r-- 1 root root    111 Jun 13 19:10 fs/notify/fsnotify.o.symversions

The reason is the $(n).symversions are generated at runtime, but
Makefile wildcard function expends and checks the file exist or not
during parsing the Makefile.

Thus fix this by use `test` shell command to check the file
existence in runtime.

Rebase from both:
1. [https://lore.kernel.org/lkml/20210616080252.32046-1-lecopzer.chen@mediatek.com/]
2. [https://lore.kernel.org/lkml/20210702032943.7865-1-lecopzer.chen@mediatek.com/]

Fixes: 38e891849003 ("kbuild: lto: fix module versioning")
Co-developed-by: Sami Tolvanen <samitolvanen@google.com>
Signed-off-by: Lecopzer Chen <lecopzer.chen@mediatek.com>
Signed-off-by: Masahiro Yamada <masahiroy@kernel.org>
6 months agokbuild: do not suppress Kconfig prompts for silent build
Masahiro Yamada [Wed, 14 Jul 2021 04:23:49 +0000 (13:23 +0900)]
kbuild: do not suppress Kconfig prompts for silent build

When a new CONFIG option is available, Kbuild shows a prompt to get
the user input.

  $ make
  [ snip ]
  Core Scheduling for SMT (SCHED_CORE) [N/y/?] (NEW)

This is the only interactive place in the build process.

Commit 174a1dcc9642 ("kbuild: sink stdout from cmd for silent build")
suppressed Kconfig prompts as well because syncconfig is invoked by
the 'cmd' macro. You cannot notice the fact that Kconfig is waiting
for the user input.

Use 'kecho' to show the equivalent short log without suppressing stdout
from sub-make.

Fixes: 174a1dcc9642 ("kbuild: sink stdout from cmd for silent build")
Reported-by: Tetsuo Handa <penguin-kernel@I-love.SAKURA.ne.jp>
Signed-off-by: Masahiro Yamada <masahiroy@kernel.org>
Tested-by: Tetsuo Handa <penguin-kernel@I-love.SAKURA.ne.jp>
6 months agoscripts/setlocalversion: fix a bug when LOCALVERSION is empty
Mikulas Patocka [Mon, 12 Jul 2021 19:35:46 +0000 (15:35 -0400)]
scripts/setlocalversion: fix a bug when LOCALVERSION is empty

The commit 042da426f8eb ("scripts/setlocalversion: simplify the short
version part") reduces indentation. Unfortunately, it also changes behavior
in a subtle way - if the user has empty "LOCALVERSION" variable, the plus
sign is appended to the kernel version. It wasn't appended before.

This patch reverts to the old behavior - we append the plus sign only if
the LOCALVERSION variable is not set.

Fixes: 042da426f8eb ("scripts/setlocalversion: simplify the short version part")
Signed-off-by: Mikulas Patocka <mpatocka@redhat.com>
Signed-off-by: Masahiro Yamada <masahiroy@kernel.org>
6 months agoperf sched: Fix record failure when CONFIG_SCHEDSTATS is not set
Yang Jihong [Tue, 13 Jul 2021 11:23:58 +0000 (19:23 +0800)]
perf sched: Fix record failure when CONFIG_SCHEDSTATS is not set

The tracepoints trace_sched_stat_{wait, sleep, iowait} are not exposed to user
if CONFIG_SCHEDSTATS is not set, "perf sched record" records the three events.
As a result, the command fails.

Before:

  #perf sched record sleep 1
  event syntax error: 'sched:sched_stat_wait'
                       \___ unknown tracepoint

  Error:  File /sys/kernel/tracing/events/sched/sched_stat_wait not found.
  Hint:   Perhaps this kernel misses some CONFIG_ setting to enable this feature?.

  Run 'perf list' for a list of valid events

   Usage: perf record [<options>] [<command>]
      or: perf record [<options>] -- <command> [<options>]

      -e, --event <event>   event selector. use 'perf list' to list available events

Solution:
  Check whether schedstat tracepoints are exposed. If no, these events are not recorded.

After:
  # perf sched record sleep 1
  [ perf record: Woken up 1 times to write data ]
  [ perf record: Captured and wrote 0.163 MB perf.data (1091 samples) ]
  # perf sched report
  run measurement overhead: 4736 nsecs
  sleep measurement overhead: 9059979 nsecs
  the run test took 999854 nsecs
  the sleep test took 8945271 nsecs
  nr_run_events:        716
  nr_sleep_events:      785
  nr_wakeup_events:     0
  ...
  ------------------------------------------------------------

Fixes: 2a09b5de235a6 ("sched/fair: do not expose some tracepoints to user if CONFIG_SCHEDSTATS is not set")
Signed-off-by: Yang Jihong <yangjihong1@huawei.com>
Cc: Alexander Shishkin <alexander.shishkin@linux.intel.com>
Cc: Jiri Olsa <jolsa@redhat.com>
Cc: Mark Rutland <mark.rutland@arm.com>
Cc: Namhyung Kim <namhyung@kernel.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Steven Rostedt (VMware) <rostedt@goodmis.org>
Cc: Yafang Shao <laoar.shao@gmail.com>
Link: http://lore.kernel.org/lkml/20210713112358.194693-1-yangjihong1@huawei.com
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
6 months agoperf probe: Fix add event failure when running 32-bit perf in a 64-bit kernel
Yang Jihong [Thu, 15 Jul 2021 06:37:23 +0000 (14:37 +0800)]
perf probe: Fix add event failure when running 32-bit perf in a 64-bit kernel

The "address" member of "struct probe_trace_point" uses long data type.
If kernel is 64-bit and perf program is 32-bit, size of "address"
variable is 32 bits.

As a result, upper 32 bits of address read from kernel are truncated, an
error occurs during address comparison in kprobe_warn_out_range().

Before:

  # perf probe -a schedule
  schedule is out of .text, skip it.
    Error: Failed to add events.

Solution:
  Change data type of "address" variable to u64 and change corresponding
address printing and value assignment.

After:

  # perf.new.new probe -a schedule
  Added new event:
    probe:schedule       (on schedule)

  You can now use it in all perf tools, such as:

          perf record -e probe:schedule -aR sleep 1

  # perf probe -l
    probe:schedule       (on schedule@kernel/sched/core.c)
  # perf record -e probe:schedule -aR sleep 1
  [ perf record: Woken up 1 times to write data ]
  [ perf record: Captured and wrote 0.156 MB perf.data (1366 samples) ]
  # perf report --stdio
  # To display the perf.data header info, please use --header/--header-only options.
  #
  #
  # Total Lost Samples: 0
  #
  # Samples: 1K of event 'probe:schedule'
  # Event count (approx.): 1366
  #
  # Overhead  Command          Shared Object      Symbol
  # ........  ...............  .................  ............
  #
       6.22%  migration/0      [kernel.kallsyms]  [k] schedule
       6.22%  migration/1      [kernel.kallsyms]  [k] schedule
       6.22%  migration/2      [kernel.kallsyms]  [k] schedule
       6.22%  migration/3      [kernel.kallsyms]  [k] schedule
       6.15%  migration/10     [kernel.kallsyms]  [k] schedule
       6.15%  migration/11     [kernel.kallsyms]  [k] schedule
       6.15%  migration/12     [kernel.kallsyms]  [k] schedule
       6.15%  migration/13     [kernel.kallsyms]  [k] schedule
       6.15%  migration/14     [kernel.kallsyms]  [k] schedule
       6.15%  migration/15     [kernel.kallsyms]  [k] schedule
       6.15%  migration/4      [kernel.kallsyms]  [k] schedule
       6.15%  migration/5      [kernel.kallsyms]  [k] schedule
       6.15%  migration/6      [kernel.kallsyms]  [k] schedule
       6.15%  migration/7      [kernel.kallsyms]  [k] schedule
       6.15%  migration/8      [kernel.kallsyms]  [k] schedule
       6.15%  migration/9      [kernel.kallsyms]  [k] schedule
       0.22%  rcu_sched        [kernel.kallsyms]  [k] schedule
  ...
  #
  # (Cannot load tips.txt file, please install perf!)
  #

Signed-off-by: Yang Jihong <yangjihong1@huawei.com>
Acked-by: Masami Hiramatsu <mhiramat@kernel.org>
Cc: Alexander Shishkin <alexander.shishkin@linux.intel.com>
Cc: Frank Ch. Eigler <fche@redhat.com>
Cc: Ian Rogers <irogers@google.com>
Cc: Jianlin Lv <jianlin.lv@arm.com>
Cc: Jin Yao <yao.jin@linux.intel.com>
Cc: Jiri Olsa <jolsa@redhat.com>
Cc: Li Huafei <lihuafei1@huawei.com>
Cc: Mark Rutland <mark.rutland@arm.com>
Cc: Namhyung Kim <namhyung@kernel.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Ravi Bangoria <ravi.bangoria@linux.ibm.com>
Cc: Srikar Dronamraju <srikar@linux.vnet.ibm.com>
Link: http://lore.kernel.org/lkml/20210715063723.11926-1-yangjihong1@huawei.com
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
6 months agoperf data: Close all files in close_dir()
Riccardo Mancini [Fri, 16 Jul 2021 14:11:20 +0000 (16:11 +0200)]
perf data: Close all files in close_dir()

When using 'perf report' in directory mode, the first file is not closed
on exit, causing a memory leak.

The problem is caused by the iterating variable never reaching 0.

Fixes: 145520631130bd64 ("perf data: Add perf_data__(create_dir|close_dir) functions")
Signed-off-by: Riccardo Mancini <rickyman7@gmail.com>
Acked-by: Namhyung Kim <namhyung@kernel.org>
Cc: Alexander Shishkin <alexander.shishkin@linux.intel.com>
Cc: Ian Rogers <irogers@google.com>
Cc: Jiri Olsa <jolsa@redhat.com>
Cc: Mark Rutland <mark.rutland@arm.com>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Zhen Lei <thunder.leizhen@huawei.com>
Link: http://lore.kernel.org/lkml/20210716141122.858082-1-rickyman7@gmail.com
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
6 months agoperf probe-file: Delete namelist in del_events() on the error path
Riccardo Mancini [Thu, 15 Jul 2021 16:07:25 +0000 (18:07 +0200)]
perf probe-file: Delete namelist in del_events() on the error path

ASan reports some memory leaks when running:

  # perf test "42: BPF filter"

This second leak is caused by a strlist not being dellocated on error
inside probe_file__del_events.

This patch adds a goto label before the deallocation and makes the error
path jump to it.

Signed-off-by: Riccardo Mancini <rickyman7@gmail.com>
Fixes: e7895e422e4da63d ("perf probe: Split del_perf_probe_events()")
Cc: Ian Rogers <irogers@google.com>
Cc: Jiri Olsa <jolsa@redhat.com>
Cc: Mark Rutland <mark.rutland@arm.com>
Cc: Namhyung Kim <namhyung@kernel.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Link: http://lore.kernel.org/lkml/174963c587ae77fa108af794669998e4ae558338.1626343282.git.rickyman7@gmail.com
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
6 months agoMerge tag 'soc-fixes-5.14-1' of git://git.kernel.org/pub/scm/linux/kernel/git/soc/soc
Linus Torvalds [Sat, 17 Jul 2021 22:58:24 +0000 (15:58 -0700)]
Merge tag 'soc-fixes-5.14-1' of git://git./linux/kernel/git/soc/soc

Pull ARM SoC fixes from Arnd Bergmann:
 "Here are the patches for this week that came as the fallout of the
  merge window:

   - Two fixes for the NVidia memory controller driver

   - multiple defconfig files get patched to turn CONFIG_FB back on
     after that is no longer selected by CONFIG_DRM

   - ffa and scmpi firmware drivers fixes, mostly addressing compiler
     and documentation warnings

   - Platform specific fixes for device tree files on ASpeed, Renesas
     and NVidia SoC, mostly for recent regressions.

   - A workaround for a regression on the USB PHY with devlink when the
     usb-nop-xceiv driver is not available until the rootfs is mounted.

   - Device tree compiler warnings in Arm Versatile-AB"

* tag 'soc-fixes-5.14-1' of git://git.kernel.org/pub/scm/linux/kernel/git/soc/soc: (35 commits)
  ARM: dts: versatile: Fix up interrupt controller node names
  ARM: multi_v7_defconfig: Make NOP_USB_XCEIV driver built-in
  ARM: configs: Update u8500_defconfig
  ARM: configs: Update Vexpress defconfig
  ARM: configs: Update Versatile defconfig
  ARM: configs: Update RealView defconfig
  ARM: configs: Update Integrator defconfig
  arm: Typo s/PCI_IXP4XX_LEGACY/IXP4XX_PCI_LEGACY/
  firmware: arm_scmi: Fix range check for the maximum number of pending messages
  firmware: arm_scmi: Avoid padding in sensor message structure
  firmware: arm_scmi: Fix kernel doc warnings about return values
  firmware: arm_scpi: Fix kernel doc warnings
  firmware: arm_scmi: Fix kernel doc warnings
  ARM: shmobile: defconfig: Restore graphical consoles
  firmware: arm_ffa: Fix a possible ffa_linux_errmap buffer overflow
  firmware: arm_ffa: Fix the comment style
  firmware: arm_ffa: Simplify probe function
  firmware: arm_ffa: Ensure drivers provide a probe function
  firmware: arm_scmi: Fix possible scmi_linux_errmap buffer overflow
  firmware: arm_scmi: Ensure drivers provide a probe function
  ...

6 months agoRevert "mm/slub: use stackdepot to save stack trace in objects"
Linus Torvalds [Sat, 17 Jul 2021 20:27:00 +0000 (13:27 -0700)]
Revert "mm/slub: use stackdepot to save stack trace in objects"

This reverts commit 788691464c29455346dc613a3b43c2fb9e5757a4.

It's not clear why, but it causes unexplained problems in entirely
unrelated xfs code.  The most likely explanation is some slab
corruption, possibly triggered due to CONFIG_SLUB_DEBUG_ON.  See [1].

It ends up having a few other problems too, like build errors on
arch/arc, and Geert reporting it using much more memory on m68k [3] (it
probably does so elsewhere too, but it is probably just more noticeable
on m68k).

The architecture issues (both build and memory use) are likely just
because this change effectively force-enabled STACKDEPOT (along with a
very bad default value for the stackdepot hash size).  But together with
the xfs issue, this all smells like "this commit was not ready" to me.

Link: https://lore.kernel.org/linux-xfs/YPE3l82acwgI2OiV@infradead.org/
Link: https://lore.kernel.org/lkml/202107150600.LkGNb4Vb-lkp@intel.com/
Link: https://lore.kernel.org/lkml/CAMuHMdW=eoVzM1Re5FVoEN87nKfiLmM2+Ah7eNu2KXEhCvbZyA@mail.gmail.com/
Reported-by: Christoph Hellwig <hch@infradead.org>
Reported-by: kernel test robot <lkp@intel.com>
Reported-by: Geert Uytterhoeven <geert@linux-m68k.org>
Cc: Andrew Morton <akpm@linux-foundation.org>
Cc: Vlastimil Babka <vbabka@suse.cz>
Cc: Randy Dunlap <rdunlap@infradead.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
6 months agoMerge tag 'scsi-fixes' of git://git.kernel.org/pub/scm/linux/kernel/git/jejb/scsi
Linus Torvalds [Sat, 17 Jul 2021 20:09:23 +0000 (13:09 -0700)]
Merge tag 'scsi-fixes' of git://git./linux/kernel/git/jejb/scsi

Pull SCSI fixes from James Bottomley:
 "One core fix for an oops which can occur if the error handling thread
  fails to start for some reason and the driver is removed.

  The other fixes are all minor ones in drivers"

* tag 'scsi-fixes' of git://git.kernel.org/pub/scm/linux/kernel/git/jejb/scsi:
  scsi: ufs: core: Add missing host_lock in ufshcd_vops_setup_xfer_req()
  scsi: mpi3mr: Fix W=1 compilation warnings
  scsi: pm8001: Clean up kernel-doc and comments
  scsi: zfcp: Report port fc_security as unknown early during remote cable pull
  scsi: core: Fix bad pointer dereference when ehandler kthread is invalid
  scsi: fas216: Fix a build error
  scsi: core: Fix the documentation of the scsi_execute() time parameter

6 months agoMerge tag '5.14-rc1-smb3-fixes' of git://git.samba.org/sfrench/cifs-2.6
Linus Torvalds [Sat, 17 Jul 2021 19:56:50 +0000 (12:56 -0700)]
Merge tag '5.14-rc1-smb3-fixes' of git://git.samba.org/sfrench/cifs-2.6

Pull cifs fixes from Steve French:
 "Eight cifs/smb3 fixes, including three for stable.

  Three are DFS related fixes, and two to fix problems pointed out by
  static checkers"

* tag '5.14-rc1-smb3-fixes' of git://git.samba.org/sfrench/cifs-2.6:
  cifs: do not share tcp sessions of dfs connections
  SMB3.1.1: fix mount failure to some servers when compression enabled
  cifs: added WARN_ON for all the count decrements
  cifs: fix missing null session check in mount
  cifs: handle reconnect of tcon when there is no cached dfs referral
  cifs: fix the out of range assignment to bit fields in parse_server_interfaces
  cifs: Do not use the original cruid when following DFS links for multiuser mounts
  cifs: use the expiry output of dns_query to schedule next resolution

6 months agoMerge tag 'linux-kselftest-kunit-fixes-5.14-rc2' of git://git.kernel.org/pub/scm...
Linus Torvalds [Sat, 17 Jul 2021 19:48:06 +0000 (12:48 -0700)]
Merge tag 'linux-kselftest-kunit-fixes-5.14-rc2' of git://git./linux/kernel/git/shuah/linux-kselftest

Pull kunit fixes from Shuah Khan:
 "Fixes to kunit tool and documentation:

   - fix asserts on older python versions

   - fixes to misleading error messages when TAP header format is
     incorrect or when file is missing

   - documentation fix: drop obsolete information about uml_abort
     coverage

   - remove unnecessary annotations"

* tag 'linux-kselftest-kunit-fixes-5.14-rc2' of git://git.kernel.org/pub/scm/linux/kernel/git/shuah/linux-kselftest:
  kunit: tool: Assert the version requirement
  kunit: tool: remove unnecessary "annotations" import
  Documentation: kunit: drop obsolete note about uml_abort for coverage
  kunit: tool: Fix error messages for cases of no tests and wrong TAP header

6 months agoMerge tag 'linux-kselftest-fixes-5.14-rc2' of git://git.kernel.org/pub/scm/linux...
Linus Torvalds [Sat, 17 Jul 2021 19:44:32 +0000 (12:44 -0700)]
Merge tag 'linux-kselftest-fixes-5.14-rc2' of git://git./linux/kernel/git/shuah/linux-kselftest

Pull kselftest fix from Shuah Khan:
 "A fix to memory-hotplug hot-remove test to stop spamming logs with
  dump_page() entries and slowing the system down to a crawl"

* tag 'linux-kselftest-fixes-5.14-rc2' of git://git.kernel.org/pub/scm/linux/kernel/git/shuah/linux-kselftest:
  selftests: memory-hotplug: avoid spamming logs with dump_page(), ratio limit hot-remove error test

6 months agoMerge tag 'trace-v5.14-5' of git://git.kernel.org/pub/scm/linux/kernel/git/rostedt...
Linus Torvalds [Sat, 17 Jul 2021 19:36:51 +0000 (12:36 -0700)]
Merge tag 'trace-v5.14-5' of git://git./linux/kernel/git/rostedt/linux-trace

Pull tracing fix from Steven Rostedt:
 "Fix the histogram logic from possibly crashing the kernel

  Working on the histogram code, I found that if you dereference a char
  pointer in a trace event that happens to point to user space, it can
  crash the kernel, as it does no checks of that pointer. I have code
  coming that will do this better, so just remove this ability to treat
  character pointers in trace events as stings in the histogram"

* tag 'trace-v5.14-5' of git://git.kernel.org/pub/scm/linux/kernel/git/rostedt/linux-trace:
  tracing: Do not reference char * as a string in histograms

6 months agoMerge tag 'devicetree-fixes-for-5.14-1' of git://git.kernel.org/pub/scm/linux/kernel...
Linus Torvalds [Sat, 17 Jul 2021 02:08:09 +0000 (19:08 -0700)]
Merge tag 'devicetree-fixes-for-5.14-1' of git://git./linux/kernel/git/robh/linux

Pull devicetree fixes from Rob Herring:

 - Drop 'resets' as required on renesas,du

 - Moving of fixed string patterns for 'properties' instead of
   'patternProperties'

 - Drop more redundant minItems/maxItems that we merged in the merge
   window

 - Indentation warning fix for sja1105

* tag 'devicetree-fixes-for-5.14-1' of git://git.kernel.org/pub/scm/linux/kernel/git/robh/linux:
  dt-bindings: display: renesas,du: Make resets optional on R-Car H1
  dt-bindings: Move fixed string 'patternProperties' to 'properties'
  dt-bindings: More dropping redundant minItems/maxItems
  dt-bindings: net: dsa: sja1105: Fix indentation warnings

6 months agoMerge tag 'arm64-fixes' of git://git.kernel.org/pub/scm/linux/kernel/git/arm64/linux
Linus Torvalds [Sat, 17 Jul 2021 02:00:53 +0000 (19:00 -0700)]
Merge tag 'arm64-fixes' of git://git./linux/kernel/git/arm64/linux

Pull arm64 fixes from Will Deacon:
 "The bulk of the diffstat consists of changes to our uaccess routines
  so that they fall back to bytewise copying prior to reporting complete
  failure when the initial (multi-byte) access faults.

  However, the most disappointing change here is that we've had to bump
  ARCH_DMA_MINALIGN back to 128 bytes thanks to Qualcomm's "Kryo" CPU,
  which ended up in the MSM8996 mobile SoC. Still, at least we're now
  aware of this design and one of the hardware designers confirmed the
  L2 cacheline size for us.

  Summary:

   - Fix instrumentation annotations for entry code

   - Ensure kernel MTE state is restored correctly on resume from suspend

   - Fix MTE fault from new strlen() routine

   - Fallback to byte-wise accesses on initial uaccess fault

   - Bump Clang requirement for BTI

   - Revert ARCH_DMA_MINALIGN back to 128 bytes (shakes fist at Qualcomm)"

* tag 'arm64-fixes' of git://git.kernel.org/pub/scm/linux/kernel/git/arm64/linux:
  arm64: entry: fix KCOV suppression
  arm64: entry: add missing noinstr
  arm64: mte: fix restoration of GCR_EL1 from suspend
  arm64: Avoid premature usercopy failure
  arm64: Restrict ARM64_BTI_KERNEL to clang 12.0.0 and newer
  Revert "arm64: cache: Lower ARCH_DMA_MINALIGN to 64 (L1_CACHE_BYTES)"
  arm64: Add missing header <asm/smp.h> in two files
  arm64: fix strlen() with CONFIG_KASAN_HW_TAGS

6 months agoARM: dts: versatile: Fix up interrupt controller node names
Sudeep Holla [Thu, 1 Jul 2021 13:21:18 +0000 (14:21 +0100)]
ARM: dts: versatile: Fix up interrupt controller node names

Once the new schema interrupt-controller/arm,vic.yaml is added, we get
the below warnings:

        arch/arm/boot/dts/versatile-ab.dt.yaml:
        intc@10140000: $nodename:0: 'intc@10140000' does not match
        '^interrupt-controller(@[0-9a-f,]+)*$'

arch/arm/boot/dts/versatile-ab.dt.yaml:
intc@10140000: 'clear-mask' does not match any of the regexes

Fix the node names for the interrupt controller to conform
to the standard node name interrupt-controller@.. Also drop invalid
clear-mask property.

Signed-off-by: Sudeep Holla <sudeep.holla@arm.com>
Acked-by: Linus Walleij <linus.walleij@linaro.org>
Link: https://lore.kernel.org/r/20210701132118.759454-1-sudeep.holla@arm.com'
Signed-off-by: Arnd Bergmann <arnd@arndb.de>
6 months agoMerge tag 'aspeed-5.14-devicetree-2' of git://git.kernel.org/pub/scm/linux/kernel...
Arnd Bergmann [Fri, 16 Jul 2021 21:04:13 +0000 (23:04 +0200)]
Merge tag 'aspeed-5.14-devicetree-2' of git://git./linux/kernel/git/joel/bmc into arm/fixes

ASPEED device tree fixes for 5.14

 - eMMC phase corrections so Tacoma and Everest can boot

 - VUART irq polarity fix for e3c246d4i, using new bindings

 - I2C address fix for Rainier power supply

 - GPIO line name fixes

* tag 'aspeed-5.14-devicetree-2' of git://git.kernel.org/pub/scm/linux/kernel/git/joel/bmc:
  ARM: dts: aspeed: everest: PSU #3 address change
  ARM: dts: everest: Add phase corrections for eMMC
  ARM: dts: tacoma: Add phase corrections for eMMC
  ARM: dts: aspeed: Update e3c246d4i vuart properties
  ARM: dts: aspeed: Fix AST2600 machines line names

Link: https://lore.kernel.org/r/CACPK8XefdPzeOUDnDgk9cHQEs-9wF_ZSPdYQRzuNOpGZTyGUKQ@mail.gmail.com
Signed-off-by: Arnd Bergmann <arnd@arndb.de>
6 months agoARM: multi_v7_defconfig: Make NOP_USB_XCEIV driver built-in
Stefan Wahren [Sat, 10 Jul 2021 11:04:55 +0000 (13:04 +0200)]
ARM: multi_v7_defconfig: Make NOP_USB_XCEIV driver built-in

The usage of usb-nop-xceiv PHY on Raspberry Pi boards with BCM283x has
been a "regression source" a lot of times. The last case is breakage of
USB mass storage boot has been commit e590474768f1 ("driver core: Set
fw_devlink=on by default") for multi_v7_defconfig. As long as
NOP_USB_XCEIV is configured as module, the dwc2 USB driver defer probing
endlessly and prevent booting from USB mass storage device. So make
the driver built-in as in bcm2835_defconfig and arm64/defconfig.

Fixes: e590474768f1 ("driver core: Set fw_devlink=on by default")
Reported-by: Ojaswin Mujoo <ojaswin98@gmail.com>
Signed-off-by: Stefan Wahren <stefan.wahren@i2se.com>
Link: https://lore.kernel.org/r/1625915095-23077-1-git-send-email-stefan.wahren@i2se.com'
Signed-off-by: Arnd Bergmann <arnd@arndb.de>
6 months agoARM: configs: Update u8500_defconfig
Linus Walleij [Mon, 12 Jul 2021 08:55:22 +0000 (10:55 +0200)]
ARM: configs: Update u8500_defconfig

The platform lost the framebuffer due to a commit solving a
circular dependency in v5.14-rc1, so add it back in by explicitly
selecting the framebuffer.

The U8500 has also gained a few systems using touchscreens from
Cypress, Melfas and Zinitix so add these at the same time as
we're updating the defconfig anyway.

Fixes: f611b1e7624c ("drm: Avoid circular dependencies for CONFIG_FB")
Signed-off-by: Linus Walleij <linus.walleij@linaro.org>
Cc: phone-devel@vger.kernel.org
Cc: Kees Cook <keescook@chromium.org>
Cc: Arnd Bergmann <arnd@kernel.org>
Cc: Stephan Gerhold <stephan@gerhold.net>
Cc: newbyte@disroot.org
Link: https://lore.kernel.org/r/20210712085522.672482-1-linus.walleij@linaro.org'
Signed-off-by: Arnd Bergmann <arnd@arndb.de>
6 months agoARM: configs: Update Vexpress defconfig
Linus Walleij [Tue, 13 Jul 2021 13:37:08 +0000 (15:37 +0200)]
ARM: configs: Update Vexpress defconfig

This updates the Versatile Express defconfig for the changes
in the v5.14-rc1 kernel:

- The Framebuffer CONFIG_FB needs to be explicitly selected
  or we don't get any framebuffer anymore. DRM has stopped to
  select FB because of circular dependency.
- CONFIG_CMA options were moved around.
- CONFIG_MODULES options were moved around.
- CONFIG_CRYPTO_HW was moved around.

Fixes: f611b1e7624c ("drm: Avoid circular dependencies for CONFIG_FB")
Signed-off-by: Linus Walleij <linus.walleij@linaro.org>
Acked-by: Sudeep Holla <sudeep.holla@arm.com>
Cc: Kees Cook <keescook@chromium.org>
Cc: Sudeep Holla <sudeep.holla@arm.com>
Link: https://lore.kernel.org/r/20210713133708.94397-1-linus.walleij@linaro.org'
Signed-off-by: Arnd Bergmann <arnd@arndb.de>
6 months agoARM: configs: Update Versatile defconfig
Linus Walleij [Wed, 14 Jul 2021 08:18:19 +0000 (10:18 +0200)]
ARM: configs: Update Versatile defconfig

This updates the Versatile defconfig for the changes
in the v5.14-rc1 kernel:

- The Framebuffer CONFIG_FB needs to be explicitly selected
  or we don't get any framebuffer anymore. DRM has stopped to
  select FB because of circular dependency.
- The CONFIG_FB_MODE_HELPERS are not needed when using DRM
  framebuffer emulation as DRM does.
- The Acorn fonts are removed, the default framebuffer font
  works fine. I don't know why this was selected in the first
  place or how the Kconfig was altered so it was removed.

Fixes: f611b1e7624c ("drm: Avoid circular dependencies for CONFIG_FB")
Signed-off-by: Linus Walleij <linus.walleij@linaro.org>
Reviewed-by: Kees Cook <keescook@chromium.org>
Cc: Kees Cook <keescook@chromium.org>
Link: https://lore.kernel.org/r/20210714081819.139210-1-linus.walleij@linaro.org'
Signed-off-by: Arnd Bergmann <arnd@arndb.de>
6 months agoARM: configs: Update RealView defconfig
Linus Walleij [Wed, 14 Jul 2021 09:00:40 +0000 (11:00 +0200)]
ARM: configs: Update RealView defconfig

This updates the RealView defconfig for the changes
in the v5.14-rc1 kernel:

- The Framebuffer CONFIG_FB needs to be explicitly selected
  or we don't get any framebuffer anymore. DRM has stopped to
  select FB because of circular dependency.
- The CONFIG_FB_MODE_HELPERS are not needed when using DRM
  framebuffer emulation as DRM does.
- Drop two unused penguin logos.

Fixes: f611b1e7624c ("drm: Avoid circular dependencies for CONFIG_FB")
Signed-off-by: Linus Walleij <linus.walleij@linaro.org>
Reviewed-by: Kees Cook <keescook@chromium.org>
Cc: Kees Cook <keescook@chromium.org>
Link: https://lore.kernel.org/r/20210714090040.182381-1-linus.walleij@linaro.org'
Signed-off-by: Arnd Bergmann <arnd@arndb.de>
6 months agoARM: configs: Update Integrator defconfig
Linus Walleij [Wed, 14 Jul 2021 12:27:03 +0000 (14:27 +0200)]
ARM: configs: Update Integrator defconfig

This updates the Integrator defconfig for the changes
in the v5.14-rc1 kernel:

- The Framebuffer CONFIG_FB needs to be explicitly selected
  or we don't get any framebuffer anymore. DRM has stopped to
  select FB because of circular dependency.
- Drop the unused Matrox FB drivers that are only used with
  specific PCI cards.

Fixes: f611b1e7624c ("drm: Avoid circular dependencies for CONFIG_FB")
Signed-off-by: Linus Walleij <linus.walleij@linaro.org>
Reviewed-by: Kees Cook <keescook@chromium.org>
Cc: Kees Cook <keescook@chromium.org>
Link: https://lore.kernel.org/r/20210714122703.212609-1-linus.walleij@linaro.org'
Signed-off-by: Arnd Bergmann <arnd@arndb.de>
6 months agoMerge tag 'scmi-fixes-5.14' of git://git.kernel.org/pub/scm/linux/kernel/git/sudeep...
Arnd Bergmann [Fri, 16 Jul 2021 21:01:24 +0000 (23:01 +0200)]
Merge tag 'scmi-fixes-5.14' of git://git./linux/kernel/git/sudeep.holla/linux into arm/fixes

ARM SCMI fixes for v5.14

A small set of fixes:
- adding check for presence of probe while registering the driver to
  prevent NULL pointer access
- dropping the duplicate check as the driver core already takes care of it
- fix for possible scmi_linux_errmap buffer overflow
- fix to avoid sensor message structure padding
- fix the range check for the maximum number of pending SCMI messages
- fix for various kernel-doc warnings

* tag 'scmi-fixes-5.14' of git://git.kernel.org/pub/scm/linux/kernel/git/sudeep.holla/linux:
  firmware: arm_scmi: Fix range check for the maximum number of pending messages
  firmware: arm_scmi: Avoid padding in sensor message structure
  firmware: arm_scmi: Fix kernel doc warnings about return values
  firmware: arm_scpi: Fix kernel doc warnings
  firmware: arm_scmi: Fix kernel doc warnings
  firmware: arm_scmi: Fix possible scmi_linux_errmap buffer overflow
  firmware: arm_scmi: Ensure drivers provide a probe function
  firmware: arm_scmi: Simplify device probe function on the bus

Link: https://lore.kernel.org/r/20210714165831.2617437-1-sudeep.holla@arm.com
Signed-off-by: Arnd Bergmann <arnd@arndb.de>
6 months agoMerge tag 'arm-ffa-fixes-5.14' of git://git.kernel.org/pub/scm/linux/kernel/git/sudee...
Arnd Bergmann [Fri, 16 Jul 2021 21:00:44 +0000 (23:00 +0200)]
Merge tag 'arm-ffa-fixes-5.14' of git://git./linux/kernel/git/sudeep.holla/linux into arm/fixes

Arm FF-A fixes for v5.14

A small set of fixes:
- adding check for presence of probe while registering the driver to
  prevent NULL pointer access
- dropping the duplicate check as the driver core already takes care of it
- fixing possible ffa_linux_errmap buffer overflow and
- fixing kernel-doc warning for comment style

* tag 'arm-ffa-fixes-5.14' of git://git.kernel.org/pub/scm/linux/kernel/git/sudeep.holla/linux:
  firmware: arm_ffa: Fix a possible ffa_linux_errmap buffer overflow
  firmware: arm_ffa: Fix the comment style
  firmware: arm_ffa: Simplify probe function
  firmware: arm_ffa: Ensure drivers provide a probe function

Link: https://lore.kernel.org/r/20210714165806.2617325-1-sudeep.holla@arm.com
Signed-off-by: Arnd Bergmann <arnd@arndb.de>
6 months agoarm: Typo s/PCI_IXP4XX_LEGACY/IXP4XX_PCI_LEGACY/
Geert Uytterhoeven [Wed, 14 Jul 2021 09:33:43 +0000 (11:33 +0200)]
arm: Typo s/PCI_IXP4XX_LEGACY/IXP4XX_PCI_LEGACY/

Kconfig symbol PCI_IXP4XX_LEGACY does not exist, but IXP4XX_PCI_LEGACY
does.

Fixes: d5d9f7ac58ea1041 ("ARM/ixp4xx: Make NEED_MACH_IO_H optional")
Signed-off-by: Geert Uytterhoeven <geert+renesas@glider.be>
Reviewed-by: Linus Walleij <linus.walleij@linaro.org>
Link: https://lore.kernel.org/r/82ce37c617293521f095a945a255456b9512769c.1626255077.git.geert+renesas@glider.be'
Signed-off-by: Arnd Bergmann <arnd@arndb.de>
6 months agoMerge tag 'tegra-for-5.14-arm64-dt-fixes' of git://git.kernel.org/pub/scm/linux/kerne...
Arnd Bergmann [Fri, 16 Jul 2021 20:57:07 +0000 (22:57 +0200)]
Merge tag 'tegra-for-5.14-arm64-dt-fixes' of git://git./linux/kernel/git/tegra/linux into arm/fixes

arm64: tegra: Device tree fixes for v5.14-rc1

This contains two late fixes for Tegra194 device tree files to restore
USB and audio functionality after enabling system-wide IOMMU support.

* tag 'tegra-for-5.14-arm64-dt-fixes' of git://git.kernel.org/pub/scm/linux/kernel/git/tegra/linux:
  arm64: tegra: Enable SMMU support for USB on Tegra194
  arm64: tegra: Enable audio IOMMU support on Tegra194

Link: https://lore.kernel.org/r/20210709150220.2543875-1-thierry.reding@gmail.com
Signed-off-by: Arnd Bergmann <arnd@arndb.de>
6 months agoMerge tag 'renesas-fixes-for-v5.14-tag1' of git://git.kernel.org/pub/scm/linux/kernel...
Arnd Bergmann [Fri, 16 Jul 2021 20:54:55 +0000 (22:54 +0200)]
Merge tag 'renesas-fixes-for-v5.14-tag1' of git://git./linux/kernel/git/geert/renesas-devel into arm/fixes

Renesas fixes for v5.14

  - Fix a clock/reset handling design issue on the new RZ/G2L SoC,
    requiring an atomic change to DT binding definitions, clock driver,
    and DTS,
  - Restore graphical consoles in the shmobile_defconfig.

* tag 'renesas-fixes-for-v5.14-tag1' of git://git.kernel.org/pub/scm/linux/kernel/git/geert/renesas-devel:
  ARM: shmobile: defconfig: Restore graphical consoles
  dt-bindings: clock: r9a07g044-cpg: Update clock/reset definitions
  clk: renesas: r9a07g044: Add P2 Clock support
  clk: renesas: r9a07g044: Fix P1 Clock
  clk: renesas: r9a07g044: Rename divider table
  clk: renesas: rzg2l: Add multi clock PM support

Link: https://lore.kernel.org/r/cover.1626253929.git.geert+renesas@glider.be
Signed-off-by: Arnd Bergmann <arnd@arndb.de>
6 months agoMerge tag 'memory-controller-drv-tegra-5.14-3' of git://git.kernel.org/pub/scm/linux...
Arnd Bergmann [Fri, 16 Jul 2021 20:50:56 +0000 (22:50 +0200)]
Merge tag 'memory-controller-drv-tegra-5.14-3' of git://git./linux/kernel/git/krzk/linux-mem-ctrl into arm/fixes

Memory controller drivers for v5.14 - Tegra SoC, late fixes

Two fixes for recent series of changes in Tegra SoC memory controller
drivers:
1. Add a stub for tegra_mc_probe_device() to fix compile testing of
   arm-smmu without TEGRA_MC.
2. Fix arm-smmu dtschema syntax.

* tag 'memory-controller-drv-tegra-5.14-3' of git://git.kernel.org/pub/scm/linux/kernel/git/krzk/linux-mem-ctrl:
  dt-bindings: arm-smmu: Fix json-schema syntax
  memory: tegra: Add compile-test stub for tegra_mc_probe_device()

Link: https://lore.kernel.org/r/20210625073604.13562-1-krzysztof.kozlowski@canonical.com
Signed-off-by: Arnd Bergmann <arnd@arndb.de>
6 months agoMerge tag 'docs-5.14-2' of git://git.lwn.net/linux
Linus Torvalds [Fri, 16 Jul 2021 20:35:04 +0000 (13:35 -0700)]
Merge tag 'docs-5.14-2' of git://git.lwn.net/linux

Pull documentation fixes from Jonathan Corbet:
 "A handful of fixes in and around documentation.

  Some funky quotes in LICENSES/dual/CC-BY-4.0 were giving spdxcheck.py
  grief; that has been fixed on both ends. Also a couple of features
  updates and one docs build fix"

* tag 'docs-5.14-2' of git://git.lwn.net/linux:
  docs/zh_CN: add a missing space character
  Documentation/features: Add THREAD_INFO_IN_TASK feature matrix
  Documentation/features: Update the ARCH_HAS_TICK_BROADCAST entry
  LICENSES/dual/CC-BY-4.0: Git rid of "smart quotes"
  scripts/spdxcheck.py: Strictly read license files in utf-8

6 months agoMerge tag 'block-5.14-2021-07-16' of git://git.kernel.dk/linux-block
Linus Torvalds [Fri, 16 Jul 2021 19:31:44 +0000 (12:31 -0700)]
Merge tag 'block-5.14-2021-07-16' of git://git.kernel.dk/linux-block

Pull block fixes from Jens Axboe:

 - NVMe fixes via Christoph:
      - fix various races in nvme-pci when shutting down just after
        probing (Casey Chen)
      - fix a net_device leak in nvme-tcp (Prabhakar Kushwaha)

 - Fix regression in xen-blkfront by cleaning up the removal state
   machine (Christoph)

 - Fix tag_set and queue cleanup ordering regression in nbd (Wang)

 - Fix tag_set and queue cleanup ordering regression in pd (Guoqing)

* tag 'block-5.14-2021-07-16' of git://git.kernel.dk/linux-block:
  xen-blkfront: sanitize the removal state machine
  nbd: fix order of cleaning up the queue and freeing the tagset
  pd: fix order of cleaning up the queue and freeing the tagset
  nvme-pci: do not call nvme_dev_remove_admin from nvme_remove
  nvme-pci: fix multiple races in nvme_setup_io_queues
  nvme-tcp: use __dev_get_by_name instead dev_get_by_name for OPT_HOST_IFACE

6 months agoMerge tag 'io_uring-5.14-2021-07-16' of git://git.kernel.dk/linux-block
Linus Torvalds [Fri, 16 Jul 2021 19:27:33 +0000 (12:27 -0700)]
Merge tag 'io_uring-5.14-2021-07-16' of git://git.kernel.dk/linux-block

Pull io_uring fixes from Jens Axboe:
 "Two small fixes: one fixing the process target of a check, and the
  other a minor issue with the drain error handling"

* tag 'io_uring-5.14-2021-07-16' of git://git.kernel.dk/linux-block:
  io_uring: fix io_drain_req()
  io_uring: use right task for exiting checks

6 months agoMerge tag 'zonefs-5.14-rc2' of git://git.kernel.org/pub/scm/linux/kernel/git/dlemoal...
Linus Torvalds [Fri, 16 Jul 2021 18:20:53 +0000 (11:20 -0700)]
Merge tag 'zonefs-5.14-rc2' of git://git./linux/kernel/git/dlemoal/zonefs

Pull zonefs fix from Damien Le Moal:
 "A single patch to remove an unnecessary NULL bio check (from
  Xianting)"

* tag 'zonefs-5.14-rc2' of git://git.kernel.org/pub/scm/linux/kernel/git/dlemoal/zonefs:
  zonefs: remove redundant null bio check

6 months agoMerge tag 'drm-fixes-2021-07-16' of git://anongit.freedesktop.org/drm/drm
Linus Torvalds [Fri, 16 Jul 2021 18:14:54 +0000 (11:14 -0700)]
Merge tag 'drm-fixes-2021-07-16' of git://anongit.freedesktop.org/drm/drm

Pull drm fixes from Dave Airlie:
 "Regular rc2 fixes though a bit more than usual at rc2 stage, people
  must have been testing early or else some fixes from last week got a
  bit laggy.

  There is one larger change in the amd fixes to amalgamate some power
  management code on the newer chips with the code from the older chips,
  it should only affects chips where support was introduced in rc1 and
  it should make future fixes easier to maintain probably a good idea to
  merge it now.

  Otherwise it's mostly fixes across the board.

  dma-buf:
   - Fix fence leak in sync_file_merge() error code

  drm/panel:
   - nt35510: Don't fail on DSI reads

  fbdev:
   - Avoid use-after-free by not deleting current video mode

  ttm:
   - Avoid NULL-ptr deref in ttm_range_man_fini()

  vmwgfx:
   - Fix a merge commit

  qxl:
   - fix a TTM regression

  amdgpu:
   - SR-IOV fixes
   - RAS fixes
   - eDP fixes
   - SMU13 code unification to facilitate fixes in the future
   - Add new renoir DID
   - Yellow Carp fixes
   - Beige Goby fixes
   - Revert a bunch of TLB fixes that caused regressions
   - Revert an LTTPR display regression

  amdkfd
   - Fix VRAM access regression
   - SVM fixes

  i915:
   - Fix -EDEADLK handling regression
   - Drop the page table optimisation"

* tag 'drm-fixes-2021-07-16' of git://anongit.freedesktop.org/drm/drm: (29 commits)
  drm/amdgpu: add another Renoir DID
  drm/ttm: add a check against null pointer dereference
  drm/i915/gtt: drop the page table optimisation
  drm/i915/gt: Fix -EDEADLK handling regression
  drm/amd/pm: Add waiting for response of mode-reset message for yellow carp
  Revert "drm/amdkfd: Add heavy-weight TLB flush after unmapping"
  Revert "drm/amdgpu: Add table_freed parameter to amdgpu_vm_bo_update"
  Revert "drm/amdkfd: Make TLB flush conditional on mapping"
  Revert "drm/amdgpu: Fix warning of Function parameter or member not described"
  Revert "drm/amdkfd: Add memory sync before TLB flush on unmap"
  drm/amd/pm: Fix BACO state setting for Beige_Goby
  drm/amdgpu: Restore msix after FLR
  drm/amdkfd: Allow CPU access for all VRAM BOs
  drm/amdgpu/display - only update eDP's backlight level when necessary
  drm/amdkfd: handle fault counters on invalid address
  drm/amdgpu: Correct the irq numbers for virtual crtc
  drm/amd/display: update header file name
  drm/amd/pm: drop smu_v13_0_1.c|h files for yellow carp
  drm/amd/display: remove faulty assert
  Revert "drm/amd/display: Always write repeater mode regardless of LTTPR"
  ...

6 months agoMerge branch 'urgent' of git://git.kernel.org/pub/scm/linux/kernel/git/paulmck/linux-rcu
Linus Torvalds [Fri, 16 Jul 2021 18:08:57 +0000 (11:08 -0700)]
Merge branch 'urgent' of git://git./linux/kernel/git/paulmck/linux-rcu

Pull RCU fixes from Paul McKenney:

 - fix regressions induced by a merge-window change in scheduler
   semantics, which means that smp_processor_id() can no longer be used
   in kthreads using simple affinity to bind themselves to a specific
   CPU.

 - fix a bug in Tasks Trace RCU that was thought to be strictly
   theoretical. However, production workloads have started hitting this,
   so these fixes need to be merged sooner rather than later.

 - fix a minor printk()-format-mismatch issue introduced during the
   merge window.

* 'urgent' of git://git.kernel.org/pub/scm/linux/kernel/git/paulmck/linux-rcu:
  rcu: Fix pr_info() formats and values in show_rcu_gp_kthreads()
  rcu-tasks: Don't delete holdouts within trc_wait_for_one_reader()
  rcu-tasks: Don't delete holdouts within trc_inspect_reader()
  refscale: Avoid false-positive warnings in ref_scale_reader()
  scftorture: Avoid false-positive warnings in scftorture_invoker()

6 months agodt-bindings: display: renesas,du: Make resets optional on R-Car H1
Geert Uytterhoeven [Wed, 14 Jul 2021 10:19:36 +0000 (12:19 +0200)]
dt-bindings: display: renesas,du: Make resets optional on R-Car H1

The "resets" property is not present on R-Car Gen1 SoCs.
Supporting it would require migrating from renesas,cpg-clocks to
renesas,cpg-mssr.

Reflect this in the DT bindings by removing the global "required:
resets".  All SoCs that do have "resets" properties already have
SoC-specific rules making it required.

Fixes: 99d66127fad25ebb ("dt-bindings: display: renesas,du: Convert binding to YAML")
Signed-off-by: Geert Uytterhoeven <geert+renesas@glider.be>
Reviewed-by: Laurent Pinchart <laurent.pinchart@ideasonboard.com>
Link: https://lore.kernel.org/r/98575791b154d80347d5b78132c1d53f5315ee62.1626257936.git.geert+renesas@glider.be
Signed-off-by: Rob Herring <robh@kernel.org>
6 months agoperf test bpf: Free obj_buf
Riccardo Mancini [Thu, 15 Jul 2021 16:07:24 +0000 (18:07 +0200)]
perf test bpf: Free obj_buf

ASan reports some memory leaks when running:

  # perf test "42: BPF filter"

The first of these leaks is caused by obj_buf never being deallocated in
__test__bpf.

This patch adds the missing free.

Signed-off-by: Riccardo Mancini <rickyman7@gmail.com>
Fixes: ba1fae431e74bb42 ("perf test: Add 'perf test BPF'")
Cc: Ian Rogers <irogers@google.com>
Cc: Jiri Olsa <jolsa@redhat.com>
Cc: Mark Rutland <mark.rutland@arm.com>
Cc: Namhyung Kim <namhyung@kernel.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Wang Nan <wangnan0@huawei.com>
Link: http://lore.kernel.org/lkml/60f3ca935fe6672e7e866276ce6264c9e26e4c87.1626343282.git.rickyman7@gmail.com
[ Added missing stdlib.h include ]
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
6 months agocifs: do not share tcp sessions of dfs connections
Paulo Alcantara [Fri, 16 Jul 2021 00:53:53 +0000 (21:53 -0300)]
cifs: do not share tcp sessions of dfs connections

Make sure that we do not share tcp sessions of dfs mounts when
mounting regular shares that connect to same server.  DFS connections
rely on a single instance of tcp in order to do failover properly in
cifs_reconnect().

Signed-off-by: Paulo Alcantara (SUSE) <pc@cjr.nz>
Signed-off-by: Steve French <stfrench@microsoft.com>
6 months agozonefs: remove redundant null bio check
Xianting Tian [Wed, 30 Jun 2021 14:33:36 +0000 (10:33 -0400)]
zonefs: remove redundant null bio check

bio_alloc() with __GFP_DIRECT_RECLAIM, which is included in
GFP_NOFS, never fails, see comments in bio_alloc_bioset().

Signed-off-by: Xianting Tian <xianting.tian@linux.alibaba.com>
Signed-off-by: Damien Le Moal <damien.lemoal@wdc.com>
6 months agoRevert "Makefile: Enable -Wimplicit-fallthrough for Clang"
Linus Torvalds [Fri, 16 Jul 2021 01:05:31 +0000 (18:05 -0700)]
Revert "Makefile: Enable -Wimplicit-fallthrough for Clang"

This reverts commit b7eb335e26a9c7f258c96b3962c283c379d3ede0.

It turns out that the problem with the clang -Wimplicit-fallthrough
warning is not about the kernel source code, but about clang itself, and
that the warning is unusable until clang fixes its broken ways.

In particular, when you enable this warning for clang, you not only get
warnings about implicit fallthroughs.  You also get this:

   warning: fallthrough annotation in unreachable code [-Wimplicit-fallthrough]

which is completely broken becasue it

 (a) doesn't even tell you where the problem is (seriously: no line
     numbers, no filename, no nothing).

 (b) is fundamentally broken anyway, because there are perfectly valid
     reasons to have a fallthrough statement even if it turns out that
     it can perhaps not be reached.

In the kernel, an example of that second case is code in the scheduler:

                switch (state) {
                case cpuset:
                        if (IS_ENABLED(CONFIG_CPUSETS)) {
                                cpuset_cpus_allowed_fallback(p);
                                state = possible;
                                break;
                        }
                        fallthrough;
                case possible:

where if CONFIG_CPUSETS is enabled you actually never hit the
fallthrough case at all.  But that in no way makes the fallthrough
wrong.

So the warning is completely broken, and enabling it for clang is a very
bad idea.

In the meantime, we can keep the gcc option enabled, and make the gcc
build use

    -Wimplicit-fallthrough=5

which means that we will at least continue to require a proper
fallthrough statement, and that gcc won't silently accept the magic
comment versions. Because gcc does this all correctly, and while the odd
"=5" part is kind of obscure, it's documented in [1]:

  "-Wimplicit-fallthrough=5 doesn’t recognize any comments as
   fallthrough comments, only attributes disable the warning"

so if clang ever fixes its bad behavior we can try enabling it there again.

Link: https://gcc.gnu.org/onlinedocs/gcc/Warning-Options.html
Cc: Kees Cook <keescook@chromium.org>
Cc: Gustavo A. R. Silva <gustavoars@kernel.org>
Cc: Nathan Chancellor <nathan@kernel.org>
Cc: Nick Desaulniers <ndesaulniers@google.com>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
6 months agoMerge tag 'drm-intel-fixes-2021-07-15' of git://anongit.freedesktop.org/drm/drm-intel...
Dave Airlie [Fri, 16 Jul 2021 00:52:58 +0000 (10:52 +1000)]
Merge tag 'drm-intel-fixes-2021-07-15' of git://anongit.freedesktop.org/drm/drm-intel into drm-fixes

Two regression fixes targeting stable:
- Fix -EDEADLK handling regression (Ville)
- Drop the page table optimisation (Matt)

Signed-off-by: Dave Airlie <airlied@redhat.com>
From: Rodrigo Vivi <rodrigo.vivi@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/YPA8y1DSCp2EbtpC@intel.com
6 months agoMerge tag 'configfs-5.13-1' of git://git.infradead.org/users/hch/configfs
Linus Torvalds [Fri, 16 Jul 2021 00:36:36 +0000 (17:36 -0700)]
Merge tag 'configfs-5.13-1' of git://git.infradead.org/users/hch/configfs

Pull configfs fix from Christoph Hellwig:

 - fix the read and write iterators (Bart Van Assche)

* tag 'configfs-5.13-1' of git://git.infradead.org/users/hch/configfs:
  configfs: fix the read and write iterators

6 months agoMerge tag 'drm-misc-fixes-2021-07-15' of git://anongit.freedesktop.org/drm/drm-misc...
Dave Airlie [Thu, 15 Jul 2021 23:51:41 +0000 (09:51 +1000)]
Merge tag 'drm-misc-fixes-2021-07-15' of git://anongit.freedesktop.org/drm/drm-misc into drm-fixes

Short summary of fixes pull (less than what git shortlog provides):

 * fbdev: Avoid use-after-free by not deleting current video mode
 * ttm: Avoid NULL-ptr deref in ttm_range_man_fini()
 * vmwgfx: Fix a merge commit

Signed-off-by: Dave Airlie <airlied@redhat.com>
From: Thomas Zimmermann <tzimmermann@suse.de>
Link: https://patchwork.freedesktop.org/patch/msgid/YO/yoFO+iSEqnIH0@linux-uq9g
6 months agoMerge tag 'pwm/for-5.14-rc2' of git://git.kernel.org/pub/scm/linux/kernel/git/thierry...
Linus Torvalds [Fri, 16 Jul 2021 00:29:44 +0000 (17:29 -0700)]
Merge tag 'pwm/for-5.14-rc2' of git://git./linux/kernel/git/thierry.reding/linux-pwm

Pull pwm fixes from Thierry Reding:
 "A couple of fixes from Uwe that I missed for v5.14-rc1"

* tag 'pwm/for-5.14-rc2' of git://git.kernel.org/pub/scm/linux/kernel/git/thierry.reding/linux-pwm:
  pwm: ep93xx: Ensure configuring period and duty_cycle isn't wrongly skipped
  pwm: berlin: Ensure configuring period and duty_cycle isn't wrongly skipped
  pwm: tiecap: Ensure configuring period and duty_cycle isn't wrongly skipped
  pwm: spear: Ensure configuring period and duty_cycle isn't wrongly skipped
  pwm: sprd: Ensure configuring period and duty_cycle isn't wrongly skipped

6 months agoSMB3.1.1: fix mount failure to some servers when compression enabled
Steve French [Thu, 15 Jul 2021 04:32:09 +0000 (23:32 -0500)]
SMB3.1.1: fix mount failure to some servers when compression enabled

When sending the compression context to some servers, they rejected
the SMB3.1.1 negotiate protocol because they expect the compression
context to have a data length of a multiple of 8.

Reviewed-by: Shyam Prasad N <sprasad@microsoft.com>
Signed-off-by: Steve French <stfrench@microsoft.com>
6 months agocifs: added WARN_ON for all the count decrements
Shyam Prasad N [Thu, 15 Jul 2021 04:00:00 +0000 (23:00 -0500)]
cifs: added WARN_ON for all the count decrements

We have a few ref counters srv_count, ses_count and
tc_count which we use for ref counting. Added a WARN_ON
during the decrement of each of these counters to make
sure that they don't go below their minimum values.

Signed-off-by: Shyam Prasad N <sprasad@microsoft.com>
Signed-off-by: Steve French <stfrench@microsoft.com>
6 months agocifs: fix missing null session check in mount
Steve French [Wed, 14 Jul 2021 00:40:33 +0000 (19:40 -0500)]
cifs: fix missing null session check in mount

Although it is unlikely to be have ended up with a null
session pointer calling cifs_try_adding_channels in cifs_mount.
Coverity correctly notes that we are already checking for
it earlier (when we return from do_dfs_failover), so at
a minimum to clarify the code we should make sure we also
check for it when we exit the loop so we don't end up calling
cifs_try_adding_channels or mount_setup_tlink with a null
ses pointer.

Addresses-Coverity: 1505608 ("Derefernce after null check")
Reviewed-by: Paulo Alcantara (SUSE) <pc@cjr.nz>
Signed-off-by: Steve French <stfrench@microsoft.com>
6 months agocifs: handle reconnect of tcon when there is no cached dfs referral
Paulo Alcantara [Mon, 12 Jul 2021 15:38:24 +0000 (12:38 -0300)]
cifs: handle reconnect of tcon when there is no cached dfs referral

When there is no cached DFS referral of tcon->dfs_path, then reconnect
to same share.

Signed-off-by: Paulo Alcantara (SUSE) <pc@cjr.nz>
Cc: <stable@vger.kernel.org>
Signed-off-by: Steve French <stfrench@microsoft.com>
6 months agoMerge tag 'amd-drm-fixes-5.14-2021-07-14' of https://gitlab.freedesktop.org/agd5f...
Dave Airlie [Thu, 15 Jul 2021 23:27:45 +0000 (09:27 +1000)]
Merge tag 'amd-drm-fixes-5.14-2021-07-14' of https://gitlab.freedesktop.org/agd5f/linux into drm-fixes

amd-drm-fixes-5.14-2021-07-14:

amdgpu:
- SR-IOV fixes
- RAS fixes
- eDP fixes
- SMU13 code unification to facilitate fixes in the future
- Add new renoir DID
- Yellow Carp fixes
- Beige Goby fixes
- Revert a bunch of TLB fixes that caused regressions
- Revert an LTTPR display regression

amdkfd
- Fix VRAM access regression
- SVM fixes

Signed-off-by: Dave Airlie <airlied@redhat.com>
From: Alex Deucher <alexander.deucher@amd.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20210714220858.5553-1-alexander.deucher@amd.com
6 months agotracing: Do not reference char * as a string in histograms
Steven Rostedt (VMware) [Thu, 15 Jul 2021 04:02:06 +0000 (00:02 -0400)]
tracing: Do not reference char * as a string in histograms

The histogram logic was allowing events with char * pointers to be used as
normal strings. But it was easy to crash the kernel with:

 # echo 'hist:keys=filename' > events/syscalls/sys_enter_openat/trigger

And open some files, and boom!

 BUG: unable to handle page fault for address: 00007f2ced0c3280
 #PF: supervisor read access in kernel mode
 #PF: error_code(0x0000) - not-present page
 PGD 1173fa067 P4D 1173fa067 PUD 1171b6067 PMD 1171dd067 PTE 0
 Oops: 0000 [#1] PREEMPT SMP
 CPU: 6 PID: 1810 Comm: cat Not tainted 5.13.0-rc5-test+ #61
 Hardware name: Hewlett-Packard HP Compaq Pro 6300 SFF/339A, BIOS K01
v03.03 07/14/2016
 RIP: 0010:strlen+0x0/0x20
 Code: f6 82 80 2a 0b a9 20 74 11 0f b6 50 01 48 83 c0 01 f6 82 80 2a 0b
a9 20 75 ef c3 66 66 2e 0f 1f 84 00 00 00 00 00 0f 1f 40 00 <80> 3f 00 74
10 48 89 f8 48 83 c0 01 80 38 00 75 f7 48 29 f8 c3

 RSP: 0018:ffffbdbf81567b50 EFLAGS: 00010246
 RAX: 0000000000000003 RBX: ffff93815cdb3800 RCX: ffff9382401a22d0
 RDX: 0000000000000100 RSI: 0000000000000000 RDI: 00007f2ced0c3280
 RBP: 0000000000000100 R08: ffff9382409ff074 R09: ffffbdbf81567c98
 R10: ffff9382409ff074 R11: 0000000000000000 R12: ffff9382409ff074
 R13: 0000000000000001 R14: ffff93815a744f00 R15: 00007f2ced0c3280
 FS:  00007f2ced0f8580(0000) GS:ffff93825a800000(0000)
knlGS:0000000000000000
 CS:  0010 DS: 0000 ES: 0000 CR0: 0000000080050033
 CR2: 00007f2ced0c3280 CR3: 0000000107069005 CR4: 00000000001706e0
 Call Trace:
  event_hist_trigger+0x463/0x5f0
  ? find_held_lock+0x32/0x90
  ? sched_clock_cpu+0xe/0xd0
  ? lock_release+0x155/0x440
  ? kernel_init_free_pages+0x6d/0x90
  ? preempt_count_sub+0x9b/0xd0
  ? kernel_init_free_pages+0x6d/0x90
  ? get_page_from_freelist+0x12c4/0x1680
  ? __rb_reserve_next+0xe5/0x460
  ? ring_buffer_lock_reserve+0x12a/0x3f0
  event_triggers_call+0x52/0xe0
  ftrace_syscall_enter+0x264/0x2c0
  syscall_trace_enter.constprop.0+0x1ee/0x210
  do_syscall_64+0x1c/0x80
  entry_SYSCALL_64_after_hwframe+0x44/0xae

Where it triggered a fault on strlen(key) where key was the filename.

The reason is that filename is a char * to user space, and the histogram
code just blindly dereferenced it, with obvious bad results.

I originally tried to use strncpy_from_user/kernel_nofault() but found
that there's other places that its dereferenced and not worth the effort.

Just do not allow "char *" to act like strings.

Link: https://lkml.kernel.org/r/20210715000206.025df9d2@rorschach.local.home
Cc: Ingo Molnar <mingo@kernel.org>
Cc: Andrew Morton <akpm@linux-foundation.org>
Cc: Masami Hiramatsu <mhiramat@kernel.org>
Cc: Tzvetomir Stoyanov <tz.stoyanov@gmail.com>
Cc: stable@vger.kernel.org
Acked-by: Namhyung Kim <namhyung@kernel.org>
Acked-by: Tom Zanussi <zanussi@kernel.org>
Fixes: 79e577cbce4c4 ("tracing: Support string type key properly")
Fixes: 5967bd5c4239 ("tracing: Let filter_assign_type() detect FILTER_PTR_STRING")
Signed-off-by: Steven Rostedt (VMware) <rostedt@goodmis.org>
6 months agoMerge tag 'Wimplicit-fallthrough-clang-5.14-rc2' of git://git.kernel.org/pub/scm...
Linus Torvalds [Thu, 15 Jul 2021 20:57:31 +0000 (13:57 -0700)]
Merge tag 'Wimplicit-fallthrough-clang-5.14-rc2' of git://git./linux/kernel/git/gustavoars/linux

Pull fallthrough fixes from Gustavo Silva:
 "This fixes many fall-through warnings when building with Clang and
  -Wimplicit-fallthrough, and also enables -Wimplicit-fallthrough for
  Clang, globally.

  It's also important to notice that since we have adopted the use of
  the pseudo-keyword macro fallthrough, we also want to avoid having
  more /* fall through */ comments being introduced. Contrary to GCC,
  Clang doesn't recognize any comments as implicit fall-through markings
  when the -Wimplicit-fallthrough option is enabled.

  So, in order to avoid having more comments being introduced, we use
  the option -Wimplicit-fallthrough=5 for GCC, which similar to Clang,
  will cause a warning in case a code comment is intended to be used as
  a fall-through marking. The patch for Makefile also enforces this.

  We had almost 4,000 of these issues for Clang in the beginning, and
  there might be a couple more out there when building some
  architectures with certain configurations. However, with the recent
  fixes I think we are in good shape and it is now possible to enable
  the warning for Clang"

* tag 'Wimplicit-fallthrough-clang-5.14-rc2' of git://git.kernel.org/pub/scm/linux/kernel/git/gustavoars/linux: (27 commits)
  Makefile: Enable -Wimplicit-fallthrough for Clang
  powerpc/smp: Fix fall-through warning for Clang
  dmaengine: mpc512x: Fix fall-through warning for Clang
  usb: gadget: fsl_qe_udc: Fix fall-through warning for Clang
  powerpc/powernv: Fix fall-through warning for Clang
  MIPS: Fix unreachable code issue
  MIPS: Fix fall-through warnings for Clang
  ASoC: Mediatek: MT8183: Fix fall-through warning for Clang
  power: supply: Fix fall-through warnings for Clang
  dmaengine: ti: k3-udma: Fix fall-through warning for Clang
  s390: Fix fall-through warnings for Clang
  dmaengine: ipu: Fix fall-through warning for Clang
  iommu/arm-smmu-v3: Fix fall-through warning for Clang
  mmc: jz4740: Fix fall-through warning for Clang
  PCI: Fix fall-through warning for Clang
  scsi: libsas: Fix fall-through warning for Clang
  video: fbdev: Fix fall-through warning for Clang
  math-emu: Fix fall-through warning
  cpufreq: Fix fall-through warning for Clang
  drm/msm: Fix fall-through warning in msm_gem_new_impl()
  ...

6 months agoperf trace: Free strings in trace__parse_events_option()
Riccardo Mancini [Thu, 15 Jul 2021 16:07:23 +0000 (18:07 +0200)]
perf trace: Free strings in trace__parse_events_option()

ASan reports several memory leaks running:

  # perf test "88: Check open filename arg using perf trace + vfs_getname"

The fourth of these leaks is related to some strings never being freed
in trace__parse_events_option.

This patch adds the missing frees.

Signed-off-by: Riccardo Mancini <rickyman7@gmail.com>
Cc: Ian Rogers <irogers@google.com>
Cc: Jiri Olsa <jolsa@redhat.com>
Cc: Mark Rutland <mark.rutland@arm.com>
Cc: Namhyung Kim <namhyung@kernel.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Link: http://lore.kernel.org/lkml/34d08535b11124106b859790549991abff5a7de8.1626343282.git.rickyman7@gmail.com
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
6 months agoperf trace: Free syscall tp fields in evsel->priv
Riccardo Mancini [Thu, 15 Jul 2021 16:07:22 +0000 (18:07 +0200)]
perf trace: Free syscall tp fields in evsel->priv

ASan reports several memory leaks running:

  # perf test "88: Check open filename arg using perf trace + vfs_getname"

The third of these leaks is related to evsel->priv fields of sycalls
never being deallocated.

This patch adds the function evlist__free_syscall_tp_fields which
iterates over all evsels in evlist, matching syscalls, and calling the
missing frees.

This new function is called at the end of trace__run, right before
calling evlist__delete.

Signed-off-by: Riccardo Mancini <rickyman7@gmail.com>
Cc: Ian Rogers <irogers@google.com>
Cc: Jiri Olsa <jolsa@redhat.com>
Cc: Mark Rutland <mark.rutland@arm.com>
Cc: Namhyung Kim <namhyung@kernel.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Link: http://lore.kernel.org/lkml/46526611904ec5ff2768b59014e3afce8e0197d1.1626343282.git.rickyman7@gmail.com
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
6 months agoperf trace: Free syscall->arg_fmt
Riccardo Mancini [Thu, 15 Jul 2021 16:07:21 +0000 (18:07 +0200)]
perf trace: Free syscall->arg_fmt

ASan reports several memory leaks running:

  # perf test "88: Check open filename arg using perf trace + vfs_getname"

The second of these leaks is caused by the arg_fmt field of syscall not
being deallocated.

This patch adds a new function syscall__exit which is called on all
syscalls.table entries in trace__exit, which will free the arg_fmt
field.

Signed-off-by: Riccardo Mancini <rickyman7@gmail.com>
Cc: Ian Rogers <irogers@google.com>
Cc: Jiri Olsa <jolsa@redhat.com>
Cc: Mark Rutland <mark.rutland@arm.com>
Cc: Namhyung Kim <namhyung@kernel.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Link: http://lore.kernel.org/lkml/d68f25c043d30464ac9fa79c3399e18f429bca82.1626343282.git.rickyman7@gmail.com
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
6 months agoperf trace: Free malloc'd trace fields on exit
Riccardo Mancini [Thu, 15 Jul 2021 16:07:20 +0000 (18:07 +0200)]
perf trace: Free malloc'd trace fields on exit

ASan reports several memory leaks running:

  # perf test "88: Check open filename arg using perf trace + vfs_getname"

The first of these leaks is related to struct trace fields never being
deallocated.

This patch adds the function trace__exit, which is called at the end of
cmd_trace, replacing the existing deallocation, which is now moved
inside the new function.

This function deallocates:

 - ev_qualifier
 - ev_qualifier_ids.entries
 - syscalls.table
 - sctbl
 - perfconfig_events

Signed-off-by: Riccardo Mancini <rickyman7@gmail.com>
Cc: Ian Rogers <irogers@google.com>
Cc: Jiri Olsa <jolsa@redhat.com>
Cc: Mark Rutland <mark.rutland@arm.com>
Cc: Namhyung Kim <namhyung@kernel.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Link: http://lore.kernel.org/lkml/de5945ed5c0cb882cbfa3268567d0bff460ff016.1626343282.git.rickyman7@gmail.com
[ Removed needless initialization to zero, missing named initializers are zeroed by the compiler ]
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
6 months agoperf lzma: Close lzma stream on exit
Riccardo Mancini [Thu, 15 Jul 2021 16:07:19 +0000 (18:07 +0200)]
perf lzma: Close lzma stream on exit

ASan reports memory leaks when running:

  # perf test "88: Check open filename arg using perf trace + vfs_getname"

One of these is caused by the lzma stream never being closed inside
lzma_decompress_to_file().

This patch adds the missing lzma_end().

Signed-off-by: Riccardo Mancini <rickyman7@gmail.com>
Fixes: 80a32e5b498a7547 ("perf tools: Add lzma decompression support for kernel module")
Cc: Ian Rogers <irogers@google.com>
Cc: Jiri Olsa <jolsa@redhat.com>
Cc: Mark Rutland <mark.rutland@arm.com>
Cc: Namhyung Kim <namhyung@kernel.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Link: http://lore.kernel.org/lkml/aaf50bdce7afe996cfc06e1bbb36e4a2a9b9db93.1626343282.git.rickyman7@gmail.com
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
6 months agoperf script: Fix memory 'threads' and 'cpus' leaks on exit
Riccardo Mancini [Thu, 15 Jul 2021 16:07:18 +0000 (18:07 +0200)]
perf script: Fix memory 'threads' and 'cpus' leaks on exit

ASan reports several memory leaks while running:

  # perf test "82: Use vfs_getname probe to get syscall args filenames"

Two of these are caused by some refcounts not being decreased on
perf-script exit, namely script.threads and script.cpus.

This patch adds the missing __put calls in a new perf_script__exit
function, which is called at the end of cmd_script.

This patch concludes the fixes of all remaining memory leaks in perf
test "82: Use vfs_getname probe to get syscall args filenames".

Signed-off-by: Riccardo Mancini <rickyman7@gmail.com>
Fixes: cfc8874a48599249 ("perf script: Process cpu/threads maps")
Cc: Ian Rogers <irogers@google.com>
Cc: Jiri Olsa <jolsa@redhat.com>
Cc: Mark Rutland <mark.rutland@arm.com>
Cc: Namhyung Kim <namhyung@kernel.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Link: http://lore.kernel.org/lkml/5ee73b19791c6fa9d24c4d57f4ac1a23609400d7.1626343282.git.rickyman7@gmail.com
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
6 months agoperf script: Release zstd data
Riccardo Mancini [Thu, 15 Jul 2021 16:07:17 +0000 (18:07 +0200)]
perf script: Release zstd data

ASan reports several memory leak while running:

  # perf test "82: Use vfs_getname probe to get syscall args filenames"

One of the leaks is caused by zstd data not being released on exit in
perf-script.

This patch adds the missing zstd_fini().

Signed-off-by: Riccardo Mancini <rickyman7@gmail.com>
Fixes: b13b04d9382113f7 ("perf script: Initialize zstd_data")
Cc: Ian Rogers <irogers@google.com>
Cc: Jiri Olsa <jolsa@redhat.com>
Cc: Mark Rutland <mark.rutland@arm.com>
Cc: Milian Wolff <milian.wolff@kdab.com>
Cc: Namhyung Kim <namhyung@kernel.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Link: http://lore.kernel.org/lkml/39388e8cc2f85ca219ea18697a17b7bd8f74b693.1626343282.git.rickyman7@gmail.com
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
6 months agoperf session: Cleanup trace_event
Riccardo Mancini [Thu, 15 Jul 2021 16:07:16 +0000 (18:07 +0200)]
perf session: Cleanup trace_event

ASan reports several memory leaks when running:

  # perf test "82: Use vfs_getname probe to get syscall args filenames"

many of which are related to session->tevent.

This patch will solve this problem, then next patch will fix the
remaining memory leaks in 'perf script'.

This bug is due to a missing deallocation of the trace_event data
strutures.

This patch adds the missing trace_event__cleanup() in
perf_session__delete().

Signed-off-by: Riccardo Mancini <rickyman7@gmail.com>
Cc: Ian Rogers <irogers@google.com>
Cc: Jiri Olsa <jolsa@redhat.com>
Cc: Mark Rutland <mark.rutland@arm.com>
Cc: Namhyung Kim <namhyung@kernel.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Link: http://lore.kernel.org/lkml/fa2a3f221d90e47ce4e5b7e2d6e64c3509ddc96a.1626343282.git.rickyman7@gmail.com
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
6 months agoperf inject: Close inject.output on exit
Riccardo Mancini [Thu, 15 Jul 2021 16:07:15 +0000 (18:07 +0200)]
perf inject: Close inject.output on exit

ASan reports a memory leak when running:

  # perf test "83: Zstd perf.data compression/decompression"

which happens inside 'perf inject'.

The bug is caused by inject.output never being closed.

This patch adds the missing perf_data__close().

Signed-off-by: Riccardo Mancini <rickyman7@gmail.com>
Fixes: 6ef81c55a2b6584c ("perf session: Return error code for perf_session__new() function on failure")
Cc: Ian Rogers <irogers@google.com>
Cc: Jiri Olsa <jolsa@redhat.com>
Cc: Mamatha Inamdar <mamatha4@linux.vnet.ibm.com>
Cc: Mark Rutland <mark.rutland@arm.com>
Cc: Namhyung Kim <namhyung@kernel.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Link: http://lore.kernel.org/lkml/c06f682afa964687367cf6e92a64ceb49aec76a5.1626343282.git.rickyman7@gmail.com
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
6 months agoperf report: Free generated help strings for sort option
Riccardo Mancini [Thu, 15 Jul 2021 16:07:14 +0000 (18:07 +0200)]
perf report: Free generated help strings for sort option

ASan reports the memory leak of the strings allocated by sort_help() when
running perf report.

This patch changes the returned pointer to char* (instead of const
char*), saves it in a temporary variable, and finally deallocates it at
function exit.

Signed-off-by: Riccardo Mancini <rickyman7@gmail.com>
Fixes: 702fb9b415e7c99b ("perf report: Show all sort keys in help output")
Cc: Andi Kleen <ak@linux.intel.com>
Cc: Ian Rogers <irogers@google.com>
Cc: Jiri Olsa <jolsa@redhat.com>
Cc: Mark Rutland <mark.rutland@arm.com>
Cc: Namhyung Kim <namhyung@kernel.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Link: http://lore.kernel.org/lkml/a38b13f02812a8a6759200b9063c6191337f44d4.1626343282.git.rickyman7@gmail.com
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>