perf build: Install libtraceevent locally when building

The perf build currently has a '-Itools/lib' on the CC command
line. This causes issues as the libapi, libsubcmd, libtraceevent,
libbpf headers are all found via this path, making it impossible to
override include behavior.

Change the libtraceevent build mirroring the libbpf, libsubcmd, libapi
and libperf build, so that it is installed in a directory along with its
headers. A later change will modify the include behavior.

Similarly, the plugins are now installed into libtraceevent_plugins
except they have no header files.

Signed-off-by: Ian Rogers <[email protected]>
Acked-by: Namhyung Kim <[email protected]>
Cc: Alexander Shishkin <[email protected]>
Cc: Andrii Nakryiko <[email protected]>
Cc: Jiri Olsa <[email protected]>
Cc: Mark Rutland <[email protected]>
Cc: Masahiro Yamada <[email protected]>
Cc: Nick Desaulniers <[email protected]>
Cc: Nicolas Schier <[email protected]>
Cc: Peter Zijlstra <[email protected]>
Cc: Stephane Eranian <[email protected]>
Cc: [email protected]
Link: http://lore.kernel.org/lkml/[email protected]
Signed-off-by: Arnaldo Carvalho de Melo <[email protected]>

perf build: Install libperf locally when building

The perf build currently has a '-Itools/lib' on the CC command
line. This causes issues as the libapi, libsubcmd, libtraceevent,
libbpf headers are all found via this path, making it impossible to
override include behavior.

Change the libperf build mirroring the libbpf, libsubcmd and libapi
build, so that it is installed in a directory along with its headers. A
later change will modify the include behavior.

Signed-off-by: Ian Rogers <[email protected]>
Acked-by: Namhyung Kim <[email protected]>
Cc: Alexander Shishkin <[email protected]>
Cc: Andrii Nakryiko <[email protected]>
Cc: Jiri Olsa <[email protected]>
Cc: Mark Rutland <[email protected]>
Cc: Masahiro Yamada <[email protected]>
Cc: Nick Desaulniers <[email protected]>
Cc: Nicolas Schier <[email protected]>
Cc: Peter Zijlstra <[email protected]>
Cc: Stephane Eranian <[email protected]>
Cc: [email protected]
Link: http://lore.kernel.org/lkml/[email protected]
Signed-off-by: Arnaldo Carvalho de Melo <[email protected]>

perf build: Install libapi locally when building

The perf build currently has a '-Itools/lib' on the CC command line.
This causes issues as the libapi, libsubcmd, libtraceevent, libbpf
headers are all found via this path, making it impossible to override
include behavior.

Change the libapi build mirroring the libbpf and libsubcmd build, so
that it is installed in a directory along with its headers. A later
change will modify the include behavior.

Signed-off-by: Ian Rogers <[email protected]>
Acked-by: Namhyung Kim <[email protected]>
Cc: Alexander Shishkin <[email protected]>
Cc: Andrii Nakryiko <[email protected]>
Cc: Jiri Olsa <[email protected]>
Cc: Mark Rutland <[email protected]>
Cc: Masahiro Yamada <[email protected]>
Cc: Nick Desaulniers <[email protected]>
Cc: Nicolas Schier <[email protected]>
Cc: Peter Zijlstra <[email protected]>
Cc: Stephane Eranian <[email protected]>
Cc: [email protected]
Link: http://lore.kernel.org/lkml/[email protected]
Signed-off-by: Arnaldo Carvalho de Melo <[email protected]>

perf build: Install libsubcmd locally when building

The perf build currently has a '-Itools/lib' on the CC command
line. This causes issues as the libapi, libsubcmd, libtraceevent,
libbpf headers are all found via this path, making it impossible to
override include behavior. Change the libsubcmd build mirroring the
libbpf build, so that it is installed in a directory along with its
headers. A later change will modify the include behavior.

Signed-off-by: Ian Rogers <[email protected]>
Acked-by: Namhyung Kim <[email protected]>
Cc: Alexander Shishkin <[email protected]>
Cc: Andrii Nakryiko <[email protected]>
Cc: Jiri Olsa <[email protected]>
Cc: Mark Rutland <[email protected]>
Cc: Masahiro Yamada <[email protected]>
Cc: Nick Desaulniers <[email protected]>
Cc: Nicolas Schier <[email protected]>
Cc: Peter Zijlstra <[email protected]>
Cc: Stephane Eranian <[email protected]>
Cc: [email protected]
Link: http://lore.kernel.org/lkml/[email protected]
Signed-off-by: Arnaldo Carvalho de Melo <[email protected]>

tools lib subcmd: Add install target

This allows libsubcmd to be installed as a dependency.

Signed-off-by: Ian Rogers <[email protected]>
Acked-by: Namhyung Kim <[email protected]>
Cc: Alexander Shishkin <[email protected]>
Cc: Andrii Nakryiko <[email protected]>
Cc: Jiri Olsa <[email protected]>
Cc: Mark Rutland <[email protected]>
Cc: Masahiro Yamada <[email protected]>
Cc: Nick Desaulniers <[email protected]>
Cc: Nicolas Schier <[email protected]>
Cc: Peter Zijlstra <[email protected]>
Cc: Stephane Eranian <[email protected]>
Cc: [email protected]
Link: http://lore.kernel.org/lkml/[email protected]
Signed-off-by: Arnaldo Carvalho de Melo <[email protected]>

tools lib api: Add install target

This allows libapi to be installed as a dependency.

Signed-off-by: Ian Rogers <[email protected]>
Acked-by: Namhyung Kim <[email protected]>
Cmc: Alexander Shishkin <[email protected]>
Cc: Andrii Nakryiko <[email protected]>
Cc: Jiri Olsa <[email protected]>
Cc: Mark Rutland <[email protected]>
Cc: Masahiro Yamada <[email protected]>
Cc: Nick Desaulniers <[email protected]>
Cc: Peter Zijlstra <[email protected]>
Cc: Stephane Eranian <[email protected]>
Cc: [email protected]
Cc: nicolas schier <[email protected]>
Link: http://lore.kernel.org/lkml/[email protected]
Signed-off-by: Arnaldo Carvalho de Melo <[email protected]>

spi: dw-dma: decrease reference count in dw_spi_dma_init_mfld()

pci_get_device() will increase the reference count for the returned
pci_dev. Since 'dma_dev' is only used to filter the channel in
dw_spi_dma_chan_filer() after using it we need to call pci_dev_put() to
decrease the reference count. Also add pci_dev_put() for the error case.

Fixes: 7063c0d942a1 ("spi/dw_spi: add DMA support")
Signed-off-by: Xiongfeng Wang <[email protected]>
Acked-by: Serge Semin <[email protected]>
Link: https://lore.kernel.org/r/[email protected]
Signed-off-by: Mark Brown <[email protected]>

tcp: configurable source port perturb table size

On embedded systems with little memory and no relevant
security concerns, it is beneficial to reduce the size
of the table.

Reducing the size from 2^16 to 2^8 saves 255 KiB
of kernel RAM.

Makes the table size configurable as an expert option.

The size was previously increased from 2^8 to 2^16
in commit 4c2c8f03a5ab ("tcp: increase source port perturb table to
2^16").

Signed-off-by: Gleb Mazovetskiy <[email protected]>
Reviewed-by: Kuniyuki Iwashima <[email protected]>
Signed-off-by: David S. Miller <[email protected]>

mmc: sdhci-pci-o2micro: fix card detect fail issue caused by CD# debounce timeout

The SD card is recognized failed sometimes when resume from suspend.
Because CD# debounce time too long then card present report wrong.
Finally, card is recognized failed.

Signed-off-by: Chevron Li <[email protected]>
Cc: [email protected]
Link: https://lore.kernel.org/r/[email protected]
Signed-off-by: Ulf Hansson <[email protected]>

mmc: core: properly select voltage range without power cycle

In mmc_select_voltage(), if there is no full power cycle, the voltage
range selected at the end of the function will be on a single range
(e.g. 3.3V/3.4V). To keep a range around the selected voltage (3.2V/3.4V),
the mask shift should be reduced by 1.

This issue was triggered by using a specific SD-card (Verbatim Premium
16GB UHS-1) on an STM32MP157C-DK2 board. This board cannot do UHS modes
and there is no power cycle. And the card was failing to switch to
high-speed mode. When adding the range 3.2V/3.3V for this card with the
proposed shift change, the card can switch to high-speed mode.

Fixes: ce69d37b7d8f ("mmc: core: Prevent violation of specs while initializing cards")
Signed-off-by: Yann Gautier <[email protected]>
Cc: [email protected]
Link: https://lore.kernel.org/r/[email protected]
Signed-off-by: Ulf Hansson <[email protected]>

l2tp: Serialize access to sk_user_data with sk_callback_lock

sk->sk_user_data has multiple users, which are not compatible with each
other. Writers must synchronize by grabbing the sk->sk_callback_lock.

l2tp currently fails to grab the lock when modifying the underlying tunnel
socket fields. Fix it by adding appropriate locking.

We err on the side of safety and grab the sk_callback_lock also inside the
sk_destruct callback overridden by l2tp, even though there should be no
refs allowing access to the sock at the time when sk_destruct gets called.

v4:
- serialize write to sk_user_data in l2tp sk_destruct

v3:
- switch from sock lock to sk_callback_lock
- document write-protection for sk_user_data

v2:
- update Fixes to point to origin of the bug
- use real names in Reported/Tested-by tags

Cc: Tom Parkin <[email protected]>
Fixes: 3557baabf280 ("[L2TP]: PPP over L2TP driver core")
Reported-by: Haowei Yan <[email protected]>
Signed-off-by: Jakub Sitnicki <[email protected]>
Signed-off-by: David S. Miller <[email protected]>

perf stat: Add print_aggr_cgroup() for --for-each-cgroup and --topdown

Normally, --for-each-cgroup only works with AGGR_GLOBAL.  However
the --topdown on some cpu (e.g. Intel Skylake) converts it to the
AGGR_CORE internally.

To support those machines, add print_aggr_cgroup and handle the events
like in print_cgroup_events().

  $ perf stat -a --for-each-cgroup system.slice,user.slice --topdown sleep 1
  nmi_watchdog enabled with topdown. May give wrong results.
  Disable with echo 0 > /proc/sys/kernel/nmi_watchdog

   Performance counter stats for 'system wide':

                               retiring   bad speculation  frontend bound  backend bound
  S0-D0-C0  2  system.slice       49.0%            -46.6%           31.4%
  S0-D0-C1  2  system.slice       55.5%              8.0%           45.5%          -9.0%
  S0-D0-C2  2  system.slice       87.8%             22.1%           30.3%         -40.3%
  S0-D0-C3  2  system.slice       53.3%            -11.9%           45.2%          13.4%
  S0-D0-C0  2  user.slice        123.5%              4.0%           48.5%         -75.9%
  S0-D0-C1  2  user.slice         19.9%              6.5%           89.9%         -16.3%
  S0-D0-C2  2  user.slice         29.9%              7.9%           71.3%          -9.1
  S0-D0-C3  2  user.slice         28.0%              7.2%           43.3%          21.5%

         1.004136937 seconds time elapsed

Signed-off-by: Namhyung Kim <[email protected]>
Cc: Adrian Hunter <[email protected]>
Cc: Athira Jajeev <[email protected]>
Cc: Ian Rogers <[email protected]>
Cc: Ingo Molnar <[email protected]>
Cc: James Clark <[email protected]>
Cc: Jiri Olsa <[email protected]>
Cc: Kan Liang <[email protected]>
Cc: Peter Zijlstra <[email protected]>
Cc: Xing Zhengjun <[email protected]>
Link: https://lore.kernel.org/r/[email protected]
Signed-off-by: Arnaldo Carvalho de Melo <[email protected]>

perf stat: Support --for-each-cgroup and --metric-only

When we have events for each cgroup, the metric should be printed for
each cgroup separately.  Add print_cgroup_counter() to handle that
situation properly.

Also change print_metric_headers() not to print duplicate headers
by checking cgroups.

  $ perf stat -a --for-each-cgroup system.slice,user.slice --metric-only sleep 1

   Performance counter stats for 'system wide':

                                     GHz       insn per cycle branch-misses of all branches
   system.slice                   3.792                0.61                                3.24%
   user.slice                     3.661                2.32                                0.37%

         1.016111516 seconds time elapsed

Signed-off-by: Namhyung Kim <[email protected]>
Cc: Adrian Hunter <[email protected]>
Cc: Athira Jajeev <[email protected]>
Cc: Ian Rogers <[email protected]>
Cc: Ingo Molnar <[email protected]>
Cc: James Clark <[email protected]>
Cc: Jiri Olsa <[email protected]>
Cc: Kan Liang <[email protected]>
Cc: Peter Zijlstra <[email protected]>
Cc: Xing Zhengjun <[email protected]>
Link: https://lore.kernel.org/r/[email protected]
Signed-off-by: Arnaldo Carvalho de Melo <[email protected]>

perf stat: Factor out print_metric_{begin,end}()

For the metric-only case, add new functions to handle the start and the
end of each metric display.

Signed-off-by: Namhyung Kim <[email protected]>
Cc: Adrian Hunter <[email protected]>
Cc: Athira Jajeev <[email protected]>
Cc: Ian Rogers <[email protected]>
Cc: Ingo Molnar <[email protected]>
Cc: James Clark <[email protected]>
Cc: Jiri Olsa <[email protected]>
Cc: Kan Liang <[email protected]>
Cc: Peter Zijlstra <[email protected]>
Cc: Xing Zhengjun <[email protected]>
Link: https://lore.kernel.org/r/[email protected]
Signed-off-by: Arnaldo Carvalho de Melo <[email protected]>

perf stat: Factor out prefix display

The prefix is needed for interval mode to print timestamp at the
beginning of each line.  But the it's tricky for the metric only
mode since it doesn't print every evsel and combines the metrics
into a single line.

So it needed to pass 'first' argument to print_counter_aggrdata()
to determine if the current event is being printed at first.  This
makes the code hard to read.

Let's move the logic out of the function and do it in the outer
print loop.  This would enable further cleanups later.

Signed-off-by: Namhyung Kim <[email protected]>
Cc: Adrian Hunter <[email protected]>
Cc: Athira Jajeev <[email protected]>
Cc: Ian Rogers <[email protected]>
Cc: Ingo Molnar <[email protected]>
Cc: James Clark <[email protected]>
Cc: Jiri Olsa <[email protected]>
Cc: Kan Liang <[email protected]>
Cc: Peter Zijlstra <[email protected]>
Cc: Xing Zhengjun <[email protected]>
Link: https://lore.kernel.org/r/[email protected]
Signed-off-by: Arnaldo Carvalho de Melo <[email protected]>

perf stat: Move condition to print_footer()

Likewise, I think it'd better to have the control inside the function, and keep
the higher level function clearer.

Signed-off-by: Namhyung Kim <[email protected]>
Cc: Adrian Hunter <[email protected]>
Cc: Athira Jajeev <[email protected]>
Cc: Ian Rogers <[email protected]>
Cc: Ingo Molnar <[email protected]>
Cc: James Clark <[email protected]>
Cc: Jiri Olsa <[email protected]>
Cc: Kan Liang <[email protected]>
Cc: Peter Zijlstra <[email protected]>
Cc: Xing Zhengjun <[email protected]>
Link: https://lore.kernel.org/r/[email protected]
Signed-off-by: Arnaldo Carvalho de Melo <[email protected]>

perf stat: Rework header display

There are print_header() and print_interval() to print header lines before
actual counter values.  Also print_metric_headers() needs to be called for
the metric-only case.

Let's move all these logics to a single place including num_print_iv to
refresh the headers for interval mode.

Signed-off-by: Namhyung Kim <[email protected]>
Cc: Adrian Hunter <[email protected]>
Cc: Athira Jajeev <[email protected]>
Cc: Ian Rogers <[email protected]>
Cc: Ingo Molnar <[email protected]>
Cc: James Clark <[email protected]>
Cc: Jiri Olsa <[email protected]>
Cc: Kan Liang <[email protected]>
Cc: Peter Zijlstra <[email protected]>
Cc: Xing Zhengjun <[email protected]>
Link: https://lore.kernel.org/r/[email protected]
Signed-off-by: Arnaldo Carvalho de Melo <[email protected]>

perf stat: Remove impossible condition

The print would run only if metric_only is not set, but it's already in a
block that says it's in metric_only case.  And there's no place to change
the setting.

Signed-off-by: Namhyung Kim <[email protected]>
Cc: Adrian Hunter <[email protected]>
Cc: Athira Jajeev <[email protected]>
Cc: Ian Rogers <[email protected]>
Cc: Ingo Molnar <[email protected]>
Cc: James Clark <[email protected]>
Cc: Jiri Olsa <[email protected]>
Cc: Kan Liang <[email protected]>
Cc: Peter Zijlstra <[email protected]>
Cc: Xing Zhengjun <[email protected]>
Link: https://lore.kernel.org/r/[email protected]
Signed-off-by: Arnaldo Carvalho de Melo <[email protected]>

perf stat: Cleanup interval print alignment

Instead of using magic values, define symbolic constants and use them.
Also add aggr_header_std[] array to simplify aggr_mode handling.

Signed-off-by: Namhyung Kim <[email protected]>
Cc: Adrian Hunter <[email protected]>
Cc: Athira Jajeev <[email protected]>
Cc: Ian Rogers <[email protected]>
Cc: Ingo Molnar <[email protected]>
Cc: James Clark <[email protected]>
Cc: Jiri Olsa <[email protected]>
Cc: Kan Liang <[email protected]>
Cc: Peter Zijlstra <[email protected]>
Cc: Xing Zhengjun <[email protected]>
Link: https://lore.kernel.org/r/[email protected]
Signed-off-by: Arnaldo Carvalho de Melo <[email protected]>

perf stat: Factor out prepare_interval()

This logic does not print the time directly, but it just puts the
timestamp in the buffer as a prefix.  To reduce the confusion, factor
out the code into a separate function.

Signed-off-by: Namhyung Kim <[email protected]>
Cc: Adrian Hunter <[email protected]>
Cc: Athira Jajeev <[email protected]>
Cc: Ian Rogers <[email protected]>
Cc: Ingo Molnar <[email protected]>
Cc: James Clark <[email protected]>
Cc: Jiri Olsa <[email protected]>
Cc: Kan Liang <[email protected]>
Cc: Peter Zijlstra <[email protected]>
Cc: Xing Zhengjun <[email protected]>
Link: https://lore.kernel.org/r/[email protected]
Signed-off-by: Arnaldo Carvalho de Melo <[email protected]>

perf stat: Split print_metric_headers() function

The print_metric_headers() shows metric headers a little bit for each
mode.  Split it out to make the code clearer.

Signed-off-by: Namhyung Kim <[email protected]>
Cc: Adrian Hunter <[email protected]>
Cc: Athira Jajeev <[email protected]>
Cc: Ian Rogers <[email protected]>
Cc: Ingo Molnar <[email protected]>
Cc: James Clark <[email protected]>
Cc: Jiri Olsa <[email protected]>
Cc: Kan Liang <[email protected]>
Cc: Peter Zijlstra <[email protected]>
Cc: Xing Zhengjun <[email protected]>
Link: https://lore.kernel.org/r/[email protected]
Signed-off-by: Arnaldo Carvalho de Melo <[email protected]>

perf stat: Align cgroup names

We don't know how long cgroup name is, but at least we can align short
ones like below.

  $ perf stat -a --for-each-cgroup system.slice,user.slice true

   Performance counter stats for 'system wide':

           0.13 msec cpu-clock         system.slice  #    0.010 CPUs utilized
              4      context-switches  system.slice  #   31.989 K/sec
              1      cpu-migrations    system.slice  #    7.997 K/sec
              0      page-faults       system.slice  #    0.000 /sec
        450,673      cycles            system.slice  #    3.604 GHz             (92.41%)
        161,216      instructions      system.slice  #    0.36  insn per cycle  (92.41%)
         32,678      branches          system.slice  #  261.332 M/sec           (92.41%)
          2,628      branch-misses     system.slice  #    8.04% of all branches (92.41%)
          14.29 msec cpu-clock         user.slice    #    1.163 CPUs utilized
             35      context-switches  user.slice    #    2.449 K/sec
             12      cpu-migrations    user.slice    #  839.691 /sec
             57      page-faults       user.slice    #    3.989 K/sec
     49,683,026      cycles            user.slice    #    3.477 GHz             (99.38%)
    110,790,266      instructions      user.slice    #    2.23  insn per cycle  (99.38%)
     24,552,255      branches          user.slice    #    1.718 G/sec           (99.38%)
        127,779      branch-misses     user.slice    #    0.52% of all branches (99.38%)

    0.012289431 seconds time elapsed

Signed-off-by: Namhyung Kim <[email protected]>
Cc: Adrian Hunter <[email protected]>
Cc: Athira Jajeev <[email protected]>
Cc: Ian Rogers <[email protected]>
Cc: Ingo Molnar <[email protected]>
Cc: James Clark <[email protected]>
Cc: Jiri Olsa <[email protected]>
Cc: Kan Liang <[email protected]>
Cc: Peter Zijlstra <[email protected]>
Cc: Xing Zhengjun <[email protected]>
Link: https://lore.kernel.org/r/[email protected]
Signed-off-by: Arnaldo Carvalho de Melo <[email protected]>

perf stat: Add before_metric argument

Unfortunately, event running time, percentage and noise data are printed
in different positions in normal output than CSV/JSON.  I think it's
better to put such details in where it actually prints.

So add before_metric argument to print_noise() and print_running() and
call them twice before and after the metric.

Signed-off-by: Namhyung Kim <[email protected]>
Cc: Adrian Hunter <[email protected]>
Cc: Athira Jajeev <[email protected]>
Cc: Ian Rogers <[email protected]>
Cc: Ingo Molnar <[email protected]>
Cc: James Clark <[email protected]>
Cc: Jiri Olsa <[email protected]>
Cc: Kan Liang <[email protected]>
Cc: Peter Zijlstra <[email protected]>
Cc: Xing Zhengjun <[email protected]>
Link: https://lore.kernel.org/r/[email protected]
Signed-off-by: Arnaldo Carvalho de Melo <[email protected]>

perf stat: Handle bad events in abs_printout()

In the printout() function, it checks if the event is bad (i.e. not
counted or not supported) and print the result.  But it does the same
what abs_printout() is doing.  So add an argument to indicate the value
is ok or not and use the same function in both cases.

Signed-off-by: Namhyung Kim <[email protected]>
Cc: Adrian Hunter <[email protected]>
Cc: Athira Jajeev <[email protected]>
Cc: Ian Rogers <[email protected]>
Cc: Ingo Molnar <[email protected]>
Cc: James Clark <[email protected]>
Cc: Jiri Olsa <[email protected]>
Cc: Kan Liang <[email protected]>
Cc: Peter Zijlstra <[email protected]>
Cc: Xing Zhengjun <[email protected]>
Link: https://lore.kernel.org/r/[email protected]
Signed-off-by: Arnaldo Carvalho de Melo <[email protected]>

perf stat: Factor out print_counter_value() function

And split it for each output mode like others.  I believe it makes the
code simpler and more intuitive.  Now abs_printout() becomes just to
call sub-functions.

Signed-off-by: Namhyung Kim <[email protected]>
Cc: Adrian Hunter <[email protected]>
Cc: Athira Jajeev <[email protected]>
Cc: Ian Rogers <[email protected]>
Cc: Ingo Molnar <[email protected]>
Cc: James Clark <[email protected]>
Cc: Jiri Olsa <[email protected]>
Cc: Kan Liang <[email protected]>
Cc: Peter Zijlstra <[email protected]>
Cc: Xing Zhengjun <[email protected]>
Link: https://lore.kernel.org/r/[email protected]
Signed-off-by: Arnaldo Carvalho de Melo <[email protected]>

perf stat: Split aggr_printout() function

The aggr_printout() function is to print aggr_id and count (nr).
Split it for each output mode to simplify the code.

Signed-off-by: Namhyung Kim <[email protected]>
Cc: Adrian Hunter <[email protected]>
Cc: Athira Jajeev <[email protected]>
Cc: Ian Rogers <[email protected]>
Cc: Ingo Molnar <[email protected]>
Cc: James Clark <[email protected]>
Cc: Jiri Olsa <[email protected]>
Cc: Kan Liang <[email protected]>
Cc: Peter Zijlstra <[email protected]>
Cc: Xing Zhengjun <[email protected]>
Link: https://lore.kernel.org/r/[email protected]
Signed-off-by: Arnaldo Carvalho de Melo <[email protected]>

perf stat: Split print_cgroup() function

Likewise, split print_cgroup() for each output mode.

Signed-off-by: Namhyung Kim <[email protected]>
Cc: Adrian Hunter <[email protected]>
Cc: Athira Jajeev <[email protected]>
Cc: Ian Rogers <[email protected]>
Cc: Ingo Molnar <[email protected]>
Cc: James Clark <[email protected]>
Cc: Jiri Olsa <[email protected]>
Cc: Kan Liang <[email protected]>
Cc: Peter Zijlstra <[email protected]>
Cc: Xing Zhengjun <[email protected]>
Link: https://lore.kernel.org/r/[email protected]
Signed-off-by: Arnaldo Carvalho de Melo <[email protected]>

perf stat: Split print_noise_pct() function

Likewise, split print_noise_pct() for each output mode.  Although it's
a tiny function, more logic will be added soon so it'd be better split
it and treat it in the same way.

Signed-off-by: Namhyung Kim <[email protected]>
Cc: Adrian Hunter <[email protected]>
Cc: Athira Jajeev <[email protected]>
Cc: Ian Rogers <[email protected]>
Cc: Ingo Molnar <[email protected]>
Cc: James Clark <[email protected]>
Cc: Jiri Olsa <[email protected]>
Cc: Kan Liang <[email protected]>
Cc: Peter Zijlstra <[email protected]>
Cc: Xing Zhengjun <[email protected]>
Link: https://lore.kernel.org/r/[email protected]
Signed-off-by: Arnaldo Carvalho de Melo <[email protected]>

perf stat: Split print_running() function

To make the code more obvious and hopefully simpler, factor out the
code for each output mode - stdio, CSV, JSON.

Signed-off-by: Namhyung Kim <[email protected]>
Cc: Adrian Hunter <[email protected]>
Cc: Athira Jajeev <[email protected]>
Cc: Ian Rogers <[email protected]>
Cc: Ingo Molnar <[email protected]>
Cc: James Clark <[email protected]>
Cc: Jiri Olsa <[email protected]>
Cc: Kan Liang <[email protected]>
Cc: Peter Zijlstra <[email protected]>
Cc: Xing Zhengjun <[email protected]>
Link: https://lore.kernel.org/r/[email protected]
Signed-off-by: Arnaldo Carvalho de Melo <[email protected]>

perf stat: Clear screen only if output file is a tty

The --interval-clear option makes perf stat to clear the terminal at
each interval.  But it doesn't need to clear the screen when it saves
to a file.  Make it fail when it's enabled with the output options.

  $ perf stat -I 1 --interval-clear -o myfile true
  --interval-clear does not work with output

   Usage: perf stat [<options>] [<command>]

      -o, --output <file>   output file name
          --log-fd <n>      log output to fd, instead of stderr
          --interval-clear  clear screen in between new interval

Signed-off-by: Namhyung Kim <[email protected]>
Tested-by: Arnaldo Carvalho de Melo <[email protected]>
Cc: Adrian Hunter <[email protected]>
Cc: Athira Jajeev <[email protected]>
Cc: Ian Rogers <[email protected]>
Cc: Ingo Molnar <[email protected]>
Cc: James Clark <[email protected]>
Cc: Jiri Olsa <[email protected]>
Cc: Kan Liang <[email protected]>
Cc: Peter Zijlstra <[email protected]>
Cc: Xing Zhengjun <[email protected]>
Link: https://lore.kernel.org/r/[email protected]
Signed-off-by: Arnaldo Carvalho de Melo <[email protected]>

spi: spi-imx: Fix spi_bus_clk if requested clock is higher than input clock

In case the requested bus clock is higher than the input clock, the correct
dividers (pre = 0, post = 0) are returned from mx51_ecspi_clkdiv(), but
*fres is left uninitialized and therefore contains an arbitrary value.

This causes trouble for the recently introduced PIO polling feature as the
value in spi_imx->spi_bus_clk is used there to calculate for which
transfers to enable PIO polling.

Fix this by setting *fres even if no clock dividers are in use.

This issue was observed on Kontron BL i.MX8MM with an SPI peripheral clock set
to 50 MHz by default and a requested SPI bus clock of 80 MHz for the SPI NOR
flash.

With the fix applied the debug message from mx51_ecspi_clkdiv() now prints the
following:

spi_imx 30820000.spi: mx51_ecspi_clkdiv: fin: 50000000, fspi: 50000000,
post: 0, pre: 0

Fixes: 6fd8b8503a0d ("spi: spi-imx: Fix out-of-order CS/SCLK operation at low speeds")
Fixes: 07e759387788 ("spi: spi-imx: add PIO polling support")
Cc: Marc Kleine-Budde <[email protected]>
Cc: David Jander <[email protected]>
Cc: Fabio Estevam <[email protected]>
Cc: Mark Brown <[email protected]>
Cc: Marek Vasut <[email protected]>
Cc: [email protected]
Signed-off-by: Frieder Schrempf <[email protected]>
Tested-by: Fabio Estevam <[email protected]>
Acked-by: Marek Vasut <[email protected]>
Link: https://lore.kernel.org/r/[email protected]
Signed-off-by: Mark Brown <[email protected]>

s390: avoid using global register for current_stack_pointer

Commit 30de14b1884b ("s390: current_stack_pointer shouldn't be a
function") made current_stack_pointer a global register variable like
on many other architectures. Unfortunately on s390 it uncovers old
gcc bug which is fixed only since gcc-9.1 [gcc commit 3ad7fed1cc87
("S/390: Fix PR89775. Stackpointer save/restore instructions removed")]
and backported to gcc-8.4 and later. Due to this bug gcc versions prior
to 8.4 generate broken code which leads to stack corruptions.

Current minimal gcc version required to build the kernel is declared
as 5.1. It is not possible to fix all old gcc versions, so work
around this problem by avoiding using global register variable for
current_stack_pointer.

Fixes: 30de14b1884b ("s390: current_stack_pointer shouldn't be a function")
Reviewed-by: Heiko Carstens <[email protected]>
Signed-off-by: Vasily Gorbik <[email protected]>
Signed-off-by: Alexander Gordeev <[email protected]>

s390/dcssblk: fix deadlock when adding a DCSS

After the rework from commit 1ebe2e5f9d68 ("block: remove
GENHD_FL_EXT_DEVT"), when calling device_add_disk(), dcssblk will end up
in disk_scan_partitions(), and not break out early w/o GENHD_FL_NO_PART.
This will trigger implicit open/release via blkdev_get/put_whole()
later. dcssblk_release() will then deadlock on dcssblk_devices_sem
semaphore, which is already held from dcssblk_add_store() when calling
device_add_disk().

dcssblk does not support partitions (DCSSBLK_MINORS_PER_DISK == 1), and
never scanned partitions before. Therefore restore the previous
behavior, and explicitly disallow partition scanning by setting the
GENHD_FL_NO_PART flag. This will also prevent this deadlock scenario.

Fixes: 1ebe2e5f9d68 ("block: remove GENHD_FL_EXT_DEVT")
Cc: <[email protected]> # 5.17+
Signed-off-by: Gerald Schaefer <[email protected]>
Acked-by: Heiko Carstens <[email protected]>
Signed-off-by: Alexander Gordeev <[email protected]>

powerpc: Fix writable sections being moved into the rodata region

.data.rel.ro*  catches .data.rel.root_cpuacct, and the kernel crashes on
a store in css_clear_dir. At least we know read-only data protection is
working...

Fixes: b6adc6d6d3272 ("powerpc/build: move .data.rel.ro, .sdata2 to read-only")
Signed-off-by: Nicholas Piggin <[email protected]>
Reviewed-by: Christophe Leroy <[email protected]>
Signed-off-by: Michael Ellerman <[email protected]>
Link: https://lore.kernel.org/r/[email protected]

net: thunderbolt: Fix error handling in tbnet_init()

A problem about insmod thunderbolt-net failed is triggered with following
log given while lsmod does not show thunderbolt_net:

 insmod: ERROR: could not insert module thunderbolt-net.ko: File exists

The reason is that tbnet_init() returns tb_register_service_driver()
directly without checking its return value, if tb_register_service_driver()
failed, it returns without removing property directory, resulting the
property directory can never be created later.

 tbnet_init()
   tb_register_property_dir() # register property directory
   tb_register_service_driver()
     driver_register()
       bus_add_driver()
         priv = kzalloc(...) # OOM happened
   # return without remove property directory

Fix by remove property directory when tb_register_service_driver() returns
error.

Fixes: e69b6c02b4c3 ("net: Add support for networking over Thunderbolt cable")
Signed-off-by: Yuan Can <[email protected]>
Acked-by: Mika Westerberg <[email protected]>
Signed-off-by: David S. Miller <[email protected]>

locking: Fix qspinlock/x86 inline asm error

When compiling linux 6.1.0-rc3 configured with CONFIG_64BIT=y and
CONFIG_PARAVIRT_SPINLOCKS=y on x86_64 using LLVM 11.0, an error:
"<inline asm> error: changed section flags for .spinlock.text,
expected:: 0x6" occurred.

The reason is the .spinlock.text in kernel/locking/qspinlock.o
is used many times, but its flags are omitted in subsequent use.

LLVM 11.0 assembler didn't permit to
leave out flags in subsequent uses of the same sections.

So this patch adds the corresponding flags to avoid above error.

Fixes: 501f7f69bca1 ("locking: Add __lockfunc to slow path functions")
Signed-off-by: Guo Jin <[email protected]>
Signed-off-by: Peter Zijlstra (Intel) <[email protected]>
Reviewed-by: Nathan Chancellor <[email protected]>
Link: https://lore.kernel.org/r/[email protected]

perf/x86/intel/pt: Fix sampling using single range output

Deal with errata TGL052, ADL037 and RPL017 "Trace May Contain Incorrect
Data When Configured With Single Range Output Larger Than 4KB" by
disabling single range output whenever larger than 4KB.

Fixes: 670638477aed ("perf/x86/intel/pt: Opportunistically use single range output mode")
Signed-off-by: Adrian Hunter <[email protected]>
Signed-off-by: Peter Zijlstra (Intel) <[email protected]>
Cc: [email protected]
Link: https://lkml.kernel.org/r/[email protected]

perf/x86/amd: Fix crash due to race between amd_pmu_enable_all, perf NMI and throttling

amd_pmu_enable_all() does:

      if (!test_bit(idx, cpuc->active_mask))
              continue;

      amd_pmu_enable_event(cpuc->events[idx]);

A perf NMI of another event can come between these two steps. Perf NMI
handler internally disables and enables _all_ events, including the one
which nmi-intercepted amd_pmu_enable_all() was in process of enabling.
If that unintentionally enabled event has very low sampling period and
causes immediate successive NMI, causing the event to be throttled,
cpuc->events[idx] and cpuc->active_mask gets cleared by x86_pmu_stop().
This will result in amd_pmu_enable_event() getting called with event=NULL
when amd_pmu_enable_all() resumes after handling the NMIs. This causes a
kernel crash:

  BUG: kernel NULL pointer dereference, address: 0000000000000198
  #PF: supervisor read access in kernel mode
  #PF: error_code(0x0000) - not-present page
  [...]
  Call Trace:
   <TASK>
   amd_pmu_enable_all+0x68/0xb0
   ctx_resched+0xd9/0x150
   event_function+0xb8/0x130
   ? hrtimer_start_range_ns+0x141/0x4a0
   ? perf_duration_warn+0x30/0x30
   remote_function+0x4d/0x60
   __flush_smp_call_function_queue+0xc4/0x500
   flush_smp_call_function_queue+0x11d/0x1b0
   do_idle+0x18f/0x2d0
   cpu_startup_entry+0x19/0x20
   start_secondary+0x121/0x160
   secondary_startup_64_no_verify+0xe5/0xeb
   </TASK>

amd_pmu_disable_all()/amd_pmu_enable_all() calls inside perf NMI handler
were recently added as part of BRS enablement but I'm not sure whether
we really need them. We can just disable BRS in the beginning and enable
it back while returning from NMI. This will solve the issue by not
enabling those events whose active_masks are set but are not yet enabled
in hw pmu.

Fixes: ada543459cab ("perf/x86/amd: Add AMD Fam19h Branch Sampling support")
Reported-by: Linux Kernel Functional Testing <[email protected]>
Signed-off-by: Ravi Bangoria <[email protected]>
Signed-off-by: Peter Zijlstra (Intel) <[email protected]>
Link: https://lkml.kernel.org/r/[email protected]

Merge branch 'microchip-fixes'

Shang XiaoJing says:

====================
net: microchip: Fix potential null-ptr-deref due to create_singlethread_workqueue()

There are some functions call create_singlethread_workqueue() without
checking ret value, and the NULL workqueue_struct pointer may causes
null-ptr-deref. Will be fixed by this patch.
====================

Signed-off-by: David S. Miller <[email protected]>

net: microchip: sparx5: Fix potential null-ptr-deref in sparx_stats_init() and sparx5_start()

sparx_stats_init() calls create_singlethread_workqueue() and not
checked the ret value, which may return NULL. And a null-ptr-deref may
happen:

sparx_stats_init()
    create_singlethread_workqueue() # failed, sparx5->stats_queue is NULL
    queue_delayed_work()
        queue_delayed_work_on()
            __queue_delayed_work()  # warning here, but continue
                __queue_work()      # access wq->flags, null-ptr-deref

Check the ret value and return -ENOMEM if it is NULL. So as
sparx5_start().

Fixes: af4b11022e2d ("net: sparx5: add ethtool configuration and statistics support")
Fixes: b37a1bae742f ("net: sparx5: add mactable support")
Signed-off-by: Shang XiaoJing <[email protected]>
Signed-off-by: David S. Miller <[email protected]>

net: lan966x: Fix potential null-ptr-deref in lan966x_stats_init()

lan966x_stats_init() calls create_singlethread_workqueue() and not
checked the ret value, which may return NULL. And a null-ptr-deref may
happen:

lan966x_stats_init()
    create_singlethread_workqueue() # failed, lan966x->stats_queue is NULL
    queue_delayed_work()
        queue_delayed_work_on()
            __queue_delayed_work()  # warning here, but continue
                __queue_work()      # access wq->flags, null-ptr-deref

Check the ret value and return -ENOMEM if it is NULL.

Fixes: 12c2d0a5b8e2 ("net: lan966x: add ethtool configuration and statistics")
Signed-off-by: Shang XiaoJing <[email protected]>
Reviewed-by: Horatiu Vultur <[email protected]>
Signed-off-by: David S. Miller <[email protected]>

platform/x86: ideapad-laptop: Add module parameters to match DMI quirk tables

Add module parameters to allow setting the hw_rfkill_switch and
set_fn_lock_led feature flags for testing these on laptops which are not
on the DMI-id based allow lists for these 2 flags.

Signed-off-by: Hans de Goede <[email protected]>
Link: https://lore.kernel.org/r/[email protected]

platform/x86: ideapad-laptop: Fix interrupt storm on fn-lock toggle on some Yoga laptops

Commit 3ae86d2d4704 ("platform/x86: ideapad-laptop: Fix Legion 5 Fn lock
LED") uses the WMI event-id for the fn-lock event on some Legion 5 laptops
to manually toggle the fn-lock LED because the EC does not do it itself.
However, the same WMI ID is also sent on some Yoga laptops. Here, setting
the fn-lock state is not valid behavior, and causes the EC to spam
interrupts until the laptop is rebooted.

Add a set_fn_lock_led_list[] DMI-id list and only enable the workaround to
manually set the LED on models on this list.

Link: https://bugzilla.kernel.org/show_bug.cgi?id=212671
Cc: Meng Dong <[email protected]>
Signed-off-by: Arnav Rawat <[email protected]>
Link: https://lore.kernel.org/r/12093851.O9o76ZdvQC@fedora
[[email protected]: Check DMI-id list only once and store the result]
Signed-off-by: Hans de Goede <[email protected]>

platform/x86: hp-wmi: Ignore Smart Experience App event

Sometimes hp-wmi driver complains on system resume:
[ 483.116451] hp_wmi: Unknown event_id - 33 - 0x0

According to HP it's a feature called "HP Smart Experience App" and it's
safe to be ignored.

Signed-off-by: Kai-Heng Feng <[email protected]>
Link: https://lore.kernel.org/r/[email protected]
Reviewed-by: Hans de Goede <[email protected]>
Signed-off-by: Hans de Goede <[email protected]>

platform/surface: aggregator_registry: Add support for Surface Laptop 5

Add device nodes to enable support for battery and charger status, the
ACPI platform profile, as well as internal HID devices (including
touchpad and keyboard) on the Surface Laptop 5.

Signed-off-by: Maximilian Luz <[email protected]>
Link: https://lore.kernel.org/r/[email protected]
Signed-off-by: Hans de Goede <[email protected]>

zonefs: Remove to_attr() helper function

to_attr() in zonefs sysfs code is unused, which it causes a warning when
compiling with clang and W=1. Delete it to prevent the warning.

Signed-off-by: Damien Le Moal <[email protected]>
Reviewed-by: Johannes Thumshirn <[email protected]>

zonefs: fix zone report size in __zonefs_io_error()

When an IO error occurs, the function __zonefs_io_error() is used to
issue a zone report to obtain the latest zone information from the
device. This function gets a zone report for all zones used as storage
for a file, which is always 1 zone except for files representing
aggregated conventional zones.

The number of zones of a zone report for a file is calculated in
__zonefs_io_error() by doing a bit-shift of the inode i_zone_size field,
which is equal to or larger than the device zone size. However, this
calculation does not take into account that the last zone of a zoned
device may be smaller than the zone size reported by bdev_zone_sectors()
(which is used to set the bit shift size). As a result, if an error
occurs for an IO targetting such last smaller zone, the zone report will
ask for 0 zones, leading to an invalid zone report.

Fix this by using the fact that all files require a 1 zone report,
except if the inode i_zone_size field indicates a zone size larger than
the device zone size. This exception case corresponds to a mount with
aggregated conventional zones.

A check for this exception is added to the file inode initialization
during mount. If an invalid setup is detected, emit an error and fail
the mount (check contributed by Johannes Thumshirn).

Signed-off-by: Johannes Thumshirn <[email protected]>
Signed-off-by: Damien Le Moal <[email protected]>

cifs: Fix wrong return value checking when GETFLAGS

The return value of CIFSGetExtAttr is negative, should be checked
with -EOPNOTSUPP rather than EOPNOTSUPP.

Fixes: 64a5cfa6db94 ("Allow setting per-file compression via SMB2/3")
Signed-off-by: Zhang Xiaoxu <[email protected]>
Signed-off-by: Steve French <[email protected]>

nvmet: fix a memory leak in nvmet_auth_set_key

When changing dhchap secrets we need to release the old
secrets as well.

kmemleak complaint:
--
unreferenced object 0xffff8c7f44ed8180 (size 64):
  comm "check", pid 7304, jiffies 4295686133 (age 72034.246s)
  hex dump (first 32 bytes):
    44 48 48 43 2d 31 3a 30 30 3a 4c 64 4c 4f 64 71  DHHC-1:00:LdLOdq
    79 56 69 67 77 48 55 32 6d 5a 59 4c 7a 35 59 38  yVigwHU2mZYLz5Y8
  backtrace:
    [<00000000b6fc5071>] kstrdup+0x2e/0x60
    [<00000000f0f4633f>] 0xffffffffc0e07ee6
    [<0000000053006c05>] 0xffffffffc0dff783
    [<00000000419ae922>] configfs_write_iter+0xb1/0x120
    [<000000008183c424>] vfs_write+0x2be/0x3c0
    [<000000009005a2a5>] ksys_write+0x5f/0xe0
    [<00000000cd495c89>] do_syscall_64+0x38/0x90
    [<00000000f2a84ac5>] entry_SYSCALL_64_after_hwframe+0x63/0xcd

Fixes: db1312dd9548 ("nvmet: implement basic In-Band Authentication")
Signed-off-by: Sagi Grimberg <[email protected]>
Signed-off-by: Christoph Hellwig <[email protected]>

nvme-pci: add NVME_QUIRK_BOGUS_NID for Netac NV7000

Added a quirk to fix the Netac NV7000 SSD reporting duplicate NGUIDs.

Cc: <[email protected]>
Signed-off-by: Tiago Dias Ferreira <[email protected]>
Reviewed-by: Chaitanya Kulkarni <[email protected]>
Signed-off-by: Christoph Hellwig <[email protected]>

MAINTAINERS: Add linux-kbuild's patchwork

Add patchwork URL for Kconfig and Kbuild.

Signed-off-by: Nicolas Schier <[email protected]>
Reviewed-by: Nick Desaulniers <[email protected]>
Signed-off-by: Masahiro Yamada <[email protected]>

MAINTAINERS: Remove Michal Marek from Kbuild maintainers

Remove Michal Marek from Kbuild maintainers as there is no response from him
since October 2017.  Add an entry for Michal in CREDITS.

Michal, thanks for maintaining Kbuild for almost eight years!

Suggested-by: Nick Desaulniers <[email protected]>
Cc: Michal Marek <[email protected]>
Signed-off-by: Nicolas Schier <[email protected]>
Signed-off-by: Masahiro Yamada <[email protected]>

MAINTAINERS: Add Nathan and Nicolas to Kbuild reviewers

As suggested by Nick, add Nathan and myself to Kbuild reviewers to share more
review responsibilities.

Signed-off-by: Nicolas Schier <[email protected]>
Acked-by: Nick Desaulniers <[email protected]>
Acked-by: Nathan Chancellor <[email protected]>
Signed-off-by: Masahiro Yamada <[email protected]>

net: dsa: don't leak tagger-owned storage on switch driver unbind

In the initial commit dc452a471dba ("net: dsa: introduce tagger-owned
storage for private and shared data"), we had a call to
tag_ops->disconnect(dst) issued from dsa_tree_free(), which is called at
tree teardown time.

There were problems with connecting to a switch tree as a whole, so this
got reworked to connecting to individual switches within the tree. In
this process, tag_ops->disconnect(ds) was made to be called only from
switch.c (cross-chip notifiers emitted as a result of dynamic tag proto
changes), but the normal driver teardown code path wasn't replaced with
anything.

Solve this problem by adding a function that does the opposite of
dsa_switch_setup_tag_protocol(), which is called from the equivalent
spot in dsa_switch_teardown(). The positioning here also ensures that we
won't have any use-after-free in tagging protocol (*rcv) ops, since the
teardown sequence is as follows:

dsa_tree_teardown
-> dsa_tree_teardown_master
   -> dsa_master_teardown
      -> unsets master->dsa_ptr, making no further packets match the
         ETH_P_XDSA packet type handler
-> dsa_tree_teardown_ports
   -> dsa_port_teardown
      -> dsa_slave_destroy
         -> unregisters DSA net devices, there is even a synchronize_net()
            in unregister_netdevice_many()
-> dsa_tree_teardown_switches
   -> dsa_switch_teardown
      -> dsa_switch_teardown_tag_protocol
         -> finally frees the tagger-owned storage

Fixes: 7f2973149c22 ("net: dsa: make tagging protocols connect to individual switches from a tree")
Signed-off-by: Vladimir Oltean <[email protected]>
Reviewed-by: Saeed Mahameed <[email protected]>
Reviewed-by: Florian Fainelli <[email protected]>
Link: https://lore.kernel.org/r/[email protected]
Signed-off-by: Jakub Kicinski <[email protected]>

net/x25: Fix skb leak in x25_lapb_receive_frame()

x25_lapb_receive_frame() using skb_copy() to get a private copy of
skb, the new skb should be freed in the undersized/fragmented skb
error handling path. Otherwise there is a memory leak.

Fixes: cb101ed2c3c7 ("x25: Handle undersized/fragmented skbs")
Signed-off-by: Wei Yongjun <[email protected]>
Acked-by: Martin Schiller <[email protected]>
Link: https://lore.kernel.org/r/[email protected]
Signed-off-by: Jakub Kicinski <[email protected]>

net: ag71xx: call phylink_disconnect_phy if ag71xx_hw_enable() fail in ag71xx_open()

If ag71xx_hw_enable() fails, call phylink_disconnect_phy() to clean up.
And if phylink_of_phy_connect() fails, nothing needs to be done.
Compile tested only.

Fixes: 892e09153fa3 ("net: ag71xx: port to phylink")
Signed-off-by: Liu Jian <[email protected]>
Reviewed-by: Russell King (Oracle) <[email protected]>
Link: https://lore.kernel.org/r/[email protected]
Signed-off-by: Jakub Kicinski <[email protected]>

AMerge tag 'netfs-fixes-20221115' of git://git.kernel.org/pub/scm/linux/kernel/git/dhowells/linux-fs

Pull netfx fixes from David Howells:
 "Two fixes, affecting the functions that iterates over the pagecache
  unmarking or unlocking pages after an op is complete:

   - xas_for_each() loops must call xas_retry() first thing and
     immediately do a "continue" in the case that the extracted value is
     a special value that indicates that the walk raced with a
     modification. Fix the unlock and unmark loops to do this.

   - The maths in the unlock loop is dodgy as it could, theoretically,
     at some point in the future end up with a starting file pointer
     that is in the middle of a folio. This will cause a subtraction to
     go negative - but the number is unsigned. Fix the maths to use
     absolute file positions instead of relative page indices"

* tag 'netfs-fixes-20221115' of git://git.kernel.org/pub/scm/linux/kernel/git/dhowells/linux-fs:
  netfs: Fix dodgy maths
  netfs: Fix missing xas_retry() calls in xarray iteration

drm/display: Don't assume dual mode adaptors support i2c sub-addressing

Current dual mode adaptor ("DP++") detection code assumes that all
adaptors support i2c sub-addressing for read operations from the
DP-HDMI adaptor ID buffer.  It has been observed that multiple
adaptors do not in fact support this, and always return data starting
at register 0.  On affected adaptors, the code fails to read the proper
registers that would identify the device as a type 2 adaptor, and
handles those as type 1, limiting the TMDS clock to 165MHz, even if
the according register would announce a higher TMDS clock.
Fix this by always reading the ID buffer starting from offset 0, and
discarding any bytes before the actual offset of interest.

We tried finding authoritative documentation on whether or not this is
allowed behaviour, but since all the official VESA docs are paywalled,
the best we could come up with was the spec sheet for Texas Instruments'
SNx5DP149 chip family.[1]  It explicitly mentions that sub-addressing is
supported for register writes, but *not* for reads (See NOTE in
section 8.5.3).  Unless TI openly decided to violate the VESA spec, one
could take that as a hint that sub-addressing is in fact not mandated
by VESA.
The other two adaptors affected used the PS8409(A) and the LT8611,
according to the data returned from their ID buffers.

[1] https://www.ti.com/lit/ds/symlink/sn75dp149.pdf

Cc: [email protected]
Signed-off-by: Simon Rettberg <[email protected]>
Reviewed-by: Rafael Gieschke <[email protected]>
Signed-off-by: Ville Syrjälä <[email protected]>
Link: https://patchwork.freedesktop.org/patch/msgid/20221006113314.41101987@computer
Acked-by: Jani Nikula <[email protected]>

cifs: add check for returning value of SMB2_close_init

If the returning value of SMB2_close_init is an error-value,
exit the function.

Found by Linux Verification Center (linuxtesting.org) with SVACE.

Fixes: 352d96f3acc6 ("cifs: multichannel: move channel selection above transport layer")
Signed-off-by: Anastasia Belova <[email protected]>
Signed-off-by: Steve French <[email protected]>

Merge tag 'erofs-for-6.1-rc6-fixes' of git://git.kernel.org/pub/scm/linux/kernel/git/xiang/erofs

Pull erofs fixes from Gao Xiang:
 "Most patches randomly fix error paths or corner cases in fscache mode
  reported recently. One fixes an invalid access relating to fragments
  on crafted images.

  Summary:

   - Fix packed_inode invalid access when reading fragments on crafted
     images

   - Add a missing erofs_put_metabuf() in an error path in fscache mode

   - Fix incorrect `count' for unmapped extents in fscache mode

   - Fix use-after-free of fsid and domain_id string when remounting

   - Fix missing xas_retry() in fscache mode"

* tag 'erofs-for-6.1-rc6-fixes' of git://git.kernel.org/pub/scm/linux/kernel/git/xiang/erofs:
  erofs: fix missing xas_retry() in fscache mode
  erofs: fix use-after-free of fsid and domain_id string
  erofs: get correct count for unmapped range in fscache mode
  erofs: put metabuf in error path in fscache mode
  erofs: fix general protection fault when reading fragment

drm/amd/pm: fix SMU13 runpm hang due to unintentional workaround

The workaround designed for some specific ASICs is wrongly applied
to SMU13 ASICs. That leads to some runpm hang.

Signed-off-by: Evan Quan <[email protected]>
Reviewed-by: Hawking Zhang <[email protected]>
Reviewed-by: Feifei Xu <[email protected]>
Signed-off-by: Alex Deucher <[email protected]>
Cc: [email protected]

drm/amd/pm: enable runpm support over BACO for SMU13.0.7

Enable SMU13.0.7 runpm support.

Signed-off-by: Evan Quan <[email protected]>
Reviewed-by: Hawking Zhang <[email protected]>
Reviewed-by: Feifei Xu <[email protected]>
Signed-off-by: Alex Deucher <[email protected]>
Cc: [email protected] # 6.0.x

drm/amd/pm: enable runpm support over BACO for SMU13.0.0

Enable SMU13.0.0 runpm support.

Signed-off-by: Evan Quan <[email protected]>
Reviewed-by: Hawking Zhang <[email protected]>
Reviewed-by: Feifei Xu <[email protected]>
Signed-off-by: Alex Deucher <[email protected]>
Cc: [email protected] # 6.0.x

drm/amdgpu: there is no vbios fb on devices with no display hw (v2)

If we enable virtual display functionality on parts with
no display hardware we can end up trying to check for and
reserve the vbios FB area on devices where it doesn't exist.
Check if display hardware is actually present on the hardware
before trying to reserve the memory.

v2: move the check into common code

Acked-by: Evan Quan <[email protected]>
Signed-off-by: Alex Deucher <[email protected]>

drm/amdkfd: Fix a memory limit issue

It is to resolve a regression, which fails to allocate
VRAM due to no free memory in application, the reason
is we add check of vram_pin_size for memory limit, and
application is pinning the memory for Peerdirect, KFD
should not count it in memory limit. So removing
vram_pin_size will resolve it.

Signed-off-by: Eric Huang <[email protected]>
Reviewed-by: Felix Kuehling <[email protected]>
Signed-off-by: Alex Deucher <[email protected]>

drm/amdgpu: disable BACO support on more cards

Otherwise, some unexpected PCIE AER errors will be observed
in runtime suspend/resume cycle.

Signed-off-by: Guchun Chen <[email protected]>
Acked-by: Alex Deucher <[email protected]>
Signed-off-by: Alex Deucher <[email protected]>

drm/amd/display: don't enable DRM CRTC degamma property for DCE

DM maps DRM CRTC degamma to DPP (pre-blending) degamma block, but DCE doesn't
support programmable degamma curve anywhere. Currently, a custom degamma is
accepted by DM but just ignored by DCE driver and degamma correction isn't
actually applied. There is no way to map custom degamma in DCE, therefore, DRM
CRTC degamma property shouldn't be enabled for DCE drivers.

Reviewed-by: Rodrigo Siqueira <[email protected]>
Signed-off-by: Melissa Wen <[email protected]>
Signed-off-by: Rodrigo Siqueira <[email protected]>
Signed-off-by: Alex Deucher <[email protected]>
Cc: [email protected]

drm/amd/display: Set max for prefetch lines on dcn32

[WHY?]
Max number of lines that can be used for prefetch due to type constraints is
63.75.

[HOW?]
Enforce maximum prefetch lines as 63.75.

Reviewed-by: Jun Lei <[email protected]>
Acked-by: Tom Chung <[email protected]>
Signed-off-by: Dillon Varone <[email protected]>
Tested-by: Daniel Wheeler <[email protected]>
Signed-off-by: Alex Deucher <[email protected]>

drm/amd/display: use uclk pstate latency for fw assisted mclk validation dcn32

[WHY?]
DCN32 uses fclk pstate watermarks for dummy pstate, and must always be
supported.

[HOW?]
Validation needs to be run with fclk pstate latency set
as the dummy pstate latency to get correct prefetch and bandwidth outputs.

Reviewed-by: Jun Lei <[email protected]>
Acked-by: Tom Chung <[email protected]>
Signed-off-by: Dillon Varone <[email protected]>
Tested-by: Daniel Wheeler <[email protected]>
Signed-off-by: Alex Deucher <[email protected]>

drm/amd/display: Fix prefetch calculations for dcn32

[Description]
Prefetch calculation loop was not exiting until utilizing all of vstartup if it
failed once.  Locals need to be reset on each iteration of the loop.

Reviewed-by: Jun Lei <[email protected]>
Acked-by: Tom Chung <[email protected]>
Signed-off-by: Dillon Varone <[email protected]>
Tested-by: Daniel Wheeler <[email protected]>
Signed-off-by: Alex Deucher <[email protected]>
Cc: [email protected] # 6.0.x

drm/amd/display: Fix optc2_configure warning on dcn314

[Why]
dcn314 uses optc2_configure_crc() that wraps
optc1_configure_crc() + set additional registers
not applicable to dcn314.
It's not critical but when used leads to warning like:
WARNING: drivers/gpu/drm/amd/amdgpu/../display/dc/dc_helper.c
Call Trace:
<TASK>
generic_reg_set_ex+0x6d/0xe0 [amdgpu]
optc2_configure_crc+0x60/0x80 [amdgpu]
dc_stream_configure_crc+0x129/0x150 [amdgpu]
amdgpu_dm_crtc_configure_crc_source+0x5d/0xe0 [amdgpu]

[How]
Use optc1_configure_crc() directly

Reviewed-by: Nicholas Kazlauskas <[email protected]>
Acked-by: Tom Chung <[email protected]>
Signed-off-by: Roman Li <[email protected]>
Tested-by: Daniel Wheeler <[email protected]>
Signed-off-by: Alex Deucher <[email protected]>
Cc: [email protected] # 6.0.x

drm/amd/display: Fix calculation for cursor CAB allocation

[Why]
The cursor size (in memory) is currently incorrectly calculated,
resulting not enough CAB being allocated for static screen cursor
in MALL refresh. This results in cursor image corruption.

[How]
Use cursor pitch instead of cursor width when calculating cursor size.
Update num cache lines calculation to use the result of the cursor size
calculation instead of manually recalculating again.

Reviewed-by: Alvin Lee <[email protected]>
Acked-by: Tom Chung <[email protected]>
Signed-off-by: George Shen <[email protected]>
Tested-by: Daniel Wheeler <[email protected]>
Signed-off-by: Alex Deucher <[email protected]>
Cc: [email protected] # 6.0.x

Revert "drm: hide unregistered connectors from GETCONNECTOR IOCTL"

This reverts commit 981f09295687f856d5345e19c7084aca481c1395.

It turns out this causes logically active but disconnected DP MST
connectors to disappear from the KMS resources list, and Mutter
then assumes the connector is already disabled. Later on Mutter tries
to re-use the same CRTC but fails since on the kernel side it's still
tied to the disconnected DP MST connector.

Signed-off-by: Simon Ser <[email protected]>
Tested-by: Jonas Ådahl <[email protected]>
Reviewed-by: Lyude Paul <[email protected]>
Cc: Daniel Vetter <[email protected]>
Link: https://patchwork.freedesktop.org/patch/msgid/[email protected]

drm/amd/display: Support parsing VRAM info v3.0 from VBIOS

[Why]
For DCN3.2 and DCN3.21, VBIOS has switch to using v3.0 of the VRAM
info struct. We should read and override the VRAM info in driver with
values provided by VBIOS to support memory downbin cases.

Reviewed-by: Alvin Lee <[email protected]>
Acked-by: Tom Chung <[email protected]>
Signed-off-by: George Shen <[email protected]>
Tested-by: Daniel Wheeler <[email protected]>
Signed-off-by: Alex Deucher <[email protected]>
Cc: [email protected] # 6.0.x

x86/cpu: Restore AMD's DE_CFG MSR after resume

DE_CFG contains the LFENCE serializing bit, restore it on resume too.
This is relevant to older families due to the way how they do S3.

Unify and correct naming while at it.

Fixes: e4d0e84e4907 ("x86/cpu/AMD: Make LFENCE a serializing instruction")
Reported-by: Andrew Cooper <[email protected]>
Reported-by: Pawan Gupta <[email protected]>
Signed-off-by: Borislav Petkov <[email protected]>
Cc: <[email protected]>
Signed-off-by: Linus Torvalds <[email protected]>

drm/amd/display: Fix invalid DPIA AUX reply causing system hang

[Why]
Some DPIA AUX replies have incorrect data length from original request.
This could lead to overwriting of destination buffer if reply length is
larger, which could cause invalid access to stack since many destination
buffers are declared as local variables.

[How]
Check for invalid length from DPIA AUX replies and trigger a retry if
reply length is not the same as original request. A DRM_WARN() dmesg log
is also produced.

Reviewed-by: Roman Li <[email protected]>
Acked-by: Tom Chung <[email protected]>
Signed-off-by: Stylon Wang <[email protected]>
Tested-by: Daniel Wheeler <[email protected]>
Signed-off-by: Alex Deucher <[email protected]>
Cc: [email protected] # 6.0.x

drm/amdgpu: Add psp_13_0_10_ta firmware to modinfo

TA firmware loaded on psp v13_0_10, but it is missing in modinfo.

Signed-off-by: Candice Li <[email protected]>
Reviewed-by: Hawking Zhang <[email protected]>
Signed-off-by: Alex Deucher <[email protected]>

drm/amd/display: Add HUBP surface flip interrupt handler

On IGT, there is a test named amd_hotplug, and when the subtest basic is
executed on DCN31, we get the following error:

[drm] *ERROR* [CRTC:71:crtc-0] flip_done timed out
[drm] *ERROR* flip_done timed out
[drm] *ERROR* [CRTC:71:crtc-0] commit wait timed out
[drm] *ERROR* flip_done timed out
[drm] *ERROR* [CONNECTOR:88:DP-1] commit wait timed out
[drm] *ERROR* flip_done timed out
[drm] *ERROR* [PLANE:59:plane-3] commit wait timed out

After enable the page flip log with the below command:

 echo -n 'format "[PFLIP]" +p' > /sys/kernel/debug/dynamic_debug/control

It is possible to see that the flip was submitted, but DC never replied
back, which generates time-out issues. This is an indication that the
HUBP surface flip is missing. This commit fixes this issue by adding
hubp1_set_flip_int to DCN31.

Reviewed-by: Nicholas Kazlauskas <[email protected]>
Acked-by: Tom Chung <[email protected]>
Signed-off-by: Rodrigo Siqueira <[email protected]>
Tested-by: Daniel Wheeler <[email protected]>
Signed-off-by: Alex Deucher <[email protected]>
Cc: [email protected]

drm/amd/display: Fix access timeout to DPIA AUX at boot time

[Why]
Since introduction of patch "Query DPIA HPD status.", link detection at
boot could be accessing DPIA AUX, which will not succeed until
DMUB outbox messaging is enabled and results in below dmesg logs:

[  160.840227] [drm:amdgpu_dm_process_dmub_aux_transfer_sync [amdgpu]] *ERROR* wait_for_completion_timeout timeout!

[How]
Enable DMUB outbox messaging before link detection at boot time.

Reviewed-by: Wayne Lin <[email protected]>
Acked-by: Tom Chung <[email protected]>
Signed-off-by: Stylon Wang <[email protected]>
Tested-by: Daniel Wheeler <[email protected]>
Signed-off-by: Alex Deucher <[email protected]>
Cc: [email protected] # 6.0.x

drm/amdgpu: Fix memory leak in amdgpu_cs_pass1

When p->gang_size equals 0, amdgpu_cs_pass1() will return directly
without freeing chunk_array, which will cause a memory leak issue,
this patch fixes it.

Fixes: 4624459c84d7 ("drm/amdgpu: add gang submit frontend v6")
Reviewed-by: Luben Tuikov <[email protected]>
Signed-off-by: Dong Chenchen <[email protected]>
Signed-off-by: Alex Deucher <[email protected]>

ALSA: hda/realtek: Fix the speaker output on Samsung Galaxy Book Pro 360

Samsung Galaxy Book Pro 360 (13" 2021 NP930QBD-ke1US) with codec SSID
144d:c1a6 requires the same workaround for enabling the speaker amp
like other Samsung models with ALC298 codec.

Link: https://bugzilla.opensuse.org/show_bug.cgi?id=1205100
Cc: <[email protected]>
Link: https://lore.kernel.org/r/[email protected]
Signed-off-by: Takashi Iwai <[email protected]>

ALSA: hda/realtek: fix speakers for Samsung Galaxy Book Pro

The Samsung Galaxy Book Pro seems to have the same issue as a few
other Samsung laptops, detailed in kernel bug report 207423. Sound from
headphone jack works, but not the built-in speakers.

alsa-info: http://alsa-project.org/db/?f=b40ba609dc6ae28dc84ad404a0d8a4bbcd8bea6d

Signed-off-by: Emil Flink <[email protected]>
Cc: <[email protected]>
Link: https://lore.kernel.org/r/[email protected]
Signed-off-by: Takashi Iwai <[email protected]>

netfs: Fix dodgy maths

Fix the dodgy maths in netfs_rreq_unlock_folios().  start_page could be
inside the folio, in which case the calculation of pgpos will be come up
with a negative number (though for the moment rreq->start is rounded down
earlier and folios would have to get merged whilst locked)

Alter how this works to just frame the tracking in terms of absolute file
positions, rather than offsets from the start of the I/O request.  This
simplifies the maths and makes it easier to follow.

Fix the issue by using folio_pos() and folio_size() to calculate the end
position of the page.

Fixes: 3d3c95046742 ("netfs: Provide readahead and readpage netfs helpers")
Reported-by: Matthew Wilcox <[email protected]>
Signed-off-by: David Howells <[email protected]>
Reviewed-by: Jeff Layton <[email protected]>
Reviewed-by: Jingbo Xu <[email protected]>
cc: [email protected]
cc: [email protected]
Link: https://lore.kernel.org/r/[email protected]/
Link: https://lore.kernel.org/r/166757988611.950645.7626959069846893164.stgit@warthog.procyon.org.uk/

netfs: Fix missing xas_retry() calls in xarray iteration

netfslib has a number of places in which it performs iteration of an xarray
whilst being under the RCU read lock.  It *should* call xas_retry() as the
first thing inside of the loop and do "continue" if it returns true in case
the xarray walker passed out a special value indicating that the walk needs
to be redone from the root[*].

Fix this by adding the missing retry checks.

[*] I wonder if this should be done inside xas_find(), xas_next_node() and
    suchlike, but I'm told that's not an simple change to effect.

This can cause an oops like that below.  Note the faulting address - this
is an internal value (|0x2) returned from xarray.

BUG: kernel NULL pointer dereference, address: 0000000000000402
...
RIP: 0010:netfs_rreq_unlock+0xef/0x380 [netfs]
...
Call Trace:
 netfs_rreq_assess+0xa6/0x240 [netfs]
 netfs_readpage+0x173/0x3b0 [netfs]
 ? init_wait_var_entry+0x50/0x50
 filemap_read_page+0x33/0xf0
 filemap_get_pages+0x2f2/0x3f0
 filemap_read+0xaa/0x320
 ? do_filp_open+0xb2/0x150
 ? rmqueue+0x3be/0xe10
 ceph_read_iter+0x1fe/0x680 [ceph]
 ? new_sync_read+0x115/0x1a0
 new_sync_read+0x115/0x1a0
 vfs_read+0xf3/0x180
 ksys_read+0x5f/0xe0
 do_syscall_64+0x38/0x90
 entry_SYSCALL_64_after_hwframe+0x44/0xae

Changes:
========
ver #2)
 - Changed an unsigned int to a size_t to reduce the likelihood of an
   overflow as per Willy's suggestion.
 - Added an additional patch to fix the maths.

Fixes: 3d3c95046742 ("netfs: Provide readahead and readpage netfs helpers")
Reported-by: George Law <[email protected]>
Signed-off-by: David Howells <[email protected]>
Reviewed-by: Jeff Layton <[email protected]>
Reviewed-by: Jingbo Xu <[email protected]>
cc: Matthew Wilcox <[email protected]>
cc: [email protected]
cc: [email protected]
Link: https://lore.kernel.org/r/166749229733.107206.17482609105741691452.stgit@warthog.procyon.org.uk/
Link: https://lore.kernel.org/r/166757987929.950645.12595273010425381286.stgit@warthog.procyon.org.uk/

platform/surface: aggregator_registry: Add support for Surface Pro 9

Add device nodes to enable support for battery and charger status, the
ACPI platform profile, as well as internal and type-cover HID devices
(including sensors, touchpad, keyboard, and other miscellaneous devices)
on the Surface Pro 9.

This does not include support for a tablet-mode switch yet, as that is
now handled via the POS subsystem (unlike the Surface Pro 8, where it is
handled via the KIP subsystem) and therefore needs further changes.

While we're at it, also add the missing comment for the Surface Pro 8.

Signed-off-by: Maximilian Luz <[email protected]>
Link: https://lore.kernel.org/r/[email protected]
Reviewed-by: Hans de Goede <[email protected]>
Signed-off-by: Hans de Goede <[email protected]>

platform/surface: aggregator: Do not check for repeated unsequenced packets

Currently, we check any received packet whether we have already seen it
previously, regardless of the packet type (sequenced / unsequenced). We
do this by checking the sequence number. This assumes that sequence
numbers are valid for both sequenced and unsequenced packets. However,
this assumption appears to be incorrect.

On some devices, the sequence number field of unsequenced packets (in
particular HID input events on the Surface Pro 9) is always zero. As a
result, the current retransmission check kicks in and discards all but
the first unsequenced packet, breaking (among other things) keyboard and
touchpad input.

Note that we have, so far, only seen packets being retransmitted in
sequenced communication. In particular, this happens when there is an
ACK timeout, causing the EC (or us) to re-send the packet waiting for an
ACK. Arguably, retransmission / duplication of unsequenced packets
should not be an issue as there is no logical condition (such as an ACK
timeout) to determine when a packet should be sent again.

Therefore, remove the retransmission check for unsequenced packets
entirely to resolve the issue.

Fixes: c167b9c7e3d6 ("platform/surface: Add Surface Aggregator subsystem")
Signed-off-by: Maximilian Luz <[email protected]>
Link: https://lore.kernel.org/r/[email protected]
Reviewed-by: Hans de Goede <[email protected]>
Signed-off-by: Hans de Goede <[email protected]>

platform/x86: acer-wmi: Enable SW_TABLET_MODE on Switch V 10 (SW5-017)

Like the Acer Switch 10 (SW5-012) and Acer Switch 10 (S1003) models
the Acer Switch V 10 (SW5-017) supports reporting SW_TABLET_MODE
through acer-wmi.

Add a DMI quirk for the SW5-017 setting force_caps to ACER_CAP_KBD_DOCK
(these devices have no other acer-wmi based functionality).

Cc: Rudolf Polzer <[email protected]>
Signed-off-by: Hans de Goede <[email protected]>
Link: https://lore.kernel.org/r/[email protected]

platform/x86: asus-wmi: add missing pci_dev_put() in asus_wmi_set_xusb2pr()

pci_get_device() will increase the reference count for the returned
pci_dev. We need to use pci_dev_put() to decrease the reference count
before asus_wmi_set_xusb2pr() returns.

Signed-off-by: Xiongfeng Wang <[email protected]>
Link: https://lore.kernel.org/r/[email protected]
Reviewed-by: Hans de Goede <[email protected]>
Signed-off-by: Hans de Goede <[email protected]>

platform/x86/intel: pmc: Don't unconditionally attach Intel PMC when virtualized

The current logic in the Intel PMC driver will forcefully attach it
when detecting any CPU on the intel_pmc_core_platform_ids array,
even if the matching ACPI device is not present.

There's no checking in pmc_core_probe() to assert that the PMC device
is present, and hence on virtualized environments the PMC device
probes successfully, even if the underlying registers are not present.
Before commit 21ae43570940 ("platform/x86: intel_pmc_core: Substitute PCI
with CPUID enumeration") the driver would check for the presence of a
specific PCI device, and that prevented the driver from attaching when
running virtualized.

Fix by only forcefully attaching the PMC device when not running
virtualized.  Note that virtualized platforms can still get the device
to load if the appropriate ACPI device is present on the tables
provided to the VM.

Make an exception for the Xen initial domain, which does have full
hardware access, and hence can attach to the PMC if present.

Fixes: 21ae43570940 ("platform/x86: intel_pmc_core: Substitute PCI with CPUID enumeration")
Signed-off-by: Roger Pau Monné <[email protected]>
Acked-by: David E. Box <[email protected]>
Reviewed-by: Andy Shevchenko <[email protected]>
Link: https://lore.kernel.org/r/[email protected]
Reviewed-by: Hans de Goede <[email protected]>
Signed-off-by: Hans de Goede <[email protected]>

platform/x86: thinkpad_acpi: Enable s2idle quirk for 21A1 machine type

Previously, the s2idle quirk was only active for the 21A0 machine type
of the P14s Gen2a product. This also enables it for the second 21A1 type,
thus reducing wake-up times from s2idle.

Signed-off-by: Lennard Gäher <[email protected]>
Suggested-by: Mario Limonciello <[email protected]>
Reviewed-by: Mario Limonciello <[email protected]>
Link: https://gitlab.freedesktop.org/drm/amd/-/issues/2181
Link: https://lore.kernel.org/r/[email protected]
Reviewed-by: Hans de Goede <[email protected]>
Signed-off-by: Hans de Goede <[email protected]>

platform/x86/amd: pmc: Add new ACPI ID AMDI0009

Add new a new ACPI ID AMDI0009 used by upcoming AMD platform to the pmc
supported list of devices.

Cc: [email protected] # 6.0
Signed-off-by: Shyam Sundar S K <[email protected]>
Reviewed-by: Mario Limonciello <[email protected]>
Link: https://lore.kernel.org/r/[email protected]
Reviewed-by: Hans de Goede <[email protected]>
Signed-off-by: Hans de Goede <[email protected]>

MAINTAINERS: repair Microchip corei2c driver entry

The driver was renamed before application but the relevant change did
not propagate to the MAINTAINERS patch that was applied. Repair it.

CC: Lukas Bulwahn <[email protected]>
Signed-off-by: Conor Dooley <[email protected]>
Link: https://lore.kernel.org/r/[email protected]
Signed-off-by: Arnd Bergmann <[email protected]>

MAINTAINERS: add an entry for StarFive devicetrees

Emil looks after the downstream StarFive stuff, and agreed to look after
the upstream ones too.

CC: Emil Renner Berthing <[email protected]>
Signed-off-by: Conor Dooley <[email protected]>
Acked-by: Emil Renner Berthing <[email protected]>
Link: https://lore.kernel.org/r/[email protected]
Signed-off-by: Arnd Bergmann <[email protected]>

MAINTAINERS: generify the Microchip RISC-V entry name

These drivers work on our other FPGAs, for example the non-SoC PolarFire
connected to an FU-540 via chiplink. Make the entry a wee bit more
generic to match. While at it, remove the / from the heading so that it
matches other, neighbouring RISC-V entries.

Signed-off-by: Conor Dooley <[email protected]>
Link: https://lore.kernel.org/r/[email protected]
Signed-off-by: Arnd Bergmann <[email protected]>

MAINTAINERS: add entries for misc. RISC-V SoC drivers and devicetrees

Following some discussion both on & off list, I have volunteered to take
over maintaining the miscellaneous RISC-V devicetrees & soc drivers from
Palmer to ease his load.

So far only SiFive and Microchip have stuff in drivers/soc. For the
former, a SiFive entry exists with a dead GitHub repo - so remove that
to avoid confusion since the patches for drivers/soc & devicetrees will
be routed via my tree & other drivers go through their subsystem trees.
The Microchip directory only contains a RISC-V driver for now, but is
likely to contain drivers for other archs in the future. To that end,
change the PolarFire SoC entry to specifically mention the RISC-V driver
& the new directory level entry does not mention an architecture.

CC: Arnd Bergmann <[email protected]>
CC: Nicolas Ferre <[email protected]>
CC: Palmer Dabbelt <[email protected]>
Link: https://lore.kernel.org/linux-riscv/mhng-e4210f56-fcc3-4db8-abdb-d43b3ebe695d@palmer-ri-x1c9a/
Signed-off-by: Conor Dooley <[email protected]>
Acked-by: Nicolas Ferre <[email protected]>
Acked-by: Palmer Dabbelt <[email protected]>
Link: https://lore.kernel.org/r/[email protected]
Signed-off-by: Arnd Bergmann <[email protected]>

drbd: use after free in drbd_create_device()

The drbd_destroy_connection() frees the "connection" so use the _safe()
iterator to prevent a use after free.

Fixes: b6f85ef9538b ("drbd: Iterate over all connections")
Signed-off-by: Dan Carpenter <[email protected]>
Reviewed-by: Christoph Böhmwalder <[email protected]>
Link: https://lore.kernel.org/r/Y3Jd5iZRbNQ9w6gm@kili
Signed-off-by: Jens Axboe <[email protected]>

drm/amdgpu: use the last IB as gang leader v2

It turned out that not the last IB specified is the gang leader,
but instead the last job allocated.

This is a bit unfortunate and not very intuitive for the CS
interface, so try to fix this.

Signed-off-by: Christian König <[email protected]>
Link: https://patchwork.freedesktop.org/patch/msgid/[email protected]
Tested-by: Timur Kristóf <[email protected]>
Acked-by: Timur Kristóf <[email protected]>
Reviewed-by: Alex Deucher <[email protected]>
Fixes: 4624459c84d7 ("drm/amdgpu: add gang submit frontend v6")

perf pmu: Restructure print_pmu_events() to avoid memory allocations

Previously print_pmu_events() would compute the values to be printed,
place them in struct sevent, sort them and then print them.

Modify the code so that struct sevent holds just the PMU and event, sort
these and then in the main print loop calculate aliases for names, etc.

This avoids memory allocations for copied values as they are computed
then printed.

Signed-off-by: Ian Rogers <[email protected]>
Cc: Adrian Hunter <[email protected]>
Cc: Alexander Shishkin <[email protected]>
Cc: Caleb Biggers <[email protected]>
Cc: Jiri Olsa <[email protected]>
Cc: Kajol Jain <[email protected]>
Cc: Kan Liang <[email protected]>
Cc: Leo Yan <[email protected]>
Cc: Mark Rutland <[email protected]>
Cc: Namhyung Kim <[email protected]>
Cc: Perry Taylor <[email protected]>
Cc: Peter Zijlstra <[email protected]>
Cc: Ravi Bangoria <[email protected]>
Cc: Rob Herring <[email protected]>
Cc: Sandipan Das <[email protected]>
Cc: Stephane Eranian <[email protected]>
Cc: Weilin Wang <[email protected]>
Cc: Xin Gao <[email protected]>
Cc: Xing Zhengjun <[email protected]>
Link: http://lore.kernel.org/lkml/[email protected]
Signed-off-by: Arnaldo Carvalho de Melo <[email protected]>

perf list: Simplify symbol event printing

The current code computes an array of symbol names then sorts and prints
them. Use a strlist to create a list of names that is sorted and then
print it.

Signed-off-by: Ian Rogers <[email protected]>
Cc: Adrian Hunter <[email protected]>
Cc: Alexander Shishkin <[email protected]>
Cc: Caleb Biggers <[email protected]>
Cc: Jiri Olsa <[email protected]>
Cc: Kajol Jain <[email protected]>
Cc: Kan Liang <[email protected]>
Cc: Leo Yan <[email protected]>
Cc: Mark Rutland <[email protected]>
Cc: Namhyung Kim <[email protected]>
Cc: Perry Taylor <[email protected]>
Cc: Peter Zijlstra <[email protected]>
Cc: Ravi Bangoria <[email protected]>
Cc: Rob Herring <[email protected]>
Cc: Sandipan Das <[email protected]>
Cc: Stephane Eranian <[email protected]>
Cc: Weilin Wang <[email protected]>
Cc: Xin Gao <[email protected]>
Cc: Xing Zhengjun <[email protected]>
Link: http://lore.kernel.org/lkml/[email protected]
Signed-off-by: Arnaldo Carvalho de Melo <[email protected]>

perf list: Simplify cache event printing

The current code computes an array of cache names then sorts and prints
them. Use a strlist to create a list of names that is sorted. Keep the
hybrid names, it is unclear how to generalize it, but drop the
computation of evt_pmus that is never used.

Signed-off-by: Ian Rogers <[email protected]>
Cc: Adrian Hunter <[email protected]>
Cc: Alexander Shishkin <[email protected]>
Cc: Caleb Biggers <[email protected]>
Cc: Jiri Olsa <[email protected]>
Cc: Kajol Jain <[email protected]>
Cc: Kan Liang <[email protected]>
Cc: Kang Minchul <[email protected]>
Cc: Leo Yan <[email protected]>
Cc: Mark Rutland <[email protected]>
Cc: Namhyung Kim <[email protected]>
Cc: Perry Taylor <[email protected]>
Cc: Peter Zijlstra <[email protected]>
Cc: Ravi Bangoria <[email protected]>
Cc: Rob Herring <[email protected]>
Cc: Sandipan Das <[email protected]>
Cc: Stephane Eranian <[email protected]>
Cc: Weilin Wang <[email protected]>
Cc: Xin Gao <[email protected]>
Cc: Xing Zhengjun <[email protected]>
Link: http://lore.kernel.org/lkml/[email protected]
[ Fixed up clash with cf9f67b36303de65 ("perf print-events: Remove redundant comparison with zero")]
Signed-off-by: Arnaldo Carvalho de Melo <[email protected]>