| lists.openwall.net | lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC | |
|
Open Source and information security mailing list archives
| ||
|
Message-ID: <CAJpZYjWSGePB6GOJb=MeLzA4bMc9BwjbkbND0p1xfkmvPZfj=Q@mail.gmail.com>
Date: Mon, 9 Dec 2024 12:03:08 -0800
From: Chun-Tse Shao <ctshao@...gle.com>
To: Namhyung Kim <namhyung@...nel.org>
Cc: linux-kernel@...r.kernel.org, peterz@...radead.org, mingo@...hat.com,
acme@...nel.org, mark.rutland@....com, alexander.shishkin@...ux.intel.com,
jolsa@...nel.org, irogers@...gle.com, adrian.hunter@...el.com,
kan.liang@...ux.intel.com, linux-perf-users@...r.kernel.org
Subject: Re: [PATCH v3 2/2] perf lock: Fix parse_lock_type which only retrieve
one lock flag
Thank you Namhyung, please check my v4 patches here:
https://lore.kernel.org/linux-perf-users/20241209200104.870531-1-ctshao@google.com/T/#t
On Mon, Dec 2, 2024 at 12:49 PM Namhyung Kim <namhyung@...nel.org> wrote:
>
> Hello,
>
> On Tue, Nov 26, 2024 at 01:22:54PM -0800, Chun-Tse Shao wrote:
> > `parse_lock_type` can only add the first lock flag in `lock_type_table`
> > given input `str`. For example, for `Y rwlock`, it only adds `rwlock:R`
> > into this perf session. Another example is for `-Y mutex`, it only adds
> > the mutex without `LCB_F_SPIN` flag. The patch fixes this issue, makes
> > sure both `rwlock:R` and `rwlock:W` will be added with `-Y rwlock`, and
> > so on.
> >
> > Testing:
> > $ ./perf lock con -ab -Y mutex,rwlock -- perf bench sched pipe
> > # Running 'sched/pipe' benchmark:
> > # Executed 1000000 pipe operations between two processes
> >
> > Total time: 8.425 [sec]
> >
> > 8.425402 usecs/op
> > 118688 ops/sec
> > contended total wait max wait avg wait type caller
> >
> > 194 1.68 ms 44.16 us 8.66 us mutex pipe_read+0x57
> > 10 423.03 us 44.27 us 42.30 us rwlock:W do_exit+0x365
> > 54 254.67 us 58.87 us 4.72 us mutex pipe_write+0x50
> > 21 146.64 us 11.54 us 6.98 us mutex pipe_read+0x282
> > 10 141.27 us 20.62 us 14.13 us rwlock:W release_task+0x6f
> > 5 58.92 us 16.37 us 11.78 us mutex do_epoll_wait+0x24d
> > 3 29.81 us 17.66 us 9.94 us mutex do_epoll_ctl+0x6c1
> > 4 26.82 us 11.02 us 6.70 us mutex do_epoll_wait+0x24d
> > 2 18.32 us 12.49 us 9.16 us rwlock:W do_epoll_wait+0x255
> > 1 11.34 us 11.34 us 11.34 us rwlock:W ep_done_scan+0x2d
> > 1 11.02 us 11.02 us 11.02 us rwlock:R mm_update_next_owner+0x4e
> > 1 10.60 us 10.60 us 10.60 us rwlock:W do_epoll_ctl+0xb65
> > 1 9.90 us 9.90 us 9.90 us rwlock:W do_exit+0x365
> >
> > Fixes: d783ea8f62c4 ("perf lock contention: Simplify parse_lock_type()")
>
> I think it won't apply to earlier kernels because you changed the field
> names in the previous commit. I think you can switch the order of the
> commits so that we can have the fix first.
>
>
> > Signed-off-by: Chun-Tse Shao <ctshao@...gle.com>
> > ---
> > tools/perf/builtin-lock.c | 68 ++++++++++++++++++++++++++-------------
> > 1 file changed, 45 insertions(+), 23 deletions(-)
> >
> > diff --git a/tools/perf/builtin-lock.c b/tools/perf/builtin-lock.c
> > index c528aff1c9d5..c1870c287580 100644
> > --- a/tools/perf/builtin-lock.c
> > +++ b/tools/perf/builtin-lock.c
> > @@ -1622,19 +1622,6 @@ static const char *get_type_lock_name(unsigned int flags)
> > return "unknown";
> > }
> >
> > -static unsigned int get_type_flag(const char *str)
> > -{
> > - for (unsigned int i = 0; i < ARRAY_SIZE(lock_type_table); i++) {
> > - if (!strcmp(lock_type_table[i].flags_name, str))
> > - return lock_type_table[i].flags;
> > - }
> > - for (unsigned int i = 0; i < ARRAY_SIZE(lock_type_table); i++) {
> > - if (!strcmp(lock_type_table[i].lock_name, str))
> > - return lock_type_table[i].flags;
> > - }
> > - return UINT_MAX;
> > -}
> > -
> > static void lock_filter_finish(void)
> > {
> > zfree(&filters.types);
> > @@ -2356,29 +2343,64 @@ static int parse_lock_type(const struct option *opt __maybe_unused, const char *
> > int unset __maybe_unused)
> > {
> > char *s, *tmp, *tok;
> > - int ret = 0;
> >
> > s = strdup(str);
> > if (s == NULL)
> > return -1;
> >
> > for (tok = strtok_r(s, ", ", &tmp); tok; tok = strtok_r(NULL, ", ", &tmp)) {
> > - unsigned int flags = get_type_flag(tok);
> > + bool found = false;
> >
> > - if (flags == -1U) {
> > - pr_err("Unknown lock flags: %s\n", tok);
> > - ret = -1;
> > - break;
> > + /* `tok` is a flags name if it contains ':'. */
> > + if (strchr(tok, ':')) {
> > + for (unsigned int i = 0; i < ARRAY_SIZE(lock_type_table); i++) {
> > + if (!strcmp(lock_type_table[i].flags_name, tok) &&
> > + add_lock_type(lock_type_table[i].flags)) {
> > + found = true;
> > + break;
> > + }
> > + }
> > +
> > + if (!found) {
> > + pr_err("Unknown lock flags name: %s\n", tok);
> > + free(s);
> > + return -1;
> > + }
> > +
> > + continue;
> > }
> >
> > - if (!add_lock_type(flags)) {
> > - ret = -1;
> > - break;
> > + /*
> > + * Otherwise `tok` is a lock name.
> > + * Single lock name could contain multiple flags.
> > + */
> > + /*
> > + * By documentation, `percpu-rwmem` should be `pcpu-sem`.
> > + * For backward compatibility, we replace `pcpu-sem` with `percpu-rwmem`.
> > + */
> > + if (!strcmp(tok, "pcpu-sem"))
> > + tok = (char *)"percpu-rwsem";
>
> I think it's better to split the pcpu-sem part into a separate commit as
> it fixes a different issue. Also you'd better merge the block commits.
>
> Thanks,
> Namhyung
>
>
> > +
> > + for (unsigned int i = 0; i < ARRAY_SIZE(lock_type_table); i++) {
> > + if (!strcmp(lock_type_table[i].lock_name, tok)) {
> > + if (add_lock_type(lock_type_table[i].flags)) {
> > + found = true;
> > + } else {
> > + free(s);
> > + return -1;
> > + }
> > + }
> > + }
> > +
> > + if (!found) {
> > + pr_err("Unknown lock name: %s\n", tok);
> > + free(s);
> > + return -1;
> > }
> > }
> >
> > free(s);
> > - return ret;
> > + return 0;
> > }
> >
> > static bool add_lock_addr(unsigned long addr)
> > --
> > 2.47.0.338.g60cca15819-goog
> >
Powered by blists - more mailing lists