Received-SPF: pass (google.com: domain of linux-kernel-owner@vger.kernel.org designates 2620:137:e000::1:20 as permitted sender) client-ip=2620:137:e000::1:20;
MIME-Version: 1.0
References: <20220412154817.2728324-1-irogers@google.com> <YlX5U+Qelf/rBdau@google.com>
In-Reply-To: <YlX5U+Qelf/rBdau@google.com>
From:   Ian Rogers <irogers@google.com>
Date:   Tue, 12 Apr 2022 16:48:28 -0700
Message-ID: <CAP-5=fUDQGtNn39xDSxagwVzN95dohNrfiU_a3pU8zkiv2-6tw@mail.gmail.com>
Subject: Re: [PATCH v2 0/4] Tidy up symbol end fixup
To:     Namhyung Kim <namhyung@kernel.org>
Cc:     John Garry <john.garry@huawei.com>, Will Deacon <will@kernel.org>,
        Mathieu Poirier <mathieu.poirier@linaro.org>,
        Leo Yan <leo.yan@linaro.org>,
        Peter Zijlstra <peterz@infradead.org>,
        Ingo Molnar <mingo@redhat.com>,
        Arnaldo Carvalho de Melo <acme@kernel.org>,
        Mark Rutland <mark.rutland@arm.com>,
        Alexander Shishkin <alexander.shishkin@linux.intel.com>,
        Jiri Olsa <jolsa@kernel.org>,
        James Clark <james.clark@arm.com>,
        Alexandre Truong <alexandre.truong@arm.com>,
        German Gomez <german.gomez@arm.com>,
        Dave Marchevsky <davemarchevsky@fb.com>,
        Song Liu <songliubraving@fb.com>,
        Ravi Bangoria <ravi.bangoria@amd.com>,
        Li Huafei <lihuafei1@huawei.com>,
        =?UTF-8?Q?Martin_Li=C5=A1ka?= <mliska@suse.cz>,
        William Cohen <wcohen@redhat.com>,
        Riccardo Mancini <rickyman7@gmail.com>,
        Masami Hiramatsu <mhiramat@kernel.org>,
        Thomas Richter <tmricht@linux.ibm.com>,
        Lexi Shao <shaolexi@huawei.com>,
        Remi Bernon <rbernon@codeweavers.com>,
        Michael Petlan <mpetlan@redhat.com>,
        Denis Nikitin <denik@chromium.org>,
        linux-arm-kernel@lists.infradead.org,
        linux-perf-users@vger.kernel.org, linux-kernel@vger.kernel.org,
        Stephane Eranian <eranian@google.com>
Content-Type: text/plain; charset="UTF-8"
Precedence: bulk

On Tue, Apr 12, 2022 at 3:12 PM Namhyung Kim <namhyung@kernel.org> wrote:
>
> Hi Ian,
>
> On Tue, Apr 12, 2022 at 08:48:13AM -0700, Ian Rogers wrote:
> > Fixing up more symbol ends as introduced in:
> > https://lore.kernel.org/lkml/20220317135536.805-1-mpetlan@redhat.com/
> > caused perf annotate to run into memory limits - every symbol holds
> > all the disassembled code in the annotation, and so making symbols
> > ends further away dramatically increased memory usage (40MB to
> >  >1GB). Modify the symbol end logic so that special kernel cases aren't
> > applied in the common case.
> >
> > v2. Drops a merged patch. Fixes a build issue with libbfd enabled.
>
> How about just like this?  We can get rid of arch functions as they
> mostly do the same thing (kernel vs module boundary check).
>
> Not tested. ;-)
>
> Thanks,
> Namhyung
>
> --------------8<-------------
>
> diff --git a/tools/perf/util/symbol.c b/tools/perf/util/symbol.c
> index dea0fc495185..df41d7266d91 100644
> --- a/tools/perf/util/symbol.c
> +++ b/tools/perf/util/symbol.c
> @@ -35,6 +35,7 @@
>  #include "path.h"
>  #include <linux/ctype.h>
>  #include <linux/zalloc.h>
> +#include <internal/lib.h>  // page_size
>
>  #include <elf.h>
>  #include <limits.h>
> @@ -231,8 +226,16 @@ void symbols__fixup_end(struct rb_root_cached *symbols)
>                 prev = curr;
>                 curr = rb_entry(nd, struct symbol, rb_node);
>
> -               if (prev->end == prev->start || prev->end != curr->start)
> -                       arch__symbols__fixup_end(prev, curr);
> +               if (prev->end == prev->start) {
> +                       /* Last kernel symbol mapped to end of page */

I like the simpler logic but don't like applying this in symbol-elf
given the comment says it is kernel specific - so we could keep the
is_kernel change.

> +                       if (!strchr(prev->name, '[') != !strchr(curr->name, '['))

I find this condition not to be intention revealing. On ARM there is
also an || for the condition reversed. When this is in an is_kernel
block then I think it is clear this is kernel hack, so I think it is
good to comment on what the condition is for.

> +                               prev->end = roundup(prev->end + 1, page_size);

Currently the roundup varies per architecture, but it is not clear to
me that it matters.

> +                       else

I think we should comment here that we're extending zero sized symbols
to the start of the next symbol.

> +                               prev->end = curr->start;
> +
> +                       pr_debug4("%s sym:%s end:%#" PRIx64 "\n",
> +                                 __func__, prev->name, prev->end);
> +               }

Thanks,
Ian

>         }
>
>         /* Last entry */