public inbox for libc-alpha@sourceware.org
 help / color / mirror / Atom feed
From: "H.J. Lu" <hjl.tools@gmail.com>
To: Adhemerval Zanella <adhemerval.zanella@linaro.org>
Cc: GNU C Library <libc-alpha@sourceware.org>,
	Florian Weimer <fweimer@redhat.com>
Subject: V5 [PATCH 2/2] ld.so: Add --list-tunables to print tunable values
Date: Thu, 14 Jan 2021 14:25:09 -0800	[thread overview]
Message-ID: <CAMe9rOrWshs-PWyeyjdvDU-NexOpvV2aOtifr35VprcyC1nA0w@mail.gmail.com> (raw)
In-Reply-To: <211802d5-3c89-b153-5854-4f72d3e60b39@linaro.org>

[-- Attachment #1: Type: text/plain, Size: 10718 bytes --]

On Thu, Jan 14, 2021 at 10:35 AM Adhemerval Zanella
<adhemerval.zanella@linaro.org> wrote:
>
>
>
> On 31/10/2020 12:44, H.J. Lu via Libc-alpha wrote:
> > Pass --list-tunables to ld.so to print tunables with min and max values.
> > ---
> >  NEWS                 |  2 ++
> >  elf/Makefile         |  8 ++++++++
> >  elf/dl-main.h        |  2 +-
> >  elf/dl-tunables.c    | 36 ++++++++++++++++++++++++++++++++++++
> >  elf/dl-tunables.h    |  2 ++
> >  elf/dl-usage.c       |  7 ++++++-
> >  elf/rtld.c           | 23 +++++++++++++++++++++++
> >  manual/tunables.texi | 37 +++++++++++++++++++++++++++++++++++++
> >  8 files changed, 115 insertions(+), 2 deletions(-)
> >
> > diff --git a/NEWS b/NEWS
> > index 4307c4b1b0..a62e7307ef 100644
> > --- a/NEWS
> > +++ b/NEWS
> > @@ -9,6 +9,8 @@ Version 2.33
> >
> >  Major new features:
> >
> > +* Pass --list-tunables to ld.so to print tunable values.
> > +
>
> Maybe use a similar sentence as for argv0:
>
>   * The dynamic linker accepts the --list-tunables argument which prints
>     all the supported tunables.  This option is disable is glibc is
>     configured with tunables disabled (--enable-tunables=no).

Fixed.

> >  * The dynamic linker accepts the --argv0 argument and provides opportunity
> >    to change argv[0] string.
> >
> > diff --git a/elf/Makefile b/elf/Makefile
> > index f10cc59e7c..86b282a32b 100644
> > --- a/elf/Makefile
> > +++ b/elf/Makefile
> > @@ -44,6 +44,10 @@ dl-routines += dl-tunables
> >  tunables-type = $(addprefix TUNABLES_FRONTEND_,$(have-tunables))
> >  CPPFLAGS-dl-tunables.c += -DTUNABLES_FRONTEND=$(tunables-type)
> >
> > +ifeq (yesyes,$(build-shared)$(run-built-tests))
> > +tests-special += $(objpfx)list-tunables.out
> > +endif
> > +
> >  # Make sure that the compiler does not insert any library calls in tunables
> >  # code paths.
>
> Ok, it is enabled iff 'ifneq (no,$(have-tunables))'.
>
> >  ifeq (yes,$(have-loop-to-function))
> > @@ -1812,3 +1816,7 @@ $(objpfx)argv0test.out: tst-rtld-argv0.sh $(objpfx)ld.so \
> >              '$(test-wrapper-env)' '$(run_program_env)' \
> >              '$(rpath-link)' 'test-argv0' > $@; \
> >      $(evaluate-test)
> > +
> > +$(objpfx)list-tunables.out: $(objpfx)ld.so
> > +     $(objpfx)ld.so --list-tunables > $@; \
> > +     $(evaluate-test)
>
> Maybe at least check for a tunable with a default value which be
> equal to all architecture, like glibc.rtld.nns, glibc.malloc.perturb, etc?

I added elf/tst-rtld-list-tunables.exp and elf/tst-rtld-list-tunables.sh
to compare against rtld and malloc tunables.  It passed on i686, x32
and x86-64.

> > diff --git a/elf/dl-main.h b/elf/dl-main.h
> > index b51256d3b4..f229867b8e 100644
> > --- a/elf/dl-main.h
> > +++ b/elf/dl-main.h
> > @@ -63,7 +63,7 @@ struct audit_list
> >  enum rtld_mode
> >    {
> >      rtld_mode_normal, rtld_mode_list, rtld_mode_verify, rtld_mode_trace,
> > -    rtld_mode_help,
> > +    rtld_mode_list_tunables, rtld_mode_help,
> >    };
> >
> >  /* Aggregated state information extracted from environment variables
>
> Ok.
>
> > diff --git a/elf/dl-tunables.c b/elf/dl-tunables.c
> > index 2ba2844075..048601b704 100644
> > --- a/elf/dl-tunables.c
> > +++ b/elf/dl-tunables.c
> > @@ -398,6 +398,42 @@ __tunables_init (char **envp)
> >      }
> >  }
> >
> > +void
> > +__tunables_print (void)
> > +{
> > +  for (int i = 0; i < sizeof (tunable_list) / sizeof (tunable_t); i++)
>
> Use array_length here.

Fixed.

> > +    {
> > +      tunable_t *cur = &tunable_list[i];
>
> Maybe use a const modifier here.

Fixed.

> > +      _dl_printf ("%s: ", cur->name);
> > +      switch (cur->type.type_code)
> > +     {
> > +     case TUNABLE_TYPE_INT_32:
> > +       _dl_printf ("%d (min: %d, max: %d)\n",
> > +                   (int) cur->val.numval,
> > +                   (int) cur->type.min,
> > +                   (int) cur->type.max);
> > +       break;
> > +     case TUNABLE_TYPE_UINT_64:
> > +       _dl_printf ("0x%lx (min: 0x%lx, max: 0x%lx)\n",
> > +                   (long int) cur->val.numval,
> > +                   (long int) cur->type.min,
> > +                   (long int) cur->type.max);
> > +       break;
> > +     case TUNABLE_TYPE_SIZE_T:
> > +       _dl_printf ("0x%Zx (min: 0x%Zx, max: 0x%Zx)\n",
> > +                   (size_t) cur->val.numval,
> > +                   (size_t) cur->type.min,
> > +                   (size_t) cur->type.max);
> > +       break;
> > +     case TUNABLE_TYPE_STRING:
> > +       _dl_printf ("%s\n", cur->val.strval ? cur->val.strval : "");
> > +       break;
> > +     default:
> > +       __builtin_unreachable ();
> > +     }
> > +    }
> > +}
> > +
>
> Ok, it should be ok to use '%d' for signed and 'lx' for unsigned.
>
> >  /* Set the tunable value.  This is called by the module that the tunable exists
> >     in. */
> >  void
> > diff --git a/elf/dl-tunables.h b/elf/dl-tunables.h
> > index 550b0cc7f4..4a0c90f3e0 100644
> > --- a/elf/dl-tunables.h
> > +++ b/elf/dl-tunables.h
> > @@ -69,9 +69,11 @@ typedef struct _tunable tunable_t;
> >  # include "dl-tunable-list.h"
> >
> >  extern void __tunables_init (char **);
> > +extern void __tunables_print (void);
> >  extern void __tunable_get_val (tunable_id_t, void *, tunable_callback_t);
> >  extern void __tunable_set_val (tunable_id_t, void *, void *, void *);
> >  rtld_hidden_proto (__tunables_init)
> > +rtld_hidden_proto (__tunables_print)
> >  rtld_hidden_proto (__tunable_get_val)
> >  rtld_hidden_proto (__tunable_set_val)
> >
>
> Ok.
>
> > diff --git a/elf/dl-usage.c b/elf/dl-usage.c
> > index 796ad38b43..3ce19fb892 100644
> > --- a/elf/dl-usage.c
> > +++ b/elf/dl-usage.c
> > @@ -190,7 +190,12 @@ setting environment variables (which would be inherited by subprocesses).\n\
> >                          in LIST\n\
> >    --audit LIST          use objects named in LIST as auditors\n\
> >    --preload LIST        preload objects named in LIST\n\
> > -  --argv0 STRING        set argv[0] to STRING before running\n\
> > +  --argv0 STRING        set argv[0] to STRING before running\n"
> > +#if HAVE_TUNABLES
> > +"\
> > +  --list-tunables       list all tunables with minimum and maximum values\n"
> > +#endif
> > +"\
> >    --help                display this help and exit\n\
> >    --version             output version information and exit\n\
> >  \n\
>
> Ok.
>
> > diff --git a/elf/rtld.c b/elf/rtld.c
> > index 5d117d0d2c..33993a6f8b 100644
> > --- a/elf/rtld.c
> > +++ b/elf/rtld.c
> > @@ -49,6 +49,10 @@
> >  #include <libc-early-init.h>
> >  #include <dl-main.h>
> >
> > +#if HAVE_TUNABLES
> > +# include <dl-tunables.h>
> > +#endif
> > +
>
> You can include this header regadless of HAVE_TUNABLES (it handles it
> internally).

Fixed.

> >  #include <assert.h>
> >
> >  /* Only enables rtld profiling for architectures which provides non generic
> > @@ -1248,6 +1252,16 @@ dl_main (const ElfW(Phdr) *phdr,
> >           _dl_argc -= 2;
> >           _dl_argv += 2;
> >         }
> > +#if HAVE_TUNABLES
> > +     else if (! strcmp (_dl_argv[1], "--list-tunables"))
> > +       {
> > +         state.mode = rtld_mode_list_tunables;
> > +
> > +         ++_dl_skip_args;
> > +         --_dl_argc;
> > +         ++_dl_argv;
> > +       }
> > +#endif
> >       else if (strcmp (_dl_argv[1], "--help") == 0)
> >         {
> >           state.mode = rtld_mode_help;
> > @@ -1268,6 +1282,15 @@ dl_main (const ElfW(Phdr) *phdr,
> >       else
> >         break;
> >
> > +#if HAVE_TUNABLES
> > +      if (__builtin_expect (state.mode, rtld_mode_normal)
> > +       == rtld_mode_list_tunables)
>
> Use
>
>   if (__glibc_unlikely (state.mode == rtld_mode_list_tunables)

Fixed.

> > +     {
> > +       __tunables_print ();
> > +       _exit (0);
> > +     }
> > +#endif
> > +
> >        /* If we have no further argument the program was called incorrectly.
> >        Grant the user some education.  */
> >        if (_dl_argc < 2)
> > diff --git a/manual/tunables.texi b/manual/tunables.texi
> > index d72d7a5ec0..924dac8876 100644
> > --- a/manual/tunables.texi
> > +++ b/manual/tunables.texi
> > @@ -28,6 +28,43 @@ Finally, the set of tunables available may vary between distributions as
> >  the tunables feature allows distributions to add their own tunables under
> >  their own namespace.
> >
> > +Passing @option{--list-tunables} to the dynamic loader to print all
> > +tunables with minimum and maximum values:
> > +
> > +@example
> > +$ /lib64/ld-linux-x86-64.so.2 --list-tunables
> > +glibc.rtld.nns: 0x4 (min: 0x1, max: 0x10)
> > +glibc.elision.skip_lock_after_retries: 3 (min: -2147483648, max: 2147483647)
> > +glibc.malloc.trim_threshold: 0x0 (min: 0x0, max: 0xffffffffffffffff)
> > +glibc.malloc.perturb: 0 (min: 0, max: 255)
> > +glibc.cpu.x86_shared_cache_size: 0x100000 (min: 0x0, max: 0xffffffffffffffff)
> > +glibc.elision.tries: 3 (min: -2147483648, max: 2147483647)
> > +glibc.elision.enable: 0 (min: 0, max: 1)
> > +glibc.cpu.x86_rep_movsb_threshold: 0x800 (min: 0x100, max: 0xffffffffffffffff)
> > +glibc.malloc.mxfast: 0x0 (min: 0x0, max: 0xffffffffffffffff)
> > +glibc.elision.skip_lock_busy: 3 (min: -2147483648, max: 2147483647)
> > +glibc.malloc.top_pad: 0x0 (min: 0x0, max: 0xffffffffffffffff)
> > +glibc.cpu.x86_rep_stosb_threshold: 0x800 (min: 0x1, max: 0xffffffffffffffff)
> > +glibc.cpu.x86_non_temporal_threshold: 0xc0000 (min: 0x0, max: 0xffffffffffffffff)
> > +glibc.cpu.x86_shstk:
>
> Trailing whitespace.

Fixed.

> > +glibc.cpu.hwcap_mask: 0x6 (min: 0x0, max: 0xffffffffffffffff)
> > +glibc.malloc.mmap_max: 0 (min: -2147483648, max: 2147483647)
> > +glibc.elision.skip_trylock_internal_abort: 3 (min: -2147483648, max: 2147483647)
> > +glibc.malloc.tcache_unsorted_limit: 0x0 (min: 0x0, max: 0xffffffffffffffff)
> > +glibc.cpu.x86_ibt:
>
> Ditto.

Fixed.

> > +glibc.cpu.hwcaps:
>
> Ditto.

Fixed.

> > +glibc.elision.skip_lock_internal_abort: 3 (min: -2147483648, max: 2147483647)
> > +glibc.malloc.arena_max: 0x0 (min: 0x1, max: 0xffffffffffffffff)
> > +glibc.malloc.mmap_threshold: 0x0 (min: 0x0, max: 0xffffffffffffffff)
> > +glibc.cpu.x86_data_cache_size: 0x8000 (min: 0x0, max: 0xffffffffffffffff)
> > +glibc.malloc.tcache_count: 0x0 (min: 0x0, max: 0xffffffffffffffff)
> > +glibc.malloc.arena_test: 0x0 (min: 0x1, max: 0xffffffffffffffff)
> > +glibc.pthread.mutex_spin_count: 100 (min: 0, max: 32767)
> > +glibc.rtld.optional_static_tls: 0x200 (min: 0x0, max: 0xffffffffffffffff)
> > +glibc.malloc.tcache_max: 0x0 (min: 0x0, max: 0xffffffffffffffff)
> > +glibc.malloc.check: 0 (min: 0, max: 3)
> > +@end example
> > +
> >  @menu
> >  * Tunable names::  The structure of a tunable name
> >  * Memory Allocation Tunables::  Tunables in the memory allocation subsystem
> >

Here is the updated patch.   OK for master?

Thanks.

-- 
H.J.

[-- Attachment #2: 0001-ld.so-Add-list-tunables-to-print-tunable-values.patch --]
[-- Type: application/x-patch, Size: 11587 bytes --]

  reply	other threads:[~2021-01-14 22:25 UTC|newest]

Thread overview: 12+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2020-10-31 15:44 V4 [PATCH 0/2] " H.J. Lu
2020-10-31 15:44 ` V4 [PATCH 1/2] x86: Move x86 processor cache info to cpu_features H.J. Lu
2021-01-14 14:13   ` Adhemerval Zanella
2021-01-14 19:28     ` V5 " H.J. Lu
2020-10-31 15:44 ` V4 [PATCH 2/2] ld.so: Add --list-tunables to print tunable values H.J. Lu
2021-01-14 18:35   ` Adhemerval Zanella
2021-01-14 22:25     ` H.J. Lu [this message]
2021-01-15 12:47       ` V5 " Adhemerval Zanella
2021-02-02 10:59       ` Andreas Schwab
2021-02-02 17:36         ` [PATCH] ld.so: Unset glibc tunables for --list-tunables test H.J. Lu
2021-02-02 18:07           ` Andreas Schwab
2021-02-02 18:12             ` [PATCH] tst-rtld-list-tunables.sh: Unset glibc tunables H.J. Lu

Reply instructions:

You may reply publicly to this message via plain-text email
using any one of the following methods:

* Save the following mbox file, import it into your mail client,
  and reply-to-all from there: mbox

  Avoid top-posting and favor interleaved quoting:
  https://en.wikipedia.org/wiki/Posting_style#Interleaved_style

* Reply using the --to, --cc, and --in-reply-to
  switches of git-send-email(1):

  git send-email \
    --in-reply-to=CAMe9rOrWshs-PWyeyjdvDU-NexOpvV2aOtifr35VprcyC1nA0w@mail.gmail.com \
    --to=hjl.tools@gmail.com \
    --cc=adhemerval.zanella@linaro.org \
    --cc=fweimer@redhat.com \
    --cc=libc-alpha@sourceware.org \
    /path/to/YOUR_REPLY

  https://kernel.org/pub/software/scm/git/docs/git-send-email.html

* If your mail client supports setting the In-Reply-To header
  via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for read-only IMAP folder(s) and NNTP newsgroup(s).