DPDK patches and discussions
 help / color / mirror / Atom feed
From: Jerin Jacob <jerinjacobk@gmail.com>
To: "Morten Brørup" <mb@smartsharesystems.com>
Cc: Jerin Jacob <jerinj@marvell.com>,
	Sunil Kumar Kori <skori@marvell.com>,
	dev@dpdk.org, techboard@dpdk.org,
	Ferruh Yigit <ferruh.yigit@amd.com>
Subject: Re: [PATCH v2] eal: add build-time option to omit trace
Date: Thu, 19 Sep 2024 19:24:15 +0530	[thread overview]
Message-ID: <CALBAE1M3mZYBH0-8B17Wfr5f85UZuv1mdaCwv47U+3D2ebPq_w@mail.gmail.com> (raw)
In-Reply-To: <98CBD80474FA8B44BF855DF32C47DC35E9F702@smartserver.smartshare.dk>

On Thu, Sep 19, 2024 at 6:45 PM Morten Brørup <mb@smartsharesystems.com> wrote:
>
> > From: Jerin Jacob [mailto:jerinjacobk@gmail.com]
> > Sent: Thursday, 19 September 2024 10.52
> >
> > On Thu, Sep 19, 2024 at 1:37 PM Morten Brørup <mb@smartsharesystems.com>
> > wrote:
> > >
> > > Some applications want to omit the trace feature.
> > > Either to reduce the memory footprint, to reduce the exposed attack
> > > surface, or for other reasons.
> > >
> > > This patch adds an option in rte_config.h to include or omit trace in the
> > > build. Trace is included by default.
> > >
> > > Omitting trace works by omitting all trace points.
> > > For API and ABI compatibility, the trace feature itself remains.
> > >
> > > Signed-off-by: Morten Brørup <mb@smartsharesystems.com>
> > > ---
> > > v2:
> > > * Added/modified macros required for building applications with trace
> > >   omitted.
> > > ---
> > >  app/test/test_trace.c                      | 11 ++++++++++-
> > >  config/rte_config.h                        |  1 +
> > >  lib/eal/include/rte_trace_point.h          | 22 +++++++++++++++++++++-
> > >  lib/eal/include/rte_trace_point_register.h | 17 +++++++++++++++++
> > >  4 files changed, 49 insertions(+), 2 deletions(-)
> > >
> > > diff --git a/app/test/test_trace.c b/app/test/test_trace.c
> > > index 00809f433b..7918cc865d 100644
> > > --- a/app/test/test_trace.c
> > > +++ b/app/test/test_trace.c
> > > @@ -12,7 +12,16 @@
> > >
> > >  int app_dpdk_test_tp_count;
> > >
> > > -#ifdef RTE_EXEC_ENV_WINDOWS
> > > +#if !defined(RTE_TRACE)
> > > +
> > > +static int
> > > +test_trace(void)
> > > +{
> > > +       printf("trace omitted at build-time, skipping test\n");
> > > +       return TEST_SKIPPED;
> > > +}
> > > +
> > > +#elif defined(RTE_EXEC_ENV_WINDOWS)
> > >
> > >  static int
> > >  test_trace(void)
> > > diff --git a/config/rte_config.h b/config/rte_config.h
> > > index dd7bb0d35b..fd6f8a2f1a 100644
> > > --- a/config/rte_config.h
> > > +++ b/config/rte_config.h
> > > @@ -49,6 +49,7 @@
> > >  #define RTE_MAX_TAILQ 32
> > >  #define RTE_LOG_DP_LEVEL RTE_LOG_INFO
> > >  #define RTE_MAX_VFIO_CONTAINERS 64
> > > +#define RTE_TRACE 1
> >
> >
> > Too many ifdefs down using this, IMO, This approach will make
> > difficult to manage the code and make sure
> >  RTE_TRACE 0 works always.
>
> I agree it's not pretty, but it's the best solution I can come up with, to meet the requirements driving this patch: reducing attack surface and memory footprint.
>
> >
> > I understand your concern about to reduce the expose attach surface.
> > BTW, it will be there with telemetry too.
>
> Yes, I have telemetry in mind too. :-)
> Got to work on them one by one, though.
>
> > Disabling full feature  for the code via config/rte_config.h is new to
> > DPDK. I think, we need wider opinion on this(Adding techboard).
>
> Yes, it’s a good topic for discussion in the techboard.
>
> Currently, there are three levels of build-time configuration:
> 1. Meson, for prominent options.
> 2. config/rte_config.h for less prominent options, e.g. enabling RTE_ASSERT(), mempool statistics, etc.
> 3. Header files in various modules/drivers, for tweaking the individual module.
>
> There didn't seem to be any interest in enabling/disabling trace via Meson, so I chose the next level of configuration.
>
> > Also, DPDK never concerned with binary size.
>
> That is a big mistake. DPDK is not only for cloud.
> DPDK should also be for CPEs, SOHO firewalls, and similar network appliances. For some of these high-volume network appliances, memory footprint matters.

Could you share data with a real-world application of the elf size?
1)Without any change
2)Only disabling via __rte_trace_point_emit_header_generic() .. aka below patch.
3)Full disable.

I think, size command will spit out with different section's size.
This data can be used to take decision and to know how much % it
adding up?


>
> Also, when running many instances of a service in the cloud, power consumption matters. Bloat increases power consumption.
>
> > Other than binary size,
> > following patch and exposed the config through meson
> > will fix the other problems. IMO, That's the correct tradeoff.
>
> Yes; the optimal solution would be making EAL less bloated, i.e. purely an abstraction layer to the underlying hardware and O/S.
> This would make trace an independent module, which could be enabled/disabled through Meson.
>
> Bruce has contributed a lot in this area, and already made many modules optional.
>
> >
> >
> > #define __rte_trace_point_emit_header_generic(t) \
> > void *mem; \
> > do { \
> >   +      if (RTE_TRACE == 0) \
> >   +             return \
> >         const uint64_t val = rte_atomic_load_explicit(t,
> > rte_memory_order_acquire); \
> >         if (likely(!(val & __RTE_TRACE_FIELD_ENABLE_MASK))) \
> >                 return; \
> >         mem = __rte_trace_mem_get(val); \
> >         if (unlikely(mem == NULL)) \
> >                 return; \
> >         mem = __rte_trace_point_emit_ev_header(mem, val); \
> > } while (0)
> >
> > >
> > >  /* bsd module defines */
> > >  #define RTE_CONTIGMEM_MAX_NUM_BUFS 64
> > > diff --git a/lib/eal/include/rte_trace_point.h
> > b/lib/eal/include/rte_trace_point.h
> > > index 41e2a7f99e..fd864125a7 100644
> > > --- a/lib/eal/include/rte_trace_point.h
> > > +++ b/lib/eal/include/rte_trace_point.h
> > > @@ -42,6 +42,8 @@ typedef RTE_ATOMIC(uint64_t) rte_trace_point_t;
> > >   */
> > >  #define RTE_TRACE_POINT_ARGS
> > >
> > > +#ifdef RTE_TRACE
> > > +
> > >  /** @internal Helper macro to support RTE_TRACE_POINT and
> > RTE_TRACE_POINT_FP */
> > >  #define __RTE_TRACE_POINT(_mode, _tp, _args, ...) \
> > >  extern rte_trace_point_t __##_tp; \
> > > @@ -100,6 +102,20 @@ _tp _args \
> > >  #define RTE_TRACE_POINT_FP(tp, args, ...) \
> > >         __RTE_TRACE_POINT(fp, tp, args, __VA_ARGS__)
> > >
> > > +#else
> > > +
> > > +/** @internal Helper macro to support RTE_TRACE_POINT and
> > RTE_TRACE_POINT_FP */
> > > +#define __RTE_TRACE_POINT_VOID(_tp, ...) \
> > > +static __rte_always_inline void \
> > > +_tp(...) \
> > > +{ \
> > > +}
> > > +
> > > +#define RTE_TRACE_POINT(tp, args, ...) __RTE_TRACE_POINT_VOID(tp, args)
> > > +#define RTE_TRACE_POINT_FP(tp, args, ...) __RTE_TRACE_POINT_VOID(tp, args)
> > > +
> > > +#endif /* RTE_TRACE */
> > > +
> > >  #ifdef __DOXYGEN__
> > >
> > >  /**
> > > @@ -212,6 +228,7 @@ bool rte_trace_point_is_enabled(rte_trace_point_t *tp);
> > >  __rte_experimental
> > >  rte_trace_point_t *rte_trace_point_lookup(const char *name);
> > >
> > > +#ifdef RTE_TRACE
> > >  /**
> > >   * @internal
> > >   *
> > > @@ -230,6 +247,7 @@ __rte_trace_point_fp_is_enabled(void)
> > >         return false;
> > >  #endif
> > >  }
> > > +#endif /* RTE_TRACE */
> > >
> > >  /**
> > >   * @internal
> > > @@ -356,6 +374,8 @@ __rte_trace_point_emit_ev_header(void *mem, uint64_t in)
> > >         return RTE_PTR_ADD(mem, __RTE_TRACE_EVENT_HEADER_SZ);
> > >  }
> > >
> > > +#ifdef RTE_TRACE
> > > +
> > >  #define __rte_trace_point_emit_header_generic(t) \
> > >  void *mem; \
> > >  do { \
> > > @@ -411,7 +431,7 @@ do { \
> > >         RTE_SET_USED(len); \
> > >  } while (0)
> > >
> > > -
> > > +#endif /* RTE_TRACE */
> > >  #endif /* ALLOW_EXPERIMENTAL_API */
> > >  #endif /* _RTE_TRACE_POINT_REGISTER_H_ */
> > >
> > > diff --git a/lib/eal/include/rte_trace_point_register.h
> > b/lib/eal/include/rte_trace_point_register.h
> > > index 41260e5964..567fce534e 100644
> > > --- a/lib/eal/include/rte_trace_point_register.h
> > > +++ b/lib/eal/include/rte_trace_point_register.h
> > > @@ -18,6 +18,8 @@ extern "C" {
> > >
> > >  RTE_DECLARE_PER_LCORE(volatile int, trace_point_sz);
> > >
> > > +#ifdef RTE_TRACE
> > > +
> > >  #define RTE_TRACE_POINT_REGISTER(trace, name) \
> > >  rte_trace_point_t __rte_section("__rte_trace_point") __##trace; \
> > >  static const char __##trace##_name[] = RTE_STR(name); \
> > > @@ -56,6 +58,21 @@ do { \
> > >                 RTE_STR(uint8_t)); \
> > >  } while (0)
> > >
> > > +#else
> > > +
> > > +#define RTE_TRACE_POINT_REGISTER(...)
> > > +#define __rte_trace_point_emit_header_generic(t) RTE_SET_USED(t)
> > > +#define __rte_trace_point_emit_header_fp(t) RTE_SET_USED(t)
> > > +#define __rte_trace_point_emit(in, type) RTE_SET_USED(in)
> > > +#define rte_trace_point_emit_string(in) RTE_SET_USED(in)
> > > +#define rte_trace_point_emit_blob(in, len) \
> > > +do { \
> > > +       RTE_SET_USED(in); \
> > > +       RTE_SET_USED(len); \
> > > +} while (0)
> > > +
> > > +#endif /* RTE_TRACE */
> > > +
> > >  #ifdef __cplusplus
> > >  }
> > >  #endif
> > > --
> > > 2.43.0
> > >

  reply	other threads:[~2024-09-19 13:54 UTC|newest]

Thread overview: 37+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2024-09-18  8:55 [PATCH] " Morten Brørup
2024-09-18  9:49 ` Jerin Jacob
2024-09-18 10:23   ` Morten Brørup
2024-09-19  8:06 ` [PATCH v2] " Morten Brørup
2024-09-19  8:51   ` Jerin Jacob
2024-09-19 13:15     ` Morten Brørup
2024-09-19 13:54       ` Jerin Jacob [this message]
2024-09-19 15:35         ` Morten Brørup
2024-09-19 15:49           ` Jerin Jacob
2024-09-19 16:08             ` Morten Brørup
2024-09-19 16:34               ` Jerin Jacob
2024-09-20  9:08 ` [PATCH v3] " Morten Brørup
2024-09-23  8:39   ` Jerin Jacob
2024-09-24 13:50     ` Morten Brørup
2024-09-24 13:39 ` [PATCH v4] " Morten Brørup
2024-09-24 15:30   ` Stephen Hemminger
2024-09-24 15:41   ` Jerin Jacob
2024-09-24 15:53     ` Morten Brørup
2024-09-24 15:57       ` Jerin Jacob
2024-10-01 13:49         ` Morten Brørup
2024-10-01 14:05           ` Jerin Jacob
2024-10-01 14:14             ` Morten Brørup
2024-10-01 15:01               ` Jerin Jacob
2024-10-01 16:01                 ` Morten Brørup
2024-10-01 16:06                   ` Jerin Jacob
2024-10-01 16:15                     ` Morten Brørup
2024-10-02 16:17                       ` Jerin Jacob
2024-10-06 12:58                         ` Morten Brørup
2024-10-06 12:38 ` [PATCH v5] " Morten Brørup
2024-10-06 13:58 ` [PATCH v6] " Morten Brørup
2024-10-07  5:45   ` Jerin Jacob
2024-10-07  6:07     ` Morten Brørup
2024-10-06 14:03 ` [PATCH v7] " Morten Brørup
2024-10-06 14:09   ` Morten Brørup
2024-10-07 11:46 ` [PATCH v9] " Morten Brørup
2024-10-08  7:16   ` Morten Brørup
2024-10-08 10:15   ` David Marchand

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=CALBAE1M3mZYBH0-8B17Wfr5f85UZuv1mdaCwv47U+3D2ebPq_w@mail.gmail.com \
    --to=jerinjacobk@gmail.com \
    --cc=dev@dpdk.org \
    --cc=ferruh.yigit@amd.com \
    --cc=jerinj@marvell.com \
    --cc=mb@smartsharesystems.com \
    --cc=skori@marvell.com \
    --cc=techboard@dpdk.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 NNTP newsgroup(s).