public inbox for dtrace@lists.linux.dev
 help / color / mirror / Atom feed
From: Kris Van Hees <kris.van.hees@oracle.com>
To: Nick Alcock <nick.alcock@oracle.com>
Cc: dtrace@lists.linux.dev, dtrace-devel@oss.oracle.com
Subject: Re: [PATCH v3 0/6] usdt typed args, translators and arg remapping
Date: Thu, 31 Oct 2024 14:56:31 -0400	[thread overview]
Message-ID: <ZyPS35JNqCM9msZ6@oracle.com> (raw)
In-Reply-To: <20241030120850.256884-1-nick.alcock@oracle.com>

Can you please rebase on github's devel?  Eugene's USDT work is on there and
that impacts patches 5/6 and 6/6.

Also, 1/6 is already on devel.

Please also searc hfor any occurences of /remap/ and fix those to use the
more consistent "map" and "mapping" language.

On Wed, Oct 30, 2024 at 12:08:44PM +0000, Nick Alcock wrote:
> This series adds back support for the probe foo (...) : (...) syntax
> in USDT provider definitions.  The dtprobed side is routine, pulling
> in the DOF and adding it to a new set of dof_parser.h records fed
> through to the DOF stash and ultimately to dt_pid.c and eventually
> dt_prov_uprobe.c: the end result is that the DOF for a given probe's
> remappings ends up in the dt_uprobe_t in more or less the same form
> as it was in the DOF.
> 
> After that things get tricky, because we want to remap not only arg[N] but
> also argN references (the single existing test for USDT arg remapping
> verifies that this works), but the SDT machinery depends on argN being
> unremapped!  So we do the remapping in dt_prov_uprobe.c while reshuffling
> things into dt_argdesc_t's, and tell the higher layers that no arg remapping
> is ever being performed, then do the remapping in the trampoline by
> physically reshuffling the arguments.  This seems to work without breaking
> SDT on all supported architectures, and is surprisingly simple (at least, it
> surprised me: multiple places where I thought I'd need hundreds of lines of
> complex code turned out to need only three or four lines).
> 
> New tests are added to verify that USDT translators work when types change,
> when arg counts change, and when there are no args at all, and that dtrace
> -vln produces the right output (all previous tests for USDT -vln, translated
> or not, are currently disabled because they all use wildcards: this new one
> doesn't).
> 
> We also fix a couple of tiny error-related bugs encountered in the
> course of development, one in usdt, one related to the print action.
> 
> Changes since v2:
>   Populate dt_argdesc_t's in USDT probe discovery rather than in
>   probe_info.
> 
> Changes since v1:
>   Adapted to review comments (all comments but the stuff around
>   trampolines and arg mapping, which I don't understand well
>   enough to implement); in particular track xlated args and
>   mappings together, and move from a flags word in the
>   dof_parser struct's DIT_PROBE record to an arg count
> 
> Nick Alcock (6):
>   error: add missing EDT_PRINT entry
>   usdt: get arg types and xlations into DTrace from the DOF
>   dtprobed: stop skipping zero-tracepoint probes in dof_stash.c
>   cg: add argument mapping in the trampoline
>   usdt: typed args and arg mapping
>   usdt: fix create_underlying error path
> 
>  dtprobed/dof_stash.c                          |  21 +--
>  dtprobed/dtprobed.c                           |  10 +-
>  include/dtrace/pid.h                          |   7 +
>  libcommon/dof_parser.c                        | 150 ++++++++++++----
>  libcommon/dof_parser.h                        |  64 ++++++-
>  libdtrace/dt_cg.c                             |  36 +++-
>  libdtrace/dt_cg.h                             |   1 +
>  libdtrace/dt_error.c                          |   3 +-
>  libdtrace/dt_pid.c                            |  60 +++++++
>  libdtrace/dt_prov_uprobe.c                    | 168 +++++++++++++++++-
>  test/triggers/usdt-tst-argmap-prov.d          |   5 +-
>  test/triggers/usdt-tst-argmap.c               |   5 +-
>  .../dtrace-util/tst.ListProbesArgsUSDT.r      |  34 ++++
>  .../dtrace-util/tst.ListProbesArgsUSDT.r.p    |   2 +
>  .../dtrace-util/tst.ListProbesArgsUSDT.sh     |  83 +++++++++
>  test/unittest/usdt/err.argmap-null.d          |  40 +++++
>  test/unittest/usdt/err.argmap-null.r          |   2 +
>  test/unittest/usdt/err.argmap-null.r.p        |   2 +
>  test/unittest/usdt/tst.argmap-null.d          |  32 ++++
>  test/unittest/usdt/tst.argmap-typed-partial.d |  49 +++++
>  test/unittest/usdt/tst.argmap-typed.d         |  48 +++++
>  test/unittest/usdt/tst.argmap.d               |   3 +-
>  test/utils/showUSDT.c                         |   5 +-
>  23 files changed, 759 insertions(+), 71 deletions(-)
>  create mode 100644 test/unittest/dtrace-util/tst.ListProbesArgsUSDT.r
>  create mode 100755 test/unittest/dtrace-util/tst.ListProbesArgsUSDT.r.p
>  create mode 100755 test/unittest/dtrace-util/tst.ListProbesArgsUSDT.sh
>  create mode 100644 test/unittest/usdt/err.argmap-null.d
>  create mode 100644 test/unittest/usdt/err.argmap-null.r
>  create mode 100755 test/unittest/usdt/err.argmap-null.r.p
>  create mode 100644 test/unittest/usdt/tst.argmap-null.d
>  create mode 100644 test/unittest/usdt/tst.argmap-typed-partial.d
>  create mode 100644 test/unittest/usdt/tst.argmap-typed.d
> 
> 
> base-commit: 8a1264bf0e818c8624be250eb5174714b62ed93c
> -- 
> 2.46.0.278.g36e3a12567
> 

      parent reply	other threads:[~2024-10-31 18:56 UTC|newest]

Thread overview: 8+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2024-10-30 12:08 [PATCH v3 0/6] usdt typed args, translators and arg remapping Nick Alcock
2024-10-30 12:08 ` [PATCH REVIEWED v3 1/6] error: add missing EDT_PRINT entry Nick Alcock
2024-10-30 12:08 ` [PATCH v3 2/6] usdt: get arg types and xlations into DTrace from the DOF Nick Alcock
2024-10-30 12:08 ` [PATCH REVIEWED v3 3/6] dtprobed: stop skipping zero-tracepoint probes in dof_stash.c Nick Alcock
2024-10-30 12:08 ` [PATCH v3 4/6] cg: add argument mapping in the trampoline Nick Alcock
2024-10-30 12:08 ` [PATCH v3 5/6] usdt: typed args and arg mapping Nick Alcock
2024-10-30 12:08 ` [PATCH REVIEWED v3 6/6] usdt: fix create_underlying error path Nick Alcock
2024-10-31 18:56 ` Kris Van Hees [this message]

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=ZyPS35JNqCM9msZ6@oracle.com \
    --to=kris.van.hees@oracle.com \
    --cc=dtrace-devel@oss.oracle.com \
    --cc=dtrace@lists.linux.dev \
    --cc=nick.alcock@oracle.com \
    /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