linux-debuggers.vger.kernel.org archive mirror
 help / color / mirror / Atom feed
From: Stephen Brennan <stephen.s.brennan@oracle.com>
To: Alan Maguire <alan.maguire@oracle.com>,
	Arnaldo Carvalho de Melo <acme@kernel.org>
Cc: dwarves@vger.kernel.org, linux-debuggers@vger.kernel.org
Subject: Re: [PATCH dwarves 2/4] dwarf_loader: add "artificial" and "top_level" variable flags
Date: Fri, 13 Sep 2024 10:16:50 -0700	[thread overview]
Message-ID: <87r09nmq2l.fsf@oracle.com> (raw)
In-Reply-To: <85af9dbb-4e96-4dac-a7d7-29c83f517030@oracle.com>

Alan Maguire <alan.maguire@oracle.com> writes:

> On 12/09/2024 20:08, Stephen Brennan wrote:
>> The "artificial" flag corresponds directly to DW_AT_artificial, which
>> indicates a compiler-generated variable (e.g. __func__) which shouldn't
>> be included in the output.
>>
>
> Nice, didn't know about this! Great to have such a clear criterion for
> filtering.

I've been finding llvm-dwarfdump invaluable for learning these things.
(Aside: thanks to Omar Sandoval to introducing it to me as I've tried to
wrap my head around some small parts of DWARF.)

I discovered this attribute when exploring those __func__ variables:

    $ llvm-dwarfdump vmlinux --name=__func__ | head -n 12
    vmlinux:        file format elf64-x86-64

    0x0001bfd3: DW_TAG_variable
                  DW_AT_name        ("__func__")
                  DW_AT_type        (0x0001c0f0 "const const char[18]")
                  DW_AT_artificial  (true)

    0x0001c148: DW_TAG_variable
                  DW_AT_name        ("__func__")
                  DW_AT_type        (0x0001c552 "const const char[12]")
                  DW_AT_artificial  (true)

Then it was just a quick ctrl-F through the specification to fill in the
gaps.

>> The "top_level" flag is intended to be a better proxy for global scoped
>> variables. Currently, the DWARF loader examines the DWARF location
>
> Looking at this, it appears that top_level means it is a top-level
> compilation unit tag, i.e. not associated with a subroutine tag (I
> think?). That's kind of implicit in your explanation so I think it would
> be helpful for the log to describe what it is as well as how you're
> using it.

Yeah, I'll spell it out more explicitly in this commit message. I do add
a comment in patch 4 which explains the differences between SCOPE_GLOBAL
and top_level (shown below) but something similar should be in the log
message for this change too.

    top_level: indicates that the variable is declared at the top
      level of the CU, and thus it is globally scoped.
    artificial: indicates that the variable is a compiler-generated
      "fake" variable that doesn't appear in the source.
    scope: set by pahole to indicate the type of storage the
      variable has. GLOBAL indicates it is stored in static
      memory (as opposed to a stack variable or register)
    
    Some variables are "top_level" but not GLOBAL:
      e.g. current_stack_pointer, which is a register variable,
      despite having global CU-declarations. We don't want that,
      since no code could actually find this variable.
    Some variables are GLOBAL but not top_level:
      e.g. function static variables

>> expression, and if the location is found to be at a constant memory
>> address (not stack, register, etc), then the variable is assumed to be
>> globally scoped. However, this includes a variety of variables that
>> aren't truly globally scoped: most commonly, static variables in
>> functions definitions. Their locations may be static, but they're not
>> globally accessible in any useful way.
>> 
>> These flags will be used by the BTF encoder to select global variables.
>> 
>> Signed-off-by: Stephen Brennan <stephen.s.brennan@oracle.com>
>
>
> Reviewed-by: Alan Maguire <alan.maguire@oracle.com>

Thanks!
Stephen

  reply	other threads:[~2024-09-13 17:16 UTC|newest]

Thread overview: 14+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2024-09-12 19:08 [PATCH dwarves 0/4] Emit global variables in BTF Stephen Brennan
2024-09-12 19:08 ` [PATCH dwarves 1/4] dutil: return ELF section name when looked up by index Stephen Brennan
2024-09-13 13:26   ` Alan Maguire
2024-09-13 17:06     ` Stephen Brennan
2024-09-12 19:08 ` [PATCH dwarves 2/4] dwarf_loader: add "artificial" and "top_level" variable flags Stephen Brennan
2024-09-13 13:35   ` Alan Maguire
2024-09-13 17:16     ` Stephen Brennan [this message]
2024-09-12 19:08 ` [PATCH dwarves 3/4] btf_encoder: cache all ELF section info Stephen Brennan
2024-09-13 15:25   ` Alan Maguire
2024-09-13 17:05     ` Stephen Brennan
2024-09-15 11:26       ` Alan Maguire
2024-09-12 19:08 ` [PATCH dwarves 4/4] btf_encoder: add global_var feature to encode globals Stephen Brennan
2024-09-15 11:49   ` Alan Maguire
2024-09-20  8:18     ` Stephen Brennan

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=87r09nmq2l.fsf@oracle.com \
    --to=stephen.s.brennan@oracle.com \
    --cc=acme@kernel.org \
    --cc=alan.maguire@oracle.com \
    --cc=dwarves@vger.kernel.org \
    --cc=linux-debuggers@vger.kernel.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).