linux-kernel - Re: [PATCH 01/16] dyndbg: drop trim_prefix, obsoleted by __FILE_

lists.openwall.net		lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC
Open Source and information security mailing list archives

Hash Suite: Windows password security audit tool. GUI, reports in PDF.

[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]

Message-ID: <07db7036-b46f-c157-5737-e3d96c808f14@rasmusvillemoes.dk>
Date:   Wed, 30 Oct 2019 22:20:39 +0100
From:   Rasmus Villemoes <linux@...musvillemoes.dk>
To:     Jim Cromie <jim.cromie@...il.com>, jbaron@...mai.com,
        linux-kernel@...r.kernel.org
Cc:     greg@...ah.com, clang-built-linux@...glegroups.com,
        Jonathan Corbet <corbet@....net>,
        Randy Dunlap <rdunlap@...radead.org>, linux-doc@...r.kernel.org
Subject: Re: [PATCH 01/16] dyndbg: drop trim_prefix, obsoleted by __FILE__s
 relative path

On 29/10/2019 21.00, Jim Cromie wrote:
> Regarding:
> commit 2b6783191da7 ("dynamic_debug: add trim_prefix() to provide source-root relative paths")
> commit a73619a845d5 ("kbuild: use -fmacro-prefix-map to make __FILE__ a relative path")
> 
> 2nd commit broke dynamic-debug's "file $fullpath" query form, but
> nobody noticed because 1st commit trimmed prefixes from control-file
> output, so the click-copy-pasting of fullpaths into new queries had
> ceased; that query form became unused.
> 
> So remove the function and callers; its purpose was to strip the
> prefix from __FILE__, which is now gone.

I agree with the intent, but I wonder if this is premature. I mean, the
-fmacro-prefix-map is only for gcc 8+, so this ends up printing the full
paths when the compiler is just a little old (and the kernel was built
out-of-tree).

I don't think keeping the current workaround for a year or two more
should hurt; when int skip = strlen(__FILE__) -
strlen("lib/dynamic_debug.c"); evaluates to 0 (in-tree build, or build
with new enough gcc), I'm pretty sure gcc optimizes the rest of the
function away (it should know that strncmp(x, y, 0) gives 0, and even if
it didn't, the conditional assigns 0 to skip which it already is, so gcc
just needs to know that strncmp() is pure).

> 
> Also drop "file $fullpath" from docs, and tweak example to cite column
> 1 of control-file as valid "file $input".

That part certainly makes sense, since the $fullpath hasn't actually
been in the control file for a long time.

Rasmus