From: Lucas De Marchi <lucas.de.marchi@gmail.com>
To: Eugene Syromiatnikov <esyr@redhat.com>
Cc: linux-modules <linux-modules@vger.kernel.org>,
Yauheni Kaliuta <yauheni.kaliuta@redhat.com>
Subject: Re: [PATCH] libkmod: fix uninitialized variable usage warnings
Date: Tue, 7 Nov 2017 08:45:56 -0800 [thread overview]
Message-ID: <CAKi4VA+8-a1s3comKHUqq1kRdX+4ri3j+kb6xbwgCtXed90wPg@mail.gmail.com> (raw)
In-Reply-To: <20171107115938.GA19236@asgard.redhat.com>
On Tue, Nov 7, 2017 at 3:59 AM, Eugene Syromiatnikov <esyr@redhat.com> wrote:
> On Mon, Nov 06, 2017 at 07:24:55PM -0800, Lucas De Marchi wrote:
>> On Mon, Nov 6, 2017 at 1:57 PM, Eugene Syromiatnikov <esyr@redhat.com> wrote:
>> > There are two places where _cleanup_free_ variables are not initialised
>> > by the time function exits that have been caught by gcc:
>> >
>> > In file included from libkmod/libkmod.c:35:0:
>> > libkmod/libkmod.c: In function 'kmod_lookup_alias_is_builtin':
>> > ./shared/util.h:73:13: warning: 'line' may be used uninitialized in this function [-Wmaybe-uninitialized]
>> > free(*(void**) p);
>> > ^
>> > libkmod/libkmod.c:551:23: note: 'line' was declared here
>> > _cleanup_free_ char *line;
>> > ^
>> > In file included from libkmod/libkmod-module.c:42:0:
>> > libkmod/libkmod-module.c: In function 'kmod_module_probe_insert_module':
>> > ./shared/util.h:73:13: warning: 'cmd' may be used uninitialized in this function [-Wmaybe-uninitialized]
>> > free(*(void**) p);
>> > ^
>> > libkmod/libkmod-module.c:996:23: note: 'cmd' was declared here
>> > _cleanup_free_ char *cmd;
>> > ^
>> >
>> > This patch initializes them to NULL so free become no-op in these cases.
>> > ---
>> > libkmod/libkmod-module.c | 2 +-
>> > libkmod/libkmod.c | 2 +-
>> > 2 files changed, 2 insertions(+), 2 deletions(-)
>> >
>> > diff --git a/libkmod/libkmod-module.c b/libkmod/libkmod-module.c
>> > index 0a3ef11..6f23c1a 100644
>> > --- a/libkmod/libkmod-module.c
>> > +++ b/libkmod/libkmod-module.c
>> > @@ -995,7 +995,7 @@ static int module_do_install_commands(struct kmod_module *mod,
>> > {
>> > const char *command = kmod_module_get_install_commands(mod);
>> > char *p;
>> > - _cleanup_free_ char *cmd;
>> > + _cleanup_free_ char *cmd = NULL;
>> > int err;
>> > size_t cmdlen, options_len, varlen;
>> >
>> > diff --git a/libkmod/libkmod.c b/libkmod/libkmod.c
>> > index 69fe431..964772d 100644
>> > --- a/libkmod/libkmod.c
>> > +++ b/libkmod/libkmod.c
>> > @@ -556,7 +556,7 @@ finish:
>> >
>> > bool kmod_lookup_alias_is_builtin(struct kmod_ctx *ctx, const char *name)
>> > {
>> > - _cleanup_free_ char *line;
>> > + _cleanup_free_ char *line = NULL;
>>
>> This seems to be a bogus warning. See the line just below... there's
>> no way to exit this function without first assigning 'line'.
>
> In this instance, I initially thought that gcc is smart and derives
> "uninitialised variable" from the lookup_builtin_file call, but nope, it
> was happy only after line was initialised on declaration. On the other
> hand, documentation[1] mentions that cleanup function is also called
> during stack unwinding in case -fexceptions is enabled, and RHEL's RPM
> package indeed has this flag, at least, so this GCC's complaint still
> seems relevant.
This is likely to trigger "double assignment" warnings on other compilers.
A good question would be.... why is -fexceptions being used in a C-only library?
If it was a binary or a library that links to a C++-library, I would
understand, but in
this case it seems nonsense. /me lost
An acceptable solution would be to just merge the 2 lines so we
initialize line on
declaration.
I added -fexceptions here to reproduce the issue, but it didn't
reproduce using gcc 7.2.1
on Fedora 27. I also remember not having the issue with gcc 6. So I
still think it's
a compiler bug, although I don't mind if the solution proposed above
is applied (but mention
on commit message the exact compiler version).
Lucas De Marchi
prev parent reply other threads:[~2017-11-07 16:45 UTC|newest]
Thread overview: 4+ messages / expand[flat|nested] mbox.gz Atom feed top
2017-11-06 21:57 [PATCH] libkmod: fix uninitialized variable usage warnings Eugene Syromiatnikov
2017-11-07 3:24 ` Lucas De Marchi
2017-11-07 11:59 ` Eugene Syromiatnikov
2017-11-07 16:45 ` Lucas De Marchi [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=CAKi4VA+8-a1s3comKHUqq1kRdX+4ri3j+kb6xbwgCtXed90wPg@mail.gmail.com \
--to=lucas.de.marchi@gmail.com \
--cc=esyr@redhat.com \
--cc=linux-modules@vger.kernel.org \
--cc=yauheni.kaliuta@redhat.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;
as well as URLs for NNTP newsgroup(s).