[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]
Message-ID: <8ae4f9b0-1917-448d-a89b-11597549f89b@kernel.org>
Date: Sat, 6 Jan 2024 09:13:53 -0700
From: David Ahern <dsahern@...nel.org>
To: Breno Leitao <leitao@...ian.org>, "David S. Miller"
<davem@...emloft.net>, Eric Dumazet <edumazet@...gle.com>,
Jakub Kicinski <kuba@...nel.org>, Paolo Abeni <pabeni@...hat.com>,
Nathan Chancellor <nathan@...nel.org>,
Nick Desaulniers <ndesaulniers@...gle.com>, Bill Wendling
<morbo@...gle.com>, Justin Stitt <justinstitt@...gle.com>
Cc: leit@...a.com, "open list:NETWORKING [IPv4/IPv6]"
<netdev@...r.kernel.org>, open list <linux-kernel@...r.kernel.org>,
"open list:CLANG/LLVM BUILD SUPPORT:Keyword:b(?i:clang|llvm)b"
<llvm@...ts.linux.dev>, Wei Wang <weiwan@...gle.com>
Subject: Re: [PATCH net-next] net/ipv6: resolve warning in ip6_fib.c
On 1/5/24 10:39 AM, Breno Leitao wrote:
> In some configurations, the 'iter' variable in function
> fib6_repair_tree() is unused, resulting the following warning when
> compiled with W=1.
>
> net/ipv6/ip6_fib.c:1781:6: warning: variable 'iter' set but not used [-Wunused-but-set-variable]
> 1781 | int iter = 0;
> | ^
>
> Rewrite RT6_TRACE() macro to keep the variable in scope.
>
> Clang 18 removes the trace completely from the binary when the trace is
> disabled[1], so, no overhead is expected with this change.
>
> [1] Link: https://paste.debian.net/1303162/
>
> Signed-off-by: Breno Leitao <leitao@...ian.org>
> ---
> include/net/ip6_fib.h | 6 +-----
> 1 file changed, 1 insertion(+), 5 deletions(-)
>
> diff --git a/include/net/ip6_fib.h b/include/net/ip6_fib.h
> index 9ba6413fd2e3..d2ef60ebeb25 100644
> --- a/include/net/ip6_fib.h
> +++ b/include/net/ip6_fib.h
> @@ -30,11 +30,7 @@
>
> #define RT6_DEBUG 2
>
> -#if RT6_DEBUG >= 3
> -#define RT6_TRACE(x...) pr_debug(x)
> -#else
> -#define RT6_TRACE(x...) do { ; } while (0)
> -#endif
> +#define RT6_TRACE(x...) do { if (RT6_DEBUG > 3) pr_debug(x); } while (0)
>
> struct rt6_info;
> struct fib6_info;
I question the value of RT6_TRACE vs just using pr_debug; pr_debug has
zero cost until enabled and can be enabled by file or line. Not
requiring a kernel build is actual better.
[cc'ed Wei who added the macro]
Powered by blists - more mailing lists