[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]
Message-ID: <aEnSbBaFYgd4Gr9u@x1>
Date: Wed, 11 Jun 2025 16:01:00 -0300
From: Arnaldo Carvalho de Melo <acme@...nel.org>
To: Sebastian Andrzej Siewior <bigeasy@...utronix.de>
Cc: Namhyung Kim <namhyung@...nel.org>, linux-kernel@...r.kernel.org,
linux-perf-users@...r.kernel.org
Subject: Re: [BUG] perf tools: Build failure in v6.16-rc1
On Wed, Jun 11, 2025 at 05:06:15PM +0200, Sebastian Andrzej Siewior wrote:
> On 2025-06-11 11:55:23 [-0300], Arnaldo Carvalho de Melo wrote:
> > commit 8386dc356158fc50c55831c96b1248e01d112ebc
> > Author: Sebastian Andrzej Siewior <bigeasy@...utronix.de>
> > Date: Wed Jun 11 11:25:42 2025 +0200
> >
> > perf bench futex: Fix prctl include in musl libc
> >
> > Namhyung Kim reported:
> >
> > I've updated the perf-tools-next to v6.16-rc1 and found a build error
> > like below on alpine linux 3.18.
> >
> > In file included from bench/futex.c:6:
> > /usr/include/sys/prctl.h:88:8: error: redefinition of 'struct prctl_mm_map'
> > 88 | struct prctl_mm_map {
> > | ^~~~~~~~~~~~
> > In file included from bench/futex.c:5:
> > /linux/tools/include/uapi/linux/prctl.h:134:8: note: originally defined here
> > 134 | struct prctl_mm_map {
> > | ^~~~~~~~~~~~
> > make[4]: *** [/linux/tools/build/Makefile.build:86: /build/bench/futex.o] Error 1
> >
> > git bisect says it's the first commit introduced the failure.
> >
> > So your /usr/include/sys/prctl.h and
> > /linux/tools/include/uapi/linux/prctl.h both provide struct prctl_mm_map
> > but their include guard must be different.
> >
> > My /usr/include/sys/prctl.h is provided by glibc and contains the
> > prctl() declaration. It includes also linux/prctl.h. The
> > tools/include/uapi/linux/prctl.h is the same as
> > /usr/include/linux/prctl.h.
> >
> > The /usr/include/sys/prctl.h on alpine linux is different. This is
> > probably coming from musl. It contains the PR_* definition and the
> > prctl() declaration. So it clashes here because now the one struct is
> > available twice.
> >
> > The man page for prctl(2) says:
> >
> > | #include <linux/prctl.h> /* Definition of PR_* constants */
> > | #include <sys/prctl.h>
> >
> > so musl doesn't follow this.
> >
> > So align with the other builds.
> >
> > Signed-off-by: Sebastian Andrzej Siewior <bigeasy@...utronix.de>
> > Reported-by: Namhyung Kim <namhyung@...nel.org>
> > Link: https://lore.kernel.org/r/20250611092542.F4ooE2FL@linutronix.de
> s/Link/Closes/
ok
> > [ Remove one more in tools/perf/bench/futex-hash.c and conditionally define PR_FUTEX_HASH and friends ]
> > Signed-off-by: Arnaldo Carvalho de Melo <acme@...hat.com>
> >
> > diff --git a/tools/perf/bench/futex-hash.c b/tools/perf/bench/futex-hash.c
> > index fdf133c9520f73a4..d2d6d7f3ea331c84 100644
> > --- a/tools/perf/bench/futex-hash.c
> > +++ b/tools/perf/bench/futex-hash.c
> > @@ -18,7 +18,6 @@
> > #include <stdlib.h>
> > #include <linux/compiler.h>
> > #include <linux/kernel.h>
> > -#include <linux/prctl.h>
> > #include <linux/zalloc.h>
> > #include <sys/time.h>
> > #include <sys/mman.h>
> > diff --git a/tools/perf/bench/futex.c b/tools/perf/bench/futex.c
> > index 26382e4d8d4ce2ff..4c4fee107e5912d5 100644
> > --- a/tools/perf/bench/futex.c
> > +++ b/tools/perf/bench/futex.c
> > @@ -2,11 +2,18 @@
> > #include <err.h>
> > #include <stdio.h>
> > #include <stdlib.h>
> > -#include <linux/prctl.h>
> > #include <sys/prctl.h>
> >
> This is what I had locally and was waiting for confirmation.
>
> > #include "futex.h"
> >
> > +#ifndef PR_FUTEX_HASH
> > +#define PR_FUTEX_HASH 78
> > +# define PR_FUTEX_HASH_SET_SLOTS 1
> > +# define FH_FLAG_IMMUTABLE (1ULL << 0)
> > +# define PR_FUTEX_HASH_GET_SLOTS 2
> > +# define PR_FUTEX_HASH_GET_IMMUTABLE 3
> > +#endif // PR_FUTEX_HASH
>
> Is this needed? Aren't these defines coming from that local copy?
So, these are, as you say, in the copied linux/prctl.h, but in musl libc
we have:
/tmp/perf-6.16.0-rc1 $ grep 'struct prctl_mm_map {' /usr/include/linux/prctl.h
struct prctl_mm_map {
/tmp/perf-6.16.0-rc1 $ grep 'struct prctl_mm_map {' /usr/include/sys/prctl.h
struct prctl_mm_map {
/tmp/perf-6.16.0-rc1 $
And sys/prctl.h doesn't include linux/prctl.h, if we do it, we get
multiple definitions for 'struct prctl_mm_map'.
While in fedora (probably in all the others, haven't checked, but no
failure on them from my last container set build tests):
⬢ [acme@...lbx perf-tools]$ grep 'struct prctl_mm_map {' /usr/include/linux/prctl.h
struct prctl_mm_map {
⬢ [acme@...lbx perf-tools]$ grep 'struct prctl_mm_map {' /usr/include/sys/prctl.h
⬢ [acme@...lbx perf-tools]$
furthermore fedora's sys/prctl.h includes linux/prctl.h, while musl libc
doesn't.
I thought this would be something fixed in newer alpine versions, but
no:
toolsbuilder@...e:~$ grep FAIL dm.log.old/summary
5 19.53 alpine:3.16 : FAIL gcc version 11.2.1 20220219 (Alpine 11.2.1_git20220219)
6 20.83 alpine:3.17 : FAIL gcc version 12.2.1 20220924 (Alpine 12.2.1_git20220924-r4)
7 13.94 alpine:3.18 : FAIL gcc version 12.2.1 20220924 (Alpine 12.2.1_git20220924-r10)
8 16.60 alpine:3.19 : FAIL gcc version 13.2.1 20231014 (Alpine 13.2.1_git20231014)
9 15.72 alpine:3.20 : FAIL gcc version 13.2.1 20240309 (Alpine 13.2.1_git20240309)
10 16.38 alpine:3.22 : FAIL gcc version 14.2.0 (Alpine 14.2.0)
11 15.09 alpine:edge : FAIL gcc version 14.2.0 (Alpine 14.2.0)
toolsbuilder@...e:~$
So the easiest way out of this seems to be not to explicitely include
linux/prctl.h and define the new stuff conditionally, as I did, right?
- Arnaldo
> > void futex_set_nbuckets_param(struct bench_futex_parameters *params)
> > {
> > unsigned long flags;
>
> Sebastian
Powered by blists - more mailing lists