linux-kernel - Re: [PATCH] perf build: fix in-tree build

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] [day] [month] [year] [list]

Message-ID: <Z5kiciNxDk-G-Zhr@google.com>
Date: Tue, 28 Jan 2025 10:31:14 -0800
From: Namhyung Kim <namhyung@...nel.org>
To: Luca Ceresoli <luca.ceresoli@...tlin.com>
Cc: Peter Zijlstra <peterz@...radead.org>, Ingo Molnar <mingo@...hat.com>,
	Arnaldo Carvalho de Melo <acme@...nel.org>,
	Mark Rutland <mark.rutland@....com>,
	Alexander Shishkin <alexander.shishkin@...ux.intel.com>,
	Jiri Olsa <jolsa@...nel.org>, Ian Rogers <irogers@...gle.com>,
	Adrian Hunter <adrian.hunter@...el.com>,
	Andi Kleen <ak@...ux.intel.com>,
	Thomas Petazzoni <thomas.petazzoni@...tlin.com>,
	Alexis Lothoré <alexis.lothore@...tlin.com>,
	Arnaldo Carvalho de Melo <acme@...hat.com>,
	linux-perf-users@...r.kernel.org, linux-kernel@...r.kernel.org
Subject: Re: [PATCH] perf build: fix in-tree build

On Mon, Jan 27, 2025 at 02:48:49PM +0100, Luca Ceresoli wrote:
> Hello Namhyung Kim, 
> 
> thanks for having a look!
> 
> On Sun, 26 Jan 2025 13:00:10 -0800
> Namhyung Kim <namhyung@...nel.org> wrote:
> 
> > Hello,
> > 
> > On Fri, Jan 24, 2025 at 02:06:08PM +0100, Luca Ceresoli wrote:
> > > Building perf in-tree is broken after commit 890a1961c812 ("perf tools:
> > > Create source symlink in perf object dir") which added a 'source' symlink
> > > in the output dir pointing to the source dir.  
> > 
> > I cannot reproduce it - both `make -C tools/perf` and `cd tools/perf; make`
> > work well for me.  What do you mean by in-tree build exactly?  Can you
> > please share your command line and the error messages?
> 
> I have narrowed down my reproducer script from the initial report to a
> fairly minimal one, and here it is:
> 
> ------------------------8<------------------------
> #!/bin/sh
> 
> set -eu
> 
> OUT_DIR=${1:-$(pwd)/tools/perf}
> DESTDIR=/tmp/aaa
> 
> RET=0
> 
> echo "Kernel version: $(git describe)"
> echo "OUT_DIR = ${OUT_DIR}"
> echo "DESTDIR = ${DESTDIR}"
> echo
> 
> git clean -xdfq
> rm -fr ${DESTDIR}/
> 
> # Only for out of tree builds: clear the build dir
> if ! echo ${OUT_DIR} | grep -q "tools/perf"; then
>     rm -fr ${OUT_DIR}
>     mkdir -p ${OUT_DIR}
> fi
> 
> LINUX_MAKE_FLAGS="\
>     -C tools/perf \
>     O=${OUT_DIR} \
>     DESTDIR=${DESTDIR} \
> "
> 
> make ${LINUX_MAKE_FLAGS}
> make ${LINUX_MAKE_FLAGS} install
> 
> echo
>     
> if [ -h ${OUT_DIR}/perf ]; then
>     echo "*** ERROR: ${OUT_DIR}/perf is a symlink, should be an exectuable file: ***" >&2
>     ls -l ${OUT_DIR}/perf
>     RET=255
> fi
> 
> if ! find ${DESTDIR}/ -name perf -not -type d | xargs file | grep 'ELF.*executable'; then
>     echo "*** ERROR: perf executable not preseint in install dir ${DESTDIR} ***" >&2
>     RET=255
> fi
> 
> exit $RET
> ------------------------8<------------------------
> 
> Just put it outside of your kernel dir (or it will be removed by the
> 'git clean' command) and run it in your kernel source dir. E.g. right
> now I'm doing:
> 
> cd <kernel dir>
> git checkout v6.13
> ../build-perf
> 
> You should see it failing as:
> 
> *** ERROR: /.../linux/tools/perf/perf is a symlink, should be an exectuable file: ***
> lrwxrwxrwx 1 user user 39 Jan 27 14:10 /.../linux/tools/perf/perf -> /.../linux/tools/perf
> *** ERROR: perf executable not preseint in install dir /tmp/aaa ***
> 
> Note that in the broken case the kernel build continues and returns 0,
> but there is no perf exectuable.

Thanks for sharing this.  I've reproduced it and found it had a
symlink to a directory rather than an executable.

> 
> PS:
> 
> And after having come up with the above script, I also found that the
> build succeeds with the following change:
> 
> ------------------------8<------------------------
> @@ -24,11 +24,10 @@
>  LINUX_MAKE_FLAGS="\
>      -C tools/perf \
>      O=${OUT_DIR} \
> -    DESTDIR=${DESTDIR} \
>  "
>  
>  make ${LINUX_MAKE_FLAGS}
> -make ${LINUX_MAKE_FLAGS} install
> +make ${LINUX_MAKE_FLAGS} DESTDIR=${DESTDIR} install
> ------------------------8<------------------------
> 
> In other words:
> 
>  * This fails:
>      make ${LINUX_MAKE_FLAGS} DESTDIR=${DESTDIR}
>      make ${LINUX_MAKE_FLAGS} DESTDIR=${DESTDIR} install
>  * This succeeds:
>      make ${LINUX_MAKE_FLAGS}
>      make ${LINUX_MAKE_FLAGS} DESTDIR=${DESTDIR} install
> 
> The only difference is the presence of 'DESTDIR' in the compile command.
> 
> My understanding and expectation is that DESTDIR has no effect on the
> make process except in the 'install' target. This is clearly not
> happening here.
> 
> I have not yet found an obvious reason why 'DESTDIR' is special in the
> compilation stage, after a quick look.

Interesting, maybe some install commands were called internally.  I'll
take a look.

Thanks,
Namhyung