linux-perf-users.vger.kernel.org archive mirror
 help / color / mirror / Atom feed
From: Luca Ceresoli <luca.ceresoli@bootlin.com>
To: Andi Kleen <ak@linux.intel.com>
Cc: namhyung@kernel.org, linux-perf-users@vger.kernel.org,
	"Yann E. MORIN" <yann.morin.1998@free.fr>,
	"buildroot@buildroot.org" <buildroot@buildroot.org>,
	Thomas Petazzoni <thomas.petazzoni@bootlin.com>
Subject: Re: [PATCH v9 1/4] Create source symlink in perf object dir
Date: Mon, 25 Nov 2024 18:25:06 +0100	[thread overview]
Message-ID: <20241125182506.38af9907@booty> (raw)
In-Reply-To: <20240807231823.898979-1-ak@linux.intel.com>

Hello Andi,

+cc Yann and the Buildroot mailing list (see the end of this report).

On Wed,  7 Aug 2024 16:18:20 -0700
Andi Kleen <ak@linux.intel.com> wrote:

> Create a source symlink to the original source in the objdir.
> This is similar to what the main kernel build script does.
> 
> Signed-off-by: Andi Kleen <ak@linux.intel.com>
> ---
>  tools/perf/Makefile.perf | 2 ++
>  1 file changed, 2 insertions(+)
> 
> diff --git a/tools/perf/Makefile.perf b/tools/perf/Makefile.perf
> index 175e4c7898f0..d46892d8223b 100644
> --- a/tools/perf/Makefile.perf
> +++ b/tools/perf/Makefile.perf
> @@ -163,6 +163,8 @@ ifneq ($(OUTPUT),)
>  # for flex/bison parsers.
>  VPATH += $(OUTPUT)
>  export VPATH
> +# create symlink to the original source
> +SOURCE := $(shell ln -sf $(srctree)/tools/perf $(OUTPUT)/source)

I found that since this patch got applied in 890a1961c812 (v6.12-rc1)
building perf fails silently in some cases.

The cases I have tested:
 - make <FLAGS> -C tools/perf                       -> OK
 - make <FLAGS> -C tools/perf O=/my/out/of/tree/dir -> OK
 - make <FLAGS> -C tools/perf O=tools/perf          -> fails

The failure in the third case is silent: make exits with a return value
of 0, but there is no perf executable created.

Even 'make <FLAGS> -C tools/perf install' exits returning 0, but no
executable is installed. However the following lines are output during
'make install', which are a sign of this issue happening:

  install: omitting directory '/home/murray/w/bootlin/linux/tools/perf/perf'
  ln: failed to access '/tmp/aaa/usr/bin/perf': No such file or directory

I realize 'O=tools/perf' is "useless" in the sense that the same result
would be obtained by not setting 'O' at all. However I also believe it
is expected that O=tools/perf should work equally to not setting 'O',
and as such there is a regression.

Also, right now the Buildroot embedded Linux build system builds perf
by always passing O=$(LINUX_DIR)/tools/perf/, thus it is currently
broken for kernels 6.12-rc1 and later.

Below is the test script I wrote to automate my testing. It is a bit
more complex than expected because it needs to hide a different bug,
which would otherwise mask the one being reported here. A slightly
different version of this script was used to automate 'git bisect'.

-------------------------8<-------------------------

#!/bin/sh

# Usage:
#   build-perf-report          Build with O=tools/perf
#   build-perf-report <DIR>    Build with O=<DIR>

set -eu

# Default to O=tools/perf
OUT_DIR=${1:-tools/perf}

echo "Output dir: ${OUT_DIR}"

# This is the toolchain I initially used to reproduce the issue:
# https://toolchains.bootlin.com/downloads/releases/toolchains/aarch64/tarballs/aarch64--glibc--stable-2024.02-1.tar.bz2
#
# However commenting the two lines for a native build is equally
# reproducing the problem on my x86_64 Ubuntu 22.04 host
export ARCH=arm64
export CROSS_COMPILE="ccache ${HOME}/x-tools/aarch64--glibc--stable-2024.02-1/bin/aarch64-linux-"

LINUX_MAKE_FLAGS="\
    GIT_DIR=. \
    KCFLAGS=-Wno-attribute-alias \
    WERROR=0 \
    REGENERATE_PARSERS=1 \
    DESTDIR=/tmp/aaa \
    prefix=/usr \
    NO_GTK2=1 \
    NO_LIBPERL=1 \
    NO_LIBPYTHON=1 \
    NO_LIBBIONIC=1 \
    NO_LIBTRACEEVENT=1 \
    NO_NEWT=1 \
    NO_SLANG=1 \
    NO_LIBAUDIT=1 \
    NO_LIBUNWIND=1 \
    NO_LIBNUMA=1 \
    NO_LIBELF=1 \
    NO_DWARF=1 \
    NO_DEMANGLE=1 \
    NO_LIBCRYPTO=1 \
    NO_ZLIB=1 \
    NO_LZMA=1 \
    -C tools/perf O=${OUT_DIR}
"

echo "Kernel version: $(git describe)"

# Revert (without committing) 13d675aea6cac5bb4619ef9e3fdd90129fe6d139
# which introduces a different build failure (fixed by a later commit)
# masking the issue being examined
git checkout -- .
git show 13d675aea6cac5bb4619ef9e3fdd90129fe6d139 | git apply || :

# Clean the DESTDIR
rm -fr /tmp/aaa/

git clean -xdfq
make defconfig

# If building out of tree, create an empty build dir
if [ ${OUT_DIR} != "tools/perf" ]; then
    rm -fr ${OUT_DIR}
    mkdir -p ${OUT_DIR}
fi

make ${LINUX_MAKE_FLAGS}
make ${LINUX_MAKE_FLAGS} install

# remove the fix added above
git checkout -- .

echo "-------------------------------------------------------"
echo "perf files in O (${OUT_DIR}):"
find ${OUT_DIR} -name perf -not -type d  | xargs file
echo "-------------------------------------------------------"
echo "perf executable in DESTDIR:"
find /tmp/aaa/ -name perf -not -type d | xargs file | grep 'executable'
echo "-------------------------------------------------------"

-------------------------8<-------------------------

Best regards,
Luca

-- 
Luca Ceresoli, Bootlin
Embedded Linux and Kernel engineering
https://bootlin.com

  parent reply	other threads:[~2024-11-25 17:25 UTC|newest]

Thread overview: 13+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2024-08-07 23:18 [PATCH v9 1/4] Create source symlink in perf object dir Andi Kleen
2024-08-07 23:18 ` [PATCH v9 2/4] perf test: Support external tests for separate objdir Andi Kleen
2024-08-07 23:18 ` [PATCH v9 3/4] perf script: Fix perf script -F +metric Andi Kleen
2024-08-08  5:18   ` Namhyung Kim
2024-08-08 18:09     ` Ian Rogers
2024-08-08 20:18       ` Andi Kleen
2024-08-09 15:00         ` Ian Rogers
2024-08-07 23:18 ` [PATCH v9 4/4] Add a test case for " Andi Kleen
2024-11-25 17:25 ` Luca Ceresoli [this message]
2024-11-25 19:12   ` [PATCH v9 1/4] Create source symlink in perf object dir Andi Kleen
2024-11-26  7:34     ` Luca Ceresoli
2025-04-09 17:43 ` Florian Fainelli
2025-04-09 17:59   ` Florian Fainelli

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=20241125182506.38af9907@booty \
    --to=luca.ceresoli@bootlin.com \
    --cc=ak@linux.intel.com \
    --cc=buildroot@buildroot.org \
    --cc=linux-perf-users@vger.kernel.org \
    --cc=namhyung@kernel.org \
    --cc=thomas.petazzoni@bootlin.com \
    --cc=yann.morin.1998@free.fr \
    /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).