 |
The problem occurs in mmap_flags.sh because $arch_mman is not exported, so in these statements : egrep -q $regex ${arch_mman} && \ (egrep $regex ${arch_mman} | \ $arch_mman is set in the parent shell, so the first egrep succeeds, but because arch_mman is not exported, it is not set in the sub-process , so the second egrep hangs because ${arch_mman} expands to nothing . Suggested patch attached . mmap_flags.sh.patch (1,222 bytes) |
|
|
complete modified mmap_flags.sh. Will test on next build. |
|
|
Oops, actually attach complete modified mmap_flags.sh. Will test on next build. |
|
|
OK, this website does not allow 'mmap_flags.sh' to be attached - here is the complete modified file: <BEGIN MODIFIED mmap_flags.sh> #!/bin/sh # SPDX-License-Identifier: LGPL-2.1 if [ $# -ne 2 ] ; then [ $# -eq 1 ] && hostarch=$1 || hostarch=`uname -m | sed -e s/i.86/x86/ -e s/x86_64/x86/` header_dir=tools/include/uapi/asm-generic arch_header_dir=tools/arch/${hostarch}/include/uapi/asm else header_dir=$1 arch_header_dir=$2 fi arch_mman=${arch_header_dir}/mman.h export arch_mman if [x != "x$header_dir" ] ; then export header_dir fi if [x != "x$arch_header_dir" ] ; then export arch_header_dir fi if [ x != "x$linux_mman" ]; then export linux_mman; fi # those in egrep -vw are flags, we want just the bits printf "static const char *mmap_flags[] = {\n" regex='^[[:space:]]*#[[:space:]]*define[[:space:]]+MAP_([[:alnum:]_]+)[[:space:]]+(0x[[:xdigit:]]+)[[:space:]]*.*' egrep -q $regex ${arch_mman} && \ (egrep $regex ${arch_mman} | \ sed -r "s/$regex/\2 \1 \1 \1 \2/g" | \ xargs printf "\t[ilog2(%s) + 1] = \"%s\",\n#ifndef MAP_%s\n#define MAP_%s %s\n#endif\n") if [ x != "$linux_mman" ]; then export linux_mman; egrep -q $regex ${linux_mman} && \ (egrep $regex ${linux_mman} | \ egrep -vw 'MAP_(UNINITIALIZED|TYPE|SHARED_VALIDATE)' | \ sed -r "s/$regex/\2 \1 \1 \1 \2/g" | \ xargs printf "\t[ilog2(%s) + 1] = \"%s\",\n#ifndef MAP_%s\n#define MAP_%s %s\n#endif\n") fi ([ ! -f ${arch_mman} ] || egrep -q '#[[:space:]]*include[[:space:]]+.*uapi/asm-generic/mman.*' ${arch_mman}) && (egrep $regex ${header_dir}/mman-common.h | \ egrep -vw 'MAP_(UNINITIALIZED|TYPE|SHARED_VALIDATE)' | \ sed -r "s/$regex/\2 \1 \1 \1 \2/g" | \ xargs printf "\t[ilog2(%s) + 1] = \"%s\",\n#ifndef MAP_%s\n#define MAP_%s %s\n#endif\n") ([ ! -f ${arch_mman} ] || egrep -q '#[[:space:]]*include[[:space:]]+.*uapi/asm-generic/mman.h>.*' ${arch_mman}) && (egrep $regex ${header_dir}/mman.h | \ sed -r "s/$regex/\2 \1 \1 \1 \2/g" | \ xargs printf "\t[ilog2(%s) + 1] = \"%s\",\n#ifndef MAP_%s\n#define MAP_%s %s\n#endif\n") printf "};\n" </END MODIFIED mmap_flags.sh> |
|
|
Oops, better, WORKING patch attached. The problem was with $linux_mman , which is not set by the script, and evidently not set by its invoking environment either. mmap_flags.sh-2.patch (1,649 bytes) |
|
|
Since once again the website does not allow 'mmap_flags.sh' to be attached, here it is in full : <BEGIN MODIFIED mmap_flags.sh> #!/bin/sh # SPDX-License-Identifier: LGPL-2.1 if [ $# -ne 2 ] ; then [ $# -eq 1 ] && hostarch=$1 || hostarch=`uname -m | sed -e s/i.86/x86/ -e s/x86_64/x86/` header_dir=tools/include/uapi/asm-generic arch_header_dir=tools/arch/${hostarch}/include/uapi/asm else header_dir=$1 arch_header_dir=$2 fi export arch_mman=${arch_header_dir}/mman.h if [ x != "x$header_dir" ] ; then export header_dir="$header_dir"; fi if [ x != "x$arch_header_dir" ] ; then export arch_header_dir="$arch_header_dir"; fi if [ x != "x$linux_mman" ]; then export linux_mman="$linux_mman"; fi # those in egrep -vw are flags, we want just the bits printf "static const char *mmap_flags[] = {\n" regex='^[[:space:]]*#[[:space:]]*define[[:space:]]+MAP_([[:alnum:]_]+)[[:space:]]+(0x[[:xdigit:]]+)[[:space:]]*.*' egrep -q $regex ${arch_mman} && \ (egrep $regex ${arch_mman} | \ sed -r "s/$regex/\2 \1 \1 \1 \2/g" | \ xargs printf "\t[ilog2(%s) + 1] = \"%s\",\n#ifndef MAP_%s\n#define MAP_%s %s\n#endif\n") if [ x != "x$linux_mman" ]; then egrep -q $regex ${linux_mman} && \ (egrep $regex ${linux_mman} | \ egrep -vw 'MAP_(UNINITIALIZED|TYPE|SHARED_VALIDATE)' | \ sed -r "s/$regex/\2 \1 \1 \1 \2/g" | \ xargs printf "\t[ilog2(%s) + 1] = \"%s\",\n#ifndef MAP_%s\n#define MAP_%s %s\n#endif\n") fi ([ ! -f ${arch_mman} ] || egrep -q '#[[:space:]]*include[[:space:]]+.*uapi/asm-generic/mman.*' ${arch_mman}) && (egrep $regex ${header_dir}/mman-common.h | \ egrep -vw 'MAP_(UNINITIALIZED|TYPE|SHARED_VALIDATE)' | \ sed -r "s/$regex/\2 \1 \1 \1 \2/g" | \ xargs printf "\t[ilog2(%s) + 1] = \"%s\",\n#ifndef MAP_%s\n#define MAP_%s %s\n#endif\n") ([ ! -f ${arch_mman} ] || egrep -q '#[[:space:]]*include[[:space:]]+.*uapi/asm-generic/mman.h>.*' ${arch_mman}) && (egrep $regex ${header_dir}/mman.h | \ sed -r "s/$regex/\2 \1 \1 \1 \2/g" | \ xargs printf "\t[ilog2(%s) + 1] = \"%s\",\n#ifndef MAP_%s\n#define MAP_%s %s\n#endif\n") printf "};\n" </END MODIFIED mmap_flags.sh> |
 |
Using mock[1], the supported method of building RPM packages for Rocky Linux, is recommended for any and all package builds. % dnf install epel-release -y % dnf install mock -y % usermod -aG mock user % su - user $ wget https://kojidev.rockylinux.org/kojifiles/packages/kernel/4.18.0/513.9.1.el8_9/src/kernel-4.18.0-513.9.1.el8_9.src.rpm ## (Don't know what you're using for $RPM_OPT_FLAGS) $ mock -r rocky-8-x86_64 --without debug --without kabichk --with release --with headers --with perf --with verbose kernel-4.18.0-513.9.1.el8_9.src.rpm The result: . . . Wrote: /builddir/build/RPMS/kernel-4.18.0-513.9.1.el8.x86_64.rpm Wrote: /builddir/build/RPMS/kernel-headers-4.18.0-513.9.1.el8.x86_64.rpm Wrote: /builddir/build/RPMS/kernel-cross-headers-4.18.0-513.9.1.el8.x86_64.rpm Wrote: /builddir/build/RPMS/kernel-debuginfo-common-x86_64-4.18.0-513.9.1.el8.x86_64.rpm Wrote: /builddir/build/RPMS/perf-4.18.0-513.9.1.el8.x86_64.rpm Wrote: /builddir/build/RPMS/perf-debuginfo-4.18.0-513.9.1.el8.x86_64.rpm Wrote: /builddir/build/RPMS/python3-perf-4.18.0-513.9.1.el8.x86_64.rpm Wrote: /builddir/build/RPMS/python3-perf-debuginfo-4.18.0-513.9.1.el8.x86_64.rpm Wrote: /builddir/build/RPMS/kernel-tools-4.18.0-513.9.1.el8.x86_64.rpm Wrote: /builddir/build/RPMS/kernel-tools-libs-4.18.0-513.9.1.el8.x86_64.rpm Wrote: /builddir/build/RPMS/kernel-tools-libs-devel-4.18.0-513.9.1.el8.x86_64.rpm Wrote: /builddir/build/RPMS/kernel-tools-debuginfo-4.18.0-513.9.1.el8.x86_64.rpm Wrote: /builddir/build/RPMS/bpftool-4.18.0-513.9.1.el8.x86_64.rpm Wrote: /builddir/build/RPMS/bpftool-debuginfo-4.18.0-513.9.1.el8.x86_64.rpm Wrote: /builddir/build/RPMS/kernel-selftests-internal-4.18.0-513.9.1.el8.x86_64.rpm Wrote: /builddir/build/RPMS/kernel-core-4.18.0-513.9.1.el8.x86_64.rpm Wrote: /builddir/build/RPMS/kernel-devel-4.18.0-513.9.1.el8.x86_64.rpm Wrote: /builddir/build/RPMS/kernel-modules-4.18.0-513.9.1.el8.x86_64.rpm Wrote: /builddir/build/RPMS/kernel-modules-extra-4.18.0-513.9.1.el8.x86_64.rpm Wrote: /builddir/build/RPMS/kernel-modules-internal-4.18.0-513.9.1.el8.x86_64.rpm Wrote: /builddir/build/RPMS/kernel-debuginfo-4.18.0-513.9.1.el8.x86_64.rpm Wrote: /builddir/build/RPMS/kernel-ipaclones-internal-4.18.0-513.9.1.el8.x86_64.rpm Finish: rpmbuild kernel-4.18.0-513.9.1.el8_9.src.rpm Finish: build phase for kernel-4.18.0-513.9.1.el8_9.src.rpm INFO: Done(kernel-4.18.0-513.9.1.el8_9.src.rpm) Config(rocky-8-x86_64) 45 minutes 35 seconds INFO: Results and/or logs in: /var/lib/mock/rocky-8-x86_64/result Finish: run $ grep mmap_flags /var/lib/mock/rocky-8-x86_64/result/build.log /var/lib/mock/rocky-8-x86_64/result/build.log:/bin/sh '/builddir/build/BUILD/kernel-4.18.0-513.9.1.el8_9/linux-4.18.0-513.9.1.el8.x86_64/tools/perf/trace/beauty/mmap_flags.sh' /builddir/build/BUILD/kernel-4.18.0-513.9.1.el8_9/linux-4.18.0-513.9.1.el8.x86_64/tools/include/uapi/asm-generic /builddir/build/BUILD/kernel-4.18.0-513.9.1.el8_9/linux-4.18.0-513.9.1.el8.x86_64/tools/arch/x86/include/uapi/asm/ > trace/beauty/generated/mmap_flags_array.c -- On the surface this issue appears to be environmental to your system setup and configuration. It is highly likely that it is due to: 1) The use of rpmbuild in a non-clean environment rather than mock[1] which is what we (and upstream) use for all of our package builds. You are highly encouraged (and recommended) to do the same. 2) The use of shell redirects which are unnecessary using the standard recommended method of building packages (mock) Please replicate your issue using mock to build the kernel package. If this still fails for you, attempt to replicate this in RHEL 8.9 and then CentOS Stream 8. If you are able to successfully replicate your issue in all cases using mock, you are encouraged to report this to issues.redhat.com. Setting to needinfo. ## Footnotes [1] https://rpm-software-management.github.io/mock/ - We and upstream use mock in our build systems and everyone is encouraged to use this tool when building packages. [2] Please re-review the bug tracker guidelines in account services. It clearly states that arbitrary scripts will be rejected. If you feel that this is an issue, please state your case here: https://git.resf.org/infrastructure/meta |
|
|
If the kernel RPM cannot be built by 'rpmbuild', you should not distribute the kernel.spec file (keep it secret) . Use of Mock should not be an excuse to allow BROKEN SCRIPTS like mmap_flags.sh in the build . You have not reproduced the test conditions I was operating under in your test shown above - I was not using Mock . When you do not use Mock, but use 'rpmbuild', as all .spec files are meant to be built with, you see that the broken mmap_flags.sh script is exposed and fails the build . If a distribution does not support building the kernel RPM from source, from the command line, by running a command like 'rpmbuild' that is the handler of the build .spec file, I would not be using such a distribution at all . I will also raise this as a RHEL-8 bug - they do seem to care about so obviously broken kernel build scripts - you will doubtless get the fixed script in an update. |
 |
rpmbuild is but a tool used to build the distribution. In fact, rpmbuild is called inside of mock. The functionality of the script in a pure rpmbuild environment (i.e., not a clean build root) is not considered a bug. Obviously the build works--otherwise how would we complete the build? We are, of course, concerned about build failures, but as it stands, your issue is purely environmental as it cannot be reproduced in a clean mock environment, which is how this package is intended to be built. |
|
|
Closing due to 8.10 beta releasing soon and the issue is not reproducible using the standard, recommended toolset used to build packages for Rocky Linux. If you are able to reliably reproduce your issue using the recommended toolset on Rocky Linux and are not able to on CentOS Stream 8 and RHEL 8, please open a new bug report. |
- Anonymous
