From d9ecb92b4fbbbb0a9993017da8d044541ca35886 Mon Sep 17 00:00:00 2001 From: Masahiro Yamada Date: Tue, 3 Dec 2024 22:59:58 +0900 Subject: [PATCH 01/41] kbuild: deb-pkg: do not include empty hook directories The linux-image package currently includes empty hook directories (/etc/kernel/{pre,post}{inst,rm}.d/ by default). These directories were perhaps intended as a fail-safe in case no hook scripts exist there. However, they are really unnecessary because the run-parts command is already guarded by the following check: test -d ${debhookdir}/${script}.d && run-parts ... The only difference is that the run-parts command either runs for empty directories (resulting in a no-op) or is skipped entirely. The maintainer scripts will succeed without these dummy directories. The linux-image packages from the Debian kernel do not contain /etc/kernel/*.d/, either. Signed-off-by: Masahiro Yamada --- scripts/package/builddeb | 2 -- 1 file changed, 2 deletions(-) diff --git a/scripts/package/builddeb b/scripts/package/builddeb index ad7aba0f268e..85fe8f56bb9b 100755 --- a/scripts/package/builddeb +++ b/scripts/package/builddeb @@ -76,8 +76,6 @@ install_maint_scripts () { # so do we; recent versions of dracut and initramfs-tools will obey this. debhookdir=${KDEB_HOOKDIR:-/etc/kernel} for script in postinst postrm preinst prerm; do - mkdir -p "${pdir}${debhookdir}/${script}.d" - mkdir -p "${pdir}/DEBIAN" cat <<-EOF > "${pdir}/DEBIAN/${script}" #!/bin/sh From ac2c30f98f28a6606af89ce44bff77af5d558fe8 Mon Sep 17 00:00:00 2001 From: Johannes Schauer Marin Rodrigues Date: Tue, 3 Dec 2024 17:17:35 +0100 Subject: [PATCH 02/41] kbuild: deb-pkg: allow hooks also in /usr/share/kernel By passing an additional directory to run-parts, allow Debian and its derivatives to ship maintainer scripts in /usr while at the same time allowing the local admin to override or disable them by placing hooks of the same name in /etc. This adds support for the mechanism described in the UAPI Configuration Files Specification for kernel hooks. The same idea is also used by udev, systemd or modprobe for their config files. https://uapi-group.org/specifications/specs/configuration_files_specification/ This functionality relies on run-parts 5.21 or later. It is the responsibility of packages installing hooks into /usr/share/kernel to also declare a Depends: debianutils (>= 5.21). KDEB_HOOKDIR can be used to change the list of directories that is searched. By default, /etc/kernel and /usr/share/kernel are hook directories. Since the list of directories in KDEB_HOOKDIR is separated by spaces, the paths must not contain the space character themselves. Signed-off-by: Johannes Schauer Marin Rodrigues Signed-off-by: Masahiro Yamada --- scripts/package/builddeb | 22 ++++++++++++++++------ 1 file changed, 16 insertions(+), 6 deletions(-) diff --git a/scripts/package/builddeb b/scripts/package/builddeb index 85fe8f56bb9b..3627ca227e5a 100755 --- a/scripts/package/builddeb +++ b/scripts/package/builddeb @@ -5,10 +5,12 @@ # # Simple script to generate a deb package for a Linux kernel. All the # complexity of what to do with a kernel after it is installed or removed -# is left to other scripts and packages: they can install scripts in the -# /etc/kernel/{pre,post}{inst,rm}.d/ directories (or an alternative location -# specified in KDEB_HOOKDIR) that will be called on package install and -# removal. +# is left to other scripts and packages. Scripts can be placed into the +# preinst, postinst, prerm and postrm directories in /etc/kernel or +# /usr/share/kernel. A different list of search directories can be given +# via KDEB_HOOKDIR. Scripts in directories earlier in the list will +# override scripts of the same name in later directories. The script will +# be called on package installation and removal. set -eu @@ -74,7 +76,7 @@ install_maint_scripts () { # kernel packages, as well as kernel packages built using make-kpkg. # make-kpkg sets $INITRD to indicate whether an initramfs is wanted, and # so do we; recent versions of dracut and initramfs-tools will obey this. - debhookdir=${KDEB_HOOKDIR:-/etc/kernel} + debhookdir=${KDEB_HOOKDIR:-/etc/kernel /usr/share/kernel} for script in postinst postrm preinst prerm; do mkdir -p "${pdir}/DEBIAN" cat <<-EOF > "${pdir}/DEBIAN/${script}" @@ -88,7 +90,15 @@ install_maint_scripts () { # Tell initramfs builder whether it's wanted export INITRD=$(if_enabled_echo CONFIG_BLK_DEV_INITRD Yes No) - test -d ${debhookdir}/${script}.d && run-parts --arg="${KERNELRELEASE}" --arg="/${installed_image_path}" ${debhookdir}/${script}.d + # run-parts will error out if one of its directory arguments does not + # exist, so filter the list of hook directories accordingly. + hookdirs= + for dir in ${debhookdir}; do + test -d "\$dir/${script}.d" || continue + hookdirs="\$hookdirs \$dir/${script}.d" + done + hookdirs="\${hookdirs# }" + test -n "\$hookdirs" && run-parts --arg="${KERNELRELEASE}" --arg="/${installed_image_path}" \$hookdirs exit 0 EOF chmod 755 "${pdir}/DEBIAN/${script}" From 5f73e7d0386d970a7d0e9de5a58d53114de85033 Mon Sep 17 00:00:00 2001 From: Masahiro Yamada Date: Tue, 10 Dec 2024 19:06:17 +0900 Subject: [PATCH 03/41] kbuild: refactor cross-compiling linux-headers package Since commit 13b25489b6f8 ("kbuild: change working directory to external module directory with M="), when cross-building host programs for the linux-headers package, the "Entering directory" and "Leaving directory" messages appear multiple times, and each object path shown is relative to the working directory. This makes it difficult to track which objects are being rebuilt. In hindsight, using the external module build (M=) was not a good idea. This commit simplifies the script by leveraging the run-command target, resulting in a cleaner build log again. [Before] $ make ARCH=arm64 CROSS_COMPILE=aarch64-linux-gnu- bindeb-pkg [ snip ] Rebuilding host programs with aarch64-linux-gnu-gcc... make[5]: Entering directory '/home/masahiro/linux' make[6]: Entering directory '/home/masahiro/linux/debian/linux-headers-6.13.0-rc1+/usr/src/linux-headers-6.13.0-rc1+' HOSTCC scripts/kallsyms HOSTCC scripts/sorttable HOSTCC scripts/asn1_compiler make[6]: Leaving directory '/home/masahiro/linux/debian/linux-headers-6.13.0-rc1+/usr/src/linux-headers-6.13.0-rc1+' make[5]: Leaving directory '/home/masahiro/linux' make[5]: Entering directory '/home/masahiro/linux' make[6]: Entering directory '/home/masahiro/linux/debian/linux-headers-6.13.0-rc1+/usr/src/linux-headers-6.13.0-rc1+' HOSTCC scripts/basic/fixdep HOSTCC scripts/mod/modpost.o HOSTCC scripts/mod/file2alias.o HOSTCC scripts/mod/sumversion.o HOSTCC scripts/mod/symsearch.o HOSTLD scripts/mod/modpost make[6]: Leaving directory '/home/masahiro/linux/debian/linux-headers-6.13.0-rc1+/usr/src/linux-headers-6.13.0-rc1+' make[5]: Leaving directory '/home/masahiro/linux' [After] $ make ARCH=arm64 CROSS_COMPILE=aarch64-linux-gnu- bindeb-pkg [ snip ] HOSTCC debian/linux-headers-6.13.0-rc1+/usr/src/linux-headers-6.13.0-rc1+/scripts/basic/fixdep HOSTCC debian/linux-headers-6.13.0-rc1+/usr/src/linux-headers-6.13.0-rc1+/scripts/kallsyms HOSTCC debian/linux-headers-6.13.0-rc1+/usr/src/linux-headers-6.13.0-rc1+/scripts/sorttable HOSTCC debian/linux-headers-6.13.0-rc1+/usr/src/linux-headers-6.13.0-rc1+/scripts/asn1_compiler HOSTCC debian/linux-headers-6.13.0-rc1+/usr/src/linux-headers-6.13.0-rc1+/scripts/mod/modpost.o HOSTCC debian/linux-headers-6.13.0-rc1+/usr/src/linux-headers-6.13.0-rc1+/scripts/mod/file2alias.o HOSTCC debian/linux-headers-6.13.0-rc1+/usr/src/linux-headers-6.13.0-rc1+/scripts/mod/sumversion.o HOSTCC debian/linux-headers-6.13.0-rc1+/usr/src/linux-headers-6.13.0-rc1+/scripts/mod/symsearch.o HOSTLD debian/linux-headers-6.13.0-rc1+/usr/src/linux-headers-6.13.0-rc1+/scripts/mod/modpost Signed-off-by: Masahiro Yamada --- scripts/package/install-extmod-build | 33 ++++++++-------------------- 1 file changed, 9 insertions(+), 24 deletions(-) diff --git a/scripts/package/install-extmod-build b/scripts/package/install-extmod-build index d3c5b104c063..bb6e23c1174e 100755 --- a/scripts/package/install-extmod-build +++ b/scripts/package/install-extmod-build @@ -49,17 +49,10 @@ mkdir -p "${destdir}" # This caters to host programs that participate in Kbuild. objtool and # resolve_btfids are out of scope. if [ "${CC}" != "${HOSTCC}" ]; then - echo "Rebuilding host programs with ${CC}..." - - # This leverages external module building. - # - Clear sub_make_done to allow the top-level Makefile to redo sub-make. - # - Filter out --no-print-directory to print "Entering directory" logs - # when Make changes the working directory. - unset sub_make_done - MAKEFLAGS=$(echo "${MAKEFLAGS}" | sed s/--no-print-directory//) - - cat <<-'EOF' > "${destdir}/Kbuild" - subdir-y := scripts + cat "${destdir}/scripts/Makefile" - <<-'EOF' > "${destdir}/scripts/Kbuild" + subdir-y += basic + hostprogs-always-y += mod/modpost + mod/modpost-objs := $(addprefix mod/, modpost.o file2alias.o sumversion.o symsearch.o) EOF # HOSTCXX is not overridden. The C++ compiler is used to build: @@ -67,20 +60,12 @@ if [ "${CC}" != "${HOSTCC}" ]; then # - GCC plugins, which will not work on the installed system even after # being rebuilt. # - # Use the single-target build to avoid the modpost invocation, which - # would overwrite Module.symvers. - "${MAKE}" HOSTCC="${CC}" KBUILD_OUTPUT=. KBUILD_EXTMOD="${destdir}" scripts/ + # Clear VPATH and srcroot because the source files reside in the output + # directory. + # shellcheck disable=SC2016 # $(MAKE), $(CC), and $(build) will be expanded by Make + "${MAKE}" run-command KBUILD_RUN_COMMAND='+$(MAKE) HOSTCC=$(CC) VPATH= srcroot=. $(build)='"${destdir}"/scripts - cat <<-'EOF' > "${destdir}/scripts/Kbuild" - subdir-y := basic - hostprogs-always-y := mod/modpost - mod/modpost-objs := $(addprefix mod/, modpost.o file2alias.o sumversion.o symsearch.o) - EOF - - # Run once again to rebuild scripts/basic/ and scripts/mod/modpost. - "${MAKE}" HOSTCC="${CC}" KBUILD_OUTPUT=. KBUILD_EXTMOD="${destdir}" scripts/ - - rm -f "${destdir}/Kbuild" "${destdir}/scripts/Kbuild" + rm -f "${destdir}/scripts/Kbuild" fi find "${destdir}" \( -name '.*.cmd' -o -name '*.o' \) -delete From 1f937a4bcb0472015818f30f4d3c5546d3f09933 Mon Sep 17 00:00:00 2001 From: Masahiro Yamada Date: Tue, 10 Dec 2024 19:24:41 +0900 Subject: [PATCH 04/41] kbuild: suppress stdout from merge_config for silent builds merge_config does not respect the Make's -s (--silent) option. Let's sink the stdout from merge_config for silent builds. This commit does not cater to the direct invocation of merge_config.sh (e.g. arch/mips/Makefile). Reported-by: Leon Romanovsky Closes: https://lore.kernel.org/all/e534ce33b0e1060eb85ece8429810f087b034c88.1733234008.git.leonro@nvidia.com/ Signed-off-by: Masahiro Yamada Tested-by: Leon Romanovsky Reviewed-by: Nathan Chancellor --- scripts/Makefile.defconf | 13 +++++++------ scripts/kconfig/Makefile | 4 +++- 2 files changed, 10 insertions(+), 7 deletions(-) diff --git a/scripts/Makefile.defconf b/scripts/Makefile.defconf index 226ea3df3b4b..a44307f08e9d 100644 --- a/scripts/Makefile.defconf +++ b/scripts/Makefile.defconf @@ -1,6 +1,11 @@ # SPDX-License-Identifier: GPL-2.0 # Configuration heplers +cmd_merge_fragments = \ + $(srctree)/scripts/kconfig/merge_config.sh \ + $4 -m -O $(objtree) $(srctree)/arch/$(SRCARCH)/configs/$2 \ + $(foreach config,$3,$(srctree)/arch/$(SRCARCH)/configs/$(config).config) + # Creates 'merged defconfigs' # --------------------------------------------------------------------------- # Usage: @@ -8,9 +13,7 @@ # # Input config fragments without '.config' suffix define merge_into_defconfig - $(Q)$(CONFIG_SHELL) $(srctree)/scripts/kconfig/merge_config.sh \ - -m -O $(objtree) $(srctree)/arch/$(SRCARCH)/configs/$(1) \ - $(foreach config,$(2),$(srctree)/arch/$(SRCARCH)/configs/$(config).config) + $(call cmd,merge_fragments,$1,$2) +$(Q)$(MAKE) -f $(srctree)/Makefile olddefconfig endef @@ -22,8 +25,6 @@ endef # # Input config fragments without '.config' suffix define merge_into_defconfig_override - $(Q)$(CONFIG_SHELL) $(srctree)/scripts/kconfig/merge_config.sh \ - -Q -m -O $(objtree) $(srctree)/arch/$(SRCARCH)/configs/$(1) \ - $(foreach config,$(2),$(srctree)/arch/$(SRCARCH)/configs/$(config).config) + $(call cmd,merge_fragments,$1,$2,-Q) +$(Q)$(MAKE) -f $(srctree)/Makefile olddefconfig endef diff --git a/scripts/kconfig/Makefile b/scripts/kconfig/Makefile index a0a0be38cbdc..fb50bd4f4103 100644 --- a/scripts/kconfig/Makefile +++ b/scripts/kconfig/Makefile @@ -105,9 +105,11 @@ configfiles = $(wildcard $(srctree)/kernel/configs/$(1) $(srctree)/arch/$(SRCARC all-config-fragments = $(call configfiles,*.config) config-fragments = $(call configfiles,$@) +cmd_merge_fragments = $(srctree)/scripts/kconfig/merge_config.sh -m $(KCONFIG_CONFIG) $(config-fragments) + %.config: $(obj)/conf $(if $(config-fragments),, $(error $@ fragment does not exists on this architecture)) - $(Q)$(CONFIG_SHELL) $(srctree)/scripts/kconfig/merge_config.sh -m $(KCONFIG_CONFIG) $(config-fragments) + $(call cmd,merge_fragments) $(Q)$(MAKE) -f $(srctree)/Makefile olddefconfig PHONY += tinyconfig From 41e86fe7ebe9eaf18c394145cc91b08b5ec932a8 Mon Sep 17 00:00:00 2001 From: Masahiro Yamada Date: Wed, 18 Dec 2024 19:37:05 +0900 Subject: [PATCH 05/41] kheaders: exclude include/generated/utsversion.h from kheaders_data.tar.xz CONFIG_IKHEADERS has a reproducibility issue because the contents of kernel/kheaders_data.tar.xz can vary depending on how you build the kernel. If you build the kernel with CONFIG_IKHEADERS enabled from a pristine state, the tarball does not include include/generated/utsversion.h. $ make -s mrproper $ make -s defconfig $ scripts/config -e CONFIG_IKHEADERS $ make -s $ tar Jtf kernel/kheaders_data.tar.xz | grep utsversion However, if you build the kernel with CONFIG_IKHEADERS disabled first and then enable it later, the tarball does include include/generated/utsversion.h. $ make -s mrproper $ make -s defconfig $ make -s $ scripts/config -e CONFIG_IKHEADERS $ make -s $ tar Jtf kernel/kheaders_data.tar.xz | grep utsversion ./include/generated/utsversion.h It is not predictable whether a stale include/generated/utsversion.h remains when kheaders_data.tar.xz is generated. For better reproducibility, include/generated/utsversions.h should always be omitted. It is not necessary for the kheaders anyway. Signed-off-by: Masahiro Yamada --- kernel/gen_kheaders.sh | 4 ++++ 1 file changed, 4 insertions(+) diff --git a/kernel/gen_kheaders.sh b/kernel/gen_kheaders.sh index 383fd43ac612..a0e3fbf4afa4 100755 --- a/kernel/gen_kheaders.sh +++ b/kernel/gen_kheaders.sh @@ -83,6 +83,10 @@ for f in $dir_list; do find "$f" -name "*.h"; done | cpio --quiet -pdu $cpio_dir >/dev/null 2>&1 +# Always exclude include/generated/utsversion.h +# Otherwise, the contents of the tarball may vary depending on the build steps. +rm -f "${cpio_dir}/include/generated/utsversion.h" + # Remove comments except SDPX lines find $cpio_dir -type f -print0 | xargs -0 -P8 -n1 perl -pi -e 'BEGIN {undef $/;}; s/\/\*((?!SPDX).)*?\*\///smg;' From de0cae9273841ca019e438192d08b7358a002973 Mon Sep 17 00:00:00 2001 From: Masahiro Yamada Date: Wed, 18 Dec 2024 19:37:06 +0900 Subject: [PATCH 06/41] kheaders: avoid unnecessary process forks of grep Exclude include/generated/{utsversion.h,autoconf.h} by using the -path option to reduce the cost of forking new processes. No functional changes are intended. Signed-off-by: Masahiro Yamada --- kernel/gen_kheaders.sh | 6 +++--- 1 file changed, 3 insertions(+), 3 deletions(-) diff --git a/kernel/gen_kheaders.sh b/kernel/gen_kheaders.sh index a0e3fbf4afa4..c2eba1a0d772 100755 --- a/kernel/gen_kheaders.sh +++ b/kernel/gen_kheaders.sh @@ -48,9 +48,9 @@ all_dirs="$all_dirs $dir_list" # check include/generated/autoconf.h explicitly. # # Ignore them for md5 calculation to avoid pointless regeneration. -headers_md5="$(find $all_dirs -name "*.h" | - grep -v "include/generated/utsversion.h" | - grep -v "include/generated/autoconf.h" | +headers_md5="$(find $all_dirs -name "*.h" -a \ + ! -path include/generated/utsversion.h -a \ + ! -path include/generated/autoconf.h | xargs ls -l | md5sum | cut -d ' ' -f1)" # Any changes to this script will also cause a rebuild of the archive. From fd2a118c483472f8862cc46981a5230414cd0e67 Mon Sep 17 00:00:00 2001 From: Masahiro Yamada Date: Wed, 18 Dec 2024 19:37:07 +0900 Subject: [PATCH 07/41] kheaders: rename the 'cpio_dir' variable to 'tmpdir' The next commit will get rid of the use of 'cpio' command, as there is no strong reason to use it just for copying files. Before that, this commit renames the 'cpio_dir' variable to 'tmpdir'. No functional changes are intended. Signed-off-by: Masahiro Yamada --- kernel/gen_kheaders.sh | 18 +++++++++--------- 1 file changed, 9 insertions(+), 9 deletions(-) diff --git a/kernel/gen_kheaders.sh b/kernel/gen_kheaders.sh index c2eba1a0d772..ddfd1177567f 100755 --- a/kernel/gen_kheaders.sh +++ b/kernel/gen_kheaders.sh @@ -7,7 +7,7 @@ set -e sfile="$(readlink -f "$0")" outdir="$(pwd)" tarfile=$1 -cpio_dir=$outdir/${tarfile%/*}/.tmp_cpio_dir +tmpdir=$outdir/${tarfile%/*}/.tmp_dir dir_list=" include/ @@ -65,15 +65,15 @@ fi echo " GEN $tarfile" -rm -rf $cpio_dir -mkdir $cpio_dir +rm -rf "${tmpdir}" +mkdir "${tmpdir}" if [ "$building_out_of_srctree" ]; then ( cd $srctree for f in $dir_list do find "$f" -name "*.h"; - done | cpio --quiet -pd $cpio_dir + done | cpio --quiet -pd "${tmpdir}" ) fi @@ -81,23 +81,23 @@ fi # of tree builds having stale headers in srctree. Just silence CPIO for now. for f in $dir_list; do find "$f" -name "*.h"; -done | cpio --quiet -pdu $cpio_dir >/dev/null 2>&1 +done | cpio --quiet -pdu "${tmpdir}" >/dev/null 2>&1 # Always exclude include/generated/utsversion.h # Otherwise, the contents of the tarball may vary depending on the build steps. -rm -f "${cpio_dir}/include/generated/utsversion.h" +rm -f "${tmpdir}/include/generated/utsversion.h" # Remove comments except SDPX lines -find $cpio_dir -type f -print0 | +find "${tmpdir}" -type f -print0 | xargs -0 -P8 -n1 perl -pi -e 'BEGIN {undef $/;}; s/\/\*((?!SPDX).)*?\*\///smg;' # Create archive and try to normalize metadata for reproducibility. tar "${KBUILD_BUILD_TIMESTAMP:+--mtime=$KBUILD_BUILD_TIMESTAMP}" \ --owner=0 --group=0 --sort=name --numeric-owner --mode=u=rw,go=r,a+X \ - -I $XZ -cf $tarfile -C $cpio_dir/ . > /dev/null + -I $XZ -cf $tarfile -C "${tmpdir}/" . > /dev/null echo $headers_md5 > kernel/kheaders.md5 echo "$this_file_md5" >> kernel/kheaders.md5 echo "$(md5sum $tarfile | cut -d ' ' -f1)" >> kernel/kheaders.md5 -rm -rf $cpio_dir +rm -rf "${tmpdir}" From 82a1978d0fdc28e561bc4d98ea155dd322f33c19 Mon Sep 17 00:00:00 2001 From: Masahiro Yamada Date: Wed, 18 Dec 2024 19:37:08 +0900 Subject: [PATCH 08/41] kheaders: use 'tar' instead of 'cpio' for copying files The 'cpio' command is used solely for copying header files to the temporary directory. However, there is no strong reason to use 'cpio' for this purpose. For example, scripts/package/install-extmod-build uses the 'tar' command to copy files. This commit replaces the use of 'cpio' with 'tar' because 'tar' is already used in this script to generate kheaders_data.tar.xz anyway. Performance-wide, there is no significant difference between 'cpio' and 'tar'. [Before] $ rm -fr kheaders; mkdir kheaders $ time sh -c ' for f in include arch/x86/include do find "$f" -name "*.h" done | cpio --quiet -pd kheaders ' real 0m0.148s user 0m0.021s sys 0m0.140s [After] $ rm -fr kheaders; mkdir kheaders $ time sh -c ' for f in include arch/x86/include do find "$f" -name "*.h" done | tar -c -f - -T - | tar -xf - -C kheaders ' real 0m0.098s user 0m0.024s sys 0m0.131s Revert commit 69ef0920bdd3 ("Docs: Add cpio requirement to changes.rst") because 'cpio' is not used anywhere else during the kernel build. Please note that the built-in initramfs is created by the in-tree tool, usr/gen_init_cpio, so it does not rely on the external 'cpio' command at all. Remove 'cpio' from the package build dependencies as well. Signed-off-by: Masahiro Yamada --- Documentation/process/changes.rst | 6 ------ kernel/gen_kheaders.sh | 13 ++----------- scripts/package/PKGBUILD | 1 - scripts/package/mkdebian | 2 +- 4 files changed, 3 insertions(+), 19 deletions(-) diff --git a/Documentation/process/changes.rst b/Documentation/process/changes.rst index 82b5e378eebf..a0beca805362 100644 --- a/Documentation/process/changes.rst +++ b/Documentation/process/changes.rst @@ -59,7 +59,6 @@ iptables 1.4.2 iptables -V openssl & libcrypto 1.0.0 openssl version bc 1.06.95 bc --version Sphinx\ [#f1]_ 2.4.4 sphinx-build --version -cpio any cpio --version GNU tar 1.28 tar --version gtags (optional) 6.6.5 gtags --version mkimage (optional) 2017.01 mkimage --version @@ -536,11 +535,6 @@ mcelog - -cpio ----- - -- - Networking ********** diff --git a/kernel/gen_kheaders.sh b/kernel/gen_kheaders.sh index ddfd1177567f..55f493d83b8f 100755 --- a/kernel/gen_kheaders.sh +++ b/kernel/gen_kheaders.sh @@ -14,13 +14,6 @@ include/ arch/$SRCARCH/include/ " -if ! command -v cpio >/dev/null; then - echo >&2 "***" - echo >&2 "*** 'cpio' could not be found." - echo >&2 "***" - exit 1 -fi - # Support incremental builds by skipping archive generation # if timestamps of files being archived are not changed. @@ -73,15 +66,13 @@ if [ "$building_out_of_srctree" ]; then cd $srctree for f in $dir_list do find "$f" -name "*.h"; - done | cpio --quiet -pd "${tmpdir}" + done | tar -c -f - -T - | tar -xf - -C "${tmpdir}" ) fi -# The second CPIO can complain if files already exist which can happen with out -# of tree builds having stale headers in srctree. Just silence CPIO for now. for f in $dir_list; do find "$f" -name "*.h"; -done | cpio --quiet -pdu "${tmpdir}" >/dev/null 2>&1 +done | tar -c -f - -T - | tar -xf - -C "${tmpdir}" # Always exclude include/generated/utsversion.h # Otherwise, the contents of the tarball may vary depending on the build steps. diff --git a/scripts/package/PKGBUILD b/scripts/package/PKGBUILD index dca706617adc..0cf3a55b05e1 100644 --- a/scripts/package/PKGBUILD +++ b/scripts/package/PKGBUILD @@ -22,7 +22,6 @@ license=(GPL-2.0-only) makedepends=( bc bison - cpio flex gettext kmod diff --git a/scripts/package/mkdebian b/scripts/package/mkdebian index b038a1380b8a..b6dd98ca860b 100755 --- a/scripts/package/mkdebian +++ b/scripts/package/mkdebian @@ -205,7 +205,7 @@ Priority: optional Maintainer: $maintainer Rules-Requires-Root: no Build-Depends: debhelper-compat (= 12) -Build-Depends-Arch: bc, bison, cpio, flex, +Build-Depends-Arch: bc, bison, flex, gcc-${host_gnu} , kmod, libelf-dev:native, libssl-dev:native, libssl-dev , From 41a00051283e301f7e0009626ddf591542e30161 Mon Sep 17 00:00:00 2001 From: HONG Yifan Date: Wed, 18 Dec 2024 20:20:11 +0000 Subject: [PATCH 09/41] kheaders: prevent `find` from seeing perl temp files Symptom: The command find ... | xargs ... perl -i occasionally triggers error messages like the following, with the build still succeeding: Can't open /kernel/.tmp_dir/include/dt-bindings/clock/XXNX4nW9: No such file or directory. Analysis: With strace, the root cause has been identified to be `perl -i` creating temporary files inside ${tmpdir}, which causes `find` to see the temporary files and emit the names. `find` is likely implemented with readdir. POSIX `readdir` says: If a file is removed from or added to the directory after the most recent call to opendir() or rewinddir(), whether a subsequent call to readdir() returns an entry for that file is unspecified. So if the libc that `find` links against choose to return that entry in readdir(), a possible sequence of events is the following: 1. find emits foo.h 2. xargs executes `perl -i foo.h` 3. perl (pid=100) creates temporary file `XXXXXXXX` 4. find sees file `XXXXXXXX` and emit it 5. PID 100 exits, cleaning up the temporary file `XXXXXXXX` 6. xargs executes `perl -i XXXXXXXX` 7. perl (pid=200) tries to read the file, but it doesn't exist any more. ... triggering the error message. One can reproduce the bug with the following command (assuming PWD contains the list of headers in kheaders.tar.xz) for i in $(seq 100); do find -type f -print0 | xargs -0 -P8 -n1 perl -pi -e 'BEGIN {undef $/;}; s/\/\*((?!SPDX).)*?\*\///smg;'; done With a `find` linking against musl libc, the error message is emitted 6/100 times. The fix: This change stores the results of `find` before feeding them into xargs. find and xargs will no longer be able to see temporary files that perl creates after this change. Signed-off-by: HONG Yifan Signed-off-by: Masahiro Yamada --- kernel/gen_kheaders.sh | 9 +++++++-- 1 file changed, 7 insertions(+), 2 deletions(-) diff --git a/kernel/gen_kheaders.sh b/kernel/gen_kheaders.sh index 55f493d83b8f..c9e5dc068e85 100755 --- a/kernel/gen_kheaders.sh +++ b/kernel/gen_kheaders.sh @@ -79,8 +79,13 @@ done | tar -c -f - -T - | tar -xf - -C "${tmpdir}" rm -f "${tmpdir}/include/generated/utsversion.h" # Remove comments except SDPX lines -find "${tmpdir}" -type f -print0 | - xargs -0 -P8 -n1 perl -pi -e 'BEGIN {undef $/;}; s/\/\*((?!SPDX).)*?\*\///smg;' +# Use a temporary file to store directory contents to prevent find/xargs from +# seeing temporary files created by perl. +find "${tmpdir}" -type f -print0 > "${tmpdir}.contents.txt" +xargs -0 -P8 -n1 \ + perl -pi -e 'BEGIN {undef $/;}; s/\/\*((?!SPDX).)*?\*\///smg;' \ + < "${tmpdir}.contents.txt" +rm -f "${tmpdir}.contents.txt" # Create archive and try to normalize metadata for reproducibility. tar "${KBUILD_BUILD_TIMESTAMP:+--mtime=$KBUILD_BUILD_TIMESTAMP}" \ From ad2091dee019a68145610081a75fae3b90f0c44d Mon Sep 17 00:00:00 2001 From: Rolf Eike Beer Date: Thu, 19 Dec 2024 08:20:34 +0100 Subject: [PATCH 10/41] kconfig: qconf: use preferred form of QString API A QString constructed from a character literal of length 0, i.e. "", is not "null" for historical reasons. This does not matter here so use the preferred method isEmpty() instead. Also directly construct empty QString objects instead of passing in an empty character literal that has to be parsed into an empty object first. Signed-off-by: Rolf Eike Beer Link: https://doc.qt.io/qt-6/qstring.html#distinction-between-null-and-empty-strings Signed-off-by: Masahiro Yamada --- scripts/kconfig/qconf.cc | 8 ++++---- 1 file changed, 4 insertions(+), 4 deletions(-) diff --git a/scripts/kconfig/qconf.cc b/scripts/kconfig/qconf.cc index 6c92ef1e16ef..eaa465b0ccf9 100644 --- a/scripts/kconfig/qconf.cc +++ b/scripts/kconfig/qconf.cc @@ -1464,8 +1464,8 @@ void ConfigMainWindow::loadConfig(void) { QString str; - str = QFileDialog::getOpenFileName(this, "", configname); - if (str.isNull()) + str = QFileDialog::getOpenFileName(this, QString(), configname); + if (str.isEmpty()) return; if (conf_read(str.toLocal8Bit().constData())) @@ -1491,8 +1491,8 @@ void ConfigMainWindow::saveConfigAs(void) { QString str; - str = QFileDialog::getSaveFileName(this, "", configname); - if (str.isNull()) + str = QFileDialog::getSaveFileName(this, QString(), configname); + if (str.isEmpty()) return; if (conf_write(str.toLocal8Bit().constData())) { From 1cd9502ee9275c6176a7312863f939cca9506114 Mon Sep 17 00:00:00 2001 From: Masahiro Yamada Date: Sun, 29 Dec 2024 00:45:28 +0900 Subject: [PATCH 11/41] module: get symbol CRC back to unsigned Commit 71810db27c1c ("modversions: treat symbol CRCs as 32 bit quantities") changed the CRC fields to s32 because the __kcrctab and __kcrctab_gpl sections contained relative references to the actual CRC values stored in the .rodata section when CONFIG_MODULE_REL_CRCS=y. Commit 7b4537199a4a ("kbuild: link symbol CRCs at final link, removing CONFIG_MODULE_REL_CRCS") removed this complexity. Now, the __kcrctab and __kcrctab_gpl sections directly contain the CRC values in all cases. The genksyms tool outputs unsigned 32-bit CRC values, so u32 is preferred over s32. No functional changes are intended. Regardless of this change, the CRC value is assigned to the u32 variable 'crcval' before the comparison, as seen in kernel/module/version.c: crcval = *crc; It was previously mandatory (but now optional) in order to avoid sign extension because the following line previously compared 'unsigned long' and 's32': if (versions[i].crc == crcval) return 1; versions[i].crc is still 'unsigned long' for backward compatibility. Signed-off-by: Masahiro Yamada Reviewed-by: Petr Pavlu --- include/linux/module.h | 4 ++-- kernel/module/internal.h | 10 +++++----- kernel/module/main.c | 2 +- kernel/module/version.c | 2 +- 4 files changed, 9 insertions(+), 9 deletions(-) diff --git a/include/linux/module.h b/include/linux/module.h index 94acbacdcdf1..903ef8fe4c04 100644 --- a/include/linux/module.h +++ b/include/linux/module.h @@ -430,7 +430,7 @@ struct module { /* Exported symbols */ const struct kernel_symbol *syms; - const s32 *crcs; + const u32 *crcs; unsigned int num_syms; #ifdef CONFIG_ARCH_USES_CFI_TRAPS @@ -448,7 +448,7 @@ struct module { /* GPL-only exported symbols. */ unsigned int num_gpl_syms; const struct kernel_symbol *gpl_syms; - const s32 *gpl_crcs; + const u32 *gpl_crcs; bool using_gplonly_symbols; #ifdef CONFIG_MODULE_SIG diff --git a/kernel/module/internal.h b/kernel/module/internal.h index daef2be83902..f10dc3ea7ff8 100644 --- a/kernel/module/internal.h +++ b/kernel/module/internal.h @@ -55,8 +55,8 @@ extern const struct kernel_symbol __start___ksymtab[]; extern const struct kernel_symbol __stop___ksymtab[]; extern const struct kernel_symbol __start___ksymtab_gpl[]; extern const struct kernel_symbol __stop___ksymtab_gpl[]; -extern const s32 __start___kcrctab[]; -extern const s32 __start___kcrctab_gpl[]; +extern const u32 __start___kcrctab[]; +extern const u32 __start___kcrctab_gpl[]; struct load_info { const char *name; @@ -102,7 +102,7 @@ struct find_symbol_arg { /* Output */ struct module *owner; - const s32 *crc; + const u32 *crc; const struct kernel_symbol *sym; enum mod_license license; }; @@ -384,7 +384,7 @@ static inline void init_param_lock(struct module *mod) { } #ifdef CONFIG_MODVERSIONS int check_version(const struct load_info *info, - const char *symname, struct module *mod, const s32 *crc); + const char *symname, struct module *mod, const u32 *crc); void module_layout(struct module *mod, struct modversion_info *ver, struct kernel_param *kp, struct kernel_symbol *ks, struct tracepoint * const *tp); int check_modstruct_version(const struct load_info *info, struct module *mod); @@ -393,7 +393,7 @@ int same_magic(const char *amagic, const char *bmagic, bool has_crcs); static inline int check_version(const struct load_info *info, const char *symname, struct module *mod, - const s32 *crc) + const u32 *crc) { return 1; } diff --git a/kernel/module/main.c b/kernel/module/main.c index 5399c182b3cb..e58bff88b8d6 100644 --- a/kernel/module/main.c +++ b/kernel/module/main.c @@ -86,7 +86,7 @@ struct mod_tree_root mod_tree __cacheline_aligned = { struct symsearch { const struct kernel_symbol *start, *stop; - const s32 *crcs; + const u32 *crcs; enum mod_license license; }; diff --git a/kernel/module/version.c b/kernel/module/version.c index 53f43ac5a73e..4e5731d403af 100644 --- a/kernel/module/version.c +++ b/kernel/module/version.c @@ -13,7 +13,7 @@ int check_version(const struct load_info *info, const char *symname, struct module *mod, - const s32 *crc) + const u32 *crc) { Elf_Shdr *sechdrs = info->sechdrs; unsigned int versindex = info->index.vers; From 5963913bb57f15f198361bc7f1389c756b98f25f Mon Sep 17 00:00:00 2001 From: Masahiro Yamada Date: Sun, 29 Dec 2024 00:45:29 +0900 Subject: [PATCH 12/41] modpost: zero-pad CRC values in modversion_info array I do not think the '#' flag is useful here because adding the explicit '0x' is clearer. Add the '0' flag to zero-pad the CRC values. This change gives better alignment in the generated *.mod.c files. There is no impact to the compiled modules. [Before] $ grep -A5 modversion_info fs/efivarfs/efivarfs.mod.c static const struct modversion_info ____versions[] __used __section("__versions") = { { 0x907d14d, "blocking_notifier_chain_register" }, { 0x53d3b64, "simple_inode_init_ts" }, { 0x65487097, "__x86_indirect_thunk_rax" }, { 0x122c3a7e, "_printk" }, [After] $ grep -A5 modversion_info fs/efivarfs/efivarfs.mod.c static const struct modversion_info ____versions[] __used __section("__versions") = { { 0x0907d14d, "blocking_notifier_chain_register" }, { 0x053d3b64, "simple_inode_init_ts" }, { 0x65487097, "__x86_indirect_thunk_rax" }, { 0x122c3a7e, "_printk" }, Signed-off-by: Masahiro Yamada --- scripts/mod/modpost.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/scripts/mod/modpost.c b/scripts/mod/modpost.c index 7ea59dc4926b..dc907014108b 100644 --- a/scripts/mod/modpost.c +++ b/scripts/mod/modpost.c @@ -1832,7 +1832,7 @@ static void add_versions(struct buffer *b, struct module *mod) s->name, mod->name); break; } - buf_printf(b, "\t{ %#8x, \"%s\" },\n", + buf_printf(b, "\t{ 0x%08x, \"%s\" },\n", s->crc, s->name); } From 45c9c4101d3d2fdfa00852274bbebba65fcc3cf2 Mon Sep 17 00:00:00 2001 From: Masahiro Yamada Date: Fri, 3 Jan 2025 16:30:38 +0900 Subject: [PATCH 13/41] genksyms: fix memory leak when the same symbol is added from source When a symbol that is already registered is added again, __add_symbol() returns without freeing the symbol definition, making it unreachable. The following test cases demonstrate different memory leak points. [Test Case 1] Forward declaration with exactly the same definition $ cat foo.c #include void foo(void); void foo(void) {} EXPORT_SYMBOL(foo); [Test Case 2] Forward declaration with a different definition (e.g. attribute) $ cat foo.c #include void foo(void); __attribute__((__section__(".ref.text"))) void foo(void) {} EXPORT_SYMBOL(foo); [Test Case 3] Preserving an overridden symbol (compile with KBUILD_PRESERVE=1) $ cat foo.c #include void foo(void); void foo(void) { } EXPORT_SYMBOL(foo); $ cat foo.symref override foo void foo ( int ) The memory leaks in Test Case 1 and 2 have existed since the introduction of genksyms into the kernel tree. [1] The memory leak in Test Case 3 was introduced by commit 5dae9a550a74 ("genksyms: allow to ignore symbol checksum changes"). When multiple init_declarators are reduced to an init_declarator_list, the decl_spec must be duplicated. Otherwise, the following Test Case 4 would result in a double-free bug. [Test Case 4] $ cat foo.c #include extern int foo, bar; int foo, bar; EXPORT_SYMBOL(foo); In this case, 'foo' and 'bar' share the same decl_spec, 'int'. It must be unshared before being passed to add_symbol(). [1]: https://git.kernel.org/pub/scm/linux/kernel/git/history/history.git/commit/?id=46bd1da672d66ccd8a639d3c1f8a166048cca608 Fixes: 5dae9a550a74 ("genksyms: allow to ignore symbol checksum changes") Signed-off-by: Masahiro Yamada --- scripts/genksyms/genksyms.c | 3 +++ scripts/genksyms/parse.y | 14 ++++++++++++-- 2 files changed, 15 insertions(+), 2 deletions(-) diff --git a/scripts/genksyms/genksyms.c b/scripts/genksyms/genksyms.c index 07f9b8cfb233..8ca46f807b57 100644 --- a/scripts/genksyms/genksyms.c +++ b/scripts/genksyms/genksyms.c @@ -239,6 +239,7 @@ static struct symbol *__add_symbol(const char *name, enum symbol_type type, "unchanged\n"); } sym->is_declared = 1; + free_list(defn, NULL); return sym; } else if (!sym->is_declared) { if (sym->is_override && flag_preserve) { @@ -247,6 +248,7 @@ static struct symbol *__add_symbol(const char *name, enum symbol_type type, print_type_name(type, name); fprintf(stderr, " modversion change\n"); sym->is_declared = 1; + free_list(defn, NULL); return sym; } else { status = is_unknown_symbol(sym) ? @@ -254,6 +256,7 @@ static struct symbol *__add_symbol(const char *name, enum symbol_type type, } } else { error_with_pos("redefinition of %s", name); + free_list(defn, NULL); return sym; } break; diff --git a/scripts/genksyms/parse.y b/scripts/genksyms/parse.y index 8e9b5e69e8f0..840371d01bf4 100644 --- a/scripts/genksyms/parse.y +++ b/scripts/genksyms/parse.y @@ -152,14 +152,19 @@ simple_declaration: ; init_declarator_list_opt: - /* empty */ { $$ = NULL; } - | init_declarator_list + /* empty */ { $$ = NULL; } + | init_declarator_list { free_list(decl_spec, NULL); $$ = $1; } ; init_declarator_list: init_declarator { struct string_list *decl = *$1; *$1 = NULL; + + /* avoid sharing among multiple init_declarators */ + if (decl_spec) + decl_spec = copy_list_range(decl_spec, NULL); + add_symbol(current_name, is_typedef ? SYM_TYPEDEF : SYM_NORMAL, decl, is_extern); current_name = NULL; @@ -170,6 +175,11 @@ init_declarator_list: *$3 = NULL; free_list(*$2, NULL); *$2 = decl_spec; + + /* avoid sharing among multiple init_declarators */ + if (decl_spec) + decl_spec = copy_list_range(decl_spec, NULL); + add_symbol(current_name, is_typedef ? SYM_TYPEDEF : SYM_NORMAL, decl, is_extern); current_name = NULL; From be2fa44b5180a1f021efb40c55fdf63c249c3209 Mon Sep 17 00:00:00 2001 From: Masahiro Yamada Date: Fri, 3 Jan 2025 16:30:39 +0900 Subject: [PATCH 14/41] genksyms: fix memory leak when the same symbol is read from *.symref file When a symbol that is already registered is read again from *.symref file, __add_symbol() removes the previous one from the hash table without freeing it. [Test Case] $ cat foo.c #include void foo(void); void foo(void) {} EXPORT_SYMBOL(foo); $ cat foo.symref foo void foo ( void ) foo void foo ( void ) When a symbol is removed from the hash table, it must be freed along with its ->name and ->defn members. However, sym->name cannot be freed because it is sometimes shared with node->string, but not always. If sym->name and node->string share the same memory, free(sym->name) could lead to a double-free bug. To resolve this issue, always assign a strdup'ed string to sym->name. Fixes: 64e6c1e12372 ("genksyms: track symbol checksum changes") Signed-off-by: Masahiro Yamada --- scripts/genksyms/genksyms.c | 8 ++++++-- scripts/genksyms/genksyms.h | 2 +- scripts/genksyms/parse.y | 4 ++-- 3 files changed, 9 insertions(+), 5 deletions(-) diff --git a/scripts/genksyms/genksyms.c b/scripts/genksyms/genksyms.c index 8ca46f807b57..c5e8e0e0f949 100644 --- a/scripts/genksyms/genksyms.c +++ b/scripts/genksyms/genksyms.c @@ -272,11 +272,15 @@ static struct symbol *__add_symbol(const char *name, enum symbol_type type, break; } } + + free_list(sym->defn, NULL); + free(sym->name); + free(sym); --nsyms; } sym = xmalloc(sizeof(*sym)); - sym->name = name; + sym->name = xstrdup(name); sym->type = type; sym->defn = defn; sym->expansion_trail = NULL; @@ -483,7 +487,7 @@ static void read_reference(FILE *f) defn = def; def = read_node(f); } - subsym = add_reference_symbol(xstrdup(sym->string), sym->tag, + subsym = add_reference_symbol(sym->string, sym->tag, defn, is_extern); subsym->is_override = is_override; free_node(sym); diff --git a/scripts/genksyms/genksyms.h b/scripts/genksyms/genksyms.h index 21ed2ec2d98c..5621533dcb8e 100644 --- a/scripts/genksyms/genksyms.h +++ b/scripts/genksyms/genksyms.h @@ -32,7 +32,7 @@ struct string_list { struct symbol { struct symbol *hash_next; - const char *name; + char *name; enum symbol_type type; struct string_list *defn; struct symbol *expansion_trail; diff --git a/scripts/genksyms/parse.y b/scripts/genksyms/parse.y index 840371d01bf4..689cb6bb40b6 100644 --- a/scripts/genksyms/parse.y +++ b/scripts/genksyms/parse.y @@ -482,12 +482,12 @@ enumerator_list: enumerator: IDENT { - const char *name = strdup((*$1)->string); + const char *name = (*$1)->string; add_symbol(name, SYM_ENUM_CONST, NULL, 0); } | IDENT '=' EXPRESSION_PHRASE { - const char *name = strdup((*$1)->string); + const char *name = (*$1)->string; struct string_list *expr = copy_list_range(*$3, *$2); add_symbol(name, SYM_ENUM_CONST, expr, 0); } From f034d186bf9e2857079815e5490e2810a1a287a6 Mon Sep 17 00:00:00 2001 From: Masahiro Yamada Date: Fri, 3 Jan 2025 16:30:40 +0900 Subject: [PATCH 15/41] genksyms: reduce the indentation in the for-loop in __add_symbol() To improve readability, reduce the indentation as follows: - Use 'continue' earlier when the symbol does not match - flip !sym->is_declared to flatten the if-else chain No functional changes are intended. Signed-off-by: Masahiro Yamada --- scripts/genksyms/genksyms.c | 63 ++++++++++++++++++------------------- 1 file changed, 30 insertions(+), 33 deletions(-) diff --git a/scripts/genksyms/genksyms.c b/scripts/genksyms/genksyms.c index c5e8e0e0f949..5a90acd693f4 100644 --- a/scripts/genksyms/genksyms.c +++ b/scripts/genksyms/genksyms.c @@ -226,41 +226,38 @@ static struct symbol *__add_symbol(const char *name, enum symbol_type type, h = crc32(name) % HASH_BUCKETS; for (sym = symtab[h]; sym; sym = sym->hash_next) { - if (map_to_ns(sym->type) == map_to_ns(type) && - strcmp(name, sym->name) == 0) { - if (is_reference) - /* fall through */ ; - else if (sym->type == type && - equal_list(sym->defn, defn)) { - if (!sym->is_declared && sym->is_override) { - print_location(); - print_type_name(type, name); - fprintf(stderr, " modversion is " - "unchanged\n"); - } - sym->is_declared = 1; - free_list(defn, NULL); - return sym; - } else if (!sym->is_declared) { - if (sym->is_override && flag_preserve) { - print_location(); - fprintf(stderr, "ignoring "); - print_type_name(type, name); - fprintf(stderr, " modversion change\n"); - sym->is_declared = 1; - free_list(defn, NULL); - return sym; - } else { - status = is_unknown_symbol(sym) ? - STATUS_DEFINED : STATUS_MODIFIED; - } - } else { - error_with_pos("redefinition of %s", name); - free_list(defn, NULL); - return sym; + if (map_to_ns(sym->type) != map_to_ns(type) || + strcmp(name, sym->name)) + continue; + + if (is_reference) { + /* fall through */ ; + } else if (sym->type == type && equal_list(sym->defn, defn)) { + if (!sym->is_declared && sym->is_override) { + print_location(); + print_type_name(type, name); + fprintf(stderr, " modversion is unchanged\n"); } - break; + sym->is_declared = 1; + free_list(defn, NULL); + return sym; + } else if (sym->is_declared) { + error_with_pos("redefinition of %s", name); + free_list(defn, NULL); + return sym; + } else if (sym->is_override && flag_preserve) { + print_location(); + fprintf(stderr, "ignoring "); + print_type_name(type, name); + fprintf(stderr, " modversion change\n"); + sym->is_declared = 1; + free_list(defn, NULL); + return sym; + } else { + status = is_unknown_symbol(sym) ? + STATUS_DEFINED : STATUS_MODIFIED; } + break; } if (sym) { From 2480f53f21b21eb24a33815d4623f54fdb30cf27 Mon Sep 17 00:00:00 2001 From: Masahiro Yamada Date: Fri, 3 Jan 2025 16:30:41 +0900 Subject: [PATCH 16/41] genksyms: refactor the return points in the for-loop in __add_symbol() free_list() must be called before returning from this for-loop. Swap 'break' and the combination of free_list() and 'return'. This reduces the code and minimizes the risk of introducing memory leaks in future changes. Signed-off-by: Masahiro Yamada --- scripts/genksyms/genksyms.c | 12 ++++-------- 1 file changed, 4 insertions(+), 8 deletions(-) diff --git a/scripts/genksyms/genksyms.c b/scripts/genksyms/genksyms.c index 5a90acd693f4..41d6cfce0088 100644 --- a/scripts/genksyms/genksyms.c +++ b/scripts/genksyms/genksyms.c @@ -231,7 +231,7 @@ static struct symbol *__add_symbol(const char *name, enum symbol_type type, continue; if (is_reference) { - /* fall through */ ; + break; } else if (sym->type == type && equal_list(sym->defn, defn)) { if (!sym->is_declared && sym->is_override) { print_location(); @@ -239,25 +239,21 @@ static struct symbol *__add_symbol(const char *name, enum symbol_type type, fprintf(stderr, " modversion is unchanged\n"); } sym->is_declared = 1; - free_list(defn, NULL); - return sym; } else if (sym->is_declared) { error_with_pos("redefinition of %s", name); - free_list(defn, NULL); - return sym; } else if (sym->is_override && flag_preserve) { print_location(); fprintf(stderr, "ignoring "); print_type_name(type, name); fprintf(stderr, " modversion change\n"); sym->is_declared = 1; - free_list(defn, NULL); - return sym; } else { status = is_unknown_symbol(sym) ? STATUS_DEFINED : STATUS_MODIFIED; + break; } - break; + free_list(defn, NULL); + return sym; } if (sym) { From 2759bd908f3cc8d286e1fa64ec7ee7f5d1124837 Mon Sep 17 00:00:00 2001 From: Masahiro Yamada Date: Fri, 3 Jan 2025 16:30:42 +0900 Subject: [PATCH 17/41] genksyms: use generic macros for hash table implementation Use macros provided by hashtable.h Signed-off-by: Masahiro Yamada --- scripts/genksyms/genksyms.c | 32 ++++++++++++-------------------- scripts/genksyms/genksyms.h | 4 +++- 2 files changed, 15 insertions(+), 21 deletions(-) diff --git a/scripts/genksyms/genksyms.c b/scripts/genksyms/genksyms.c index 41d6cfce0088..e2cd3dcb469f 100644 --- a/scripts/genksyms/genksyms.c +++ b/scripts/genksyms/genksyms.c @@ -18,12 +18,12 @@ #include #include +#include + #include "genksyms.h" /*----------------------------------------------------------------------*/ -#define HASH_BUCKETS 4096 - -static struct symbol *symtab[HASH_BUCKETS]; +static HASHTABLE_DEFINE(symbol_hashtable, 1U << 12); static FILE *debugfile; int cur_line = 1; @@ -151,14 +151,14 @@ static enum symbol_type map_to_ns(enum symbol_type t) struct symbol *find_symbol(const char *name, enum symbol_type ns, int exact) { - unsigned long h = crc32(name) % HASH_BUCKETS; struct symbol *sym; - for (sym = symtab[h]; sym; sym = sym->hash_next) + hash_for_each_possible(symbol_hashtable, sym, hnode, crc32(name)) { if (map_to_ns(sym->type) == map_to_ns(ns) && strcmp(name, sym->name) == 0 && sym->is_declared) break; + } if (exact && sym && sym->type != ns) return NULL; @@ -224,8 +224,8 @@ static struct symbol *__add_symbol(const char *name, enum symbol_type type, return NULL; } - h = crc32(name) % HASH_BUCKETS; - for (sym = symtab[h]; sym; sym = sym->hash_next) { + h = crc32(name); + hash_for_each_possible(symbol_hashtable, sym, hnode, h) { if (map_to_ns(sym->type) != map_to_ns(type) || strcmp(name, sym->name)) continue; @@ -257,14 +257,7 @@ static struct symbol *__add_symbol(const char *name, enum symbol_type type, } if (sym) { - struct symbol **psym; - - for (psym = &symtab[h]; *psym; psym = &(*psym)->hash_next) { - if (*psym == sym) { - *psym = sym->hash_next; - break; - } - } + hash_del(&sym->hnode); free_list(sym->defn, NULL); free(sym->name); @@ -280,8 +273,7 @@ static struct symbol *__add_symbol(const char *name, enum symbol_type type, sym->visited = NULL; sym->is_extern = is_extern; - sym->hash_next = symtab[h]; - symtab[h] = sym; + hash_add(symbol_hashtable, &sym->hnode, h); sym->is_declared = !is_reference; sym->status = status; @@ -832,9 +824,9 @@ int main(int argc, char **argv) } if (flag_debug) { - fprintf(debugfile, "Hash table occupancy %d/%d = %g\n", - nsyms, HASH_BUCKETS, - (double)nsyms / (double)HASH_BUCKETS); + fprintf(debugfile, "Hash table occupancy %d/%zd = %g\n", + nsyms, HASH_SIZE(symbol_hashtable), + (double)nsyms / HASH_SIZE(symbol_hashtable)); } if (dumpfile) diff --git a/scripts/genksyms/genksyms.h b/scripts/genksyms/genksyms.h index 5621533dcb8e..8c45ada59ece 100644 --- a/scripts/genksyms/genksyms.h +++ b/scripts/genksyms/genksyms.h @@ -14,6 +14,8 @@ #include +#include + enum symbol_type { SYM_NORMAL, SYM_TYPEDEF, SYM_ENUM, SYM_STRUCT, SYM_UNION, SYM_ENUM_CONST @@ -31,7 +33,7 @@ struct string_list { }; struct symbol { - struct symbol *hash_next; + struct hlist_node hnode; char *name; enum symbol_type type; struct string_list *defn; From a56fece7f302ff1eb49535e66bdd5d03ced0ca20 Mon Sep 17 00:00:00 2001 From: Masahiro Yamada Date: Fri, 3 Jan 2025 16:30:43 +0900 Subject: [PATCH 18/41] genksyms: use uint32_t instead of unsigned long for calculating CRC Currently, 'unsigned long' is used for intermediate variables when calculating CRCs. The size of 'long' differs depending on the architecture: it is 32 bits on 32-bit architectures and 64 bits on 64-bit architectures. The CRC values generated by genksyms represent the compatibility of exported symbols. Therefore, reproducibility is important. In other words, we need to ensure that the output is the same when the kernel source is identical, regardless of whether genksyms is running on a 32-bit or 64-bit build machine. Fortunately, the output from genksyms is not affected by the build machine's architecture because only the lower 32 bits of the 'unsigned long' variables are used. To make it even clearer that the CRC calculation is independent of the build machine's architecture, this commit explicitly uses the fixed-width type, uint32_t. Signed-off-by: Masahiro Yamada --- scripts/genksyms/genksyms.c | 15 ++++++++------- 1 file changed, 8 insertions(+), 7 deletions(-) diff --git a/scripts/genksyms/genksyms.c b/scripts/genksyms/genksyms.c index e2cd3dcb469f..8b0d7ac73dbb 100644 --- a/scripts/genksyms/genksyms.c +++ b/scripts/genksyms/genksyms.c @@ -12,6 +12,7 @@ #include #include +#include #include #include #include @@ -60,7 +61,7 @@ static void print_type_name(enum symbol_type type, const char *name); /*----------------------------------------------------------------------*/ -static const unsigned int crctab32[] = { +static const uint32_t crctab32[] = { 0x00000000U, 0x77073096U, 0xee0e612cU, 0x990951baU, 0x076dc419U, 0x706af48fU, 0xe963a535U, 0x9e6495a3U, 0x0edb8832U, 0x79dcb8a4U, 0xe0d5e91eU, 0x97d2d988U, 0x09b64c2bU, 0x7eb17cbdU, 0xe7b82d07U, @@ -115,19 +116,19 @@ static const unsigned int crctab32[] = { 0x2d02ef8dU }; -static unsigned long partial_crc32_one(unsigned char c, unsigned long crc) +static uint32_t partial_crc32_one(uint8_t c, uint32_t crc) { return crctab32[(crc ^ c) & 0xff] ^ (crc >> 8); } -static unsigned long partial_crc32(const char *s, unsigned long crc) +static uint32_t partial_crc32(const char *s, uint32_t crc) { while (*s) crc = partial_crc32_one(*s++, crc); return crc; } -static unsigned long crc32(const char *s) +static uint32_t crc32(const char *s) { return partial_crc32(s, 0xffffffff) ^ 0xffffffff; } @@ -517,7 +518,7 @@ static void print_list(FILE * f, struct string_list *list) } } -static unsigned long expand_and_crc_sym(struct symbol *sym, unsigned long crc) +static uint32_t expand_and_crc_sym(struct symbol *sym, uint32_t crc) { struct string_list *list = sym->defn; struct string_list **e, **b; @@ -624,7 +625,7 @@ static unsigned long expand_and_crc_sym(struct symbol *sym, unsigned long crc) void export_symbol(const char *name) { struct symbol *sym; - unsigned long crc; + uint32_t crc; int has_changed = 0; sym = find_symbol(name, SYM_NORMAL, 0); @@ -672,7 +673,7 @@ void export_symbol(const char *name) if (flag_dump_defs) fputs(">\n", debugfile); - printf("#SYMVER %s 0x%08lx\n", name, crc); + printf("#SYMVER %s 0x%08lx\n", name, (unsigned long)crc); } /*----------------------------------------------------------------------*/ From f28568841ae0a0dd48dfc5400aaebedf10a54d10 Mon Sep 17 00:00:00 2001 From: Sami Tolvanen Date: Fri, 3 Jan 2025 20:45:23 +0000 Subject: [PATCH 19/41] tools: Add gendwarfksyms Add a basic DWARF parser, which uses libdw to traverse the debugging information in an object file and looks for functions and variables. In follow-up patches, this will be expanded to produce symbol versions for CONFIG_MODVERSIONS from DWARF. Signed-off-by: Sami Tolvanen Reviewed-by: Petr Pavlu Signed-off-by: Masahiro Yamada --- MAINTAINERS | 7 ++ kernel/module/Kconfig | 8 ++ scripts/Makefile | 1 + scripts/gendwarfksyms/.gitignore | 2 + scripts/gendwarfksyms/Makefile | 8 ++ scripts/gendwarfksyms/dwarf.c | 166 ++++++++++++++++++++++++++ scripts/gendwarfksyms/gendwarfksyms.c | 128 ++++++++++++++++++++ scripts/gendwarfksyms/gendwarfksyms.h | 95 +++++++++++++++ scripts/gendwarfksyms/symbols.c | 98 +++++++++++++++ 9 files changed, 513 insertions(+) create mode 100644 scripts/gendwarfksyms/.gitignore create mode 100644 scripts/gendwarfksyms/Makefile create mode 100644 scripts/gendwarfksyms/dwarf.c create mode 100644 scripts/gendwarfksyms/gendwarfksyms.c create mode 100644 scripts/gendwarfksyms/gendwarfksyms.h create mode 100644 scripts/gendwarfksyms/symbols.c diff --git a/MAINTAINERS b/MAINTAINERS index 30cbc3d44cd5..1ec4753b1457 100644 --- a/MAINTAINERS +++ b/MAINTAINERS @@ -9550,6 +9550,13 @@ W: https://linuxtv.org T: git git://linuxtv.org/media.git F: drivers/media/radio/radio-gemtek* +GENDWARFKSYMS +M: Sami Tolvanen +L: linux-modules@vger.kernel.org +L: linux-kbuild@vger.kernel.org +S: Maintained +F: scripts/gendwarfksyms/ + GENERIC ARCHITECTURE TOPOLOGY M: Sudeep Holla L: linux-kernel@vger.kernel.org diff --git a/kernel/module/Kconfig b/kernel/module/Kconfig index 7b329057997a..4637f063d0fc 100644 --- a/kernel/module/Kconfig +++ b/kernel/module/Kconfig @@ -169,6 +169,14 @@ config MODVERSIONS make them incompatible with the kernel you are running. If unsure, say N. +config GENDWARFKSYMS + bool "gendwarfksyms (from debugging information)" + depends on DEBUG_INFO + # Requires full debugging information, split DWARF not supported. + depends on !DEBUG_INFO_REDUCED && !DEBUG_INFO_SPLIT + # Requires ELF object files. + depends on !LTO + config ASM_MODVERSIONS bool default HAVE_ASM_MODVERSIONS && MODVERSIONS diff --git a/scripts/Makefile b/scripts/Makefile index 6bcda4b9d054..d7fec46d38c0 100644 --- a/scripts/Makefile +++ b/scripts/Makefile @@ -54,6 +54,7 @@ targets += module.lds subdir-$(CONFIG_GCC_PLUGINS) += gcc-plugins subdir-$(CONFIG_MODVERSIONS) += genksyms +subdir-$(CONFIG_GENDWARFKSYMS) += gendwarfksyms subdir-$(CONFIG_SECURITY_SELINUX) += selinux subdir-$(CONFIG_SECURITY_IPE) += ipe diff --git a/scripts/gendwarfksyms/.gitignore b/scripts/gendwarfksyms/.gitignore new file mode 100644 index 000000000000..0927f8d3cd96 --- /dev/null +++ b/scripts/gendwarfksyms/.gitignore @@ -0,0 +1,2 @@ +# SPDX-License-Identifier: GPL-2.0 +/gendwarfksyms diff --git a/scripts/gendwarfksyms/Makefile b/scripts/gendwarfksyms/Makefile new file mode 100644 index 000000000000..9f8fec4fd39b --- /dev/null +++ b/scripts/gendwarfksyms/Makefile @@ -0,0 +1,8 @@ +# SPDX-License-Identifier: GPL-2.0 +hostprogs-always-y += gendwarfksyms + +gendwarfksyms-objs += gendwarfksyms.o +gendwarfksyms-objs += dwarf.o +gendwarfksyms-objs += symbols.o + +HOSTLDLIBS_gendwarfksyms := -ldw -lelf diff --git a/scripts/gendwarfksyms/dwarf.c b/scripts/gendwarfksyms/dwarf.c new file mode 100644 index 000000000000..81df3e2ad3ae --- /dev/null +++ b/scripts/gendwarfksyms/dwarf.c @@ -0,0 +1,166 @@ +// SPDX-License-Identifier: GPL-2.0 +/* + * Copyright (C) 2024 Google LLC + */ + +#include "gendwarfksyms.h" + +static bool get_ref_die_attr(Dwarf_Die *die, unsigned int id, Dwarf_Die *value) +{ + Dwarf_Attribute da; + + /* dwarf_formref_die returns a pointer instead of an error value. */ + return dwarf_attr(die, id, &da) && dwarf_formref_die(&da, value); +} + +#define DEFINE_GET_STRING_ATTR(attr) \ + static const char *get_##attr##_attr(Dwarf_Die *die) \ + { \ + Dwarf_Attribute da; \ + if (dwarf_attr(die, DW_AT_##attr, &da)) \ + return dwarf_formstring(&da); \ + return NULL; \ + } + +DEFINE_GET_STRING_ATTR(name) +DEFINE_GET_STRING_ATTR(linkage_name) + +static const char *get_symbol_name(Dwarf_Die *die) +{ + const char *name; + + /* rustc uses DW_AT_linkage_name for exported symbols */ + name = get_linkage_name_attr(die); + if (!name) + name = get_name_attr(die); + + return name; +} + +static bool match_export_symbol(struct state *state, Dwarf_Die *die) +{ + Dwarf_Die *source = die; + Dwarf_Die origin; + + /* If the DIE has an abstract origin, use it for type information. */ + if (get_ref_die_attr(die, DW_AT_abstract_origin, &origin)) + source = &origin; + + state->sym = symbol_get(get_symbol_name(die)); + + /* Look up using the origin name if there are no matches. */ + if (!state->sym && source != die) + state->sym = symbol_get(get_symbol_name(source)); + + state->die = *source; + return !!state->sym; +} + +/* + * Type string processing + */ +static void process(const char *s) +{ + s = s ?: ""; + + if (dump_dies) + fputs(s, stderr); +} + +bool match_all(Dwarf_Die *die) +{ + return true; +} + +int process_die_container(struct state *state, Dwarf_Die *die, + die_callback_t func, die_match_callback_t match) +{ + Dwarf_Die current; + int res; + + res = checkp(dwarf_child(die, ¤t)); + while (!res) { + if (match(¤t)) { + /* <0 = error, 0 = continue, >0 = stop */ + res = checkp(func(state, ¤t)); + if (res) + return res; + } + + res = checkp(dwarf_siblingof(¤t, ¤t)); + } + + return 0; +} + +/* + * Exported symbol processing + */ +static void process_symbol(struct state *state, Dwarf_Die *die, + die_callback_t process_func) +{ + debug("%s", state->sym->name); + check(process_func(state, die)); + if (dump_dies) + fputs("\n", stderr); +} + +static int __process_subprogram(struct state *state, Dwarf_Die *die) +{ + process("subprogram"); + return 0; +} + +static void process_subprogram(struct state *state, Dwarf_Die *die) +{ + process_symbol(state, die, __process_subprogram); +} + +static int __process_variable(struct state *state, Dwarf_Die *die) +{ + process("variable "); + return 0; +} + +static void process_variable(struct state *state, Dwarf_Die *die) +{ + process_symbol(state, die, __process_variable); +} + +static int process_exported_symbols(struct state *unused, Dwarf_Die *die) +{ + int tag = dwarf_tag(die); + + switch (tag) { + /* Possible containers of exported symbols */ + case DW_TAG_namespace: + case DW_TAG_class_type: + case DW_TAG_structure_type: + return check(process_die_container( + NULL, die, process_exported_symbols, match_all)); + + /* Possible exported symbols */ + case DW_TAG_subprogram: + case DW_TAG_variable: { + struct state state; + + if (!match_export_symbol(&state, die)) + return 0; + + if (tag == DW_TAG_subprogram) + process_subprogram(&state, &state.die); + else + process_variable(&state, &state.die); + + return 0; + } + default: + return 0; + } +} + +void process_cu(Dwarf_Die *cudie) +{ + check(process_die_container(NULL, cudie, process_exported_symbols, + match_all)); +} diff --git a/scripts/gendwarfksyms/gendwarfksyms.c b/scripts/gendwarfksyms/gendwarfksyms.c new file mode 100644 index 000000000000..a1d13353c6bc --- /dev/null +++ b/scripts/gendwarfksyms/gendwarfksyms.c @@ -0,0 +1,128 @@ +// SPDX-License-Identifier: GPL-2.0 +/* + * Copyright (C) 2024 Google LLC + */ + +#include +#include +#include +#include +#include +#include +#include "gendwarfksyms.h" + +/* + * Options + */ + +/* Print debugging information to stderr */ +int debug; +/* Dump DIE contents */ +int dump_dies; + +static void usage(void) +{ + fputs("Usage: gendwarfksyms [options] elf-object-file ... < symbol-list\n\n" + "Options:\n" + " -d, --debug Print debugging information\n" + " --dump-dies Dump DWARF DIE contents\n" + " -h, --help Print this message\n" + "\n", + stderr); +} + +static int process_module(Dwfl_Module *mod, void **userdata, const char *name, + Dwarf_Addr base, void *arg) +{ + Dwarf_Addr dwbias; + Dwarf_Die cudie; + Dwarf_CU *cu = NULL; + Dwarf *dbg; + int res; + + debug("%s", name); + dbg = dwfl_module_getdwarf(mod, &dwbias); + + do { + res = dwarf_get_units(dbg, cu, &cu, NULL, NULL, &cudie, NULL); + if (res < 0) + error("dwarf_get_units failed: no debugging information?"); + if (res == 1) + break; /* No more units */ + + process_cu(&cudie); + } while (cu); + + return DWARF_CB_OK; +} + +static const Dwfl_Callbacks callbacks = { + .section_address = dwfl_offline_section_address, + .find_debuginfo = dwfl_standard_find_debuginfo, +}; + +int main(int argc, char **argv) +{ + unsigned int n; + int opt; + + static const struct option opts[] = { + { "debug", 0, NULL, 'd' }, + { "dump-dies", 0, &dump_dies, 1 }, + { "help", 0, NULL, 'h' }, + { 0, 0, NULL, 0 } + }; + + while ((opt = getopt_long(argc, argv, "dh", opts, NULL)) != EOF) { + switch (opt) { + case 0: + break; + case 'd': + debug = 1; + break; + case 'h': + usage(); + return 0; + default: + usage(); + return 1; + } + } + + if (optind >= argc) { + usage(); + error("no input files?"); + } + + symbol_read_exports(stdin); + + for (n = optind; n < argc; n++) { + Dwfl *dwfl; + int fd; + + fd = open(argv[n], O_RDONLY); + if (fd == -1) + error("open failed for '%s': %s", argv[n], + strerror(errno)); + + dwfl = dwfl_begin(&callbacks); + if (!dwfl) + error("dwfl_begin failed for '%s': %s", argv[n], + dwarf_errmsg(-1)); + + if (!dwfl_report_offline(dwfl, argv[n], argv[n], fd)) + error("dwfl_report_offline failed for '%s': %s", + argv[n], dwarf_errmsg(-1)); + + dwfl_report_end(dwfl, NULL, NULL); + + if (dwfl_getmodules(dwfl, &process_module, NULL, 0)) + error("dwfl_getmodules failed for '%s'", argv[n]); + + dwfl_end(dwfl); + } + + symbol_free(); + + return 0; +} diff --git a/scripts/gendwarfksyms/gendwarfksyms.h b/scripts/gendwarfksyms/gendwarfksyms.h new file mode 100644 index 000000000000..5c8288c71fdd --- /dev/null +++ b/scripts/gendwarfksyms/gendwarfksyms.h @@ -0,0 +1,95 @@ +/* SPDX-License-Identifier: GPL-2.0 */ +/* + * Copyright (C) 2024 Google LLC + */ + +#include +#include +#include +#include +#include + +#include +#include +#include + +#ifndef __GENDWARFKSYMS_H +#define __GENDWARFKSYMS_H + +/* + * Options -- in gendwarfksyms.c + */ +extern int debug; +extern int dump_dies; + +/* + * Output helpers + */ +#define __PREFIX "gendwarfksyms: " +#define __println(prefix, format, ...) \ + fprintf(stderr, prefix __PREFIX "%s: " format "\n", __func__, \ + ##__VA_ARGS__) + +#define debug(format, ...) \ + do { \ + if (debug) \ + __println("", format, ##__VA_ARGS__); \ + } while (0) + +#define warn(format, ...) __println("warning: ", format, ##__VA_ARGS__) +#define error(format, ...) \ + do { \ + __println("error: ", format, ##__VA_ARGS__); \ + exit(1); \ + } while (0) + +/* + * Error handling helpers + */ +#define __check(expr, test) \ + ({ \ + int __res = expr; \ + if (test) \ + error("`%s` failed: %d", #expr, __res); \ + __res; \ + }) + +/* Error == non-zero values */ +#define check(expr) __check(expr, __res) +/* Error == negative values */ +#define checkp(expr) __check(expr, __res < 0) + +/* + * symbols.c + */ + +struct symbol { + const char *name; + struct hlist_node name_hash; +}; + +typedef void (*symbol_callback_t)(struct symbol *, void *arg); + +void symbol_read_exports(FILE *file); +struct symbol *symbol_get(const char *name); +void symbol_free(void); + +/* + * dwarf.c + */ + +struct state { + struct symbol *sym; + Dwarf_Die die; +}; + +typedef int (*die_callback_t)(struct state *state, Dwarf_Die *die); +typedef bool (*die_match_callback_t)(Dwarf_Die *die); +bool match_all(Dwarf_Die *die); + +int process_die_container(struct state *state, Dwarf_Die *die, + die_callback_t func, die_match_callback_t match); + +void process_cu(Dwarf_Die *cudie); + +#endif /* __GENDWARFKSYMS_H */ diff --git a/scripts/gendwarfksyms/symbols.c b/scripts/gendwarfksyms/symbols.c new file mode 100644 index 000000000000..592eacf72694 --- /dev/null +++ b/scripts/gendwarfksyms/symbols.c @@ -0,0 +1,98 @@ +// SPDX-License-Identifier: GPL-2.0 +/* + * Copyright (C) 2024 Google LLC + */ + +#include "gendwarfksyms.h" + +#define SYMBOL_HASH_BITS 12 + +/* name -> struct symbol */ +static HASHTABLE_DEFINE(symbol_names, 1 << SYMBOL_HASH_BITS); + +static unsigned int for_each(const char *name, symbol_callback_t func, + void *data) +{ + struct hlist_node *tmp; + struct symbol *match; + + if (!name || !*name) + return 0; + + hash_for_each_possible_safe(symbol_names, match, tmp, name_hash, + hash_str(name)) { + if (strcmp(match->name, name)) + continue; + + if (func) + func(match, data); + + return 1; + } + + return 0; +} + +static bool is_exported(const char *name) +{ + return for_each(name, NULL, NULL) > 0; +} + +void symbol_read_exports(FILE *file) +{ + struct symbol *sym; + char *line = NULL; + char *name = NULL; + size_t size = 0; + int nsym = 0; + + while (getline(&line, &size, file) > 0) { + if (sscanf(line, "%ms\n", &name) != 1) + error("malformed input line: %s", line); + + if (is_exported(name)) { + /* Ignore duplicates */ + free(name); + continue; + } + + sym = xcalloc(1, sizeof(struct symbol)); + sym->name = name; + + hash_add(symbol_names, &sym->name_hash, hash_str(sym->name)); + ++nsym; + + debug("%s", sym->name); + } + + free(line); + debug("%d exported symbols", nsym); +} + +static void get_symbol(struct symbol *sym, void *arg) +{ + struct symbol **res = arg; + + *res = sym; +} + +struct symbol *symbol_get(const char *name) +{ + struct symbol *sym = NULL; + + for_each(name, get_symbol, &sym); + return sym; +} + +void symbol_free(void) +{ + struct hlist_node *tmp; + struct symbol *sym; + + hash_for_each_safe(symbol_names, sym, tmp, name_hash) { + free((void *)sym->name); + free(sym); + } + + hash_init(symbol_names); +} From e982abf43749529687dd1d07fa4f495902910cf2 Mon Sep 17 00:00:00 2001 From: Sami Tolvanen Date: Fri, 3 Jan 2025 20:45:24 +0000 Subject: [PATCH 20/41] gendwarfksyms: Add address matching The compiler may choose not to emit type information in DWARF for all aliases, but it's possible for each alias to be exported separately. To ensure we find type information for the aliases as well, read {section, address} tuples from the symbol table and match symbols also by address. Signed-off-by: Sami Tolvanen Reviewed-by: Petr Pavlu Signed-off-by: Masahiro Yamada --- scripts/gendwarfksyms/gendwarfksyms.c | 2 + scripts/gendwarfksyms/gendwarfksyms.h | 13 +++ scripts/gendwarfksyms/symbols.c | 161 ++++++++++++++++++++++++++ 3 files changed, 176 insertions(+) diff --git a/scripts/gendwarfksyms/gendwarfksyms.c b/scripts/gendwarfksyms/gendwarfksyms.c index a1d13353c6bc..cd8bfe973a5c 100644 --- a/scripts/gendwarfksyms/gendwarfksyms.c +++ b/scripts/gendwarfksyms/gendwarfksyms.c @@ -105,6 +105,8 @@ int main(int argc, char **argv) error("open failed for '%s': %s", argv[n], strerror(errno)); + symbol_read_symtab(fd); + dwfl = dwfl_begin(&callbacks); if (!dwfl) error("dwfl_begin failed for '%s': %s", argv[n], diff --git a/scripts/gendwarfksyms/gendwarfksyms.h b/scripts/gendwarfksyms/gendwarfksyms.h index 5c8288c71fdd..cb9fd78a58da 100644 --- a/scripts/gendwarfksyms/gendwarfksyms.h +++ b/scripts/gendwarfksyms/gendwarfksyms.h @@ -63,14 +63,27 @@ extern int dump_dies; * symbols.c */ +static inline unsigned int addr_hash(uintptr_t addr) +{ + return hash_ptr((const void *)addr); +} + +struct symbol_addr { + uint32_t section; + Elf64_Addr address; +}; + struct symbol { const char *name; + struct symbol_addr addr; + struct hlist_node addr_hash; struct hlist_node name_hash; }; typedef void (*symbol_callback_t)(struct symbol *, void *arg); void symbol_read_exports(FILE *file); +void symbol_read_symtab(int fd); struct symbol *symbol_get(const char *name); void symbol_free(void); diff --git a/scripts/gendwarfksyms/symbols.c b/scripts/gendwarfksyms/symbols.c index 592eacf72694..98febb524dd5 100644 --- a/scripts/gendwarfksyms/symbols.c +++ b/scripts/gendwarfksyms/symbols.c @@ -7,9 +7,38 @@ #define SYMBOL_HASH_BITS 12 +/* struct symbol_addr -> struct symbol */ +static HASHTABLE_DEFINE(symbol_addrs, 1 << SYMBOL_HASH_BITS); /* name -> struct symbol */ static HASHTABLE_DEFINE(symbol_names, 1 << SYMBOL_HASH_BITS); +static inline unsigned int symbol_addr_hash(const struct symbol_addr *addr) +{ + return hash_32(addr->section ^ addr_hash(addr->address)); +} + +static unsigned int __for_each_addr(struct symbol *sym, symbol_callback_t func, + void *data) +{ + struct hlist_node *tmp; + struct symbol *match = NULL; + unsigned int processed = 0; + + hash_for_each_possible_safe(symbol_addrs, match, tmp, addr_hash, + symbol_addr_hash(&sym->addr)) { + if (match == sym) + continue; /* Already processed */ + + if (match->addr.section == sym->addr.section && + match->addr.address == sym->addr.address) { + func(match, data); + ++processed; + } + } + + return processed; +} + static unsigned int for_each(const char *name, symbol_callback_t func, void *data) { @@ -24,9 +53,13 @@ static unsigned int for_each(const char *name, symbol_callback_t func, if (strcmp(match->name, name)) continue; + /* Call func for the match, and all address matches */ if (func) func(match, data); + if (match->addr.section != SHN_UNDEF) + return __for_each_addr(match, func, data) + 1; + return 1; } @@ -58,6 +91,7 @@ void symbol_read_exports(FILE *file) sym = xcalloc(1, sizeof(struct symbol)); sym->name = name; + sym->addr.section = SHN_UNDEF; hash_add(symbol_names, &sym->name_hash, hash_str(sym->name)); ++nsym; @@ -84,6 +118,132 @@ struct symbol *symbol_get(const char *name) return sym; } +typedef void (*elf_symbol_callback_t)(const char *name, GElf_Sym *sym, + Elf32_Word xndx, void *arg); + +static void elf_for_each_global(int fd, elf_symbol_callback_t func, void *arg) +{ + size_t sym_size; + GElf_Shdr shdr_mem; + GElf_Shdr *shdr; + Elf_Data *xndx_data = NULL; + Elf_Scn *scn; + Elf *elf; + + if (elf_version(EV_CURRENT) != EV_CURRENT) + error("elf_version failed: %s", elf_errmsg(-1)); + + elf = elf_begin(fd, ELF_C_READ_MMAP, NULL); + if (!elf) + error("elf_begin failed: %s", elf_errmsg(-1)); + + scn = elf_nextscn(elf, NULL); + + while (scn) { + shdr = gelf_getshdr(scn, &shdr_mem); + if (!shdr) + error("gelf_getshdr failed: %s", elf_errmsg(-1)); + + if (shdr->sh_type == SHT_SYMTAB_SHNDX) { + xndx_data = elf_getdata(scn, NULL); + if (!xndx_data) + error("elf_getdata failed: %s", elf_errmsg(-1)); + break; + } + + scn = elf_nextscn(elf, scn); + } + + sym_size = gelf_fsize(elf, ELF_T_SYM, 1, EV_CURRENT); + scn = elf_nextscn(elf, NULL); + + while (scn) { + shdr = gelf_getshdr(scn, &shdr_mem); + if (!shdr) + error("gelf_getshdr failed: %s", elf_errmsg(-1)); + + if (shdr->sh_type == SHT_SYMTAB) { + unsigned int nsyms; + unsigned int n; + Elf_Data *data = elf_getdata(scn, NULL); + + if (!data) + error("elf_getdata failed: %s", elf_errmsg(-1)); + + if (shdr->sh_entsize != sym_size) + error("expected sh_entsize (%lu) to be %zu", + shdr->sh_entsize, sym_size); + + nsyms = shdr->sh_size / shdr->sh_entsize; + + for (n = 1; n < nsyms; ++n) { + const char *name = NULL; + Elf32_Word xndx = 0; + GElf_Sym sym_mem; + GElf_Sym *sym; + + sym = gelf_getsymshndx(data, xndx_data, n, + &sym_mem, &xndx); + if (!sym) + error("gelf_getsymshndx failed: %s", + elf_errmsg(-1)); + + if (GELF_ST_BIND(sym->st_info) == STB_LOCAL) + continue; + + if (sym->st_shndx != SHN_XINDEX) + xndx = sym->st_shndx; + + name = elf_strptr(elf, shdr->sh_link, + sym->st_name); + if (!name) + error("elf_strptr failed: %s", + elf_errmsg(-1)); + + /* Skip empty symbol names */ + if (*name) + func(name, sym, xndx, arg); + } + } + + scn = elf_nextscn(elf, scn); + } + + check(elf_end(elf)); +} + +static void set_symbol_addr(struct symbol *sym, void *arg) +{ + struct symbol_addr *addr = arg; + + if (sym->addr.section == SHN_UNDEF) { + sym->addr = *addr; + hash_add(symbol_addrs, &sym->addr_hash, + symbol_addr_hash(&sym->addr)); + + debug("%s -> { %u, %lx }", sym->name, sym->addr.section, + sym->addr.address); + } else if (sym->addr.section != addr->section || + sym->addr.address != addr->address) { + warn("multiple addresses for symbol %s?", sym->name); + } +} + +static void elf_set_symbol_addr(const char *name, GElf_Sym *sym, + Elf32_Word xndx, void *arg) +{ + struct symbol_addr addr = { .section = xndx, .address = sym->st_value }; + + /* Set addresses for exported symbols */ + if (addr.section != SHN_UNDEF) + for_each(name, set_symbol_addr, &addr); +} + +void symbol_read_symtab(int fd) +{ + elf_for_each_global(fd, elf_set_symbol_addr, NULL); +} + void symbol_free(void) { struct hlist_node *tmp; @@ -94,5 +254,6 @@ void symbol_free(void) free(sym); } + hash_init(symbol_addrs); hash_init(symbol_names); } From 5b7780e86857f70249df2c4f8982cad3ba931eee Mon Sep 17 00:00:00 2001 From: Sami Tolvanen Date: Fri, 3 Jan 2025 20:45:25 +0000 Subject: [PATCH 21/41] gendwarfksyms: Expand base_type Start making gendwarfksyms more useful by adding support for expanding DW_TAG_base_type types and basic DWARF attributes. Example: $ echo loops_per_jiffy | \ scripts/gendwarfksyms/gendwarfksyms \ --debug --dump-dies vmlinux.o ... gendwarfksyms: process_symbol: loops_per_jiffy variable base_type unsigned long byte_size(8) encoding(7) ... Signed-off-by: Sami Tolvanen Reviewed-by: Petr Pavlu Signed-off-by: Masahiro Yamada --- scripts/gendwarfksyms/dwarf.c | 160 ++++++++++++++++++++++++++++++++++ 1 file changed, 160 insertions(+) diff --git a/scripts/gendwarfksyms/dwarf.c b/scripts/gendwarfksyms/dwarf.c index 81df3e2ad3ae..74e75b8ec891 100644 --- a/scripts/gendwarfksyms/dwarf.c +++ b/scripts/gendwarfksyms/dwarf.c @@ -3,8 +3,21 @@ * Copyright (C) 2024 Google LLC */ +#include +#include #include "gendwarfksyms.h" +#define DEFINE_GET_ATTR(attr, type) \ + static bool get_##attr##_attr(Dwarf_Die *die, unsigned int id, \ + type *value) \ + { \ + Dwarf_Attribute da; \ + return dwarf_attr(die, id, &da) && \ + !dwarf_form##attr(&da, value); \ + } + +DEFINE_GET_ATTR(udata, Dwarf_Word) + static bool get_ref_die_attr(Dwarf_Die *die, unsigned int id, Dwarf_Die *value) { Dwarf_Attribute da; @@ -67,6 +80,109 @@ static void process(const char *s) fputs(s, stderr); } +#define MAX_FMT_BUFFER_SIZE 128 + +static void process_fmt(const char *fmt, ...) +{ + char buf[MAX_FMT_BUFFER_SIZE]; + va_list args; + + va_start(args, fmt); + + if (checkp(vsnprintf(buf, sizeof(buf), fmt, args)) >= sizeof(buf)) + error("vsnprintf overflow: increase MAX_FMT_BUFFER_SIZE"); + + process(buf); + va_end(args); +} + +#define MAX_FQN_SIZE 64 + +/* Get a fully qualified name from DWARF scopes */ +static char *get_fqn(Dwarf_Die *die) +{ + const char *list[MAX_FQN_SIZE]; + Dwarf_Die *scopes = NULL; + bool has_name = false; + char *fqn = NULL; + char *p; + int count = 0; + int len = 0; + int res; + int i; + + res = checkp(dwarf_getscopes_die(die, &scopes)); + if (!res) { + list[count] = get_name_attr(die); + + if (!list[count]) + return NULL; + + len += strlen(list[count]); + count++; + + goto done; + } + + for (i = res - 1; i >= 0 && count < MAX_FQN_SIZE; i--) { + if (dwarf_tag(&scopes[i]) == DW_TAG_compile_unit) + continue; + + list[count] = get_name_attr(&scopes[i]); + + if (list[count]) { + has_name = true; + } else { + list[count] = ""; + has_name = false; + } + + len += strlen(list[count]); + count++; + + if (i > 0) { + list[count++] = "::"; + len += 2; + } + } + + free(scopes); + + if (count == MAX_FQN_SIZE) + warn("increase MAX_FQN_SIZE: reached the maximum"); + + /* Consider the DIE unnamed if the last scope doesn't have a name */ + if (!has_name) + return NULL; +done: + fqn = xmalloc(len + 1); + *fqn = '\0'; + + p = fqn; + for (i = 0; i < count; i++) + p = stpcpy(p, list[i]); + + return fqn; +} + +static void process_fqn(Dwarf_Die *die) +{ + process(" "); + process(get_fqn(die) ?: ""); +} + +#define DEFINE_PROCESS_UDATA_ATTRIBUTE(attribute) \ + static void process_##attribute##_attr(Dwarf_Die *die) \ + { \ + Dwarf_Word value; \ + if (get_udata_attr(die, DW_AT_##attribute, &value)) \ + process_fmt(" " #attribute "(%" PRIu64 ")", value); \ + } + +DEFINE_PROCESS_UDATA_ATTRIBUTE(alignment) +DEFINE_PROCESS_UDATA_ATTRIBUTE(byte_size) +DEFINE_PROCESS_UDATA_ATTRIBUTE(encoding) + bool match_all(Dwarf_Die *die) { return true; @@ -93,6 +209,49 @@ int process_die_container(struct state *state, Dwarf_Die *die, return 0; } +static int process_type(struct state *state, Dwarf_Die *die); + +static void process_type_attr(struct state *state, Dwarf_Die *die) +{ + Dwarf_Die type; + + if (get_ref_die_attr(die, DW_AT_type, &type)) { + check(process_type(state, &type)); + return; + } + + /* Compilers can omit DW_AT_type -- print out 'void' to clarify */ + process("base_type void"); +} + +static void process_base_type(struct state *state, Dwarf_Die *die) +{ + process("base_type"); + process_fqn(die); + process_byte_size_attr(die); + process_encoding_attr(die); + process_alignment_attr(die); +} + +#define PROCESS_TYPE(type) \ + case DW_TAG_##type##_type: \ + process_##type##_type(state, die); \ + break; + +static int process_type(struct state *state, Dwarf_Die *die) +{ + int tag = dwarf_tag(die); + + switch (tag) { + PROCESS_TYPE(base) + default: + debug("unimplemented type: %x", tag); + break; + } + + return 0; +} + /* * Exported symbol processing */ @@ -119,6 +278,7 @@ static void process_subprogram(struct state *state, Dwarf_Die *die) static int __process_variable(struct state *state, Dwarf_Die *die) { process("variable "); + process_type_attr(state, die); return 0; } From 0c1c76274e88c420779c3aea077f9812bd16edaa Mon Sep 17 00:00:00 2001 From: Sami Tolvanen Date: Fri, 3 Jan 2025 20:45:26 +0000 Subject: [PATCH 22/41] gendwarfksyms: Add a cache for processed DIEs Basic types in DWARF repeat frequently and traversing the DIEs using libdw is relatively slow. Add a simple hashtable based cache for the processed DIEs. Signed-off-by: Sami Tolvanen Reviewed-by: Petr Pavlu Signed-off-by: Masahiro Yamada --- scripts/gendwarfksyms/Makefile | 1 + scripts/gendwarfksyms/die.c | 143 ++++++++++++++++++++++++++ scripts/gendwarfksyms/dwarf.c | 136 +++++++++++++++++------- scripts/gendwarfksyms/gendwarfksyms.c | 6 ++ scripts/gendwarfksyms/gendwarfksyms.h | 63 +++++++++++- 5 files changed, 308 insertions(+), 41 deletions(-) create mode 100644 scripts/gendwarfksyms/die.c diff --git a/scripts/gendwarfksyms/Makefile b/scripts/gendwarfksyms/Makefile index 9f8fec4fd39b..c0d4ce50fc27 100644 --- a/scripts/gendwarfksyms/Makefile +++ b/scripts/gendwarfksyms/Makefile @@ -2,6 +2,7 @@ hostprogs-always-y += gendwarfksyms gendwarfksyms-objs += gendwarfksyms.o +gendwarfksyms-objs += die.o gendwarfksyms-objs += dwarf.o gendwarfksyms-objs += symbols.o diff --git a/scripts/gendwarfksyms/die.c b/scripts/gendwarfksyms/die.c new file mode 100644 index 000000000000..b7d900c6a9c8 --- /dev/null +++ b/scripts/gendwarfksyms/die.c @@ -0,0 +1,143 @@ +// SPDX-License-Identifier: GPL-2.0 +/* + * Copyright (C) 2024 Google LLC + */ + +#include +#include "gendwarfksyms.h" + +#define DIE_HASH_BITS 15 + +/* {die->addr, state} -> struct die * */ +static HASHTABLE_DEFINE(die_map, 1 << DIE_HASH_BITS); + +static unsigned int map_hits; +static unsigned int map_misses; + +static inline unsigned int die_hash(uintptr_t addr, enum die_state state) +{ + return hash_32(addr_hash(addr) ^ (unsigned int)state); +} + +static void init_die(struct die *cd) +{ + cd->state = DIE_INCOMPLETE; + cd->fqn = NULL; + cd->tag = -1; + cd->addr = 0; + INIT_LIST_HEAD(&cd->fragments); +} + +static struct die *create_die(Dwarf_Die *die, enum die_state state) +{ + struct die *cd; + + cd = xmalloc(sizeof(struct die)); + init_die(cd); + cd->addr = (uintptr_t)die->addr; + + hash_add(die_map, &cd->hash, die_hash(cd->addr, state)); + return cd; +} + +int __die_map_get(uintptr_t addr, enum die_state state, struct die **res) +{ + struct die *cd; + + hash_for_each_possible(die_map, cd, hash, die_hash(addr, state)) { + if (cd->addr == addr && cd->state == state) { + *res = cd; + return 0; + } + } + + return -1; +} + +struct die *die_map_get(Dwarf_Die *die, enum die_state state) +{ + struct die *cd; + + if (__die_map_get((uintptr_t)die->addr, state, &cd) == 0) { + map_hits++; + return cd; + } + + map_misses++; + return create_die(die, state); +} + +static void reset_die(struct die *cd) +{ + struct die_fragment *tmp; + struct die_fragment *df; + + list_for_each_entry_safe(df, tmp, &cd->fragments, list) { + if (df->type == FRAGMENT_STRING) + free(df->data.str); + free(df); + } + + if (cd->fqn && *cd->fqn) + free(cd->fqn); + init_die(cd); +} + +void die_map_free(void) +{ + struct hlist_node *tmp; + unsigned int stats[DIE_LAST + 1]; + struct die *cd; + int i; + + memset(stats, 0, sizeof(stats)); + + hash_for_each_safe(die_map, cd, tmp, hash) { + stats[cd->state]++; + reset_die(cd); + free(cd); + } + hash_init(die_map); + + if (map_hits + map_misses > 0) + debug("hits %u, misses %u (hit rate %.02f%%)", map_hits, + map_misses, + (100.0f * map_hits) / (map_hits + map_misses)); + + for (i = 0; i <= DIE_LAST; i++) + debug("%s: %u entries", die_state_name(i), stats[i]); +} + +static struct die_fragment *append_item(struct die *cd) +{ + struct die_fragment *df; + + df = xmalloc(sizeof(struct die_fragment)); + df->type = FRAGMENT_EMPTY; + list_add_tail(&df->list, &cd->fragments); + return df; +} + +void die_map_add_string(struct die *cd, const char *str) +{ + struct die_fragment *df; + + if (!cd) + return; + + df = append_item(cd); + df->data.str = xstrdup(str); + df->type = FRAGMENT_STRING; +} + +void die_map_add_die(struct die *cd, struct die *child) +{ + struct die_fragment *df; + + if (!cd) + return; + + df = append_item(cd); + df->data.addr = child->addr; + df->type = FRAGMENT_DIE; +} diff --git a/scripts/gendwarfksyms/dwarf.c b/scripts/gendwarfksyms/dwarf.c index 74e75b8ec891..f40e23a547da 100644 --- a/scripts/gendwarfksyms/dwarf.c +++ b/scripts/gendwarfksyms/dwarf.c @@ -72,17 +72,19 @@ static bool match_export_symbol(struct state *state, Dwarf_Die *die) /* * Type string processing */ -static void process(const char *s) +static void process(struct die *cache, const char *s) { s = s ?: ""; if (dump_dies) fputs(s, stderr); + + die_map_add_string(cache, s); } #define MAX_FMT_BUFFER_SIZE 128 -static void process_fmt(const char *fmt, ...) +static void process_fmt(struct die *cache, const char *fmt, ...) { char buf[MAX_FMT_BUFFER_SIZE]; va_list args; @@ -92,7 +94,7 @@ static void process_fmt(const char *fmt, ...) if (checkp(vsnprintf(buf, sizeof(buf), fmt, args)) >= sizeof(buf)) error("vsnprintf overflow: increase MAX_FMT_BUFFER_SIZE"); - process(buf); + process(cache, buf); va_end(args); } @@ -165,18 +167,28 @@ done: return fqn; } -static void process_fqn(Dwarf_Die *die) +static void update_fqn(struct die *cache, Dwarf_Die *die) { - process(" "); - process(get_fqn(die) ?: ""); + if (!cache->fqn) + cache->fqn = get_fqn(die) ?: ""; } -#define DEFINE_PROCESS_UDATA_ATTRIBUTE(attribute) \ - static void process_##attribute##_attr(Dwarf_Die *die) \ - { \ - Dwarf_Word value; \ - if (get_udata_attr(die, DW_AT_##attribute, &value)) \ - process_fmt(" " #attribute "(%" PRIu64 ")", value); \ +static void process_fqn(struct die *cache, Dwarf_Die *die) +{ + update_fqn(cache, die); + if (*cache->fqn) + process(cache, " "); + process(cache, cache->fqn); +} + +#define DEFINE_PROCESS_UDATA_ATTRIBUTE(attribute) \ + static void process_##attribute##_attr(struct die *cache, \ + Dwarf_Die *die) \ + { \ + Dwarf_Word value; \ + if (get_udata_attr(die, DW_AT_##attribute, &value)) \ + process_fmt(cache, " " #attribute "(%" PRIu64 ")", \ + value); \ } DEFINE_PROCESS_UDATA_ATTRIBUTE(alignment) @@ -188,8 +200,9 @@ bool match_all(Dwarf_Die *die) return true; } -int process_die_container(struct state *state, Dwarf_Die *die, - die_callback_t func, die_match_callback_t match) +int process_die_container(struct state *state, struct die *cache, + Dwarf_Die *die, die_callback_t func, + die_match_callback_t match) { Dwarf_Die current; int res; @@ -198,7 +211,7 @@ int process_die_container(struct state *state, Dwarf_Die *die, while (!res) { if (match(¤t)) { /* <0 = error, 0 = continue, >0 = stop */ - res = checkp(func(state, ¤t)); + res = checkp(func(state, cache, ¤t)); if (res) return res; } @@ -209,39 +222,78 @@ int process_die_container(struct state *state, Dwarf_Die *die, return 0; } -static int process_type(struct state *state, Dwarf_Die *die); +static int process_type(struct state *state, struct die *parent, + Dwarf_Die *die); -static void process_type_attr(struct state *state, Dwarf_Die *die) +static void process_type_attr(struct state *state, struct die *cache, + Dwarf_Die *die) { Dwarf_Die type; if (get_ref_die_attr(die, DW_AT_type, &type)) { - check(process_type(state, &type)); + check(process_type(state, cache, &type)); return; } /* Compilers can omit DW_AT_type -- print out 'void' to clarify */ - process("base_type void"); + process(cache, "base_type void"); } -static void process_base_type(struct state *state, Dwarf_Die *die) +static void process_base_type(struct state *state, struct die *cache, + Dwarf_Die *die) { - process("base_type"); - process_fqn(die); - process_byte_size_attr(die); - process_encoding_attr(die); - process_alignment_attr(die); + process(cache, "base_type"); + process_fqn(cache, die); + process_byte_size_attr(cache, die); + process_encoding_attr(cache, die); + process_alignment_attr(cache, die); } -#define PROCESS_TYPE(type) \ - case DW_TAG_##type##_type: \ - process_##type##_type(state, die); \ +static void process_cached(struct state *state, struct die *cache, + Dwarf_Die *die) +{ + struct die_fragment *df; + Dwarf_Die child; + + list_for_each_entry(df, &cache->fragments, list) { + switch (df->type) { + case FRAGMENT_STRING: + process(NULL, df->data.str); + break; + case FRAGMENT_DIE: + if (!dwarf_die_addr_die(dwarf_cu_getdwarf(die->cu), + (void *)df->data.addr, &child)) + error("dwarf_die_addr_die failed"); + check(process_type(state, NULL, &child)); + break; + default: + error("empty die_fragment"); + } + } +} + +#define PROCESS_TYPE(type) \ + case DW_TAG_##type##_type: \ + process_##type##_type(state, cache, die); \ break; -static int process_type(struct state *state, Dwarf_Die *die) +static int process_type(struct state *state, struct die *parent, Dwarf_Die *die) { + struct die *cache; int tag = dwarf_tag(die); + /* + * If we have the DIE already cached, use it instead of walking + * through DWARF. + */ + cache = die_map_get(die, DIE_COMPLETE); + + if (cache->state == DIE_COMPLETE) { + process_cached(state, cache, die); + die_map_add_die(parent, cache); + return 0; + } + switch (tag) { PROCESS_TYPE(base) default: @@ -249,6 +301,11 @@ static int process_type(struct state *state, Dwarf_Die *die) break; } + /* Update cache state and append to the parent (if any) */ + cache->tag = tag; + cache->state = DIE_COMPLETE; + die_map_add_die(parent, cache); + return 0; } @@ -259,14 +316,15 @@ static void process_symbol(struct state *state, Dwarf_Die *die, die_callback_t process_func) { debug("%s", state->sym->name); - check(process_func(state, die)); + check(process_func(state, NULL, die)); if (dump_dies) fputs("\n", stderr); } -static int __process_subprogram(struct state *state, Dwarf_Die *die) +static int __process_subprogram(struct state *state, struct die *cache, + Dwarf_Die *die) { - process("subprogram"); + process(cache, "subprogram"); return 0; } @@ -275,10 +333,11 @@ static void process_subprogram(struct state *state, Dwarf_Die *die) process_symbol(state, die, __process_subprogram); } -static int __process_variable(struct state *state, Dwarf_Die *die) +static int __process_variable(struct state *state, struct die *cache, + Dwarf_Die *die) { - process("variable "); - process_type_attr(state, die); + process(cache, "variable "); + process_type_attr(state, cache, die); return 0; } @@ -287,7 +346,8 @@ static void process_variable(struct state *state, Dwarf_Die *die) process_symbol(state, die, __process_variable); } -static int process_exported_symbols(struct state *unused, Dwarf_Die *die) +static int process_exported_symbols(struct state *unused, struct die *cache, + Dwarf_Die *die) { int tag = dwarf_tag(die); @@ -297,7 +357,7 @@ static int process_exported_symbols(struct state *unused, Dwarf_Die *die) case DW_TAG_class_type: case DW_TAG_structure_type: return check(process_die_container( - NULL, die, process_exported_symbols, match_all)); + NULL, cache, die, process_exported_symbols, match_all)); /* Possible exported symbols */ case DW_TAG_subprogram: @@ -321,6 +381,6 @@ static int process_exported_symbols(struct state *unused, Dwarf_Die *die) void process_cu(Dwarf_Die *cudie) { - check(process_die_container(NULL, cudie, process_exported_symbols, + check(process_die_container(NULL, NULL, cudie, process_exported_symbols, match_all)); } diff --git a/scripts/gendwarfksyms/gendwarfksyms.c b/scripts/gendwarfksyms/gendwarfksyms.c index cd8bfe973a5c..3809db840c06 100644 --- a/scripts/gendwarfksyms/gendwarfksyms.c +++ b/scripts/gendwarfksyms/gendwarfksyms.c @@ -43,6 +43,10 @@ static int process_module(Dwfl_Module *mod, void **userdata, const char *name, debug("%s", name); dbg = dwfl_module_getdwarf(mod, &dwbias); + /* + * Look for exported symbols in each CU, follow the DIE tree, and add + * the entries to die_map. + */ do { res = dwarf_get_units(dbg, cu, &cu, NULL, NULL, &cudie, NULL); if (res < 0) @@ -53,6 +57,8 @@ static int process_module(Dwfl_Module *mod, void **userdata, const char *name, process_cu(&cudie); } while (cu); + die_map_free(); + return DWARF_CB_OK; } diff --git a/scripts/gendwarfksyms/gendwarfksyms.h b/scripts/gendwarfksyms/gendwarfksyms.h index cb9fd78a58da..601f877bc8ca 100644 --- a/scripts/gendwarfksyms/gendwarfksyms.h +++ b/scripts/gendwarfksyms/gendwarfksyms.h @@ -87,6 +87,61 @@ void symbol_read_symtab(int fd); struct symbol *symbol_get(const char *name); void symbol_free(void); +/* + * die.c + */ + +enum die_state { + DIE_INCOMPLETE, + DIE_COMPLETE, + DIE_LAST = DIE_COMPLETE +}; + +enum die_fragment_type { + FRAGMENT_EMPTY, + FRAGMENT_STRING, + FRAGMENT_DIE +}; + +struct die_fragment { + enum die_fragment_type type; + union { + char *str; + uintptr_t addr; + } data; + struct list_head list; +}; + +#define CASE_CONST_TO_STR(name) \ + case name: \ + return #name; + +static inline const char *die_state_name(enum die_state state) +{ + switch (state) { + CASE_CONST_TO_STR(DIE_INCOMPLETE) + CASE_CONST_TO_STR(DIE_COMPLETE) + } + + error("unexpected die_state: %d", state); +} + +struct die { + enum die_state state; + char *fqn; + int tag; + uintptr_t addr; + struct list_head fragments; + struct hlist_node hash; +}; + +int __die_map_get(uintptr_t addr, enum die_state state, struct die **res); +struct die *die_map_get(Dwarf_Die *die, enum die_state state); +void die_map_add_string(struct die *pd, const char *str); +void die_map_add_linebreak(struct die *pd, int linebreak); +void die_map_add_die(struct die *pd, struct die *child); +void die_map_free(void); + /* * dwarf.c */ @@ -96,12 +151,14 @@ struct state { Dwarf_Die die; }; -typedef int (*die_callback_t)(struct state *state, Dwarf_Die *die); +typedef int (*die_callback_t)(struct state *state, struct die *cache, + Dwarf_Die *die); typedef bool (*die_match_callback_t)(Dwarf_Die *die); bool match_all(Dwarf_Die *die); -int process_die_container(struct state *state, Dwarf_Die *die, - die_callback_t func, die_match_callback_t match); +int process_die_container(struct state *state, struct die *cache, + Dwarf_Die *die, die_callback_t func, + die_match_callback_t match); void process_cu(Dwarf_Die *cudie); From 06b8b036ab9c1e70a562705a398bcd271e0b5ebf Mon Sep 17 00:00:00 2001 From: Sami Tolvanen Date: Fri, 3 Jan 2025 20:45:27 +0000 Subject: [PATCH 23/41] gendwarfksyms: Expand type modifiers and typedefs Add support for expanding DWARF type modifiers, such as pointers, const values etc., and typedefs. These types all have DW_AT_type attribute pointing to the underlying type, and thus produce similar output. Also add linebreaks and indentation to debugging output to make it more readable. Signed-off-by: Sami Tolvanen Reviewed-by: Petr Pavlu Signed-off-by: Masahiro Yamada --- scripts/gendwarfksyms/die.c | 12 +++++ scripts/gendwarfksyms/dwarf.c | 67 +++++++++++++++++++++++++++ scripts/gendwarfksyms/gendwarfksyms.h | 5 ++ 3 files changed, 84 insertions(+) diff --git a/scripts/gendwarfksyms/die.c b/scripts/gendwarfksyms/die.c index b7d900c6a9c8..0d70e02d02b5 100644 --- a/scripts/gendwarfksyms/die.c +++ b/scripts/gendwarfksyms/die.c @@ -130,6 +130,18 @@ void die_map_add_string(struct die *cd, const char *str) df->type = FRAGMENT_STRING; } +void die_map_add_linebreak(struct die *cd, int linebreak) +{ + struct die_fragment *df; + + if (!cd) + return; + + df = append_item(cd); + df->data.linebreak = linebreak; + df->type = FRAGMENT_LINEBREAK; +} + void die_map_add_die(struct die *cd, struct die *child) { struct die_fragment *df; diff --git a/scripts/gendwarfksyms/dwarf.c b/scripts/gendwarfksyms/dwarf.c index f40e23a547da..3e08a32b7b16 100644 --- a/scripts/gendwarfksyms/dwarf.c +++ b/scripts/gendwarfksyms/dwarf.c @@ -7,6 +7,17 @@ #include #include "gendwarfksyms.h" +static bool do_linebreak; +static int indentation_level; + +/* Line breaks and indentation for pretty-printing */ +static void process_linebreak(struct die *cache, int n) +{ + indentation_level += n; + do_linebreak = true; + die_map_add_linebreak(cache, n); +} + #define DEFINE_GET_ATTR(attr, type) \ static bool get_##attr##_attr(Dwarf_Die *die, unsigned int id, \ type *value) \ @@ -76,6 +87,12 @@ static void process(struct die *cache, const char *s) { s = s ?: ""; + if (dump_dies && do_linebreak) { + fputs("\n", stderr); + for (int i = 0; i < indentation_level; i++) + fputs(" ", stderr); + do_linebreak = false; + } if (dump_dies) fputs(s, stderr); @@ -239,6 +256,40 @@ static void process_type_attr(struct state *state, struct die *cache, process(cache, "base_type void"); } +/* Container types with DW_AT_type */ +static void __process_type(struct state *state, struct die *cache, + Dwarf_Die *die, const char *type) +{ + process(cache, type); + process_fqn(cache, die); + process(cache, " {"); + process_linebreak(cache, 1); + process_type_attr(state, cache, die); + process_linebreak(cache, -1); + process(cache, "}"); + process_byte_size_attr(cache, die); + process_alignment_attr(cache, die); +} + +#define DEFINE_PROCESS_TYPE(type) \ + static void process_##type##_type(struct state *state, \ + struct die *cache, Dwarf_Die *die) \ + { \ + __process_type(state, cache, die, #type "_type"); \ + } + +DEFINE_PROCESS_TYPE(atomic) +DEFINE_PROCESS_TYPE(const) +DEFINE_PROCESS_TYPE(immutable) +DEFINE_PROCESS_TYPE(packed) +DEFINE_PROCESS_TYPE(pointer) +DEFINE_PROCESS_TYPE(reference) +DEFINE_PROCESS_TYPE(restrict) +DEFINE_PROCESS_TYPE(rvalue_reference) +DEFINE_PROCESS_TYPE(shared) +DEFINE_PROCESS_TYPE(volatile) +DEFINE_PROCESS_TYPE(typedef) + static void process_base_type(struct state *state, struct die *cache, Dwarf_Die *die) { @@ -260,6 +311,9 @@ static void process_cached(struct state *state, struct die *cache, case FRAGMENT_STRING: process(NULL, df->data.str); break; + case FRAGMENT_LINEBREAK: + process_linebreak(NULL, df->data.linebreak); + break; case FRAGMENT_DIE: if (!dwarf_die_addr_die(dwarf_cu_getdwarf(die->cu), (void *)df->data.addr, &child)) @@ -295,7 +349,20 @@ static int process_type(struct state *state, struct die *parent, Dwarf_Die *die) } switch (tag) { + /* Type modifiers */ + PROCESS_TYPE(atomic) + PROCESS_TYPE(const) + PROCESS_TYPE(immutable) + PROCESS_TYPE(packed) + PROCESS_TYPE(pointer) + PROCESS_TYPE(reference) + PROCESS_TYPE(restrict) + PROCESS_TYPE(rvalue_reference) + PROCESS_TYPE(shared) + PROCESS_TYPE(volatile) + /* Other types */ PROCESS_TYPE(base) + PROCESS_TYPE(typedef) default: debug("unimplemented type: %x", tag); break; diff --git a/scripts/gendwarfksyms/gendwarfksyms.h b/scripts/gendwarfksyms/gendwarfksyms.h index 601f877bc8ca..832d05b4fc1c 100644 --- a/scripts/gendwarfksyms/gendwarfksyms.h +++ b/scripts/gendwarfksyms/gendwarfksyms.h @@ -59,6 +59,9 @@ extern int dump_dies; /* Error == negative values */ #define checkp(expr) __check(expr, __res < 0) +/* Consistent aliases (DW_TAG__type) for DWARF tags */ +#define DW_TAG_typedef_type DW_TAG_typedef + /* * symbols.c */ @@ -100,6 +103,7 @@ enum die_state { enum die_fragment_type { FRAGMENT_EMPTY, FRAGMENT_STRING, + FRAGMENT_LINEBREAK, FRAGMENT_DIE }; @@ -107,6 +111,7 @@ struct die_fragment { enum die_fragment_type type; union { char *str; + int linebreak; uintptr_t addr; } data; struct list_head list; From 220a0857f3a89e0dce3fc7c38d981df41c4537a7 Mon Sep 17 00:00:00 2001 From: Sami Tolvanen Date: Fri, 3 Jan 2025 20:45:28 +0000 Subject: [PATCH 24/41] gendwarfksyms: Expand subroutine_type Add support for expanding DW_TAG_subroutine_type and the parameters in DW_TAG_formal_parameter. Use this to also expand subprograms. Example output with --dump-dies: subprogram ( formal_parameter pointer_type { const_type { base_type char byte_size(1) encoding(6) } } ) -> base_type unsigned long byte_size(8) encoding(7) Signed-off-by: Sami Tolvanen Reviewed-by: Petr Pavlu Signed-off-by: Masahiro Yamada --- scripts/gendwarfksyms/dwarf.c | 84 ++++++++++++++++++++++++++- scripts/gendwarfksyms/gendwarfksyms.h | 4 ++ 2 files changed, 85 insertions(+), 3 deletions(-) diff --git a/scripts/gendwarfksyms/dwarf.c b/scripts/gendwarfksyms/dwarf.c index 3e08a32b7b16..7d8a4eb6c387 100644 --- a/scripts/gendwarfksyms/dwarf.c +++ b/scripts/gendwarfksyms/dwarf.c @@ -212,6 +212,15 @@ DEFINE_PROCESS_UDATA_ATTRIBUTE(alignment) DEFINE_PROCESS_UDATA_ATTRIBUTE(byte_size) DEFINE_PROCESS_UDATA_ATTRIBUTE(encoding) +/* Match functions -- die_match_callback_t */ +#define DEFINE_MATCH(type) \ + static bool match_##type##_type(Dwarf_Die *die) \ + { \ + return dwarf_tag(die) == DW_TAG_##type##_type; \ + } + +DEFINE_MATCH(formal_parameter) + bool match_all(Dwarf_Die *die) { return true; @@ -224,19 +233,28 @@ int process_die_container(struct state *state, struct die *cache, Dwarf_Die current; int res; + /* Track the first item in lists. */ + if (state) + state->first_list_item = true; + res = checkp(dwarf_child(die, ¤t)); while (!res) { if (match(¤t)) { /* <0 = error, 0 = continue, >0 = stop */ res = checkp(func(state, cache, ¤t)); if (res) - return res; + goto out; } res = checkp(dwarf_siblingof(¤t, ¤t)); } - return 0; + res = 0; +out: + if (state) + state->first_list_item = false; + + return res; } static int process_type(struct state *state, struct die *parent, @@ -256,6 +274,40 @@ static void process_type_attr(struct state *state, struct die *cache, process(cache, "base_type void"); } +static void process_list_comma(struct state *state, struct die *cache) +{ + if (state->first_list_item) { + state->first_list_item = false; + } else { + process(cache, " ,"); + process_linebreak(cache, 0); + } +} + +/* Comma-separated with DW_AT_type */ +static void __process_list_type(struct state *state, struct die *cache, + Dwarf_Die *die, const char *type) +{ + const char *name = get_name_attr(die); + + process_list_comma(state, cache); + process(cache, type); + process_type_attr(state, cache, die); + if (name) { + process(cache, " "); + process(cache, name); + } +} + +#define DEFINE_PROCESS_LIST_TYPE(type) \ + static void process_##type##_type(struct state *state, \ + struct die *cache, Dwarf_Die *die) \ + { \ + __process_list_type(state, cache, die, #type " "); \ + } + +DEFINE_PROCESS_LIST_TYPE(formal_parameter) + /* Container types with DW_AT_type */ static void __process_type(struct state *state, struct die *cache, Dwarf_Die *die, const char *type) @@ -290,6 +342,29 @@ DEFINE_PROCESS_TYPE(shared) DEFINE_PROCESS_TYPE(volatile) DEFINE_PROCESS_TYPE(typedef) +static void __process_subroutine_type(struct state *state, struct die *cache, + Dwarf_Die *die, const char *type) +{ + process(cache, type); + process(cache, " ("); + process_linebreak(cache, 1); + /* Parameters */ + check(process_die_container(state, cache, die, process_type, + match_formal_parameter_type)); + process_linebreak(cache, -1); + process(cache, ")"); + process_linebreak(cache, 0); + /* Return type */ + process(cache, "-> "); + process_type_attr(state, cache, die); +} + +static void process_subroutine_type(struct state *state, struct die *cache, + Dwarf_Die *die) +{ + __process_subroutine_type(state, cache, die, "subroutine_type"); +} + static void process_base_type(struct state *state, struct die *cache, Dwarf_Die *die) { @@ -360,8 +435,11 @@ static int process_type(struct state *state, struct die *parent, Dwarf_Die *die) PROCESS_TYPE(rvalue_reference) PROCESS_TYPE(shared) PROCESS_TYPE(volatile) + /* Subtypes */ + PROCESS_TYPE(formal_parameter) /* Other types */ PROCESS_TYPE(base) + PROCESS_TYPE(subroutine) PROCESS_TYPE(typedef) default: debug("unimplemented type: %x", tag); @@ -391,7 +469,7 @@ static void process_symbol(struct state *state, Dwarf_Die *die, static int __process_subprogram(struct state *state, struct die *cache, Dwarf_Die *die) { - process(cache, "subprogram"); + __process_subroutine_type(state, cache, die, "subprogram"); return 0; } diff --git a/scripts/gendwarfksyms/gendwarfksyms.h b/scripts/gendwarfksyms/gendwarfksyms.h index 832d05b4fc1c..0746a36f4924 100644 --- a/scripts/gendwarfksyms/gendwarfksyms.h +++ b/scripts/gendwarfksyms/gendwarfksyms.h @@ -60,6 +60,7 @@ extern int dump_dies; #define checkp(expr) __check(expr, __res < 0) /* Consistent aliases (DW_TAG__type) for DWARF tags */ +#define DW_TAG_formal_parameter_type DW_TAG_formal_parameter #define DW_TAG_typedef_type DW_TAG_typedef /* @@ -154,6 +155,9 @@ void die_map_free(void); struct state { struct symbol *sym; Dwarf_Die die; + + /* List expansion */ + bool first_list_item; }; typedef int (*die_callback_t)(struct state *state, struct die *cache, From c772f1d1eaac608c083ee79fd5cfbe879958eb3e Mon Sep 17 00:00:00 2001 From: Sami Tolvanen Date: Fri, 3 Jan 2025 20:45:29 +0000 Subject: [PATCH 25/41] gendwarfksyms: Expand array_type Add support for expanding DW_TAG_array_type, and the subrange type indicating array size. Example source code: const char *s[34]; Output with --dump-dies: variable array_type[34] { pointer_type { const_type { base_type char byte_size(1) encoding(6) } } byte_size(8) } Signed-off-by: Sami Tolvanen Reviewed-by: Petr Pavlu Signed-off-by: Masahiro Yamada --- scripts/gendwarfksyms/dwarf.c | 30 ++++++++++++++++++++++++++++++ 1 file changed, 30 insertions(+) diff --git a/scripts/gendwarfksyms/dwarf.c b/scripts/gendwarfksyms/dwarf.c index 7d8a4eb6c387..46ce17b2459b 100644 --- a/scripts/gendwarfksyms/dwarf.c +++ b/scripts/gendwarfksyms/dwarf.c @@ -220,6 +220,7 @@ DEFINE_PROCESS_UDATA_ATTRIBUTE(encoding) } DEFINE_MATCH(formal_parameter) +DEFINE_MATCH(subrange) bool match_all(Dwarf_Die *die) { @@ -342,6 +343,33 @@ DEFINE_PROCESS_TYPE(shared) DEFINE_PROCESS_TYPE(volatile) DEFINE_PROCESS_TYPE(typedef) +static void process_subrange_type(struct state *state, struct die *cache, + Dwarf_Die *die) +{ + Dwarf_Word count = 0; + + if (get_udata_attr(die, DW_AT_count, &count)) + process_fmt(cache, "[%" PRIu64 "]", count); + else if (get_udata_attr(die, DW_AT_upper_bound, &count)) + process_fmt(cache, "[%" PRIu64 "]", count + 1); + else + process(cache, "[]"); +} + +static void process_array_type(struct state *state, struct die *cache, + Dwarf_Die *die) +{ + process(cache, "array_type"); + /* Array size */ + check(process_die_container(state, cache, die, process_type, + match_subrange_type)); + process(cache, " {"); + process_linebreak(cache, 1); + process_type_attr(state, cache, die); + process_linebreak(cache, -1); + process(cache, "}"); +} + static void __process_subroutine_type(struct state *state, struct die *cache, Dwarf_Die *die, const char *type) { @@ -437,7 +465,9 @@ static int process_type(struct state *state, struct die *parent, Dwarf_Die *die) PROCESS_TYPE(volatile) /* Subtypes */ PROCESS_TYPE(formal_parameter) + PROCESS_TYPE(subrange) /* Other types */ + PROCESS_TYPE(array) PROCESS_TYPE(base) PROCESS_TYPE(subroutine) PROCESS_TYPE(typedef) From f6bb92455a5e5b2241d2e1f3e240c5fc036c55cb Mon Sep 17 00:00:00 2001 From: Sami Tolvanen Date: Fri, 3 Jan 2025 20:45:30 +0000 Subject: [PATCH 26/41] gendwarfksyms: Expand structure types Recursively expand DWARF structure types, i.e. structs, unions, and enums. Also include relevant DWARF attributes in type strings to encode structure layout, for example. Example output with --dump-dies: subprogram ( formal_parameter structure_type &str { member pointer_type { base_type u8 byte_size(1) encoding(7) } data_ptr data_member_location(0) , member base_type usize byte_size(8) encoding(7) length data_member_location(8) } byte_size(16) alignment(8) msg ) -> base_type void Signed-off-by: Sami Tolvanen Reviewed-by: Petr Pavlu Signed-off-by: Masahiro Yamada --- scripts/gendwarfksyms/dwarf.c | 138 +++++++++++++++++++++++++- scripts/gendwarfksyms/gendwarfksyms.h | 5 + 2 files changed, 141 insertions(+), 2 deletions(-) diff --git a/scripts/gendwarfksyms/dwarf.c b/scripts/gendwarfksyms/dwarf.c index 46ce17b2459b..6ec1138c459f 100644 --- a/scripts/gendwarfksyms/dwarf.c +++ b/scripts/gendwarfksyms/dwarf.c @@ -208,9 +208,14 @@ static void process_fqn(struct die *cache, Dwarf_Die *die) value); \ } +DEFINE_PROCESS_UDATA_ATTRIBUTE(accessibility) DEFINE_PROCESS_UDATA_ATTRIBUTE(alignment) +DEFINE_PROCESS_UDATA_ATTRIBUTE(bit_size) DEFINE_PROCESS_UDATA_ATTRIBUTE(byte_size) DEFINE_PROCESS_UDATA_ATTRIBUTE(encoding) +DEFINE_PROCESS_UDATA_ATTRIBUTE(data_bit_offset) +DEFINE_PROCESS_UDATA_ATTRIBUTE(data_member_location) +DEFINE_PROCESS_UDATA_ATTRIBUTE(discr_value) /* Match functions -- die_match_callback_t */ #define DEFINE_MATCH(type) \ @@ -219,7 +224,9 @@ DEFINE_PROCESS_UDATA_ATTRIBUTE(encoding) return dwarf_tag(die) == DW_TAG_##type##_type; \ } +DEFINE_MATCH(enumerator) DEFINE_MATCH(formal_parameter) +DEFINE_MATCH(member) DEFINE_MATCH(subrange) bool match_all(Dwarf_Die *die) @@ -298,6 +305,10 @@ static void __process_list_type(struct state *state, struct die *cache, process(cache, " "); process(cache, name); } + process_accessibility_attr(cache, die); + process_bit_size_attr(cache, die); + process_data_bit_offset_attr(cache, die); + process_data_member_location_attr(cache, die); } #define DEFINE_PROCESS_LIST_TYPE(type) \ @@ -308,6 +319,7 @@ static void __process_list_type(struct state *state, struct die *cache, } DEFINE_PROCESS_LIST_TYPE(formal_parameter) +DEFINE_PROCESS_LIST_TYPE(member) /* Container types with DW_AT_type */ static void __process_type(struct state *state, struct die *cache, @@ -340,6 +352,7 @@ DEFINE_PROCESS_TYPE(reference) DEFINE_PROCESS_TYPE(restrict) DEFINE_PROCESS_TYPE(rvalue_reference) DEFINE_PROCESS_TYPE(shared) +DEFINE_PROCESS_TYPE(template_type_parameter) DEFINE_PROCESS_TYPE(volatile) DEFINE_PROCESS_TYPE(typedef) @@ -393,6 +406,107 @@ static void process_subroutine_type(struct state *state, struct die *cache, __process_subroutine_type(state, cache, die, "subroutine_type"); } +static void process_variant_type(struct state *state, struct die *cache, + Dwarf_Die *die) +{ + process_list_comma(state, cache); + process(cache, "variant {"); + process_linebreak(cache, 1); + check(process_die_container(state, cache, die, process_type, + match_member_type)); + process_linebreak(cache, -1); + process(cache, "}"); + process_discr_value_attr(cache, die); +} + +static void process_variant_part_type(struct state *state, struct die *cache, + Dwarf_Die *die) +{ + process_list_comma(state, cache); + process(cache, "variant_part {"); + process_linebreak(cache, 1); + check(process_die_container(state, cache, die, process_type, + match_all)); + process_linebreak(cache, -1); + process(cache, "}"); +} + +static int ___process_structure_type(struct state *state, struct die *cache, + Dwarf_Die *die) +{ + switch (dwarf_tag(die)) { + case DW_TAG_member: + case DW_TAG_variant_part: + return check(process_type(state, cache, die)); + case DW_TAG_class_type: + case DW_TAG_enumeration_type: + case DW_TAG_structure_type: + case DW_TAG_template_type_parameter: + case DW_TAG_union_type: + case DW_TAG_subprogram: + /* Skip non-member types, including member functions */ + return 0; + default: + error("unexpected structure_type child: %x", dwarf_tag(die)); + } +} + +static void __process_structure_type(struct state *state, struct die *cache, + Dwarf_Die *die, const char *type, + die_callback_t process_func, + die_match_callback_t match_func) +{ + process(cache, type); + process_fqn(cache, die); + process(cache, " {"); + process_linebreak(cache, 1); + + check(process_die_container(state, cache, die, process_func, + match_func)); + + process_linebreak(cache, -1); + process(cache, "}"); + + process_byte_size_attr(cache, die); + process_alignment_attr(cache, die); +} + +#define DEFINE_PROCESS_STRUCTURE_TYPE(structure) \ + static void process_##structure##_type( \ + struct state *state, struct die *cache, Dwarf_Die *die) \ + { \ + __process_structure_type(state, cache, die, \ + #structure "_type", \ + ___process_structure_type, \ + match_all); \ + } + +DEFINE_PROCESS_STRUCTURE_TYPE(class) +DEFINE_PROCESS_STRUCTURE_TYPE(structure) +DEFINE_PROCESS_STRUCTURE_TYPE(union) + +static void process_enumerator_type(struct state *state, struct die *cache, + Dwarf_Die *die) +{ + Dwarf_Word value; + + process_list_comma(state, cache); + process(cache, "enumerator"); + process_fqn(cache, die); + + if (get_udata_attr(die, DW_AT_const_value, &value)) { + process(cache, " = "); + process_fmt(cache, "%" PRIu64, value); + } +} + +static void process_enumeration_type(struct state *state, struct die *cache, + Dwarf_Die *die) +{ + __process_structure_type(state, cache, die, "enumeration_type", + process_type, match_enumerator_type); +} + static void process_base_type(struct state *state, struct die *cache, Dwarf_Die *die) { @@ -403,6 +517,16 @@ static void process_base_type(struct state *state, struct die *cache, process_alignment_attr(cache, die); } +static void process_unspecified_type(struct state *state, struct die *cache, + Dwarf_Die *die) +{ + /* + * These can be emitted for stand-alone assembly code, which means we + * might run into them in vmlinux.o. + */ + process(cache, "unspecified_type"); +} + static void process_cached(struct state *state, struct die *cache, Dwarf_Die *die) { @@ -463,17 +587,27 @@ static int process_type(struct state *state, struct die *parent, Dwarf_Die *die) PROCESS_TYPE(rvalue_reference) PROCESS_TYPE(shared) PROCESS_TYPE(volatile) + /* Container types */ + PROCESS_TYPE(class) + PROCESS_TYPE(structure) + PROCESS_TYPE(union) + PROCESS_TYPE(enumeration) /* Subtypes */ + PROCESS_TYPE(enumerator) PROCESS_TYPE(formal_parameter) + PROCESS_TYPE(member) PROCESS_TYPE(subrange) + PROCESS_TYPE(template_type_parameter) + PROCESS_TYPE(variant) + PROCESS_TYPE(variant_part) /* Other types */ PROCESS_TYPE(array) PROCESS_TYPE(base) PROCESS_TYPE(subroutine) PROCESS_TYPE(typedef) + PROCESS_TYPE(unspecified) default: - debug("unimplemented type: %x", tag); - break; + error("unexpected type: %x", tag); } /* Update cache state and append to the parent (if any) */ diff --git a/scripts/gendwarfksyms/gendwarfksyms.h b/scripts/gendwarfksyms/gendwarfksyms.h index 0746a36f4924..1796f71b3a34 100644 --- a/scripts/gendwarfksyms/gendwarfksyms.h +++ b/scripts/gendwarfksyms/gendwarfksyms.h @@ -60,8 +60,13 @@ extern int dump_dies; #define checkp(expr) __check(expr, __res < 0) /* Consistent aliases (DW_TAG__type) for DWARF tags */ +#define DW_TAG_enumerator_type DW_TAG_enumerator #define DW_TAG_formal_parameter_type DW_TAG_formal_parameter +#define DW_TAG_member_type DW_TAG_member +#define DW_TAG_template_type_parameter_type DW_TAG_template_type_parameter #define DW_TAG_typedef_type DW_TAG_typedef +#define DW_TAG_variant_part_type DW_TAG_variant_part +#define DW_TAG_variant_type DW_TAG_variant /* * symbols.c From f936c129fd4c3ce495768374ea48e5b736655046 Mon Sep 17 00:00:00 2001 From: Sami Tolvanen Date: Fri, 3 Jan 2025 20:45:31 +0000 Subject: [PATCH 27/41] gendwarfksyms: Limit structure expansion Expand each structure type only once per exported symbol. This is necessary to support self-referential structures, which would otherwise result in infinite recursion, and it's sufficient for catching ABI changes. Types defined in .c files are opaque to external users and thus cannot affect the ABI. Consider type definitions in .c files to be declarations to prevent opaque types from changing symbol versions. Signed-off-by: Sami Tolvanen Reviewed-by: Petr Pavlu Signed-off-by: Masahiro Yamada --- scripts/gendwarfksyms/Makefile | 1 + scripts/gendwarfksyms/cache.c | 51 +++++++++++ scripts/gendwarfksyms/dwarf.c | 125 ++++++++++++++++++++++++-- scripts/gendwarfksyms/gendwarfksyms.h | 46 ++++++++++ 4 files changed, 215 insertions(+), 8 deletions(-) create mode 100644 scripts/gendwarfksyms/cache.c diff --git a/scripts/gendwarfksyms/Makefile b/scripts/gendwarfksyms/Makefile index c0d4ce50fc27..c06145d84df8 100644 --- a/scripts/gendwarfksyms/Makefile +++ b/scripts/gendwarfksyms/Makefile @@ -2,6 +2,7 @@ hostprogs-always-y += gendwarfksyms gendwarfksyms-objs += gendwarfksyms.o +gendwarfksyms-objs += cache.o gendwarfksyms-objs += die.o gendwarfksyms-objs += dwarf.o gendwarfksyms-objs += symbols.o diff --git a/scripts/gendwarfksyms/cache.c b/scripts/gendwarfksyms/cache.c new file mode 100644 index 000000000000..c9c19b86a686 --- /dev/null +++ b/scripts/gendwarfksyms/cache.c @@ -0,0 +1,51 @@ +// SPDX-License-Identifier: GPL-2.0 +/* + * Copyright (C) 2024 Google LLC + */ + +#include "gendwarfksyms.h" + +struct cache_item { + unsigned long key; + int value; + struct hlist_node hash; +}; + +void cache_set(struct cache *cache, unsigned long key, int value) +{ + struct cache_item *ci; + + ci = xmalloc(sizeof(struct cache_item)); + ci->key = key; + ci->value = value; + hash_add(cache->cache, &ci->hash, hash_32(key)); +} + +int cache_get(struct cache *cache, unsigned long key) +{ + struct cache_item *ci; + + hash_for_each_possible(cache->cache, ci, hash, hash_32(key)) { + if (ci->key == key) + return ci->value; + } + + return -1; +} + +void cache_init(struct cache *cache) +{ + hash_init(cache->cache); +} + +void cache_free(struct cache *cache) +{ + struct hlist_node *tmp; + struct cache_item *ci; + + hash_for_each_safe(cache->cache, ci, tmp, hash) { + free(ci); + } + + hash_init(cache->cache); +} diff --git a/scripts/gendwarfksyms/dwarf.c b/scripts/gendwarfksyms/dwarf.c index 6ec1138c459f..6b30e45a4e82 100644 --- a/scripts/gendwarfksyms/dwarf.c +++ b/scripts/gendwarfksyms/dwarf.c @@ -27,6 +27,7 @@ static void process_linebreak(struct die *cache, int n) !dwarf_form##attr(&da, value); \ } +DEFINE_GET_ATTR(flag, bool) DEFINE_GET_ATTR(udata, Dwarf_Word) static bool get_ref_die_attr(Dwarf_Die *die, unsigned int id, Dwarf_Die *value) @@ -80,6 +81,55 @@ static bool match_export_symbol(struct state *state, Dwarf_Die *die) return !!state->sym; } +/* DW_AT_decl_file -> struct srcfile */ +static struct cache srcfile_cache; + +static bool is_definition_private(Dwarf_Die *die) +{ + Dwarf_Word filenum; + Dwarf_Files *files; + Dwarf_Die cudie; + const char *s; + int res; + + /* + * Definitions in .c files cannot change the public ABI, + * so consider them private. + */ + if (!get_udata_attr(die, DW_AT_decl_file, &filenum)) + return false; + + res = cache_get(&srcfile_cache, filenum); + if (res >= 0) + return !!res; + + if (!dwarf_cu_die(die->cu, &cudie, NULL, NULL, NULL, NULL, NULL, NULL)) + error("dwarf_cu_die failed: '%s'", dwarf_errmsg(-1)); + + if (dwarf_getsrcfiles(&cudie, &files, NULL)) + error("dwarf_getsrcfiles failed: '%s'", dwarf_errmsg(-1)); + + s = dwarf_filesrc(files, filenum, NULL, NULL); + if (!s) + error("dwarf_filesrc failed: '%s'", dwarf_errmsg(-1)); + + s = strrchr(s, '.'); + res = s && !strcmp(s, ".c"); + cache_set(&srcfile_cache, filenum, res); + + return !!res; +} + +static bool is_kabi_definition(Dwarf_Die *die) +{ + bool value; + + if (get_flag_attr(die, DW_AT_declaration, &value) && value) + return false; + + return !is_definition_private(die); +} + /* * Type string processing */ @@ -456,19 +506,27 @@ static void __process_structure_type(struct state *state, struct die *cache, die_callback_t process_func, die_match_callback_t match_func) { + bool expand; + process(cache, type); process_fqn(cache, die); process(cache, " {"); process_linebreak(cache, 1); - check(process_die_container(state, cache, die, process_func, - match_func)); + expand = state->expand.expand && is_kabi_definition(die); + + if (expand) { + check(process_die_container(state, cache, die, process_func, + match_func)); + } process_linebreak(cache, -1); process(cache, "}"); - process_byte_size_attr(cache, die); - process_alignment_attr(cache, die); + if (expand) { + process_byte_size_attr(cache, die); + process_alignment_attr(cache, die); + } } #define DEFINE_PROCESS_STRUCTURE_TYPE(structure) \ @@ -553,6 +611,30 @@ static void process_cached(struct state *state, struct die *cache, } } +static void state_init(struct state *state) +{ + state->expand.expand = true; + cache_init(&state->expansion_cache); +} + +static void expansion_state_restore(struct expansion_state *state, + struct expansion_state *saved) +{ + state->expand = saved->expand; +} + +static void expansion_state_save(struct expansion_state *state, + struct expansion_state *saved) +{ + expansion_state_restore(saved, state); +} + +static bool is_expanded_type(int tag) +{ + return tag == DW_TAG_class_type || tag == DW_TAG_structure_type || + tag == DW_TAG_union_type || tag == DW_TAG_enumeration_type; +} + #define PROCESS_TYPE(type) \ case DW_TAG_##type##_type: \ process_##type##_type(state, cache, die); \ @@ -560,18 +642,39 @@ static void process_cached(struct state *state, struct die *cache, static int process_type(struct state *state, struct die *parent, Dwarf_Die *die) { + enum die_state want_state = DIE_COMPLETE; struct die *cache; + struct expansion_state saved; int tag = dwarf_tag(die); + expansion_state_save(&state->expand, &saved); + /* - * If we have the DIE already cached, use it instead of walking + * Structures and enumeration types are expanded only once per + * exported symbol. This is sufficient for detecting ABI changes + * within the structure. + */ + if (is_expanded_type(tag)) { + if (cache_was_expanded(&state->expansion_cache, die->addr)) + state->expand.expand = false; + + if (state->expand.expand) + cache_mark_expanded(&state->expansion_cache, die->addr); + else + want_state = DIE_UNEXPANDED; + } + + /* + * If we have want_state already cached, use it instead of walking * through DWARF. */ - cache = die_map_get(die, DIE_COMPLETE); + cache = die_map_get(die, want_state); - if (cache->state == DIE_COMPLETE) { + if (cache->state == want_state) { process_cached(state, cache, die); die_map_add_die(parent, cache); + + expansion_state_restore(&state->expand, &saved); return 0; } @@ -612,9 +715,10 @@ static int process_type(struct state *state, struct die *parent, Dwarf_Die *die) /* Update cache state and append to the parent (if any) */ cache->tag = tag; - cache->state = DIE_COMPLETE; + cache->state = want_state; die_map_add_die(parent, cache); + expansion_state_restore(&state->expand, &saved); return 0; } @@ -676,11 +780,14 @@ static int process_exported_symbols(struct state *unused, struct die *cache, if (!match_export_symbol(&state, die)) return 0; + state_init(&state); + if (tag == DW_TAG_subprogram) process_subprogram(&state, &state.die); else process_variable(&state, &state.die); + cache_free(&state.expansion_cache); return 0; } default: @@ -692,4 +799,6 @@ void process_cu(Dwarf_Die *cudie) { check(process_die_container(NULL, NULL, cudie, process_exported_symbols, match_all)); + + cache_free(&srcfile_cache); } diff --git a/scripts/gendwarfksyms/gendwarfksyms.h b/scripts/gendwarfksyms/gendwarfksyms.h index 1796f71b3a34..941c4134da8e 100644 --- a/scripts/gendwarfksyms/gendwarfksyms.h +++ b/scripts/gendwarfksyms/gendwarfksyms.h @@ -102,6 +102,7 @@ void symbol_free(void); enum die_state { DIE_INCOMPLETE, + DIE_UNEXPANDED, DIE_COMPLETE, DIE_LAST = DIE_COMPLETE }; @@ -131,6 +132,7 @@ static inline const char *die_state_name(enum die_state state) { switch (state) { CASE_CONST_TO_STR(DIE_INCOMPLETE) + CASE_CONST_TO_STR(DIE_UNEXPANDED) CASE_CONST_TO_STR(DIE_COMPLETE) } @@ -153,16 +155,60 @@ void die_map_add_linebreak(struct die *pd, int linebreak); void die_map_add_die(struct die *pd, struct die *child); void die_map_free(void); +/* + * cache.c + */ + +#define CACHE_HASH_BITS 10 + +/* A cache for addresses we've already seen. */ +struct cache { + HASHTABLE_DECLARE(cache, 1 << CACHE_HASH_BITS); +}; + +void cache_set(struct cache *cache, unsigned long key, int value); +int cache_get(struct cache *cache, unsigned long key); +void cache_init(struct cache *cache); +void cache_free(struct cache *cache); + +static inline void __cache_mark_expanded(struct cache *cache, uintptr_t addr) +{ + cache_set(cache, addr, 1); +} + +static inline bool __cache_was_expanded(struct cache *cache, uintptr_t addr) +{ + return cache_get(cache, addr) == 1; +} + +static inline void cache_mark_expanded(struct cache *cache, void *addr) +{ + __cache_mark_expanded(cache, (uintptr_t)addr); +} + +static inline bool cache_was_expanded(struct cache *cache, void *addr) +{ + return __cache_was_expanded(cache, (uintptr_t)addr); +} + /* * dwarf.c */ +struct expansion_state { + bool expand; +}; + struct state { struct symbol *sym; Dwarf_Die die; /* List expansion */ bool first_list_item; + + /* Structure expansion */ + struct expansion_state expand; + struct cache expansion_cache; }; typedef int (*die_callback_t)(struct state *state, struct die *cache, From d2ffdc1c9a0ee71b30e25fbe3e2a37bf4c146085 Mon Sep 17 00:00:00 2001 From: Sami Tolvanen Date: Fri, 3 Jan 2025 20:45:32 +0000 Subject: [PATCH 28/41] gendwarfksyms: Add die_map debugging Debugging the DWARF processing can be somewhat challenging, so add more detailed debugging output for die_map operations. Add the --dump-die-map flag, which adds color coded tags to the output for die_map changes. Signed-off-by: Sami Tolvanen Reviewed-by: Petr Pavlu Signed-off-by: Masahiro Yamada --- scripts/gendwarfksyms/dwarf.c | 15 +++++++++++++++ scripts/gendwarfksyms/gendwarfksyms.c | 7 +++++++ scripts/gendwarfksyms/gendwarfksyms.h | 13 +++++++++++++ 3 files changed, 35 insertions(+) diff --git a/scripts/gendwarfksyms/dwarf.c b/scripts/gendwarfksyms/dwarf.c index 6b30e45a4e82..364ff4892d5c 100644 --- a/scripts/gendwarfksyms/dwarf.c +++ b/scripts/gendwarfksyms/dwarf.c @@ -146,6 +146,8 @@ static void process(struct die *cache, const char *s) if (dump_dies) fputs(s, stderr); + if (cache) + die_debug_r("cache %p string '%s'", cache, s); die_map_add_string(cache, s); } @@ -594,6 +596,8 @@ static void process_cached(struct state *state, struct die *cache, list_for_each_entry(df, &cache->fragments, list) { switch (df->type) { case FRAGMENT_STRING: + die_debug_b("cache %p STRING '%s'", cache, + df->data.str); process(NULL, df->data.str); break; case FRAGMENT_LINEBREAK: @@ -603,6 +607,8 @@ static void process_cached(struct state *state, struct die *cache, if (!dwarf_die_addr_die(dwarf_cu_getdwarf(die->cu), (void *)df->data.addr, &child)) error("dwarf_die_addr_die failed"); + die_debug_b("cache %p DIE addr %" PRIxPTR " tag %x", + cache, df->data.addr, dwarf_tag(&child)); check(process_type(state, NULL, &child)); break; default: @@ -671,6 +677,9 @@ static int process_type(struct state *state, struct die *parent, Dwarf_Die *die) cache = die_map_get(die, want_state); if (cache->state == want_state) { + die_debug_g("cached addr %p tag %x -- %s", die->addr, tag, + die_state_name(cache->state)); + process_cached(state, cache, die); die_map_add_die(parent, cache); @@ -678,6 +687,9 @@ static int process_type(struct state *state, struct die *parent, Dwarf_Die *die) return 0; } + die_debug_g("addr %p tag %x -- %s -> %s", die->addr, tag, + die_state_name(cache->state), die_state_name(want_state)); + switch (tag) { /* Type modifiers */ PROCESS_TYPE(atomic) @@ -713,6 +725,9 @@ static int process_type(struct state *state, struct die *parent, Dwarf_Die *die) error("unexpected type: %x", tag); } + die_debug_r("parent %p cache %p die addr %p tag %x", parent, cache, + die->addr, tag); + /* Update cache state and append to the parent (if any) */ cache->tag = tag; cache->state = want_state; diff --git a/scripts/gendwarfksyms/gendwarfksyms.c b/scripts/gendwarfksyms/gendwarfksyms.c index 3809db840c06..bf282e33e00c 100644 --- a/scripts/gendwarfksyms/gendwarfksyms.c +++ b/scripts/gendwarfksyms/gendwarfksyms.c @@ -19,6 +19,8 @@ int debug; /* Dump DIE contents */ int dump_dies; +/* Print debugging information about die_map changes */ +int dump_die_map; static void usage(void) { @@ -26,6 +28,7 @@ static void usage(void) "Options:\n" " -d, --debug Print debugging information\n" " --dump-dies Dump DWARF DIE contents\n" + " --dump-die-map Print debugging information about die_map changes\n" " -h, --help Print this message\n" "\n", stderr); @@ -75,6 +78,7 @@ int main(int argc, char **argv) static const struct option opts[] = { { "debug", 0, NULL, 'd' }, { "dump-dies", 0, &dump_dies, 1 }, + { "dump-die-map", 0, &dump_die_map, 1 }, { "help", 0, NULL, 'h' }, { 0, 0, NULL, 0 } }; @@ -95,6 +99,9 @@ int main(int argc, char **argv) } } + if (dump_die_map) + dump_dies = 1; + if (optind >= argc) { usage(); error("no input files?"); diff --git a/scripts/gendwarfksyms/gendwarfksyms.h b/scripts/gendwarfksyms/gendwarfksyms.h index 941c4134da8e..251832dac599 100644 --- a/scripts/gendwarfksyms/gendwarfksyms.h +++ b/scripts/gendwarfksyms/gendwarfksyms.h @@ -21,6 +21,7 @@ */ extern int debug; extern int dump_dies; +extern int dump_die_map; /* * Output helpers @@ -43,6 +44,18 @@ extern int dump_dies; exit(1); \ } while (0) +#define __die_debug(color, format, ...) \ + do { \ + if (dump_dies && dump_die_map) \ + fprintf(stderr, \ + "\033[" #color "m<" format ">\033[39m", \ + __VA_ARGS__); \ + } while (0) + +#define die_debug_r(format, ...) __die_debug(91, format, __VA_ARGS__) +#define die_debug_g(format, ...) __die_debug(92, format, __VA_ARGS__) +#define die_debug_b(format, ...) __die_debug(94, format, __VA_ARGS__) + /* * Error handling helpers */ From ab4439981f8549b013f4ea0b274b7c77c88ab4bc Mon Sep 17 00:00:00 2001 From: Sami Tolvanen Date: Fri, 3 Jan 2025 20:45:33 +0000 Subject: [PATCH 29/41] gendwarfksyms: Add symtypes output Add support for producing genksyms-style symtypes files. Process die_map to find the longest expansions for each type, and use symtypes references in type definitions. The basic file format is similar to genksyms, with two notable exceptions: 1. Type names with spaces (common with Rust) in references are wrapped in single quotes. E.g.: s#'core::result::Result' 2. The actual type definition is the simple parsed DWARF format we output with --dump-dies, not the preprocessed C-style format genksyms produces. Signed-off-by: Sami Tolvanen Reviewed-by: Petr Pavlu Signed-off-by: Masahiro Yamada --- scripts/gendwarfksyms/Makefile | 1 + scripts/gendwarfksyms/die.c | 11 + scripts/gendwarfksyms/dwarf.c | 1 + scripts/gendwarfksyms/gendwarfksyms.c | 33 ++- scripts/gendwarfksyms/gendwarfksyms.h | 19 ++ scripts/gendwarfksyms/symbols.c | 4 +- scripts/gendwarfksyms/types.c | 363 ++++++++++++++++++++++++++ 7 files changed, 429 insertions(+), 3 deletions(-) create mode 100644 scripts/gendwarfksyms/types.c diff --git a/scripts/gendwarfksyms/Makefile b/scripts/gendwarfksyms/Makefile index c06145d84df8..6540282dc746 100644 --- a/scripts/gendwarfksyms/Makefile +++ b/scripts/gendwarfksyms/Makefile @@ -6,5 +6,6 @@ gendwarfksyms-objs += cache.o gendwarfksyms-objs += die.o gendwarfksyms-objs += dwarf.o gendwarfksyms-objs += symbols.o +gendwarfksyms-objs += types.o HOSTLDLIBS_gendwarfksyms := -ldw -lelf diff --git a/scripts/gendwarfksyms/die.c b/scripts/gendwarfksyms/die.c index 0d70e02d02b5..66bd4c9bc952 100644 --- a/scripts/gendwarfksyms/die.c +++ b/scripts/gendwarfksyms/die.c @@ -22,6 +22,7 @@ static inline unsigned int die_hash(uintptr_t addr, enum die_state state) static void init_die(struct die *cd) { cd->state = DIE_INCOMPLETE; + cd->mapped = false; cd->fqn = NULL; cd->tag = -1; cd->addr = 0; @@ -83,6 +84,16 @@ static void reset_die(struct die *cd) init_die(cd); } +void die_map_for_each(die_map_callback_t func, void *arg) +{ + struct hlist_node *tmp; + struct die *cd; + + hash_for_each_safe(die_map, cd, tmp, hash) { + func(cd, arg); + } +} + void die_map_free(void) { struct hlist_node *tmp; diff --git a/scripts/gendwarfksyms/dwarf.c b/scripts/gendwarfksyms/dwarf.c index 364ff4892d5c..a9966a23167a 100644 --- a/scripts/gendwarfksyms/dwarf.c +++ b/scripts/gendwarfksyms/dwarf.c @@ -745,6 +745,7 @@ static void process_symbol(struct state *state, Dwarf_Die *die, { debug("%s", state->sym->name); check(process_func(state, NULL, die)); + state->sym->state = SYMBOL_MAPPED; if (dump_dies) fputs("\n", stderr); } diff --git a/scripts/gendwarfksyms/gendwarfksyms.c b/scripts/gendwarfksyms/gendwarfksyms.c index bf282e33e00c..1d30f42cbd14 100644 --- a/scripts/gendwarfksyms/gendwarfksyms.c +++ b/scripts/gendwarfksyms/gendwarfksyms.c @@ -21,6 +21,11 @@ int debug; int dump_dies; /* Print debugging information about die_map changes */ int dump_die_map; +/* Print out type strings (i.e. type_map) */ +int dump_types; +/* Write a symtypes file */ +int symtypes; +static const char *symtypes_file; static void usage(void) { @@ -29,6 +34,8 @@ static void usage(void) " -d, --debug Print debugging information\n" " --dump-dies Dump DWARF DIE contents\n" " --dump-die-map Print debugging information about die_map changes\n" + " --dump-types Dump type strings\n" + " -T, --symtypes file Write a symtypes file\n" " -h, --help Print this message\n" "\n", stderr); @@ -41,6 +48,7 @@ static int process_module(Dwfl_Module *mod, void **userdata, const char *name, Dwarf_Die cudie; Dwarf_CU *cu = NULL; Dwarf *dbg; + FILE *symfile = arg; int res; debug("%s", name); @@ -60,6 +68,10 @@ static int process_module(Dwfl_Module *mod, void **userdata, const char *name, process_cu(&cudie); } while (cu); + /* + * Use die_map to expand type strings and write them to `symfile`. + */ + generate_symtypes(symfile); die_map_free(); return DWARF_CB_OK; @@ -72,6 +84,7 @@ static const Dwfl_Callbacks callbacks = { int main(int argc, char **argv) { + FILE *symfile = NULL; unsigned int n; int opt; @@ -79,17 +92,23 @@ int main(int argc, char **argv) { "debug", 0, NULL, 'd' }, { "dump-dies", 0, &dump_dies, 1 }, { "dump-die-map", 0, &dump_die_map, 1 }, + { "dump-types", 0, &dump_types, 1 }, + { "symtypes", 1, NULL, 'T' }, { "help", 0, NULL, 'h' }, { 0, 0, NULL, 0 } }; - while ((opt = getopt_long(argc, argv, "dh", opts, NULL)) != EOF) { + while ((opt = getopt_long(argc, argv, "dT:h", opts, NULL)) != EOF) { switch (opt) { case 0: break; case 'd': debug = 1; break; + case 'T': + symtypes = 1; + symtypes_file = optarg; + break; case 'h': usage(); return 0; @@ -109,6 +128,13 @@ int main(int argc, char **argv) symbol_read_exports(stdin); + if (symtypes_file) { + symfile = fopen(symtypes_file, "w"); + if (!symfile) + error("fopen failed for '%s': %s", symtypes_file, + strerror(errno)); + } + for (n = optind; n < argc; n++) { Dwfl *dwfl; int fd; @@ -131,12 +157,15 @@ int main(int argc, char **argv) dwfl_report_end(dwfl, NULL, NULL); - if (dwfl_getmodules(dwfl, &process_module, NULL, 0)) + if (dwfl_getmodules(dwfl, &process_module, symfile, 0)) error("dwfl_getmodules failed for '%s'", argv[n]); dwfl_end(dwfl); } + if (symfile) + check(fclose(symfile)); + symbol_free(); return 0; diff --git a/scripts/gendwarfksyms/gendwarfksyms.h b/scripts/gendwarfksyms/gendwarfksyms.h index 251832dac599..98d5b2315f21 100644 --- a/scripts/gendwarfksyms/gendwarfksyms.h +++ b/scripts/gendwarfksyms/gendwarfksyms.h @@ -22,6 +22,8 @@ extern int debug; extern int dump_dies; extern int dump_die_map; +extern int dump_types; +extern int symtypes; /* * Output helpers @@ -90,6 +92,11 @@ static inline unsigned int addr_hash(uintptr_t addr) return hash_ptr((const void *)addr); } +enum symbol_state { + SYMBOL_UNPROCESSED, + SYMBOL_MAPPED, +}; + struct symbol_addr { uint32_t section; Elf64_Addr address; @@ -100,6 +107,8 @@ struct symbol { struct symbol_addr addr; struct hlist_node addr_hash; struct hlist_node name_hash; + enum symbol_state state; + uintptr_t die_addr; }; typedef void (*symbol_callback_t)(struct symbol *, void *arg); @@ -154,6 +163,7 @@ static inline const char *die_state_name(enum die_state state) struct die { enum die_state state; + bool mapped; char *fqn; int tag; uintptr_t addr; @@ -161,10 +171,13 @@ struct die { struct hlist_node hash; }; +typedef void (*die_map_callback_t)(struct die *, void *arg); + int __die_map_get(uintptr_t addr, enum die_state state, struct die **res); struct die *die_map_get(Dwarf_Die *die, enum die_state state); void die_map_add_string(struct die *pd, const char *str); void die_map_add_linebreak(struct die *pd, int linebreak); +void die_map_for_each(die_map_callback_t func, void *arg); void die_map_add_die(struct die *pd, struct die *child); void die_map_free(void); @@ -235,4 +248,10 @@ int process_die_container(struct state *state, struct die *cache, void process_cu(Dwarf_Die *cudie); +/* + * types.c + */ + +void generate_symtypes(FILE *file); + #endif /* __GENDWARFKSYMS_H */ diff --git a/scripts/gendwarfksyms/symbols.c b/scripts/gendwarfksyms/symbols.c index 98febb524dd5..0d2ce7284a53 100644 --- a/scripts/gendwarfksyms/symbols.c +++ b/scripts/gendwarfksyms/symbols.c @@ -92,6 +92,7 @@ void symbol_read_exports(FILE *file) sym = xcalloc(1, sizeof(struct symbol)); sym->name = name; sym->addr.section = SHN_UNDEF; + sym->state = SYMBOL_UNPROCESSED; hash_add(symbol_names, &sym->name_hash, hash_str(sym->name)); ++nsym; @@ -107,7 +108,8 @@ static void get_symbol(struct symbol *sym, void *arg) { struct symbol **res = arg; - *res = sym; + if (sym->state == SYMBOL_UNPROCESSED) + *res = sym; } struct symbol *symbol_get(const char *name) diff --git a/scripts/gendwarfksyms/types.c b/scripts/gendwarfksyms/types.c new file mode 100644 index 000000000000..21d7a34228eb --- /dev/null +++ b/scripts/gendwarfksyms/types.c @@ -0,0 +1,363 @@ +// SPDX-License-Identifier: GPL-2.0 +/* + * Copyright (C) 2024 Google LLC + */ + +#define _GNU_SOURCE +#include +#include + +#include "gendwarfksyms.h" + +static struct cache expansion_cache; + +/* + * A simple linked list of shared or owned strings to avoid copying strings + * around when not necessary. + */ +struct type_list_entry { + const char *str; + void *owned; + struct list_head list; +}; + +static void type_list_free(struct list_head *list) +{ + struct type_list_entry *entry; + struct type_list_entry *tmp; + + list_for_each_entry_safe(entry, tmp, list, list) { + if (entry->owned) + free(entry->owned); + free(entry); + } + + INIT_LIST_HEAD(list); +} + +static int type_list_append(struct list_head *list, const char *s, void *owned) +{ + struct type_list_entry *entry; + + if (!s) + return 0; + + entry = xmalloc(sizeof(struct type_list_entry)); + entry->str = s; + entry->owned = owned; + list_add_tail(&entry->list, list); + + return strlen(entry->str); +} + +static void type_list_write(struct list_head *list, FILE *file) +{ + struct type_list_entry *entry; + + list_for_each_entry(entry, list, list) { + if (entry->str) + checkp(fputs(entry->str, file)); + } +} + +/* + * An expanded type string in symtypes format. + */ +struct type_expansion { + char *name; + size_t len; + struct list_head expanded; + struct hlist_node hash; +}; + +static void type_expansion_init(struct type_expansion *type) +{ + type->name = NULL; + type->len = 0; + INIT_LIST_HEAD(&type->expanded); +} + +static inline void type_expansion_free(struct type_expansion *type) +{ + free(type->name); + type->name = NULL; + type->len = 0; + type_list_free(&type->expanded); +} + +static void type_expansion_append(struct type_expansion *type, const char *s, + void *owned) +{ + type->len += type_list_append(&type->expanded, s, owned); +} + +/* + * type_map -- the longest expansions for each type. + * + * const char *name -> struct type_expansion * + */ +#define TYPE_HASH_BITS 12 +static HASHTABLE_DEFINE(type_map, 1 << TYPE_HASH_BITS); + +static int type_map_get(const char *name, struct type_expansion **res) +{ + struct type_expansion *e; + + hash_for_each_possible(type_map, e, hash, hash_str(name)) { + if (!strcmp(name, e->name)) { + *res = e; + return 0; + } + } + + return -1; +} + +static void type_map_add(const char *name, struct type_expansion *type) +{ + struct type_expansion *e; + + if (type_map_get(name, &e)) { + e = xmalloc(sizeof(struct type_expansion)); + type_expansion_init(e); + e->name = xstrdup(name); + + hash_add(type_map, &e->hash, hash_str(e->name)); + + if (dump_types) + debug("adding %s", e->name); + } else { + /* Use the longest available expansion */ + if (type->len <= e->len) + return; + + type_list_free(&e->expanded); + + if (dump_types) + debug("replacing %s", e->name); + } + + /* Take ownership of type->expanded */ + list_replace_init(&type->expanded, &e->expanded); + e->len = type->len; + + if (dump_types) { + checkp(fputs(e->name, stderr)); + checkp(fputs(" ", stderr)); + type_list_write(&e->expanded, stderr); + checkp(fputs("\n", stderr)); + } +} + +static void type_map_write(FILE *file) +{ + struct type_expansion *e; + struct hlist_node *tmp; + + if (!file) + return; + + hash_for_each_safe(type_map, e, tmp, hash) { + checkp(fputs(e->name, file)); + checkp(fputs(" ", file)); + type_list_write(&e->expanded, file); + checkp(fputs("\n", file)); + } +} + +static void type_map_free(void) +{ + struct type_expansion *e; + struct hlist_node *tmp; + + hash_for_each_safe(type_map, e, tmp, hash) { + type_expansion_free(e); + free(e); + } + + hash_init(type_map); +} + +/* + * Type reference format: #, where prefix: + * s -> structure + * u -> union + * e -> enum + * t -> typedef + * + * Names with spaces are additionally wrapped in single quotes. + */ +static char get_type_prefix(int tag) +{ + switch (tag) { + case DW_TAG_class_type: + case DW_TAG_structure_type: + return 's'; + case DW_TAG_union_type: + return 'u'; + case DW_TAG_enumeration_type: + return 'e'; + case DW_TAG_typedef_type: + return 't'; + default: + return 0; + } +} + +static char *get_type_name(struct die *cache) +{ + const char *quote; + char prefix; + char *name; + + if (cache->state == DIE_INCOMPLETE) { + warn("found incomplete cache entry: %p", cache); + return NULL; + } + if (!cache->fqn || !*cache->fqn) + return NULL; + + prefix = get_type_prefix(cache->tag); + if (!prefix) + return NULL; + + /* Wrap names with spaces in single quotes */ + quote = strstr(cache->fqn, " ") ? "'" : ""; + + /* #\0 */ + if (asprintf(&name, "%c#%s%s%s", prefix, quote, cache->fqn, quote) < 0) + error("asprintf failed for '%s'", cache->fqn); + + return name; +} + +static void __type_expand(struct die *cache, struct type_expansion *type, + bool recursive); + +static void type_expand_child(struct die *cache, struct type_expansion *type, + bool recursive) +{ + struct type_expansion child; + char *name; + + name = get_type_name(cache); + if (!name) { + __type_expand(cache, type, recursive); + return; + } + + if (recursive && !__cache_was_expanded(&expansion_cache, cache->addr)) { + __cache_mark_expanded(&expansion_cache, cache->addr); + type_expansion_init(&child); + __type_expand(cache, &child, true); + type_map_add(name, &child); + type_expansion_free(&child); + } + + type_expansion_append(type, name, name); +} + +static void __type_expand(struct die *cache, struct type_expansion *type, + bool recursive) +{ + struct die_fragment *df; + struct die *child; + + list_for_each_entry(df, &cache->fragments, list) { + switch (df->type) { + case FRAGMENT_STRING: + type_expansion_append(type, df->data.str, NULL); + break; + case FRAGMENT_DIE: + /* Use a complete die_map expansion if available */ + if (__die_map_get(df->data.addr, DIE_COMPLETE, + &child) && + __die_map_get(df->data.addr, DIE_UNEXPANDED, + &child)) + error("unknown child: %" PRIxPTR, + df->data.addr); + + type_expand_child(child, type, recursive); + break; + case FRAGMENT_LINEBREAK: + /* + * Keep whitespace in the symtypes format, but avoid + * repeated spaces. + */ + if (list_is_last(&df->list, &cache->fragments) || + list_next_entry(df, list)->type != + FRAGMENT_LINEBREAK) + type_expansion_append(type, " ", NULL); + break; + default: + error("empty die_fragment in %p", cache); + } + } +} + +static void type_expand(struct die *cache, struct type_expansion *type, + bool recursive) +{ + type_expansion_init(type); + __type_expand(cache, type, recursive); + cache_free(&expansion_cache); +} + +static void expand_type(struct die *cache, void *arg) +{ + struct type_expansion type; + char *name; + + if (cache->mapped) + return; + + cache->mapped = true; + + /* + * Skip unexpanded die_map entries if there's a complete + * expansion available for this DIE. + */ + if (cache->state == DIE_UNEXPANDED && + !__die_map_get(cache->addr, DIE_COMPLETE, &cache)) { + if (cache->mapped) + return; + + cache->mapped = true; + } + + name = get_type_name(cache); + if (!name) + return; + + debug("%s", name); + type_expand(cache, &type, true); + type_map_add(name, &type); + + type_expansion_free(&type); + free(name); +} + +void generate_symtypes(FILE *file) +{ + cache_init(&expansion_cache); + + /* + * die_map processing: + * + * 1. die_map contains all types referenced in exported symbol + * signatures, but can contain duplicates just like the original + * DWARF, and some references may not be fully expanded depending + * on how far we processed the DIE tree for that specific symbol. + * + * For each die_map entry, find the longest available expansion, + * and add it to type_map. + */ + die_map_for_each(expand_type, NULL); + + /* + * 2. If a symtypes file is requested, write type_map contents to + * the file. + */ + type_map_write(file); + type_map_free(); +} From 71378888018833a1cdcbf72f1e95d7c010542d8b Mon Sep 17 00:00:00 2001 From: Sami Tolvanen Date: Fri, 3 Jan 2025 20:45:34 +0000 Subject: [PATCH 30/41] gendwarfksyms: Add symbol versioning Calculate symbol versions from the fully expanded type strings in type_map, and output the versions in a genksyms-compatible format. Signed-off-by: Sami Tolvanen Reviewed-by: Petr Pavlu Signed-off-by: Masahiro Yamada --- scripts/gendwarfksyms/Makefile | 2 +- scripts/gendwarfksyms/dwarf.c | 25 +++++- scripts/gendwarfksyms/gendwarfksyms.c | 10 ++- scripts/gendwarfksyms/gendwarfksyms.h | 13 ++- scripts/gendwarfksyms/symbols.c | 53 +++++++++++ scripts/gendwarfksyms/types.c | 122 +++++++++++++++++++++++++- 6 files changed, 216 insertions(+), 9 deletions(-) diff --git a/scripts/gendwarfksyms/Makefile b/scripts/gendwarfksyms/Makefile index 6540282dc746..e889b958957b 100644 --- a/scripts/gendwarfksyms/Makefile +++ b/scripts/gendwarfksyms/Makefile @@ -8,4 +8,4 @@ gendwarfksyms-objs += dwarf.o gendwarfksyms-objs += symbols.o gendwarfksyms-objs += types.o -HOSTLDLIBS_gendwarfksyms := -ldw -lelf +HOSTLDLIBS_gendwarfksyms := -ldw -lelf -lz diff --git a/scripts/gendwarfksyms/dwarf.c b/scripts/gendwarfksyms/dwarf.c index a9966a23167a..bdf899d60707 100644 --- a/scripts/gendwarfksyms/dwarf.c +++ b/scripts/gendwarfksyms/dwarf.c @@ -740,12 +740,33 @@ static int process_type(struct state *state, struct die *parent, Dwarf_Die *die) /* * Exported symbol processing */ +static struct die *get_symbol_cache(struct state *state, Dwarf_Die *die) +{ + struct die *cache; + + cache = die_map_get(die, DIE_SYMBOL); + + if (cache->state != DIE_INCOMPLETE) + return NULL; /* We already processed a symbol for this DIE */ + + cache->tag = dwarf_tag(die); + return cache; +} + static void process_symbol(struct state *state, Dwarf_Die *die, die_callback_t process_func) { + struct die *cache; + + symbol_set_die(state->sym, die); + + cache = get_symbol_cache(state, die); + if (!cache) + return; + debug("%s", state->sym->name); - check(process_func(state, NULL, die)); - state->sym->state = SYMBOL_MAPPED; + check(process_func(state, cache, die)); + cache->state = DIE_SYMBOL; if (dump_dies) fputs("\n", stderr); } diff --git a/scripts/gendwarfksyms/gendwarfksyms.c b/scripts/gendwarfksyms/gendwarfksyms.c index 1d30f42cbd14..b0e13c37c6c2 100644 --- a/scripts/gendwarfksyms/gendwarfksyms.c +++ b/scripts/gendwarfksyms/gendwarfksyms.c @@ -23,6 +23,8 @@ int dump_dies; int dump_die_map; /* Print out type strings (i.e. type_map) */ int dump_types; +/* Print out expanded type strings used for symbol versions */ +int dump_versions; /* Write a symtypes file */ int symtypes; static const char *symtypes_file; @@ -35,6 +37,7 @@ static void usage(void) " --dump-dies Dump DWARF DIE contents\n" " --dump-die-map Print debugging information about die_map changes\n" " --dump-types Dump type strings\n" + " --dump-versions Dump expanded type strings used for symbol versions\n" " -T, --symtypes file Write a symtypes file\n" " -h, --help Print this message\n" "\n", @@ -69,9 +72,10 @@ static int process_module(Dwfl_Module *mod, void **userdata, const char *name, } while (cu); /* - * Use die_map to expand type strings and write them to `symfile`. + * Use die_map to expand type strings, write them to `symfile`, and + * calculate symbol versions. */ - generate_symtypes(symfile); + generate_symtypes_and_versions(symfile); die_map_free(); return DWARF_CB_OK; @@ -93,6 +97,7 @@ int main(int argc, char **argv) { "dump-dies", 0, &dump_dies, 1 }, { "dump-die-map", 0, &dump_die_map, 1 }, { "dump-types", 0, &dump_types, 1 }, + { "dump-versions", 0, &dump_versions, 1 }, { "symtypes", 1, NULL, 'T' }, { "help", 0, NULL, 'h' }, { 0, 0, NULL, 0 } @@ -166,6 +171,7 @@ int main(int argc, char **argv) if (symfile) check(fclose(symfile)); + symbol_print_versions(); symbol_free(); return 0; diff --git a/scripts/gendwarfksyms/gendwarfksyms.h b/scripts/gendwarfksyms/gendwarfksyms.h index 98d5b2315f21..203534abcd35 100644 --- a/scripts/gendwarfksyms/gendwarfksyms.h +++ b/scripts/gendwarfksyms/gendwarfksyms.h @@ -23,6 +23,7 @@ extern int debug; extern int dump_dies; extern int dump_die_map; extern int dump_types; +extern int dump_versions; extern int symtypes; /* @@ -95,6 +96,7 @@ static inline unsigned int addr_hash(uintptr_t addr) enum symbol_state { SYMBOL_UNPROCESSED, SYMBOL_MAPPED, + SYMBOL_PROCESSED }; struct symbol_addr { @@ -109,6 +111,7 @@ struct symbol { struct hlist_node name_hash; enum symbol_state state; uintptr_t die_addr; + unsigned long crc; }; typedef void (*symbol_callback_t)(struct symbol *, void *arg); @@ -116,6 +119,10 @@ typedef void (*symbol_callback_t)(struct symbol *, void *arg); void symbol_read_exports(FILE *file); void symbol_read_symtab(int fd); struct symbol *symbol_get(const char *name); +void symbol_set_die(struct symbol *sym, Dwarf_Die *die); +void symbol_set_crc(struct symbol *sym, unsigned long crc); +void symbol_for_each(symbol_callback_t func, void *arg); +void symbol_print_versions(void); void symbol_free(void); /* @@ -126,7 +133,8 @@ enum die_state { DIE_INCOMPLETE, DIE_UNEXPANDED, DIE_COMPLETE, - DIE_LAST = DIE_COMPLETE + DIE_SYMBOL, + DIE_LAST = DIE_SYMBOL }; enum die_fragment_type { @@ -156,6 +164,7 @@ static inline const char *die_state_name(enum die_state state) CASE_CONST_TO_STR(DIE_INCOMPLETE) CASE_CONST_TO_STR(DIE_UNEXPANDED) CASE_CONST_TO_STR(DIE_COMPLETE) + CASE_CONST_TO_STR(DIE_SYMBOL) } error("unexpected die_state: %d", state); @@ -252,6 +261,6 @@ void process_cu(Dwarf_Die *cudie); * types.c */ -void generate_symtypes(FILE *file); +void generate_symtypes_and_versions(FILE *file); #endif /* __GENDWARFKSYMS_H */ diff --git a/scripts/gendwarfksyms/symbols.c b/scripts/gendwarfksyms/symbols.c index 0d2ce7284a53..4c499ba6c86d 100644 --- a/scripts/gendwarfksyms/symbols.c +++ b/scripts/gendwarfksyms/symbols.c @@ -66,6 +66,36 @@ static unsigned int for_each(const char *name, symbol_callback_t func, return 0; } +static void set_crc(struct symbol *sym, void *data) +{ + unsigned long *crc = data; + + if (sym->state == SYMBOL_PROCESSED && sym->crc != *crc) + warn("overriding version for symbol %s (crc %lx vs. %lx)", + sym->name, sym->crc, *crc); + + sym->state = SYMBOL_PROCESSED; + sym->crc = *crc; +} + +void symbol_set_crc(struct symbol *sym, unsigned long crc) +{ + if (for_each(sym->name, set_crc, &crc) == 0) + error("no matching symbols: '%s'", sym->name); +} + +static void set_die(struct symbol *sym, void *data) +{ + sym->die_addr = (uintptr_t)((Dwarf_Die *)data)->addr; + sym->state = SYMBOL_MAPPED; +} + +void symbol_set_die(struct symbol *sym, Dwarf_Die *die) +{ + if (for_each(sym->name, set_die, die) == 0) + error("no matching symbols: '%s'", sym->name); +} + static bool is_exported(const char *name) { return for_each(name, NULL, NULL) > 0; @@ -120,6 +150,16 @@ struct symbol *symbol_get(const char *name) return sym; } +void symbol_for_each(symbol_callback_t func, void *arg) +{ + struct hlist_node *tmp; + struct symbol *sym; + + hash_for_each_safe(symbol_names, sym, tmp, name_hash) { + func(sym, arg); + } +} + typedef void (*elf_symbol_callback_t)(const char *name, GElf_Sym *sym, Elf32_Word xndx, void *arg); @@ -246,6 +286,19 @@ void symbol_read_symtab(int fd) elf_for_each_global(fd, elf_set_symbol_addr, NULL); } +void symbol_print_versions(void) +{ + struct hlist_node *tmp; + struct symbol *sym; + + hash_for_each_safe(symbol_names, sym, tmp, name_hash) { + if (sym->state != SYMBOL_PROCESSED) + warn("no information for symbol %s", sym->name); + + printf("#SYMVER %s 0x%08lx\n", sym->name, sym->crc); + } +} + void symbol_free(void) { struct hlist_node *tmp; diff --git a/scripts/gendwarfksyms/types.c b/scripts/gendwarfksyms/types.c index 21d7a34228eb..6c03265f4d10 100644 --- a/scripts/gendwarfksyms/types.c +++ b/scripts/gendwarfksyms/types.c @@ -6,6 +6,7 @@ #define _GNU_SOURCE #include #include +#include #include "gendwarfksyms.h" @@ -178,6 +179,33 @@ static void type_map_free(void) hash_init(type_map); } +/* + * CRC for a type, with an optional fully expanded type string for + * debugging. + */ +struct version { + struct type_expansion type; + unsigned long crc; +}; + +static void version_init(struct version *version) +{ + version->crc = crc32(0, NULL, 0); + type_expansion_init(&version->type); +} + +static void version_free(struct version *version) +{ + type_expansion_free(&version->type); +} + +static void version_add(struct version *version, const char *s) +{ + version->crc = crc32(version->crc, (void *)s, strlen(s)); + if (dump_versions) + type_expansion_append(&version->type, s, NULL); +} + /* * Type reference format: #, where prefix: * s -> structure @@ -187,6 +215,12 @@ static void type_map_free(void) * * Names with spaces are additionally wrapped in single quotes. */ +static inline bool is_type_prefix(const char *s) +{ + return (s[0] == 's' || s[0] == 'u' || s[0] == 'e' || s[0] == 't') && + s[1] == '#'; +} + static char get_type_prefix(int tag) { switch (tag) { @@ -214,6 +248,8 @@ static char *get_type_name(struct die *cache) warn("found incomplete cache entry: %p", cache); return NULL; } + if (cache->state == DIE_SYMBOL) + return NULL; if (!cache->fqn || !*cache->fqn) return NULL; @@ -231,6 +267,39 @@ static char *get_type_name(struct die *cache) return name; } +static void __calculate_version(struct version *version, struct list_head *list) +{ + struct type_list_entry *entry; + struct type_expansion *e; + + /* Calculate a CRC over an expanded type string */ + list_for_each_entry(entry, list, list) { + if (is_type_prefix(entry->str)) { + check(type_map_get(entry->str, &e)); + + /* + * It's sufficient to expand each type reference just + * once to detect changes. + */ + if (cache_was_expanded(&expansion_cache, e)) { + version_add(version, entry->str); + } else { + cache_mark_expanded(&expansion_cache, e); + __calculate_version(version, &e->expanded); + } + } else { + version_add(version, entry->str); + } + } +} + +static void calculate_version(struct version *version, struct list_head *list) +{ + version_init(version); + __calculate_version(version, list); + cache_free(&expansion_cache); +} + static void __type_expand(struct die *cache, struct type_expansion *type, bool recursive); @@ -337,7 +406,49 @@ static void expand_type(struct die *cache, void *arg) free(name); } -void generate_symtypes(FILE *file) +static void expand_symbol(struct symbol *sym, void *arg) +{ + struct type_expansion type; + struct version version; + struct die *cache; + + /* + * No need to expand again unless we want a symtypes file entry + * for the symbol. Note that this means `sym` has the same address + * as another symbol that was already processed. + */ + if (!symtypes && sym->state == SYMBOL_PROCESSED) + return; + + if (__die_map_get(sym->die_addr, DIE_SYMBOL, &cache)) + return; /* We'll warn about missing CRCs later. */ + + type_expand(cache, &type, false); + + /* If the symbol already has a version, don't calculate it again. */ + if (sym->state != SYMBOL_PROCESSED) { + calculate_version(&version, &type.expanded); + symbol_set_crc(sym, version.crc); + debug("%s = %lx", sym->name, version.crc); + + if (dump_versions) { + checkp(fputs(sym->name, stderr)); + checkp(fputs(" ", stderr)); + type_list_write(&version.type.expanded, stderr); + checkp(fputs("\n", stderr)); + } + + version_free(&version); + } + + /* These aren't needed in type_map unless we want a symtypes file. */ + if (symtypes) + type_map_add(sym->name, &type); + + type_expansion_free(&type); +} + +void generate_symtypes_and_versions(FILE *file) { cache_init(&expansion_cache); @@ -355,7 +466,14 @@ void generate_symtypes(FILE *file) die_map_for_each(expand_type, NULL); /* - * 2. If a symtypes file is requested, write type_map contents to + * 2. For each exported symbol, expand the die_map type, and use + * type_map expansions to calculate a symbol version from the + * fully expanded type string. + */ + symbol_for_each(expand_symbol, NULL); + + /* + * 3. If a symtypes file is requested, write type_map contents to * the file. */ type_map_write(file); From 936cf61c3ef5d6dad714d6c01a85704027dddeb9 Mon Sep 17 00:00:00 2001 From: Sami Tolvanen Date: Fri, 3 Jan 2025 20:45:35 +0000 Subject: [PATCH 31/41] gendwarfksyms: Add support for kABI rules Distributions that want to maintain a stable kABI need the ability to make ABI compatible changes to kernel without affecting symbol versions, either because of LTS updates or backports. With genksyms, developers would typically hide these changes from version calculation with #ifndef __GENKSYMS__, which would result in the symbol version not changing even though the actual type has changed. When we process precompiled object files, this isn't an option. To support this use case, add a --stable command line flag that gates kABI stability features that are not needed in mainline kernels, but can be useful for distributions, and add support for kABI rules, which can be used to restrict gendwarfksyms output. The rules are specified as a set of null-terminated strings stored in the .discard.gendwarfksyms.kabi_rules section. Each rule consists of four strings as follows: "version\0type\0target\0value" The version string ensures the structure can be changed in a backwards compatible way. The type string indicates the type of the rule, and target and value strings contain rule-specific data. Initially support two simple rules: 1. Declaration-only types A type declaration can change into a full definition when additional includes are pulled in to the TU, which changes the versions of any symbol that references the type. Add support for defining declaration-only types whose definition is not expanded during versioning. 2. Ignored enumerators It's possible to add new enum fields without changing the ABI, but as the fields are included in symbol versioning, this would change the versions. Add support for ignoring specific fields. 3. Overridden enumerator values Add support for overriding enumerator values when calculating versions. This may be needed when the last field of the enum is used as a sentinel and new fields must be added before it. Add examples for using the rules under the examples/ directory. Signed-off-by: Sami Tolvanen Signed-off-by: Masahiro Yamada --- scripts/gendwarfksyms/Makefile | 1 + scripts/gendwarfksyms/dwarf.c | 25 +- scripts/gendwarfksyms/examples/kabi.h | 70 +++++ scripts/gendwarfksyms/examples/kabi_ex.c | 14 + scripts/gendwarfksyms/examples/kabi_ex.h | 64 +++++ scripts/gendwarfksyms/gendwarfksyms.c | 11 +- scripts/gendwarfksyms/gendwarfksyms.h | 14 + scripts/gendwarfksyms/kabi.c | 336 +++++++++++++++++++++++ 8 files changed, 531 insertions(+), 4 deletions(-) create mode 100644 scripts/gendwarfksyms/examples/kabi.h create mode 100644 scripts/gendwarfksyms/examples/kabi_ex.c create mode 100644 scripts/gendwarfksyms/examples/kabi_ex.h create mode 100644 scripts/gendwarfksyms/kabi.c diff --git a/scripts/gendwarfksyms/Makefile b/scripts/gendwarfksyms/Makefile index e889b958957b..6334c7d3c4d5 100644 --- a/scripts/gendwarfksyms/Makefile +++ b/scripts/gendwarfksyms/Makefile @@ -5,6 +5,7 @@ gendwarfksyms-objs += gendwarfksyms.o gendwarfksyms-objs += cache.o gendwarfksyms-objs += die.o gendwarfksyms-objs += dwarf.o +gendwarfksyms-objs += kabi.o gendwarfksyms-objs += symbols.o gendwarfksyms-objs += types.o diff --git a/scripts/gendwarfksyms/dwarf.c b/scripts/gendwarfksyms/dwarf.c index bdf899d60707..17f7e6b9a7ff 100644 --- a/scripts/gendwarfksyms/dwarf.c +++ b/scripts/gendwarfksyms/dwarf.c @@ -120,13 +120,16 @@ static bool is_definition_private(Dwarf_Die *die) return !!res; } -static bool is_kabi_definition(Dwarf_Die *die) +static bool is_kabi_definition(struct die *cache, Dwarf_Die *die) { bool value; if (get_flag_attr(die, DW_AT_declaration, &value) && value) return false; + if (kabi_is_declonly(cache->fqn)) + return false; + return !is_definition_private(die); } @@ -515,9 +518,10 @@ static void __process_structure_type(struct state *state, struct die *cache, process(cache, " {"); process_linebreak(cache, 1); - expand = state->expand.expand && is_kabi_definition(die); + expand = state->expand.expand && is_kabi_definition(cache, die); if (expand) { + state->expand.current_fqn = cache->fqn; check(process_die_container(state, cache, die, process_func, match_func)); } @@ -548,13 +552,26 @@ DEFINE_PROCESS_STRUCTURE_TYPE(union) static void process_enumerator_type(struct state *state, struct die *cache, Dwarf_Die *die) { + bool overridden = false; Dwarf_Word value; + if (stable) { + /* Get the fqn before we process anything */ + update_fqn(cache, die); + + if (kabi_is_enumerator_ignored(state->expand.current_fqn, + cache->fqn)) + return; + + overridden = kabi_get_enumerator_value( + state->expand.current_fqn, cache->fqn, &value); + } + process_list_comma(state, cache); process(cache, "enumerator"); process_fqn(cache, die); - if (get_udata_attr(die, DW_AT_const_value, &value)) { + if (overridden || get_udata_attr(die, DW_AT_const_value, &value)) { process(cache, " = "); process_fmt(cache, "%" PRIu64, value); } @@ -620,6 +637,7 @@ static void process_cached(struct state *state, struct die *cache, static void state_init(struct state *state) { state->expand.expand = true; + state->expand.current_fqn = NULL; cache_init(&state->expansion_cache); } @@ -627,6 +645,7 @@ static void expansion_state_restore(struct expansion_state *state, struct expansion_state *saved) { state->expand = saved->expand; + state->current_fqn = saved->current_fqn; } static void expansion_state_save(struct expansion_state *state, diff --git a/scripts/gendwarfksyms/examples/kabi.h b/scripts/gendwarfksyms/examples/kabi.h new file mode 100644 index 000000000000..fcd0300e5b58 --- /dev/null +++ b/scripts/gendwarfksyms/examples/kabi.h @@ -0,0 +1,70 @@ +/* SPDX-License-Identifier: GPL-2.0 */ +/* + * Copyright (C) 2024 Google LLC + * + * Example macros for maintaining kABI stability. + * + * This file is based on android_kabi.h, which has the following notice: + * + * Heavily influenced by rh_kabi.h which came from the RHEL/CENTOS kernel + * and was: + * Copyright (c) 2014 Don Zickus + * Copyright (c) 2015-2018 Jiri Benc + * Copyright (c) 2015 Sabrina Dubroca, Hannes Frederic Sowa + * Copyright (c) 2016-2018 Prarit Bhargava + * Copyright (c) 2017 Paolo Abeni, Larry Woodman + */ + +#ifndef __KABI_H__ +#define __KABI_H__ + +/* Kernel macros for userspace testing. */ +#ifndef __aligned +#define __aligned(x) __attribute__((__aligned__(x))) +#endif +#ifndef __used +#define __used __attribute__((__used__)) +#endif +#ifndef __section +#define __section(section) __attribute__((__section__(section))) +#endif +#ifndef __PASTE +#define ___PASTE(a, b) a##b +#define __PASTE(a, b) ___PASTE(a, b) +#endif +#ifndef __stringify +#define __stringify_1(x...) #x +#define __stringify(x...) __stringify_1(x) +#endif + +#define __KABI_RULE(hint, target, value) \ + static const char __PASTE(__gendwarfksyms_rule_, \ + __COUNTER__)[] __used __aligned(1) \ + __section(".discard.gendwarfksyms.kabi_rules") = \ + "1\0" #hint "\0" #target "\0" #value + +/* + * KABI_DECLONLY(fqn) + * Treat the struct/union/enum fqn as a declaration, i.e. even if + * a definition is available, don't expand the contents. + */ +#define KABI_DECLONLY(fqn) __KABI_RULE(declonly, fqn, ) + +/* + * KABI_ENUMERATOR_IGNORE(fqn, field) + * When expanding enum fqn, skip the provided field. This makes it + * possible to hide added enum fields from versioning. + */ +#define KABI_ENUMERATOR_IGNORE(fqn, field) \ + __KABI_RULE(enumerator_ignore, fqn field, ) + +/* + * KABI_ENUMERATOR_VALUE(fqn, field, value) + * When expanding enum fqn, use the provided value for the + * specified field. This makes it possible to override enumerator + * values when calculating versions. + */ +#define KABI_ENUMERATOR_VALUE(fqn, field, value) \ + __KABI_RULE(enumerator_value, fqn field, value) + +#endif /* __KABI_H__ */ diff --git a/scripts/gendwarfksyms/examples/kabi_ex.c b/scripts/gendwarfksyms/examples/kabi_ex.c new file mode 100644 index 000000000000..799552ea6679 --- /dev/null +++ b/scripts/gendwarfksyms/examples/kabi_ex.c @@ -0,0 +1,14 @@ +// SPDX-License-Identifier: GPL-2.0 +/* + * kabi_ex.c + * + * Copyright (C) 2024 Google LLC + * + * Examples for kABI stability features with --stable. See kabi_ex.h + * for details. + */ + +#include "kabi_ex.h" + +struct s e0; +enum e e1; diff --git a/scripts/gendwarfksyms/examples/kabi_ex.h b/scripts/gendwarfksyms/examples/kabi_ex.h new file mode 100644 index 000000000000..fca1e07c78e2 --- /dev/null +++ b/scripts/gendwarfksyms/examples/kabi_ex.h @@ -0,0 +1,64 @@ +/* SPDX-License-Identifier: GPL-2.0 */ +/* + * kabi_ex.h + * + * Copyright (C) 2024 Google LLC + * + * Examples for kABI stability features with --stable. + */ + +/* + * The comments below each example contain the expected gendwarfksyms + * output, which can be verified using LLVM's FileCheck tool: + * + * https://llvm.org/docs/CommandGuide/FileCheck.html + * + * Usage: + * + * $ gcc -g -c examples/kabi_ex.c -o examples/kabi_ex.o + * + * $ nm examples/kabi_ex.o | awk '{ print $NF }' | \ + * ./gendwarfksyms --stable --dump-dies \ + * examples/kabi_ex.o 2>&1 >/dev/null | \ + * FileCheck examples/kabi_ex.h --check-prefix=STABLE + */ + +#ifndef __KABI_EX_H__ +#define __KABI_EX_H__ + +#include "kabi.h" + +/* + * Example: kABI rules + */ + +struct s { + int a; +}; + +KABI_DECLONLY(s); + +/* + * STABLE: variable structure_type s { + * STABLE-NEXT: } + */ + +enum e { + A, + B, + C, + D, +}; + +KABI_ENUMERATOR_IGNORE(e, B); +KABI_ENUMERATOR_IGNORE(e, C); +KABI_ENUMERATOR_VALUE(e, D, 123456789); + +/* + * STABLE: variable enumeration_type e { + * STABLE-NEXT: enumerator A = 0 , + * STABLE-NEXT: enumerator D = 123456789 + * STABLE-NEXT: } byte_size(4) + */ + +#endif /* __KABI_EX_H__ */ diff --git a/scripts/gendwarfksyms/gendwarfksyms.c b/scripts/gendwarfksyms/gendwarfksyms.c index b0e13c37c6c2..08ae61eb327e 100644 --- a/scripts/gendwarfksyms/gendwarfksyms.c +++ b/scripts/gendwarfksyms/gendwarfksyms.c @@ -25,6 +25,8 @@ int dump_die_map; int dump_types; /* Print out expanded type strings used for symbol versions */ int dump_versions; +/* Support kABI stability features */ +int stable; /* Write a symtypes file */ int symtypes; static const char *symtypes_file; @@ -38,6 +40,7 @@ static void usage(void) " --dump-die-map Print debugging information about die_map changes\n" " --dump-types Dump type strings\n" " --dump-versions Dump expanded type strings used for symbol versions\n" + " -s, --stable Support kABI stability features\n" " -T, --symtypes file Write a symtypes file\n" " -h, --help Print this message\n" "\n", @@ -98,18 +101,22 @@ int main(int argc, char **argv) { "dump-die-map", 0, &dump_die_map, 1 }, { "dump-types", 0, &dump_types, 1 }, { "dump-versions", 0, &dump_versions, 1 }, + { "stable", 0, NULL, 's' }, { "symtypes", 1, NULL, 'T' }, { "help", 0, NULL, 'h' }, { 0, 0, NULL, 0 } }; - while ((opt = getopt_long(argc, argv, "dT:h", opts, NULL)) != EOF) { + while ((opt = getopt_long(argc, argv, "dsT:h", opts, NULL)) != EOF) { switch (opt) { case 0: break; case 'd': debug = 1; break; + case 's': + stable = 1; + break; case 'T': symtypes = 1; symtypes_file = optarg; @@ -150,6 +157,7 @@ int main(int argc, char **argv) strerror(errno)); symbol_read_symtab(fd); + kabi_read_rules(fd); dwfl = dwfl_begin(&callbacks); if (!dwfl) @@ -166,6 +174,7 @@ int main(int argc, char **argv) error("dwfl_getmodules failed for '%s'", argv[n]); dwfl_end(dwfl); + kabi_free(); } if (symfile) diff --git a/scripts/gendwarfksyms/gendwarfksyms.h b/scripts/gendwarfksyms/gendwarfksyms.h index 203534abcd35..c0207ca10e19 100644 --- a/scripts/gendwarfksyms/gendwarfksyms.h +++ b/scripts/gendwarfksyms/gendwarfksyms.h @@ -24,6 +24,7 @@ extern int dump_dies; extern int dump_die_map; extern int dump_types; extern int dump_versions; +extern int stable; extern int symtypes; /* @@ -232,6 +233,7 @@ static inline bool cache_was_expanded(struct cache *cache, void *addr) struct expansion_state { bool expand; + const char *current_fqn; }; struct state { @@ -263,4 +265,16 @@ void process_cu(Dwarf_Die *cudie); void generate_symtypes_and_versions(FILE *file); +/* + * kabi.c + */ + +bool kabi_is_enumerator_ignored(const char *fqn, const char *field); +bool kabi_get_enumerator_value(const char *fqn, const char *field, + unsigned long *value); +bool kabi_is_declonly(const char *fqn); + +void kabi_read_rules(int fd); +void kabi_free(void); + #endif /* __GENDWARFKSYMS_H */ diff --git a/scripts/gendwarfksyms/kabi.c b/scripts/gendwarfksyms/kabi.c new file mode 100644 index 000000000000..66f01fcd1607 --- /dev/null +++ b/scripts/gendwarfksyms/kabi.c @@ -0,0 +1,336 @@ +// SPDX-License-Identifier: GPL-2.0 +/* + * Copyright (C) 2024 Google LLC + */ + +#define _GNU_SOURCE +#include +#include + +#include "gendwarfksyms.h" + +#define KABI_RULE_SECTION ".discard.gendwarfksyms.kabi_rules" +#define KABI_RULE_VERSION "1" + +/* + * The rule section consists of four null-terminated strings per + * entry: + * + * 1. version + * Entry format version. Must match KABI_RULE_VERSION. + * + * 2. type + * Type of the kABI rule. Must be one of the tags defined below. + * + * 3. target + * Rule-dependent target, typically the fully qualified name of + * the target DIE. + * + * 4. value + * Rule-dependent value. + */ +#define KABI_RULE_MIN_ENTRY_SIZE \ + (/* version\0 */ 2 + /* type\0 */ 2 + /* target\0" */ 1 + \ + /* value\0 */ 1) +#define KABI_RULE_EMPTY_VALUE "" + +/* + * Rule: declonly + * - For the struct/enum/union in the target field, treat it as a + * declaration only even if a definition is available. + */ +#define KABI_RULE_TAG_DECLONLY "declonly" + +/* + * Rule: enumerator_ignore + * - For the enum_field in the target field, ignore the enumerator. + */ +#define KABI_RULE_TAG_ENUMERATOR_IGNORE "enumerator_ignore" + +/* + * Rule: enumerator_value + * - For the fqn_field in the target field, set the value to the + * unsigned integer in the value field. + */ +#define KABI_RULE_TAG_ENUMERATOR_VALUE "enumerator_value" + +enum kabi_rule_type { + KABI_RULE_TYPE_UNKNOWN, + KABI_RULE_TYPE_DECLONLY, + KABI_RULE_TYPE_ENUMERATOR_IGNORE, + KABI_RULE_TYPE_ENUMERATOR_VALUE, +}; + +#define RULE_HASH_BITS 7 + +struct rule { + enum kabi_rule_type type; + const char *target; + const char *value; + struct hlist_node hash; +}; + +/* { type, target } -> struct rule */ +static HASHTABLE_DEFINE(rules, 1 << RULE_HASH_BITS); + +static inline unsigned int rule_values_hash(enum kabi_rule_type type, + const char *target) +{ + return hash_32(type) ^ hash_str(target); +} + +static inline unsigned int rule_hash(const struct rule *rule) +{ + return rule_values_hash(rule->type, rule->target); +} + +static inline const char *get_rule_field(const char **pos, ssize_t *left) +{ + const char *start = *pos; + size_t len; + + if (*left <= 0) + error("unexpected end of kABI rules"); + + len = strnlen(start, *left) + 1; + *pos += len; + *left -= len; + + return start; +} + +void kabi_read_rules(int fd) +{ + GElf_Shdr shdr_mem; + GElf_Shdr *shdr; + Elf_Data *rule_data = NULL; + Elf_Scn *scn; + Elf *elf; + size_t shstrndx; + const char *rule_str; + ssize_t left; + int i; + + const struct { + enum kabi_rule_type type; + const char *tag; + } rule_types[] = { + { + .type = KABI_RULE_TYPE_DECLONLY, + .tag = KABI_RULE_TAG_DECLONLY, + }, + { + .type = KABI_RULE_TYPE_ENUMERATOR_IGNORE, + .tag = KABI_RULE_TAG_ENUMERATOR_IGNORE, + }, + { + .type = KABI_RULE_TYPE_ENUMERATOR_VALUE, + .tag = KABI_RULE_TAG_ENUMERATOR_VALUE, + }, + }; + + if (!stable) + return; + + if (elf_version(EV_CURRENT) != EV_CURRENT) + error("elf_version failed: %s", elf_errmsg(-1)); + + elf = elf_begin(fd, ELF_C_READ_MMAP, NULL); + if (!elf) + error("elf_begin failed: %s", elf_errmsg(-1)); + + if (elf_getshdrstrndx(elf, &shstrndx) < 0) + error("elf_getshdrstrndx failed: %s", elf_errmsg(-1)); + + scn = elf_nextscn(elf, NULL); + + while (scn) { + const char *sname; + + shdr = gelf_getshdr(scn, &shdr_mem); + if (!shdr) + error("gelf_getshdr failed: %s", elf_errmsg(-1)); + + sname = elf_strptr(elf, shstrndx, shdr->sh_name); + if (!sname) + error("elf_strptr failed: %s", elf_errmsg(-1)); + + if (!strcmp(sname, KABI_RULE_SECTION)) { + rule_data = elf_getdata(scn, NULL); + if (!rule_data) + error("elf_getdata failed: %s", elf_errmsg(-1)); + break; + } + + scn = elf_nextscn(elf, scn); + } + + if (!rule_data) { + debug("kABI rules not found"); + check(elf_end(elf)); + return; + } + + rule_str = rule_data->d_buf; + left = shdr->sh_size; + + if (left < KABI_RULE_MIN_ENTRY_SIZE) + error("kABI rule section too small: %zd bytes", left); + + if (rule_str[left - 1] != '\0') + error("kABI rules are not null-terminated"); + + while (left > KABI_RULE_MIN_ENTRY_SIZE) { + enum kabi_rule_type type = KABI_RULE_TYPE_UNKNOWN; + const char *field; + struct rule *rule; + + /* version */ + field = get_rule_field(&rule_str, &left); + + if (strcmp(field, KABI_RULE_VERSION)) + error("unsupported kABI rule version: '%s'", field); + + /* type */ + field = get_rule_field(&rule_str, &left); + + for (i = 0; i < ARRAY_SIZE(rule_types); i++) { + if (!strcmp(field, rule_types[i].tag)) { + type = rule_types[i].type; + break; + } + } + + if (type == KABI_RULE_TYPE_UNKNOWN) + error("unsupported kABI rule type: '%s'", field); + + rule = xmalloc(sizeof(struct rule)); + + rule->type = type; + rule->target = xstrdup(get_rule_field(&rule_str, &left)); + rule->value = xstrdup(get_rule_field(&rule_str, &left)); + + hash_add(rules, &rule->hash, rule_hash(rule)); + + debug("kABI rule: type: '%s', target: '%s', value: '%s'", field, + rule->target, rule->value); + } + + if (left > 0) + warn("unexpected data at the end of the kABI rules section"); + + check(elf_end(elf)); +} + +bool kabi_is_declonly(const char *fqn) +{ + struct rule *rule; + + if (!stable) + return false; + if (!fqn || !*fqn) + return false; + + hash_for_each_possible(rules, rule, hash, + rule_values_hash(KABI_RULE_TYPE_DECLONLY, fqn)) { + if (rule->type == KABI_RULE_TYPE_DECLONLY && + !strcmp(fqn, rule->target)) + return true; + } + + return false; +} + +static char *get_enumerator_target(const char *fqn, const char *field) +{ + char *target = NULL; + + if (asprintf(&target, "%s %s", fqn, field) < 0) + error("asprintf failed for '%s %s'", fqn, field); + + return target; +} + +static unsigned long get_ulong_value(const char *value) +{ + unsigned long result = 0; + char *endptr = NULL; + + errno = 0; + result = strtoul(value, &endptr, 10); + + if (errno || *endptr) + error("invalid unsigned value '%s'", value); + + return result; +} + +bool kabi_is_enumerator_ignored(const char *fqn, const char *field) +{ + bool match = false; + struct rule *rule; + char *target; + + if (!stable) + return false; + if (!fqn || !*fqn || !field || !*field) + return false; + + target = get_enumerator_target(fqn, field); + + hash_for_each_possible( + rules, rule, hash, + rule_values_hash(KABI_RULE_TYPE_ENUMERATOR_IGNORE, target)) { + if (rule->type == KABI_RULE_TYPE_ENUMERATOR_IGNORE && + !strcmp(target, rule->target)) { + match = true; + break; + } + } + + free(target); + return match; +} + +bool kabi_get_enumerator_value(const char *fqn, const char *field, + unsigned long *value) +{ + bool match = false; + struct rule *rule; + char *target; + + if (!stable) + return false; + if (!fqn || !*fqn || !field || !*field) + return false; + + target = get_enumerator_target(fqn, field); + + hash_for_each_possible(rules, rule, hash, + rule_values_hash(KABI_RULE_TYPE_ENUMERATOR_VALUE, + target)) { + if (rule->type == KABI_RULE_TYPE_ENUMERATOR_VALUE && + !strcmp(target, rule->target)) { + *value = get_ulong_value(rule->value); + match = true; + break; + } + } + + free(target); + return match; +} + +void kabi_free(void) +{ + struct hlist_node *tmp; + struct rule *rule; + + hash_for_each_safe(rules, rule, tmp, hash) { + free((void *)rule->target); + free((void *)rule->value); + free(rule); + } + + hash_init(rules); +} From a93694188127a5f7ba3baa2f98b275ce388a5246 Mon Sep 17 00:00:00 2001 From: Sami Tolvanen Date: Fri, 3 Jan 2025 20:45:36 +0000 Subject: [PATCH 32/41] gendwarfksyms: Add support for reserved and ignored fields Distributions that want to maintain a stable kABI need the ability to make ABI compatible changes to kernel data structures without affecting symbol versions, either because of LTS updates or backports. With genksyms, developers would typically hide these changes from version calculation with #ifndef __GENKSYMS__, which would result in the symbol version not changing even though the actual type has changed. When we process precompiled object files, this isn't an option. Change union processing to recognize field name prefixes that allow the user to ignore the union completely during symbol versioning with a __kabi_ignored prefix in a field name, or to replace the type of a placeholder field using a __kabi_reserved field name prefix. For example, assume we want to add a new field to an existing alignment hole in a data structure, and ignore the new field when calculating symbol versions: struct struct1 { int a; /* a 4-byte alignment hole */ unsigned long b; }; To add `int n` to the alignment hole, we can add a union that includes a __kabi_ignored field that causes gendwarfksyms to ignore the entire union: struct struct1 { int a; union { char __kabi_ignored_0; int n; }; unsigned long b; }; With --stable, both structs produce the same symbol version. Alternatively, when a distribution expects future modification to a data structure, they can explicitly add reserved fields: struct struct2 { long a; long __kabi_reserved_0; /* reserved for future use */ }; To take the field into use, we can again replace it with a union, with one of the fields keeping the __kabi_reserved name prefix to indicate the original type: struct struct2 { long a; union { long __kabi_reserved_0; struct { int b; int v; }; }; Here gendwarfksyms --stable replaces the union with the type of the placeholder field when calculating versions. Signed-off-by: Sami Tolvanen Reviewed-by: Petr Pavlu Signed-off-by: Masahiro Yamada --- scripts/gendwarfksyms/dwarf.c | 248 ++++++++++++++++++++++- scripts/gendwarfksyms/examples/kabi.h | 87 ++++++++ scripts/gendwarfksyms/examples/kabi_ex.c | 16 ++ scripts/gendwarfksyms/examples/kabi_ex.h | 199 ++++++++++++++++++ scripts/gendwarfksyms/gendwarfksyms.h | 9 + 5 files changed, 558 insertions(+), 1 deletion(-) diff --git a/scripts/gendwarfksyms/dwarf.c b/scripts/gendwarfksyms/dwarf.c index 17f7e6b9a7ff..746a89d9e3d4 100644 --- a/scripts/gendwarfksyms/dwarf.c +++ b/scripts/gendwarfksyms/dwarf.c @@ -3,10 +3,33 @@ * Copyright (C) 2024 Google LLC */ +#include #include #include #include "gendwarfksyms.h" +/* See get_union_kabi_status */ +#define KABI_PREFIX "__kabi_" +#define KABI_PREFIX_LEN (sizeof(KABI_PREFIX) - 1) +#define KABI_RESERVED_PREFIX "reserved" +#define KABI_RESERVED_PREFIX_LEN (sizeof(KABI_RESERVED_PREFIX) - 1) +#define KABI_RENAMED_PREFIX "renamed" +#define KABI_RENAMED_PREFIX_LEN (sizeof(KABI_RENAMED_PREFIX) - 1) +#define KABI_IGNORED_PREFIX "ignored" +#define KABI_IGNORED_PREFIX_LEN (sizeof(KABI_IGNORED_PREFIX) - 1) + +static inline bool is_kabi_prefix(const char *name) +{ + return name && !strncmp(name, KABI_PREFIX, KABI_PREFIX_LEN); +} + +enum kabi_status { + /* >0 to stop DIE processing */ + KABI_NORMAL = 1, + KABI_RESERVED, + KABI_IGNORED, +}; + static bool do_linebreak; static int indentation_level; @@ -353,13 +376,23 @@ static void __process_list_type(struct state *state, struct die *cache, { const char *name = get_name_attr(die); + if (stable) { + if (is_kabi_prefix(name)) + name = NULL; + state->kabi.orig_name = NULL; + } + process_list_comma(state, cache); process(cache, type); process_type_attr(state, cache, die); + + if (stable && state->kabi.orig_name) + name = state->kabi.orig_name; if (name) { process(cache, " "); process(cache, name); } + process_accessibility_attr(cache, die); process_bit_size_attr(cache, die); process_data_bit_offset_attr(cache, die); @@ -486,11 +519,208 @@ static void process_variant_part_type(struct state *state, struct die *cache, process(cache, "}"); } +static int get_kabi_status(Dwarf_Die *die, const char **suffix) +{ + const char *name = get_name_attr(die); + + if (suffix) + *suffix = NULL; + + if (is_kabi_prefix(name)) { + name += KABI_PREFIX_LEN; + + if (!strncmp(name, KABI_RESERVED_PREFIX, + KABI_RESERVED_PREFIX_LEN)) + return KABI_RESERVED; + if (!strncmp(name, KABI_IGNORED_PREFIX, + KABI_IGNORED_PREFIX_LEN)) + return KABI_IGNORED; + + if (!strncmp(name, KABI_RENAMED_PREFIX, + KABI_RENAMED_PREFIX_LEN)) { + if (suffix) { + name += KABI_RENAMED_PREFIX_LEN; + *suffix = name; + } + return KABI_RESERVED; + } + } + + return KABI_NORMAL; +} + +static int check_struct_member_kabi_status(struct state *state, + struct die *__unused, Dwarf_Die *die) +{ + int res; + + assert(dwarf_tag(die) == DW_TAG_member_type); + + /* + * If the union member is a struct, expect the __kabi field to + * be the first member of the structure, i.e..: + * + * union { + * type new_member; + * struct { + * type __kabi_field; + * } + * }; + */ + res = get_kabi_status(die, &state->kabi.orig_name); + + if (res == KABI_RESERVED && + !get_ref_die_attr(die, DW_AT_type, &state->kabi.placeholder)) + error("structure member missing a type?"); + + return res; +} + +static int check_union_member_kabi_status(struct state *state, + struct die *__unused, Dwarf_Die *die) +{ + Dwarf_Die type; + int res; + + assert(dwarf_tag(die) == DW_TAG_member_type); + + if (!get_ref_die_attr(die, DW_AT_type, &type)) + error("union member missing a type?"); + + /* + * We expect a union with two members. Check if either of them + * has a __kabi name prefix, i.e.: + * + * union { + * ... + * type memberN; // <- type, N = {0,1} + * ... + * }; + * + * The member can also be a structure type, in which case we'll + * check the first structure member. + * + * In any case, stop processing after we've seen two members. + */ + res = get_kabi_status(die, &state->kabi.orig_name); + + if (res == KABI_RESERVED) + state->kabi.placeholder = type; + if (res != KABI_NORMAL) + return res; + + if (dwarf_tag(&type) == DW_TAG_structure_type) + res = checkp(process_die_container( + state, NULL, &type, check_struct_member_kabi_status, + match_member_type)); + + if (res <= KABI_NORMAL && ++state->kabi.members < 2) + return 0; /* Continue */ + + return res; +} + +static int get_union_kabi_status(Dwarf_Die *die, Dwarf_Die *placeholder, + const char **orig_name) +{ + struct state state; + int res; + + if (!stable) + return KABI_NORMAL; + + /* + * To maintain a stable kABI, distributions may choose to reserve + * space in structs for later use by adding placeholder members, + * for example: + * + * struct s { + * u32 a; + * // an 8-byte placeholder for future use + * u64 __kabi_reserved_0; + * }; + * + * When the reserved member is taken into use, the type change + * would normally cause the symbol version to change as well, but + * if the replacement uses the following convention, gendwarfksyms + * continues to use the placeholder type for versioning instead, + * thus maintaining the same symbol version: + * + * struct s { + * u32 a; + * union { + * // placeholder replaced with a new member `b` + * struct t b; + * struct { + * // the placeholder type that is still + * // used for versioning + * u64 __kabi_reserved_0; + * }; + * }; + * }; + * + * I.e., as long as the replaced member is in a union, and the + * placeholder has a __kabi_reserved name prefix, we'll continue + * to use the placeholder type (here u64) for version calculation + * instead of the union type. + * + * It's also possible to ignore new members from versioning if + * they've been added to alignment holes, for example, by + * including them in a union with another member that uses the + * __kabi_ignored name prefix: + * + * struct s { + * u32 a; + * // an alignment hole is used to add `n` + * union { + * u32 n; + * // hide the entire union member from versioning + * u8 __kabi_ignored_0; + * }; + * u64 b; + * }; + * + * Note that the user of this feature is responsible for ensuring + * that the structure actually remains ABI compatible. + */ + memset(&state.kabi, 0, sizeof(struct kabi_state)); + + res = checkp(process_die_container(&state, NULL, die, + check_union_member_kabi_status, + match_member_type)); + + if (res == KABI_RESERVED) { + if (placeholder) + *placeholder = state.kabi.placeholder; + if (orig_name) + *orig_name = state.kabi.orig_name; + } + + return res; +} + +static bool is_kabi_ignored(Dwarf_Die *die) +{ + Dwarf_Die type; + + if (!stable) + return false; + + if (!get_ref_die_attr(die, DW_AT_type, &type)) + error("member missing a type?"); + + return dwarf_tag(&type) == DW_TAG_union_type && + checkp(get_union_kabi_status(&type, NULL, NULL)) == KABI_IGNORED; +} + static int ___process_structure_type(struct state *state, struct die *cache, Dwarf_Die *die) { switch (dwarf_tag(die)) { case DW_TAG_member: + if (is_kabi_ignored(die)) + return 0; + return check(process_type(state, cache, die)); case DW_TAG_variant_part: return check(process_type(state, cache, die)); case DW_TAG_class_type: @@ -547,7 +777,23 @@ static void __process_structure_type(struct state *state, struct die *cache, DEFINE_PROCESS_STRUCTURE_TYPE(class) DEFINE_PROCESS_STRUCTURE_TYPE(structure) -DEFINE_PROCESS_STRUCTURE_TYPE(union) + +static void process_union_type(struct state *state, struct die *cache, + Dwarf_Die *die) +{ + Dwarf_Die placeholder; + + int res = checkp(get_union_kabi_status(die, &placeholder, + &state->kabi.orig_name)); + + if (res == KABI_RESERVED) + check(process_type(state, cache, &placeholder)); + if (res > KABI_NORMAL) + return; + + __process_structure_type(state, cache, die, "union_type", + ___process_structure_type, match_all); +} static void process_enumerator_type(struct state *state, struct die *cache, Dwarf_Die *die) diff --git a/scripts/gendwarfksyms/examples/kabi.h b/scripts/gendwarfksyms/examples/kabi.h index fcd0300e5b58..97a5669b083d 100644 --- a/scripts/gendwarfksyms/examples/kabi.h +++ b/scripts/gendwarfksyms/examples/kabi.h @@ -43,6 +43,28 @@ __section(".discard.gendwarfksyms.kabi_rules") = \ "1\0" #hint "\0" #target "\0" #value +#define __KABI_NORMAL_SIZE_ALIGN(_orig, _new) \ + union { \ + _Static_assert( \ + sizeof(struct { _new; }) <= sizeof(struct { _orig; }), \ + __FILE__ ":" __stringify(__LINE__) ": " __stringify( \ + _new) " is larger than " __stringify(_orig)); \ + _Static_assert( \ + __alignof__(struct { _new; }) <= \ + __alignof__(struct { _orig; }), \ + __FILE__ ":" __stringify(__LINE__) ": " __stringify( \ + _orig) " is not aligned the same as " __stringify(_new)); \ + } + +#define __KABI_REPLACE(_orig, _new) \ + union { \ + _new; \ + struct { \ + _orig; \ + }; \ + __KABI_NORMAL_SIZE_ALIGN(_orig, _new); \ + } + /* * KABI_DECLONLY(fqn) * Treat the struct/union/enum fqn as a declaration, i.e. even if @@ -67,4 +89,69 @@ #define KABI_ENUMERATOR_VALUE(fqn, field, value) \ __KABI_RULE(enumerator_value, fqn field, value) +/* + * KABI_RESERVE + * Reserve some "padding" in a structure for use by LTS backports. + * This is normally placed at the end of a structure. + * number: the "number" of the padding variable in the structure. Start with + * 1 and go up. + */ +#define KABI_RESERVE(n) unsigned long __kabi_reserved##n + +/* + * KABI_RESERVE_ARRAY + * Same as _BACKPORT_RESERVE but allocates an array with the specified + * size in bytes. + */ +#define KABI_RESERVE_ARRAY(n, s) \ + unsigned char __aligned(8) __kabi_reserved##n[s] + +/* + * KABI_IGNORE + * Add a new field that's ignored in versioning. + */ +#define KABI_IGNORE(n, _new) \ + union { \ + _new; \ + unsigned char __kabi_ignored##n; \ + } + +/* + * KABI_REPLACE + * Replace a field with a compatible new field. + */ +#define KABI_REPLACE(_oldtype, _oldname, _new) \ + __KABI_REPLACE(_oldtype __kabi_renamed##_oldname, struct { _new; }) + +/* + * KABI_USE(number, _new) + * Use a previous padding entry that was defined with KABI_RESERVE + * number: the previous "number" of the padding variable + * _new: the variable to use now instead of the padding variable + */ +#define KABI_USE(number, _new) __KABI_REPLACE(KABI_RESERVE(number), _new) + +/* + * KABI_USE2(number, _new1, _new2) + * Use a previous padding entry that was defined with KABI_RESERVE for + * two new variables that fit into 64 bits. This is good for when you do not + * want to "burn" a 64bit padding variable for a smaller variable size if not + * needed. + */ +#define KABI_USE2(number, _new1, _new2) \ + __KABI_REPLACE( \ + KABI_RESERVE(number), struct { \ + _new1; \ + _new2; \ + }) +/* + * KABI_USE_ARRAY(number, bytes, _new) + * Use a previous padding entry that was defined with KABI_RESERVE_ARRAY + * number: the previous "number" of the padding variable + * bytes: the size in bytes reserved for the array + * _new: the variable to use now instead of the padding variable + */ +#define KABI_USE_ARRAY(number, bytes, _new) \ + __KABI_REPLACE(KABI_RESERVE_ARRAY(number, bytes), _new) + #endif /* __KABI_H__ */ diff --git a/scripts/gendwarfksyms/examples/kabi_ex.c b/scripts/gendwarfksyms/examples/kabi_ex.c index 799552ea6679..0b7ffd830541 100644 --- a/scripts/gendwarfksyms/examples/kabi_ex.c +++ b/scripts/gendwarfksyms/examples/kabi_ex.c @@ -12,3 +12,19 @@ struct s e0; enum e e1; + +struct ex0a ex0a; +struct ex0b ex0b; +struct ex0c ex0c; + +struct ex1a ex1a; +struct ex1b ex1b; +struct ex1c ex1c; + +struct ex2a ex2a; +struct ex2b ex2b; +struct ex2c ex2c; + +struct ex3a ex3a; +struct ex3b ex3b; +struct ex3c ex3c; diff --git a/scripts/gendwarfksyms/examples/kabi_ex.h b/scripts/gendwarfksyms/examples/kabi_ex.h index fca1e07c78e2..1736e0f65208 100644 --- a/scripts/gendwarfksyms/examples/kabi_ex.h +++ b/scripts/gendwarfksyms/examples/kabi_ex.h @@ -59,6 +59,205 @@ KABI_ENUMERATOR_VALUE(e, D, 123456789); * STABLE-NEXT: enumerator A = 0 , * STABLE-NEXT: enumerator D = 123456789 * STABLE-NEXT: } byte_size(4) +*/ + +/* + * Example: Reserved fields + */ +struct ex0a { + int a; + KABI_RESERVE(0); + KABI_RESERVE(1); +}; + +/* + * STABLE: variable structure_type ex0a { + * STABLE-NEXT: member base_type int byte_size(4) encoding(5) a data_member_location(0) , + * STABLE-NEXT: member base_type [[ULONG:long unsigned int|unsigned long]] byte_size(8) encoding(7) data_member_location(8) , + * STABLE-NEXT: member base_type [[ULONG]] byte_size(8) encoding(7) data_member_location(16) + * STABLE-NEXT: } byte_size(24) + */ + +struct ex0b { + int a; + KABI_RESERVE(0); + KABI_USE2(1, int b, int c); +}; + +/* + * STABLE: variable structure_type ex0b { + * STABLE-NEXT: member base_type int byte_size(4) encoding(5) a data_member_location(0) , + * STABLE-NEXT: member base_type [[ULONG]] byte_size(8) encoding(7) data_member_location(8) , + * STABLE-NEXT: member base_type [[ULONG]] byte_size(8) encoding(7) data_member_location(16) + * STABLE-NEXT: } byte_size(24) + */ + +struct ex0c { + int a; + KABI_USE(0, void *p); + KABI_USE2(1, int b, int c); +}; + +/* + * STABLE: variable structure_type ex0c { + * STABLE-NEXT: member base_type int byte_size(4) encoding(5) a data_member_location(0) , + * STABLE-NEXT: member base_type [[ULONG]] byte_size(8) encoding(7) data_member_location(8) , + * STABLE-NEXT: member base_type [[ULONG]] byte_size(8) encoding(7) data_member_location(16) + * STABLE-NEXT: } byte_size(24) + */ + +/* + * Example: A reserved array + */ + +struct ex1a { + unsigned int a; + KABI_RESERVE_ARRAY(0, 64); +}; + +/* + * STABLE: variable structure_type ex1a { + * STABLE-NEXT: member base_type unsigned int byte_size(4) encoding(7) a data_member_location(0) , + * STABLE-NEXT: member array_type[64] { + * STABLE-NEXT: base_type unsigned char byte_size(1) encoding(8) + * STABLE-NEXT: } data_member_location(8) + * STABLE-NEXT: } byte_size(72) + */ + +struct ex1b { + unsigned int a; + KABI_USE_ARRAY( + 0, 64, struct { + void *p; + KABI_RESERVE_ARRAY(1, 56); + }); +}; + +/* + * STABLE: variable structure_type ex1b { + * STABLE-NEXT: member base_type unsigned int byte_size(4) encoding(7) a data_member_location(0) , + * STABLE-NEXT: member array_type[64] { + * STABLE-NEXT: base_type unsigned char byte_size(1) encoding(8) + * STABLE-NEXT: } data_member_location(8) + * STABLE-NEXT: } byte_size(72) + */ + +struct ex1c { + unsigned int a; + KABI_USE_ARRAY(0, 64, void *p[8]); +}; + +/* + * STABLE: variable structure_type ex1c { + * STABLE-NEXT: member base_type unsigned int byte_size(4) encoding(7) a data_member_location(0) , + * STABLE-NEXT: member array_type[64] { + * STABLE-NEXT: base_type unsigned char byte_size(1) encoding(8) + * STABLE-NEXT: } data_member_location(8) + * STABLE-NEXT: } byte_size(72) + */ + +/* + * Example: An ignored field added to an alignment hole + */ + +struct ex2a { + int a; + unsigned long b; + int c; + unsigned long d; +}; + +/* + * STABLE: variable structure_type ex2a { + * STABLE-NEXT: member base_type int byte_size(4) encoding(5) a data_member_location(0) , + * STABLE-NEXT: member base_type [[ULONG:long unsigned int|unsigned long]] byte_size(8) encoding(7) b data_member_location(8) + * STABLE-NEXT: member base_type int byte_size(4) encoding(5) c data_member_location(16) , + * STABLE-NEXT: member base_type [[ULONG]] byte_size(8) encoding(7) d data_member_location(24) + * STABLE-NEXT: } byte_size(32) + */ + +struct ex2b { + int a; + KABI_IGNORE(0, unsigned int n); + unsigned long b; + int c; + unsigned long d; +}; + +_Static_assert(sizeof(struct ex2a) == sizeof(struct ex2b), "ex2a size doesn't match ex2b"); + +/* + * STABLE: variable structure_type ex2b { + * STABLE-NEXT: member base_type int byte_size(4) encoding(5) a data_member_location(0) , + * STABLE-NEXT: member base_type [[ULONG]] byte_size(8) encoding(7) b data_member_location(8) + * STABLE-NEXT: member base_type int byte_size(4) encoding(5) c data_member_location(16) , + * STABLE-NEXT: member base_type [[ULONG]] byte_size(8) encoding(7) d data_member_location(24) + * STABLE-NEXT: } byte_size(32) + */ + +struct ex2c { + int a; + KABI_IGNORE(0, unsigned int n); + unsigned long b; + int c; + KABI_IGNORE(1, unsigned int m); + unsigned long d; +}; + +_Static_assert(sizeof(struct ex2a) == sizeof(struct ex2c), "ex2a size doesn't match ex2c"); + +/* + * STABLE: variable structure_type ex2c { + * STABLE-NEXT: member base_type int byte_size(4) encoding(5) a data_member_location(0) , + * STABLE-NEXT: member base_type [[ULONG]] byte_size(8) encoding(7) b data_member_location(8) + * STABLE-NEXT: member base_type int byte_size(4) encoding(5) c data_member_location(16) , + * STABLE-NEXT: member base_type [[ULONG]] byte_size(8) encoding(7) d data_member_location(24) + * STABLE-NEXT: } byte_size(32) + */ + + +/* + * Example: A replaced field + */ + +struct ex3a { + unsigned long a; + unsigned long unused; +}; + +/* + * STABLE: variable structure_type ex3a { + * STABLE-NEXT: member base_type [[ULONG:long unsigned int|unsigned long]] byte_size(8) encoding(7) a data_member_location(0) + * STABLE-NEXT: member base_type [[ULONG]] byte_size(8) encoding(7) unused data_member_location(8) + * STABLE-NEXT: } byte_size(16) + */ + +struct ex3b { + unsigned long a; + KABI_REPLACE(unsigned long, unused, unsigned long renamed); +}; + +_Static_assert(sizeof(struct ex3a) == sizeof(struct ex3b), "ex3a size doesn't match ex3b"); + +/* + * STABLE: variable structure_type ex3b { + * STABLE-NEXT: member base_type [[ULONG]] byte_size(8) encoding(7) a data_member_location(0) + * STABLE-NEXT: member base_type [[ULONG]] byte_size(8) encoding(7) unused data_member_location(8) + * STABLE-NEXT: } byte_size(16) + */ + +struct ex3c { + unsigned long a; + KABI_REPLACE(unsigned long, unused, long replaced); +}; + +_Static_assert(sizeof(struct ex3a) == sizeof(struct ex3c), "ex3a size doesn't match ex3c"); + +/* + * STABLE: variable structure_type ex3c { + * STABLE-NEXT: member base_type [[ULONG]] byte_size(8) encoding(7) a data_member_location(0) + * STABLE-NEXT: member base_type [[ULONG]] byte_size(8) encoding(7) unused data_member_location(8) + * STABLE-NEXT: } byte_size(16) */ #endif /* __KABI_EX_H__ */ diff --git a/scripts/gendwarfksyms/gendwarfksyms.h b/scripts/gendwarfksyms/gendwarfksyms.h index c0207ca10e19..fe49730fe623 100644 --- a/scripts/gendwarfksyms/gendwarfksyms.h +++ b/scripts/gendwarfksyms/gendwarfksyms.h @@ -236,6 +236,12 @@ struct expansion_state { const char *current_fqn; }; +struct kabi_state { + int members; + Dwarf_Die placeholder; + const char *orig_name; +}; + struct state { struct symbol *sym; Dwarf_Die die; @@ -246,6 +252,9 @@ struct state { /* Structure expansion */ struct expansion_state expand; struct cache expansion_cache; + + /* Reserved or ignored members */ + struct kabi_state kabi; }; typedef int (*die_callback_t)(struct state *state, struct die *cache, From fa624569b70d8015775592ae7e2c514009367541 Mon Sep 17 00:00:00 2001 From: Sami Tolvanen Date: Fri, 3 Jan 2025 20:45:37 +0000 Subject: [PATCH 33/41] gendwarfksyms: Add support for symbol type pointers The compiler may choose not to emit type information in DWARF for external symbols. Clang, for example, does this for symbols not defined in the current TU. To provide a way to work around this issue, add support for __gendwarfksyms_ptr_ pointers that force the compiler to emit the necessary type information in DWARF also for the missing symbols. Example usage: #define GENDWARFKSYMS_PTR(sym) \ static typeof(sym) *__gendwarfksyms_ptr_##sym __used \ __section(".discard.gendwarfksyms") = &sym; extern int external_symbol(void); GENDWARFKSYMS_PTR(external_symbol); Signed-off-by: Sami Tolvanen Reviewed-by: Petr Pavlu Signed-off-by: Masahiro Yamada --- scripts/gendwarfksyms/dwarf.c | 55 +++++++++++++++++++++- scripts/gendwarfksyms/examples/symbolptr.c | 33 +++++++++++++ scripts/gendwarfksyms/gendwarfksyms.h | 7 +++ scripts/gendwarfksyms/symbols.c | 27 +++++++++++ 4 files changed, 121 insertions(+), 1 deletion(-) create mode 100644 scripts/gendwarfksyms/examples/symbolptr.c diff --git a/scripts/gendwarfksyms/dwarf.c b/scripts/gendwarfksyms/dwarf.c index 746a89d9e3d4..534d9aa7c114 100644 --- a/scripts/gendwarfksyms/dwarf.c +++ b/scripts/gendwarfksyms/dwarf.c @@ -1061,6 +1061,31 @@ static void process_variable(struct state *state, Dwarf_Die *die) process_symbol(state, die, __process_variable); } +static void save_symbol_ptr(struct state *state) +{ + Dwarf_Die ptr_type; + Dwarf_Die type; + + if (!get_ref_die_attr(&state->die, DW_AT_type, &ptr_type) || + dwarf_tag(&ptr_type) != DW_TAG_pointer_type) + error("%s must be a pointer type!", + get_symbol_name(&state->die)); + + if (!get_ref_die_attr(&ptr_type, DW_AT_type, &type)) + error("%s pointer missing a type attribute?", + get_symbol_name(&state->die)); + + /* + * Save the symbol pointer DIE in case the actual symbol is + * missing from the DWARF. Clang, for example, intentionally + * omits external symbols from the debugging information. + */ + if (dwarf_tag(&type) == DW_TAG_subroutine_type) + symbol_set_ptr(state->sym, &type); + else + symbol_set_ptr(state->sym, &ptr_type); +} + static int process_exported_symbols(struct state *unused, struct die *cache, Dwarf_Die *die) { @@ -1084,7 +1109,9 @@ static int process_exported_symbols(struct state *unused, struct die *cache, state_init(&state); - if (tag == DW_TAG_subprogram) + if (is_symbol_ptr(get_symbol_name(&state.die))) + save_symbol_ptr(&state); + else if (tag == DW_TAG_subprogram) process_subprogram(&state, &state.die); else process_variable(&state, &state.die); @@ -1097,10 +1124,36 @@ static int process_exported_symbols(struct state *unused, struct die *cache, } } +static void process_symbol_ptr(struct symbol *sym, void *arg) +{ + struct state state; + Dwarf *dwarf = arg; + + if (sym->state != SYMBOL_UNPROCESSED || !sym->ptr_die_addr) + return; + + debug("%s", sym->name); + state_init(&state); + state.sym = sym; + + if (!dwarf_die_addr_die(dwarf, (void *)sym->ptr_die_addr, &state.die)) + error("dwarf_die_addr_die failed for symbol ptr: '%s'", + sym->name); + + if (dwarf_tag(&state.die) == DW_TAG_subroutine_type) + process_subprogram(&state, &state.die); + else + process_variable(&state, &state.die); + + cache_free(&state.expansion_cache); +} + void process_cu(Dwarf_Die *cudie) { check(process_die_container(NULL, NULL, cudie, process_exported_symbols, match_all)); + symbol_for_each(process_symbol_ptr, dwarf_cu_getdwarf(cudie->cu)); + cache_free(&srcfile_cache); } diff --git a/scripts/gendwarfksyms/examples/symbolptr.c b/scripts/gendwarfksyms/examples/symbolptr.c new file mode 100644 index 000000000000..88bc1bd60da8 --- /dev/null +++ b/scripts/gendwarfksyms/examples/symbolptr.c @@ -0,0 +1,33 @@ +// SPDX-License-Identifier: GPL-2.0 +/* + * Copyright (C) 2024 Google LLC + * + * Example for symbol pointers. When compiled with Clang, gendwarfkyms + * uses a symbol pointer for `f`. + * + * $ clang -g -c examples/symbolptr.c -o examples/symbolptr.o + * $ echo -e "f\ng\np" | ./gendwarfksyms -d examples/symbolptr.o + */ + +/* Kernel macros for userspace testing. */ +#ifndef __used +#define __used __attribute__((__used__)) +#endif +#ifndef __section +#define __section(section) __attribute__((__section__(section))) +#endif + +#define __GENDWARFKSYMS_EXPORT(sym) \ + static typeof(sym) *__gendwarfksyms_ptr_##sym __used \ + __section(".discard.gendwarfksyms") = &sym; + +extern void f(unsigned int arg); +void g(int *arg); +void g(int *arg) {} + +struct s; +extern struct s *p; + +__GENDWARFKSYMS_EXPORT(f); +__GENDWARFKSYMS_EXPORT(g); +__GENDWARFKSYMS_EXPORT(p); diff --git a/scripts/gendwarfksyms/gendwarfksyms.h b/scripts/gendwarfksyms/gendwarfksyms.h index fe49730fe623..197a1a8123c6 100644 --- a/scripts/gendwarfksyms/gendwarfksyms.h +++ b/scripts/gendwarfksyms/gendwarfksyms.h @@ -89,6 +89,10 @@ extern int symtypes; * symbols.c */ +/* See symbols.c:is_symbol_ptr */ +#define SYMBOL_PTR_PREFIX "__gendwarfksyms_ptr_" +#define SYMBOL_PTR_PREFIX_LEN (sizeof(SYMBOL_PTR_PREFIX) - 1) + static inline unsigned int addr_hash(uintptr_t addr) { return hash_ptr((const void *)addr); @@ -112,14 +116,17 @@ struct symbol { struct hlist_node name_hash; enum symbol_state state; uintptr_t die_addr; + uintptr_t ptr_die_addr; unsigned long crc; }; typedef void (*symbol_callback_t)(struct symbol *, void *arg); +bool is_symbol_ptr(const char *name); void symbol_read_exports(FILE *file); void symbol_read_symtab(int fd); struct symbol *symbol_get(const char *name); +void symbol_set_ptr(struct symbol *sym, Dwarf_Die *ptr); void symbol_set_die(struct symbol *sym, Dwarf_Die *die); void symbol_set_crc(struct symbol *sym, unsigned long crc); void symbol_for_each(symbol_callback_t func, void *arg); diff --git a/scripts/gendwarfksyms/symbols.c b/scripts/gendwarfksyms/symbols.c index 4c499ba6c86d..327f87389c34 100644 --- a/scripts/gendwarfksyms/symbols.c +++ b/scripts/gendwarfksyms/symbols.c @@ -39,6 +39,20 @@ static unsigned int __for_each_addr(struct symbol *sym, symbol_callback_t func, return processed; } +/* + * For symbols without debugging information (e.g. symbols defined in other + * TUs), we also match __gendwarfksyms_ptr_ symbols, which the + * kernel uses to ensure type information is present in the TU that exports + * the symbol. A __gendwarfksyms_ptr pointer must have the same type as the + * exported symbol, e.g.: + * + * typeof(symname) *__gendwarf_ptr_symname = &symname; + */ +bool is_symbol_ptr(const char *name) +{ + return name && !strncmp(name, SYMBOL_PTR_PREFIX, SYMBOL_PTR_PREFIX_LEN); +} + static unsigned int for_each(const char *name, symbol_callback_t func, void *data) { @@ -47,6 +61,8 @@ static unsigned int for_each(const char *name, symbol_callback_t func, if (!name || !*name) return 0; + if (is_symbol_ptr(name)) + name += SYMBOL_PTR_PREFIX_LEN; hash_for_each_possible_safe(symbol_names, match, tmp, name_hash, hash_str(name)) { @@ -84,6 +100,17 @@ void symbol_set_crc(struct symbol *sym, unsigned long crc) error("no matching symbols: '%s'", sym->name); } +static void set_ptr(struct symbol *sym, void *data) +{ + sym->ptr_die_addr = (uintptr_t)((Dwarf_Die *)data)->addr; +} + +void symbol_set_ptr(struct symbol *sym, Dwarf_Die *ptr) +{ + if (for_each(sym->name, set_ptr, ptr) == 0) + error("no matching symbols: '%s'", sym->name); +} + static void set_die(struct symbol *sym, void *data) { sym->die_addr = (uintptr_t)((Dwarf_Die *)data)->addr; From d7476f24c9aa93d02ef3fd8d587a6114387b7667 Mon Sep 17 00:00:00 2001 From: Sami Tolvanen Date: Fri, 3 Jan 2025 20:45:38 +0000 Subject: [PATCH 34/41] export: Add __gendwarfksyms_ptr_ references to exported symbols With gendwarfksyms, we need each TU where the EXPORT_SYMBOL() macro is used to also contain DWARF type information for the symbols it exports. However, as a TU can also export external symbols and compilers may choose not to emit debugging information for symbols not defined in the current TU, the missing types will result in missing symbol versions. Stand-alone assembly code also doesn't contain type information for exported symbols, so we need to compile a temporary object file with asm-prototypes.h instead, and similarly need to ensure the DWARF in the temporary object file contains the necessary types. To always emit type information for external exports, add explicit __gendwarfksyms_ptr_ references to them in EXPORT_SYMBOL(). gendwarfksyms will use the type information for __gendwarfksyms_ptr_* if needed. Discard the pointers from the final binary to avoid further bloat. Signed-off-by: Sami Tolvanen Signed-off-by: Masahiro Yamada --- include/linux/export.h | 15 +++++++++++++++ 1 file changed, 15 insertions(+) diff --git a/include/linux/export.h b/include/linux/export.h index 2633df4d31e6..a8c23d945634 100644 --- a/include/linux/export.h +++ b/include/linux/export.h @@ -52,9 +52,24 @@ #else +#ifdef CONFIG_GENDWARFKSYMS +/* + * With CONFIG_GENDWARFKSYMS, ensure the compiler emits debugging + * information for all exported symbols, including those defined in + * different TUs, by adding a __gendwarfksyms_ptr_ pointer + * that's discarded during the final link. + */ +#define __GENDWARFKSYMS_EXPORT(sym) \ + static typeof(sym) *__gendwarfksyms_ptr_##sym __used \ + __section(".discard.gendwarfksyms") = &sym; +#else +#define __GENDWARFKSYMS_EXPORT(sym) +#endif + #define __EXPORT_SYMBOL(sym, license, ns) \ extern typeof(sym) sym; \ __ADDRESSABLE(sym) \ + __GENDWARFKSYMS_EXPORT(sym) \ asm(__stringify(___EXPORT_SYMBOL(sym, license, ns))) #endif From 9c3681f9b9fd12cdbc4a542df599f1837512f3d5 Mon Sep 17 00:00:00 2001 From: Sami Tolvanen Date: Fri, 3 Jan 2025 20:45:39 +0000 Subject: [PATCH 35/41] kbuild: Add gendwarfksyms as an alternative to genksyms When MODVERSIONS is enabled, allow selecting gendwarfksyms as the implementation, but default to genksyms. Signed-off-by: Sami Tolvanen Signed-off-by: Masahiro Yamada --- kernel/module/Kconfig | 22 ++++++++++++++++++++++ scripts/Makefile | 2 +- scripts/Makefile.build | 35 +++++++++++++++++++++++++++++------ 3 files changed, 52 insertions(+), 7 deletions(-) diff --git a/kernel/module/Kconfig b/kernel/module/Kconfig index 4637f063d0fc..d443fc504ffc 100644 --- a/kernel/module/Kconfig +++ b/kernel/module/Kconfig @@ -169,6 +169,22 @@ config MODVERSIONS make them incompatible with the kernel you are running. If unsure, say N. +choice + prompt "Module versioning implementation" + depends on MODVERSIONS + help + Select the tool used to calculate symbol versions for modules. + + If unsure, select GENKSYMS. + +config GENKSYMS + bool "genksyms (from source code)" + help + Calculate symbol versions from pre-processed source code using + genksyms. + + If unsure, say Y. + config GENDWARFKSYMS bool "gendwarfksyms (from debugging information)" depends on DEBUG_INFO @@ -176,6 +192,12 @@ config GENDWARFKSYMS depends on !DEBUG_INFO_REDUCED && !DEBUG_INFO_SPLIT # Requires ELF object files. depends on !LTO + help + Calculate symbol versions from DWARF debugging information using + gendwarfksyms. Requires DEBUG_INFO to be enabled. + + If unsure, say N. +endchoice config ASM_MODVERSIONS bool diff --git a/scripts/Makefile b/scripts/Makefile index d7fec46d38c0..8533f4498885 100644 --- a/scripts/Makefile +++ b/scripts/Makefile @@ -53,7 +53,7 @@ hostprogs += unifdef targets += module.lds subdir-$(CONFIG_GCC_PLUGINS) += gcc-plugins -subdir-$(CONFIG_MODVERSIONS) += genksyms +subdir-$(CONFIG_GENKSYMS) += genksyms subdir-$(CONFIG_GENDWARFKSYMS) += gendwarfksyms subdir-$(CONFIG_SECURITY_SELINUX) += selinux subdir-$(CONFIG_SECURITY_IPE) += ipe diff --git a/scripts/Makefile.build b/scripts/Makefile.build index c16e4cf54d77..81d9dacad03c 100644 --- a/scripts/Makefile.build +++ b/scripts/Makefile.build @@ -107,13 +107,24 @@ cmd_cpp_i_c = $(CPP) $(c_flags) -o $@ $< $(obj)/%.i: $(obj)/%.c FORCE $(call if_changed_dep,cpp_i_c) +getexportsymbols = $(NM) $@ | sed -n 's/.* __export_symbol_\(.*\)/$(1)/p' + +gendwarfksyms = $(objtree)/scripts/gendwarfksyms/gendwarfksyms \ + $(if $(KBUILD_SYMTYPES), --symtypes $(@:.o=.symtypes)) \ + $(if $(KBUILD_GENDWARFKSYMS_STABLE), --stable) + genksyms = $(objtree)/scripts/genksyms/genksyms \ $(if $(KBUILD_SYMTYPES), -T $(@:.o=.symtypes)) \ $(if $(KBUILD_PRESERVE), -p) \ $(addprefix -r , $(wildcard $(@:.o=.symref))) # These mirror gensymtypes_S and co below, keep them in synch. +ifdef CONFIG_GENDWARFKSYMS +cmd_gensymtypes_c = $(if $(skip_gendwarfksyms),, \ + $(call getexportsymbols,\1) | $(gendwarfksyms) $@) +else cmd_gensymtypes_c = $(CPP) -D__GENKSYMS__ $(c_flags) $< | $(genksyms) +endif # CONFIG_GENDWARFKSYMS # LLVM assembly # Generate .ll files from .c @@ -286,14 +297,26 @@ $(obj)/%.rs: $(obj)/%.rs.S FORCE # This is convoluted. The .S file must first be preprocessed to run guards and # expand names, then the resulting exports must be constructed into plain # EXPORT_SYMBOL(symbol); to build our dummy C file, and that gets preprocessed -# to make the genksyms input. +# to make the genksyms input or compiled into an object for gendwarfksyms. # # These mirror gensymtypes_c and co above, keep them in synch. -cmd_gensymtypes_S = \ - { echo "\#include " ; \ - echo "\#include " ; \ - $(NM) $@ | sed -n 's/.* __export_symbol_\(.*\)/EXPORT_SYMBOL(\1);/p' ; } | \ - $(CPP) -D__GENKSYMS__ $(c_flags) -xc - | $(genksyms) +getasmexports = \ + { echo "\#include " ; \ + echo "\#include " ; \ + echo "\#include " ; \ + $(call getexportsymbols,EXPORT_SYMBOL(\1);) ; } + +ifdef CONFIG_GENDWARFKSYMS +cmd_gensymtypes_S = \ + $(getasmexports) | \ + $(CC) $(c_flags) -c -o $(@:.o=.gendwarfksyms.o) -xc -; \ + $(call getexportsymbols,\1) | \ + $(gendwarfksyms) $(@:.o=.gendwarfksyms.o) +else +cmd_gensymtypes_S = \ + $(getasmexports) | \ + $(CPP) -D__GENKSYMS__ $(c_flags) -xc - | $(genksyms) +endif # CONFIG_GENDWARFKSYMS quiet_cmd_cpp_s_S = CPP $(quiet_modtag) $@ cmd_cpp_s_S = $(CPP) $(a_flags) -o $@ $< From 8c6d7b417f0fe69d7e29501db801838a54c6764b Mon Sep 17 00:00:00 2001 From: Sami Tolvanen Date: Fri, 3 Jan 2025 20:45:40 +0000 Subject: [PATCH 36/41] Documentation/kbuild: Add DWARF module versioning Add documentation for gendwarfksyms changes, and the kABI stability features that can be useful for distributions even though they're not used in mainline kernels. Signed-off-by: Sami Tolvanen Signed-off-by: Masahiro Yamada --- Documentation/kbuild/gendwarfksyms.rst | 308 +++++++++++++++++++++++++ Documentation/kbuild/index.rst | 1 + 2 files changed, 309 insertions(+) create mode 100644 Documentation/kbuild/gendwarfksyms.rst diff --git a/Documentation/kbuild/gendwarfksyms.rst b/Documentation/kbuild/gendwarfksyms.rst new file mode 100644 index 000000000000..e4beaae7e456 --- /dev/null +++ b/Documentation/kbuild/gendwarfksyms.rst @@ -0,0 +1,308 @@ +======================= +DWARF module versioning +======================= + +1. Introduction +=============== + +When CONFIG_MODVERSIONS is enabled, symbol versions for modules +are typically calculated from preprocessed source code using the +**genksyms** tool. However, this is incompatible with languages such +as Rust, where the source code has insufficient information about +the resulting ABI. With CONFIG_GENDWARFKSYMS (and CONFIG_DEBUG_INFO) +selected, **gendwarfksyms** is used instead to calculate symbol versions +from the DWARF debugging information, which contains the necessary +details about the final module ABI. + +1.1. Usage +========== + +gendwarfksyms accepts a list of object files on the command line, and a +list of symbol names (one per line) in standard input:: + + Usage: gendwarfksyms [options] elf-object-file ... < symbol-list + + Options: + -d, --debug Print debugging information + --dump-dies Dump DWARF DIE contents + --dump-die-map Print debugging information about die_map changes + --dump-types Dump type strings + --dump-versions Dump expanded type strings used for symbol versions + -s, --stable Support kABI stability features + -T, --symtypes file Write a symtypes file + -h, --help Print this message + + +2. Type information availability +================================ + +While symbols are typically exported in the same translation unit (TU) +where they're defined, it's also perfectly fine for a TU to export +external symbols. For example, this is done when calculating symbol +versions for exports in stand-alone assembly code. + +To ensure the compiler emits the necessary DWARF type information in the +TU where symbols are actually exported, gendwarfksyms adds a pointer +to exported symbols in the `EXPORT_SYMBOL()` macro using the following +macro:: + + #define __GENDWARFKSYMS_EXPORT(sym) \ + static typeof(sym) *__gendwarfksyms_ptr_##sym __used \ + __section(".discard.gendwarfksyms") = &sym; + + +When a symbol pointer is found in DWARF, gendwarfksyms can use its +type for calculating symbol versions even if the symbol is defined +elsewhere. The name of the symbol pointer is expected to start with +`__gendwarfksyms_ptr_`, followed by the name of the exported symbol. + +3. Symtypes output format +========================= + +Similarly to genksyms, gendwarfksyms supports writing a symtypes +file for each processed object that contain types for exported +symbols and each referenced type that was used in calculating symbol +versions. These files can be useful when trying to determine what +exactly caused symbol versions to change between builds. To generate +symtypes files during a kernel build, set `KBUILD_SYMTYPES=1`. + +Matching the existing format, the first column of each line contains +either a type reference or a symbol name. Type references have a +one-letter prefix followed by "#" and the name of the type. Four +reference types are supported:: + + e# = enum + s# = struct + t# = typedef + u# = union + +Type names with spaces in them are wrapped in single quotes, e.g.:: + + s#'core::result::Result' + +The rest of the line contains a type string. Unlike with genksyms that +produces C-style type strings, gendwarfksyms uses the same simple parsed +DWARF format produced by **--dump-dies**, but with type references +instead of fully expanded strings. + +4. Maintaining a stable kABI +============================ + +Distribution maintainers often need the ability to make ABI compatible +changes to kernel data structures due to LTS updates or backports. Using +the traditional `#ifndef __GENKSYMS__` to hide these changes from symbol +versioning won't work when processing object files. To support this +use case, gendwarfksyms provides kABI stability features designed to +hide changes that won't affect the ABI when calculating versions. These +features are all gated behind the **--stable** command line flag and are +not used in the mainline kernel. To use stable features during a kernel +build, set `KBUILD_GENDWARFKSYMS_STABLE=1`. + +Examples for using these features are provided in the +**scripts/gendwarfksyms/examples** directory, including helper macros +for source code annotation. Note that as these features are only used to +transform the inputs for symbol versioning, the user is responsible for +ensuring that their changes actually won't break the ABI. + +4.1. kABI rules +=============== + +kABI rules allow distributions to fine-tune certain parts +of gendwarfksyms output and thus control how symbol +versions are calculated. These rules are defined in the +`.discard.gendwarfksyms.kabi_rules` section of the object file and +consist of simple null-terminated strings with the following structure:: + + version\0type\0target\0value\0 + +This string sequence is repeated as many times as needed to express all +the rules. The fields are as follows: + +- `version`: Ensures backward compatibility for future changes to the + structure. Currently expected to be "1". +- `type`: Indicates the type of rule being applied. +- `target`: Specifies the target of the rule, typically the fully + qualified name of the DWARF Debugging Information Entry (DIE). +- `value`: Provides rule-specific data. + +The following helper macro, for example, can be used to specify rules +in the source code:: + + #define __KABI_RULE(hint, target, value) \ + static const char __PASTE(__gendwarfksyms_rule_, \ + __COUNTER__)[] __used __aligned(1) \ + __section(".discard.gendwarfksyms.kabi_rules") = \ + "1\0" #hint "\0" #target "\0" #value + + +Currently, only the rules discussed in this section are supported, but +the format is extensible enough to allow further rules to be added as +need arises. + +4.1.1. Managing definition visibility +===================================== + +A declaration can change into a full definition when additional includes +are pulled into the translation unit. This changes the versions of any +symbol that references the type even if the ABI remains unchanged. As +it may not be possible to drop includes without breaking the build, the +`declonly` rule can be used to specify a type as declaration-only, even +if the debugging information contains the full definition. + +The rule fields are expected to be as follows: + +- `type`: "declonly" +- `target`: The fully qualified name of the target data structure + (as shown in **--dump-dies** output). +- `value`: This field is ignored. + +Using the `__KABI_RULE` macro, this rule can be defined as:: + + #define KABI_DECLONLY(fqn) __KABI_RULE(declonly, fqn, ) + +Example usage:: + + struct s { + /* definition */ + }; + + KABI_DECLONLY(s); + +4.1.2. Adding enumerators +========================= + +For enums, all enumerators and their values are included in calculating +symbol versions, which becomes a problem if we later need to add more +enumerators without changing symbol versions. The `enumerator_ignore` +rule allows us to hide named enumerators from the input. + +The rule fields are expected to be as follows: + +- `type`: "enumerator_ignore" +- `target`: The fully qualified name of the target enum + (as shown in **--dump-dies** output) and the name of the + enumerator field separated by a space. +- `value`: This field is ignored. + +Using the `__KABI_RULE` macro, this rule can be defined as:: + + #define KABI_ENUMERATOR_IGNORE(fqn, field) \ + __KABI_RULE(enumerator_ignore, fqn field, ) + +Example usage:: + + enum e { + A, B, C, D, + }; + + KABI_ENUMERATOR_IGNORE(e, B); + KABI_ENUMERATOR_IGNORE(e, C); + +If the enum additionally includes an end marker and new values must +be added in the middle, we may need to use the old value for the last +enumerator when calculating versions. The `enumerator_value` rule allows +us to override the value of an enumerator for version calculation: + +- `type`: "enumerator_value" +- `target`: The fully qualified name of the target enum + (as shown in **--dump-dies** output) and the name of the + enumerator field separated by a space. +- `value`: Integer value used for the field. + +Using the `__KABI_RULE` macro, this rule can be defined as:: + + #define KABI_ENUMERATOR_VALUE(fqn, field, value) \ + __KABI_RULE(enumerator_value, fqn field, value) + +Example usage:: + + enum e { + A, B, C, LAST, + }; + + KABI_ENUMERATOR_IGNORE(e, C); + KABI_ENUMERATOR_VALUE(e, LAST, 2); + +4.3. Adding structure members +============================= + +Perhaps the most common ABI compatible change is adding a member to a +kernel data structure. When changes to a structure are anticipated, +distribution maintainers can pre-emptively reserve space in the +structure and take it into use later without breaking the ABI. If +changes are needed to data structures without reserved space, existing +alignment holes can potentially be used instead. While kABI rules could +be added for these type of changes, using unions is typically a more +natural method. This section describes gendwarfksyms support for using +reserved space in data structures and hiding members that don't change +the ABI when calculating symbol versions. + +4.3.1. Reserving space and replacing members +============================================ + +Space is typically reserved for later use by appending integer types, or +arrays, to the end of the data structure, but any type can be used. Each +reserved member needs a unique name, but as the actual purpose is usually +not known at the time the space is reserved, for convenience, names that +start with `__kabi_` are left out when calculating symbol versions:: + + struct s { + long a; + long __kabi_reserved_0; /* reserved for future use */ + }; + +The reserved space can be taken into use by wrapping the member in a +union, which includes the original type and the replacement member:: + + struct s { + long a; + union { + long __kabi_reserved_0; /* original type */ + struct b b; /* replaced field */ + }; + }; + +If the `__kabi_` naming scheme was used when reserving space, the name +of the first member of the union must start with `__kabi_reserved`. This +ensures the original type is used when calculating versions, but the name +is again left out. The rest of the union is ignored. + +If we're replacing a member that doesn't follow this naming convention, +we also need to preserve the original name to avoid changing versions, +which we can do by changing the first union member's name to start with +`__kabi_renamed` followed by the original name. + +The examples include `KABI_(RESERVE|USE|REPLACE)*` macros that help +simplify the process and also ensure the replacement member is correctly +aligned and its size won't exceed the reserved space. + +4.3.2. Hiding members +===================== + +Predicting which structures will require changes during the support +timeframe isn't always possible, in which case one might have to resort +to placing new members into existing alignment holes:: + + struct s { + int a; + /* a 4-byte alignment hole */ + unsigned long b; + }; + + +While this won't change the size of the data structure, one needs to +be able to hide the added members from symbol versioning. Similarly +to reserved fields, this can be accomplished by wrapping the added +member to a union where one of the fields has a name starting with +`__kabi_ignored`:: + + struct s { + int a; + union { + char __kabi_ignored_0; + int n; + }; + unsigned long b; + }; + +With **--stable**, both versions produce the same symbol version. diff --git a/Documentation/kbuild/index.rst b/Documentation/kbuild/index.rst index cee2f99f734b..e82af05cd652 100644 --- a/Documentation/kbuild/index.rst +++ b/Documentation/kbuild/index.rst @@ -21,6 +21,7 @@ Kernel Build System reproducible-builds gcc-plugins llvm + gendwarfksyms .. only:: subproject and html From 54ac1ac8edeb74ff87fc880d1ee58785bdcbe323 Mon Sep 17 00:00:00 2001 From: Matthew Maurer Date: Fri, 3 Jan 2025 17:37:01 +0000 Subject: [PATCH 37/41] modules: Support extended MODVERSIONS info Adds a new format for MODVERSIONS which stores each field in a separate ELF section. This initially adds support for variable length names, but could later be used to add additional fields to MODVERSIONS in a backwards compatible way if needed. Any new fields will be ignored by old user tooling, unlike the current format where user tooling cannot tolerate adjustments to the format (for example making the name field longer). Since PPC munges its version records to strip leading dots, we reproduce the munging for the new format. Other architectures do not appear to have architecture-specific usage of this information. Reviewed-by: Sami Tolvanen Signed-off-by: Matthew Maurer Signed-off-by: Masahiro Yamada --- arch/powerpc/kernel/module_64.c | 24 ++++++++- kernel/module/internal.h | 11 ++++ kernel/module/main.c | 92 ++++++++++++++++++++++++++++++--- kernel/module/version.c | 45 ++++++++++++++++ 4 files changed, 162 insertions(+), 10 deletions(-) diff --git a/arch/powerpc/kernel/module_64.c b/arch/powerpc/kernel/module_64.c index 45dac7b46aa3..34a5aec4908f 100644 --- a/arch/powerpc/kernel/module_64.c +++ b/arch/powerpc/kernel/module_64.c @@ -369,6 +369,24 @@ static void dedotify_versions(struct modversion_info *vers, } } +/* Same as normal versions, remove a leading dot if present. */ +static void dedotify_ext_version_names(char *str_seq, unsigned long size) +{ + unsigned long out = 0; + unsigned long in; + char last = '\0'; + + for (in = 0; in < size; in++) { + /* Skip one leading dot */ + if (last == '\0' && str_seq[in] == '.') + in++; + last = str_seq[in]; + str_seq[out++] = last; + } + /* Zero the trailing portion of the names table for robustness */ + memset(&str_seq[out], 0, size - out); +} + /* * Undefined symbols which refer to .funcname, hack to funcname. Make .TOC. * seem to be defined (value set later). @@ -438,10 +456,12 @@ int module_frob_arch_sections(Elf64_Ehdr *hdr, me->arch.toc_section = i; if (sechdrs[i].sh_addralign < 8) sechdrs[i].sh_addralign = 8; - } - else if (strcmp(secstrings+sechdrs[i].sh_name,"__versions")==0) + } else if (strcmp(secstrings + sechdrs[i].sh_name, "__versions") == 0) dedotify_versions((void *)hdr + sechdrs[i].sh_offset, sechdrs[i].sh_size); + else if (strcmp(secstrings + sechdrs[i].sh_name, "__version_ext_names") == 0) + dedotify_ext_version_names((void *)hdr + sechdrs[i].sh_offset, + sechdrs[i].sh_size); if (sechdrs[i].sh_type == SHT_SYMTAB) dedotify((void *)hdr + sechdrs[i].sh_offset, diff --git a/kernel/module/internal.h b/kernel/module/internal.h index f10dc3ea7ff8..887838589020 100644 --- a/kernel/module/internal.h +++ b/kernel/module/internal.h @@ -86,6 +86,8 @@ struct load_info { unsigned int vers; unsigned int info; unsigned int pcpu; + unsigned int vers_ext_crc; + unsigned int vers_ext_name; } index; }; @@ -389,6 +391,15 @@ void module_layout(struct module *mod, struct modversion_info *ver, struct kerne struct kernel_symbol *ks, struct tracepoint * const *tp); int check_modstruct_version(const struct load_info *info, struct module *mod); int same_magic(const char *amagic, const char *bmagic, bool has_crcs); +struct modversion_info_ext { + size_t remaining; + const u32 *crc; + const char *name; +}; +void modversion_ext_start(const struct load_info *info, struct modversion_info_ext *ver); +void modversion_ext_advance(struct modversion_info_ext *ver); +#define for_each_modversion_info_ext(ver, info) \ + for (modversion_ext_start(info, &ver); ver.remaining > 0; modversion_ext_advance(&ver)) #else /* !CONFIG_MODVERSIONS */ static inline int check_version(const struct load_info *info, const char *symname, diff --git a/kernel/module/main.c b/kernel/module/main.c index e58bff88b8d6..0830892070fe 100644 --- a/kernel/module/main.c +++ b/kernel/module/main.c @@ -2073,6 +2073,82 @@ static int elf_validity_cache_index_str(struct load_info *info) return 0; } +/** + * elf_validity_cache_index_versions() - Validate and cache version indices + * @info: Load info to cache version indices in. + * Must have &load_info->sechdrs and &load_info->secstrings populated. + * @flags: Load flags, relevant to suppress version loading, see + * uapi/linux/module.h + * + * If we're ignoring modversions based on @flags, zero all version indices + * and return validity. Othewrise check: + * + * * If "__version_ext_crcs" is present, "__version_ext_names" is present + * * There is a name present for every crc + * + * Then populate: + * + * * &load_info->index.vers + * * &load_info->index.vers_ext_crc + * * &load_info->index.vers_ext_names + * + * if present. + * + * Return: %0 if valid, %-ENOEXEC on failure. + */ +static int elf_validity_cache_index_versions(struct load_info *info, int flags) +{ + unsigned int vers_ext_crc; + unsigned int vers_ext_name; + size_t crc_count; + size_t remaining_len; + size_t name_size; + char *name; + + /* If modversions were suppressed, pretend we didn't find any */ + if (flags & MODULE_INIT_IGNORE_MODVERSIONS) { + info->index.vers = 0; + info->index.vers_ext_crc = 0; + info->index.vers_ext_name = 0; + return 0; + } + + vers_ext_crc = find_sec(info, "__version_ext_crcs"); + vers_ext_name = find_sec(info, "__version_ext_names"); + + /* If we have one field, we must have the other */ + if (!!vers_ext_crc != !!vers_ext_name) { + pr_err("extended version crc+name presence does not match"); + return -ENOEXEC; + } + + /* + * If we have extended version information, we should have the same + * number of entries in every section. + */ + if (vers_ext_crc) { + crc_count = info->sechdrs[vers_ext_crc].sh_size / sizeof(u32); + name = (void *)info->hdr + + info->sechdrs[vers_ext_name].sh_offset; + remaining_len = info->sechdrs[vers_ext_name].sh_size; + + while (crc_count--) { + name_size = strnlen(name, remaining_len) + 1; + if (name_size > remaining_len) { + pr_err("more extended version crcs than names"); + return -ENOEXEC; + } + remaining_len -= name_size; + name += name_size; + } + } + + info->index.vers = find_sec(info, "__versions"); + info->index.vers_ext_crc = vers_ext_crc; + info->index.vers_ext_name = vers_ext_name; + return 0; +} + /** * elf_validity_cache_index() - Resolve, validate, cache section indices * @info: Load info to read from and update. @@ -2087,9 +2163,7 @@ static int elf_validity_cache_index_str(struct load_info *info) * * elf_validity_cache_index_mod() * * elf_validity_cache_index_sym() * * elf_validity_cache_index_str() - * - * If versioning is not suppressed via flags, load the version index from - * a section called "__versions" with no validation. + * * elf_validity_cache_index_versions() * * If CONFIG_SMP is enabled, load the percpu section by name with no * validation. @@ -2112,11 +2186,9 @@ static int elf_validity_cache_index(struct load_info *info, int flags) err = elf_validity_cache_index_str(info); if (err < 0) return err; - - if (flags & MODULE_INIT_IGNORE_MODVERSIONS) - info->index.vers = 0; /* Pretend no __versions section! */ - else - info->index.vers = find_sec(info, "__versions"); + err = elf_validity_cache_index_versions(info, flags); + if (err < 0) + return err; info->index.pcpu = find_pcpusec(info); @@ -2327,6 +2399,10 @@ static int rewrite_section_headers(struct load_info *info, int flags) /* Track but don't keep modinfo and version sections. */ info->sechdrs[info->index.vers].sh_flags &= ~(unsigned long)SHF_ALLOC; + info->sechdrs[info->index.vers_ext_crc].sh_flags &= + ~(unsigned long)SHF_ALLOC; + info->sechdrs[info->index.vers_ext_name].sh_flags &= + ~(unsigned long)SHF_ALLOC; info->sechdrs[info->index.info].sh_flags &= ~(unsigned long)SHF_ALLOC; return 0; diff --git a/kernel/module/version.c b/kernel/module/version.c index 4e5731d403af..3718a8868321 100644 --- a/kernel/module/version.c +++ b/kernel/module/version.c @@ -19,11 +19,28 @@ int check_version(const struct load_info *info, unsigned int versindex = info->index.vers; unsigned int i, num_versions; struct modversion_info *versions; + struct modversion_info_ext version_ext; /* Exporting module didn't supply crcs? OK, we're already tainted. */ if (!crc) return 1; + /* If we have extended version info, rely on it */ + if (info->index.vers_ext_crc) { + for_each_modversion_info_ext(version_ext, info) { + if (strcmp(version_ext.name, symname) != 0) + continue; + if (*version_ext.crc == *crc) + return 1; + pr_debug("Found checksum %X vs module %X\n", + *crc, *version_ext.crc); + goto bad_version; + } + pr_warn_once("%s: no extended symbol version for %s\n", + info->name, symname); + return 1; + } + /* No versions at all? modprobe --force does this. */ if (versindex == 0) return try_to_force_load(mod, symname) == 0; @@ -87,6 +104,34 @@ int same_magic(const char *amagic, const char *bmagic, return strcmp(amagic, bmagic) == 0; } +void modversion_ext_start(const struct load_info *info, + struct modversion_info_ext *start) +{ + unsigned int crc_idx = info->index.vers_ext_crc; + unsigned int name_idx = info->index.vers_ext_name; + Elf_Shdr *sechdrs = info->sechdrs; + + /* + * Both of these fields are needed for this to be useful + * Any future fields should be initialized to NULL if absent. + */ + if (crc_idx == 0 || name_idx == 0) { + start->remaining = 0; + return; + } + + start->crc = (const u32 *)sechdrs[crc_idx].sh_addr; + start->name = (const char *)sechdrs[name_idx].sh_addr; + start->remaining = sechdrs[crc_idx].sh_size / sizeof(*start->crc); +} + +void modversion_ext_advance(struct modversion_info_ext *vers) +{ + vers->remaining--; + vers->crc++; + vers->name += strlen(vers->name) + 1; +} + /* * Generate the signature for all relevant module structures here. * If these change, we don't want to try to parse the module. From fc7d5e3210ae083a29ce224ffce18eaf3d1c645a Mon Sep 17 00:00:00 2001 From: Matthew Maurer Date: Fri, 3 Jan 2025 17:37:02 +0000 Subject: [PATCH 38/41] modpost: Produce extended MODVERSIONS information Generate both the existing modversions format and the new extended one when running modpost. Presence of this metadata in the final .ko is guarded by CONFIG_EXTENDED_MODVERSIONS. We no longer generate an error on long symbols in modpost if CONFIG_EXTENDED_MODVERSIONS is set, as they can now be appropriately encoded in the extended section. These symbols will be skipped in the previous encoding. An error will still be generated if CONFIG_EXTENDED_MODVERSIONS is not set. Reviewed-by: Sami Tolvanen Signed-off-by: Matthew Maurer Signed-off-by: Masahiro Yamada --- kernel/module/Kconfig | 10 +++++++ scripts/Makefile.modpost | 1 + scripts/mod/modpost.c | 62 +++++++++++++++++++++++++++++++++++++--- 3 files changed, 69 insertions(+), 4 deletions(-) diff --git a/kernel/module/Kconfig b/kernel/module/Kconfig index d443fc504ffc..9568b629a03c 100644 --- a/kernel/module/Kconfig +++ b/kernel/module/Kconfig @@ -207,6 +207,16 @@ config ASM_MODVERSIONS assembly. This can be enabled only when the target architecture supports it. +config EXTENDED_MODVERSIONS + bool "Extended Module Versioning Support" + depends on MODVERSIONS + help + This enables extended MODVERSIONs support, allowing long symbol + names to be versioned. + + The most likely reason you would enable this is to enable Rust + support. If unsure, say N. + config MODULE_SRCVERSION_ALL bool "Source checksum for all modules" help diff --git a/scripts/Makefile.modpost b/scripts/Makefile.modpost index ab0e94ea6249..40426fc63509 100644 --- a/scripts/Makefile.modpost +++ b/scripts/Makefile.modpost @@ -43,6 +43,7 @@ MODPOST = $(objtree)/scripts/mod/modpost modpost-args = \ $(if $(CONFIG_MODULES),-M) \ $(if $(CONFIG_MODVERSIONS),-m) \ + $(if $(CONFIG_EXTENDED_MODVERSIONS),-x) \ $(if $(CONFIG_MODULE_SRCVERSION_ALL),-a) \ $(if $(CONFIG_SECTION_MISMATCH_WARN_ONLY),,-E) \ $(if $(KBUILD_MODPOST_WARN),-w) \ diff --git a/scripts/mod/modpost.c b/scripts/mod/modpost.c index dc907014108b..38ff3dd4a9a1 100644 --- a/scripts/mod/modpost.c +++ b/scripts/mod/modpost.c @@ -33,6 +33,8 @@ static bool module_enabled; static bool modversions; /* Is CONFIG_MODULE_SRCVERSION_ALL set? */ static bool all_versions; +/* Is CONFIG_EXTENDED_MODVERSIONS set? */ +static bool extended_modversions; /* If we are modposting external module set to 1 */ static bool external_module; /* Only warn about unresolved symbols */ @@ -1805,6 +1807,49 @@ static void add_exported_symbols(struct buffer *buf, struct module *mod) } } +/** + * Record CRCs for unresolved symbols, supporting long names + */ +static void add_extended_versions(struct buffer *b, struct module *mod) +{ + struct symbol *s; + + if (!extended_modversions) + return; + + buf_printf(b, "\n"); + buf_printf(b, "static const u32 ____version_ext_crcs[]\n"); + buf_printf(b, "__used __section(\"__version_ext_crcs\") = {\n"); + list_for_each_entry(s, &mod->unresolved_symbols, list) { + if (!s->module) + continue; + if (!s->crc_valid) { + warn("\"%s\" [%s.ko] has no CRC!\n", + s->name, mod->name); + continue; + } + buf_printf(b, "\t0x%08x,\n", s->crc); + } + buf_printf(b, "};\n"); + + buf_printf(b, "static const char ____version_ext_names[]\n"); + buf_printf(b, "__used __section(\"__version_ext_names\") =\n"); + list_for_each_entry(s, &mod->unresolved_symbols, list) { + if (!s->module) + continue; + if (!s->crc_valid) + /* + * We already warned on this when producing the crc + * table. + * We need to skip its name too, as the indexes in + * both tables need to align. + */ + continue; + buf_printf(b, "\t\"%s\\0\"\n", s->name); + } + buf_printf(b, ";\n"); +} + /** * Record CRCs for unresolved symbols **/ @@ -1828,9 +1873,14 @@ static void add_versions(struct buffer *b, struct module *mod) continue; } if (strlen(s->name) >= MODULE_NAME_LEN) { - error("too long symbol \"%s\" [%s.ko]\n", - s->name, mod->name); - break; + if (extended_modversions) { + /* this symbol will only be in the extended info */ + continue; + } else { + error("too long symbol \"%s\" [%s.ko]\n", + s->name, mod->name); + break; + } } buf_printf(b, "\t{ 0x%08x, \"%s\" },\n", s->crc, s->name); @@ -1961,6 +2011,7 @@ static void write_mod_c_file(struct module *mod) add_header(&buf, mod); add_exported_symbols(&buf, mod); add_versions(&buf, mod); + add_extended_versions(&buf, mod); add_depends(&buf, mod); buf_printf(&buf, "\n"); @@ -2126,7 +2177,7 @@ int main(int argc, char **argv) LIST_HEAD(dump_lists); struct dump_list *dl, *dl2; - while ((opt = getopt(argc, argv, "ei:MmnT:to:au:WwENd:")) != -1) { + while ((opt = getopt(argc, argv, "ei:MmnT:to:au:WwENd:x")) != -1) { switch (opt) { case 'e': external_module = true; @@ -2175,6 +2226,9 @@ int main(int argc, char **argv) case 'd': missing_namespace_deps = optarg; break; + case 'x': + extended_modversions = true; + break; default: exit(1); } From e8639b7ef0f871753b4262ec0eacd3da29eebcee Mon Sep 17 00:00:00 2001 From: Matthew Maurer Date: Fri, 3 Jan 2025 17:37:03 +0000 Subject: [PATCH 39/41] modpost: Allow extended modversions without basic MODVERSIONS If you know that your kernel modules will only ever be loaded by a newer kernel, you can disable BASIC_MODVERSIONS to save space. This also allows easy creation of test modules to see how tooling will respond to modules that only have the new format. Signed-off-by: Matthew Maurer Signed-off-by: Masahiro Yamada --- kernel/module/Kconfig | 15 +++++++++++++++ scripts/Makefile.modpost | 1 + scripts/mod/modpost.c | 9 +++++++-- 3 files changed, 23 insertions(+), 2 deletions(-) diff --git a/kernel/module/Kconfig b/kernel/module/Kconfig index 9568b629a03c..4538f3af63e1 100644 --- a/kernel/module/Kconfig +++ b/kernel/module/Kconfig @@ -217,6 +217,21 @@ config EXTENDED_MODVERSIONS The most likely reason you would enable this is to enable Rust support. If unsure, say N. +config BASIC_MODVERSIONS + bool "Basic Module Versioning Support" + depends on MODVERSIONS + default y + help + This enables basic MODVERSIONS support, allowing older tools or + kernels to potentially load modules. + + Disabling this may cause older `modprobe` or `kmod` to be unable + to read MODVERSIONS information from built modules. With this + disabled, older kernels may treat this module as unversioned. + + This is enabled by default when MODVERSIONS are enabled. + If unsure, say Y. + config MODULE_SRCVERSION_ALL bool "Source checksum for all modules" help diff --git a/scripts/Makefile.modpost b/scripts/Makefile.modpost index 40426fc63509..d7d45067d08b 100644 --- a/scripts/Makefile.modpost +++ b/scripts/Makefile.modpost @@ -43,6 +43,7 @@ MODPOST = $(objtree)/scripts/mod/modpost modpost-args = \ $(if $(CONFIG_MODULES),-M) \ $(if $(CONFIG_MODVERSIONS),-m) \ + $(if $(CONFIG_BASIC_MODVERSIONS),-b) \ $(if $(CONFIG_EXTENDED_MODVERSIONS),-x) \ $(if $(CONFIG_MODULE_SRCVERSION_ALL),-a) \ $(if $(CONFIG_SECTION_MISMATCH_WARN_ONLY),,-E) \ diff --git a/scripts/mod/modpost.c b/scripts/mod/modpost.c index 38ff3dd4a9a1..e18ae7dc8140 100644 --- a/scripts/mod/modpost.c +++ b/scripts/mod/modpost.c @@ -33,6 +33,8 @@ static bool module_enabled; static bool modversions; /* Is CONFIG_MODULE_SRCVERSION_ALL set? */ static bool all_versions; +/* Is CONFIG_BASIC_MODVERSIONS set? */ +static bool basic_modversions; /* Is CONFIG_EXTENDED_MODVERSIONS set? */ static bool extended_modversions; /* If we are modposting external module set to 1 */ @@ -1857,7 +1859,7 @@ static void add_versions(struct buffer *b, struct module *mod) { struct symbol *s; - if (!modversions) + if (!basic_modversions) return; buf_printf(b, "\n"); @@ -2177,7 +2179,7 @@ int main(int argc, char **argv) LIST_HEAD(dump_lists); struct dump_list *dl, *dl2; - while ((opt = getopt(argc, argv, "ei:MmnT:to:au:WwENd:x")) != -1) { + while ((opt = getopt(argc, argv, "ei:MmnT:to:au:WwENd:xb")) != -1) { switch (opt) { case 'e': external_module = true; @@ -2226,6 +2228,9 @@ int main(int argc, char **argv) case 'd': missing_namespace_deps = optarg; break; + case 'b': + basic_modversions = true; + break; case 'x': extended_modversions = true; break; From 272f8a6d625a0cf7fba9c5af5202edc84dee326c Mon Sep 17 00:00:00 2001 From: Matthew Maurer Date: Fri, 3 Jan 2025 17:37:04 +0000 Subject: [PATCH 40/41] Documentation/kbuild: Document storage of symbol information Document where exported and imported symbols are kept, format options, and limitations. Signed-off-by: Matthew Maurer Signed-off-by: Masahiro Yamada --- Documentation/kbuild/modules.rst | 20 ++++++++++++++++++++ 1 file changed, 20 insertions(+) diff --git a/Documentation/kbuild/modules.rst b/Documentation/kbuild/modules.rst index 101de236cd0c..a42f00d8cb90 100644 --- a/Documentation/kbuild/modules.rst +++ b/Documentation/kbuild/modules.rst @@ -423,6 +423,26 @@ Symbols From the Kernel (vmlinux + modules) 1) It lists all exported symbols from vmlinux and all modules. 2) It lists the CRC if CONFIG_MODVERSIONS is enabled. +Version Information Formats +--------------------------- + + Exported symbols have information stored in __ksymtab or __ksymtab_gpl + sections. Symbol names and namespaces are stored in __ksymtab_strings, + using a format similar to the string table used for ELF. If + CONFIG_MODVERSIONS is enabled, the CRCs corresponding to exported + symbols will be added to the __kcrctab or __kcrctab_gpl. + + If CONFIG_BASIC_MODVERSIONS is enabled (default with + CONFIG_MODVERSIONS), imported symbols will have their symbol name and + CRC stored in the __versions section of the importing module. This + mode only supports symbols of length up to 64 bytes. + + If CONFIG_EXTENDED_MODVERSIONS is enabled (required to enable both + CONFIG_MODVERSIONS and CONFIG_RUST at the same time), imported symbols + will have their symbol name recorded in the __version_ext_names + section as a series of concatenated, null-terminated strings. CRCs for + these symbols will be recorded in the __version_ext_crcs section. + Symbols and External Modules ---------------------------- From be3d3e40bb5691ba880304222fa7baeee6cd71cc Mon Sep 17 00:00:00 2001 From: Sami Tolvanen Date: Fri, 3 Jan 2025 17:37:05 +0000 Subject: [PATCH 41/41] rust: Use gendwarfksyms + extended modversions for CONFIG_MODVERSIONS Previously, two things stopped Rust from using MODVERSIONS: 1. Rust symbols are occasionally too long to be represented in the original versions table 2. Rust types cannot be properly hashed by the existing genksyms approach because: * Looking up type definitions in Rust is more complex than C * Type layout is potentially dependent on the compiler in Rust, not just the source type declaration. CONFIG_EXTENDED_MODVERSIONS addresses the first point, and CONFIG_GENDWARFKSYMS the second. If Rust wants to use MODVERSIONS, allow it to do so by selecting both features. Signed-off-by: Sami Tolvanen Co-developed-by: Matthew Maurer Signed-off-by: Matthew Maurer Signed-off-by: Masahiro Yamada --- init/Kconfig | 3 ++- rust/Makefile | 34 ++++++++++++++++++++++++++++++++-- 2 files changed, 34 insertions(+), 3 deletions(-) diff --git a/init/Kconfig b/init/Kconfig index a20e6efd3f0f..2cfbefe0933e 100644 --- a/init/Kconfig +++ b/init/Kconfig @@ -1956,7 +1956,8 @@ config RUST bool "Rust support" depends on HAVE_RUST depends on RUST_IS_AVAILABLE - depends on !MODVERSIONS + select EXTENDED_MODVERSIONS if MODVERSIONS + depends on !MODVERSIONS || GENDWARFKSYMS depends on !GCC_PLUGIN_RANDSTRUCT depends on !RANDSTRUCT depends on !DEBUG_INFO_BTF || PAHOLE_HAS_LANG_EXCLUDE diff --git a/rust/Makefile b/rust/Makefile index a40a3936126d..80f970ad81f7 100644 --- a/rust/Makefile +++ b/rust/Makefile @@ -329,10 +329,11 @@ $(obj)/bindings/bindings_helpers_generated.rs: private bindgen_target_extra = ; $(obj)/bindings/bindings_helpers_generated.rs: $(src)/helpers/helpers.c FORCE $(call if_changed_dep,bindgen) +rust_exports = $(NM) -p --defined-only $(1) | awk '$$2~/(T|R|D|B)/ && $$3!~/__cfi/ { printf $(2),$(3) }' + quiet_cmd_exports = EXPORTS $@ cmd_exports = \ - $(NM) -p --defined-only $< \ - | awk '$$2~/(T|R|D|B)/ && $$3!~/__cfi/ {printf "EXPORT_SYMBOL_RUST_GPL(%s);\n",$$3}' > $@ + $(call rust_exports,$<,"EXPORT_SYMBOL_RUST_GPL(%s);\n",$$3) > $@ $(obj)/exports_core_generated.h: $(obj)/core.o FORCE $(call if_changed,exports) @@ -401,11 +402,36 @@ ifneq ($(or $(CONFIG_ARM64),$(and $(CONFIG_RISCV),$(CONFIG_64BIT))),) __ashlti3 __lshrti3 endif +ifdef CONFIG_MODVERSIONS +cmd_gendwarfksyms = $(if $(skip_gendwarfksyms),, \ + $(call rust_exports,$@,"%s\n",$$3) | \ + scripts/gendwarfksyms/gendwarfksyms \ + $(if $(KBUILD_GENDWARFKSYMS_STABLE), --stable) \ + $(if $(KBUILD_SYMTYPES), --symtypes $(@:.o=.symtypes),) \ + $@ >> $(dot-target).cmd) +endif + define rule_rustc_library $(call cmd_and_fixdep,rustc_library) $(call cmd,gen_objtooldep) + $(call cmd,gendwarfksyms) endef +define rule_rust_cc_library + $(call if_changed_rule,cc_o_c) + $(call cmd,force_checksrc) + $(call cmd,gendwarfksyms) +endef + +# helpers.o uses the same export mechanism as Rust libraries, so ensure symbol +# versions are calculated for the helpers too. +$(obj)/helpers/helpers.o: $(src)/helpers/helpers.c $(recordmcount_source) FORCE + +$(call if_changed_rule,rust_cc_library) + +# Disable symbol versioning for exports.o to avoid conflicts with the actual +# symbol versions generated from Rust objects. +$(obj)/exports.o: private skip_gendwarfksyms = 1 + $(obj)/core.o: private skip_clippy = 1 $(obj)/core.o: private skip_flags = -Wunreachable_pub $(obj)/core.o: private rustc_objcopy = $(foreach sym,$(redirect-intrinsics),--redefine-sym $(sym)=__rust$(sym)) @@ -417,13 +443,16 @@ ifneq ($(or $(CONFIG_X86_64),$(CONFIG_X86_32)),) $(obj)/core.o: scripts/target.json endif +$(obj)/compiler_builtins.o: private skip_gendwarfksyms = 1 $(obj)/compiler_builtins.o: private rustc_objcopy = -w -W '__*' $(obj)/compiler_builtins.o: $(src)/compiler_builtins.rs $(obj)/core.o FORCE +$(call if_changed_rule,rustc_library) +$(obj)/build_error.o: private skip_gendwarfksyms = 1 $(obj)/build_error.o: $(src)/build_error.rs $(obj)/compiler_builtins.o FORCE +$(call if_changed_rule,rustc_library) +$(obj)/ffi.o: private skip_gendwarfksyms = 1 $(obj)/ffi.o: $(src)/ffi.rs $(obj)/compiler_builtins.o FORCE +$(call if_changed_rule,rustc_library) @@ -435,6 +464,7 @@ $(obj)/bindings.o: $(src)/bindings/lib.rs \ +$(call if_changed_rule,rustc_library) $(obj)/uapi.o: private rustc_target_flags = --extern ffi +$(obj)/uapi.o: private skip_gendwarfksyms = 1 $(obj)/uapi.o: $(src)/uapi/lib.rs \ $(obj)/ffi.o \ $(obj)/uapi/uapi_generated.rs FORCE