| lists.openwall.net | lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC | |
|
Open Source and information security mailing list archives
| ||
|
Message-ID: <f50cd6ae-0d99-4836-b268-c9c86da448d8@alu.unizg.hr> Date: Sat, 23 Nov 2024 02:53:09 +0100 From: Mirsad Todorovac <mtodorovac69@...il.com> To: Thomas Weißschuh <thomas@...ch.de>, Mirsad Todorovac <mtodorovac69@...il.com> Cc: Linux Kernel Build System <linux-kbuild@...r.kernel.org>, Masahiro Yamada <masahiroy@...nel.org>, Linux Kernel Mailing List <linux-kernel@...r.kernel.org>, Bagas Sanjaya <bagasdotme@...il.com> Subject: Re: [PROBLEM] make bindeb-pkg: When just one source line in a single driver is changed, all driver are rebuilt? Hi, Mr. Weißschuh, On 11/21/24 22:43, Thomas Weißschuh wrote: > Hi Mirsad, > > On 2024-11-21 22:18:58+0100, Mirsad Todorovac wrote: >> I am perplexed now at 6.12 stable vanilla tree, and I've noticed that after changing just one >> line in ./drivers/gpu/drm/xe/tests/xe_migrate.c:226:5-11, all unrelated drivers are being BTF'd, SIGNED, >> and INSTALL-ed: >> >> SIGN debian/linux-image-6.12.0-dirty/lib/modules/6.12.0-dirty/kernel/drivers/hwmon/aht10.ko >> INSTALL debian/linux-image-6.12.0-dirty/lib/modules/6.12.0-dirty/kernel/drivers/hwmon/coretemp.ko >> STRIP debian/linux-image-6.12.0-dirty/lib/modules/6.12.0-dirty/kernel/drivers/hwmon/atxp1.ko >> INSTALL debian/linux-image-6.12.0-dirty/lib/modules/6.12.0-dirty/kernel/drivers/hwmon/corsair-cpro.ko >> SIGN debian/linux-image-6.12.0-dirty/lib/modules/6.12.0-dirty/kernel/drivers/hwmon/adt7475.ko >> STRIP debian/linux-image-6.12.0-dirty/lib/modules/6.12.0-dirty/kernel/drivers/hwmon/axi-fan-control.ko >> SIGN debian/linux-image-6.12.0-dirty/lib/modules/6.12.0-dirty/kernel/drivers/hwmon/as370-hwmon.ko >> STRIP debian/linux-image-6.12.0-dirty/lib/modules/6.12.0-dirty/kernel/drivers/hwmon/chipcap2.ko >> SIGN debian/linux-image-6.12.0-dirty/lib/modules/6.12.0-dirty/kernel/drivers/hwmon/aquacomputer_d5next.ko >> STRIP debian/linux-image-6.12.0-dirty/lib/modules/6.12.0-dirty/kernel/drivers/hwmon/coretemp.ko >> SIGN debian/linux-image-6.12.0-dirty/lib/modules/6.12.0-dirty/kernel/drivers/hwmon/applesmc.ko >> INSTALL debian/linux-image-6.12.0-dirty/lib/modules/6.12.0-dirty/kernel/drivers/hwmon/corsair-psu.ko >> STRIP debian/linux-image-6.12.0-dirty/lib/modules/6.12.0-dirty/kernel/drivers/hwmon/corsair-cpro.ko >> SIGN debian/linux-image-6.12.0-dirty/lib/modules/6.12.0-dirty/kernel/drivers/hwmon/asus_rog_ryujin.ko >> INSTALL debian/linux-image-6.12.0-dirty/lib/modules/6.12.0-dirty/kernel/drivers/hwmon/cros_ec_hwmon.ko >> SIGN debian/linux-image-6.12.0-dirty/lib/modules/6.12.0-dirty/kernel/drivers/hwmon/asc7621.ko >> INSTALL debian/linux-image-6.12.0-dirty/lib/modules/6.12.0-dirty/kernel/drivers/hwmon/da9052-hwmon.ko >> SIGN debian/linux-image-6.12.0-dirty/lib/modules/6.12.0-dirty/kernel/drivers/hwmon/atxp1.ko >> SIGN debian/linux-image-6.12.0-dirty/lib/modules/6.12.0-dirty/kernel/drivers/hwmon/axi-fan-control.ko >> STRIP debian/linux-image-6.12.0-dirty/lib/modules/6.12.0-dirty/kernel/drivers/hwmon/corsair-psu.ko >> INSTALL debian/linux-image-6.12.0-dirty/lib/modules/6.12.0-dirty/kernel/drivers/hwmon/da9055-hwmon.ko >> SIGN debian/linux-image-6.12.0-dirty/lib/modules/6.12.0-dirty/kernel/drivers/hwmon/chipcap2.ko >> STRIP debian/linux-image-6.12.0-dirty/lib/modules/6.12.0-dirty/kernel/drivers/hwmon/cros_ec_hwmon.ko >> STRIP debian/linux-image-6.12.0-dirty/lib/modules/6.12.0-dirty/kernel/drivers/hwmon/da9052-hwmon.ko >> SIGN debian/linux-image-6.12.0-dirty/lib/modules/6.12.0-dirty/kernel/drivers/hwmon/coretemp.ko >> SIGN debian/linux-image-6.12.0-dirty/lib/modules/6.12.0-dirty/kernel/drivers/hwmon/corsair-cpro.ko >> INSTALL debian/linux-image-6.12.0-dirty/lib/modules/6.12.0-dirty/kernel/drivers/hwmon/dell-smm-hwmon.ko > This happens when the vermagic is changed. The vermagic is linked into > each module. And after relinking all the other steps also need to be > done. > Building the Debian package increases the build number which ends up in > the vermagic and therefore all modules. > Also there is CONFIG_LOCALVERSION_AUTO which may be involved. Yes, I use CONFIG_LOCALVERSION_AUTO because it saves a lot of trouble in bisecting. This vermagic manipulation is beyond my grasp. I understand that modules must match kernel version, for internal ABI often changes. But not between builds of the same source? Am I thinking right? >> Is there an option to rebuild just the necessary driver that was modified? > With the whole Debian packaging: No idea. > With plain kbuild: make $MAKEFLAGS $PATH_TO_MODULE.ko This could speed up some fixes, thanks :-) >> 2. >> >> >From the list of files that were compiled, it seems that drivers/gpu/drm/xe/tests/xe_migrate.c was >> not rebult: >> >> $ grep CC ../err-6.12.0-tiny-pt-02b.log >> CC init/version.o >> CC init/version-timestamp.o >> CC arch/x86/boot/version.o >> CC arch/x86/boot/compressed/kaslr.o >> CC arch/x86/boot/compressed/misc.o >> HOSTCC debian/linux-headers-6.12.0-dirty/usr/src/linux-headers-6.12.0-dirty/scripts/kallsyms >> HOSTCC debian/linux-headers-6.12.0-dirty/usr/src/linux-headers-6.12.0-dirty/scripts/sorttable >> HOSTCC debian/linux-headers-6.12.0-dirty/usr/src/linux-headers-6.12.0-dirty/scripts/asn1_compiler >> HOSTCC debian/linux-headers-6.12.0-dirty/usr/src/linux-headers-6.12.0-dirty/scripts/genksyms/genksyms.o >> HOSTCC debian/linux-headers-6.12.0-dirty/usr/src/linux-headers-6.12.0-dirty/scripts/selinux/genheaders/genheaders >> HOSTCC debian/linux-headers-6.12.0-dirty/usr/src/linux-headers-6.12.0-dirty/scripts/selinux/mdp/mdp >> YACC debian/linux-headers-6.12.0-dirty/usr/src/linux-headers-6.12.0-dirty/scripts/genksyms/parse.tab.[ch] >> HOSTCC debian/linux-headers-6.12.0-dirty/usr/src/linux-headers-6.12.0-dirty/scripts/sign-file >> HOSTCC debian/linux-headers-6.12.0-dirty/usr/src/linux-headers-6.12.0-dirty/scripts/insert-sys-cert >> HOSTCC debian/linux-headers-6.12.0-dirty/usr/src/linux-headers-6.12.0-dirty/scripts/genksyms/parse.tab.o >> HOSTCC debian/linux-headers-6.12.0-dirty/usr/src/linux-headers-6.12.0-dirty/scripts/genksyms/lex.lex.o >> HOSTCC debian/linux-headers-6.12.0-dirty/usr/src/linux-headers-6.12.0-dirty/scripts/mod/modpost.o >> HOSTCC debian/linux-headers-6.12.0-dirty/usr/src/linux-headers-6.12.0-dirty/scripts/mod/file2alias.o >> HOSTCC debian/linux-headers-6.12.0-dirty/usr/src/linux-headers-6.12.0-dirty/scripts/mod/sumversion.o >> HOSTCC debian/linux-headers-6.12.0-dirty/usr/src/linux-headers-6.12.0-dirty/scripts/mod/symsearch.o >> HOSTCC debian/linux-headers-6.12.0-dirty/usr/src/linux-headers-6.12.0-dirty/scripts/basic/fixdep >> >> But they were only BTF + INSTALL + STRIP + SIGN'ed: >> >> $ grep drivers/gpu/drm/xe/ ../err-6.12.0-tiny-pt-02b.log >> BTF [M] drivers/gpu/drm/xe/xe.ko >> INSTALL debian/linux-image-6.12.0-dirty/lib/modules/6.12.0-dirty/kernel/drivers/gpu/drm/xe/xe.ko >> STRIP debian/linux-image-6.12.0-dirty/lib/modules/6.12.0-dirty/kernel/drivers/gpu/drm/xe/xe.ko >> SIGN debian/linux-image-6.12.0-dirty/lib/modules/6.12.0-dirty/kernel/drivers/gpu/drm/xe/xe.ko >> >> Nothing suggests that "drivers/gpu/drm/xe/tests/xe_migrate.c" is being recompiled, though it was modified, >> or at least "touch"-ed. > Do you have CONFIG_DRM_XE_KUNIT_TEST enabled? > It probably does not make sense to use Debian packaging for a kunit > test. Just use kunit.py directly. Thanks. > >> Is this the expected behaviour? >> >> (I have deliberately skipped "make clean", to see if I can save some building time and grid power.) Best regards, Mirsad -- "Did I succeed in making a coherent view on the world?"
Powered by blists - more mailing lists