I use root on ZFS and use dracut to import my root dataset. Starting from 5.2.20 I can't reboot my laptop normally. when the system finishes bringing itself down and the only thing that is left is reboot, I'm dropped to a shell with a message that reboot command is not available. I tried to research the issue myself and here is what lsinird gives me. PF16W6Y2 /usr/src/linux # lsinitrd /boot/initramfs-5.3.11-gentoo.img |grep reboot lrwxrwxrwx 1 root root 13 Nov 13 18:53 lib/systemd/system/ctrl-alt-del.target -> reboot.target -rw-r--r-- 1 root root 583 Sep 3 12:27 lib/systemd/system/reboot.target -rw-r--r-- 1 root root 551 Sep 3 12:27 lib/systemd/system/systemd-reboot.service lrwxrwxrwx 1 root root 16 Nov 13 18:53 sbin/reboot -> ../bin/systemctl
I tried to reboot to older kernels/initramfs (from Aug 2019) that were working fine, and I noticed that they also have this problem.
Created attachment 596004 [details] dracut screen
Created attachment 596006 [details] rdsosreport
Yes, I can reproduce this with app-arch/cpio-2.13. CCing @base-system * cpio-2.12-r1: $ sudo rm -rf -- /run/initramfs/* && sudo touch /run/initramfs/.need_shutdown $ sudo LANG=C /usr/lib/dracut/dracut-initramfs-restore gzip: stdin: not in gzip format cpio: premature end of archive $ LANG=C ls -l /run/initramfs/sbin/ total 2912 -rwxr-xr-x 1 root root 116632 Nov 14 19:02 blkid -rwxr-xr-x 1 root root 770 Nov 14 19:02 crypt-run-generator lrwxrwxrwx 1 root root 11 Nov 14 19:02 depmod -> ../bin/kmod -r-xr-xr-x 1 root root 42872 Nov 14 19:02 dmeventd -r-xr-xr-x 1 root root 154000 Nov 14 19:02 dmsetup -rwxr-xr-x 1 root root 51112 Nov 14 19:02 fsck -rwxr-xr-x 1 root root 1968 Nov 14 19:02 fsck.xfs lrwxrwxrwx 1 root root 16 Nov 14 19:02 halt -> ../bin/systemctl lrwxrwxrwx 1 root root 22 Nov 14 19:02 init -> ../lib/systemd/systemd -rwxr-xr-x 1 root root 1163 Nov 14 19:02 initqueue lrwxrwxrwx 1 root root 11 Nov 14 19:02 insmod -> ../bin/kmod -rwxr-xr-x 1 root root 193 Nov 14 19:02 insmodpost.sh -rwxr-xr-x 1 root root 492 Nov 14 19:02 loginit -rwxr-xr-x 1 root root 108520 Nov 14 19:02 losetup -r-xr-xr-x 1 root root 2135432 Nov 14 19:02 lvm -rwxr-xr-x 1 root root 3527 Nov 14 19:02 lvm_scan lrwxrwxrwx 1 root root 11 Nov 14 19:02 modprobe -> ../bin/kmod -rwxr-xr-x 1 root root 14120 Nov 14 19:02 nologin lrwxrwxrwx 1 root root 16 Nov 14 19:02 poweroff -> ../bin/systemctl -rwxr-xr-x 1 root root 1346 Nov 14 19:02 rdsosreport lrwxrwxrwx 1 root root 16 Nov 14 19:02 reboot -> ../bin/systemctl lrwxrwxrwx 1 root root 11 Nov 14 19:02 rmmod -> ../bin/kmod -rwxr-xr-x 1 root root 22384 Nov 14 19:02 swapoff -rwxr-xr-x 1 root root 6052 Nov 14 19:02 tracekomem -rwxr-xr-x 1 root root 281336 Nov 14 19:02 xfs_repair * cpio-2.13: $ sudo rm -rf -- /run/initramfs/* && sudo touch /run/initramfs/.need_shutdown $ sudo LANG=C /usr/lib/dracut/dracut-initramfs-restore gzip: stdin: not in gzip format cpio: premature end of archive cpio: Removing leading `../' from hard link targets cpio: Removing leading `/' from hard link targets cpio: Removing leading `../../' from hard link targets $ LANG=C ls -l /run/initramfs/sbin/ total 2912 -rwxr-xr-x 1 root root 116632 Nov 14 18:58 blkid -rwxr-xr-x 1 root root 770 Nov 14 18:58 crypt-run-generator lrwxrwxrwx 1 root root 8 Nov 14 18:58 depmod -> bin/kmod -r-xr-xr-x 1 root root 42872 Nov 14 18:58 dmeventd -r-xr-xr-x 1 root root 154000 Nov 14 18:58 dmsetup -rwxr-xr-x 1 root root 51112 Nov 14 18:58 fsck -rwxr-xr-x 1 root root 1968 Nov 14 18:58 fsck.xfs lrwxrwxrwx 1 root root 13 Nov 14 18:58 halt -> bin/systemctl lrwxrwxrwx 1 root root 19 Nov 14 18:58 init -> lib/systemd/systemd -rwxr-xr-x 1 root root 1163 Nov 14 18:58 initqueue lrwxrwxrwx 1 root root 8 Nov 14 18:58 insmod -> bin/kmod -rwxr-xr-x 1 root root 193 Nov 14 18:58 insmodpost.sh -rwxr-xr-x 1 root root 492 Nov 14 18:58 loginit -rwxr-xr-x 1 root root 108520 Nov 14 18:58 losetup -r-xr-xr-x 1 root root 2135432 Nov 14 18:58 lvm -rwxr-xr-x 1 root root 3527 Nov 14 18:58 lvm_scan lrwxrwxrwx 1 root root 8 Nov 14 18:58 modprobe -> bin/kmod -rwxr-xr-x 1 root root 14120 Nov 14 18:58 nologin lrwxrwxrwx 1 root root 13 Nov 14 18:58 poweroff -> bin/systemctl -rwxr-xr-x 1 root root 1346 Nov 14 18:58 rdsosreport lrwxrwxrwx 1 root root 13 Nov 14 18:58 reboot -> bin/systemctl lrwxrwxrwx 1 root root 8 Nov 14 18:58 rmmod -> bin/kmod -rwxr-xr-x 1 root root 22384 Nov 14 18:58 swapoff -rwxr-xr-x 1 root root 6052 Nov 14 18:58 tracekomem -rwxr-xr-x 1 root root 281336 Nov 14 18:58 xfs_repair
The bug has been referenced in the following commit(s): https://gitweb.gentoo.org/repo/gentoo.git/commit/?id=b970f65cf4d322f6b265517c2d6436733de37abd commit b970f65cf4d322f6b265517c2d6436733de37abd Author: Lars Wendler <polynomial-c@gentoo.org> AuthorDate: 2019-11-14 16:15:55 +0000 Commit: Lars Wendler <polynomial-c@gentoo.org> CommitDate: 2019-11-14 16:15:55 +0000 Revert "app-arch/cpio: arm64 stable (bug #699456)" This reverts commit 2b5ae3f07795864449f2d07859fdfd0433d3a021. Bug: https://bugs.gentoo.org/700020 Signed-off-by: Lars Wendler <polynomial-c@gentoo.org> app-arch/cpio/cpio-2.13.ebuild | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) https://gitweb.gentoo.org/repo/gentoo.git/commit/?id=129584525d0aeb98551c4d53320932d684ebf2ed commit 129584525d0aeb98551c4d53320932d684ebf2ed Author: Lars Wendler <polynomial-c@gentoo.org> AuthorDate: 2019-11-14 16:15:24 +0000 Commit: Lars Wendler <polynomial-c@gentoo.org> CommitDate: 2019-11-14 16:15:24 +0000 Revert "app-arch/cpio: x86 stable wrt bug #699456" This reverts commit 0321d4558c0fd573f8378e27b46d2a8cfd59d4cc. Bug: https://bugs.gentoo.org/700020 Signed-off-by: Lars Wendler <polynomial-c@gentoo.org> app-arch/cpio/cpio-2.13.ebuild | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-)
Exact commands executed by dracut-initramfs-restore.sh: https://github.com/dracutdevs/dracut/blob/ff2043fb4244b0ba740724921e9667aad19f262a/dracut-initramfs-restore.sh#L28 Removing --no-absolute-filenames arg fixes this issue for me. But I don't think that stripping '../' from symlink target is expected behaviour even with --no-absolute-filenames.
The bug has been referenced in the following commit(s): https://gitweb.gentoo.org/repo/gentoo.git/commit/?id=ff0c82b832d026085ef06fb3186f52d12d6ff7dc commit ff0c82b832d026085ef06fb3186f52d12d6ff7dc Author: Lars Wendler <polynomial-c@gentoo.org> AuthorDate: 2019-11-14 16:27:58 +0000 Commit: Lars Wendler <polynomial-c@gentoo.org> CommitDate: 2019-11-14 16:28:40 +0000 package.mask: Masked =app-arch/cpio-2.13 Bug: https://bugs.gentoo.org/700020 Signed-off-by: Lars Wendler <polynomial-c@gentoo.org> profiles/package.mask | 6 ++++++ 1 file changed, 6 insertions(+)
(In reply to Alexander Tsoy from comment #6) > Removing --no-absolute-filenames arg fixes this issue for me. But I don't > think that stripping '../' from symlink target is expected behaviour even > with --no-absolute-filenames. Looks like this is exactly what fix for CVE-2015-1197 is doing.
Probably off topic but I had to patch kernel usr/Makefile to generate a valid initramfs. I took some time to figure it out. --- usr/Makefile.orig 2017-02-19 23:34:00.000000000 +0100 +++ usr/Makefile 2017-02-22 23:44:24.554921038 +0100 @@ -43,7 +43,7 @@ targets := $(datafile_y) # do not try to update files included in initramfs -$(deps_initramfs): ; +$(deps_initramfs): ; $(deps_initramfs): klibcdirs # We rebuild initramfs_data.cpio if: @@ -52,5 +52,6 @@ # 3) If gen_init_cpio are newer than initramfs_data.cpio # 4) arguments to gen_initramfs.sh changes $(obj)/$(datafile_y): $(obj)/gen_init_cpio $(deps_initramfs) klibcdirs - $(Q)$(initramfs) -l $(ramfs-input) > $(obj)/$(datafile_d_y) + $(Q)$(initramfs) -l $(ramfs-input) | \ + sed '2,$$s/:/\\:/g' > $(obj)/$(datafile_d_y) $(call if_changed,initfs)
(In reply to Jimmy.Jazz from comment #9) Yes, that is completely off-topic.
Sent mail to cpio upstream. https://lists.gnu.org/archive/html/bug-cpio/2019-11/msg00013.html
Upstream confirms this is a bug, and is working on a fix. In the mean time, they suggested reverting this commit: https://git.savannah.gnu.org/cgit/cpio.git/commit/?id=45b0ee2b407913c533f7ded8d6f8cbeec16ff6ca
The bug has been closed via the following commit(s): https://gitweb.gentoo.org/repo/gentoo.git/commit/?id=30d0bdb974112f7857d6e50efb7d6b4b2b1ec295 commit 30d0bdb974112f7857d6e50efb7d6b4b2b1ec295 Author: Sam James <sam@gentoo.org> AuthorDate: 2022-10-18 18:40:04 +0000 Commit: Sam James <sam@gentoo.org> CommitDate: 2022-10-18 18:41:04 +0000 app-arch/cpio: patch regressions in 2.13, allowing CVE-2021-38185 fix (unkeyworded) To be keyworded after testing on more machines. Bug: https://bugs.gentoo.org/699456 Bug: https://bugs.gentoo.org/807088 Bug: https://bugs.gentoo.org/854192 Closes: https://bugs.gentoo.org/700020 Signed-off-by: Sam James <sam@gentoo.org> app-arch/cpio/Manifest | 1 + app-arch/cpio/cpio-2.13-r1.ebuild | 39 ++++++++++++++++++++++ .../files/cpio-2.13-sysmacros-glibc-2.26.patch | 12 +++++++ 3 files changed, 52 insertions(+)
Oh my. This bit isn't fixed.
The bug has been closed via the following commit(s): https://gitweb.gentoo.org/repo/gentoo.git/commit/?id=a52ec56f85b11ee1faceddac7874666ad6d2b164 commit a52ec56f85b11ee1faceddac7874666ad6d2b164 Author: Sam James <sam@gentoo.org> AuthorDate: 2022-10-18 19:11:52 +0000 Commit: Sam James <sam@gentoo.org> CommitDate: 2022-10-18 19:12:00 +0000 app-arch/cpio: revert CVE-2015-1197 fix for --no-absolute-filenames At least we can have the fix for CVE-2021-38185. Bug: https://bugs.gentoo.org/699456 Bug: https://bugs.gentoo.org/807088 Closes: https://bugs.gentoo.org/700020 Signed-off-by: Sam James <sam@gentoo.org> .../{cpio-2.13-r1.ebuild => cpio-2.13-r2.ebuild} | 1 + ...e-filenames-revert-CVE-2015-1197-handling.patch | 47 ++++++++++++++++++++++ 2 files changed, 48 insertions(+)
Upstream patches are now available: openwall[dot]com/lists/oss-security/2023/12/21/8 (my account is too new to post the link)
(In reply to eslerm from comment #16) > Upstream patches are now available: > openwall[dot]com/lists/oss-security/2023/12/21/8 > > (my account is too new to post the link) I think this is all handled already - see bug 699456, i.e. 2.14 should be OK?
