Kraków, Poland
Comment by ~nabijaczleweli on ~sircmpwn/git.sr.ht
Comment by ~nabijaczleweli on ~nabijaczleweli/tzpfms
I evaluated depending on 91tpm2-tss here but it's impossible since it hard-depends on tpm2(1) which we don't care about and don't depend on, so this would be breaking for no reason. Enabling it doesn't interfere with anything though.
Released in
v0.4.1
.
Comment by ~nabijaczleweli on ~nabijaczleweli/tzpfms
наб referenced this ticket in commit 82b5118.
RESOLVED FIXED
RESOLVED FIXEDComment by ~nabijaczleweli on ~nabijaczleweli/tzpfms
наб referenced this ticket in commit 588cce3.
Comment by ~nabijaczleweli on ~nabijaczleweli/tzpfms
наб referenced this ticket in commit 82b5118.
REPORTED
RESOLVED FIXEDComment by ~nabijaczleweli on ~nabijaczleweli/tzpfms
~projectgus, can you try to apply the above to your system directly by, in
/usr/lib/dracut/modules.d/91tzpfms/module-setup.sh
, replacing the line that starts with "inst_library $(find /usr/lib
" withinst_libdir_file 'libtss2-tcti*.so*'and disabling 91tpm2-tss?
Comment by ~nabijaczleweli on ~nabijaczleweli/tzpfms
наб referenced this ticket in commit 588cce3.
Comment by ~nabijaczleweli on ~nabijaczleweli/tzpfms
Thanks for reporting this, I don't really (have a way to) test this continuously on sid.
The two users are:
$ git grep inst_library initrd/dracut/module-setup.sh: inst_library $(find /usr/lib -name 'libtss2-tcti*.so*') # TODO: there's got to be a better way™! initrd/dracut/module-setup.sh: INSTALL_TPM1X{inst_binary tcsd; inst_binary ip; inst_binary ss, initdir, inst_simple, inst_simple, inst_simple, inst_library}and the latter is
databases="$(awk '/^(group|hosts)/ {for(i = 2; i <= NF; ++i) if($i !~ /[^a-z0-9_-]/) db[$i]=0} END {for(d in db) print d}' /etc/nsswitch.conf)" for db in $databases; do for f in /lib/*/"libnss_$db"*; do INST_LIB "$f" done donefzifdso is not affected.
I think a solid replacement would be inst_libdir_file since upstream basically uses it like that
modules.d/95cifs/module-setup.sh: inst_libdir_file 'libcap-ng.so*' modules.d/95cifs/module-setup.sh: inst_libdir_file -n "$_nsslibs" 'libnss_*.so*' modules.d/95fcoe/module-setup.sh: inst_libdir_file 'libhbalinux.so*' modules.d/95iscsi/module-setup.sh: inst_libdir_file 'libgcc_s.so*'and we save on looping ourselves. Funnily enough, 95cifs is doing basically the same thing with nsswitch.conf as the loops above.
Bookworm dracut ships a
inst_libdir_file()
identical to upstream trunk, so this is convenient.You do absolutely need the right
libtss2-tcti-$plugin.so
for your device, so 91tpm2-tss pulling it in is definitely saving you here.Depending on dracut's TPM2 module is definitely in the TODO list, but last time I looked it was barely nascent, then the fork, ... I'll have to see if we can have it now.
Ticket created by ~nabijaczleweli on ~sircmpwn/git.sr.ht
I created https://git.sr.ht/~nabijaczleweli/pre-rename with visibility private.
This means that if, say, ~nabijaczleweli-evil-twin guesses
git@git.sr.ht:~nabijaczleweli/pre-rename
it gets "Access denied." (as opposed to "Repository not found." but whatever). This makes the existence of any repository queriable at unlisted-equivalent visibility.Then I renamed it to https://git.sr.ht/~nabijaczleweli/post-rename.
git clone git@git.sr.ht:~nabijaczleweli/pre-rename
from both ~nabijaczleweli and ~nabijaczleweli-evil-twin getsNOTICE: This repository has moved. Please update your remote to: https://git.sr.ht/~nabijaczleweli/post-rename fatal: Could not read from remote repository. Please make sure you have the correct access rights and the repository exists.
which I consider to be an unacceptable information leak; the ACL of the new repository should affect the ACL of the redirect (as is clearly the intent in
git.sr.ht-shell
I think from a quick skim).
Comment by ~nabijaczleweli on ~nabijaczleweli/tzpfms
наб referenced this ticket in commit 44f9b25.
RESOLVED FIXED
RESOLVED FIXED