Bug 2211583 - dracut generates initramfs for non-existing kernels
Summary: dracut generates initramfs for non-existing kernels
Keywords:
Status: NEW
Alias: None
Product: Red Hat Enterprise Linux 8
Classification: Red Hat
Component: dracut
Version: 8.8
Hardware: All
OS: Linux
medium
medium
Target Milestone: rc
: ---
Assignee: dracut-maint-list
QA Contact: qe-baseos-daemons
URL:
Whiteboard:
Depends On:
Blocks:
TreeView+ depends on / blocked
 
Reported: 2023-06-01 06:31 UTC by Renaud Métrich
Modified: 2023-08-07 15:50 UTC (History)
1 user (show)

Fixed In Version:
Doc Type: If docs needed, set a value
Doc Text:
Clone Of:
Environment:
Last Closed:
Type: Bug
Target Upstream Version:
Embargoed:


Attachments (Terms of Use)


Links
System ID Private Priority Status Summary Last Updated
Red Hat Issue Tracker RHELPLAN-158766 0 None None None 2023-06-01 06:38:13 UTC

Description Renaud Métrich 2023-06-01 06:31:25 UTC
Description of problem:

We see a couple of customers getting their /boot filled with initramfs files for non-existing kernels (as shown by "rpm -q kernel-core").
It appears that "dracut --regenerate-all" bases itself on the existence of /lib/modules/<KVERS> directory, which leads to creating non-usable initramfs files for removed kernels that still have the directory existing, for example because of presence of 3rd party kernel modules.

Filing as Medium/Medium because this leads to space issues in /boot and there is no simple recovery but deleting.

Version-Release number of selected component (if applicable):

dracut-049

How reproducible:

Always

Steps to Reproduce:
1. Create directory for some non-installed kernel

  # mkdir /lib/modules/4.18.0-372.9.1.el8.x86_64

2. Execute dracut

  # dracut /tmp/foo.img 4.18.0-372.9.1.el8.x86_64

Actual results:

/tmp/foo.img created

Expected results:

Some error

Additional info:

I think dracut should check that there is really some kernel available, e.g. through checking presence of /boot/vmlinuz-<KVERS>, or failing if some critical kernel files are missing.

Comment 1 Renaud Métrich 2023-08-07 15:50:20 UTC
It's unclear when the kernel removal leaves files behind, it looks like it's due to some 3rd party script executing.
Anyway, for the hardening of dracut, please find below an example of listing making dracut generate a dumb initramfs "polluting" /boot:
-------- 8< ---------------- 8< ---------------- 8< ---------------- 8< --------
# ll /lib/modules/4.18.0-193.el8.x86_64
total 24
drwxrwxrwx. 3 1002 1012 21 May 21  2022 kernel
-rw-rw-rw-. 1 1002 1012 45 May 19 08:54 modules.alias
-rw-rw-rw-. 1 1002 1012 12 May 19 08:54 modules.alias.bin
-rw-rw-rw-. 1 1002 1012  0 May 19 08:54 modules.builtin.bin
-rw-rw-rw-. 1 1002 1012  0 May 19 08:54 modules.dep
-rw-rw-rw-. 1 1002 1012 12 May 19 08:54 modules.dep.bin
-rw-rw-rw-. 1 1002 1012  0 May 19 08:54 modules.devname
-rw-rw-rw-. 1 1002 1012 55 May 19 08:54 modules.softdep
-rw-rw-rw-. 1 1002 1012 49 May 19 08:54 modules.symbols
-rw-rw-rw-. 1 1002 1012 12 May 19 08:54 modules.symbols.bin
-------- 8< ---------------- 8< ---------------- 8< ---------------- 8< --------

Here above all non-binary files are "empty" or just contain comments:
-------- 8< ---------------- 8< ---------------- 8< ---------------- 8< --------
# cat modules.alias modules.softdep modules.symbols
# Aliases extracted from modules themselves.
# Soft dependencies extracted from modules themselves.
# Aliases for symbols, used by symbol_request().
-------- 8< ---------------- 8< ---------------- 8< ---------------- 8< --------

This leads to generating a large but useless initramfs:
-------- 8< ---------------- 8< ---------------- 8< ---------------- 8< --------
# dracut /tmp/test.img 4.18.0-193.el8.x86_64
depmod: WARNING: could not open /var/tmp/dracut.YTEvX1/initramfs/lib/modules/4.18.0-193.el8.x86_64/modules.order: No such file or directory
depmod: WARNING: could not open /var/tmp/dracut.YTEvX1/initramfs/lib/modules/4.18.0-193.el8.x86_64/modules.builtin: No such file or directory
Warning no default label for /tmp/test.img

# ll /tmp/test.img
-rw-------. 1 root root 25489249 Aug  7 17:49 /tmp/test.img
-------- 8< ---------------- 8< ---------------- 8< ---------------- 8< --------


Note You need to log in before you can comment on or make changes to this bug.