View Issue Details
| ID | Project | Category | View Status | Date Submitted | Last Update |
|---|---|---|---|---|---|
| 0001060 | channel: elrepo/el8 | kmod-mpt3sas | public | 2020-12-15 11:03 | 2021-03-03 19:20 |
| Reporter | gbrandl | Assigned To | pperry | ||
| Priority | normal | Severity | major | Reproducibility | have not tried |
| Status | resolved | Resolution | fixed | ||
| Platform | x86_64 | OS | CentOS | OS Version | 8 |
| Summary | 0001060: dracut not called in posttrans since /boot/symvers-* is missing | ||||
| Description | In the `posttrans` snippet of the mpt3sas module (8.2 and 8.3 build), the following snippet is responsible for calling dracut to add the module to initramfs: ``` # The same check as in weak-modules: we assume that the kernel present # if the symvers file exists. if [ -e "/boot/symvers-$k.gz" ]; then /usr/bin/dracut -f "$tmp_initramfs" "$k" || exit 1 ``` However, on my CentOS system (installed as 8.2, then upgraded to 8.3), no `/boot/symvers-*` files exist although the kernels are present. Looking into `/usr/sbin/weak-modules`, parts of which this snippet apparently reimplements, I see (line 552): ``` # find_symvers_file: # Since /boot/ files population process is now controlled by systemd's # kernel-install bash script and its plug-ins, it might be the case # that, while present, symvers file is not populated in /boot. # Let's also check for /lib/modules/$kver/symvers.gz, since that's where # it is populated from. ``` So it seems that the check in the posttrans is not correct anymore for this version of CentOS, and needs to be adapted to also check in `/lib/modules`? | ||||
| Tags | No tags attached. | ||||
|
|
Interesting - thank you. Looks like you are correct - symvers.gz is now located at /lib/modules/<kversion>/symvers.gz |
|
|
Apologies for the delay... We are proposing the following simple patch: # The same check as in weak-modules: we assume that the kernel present # if the symvers file exists. - if [ -e "/boot/symvers-$k.gz" ]; then + if [ -e "/lib/modules/%{kmod_kernel_version}.x86_64/symvers.gz" ]; then /usr/bin/dracut -f "$tmp_initramfs" "$k" || exit 1 I will try to get this (and other affected storage) package(s) fixed and released over the weekend. Thank you again for this bug report - we really appreciate it. |
|
|
Great! BTW, it looks like this snippet is not only in storage packages, I also found it in rtl8187 and e100, for example. |
|
|
Correct, it affects ALL packages, as it's part of the standard template we use. However, not all drivers need to be in the initramfs to enable the system to boot so we will concentrate on rebuilding those packages affected most first, and queue the fix to all other packages as part of their next rebuild. Users can also work around temporarily by manually rebuilding their initramfs file if needed. If you find the issue is affecting you with other packages, please say (or open a bug), and I'll happily fix and rebuild sooner, but I didn't want to invoke a mass rebuild of every package unless absolutely necessary. Again, we really appreciate you highlighting this issue, which crept in following a change in the way the disto kernel packages symvers. |
|
|
I'm happy with mpt3sas being fixed. Thanks for your quick reaction and handling of the bug! |
|
|
I've built updated kmod-mpt3sas packages with the following patch and released to the main repository. Packages are currently syncing to the mirrors and should show up shortly. # The same check as in weak-modules: we assume that the kernel present # if the symvers file exists. - if [ -e "/boot/symvers-$k.gz" ]; then + if [ -e "$k_dir/symvers.gz" ]; then /usr/bin/dracut -f "$tmp_initramfs" "$k" || exit 1 cmp -s "$tmp_initramfs" "$dst_initramfs" I've tested locally, and the initramfs image is now updated for me. Please can you test and let me know if you feel it's fixed the issue for you. Many thanks. |
|
|
Positive: Installed the -2 package on a fresh system and it did regenerate initramfs. |
|
|
Great, thank you for the feedback - much appreciated. Closing as fixed. Please feel free to open further reports if you have issues. |
|
|
Red Hat kmod packages have the same issue. Filed a bug here: https://bugzilla.redhat.com/show_bug.cgi?id=1909441 |
|
|
The RHBZ has been closed as 'fixed'. A solution article is now available: https://access.redhat.com/solutions/5681371 |
|
|
RH fixed the bug as follows (using kmod-redhat-mlx5_core as an example): $ diff kmod-redhat-mlx5_core.spec.old2 kmod-redhat-mlx5_core.spec 6c6 < %define kmod_rpm_release 2 --- > %define kmod_rpm_release 3 475c475 < if [ -e "/boot/symvers-$k.gz" ]; then --- > if [ -e "/boot/symvers-$k.gz" ] || [ -e "$k_dir/symvers.gz" ]; then 924a925,928 > * Wed Jan 13 2021 Eugene Syromiatnikov <esyr@redhat.com> 5.0_0_dup8.2-3 > - Check for symvers.gz presence in /lib/modules/KVER in addition to boot > (#1915479). |
| Date Modified | Username | Field | Change |
|---|---|---|---|
| 2020-12-15 11:03 | gbrandl | New Issue | |
| 2020-12-15 11:03 | gbrandl | Status | new => assigned |
| 2020-12-15 11:03 | gbrandl | Assigned To | => pperry |
| 2020-12-15 11:48 | pperry | Note Added: 0007332 | |
| 2020-12-18 03:56 | pperry | Note Added: 0007356 | |
| 2020-12-18 04:26 | gbrandl | Note Added: 0007357 | |
| 2020-12-18 08:30 | pperry | Note Added: 0007358 | |
| 2020-12-18 08:37 | gbrandl | Note Added: 0007359 | |
| 2020-12-18 15:46 | pperry | Note Added: 0007361 | |
| 2020-12-19 03:09 | gbrandl | Note Added: 0007363 | |
| 2020-12-19 06:44 | pperry | Note Added: 0007364 | |
| 2020-12-19 06:44 | pperry | Status | assigned => resolved |
| 2020-12-19 06:44 | pperry | Resolution | open => fixed |
| 2020-12-19 19:54 | toracat | Note Added: 0007366 | |
| 2021-02-23 12:37 | toracat | Note Added: 0007471 | |
| 2021-03-03 19:20 | toracat | Note Added: 0007487 |
