| Summary: | Classic installer summary gives incorrect description for sound cards (due to corrupt entries in modules.description) | ||
|---|---|---|---|
| Product: | Mageia | Reporter: | Martin Whitaker <mageia> |
| Component: | Installer | Assignee: | Thomas Backlund <tmb> |
| Status: | RESOLVED FIXED | QA Contact: | |
| Severity: | major | ||
| Priority: | Normal | CC: | thierry.vignaud, tmb, westel |
| Version: | Cauldron | ||
| Target Milestone: | --- | ||
| Hardware: | x86_64 | ||
| OS: | Linux | ||
| Whiteboard: | |||
| Source RPM: | kernel | CVE: | |
| Status comment: | |||
| Attachments: |
Installer report.bug.xz (from machine with two sound devices)
Screenshot of summary screen |
||
|
Description
Martin Whitaker
2015-05-16 15:19:05 CEST
Samuel Verschelde
2015-05-16 17:45:10 CEST
CC:
(none) =>
thierry.vignaud, tmb Please attach your /root/drakx/report.bug.xz file. Keywords:
(none) =>
NEEDINFO Created attachment 6563 [details]
Installer report.bug.xz (from machine with two sound devices)
Martin Whitaker
2015-05-16 21:23:08 CEST
Keywords:
NEEDINFO =>
(none) There's no reference to Domex or DMX3191D in the report.bug.xz... Keywords:
(none) =>
NEEDINFO I know, that's why I didn't attach it in the first place... It only appears on the summary screen. If you point me at the perl module that generates this screen, I'll try to trace what's happening.
Samuel Verschelde
2015-05-17 00:27:47 CEST
Keywords:
NEEDINFO =>
(none)
Samuel Verschelde
2015-05-17 00:29:05 CEST
Whiteboard:
(none) =>
NEEDINFO
Samuel Verschelde
2015-05-17 00:29:14 CEST
Keywords:
(none) =>
NEEDINFO it's install::steps_interactive. But it should be seen in lspcidrake output... It's caused by a bogus entry in modules.description, which appeared between 3.19.4 and 3.19.6: % grep snd-hda-intel /lib/modules/3.19.4-desktop-2.mga5/modules.description snd-hda-intel Intel HDA driver % grep snd-hda-intel /lib/modules/3.19.6-desktop-2.mga5/modules.description snd-hda-intel Intel HDA driver snd-hda-intel Domex DMX3191D SCSI driver which is still there, albeit different, in the latest kernel: % grep snd-hda-intel /lib/modules/3.19.8-desktop-1.mga5/modules.description snd-hda-intel Intel HDA driver snd-hda-intel EMC CX/AX/FC-family driver The summary screen is showing the second (bogus) entry. Keywords:
NEEDINFO =>
(none)
Thierry Vignaud
2015-05-17 16:55:30 CEST
Priority:
Normal =>
Low This appears to have been fixed in the 3.19.8-desktop-2.mga5 - there are now no entries for snd-hda-intel in modules.description, and the installer falls back to getting the description from elsewhere (probably /usr/share/pci.ids). Status:
NEW =>
RESOLVED Using 5.0.7-desktop586-4.mga7, the sound card description is: urt/rpmbuild/BUILDROOT/kernel-5.0.7-4.mga7.i386/lib/modules/5.0.7-desktop586-4.mga7/./kernel/drivers/media/rc/keymaps/rc-reddo.ko.xz This appears to be intermittent and random corruption of entries in modules.description - looking through that file, I can often find bad entries. snd-hda-intel does seem to get hit quite often. It was wrong on the Mageia 6 64-bit CI ISO too, but not so noticeably! Priority:
Low =>
Normal Created attachment 10924 [details]
Screenshot of summary screen
To show why I think the severity is major.
For the 64-bit kernels installed on my desktop: % grep snd-hda-intel /lib/modules/*/modules.description /lib/modules/5.0.6-desktop-1.mga7/modules.description:snd-hda-intel Intel HDA driver /lib/modules/5.0.7-desktop-2.mga7/modules.description:snd-hda-intel Intel HDA driver /lib/modules/5.0.7-desktop-5.mga7/modules.description:snd-hda-intel Intel HDA driver /lib/modules/5.0.7-desktop-5.mga7/modules.description:snd-hda-intel Sharp QM1D1C0042 tuner /lib/modules/5.0.9-desktop-1.mga7/modules.description:snd-hda-intel Intel HDA driver /lib/modules/5.0.9-desktop-1.mga7/modules.description:snd-hda-intel Opticon USB barcode to serial driver (1D) With the latest desktop586 kernel, there are two entries for snd-hda-intel: [root@localhost 5.1.1-desktop586-1.mga7]# grep snd-hda-intel modules.description snd-hda-intel Intel HDA driver snd-hda-intel Analog Devices AD7190, AD7192, AD7193, AD7195 ADC Unfortunately the installer displays the second one :-( Then that's a bug in kernel process for generating modules.description $ egrep snd-hda-intel lib/modules/5.1.*/modules.description lib/modules/5.1.1-desktop586-1.mga7/modules.description:snd-hda-intel Intel HDA driver lib/modules/5.1.1-desktop586-1.mga7/modules.description:snd-hda-intel Analog Devices AD7190, AD7192, AD7193, AD7195 ADC lib/modules/5.1.2-desktop586-1.mga7/modules.description:snd-hda-intel Intel HDA driver Same for comment #8 : it's a bug if kernel build system generates corrupted files... Which is a Mageia specific: http://svnweb.mageia.org/packages/cauldron/kernel/current/SPECS/kernel.spec?revision=1397727&view=markup#l1263 We would need to fix that. An issue might be "%kxargs" expanding to "xargs -P 24" on BS I guess having 24 modinfo processes writing at once to the same perl one liner is causing mixed lines Using GNU parallel might be a solution... No need for parallell xargs when generating the modules.description. Running it as a single xargs on a kernel tree takes around 8 seconds on spinning rust... So I've switched from kxargs to xargs for next kernel build Status:
REOPENED =>
RESOLVED |