FS#32671 - [kmod] depmod fails with modules in extramodules
Attached to Project:
Arch Linux
Opened by Jonathan Frazier (wide-eye) - Thursday, 15 November 2012, 07:43 GMT
Last edited by Gerardo Exequiel Pozzi (djgera) - Saturday, 17 November 2012, 04:28 GMT
Opened by Jonathan Frazier (wide-eye) - Thursday, 15 November 2012, 07:43 GMT
Last edited by Gerardo Exequiel Pozzi (djgera) - Saturday, 17 November 2012, 04:28 GMT
|
Details
using kmod 11-1, depmod errors with modules in extramodules.
it works with kmod 10-2.
depmod: tools/depmod.c:980: mod_free: Assertion `mod->kmod == ((void *)0)' failed. 1. cp /usr/lib/modules/3.6.6-1-ARCH/kernel/drivers/input/mouse/psmouse.ko.gz /usr/lib/modules/extramodules-3.6-ARCH/ 2. depmod -vvv gives: (snip 150k lines) depmod: DEBUG: try extramodules/psmouse.ko.gz (psmouse) depmod: DEBUG: comparing priorities of kernel/drivers/input/mouse/psmouse.ko.gz and extramodules/psmouse.ko.gz depmod: DEBUG: search built-in depmod: DEBUG: search extramodules depmod: DEBUG: search updates depmod: DEBUG: priorities: built-in: 0, old: 1, new: 0 depmod: DEBUG: Replace lower priority kernel/drivers/input/mouse/psmouse.ko.gz with new module extramodules/psmouse.ko.gz depmod: DEBUG: del 0x1f22980 kmod=0x1f22840, path=/lib/modules/3.6.6-1-ARCH/kernel/drivers/input/mouse/psmouse.ko.gz depmod: DEBUG: free 0x1f22980 kmod=0x1f22840, path=/lib/modules/3.6.6-1-ARCH/kernel/drivers/input/mouse/psmouse.ko.gz depmod: tools/depmod.c:980: mod_free: Assertion `mod->kmod == ((void *)0)' failed. |
This task depends upon
Closed by Gerardo Exequiel Pozzi (djgera)
Saturday, 17 November 2012, 04:28 GMT
Reason for closing: Fixed
Saturday, 17 November 2012, 04:28 GMT
Reason for closing: Fixed
depmod: DEBUG: Replace lower priority kernel/drivers/platform/x86/hdaps.ko.gz with new module extramodules/hdaps.ko.gz
depmod: DEBUG: del 0xb83c50 kmod=0xb83b10, path=/lib/modules/3.6.6-1-ARCH/kernel/drivers/platform/x86/hdaps.ko.gz
depmod: DEBUG: free 0xb83c50 kmod=0xb83b10, path=/lib/modules/3.6.6-1-ARCH/kernel/drivers/platform/x86/hdaps.ko.gz
depmod: tools/depmod.c:980: mod_free: Assertion `mod->kmod == ((void *)0)' failed.
Not really related, but you really should be using updates/ and not extramodules/ if you're rebuilding your own modules to supersede the in tree flavor.
this is tp_smapi from [community], version 0.41-11. It fails during installation with the same error, the above output is from depmod -vvv. This did not happen before. I should mention that this is a fresh arch installation.
I apologize if I am misunderstanding something here, but installing tp_smapi triggers this bug on all my systems (which did not happen before). It is strange that you cannot reproduce this.
commit e4a7352ad3c16f0b4b658218d9997e3570a70772
Author: Lucas De Marchi <...>
Date: Tue Oct 30 03:46:12 2012 -0200
depmod: unref kmod_module once we don't need it anymore
after reverting this change depmod gets through this section of code but has later memory allocation errors.
example error:
ERROR: failed to load symbols from /lib/modules/3.6.6-1-ARCH/kernel/drivers/ata/pata_opti.ko.gz: Cannot allocate memory
"working" log. identical to kmod 10-2 for this section.
depmod: DEBUG: try extramodules/hdaps.ko.gz (hdaps)
depmod: DEBUG: comparing priorities of kernel/drivers/platform/x86/hdaps.ko.gz and extramodules/hdaps.ko.gz
depmod: DEBUG: search built-in
depmod: DEBUG: search extramodules
depmod: DEBUG: search updates
depmod: DEBUG: priorities: built-in: 0, old: 1, new: 0
depmod: DEBUG: Replace lower priority kernel/drivers/platform/x86/hdaps.ko.gz with new module extramodules/hdaps.ko.gz
depmod: DEBUG: del 0x8d7c5a8 kmod=0x8d7c4f8, path=/lib/modules/3.6.6-1-ARCH/kernel/drivers/platform/x86/hdaps.ko.gz
depmod: DEBUG: free 0x8d7c5a8 kmod=0x8d7c4f8, path=/lib/modules/3.6.6-1-ARCH/kernel/drivers/platform/x86/hdaps.ko.gz
depmod: DEBUG: add 0x8d893e8 kmod=0x8d7c4f8, path=/lib/modules/3.6.6-1-ARCH/extramodules/hdaps.ko.gz
However, with kmod 10-2, depmod picks up hdaps.ko.gz in extramodules/ just fine, even though the kernel tree module with the same name still exists.
But yes, agree that the module is in the wrong directory. That's a separate bug.