Am Montag 08 November 2004 13:34 schrieb Karsten Wiese:Hi
This bug is triggered by logging on to bash (runlevel 3),
typing "cat /proc/acpi", then <TAB> gives the correct "/" to complete,
the 2nd <TAB> has no visual effect, the 3rd <TAB> generates this oops:
Unable to handle kernel paging request at virtual address f89e7b00
*pde = 37ff1067
*pte = 00000000
Oops: 0000 [#1]
Modules linked in: binfmt_misc video ohci1394 ieee1394 uhci_hcd intel_agp agpgart i2c_i801 i2c_core snd_emu10k1 snd_rawmidi snd_seq_device snd_ac97_codec snd_pcm snd_timer snd_page_alloc snd_util_mem snd_hwdep snd soundcore ext3 jbd ata_piix libata sd_mod scsi_mod
EIP: 0060:[<c0187452>] Not tainted VLI
EFLAGS: 00010286 (2.6.10-rc1-mm3)
EIP is at proc_get_inode+0xa0/0x184
Found out, what happened:
By accident I had ibm_acpi.ko built. Its name "ibm" was still present under "/proc/acpi".
This is not an ibm(-laptop)-machine, so ibm_acpi.ko is useless here.
"Unable to handle kernel paging request at virtual address f89e7b00":
this address corresponds to the "struct module *" of ibm_acpi.ko, which was not loaded anymore.
So the real bug here is that there is a non NULL "struct module *", where the corresponding module is unloaded.
Or so I guess.....