Oops on megaraid.

Kernel: 2.6.8.1 vanilla
Card: Dell PERC DC/2 (megaraid)

The oops happened when I attempted to unload the megaraid module. Before doing this, I ran the dellmgr program to reconfigure the raid.

Before the "oops" text, I saw:
Badness in remove_proc_entry at fs/proc/generic.c:688

 [<c017c3fa>] remove_proc_entry+0x10a/0x150
 [<d88f6e3e>] megaraid_exit+0x2e/0x3e [megaraid]
 [<c012c690>] sys_delete_module+0x150/0x1a0
 [<c0142a00>] do_munmap+0x140/0x190
 [<c010513b>] syscall_call+0x7/0xb

That above explains the few lines above "Unable to handle kernel paging..."

This system boots via NFS and / is nfsroot. There is no swap setup on the machine since there are no local disks. The raid there for testing before moving it to another machine. The system has 384mb of physical ram.

Oops follows:

ksymoops 2.4.9 on i686 2.6.8.1. Options used

     -v /lib/modules/2.6.8.1/build/vmlinux (specified)
     -k /proc/ksyms (default)
     -l /proc/modules (default)
     -o /lib/modules/2.6.8.1/ (default)
     -m /boot/System.map-2.6.8.1 (default)

Error (regular_file): read_ksyms stat /proc/ksyms failed No modules in ksyms, skipping objects
No ksyms, skipping lsmod

 [<c017c3fa>] remove_proc_entry+0x10a/0x150
 [<d88f6e3e>] megaraid_exit+0x2e/0x3e [megaraid]
 [<c012c690>] sys_delete_module+0x150/0x1a0
 [<c0142a00>] do_munmap+0x140/0x190
 [<c010513b>] syscall_call+0x7/0xb
Unable to handle kernel paging request at virtual address 4b87ad72
c017b52e
*pde = 00000000
Oops: 0000 [#1]
CPU:    0
EIP:    0060:[<c017b52e>]    Not tainted

Using defaults from ksymoops -t elf32-i386 -a i386
EFLAGS: 00010292 (2.6.8.1)
eax: 4b87ad6e ebx: d7b68cb4 ecx: 00000004 edx: d7b68c80
esi: 00000004 edi: 4b87ad6e ebp: 000040f8 esp: d23d3e8c
ds: 007b es: 007b ss: 0068

Stack: 00000004 4b87ad6e c017c33c 00000004 d23d3ed4 4b87ad6e d23d3ec8 d23d3ed4 
       d77d9000 d7e6c1d4 d75c4074 d88f6d6a d23d3ed4 d7b68c80 00000000 177db000 
       0000007e d7e6c000 30616268 00000000 00000000 d704dbcc c4087800 d88fab44 
Call Trace:
 [<c017c33c>] remove_proc_entry+0x4c/0x150
 [<d88f6d6a>] megaraid_remove_one+0x34a/0x370 [megaraid]
 [<c01aca7b>] pci_device_remove+0x3b/0x40
 [<c01e5a96>] device_release_driver+0x66/0x70
 [<c01e5acb>] driver_detach+0x2b/0x40
 [<c01e5f7c>] bus_remove_driver+0x4c/0x90
 [<c01e64e3>] driver_unregister+0x13/0x30
 [<c01acce6>] pci_unregister_driver+0x16/0x30
 [<d88f6e4a>] megaraid_exit+0x3a/0x3e [megaraid]
 [<c012c690>] sys_delete_module+0x150/0x1a0
 [<c0142a00>] do_munmap+0x140/0x190
 [<c010513b>] syscall_call+0x7/0xb
Code: 0f b7 48 04 3b 4c 24 0c 74 0d 31 c0 8b 34 24 8b 7c 24 04 83 

Trace; c017c33c <remove_proc_entry+4c/150>
Trace; d88f6d6a <pg0+185ebd6a/3fcf3000>
Trace; c01aca7b <pci_device_remove+3b/40>
Trace; c01e5a96 <device_release_driver+66/70>
Trace; c01e5acb <driver_detach+2b/40>
Trace; c01e5f7c <bus_remove_driver+4c/90>
Trace; c01e64e3 <driver_unregister+13/30>
Trace; c01acce6 <pci_unregister_driver+16/30>
Trace; d88f6e4a <pg0+185ebe4a/3fcf3000>
Trace; c012c690 <sys_delete_module+150/1a0>
Trace; c0142a00 <do_munmap+140/190>
Trace; c010513b <syscall_call+7/b>

Code; c017b52e <proc_match+e/40>
00000000 <_EIP>:
Code; c017b52e <proc_match+e/40> <=====
0: 0f b7 48 04 movzwl 0x4(%eax),%ecx <=====
Code; c017b532 <proc_match+12/40>
4: 3b 4c 24 0c cmp 0xc(%esp),%ecx
Code; c017b536 <proc_match+16/40>
8: 74 0d je 17 <_EIP+0x17>
Code; c017b538 <proc_match+18/40>
a: 31 c0 xor %eax,%eax
Code; c017b53a <proc_match+1a/40>
c: 8b 34 24 mov (%esp),%esi
Code; c017b53d <proc_match+1d/40>
f: 8b 7c 24 04 mov 0x4(%esp),%edi
Code; c017b541 <proc_match+21/40>
13: 83 00 00 addl $0x0,(%eax)

1 error issued. Results may not be reliable.

Modules loaded:

Module                  Size  Used by
reiserfs              244368  0 
dm_mirror              20404  2 
aic7xxx               195544  0 
sd_mod                 16320  0 
megaraid               42376  0 
snd_ymfpci             55236  0 
snd_ac97_codec         66948  1 snd_ymfpci
snd_pcm                86152  1 snd_ymfpci
snd_opl3_lib            8896  1 snd_ymfpci
snd_timer              21316  3 snd_ymfpci,snd_pcm,snd_opl3_lib
snd_hwdep               7076  1 snd_opl3_lib
snd_page_alloc          9032  2 snd_ymfpci,snd_pcm
snd_mpu401_uart         6144  1 snd_ymfpci
snd_rawmidi            20164  1 snd_mpu401_uart
snd_seq_device          6440  2 snd_opl3_lib,snd_rawmidi
snd                    47972  9
snd_ymfpci,snd_ac97_codec,snd_pcm,snd_opl3_lib,snd_timer,snd_hwdep,snd_mpu401_uart,snd_rawmidi,snd_seq_device
ymfpci                 47872  0 
ac97_codec             17004  1 ymfpci
soundcore               7328  2 snd,ymfpci
usbhid                 29664  0 
uhci_hcd               29936  0 
usbcore               103908  4 usbhid,uhci_hcd
piix                   11488  0 [permanent]
ide_core              115868  1 piix
evdev                   7296  0 
sr_mod                 13668  0 
scsi_mod               70432  4 aic7xxx,sd_mod,megaraid,sr_mod
cdrom                  37756  1 sr_mod
isofs                  24228  0 
dm_mod                 56124  1 dm_mirror

 Lab tests show that use of micro$oft causes cancer in lab animals

Someone's removing non-empty directory in procfs. Let's see... Indeed.

#ifdef CONFIG_PROC_FS
        remove_proc_entry("megaraid", &proc_root);

#endif

        pci_unregister_driver(&megaraid_pci_driver);

so we remove /proc/megaraid and then procees to remove controllers found by driver. Each of those has a subdirectory in /proc/megaraid...

Fix is trivial:

diff -urN RC10-bk6-base/drivers/scsi/megaraid.c RC10-bk6-current/drivers/scsi/megaraid.c
@@ -5109,11 +5109,11 @@
 	 */
 	unregister_chrdev(major, "megadev");
 
+	pci_unregister_driver(&megaraid_pci_driver);
+
 #ifdef CONFIG_PROC_FS
 	remove_proc_entry("megaraid", &proc_root);
 #endif
-
-	pci_unregister_driver(&megaraid_pci_driver);
 }
 
 module_init(megaraid_init);

Cool, thanks.

 Lab tests show that use of micro$oft causes cancer in lab animals