Re: [PATCH v3] efivars: prevent oops on unload when efi is not enabled

[Randy Dunlap <randy.dunlap@oracle.com>]

On 05/06/11 13:27, Mike Waychison wrote:
> From: Randy Dunlap <randy.dunlap@oracle.com>
> 
> efivars_exit() should check for efi_enabled and not undo
> allocations when efi is not enabled.  Otherwise there is an Oops
> during module unload:
> 
> calling  efivars_init+0x0/0x1000 [efivars] @ 2810
> EFI Variables Facility v0.08 2004-May-17
> initcall efivars_init+0x0/0x1000 [efivars] returned 0 after 5120 usecs
> Oops: 0000 [#1] SMP DEBUG_PAGEALLOC
> last sysfs file: /sys/module/firmware_class/initstate
> CPU 1
> Modules linked in: efivars(-) af_packet tun nfsd lockd nfs_acl auth_rpcgss sunrpc ipt_REJECT nf_conntrack_ipv4 nf_defrag_ipv4 iptable_filter ip_tables ip6t_REJECT xt_tcpudp nf_conntrack_ipv6 nf_defrag_ipv6 xt_state nf_conntrack ip6table_filter ip6_tables x_tables ipv6 cpufreq_ondemand acpi_cpufreq freq_table mperf binfmt_misc dm_mirror dm_region_hash dm_log dm_multipath scsi_dh dm_mod snd_hda_codec_analog snd_hda_intel snd_hda_codec snd_hwdep mousedev snd_seq joydev snd_seq_device mac_hid evdev snd_pcm usbkbd usbmouse usbhid snd_timer hid tg3 snd sr_mod pcspkr rtc_cmos soundcore cdrom iTCO_wdt processor sg dcdbas i2c_i801 rtc_core iTCO_vendor_support intel_agp snd_page_alloc thermal_sys rtc_lib intel_gtt 8250_pnp button hwmon unix ide_pci_generic ide_core ata_generic pata_acpi ata_piix sd_mod crc_t10dif ext3 jbd mbcache uhci_hcd ohci_hcd ssb mmc_core pcmcia pcmcia_core firmware_class ehci_hcd usbcore [last unloaded: dell_rbu]
> 
> Pid: 2812, comm: rmmod Not tainted 2.6.39-rc6 #1 Dell Inc.                 OptiPlex 745                 /0TY565
> RIP: 0010:[<ffffffffa06a17f6>]  [<ffffffffa06a17f6>] unregister_efivars+0x28/0x12c [efivars]
> RSP: 0018:ffff88005eedde98  EFLAGS: 00010283
> RAX: ffffffffa06a23fc RBX: ffffffffa06a44c0 RCX: ffff88007c227a50
> RDX: 0000000000000000 RSI: 00000055ac13db78 RDI: ffffffffa06a44c0
> RBP: ffff88005eeddec8 R08: 0000000000000000 R09: ffff88005eeddd78
> R10: ffffffffa06a4220 R11: ffff88005eeddd78 R12: fffffffffffff7d0
> R13: 00007fff5a3aaec0 R14: 0000000000000000 R15: ffffffffa06a4508
> FS:  00007fa8dcc4a6f0(0000) GS:ffff88007c200000(0000) knlGS:0000000000000000
> CS:  0010 DS: 0000 ES: 0000 CR0: 000000008005003b
> CR2: 0000000000000000 CR3: 000000005d148000 CR4: 00000000000006e0
> DR0: 0000000000000000 DR1: 0000000000000000 DR2: 0000000000000000
> DR3: 0000000000000000 DR6: 00000000ffff0ff0 DR7: 0000000000000400
> Process rmmod (pid: 2812, threadinfo ffff88005eedc000, task ffff88006754b000)
> Stack:
>  ffff88005eeddec8 ffffffffa06a4220 0000000000000000 00007fff5a3aaec0
>  0000000000000000 0000000000000001 ffff88005eedded8 ffffffffa06a2418
>  ffff88005eeddf78 ffffffff810d3598 ffffffffa06a4220 0000000000000880
> Call Trace:
>  [<ffffffffa06a2418>] efivars_exit+0x1c/0xc04 [efivars]
>  [<ffffffff810d3598>] sys_delete_module+0x2d6/0x368
>  [<ffffffff812d1db9>] ? lockdep_sys_exit_thunk+0x35/0x67
>  [<ffffffff810fcba1>] ? audit_syscall_entry+0x172/0x1a5
>  [<ffffffff81575082>] system_call_fastpath+0x16/0x1b
> Code: 5c c9 c3 55 48 89 e5 41 57 41 56 41 55 41 54 53 48 83 ec 08 0f 1f 44 00 00 4c 8b 67 48 48 89 fb 4c 8d 7f 48 49 81 ec 30 08 00 00 <4d> 8b ac 24 30 08 00 00 49 81 ed 30 08 00 00 eb 59 48 89 df 48
> RIP  [<ffffffffa06a17f6>] unregister_efivars+0x28/0x12c [efivars]
>  RSP <ffff88005eedde98>
> CR2: 0000000000000000
>  ---[ end trace aa99b99090f70baa ]---
> 
> Matt apparently removed such a check in 2004 (with no reason given):
>  *  17 May 2004 - Matt Domsch <Matt_Domsch@dell.com>
>  *   remove check for efi_enabled in exit
> but there have been several changes since then.
> 
> Signed-off-by: Randy Dunlap <randy.dunlap@oracle.com>
> Signed-off-by: Mike Waychison <mikew@google.com>

Tested-by: Randy Dunlap <randy.dunlap@oracle.com>

Thanks for the fix.

> Cc: Matt Domsch <Matt_Domsch@dell.com>
> Cc: <matthew.e.tolentino@intel.com>
> Cc: Greg Kroah-Hartman <gregkh@suse.de>
> ---
> v3:
> - Fix typo that I forgot to commit.
> v2:
> - Updated to conditionally unregister efivars for real EFI in the module
>   unload itself -- mikew
> ---
>  drivers/firmware/efivars.c |    6 ++++--
>  1 files changed, 4 insertions(+), 2 deletions(-)
> 
> diff --git a/drivers/firmware/efivars.c b/drivers/firmware/efivars.c
> index ff0c373..2934a13 100644
> --- a/drivers/firmware/efivars.c
> +++ b/drivers/firmware/efivars.c
> @@ -820,8 +820,10 @@ efivars_init(void)
>  static void __exit
>  efivars_exit(void)
>  {
> -	unregister_efivars(&__efivars);
> -	kobject_put(efi_kobj);
> +	if (efi_enabled) {
> +		unregister_efivars(&__efivars);
> +		kobject_put(efi_kobj);
> +	}
>  }
>  
>  module_init(efivars_init);


-- 
~Randy
*** Remember to use Documentation/SubmitChecklist when testing your code ***