Re: kernel 3.14.2 oops: seems related to EFI

[Matt Fleming <matt-HNK1S37rvNbeXh+fF434Mdi2O/JbrIOy@public.gmane.org>]

On Mon, 19 May, at 09:09:58AM, Francis Moreau wrote:
> On 05/18/2014 03:42 PM, Borislav Petkov wrote:
> > On Sat, May 17, 2014 at 05:25:47PM +0200, Francis Moreau wrote:
> >> [  +0.018677] general protection fault: 0000 [#1] PREEMPT SMP
> >> [  +0.000068] Modules linked in: usb_storage tun raid1 md_mod loop fuse
> >> joydev coretemp hwmon arc4 intel_rapl x86_pkg_temp_thermal
> >> intel_powerclamp kvm_intel nls_iso8859_1 nls_cp437 iTCO_wdt kvm vfat fat
> >> iTCO_vendor_support iwldvm uvcvideo led_class crct10dif_pclmul
> >> crc32_pclmul crc32c_intel ghash_clmulni_intel mac80211 videobuf2_vmalloc
> >> videobuf2_memops videobuf2_core aesni_intel videodev aes_x86_64
> >> snd_hda_codec_hdmi lrw gf128mul mousedev glue_helper btusb
> >> snd_hda_codec_via ablk_helper media cryptd iwlwifi snd_hda_codec_generic
> >> bluetooth psmouse microcode i2c_i801 serio_raw cfg80211 6lowpan_iphc
> >> rtsx_pci_ms r8169 memstick rfkill lpc_ich mii snd_hda_intel
> >> snd_hda_codec thermal snd_hwdep wmi snd_pcm tpm_infineon snd_timer
> >> tpm_tis mei_me snd tpm mei shpchp evdev soundcore processor battery
> >> mac_hid ac
> >> [  +0.000803]  ext4 crc16 mbcache jbd2 hid_generic usbhid hid bcache
> >> sd_mod sr_mod crc_t10dif cdrom crct10dif_common rtsx_pci_sdmmc mmc_core
> >> atkbd libps2 ahci libahci ehci_pci libata xhci_hcd ehci_hcd scsi_mod
> >> rtsx_pci usbcore usb_common i8042 serio i915 video button intel_gtt
> >> i2c_algo_bit drm_kms_helper drm i2c_core
> >> [  +0.000328] CPU: 0 PID: 30835 Comm: systemd-udevd Not tainted
> >> 3.14.2-1-ARCH #1
> >> [  +0.000064] Hardware name: CLEVO CO.                        W55xEU
> >>                       /W55xEU                          , BIOS 4.6.5
> >> 03/05/2013
> >> [  +0.000102] task: ffff880405ee6bf0 ti: ffff880400f4a000 task.ti:
> >> ffff880400f4a000
> >> [  +0.000060] RIP: 0010:[<ffffffff810655af>]  [<ffffffff810655af>]
> >> efi_call5+0x6f/0xf0
> >> [  +0.000071] RSP: 0018:ffff880400f4bdb0  EFLAGS: 00010002
> >> [  +0.000045] RAX: 0000000080050033 RBX: ffff8804040e3000 RCX:
> >> ffff8804040e3000
> >> [  +0.000055] RDX: ffff8804040e3400 RSI: ffff8804040e3000 RDI:
> >> bff7fffff7afffff
> > 
> > So you get a #GP while executing call *rdi and %rdi is supposed to
> > contain ->get_variable. But instead it contains some very funky shit:
> > 
> > 0xbff7fffff7afffff
> > 
> > Who made it contain that nuisance of a pointer which thinks it is
> > ->get_variable, huh? If only I could get my hands on that guy! :-P
> > 
> > Ok, seriously, how reproducible is this?
> 
> I don't really know how to reproduce this, I only can say that it
> usually happens while partitioning the loop device or perhaps when the
> kernel reads the partition table afterwards.
 
It looks like it's oopsing as a result of systemd-udevd trying to
read a variable via the efivarfs mount,

 Call Trace:
  [<ffffffff81064901>] ? virt_efi_get_variable+0x51/0x80
  [<ffffffff813dbb71>] efivar_entry_size+0x41/0x80
  [<ffffffff81245659>] efivarfs_file_read+0x49/0x100
  [<ffffffff811ba797>] vfs_read+0x97/0x160
  [<ffffffff811bb2e9>] SyS_read+0x59/0xd0
  [<ffffffff81517629>] system_call_fastpath+0x16/0x1b

-- 
Matt Fleming, Intel Open Source Technology Center