Re: Problem with efibootmgr on asus 1215b

From: Matt Fleming
Date: Fri Jan 20 2012 - 06:48:41 EST


(Adding Matthew Garrett to Cc..)

On Sun, 2012-01-01 at 06:20 +0000, Vasco Dias wrote:
> Hi,
>
> the problem is that after booting the system from uefi (with the uefi
> shell) and after loading
> the module efivars i can't create a new boot entry because the system
> just locks completely.

[...]

> I'm reporting here on the list after asking about it on the arch linux
> forums ( https://bbs.archlinux.org/viewtopic.php?pid=1024451 ).

Your report on the above forum shows a page fault in the dmesg. Are both
the page fault and lockup reproducible?

I think that the page fault is the most interesting error,

[ 313.519164] EFI Variables Facility v0.08 2004-May-17
[ 441.370854] BUG: unable to handle kernel paging request at 00000000ffe1fbc8
[ 441.370854] IP: [<ffff8800a7d44c1f>] 0xffff8800a7d44c1e
[ 441.370854] PGD 131775067 PUD 0
[ 441.370854] Oops: 0000 [#1] PREEMPT SMP
[ 441.370854] CPU 1
[ 441.370854] Modules linked in: nls_cp437 vfat fat efivars ipv6 fuse btusb bluetooth uvcvideo videodev media v4l2_compat_ioctl32 joydev snd_hda_codec_realtek bcma arc4 snd_hda_codec_hdmi brcmsmac(C) snd_hda_intel snd_hda_codec snd_hwdep snd_pcm brcmutil(C) ohci_hcd snd_timer mac80211 snd eeepc_wmi i2c_piix4 asus_wmi soundcore ehci_hcd cfg80211 sparse_keymap serio_raw pci_hotplug snd_page_alloc rfkill crc_ccitt xhci_hcd usbcore pcspkr psmouse evdev sp5100_tco k10temp atl1c wmi processor battery video ac button ext4 mbcache jbd2 crc16 sd_mod ahci libahci libata scsi_mod radeon ttm drm_kms_helper drm i2c_algo_bit i2c_core
[ 441.416767]
[ 441.416767] Pid: 1217, comm: efibootmgr Tainted: G C 3.0-ARCH #1 ASUSTeK Computer INC. 1215B/1215B
[ 441.416767] RIP: 0010:[<ffff8800a7d44c1f>] [<ffff8800a7d44c1f>] 0xffff8800a7d44c1e
[ 441.416767] RSP: 0018:ffff880132463ab0 EFLAGS: 00010082
[ 441.416767] RAX: 00000000ffe1fbc8 RBX: 000000000000000a RCX: 00000000ffe1fbc8
[ 441.416767] RDX: 0000000000000000 RSI: ffff880132463bea RDI: ffff880132463b40
[ 441.416767] RBP: 00000000ffe1fbc8 R08: 0000000000000000 R09: 0000000000000038
[ 441.416767] R10: 00000000000000ff R11: ffffc9001101fbca R12: 0000000000000001
[ 441.416767] R13: 0000000000000000 R14: ffffc9001101fbc8 R15: ffff880132463be0
[ 441.416767] FS: 00007fb40f3d2700(0000) GS:ffff88013ed00000(0000) knlGS:0000000000000000
[ 441.416767] CS: 0010 DS: 0000 ES: 0000 CR0: 0000000080050033
[ 441.416767] CR2: 00000000ffe1fbc8 CR3: 00000001328c6000 CR4: 00000000000006e0
[ 441.416767] DR0: 0000000000000000 DR1: 0000000000000000 DR2: 0000000000000000
[ 441.416767] DR3: 0000000000000000 DR6: 00000000ffff0ff0 DR7: 0000000000000400
[ 441.416767] Process efibootmgr (pid: 1217, threadinfo ffff880132462000, task ffff88013210cf10)
[ 441.416767] Stack:
[ 441.416767] ffffc90011010000 ffff880132463b78 0000000000010000 ffff8800a7d4401d
[ 441.416767] 0000000000000070 ffff88013249b400 0000000000000084 ffff8800a7d45738
[ 441.416767] ffffc9001101fbc8 ffff8800a7d4331c 000000000000000a ffffc9001101fbc8
[ 441.416767] Call Trace:
[ 441.416767] [<ffffffff811c71ed>] ? open+0x10d/0x1b0
[ 441.416767] [<ffffffff81155a93>] ? __dentry_open+0x323/0x390
[ 441.416767] [<ffffffff811c70e0>] ? bin_vma_open+0x70/0x70
[ 441.416767] [<ffffffff8104378b>] ? efi_call5+0x4b/0x80
[ 441.416767] [<ffffffff810431cf>] ? virt_efi_set_variable+0x2f/0x40
[ 441.416767] [<ffffffffa060fea3>] ? efivar_create+0x1e3/0x280 [efivars]
[ 441.416767] [<ffffffff811c73a4>] ? write+0x114/0x190
[ 441.416767] [<ffffffff81157a5f>] ? vfs_write+0xaf/0x180
[ 441.416767] [<ffffffff81157d8a>] ? sys_write+0x4a/0x90
[ 441.416767] [<ffffffff813f4c42>] ? system_call_fastpath+0x16/0x1b
[ 441.416767] Code: ec 01 75 f0 41 bc 01 00 00 00 e8 e5 fb ff ff e8 e4 fc ff ff 33 c0 44 0f b7 c0 66 3b c3 73 20 41 0f b7 c0 41 0f b7 d0 03 c5 8b c8 <8a> 00 42 38 04 3a 75 0a 66 45 03 c4 66 44 3b c3 72 e2 33 c0 66
[ 441.416767] RIP [<ffff8800a7d44c1f>] 0xffff8800a7d44c1e
[ 441.416767] RSP <ffff880132463ab0>
[ 441.416767] CR2: 00000000ffe1fbc8
[ 441.416767] ---[ end trace 8bfd3b1218aba64c ]---

As Mike said, it looks like the runtime code is trying to access an MMIO
address, which is weird.

Mike, you also commented that,

"It would appear that the efi-provided runtime services code is
trying to do a mmio read from 0xffe1fbc8, which is in the IO hole
(reading from *some* device, unsure which). Either way, it looks
like the runtime service code is probably 32bit, and isn't
thunking out of paging mode, so the mmio access is totally bogus."

What lead you to think that the runtime service code is 32bit? If that
were the case I wouldn't have expected the machine to make it to
userland. But I'm honestly not quite sure what's going on.

Perhaps we should blacklist this EFI firmware from using efivars
(assuming the rest of the firmware implementation is sound)?

--
Matt Fleming, Intel Open Source Technology Center

--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at http://vger.kernel.org/majordomo-info.html
Please read the FAQ at http://www.tux.org/lkml/