Skip to contentSkip to conversation

2.6.27-rc8-git8: oops in process hald-addon-dell-backlight (intel iommu related?)

DISCUSSION Frederik Himpe 2 participants 5 messages Oct 8, 2008 – Oct 25, 2008

lore
Share & tools
lore
#0 lore raw

On my Dell Latitude E6400 system with Intel chipset and graphics, I get this oops in my 2.6.27-rc8-git8 based kernel, which probably happens when hald-addon-dell-backlight is started.

Complete dmesg and lspci info:
https://qa.mandriva.com/show_bug.cgi?id=44711

iwlagn 0000:0c:00.0: PCI INT A -> GSI 17 (level, low) -> IRQ 17 iwlagn 0000:0c:00.0: restoring config space at offset 0x1 (was 0x100102, writing 0x100106)
firmware: requesting iwlwifi-5000-1.ucode
iwlagn: Radio disabled by HW RF Kill switch ADDRCONF(NETDEV_UP): wlan0: link is not ready
wlan0: Failed to config new SSID to the low-level driver

BUG: unable to handle kernel NULL pointer dereference at 0000000000000038
IP: [<ffffffff8037cf1f>] pci_find_upstream_pcie_bridge+0x1f/0xa0
PGD 11addf067 PUD 11b1c1067 PMD 0 
Oops: 0000 [1] SMP 
CPU 1 
Modules linked in: ipv6 binfmt_misc fuse nls_utf8 nls_cp437 vfat fat ext3 

jbd loop cpufreq_ondemand cpufreq_conservative cpufreq_powersave acpi_cpufreq freq_table nvram i8k ohci1394 ieee1394 pcmcia snd_hda_intel snd_hwdep firewire_ohci firewire_core arc4 sdhci_pci crc_itu_t sdhci ecb video output crypto_blkcipher snd_seq_dummy mmc_core snd_seq_oss snd_seq_midi_event snd_seq yenta_socket rsrc_nonstatic pcmcia_core snd_seq_device snd_pcm_oss snd_pcm snd_timer iwlagn snd_page_alloc snd_mixer_oss iwlcore i2c_i801 thermal wmi button battery ac snd e1000e rfkill led_class i2c_core processor soundcore evdev sr_mod rtc_cmos sg dcdbas mac80211 cfg80211 dm_snapshot dm_zero dm_mirror dm_log dm_mod ata_piix ahci libata dock sd_mod scsi_mod crc_t10dif xfs uhci_hcd ohci_hcd ehci_hcd usbcore [last unloaded: scsi_wait_scan]
Pid: 2685, comm: hald-addon-dell Not tainted 2.6.27-tmb- desktop-0.rc8.8.2mdv #1

RIP: 0010:[<ffffffff8037cf1f>]  [<ffffffff8037cf1f>] 
pci_find_upstream_pcie_bridge+0x1f/0xa0
RSP: 0018:ffff88011b453c98  EFLAGS: 00010246
RAX: 0000000000000000 RBX: 0000000000000000 RCX: 0000000000000000
RDX: 0000000000000000 RSI: 0000000000000030 RDI: ffff88011b7e4f80
RBP: ffff88011b453c98 R08: 0000000000000000 R09: ffff88011cc5f000
R10: 0000000000000000 R11: 0000000000000246 R12: 0000000000001000
R13: 0000000000000030 R14: ffff88011b7e4f80 R15: ffff88011b7e5010
FS:  00007f574f1ca700(0000) GS:ffff88011fc02980(0000) 

knlGS:0000000000000000
CS: 0010 DS: 0000 ES: 0000 CR0: 000000008005003b
CR2: 0000000000000038 CR3: 000000011b1b4000 CR4: 00000000000006e0
DR0: 0000000000000000 DR1: 0000000000000000 DR2: 0000000000000000
DR3: 0000000000000000 DR6: 00000000ffff0ff0 DR7: 0000000000000400 Process hald-addon-dell (pid: 2685, threadinfo ffff88011b452000, task ffff88011adfd700)

Stack:  ffff88011b453d28 ffffffff80386808 ffff88011adfd700 0000000000000000
 0000000000000000 ffff88011b453e70 ffff88011b453d68 ffffffff8029574d
 0000000000000000 80ff880000000000 0000000000000000 ffff880000002a00
Call Trace:
 [<ffffffff80386808>] get_domain_for_dev+0x58/0x620
 [<ffffffff8029574d>] ? __alloc_pages_internal+0xed/0x4f0
 [<ffffffff803870f8>] get_valid_domain_for_dev+0x18/0x150
 [<ffffffff803875e7>] intel_map_single+0x47/0x160
 [<ffffffff802b8726>] ? alloc_pages_current+0x76/0xf0
 [<ffffffff8038777f>] intel_alloc_coherent+0x7f/0xb0
 [<ffffffff80212dfa>] dma_alloc_coherent+0x28a/0x310
 [<ffffffffa01a258a>] smi_data_buf_realloc+0x5a/0xf0 [dcdbas]
 [<ffffffff804da2b1>] ? mutex_lock+0x11/0x30
 [<ffffffffa01a26ca>] smi_data_buf_size_store+0x3a/0x60 [dcdbas]
 [<ffffffff803eefee>] dev_attr_store+0x1e/0x20
 [<ffffffff80325d85>] sysfs_write_file+0xc5/0x140
 [<ffffffff802c7a0b>] vfs_write+0xcb/0x170
 [<ffffffff802c7ba0>] sys_write+0x50/0x90
 [<ffffffff8020c69a>] system_call_fastpath+0x16/0x1b


Code: 66 66 66 2e 0f 1f 84 00 00 00 00 00 55 f6 87 89 05 00 00 04 48 89 e5 

75 3b 31 d2 eb 0a 0f 1f 80 00 00 00 00 48 89 fa 48 8b 47 10 <48> 8b 78 38 48 85 ff 74 30 f6 87 89 05 00 00 04 74 e7 80 7f 4a
RIP [<ffffffff8037cf1f>] pci_find_upstream_pcie_bridge+0x1f/0xa0
RSP <ffff88011b453c98>
CR2: 0000000000000038

---[ end trace 200eed5d7163b4d1 ]---

NET: Registered protocol family 17
0000:00:19.0: eth0: Link is Up 100 Mbps Full Duplex, Flow Control: RX/TX 0000:00:19.0: eth0: 10/100 speed: disabling TSO ADDRCONF(NETDEV_CHANGE): eth0: link becomes ready

#1 lore raw
Quoted 6 lines · Frederik Himpe

On my Dell Latitude E6400 system with Intel chipset and graphics, I get this oops in my 2.6.27-rc8-git8 based kernel, which probably happens when hald-addon-dell-backlight is started.

Complete dmesg and lspci info:
https://qa.mandriva.com/show_bug.cgi?id=44711

Booting with intel_iommu=off seems to fix this issue, and seems to make my e1000e based network card stop from giving "transmit timed out" errors after which it stopped working.

Signature · 1 line
Frederik Himpe
#2 lore raw
Quoted 6 lines · Frederik Himpe
Quoted 4 lines · Frederik Himpe

On my Dell Latitude E6400 system with Intel chipset and graphics, I get this oops in my 2.6.27-rc8-git8 based kernel, which probably happens when hald-addon-dell-backlight is started.

Quoted 6 lines
Quoted 2 lines

Complete dmesg and lspci info:
https://qa.mandriva.com/show_bug.cgi?id=44711

Booting with intel_iommu=off seems to fix this issue, and seems to make my e1000e based network card stop from giving "transmit timed out" errors after which it stopped working.

Is it necessary to boot 2.6.27 final on this box with intel_iommu=off, and if so, is it still necessary in mainline?

Rafael

#3 lore raw
Quoted 10 lines · Rafael J. Wysocki
Quoted 9 lines · Frederik Himpe
Quoted 6 lines · Frederik Himpe

On my Dell Latitude E6400 system with Intel chipset and graphics, I get this oops in my 2.6.27-rc8-git8 based kernel, which probably happens when hald-addon-dell-backlight is started.

Complete dmesg and lspci info:
https://qa.mandriva.com/show_bug.cgi?id=44711

Quoted 6 lines
Quoted 3 lines

Booting with intel_iommu=off seems to fix this issue, and seems to make my e1000e based network card stop from giving "transmit timed out" errors after which it stopped working.

Is it necessary to boot 2.6.27 final on this box with intel_iommu=off, and if so, is it still necessary in mainline?

I just tested stock 2.6.27.3 and this machine (Dell Latitude E6400) is totally unusable without intel_iommu=off if Intel VT is enabled in the BIOS. During the boot process, stack traces start scrolling over the screen. In the end, it goes on starting X, and from that moment, the machine locks up completely. Booting in runlevel 3, I see lots of stack traces (too much to even be able to read them).

In the past (with Mandriva's 2.6.27-rc) it would at least succeed in booting (with one or a few backtraces), but even then the e1000e network card would timeout a few seconds after when starting to transfer some data, after which the network was totally unusable until a reboot. A reader of my blog confirmed having the same problem on this same type of laptop and also solving it by using intel_iommu=off.

I entered some more information here:
http://bugzilla.kernel.org/show_bug.cgi?id=11821

Signature · 1 line
Frederik Himpe <[email protected]>
#4 lore raw
Quoted 32 lines · Frederik Himpe
Quoted 10 lines · Rafael J. Wysocki
Quoted 9 lines · Frederik Himpe
Quoted 6 lines · Frederik Himpe

On my Dell Latitude E6400 system with Intel chipset and graphics, I get this oops in my 2.6.27-rc8-git8 based kernel, which probably happens when hald-addon-dell-backlight is started.

Complete dmesg and lspci info:
https://qa.mandriva.com/show_bug.cgi?id=44711

Quoted 6 lines
Quoted 3 lines

Booting with intel_iommu=off seems to fix this issue, and seems to make my e1000e based network card stop from giving "transmit timed out" errors after which it stopped working.

Is it necessary to boot 2.6.27 final on this box with intel_iommu=off, and if so, is it still necessary in mainline?

I just tested stock 2.6.27.3 and this machine (Dell Latitude E6400) is totally unusable without intel_iommu=off if Intel VT is enabled in the BIOS. During the boot process, stack traces start scrolling over the screen. In the end, it goes on starting X, and from that moment, the machine locks up completely. Booting in runlevel 3, I see lots of stack traces (too much to even be able to read them).

In the past (with Mandriva's 2.6.27-rc) it would at least succeed in booting (with one or a few backtraces), but even then the e1000e network card would timeout a few seconds after when starting to transfer some data, after which the network was totally unusable until a reboot. A reader of my blog confirmed having the same problem on this same type of laptop and also solving it by using intel_iommu=off.

Quoted 2 lines

I entered some more information here:
http://bugzilla.kernel.org/show_bug.cgi?id=11821

OK, thanks.

I won't consider this as a regression, then.

Rafael