[Workaround] V7.1 crashes on use of SENSORS command

[dgscos]

The following is the back trace from the crash:

Version: 3.10.0-229.1.2.el7.x86_64
BUG: unable to handle kernel NULL pointer dereference at 00000000000001d8
IP: [<ffffffffa0215fd2>] radeon_hwmon_show_temp+0x32/0x70 [radeon]
PGD 0
Oops: 0000 [#1] SMP
Modules linked in: bnep bluetooth rfkill fuse xt_CHECKSUM ipt_MASQUERADE nf_nat_masquerade_ipv4 tun ip6t_rpfilter ip6t_REJECT ipt_REJECT xt_conntrack ebtable_nat ebtable_broute bridge stp llc ebtable_filter ebtables ip6table_nat nf_conntrack_ipv6 nf_defrag_ipv6 nf_nat_ipv6 ip6table_mangle ip6table_security ip6table_raw ip6table_filter ip6_tables iptable_nat nf_conntrack_ipv4 nf_defrag_ipv4 nf_nat_ipv4 nf_nat nf_conntrack iptable_mangle iptable_security iptable_raw iptable_filter ip_tables ext4 mbcache jbd2 kvm_amd kvm dm_mirror dm_region_hash dm_log dm_mod snd_hda_codec_realtek snd_hda_codec_generic snd_hda_codec_hdmi snd_hda_intel snd_hda_controller snd_hda_codec snd_hwdep snd_seq snd_seq_device snd_pcm edac_mce_amd edac_core shpchp pcspkr serio_raw k10temp acpi_cpufreq tpm_infineon
snd_timer snd soundcore sp5100_tco i2c_piix4 wmi nfsd auth_rpcgss nfs_acl lockd uinput sunrpc xfs libcrc32c usb_storage sd_mod sr_mod crc_t10dif cdrom crct10dif_common ata_generic pata_acpi radeon i2c_algo_bit drm_kms_helper ahci pata_jmicron firewire_ohci libahci pata_atiixp ttm drm libata r8169 firewire_core i2c_core mii crc_itu_t [last unloaded: i2c_dev]
CPU: 4 PID: 3856 Comm: sensors Not tainted 3.10.0-229.1.2.el7.x86_64 #1
Hardware name: MSI MS-7640/890FXA-GD70 (MS-7640) , BIOS V1.6 05/18/2010
task: ffff88041d5d96c0 ti: ffff8803e4908000 task.ti: ffff8803e4908000
RIP: 0010:[<ffffffffa0215fd2>] [<ffffffffa0215fd2>] radeon_hwmon_show_temp+0x32/0x70 [radeon]
RSP: 0018:ffff8803e490be88 EFLAGS: 00010246
RAX: ffff88041da0c000 RBX: ffff8800b8744000 RCX: 0000000000000000
RDX: 0000000000000000 RSI: ffffffffa02eb460 RDI: ffff8804216e6c00
RBP: ffff8803e490be90 R08: ffffffff8183c4e0 R09: ffffea0002e1d140
R10: 00000000000041c6 R11: 0000000000000246 R12: ffff8803e490bf48
R13: 0000000000001000 R14: ffff8800b87155a0 R15: ffff88041cad74d0
FS: 00007f9c88bee740(0000) GS:ffff880437d00000(0000) knlGS:0000000000000000
CS: 0010 DS: 0000 ES: 0000 CR0: 0000000080050033
CR2: 00000000000001d8 CR3: 00000003e3892000 CR4: 00000000000007e0
DR0: 0000000000000000 DR1: 0000000000000000 DR2: 0000000000000000
DR3: 0000000000000000 DR6: 00000000ffff0ff0 DR7: 0000000000000400
Stack:
ffffffffa02eb460 ffff8803e490beb0 ffffffff813cec30 ffff8803e490beb0
ffff8800b87155c0 ffff8803e490bf00 ffffffff8123d38a ffff8804216e6c10
ffffffff8168eaf0 00007f9c88c18000 ffff8803e7dace00 00007f9c88c18000
Call Trace:
[<ffffffff813cec30>] dev_attr_show+0x20/0x60
[<ffffffff8123d38a>] sysfs_read_file+0x9a/0x1a0
[<ffffffff811c6acc>] vfs_read+0x9c/0x170
[<ffffffff811c75f8>] SyS_read+0x58/0xb0
[<ffffffff81614a29>] system_call_fastpath+0x16/0x1b
Code: 89 e5 53 48 89 d3 e8 7e cd 1b e1 f6 80 6b 01 00 00 02 48 8b 50 08 74 0a 8b 92 60 05 00 00 85 d2 75 37 48 8b 90 e0 16 00 00 31 c9 <48> 8b b2 d8 01 00 00 48 85 f6 74 07 48 89 c7 ff d6 89 c1 48 c7
RIP [<ffffffffa0215fd2>] radeon_hwmon_show_temp+0x32/0x70 [radeon]
RSP <ffff8803e490be88>

Note: the "sensors" command works fine on the previous kernel, 3.10.0-123...

I am not a developer (kernel or otherwise), but this looks to my untrained eye very much like a bug in either the V7.1 kernel or the radeon driver. Everything else seems to be working on V7.1.

Anyone have any ideas on a workaround?

Thanks.

DGS.

[TrevorH]

Did you try it on the original 7.1 kernel? (I suspect it will still fail but it would tell you if it broke between 7.0 and 7.1 or between 3.10.0-229.el7 and 3.10.0-229.1.2.el7).

You could try the ELRepo packaged version of the ATI drivers instead of the opensource radeon driver. Otherwise, I think your best bet is to open a problem report on bugzilla.redhat.com and report the issue there since the fix would need to come from upstream.

[dgscos]

I have followed your suggestion and created a problem report on bugzilla.redhat.com.

I investigated fglrx drivers on ELRepo. There are none for V7. There are "a bunch" for V6. Do the V6 fglrx drivers
work on V7? Since this is my "home production machine", I am reluctant to experiment with graphic drivers. I have commented out the sensors commands in the appropriate scripts, and for the time being will wait to see if Red Hat fixes this issue.

Thanks for your help.

[TrevorH]

There is a kmod-fglrx package for el7 but it's currently in the elrepo-testing repository, waiting enough feedback from successful users for it to be promoted to the main repo.

[dgscos]

It took me awhile to setup a test CentOS V7.1 machine and install the fglrx drivers from elrepo-testing, but I can report that using the fglrx drivers seems to be a good workaround. Use of the sensors command now does not crash the OS and correctly reports the CPU and graphics card temps.

Thus, it appears that the bug is in the radeon driver and not the kernel (3.10.0-229.1.2).

Thanks.

[toracat]

It is nice to know that kmod-fglrx can be used as a workaround. There is a bug report on the CentOS bug tracker. I've update it with this info.

http://bugs.centos.org/view.php?id=8449

[toracat]

Oops, that bug report is "private". I'm trying to find out if this can be made public.

[toracat]

The reason why it was made private is that this is a potential security issue (non-root user can crash the system). The bug is now viewable.