MCE现象
Intel在Pentium 4、Xenon和P6系列处理器中实现了机器检查（Machinecheck）架构，提供能够检测和报告硬件（机器）的错误机制，如系统总线错误、ECC错误、奇偶校验错误、缓存错误、TLB错误等。它包括一直MSR（Model-Specific Registers）寄存器，用来设置机器检查和额外的bank MSR记录错误。
当机器检查到不可纠正的machine-check错误时，就触发一个machine-check异常。machine-check架构不允许在出现MCE后处理器重启，但MCE处理程序可以从MSR寄存器收集相关信息。
CPU 7: Machine Check Exception: 5 Bank 0: b200004010000400
RIP !INEXACT! 10:<ffffffff8010f16e> {mwait_idle+0x5e/0x90}
TSC 1952dbeebcc8
Kernel panic: Machine check
Reconfiguring memory bank information….
This may take a while….
done waiting: 3 cpus not responding
Warning: Non-empty request queue
I/O requests in flight at dump time
CPU 7: Machine Check Exception: 4 Bank 0: f200004040000400
RIP !INEXACT! 10:<ffffffff8011ef69>
MCE错误判断原则
凡是内核死机打印“Machine Check Exception“或内核栈信息中打印有do_machine_check()函数，均为MCE问题。
MCE错误来源
- PCI-E设备信号质量/时钟
CPU芯片损坏/设计BUG
CPU Cache损坏或其它故障
CPU可能的缺陷
如CPU生产制造过程中带来的缺陷
- 内存坏/接触不良
- BIOS配置不当
- OS/MCE中断程序Bug
- 环境因素，如温度/湿度
![x86服务器MCE（Machine Check Exception）问题-LMLPHP]( "x86服务器MCE（Machine Check Exception）问题-LMLPHP")
MCE错误码解析
![x86服务器MCE（Machine Check Exception）问题-LMLPHP]( "x86服务器MCE（Machine Check Exception）问题-LMLPHP")
以上面MCE错误为例，Machine Check Exception和Bank 0(5)的值分别对应IA32_MCG_STATUS MSR、IA32_MCi_STATUS寄存器。
则对应的寄存器值为：
IA32_MCG_STATUS MSR寄存器的值为0000000000000004
IA32_MC0_STATUS MSR的值为f200000410000800
IA32_MC5_STATUS MSR的值为f200001044100e0f
根据MSR的值，对照Intel编程手册和Intel其他资料，就可以比较容易找出MCE原因。
dmesg显示
1
2
3
4
5
6
7
8
| ...
sbridge: HANDLING MCE MEMORY ERROR
CPU 0: Machine Check Exception: 0 Bank 5: 8c00004000010093
TSC 0 ADDR 67081b300 MISC 2140040486 PROCESSOR 0:206d7 TIME 1441181676 SOCKET 0 APIC 0
EDAC MC0: CE row 2, channel 0, label "CPU_SrcID#0_Channel#3_DIMM#0": 1 Unknown error(s): memory read on FATAL area : cpu=0 Err=0001:0093 (ch=3), addr= 0x67081b300 => socket=0, Channel=3(mask=8), rank=0
...
|
保存4行log为mlog
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
| # mcelog --ascii < /tmp/mlog
WARNING: with --dmi mcelog --ascii must run on the same machine with the
same BIOS/memory configuration as where the machine check occurred.
sbridge: HANDLING MCE MEMORY ERROR
CPU 0: Machine Check Exception: 0 Bank 5: 8c00004000010093
HARDWARE ERROR. This is *NOT* a software problem!
Please contact your hardware vendor
Wed Sep 2 16:14:36 2015
CPU 0 BANK 5 MISC 2140040486 ADDR 67081b300
STATUS 8c00004000010093 MCGSTATUS 0
CPUID Vendor Intel Family 6 Model 45
WARNING: SMBIOS data is often unreliable. Take with a grain of salt!
<24> DIMM 1333 Mhz Res13 Width 72 Data Width 64 Size 16 GB
Device Locator: Node0_Channel2_Dimm0
Bank Locator: Node0_Bank0
Manufacturer: Hynix Semiconducto
Serial Number: 40743B5A
Asset Tag: Dimm2_AssetTag
Part Number: HMT42GR7BFR4A-PB
TSC 0 ADDR 67081b300 MISC 2140040486 PROCESSOR 0:206d7 TIME 1441181676 SOCKET 0 APIC 0
EDAC MC0: CE row 2, channel 0, label "CPU_SrcID#0_Channel#3_DIMM#0": 1 Unknown error(s): memory read on FATAL area : cpu=0 Err=0001:0093 (ch=3), addr = 0x67081b300 => socket=0, Channel=3(mask=8), rank=0
|
根据
Part Number: HMT42GR7BFR4A-PB
Serial Number: 40743B5A
在lshw中找相应硬件
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
| ...
*-memory:0
description: System Memory
physical id: 2d
slot: System board or motherboard
*-bank:0
description: DIMM 1333 MHz (0.8 ns)
product: HMT42GR7BFR4A-PB
vendor: Hynix Semiconducto
physical id: 0
serial: 905D21AE
slot: Node0_Channel1_Dimm0
size: 16GiB
width: 64 bits
clock: 1333MHz (0.8ns)
*-bank:1
description: DIMM Synchronous [empty]
product: A1_Dimm1_PartNumber
vendor: Dimm1_Manufacturer
physical id: 1
serial: Dimm1_SerNum
slot: Node0_Channel1_Dimm1
width: 64 bits
*-bank:2
description: DIMM 1333 MHz (0.8 ns)
product: HMT42GR7BFR4A-PB
vendor: Hynix Semiconducto
physical id: 2
serial: 40743B5A
slot: Node0_Channel2_Dimm0
size: 16GiB
width: 64 bits
clock: 1333MHz (0.8ns)
...
|