dump / lcrash dump crash dump 2

Size: px

Start display at page:

Download "dump / lcrash dump crash dump 2"

なおちかおとじま
5 years ago
Views:

1 Linux crash dump Linux Kernel crash dump Asianux

2 dump / lcrash dump crash dump 2

3 Dump? Linux / Linux IA +Linux 3

4 Dump (1) ( ) Kernel Panic Linux Linux 4

5 Dump (2) Kernel Panic Linux Linux (Dump) Kernelpanic =>Dump 5

6 Dump (1) Linux (OS) Kernel Panic ( ) 6

7 Dump (2) KernelPanic CPU,,GPU,,DISK,RAID! Kernel bug ( )! 0! Oops! 7

8 Dump LKCD(Linux Kernel Crash Dump) MiracleLinux Ver2.1,Ver3 diskdump,netdump MiracleLinux Ver4,RHEL4 Kdump Asianux Server 3,RHEL5 Dump 8

9 LKCD LKCD Swap dump dump dump /var/log/dump swap Dump 9

10 diskdump dump dump /etc/sysconfig/diskdump # service diskdump initalformat # service diskdump start # chkconfig diskdump on Dump 10

11 netdump ssh dump netdump SSH netdump Dump 11

12 Kdump(1) dump kernel 1. dump Dump 4. dump 12

13 Kdump(2) dump kernel 13

14 Kdump(3) /boot/grub/menu.lst crashkernel= 14

15 (1) panic # cat /proc/sys/kernel/panic reboot ( 0: ) : :panic /var/log/message netconsole, 15

16 (2) oops oops:kernel # cat /proc/sys/kernel/panic_on_oops oops panic (1) oops (0) : :.. oops Panic 16

17 (3) nmi_watchdog nmi_watchdog=1 nmi_watchdog=2 /proc/interrupts NMI Oops (panic_on_oops panic,dump) : sysrq /proc sync,dump, /etc/sysctl.conf kernel.sysrq = 1 #sysctl -p # echo 1 > /proc/sys/kernel/sysrq 17

18 (4) /var dump Disk LKCD : /var/log/dump diskdump,kdump : /var/crash netdump : netdump /var/crash 18

19 Dump NMI ( ) Sysq-key Alt-SysRq-s 'S'ync Alt-SysRq-c 'C'rach Alt-SysRq-b re'b'oot Sysrq-trigger # sync;sync;sync # echo 1 > /proc/sys/kernel/sysrq # echo c > /proc/sysrq-trigger 19

20 (1) 20

21 (2) 21

22 dump LKCD dump.x( ) map.x( ) kerntypes.x Kdump,diskdump,netdump vmcore( ) 22

23 dump lcrash lkcd MiracleLinux Ver2.1,Ver3.0 crash Kdump,diskdump,netdump MiracleLinux Ver4.0 Asianux Server 3 RHEL4,5 23

24 dump IA (x86_64) C kernel 24

25 CPU CPU 32bit?x86_64 32bit=i386 i686,x86_64=amd64,intel 64(EMT64) CPU [(E R)AX] [(E R)SP] [(E R)IP] [(E R)FLAGS] a1 a2 a3 a4 a5 a6 a7 1 a5 r1 2 a6 r2 3 r1 r2 r2 4 r2 a7 1 [1] 2 [1] 3 [a3] Intel 25

26 C C GCC 26

27 kernel kernel kernel kernel bugzilla LKML 27

28 Dump (lcrash) 1. Dump 2. analysis.x 3. lcrash dump 4. (stat,report) kernel 5. (bt) panic 6. (ps) 7. (dis) (print ) 28

29 1 cron whatis,locate /var/log/dump dump LKCD dump 29

30 analysis.x( ) 30

31 1 lcrash 31

32 lcrash stat,report( ) LKST 32

33 lcrash ps( ) 33

34 lcrash bt( ) Dump 34

35 nvdevmgr=> panic sg SCSI kernel bug 35

36 ebx dis( ) panic generic_unplug_device+11 ebx 36

37 generic_unplug_device scsi_do_req lcrash dis( 2) sg_common_write generic_unplug_device 37

38 (sg.c) 38

39 / (1) (src.rpm) sg.c e.25.50ml kernel RHEL e.72 sg.c * Mon Jun Tom Coughlan <coughlan@redhat.com> - Fix bug 75669, SG driver kernel oops due to a null pointer in sg_common_write. Fixed based on upstream. Patch added to scsifixes.patch. 39

40 / (1) e.72 spec ======================================== * Mon Jun Tom Coughlan <coughlan@redhat.com> - Fix bug 75669, SG driver kernel oops due to a null pointer in sg_common_write. Fixed based on upstream. Patch added to scsifixes.patch. ======================================== 40

41 SPEC bugzilla 41

42 1 Kernel Kernel 42

43 crash dump (1) debug vmlinux(debug kernel ) dump kernel debug kernel ) kernel-pae-debuginfo ax.i686.rpm kernel-debuginfo ax.x86_64.rpm System.map( kernel ) dump kernel vmcore(dump : ) :/var/crash/ /vmcore debug kernel 43

44 crash dump (2) dump (crash )dump dump (32bit,x86_64 ) 44

45 Dump (crash) 1. Dump 2. crash dump 3. (log) kernel 4. (bt) panic 5. (ps) 6. (dis) 7. (gdb list) 8. (struct xx ) 45

46 2 panic! 46

47 (1) crash 47

48 log( ) nullpointer panic 48

49 bt( ) dump 49

50 dis ( ) 50

51 gdb list ( ) 51

52 2 C shmcdump.c ( ) 52

53 kernel panic dump dump LKCD diskdump/netdump kdump kernel panic Linux 53

54 54

55 / Miracle Linux Support MiracleLinux Asianux blog Alicia(Advanced LInux Crash-dump Interactive Analyzer) intel google 3/3 ( ) 55

Linux障害解析の状況と今後の展望

Linux障害解析の状況と今後の展望 Linux 障害解析の状況と今後の展望 Oct.15, 2004 NTT Data 先端技術株式会社鈴木幸市 koichi@intellilink.co.jp エンタープライズ環境で Linux の利用を促進するうえでカーネル障害解析は重要な位置を占めるカーネルダンプとしては LKCD や NetDump が知られているがより高い確実性信頼性を目指した Linux カーネルダンプ取得ツールの提案もなされてきている