根據(jù)linux Oops定位錯誤代碼行

zhuqing_739

這幾天一直在調試atmel at91sam9x25的串口，用著用著總會導致Oops，Oops內(nèi)容如下：
[ 1023.510000] Unable to handle kernel NULL pointer dereference at virtual address 00000000
[ 1023.520000] pgd = c0004000
[ 1023.520000] [00000000] *pgd=00000000
[ 1023.520000] Internal error: Oops: 17 [#1]
[ 1023.520000] last sysfs file: /sys/devices/virtual/misc/at91flash/dev
[ 1023.520000] Modules linked in: at91flash at91gpio at91mc323 ds18b20 at91adc
[ 1023.520000] CPU: 0    Tainted: G        W    (2.6.39 #35)
[ 1023.520000] PC is at atmel_tasklet_func+0x104/0x690
[ 1023.520000] LR is at atmel_tasklet_func+0x10/0x690
[ 1023.520000] pc : [<c01a33ac>]    lr : [<c01a32b8>]    psr: 20000013
[ 1023.520000] sp : c7825f58  ip : 60000093  fp : 00000000
[ 1023.520000] r10: 00000006  r9 : 00000000  r8 : 0000000a
[ 1023.520000] r7 : 00000000  r6 : c7824000  r5 : c78a2484  r4 : c03c0cb8
[ 1023.520000] r3 : 0000004c  r2 : 0000004c  r1 : 60000013  r0 : 00000001
[ 1023.520000] Flags: nzCv  IRQs on  FIQs on  Mode SVC_32  ISA ARM  Segment kernel
[ 1023.520000] Control: 0005317f  Table: 27b40000  DAC: 00000017
[ 1023.520000] Process ksoftirqd/0 (pid: 3, stack limit = 0xc7824270)
[ 1023.520000] Stack: (0xc7825f58 to 0xc7826000)
[ 1023.520000] 5f40:                                                       00000001 c7824000
[ 1023.520000] 5f60: 00000100 0000000a 00000000 00000006 c7825f8c 00000000 00000001 c7824000
[ 1023.520000] 5f80: 00000100 0000000a 00000006 c0045cf8 c03b995c c00461d8 c7aa6ae0 00000000
[ 1023.520000] 5fa0: 60000093 00000000 c7824000 c0046274 00000013 00000000 00000000 c00462e0
[ 1023.520000] 5fc0: 00000000 c7819f70 00000000 c00570e0 00000000 00000000 00000000 00000000
[ 1023.520000] 5fe0: c7825fe0 c7825fe0 c7819f70 c0057060 c0030b14 c0030b14 ffffffff ffffffff
[ 1023.520000] [<c01a33ac>] (atmel_tasklet_func+0x104/0x690) from [<c0045cf8>] (tasklet_action+0x84/0xe
[ 1023.520000] [<c0045cf8>] (tasklet_action+0x84/0xe from [<c00461d8>] (__do_softirq+0x88/0x124)
[ 1023.520000] [<c00461d8>] (__do_softirq+0x88/0x124) from [<c00462e0>] (run_ksoftirqd+0x6c/0x12
[ 1023.520000] [<c00462e0>] (run_ksoftirqd+0x6c/0x12 from [<c00570e0>] (kthread+0x80/0x8
[ 1023.520000] [<c00570e0>] (kthread+0x80/0x8 from [<c0030b14>] (kernel_thread_exit+0x0/0x
[ 1023.520000] Code: 1a000002 e59f057c e59f157c ebfa3d49 (e5973000)
[ 1023.710000] ---[ end trace 786b41cd25d3b661 ]---
[ 1023.710000] Kernel panic - not syncing: Fatal exception in interrupt
[ 1023.720000] [<c0034b10>] (unwind_backtrace+0x0/0xe0) from [<c02a8af8>] (panic+0x50/0x170)
[ 1023.720000] [<c02a8af8>] (panic+0x50/0x170) from [<c0032e00>] (die+0x184/0x1c4)
[ 1023.730000] [<c0032e00>] (die+0x184/0x1c4) from [<c0035aa8>] (__do_kernel_fault+0x64/0x84)
[ 1023.740000] [<c0035aa8>] (__do_kernel_fault+0x64/0x84) from [<c0035c7c>] (do_page_fault+0x1b4/0x1c
[ 1023.750000] [<c0035c7c>] (do_page_fault+0x1b4/0x1c from [<c002a240>] (do_DataAbort+0x30/0x9
[ 1023.760000] [<c002a240>] (do_DataAbort+0x30/0x98) from [<c002f86c>] (__dabt_svc+0x4c/0x60)
[ 1023.770000] Exception stack(0xc7825f10 to 0xc7825f58)
[ 1023.770000] 5f00:                                     00000001 60000013 0000004c 0000004c
[ 1023.780000] 5f20: c03c0cb8 c78a2484 c7824000 00000000 0000000a 00000000 00000006 00000000
[ 1023.790000] 5f40: 60000093 c7825f58 c01a32b8 c01a33ac 20000013 ffffffff
[ 1023.790000] [<c002f86c>] (__dabt_svc+0x4c/0x60) from [<c01a33ac>] (atmel_tasklet_func+0x104/0x690)
[ 1023.800000] [<c01a33ac>] (atmel_tasklet_func+0x104/0x690) from [<c0045cf8>] (tasklet_action+0x84/0xe8)
[ 1023.810000] [<c0045cf8>] (tasklet_action+0x84/0xe8) from [<c00461d8>] (__do_softirq+0x88/0x124)
[ 1023.820000] [<c00461d8>] (__do_softirq+0x88/0x124) from [<c00462e0>] (run_ksoftirqd+0x6c/0x128)
[ 1023.830000] [<c00462e0>] (run_ksoftirqd+0x6c/0x128) from [<c00570e0>] (kthread+0x80/0x88)
[ 1023.840000] [<c00570e0>] (kthread+0x80/0x88) from [<c0030b14>] (kernel_thread_exit+0x0/0x8)
注意上述紅色的地方。
下面就來顯示如何定位出出錯代碼行:
1.首先，編譯時打開complie with debug info選項，步則如下
make ARCH=arm CROSS_COMPILE=arm-none-linux-gnueabi- menuconfig

進入 Kernel hacking

選擇 Compile the kernel with debug info
然后，保存，退出。
接著 make ARCH=arm CROSS_COMPILE=arm-none-linux-gnueabi-
編譯， 等編譯完成。
2.利用arm-none-linux-gnueabi-gdb 調試，如下：
arm-none-linux-gnueabi-gdb vmlinux

對應著Oops 消息里面的這一行
[ 1023.520000] LR is at atmel_tasklet_func+0x10/0x690
在gdb下鍵入命令 ： l *at atmel_tasklet_func+0x10

這樣就找到了出錯的代碼行。在這里鄙視一下atmel提供的內(nèi)核，竟然還有bug，fuck it！
從這里可以看出是由于串口的dma導致Oops的，于是我去掉了串口的dma傳輸。方法如下：

去掉之后還沒有發(fā)現(xiàn)上述的Oops出現(xiàn)。

灌水菜鳥

CROSS_COMPILE 的 objdump不行么？

npuazm

atmel DMA 傳輸，是什么bug？不能在fix掉嗎？
感謝樓主的share

瀚海書香

回復 1# zhuqing_739
不錯，多謝分享！
串口傳輸?shù)拇_是不同需要DMA的。

   

tempname2

addr2line.

tempname2

最新內(nèi)核么？LZ跟一跟啊，這么好一個BUG。

linuxfellow

回復 1# zhuqing_739
有時不一定內(nèi)核代碼有問題，DMA需要用物理地址，配置是這一點一定要注意。如果再深入一步， 找到DMA的問題，那么對DMA也能完全懂了。

   

victure83

不錯，學習

25641463

我用的比較搓的方法，需要對反匯編看的比較熟才行。
就是把導致oops的.ko拖到 ida里，然后搜索一下oops打印的 函數(shù)名+偏移，定位到匯編行，然后看一眼附近的匯編代碼，再看源碼。。。。