Linux fdisk -l引起内核死锁分析

Linux fdisk -l查看分区，结果打印一个分区后引起内核死锁。

打印

root@www:~# fdisk -l

Disk /dev/mtdblock0: 2 MB, 2621440 bytes
255 heads, 63 sectors/track, 0 cylinders
Units = cylinders of 16065 * 512 = 8225280 bytes

Disk /dev/mtdblock0 doesn't contain a valid partition table


INFO: task update_mcu.sh:254 blocked for more than 120 seconds.
      Not tainted 3.18.20 #2
"echo 0 > /proc/sys/kernel/hung_task_timeout_secs" disables this message.
update_mcu.sh   D c0887e9c     0   254      1 0x00000000
Backtrace: 
[<c0887b60>] (__schedule) from [<c08880f4>] (schedule+0xa0/0xa4)
 r10:019b00ec r9:00000002 r8:ffffffff r7:cb0c2200 r6:ce94d448 r5:cd81a018
 r4:cd81a000
[<c0888054>] (schedule) from [<c08883e4>] (schedule_preempt_disabled+0x30/0x40)
[<c08883b4>] (schedule_preempt_disabled) from [<c0889c3c>] (__mutex_lock_slowpath+0xc0/0x188)
 r4:ce94d444 r3:00000001
[<c0889b7c>] (__mutex_lock_slowpath) from [<c0889d3c>] (mutex_lock+0x38/0x3c)
 r9:ce94d444 r8:cf844590 r7:cf844500 r6:cd81b938 r5:cf8449f4 r4:ce94d010
[<c0889d04>] (mutex_lock) from [<c03c8f74>] (msm_nand_read_oob+0x3f0/0x8fc)
[<c03c8b84>] (msm_nand_read_oob) from [<c03c95c8>] (msm_nand_read_partial_page+0x148/0x22c)

分析

可以看到是内核死锁超时2分钟打印出backtrace，死锁函数msm_nand_read_oob中的mutex_lock锁等待超时死锁，通过Linux下使用额外增加文件方式查看debug信息添加开关代码如下：

static int msm_nand_read_oob(struct mtd_info *mtd, loff_t from,
			     struct mtd_oob_ops *ops)
{

		if(1 == io_run_num) {
			printk(KERN_ERR "eason in\n");
		}
		mutex_lock(&info->lock);
		if(1 == io_run_num) {
			printk(KERN_ERR "eason out\n");
		}

fdisk -l死锁后打印如下：

eason in
eason out
QCMAP:bringup v6
enter sleep, msm_uport->obs: 0
msm_hs_pm_resume(): jun test1
resule from sleep, msm_uport->obs: 0
eason in
enter sleep, msm_uport->obs: 0
没有out出

移远回复说不支持查看nand flash，如果有其他设备可以使用fdisk /dev/*进行查看，原因未告知，且暂不解决；我个人认为这不合理的，其他设备上面fdisk就可以查看flash，且导致了内核死锁应该就是可以查看原因和解决的。

原因待继续查。。。