记录linuxtty的一次软锁排查2

在复现tty的死锁问题的时候，文洋兄使用了如下的方式：

#include
#include
#include
#define TIOCVHANGUP 0x5437
int main(int argc,char* argv[])
{
int fd;
if(argc <2)
{
printf("error,you should input tty as a parameter\r\n");
return 1;
}
fd = open(argv[1], O_WRONLY | O_NOCTTY);

          if(fd<0)
          {
                return 1;
          }

write(fd, "test tty\n ", 20);
ioctl(fd, TIOCVHANGUP, 0);
//sleep(1);
close(fd);
return 0;
}

编译成gcc -g -o main.o main.c ,然后使用脚本呼叫：

#!/bin/bash
while [ 1 ]
do
./main.o /dev/tty4
done

之所以使用脚本而不是在c中while处理，是因为在进程exit的时候，会有些tty的处理,我们希望尽可能地覆盖测试，所以甚至都没有加sleep来延时。

结果复现出来下面的软锁故障，堆栈如下：

[517571.855382] INFO: task systemd:1 blocked for more than 120 seconds.
[517571.856127] "echo 0 > /proc/sys/kernel/hung_task_timeout_secs" disables this message.
[517571.856846] systemd D ffff881fffc347c0 0 1 0 0x00000000
[517571.856852] ffff881fd35c7b50 0000000000000086 ffff881fd35c7fd8 ffff881fd35c7fd8
[517571.856859] ffff881fd35c7fd8 00000000000147c0 ffff881fd313c500 ffff883f5ee2ac80
[517571.856863] ffff883f5ee2ac84 ffff883fd1630000 00000000ffffffff ffff883f5ee2ac88
[517571.856867] Call Trace:
[517571.856880] [] schedule_preempt_disabled+0x29/0x70
[517571.856883] [] __mutex_lock_slowpath+0xc5/0x1c0
[517571.856888] [] mutex_lock+0x1f/0x2f
[517571.856890] [] tty_lock_nested.isra.0+0x38/0x90
[517571.856892] [] tty_lock+0xe/0x10
[517571.856899] [] tty_open+0xcc/0x620
[517571.856906] [] chrdev_open+0xa1/0x1e0
[517571.856912] [] do_dentry_open+0x1a7/0x2e0
[517571.856916] [] ? cdev_put+0x30/0x30
[517571.856918] [] vfs_open+0x39/0x70
[517571.856922] [] do_last+0x1ed/0x1270
[517571.856925] [] path_openat+0xc2/0x490
[517571.856930] [] ? __wake_up_common+0x58/0x90
[517571.856935] [] do_filp_open+0x4b/0xb0
[517571.856941] [] ? __alloc_fd+0xa7/0x130
[517571.856945] [] do_sys_open+0xf3/0x1f0
[517571.856949] [] SyS_open+0x1e/0x20
[517571.856955] [] system_call_fastpath+0x16/0x1b

从堆栈看，显然又是在等锁超时了。反汇编找到这把锁是关键。

void __lockfunc tty_lock(struct tty_struct *tty)
{
return tty_lock_nested(tty, TTY_MUTEX_NORMAL);
}

static void __lockfunc tty_lock_nested(struct tty_struct *tty,
unsigned int subclass)
{
if (tty->magic != TTY_MAGIC) {
pr_err("L Bad %p\n", tty);
WARN_ON(1);
return;
}
tty_kref_get(tty);
mutex_lock_nested(&tty->legacy_mutex, subclass);--------------传入锁的指针
}

由于CONFIG_DEBUG_LOCK_ALLOC并没有配置，所以mutex_lock_nested就是mutex_lock。和堆栈是匹配的。

# define mutex_lock_nested(lock, subclass) mutex_lock(lock)

crash> dis -l tty_lock_nested
/usr/src/debug/kernel-3.10.0-327.22.2.el7/linux-3.10.0-327.22.2.el7.x86_64/drivers/tty/tty_mutex.c: 18
0xffffffff81640dc0 : nopl 0x0(%rax,%rax,1) [FTRACE NOP]
0xffffffff81640dc5 5>: push %rbp
0xffffffff81640dc6 6>: mov %rsp,%rbp
0xffffffff81640dc9 9>: push %rbx
/usr/src/debug/kernel-3.10.0-327.22.2.el7/linux-3.10.0-327.22.2.el7.x86_64/drivers/tty/tty_mutex.c: 21
0xffffffff81640dca 10>: cmpl $0x5401,(%rdi)
/usr/src/debug/kernel-3.10.0-327.22.2.el7/linux-3.10.0-327.22.2.el7.x86_64/drivers/tty/tty_mutex.c: 18
0xffffffff81640dd0 16>: mov %rdi,%rbx
/usr/src/debug/kernel-3.10.0-327.22.2.el7/linux-3.10.0-327.22.2.el7.x86_64/drivers/tty/tty_mutex.c: 21
0xffffffff81640dd3 19>: jne 0xffffffff81640dfb 59>
/usr/src/debug/kernel-3.10.0-327.22.2.el7/linux-3.10.0-327.22.2.el7.x86_64/include/linux/tty.h: 388
0xffffffff81640dd5 21>: test %rdi,%rdi
0xffffffff81640dd8 24>: je 0xffffffff81640dec 44>
/usr/src/debug/kernel-3.10.0-327.22.2.el7/linux-3.10.0-327.22.2.el7.x86_64/arch/x86/include/asm/atomic.h: 176
0xffffffff81640dda 26>: mov $0x1,%eax
0xffffffff81640ddf 31>: lock xadd %eax,0x4(%rdi)
0xffffffff81640de4 36>: add $0x1,%eax
/usr/src/debug/kernel-3.10.0-327.22.2.el7/linux-3.10.0-327.22.2.el7.x86_64/include/linux/kref.h: 47
0xffffffff81640de7 39>: cmp $0x1,%eax
0xffffffff81640dea 42>: jle 0xffffffff81640e1f 95>
/usr/src/debug/kernel-3.10.0-327.22.2.el7/linux-3.10.0-327.22.2.el7.x86_64/drivers/tty/tty_mutex.c: 27
0xffffffff81640dec 44>: lea 0x80(%rbx),%rdi------------------传入的参数是一把锁的地址，即&tty->legacy_mutex，rbx就是tty的指针了。

0xffffffff81640df3 51>: callq 0xffffffff8163c860 --------------------调用mutex_lock

crash> dis -l mutex_lock
/usr/src/debug/kernel-3.10.0-327.22.2.el7/linux-3.10.0-327.22.2.el7.x86_64/kernel/mutex.c: 103
0xffffffff8163c860 : nopl 0x0(%rax,%rax,1) [FTRACE NOP]
0xffffffff8163c865 5>: push %rbp
0xffffffff8163c866 6>: mov %rsp,%rbp
0xffffffff8163c869 9>: push %rbx--------------------------------------------------rbx压栈,所以rbp后面就是rbx的值

所以我们能够通过堆栈分析出tty的指针来，rbx的压栈的位置是在rbp之后。

ffff881fd35c7bc0: ffff881fd35c7bd8 ffffffff8163c87f
#3 [ffff881fd35c7bc8] mutex_lock at ffffffff8163c87f
ffff881fd35c7bd0: ffff883f5ee2ac00 ffff881fd35c7bf0 -----------------------ffff883f5ee2ac00就是rbx的值，也就是tty指针
ffff881fd35c7be0: ffffffff81640df8
#4 [ffff881fd35c7be0] tty_lock_nested at ffffffff81640df8
ffff881fd35c7be8: ffff88211f6a3200 ffff881fd35c7c00
ffff881fd35c7bf8: ffffffff81640e5e

现在，需要找到持有这把锁的owner是谁。

crash> struct tty_struct.legacy_mutex ffff883f5ee2ac00
legacy_mutex = {
count = {
counter = -1
},
wait_lock = {
{
rlock = {
raw_lock = {
{
head_tail = 524296,
tickets = {
head = 8,
tail = 8
}
}
}
}
}
},
wait_list = {
next = 0xffff881fd35c7b70,
prev = 0xffff881fd35c7b70
},
owner = 0xffff880190f5c500, -----------------持有锁

查看对应的task：

crash> task 0xffff880190f5c500
PID: 5628 TASK: ffff880190f5c500 CPU: 47 COMMAND: "main.o"------------就是我们编译的测试命令

确认下是不是我们的tty4.

crash> struct tty_strt.name ffff883f5ee2ac00
name = "tty4\000\000\000\000\000\000\000\000\000\000\000\000\000\000\000\000\000\000\000\000\000\000\000\000\000\000\000\000\000\000\000\000\000\000\000\000\000\000\000\000\000\000\000\000\000\000\000\000\000\000\000\000\000\000\000\000\000\000\000"

确定无误后，看看进程打开的文件列表：

crash> files 5628
PID: 5628 TASK: ffff880190f5c500 CPU: 47 COMMAND: "main.o"
ROOT: / CWD: /home/caq
FD FILE DENTRY INODE TYPE PATH
0 ffff881f0e31a600 ffff880dd37f8000 ffff8801713fcea0 CHR /dev/pts/45
1 ffff881f0e31a600 ffff880dd37f8000 ffff8801713fcea0 CHR /dev/pts/45
2 ffff881f0e31a600 ffff880dd37f8000 ffff8801713fcea0 CHR /dev/pts/45
3 ffff881a00324400 ffff883fd1010fc0 ffff883fd0b73820 CHR /dev/tty4

查看对应的tty的属性：

crash> struct file.private_data ffff881a00324400
private_data = 0xffff883f6101e840
crash> struct tty_file_private.tty 0xffff883f6101e840
tty = 0xffff883f5ee2ac00
crash> struct tty_struct.disc_data 0xffff883f5ee2ac00----------------这个 0xffff883f5ee2ac00 也就是在前面反汇编找到的tty指针

disc_data = 0xffff883f9a1d8c00
crash> struct n_tty_data.icanon 0xffff883f9a1d8c00 icanon = 1 '\001'

当然也可以使用tty来直接查看。

最后殊途同归，还是同一个问题，属性导致的。

我们继续来看到底有多少进程被阻塞了：

# grep mutex_lock -A 5 -B 5 caq_all_bt.txt |grep tty_open |wc -l
90

# grep mutex_lock -A 5 -B 5 caq_all_bt.txt |grep tty_open
#6 [ffff881fd35c7c08] tty_open at ffffffff813b204c-----------------只有1号进程阻塞在这
#3 [ffff8820c222bc08] tty_open at ffffffff813b1ff7-----------------其余全部阻塞在这
#3 [ffff882aaa9b3c08] tty_open at ffffffff813b1ff7
#3 [ffff883f20ca7c08] tty_open at ffffffff813b1ff7
#3 [ffff882098d2bc08] tty_open at ffffffff813b1ff7
#3 [ffff88147ff87c08] tty_open at ffffffff813b1ff7
#3 [ffff8820ff4cbc08] tty_open at ffffffff813b1ff7
#3 [ffff88106e5c7c08] tty_open at ffffffff813b1ff7
#3 [ffff880192813c08] tty_open at ffffffff813b1ff7
#3 [ffff880164ccbc08] tty_open at ffffffff813b1ff7
#3 [ffff882093c13c08] tty_open at ffffffff813b1ff7
#3 [ffff8814221b7c08] tty_open at ffffffff813b1ff7
#3 [ffff883f3c74fc08] tty_open at ffffffff813b1ff7
#3 [ffff88136e433c08] tty_open at ffffffff813b1ff7
#3 [ffff882141f37c08] tty_open at ffffffff813b1ff7
#3 [ffff8820db4ebc08] tty_open at ffffffff813b1ff7
#3 [ffff88149471fc08] tty_open at ffffffff813b1ff7
#3 [ffff8801a4417c08] tty_open at ffffffff813b1ff7
#3 [ffff883f0acd3c08] tty_open at ffffffff813b1ff7
#3 [ffff883ebce9fc08] tty_open at ffffffff813b1ff7
#3 [ffff88208bfd3c08] tty_open at ffffffff813b1ff7
#3 [ffff882087d0bc08] tty_open at ffffffff813b1ff7
#3 [ffff8820d556bc08] tty_open at ffffffff813b1ff7
#3 [ffff8820c235bc08] tty_open at ffffffff813b1ff7
#3 [ffff8820e7ce3c08] tty_open at ffffffff813b1ff7
#3 [ffff88210c25fc08] tty_open at ffffffff813b1ff7
#3 [ffff8820ebe2fc08] tty_open at ffffffff813b1ff7
#3 [ffff8820e82c7c08] tty_open at ffffffff813b1ff7
#3 [ffff88212af2fc08] tty_open at ffffffff813b1ff7
#3 [ffff881ad4ef7c08] tty_open at ffffffff813b1ff7
#3 [ffff883f1a8afc08] tty_open at ffffffff813b1ff7
#3 [ffff88146efb3c08] tty_open at ffffffff813b1ff7
#3 [ffff8801c557fc08] tty_open at ffffffff813b1ff7
#3 [ffff88044e66fc08] tty_open at ffffffff813b1ff7
#3 [ffff8801664dbc08] tty_open at ffffffff813b1ff7
#3 [ffff8801a1fefc08] tty_open at ffffffff813b1ff7
#3 [ffff8801850c7c08] tty_open at ffffffff813b1ff7
#3 [ffff8801c6563c08] tty_open at ffffffff813b1ff7
#3 [ffff8801751dfc08] tty_open at ffffffff813b1ff7
#3 [ffff8801272fbc08] tty_open at ffffffff813b1ff7
#3 [ffff880173073c08] tty_open at ffffffff813b1ff7
#3 [ffff880179ccbc08] tty_open at ffffffff813b1ff7
#3 [ffff8813895f7c08] tty_open at ffffffff813b1ff7
#3 [ffff88152025fc08] tty_open at ffffffff813b1ff7
#3 [ffff88019e403c08] tty_open at ffffffff813b1ff7
#3 [ffff8801504f3c08] tty_open at ffffffff813b1ff7
#3 [ffff88017841fc08] tty_open at ffffffff813b1ff7
#3 [ffff88018e80fc08] tty_open at ffffffff813b1ff7
#3 [ffff881345b57c08] tty_open at ffffffff813b1ff7
#3 [ffff881f2c0ffc08] tty_open at ffffffff813b1ff7
#3 [ffff88049b78bc08] tty_open at ffffffff813b1ff7
#3 [ffff8801aff13c08] tty_open at ffffffff813b1ff7
#3 [ffff880186f77c08] tty_open at ffffffff813b1ff7
#3 [ffff8814fd963c08] tty_open at ffffffff813b1ff7
#3 [ffff8803d37dbc08] tty_open at ffffffff813b1ff7
#3 [ffff8801cacfbc08] tty_open at ffffffff813b1ff7
#3 [ffff8801d6937c08] tty_open at ffffffff813b1ff7
#3 [ffff8805689d3c08] tty_open at ffffffff813b1ff7
#3 [ffff883f8b9d7c08] tty_open at ffffffff813b1ff7
#3 [ffff883f7d873c08] tty_open at ffffffff813b1ff7
#3 [ffff8801fd47bc08] tty_open at ffffffff813b1ff7
#3 [ffff881387ecfc08] tty_open at ffffffff813b1ff7
#3 [ffff88145225fc08] tty_open at ffffffff813b1ff7
#3 [ffff88055235bc08] tty_open at ffffffff813b1ff7
#3 [ffff8803d2297c08] tty_open at ffffffff813b1ff7
#3 [ffff881432223c08] tty_open at ffffffff813b1ff7
#3 [ffff880d100cbc08] tty_open at ffffffff813b1ff7
#3 [ffff88018e9e3c08] tty_open at ffffffff813b1ff7
#3 [ffff8813879d7c08] tty_open at ffffffff813b1ff7
#3 [ffff88021a327c08] tty_open at ffffffff813b1ff7
#3 [ffff88021747bc08] tty_open at ffffffff813b1ff7
#3 [ffff88016bb43c08] tty_open at ffffffff813b1ff7
#3 [ffff880152223c08] tty_open at ffffffff813b1ff7
#3 [ffff8801acbcbc08] tty_open at ffffffff813b1ff7
#3 [ffff88018a2dfc08] tty_open at ffffffff813b1ff7
#3 [ffff88018821bc08] tty_open at ffffffff813b1ff7
#3 [ffff883ea5b9bc08] tty_open at ffffffff813b1ff7
#3 [ffff880242e8fc08] tty_open at ffffffff813b1ff7
#3 [ffff88136ce7fc08] tty_open at ffffffff813b1ff7
#3 [ffff880186217c08] tty_open at ffffffff813b1ff7
#3 [ffff8801685b3c08] tty_open at ffffffff813b1ff7
#3 [ffff883edb1bbc08] tty_open at ffffffff813b1ff7
#3 [ffff883efc4dfc08] tty_open at ffffffff813b1ff7
#3 [ffff8820ecaffc08] tty_open at ffffffff813b1ff7
#3 [ffff883e77557c08] tty_open at ffffffff813b1ff7
#3 [ffff8813dcbdfc08] tty_open at ffffffff813b1ff7
#3 [ffff8801544dfc08] tty_open at ffffffff813b1ff7
#3 [ffff8820d552fc08] tty_open at ffffffff813b1ff7
#3 [ffff8801dab0fc08] tty_open at ffffffff813b1ff7
#3 [ffff883fa1f83c08] tty_open at ffffffff813b1ff7

这90个中，只有一个是#6 [ffff881fd35c7c08] tty_open at ffffffff813b204c，其他都是阻塞在tty_open at ffffffff813b1ff7，根据反汇编的行号，说明89个进程在

mutex_lock(&tty_mutex);阻塞。这是一把大锁。

这89个进程阻塞的原因是1号进程拿到了tty_mutex这把大的互斥锁。

然后1号进程被阻塞在

if (tty) {
tty_lock(tty);--------------------1号进程阻塞在这，即阻塞在tty->legacy_mutex 锁。

retval = tty_reopen(tty); if (retval <0) { tty_unlock(tty); tty = ERR_PTR(retval); }

1号进程阻塞是因为5628进程，来看一下5628进程的堆栈：

#0 [ffff883edb11fbd0] __schedule at ffffffff8163df9b
#1 [ffff883edb11fc38] schedule at ffffffff8163e879
#2 [ffff883edb11fc48] schedule_timeout at ffffffff8163c329
#3 [ffff883edb11fcf8] ldsem_down_write at ffffffff8164061a
#4 [ffff883edb11fd68] tty_ldisc_lock_pair_timeout at ffffffff81640cd8
#5 [ffff883edb11fd98] tty_ldisc_hangup at ffffffff813b8dc4
#6 [ffff883edb11fdc0] __tty_hangup at ffffffff813b0594
#7 [ffff883edb11fe10] tty_ioctl at ffffffff813b2e55
#8 [ffff883edb11feb8] do_vfs_ioctl at ffffffff811f4465
#9 [ffff883edb11ff30] sys_ioctl at ffffffff811f46e1
#10 [ffff883edb11ff80] system_call_fastpath at ffffffff81649909
RIP: 00007f5438b3f537 RSP: 00007ffef141f478 RFLAGS: 00010206
RAX: 0000000000000010 RBX: ffffffff81649909 RCX: 00007f5438b39c90
RDX: 0000000000000000 RSI: 0000000000005437 RDI: 0000000000000003
RBP: 00007ffef141f4a0 R8: 00007f5438e0ce80 R9: 0000000000000000
R10: 00007ffef141f200 R11: 0000000000000206 R12: 0000000000000000
R13: 0000000000000000 R14: 00007ffef141f580 R15: 0000000000400560
ORIG_RAX: 0000000000000010 CS: 0033 SS: 002b

从下面的__tty_hangup的代码看出，调用tty_ldisc_hangup 前，因为调用了tty_lock(tty);那么确实持有了一把tty->legacy_mutex .

static void __tty_hangup(struct tty_struct *tty, int exit_session)
{
struct file *cons_filp = NULL;
struct file *filp, *f = NULL;
struct tty_file_private *priv;
int closecount = 0, n;
int refs;
if (!tty)
return;
spin_lock(&redirect_lock);
if (redirect && file_tty(redirect) == tty) {
f = redirect;
redirect = NULL;
}
spin_unlock(&redirect_lock);
tty_lock(tty);-----------------------------------------加锁
/* some functions below drop BTM, so we need this bit */
set_bit(TTY_HUPPING, &tty->flags);
/* inuse_filps is protected by the single tty lock,
this really needs to change if we want to flush the
workqueue with the lock held */
check_tty_count(tty, "tty_hangup");
spin_lock(&tty_files_lock);
/* This breaks for file handles being sent over AF_UNIX sockets ? */
list_for_each_entry(priv, &tty->tty_files, list) {
filp = priv->file;
if (filp->f_op->write == redirected_tty_write)
cons_filp = filp;
if (filp->f_op->write != tty_write)
continue;
closecount++;
__tty_fasync(-1, filp, 0); /* can't block */
filp->f_op = &hung_up_tty_fops;
}
spin_unlock(&tty_files_lock);
refs = tty_signal_session_leader(tty, exit_session);
/* Account for the p->signal references we killed */
while (refs--)
tty_kref_put(tty);
/*
* it drops BTM and thus races with reopen
* we protect the race by TTY_HUPPING
*/
tty_ldisc_hangup(tty);-----------------------阻塞，阻塞的原因上面已经描述了。
spin_lock_irq(&tty->ctrl_lock);
clear_bit(TTY_THROTTLED, &tty->flags);
clear_bit(TTY_PUSH, &tty->flags);
clear_bit(TTY_DO_WRITE_WAKEUP, &tty->flags);
put_pid(tty->session);
put_pid(tty->pgrp);
tty->session = NULL;
tty->pgrp = NULL;
tty->ctrl_status = 0;
spin_unlock_irq(&tty->ctrl_lock);
/*
* If one of the devices matches a console pointer, we
* cannot just call hangup() because that will cause
* tty->count and state->count to go out of sync.
* So we just call close() the right number of times.
*/
if (cons_filp) {
if (tty->ops->close)
for (n = 0; n )
tty->ops->close(tty, cons_filp);
} else if (tty->ops->hangup)
(tty->ops->hangup)(tty);
/*
* We don't want to have driver/ldisc interactions beyond
* the ones we did here. The driver layer expects no
* calls after ->hangup() from the ldisc side. However we
* can't yet guarantee all that.
*/
set_bit(TTY_HUPPED, &tty->flags);
clear_bit(TTY_HUPPING, &tty->flags);
tty_unlock(tty);-------------------------------导致没有走到这放锁。
if (f)
fput(f);
}

本来以为分析已经完成了，结果看了一下tty_ldisc_hangup的代码，又推翻了自己的判断。下面，我们先来看一下tty_ldisc_hangup运行到哪行代码。

crash> dis -l ffffffff813b8dc4
/usr/src/debug/kernel-3.10.0-327.22.2.el7/linux-3.10.0-327.22.2.el7.x86_64/drivers/tty/tty_ldisc.c: 690
0xffffffff813b8dc4 196>: cmpq $0x0,0x50(%rbx)

690行刚好就是tty_ldisc_lock_pair，也就是tty_ldisc_lock_pair_timeout(tty, tty2, MAX_SCHEDULE_TIMEOUT);

我们看下tty_ldisc_hangup的代码：

void tty_ldisc_hangup(struct tty_struct *tty)
{
struct tty_ldisc *ld;
int reset = tty->driver->flags & TTY_DRIVER_RESET_TERMIOS;
int err = 0;
tty_ldisc_debug(tty, "closing ldisc: %p\n", tty->ldisc);
ld = tty_ldisc_ref(tty);
if (ld != NULL) {
if (ld->ops->flush_buffer)
ld->ops->flush_buffer(tty);
tty_driver_flush_buffer(tty);
if ((test_bit(TTY_DO_WRITE_WAKEUP, &tty->flags)) &&
ld->ops->write_wakeup)
ld->ops->write_wakeup(tty);
if (ld->ops->hangup)
ld->ops->hangup(tty);
tty_ldisc_deref(ld);
}
wake_up_interruptible_poll(&tty->write_wait, POLLOUT);
wake_up_interruptible_poll(&tty->read_wait, POLLIN);
tty_unlock(tty);------------------------这里明明释放了锁
/*
* Shutdown the current line discipline, and reset it to
* N_TTY if need be.
*
* Avoid racing set_ldisc or tty_ldisc_release
*/
tty_ldisc_lock_pair(tty, tty->link);--------------------690行，也就是tty_ldisc_lock_pair_timeout(tty, tty2, MAX_SCHEDULE_TIMEOUT);跟堆栈一致。
tty_lock(tty);--------------------------重新加上锁
if (tty->ldisc) {
/* At this point we have a halted ldisc; we want to close it and
reopen a new ldisc. We could defer the reopen to the next
open but it means auditing a lot of other paths so this is
a FIXME */
if (reset == 0) {
if (!tty_ldisc_reinit(tty, tty->termios.c_line))
err = tty_ldisc_open(tty, tty->ldisc);
else
err = 1;
}
/* If the re-open fails or we reset then go to N_TTY. The
N_TTY open cannot fail */
if (reset || err) {
BUG_ON(tty_ldisc_reinit(tty, N_TTY));
WARN_ON(tty_ldisc_open(tty, tty->ldisc));
}
}
tty_ldisc_enable_pair(tty, tty->link);
if (reset)
tty_reset_termios(tty);
tty_ldisc_debug(tty, "re-opened ldisc: %p\n", tty->ldisc);
}

这说明，明明5628进程释放了tty->legacy_mutex啊，为什么1号进程的互斥锁的owner还指向它呢？这个留在下次单独对互斥信号来描述。

我们再次回到那把tty->legacy_mutex锁，

wait_list = {
next = 0xffff881fd35c7b70,
prev = 0xffff881fd35c7b70
},

list -s mutex_waiter.task 0xffff881fd35c7b70
ffff881fd35c7b70
task = 0xffff883fd1630000
ffff883f5ee2ac88
task = 0xffff880190f5c500

crash> task 0xffff883fd1630000
PID: 1 TASK: ffff883fd1630000 CPU: 1 COMMAND: "systemd"

crash> task 0xffff880190f5c500
PID: 5628 TASK: ffff880190f5c500 CPU: 47 COMMAND: "main.o"

5628怎么可能既是owner，又是waiter呢？这个问题我们放到后面来解释。

crash> struct tty_struct.link ffff883f5ee2ac00
link = 0x0

所以后面的调用链就是：tty_ldisc_lock_pair(tty, tty->link);---->tty_ldisc_lock_pair_timeout(0xffff883f5ee2ac00,0,MAX_SCHEDULE_TIMEOUT)--->tty_ldisc_lock--->ldsem_down_write

5628阻塞在线路规程的锁，也就是tty->ldisc_sem,这个是一把读写锁，在没打开debug的情况下，是没有owner成员的。

crash> struct tty_struct.ldisc_sem ffff883f5ee2ac00
ldisc_sem = {
count = -8589934591,
wait_lock = {
raw_lock = {
{
head_tail = 655370,
tickets = {
head = 10,
tail = 10
}
}
}
},
wait_readers = 1,
read_wait = {
next = 0xffff8801846d3df0,
prev = 0xffff8801846d3df0
},
write_wait = {
next = 0xffff883edb11fd10,
prev = 0xffff883edb11fd10
}
}

要找到owner，又得人肉遍历堆栈了。和《记录linux tty的一次软锁排查》一样，也是占用了锁，但本来的意愿是占用200ms超时，由于属性被修改，导致了占用无限时间。

那么，很显然，这个测试脚本，可以测试《记录linux tty的一次软锁排查》中的修改是否已经ok。

修改脚本如下：

#!/bin/bash
while [ 1 ]
do
for i in {1..64}
do
./main.o /dev/tty$i
done
done

 之前，在未修改fd为noblock的时候，是必现，改完之后，暴力测试一天都正常。

下面，针对前面所说的为什么wait里面看到的task和owner是同一个这个问题，再进行下解释。

tty_init_dev初始化一个tty的时候，调用initialize_tty_struct------>mutex_init(&tty->legacy_mutex);---->__mutex_init,代码如下

__mutex_init(struct mutex *lock, const char *name, struct lock_class_key *key)
{
atomic_set(&lock->count, 1);
spin_lock_init(&lock->wait_lock);
INIT_LIST_HEAD(&lock->wait_list);
mutex_clear_owner(lock);
#ifdef CONFIG_MUTEX_SPIN_ON_OWNER
lock->osq = NULL;
#endif
debug_mutex_init(lock, name, key);
}

此时，lock的wait_list只包含一个头结点，也就是&lock->wait_list，也就是0xffff881fd35c7b70

此时如果用list -s mutex_waiter.task 0xffff881fd35c7b70 去查看，那么对应的task是NULL。

我们来看mutex_waiter的结构：

struct mutex_waiter {
struct list_head list;
struct task_struct *task;
#ifdef CONFIG_DEBUG_MUTEXES
void *magic;
#endif
};

本来lock->wait_list把mutex_waiter 串起来，而在struct mutex结构中，owner成员刚好就位于struct list_head wait_list的后面，所以当owner获取锁之后，设置owner指针，刚好

就是和mutex_waiter 中设置task一样，所以这次看到的互斥锁的list中，使用list方法查看，会出现owner和wait指向同一个task的现象。