Re: [PATCH][v2] rtnetlink: add rtnl_lock debug log

On Sat, May 8, 2021 at 12:11 PM Rocco Yue [off-list ref] wrote:

We often encounter system hangs caused by certain process
holding rtnl_lock for a long time. Even if there is a lock
detection mechanism in Linux, it is a bit troublesome and
affects the system performance. We hope to add a lightweight
debugging mechanism for detecting rtnl_lock.

Up to now, we have discovered and solved some potential bugs
through this lightweight rtnl_lock debugging mechanism, which
is helpful for us.

When you say Y for RTNL_LOCK_DEBUG, then the kernel will detect
if any function hold rtnl_lock too long and some key information
will be printed out to help locate the problem.

i.e: from the following logs, we can clearly know that the pid=2206
RfxSender_4 process holds rtnl_lock for a long time, causing the
system to hang. And we can also speculate that the delay operation
may be performed in devinet_ioctl(), resulting in rtnl_lock was
not released in time.

<6>[   40.191481][    C6] rtnetlink: -- rtnl_print_btrace start --

You don't seem to get it. It's a quite long trace for the commit
message. Do you need all those lines below? Why?

<6>[   40.191494][    C6] rtnetlink: RfxSender_4[2206][R] hold rtnl_lock
more than 2 sec, start time: 38181400013
<4>[   40.191510][    C6]  devinet_ioctl+0x1fc/0x75c
<4>[   40.191517][    C6]  inet_ioctl+0xb8/0x1f8
<4>[   40.191527][    C6]  sock_do_ioctl+0x70/0x2ac
<4>[   40.191533][    C6]  sock_ioctl+0x5dc/0xa74
<4>[   40.191541][    C6]  __arm64_sys_ioctl+0x178/0x1fc
<4>[   40.191548][    C6]  el0_svc_common+0xc0/0x24c
<4>[   40.191555][    C6]  el0_svc+0x28/0x88
<4>[   40.191560][    C6]  el0_sync_handler+0x8c/0xf0
<4>[   40.191566][    C6]  el0_sync+0x198/0x1c0
<6>[   40.191571][    C6] Call trace:
<6>[   40.191586][    C6]  rtnl_print_btrace+0xf0/0x124
<6>[   40.191595][    C6]  call_timer_fn+0x5c/0x3b4
<6>[   40.191602][    C6]  expire_timers+0xe0/0x49c
<6>[   40.191609][    C6]  __run_timers+0x34c/0x48c
<6>[   40.191616][    C6]  run_timer_softirq+0x28/0x58
<6>[   40.191621][    C6]  efi_header_end+0x168/0x690
<6>[   40.191628][    C6]  __irq_exit_rcu+0x108/0x124
<6>[   40.191635][    C6]  __handle_domain_irq+0x130/0x1b4
<6>[   40.191643][    C6]  gic_handle_irq.29882+0x6c/0x2d8
<6>[   40.191648][    C6]  el1_irq+0xdc/0x1c0
<6>[   40.191656][    C6]  __delay+0xc0/0x180
<6>[   40.191663][    C6]  devinet_ioctl+0x21c/0x75c
<6>[   40.191668][    C6]  inet_ioctl+0xb8/0x1f8
<6>[   40.191675][    C6]  sock_do_ioctl+0x70/0x2ac
<6>[   40.191682][    C6]  sock_ioctl+0x5dc/0xa74
<6>[   40.191688][    C6]  __arm64_sys_ioctl+0x178/0x1fc
<6>[   40.191694][    C6]  el0_svc_common+0xc0/0x24c
<6>[   40.191699][    C6]  el0_svc+0x28/0x88
<6>[   40.191705][    C6]  el0_sync_handler+0x8c/0xf0
<6>[   40.191710][    C6]  el0_sync+0x198/0x1c0
<6>[   40.191715][    C6] rtnetlink: -- rtnl_print_btrace end --

<6>[   42.181879][ T2206] rtnetlink: rtnl_lock is held by [2206] from
[38181400013] to [42181875177]

-- 
With Best Regards,
Andy Shevchenko