[PATCH v2] riscv: fix oops caused by irq on/off tracer

[Changbin Du]

On Thu, Feb 10, 2022 at 01:32:59AM +0800, Jisheng Zhang wrote:
[snip]
> Hi Changbin,
> 
> I read the code and find that current riscv frame records during
> exception isn't as completed as other architectures. riscv only
> records frames from the ret_from_exception(). If we add completed
What do you mean for 'record'?

> frame records as other arch do, then the issue you saw can also
> be fixed at the same time.
> 
I don't think so. The problem is __builtin_return_address(1) trigger page fault
here.

> However, I'm not sure what's the best choice now.
> 
> A simple demo to this incomplete frames:
> add dump_stack() in any ISR, then
> 
> in riscv:
> [    2.961294] Call Trace:
> [    2.961460] [<ffffffff8000485e>] dump_backtrace+0x1c/0x24
> [    2.961823] [<ffffffff805ed980>] show_stack+0x2c/0x38
> [    2.962153] [<ffffffff805f292e>] dump_stack_lvl+0x40/0x58
> [    2.962483] [<ffffffff805f295a>] dump_stack+0x14/0x1c
> [    2.962792] [<ffffffff805f31a0>] serial8250_interrupt+0x20/0x82
> [    2.963139] [<ffffffff80053032>] __handle_irq_event_percpu+0x4c/0x106
> [    2.963526] [<ffffffff80053170>] handle_irq_event+0x38/0x80
> [    2.963856] [<ffffffff80056a32>] handle_fasteoi_irq+0x96/0x188
> [    2.964198] [<ffffffff800526ce>] generic_handle_domain_irq+0x28/0x3a
> [    2.964567] [<ffffffff802f0ae4>] plic_handle_irq+0x88/0xec
> [    2.964896] [<ffffffff800526ce>] generic_handle_domain_irq+0x28/0x3a
> [    2.965264] [<ffffffff802f08e4>] riscv_intc_irq+0x34/0x5c
> [    2.965584] [<ffffffff805f6dc8>] generic_handle_arch_irq+0x4a/0x74
> [    2.966068] [<ffffffff80002fe8>] ret_from_exception+0x0/0xc
> 
> in x86:
> [    1.191274] Call Trace:
> [    1.192223]  <IRQ>
> [    1.192758]  dump_stack_lvl+0x45/0x59
> [    1.192982]  serial8250_interrupt+0x24/0x88
> [    1.193105]  __handle_irq_event_percpu+0x66/0x1b0
> [    1.193239]  handle_irq_event+0x34/0x70
> [    1.193345]  handle_edge_irq+0x85/0x1e0
> [    1.193455]  __common_interrupt+0x38/0x90
> [    1.193573]  common_interrupt+0x73/0x90
> [    1.193809]  </IRQ>
> [    1.193889]  <TASK>
> [    1.193956]  asm_common_interrupt+0x1b/0x40
> [    1.194318] RIP: 0010:_raw_spin_unlock_irqrestore+0x1b/0x40
> [    1.194566] Code: 24 be 01 02 00 00 e9 54 20 bf ff 0f 1f 40 00 0f 1f
> 44 00 00 f7 c6 00f
> [    1.195137] RSP: 0000:ffff888000243b68 EFLAGS: 00000246
> [    1.195314] RAX: 0000000000000000 RBX: ffffffff82025840 RCX:
> 0000000000000000
> [    1.195482] RDX: 0000000000000001 RSI: 0000000000000000 RDI:
> 0000000000000001
> [    1.195645] RBP: 0000000000000202 R08: ffffffffffffffff R09:
> 0000000000000000
> [    1.195808] R10: 00000000000000eb R11: 0000000000000000 R12:
> 0000000000000000
> [    1.195972] R13: 0000000000000040 R14: 0000000000000000 R15:
> ffff888000c39000
> [    1.196245]  ? _raw_spin_unlock_irqrestore+0x15/0x40
> [    1.196373]  serial8250_do_startup+0x42d/0x600
> [    1.196502]  uart_port_startup+0x11b/0x270
> [    1.196619]  uart_port_activate+0x3f/0x60
> [    1.196729]  tty_port_open+0x7e/0xd0
> [    1.196835]  ? _raw_spin_unlock+0x12/0x30
> [    1.196942]  uart_open+0x1a/0x30
> [    1.197036]  tty_open+0x153/0x7c0
> [    1.197144]  chrdev_open+0xbf/0x230
> [    1.197253]  ? cdev_device_add+0x90/0x90
> [    1.197359]  do_dentry_open+0x13c/0x360
> [    1.197470]  path_openat+0xb0c/0xe00
> [    1.197577]  ? update_load_avg+0x5f/0x640
> [    1.197691]  ? finish_task_switch.isra.0+0xac/0x240
> [    1.197821]  do_filp_open+0xb2/0x150
> [    1.197935]  ? preempt_schedule_thunk+0x16/0x18
> [    1.198049]  ? preempt_schedule_common+0x90/0xd0
> [    1.198167]  ? preempt_schedule_thunk+0x16/0x18
> [    1.198291]  file_open_name+0xf1/0x1b0
> [    1.198397]  filp_open+0x2c/0x50
> [    1.198495]  console_on_rootfs+0x19/0x52
> [    1.198648]  kernel_init_freeable+0x19a/0x1c7
> [    1.198765]  ? rest_init+0xc0/0xc0
> [    1.198867]  kernel_init+0x16/0x110
> [    1.198965]  ret_from_fork+0x1f/0x30
> [    1.199131]  </TASK>
> 
As I said before, this issue is not related to stackdump.

Besides, you can see more calltrace on x86 that because x86 iterate all stacks
(kernel, irq or exception) when dumping stacktrace. While RISCV only show
calltrace of current stack.


-- 
Cheers,
Changbin Du