Acpi is used to find hardware so you can not disable it. My custom kernel 4.14.0-rc2 from kernel.org works fine with the Ryzen 5 1600 and the Asus Prime B350M-K motherboard. To prevent random kernel lock ups, enable RCU_NOCB_CPU and boot the kernel with the rcu_nocbs=0-X command line parameter. X is the cpu thread count -1.

Bug is in your distribution kernel, so remove this bug and make your own kernel from the latest code from kernel.org.

I compile kernel 4.14.0-rc4 from kernel.org. Bug the bug exits. this is the Result:

page_fault +0x22/0x30
RIP: 0010:cfq_completed_request +0x48/0xa00
RSP: 0018:ffff88011ef43d78 EFLAGS: 00010002
RAX: 0000000177dbd2ea RBX: 0000000000000000 RCX: 0000000000000018
RDX: 00002e421dab6dc8 RSI: 001e427deca0eb7d RDI: ffffffff81c28380
RBP: ffff88011ef43dc0 R08: 0000000000000000 R09: ffffffff81186527
R10: 0000000000000000 R11: 0000000000000001 R12: 0000000000027003
R13: ffff880117515400 R14: 0000000177dbd2ea R15: ffff8801171193b0
  ? mempool_free_slab +0x17/0x20
  ? cfq_completed_request +0x45/0xa00
  ? mempool_free_slab +0x17/0x20
  ? mempool_free +0x29/0x80
  elv_completed_request +0x51/0x60
  __blk_put_request +0x51/0x190
  blk_finish_request +0x106/0x110 
  scsi_end_request +0x106/0x200 [scsi_mod]
  scsi_io_completion +0x22b/0x610 [scsi_mod]
  ? rebalance_domains +0xf0/0x2b0
  scsi_finish_command +0xd3/0xf0 [scsi_mod]
  scsi_softirq_done +0x127/0x130 [scsi_mod]
  blk_done_softirq +0x83/0xc0
  __do_softirq +0xcd/0x2f4
  irq_exit +0xb6/0xc0
  smp_call_function_single_interrupt +0x3e/0xf0
  call_function_single_interrupt +0x93/0xa0
  </IRQ>
  RIP: 0010:native_safe_halt +0x6/0x10
  RSP: 0018:ffffc9000079fdf0 EFLAGS: 000046 ORIG_RAX: fffffffffffff04
  RAX: 0000000000000009 RBX: ffff8801194fbc00 RCX: 000000000000001f
  RDX: 4ec4ec4ec4ec4ec5 RSI: 0000000000000034 RDI: ffff8801194fbc64
  RBP: ffffc9000079fdf0 R08: 00000030eee5ac8a R09: 0000000000000008
  R10: ffffc9000079fe40 R11: 0000000000000cf9 R12: ffff8801194fbc64
  R13: 0000000000000001 R14: 0000000000000001 R15: 0000000000000001
   acpi_safe_halt.part .0+0xe/0x20
   acpi_idle_do_entry +0x38/0x50
   acpi_idle_enter +0x114/0x2e0
   ? sched_clock +0x9/0x10
   cpuidle_enter_state +0xfd/0x2f0
   cpuidle_enter +0x17/0x20
   call_cpuidel +0x23/0x40
   do_idle +0x181/0x1d0
   cpu_startup_entry +0x73/0x80
   start_secondary +0x179/0x1c0
   secondary_startup_64 +0xa5/0xa5
 Code: 0f a3 05 ad bc cb 00 73 14 48 8b 05 f4 ea 95 00 be fd 00 00 00 ff
     .........
---[ end trace 3396a394afcaa20d ]----

Use the deadline I/O scheduler, especially for a ssd.

(In reply to fin4478 from comment #3)
> Use the deadline I/O scheduler, especially for a ssd.

How to use the deadline I/O scheduler? I compile the lastest kernel, and why ssd  works well with apci=off parameters.

I search google, Should I add parameters elevator=deadline? I'll try this after work. Thanks

(In reply to ldk from comment #4)
> (In reply to fin4478 from comment #3)
> > Use the deadline I/O scheduler, especially for a ssd.
> 
> How to use the deadline I/O scheduler? I compile the lastest kernel, and why
> ssd  works well with apci=off parameters.

Configure kernel with the make xconfig command. There is a search feature and you can find where is deadline. Set it is to default I/O scheduler.

I found the reason why acpi failed. The bios's immou default value was auto. diabled it, everything is fine. I'll search some article about immou and configuation it better. 

Thank you fin4478@hotmail.com.


Now this bug can be closed. Maybe kernel should take into account this situation. At least should give this error reason directly, Thanks again.

Sorry, The world "immou" is "IOMMU".