--- /dev/null
+ NVMe Fault Injection
+ ====================
+ Linux's fault injection framework provides a systematic way to support
+ error injection via debugfs in the /sys/kernel/debug directory. When
+ enabled, the default NVME_SC_INVALID_OPCODE with no retry will be
+ injected into the nvme_end_request. Users can change the default status
+ code and no retry flag via the debugfs. The list of Generic Command
+ Status can be found in include/linux/nvme.h
+ 
+ Following examples show how to inject an error into the nvme.
+ 
+ First, enable CONFIG_FAULT_INJECTION_DEBUG_FS kernel config,
+ recompile the kernel. After booting up the kernel, do the
+ following.
+ 
+ Example 1: Inject default status code with no retry
+ ---------------------------------------------------
+ 
+ ::
+ 
+   mount /dev/nvme0n1 /mnt
+   echo 1 > /sys/kernel/debug/nvme0n1/fault_inject/times
+   echo 100 > /sys/kernel/debug/nvme0n1/fault_inject/probability
+   cp a.file /mnt
+ 
+ Expected Result::
+ 
+   cp: cannot stat ‘/mnt/a.file’: Input/output error
+ 
+ Message from dmesg::
+ 
+   FAULT_INJECTION: forcing a failure.
+   name fault_inject, interval 1, probability 100, space 0, times 1
+   CPU: 0 PID: 0 Comm: swapper/0 Not tainted 4.15.0-rc8+ #2
+   Hardware name: innotek GmbH VirtualBox/VirtualBox,
+   BIOS VirtualBox 12/01/2006
+   Call Trace:
+     <IRQ>
+     dump_stack+0x5c/0x7d
+     should_fail+0x148/0x170
+     nvme_should_fail+0x2f/0x50 [nvme_core]
+     nvme_process_cq+0xe7/0x1d0 [nvme]
+     nvme_irq+0x1e/0x40 [nvme]
+     __handle_irq_event_percpu+0x3a/0x190
+     handle_irq_event_percpu+0x30/0x70
+     handle_irq_event+0x36/0x60
+     handle_fasteoi_irq+0x78/0x120
+     handle_irq+0xa7/0x130
+     ? tick_irq_enter+0xa8/0xc0
+     do_IRQ+0x43/0xc0
+     common_interrupt+0xa2/0xa2
+     </IRQ>
+   RIP: 0010:native_safe_halt+0x2/0x10
+   RSP: 0018:ffffffff82003e90 EFLAGS: 00000246 ORIG_RAX: ffffffffffffffdd
+   RAX: ffffffff817a10c0 RBX: ffffffff82012480 RCX: 0000000000000000
+   RDX: 0000000000000000 RSI: 0000000000000000 RDI: 0000000000000000
+   RBP: 0000000000000000 R08: 000000008e38ce64 R09: 0000000000000000
+   R10: 0000000000000000 R11: 0000000000000000 R12: ffffffff82012480
+   R13: ffffffff82012480 R14: 0000000000000000 R15: 0000000000000000
+     ? __sched_text_end+0x4/0x4
+     default_idle+0x18/0xf0
+     do_idle+0x150/0x1d0
+     cpu_startup_entry+0x6f/0x80
+     start_kernel+0x4c4/0x4e4
+     ? set_init_arg+0x55/0x55
+     secondary_startup_64+0xa5/0xb0
+     print_req_error: I/O error, dev nvme0n1, sector 9240
+   EXT4-fs error (device nvme0n1): ext4_find_entry:1436:
+   inode #2: comm cp: reading directory lblock 0
+ 
+ Example 2: Inject default status code with retry
+ ------------------------------------------------
+ 
+ ::
+ 
+   mount /dev/nvme0n1 /mnt
+   echo 1 > /sys/kernel/debug/nvme0n1/fault_inject/times
+   echo 100 > /sys/kernel/debug/nvme0n1/fault_inject/probability
+   echo 1 > /sys/kernel/debug/nvme0n1/fault_inject/status
+   echo 0 > /sys/kernel/debug/nvme0n1/fault_inject/dont_retry
+ 
+   cp a.file /mnt
+ 
+ Expected Result::
+ 
+   command success without error
+ 
+ Message from dmesg::
+ 
+   FAULT_INJECTION: forcing a failure.
+   name fault_inject, interval 1, probability 100, space 0, times 1
+   CPU: 1 PID: 0 Comm: swapper/1 Not tainted 4.15.0-rc8+ #4
+   Hardware name: innotek GmbH VirtualBox/VirtualBox, BIOS VirtualBox 12/01/2006
+   Call Trace:
+     <IRQ>
+     dump_stack+0x5c/0x7d
+     should_fail+0x148/0x170
+     nvme_should_fail+0x30/0x60 [nvme_core]
+     nvme_loop_queue_response+0x84/0x110 [nvme_loop]
+     nvmet_req_complete+0x11/0x40 [nvmet]
+     nvmet_bio_done+0x28/0x40 [nvmet]
+     blk_update_request+0xb0/0x310
+     blk_mq_end_request+0x18/0x60
+     flush_smp_call_function_queue+0x3d/0xf0
+     smp_call_function_single_interrupt+0x2c/0xc0
+     call_function_single_interrupt+0xa2/0xb0
+     </IRQ>
+   RIP: 0010:native_safe_halt+0x2/0x10
+   RSP: 0018:ffffc9000068bec0 EFLAGS: 00000246 ORIG_RAX: ffffffffffffff04
+   RAX: ffffffff817a10c0 RBX: ffff88011a3c9680 RCX: 0000000000000000
+   RDX: 0000000000000000 RSI: 0000000000000000 RDI: 0000000000000000
+   RBP: 0000000000000001 R08: 000000008e38c131 R09: 0000000000000000
+   R10: 0000000000000000 R11: 0000000000000000 R12: ffff88011a3c9680
+   R13: ffff88011a3c9680 R14: 0000000000000000 R15: 0000000000000000
+     ? __sched_text_end+0x4/0x4
+     default_idle+0x18/0xf0
+     do_idle+0x150/0x1d0
+     cpu_startup_entry+0x6f/0x80
+     start_secondary+0x187/0x1e0
+     secondary_startup_64+0xa5/0xb0
++
++Example 3: Inject an error into the 10th admin command
++------------------------------------------------------
++
++::
++
++  echo 100 > /sys/kernel/debug/nvme0/fault_inject/probability
++  echo 10 > /sys/kernel/debug/nvme0/fault_inject/space
++  echo 1 > /sys/kernel/debug/nvme0/fault_inject/times
++  nvme reset /dev/nvme0
++
++Expected Result::
++
++  After NVMe controller reset, the reinitialization may or may not succeed.
++  It depends on which admin command is actually forced to fail.
++
++Message from dmesg::
++
++  nvme nvme0: resetting controller
++  FAULT_INJECTION: forcing a failure.
++  name fault_inject, interval 1, probability 100, space 1, times 1
++  CPU: 0 PID: 0 Comm: swapper/0 Not tainted 5.2.0-rc2+ #2
++  Hardware name: MSI MS-7A45/B150M MORTAR ARCTIC (MS-7A45), BIOS 1.50 04/25/2017
++  Call Trace:
++   <IRQ>
++   dump_stack+0x63/0x85
++   should_fail+0x14a/0x170
++   nvme_should_fail+0x38/0x80 [nvme_core]
++   nvme_irq+0x129/0x280 [nvme]
++   ? blk_mq_end_request+0xb3/0x120
++   __handle_irq_event_percpu+0x84/0x1a0
++   handle_irq_event_percpu+0x32/0x80
++   handle_irq_event+0x3b/0x60
++   handle_edge_irq+0x7f/0x1a0
++   handle_irq+0x20/0x30
++   do_IRQ+0x4e/0xe0
++   common_interrupt+0xf/0xf
++   </IRQ>
++  RIP: 0010:cpuidle_enter_state+0xc5/0x460
++  Code: ff e8 8f 5f 86 ff 80 7d c7 00 74 17 9c 58 0f 1f 44 00 00 f6 c4 02 0f 85 69 03 00 00 31 ff e8 62 aa 8c ff fb 66 0f 1f 44 00 00 <45> 85 ed 0f 88 37 03 00 00 4c 8b 45 d0 4c 2b 45 b8 48 ba cf f7 53
++  RSP: 0018:ffffffff88c03dd0 EFLAGS: 00000246 ORIG_RAX: ffffffffffffffdc
++  RAX: ffff9dac25a2ac80 RBX: ffffffff88d53760 RCX: 000000000000001f
++  RDX: 0000000000000000 RSI: 000000002d958403 RDI: 0000000000000000
++  RBP: ffffffff88c03e18 R08: fffffff75e35ffb7 R09: 00000a49a56c0b48
++  R10: ffffffff88c03da0 R11: 0000000000001b0c R12: ffff9dac25a34d00
++  R13: 0000000000000006 R14: 0000000000000006 R15: ffffffff88d53760
++   cpuidle_enter+0x2e/0x40
++   call_cpuidle+0x23/0x40
++   do_idle+0x201/0x280
++   cpu_startup_entry+0x1d/0x20
++   rest_init+0xaa/0xb0
++   arch_call_rest_init+0xe/0x1b
++   start_kernel+0x51c/0x53b
++   x86_64_start_reservations+0x24/0x26
++   x86_64_start_kernel+0x74/0x77
++   secondary_startup_64+0xa4/0xb0
++  nvme nvme0: Could not set queue count (16385)
++  nvme nvme0: IO queues not created