> On May 10, 2022, at 8:34 PM, Shinichiro Kawasaki <shinichiro.kawasaki@xxxxxxx> wrote:
>
> On May 10, 2022 / 16:48, Sagi Grimberg wrote:
>>
>>>> On 5/10/22 19:43, Alan Adamson wrote:
>>>>> Test nvme error logging by injecting errors. Kernel must have FAULT_INJECTION
>>>>> and FAULT_INJECTION_DEBUG_FS configured to use error injector. Tests can be
>>>>> run with or without NVME_VERBOSE_ERRORS configured.
>>>>>
>>>>> These test verify the functionality delivered by the follow:
>>>>> commit bd83fe6f2cd2 ("nvme: add verbose error logging")
>>>>>
>>>>> Signed-off-by: Alan Adamson <alan.adamson@xxxxxxxxxx>
>>>>> Signed-off-by: Chaitanya Kulkarni <kch@xxxxxxxxxx>
>>>>> ---
>>>>> tests/nvme/039 | 185 +++++++++++++++++++++++++++++++++++++++++++++
>>>>> tests/nvme/039.out | 7 ++
>>>>> 2 files changed, 192 insertions(+)
>>>>> create mode 100755 tests/nvme/039
>>>>> create mode 100644 tests/nvme/039.out
>>>>>
>>>>> diff --git a/tests/nvme/039 b/tests/nvme/039
>>>>> new file mode 100755
>>>>> index 000000000000..e6d45a6e3fe5
>>>>> --- /dev/null
>>>>> +++ b/tests/nvme/039
>>>>> @@ -0,0 +1,185 @@
>>>>> +#!/bin/bash
>>>>> +# SPDX-License-Identifier: GPL-3.0+
>>>>> +# Copyright (C) 2022 Oracle and/or its affiliates
>>>>> +#
>>>>> +# Test nvme error logging by injecting errors. Kernel must have FAULT_INJECTION
>>>>> +# and FAULT_INJECTION_DEBUG_FS configured to use error injector. Tests can be
>>>>> +# run with or without NVME_VERBOSE_ERRORS configured.
>>>>> +#
>>>>> +# Test for commit bd83fe6f2cd2 ("nvme: add verbose error logging").
>>>>> +
>>>>> +. tests/nvme/rc
>>>>> +DESCRIPTION="test error logging"
>>>>> +QUICK=1
>>>>> +
>>>>> +requires() {
>>>>> + _nvme_requires
>>>>> + _have_kernel_option FAULT_INJECTION && \
>>>>> + _have_kernel_option FAULT_INJECTION_DEBUG_FS
>>>>> +}
>>>>> +
>>>>> +declare -A NS_DEV_FAULT_INJECT_SAVE
>>>>> +declare -A CTRL_DEV_FAULT_INJECT_SAVE
>>>>> +
>>>>> +save_err_inject_attr()
>>>>> +{
>>>>> + local a
>>>>> +
>>>>> + for a in /sys/kernel/debug/"${ns_dev}"/fault_inject/*; do
>>>>> + NS_DEV_FAULT_INJECT_SAVE[${a}]=$(<"${a}")
>>>>> + done
>>>>> + for a in /sys/kernel/debug/"${ctrl_dev}"/fault_inject/*; do
>>>>> + CTRL_DEV_FAULT_INJECT_SAVE[${a}]=$(<"${a}")
>>>>> + done
>>>>> +}
>>>>> +
>>>>> +restore_err_inject_attr()
>>>>> +{
>>>>> + local a
>>>>> +
>>>>> + for a in /sys/kernel/debug/"${ns_dev}"/fault_inject/*; do
>>>>> + echo "${NS_DEV_FAULT_INJECT_SAVE[${a}]}" > "${a}"
>>>>> + done
>>>>> + for a in /sys/kernel/debug/"${ctrl_dev}"/fault_inject/*; do
>>>>> + echo "${CTRL_DEV_FAULT_INJECT_SAVE[${a}]}" > "${a}"
>>>>> + done
>>>>> +}
>>>>> +
>>>>> +set_verbose_prob_retry()
>>>>> +{
>>>>> + echo 0 > /sys/kernel/debug/"$1"/fault_inject/verbose
>>>>> + echo 100 > /sys/kernel/debug/"$1"/fault_inject/probability
>>>>> + echo 1 > /sys/kernel/debug/"$1"/fault_inject/dont_retry
>>>>> +}
>>>>> +
>>>>> +set_status_time()
>>>>> +{
>>>>> + echo "$1" > /sys/kernel/debug/"$3"/fault_inject/status
>>>>> + echo "$2" > /sys/kernel/debug/"$3"/fault_inject/times
>>>>> +}
>>>>> +
>>>>> +inject_unrec_read_err()
>>>>> +{
>>>>> + # Inject a 'Unrecovered Read Error' error on a READ
>>>>> + set_status_time 0x281 1 "$1"
>>>>> +
>>>>> + dd if=/dev/"$1" of=/dev/null bs=512 count=1 iflag=direct \
>>>>> + 2> /dev/null 1>&2
>>>>> +
>>>>> + if ${nvme_verbose_errors}; then
>>>>> + dmesg -t | tail -2 | grep "Unrecovered Read Error (" | \
>>>>> + sed 's/nvme.*://g'
>>>>> + else
>>>>> + dmesg -t | tail -2 | grep "Cmd(" | sed 's/I\/O Cmd/Read/g' | \
>>>>> + sed 's/I\/O Error/Unrecovered Read Error/g' | \
>>>>> + sed 's/nvme.*://g'
>>>>> + fi
>>>>> +}
>>>>> +
>>>>> +inject_invalid_read_err()
>>>>> +{
>>>>> + # Inject a valid invalid error status (0x375) on a READ
>>>>> + set_status_time 0x375 1 "$1"
>>>>> +
>>>>> + dd if=/dev/"$1" of=/dev/null bs=512 count=1 iflag=direct \
>>>>> + 2> /dev/null 1>&2
>>>>> +
>>>>> + if ${nvme_verbose_errors}; then
>>>>> + dmesg -t | tail -2 | grep "Unknown (" | \
>>>>> + sed 's/nvme.*://g'
>>>>> + else
>>>>> + dmesg -t | tail -2 | grep "Cmd(" | sed 's/I\/O Cmd/Read/g' | \
>>>>> + sed 's/I\/O Error/Unknown/g' | \
>>>>> + sed 's/nvme.*://g'
>>>>> + fi
>>>>> +}
>>>>> +
>>>>> +inject_write_fault()
>>>>> +{
>>>>> + # Inject a 'Write Fault' error on a WRITE
>>>>> + set_status_time 0x280 1 "$1"
>>>>> +
>>>>> + dd if=/dev/zero of=/dev/"$1" bs=512 count=1 oflag=direct \
>>>>> + 2> /dev/null 1>&2
>>>>> +
>>>>> + if ${nvme_verbose_errors}; then
>>>>> + dmesg -t | tail -2 | grep "Write Fault (" | \
>>>>> + sed 's/nvme.*://g'
>>>>> + else
>>>>> + dmesg -t | tail -2 | grep "Cmd(" | sed 's/I\/O Cmd/Write/g' | \
>>>>> + sed 's/I\/O Error/Write Fault/g' | \
>>>>> + sed 's/nvme.*://g'
>>>>> + fi
>>>>> +}
>>>>> +
>>>>> +inject_id_admin()
>>>>> +{
>>>>> + # Inject a valid (Identify) Admin command
>>>>> + set_status_time 0x286 1000 "$1"
>>>>> +
>>>>> + nvme admin-passthru /dev/"$1" --opcode=0x06 --data-len=4096 \
>>>>> + --cdw10=1 -r 2> /dev/null 1>&2
>>>>> +
>>>>> + if ${nvme_verbose_errors}; then
>>>>> + dmesg -t | tail -1 | grep "Access Denied (" | \
>>>>> + sed 's/nvme.*://g'
>>>>> + else
>>>>> + dmesg -t | tail -1 | grep "Admin Cmd(" | \
>>>>> + sed 's/Admin Cmd/Identify/g' | \
>>>>> + sed 's/I\/O Error/Access Denied/g' | \
>>>>> + sed 's/nvme.*://g'
>>>>> + fi
>>>>> +}
>>>>> +
>>>>> +inject_invalid_cmd()
>>>>> +{
>>>>> + # Inject an invalid command (0x96)
>>>>> + set_status_time 0x1 1 "$1"
>>>>> +
>>>>> + nvme admin-passthru /dev/"$1" --opcode=0x96 --data-len=4096 \
>>>>> + --cdw10=1 -r 2> /dev/null 1>&2
>>>>> + if ${nvme_verbose_errors}; then
>>>>> + dmesg -t | tail -1 | grep "Invalid Command Opcode (" | \
>>>>> + sed 's/nvme.*://g'
>>>>> + else
>>>>> + dmesg -t | tail -1 | grep "Admin Cmd(" | \
>>>>> + sed 's/Admin Cmd/Unknown/g' | \
>>>>> + sed 's/I\/O Error/Invalid Command Opcode/g' | \
>>>>> + sed 's/nvme.*://g'
>>>>> + fi
>>>>> +}
>>>>> +
>>>>
>>>> All of the above seems like they belong in common code...
>>>
>>> So far, this nvme/039 is the only one user of the helper functions above. Do you
>>> foresee that future new test cases in nvmeof-mp or srp group will use them?
>>>
>>
>> I can absolutely see other tests inject errors. I meant that this code
>> should live in nvme/rc
Should the helpers inject errors itself (_nvme_inject_write_fault), or just provide the functions to setup the
error injector (_nvme_set_inject) and let the test specify the error status and do the IO that causes the injected
error?
Alan
>
> Thanks for clarification. So we expect two patches: one to add the helper
> functions to nvme/rc, and the other to add this test case.
>
> The inject_*() functions refer nvme_verbose_errors for dmesg filter, which is
> defined in test_device(). That part needs a bit of rework to move to nvme/rc.
>
> --
> Best Regards,
> Shin'ichiro Kawasaki
