qemu-devel
[Top][All Lists]
Advanced

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

[RFC PATCH 0/3] support command retry


From: Minwoo Im
Subject: [RFC PATCH 0/3] support command retry
Date: Thu, 11 Feb 2021 04:52:49 +0900

Hello,

This series is RFC about supporting command retry feature in NVMe device
model.  The background to propose this feature is that in kernel
development and testing, retry scheme has not been able to be covered in
QEMU NVMe model device.  If we are able to control the retry scheme
fromt he device side, it would be nice for kernel developers to test.

We have been putting NVME_DNR in the CQ entry status field for
all error cases.  This series added a control for the command retry
based on the 'cmd-retry-delay' parameter which is newly added.  If it's
given to positive value, Command Retry Delay Time1(CRDT1) in the
Identify Controller data structure will be set in 100msec units.
Accordingly, it will cause host to Set Feature with Host Behavior(0x16)
to enable the Advanced Command Retry Enable(ACRE) feature to support
command retry with defined delay.  If 'cmd-retry-delay' param is given
to 0, then command failures will be retried directly without delay.

This series just considered Command Interrupted status code first which
is mainly about the ACRE feature addition.  nvme_should_retry() helper
will decide command should be retried or not.

But, we don't have any use-cases specified for the Command Interrupted
status code in the device model.  So, I proposed [3/3] patch by adding
'nvme_inject_state' HMP command to make users to give pre-defined state
to the controller device by injecting it via QEMU monitor.

Usage:

  # Configure the nvme0 device to be retried every 1sec(1000msec)
  -device nvme,id=nvme0,cmd-retry-delay=1000,...

  (qemu) nvme_inject_state nvme0 cmd-interrupted
  -device nvme,id=nvme0: state cmd-interrupted injected
  (qemu)

  # Then from now on, controller will interrupt all the commands
  # to be processed with Command Interrupted status code.  Then host
  # will retry based on the delay.

Thanks,

Minwoo Im (3):
  hw/block/nvme: set NVME_DNR in a single place
  hw/block/nvme: support command retry delay
  hw/block/nvme: add nvme_inject_state HMP command

 hmp-commands.hx       |  13 ++
 hw/block/nvme.c       | 304 +++++++++++++++++++++++++++++-------------
 hw/block/nvme.h       |  10 ++
 include/block/nvme.h  |  13 +-
 include/monitor/hmp.h |   1 +
 5 files changed, 244 insertions(+), 97 deletions(-)

-- 
2.17.1




reply via email to

[Prev in Thread] Current Thread [Next in Thread]