[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]
[RFC PATCH 00/11] hw/nvme: reimplement all multi-aio commands with custo
From: |
Klaus Jensen |
Subject: |
[RFC PATCH 00/11] hw/nvme: reimplement all multi-aio commands with custom aiocbs |
Date: |
Fri, 4 Jun 2021 08:52:26 +0200 |
From: Klaus Jensen <k.jensen@samsung.com>
This series reimplements flush, dsm, copy, zone reset and format nvm to
allow cancellation. I posted an RFC back in March ("hw/block/nvme:
convert ad-hoc aio tracking to aiocb") and I've applied some feedback
from Stefan and reimplemented the remaining commands.
The basic idea is to define custom AIOCBs for these commands. The custom
AIOCB takes care of issuing all the "nested" AIOs one by one instead of
blindly sending them off simultaneously without tracking the returned
aiocbs.
I've kept the RFC since I'm still new to using the block layer like
this. I was hoping that Stefan could find some time to look over this -
this is a huge series, so I don't expect non-nvme folks to spend a large
amount of time on it, but I would really like feedback on my approach in
the reimplementation of flush and format. Those commands are special in
that may issue AIOs to multiple namespaces and thus, to multiple block
backends. Since this device does not support iothreads, I've opted for
simply always returning the main loop aio context, but I wonder if this
is acceptable or not. It might be the case that this should contain an
assert of some kind, in case someone starts adding iothread support.
Klaus Jensen (11):
hw/nvme: reimplement flush to allow cancellation
hw/nvme: add nvme_block_status_all helper
hw/nvme: reimplement dsm to allow cancellation
hw/nvme: save reftag when generating pi
hw/nvme: remove assert from nvme_get_zone_by_slba
hw/nvme: use prinfo directly in nvme_check_prinfo and nvme_dif_check
hw/nvme: add dw0/1 to the req completion trace event
hw/nvme: reimplement the copy command to allow aio cancellation
hw/nvme: reimplement zone reset to allow cancellation
hw/nvme: reimplement format nvm to allow cancellation
Partially revert "hw/block/nvme: drain namespaces on sq deletion"
hw/nvme/nvme.h | 10 +-
include/block/nvme.h | 8 +
hw/nvme/ctrl.c | 1861 ++++++++++++++++++++++++------------------
hw/nvme/dif.c | 64 +-
hw/nvme/trace-events | 21 +-
5 files changed, 1102 insertions(+), 862 deletions(-)
--
2.31.1
- [RFC PATCH 00/11] hw/nvme: reimplement all multi-aio commands with custom aiocbs,
Klaus Jensen <=
- [RFC PATCH 01/11] hw/nvme: reimplement flush to allow cancellation, Klaus Jensen, 2021/06/04
- [RFC PATCH 02/11] hw/nvme: add nvme_block_status_all helper, Klaus Jensen, 2021/06/04
- [RFC PATCH 03/11] hw/nvme: reimplement dsm to allow cancellation, Klaus Jensen, 2021/06/04
- [RFC PATCH 04/11] hw/nvme: save reftag when generating pi, Klaus Jensen, 2021/06/04
- [RFC PATCH 05/11] hw/nvme: remove assert from nvme_get_zone_by_slba, Klaus Jensen, 2021/06/04
- [RFC PATCH 06/11] hw/nvme: use prinfo directly in nvme_check_prinfo and nvme_dif_check, Klaus Jensen, 2021/06/04
- [RFC PATCH 07/11] hw/nvme: add dw0/1 to the req completion trace event, Klaus Jensen, 2021/06/04
- [RFC PATCH 08/11] hw/nvme: reimplement the copy command to allow aio cancellation, Klaus Jensen, 2021/06/04
- [RFC PATCH 09/11] hw/nvme: reimplement zone reset to allow cancellation, Klaus Jensen, 2021/06/04
- [RFC PATCH 10/11] hw/nvme: reimplement format nvm to allow cancellation, Klaus Jensen, 2021/06/04