[PATCHSET,v2,0/4] Enable NO_OFFLOAD support

Message ID	20230420183135.119618-1-axboe@kernel.dk (mailing list archive)
Headers	show Return-Path: <io-uring-owner@vger.kernel.org> From: Jens Axboe <axboe@kernel.dk> To: io-uring@vger.kernel.org Subject: [PATCHSET v2 0/4] Enable NO_OFFLOAD support Date: Thu, 20 Apr 2023 12:31:31 -0600 Message-Id: <20230420183135.119618-1-axboe@kernel.dk> MIME-Version: 1.0 Content-Transfer-Encoding: 8bit Precedence: bulk
Series	Enable NO_OFFLOAD support \| expand [PATCHSET,v2,0/4] Enable NO_OFFLOAD support [1/4] io_uring: add support for NO_OFFLOAD [2/4] Revert "io_uring: always go async for unsupported fadvise flags" [3/4] Revert "io_uring: for requests that require async, force it" [4/4] io_uring: mark opcodes that always need io-wq punt

Message ID

20230420183135.119618-1-axboe@kernel.dk (mailing list archive)

Headers

From: Jens Axboe <axboe@kernel.dk>
To: io-uring@vger.kernel.org
Subject: [PATCHSET v2 0/4] Enable NO_OFFLOAD support
Date: Thu, 20 Apr 2023 12:31:31 -0600
Message-Id: <20230420183135.119618-1-axboe@kernel.dk>
MIME-Version: 1.0
Content-Transfer-Encoding: 8bit
Precedence: bulk

Series

Enable NO_OFFLOAD support | expand

Message

Jens Axboe April 20, 2023, 6:31 p.m. UTC

This series enables support for forcing no-offload for requests that
otherwise would have been punted to io-wq. In essence, it bypasses
the normal non-blocking issue in favor of just letting the issue block.
This is only done for requests that would've otherwise hit io-wq in
the offload path, anything pollable will still be doing non-blocking
issue. See patch 3 for details.

Git tree: https://git.kernel.dk/cgit/linux/log/?h=io_uring-sync-issue

Since v1:
- Get rid of per-io_kiocb state, only apply this logic off the direct
  and initial submission path. Add IO_URING_F_NO_OFFLOAD for that.
- Due to the above, drop the two first prep patches as they are no
  longer needed.
- Make it mutually exclusive with IORING_SETUP_IOPOLL.