public inbox for [email protected]
 help / color / mirror / Atom feed
From: Jens Axboe <[email protected]>
To: Pavel Begunkov <[email protected]>, [email protected]
Cc: [email protected], [email protected],
	[email protected], [email protected]
Subject: Re: [PATCH 15/15] io_uring: support true async buffered reads, if file provides it
Date: Tue, 23 Jun 2020 08:38:30 -0600	[thread overview]
Message-ID: <[email protected]> (raw)
In-Reply-To: <[email protected]>

On 6/23/20 6:39 AM, Pavel Begunkov wrote:
> On 18/06/2020 17:43, Jens Axboe wrote:
>> If the file is flagged with FMODE_BUF_RASYNC, then we don't have to punt
>> the buffered read to an io-wq worker. Instead we can rely on page
>> unlocking callbacks to support retry based async IO. This is a lot more
>> efficient than doing async thread offload.
>>
>> The retry is done similarly to how we handle poll based retry. From
>> the unlock callback, we simply queue the retry to a task_work based
>> handler.
>>
>> Signed-off-by: Jens Axboe <[email protected]>
>> ---
>>  fs/io_uring.c | 145 +++++++++++++++++++++++++++++++++++++++++++++++---
>>  1 file changed, 137 insertions(+), 8 deletions(-)
>>
> ...
>>  static int io_read(struct io_kiocb *req, bool force_nonblock)
>>  {
>>  	struct iovec inline_vecs[UIO_FASTIOV], *iovec = inline_vecs;
>> @@ -2784,10 +2907,7 @@ static int io_read(struct io_kiocb *req, bool force_nonblock)
>>  		unsigned long nr_segs = iter.nr_segs;
>>  		ssize_t ret2 = 0;
>>  
>> -		if (req->file->f_op->read_iter)
>> -			ret2 = call_read_iter(req->file, kiocb, &iter);
>> -		else
>> -			ret2 = loop_rw_iter(READ, req->file, kiocb, &iter);
>> +		ret2 = io_iter_do_read(req, &iter);
>>  
>>  		/* Catch -EAGAIN return for forced non-blocking submission */
>>  		if (!force_nonblock || (ret2 != -EAGAIN && ret2 != -EIO)) {
>> @@ -2799,17 +2919,26 @@ static int io_read(struct io_kiocb *req, bool force_nonblock)
>>  			ret = io_setup_async_rw(req, io_size, iovec,
>>  						inline_vecs, &iter);
>>  			if (ret)
>> -				goto out_free;
>> +				goto out;
>>  			/* any defer here is final, must blocking retry */
>>  			if (!(req->flags & REQ_F_NOWAIT) &&
>>  			    !file_can_poll(req->file))
>>  				req->flags |= REQ_F_MUST_PUNT;
>> +			/* if we can retry, do so with the callbacks armed */
>> +			if (io_rw_should_retry(req)) {
>> +				ret2 = io_iter_do_read(req, &iter);
>> +				if (ret2 == -EIOCBQUEUED) {
>> +					goto out;
>> +				} else if (ret2 != -EAGAIN) {
>> +					kiocb_done(kiocb, ret2);
>> +					goto out;
>> +				}
>> +			}
>> +			kiocb->ki_flags &= ~IOCB_WAITQ;
>>  			return -EAGAIN;
>>  		}
>>  	}
>> -out_free:
>> -	kfree(iovec);
>> -	req->flags &= ~REQ_F_NEED_CLEANUP;
> 
> This looks fishy. For instance, if it fails early on rw_verify_area(), how would
> it free yet on-stack iovec? Is it handled somehow?

This was tweaked and rebased on top of the REQ_F_NEED_CLEANUP change,
it should be correct in the tree:

https://git.kernel.dk/cgit/linux-block/tree/fs/io_uring.c?h=for-5.9/io_uring#n2908

-- 
Jens Axboe


  reply	other threads:[~2020-06-23 14:38 UTC|newest]

Thread overview: 38+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2020-06-18 14:43 Jens Axboe
2020-06-18 14:43 ` [PATCH 01/15] block: provide plug based way of signaling forced no-wait semantics Jens Axboe
2020-06-18 14:43 ` [PATCH 02/15] io_uring: always plug for any number of IOs Jens Axboe
2020-06-18 14:43 ` [PATCH 03/15] io_uring: catch -EIO from buffered issue request failure Jens Axboe
2020-06-18 14:43 ` [PATCH 04/15] io_uring: re-issue block requests that failed because of resources Jens Axboe
2020-06-19 14:12   ` Pavel Begunkov
2020-06-19 14:22     ` Jens Axboe
2020-06-19 14:30       ` Pavel Begunkov
2020-06-19 14:36         ` Jens Axboe
2020-06-18 14:43 ` [PATCH 05/15] mm: allow read-ahead with IOCB_NOWAIT set Jens Axboe
2020-06-24  1:02   ` Dave Chinner
2020-06-24  1:46     ` Matthew Wilcox
2020-06-24 15:00       ` Jens Axboe
2020-06-24 15:35         ` Jens Axboe
2020-06-24 16:41           ` Matthew Wilcox
2020-06-24 16:44             ` Jens Axboe
2020-07-07 11:38               ` Andreas Grünbacher
2020-07-07 14:31                 ` Jens Axboe
2020-08-10 22:56               ` Dave Chinner
2020-08-10 23:03                 ` Jens Axboe
2020-06-24  4:38   ` Dave Chinner
2020-06-24 15:01     ` Jens Axboe
2020-06-18 14:43 ` [PATCH 06/15] mm: abstract out wake_page_match() from wake_page_function() Jens Axboe
2020-06-18 14:43 ` [PATCH 07/15] mm: add support for async page locking Jens Axboe
2020-07-07 11:32   ` Andreas Grünbacher
2020-07-07 14:32     ` Jens Axboe
2020-06-18 14:43 ` [PATCH 08/15] mm: support async buffered reads in generic_file_buffered_read() Jens Axboe
2020-06-18 14:43 ` [PATCH 09/15] fs: add FMODE_BUF_RASYNC Jens Axboe
2020-06-18 14:43 ` [PATCH 10/15] block: flag block devices as supporting IOCB_WAITQ Jens Axboe
2020-06-18 14:43 ` [PATCH 11/15] xfs: flag files as supporting buffered async reads Jens Axboe
2020-06-18 14:43 ` [PATCH 12/15] btrfs: " Jens Axboe
2020-06-19 11:11   ` David Sterba
2020-06-18 14:43 ` [PATCH 13/15] ext4: flag " Jens Axboe
2020-06-18 14:43 ` [PATCH 14/15] mm: add kiocb_wait_page_queue_init() helper Jens Axboe
2020-06-18 14:43 ` [PATCH 15/15] io_uring: support true async buffered reads, if file provides it Jens Axboe
2020-06-23 12:39   ` Pavel Begunkov
2020-06-23 14:38     ` Jens Axboe [this message]
2020-06-18 14:45 ` [PATCHSET v7 0/12] Add support for async buffered reads Jens Axboe

Reply instructions:

You may reply publicly to this message via plain-text email
using any one of the following methods:

* Save the following mbox file, import it into your mail client,
  and reply-to-all from there: mbox

  Avoid top-posting and favor interleaved quoting:
  https://en.wikipedia.org/wiki/Posting_style#Interleaved_style

* Reply using the --to, --cc, and --in-reply-to
  switches of git-send-email(1):

  git send-email \
    [email protected] \
    [email protected] \
    [email protected] \
    [email protected] \
    [email protected] \
    [email protected] \
    [email protected] \
    [email protected] \
    /path/to/YOUR_REPLY

  https://kernel.org/pub/software/scm/git/docs/git-send-email.html

* If your mail client supports setting the In-Reply-To header
  via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox