Flexible I/O Tester development
 help / color / mirror / Atom feed
From: Alexey Dobriyan <adobriyan@gmail.com>
To: Jens Axboe <axboe@kernel.dk>
Cc: fio@vger.kernel.org
Subject: Re: [PATCH] fio: add NVMe engine
Date: Fri, 27 Mar 2020 09:19:21 +0300	[thread overview]
Message-ID: <20200327061921.GA3260@avx2> (raw)
In-Reply-To: <9d18acbf-f878-c685-6272-a77eb5f98bcc@kernel.dk>

On Thu, Mar 26, 2020 at 04:05:08PM -0600, Jens Axboe wrote:
> On 3/26/20 2:44 PM, Alexey Dobriyan wrote:
> > Add simple iodepth=1 NVMe engine:
> > 
> > 	ioengine=nvme
> > 
> > It works via standard Linux NVMe ioctls.
> > 
> > It will be used for testing upcoming ZNS stuff.
> > 
> > Currently Linux doesn't recognize NVMe ZNS devices as zoned block
> > devices so zone ioctls (BLKRESETZONE et al) can't be used.
> > 
> > Passthrough ioctls should allow Zone Append and whatever commands
> > new specs bring.
> > 
> > Support read, write, fsync, fdatasync.
> > Don't support sync_file_range obviously.
> > Don't support trim for now, until I figure all qemu options and
> > the story behind broken qemu trim support.
> 
> Needs a bit of cleanup, but otherwise simple enough. A few comments:
> 
> > +static inline uint32_t get_nsid(const struct fio_file *f)
> > +{
> > +	return (uintptr_t)f->engine_data;
> > +}
> 
> You could pack both nsid and lba_shift in there, and maybe use
> engine_pos instead to avoid the cast.

lba_shift should be generic (live in struct fio_file). Maybe something
else will use it. If my zone fix goes in then 2 64-bit divisions can be
simplified!

> > +static int fio_nvme_read(struct fio_file *f, struct io_u *io_u)
> > +{
> > +	fio_unused uint32_t nsid = get_nsid(f);
> > +	struct nvme_user_io cmd = {};
> > +
> > +	//printf("R %u %llu/%llu\n", nsid, io_u->offset, io_u->xfer_buflen);
> 
> Leftover debug

dprint(FD_IO is pretty chatty, but OK.

> 
> > +static int fio_nvme_write(struct fio_file *f, struct io_u *io_u)
> > +{
> > +	fio_unused uint32_t nsid = get_nsid(f);
> > +	struct nvme_user_io cmd = {};
> > +
> > +	//printf("W %u %llu/%llu\n", nsid, io_u->offset, io_u->xfer_buflen);
> 
> Ditto
> 
> > +static int fio_nvme_flush(struct fio_file *f)
> > +{
> > +	uint32_t nsid = get_nsid(f);
> > +	struct nvme_passthru_cmd cmd = {};
> > +
> > +	//printf("F %u\n", nsid);
> 
> Ditto


  reply	other threads:[~2020-03-27  6:19 UTC|newest]

Thread overview: 15+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2020-03-26 20:44 [PATCH] fio: add NVMe engine Alexey Dobriyan
2020-03-26 22:05 ` Jens Axboe
2020-03-27  6:19   ` Alexey Dobriyan [this message]
2020-03-27 14:45     ` Jens Axboe
2020-03-27  0:56 ` Damien Le Moal
2020-03-27  6:14   ` Alexey Dobriyan
2020-03-27 14:25     ` Keith Busch
2020-03-27 14:26       ` Keith Busch
2020-03-27 19:06         ` Alexey Dobriyan
2020-03-27 21:05           ` Keith Busch
2020-03-27 14:47       ` Jens Axboe
2020-03-27 19:01         ` Alexey Dobriyan
2020-03-27 21:25           ` Jens Axboe
2020-03-27 21:58             ` Keith Busch
2020-03-28 13:41               ` 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 \
    --in-reply-to=20200327061921.GA3260@avx2 \
    --to=adobriyan@gmail.com \
    --cc=axboe@kernel.dk \
    --cc=fio@vger.kernel.org \
    /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