From: Weiping Zhang <zwp10758@gmail.com>
To: Max Gurtovoy <maxg@mellanox.com>
Cc: Jens Axboe <axboe@kernel.dk>,
sagi@grimberg.me, Weiping Zhang <zhangweiping@didiglobal.com>,
linux-nvme@lists.infradead.org,
Christoph Hellwig <hch@infradead.org>,
Keith Busch <kbusch@kernel.org>
Subject: Re: [PATCH v2] nvme: align io queue count with allocted nvme_queue in nvme_probe
Date: Fri, 24 Apr 2020 12:25:19 +0800 [thread overview]
Message-ID: <CAA70yB6iHXRjr-dYWQKh7UP2-vmA8rk-7=h3DprvQENFqn3YFA@mail.gmail.com> (raw)
In-Reply-To: <de3cdc57-a97a-f800-26dd-f9439f97bba5@mellanox.com>
On Thu, Apr 23, 2020 at 6:25 PM Max Gurtovoy <maxg@mellanox.com> wrote:
>
>
> On 4/23/2020 10:59 AM, Weiping Zhang wrote:
> > Since the commit 147b27e4bd0 "nvme-pci: allocate device queues storage space at probe"
> > nvme_alloc_queue will not alloc struct nvme_queue any more.
> > If user change write/poll_queues to larger than the number of
> > allocated queue in nvme_probe, nvme_alloc_queue will touch
> > the memory out of boundary.
> >
> > This patch add nr_allocated_queues for struct nvme_dev to record how
> > many queues alloctaed in nvme_probe, then nvme driver will not use
> > more queues than nr_allocated_queues when user update queue count
> > and do a controller reset.
> >
> > Since global module parameter can be changed at rumtime, so it's not
> > safe to use these two parameter directly in the following functions:
> > nvme_dbbuf_dma_alloc
> > nvme_dbbuf_dma_free
> > nvme_calc_irq_sets
> > nvme_setup_io_queues
> >
> > This patch also add nr_write_queues, nr_poll_queues for
> > struct nvme_dev and io_queues_reload for struct nvme_ctrl, that allow
> > per-controller reload module parmater when reset controller. The nvme
> > driver will not reload module parameter(write/poll_queues) by default
> > when reset controller. If user want to reload them, they should enable
> > it by echo 1 > /sys/block/<nvme_disk>/device/io_queues_reload.
> >
> > By now, nvme pci driver allow user change io queue count for each
> > type(write, read, poll) within nr_allocated_queue, that's to say, if
> > user want to change queue dynamically by reset controller, they should
> > setup io queues as many as possiable when laod nvme module, and then
> > tune io queue count for each type.
>
> typo: laod --> load
>
OK, fix it in V3.
>
> > Signed-off-by: Weiping Zhang <zhangweiping@didiglobal.com>
> > ---
> > Changes since V1:
> > * don't use module parameter nvme_dbbuf_dma_free, nvme_dbbuf_dma_alloc
> > and nvme_calc_irq_sets.
> > * add per-controller sysfs file io_queues_reload to enable/disable
> > reload global module parameter.
> >
> > drivers/nvme/host/core.c | 29 +++++++++++++++++++++
> > drivers/nvme/host/nvme.h | 1 +
> > drivers/nvme/host/pci.c | 55 +++++++++++++++++++++++-----------------
> > 3 files changed, 62 insertions(+), 23 deletions(-)
> >
> > diff --git a/drivers/nvme/host/core.c b/drivers/nvme/host/core.c
> > index dfb064b4334f..80172192a9d8 100644
> > --- a/drivers/nvme/host/core.c
> > +++ b/drivers/nvme/host/core.c
> > @@ -3357,6 +3357,34 @@ static ssize_t nvme_sysfs_show_address(struct device *dev,
> > }
> > static DEVICE_ATTR(address, S_IRUGO, nvme_sysfs_show_address, NULL);
> >
> > +static ssize_t nvme_sysfs_io_queues_reload_show(struct device *dev,
> > + struct device_attribute *attr,
> > + char *buf)
> > +{
> > + struct nvme_ctrl *ctrl = dev_get_drvdata(dev);
> > +
> > + return snprintf(buf, PAGE_SIZE, "%d\n",
> > + ctrl->io_queues_reload ? 1 : 0);
> > +}
> > +
> > +static ssize_t nvme_sysfs_io_queues_reload_store(struct device *dev,
> > + struct device_attribute *attr, const char *buf,
> > + size_t count)
> > +{
> > + struct nvme_ctrl *ctrl = dev_get_drvdata(dev);
> > + bool val;
> > +
> > + if (kstrtobool(buf, &val))
> > + return -EINVAL;
> > + ctrl->io_queues_reload = val;
> > +
> > + return count;
> > +}
> > +
> > +static DEVICE_ATTR(io_queues_reload, S_IRUGO | S_IWUSR,
> > + nvme_sysfs_io_queues_reload_show,
> > + nvme_sysfs_io_queues_reload_store);
> > +
> > static struct attribute *nvme_dev_attrs[] = {
> > &dev_attr_reset_controller.attr,
> > &dev_attr_rescan_controller.attr,
> > @@ -3374,6 +3402,7 @@ static struct attribute *nvme_dev_attrs[] = {
> > &dev_attr_sqsize.attr,
> > &dev_attr_hostnqn.attr,
> > &dev_attr_hostid.attr,
> > + &dev_attr_io_queues_reload.attr,
> > NULL
> > };
>
> Well for fabrics controllers it doesn't mean anything.
>
> maybe we can do it non-visible for fabrics ?
>
Make sense, fix in v3.
Thanks
> _______________________________________________
> linux-nvme mailing list
> linux-nvme@lists.infradead.org
> http://lists.infradead.org/mailman/listinfo/linux-nvme
_______________________________________________
linux-nvme mailing list
linux-nvme@lists.infradead.org
http://lists.infradead.org/mailman/listinfo/linux-nvme
prev parent reply other threads:[~2020-04-24 4:25 UTC|newest]
Thread overview: 3+ messages / expand[flat|nested] mbox.gz Atom feed top
2020-04-23 7:59 [PATCH v2] nvme: align io queue count with allocted nvme_queue in nvme_probe Weiping Zhang
2020-04-23 10:24 ` Max Gurtovoy
2020-04-24 4:25 ` Weiping Zhang [this message]
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='CAA70yB6iHXRjr-dYWQKh7UP2-vmA8rk-7=h3DprvQENFqn3YFA@mail.gmail.com' \
--to=zwp10758@gmail.com \
--cc=axboe@kernel.dk \
--cc=hch@infradead.org \
--cc=kbusch@kernel.org \
--cc=linux-nvme@lists.infradead.org \
--cc=maxg@mellanox.com \
--cc=sagi@grimberg.me \
--cc=zhangweiping@didiglobal.com \
/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;
as well as URLs for NNTP newsgroup(s).