From: Dan Williams <dan.j.williams@intel.com>
To: alexander.h.duyck@linux.intel.com
Cc: "Brown, Len" <len.brown@intel.com>,
bvanassche@acm.org,
Linux-pm mailing list <linux-pm@vger.kernel.org>,
Greg KH <gregkh@linuxfoundation.org>,
linux-nvdimm <linux-nvdimm@lists.01.org>,
jiangshanlai@gmail.com,
Linux Kernel Mailing List <linux-kernel@vger.kernel.org>,
Pavel Machek <pavel@ucw.cz>,
zwisler@kernel.org, Tejun Heo <tj@kernel.org>,
Andrew Morton <akpm@linux-foundation.org>,
"Rafael J. Wysocki" <rafael@kernel.org>
Subject: Re: [driver-core PATCH v6 6/9] driver core: Probe devices asynchronously instead of the driver
Date: Mon, 26 Nov 2018 18:48:05 -0800 [thread overview]
Message-ID: <CAPcyv4g0LNJZ38NKMXsBY+GgmGbHR4VrTM1gkm3m4bg6fOrfsw@mail.gmail.com> (raw)
In-Reply-To: <154170043123.12967.3591757325647337726.stgit@ahduyck-desk1.jf.intel.com>
On Thu, Nov 8, 2018 at 10:07 AM Alexander Duyck
<alexander.h.duyck@linux.intel.com> wrote:
>
> Probe devices asynchronously instead of the driver. This results in us
> seeing the same behavior if the device is registered before the driver or
> after. This way we can avoid serializing the initialization should the
> driver not be loaded until after the devices have already been added.
>
> The motivation behind this is that if we have a set of devices that
> take a significant amount of time to load we can greatly reduce the time to
> load by processing them in parallel instead of one at a time. In addition,
> each device can exist on a different node so placing a single thread on one
> CPU to initialize all of the devices for a given driver can result in poor
> performance on a system with multiple nodes.
Do you have numbers on effects of this change individually? Is this
change necessary for the libnvdimm init speedup, or is it independent?
> I am using the driver_data member of the device struct to store the driver
> pointer while we wait on the deferred probe call. This should be safe to do
> as the value will either be set to NULL on a failed probe or driver load
> followed by unload, or the driver value itself will be set on a successful
> driver load. In addition I have used the async_probe flag to add additional
> protection as it will be cleared if someone overwrites the driver_data
> member as a part of loading the driver.
I would not put it past a device-driver to call dev_get_drvdata()
before dev_set_drvdata(), to check "has this device already been
initialized". So I don't think it is safe to assume that the core can
stash this information in ->driver_data. Why not put this
infrastructure in struct device_private?
_______________________________________________
Linux-nvdimm mailing list
Linux-nvdimm@lists.01.org
https://lists.01.org/mailman/listinfo/linux-nvdimm
WARNING: multiple messages have this Message-ID (diff)
From: Dan Williams <dan.j.williams@intel.com>
To: alexander.h.duyck@linux.intel.com
Cc: Linux Kernel Mailing List <linux-kernel@vger.kernel.org>,
Greg KH <gregkh@linuxfoundation.org>,
linux-nvdimm <linux-nvdimm@lists.01.org>,
Tejun Heo <tj@kernel.org>,
Andrew Morton <akpm@linux-foundation.org>,
Linux-pm mailing list <linux-pm@vger.kernel.org>,
jiangshanlai@gmail.com, "Rafael J. Wysocki" <rafael@kernel.org>,
"Brown, Len" <len.brown@intel.com>, Pavel Machek <pavel@ucw.cz>,
zwisler@kernel.org, Dave Jiang <dave.jiang@intel.com>,
bvanassche@acm.org
Subject: Re: [driver-core PATCH v6 6/9] driver core: Probe devices asynchronously instead of the driver
Date: Mon, 26 Nov 2018 18:48:05 -0800 [thread overview]
Message-ID: <CAPcyv4g0LNJZ38NKMXsBY+GgmGbHR4VrTM1gkm3m4bg6fOrfsw@mail.gmail.com> (raw)
In-Reply-To: <154170043123.12967.3591757325647337726.stgit@ahduyck-desk1.jf.intel.com>
On Thu, Nov 8, 2018 at 10:07 AM Alexander Duyck
<alexander.h.duyck@linux.intel.com> wrote:
>
> Probe devices asynchronously instead of the driver. This results in us
> seeing the same behavior if the device is registered before the driver or
> after. This way we can avoid serializing the initialization should the
> driver not be loaded until after the devices have already been added.
>
> The motivation behind this is that if we have a set of devices that
> take a significant amount of time to load we can greatly reduce the time to
> load by processing them in parallel instead of one at a time. In addition,
> each device can exist on a different node so placing a single thread on one
> CPU to initialize all of the devices for a given driver can result in poor
> performance on a system with multiple nodes.
Do you have numbers on effects of this change individually? Is this
change necessary for the libnvdimm init speedup, or is it independent?
> I am using the driver_data member of the device struct to store the driver
> pointer while we wait on the deferred probe call. This should be safe to do
> as the value will either be set to NULL on a failed probe or driver load
> followed by unload, or the driver value itself will be set on a successful
> driver load. In addition I have used the async_probe flag to add additional
> protection as it will be cleared if someone overwrites the driver_data
> member as a part of loading the driver.
I would not put it past a device-driver to call dev_get_drvdata()
before dev_set_drvdata(), to check "has this device already been
initialized". So I don't think it is safe to assume that the core can
stash this information in ->driver_data. Why not put this
infrastructure in struct device_private?
WARNING: multiple messages have this Message-ID (diff)
From: Dan Williams <dan.j.williams-ral2JQCrhuEAvxtiuMwx3w@public.gmane.org>
To: alexander.h.duyck-VuQAYsv1563Yd54FQh9/CA@public.gmane.org
Cc: "Brown, Len" <len.brown-ral2JQCrhuEAvxtiuMwx3w@public.gmane.org>,
bvanassche-HInyCGIudOg@public.gmane.org,
Linux-pm mailing list
<linux-pm-u79uwXL29TY76Z2rM5mHXA@public.gmane.org>,
Greg KH
<gregkh-hQyY1W1yCW8ekmWlsbkhG0B+6BGkLq7r@public.gmane.org>,
linux-nvdimm
<linux-nvdimm-hn68Rpc1hR1g9hUCZPvPmw@public.gmane.org>,
jiangshanlai-Re5JQEeQqe8AvxtiuMwx3w@public.gmane.org,
Linux Kernel Mailing List
<linux-kernel-u79uwXL29TY76Z2rM5mHXA@public.gmane.org>,
Pavel Machek <pavel-+ZI9xUNit7I@public.gmane.org>,
zwisler-DgEjT+Ai2ygdnm+yROfE0A@public.gmane.org,
Tejun Heo <tj-DgEjT+Ai2ygdnm+yROfE0A@public.gmane.org>,
Andrew Morton
<akpm-de/tnXTf+JLsfHDXvbKv3WD2FQJk+8+b@public.gmane.org>,
"Rafael J. Wysocki"
<rafael-DgEjT+Ai2ygdnm+yROfE0A@public.gmane.org>
Subject: Re: [driver-core PATCH v6 6/9] driver core: Probe devices asynchronously instead of the driver
Date: Mon, 26 Nov 2018 18:48:05 -0800 [thread overview]
Message-ID: <CAPcyv4g0LNJZ38NKMXsBY+GgmGbHR4VrTM1gkm3m4bg6fOrfsw@mail.gmail.com> (raw)
In-Reply-To: <154170043123.12967.3591757325647337726.stgit-+uVpp3jiz/RcxmDmkzA3yGt3HXsI98Cx0E9HWUfgJXw@public.gmane.org>
On Thu, Nov 8, 2018 at 10:07 AM Alexander Duyck
<alexander.h.duyck-VuQAYsv1563Yd54FQh9/CA@public.gmane.org> wrote:
>
> Probe devices asynchronously instead of the driver. This results in us
> seeing the same behavior if the device is registered before the driver or
> after. This way we can avoid serializing the initialization should the
> driver not be loaded until after the devices have already been added.
>
> The motivation behind this is that if we have a set of devices that
> take a significant amount of time to load we can greatly reduce the time to
> load by processing them in parallel instead of one at a time. In addition,
> each device can exist on a different node so placing a single thread on one
> CPU to initialize all of the devices for a given driver can result in poor
> performance on a system with multiple nodes.
Do you have numbers on effects of this change individually? Is this
change necessary for the libnvdimm init speedup, or is it independent?
> I am using the driver_data member of the device struct to store the driver
> pointer while we wait on the deferred probe call. This should be safe to do
> as the value will either be set to NULL on a failed probe or driver load
> followed by unload, or the driver value itself will be set on a successful
> driver load. In addition I have used the async_probe flag to add additional
> protection as it will be cleared if someone overwrites the driver_data
> member as a part of loading the driver.
I would not put it past a device-driver to call dev_get_drvdata()
before dev_set_drvdata(), to check "has this device already been
initialized". So I don't think it is safe to assume that the core can
stash this information in ->driver_data. Why not put this
infrastructure in struct device_private?
next prev parent reply other threads:[~2018-11-27 2:48 UTC|newest]
Thread overview: 107+ messages / expand[flat|nested] mbox.gz Atom feed top
2018-11-08 18:06 [driver-core PATCH v6 0/9] Add NUMA aware async_schedule calls Alexander Duyck
2018-11-08 18:06 ` Alexander Duyck
2018-11-08 18:06 ` [driver-core PATCH v6 1/9] workqueue: Provide queue_work_node to queue work near a given NUMA node Alexander Duyck
2018-11-08 18:06 ` Alexander Duyck
2018-11-27 1:01 ` Dan Williams
2018-11-27 1:01 ` Dan Williams
2018-11-08 18:06 ` [driver-core PATCH v6 2/9] async: Add support for queueing on specific " Alexander Duyck
2018-11-08 18:06 ` Alexander Duyck
2018-11-08 23:36 ` Bart Van Assche
2018-11-08 23:36 ` Bart Van Assche
2018-11-08 23:36 ` Bart Van Assche
2018-11-11 19:32 ` Greg KH
2018-11-11 19:32 ` Greg KH
2018-11-11 19:32 ` Greg KH
2018-11-11 19:53 ` Dan Williams
2018-11-11 19:53 ` Dan Williams
2018-11-11 19:53 ` Dan Williams
2018-11-11 20:35 ` Greg KH
2018-11-11 20:35 ` Greg KH
2018-11-11 20:35 ` Greg KH
2018-11-11 22:17 ` Dan Williams
2018-11-11 22:17 ` Dan Williams
2018-11-11 22:17 ` Dan Williams
2018-11-11 23:27 ` Alexander Duyck
2018-11-11 23:27 ` Alexander Duyck
2018-11-11 23:27 ` Alexander Duyck
2018-11-11 19:59 ` Pavel Machek
2018-11-11 20:33 ` Greg KH
2018-11-11 20:33 ` Greg KH
2018-11-11 21:24 ` Bart Van Assche
2018-11-11 21:24 ` Bart Van Assche
2018-11-13 22:10 ` Pavel Machek
2018-11-27 1:10 ` Dan Williams
2018-11-27 1:10 ` Dan Williams
2018-11-27 1:10 ` Dan Williams
2018-11-08 18:06 ` [driver-core PATCH v6 3/9] device core: Consolidate locking and unlocking of parent and device Alexander Duyck
2018-11-08 18:06 ` Alexander Duyck
2018-11-08 18:06 ` Alexander Duyck
2018-11-08 22:43 ` jane.chu
2018-11-08 22:43 ` jane.chu-QHcLZuEGTsvQT0dZR+AlfA
2018-11-08 22:43 ` jane.chu
2018-11-08 22:48 ` Alexander Duyck
2018-11-08 22:48 ` Alexander Duyck
2018-11-27 1:44 ` Dan Williams
2018-11-27 1:44 ` Dan Williams
2018-11-08 18:07 ` [driver-core PATCH v6 4/9] driver core: Move async_synchronize_full call Alexander Duyck
2018-11-08 18:07 ` Alexander Duyck
2018-11-27 2:11 ` Dan Williams
2018-11-27 2:11 ` Dan Williams
2018-11-27 17:38 ` Alexander Duyck
2018-11-27 17:38 ` Alexander Duyck
2018-11-27 20:35 ` Dan Williams
2018-11-27 20:35 ` Dan Williams
2018-11-27 21:36 ` Alexander Duyck
2018-11-27 21:36 ` Alexander Duyck
2018-11-27 22:26 ` Dan Williams
2018-11-27 22:26 ` Dan Williams
2018-11-27 22:26 ` Dan Williams
2018-11-08 18:07 ` [driver-core PATCH v6 5/9] driver core: Establish clear order of operations for deferred probe and remove Alexander Duyck
2018-11-08 18:07 ` Alexander Duyck
2018-11-08 18:07 ` Alexander Duyck
2018-11-08 23:47 ` Bart Van Assche
2018-11-08 23:47 ` Bart Van Assche
2018-11-08 23:47 ` Bart Van Assche
2018-11-08 18:07 ` [driver-core PATCH v6 6/9] driver core: Probe devices asynchronously instead of the driver Alexander Duyck
2018-11-08 18:07 ` Alexander Duyck
2018-11-08 18:07 ` Alexander Duyck
2018-11-08 23:59 ` Bart Van Assche
2018-11-08 23:59 ` Bart Van Assche
2018-11-27 2:48 ` Dan Williams [this message]
2018-11-27 2:48 ` Dan Williams
2018-11-27 2:48 ` Dan Williams
2018-11-27 17:57 ` Alexander Duyck
2018-11-27 17:57 ` Alexander Duyck
2018-11-27 18:32 ` Dan Williams
2018-11-27 18:32 ` Dan Williams
2018-11-08 18:07 ` [driver-core PATCH v6 7/9] driver core: Attach devices on CPU local to device node Alexander Duyck
2018-11-08 18:07 ` Alexander Duyck
2018-11-27 4:50 ` Dan Williams
2018-11-27 4:50 ` Dan Williams
2018-11-27 4:50 ` Dan Williams
2018-11-08 18:07 ` [driver-core PATCH v6 8/9] PM core: Use new async_schedule_dev command Alexander Duyck
2018-11-08 18:07 ` Alexander Duyck
2018-11-08 18:07 ` Alexander Duyck
2018-11-27 4:52 ` Dan Williams
2018-11-27 4:52 ` Dan Williams
2018-11-27 4:52 ` Dan Williams
2018-11-08 18:07 ` [driver-core PATCH v6 9/9] libnvdimm: Schedule device registration on node local to the device Alexander Duyck
2018-11-08 18:07 ` Alexander Duyck
2018-11-27 2:21 ` Dan Williams
2018-11-27 2:21 ` Dan Williams
2018-11-27 2:21 ` Dan Williams
2018-11-27 18:04 ` Alexander Duyck
2018-11-27 19:34 ` Dan Williams
2018-11-27 19:34 ` Dan Williams
2018-11-27 20:33 ` Bart Van Assche
2018-11-27 20:33 ` Bart Van Assche
2018-11-27 20:33 ` Bart Van Assche
2018-11-27 20:50 ` Dan Williams
2018-11-27 20:50 ` Dan Williams
2018-11-27 20:50 ` Dan Williams
2018-11-27 21:22 ` Bart Van Assche
2018-11-27 21:22 ` Bart Van Assche
2018-11-27 21:22 ` Bart Van Assche
2018-11-27 22:34 ` Dan Williams
2018-11-27 22:34 ` Dan Williams
2018-11-27 22:34 ` Dan Williams
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=CAPcyv4g0LNJZ38NKMXsBY+GgmGbHR4VrTM1gkm3m4bg6fOrfsw@mail.gmail.com \
--to=dan.j.williams@intel.com \
--cc=akpm@linux-foundation.org \
--cc=alexander.h.duyck@linux.intel.com \
--cc=bvanassche@acm.org \
--cc=gregkh@linuxfoundation.org \
--cc=jiangshanlai@gmail.com \
--cc=len.brown@intel.com \
--cc=linux-kernel@vger.kernel.org \
--cc=linux-nvdimm@lists.01.org \
--cc=linux-pm@vger.kernel.org \
--cc=pavel@ucw.cz \
--cc=rafael@kernel.org \
--cc=tj@kernel.org \
--cc=zwisler@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 an external index of several public inboxes,
see mirroring instructions on how to clone and mirror
all data and code used by this external index.