From: Bastien Nocera <hadess@hadess.net>
To: Alan Stern <stern@rowland.harvard.edu>
Cc: linux-usb@vger.kernel.org,
Greg Kroah-Hartman <gregkh@linuxfoundation.org>,
Benjamin Tissoires <benjamin.tissoires@redhat.com>
Subject: Re: [PATCH 2/5] USB: Make it possible to "subclass" usb_device_driver
Date: Thu, 10 Oct 2019 10:10:12 +0200 [thread overview]
Message-ID: <28db57b47c9f0e7a7a8ad1b6c4bac70cdf96b450.camel@hadess.net> (raw)
In-Reply-To: <Pine.LNX.4.44L0.1910091025500.1603-100000@iolanthe.rowland.org>
More replies inline (which I always miss)
On Wed, 2019-10-09 at 10:34 -0400, Alan Stern wrote:
> On Wed, 9 Oct 2019, Bastien Nocera wrote:
>
> > The kernel currenly has only 2 usb_device_drivers, one generic one,
> one
> > that completely replaces the generic one to make USB devices usable
> over
> > a network.
>
> Presumably your first driver is in generic.c. Where is the second
> one?
>
> > Use the newly exported generic driver functions when a driver
> declares
> > to want them run, in addition to its own code. This makes it
> possible to
> > write drivers that extend the generic USB driver.
> >
> > Signed-off-by: Bastien Nocera <hadess@hadess.net>
>
> This has a few problems. The biggest one is that the device core
> does
> not guarantee any order of driver probing. If generic.c is probed
> first, the subclass driver will never get probed -- which is a
> pretty
> fatal flaw.
>
> > ---
> > drivers/usb/core/driver.c | 36 ++++++++++++++++++++++++++++++-----
> -
> > include/linux/usb.h | 1 +
> > 2 files changed, 31 insertions(+), 6 deletions(-)
> >
> > diff --git a/drivers/usb/core/driver.c b/drivers/usb/core/driver.c
> > index 2b27d232d7a7..863e380a272b 100644
> > --- a/drivers/usb/core/driver.c
> > +++ b/drivers/usb/core/driver.c
> > @@ -261,10 +261,17 @@ static int usb_probe_device(struct device
> *dev)
> > */
> > if (!udriver->supports_autosuspend)
> > error = usb_autoresume_device(udev);
> > + if (error)
> > + return error;
> >
> > - if (!error)
> > - error = udriver->probe(udev);
> > - return error;
> > + if (udriver->generic_init)
> > + error = usb_generic_driver_probe(udev);
> > + if (error)
> > + return error;
> > +
> > + if (udriver->probe)
> > + return udriver->probe(udev);
> > + return 0;
> > }
> >
> > /* called from driver core with dev locked */
> > @@ -273,7 +280,10 @@ static int usb_unbind_device(struct device
> *dev)
> > struct usb_device *udev = to_usb_device(dev);
> > struct usb_device_driver *udriver = to_usb_device_driver(dev-
> >driver);
> >
> > - udriver->disconnect(udev);
> > + if (udriver->generic_init)
> > + usb_generic_driver_disconnect(udev);
> > + if (udriver->disconnect)
> > + udriver->disconnect(udev);
>
> The order is wrong. The disconnects should always be done in
> reverse
> order of probing. This is true whenever you have a destructor for a
> subclass; the subclasses destructor runs before the superclass's
> destructor.
Fixed. Fixed in the suspend function as well.
> > if (!udriver->supports_autosuspend)
> > usb_autosuspend_device(udev);
> > return 0;
> > @@ -886,6 +896,14 @@ int usb_register_device_driver(struct
> usb_device_driver *new_udriver,
> > if (usb_disabled())
> > return -ENODEV;
> >
> > + if (new_udriver->probe == NULL &&
> > + !new_udriver->generic_init) {
>
> There's no point adding this extra test. Even subclass drivers
> should
> have a probe function.
Removed.
> > + printk(KERN_ERR "%s: error %d registering device "
> > + " driver %s, no probe() function\n",
>
> Don't split character strings. They are an exception to the 80-
> column
> limit.
I was using the error message just below in the function as an example.
A bad one apparently. This is gone in any case.
> > + usbcore_name, retval, new_udriver->name);
> > + return -EINVAL;
> > + }
> > +
> > new_udriver->drvwrap.for_devices = 1;
> > new_udriver->drvwrap.driver.name = new_udriver->name;
> > new_udriver->drvwrap.driver.bus = &usb_bus_type;
> > @@ -1149,7 +1167,10 @@ static int usb_suspend_device(struct
> usb_device *udev, pm_message_t msg)
> > udev->do_remote_wakeup = 0;
> > udriver = &usb_generic_driver;
> > }
> > - status = udriver->suspend(udev, msg);
> > + if (udriver->generic_init)
> > + status = usb_generic_driver_suspend (udev, msg);
> > + if (status == 0 && udriver->suspend)
> > + status = udriver->suspend(udev, msg);
>
> Again, the order is wrong. Suspend the subclass driver first.
Done, as mentioned above.
> > done:
> > dev_vdbg(&udev->dev, "%s: status %d\n", __func__, status);
> > @@ -1181,7 +1202,10 @@ static int usb_resume_device(struct
> usb_device *udev, pm_message_t msg)
> > udev->reset_resume = 1;
> >
> > udriver = to_usb_device_driver(udev->dev.driver);
> > - status = udriver->resume(udev, msg);
> > + if (udriver->generic_init)
> > + status = usb_generic_driver_resume (udev, msg);
> > + if (status == 0 && udriver->resume)
> > + status = udriver->resume(udev, msg);
> >
> > done:
> > dev_vdbg(&udev->dev, "%s: status %d\n", __func__, status);
> > diff --git a/include/linux/usb.h b/include/linux/usb.h
> > index e656e7b4b1e4..fb9ad3511e55 100644
> > --- a/include/linux/usb.h
> > +++ b/include/linux/usb.h
> > @@ -1242,6 +1242,7 @@ struct usb_device_driver {
> > const struct attribute_group **dev_groups;
> > struct usbdrv_wrap drvwrap;
> > unsigned int supports_autosuspend:1;
> > + unsigned int generic_init:1;
>
> How about using a name that actually says something about the
> driver?
> Such as generic_subclass? Or subclass_of_generic?
>
> "init" has nothing to do with anything.
generic_subclass it will be. I've also documented it in the header.
> > };
> > #define to_usb_device_driver(d) container_of(d, struct
> usb_device_driver, \
> > drvwrap.driver)
>
> Alan Stern
>
next prev parent reply other threads:[~2019-10-10 8:10 UTC|newest]
Thread overview: 21+ messages / expand[flat|nested] mbox.gz Atom feed top
2019-10-09 13:43 [PATCH 0/5] Add Apple MFi fastcharge USB device driver Bastien Nocera
2019-10-09 13:43 ` [PATCH 1/5] USB: Export generic USB device driver functions Bastien Nocera
2019-10-09 13:43 ` [PATCH 2/5] USB: Make it possible to "subclass" usb_device_driver Bastien Nocera
2019-10-09 14:34 ` Alan Stern
2019-10-09 14:41 ` Alan Stern
2019-10-10 8:10 ` Bastien Nocera [this message]
2019-10-10 9:58 ` kbuild test robot
2019-10-09 13:43 ` [PATCH 3/5] USB: Implement usb_device_match_id() Bastien Nocera
2019-10-09 14:36 ` Alan Stern
2019-10-09 15:40 ` Bastien Nocera
2019-10-09 17:29 ` Alan Stern
2019-10-09 13:43 ` [PATCH 4/5] USB: Select better matching USB drivers when available Bastien Nocera
2019-10-09 14:43 ` Alan Stern
2019-10-09 15:35 ` Bastien Nocera
2019-10-09 17:28 ` Alan Stern
2019-10-09 18:24 ` Bastien Nocera
2019-10-09 18:45 ` Alan Stern
2019-10-10 8:26 ` Bastien Nocera
2019-10-10 14:19 ` Alan Stern
2019-10-10 10:33 ` kbuild test robot
2019-10-09 13:43 ` [PATCH 5/5] USB: Add driver to control USB fast charge for iOS devices Bastien Nocera
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=28db57b47c9f0e7a7a8ad1b6c4bac70cdf96b450.camel@hadess.net \
--to=hadess@hadess.net \
--cc=benjamin.tissoires@redhat.com \
--cc=gregkh@linuxfoundation.org \
--cc=linux-usb@vger.kernel.org \
--cc=stern@rowland.harvard.edu \
/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).