lists.openwall.net   lists  /  announce  owl-users  owl-dev  john-users  john-dev  passwdqc-users  yescrypt  popa3d-users  /  oss-security  kernel-hardening  musl  sabotage  tlsify  passwords  /  crypt-dev  xvendor  /  Bugtraq  Full-Disclosure  linux-kernel  linux-netdev  linux-ext4  linux-hardening  linux-cve-announce  PHC 
Open Source and information security mailing list archives
 
Hash Suite: Windows password security audit tool. GUI, reports in PDF.
[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]
Message-ID: <3773c98d27b5043dae7d293bc4ab8268ae204c16.camel@analog.com>
Date:   Mon, 25 May 2020 07:24:51 +0000
From:   "Ardelean, Alexandru" <alexandru.Ardelean@...log.com>
To:     "jic23@...nel.org" <jic23@...nel.org>
CC:     "linux-kernel@...r.kernel.org" <linux-kernel@...r.kernel.org>,
        "linux-iio@...r.kernel.org" <linux-iio@...r.kernel.org>,
        "lars@...afoo.de" <lars@...afoo.de>
Subject: Re: [RFC PATCH 07/14] iio: core: add simple centralized mechanism for
 ioctl() handlers

On Sun, 2020-05-24 at 17:45 +0100, Jonathan Cameron wrote:
> [External]
> 
> On Fri, 8 May 2020 16:53:41 +0300
> Alexandru Ardelean <alexandru.ardelean@...log.com> wrote:
> 
> > The aim of this is to reduce the organization violation of ioctl() calls in
> > IIO core. Currently, since the chardev is split across files, event ioctl()
> > calls need to be called in buffer ioctl() calls.
> > 
> > The 'industrialio-core.c' file will provide a 'iio_device_ioctl()' which
> > will iterate over a list of ioctls registered with the IIO device. These
> > can be event ioctl() or buffer ioctl() calls, or something else.
> > This is needed, since there is currently one chardev per IIO device and
> > that is used for both event handling and reading from the buffer.
> > 
> > Each ioctl() will have to return a IIO_IOCTL_UNHANDLED code (which is
> > positive 1), if the ioctl() did not handle the call in any. This eliminates
> > any potential ambiguities; if we were to have used error codes it would
> > have been uncertain whether they were actual errors, or whether
> > the registered ioctl() doesn't service the command.
> > 
> > If any ioctl() returns 0, it was considered that it was serviced
> > successfully and the loop will exit.
> > 
> > One assumption for all registered ioctl() handlers is that they are
> > statically allocated, so the iio_device_unregister() which just remove all
> > of them from the device's ioctl() handler list.
> > 
> > Also, something that is a bit hard to do [at this point] and may not be
> > worth the effort of doing, is to check whether registered ioctl()
> > calls/commands overlap. This should be unlikely to happen, and should get
> > caught at review time. Though, new ioctl() calls would likely not be added
> > too often.
> > 
> > Signed-off-by: Alexandru Ardelean <alexandru.ardelean@...log.com>
> 
> We seem to have dropped the locking in here.   What am I missing that
> stops us racing a remove with the ioctl?  If there is a reason that
> can't race, please add comments there so I don't wonders sometime in
> the future.
> 

Yeah.
My bad about the locking.
I have too many branches for this stuff.
But for the chardev split, there's a branch that should have the locking in
place.
I'll re-check that, when sending only the chardev split anyway.


> The check on iio_dev->info means we won't start the ioctl if the
> remove has been called, but if we switch immediately after that,
> anything can happen before we start calling the ioctls.
> 
> J
> 
> > ---
> >  drivers/iio/iio_core.h          | 14 ++++++++++++++
> >  drivers/iio/industrialio-core.c | 33 +++++++++++++++++++++++++++++++++
> >  include/linux/iio/iio.h         |  2 ++
> >  3 files changed, 49 insertions(+)
> > 
> > diff --git a/drivers/iio/iio_core.h b/drivers/iio/iio_core.h
> > index a527a66be9e5..34c3e19229d8 100644
> > --- a/drivers/iio/iio_core.h
> > +++ b/drivers/iio/iio_core.h
> > @@ -17,6 +17,20 @@ struct iio_dev;
> >  
> >  extern struct device_type iio_device_type;
> >  
> > +#define IIO_IOCTL_UNHANDLED	1
> > +struct iio_ioctl_handler {
> > +	struct list_head entry;
> > +	long (*ioctl)(struct iio_dev *indio_dev, struct file *filp,
> > +		      unsigned int cmd, unsigned long arg);
> > +};
> > +
> > +long iio_device_ioctl(struct iio_dev *indio_dev, struct file *filp,
> > +		      unsigned int cmd, unsigned long arg);
> > +
> > +void iio_device_ioctl_handler_register(struct iio_dev *indio_dev,
> > +				       struct iio_ioctl_handler *h);
> > +void iio_device_ioctl_handler_unregister(struct iio_ioctl_handler *h);
> > +
> >  int __iio_add_chan_devattr(const char *postfix,
> >  			   struct iio_chan_spec const *chan,
> >  			   ssize_t (*func)(struct device *dev,
> > diff --git a/drivers/iio/industrialio-core.c b/drivers/iio/industrialio-
> > core.c
> > index 32e045c7f0c1..5df3af5e7dcb 100644
> > --- a/drivers/iio/industrialio-core.c
> > +++ b/drivers/iio/industrialio-core.c
> > @@ -1534,6 +1534,7 @@ struct iio_dev *iio_device_alloc(int sizeof_priv)
> >  	}
> >  	dev_set_name(&dev->dev, "iio:device%d", dev->id);
> >  	INIT_LIST_HEAD(&dev->buffer_list);
> > +	INIT_LIST_HEAD(&dev->ioctl_handlers);
> >  
> >  	return dev;
> >  }
> > @@ -1587,6 +1588,33 @@ struct iio_dev *devm_iio_device_alloc(struct device
> > *dev, int sizeof_priv)
> >  }
> >  EXPORT_SYMBOL_GPL(devm_iio_device_alloc);
> >  
> > +void iio_device_ioctl_handler_register(struct iio_dev *indio_dev,
> > +				       struct iio_ioctl_handler *h)
> > +{
> > +	/* this assumes that all ioctl() handlers are statically allocated */
> > +	list_add_tail(&h->entry, &indio_dev->ioctl_handlers);
> > +}
> > +
> > +long iio_device_ioctl(struct iio_dev *indio_dev, struct file *filp,
> > +		      unsigned int cmd, unsigned long arg)
> > +{
> > +	struct iio_ioctl_handler *h;
> > +	int ret;
> > +
> > +	if (!indio_dev->info)
> > +		return -ENODEV;
> 
> The locking is gone?  
> > +
> > +	list_for_each_entry(h, &indio_dev->ioctl_handlers, entry) {
> > +		ret = h->ioctl(indio_dev, filp, cmd, arg);
> > +		if (ret == 0)
> > +			return 0;
> > +		if (ret != IIO_IOCTL_UNHANDLED)
> > +			return ret;
> > +	}
> > +
> > +	return -EINVAL;
> > +}
> > +
> >  static int iio_check_unique_scan_index(struct iio_dev *indio_dev)
> >  {
> >  	int i, j;
> > @@ -1722,6 +1750,8 @@ EXPORT_SYMBOL(__iio_device_register);
> >   **/
> >  void iio_device_unregister(struct iio_dev *indio_dev)
> >  {
> > +	struct iio_ioctl_handler *h, *t;
> > +
> >  	cdev_device_del(indio_dev->chrdev, &indio_dev->dev);
> >  	iio_device_free_chrdev_id(&indio_dev->dev);
> >  
> > @@ -1731,6 +1761,9 @@ void iio_device_unregister(struct iio_dev *indio_dev)
> >  
> >  	iio_disable_all_buffers(indio_dev);
> >  
> > +	list_for_each_entry_safe(h, t, &indio_dev->ioctl_handlers, entry)
> > +		list_del(&h->entry);
> > +
> >  	indio_dev->info = NULL;
> >  
> >  	iio_device_wakeup_eventset(indio_dev);
> > diff --git a/include/linux/iio/iio.h b/include/linux/iio/iio.h
> > index 52992be44e9e..b6ca8d85629e 100644
> > --- a/include/linux/iio/iio.h
> > +++ b/include/linux/iio/iio.h
> > @@ -488,6 +488,7 @@ struct iio_buffer_setup_ops {
> >   * @currentmode:	[DRIVER] current operating mode
> >   * @dev:		[DRIVER] device structure, should be assigned a parent
> >   *			and owner
> > + * @ioctl_handlers:	[INTERN] list of registered ioctl handlers
> >   * @event_interface:	[INTERN] event chrdevs associated with interrupt
> > lines
> >   * @buffer:		[DRIVER] any buffer present
> >   * @buffer_list:	[INTERN] list of all buffers currently attached
> > @@ -529,6 +530,7 @@ struct iio_dev {
> >  	int				modes;
> >  	int				currentmode;
> >  	struct device			dev;
> > +	struct list_head		ioctl_handlers;
> >  
> >  	struct iio_event_interface	*event_interface;
> >  

Powered by blists - more mailing lists

Powered by Openwall GNU/*/Linux Powered by OpenVZ