Re: [PATCH RFC v2 3/3] drm/dp: Add a drm_aux-dev module for reading/writing dpcd registers.

[Rafael Antognolli <rafael.antognolli@intel.com>]

On Tue, Sep 22, 2015 at 02:17:51PM +0200, Daniel Vetter wrote:
> On Tue, Sep 22, 2015 at 03:00:54PM +0300, Ville Syrjälä wrote:
> > On Tue, Sep 15, 2015 at 04:55:04PM -0700, Rafael Antognolli wrote:
> > > This module is heavily based on i2c-dev. Once loaded, it provides one
> > > dev node per DP AUX channel, named drm_aux-N.
> > > 
> > > It's possible to know which connector owns this aux channel by looking
> > > at the respective sysfs /sys/class/drm_aux-dev/drm_aux-N/connector, if
> > > the connector device pointer was correctly set in the aux helper struct.
> > > 
> > > Two main operations are provided on the registers: read and write. The
> > > address of the register to be read or written is given using lseek.
> > > Reading or writing does not update the offset of the file.
> > > 
> > > Signed-off-by: Rafael Antognolli <rafael.antognolli@intel.com>
> > > ---
> > >  drivers/gpu/drm/Kconfig       |   4 +
> > >  drivers/gpu/drm/Makefile      |   1 +
> > >  drivers/gpu/drm/drm_aux-dev.c | 326 ++++++++++++++++++++++++++++++++++++++++++
> > >  3 files changed, 331 insertions(+)
> > >  create mode 100644 drivers/gpu/drm/drm_aux-dev.c
> > > 
> > > diff --git a/drivers/gpu/drm/Kconfig b/drivers/gpu/drm/Kconfig
> > > index 1a0a8df..eae847c 100644
> > > --- a/drivers/gpu/drm/Kconfig
> > > +++ b/drivers/gpu/drm/Kconfig
> > > @@ -25,6 +25,10 @@ config DRM_MIPI_DSI
> > >  	bool
> > >  	depends on DRM
> > >  
> > > +config DRM_AUX_CHARDEV
> > > +	tristate "DRM DP AUX Interface"
> > > +	depends on DRM
> > > +
> > >  config DRM_KMS_HELPER
> > >  	tristate
> > >  	depends on DRM
> > > diff --git a/drivers/gpu/drm/Makefile b/drivers/gpu/drm/Makefile
> > > index 45e7719..a1a94306 100644
> > > --- a/drivers/gpu/drm/Makefile
> > > +++ b/drivers/gpu/drm/Makefile
> > > @@ -32,6 +32,7 @@ CFLAGS_drm_trace_points.o := -I$(src)
> > >  
> > >  obj-$(CONFIG_DRM)	+= drm.o
> > >  obj-$(CONFIG_DRM_MIPI_DSI) += drm_mipi_dsi.o
> > > +obj-$(CONFIG_DRM_AUX_CHARDEV) += drm_aux-dev.o
> > >  obj-$(CONFIG_DRM_TTM)	+= ttm/
> > >  obj-$(CONFIG_DRM_TDFX)	+= tdfx/
> > >  obj-$(CONFIG_DRM_R128)	+= r128/
> > > diff --git a/drivers/gpu/drm/drm_aux-dev.c b/drivers/gpu/drm/drm_aux-dev.c
> > > new file mode 100644
> > > index 0000000..fcc334a
> > > --- /dev/null
> > > +++ b/drivers/gpu/drm/drm_aux-dev.c
> > > @@ -0,0 +1,326 @@
> > > +/*
> > > + * Copyright © 2015 Intel Corporation
> > > + *
> > > + * Permission is hereby granted, free of charge, to any person obtaining a
> > > + * copy of this software and associated documentation files (the "Software"),
> > > + * to deal in the Software without restriction, including without limitation
> > > + * the rights to use, copy, modify, merge, publish, distribute, sublicense,
> > > + * and/or sell copies of the Software, and to permit persons to whom the
> > > + * Software is furnished to do so, subject to the following conditions:
> > > + *
> > > + * The above copyright notice and this permission notice (including the next
> > > + * paragraph) shall be included in all copies or substantial portions of the
> > > + * Software.
> > > + *
> > > + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
> > > + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
> > > + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
> > > + * THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
> > > + * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING
> > > + * FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS
> > > + * IN THE SOFTWARE.
> > > + *
> > > + * Authors:
> > > + *    Rafael Antognolli <rafael.antognolli@intel.com>
> > > + *
> > > + */
> > > +
> > > +#include <linux/device.h>
> > > +#include <linux/fs.h>
> > > +#include <linux/slab.h>
> > > +#include <linux/init.h>
> > > +#include <linux/kernel.h>
> > > +#include <linux/module.h>
> > > +#include <asm/uaccess.h>
> > > +#include <drm/drm_dp_helper.h>
> > > +#include <drm/drm_crtc.h>
> > > +
> > > +struct drm_aux_dev {
> > > +	struct list_head list;
> > > +	unsigned index;
> > > +	struct drm_dp_aux *aux;
> > > +	struct device *dev;
> > > +};
> > > +
> > > +#define DRM_AUX_MINORS	256
> > > +static int drm_aux_dev_count = 0;
> > > +static LIST_HEAD(drm_aux_dev_list);
> > > +static DEFINE_SPINLOCK(drm_aux_dev_list_lock);
> > > +
> > > +static struct drm_aux_dev *drm_aux_dev_get_by_minor(unsigned index)
> > > +{
> > > +	struct drm_aux_dev *aux_dev;
> > > +
> > > +	spin_lock(&drm_aux_dev_list_lock);
> > > +	list_for_each_entry(aux_dev, &drm_aux_dev_list, list) {
> > > +		if (aux_dev->index == index)
> > > +			goto found;
> > > +	}
> > > +
> > > +	aux_dev = NULL;
> > > +found:
> > > +	spin_unlock(&drm_aux_dev_list_lock);
> > > +	return aux_dev;
> > > +}
> > > +
> > > +static struct drm_aux_dev *drm_aux_dev_get_by_aux(struct drm_dp_aux *aux)
> > > +{
> > > +	struct drm_aux_dev *aux_dev;
> > > +
> > > +	spin_lock(&drm_aux_dev_list_lock);
> > > +	list_for_each_entry(aux_dev, &drm_aux_dev_list, list) {
> > > +		if (aux_dev->aux == aux)
> > > +			goto found;
> > > +	}
> > > +
> > > +	aux_dev = NULL;
> > > +found:
> > > +	spin_unlock(&drm_aux_dev_list_lock);
> > > +	return aux_dev;
> > > +}
> > > +
> > > +static struct drm_aux_dev *get_free_drm_aux_dev(struct drm_dp_aux *aux)
> > > +{
> > > +	struct drm_aux_dev *aux_dev;
> > > +	int index;
> > > +
> > > +	spin_lock(&drm_aux_dev_list_lock);
> > > +	index = drm_aux_dev_count;
> > > +	spin_unlock(&drm_aux_dev_list_lock);
> > > +	if (index >= DRM_AUX_MINORS) {
> > > +		printk(KERN_ERR "i2c-dev: Out of device minors (%d)\n",
> > > +		       index);
> > > +		return ERR_PTR(-ENODEV);
> > > +	}
> > > +
> > > +	aux_dev = kzalloc(sizeof(*aux_dev), GFP_KERNEL);
> > > +	if (!aux_dev)
> > > +		return ERR_PTR(-ENOMEM);
> > > +	aux_dev->aux = aux;
> > > +	aux_dev->index = index;
> > > +
> > > +	spin_lock(&drm_aux_dev_list_lock);
> > > +	drm_aux_dev_count++;
> > > +	list_add_tail(&aux_dev->list, &drm_aux_dev_list);
> > > +	spin_unlock(&drm_aux_dev_list_lock);
> > > +	return aux_dev;
> > > +}
> > > +
> > > +static void return_drm_aux_dev(struct drm_aux_dev *aux_dev)
> > > +{
> > > +	spin_lock(&drm_aux_dev_list_lock);
> > > +	list_del(&aux_dev->list);
> > > +	spin_unlock(&drm_aux_dev_list_lock);
> > > +	kfree(aux_dev);
> > > +}
> > > +
> > > +static ssize_t name_show(struct device *dev,
> > > +			 struct device_attribute *attr, char *buf)
> > > +{
> > > +	struct drm_aux_dev *aux_dev = drm_aux_dev_get_by_minor(MINOR(dev->devt));
> > > +
> > > +	if (!aux_dev)
> > > +		return -ENODEV;
> > > +	return sprintf(buf, "%s\n", aux_dev->aux->name);
> > > +}
> > > +static DEVICE_ATTR_RO(name);
> > > +
> > > +static ssize_t connector_show(struct device *dev,
> > > +			      struct device_attribute *attr, char *buf)
> > > +{
> > > +	struct drm_aux_dev *aux_dev = drm_aux_dev_get_by_minor(MINOR(dev->devt));
> > > +	struct drm_dp_aux *aux;
> > > +	struct device *conn_dev;
> > > +	struct drm_connector *connector = NULL;
> > > +
> > > +	if (!aux_dev)
> > > +		return -ENODEV;
> > > +	aux = aux_dev->aux;
> > > +	conn_dev = aux->connector;
> > > +	if (!conn_dev)
> > > +		return sprintf(buf, "unknown\n");
> > > +
> > > +	connector = dev_get_drvdata(aux->connector);
> > > +
> > > +	return sprintf(buf, "%s\n", connector->name);
> > > +}
> > > +static DEVICE_ATTR_RO(connector);
> > > +
> > > +static struct attribute *drm_aux_attrs[] = {
> > > +	&dev_attr_name.attr,
> > > +	&dev_attr_connector.attr,
> > > +	NULL,
> > > +};
> > > +ATTRIBUTE_GROUPS(drm_aux);
> > > +
> > > +static int auxdev_open(struct inode *inode, struct file *file)
> > > +{
> > > +	unsigned int minor = iminor(inode);
> > > +	struct drm_aux_dev *aux_dev;
> > > +
> > > +	aux_dev = drm_aux_dev_get_by_minor(minor);
> > > +	if (!aux_dev)
> > > +		return -ENODEV;
> > > +
> > > +	file->private_data = aux_dev;
> > > +	return 0;
> > > +}
> > > +
> > > +static ssize_t auxdev_read(struct file *file, char __user *buf, size_t count,
> > > +			   loff_t *offset)
> > > +{
> > > +	char *localbuf;
> > > +	ssize_t res;
> > > +	struct drm_aux_dev *aux_dev = file->private_data;
> > > +
> > > +	localbuf = memdup_user(buf, count);
> > > +	if (IS_ERR(localbuf))
> > > +		return PTR_ERR(buf);
> > > +
> > > +	res = drm_dp_dpcd_read(aux_dev->aux, *offset, localbuf, count);
> > > +	if (res < 0)
> > > +		goto finish;
> > 
> > read/write will need to do the access in 16 byte chunks. That's the max
> > amount of data that can be transferred with a single AUX operation. I
> > suppose that would also avoid the need for allocating large buffers
> > since you can just have a 16 byte buffer on the stack and keep reusing
> > it while you iterate through the total count. I'm not sure what are
> > the expected semantics of read/write if you do it in chunks and hit
> > a copy_{to,from}_user() failure somewhere down the line. But I suppose
> > access_ok() upfront for the whole user buffer could avoid that issue.
> 
> Iirc short reads are ok in all cases, so we could even punt the restarting
> to userspace by just doing short reads/writes (like sockets do).
> 
> > Also the size of the DPCD address space is 1<<20 bytes, so you'll need to
> > limit the access to stay within that. 'size = lseek(fd, 0, SEEK_END)'
> > can be used to get the file size, so you should allow the offset to
> > reach 1<<20, but not beyond. read/write should process as much much of
> > the data that they can given the current file offset, and return how
> > many bytes made it in/out (can also be 0 if offset was already at 1<<20).
> 
> Hm, sounds like a few testcases for cornercases would be neat. Best would
> be to do that in igt and iterate over all available dp_aux dev nodes (igt
> is now generic, so makes sense to put that stuff there). Something like
> kms_dp_aux_dev would be a nice testcase name.

OK, I sent an updated patch set including all (I think) changes
suggested by you and Ville.

I'm working on the igt test, but I am not sure exactly what cornercases
to test, and specially how to test them. I can read a couple of
registers that I have an idea of what to expect from them, but not
corner cases. What do you have in mind?

Also, I am not sure how to identify that a given aux channel doesn't
have a display/panel attached to it, so even a read operation just fails
with a connection timeout. Should I just skip this aux channel on this
case, and not consider it a failure, when iterating over them?

Thanks,
Rafael
_______________________________________________
dri-devel mailing list
dri-devel@lists.freedesktop.org
http://lists.freedesktop.org/mailman/listinfo/dri-devel