[patch] rio500 devfs support

[patch] rio500 devfs support

[Gregory T. Norris]

[-- Attachment #1.1: Type: text/plain, Size: 292 bytes --]

The attached diff adds devfs support to the rio500 driver, so that
/dev/usb/rio500 gets created automagically.  It was generated against
2.4.5, but probably applies fine against any recent kernel.  Comments
are welcome (but be gentle, this is my first attempt at a kernel
patch :-).

Cheers!

[-- Attachment #1.2: rio500-devfs.diff --]
[-- Type: text/plain, Size: 1047 bytes --]

diff -urN linux-2.4.5.orig/drivers/usb/rio500.c linux-2.4.5/drivers/usb/rio500.c
--- linux-2.4.5.orig/drivers/usb/rio500.c	Mon Jun 18 17:10:39 2001
+++ linux-2.4.5/drivers/usb/rio500.c	Tue Jun 19 17:12:26 2001
@@ -38,6 +38,7 @@
 #include <linux/spinlock.h>
 #include <linux/usb.h>
 #include <linux/smp_lock.h>
+#include <linux/devfs_fs_kernel.h>
 
 #include "rio500_usb.h"
 
@@ -70,6 +71,7 @@
 };
 
 static struct rio_usb_data rio_instance;
+static devfs_handle_t rio500_devfs_handle;
 
 static int open_rio(struct inode *inode, struct file *file)
 {
@@ -492,6 +494,12 @@
 	if (usb_register(&rio_driver) < 0)
 		return -1;
 
+	rio500_devfs_handle = devfs_register(NULL, "usb/rio500",
+					DEVFS_FL_DEFAULT,
+					USB_MAJOR, RIO_MINOR,
+					S_IFCHR | S_IRUSR | S_IWUSR | S_IRGRP | S_IWGRP,
+					&usb_rio_fops, NULL);
+
 	info(DRIVER_VERSION " " DRIVER_AUTHOR);
 	info(DRIVER_DESC);
 
@@ -506,6 +514,7 @@
 	rio->present = 0;
 	usb_deregister(&rio_driver);
 
+	devfs_unregister(rio500_devfs_handle);
 
 }
 

[-- Attachment #2: Type: application/pgp-signature, Size: 232 bytes --]

Re: [patch] rio500 devfs support

[Gregory T. Norris]

[-- Attachment #1.1: Type: text/plain, Size: 668 bytes --]

Here's an updated version of the patch - the only real difference is
that rio500.c will printk an error message if devfs_register() fails. 
I left that out originally because devfs logs the error, but it's
probably a good idea to indicate which driver made the request.

Cheers!

On Tue, Jun 19, 2001 at 05:52:24PM -0500, Gregory T. Norris wrote:
> The attached diff adds devfs support to the rio500 driver, so that
> /dev/usb/rio500 gets created automagically.  It was generated against
> 2.4.5, but probably applies fine against any recent kernel.  Comments
> are welcome (but be gentle, this is my first attempt at a kernel
> patch :-).
> 
> Cheers!

[-- Attachment #1.2: rio500-devfs-02.diff --]
[-- Type: text/plain, Size: 1207 bytes --]

diff -urN linux-2.4.5.orig/drivers/usb/rio500.c linux-2.4.5/drivers/usb/rio500.c
--- linux-2.4.5.orig/drivers/usb/rio500.c	Sun Jun 24 16:29:35 2001
+++ linux-2.4.5/drivers/usb/rio500.c	Sun Jun 24 16:45:08 2001
@@ -38,6 +38,7 @@
 #include <linux/spinlock.h>
 #include <linux/usb.h>
 #include <linux/smp_lock.h>
+#include <linux/devfs_fs_kernel.h>
 
 #include "rio500_usb.h"
 
@@ -70,6 +71,7 @@
 };
 
 static struct rio_usb_data rio_instance;
+static devfs_handle_t rio500_devfs_handle;
 
 static int open_rio(struct inode *inode, struct file *file)
 {
@@ -492,6 +494,14 @@
 	if (usb_register(&rio_driver) < 0)
 		return -1;
 
+	rio500_devfs_handle = devfs_register(NULL, "usb/rio500",
+					DEVFS_FL_DEFAULT,
+					USB_MAJOR, RIO_MINOR,
+					S_IFCHR | S_IRUSR | S_IWUSR | S_IRGRP | S_IWGRP,
+					&usb_rio_fops, NULL);
+	if (rio500_devfs_handle == NULL)
+		printk(KERN_WARNING __FILE__ ": unable to register /dev/usb/rio500 with devfs\n");
+
 	info(DRIVER_VERSION " " DRIVER_AUTHOR);
 	info(DRIVER_DESC);
 
@@ -506,6 +516,8 @@
 	rio->present = 0;
 	usb_deregister(&rio_driver);
 
+	if (rio500_devfs_handle != NULL)
+		devfs_unregister(rio500_devfs_handle);
 
 }
 

[-- Attachment #2: Type: application/pgp-signature, Size: 232 bytes --]

Re: [patch] rio500 devfs support

[Richard Gooch]

Gregory T. Norris writes:
> 
> --qlTNgmc+xy1dBmNv
> Content-Type: multipart/mixed; boundary="0F1p//8PRICkK4MW"
> Content-Disposition: inline

Yuk! MIME!

> --0F1p//8PRICkK4MW
> Content-Type: text/plain; charset=us-ascii
> Content-Disposition: inline
> Content-Transfer-Encoding: quoted-printable

Horrors! Quoted-printables!

> Here's an updated version of the patch - the only real difference is
> that rio500.c will printk an error message if devfs_register() fails.=20
> I left that out originally because devfs logs the error, but it's
> probably a good idea to indicate which driver made the request.

No, it's a bad idea to test the error from devfs_register() unless you
*really* have a good reason. Most people who think they have a good
reason actually don't, they're just confused :-)

The reason you don't want to test the return value is that way the
driver works fine with CONFIG_DEVFS=n. Otherwise, you have a driver
that doesn't work with devfs, or you have to put ugly #ifdef's in the
code.

				Regards,

					Richard....
Permanent: rgooch@atnf.csiro.au
Current:   rgooch@ras.ucalgary.ca

Re: [patch] rio500 devfs support

[Gregory T. Norris]

[-- Attachment #1: Type: text/plain, Size: 654 bytes --]

Ok, back to the original version then.  Thanx!

On Mon, Jun 25, 2001 at 12:16:51AM -0600, Richard Gooch wrote:
> No, it's a bad idea to test the error from devfs_register() unless you
> *really* have a good reason. Most people who think they have a good
> reason actually don't, they're just confused :-)
> 
> The reason you don't want to test the return value is that way the
> driver works fine with CONFIG_DEVFS=n. Otherwise, you have a driver
> that doesn't work with devfs, or you have to put ugly #ifdef's in the
> code.
> 
> 				Regards,
> 
> 					Richard....
> Permanent: rgooch@atnf.csiro.au
> Current:   rgooch@ras.ucalgary.ca

[-- Attachment #2: Type: application/pgp-signature, Size: 232 bytes --]

Re: [patch] rio500 devfs support

[Greg KH]

[-- Attachment #1: Type: text/plain, Size: 570 bytes --]

On Tue, Jun 19, 2001 at 05:52:24PM -0500, Gregory T. Norris wrote:
> The attached diff adds devfs support to the rio500 driver, so that
> /dev/usb/rio500 gets created automagically.  It was generated against
> 2.4.5, but probably applies fine against any recent kernel.  Comments
> are welcome (but be gentle, this is my first attempt at a kernel
> patch :-).

Here's a small change that makes the node a child of the usb devfs
entry.  It also enables the node to only be present when the device is
actually present.  The patch is against 2.4.6-pre5.

thanks,

greg k-h

[-- Attachment #2: usb-rio500-2.4.6-pre5.patch --]
[-- Type: text/plain, Size: 2405 bytes --]

diff -Nru a/drivers/usb/rio500.c b/drivers/usb/rio500.c
--- a/drivers/usb/rio500.c	Mon Jun 25 10:31:03 2001
+++ b/drivers/usb/rio500.c	Mon Jun 25 10:31:03 2001
@@ -38,13 +38,14 @@
 #include <linux/spinlock.h>
 #include <linux/usb.h>
 #include <linux/smp_lock.h>
+#include <linux/devfs_fs_kernel.h>
 
 #include "rio500_usb.h"
 
 /*
  * Version Information
  */
-#define DRIVER_VERSION "v1.0.0"
+#define DRIVER_VERSION "v1.1"
 #define DRIVER_AUTHOR "Cesar Miquel <miquel@df.uba.ar>"
 #define DRIVER_DESC "USB Rio 500 driver"
 
@@ -60,6 +61,7 @@
 
 struct rio_usb_data {
         struct usb_device *rio_dev;     /* init: probe_rio */
+        devfs_handle_t devfs;           /* devfs device */
         unsigned int ifnum;             /* Interface number of the USB device */
         int isopen;                     /* nz if open */
         int present;                    /* Device is present on the bus */
@@ -69,6 +71,8 @@
 	struct semaphore lock;          /* general race avoidance */
 };
 
+extern devfs_handle_t usb_devfs_handle;	/* /dev/usb dir. */
+
 static struct rio_usb_data rio_instance;
 
 static int open_rio(struct inode *inode, struct file *file)
@@ -416,6 +420,15 @@
 	return read_count;
 }
 
+static struct
+file_operations usb_rio_fops = {
+	read:		read_rio,
+	write:		write_rio,
+	ioctl:		ioctl_rio,
+	open:		open_rio,
+	release:	close_rio,
+};
+
 static void *probe_rio(struct usb_device *dev, unsigned int ifnum,
 		       const struct usb_device_id *id)
 {
@@ -439,6 +452,14 @@
 	}
 	dbg("probe_rio: ibuf address:%p", rio->ibuf);
 
+	rio->devfs = devfs_register(usb_devfs_handle, "rio500",
+				    DEVFS_FL_DEFAULT, USB_MAJOR,
+				    RIO_MINOR,
+				    S_IFCHR | S_IRUSR | S_IWUSR | S_IRGRP |
+				    S_IWGRP, &usb_rio_fops, NULL);
+	if (rio->devfs == NULL)
+		dbg("probe_rio: device node registration failed");
+	
 	init_MUTEX(&(rio->lock));
 
 	return rio;
@@ -448,6 +469,8 @@
 {
 	struct rio_usb_data *rio = (struct rio_usb_data *) ptr;
 
+	devfs_unregister(rio->devfs);
+
 	if (rio->isopen) {
 		rio->isopen = 0;
 		/* better let it finish - the release will do whats needed */
@@ -461,15 +484,6 @@
 
 	rio->present = 0;
 }
-
-static struct
-file_operations usb_rio_fops = {
-	read:		read_rio,
-	write:		write_rio,
-	ioctl:		ioctl_rio,
-	open:		open_rio,
-	release:	close_rio,
-};
 
 static struct usb_device_id rio_table [] = {
 	{ USB_DEVICE(0x0841, 1) }, 		/* Rio 500 */

Re: [patch] rio500 devfs support

[Gregory T. Norris]

[-- Attachment #1: Type: text/plain, Size: 710 bytes --]

I was thinking of doing some similar updates this evening after work. 
Darn it... now I have to find something else to do!  :-)

Going by this morning's comments from Richard Gooch, it sounds like the

	if (rio->devfs == NULL)
		dbg("probe_rio: device node registration failed");

part should probably be removed... it looks good to me otherwise, tho. 
I'll try it out tonight and post the results.

Cheers!

On Mon, Jun 25, 2001 at 10:35:21AM -0700, Greg KH wrote:
> Here's a small change that makes the node a child of the usb devfs
> entry.  It also enables the node to only be present when the device is
> actually present.  The patch is against 2.4.6-pre5.
> 
> thanks,
> 
> greg k-h

[-- Attachment #2: Type: application/pgp-signature, Size: 232 bytes --]

Re: [patch] rio500 devfs support

[Greg KH]

On Mon, Jun 25, 2001 at 05:12:01PM -0500, Gregory T. Norris wrote:
> I was thinking of doing some similar updates this evening after work. 
> Darn it... now I have to find something else to do!  :-)
> 
> Going by this morning's comments from Richard Gooch, it sounds like the
> 
> 	if (rio->devfs == NULL)
> 		dbg("probe_rio: device node registration failed");

Just don't have it be using a call to "err()" like the printer.c file
does.  Loads of users wondering why they have an error message in their
logs, yet their printers still work.  That's why I made it dbg().

But yes, I agree that it doesn't have to be there at all.

greg k-h

Re: [patch] rio500 devfs support

[Gregory T. Norris]

[-- Attachment #1: Type: text/plain, Size: 130 bytes --]

It's working beautifully here.  I'll forward the patch to the
maintainer, since I have no idea if he's seen this thread.

Cheers!

[-- Attachment #2: Type: application/pgp-signature, Size: 232 bytes --]