Skip to contentSkip to conversation

staging: comedi: replace ENOSYS by proper error codes

PATCH In review [email protected] 4 participants 6 messages Jun 12, 2015 – Jun 14, 2015 Staging

lore
Share & tools
lore
#0 lore raw

PATCH 1/1 staging: comedi: replace ENOSYS by proper error codes

From: Julien Dehee <[email protected]>

comedi/comedi_fops.c
use ENODEV following open manual

comedi/drivers.c
use ENOTTY following ioctl manual

drivers/serial2002.c
use ENOTTY following ioctl manual

---
 drivers/staging/comedi/comedi_fops.c        | 2 +-
 drivers/staging/comedi/drivers.c            | 2 +-
 drivers/staging/comedi/drivers/serial2002.c | 2 +-
 3 files changed, 3 insertions(+), 3 deletions(-)
drivers/staging/comedi/comedi_fops.c +1 -1
index 146ab00..6896a1f 100644
@@ -2611,7 +2611,7 @@ static int comedi_open(struct inode *inode, struct file *file)
 	}
 	if (dev->attached && dev->use_count == 0) {
 		if (!try_module_get(dev->driver->module)) {
-			rc = -ENOSYS;
+			rc = -ENODEV;
 			goto out;
 		}
 		if (dev->open) {
drivers/staging/comedi/drivers.c +1 -1
index ed0b60c..db89096 100644
@@ -820,7 +820,7 @@ int comedi_device_attach(struct comedi_device *dev, struct comedi_devconfig *it)
 			 "driver '%s' does not support attach using comedi_config\n",
 			 driv->driver_name);
 		module_put(driv->module);
-		ret = -ENOSYS;
+		ret = -ENOTTY;
 		goto out;
 	}
 	dev->driver = driv;
drivers/staging/comedi/drivers/serial2002.c +1 -1
index 83da162..929bf20 100644
@@ -101,7 +101,7 @@ static long serial2002_tty_ioctl(struct file *f, unsigned op,
 	if (f->f_op->unlocked_ioctl)
 		return f->f_op->unlocked_ioctl(f, op, param);
 
-	return -ENOSYS;
+	return -ENOTTY;
 }
 
 static int serial2002_tty_write(struct file *f, unsigned char *buf, int count)
Signature · 1 line
1.9.1
maintainer
#1 lore raw
Quoted 10 lines · On Fri, Jun 12, 2015 at 10:20:38PM +0200, [email protected] wrote:

From: Julien Dehee <[email protected]>

comedi/comedi_fops.c
use ENODEV following open manual

comedi/drivers.c
use ENOTTY following ioctl manual

drivers/serial2002.c
use ENOTTY following ioctl manual

What do you mean by "ioctl manual"?

And shouldn't this be 3 different patches as they do different things to different drivers?

thanks,

greg k-h

#2 lore raw
Quoted 5 lines · On Fri, 12 Jun 2015 13:43:27 -0700 Greg KH <[email protected]> wrote:
Quoted 4 lines · On Fri, Jun 12, 2015 at 10:20:38PM +0200, [email protected] wrote:

From: Julien Dehee <[email protected]>

comedi/comedi_fops.c
use ENODEV following open manual

That should probably be ENXIO by a strict reading, but Linux has always used ENODEV 8)

Quoted 8 lines
Quoted 6 lines

comedi/drivers.c
use ENOTTY following ioctl manual

drivers/serial2002.c
use ENOTTY following ioctl manual

What do you mean by "ioctl manual"?

man 2 ioctl

Unknown ioctls on a device should error with ENOTTY. It's one of those crazy pieces of Unix history.

maintainer
#3 lore raw
Quoted 19 lines · One Thousand Gnomes
Quoted 5 lines · On Fri, 12 Jun 2015 13:43:27 -0700 Greg KH <[email protected]> wrote:
Quoted 4 lines · On Fri, Jun 12, 2015 at 10:20:38PM +0200, [email protected] wrote:

From: Julien Dehee <[email protected]>

comedi/comedi_fops.c
use ENODEV following open manual

That should probably be ENXIO by a strict reading, but Linux has always used ENODEV 8)

Quoted 7 lines
Quoted 6 lines

comedi/drivers.c
use ENOTTY following ioctl manual

drivers/serial2002.c
use ENOTTY following ioctl manual

Quoted 6 lines
Quoted 1 line

What do you mean by "ioctl manual"?

man 2 ioctl

Unknown ioctls on a device should error with ENOTTY. It's one of those crazy pieces of Unix history.

Note that the man 2 ioctl interface is not always what the kernel exposes, but rather, what your libc exposes to other programs. So setting these kernel values might not change what you get all the way through, have you tested it?

thanks,

greg k-h

maintainer
#4 lore raw
Quoted 24 lines · Greg KH
Quoted 23 lines · One Thousand Gnomes
Quoted 5 lines · On Fri, 12 Jun 2015 13:43:27 -0700 Greg KH <[email protected]> wrote:
Quoted 4 lines · On Fri, Jun 12, 2015 at 10:20:38PM +0200, [email protected] wrote:

From: Julien Dehee <[email protected]>

comedi/comedi_fops.c
use ENODEV following open manual

That should probably be ENXIO by a strict reading, but Linux has always used ENODEV 8)

Quoted 8 lines
Quoted 6 lines

comedi/drivers.c
use ENOTTY following ioctl manual

drivers/serial2002.c
use ENOTTY following ioctl manual

What do you mean by "ioctl manual"?

man 2 ioctl

Quoted 7 lines
Quoted 2 lines

Unknown ioctls on a device should error with ENOTTY. It's one of those crazy pieces of Unix history.

Note that the man 2 ioctl interface is not always what the kernel exposes, but rather, what your libc exposes to other programs. So setting these kernel values might not change what you get all the way through, have you tested it?

We also have the issue of the existing userspace comedi code, I don't know if it's handling the ioctl errors differently than ioctl(2) says, but that should be checked as well before I can take this patch.

thanks,

greg k-h

#5 lore raw
In reply to [email protected]
Quoted 23 lines · On 12/06/15 21:20, [email protected] wrote:

From: Julien Dehee <[email protected]>

comedi/comedi_fops.c
use ENODEV following open manual

comedi/drivers.c
use ENOTTY following ioctl manual

drivers/serial2002.c
use ENOTTY following ioctl manual

---
  drivers/staging/comedi/comedi_fops.c        | 2 +-
  drivers/staging/comedi/drivers.c            | 2 +-
  drivers/staging/comedi/drivers/serial2002.c | 2 +-
  3 files changed, 3 insertions(+), 3 deletions(-)
drivers/staging/comedi/comedi_fops.c +0 -0
index 146ab00..6896a1f 100644

drivers/staging/comedi/comedi_fops.c static int comedi_open(struct inode *inode, struct file *file) · line 2615

Quoted 6 lines
drivers/staging/comedi/comedi_fops.c +1 -1
@@ -2611,7 +2611,7 @@ static int comedi_open(struct inode *inode, struct file *file)
  	}
  	if (dev->attached && dev->use_count == 0) {
  		if (!try_module_get(dev->driver->module)) {
-			rc = -ENOSYS;
+			rc = -ENODEV;
  			goto out;

Seems reasonable as driver for the device is being unloaded.

drivers/staging/comedi/drivers.c int comedi_device_attach(struct comedi_device *dev, struct comedi_devconfig *it) · line 824

Quoted 13 lines
drivers/staging/comedi/comedi_fops.c +0 -0
  		}
  		if (dev->open) {
drivers/staging/comedi/drivers.c +1 -1
index ed0b60c..db89096 100644
@@ -820,7 +820,7 @@ int comedi_device_attach(struct comedi_device *dev, struct comedi_devconfig *it)
  			 "driver '%s' does not support attach using comedi_config\n",
  			 driv->driver_name);
  		module_put(driv->module);
-		ret = -ENOSYS;
+		ret = -ENOTTY;
  		goto out;

I don't think ENOTTY is the correct error code here. the COMEDI_DEVCONFIG ioctl code is an "appropriate" ioctl to send to a comedi device. I'd suggest ENOTSUP as a replacement.

drivers/staging/comedi/drivers/serial2002.c static long serial2002_tty_ioctl(struct file *f, unsigned op, · line 104

Quoted 12 lines
drivers/staging/comedi/drivers.c +0 -0
  	}
  	dev->driver = driv;
drivers/staging/comedi/drivers/serial2002.c +1 -1
index 83da162..929bf20 100644
@@ -101,7 +101,7 @@ static long serial2002_tty_ioctl(struct file *f, unsigned op,
  	if (f->f_op->unlocked_ioctl)
  		return f->f_op->unlocked_ioctl(f, op, param);
 
-	return -ENOSYS;
+	return -ENOTTY;

That is reasonable. It shouldn't happen anyway if 'f' really is linked to a tty device. And the return value is currently ignored anyway!

drivers/staging/comedi/drivers/serial2002.c static long serial2002_tty_ioctl(struct file *f, unsigned op, · line 107

Quoted 4 lines
drivers/staging/comedi/drivers/serial2002.c +0 -0
  }
 
  static int serial2002_tty_write(struct file *f, unsigned char *buf, int count)

As Greg mentioned, it should be three different patches.

Signature · 2 lines
-=( Ian Abbott @ MEV Ltd.    E-mail: <[email protected]> )=-
-=(                          Web: http://www.mev.co.uk/  )=-