[PATCH] remove generic symlink from sg

[PATCH] remove generic symlink from sg

[James Bottomley]

This has been redundant ever since James Smart added the class back
links since every scsi_device now gets one of these of the form
scsi_generic:sg<n>.

I'm open to going through the deprecated features business for this,
since this would represent a user visible change of the scsi sysfs
interface if you wish.

James

diff --git a/drivers/scsi/sg.c b/drivers/scsi/sg.c
--- a/drivers/scsi/sg.c
+++ b/drivers/scsi/sg.c
@@ -1471,11 +1471,6 @@ sg_add(struct class_device *cl_dev, stru
 			printk(KERN_WARNING "sg_add: "
 				"class_device_create failed\n");
 		class_set_devdata(sg_class_member, sdp);
-		error = sysfs_create_link(&scsidp->sdev_gendev.kobj, 
-					  &sg_class_member->kobj, "generic");
-		if (error)
-			printk(KERN_ERR "sg_add: unable to make symlink "
-					"'generic' back to sg%d\n", k);
 	} else
 		printk(KERN_WARNING "sg_add: sg_sys INvalid\n");
 
@@ -1544,7 +1539,6 @@ sg_remove(struct class_device *cl_dev, s
 	write_unlock_irqrestore(&sg_dev_arr_lock, iflags);
 
 	if (sdp) {
-		sysfs_remove_link(&scsidp->sdev_gendev.kobj, "generic");
 		class_device_destroy(sg_sysfs_class, MKDEV(SCSI_GENERIC_MAJOR, k));
 		cdev_del(sdp->cdev);
 		sdp->cdev = NULL;

Re: [PATCH] remove generic symlink from sg

[Douglas Gilbert]

James Bottomley wrote:
> This has been redundant ever since James Smart added the class back
> links since every scsi_device now gets one of these of the form
> scsi_generic:sg<n>.

James,
Careful with the spin:-) It is arguable which one is
redundant. I am testing this patch and it breaks
lsscsi and sg_map26 (again). Could you tell me which
version of the lk 2.6 series the scsi_generic:sg<n>
link appeared in? Because if I switch my code to
search for a scsi_generic:sg<n> symlink rather than
following the generic symlink then I will break
backward compatibility prior to that version (unless
I cope with both). At least one major distro has an
"enterprise" edition based on lk 2.6.5

> I'm open to going through the deprecated features business for this,
> since this would represent a user visible change of the scsi sysfs
> interface if you wish.

I can see from my lsscsi code and the st driver that
tape devices still have a plain old "tape" symlink.
Should I be expecting that one to break in the near
future?

Doug Gilbert

Re: [PATCH] remove generic symlink from sg

[James Bottomley]

On Wed, 2006-03-22 at 10:25 -0500, Douglas Gilbert wrote:
> Careful with the spin:-) It is arguable which one is
> redundant. I am testing this patch and it breaks
> lsscsi and sg_map26 (again). Could you tell me which
> version of the lk 2.6 series the scsi_generic:sg<n>
> link appeared in? Because if I switch my code to
> search for a scsi_generic:sg<n> symlink rather than
> following the generic symlink then I will break
> backward compatibility prior to that version (unless
> I cope with both). At least one major distro has an
> "enterprise" edition based on lk 2.6.5

It was this patch

76d1ce00bdd76c2987fbfb763cd40447413a55b3
[PATCH] Driver core: link device and all class devices derived from it.

which added it on 10 July 05 according to the log.

I'd suggest not checking for version, but instead check for
scsi_generic: as the beginning of the link and then fallback if it's not
found.

> > I'm open to going through the deprecated features business for this,
> > since this would represent a user visible change of the scsi sysfs
> > interface if you wish.
> 
> I can see from my lsscsi code and the st driver that
> tape devices still have a plain old "tape" symlink.
> Should I be expecting that one to break in the near
> future?

Presumably if that link is likewise redundant.

James