1. creating proc entry and not saving pointer to PDE and checking it

   is not going to work.

2. if proc entry wasn't created, no reason to remove it on error path.
Signed-off-by: Alexey Dobriyan <adobriyan@gmail.com>

---
 drivers/isdn/hardware/eicon/divasmain.c |    1 -

 drivers/isdn/hardware/eicon/divasproc.c |    4 ++--

 2 files changed, 2 insertions(+), 3 deletions(-)
--- a/drivers/isdn/hardware/eicon/divasmain.c

+++ b/drivers/isdn/hardware/eicon/divasmain.c

@@ -806,7 +806,6 @@ static int DIVA_INIT_FUNCTION divas_init(void)
 	if (!create_divas_proc()) {

 #ifdef MODULE

-		remove_divas_proc();

 		divas_unregister_chrdev();

 		divasfunc_exit();

 #endif

--- a/drivers/isdn/hardware/eicon/divasproc.c

+++ b/drivers/isdn/hardware/eicon/divasproc.c

@@ -125,8 +125,8 @@ static const struct file_operations divas_fops = {
 int create_divas_proc(void)

 {

-	proc_create(divas_proc_name, S_IFREG | S_IRUGO, proc_net_eicon,

-		    &divas_fops);

+	divas_proc_entry = proc_create(divas_proc_name, S_IFREG | S_IRUGO,

+					proc_net_eicon, &divas_fops);

 	if (!divas_proc_entry)

 		return (0);
--

I don't know where you found this. I have look even in older versions, but

the pointer divas_proc_entry is set by proc_create(). The patch to

divasproc.c is wrong, it exists from the beginning of the driver.
Agreed here. The reason for doing remove_divas_proc() anyway, was to

make sure all possible settings done in !create_divas_proc() are

reverted. But since nothing else is done here, it is okay to remove it.
--

--

Ah, okay. So someone removed the pointer between 2.6.25 and 2.6.26 then.

In that case your patch is correct of course.
--