[lkml]   [2011]   [May]   [21]   [last100]   RSS Feed
Views: [wrap][no wrap]   [headers]  [forward] 
Messages in this thread
SubjectRe: [PATCH 00/23] add register_chrdev_ids() to char_dev.c, API
On Thu, May 19, 2011 at 4:44 PM, Greg KH <> wrote:
> On Thu, May 19, 2011 at 03:33:03PM -0600, Jim Cromie wrote:
>> over on kernelnewbies, gregkh said:
>>      The chardev stuff is a mess, I keep meaning for years to clean it
>>      up.  Any proposals on a sane interface for this stuff is greatly
>>      appreciated.
>> this is a 1st step.
>> register_chrdev_ids() replaces and deprecates register_chrdev_region()
>> and alloc_chrdev_region() with a single function that works for both
>> dynamic and static major numbers.
>> Like alloc_chrdev_region(), 1st arg is a dev_t*, but its an in/out
>> parameter, and expects both major and minor to be preset, and thus the
>> separate minor arg is dropped.  If major == 0, a dynamic major is
>> reserved, saved into 1st arg, and thus available to caller afterwards.
>> [PATCH 01/23] add register_chrdev_ids() to char_dev.c, API
>> [PATCH 02/23] reimplement alloc_chrdev_region with
>> [PATCH 03/23] use register_chrdev_ids to replace
>> [PATCH 04/23] use register_chrdev_ids in drivers/tty/
>> [PATCH 05/23] use register_chrdev_ids in drivers/infiniband/
>> [PATCH 06/23] use register_chrdev_ids in drivers/media/
>> [PATCH 07/23] use register_chrdev_ids in drivers/s390/
>> [PATCH 08/23] use register_chrdev_ids in drivers/scsi/
>> [PATCH 09/23] use register_chrdev_ids in drivers/staging/
>> Ive held back the rest, no point in spamming.
> It's a nice first step, but that's the easy part, what is your 2nd
> through 4th one going to be?  :)
> I'd also like to sanatize the function namespace a bit as well, how
> about chrdev_register_ids() instead?

that seems sensible, modern.
also have register_chrdev(), which I presume should also be fixed.

> Ideally, we could drop down to a single register/unregister pair of
> functions, that are easy to use and understand.

__register_chrdev() does more stuff, mainly around cdevs, fops.
If fops was passed as NULL, we just do the __register_chardev_region()
and return early, skipping the cdev_alloc() and everything afterwards,
thus yielding register_chrdev_ids() behavior.

> Do you think you can
> get there with this intermediate step or do you want to step back and
> rethink this?

hmm. If above is right, theres no need for the new api fn I added,
and probably should also drop the __ on both (un)?register_chardev.
So thats step 2 :) Any ideas for 3 ?

btw, I think theres a major/minor error in the linuxdoc for the count
param in some of these register-* functions. I'll take a closer look,
and send a patch RSN if needed, even if fn is going away later.

> thanks,
> greg k-h
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to
More majordomo info at
Please read the FAQ at

 \ /
  Last update: 2011-05-21 07:17    [W:0.068 / U:4.428 seconds]
©2003-2020 Jasper Spaans|hosted at Digital Ocean and TransIP|Read the blog|Advertise on this site