unlocked_ioctl explanation

unlocked_ioctl explanation

[Ezequiel García]

I am aware that ioctl has been superseeded by unlocked_ioctl. I've been looking through patches and this seems to be BKL related.

Could someone explain further the differences and reasons for the new ioctl prototype and name?

I guess 'unlocked' means BKL is no longer held inside ioctl, right? 

What precautions should we take when using unlocked_ioctl ?

Greetings,
Ezequiel.
-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://lists.kernelnewbies.org/pipermail/kernelnewbies/attachments/20110517/c5eeba97/attachment.html 

unlocked_ioctl explanation

[Dave Hylands]

Hi Ezequiel,

2011/5/17 Ezequiel Garc?a <elezegarcia@yahoo.com.ar>
>
> I am aware that ioctl has been superseeded by unlocked_ioctl. I've been looking through patches and this seems to be BKL related.
>
> Could someone explain further the differences and reasons for the new ioctl prototype and name?
>
> I guess 'unlocked' means BKL is no longer held inside ioctl, right?

Exaclty.

> What precautions should we take when using unlocked_ioctl ?

Well, since there is no BKL, your ioctl can get preempted by another
process and that other process just might happen to make an ioctl call
to your driver.

So your driver has to protect itself from being called by 2 threads
simultaneously, if that's a problem.

Some drivers won't need any extra protection. Unfortunately, it really
depends on exactly what your driver code does.

--
Dave Hylands
Shuswap, BC, Canada
http://www.davehylands.com

unlocked_ioctl explanation

[Ezequiel García]

Thanks Dave for your answer. I guess the same question has been answered several times before. For the interest reader, I've found these:

http://unix.stackexchange.com/questions/4711/what-is-the-difference-between-ioctl-unlocked-ioctl-and-compat-ioctl

http://lwn.net/Articles/119652/

http://www.mail-archive.com/kernelnewbies at nl.linux.org/msg00269.html

I have a remaining question, though. In the lwn article I read this:

"The ioctl() system call has long been out of favor among the
kernel developers, who see it as a completely uncontrolled entry point into
the kernel".

Is this pointing that ioctl() is planning to get removed from the kernel ? In that case what's the currently preferred mechanism for giving 'control commands' to device drivers ?

Thanks again and greetings,
Ezequiel.

--- El mi? 18-may-11, Dave Hylands <dhylands@gmail.com> escribi?:

De: Dave Hylands <dhylands@gmail.com>
Asunto: Re: unlocked_ioctl explanation
Para: "Ezequiel Garc?a" <elezegarcia@yahoo.com.ar>
Cc: kernelnewbies at kernelnewbies.org
Fecha: mi?rcoles, 18 de mayo de 2011, 3:20

Hi Ezequiel,

2011/5/17 Ezequiel Garc?a <elezegarcia@yahoo.com.ar>
>
> I am aware that ioctl has been superseeded by unlocked_ioctl. I've been looking through patches and this seems to be BKL related.
>
> Could someone explain further the differences and reasons for the new ioctl prototype and name?
>
> I guess 'unlocked' means BKL is no longer held inside ioctl, right?

Exaclty.

> What precautions should we take when using unlocked_ioctl ?

Well, since there is no BKL, your ioctl can get preempted by another
process and that other process just might happen to make an ioctl call
to your driver.

So your driver has to protect itself from being called by 2 threads
simultaneously, if that's a problem.

Some drivers won't need any extra protection. Unfortunately, it really
depends on exactly what your driver code does.

--
Dave Hylands
Shuswap, BC, Canada
http://www.davehylands.com

_______________________________________________
Kernelnewbies mailing list
Kernelnewbies at kernelnewbies.org
http://lists.kernelnewbies.org/mailman/listinfo/kernelnewbies
-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://lists.kernelnewbies.org/pipermail/kernelnewbies/attachments/20110519/4d8baec5/attachment.html 

unlocked_ioctl explanation

[Daniel Baluta]

2011/5/19 Ezequiel Garc?a <elezegarcia@yahoo.com.ar>
>
> Thanks Dave for your answer. I guess the same question has been answered several times before. For the interest reader, I've found these:
>
> http://unix.stackexchange.com/questions/4711/what-is-the-difference-between-ioctl-unlocked-ioctl-and-compat-ioctl
>
> http://lwn.net/Articles/119652/
>
> http://www.mail-archive.com/kernelnewbies at nl.linux.org/msg00269.html
>
> I have a remaining question, though. In the lwn article I read this:
>
> "The ioctl() system call has long been out of favor among the kernel developers, who see it as a completely uncontrolled entry point into the kernel".
>
> Is this pointing that ioctl() is planning to get removed from the kernel ? In that case what's the currently preferred mechanism for giving 'control commands' to device drivers ?

I would bet for 'sysfs'.

thanks,
Daniel.