Re: [PATCH 7/9] drivers/video: Correct code taking the size of a

Hi Andrew,

thanks for your comment (and for taking the patch).
You can count it as Acked-by:me as it is technically correct. I also 
tested it but as I don't know any program that uses this interface (and 
not wanting to write my own as I do not want to encourage anyone to use 
this interface) I can't say whether it now works as intended.
However this patch is certainly a step in the right direction as writing 
random numbers to hardware is rarely correct :-)

Andrew Morton schrieb:

On Sun, 13 Dec 2009 13:52:59 +0100 Florian Tobias Schandinat [off-list ref] wrote:

quoted

Hi Julia, Andrew,

Julia Lawall schrieb:

quoted

From: Julia Lawall <redacted>

sizeof(viafb_gamma_table) is just the size of the pointer.  This is changed
to the size used when calling kmalloc to initialize the pointer.

this is the second patch addressing this issue ending up in my mailbox. 
At least this one is technically correct so feel free to upstream it.
However I vote for removing this ioctl hell from viafb as most of them 
duplicate framebuffer functionality or have unknown (not clearly 
defined) functionality or at least solve a generic problem with a custom 
ioctl (which I consider bad). I had a patch ready to move this stuff to 
an extra file and print a warning that it is subject to be removed. I 
feel a bit uncomfortable about repairing broken stuff prior to removing it.
Any comments on this subject?

I favour both repairing and removing broken stuff ;)

We may as well fix it if problems are known.  Perhaps someone is
hitting the problem at runtime in an older kernel and needs a patch to
backport.  Perhaps we later decide to revert the removal, thus
reinstating the known bug.

You are right.
I'll try to send a patch that at least labels these ioctls 
unstable/depreciated as soon as time permits so that nobody will start 
using them.


Regards,

Florian Tobias Schandinat