brcm80211 typedef bug

Greg KH gregkh at suse.de
Fri Sep 17 14:06:53 UTC 2010


On Fri, Sep 17, 2010 at 02:37:16AM -0700, Henry Ptasinski wrote:
> Our general rule is that explicitly-sized types are used when they
> relate to something that has a fixed size (register, frame field,
> etc), and natural types otherwise.

What do you mean by "explicitly-sized types" here?  It looks like you do
have an explicit sized type, but in reality, you don't.

> I'll have to do a little more digging on the 64-bit question. It
> certainly doesn't appear to be generally safe, but would be ok for
> some 64 bit platforms. (Ok on x86 family, not ok on sparc if I recall
> correctly).

Why would it be ok on x86?  What's the size of an unsigned int on a
64bit x86 kernel these days?  For some reason I thought it was 64bits,
but hm, maybe it's always 32?  I can't remember...

Oh yeah, I wrote that chapter in LDD3, unsigned int is ok, it's unsigned
long that gets messed up with the sizes.  Ok, this is fine, but why
didn't you all just use the correct kernel types in the first place?

thanks,

greg k-h



More information about the devel mailing list