Re: [PATCH 2/2] wil6210: atomic I/O for the card memory

[Vladimir Kondratiev <qca_vkondrat@qca.qualcomm.com>]

On Friday, November 14, 2014 03:29:52 PM Kalle Valo wrote:
> Vladimir Kondratiev <qca_vkondrat@qca.qualcomm.com> writes:
> 
> > Introduce netdev IOCTLs, to be used by the debug tools.
> >
> > Allows to read/write single dword value or
> > memory block, aligned to dword
> > Different address modes supported:
> > - BAR offset
> > - Firmware "linker" address
> > - target's AHB bus
> >
> > Signed-off-by: Vladimir Kondratiev <qca_vkondrat@qca.qualcomm.com>
> 
> An ioctl interface for a wireless driver? IMHO that would have been ok
> in 2004, but not in 2014. Isn't there really better way to implement
> this?

Functionality implemented through ioctl is hardware access used for
debug/monitoring purposes. I did not found any alternative way to do this.

Any ideas? I need to implement read/write that takes address,
value (for write) and few options like address interpretation flavor
(hardware has several types of addresses)

> 
> > +/* Numbers SIOCDEVPRIVATE and SIOCDEVPRIVATE + 1
> > + * are used by Android devices to implement PNO (preferred network offload).
> > + * Albeit it is temporary solution, use different numbers to avoid conflicts
> > + */
> 
> Comments like this make me even more worried that that this is just yet
> another way to implement wext iwpriv interface.
It just happens wifi drivers used for android, already use these 2 ioctls
for PNO. And, this (PNO) is clearly something that should be done
with nl80211 instead of ioctl.