| |||||||||||||
Wow?Wow?Posted Sep 28, 2008 7:59 UTC (Sun) by johill (subscriber, #25196)In reply to: Wow? by tialaramex Parent article: Atheros releases ath5k HAL code
You're going to have to back up that claim a bit more than with handwaving, and I think you're talking about the 6k fullmac device rather than the 5k (and now 9k) halfmac device, because I'm fairly certain they neither include a ROM nor a generic MAC processor but rather contain the MAC in the HDL.
(Log in to post comments)
Wow? Posted Sep 28, 2008 11:02 UTC (Sun) by tialaramex (subscriber, #21167) [Link]
Atheros themselves don't seem to use the 5k / 6k / 9k nomenclature, but assuming that this isn't one of those situations (like with the Radeon GPUs we've also touched on in this thread) where a 6k can be a device numbered beginning with a 5, while a 9k is one beginning with a 3, then I think the Atheros pages speak for themselves, here's the statement used in the glossy for each of the 5xxx series designs I could be bothered to look at.
"A single driver and firmware code base supports all Atheros chipsets"
Let me reiterate, there is no reason in general why we should care about firmware or indeed care whether a binary blob we find on a device is in fact firmware (as opposed to a manufacturer's serial number, built-in presets, a look-up table to simplify some mathematics, or a vanity logo). There's no reason why the host driver should care how the host interface is implemented - what we need, to secure Free Software, is just the host interface itself, and Atheros, like Intel have done enough to provide us with that (though doubtless Theo doesn't agree with me). For all I know, the AHCI implementation in my laptop is just a painstaking gate-level design from scratch, while the implementation in the chipset for my server actually uses a tiny embedded microprocessor and forty thousand lines of C++. I doubt it in both cases, but more importantly I don't see any reason we should care - they both speak AHCI just fine.
Wow? Posted Sep 28, 2008 16:24 UTC (Sun) by johill (subscriber, #25196) [Link]
Yeah, I don't know what numbering Atheros uses, it's very confusing.
Anyhow, I agree with you, it doesn't matter much to me; although we have almost successfully reverse engineered Broadcom's firmware, which isn't possible if it's burned into a ROM. It just seems weird to me to burn it into ROM if you could save that cost which is why I always believed it was simply in the HDL. But yeah, it doesn't matter much since you can always treat firmware+cpu as one thing and combine it into new HDL too.
Wow? Posted Sep 29, 2008 20:15 UTC (Mon) by proski (subscriber, #104) [Link] We should not care about firmware that doesn't need to be distributed (unless we are trying to improve it or hack the hardware). But we should care about the firmware that needs to be loaded by the host system. That firmware needs to be distributed with the operating system. Operating systems have different policies, which may be incompatible with the license for the firmware. Even if the firmware is allowed to be copied freely, it may still be incompatible with a distro that aims to provide sources for every binary, whether it runs on the host CPU or not.Atheros cards would satisfy every distribution because no blobs need to be distributed at all. Sure, there are tables for register initialization in the driver, but I don't think anyone would treat them as binary blobs.
|
|||||||||||||