Bob, thank you for this exhaustive reply!
1) Are there Wi-Fi adapters which do not need firmware? I guess there are if manufacturer does not use semi-general purpose hardware? 2) Is the RAM built into the Wi-Fi card chipset? If I inspect my Ralink W-Fi card, which loads the 4096 byte /lib/firmware/rt2870.bin firmware file once the interface is brought up, then I don't see a separate RAM chip: https://www.dropbox.com/s/9z5rir3byfr9sg2/TEW-624UB-front.png https://www.dropbox.com/s/8p4iw8dpnfiphqb/TEW-624UB-back.png So in this case the RAM which stores the rt2870.bin firmware file is probably part of the RT2870F chip? regards, Martin 2013/1/2 Bob Proulx <b...@proulx.com>: > Martin T wrote: >> some Wi-Fi adapters(for example Intel ipw2200 family and many Ralink >> cards) require both firmware and drivers in order to operate properly. > > Yes. > >> 1) As I understand, firmware is usually a closed-source binary image >> provided by Wi-Fi card manufacturer? > > Yes. And because of this it doesn't meet the Debian Free Software > Guidelines and can't be part of Debian. Packages of this firmware > can only go in the non-free repository. > >> 2) What happens with the firmware when card becomes operational? I >> mean by definition it should be written to device non-volatile >> memory(for example flash memory), but I doubt that this is the case >> for Wi-Fi adapters.. Or is it? > > No. The image is simply loaded into the adapter's ram. After the > device loses power the memory evaporates. When power is applied again > the device is once again blank or back to the default power on state > and the firmware must be loaded again. > >> 3) Last but not least, which aspects are controlled by Wi-Fi card >> firmware and what role plays the Wi-Fi adapter driver? > > Most devices today are designed using semi-general purpose hardware > which includes a firmware part to configure and program the device. > Vendors do this because it allows them to ship hardware early and to > fix hardware bugs by making changes to the firmware later. They can > do this very late and often after shipping the product. They simply > update the device driver and put it on their web site and their > customers load the new driver. This can effectively fix an > incorrectly designed piece of hardware in the field. And that is why > vendors like to design devices that use a firmware component. > > That they are programmable does not make them a computer. They are > usually simply logic devices with memory to program the operation of > them. There are various techniques such as state machines implemented > with FPGAs and so forth. Programmable state machines almost always > play a huge part in any electronic chip design. > > There isn't a simple rule that says every adapter behaves a certain > way and the firmware does a certain thing on every one of them. Every > vendor designs their device independently. Every one of them may be > different. Which makes it hard to produce a generally correct answer. > All that can really be said is that the firmware is used by the device > to set itself up for the operation it was designed to do. > > After loading the device firmware then the kernel device driver has > the task of interacting with the device. It provides the software > layer that translates between a physical hardware device and the > kernel's software. > > Bob > Speaking as a past chip designer. -- To UNSUBSCRIBE, email to debian-user-requ...@lists.debian.org with a subject of "unsubscribe". Trouble? Contact listmas...@lists.debian.org Archive: http://lists.debian.org/cajx5yvfdncsygshjs-1dp6uqun7ampcbc7jurzcxlhjlp8x...@mail.gmail.com
