Re: firmware loading

["km b" <kmb810@xxxxxxxxx>]

On 5/12/07, Matthew Dillon <dillon@apollo.backplane.com> wrote:

>     The root filesystem will not be mounted during driver attach so
>     the kernel will not have access to the filesystem.  But there are
>     a couple of other possibilities:
>
>     (1) You could embed the firmware as a const char [] array in the
>     source code itself.  It would depend how big the firmware image is...
>     if it's really big I'd rather not embed it in source code.
>
>     There might also be a copyright issue with the firmware that prevents
>     it from being embeddable in the source code.

i thought of it but i can't understand the implications of Intel's
firmware license.

>     (2) The firmware could be loaded into kernel memory as a module by
>     the boot loader.  The boot loader does have access to the root
>     filesystem (usually).  The driver would then be able to locate and
>     access it at attach time.
>
>     (3) There is a kernel firmware loading facility in
>     sys/kern/kern_firmware.c.  Again, the file cannot be loaded at
>     attach time but this way you could have an ioctl that just passes
>     the filename to the driver as part of the RC boot sequence
>     and the driver could then use the kern_firmware.c facility to
>     actually get an image map into kernel memory.
>
>     Method #1 or #3 is probably the best way to do it.  Method #2 requires
>     more configuration glue and changes in the loader config.

I think (3) is the best way to do it as it would relieve me from
writing the user-space utility.

cheers
kmb

--
Something is wrong up on cloud # 9!