[geeks] Apple applications phoning home

[Shannon Hendrix]

On Oct 23, 2007, at 1:28 PM, der Mouse wrote:

> That should not be able to crash the kernel any more than how you move
> the mouse should be.  It's input from outside; it should be vetted for
> sanity before it's believed, and bad data should not cause a crash.

It's not just bad data: USB devices often ship with drivers that  
Apple didn't write, this they are part of the running kernel.

I'm not exactly sure how they interact with the USB IOKit layers.   
You would think the IOKit drivers for the USB hardware could provide  
an isolated interface for their party drivers, but obviously it isn't  
perfect.

In Apple's defense, no one else seems immune to USB issues so far  
either.  FreeBSD, Linux, and Solaris will also come down hard with  
certain USB situations, even on server hardware.

I've seen FreeBSD, Linux, and Solaris all crash instantly when USB  
devices wake up, or just when they are plugged in.

It seems that getting USB right is rather difficult.

Why is the USB controller hardware so fussy, and why is it able to  
have that much effect on system hardware?

-- 
"Where some they sell their dreams for small desires."