your RELENG_5 usb patches (was: Re: cvs commit: src/sys/dev/usb ehci.c ehci_pci.c ehcivar.h)

[Ian Dowse]

In message <20060215190920.GA942 at saturn.kn-bremen.de>, Juergen Lock writes:
>I finally got around testing this patch on today's RELENG_5.  I built
>my custom kernel (mostly based on GENERIC but with ehci), booted that,
>copied the kernel to an 16 MB sdcard i had lying around using an usb
>multi-cardreader, which worked.  umounted, disconnected, reconnected,
>mounted again and verified the kernel, which worked as well.  Then I
>noticed this in dmesg:
>	da3: 1.000MB/s transfers
>so it wasn't using ehci apparently.  I thought, maybe this is
>intermittent too like with my externally powered devices, so I
>disconnected and reconnected again, but this time da3s1 didn't appear.
>(dmesg also shows da1, sometimes with 1 and sometimes with 40 MB/s,
>but I never saw a /dev/da1s1 appear.)  I wanted to try again, and when
>disconnecting I got the following panic:

I'm not sure about the intermittent probing behaviour, but there
are known bugs in CAM that cause it to crash in camisr() if a device
goes away while being probed. There have been some improvements in
this area in -CURRENT, so if you like you could try a CAM patch
too:

	http://people.freebsd.org/~iedowse/cam_remove_releng_5.diff

Note that this is just a blind backport - I haven't even tested
that it compiles.

Ian