I sent a bug report some time ago to linux-kernel about the EHCI-HCD driver triggering an instant reboot. Greg KH suggested I post something here. Then I got side-tracked; well, finally...
I have a SIIG USB2.0+FireWire combo card (NEC USB2.0 EHCI rev 2/USB OHCI
rev 65, and VIA 1394 rev 67, sitting behind a Hint SB1-SE33 PCI-PCI
bridge) plugged into a Soltek DRV2 Athlon XP motherboard.
The OHCI driver works just fine with this card in USB 1 mode. Inserting
the EHCI-HCD driver (all released versions from kernels 2.4.19pre7
through 2.5.24) also works fine, but only for *short* transfers of at
most a few hundred bytes. Anything more than that, and the machine
instantly reboots, with no log entries appended (despite verbose debug
enabled). So,
linux:~# dd if=/dev/sda of=/dev/null
<Time passes...>
linux:~# modprobe ehci-hcd
linux:~# dd if=/dev/sda of=/dev/null
<instant reboot>
Since I gather other people have had good success with EHCI-HCD and the
NEC controller chip, I'm wondering if the driver is interacting with the
Hint PCI bridge in some odd fashion (else shared IRQs).
Attached are some files from /proc describing the hardware found, as
well as an excert from syslog with usb-storage verbose-debug enabled
(right up to the moment of instant-reboot).
Kris
proc-interrupts.txt.gz
Description: GNU Zip compressed data
proc-pci.txt.gz
Description: GNU Zip compressed data
syslog.txt.gz
Description: GNU Zip compressed data
