J?rgen Keil wrote: >>We can use lspci to verify the bug (with a card inserted in the lower socket): >># lspci -M >> >> > >lspci? Isn't that a linux utility? > >Neither my solaris 10 x86 nor my snv_25 x86 (+ bfu snv_26) has it. > > > I'll send you a lspci package offline because of the attachment size.
>Now that I just found out that hotpluging the cardbus card first into pcic#0 >socket then into pcic#1 socket works, I checked the pci devices with >/usr/X11/bin/scanpci -v: in both cases the firewire card is detected at >pci bus/device/function 3/0/0. Hmm, I guess I now need a second cardbus >card, to check if pci bus #4 is used when one cardbus slot is in use and the >other one receives a socket insert event. :-) > >Attached is a /var/adm/messages file when moving the firewire cardbus card >between pcic#0 -> pcic#1 -> pcic#0 -> pcic#1. > >Why does this work? After moving the card from pcic#0 -> pcic#1 >both bridges should be configured for pci bus #3, correct? >Is the pcic#0 cardbus bridge disabled after a card remove event, so >that when the card is inserted on pcic#1 we can use pci bus 3 because >only pcic#1 is enabled at this time? > > Exactly. When the card is removed, the bus number is cleared on the bridge. The nexus driver is doing "dynamic bus numbering" for its children. This can be problematic. Since the socket isn't removable, only the card is removable, we'd better use fixed bus numbers. You may use lspci to demonstrate the bus number problem. I'll fix this in a few days. Vincent.
