On 17.09.2012 17:30, Chlon Michaël wrote: > Good Evening, > > > Well, in fact, it is *an issue with shared IRQ*. > I wonder if there is a mean to pass to the kernel a fixed IRQ for > this PCI device, at boot for example. I haven't find anything ... > I have also stop my server, change physically the port of the > WiFi card, but after reboot it's remain shared ... with an > another IRQ, but shared! :'( > Here there is: > > $ cat /proc/interrupts > 21: 0 0 0 0 0 0 0 0 IO-APIC-fasteoi uhci_hcd:usb4, uhci_hcd:usb7, ath
Ok. Please provide lspci -vvvxxx output for this device. Take a look at lspci, note the device number (left column), and do lspci -vvvxxx -s <device> like, for a random my device, "lspci -xxxvvv -s 02:00.0", and send the result. It might be that the device itself does not support irq sharing in this mode. qemu-kvm wants devices which either support MSI (in this case everything should just work, and is fast, if the device is not buggy to start with), or by disabling using legacy INTx by config register setup. You can check if the device supports the 2nd mode by writing bit 10 (0x0400) to CONTROL register using setpci and verifying (using setpci again) it is set by reading it again. For this: setpci -s <device> CONTROL this will return a hexadecimal number. Or it with 0x0400, eg, like this: printf "%x\n" $((0x<prev_value> | 0x0400)) and write it back: setpci -s <device> CONTROL=<newvalue> and read it again: setpci -s <device> CONTROL If the new value is the same what you set above, it has some chance to work. Now, There are two parameters for -device pci-assign -- prefer_msi and share_intx. You can experiment with the two - prefer_msi is now off, but if your device supports msi, you can turn it on. share_intx is also an on/off switch. Please post your findings. Thanks, /mjt -- To UNSUBSCRIBE, email to debian-bugs-dist-requ...@lists.debian.org with a subject of "unsubscribe". Trouble? Contact listmas...@lists.debian.org
