(Sorry for the delay.) Sarah Sharp wrote: > On Wed, Jun 09, 2010 at 11:16:10AM +0200, Clemens Ladisch wrote: > > MSI is just a minor variation of a normal interrupt, but [my old patch] > > cannot handle MSI-X as it isn't known how many interrupts are acceptable, > > or what the dev parameters in the request_irq() calls should be. > > Hmm, why not add new API in the same spot to ask the host how many MSI-X > interrupters it can support? If it says zero, then you can see if it > supports MSI with a call to is_pci_msi_allowed(). The kernel might not be able to allocate that many interrupts, so in theory, the pci_enable_msi(x) calls belong into the HCD so that the driver can negotiate with the kernel how many interrupts to actually use. > > It appears that common MSI code is easy, but that implementing MSI-X > > still requires replacing most of the generic hcd interrupt handling code > > (either with a callback or by replacing the interrupt later). > > Perhaps the HCD could provide an MSI/MSI-X call back function? That > would fit with Dong's current patch. Yes, the HC driver should be able to decide whether to use MSI-X or MSI. > There's also issues with mapping IRQ to internal data structures. The > xHCI host is going to need to be able to map the irq number for the > MSI-X vector into a specific event ring that needs to be serviced. If the HCD does the request_irq calls by itself, it can just use whatever dev parameter it likes. I've decided not to add another callback but to leave this code in the start callback. These patches are on top of Dong's; the xHCI stuff is untested. drivers/usb/core/hcd-pci.c | 36 +++++++-- drivers/usb/core/hcd.c | 5 - drivers/usb/host/ehci-pci.c | 72 ++++++++++++++++-- drivers/usb/host/ohci-pci.c | 51 +++++++++++++ drivers/usb/host/xhci-pci.c | 83 +++++++++++++++++++++ drivers/usb/host/xhci.c | 137 +++++------------------------------- include/linux/usb/hcd.h | 11 ++ 7 files changed, 258 insertions(+), 137 deletions(-) -- To unsubscribe from this list: send the line "unsubscribe linux-usb" in the body of a message to majordomo@xxxxxxxxxxxxxxx More majordomo info at http://vger.kernel.org/majordomo-info.html
