On Sat, 22 Aug 2009, Wu Zhangjin wrote: > On Sat, 2009-08-22 at 00:06 +0200, Rafael J. Wysocki wrote: > > On Friday 21 August 2009, Wu Zhangjin wrote: > > > Hi All, > > > > > > After reading some documents in Documentation/ and the Internet, I found > > > there is an interface /proc/acpi/wakeup to set the wakeup devices on > > > machines support ACPI. but my machine does not support ACPI, > > > no /proc/acpi/wakeup there. > > > > > > my machine is FuLoong2F, which is loongson based machine, it does not > > > support ACPI, but currently, I have made it support basic standby mode > > > for it has a 'wait' mode(after setting it's frequency to ZERO) and can > > > be waked up via an external interrupt. and here is the basic framework: > > > > > > 0. suspend: > > > > > > arch_suspend_disable_irqs(disable all interrupts) > > > > > > ... > > > > > > fuloong_pm_enter() > > > > > > ... > > > > > > arch_suspend_enable_irqs(enable all interrupts) > > > > > > 1. fuloong_pm_enter: > > > > > > setup_wakeup_interrupts() > > > > > > loongson_pm_enter(put loongson into 'wait' mode) > > > > > > 3. setup_wakeup_interrupts > > > > > > if the interrupt line is connected to loongson directly, no setup need > > > to do, but if there is an interrupt line connected to loongson > > > indirectly(i.e via i8259A), I need to get through the interrupt path via > > > enable(un-mask) the interrupts in interrupt controller and resuming the > > > devices earlier manually or enable the interrupt bit on the devices and > > > even or set a flag of the device driver to tell the system not suspend > > > this device. > > > > > > I have tried the first(resuming earlier) and second method(enable the > > > interrupt bit) on the 8042 keyboard o YeeLoong(another loongson-based > > > machine, a netbook, i8042 is connected to i8259A directly), it works > > > well, but not tried the third method yet. > > > > > > I guess the /proc/acpi/wakeup based method in user-space should be > > > something like this, but is there any relative method in kernel space? > > > > > > there are only USB interfaces on FuLoong2F machines, I guess it will be > > > very hard to get through the interrupt path via the above two methods(I > > > have tried to enable the USB interrupt in south bridge, but not work), > > > so, is there an easier method to do it? > > > > Is your USB controller a PCI one? > > Yes. > > and I just checked the file: drivers/usb/host/ohci-pci.c, there are two > functions ohci_pci_suspend/ohci_pci_resume there, perhaps I can call > these two functions directly to resume it earlier and suspend it again > before the real resuming, any easier methods? is there a flag to tell > the system not suspend it or just not disable the interrupt request? I think the interface you're looking for is /sys/bus/usb/devices/usbN/../power/wakeup where N is the USB bus number. This file defaults to "disabled"; if you write "enabled" to it then the USB host controller will be enabled as a wakeup source. This should work even without ACPI; it uses the PCI PME mechanism. However you need to realize something: If you enable USB wakeups then unplugging your USB keyboard or mouse will cause the computer to wake up. Alan Stern _______________________________________________ linux-pm mailing list linux-pm@xxxxxxxxxxxxxxxxxxxxxxxxxx https://lists.linux-foundation.org/mailman/listinfo/linux-pm
