>> This patch addresses the issue fully described here: >> http://lists.nongnu.org/archive/html/qemu-devel/2013-02/msg01804.html >> >> Linux kernels prior to 2.6.36 do not disable the PCI device during >> enumeration process. Since lower and higher parts of a 64bit BAR >> are programmed separately this leads to qemu receiving a request to occupy >> a completely wrong address region for a short period of time. >> We have found that the boot process screws up completely if kvm-apic range >> is overlapped even for a short period of time (it is fine for other >> regions though). > > So the issue it that we hide the MSI target region from the device > models for a "short periods of time"? How short is this? It would be >good to fully understand what goes "completely wrong" to avoid that we > miss to fix something else (IO-APIC, HPET) or even paper over a problem > of the memory subsystem. As far as I can understand the region corruption and recovery happens within one call of pci_read_bases of linux kernel enumeration: (http://lxr.linux.no/#linux+v2.6.21/drivers/pci/probe.c#L173) Unfortunately I cannot even guess what is happening with kvm-apic-msi, or why it does not tolerate overlaps. This is actually a still open qestion.
>> >> This patch raises the priority of the kvm-apic memory region, so it is >> never pushed out by PCI devices. The patch is quite safe as it does not >> touch memory manager. I have fixed styling issues. Please use this version. Signed-off-by: Alexey Korolev <akoro...@gmail.com> Signed-off-by: Michael S. Tsirkin <m...@redhat.com> --- hw/sysbus.c | 26 ++++++++++++++++++++++---- hw/sysbus.h | 2 ++ target-i386/cpu.c | 3 ++- 3 files changed, 26 insertions(+), 5 deletions(-) diff --git a/hw/sysbus.c b/hw/sysbus.c index 6d9d1df..242eb1e 100644 --- a/hw/sysbus.c +++ b/hw/sysbus.c @@ -48,7 +48,8 @@ void sysbus_connect_irq(SysBusDevice *dev, int n, qemu_irq irq) } } -void sysbus_mmio_map(SysBusDevice *dev, int n, hwaddr addr) +static void sysbus_mmio_map_common(SysBusDevice *dev, int n, hwaddr addr, + bool may_overlap, unsigned priority) { assert(n >= 0 && n < dev->num_mmio); @@ -61,11 +62,28 @@ void sysbus_mmio_map(SysBusDevice *dev, int n, hwaddr addr) memory_region_del_subregion(get_system_memory(), dev->mmio[n].memory); } dev->mmio[n].addr = addr; - memory_region_add_subregion(get_system_memory(), - addr, - dev->mmio[n].memory); + if (may_overlap) { + memory_region_add_subregion_overlap(get_system_memory(), + addr, + dev->mmio[n].memory, + priority); + } else { + memory_region_add_subregion(get_system_memory(), + addr, + dev->mmio[n].memory); + } +} + +void sysbus_mmio_map(SysBusDevice *dev, int n, hwaddr addr) +{ + sysbus_mmio_map_common(dev, n, addr, false, 0); } +void sysbus_mmio_map_overlap(SysBusDevice *dev, int n, hwaddr addr, + unsigned priority) +{ + sysbus_mmio_map_common(dev, n, addr, true, priority); +} /* Request an IRQ source. The actual IRQ object may be populated later. */ void sysbus_init_irq(SysBusDevice *dev, qemu_irq *p) diff --git a/hw/sysbus.h b/hw/sysbus.h index a7fcded..2100bd7 100644 --- a/hw/sysbus.h +++ b/hw/sysbus.h @@ -56,6 +56,8 @@ void sysbus_init_ioports(SysBusDevice *dev, pio_addr_t ioport, pio_addr_t size); void sysbus_connect_irq(SysBusDevice *dev, int n, qemu_irq irq); void sysbus_mmio_map(SysBusDevice *dev, int n, hwaddr addr); +void sysbus_mmio_map_overlap(SysBusDevice *dev, int n, hwaddr addr, + unsigned priority); void sysbus_add_memory(SysBusDevice *dev, hwaddr addr, MemoryRegion *mem); void sysbus_add_memory_overlap(SysBusDevice *dev, hwaddr addr, diff --git a/target-i386/cpu.c b/target-i386/cpu.c index 5582e5f..4b72094 100644 --- a/target-i386/cpu.c +++ b/target-i386/cpu.c @@ -2088,7 +2088,8 @@ static void x86_cpu_apic_init(X86CPU *cpu, Error **errp) /* NOTE: the APIC is directly connected to the CPU - it is not on the global memory bus. */ /* XXX: what if the base changes? */ - sysbus_mmio_map(SYS_BUS_DEVICE(env->apic_state), 0, MSI_ADDR_BASE); + sysbus_mmio_map_overlap(SYS_BUS_DEVICE(env->apic_state), 0, + MSI_ADDR_BASE, 1000); apic_mapped = 1; } } -- 1.7.9.5