Re: [PATCH] xen: start PCI hole at 0xe0000000 (same as pc_init1 and qemu-xen-traditional)

[Paolo Bonzini <pbonzini@redhat.com>]

Il 31/05/2013 12:58, Stefano Stabellini ha scritto:
> On Thu, 30 May 2013, Stefano Stabellini wrote:
>> On Thu, 30 May 2013, Andreas Färber wrote:
>>> Am 30.05.2013 15:20, schrieb Stefano Stabellini:
>>>> We are currently setting the PCI hole to start at HVM_BELOW_4G_RAM_END,
>>>> that is 0xf0000000.
>>>> Start the PCI hole at 0xe0000000 instead, that is the same value used by
>>>> pc_init1 and qemu-xen-traditional.
>>>>
>>>> This fixes a bug with device assignment of graphic cards with large
>>>> amount of video memory (http://marc.info/?l=xen-devel&m=136242365010750).
>>>>
>>>> Signed-off-by: Stefano Stabellini <stefano.stabellini@eu.citrix.com>
>>>> CC: qemu-stable@nongnu.org
>>>> CC: pbonzini@redhat.com
>>>> CC: afaerber@suse.de
>>>> CC: hanweidong@huawei.com
>>>> CC: arei.gonglei@huawei.com
>>>> ---
>>>>  hw/i386/pc_piix.c    |    6 +++---
>>>>  include/hw/i386/pc.h |    3 +++
>>>>  xen-all.c            |   12 ++++++------
>>>>  3 files changed, 12 insertions(+), 9 deletions(-)
>>>>
>>>> diff --git a/hw/i386/pc_piix.c b/hw/i386/pc_piix.c
>>>> index 43ab480..06416ce 100644
>>>> --- a/hw/i386/pc_piix.c
>>>> +++ b/hw/i386/pc_piix.c
>>>> @@ -102,9 +102,9 @@ static void pc_init1(MemoryRegion *system_memory,
>>>>          kvmclock_create();
>>>>      }
>>>>  
>>>> -    if (ram_size >= 0xe0000000 ) {
>>>> -        above_4g_mem_size = ram_size - 0xe0000000;
>>>> -        below_4g_mem_size = 0xe0000000;
>>>> +    if (ram_size >= QEMU_BELOW_4G_RAM_END ) {
>>>> +        above_4g_mem_size = ram_size - QEMU_BELOW_4G_RAM_END;
>>>> +        below_4g_mem_size = QEMU_BELOW_4G_RAM_END;
>>>>      } else {
>>>>          above_4g_mem_size = 0;
>>>>          below_4g_mem_size = ram_size;
>>>> diff --git a/include/hw/i386/pc.h b/include/hw/i386/pc.h
>>>> index 663426c..626442b 100644
>>>> --- a/include/hw/i386/pc.h
>>>> +++ b/include/hw/i386/pc.h
>>>> @@ -77,6 +77,9 @@ extern int fd_bootchk;
>>>>  void pc_register_ferr_irq(qemu_irq irq);
>>>>  void pc_acpi_smi_interrupt(void *opaque, int irq, int level);
>>>>  
>>>> +#define QEMU_BELOW_4G_RAM_END       0xe0000000
>>>> +#define QEMU_BELOW_4G_MMIO_LENGTH   ((1ULL << 32) - QEMU_BELOW_4G_RAM_END)
>>>> +
>>>>  void pc_cpus_init(const char *cpu_model, DeviceState *icc_bridge);
>>>>  void pc_hot_add_cpu(const int64_t id, Error **errp);
>>>>  void pc_acpi_init(const char *default_dsdt);
>>>> diff --git a/xen-all.c b/xen-all.c
>>>> index 539a154..5ca0c26 100644
>>>> --- a/xen-all.c
>>>> +++ b/xen-all.c
>>>> @@ -161,18 +161,18 @@ static void xen_ram_init(ram_addr_t ram_size)
>>>>      ram_addr_t block_len;
>>>>  
>>>>      block_len = ram_size;
>>>> -    if (ram_size >= HVM_BELOW_4G_RAM_END) {
>>>> +    if (ram_size >= QEMU_BELOW_4G_RAM_END) {
>>>
>>> Is the HVM_ constant still used elsewhere? I don't see it being removed.
>>>
>>> Otherwise looks straightforward.
>>
>> It's defined in the Xen headers so we don't have to remove it here.
> 
> If you are happy with it, should I just go ahead and include this patch
> in the next PULL request?
> 

Yes, please.

Paolo