Re: [PATCH 8/9] coalesce mmio regions with an explicit call

[Glauber Costa <glommer@redhat.com>]

On Tue, Sep 23, 2008 at 10:29:48AM +0300, Avi Kivity wrote:
> Glauber Costa wrote:
>> On Sat, Sep 20, 2008 at 11:39:44AM -0700, Avi Kivity wrote:
>>   
>>> Glauber Costa wrote:
>>>     
>>>> Remove explicit calls to mmio coalescing. Rather,
>>>> include it in the registration functions.
>>>>
>>>> index 5ae3960..2d97b34 100644
>>>> --- a/qemu/hw/e1000.c
>>>> +++ b/qemu/hw/e1000.c
>>>> @@ -942,18 +942,6 @@ e1000_mmio_map(PCIDevice *pci_dev, int region_num,
>>>>       d->mmio_base = addr;
>>>>      cpu_register_physical_memory(addr, PNPMMIO_SIZE, d->mmio_index);
>>>> -
>>>> -    if (kvm_enabled()) {
>>>> -	int i;
>>>> -        uint32_t excluded_regs[] = {
>>>> -            E1000_MDIC, E1000_ICR, E1000_ICS, E1000_IMS,
>>>> -            E1000_IMC, E1000_TCTL, E1000_TDT, PNPMMIO_SIZE
>>>> -        };
>>>> -        qemu_kvm_register_coalesced_mmio(addr, excluded_regs[0]);
>>>> -        for (i = 0; excluded_regs[i] != PNPMMIO_SIZE; i++)
>>>> -            qemu_kvm_register_coalesced_mmio(addr + excluded_regs[i] + 4,
>>>> -                             excluded_regs[i + 1] - excluded_regs[i] - 4);
>>>> -    }
>>>>  }
>>>>          
>>> Where did all of this go?
>>>     
>>
>> All the region is coalesced (not just the pieces) automatically during memory registration.
>> Or not at all, in case coalescing is disabled.
>>   
>
> You can't coalesce the registers which trigger device action.  You'll  
> destroy latency and/or functionality.

which kills the goal of getting rid of explicit kvm code.

So maybe the solution here is to add calls in qemu to a memory
coalescing function that in the raw qemu / kqemu case just don't
do anything?

Anthony, do you have an opinion about it ?