[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: [Xen-devel] [PATCH] Speed up superpage mapping, take 2

  • To: Jan Beulich <JBeulich@xxxxxxxxxx>, Dave McCracken <dcm@xxxxxxxx>
  • From: Keir Fraser <keir.fraser@xxxxxxxxxxxxx>
  • Date: Wed, 26 May 2010 08:45:51 +0100
  • Cc: Xen Developers List <xen-devel@xxxxxxxxxxxxxxxxxxx>
  • Delivery-date: Wed, 26 May 2010 00:46:57 -0700
  • List-id: Xen developer discussion <xen-devel.lists.xensource.com>
  • Thread-index: Acr8phy9B5mC1g9yTxGHZg2YbubIPgAAVms6
  • Thread-topic: [Xen-devel] [PATCH] Speed up superpage mapping, take 2

On 26/05/2010 08:36, "Jan Beulich" <JBeulich@xxxxxxxxxx> wrote:

>> -#define __HYPERVISOR_VIRT_START_PAE    0xF5800000
>> -#define __MACH2PHYS_VIRT_START_PAE     0xF5800000
>> -#define __MACH2PHYS_VIRT_END_PAE       0xF6800000
>> +#define __HYPERVISOR_VIRT_START_PAE    0xF5600000
>> +#define __MACH2PHYS_VIRT_START_PAE     0xF5600000
>> +#define __MACH2PHYS_VIRT_END_PAE       0xF6600000
>>  #define HYPERVISOR_VIRT_START_PAE      \
>>      mk_unsigned_long(__HYPERVISOR_VIRT_START_PAE)
>>  #define MACH2PHYS_VIRT_START_PAE       \
> is absolutely forbidden: You cannot lower the start address of
> the hypervisor address space hole - no kernel can be expected
> to be able to deal with that. I think you will need to make the
> feature 64-bit specific, unless you can find a 2Mb range within
> the existing hole (but I doubt there is any room left), or unless
> you map the table dynamically as needed for 32-bits.

Sixty-four-bit-only is *fine* by me. But yes, the patch will need another
rev for this.

 -- Keir

Xen-devel mailing list



Lists.xenproject.org is hosted with RackSpace, monitoring our
servers 24x7x365 and backed by RackSpace's Fanatical Support®.