Re: [PATCH V2 0/5] x86: Create direct mappings for E820_RAM only

[WANG Chao <chaowang@redhat.com>]

On 08/15/2012 06:54 AM, Jacob Shin wrote:
> On Tue, Aug 14, 2012 at 04:34:39PM +0800, Dave Young wrote:
>> On 08/14/2012 05:46 AM, Jacob Shin wrote:
>>
>>> Currently kernel direct mappings are created for all pfns between
>>> [ 0 to max_low_pfn ) and [ 4GB to max_pfn ). When we introduce memory
>>> holes, we end up mapping memory ranges that are not backed by physical
>>> DRAM. This is fine for lower memory addresses which can be marked as UC
>>> by fixed/variable range MTRRs, however we run in to trouble with high
>>> addresses.
>>>
>>> The following patchset creates direct mappings only for E820_RAM regions
>>> between 0 ~ max_low_pfn and 4GB ~ max_pfn. And leaves non-E820_RAM and
>>> memory holes unmapped.
>>
>>
>> Hi,
>>
>> Chaowang did some kdump test in a kvm guest with this patchset, 2nd
>> kenrel just reboot after some ACPI printk, see below dmesg of 2nd kernel:
>
> Hello, thanks for testing, since I have not tested under KVM .. I also have
> not tested passing in user supplied memory maps as your kernel log suggests.
>
> Looking into this, it seems like we get a page fault while trying to set up
> fixmap for the APIC. I think the fixmap is set up even before we get to
> setup_arch(), and it is sitting in memory that is not marked as usable by
> your user supplied e820.
>
> Could you give V3 a try? I just sent it out a minute ago, this version
> won't try to remap what has already been mapped as part of the boot process
> before we get to setup_arch, it'll just take what its given.
>

Hi, Jacob

I just tried v3 patchset in my x86_64 kvm guest, it was booting 
successfully and the issue mentioned is gone.

-WANG Chao