Re: [PATCH] memremap: Fix NULL pointer BUG in get_zone_device_page()

[Dan Williams]

On Tue, Aug 23, 2016 at 6:29 PM, Kani, Toshimitsu <toshi.kani@xxxxxxx> wrote:
>> On Tue, Aug 23, 2016 at 4:47 PM, Kani, Toshimitsu <toshi.kani@xxxxxxx>
>> wrote:
>> > On Tue, 2016-08-23 at 15:32 -0700, Dan Williams wrote:
>> >> On Tue, Aug 23, 2016 at 11:43 AM, Toshi Kani <toshi.kani@xxxxxxx>
>> >> wrote:
>> >  :
>> >> I'm not sure about this fix.  The point of honoring
>> >> vmem_altmap_offset() is because a portion of the resource that is
>> >> passed to devm_memremap_pages() also contains the metadata info
>> block
>> >> for the device.  The offset says "use everything past this point for
>> >> pages".  This may work for avoiding a crash, but it may corrupt info
>> >> block metadata in the process.  Can you provide more information
>> >> about the failing scenario to be sure that we are not triggering a
>> >> fault on an address that is not meant to have a page mapping?  I.e.
>> >> what is the host physical address of the page that caused this fault,
>> >> and is it valid?
>> >
>> > The fault address in question was the 2nd page of an NVDIMM range.  I
>> > assumed this fault address was valid and needed to be handled.  Here is
>> > some info about the base and patched cases.  Let me know if you need
>> > more info.
>> >
>> > Base
>> > ====
>> >
>> > The following NVDIMM range was set to /dev/dax.
>>
>> With ndctl create-namespace or manually via sysfs?  Specifically I'm
>> looking for what the 'align' attribute was set to when the
>> configuration was established.  Can you provide a dump of the sysfs
>> attributes for the /dev/dax parent device?
>
> I used the ndctl command below.
> ndctl create-namespace -f -e namespace0.0 -m dax
>
> Here is additional info from my note for the base case.
>
> p {struct dev_pagemap} 0xffff88046d0453f0
> $3 = {
>   altmap = 0xffff88046d045410,
>   res = 0xffff88046d0453a8,
>   ref = 0xffff88046d0452f0,
>   dev = 0xffff880464790410
> }
>
> crash> p {struct vmem_altmap} 0xffff88046d045410
> $6 = {
>   base_pfn = 0x480000,
>   reserve = 0x2,        // PHYS_PFN(SZ_8K)
>   free = 0x101fe,
>   align = 0x1fe,
>   alloc = 0x10000
> }

Ah, so, on second look the 0x490200000 data offset looks correct.  The
total size of the address range is 16GB which equates to 256MB needed
for struct page, plus 2MB more to re-align the data on the next 2MB
boundary.

The question now is why is the guest faulting on an access to an
address less than 0x490200000?