Re: [RFC PATCH] KVM: remove the writable page for read fault case in hva_to_pfn_slow()

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

 



On Tue, 2018-10-09 at 15:32 +0000, richard.weiyang@xxxxxxxxx wrote:
> > If the PTE is indeed writable, we can also create the shadow PTE with
> > write permissions even though we're handling a read fault.????Marking
> > the SPTE as writable means we don't take another page fault if/when
> > the guest writes the page, e.g. in a RMW scenario we take one page
> > fault instead of two.
> > 
> This makes sence to me.
> 
> As you mentioned in the last sentense of previous paragraph, a read-only
> PTE with writable VMA means the PTE is indeed writalbe. In this case we
> could set the SPTE as write to reduce another page fault in a RMW
> scenario.
> 
> After all I still have one confusion:
> 
>   My confusion is why the 2nd __get_user_pages_fast() would succeed
>   while the 1st failed with same parameters passed. Do we touch the PTE
>   mapping during get_user_pages_unlocked()?

Yep, get_user_pages_unlocked() will fault-in the page if necessary.



[Index of Archives]     [KVM ARM]     [KVM ia64]     [KVM ppc]     [Virtualization Tools]     [Spice Development]     [Libvirt]     [Libvirt Users]     [Linux USB Devel]     [Linux Audio Users]     [Yosemite Questions]     [Linux Kernel]     [Linux SCSI]     [XFree86]

  Powered by Linux