[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]
RE: [PATCH] remove VHPT_ADDR (was Re: [Xen-ia64-devel] [PATCH] per vcpu vhpt)
- To: "Isaku Yamahata" <yamahata@xxxxxxxxxxxxx>
- From: "Xu, Anthony" <anthony.xu@xxxxxxxxx>
- Date: Wed, 11 Oct 2006 13:49:59 +0800
- Cc: xen-ia64-devel@xxxxxxxxxxxxxxxxxxx
- Delivery-date: Tue, 10 Oct 2006 22:50:26 -0700
- List-id: Discussion of the ia64 port of Xen <xen-ia64-devel.lists.xensource.com>
- Thread-index: Acbs4CZM0vFEdgg7QIizbQkPjWa9ZAAGMGuA
- Thread-topic: [PATCH] remove VHPT_ADDR (was Re: [Xen-ia64-devel] [PATCH] per vcpu vhpt)
Does this patch work?
Is it for checkin or review?
>From: Isaku Yamahata [mailto:yamahata@xxxxxxxxxxxxx]
>Sent: 2006年10月11日 10:51
>To: Xu, Anthony
>Subject: [PATCH] remove VHPT_ADDR (was Re: [Xen-ia64-devel] [PATCH] per vcpu
>Hi. I dig out the old patch and updated it.
>On Tue, Oct 10, 2006 at 09:31:26AM +0800, Xu, Anthony wrote:
>> >> >non-current vcpu's case must be handled.
>> >> I don't see any issue here, if we use option1, we use identity
>> >> for VHPT, and so we can use 16 MB TR mapping, of cause, 64KB VHPT is
>> >> covered by this 16M TR.
>> >In fact I had such a patch for the very early version of
>> >the pervcpu vhpt patch.
>> >But I discarded it for some reason. I don't remember clearly why.
>> >Probably I wanted to avoid introducing an implicit dependency
>> >and increasing vcpu context switch overhead.
>> As I know, IA64/LINUX is using this mechanism, the task_struct is 32k,
>> but it uses 16M TR mapping if it is not covered 64M kernel TR.
>> Xen-ia64-devel mailing list
Xen-ia64-devel mailing list