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

[Xen-devel] Few questions about migration

  • To: xen-devel <xen-devel@xxxxxxxxxxxxxxxxxxx>
  • From: "Amitayu Das" <amitayudas@xxxxxxxxx>
  • Date: Thu, 2 Mar 2006 17:14:11 -0500
  • Delivery-date: Thu, 02 Mar 2006 22:14:44 +0000
  • Domainkey-signature: a=rsa-sha1; q=dns; c=nofws; s=beta; d=gmail.com; h=received:message-id:date:from:to:subject:mime-version:content-type; b=NXwmhnaczBOpFm+lYB80QDmELRhCeplhJ+lF/umIqgqSoOQEdphMYS4Jm8AgsAv7knFDe4gj0BMNtTC8+nvU7zXSw5syZ2cBY7gWTNTkVjCp8ZpzJ2yy89raQIvOKdjPOaAamyNzN2nyY5eZfZqQ1htXZwMuVxqBDUU8IVJcnms=
  • List-id: Xen developer discussion <xen-devel.lists.xensource.com>


I had few questions regarding the mechanism employed by the hypervisor for figuring out the pages modified during a pre-copy round of a migration of a VM.  Please assume that a migration of a VM is going on in the backdrop and a process (running inside that VM) is accessing (reading/writing) into pages in memory.

1. The hypervisor maintains shadow page-table with *read-only" permission for all the pages. When a page is being written by the afore-menioned process (and assume the page is writable),  what is the way for the hypervisor to know about it? There may be two possibilities as mentioned below (please let me know which one is correct):
       a. The guest-OS does not raise any trap (as the page is writable); but, because a page was modified, he informs the hypervisor. Is this correct?
       b. Following the writing, the TLB has to be updated. Because of that the hypervisor comes to know about the modification. Is this correct?

Put in other words, what exactly happens to let hypervisor track page-modification with the help of shadow page-table or dirty-bitmap?

2. One may want to update TLB entries for the pages updated during a pre-copy round in a batch for better performance? Then, is it the case that hypervisor comes to know about all the modified pages during the end of a pre-copy round when the updation of TLB is over? Pleaes note that this would be possible for only those pages whose entries were already available in TLB. Now, what'd happen if the writing process has resulted in a TLB miss in the first place?

Thanks in advance,

Xen-devel mailing list



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