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

[Xen-devel] When can ptes be written by Xen?

  • To: xen-devel@xxxxxxxxxxxxxxxxxxx
  • From: "Jacob Gorm Hansen" <jacobg@xxxxxxx>
  • Date: Fri, 30 Jun 2006 15:17:34 +0200
  • Delivery-date: Fri, 30 Jun 2006 06:17:57 -0700
  • Domainkey-signature: a=rsa-sha1; q=dns; c=nofws; s=beta; d=gmail.com; h=received:message-id:date:from:sender:to:subject:mime-version:content-type:content-transfer-encoding:content-disposition:x-google-sender-auth; b=mSjVlqeXKYtm7cDss0MHY848wD0Etu37+wb2KkRlz2qe4vomQIJtB5gDwcItO1Jmp63X0jw+JoPE3zmdZ1MyuGigJgXRHkBKif1/8Bkq6rSMNo4GJMrvVqpeCoU7LtW0dJvQ+rWP8qGuLUvfzrCW5I/gx9fxdb4LSNDd/p/1ByQ=
  • List-id: Xen developer discussion <xen-devel.lists.xensource.com>


I am having a weird consistency problem with a XenLinux page that
changes even though it is mapped read-only.

The offending page looks like a pte, and it has PG_foreign set in it's
struct page. The page is zero-filled, but an 'invisible' write then
changes it to contain a value like:


-- which to me looks like a user-accessible mapping of the mfn 0xc642
(this is x86-32). However, the page is not listed in any pgd in the
system, though it's mfn does map to a pfn belonging to the domain.

Before I do anything I attempt to make sure writable page tables are
flushed by executing a dummy mmu-update. I trap update_va_mapping()
calls, so that should not be the source of the error.

I can fix the problem by scanning the page array for all pages with
the foreign bit set and giving these special treatment. However that
will scale badly with the number of pages.

Under what conditions would Xen write to a pte like above?

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®.