usage by means of fadvise64() to tell the OS to discard the cache
pages used for the save/dump file.
Signed-off-by: Simon Graham <Simon.Graham@stratus.com>
The old format is still supported and the old implementation is left
as is with the slight adjustment to coexist with new implementation.
Signed-off-by: Isaku Yamahata <yamahata@valinux.co.jp>
added PFN-GMFN table for non-auto translated physmap
added PFN table for auto translated physmap.
HVM domain support.
IA64 support
Signed-off-by: Isaku Yamahata <yamahata@valinux.co.jp>
Use the guest's own p2m table instead of xc_get_pfn_list(), which
cannot handle PFNs with no MFN.
Dump a zeroed page for PFNs with no MFN.
Clearly deprecate xc_get_pfn_list().
Do not include a P2M table with HVM domains.
Refuse to dump HVM until we can map its pages with PFNs.
Signed-off-by: John Levon <john.levon@sun.com>
Hide the (default or user specified) DMA width from anything outside
the heap allocator. I/O-capable guests can now request any width for
the memory they want exchanged/added.
Signed-off-by: Jan Beulich <jbeulich@novell.com>
Replace the 3-zone scheme of the heap allocator with one with one
where zones are distinguished by their bit widths.
Signed-off-by: Jan Beulich <jbeulich@novell.com>
This patch doesn't introduce functional changes, but simply moves code
around to make the unused (outside of the page allocator) heap alloc
functions taking an explicit zone parameter static without having to
forward-prototype them in their source file.
Signed-off-by: Jan Beulich <jbeulich@novell.com>
Additionally a default line is added (key 0x00000000) which is used
when no specific entry in the format file is found.
Signed-off-by: Thomas Friebel <thomas.friebel@amd.com>
Get rid of USER|RW|ACCESSED L3 PTE flags which may be present
in a PAE guest when running on a 64bit hypervisor at save time.
This prevents the restore operation from getting confused by
these flags.
Reset disallowed L3E flags to match PAE case.
Signed-off-by: Emmanuel Ackaouy <ack@xensource.com>
Fix the VMX and SVM handlers to reflect the new semantics (which is
what is actually required by the one caller, in shadow fault path).
Signed-off-by: Keir Fraser <keir@xensource.com>
This conserves low memory.
Provide new function alloc_boot_low_pages() for those callers who
actually require lowmem pages (e.g., below 4GB).
Based on a patch by Chris Lalancette <clalance@redhat.com>
Signed-off-by: Keir Fraser <keir@xensource.com>
handler, where we can quite conveniently check whether the access
segment is SS. This will only give a false positive for non-stack
writes that override the destination segment to SS, which it is
probably safe to assume will never happen when the destination is a
legitimate page table!
For now, instead of bailing and unshadowing just increment a perfctr
and we'll see if that increases fast under any reasonable workload.
Signed-off-by: Keir Fraser <keir@xensource.com>
