this macro where newlines have been appended -- the macro does this itself so
there is no need to add another one.
Signed-off-by: Ewan Mellor <ewan@xensource.com>
structure that is 64 bytes on x86. This ensures that indexing into
the array is fast (power-of-two size) and that accesses are cache
friendly (cache line size is usually 32 or 64 bytes).
Rename vcpu_data to vcpu_info, vcpu_time to vcpu_info.time.
Signed-off-by: Keir Fraser <keir@xensource.com>
(live) restore. This should now enable live migrate of fork-intensive (or
other page-table-intensive) workloads.
Signed-off-by: Steven Hand <steven@xensource.com>
Valgrind cannot see when a hypercall has initialised entries in a data
structure, so appropriate memsets have been placed before using dom0_op_t,
privcmd_hypercall_t, and a couple of miscellaneous blocks passed into this
layer. This initialises the block so that valgrind considers it to be valid,
but the data therein will be immediately overwritten by the hypercall, all
being well.
These changes are semantically neutral if -DVALGRIND is not set.
Signed-off-by: Ewan Mellor <ewan@xensource.com>
This includes quite a few cleanups / refactoring of the old code, some
of which is intended to prepare for 64-bit save/restore.
Signed-off-by: Steven Hand <steven@xensource.com>
recreation semantics.
Remove the unpause at the end of xc_linux_restore, and move it to
XendDomainInfo. This is necessary because xenstored now allocates the domain
path when the domain is introduced, which means that the new domain cannot
start running until that introduce is performed and the new devices
configured.
Give restore a separate completion phase in which domain details are stored.
This is required because the domain path is no longer available until after
the introduceDomain call.
TODO: Split the domain introduction into two so that the domain path is
available earlier. At the moment, the domain <-> store channel details are
passed in to xenstored when the domain is introduced, but in the case of
restore it is necessary to wait until the restore is completed before the
channel MFN is available.
Change the interface between XendDomainInfo and XendCheckpoint/image to not
have hideous callbacks through setConsoleRef and setStoreRef. Instead,
image.createImage explicitly returns those values, and XendCheckpoint passes
them through to completeRestore.
Move the purging of the domain path corresponding to a new domain from Xend
to xenstored, since xenstored is now in charge of this path. With the domain
path creation moved to xenstored, Xend cannot remove the path, because watches
may have fired on it already.
Fix the printf statement in xenstored in verbose mode that details the
messages being written. This statement was printing the buffer using %s, but
this buffer has an explicit length field, so we were seeing garbage after the
correct details.
Signed-off-by: Ewan Mellor <ewan@xensource.com>
change to watch out for is that the suspend record, on
save/restore, is now passed via register %edx not %esi
(both i386 and x86/64).
Signed-off-by: Keir Fraser <keir@xensource.com>
PAE should be a *lot* more robust on systems that actually have more than 4GB thanks to all the various patches that went in today. I find it astounding that it ever appeared to work at all!
Signed-off-by: ian@xensource.com
- translate console mfn to pfn
- unbind/bind console evtchn
- recover console mfn on restore
- trigger xenconsoled refresh after mfn is put in store
Signed-off-by: Christian Limpach <Christian.Limpach@cl.cam.ac.uk>
was failing to unbind its irq handler correctly which (on SMP) caused the
guest to BUG() on resume when trying to request its callfunc_irq. Plus a
few small cosmetic changes.
Signed-off-by: Steven Hand <steven@xensource.com>
/bin/bash: ❌ command not found
- Add xenbus driver suspend/resume functions
- Change xenbus irq to not be shared
- Translate store mfn in suspend record to pfn and back
- Make tools re-introduce the domain to the store
Signed-off-by: Rusty Russel <rusty@rustcorp.com.au>
Signed-off-by: Christian Limpach <Christian.Limpach@cl.cam.ac.uk>
Rationalise x86 CRn guest state into a ctrlreg array in the per-vcpu
context structure. Most noticeably this means the pt_base field has
gone away -- replaced by ctrlreg[3] (CR3). VCPU_guest_stts is also
gone -- it was never arch-independent anyway.
Signed-off-by: Keir Fraser <keir@xensource.com>
Implement the parts of vm save which need interaction with xend
as part of xend, instead of using xfrd.
Execute xc_linux_save in a seperate process so that it can't
crash xend. Also handle errors passed from xc_linux_save.
xen_domain.c:
Disable save in xfrd.
xc_save.c:
new file
Makefile:
Add xc_save.
XendDomainInfo.py:
Add suspended state and threading Condition with notification, allowing
easy waiting for state changes.
XendDomain.py:
Implement the parts of vm save which need interaction with xend
as part of xend, instead of using xfrd. Set state to "suspended"
when detecting a suspended domain.
Fix reading output from subprocesses.
Fix ValueError in xen_domain().
xc.c:
Remove python binding for xc_linux_save.
xc_linux_save.c:
Implement the parts of vm save which need interaction with xend
as part of xend, instead of using xfrd. Also run xc_linux_save
in a seperate process.
xc_linux_restore.c:
Flush output so that xend picks it up timely.
Also disable debug output again.
xc.h:
Update xc_linux_save prototype and fix comments for xc_linux_{save,restore}.
ignore:
Add tools/xcutils/xc_save.
xpopen.py:
Exit with 127 if exec fails.
Signed-off-by: Christian Limpach <Christian.Limpach@cl.cam.ac.uk>
Execute xc_linux_restore in a seperate process so that it can't
crash xend. Also handle errors passed from xc_linux_restore and
log info messages from xc_linux_restore.
XendDomain.py:
Popen xc_restore instead of calling xc_linux_restore directly.
xc.c:
Add pyxc_handle exporting the file descriptor to the control interface.
Remove xc_linux_restore -- replaced by popen of xc_restore directly
from python.
xc_linux_restore.c:
Enable debug output.
xpopen.py:
Add xpopen functionality:
Optionally exclude a list of file descriptors from being closed, allowing
access to those file descriptors from the command.
Remove unused parts.
xpopen.py, Makefile, xc_restore.c:
new file
Makefile:
Add xcutils subdir.
ignore:
Add tools/xcutils/xc_restore.
Signed-off-by: Christian Limpach <Christian.Limpach@cl.cam.ac.uk>
XendDomain.py, xc.c, xc_linux_restore.c, xc.h:
Move read of pfn to mfn frame list into xc_linux_restore.
Signed-off-by: Christian Limpach <Christian.Limpach@cl.cam.ac.uk>
XendDomainInfo.py, XendDomain.py, xc.c, xc_linux_restore.c, xc.h:
Implement the parts of vm restore which need interaction with xend
as part of xend, instead of using xfrd. The restore functionality
using xfrd was broken anyway since xend didn't handle the callback
channel from xfrd correctly.
xen_domain.c:
Disable restore in xfrd.
Signed-off-by: Christian Limpach <Christian.Limpach@cl.cam.ac.uk>
XendDomainInfo.py, xc.c, xc_linux_restore.c, xc_domain.c, xc.h:
Split pincpu, setcpuweight, setmaxmem and memory increase reservation
out of xc_domain_create. Add glue to get the missing functions exposed
to python and use.
Signed-off-by: Christian Limpach <Christian.Limpach@cl.cam.ac.uk>
Cleaner GDT interface. Xen now reserves the last 2 pages of a full-size
(16 page) GDT. A guest can register a GDT with up to 7k entries. Xen
automatically makes the GDT up to max size with empty page mappings,
and turns page faults in that area to #GP(selector).
Signed-off-by: Keir Fraser <keir@xensource.com>
Remove generic 'fast trap' from the Xen public interfaces. Internally
within Xen, we automatically set up direct trap for int 0x80, which is
the only 'fast trap' vector that we ever supported anyway.
Signed-off-by: Keir Fraser <keir@xensource.com>
Split out context fetching into separate DOM0 op
make GETDOMAININFO a little more sensible with respect to MP
make coredump dump all cpu contexts
Signed-off-by: Kip Macy <kmacy@netapp.com>
Signed-off-by: Christian Limpach <Christian.Limpach@cl.cam.ac.uk>
Batch p.t. pin requests during guest restore (after being suspended, or
after migration). Allow L1 tables to be pinned even after their va
backptr is fixed (no longer mutable).
Signed-off-by: Keir Fraser <keir@xensource.com>
