Cleaning the python directory should completely remove the build/
directory, otherwise subsequent builds may be short-circuited and a
stale build installed.
Signed-off-by: Andrew Cooper <andrew.cooper3@citrix.com>
Acked-by: Ian Campbell <ian.campbell@citrix.com>
Committed-by: Ian Campbell <ian.campbell@citrix.com>
This avoids the hypercall buffer becoming CoW on fork.
In multi-threads and multi-processes environment, e.g. the process has two
threads, thread A may call hypercall, thread B may call fork() to create child
process. After forking, all pages of the process including hypercall buffers
are cow. It will cause a write protection and return EFAULT error if hypervisor
calls copy_to_user in hypercall in thread A context,
Fix:
1. Before hypercall: use MADV_DONTFORK of madvise syscall to make the hypercall
buffer not to be copied to child process after fork.
2. After hypercall: undo the effect of MADV_DONTFORK for the hypercall buffer
by using MADV_DOFORK of madvise syscall.
3. Use mmap/nunmap for memory alloc/free instead of malloc/free to bypass libc.
Note:
Child processes must not use the opened xc_{interface,evtchn,gnttab,gntshr}
handle that inherits from parents. They should reopen the handle if they want
to interact with xc. Otherwise, it may cause segment fault to access hypercall
buffer caches of the handle.
Signed-off-by: Zhenguo Wang <wangzhenguo@huawei.com>
Signed-off-by: Xiaowei Yang <xiaowei.yang@huawei.com>
Acked-by: Ian Campbell <ian.campbell@citrix.com>
[ ijc -- s/ptr/p/ to fix build & tweaked the wording of the comments
slightly. ]
Committed-by: Ian Campbell <ian.campbell@citrix.com>
When an external agent (e.g. a monitoring daemon) happens to access the
memory of a PoD guest prior to setting the PoD target, that access must
fail for there not being any page in the PoD cache, and only the space
above the low 2Mb gets scanned for victim pages (while only the low 2Mb
got real pages populated so far).
To accomodate for this
- set the PoD target first
- do all physmap population in PoD mode (i.e. not just large [2Mb or
1Gb] pages)
- slightly lift the restrictions enforced by p2m_pod_set_mem_target()
to accomodate for the changed tools behavior
Tested-by: Jürgen Groß <juergen.gross@ts.fujitsu.com>
(in a 4.0.x based incarnation)
Signed-off-by: Jan Beulich <jbeulich@suse.com>
Acked-by: George Dunlap <george.dunlap@eu.citrix.com>
NetBSD doesn't have a gntdev, but we should not print an error when
falling back to the previous implementation.
Signed-off-by: Roger Pau Monne <roger.pau@citrix.com>
Acked-by: Ian Campbell <ian.campbell@citrix.com>
Committed-by: Ian Campbell <ian.campbell@citrix.com>
Make some NUMA node information available to the toolstack. Achieve
this by means of xc_numainfo(), which exposes memory size and amount
of free memory of each node, as well as the relative distances of
each node to all the others.
For properly exposing distances we need the IDL to support arrays.
Signed-off-by: Dario Faggioli <dario.faggioli@citrix.com>
Acked-by: Ian Campbell <ian.campbell@citrix.com>
Committed-by: Ian Campbell <ian.campbell@citrix.com>
In xc_domain_save the local variable `ob' is initialised to NULL.
There are then various startup actions. Some of these `goto out' on
failure; for example the call to callbacks->switch_qemu_logdirty on
l.978. However, out is used both by success and error paths. So it
attempts (l.2043) to flush the current output buffer. If ob has not
yet been assigned a non-NULL value, this segfaults. So make the call
to outbuf_flush conditional on ob.
Signed-off-by: Ian Jackson <ian.jackson@eu.citrix.com>
Acked-by: Ian Campbell <ian.campbell@citrix.com>
Committed-by: Ian Jackson <Ian.Jackson@eu.citrix.com>
Up until the point where we think this is a bzImage there is no point in
printing panicy messages -- some other loader will have a go (probably the
compressed ELF one)
Signed-off-by: Ian Campbell <ian.campbell@citrix.com>
Acked-by: Ian Jackson <ian.jackson@eu.citrix.com>
Committed-by: Ian Jackson <ian.jackson@eu.citrix.com>
Introduce an empty implementation of the arch specific ARM functions in
xc_dom_arm.c.
Provide empty implementations of xc_domain_save and xc_domain_restore
when CONFIG_MIGRATE is not set.
Move xc_hvm_build.c to xc_hvm_build_x86.c because the implementation is
x86 specific, introduce xc_hvm_build_arm.c with empty stubs.
Signed-off-by: Stefano Stabellini <stefano.stabellini@eu.citrix.com>
Acked-by: Ian Jackson <Ian.Jackson@eu.citrix.com>
Acked-by: Ian Campbell <Ian.Campbell@citrix.com>
Committed-by: Ian Campbell <ian.campbell@citrix.com>
* Add libxl callback functions for Remus checkpoint suspend, postflush
(aka resume) and checkpoint commit callbacks.
* suspend callback is a stub that just bounces off
libxl__domain_suspend_common_callback - which suspends the domain and
saves the devices model state to a file.
* resume callback currently just resumes the domain (and the device model).
* commit callback just writes out the saved device model state to the
network and sleeps for the checkpoint interval.
* Introduce a new public API, libxl_domain_remus_start (currently a stub)
that sets up the network and disk buffer and initiates continuous
checkpointing.
* Future patches will augment these callbacks/functions with more functionalities
like issuing network buffer plug/unplug commands, disk checkpoint commands, etc.
Signed-off-by: Shriram Rajagopalan <rshriram@cs.ubc.ca>
Acked-by: Ian Campbell <ian.campbell@citrix.com>
Committed-by: Ian Campbell <ian.campbell@citrix.com>
Legacy (non-pvops) gntdev drivers may require this operation to be
performed when the number of grants intended to be used simultaneously
exceeds a certain driver specific default limit, and qemu's qdisk
driver is an example of needing to do so.
Signed-off-by: Jan Beulich <jbeulich@suse.com>
Committed-by: Ian Jackson <ian.jackson@eu.citrix.com>
Introduce a new save_id to save/restore toolstack specific extra
information.
Signed-off-by: Stefano Stabellini <stefano.stabellini@eu.citrix.com>
Committed-by: Ian Jackson <ian.jackson.citrix.com>
(also make note about AMD support's experimental status)
Signed-off-by: Andres Lagar-Cavilla <andres@lagarcavilla>
Acked-by: Tim Deegan <tim@xen.org>
Committed-by: Ian Jackson <ian.jackson@eu.citrix.com>
Replace alloca() with mmap() for array sizes greater than a page in
xc_linux_osdep.c.
When mapping in large amounts of pages (in the GB range) from a guest
in to Dom0 using xc_map_foreign_bulk(), a segfault occurs in the libxc
client application. This is because the pfn array in
linux_privcmd_map_foreign_bulk() is being allocated using alloca() and
the subsequent memcpy causes the stack to blow. This patch replaces
the alloca() with mmap() for pfn array sizes greater than a page.
Fix an error print with the correct function name.
Do the same for the map array in linux_gnttab_grant_map()
Signed-off-by: Aravindh Puthiyaparambil <aravindh@virtuata.com>
Acked-by: Andres Lagar-Cavilla <andres@lagarcavilla.org>
Acked-by: Ian Campbell <ian.campbell@citrix.com>
Committed-by: Ian Jackson <ian.jackson@eu.citrix.com>
It is not correct to say -lpthread. The correct option is -pthread,
which may have sundry other effects on code generation etc. It needs
to be passed both to compilation and linking.
Fix the configure test to test -pthread, and plumb the resulting flag
through to PTHREAD_{CFLAGS,LDFLAGS} in Tools.mk; also substitute
PTHREAD_LIBS (although this will currently always be empty).
Remove PTHREAD_LIBS setting from StdGNU.mk.
Fix the one user (libxc) to use PTHREAD_{CFLAGS,LDFLAGS} too.
There are still some other users in tree which pass -pthread or
-lpthread by adding it as a literal to their own compiler options.
These will be fixed in a later patch.
Signed-off-by: Ian Jackson <ian.jackson@eu.citrix.com>
Cc: Roger Pau Monne <roger.pau@entel.upc.edu>
Acked-by: Roger Pau Monne <roger.pau@entel.upc.edu>
Committed-by: Ian Jackson <Ian.Jackson@eu.citrix.com>
One of the bottlenecks with foreign page-in request is the poor retry
handling in linux_privcmd_map_foreign_bulk(). It sends one request per
paged gfn at a time and it waits until the gfn is accessible. This
causes long delays in mmap requests from qemu-dm and xc_save.
Instead of sending one request at a time, walk the entire gfn list and
send batches of mmap requests. They will eventually end up in the pager's
request ring (if it has room again), and will fill up this ring so that
in turn the pager can also process page-in in batches.
Signed-off-by: Olaf Hering <olaf@aepfle.de>
Committed-by: Ian Jackson <ian.jackson.citrix.com>
This solves a long-standing issue in which the pages backing these rings were
pages belonging to dom0 user-space processes. Thus, if the process would die
unexpectedly, Xen would keep posting events to a page now belonging to some
other process.
We update all API-consumers in tree (xenpaging and xen-access).
This is an API/ABI change, so please speak up if it breaks your accumptions.
The patch touches tools, hypervisor x86/hvm bits, and hypervisor x86/mm bits.
Signed-off-by: Andres Lagar-Cavilla <andres@lagarcavilla.org>
Acked-by: Tim Deegan <tim@xen.org>
Acked-by: Ian Campbell <ian.campbell@citrix.com>
Committed-by: Tim Deegan <tim@xen.org>
Don't use the superfluous shared page, return the event channel directly as
part of the domctl struct, instead.
In-tree consumers (xenpaging, xen-access) updated. This is an ABI/API change,
so please voice any concerns.
Known pending issues:
- pager could die and its ring page could be used by some other process, yet
Xen retains the mapping to it.
- use a saner interface for the paging_load buffer.
This change also affects the x86/mm bits in the hypervisor that process the
mem_event setup domctl.
Signed-off-by: Andres Lagar-Cavilla <andres@lagarcavilla.org>
Acked-by: Tim Deegan <tim@xen.org>
Acked-by: Olaf Hering <olaf@aepfle.de>
Acked-by: Ian Campbell <ian.campbell@citrix.com>
Committed-by: Tim Deegan <tim@xen.org>
Introduce an empty implementation of the arch specific ARM functions in
xc_core_arm.c and xc_core_arm.h; define barriers on ARM.
Signed-off-by: Stefano Stabellini <stefano.stabellini@eu.citrix.com>
Committed-by: Ian Jackson <ian.jackson.citrix.com>
Acked-by: Ian Campbell <ian.campbell@citrix.com>
Acked-by: Ian Jackson <ian.jackson@eu.citrix.com>
With changeset 24364:0964341efd65 an event channel based notification of
new responses in the ringbuffer is implemented. This makes the memevent
interface obsolete. Currently a call to p2m_mem_paging_resume() is
triggered twice by xenpaging, once per memevent and once per even
channel. In practice this double call does not lead to issues because
p2m_mem_paging_resume() processes only available events.
xenpaging used the event channel notification since the beginning, but
it was a no-op until changeset mentioned above. This change removes the
unneeded XENMEM_paging_op_resume functionality. Pagers are notified via
an event channel of new requests, and now they are required to notify
the hypervisor about their responses also with an event channel.
Signed-off-by: Olaf Hering <olaf@aepfle.de>
Acked-by: Tim Deegan <tim@xen.org>
Acked-by: Ian Jackson <ian.jackson@eu.citrix.com>
Committed-by: Ian Jackson <ian.jackson@eu.citrix.com>
Add a parameter for the MMIO hole size when building a HVM domain.
This is useful for specifying a larger than normal MMIO hole to ensure
that no PCIe device's MMIO region overlaps with RAM in a guest's
physical address space. This is needed on certain systems with PCIe
switches with a broken ACS capability. On these systems, if a device
downstream of the switch attempts a DMA to a guest physical address
that overlaps with the MMIO region of another downstream device, then
the switch routes the transfer directly to the device and the read or
write never hits RAM.
Signed-off-by: David Vrabel <david.vrabel@citrix.com>
Acked-by: George Dunlap <george.dunlap@eu.citrix.com>
Committed-by: Ian Jackson <ian.jackson@eu.citrix.com>
To allow new parameters to be added to the xc_hvm_build*() family of
functions, pass them in a structure. Make the other variants fill in
the structure and call xc_hvm_build() (except for xc_hvm_build_mem()
which had no users and is removed).
The units of the mem_size and mem_target arguments are in bytes (not
MiB like the old functions).
Signed-off-by: David Vrabel <david.vrabel@citrix.com>
Acked-by: George Dunlap <george.dunlap@eu.citrix.com>
Acked-by: Ian Jackson <ian.jackson@eu.citrix.com>
Committed-by: Ian Jackson <ian.jackson@eu.citrix.com>
This allows migration to a host with less MCA banks than the source
host had, while without this patch accesses to the excess banks' MSRs
caused #GP-s in the guest after migration (and it depended on the guest
kernel whether this would be fatal).
A fundamental question is whether we should also save/restore MCG_CTL
and MCi_CTL, as the HVM save record would better be defined to the
complete state that needs saving from the beginning (I'm unsure whether
the save/restore logic allows for future extension of an existing
record).
Of course, this change is expected to make migration from new to older
Xen impossible (again I'm unsure what the save/restore logic does with
records it doesn't even know about).
The (trivial) tools side change may seem unrelated, but the code should
have been that way from the beginning to allow the hypervisor to look
at currently unused ext_vcpucontext fields without risking to read
garbage when those fields get a meaning assigned in the future. This
isn't being enforced here - should it be? (Obviously, for backwards
compatibility, the hypervisor must assume these fields to be clear only
when the extended context's size exceeds the old original one.)
A future addition to this change might be to allow configuration of the
number of banks and other MCA capabilities for a guest before it starts
(i.e. to not inherits the values seen on the first host it runs on).
Signed-off-by: Jan Beulich <jbeulich@suse.com>
Acked-by: Keir Fraser <keir@xen.org>
NetBSD libxc hypercall implementation was returning -errno on error,
instead of the actual error value from ioctl. Returning error is
easier to understand, and the caller can always check errno.
Signed-off-by: Roger Pau Monne <roger.pau@entel.upc.edu>
Acked-by: Ian Campbell <ian.campbell.com>
Committed-by: Ian Jackson <ian.jackson.citrix.com>
Reported-by: Olaf Hering <olaf@aepfle.de>
Added a comment that explains why NetBSD return hypercall->retval on
success.
Signed-off-by: Roger Pau Monne <roger.pau@entel.upc.edu>
Acked-by: Ian Campbell <ian.campbell.com>
Committed-by: Ian Jackson <ian.jackson.citrix.com>
Reported-by: Olaf Hering <olaf@aepfle.de>
Remove costly mem_sharing audits from the inline path, and instead make them
callable as a memop.
Have the audit function return the number of errors detected.
Update memshrtool to be able to trigger audits.
Set sharing audits as enabled by default.
Signed-off-by: Andres Lagar-Cavilla <andres@lagarcavilla.org>
Signed-off-by: Adin Scannell <adin@scannell.ca>
Acked-by: Tim Deegan <tim@xen.org>
Acked-by: Ian Jackson <ian.jackson@eu.citrix.com>
Committed-by: Tim Deegan <tim@xen.org>
Per page operations in the paging, sharing, and access tracking subsystems are
all implemented with domctls (e.g. a domctl to evict one page, or to share one
page).
Under heavy load, the domctl path reveals a lack of scalability. The domctl
lock serializes dom0's vcpus in the hypervisor. When performing thousands of
per-page operations on dozens of domains, these vcpus will spin in the
hypervisor. Beyond the aggressive locking, an added inefficiency of blocking
vcpus in the domctl lock is that dom0 is prevented from re-scheduling any of
its other work-starved processes.
We retain the domctl interface for setting up and tearing down
paging/sharing/mem access for a domain. But we migrate all the per page
operations to use the memory_op hypercalls (e.g XENMEM_*).
Signed-off-by: Andres Lagar-Cavilla <andres@lagarcavilla>
Signed-off-by: Adin Scannell <adin@scannell.ca>
Acked-by: Tim Deegan <tim@xen.org>
Acked-by: Ian Jackson <ian.jackson@eu.citrix.com>
Committed-by: Tim Deegan <tim@xen.org>
In some cases guests should not provide workarounds for errata even when the
physical processor is affected. For example, because of erratum 400 on family
10h processors a Linux guest will read an MSR (resulting in VMEXIT) before
going to idle in order to avoid getting stuck in a non-C0 state. This is not
necessary: HLT and IO instructions are intercepted and therefore there is no
reason for erratum 400 workaround in the guest.
This patch allows us to present a guest with certain errata as fixed,
regardless of the state of actual hardware.
Signed-off-by: Boris Ostrovsky <boris.ostrovsky@amd.com>
Acked-by: Christoph Egger <Christoph.Egger@amd.com>
Signed-off-by: Jan Beulich <jbeulich@suse.com>
Acked-by: Keir Fraser <keir@xen.org>
Committed-by: Jan Beulich <jbeulich@suse.com>
Instead of placing string parsing inside the hypervisor, use binary
structures like other Xen hypercalls do.
Signed-off-by: Daniel De Graaf <dgdegra@tycho.nsa.gov>
Committed-by: Keir Fraser <keir@xen.org>
This library has been deprecated since July 2010; remove the in-tree
users and library.
Signed-off-by: Daniel De Graaf <dgdegra@tycho.nsa.gov>
Committed-by: Keir Fraser <keir@xen.org>
All bitmaps maintained by xc_bitops.h are used in single threaded
applications. So nothing will change the bitmaps content, adding
volatile adds just unneeded memory reloads.
xenpaging uses bitmaps alot and using non-volatile versions will
slightly improve performance.
Signed-off-by: Olaf Hering <olaf@aepfle.de>
Committed-by: Keir Fraser <keir@xen.org>
Changeset 23577:607474aeefe1 introduced an error in
xc_ia64_send_vcpumap(), bitmap_alloc() was not used correctly.
Signed-off-by: Olaf Hering <olaf@aepfle.de>
Committed-by: Keir Fraser <keir@xen.org>
mlock and munlock are implemented as macros in mini-os. Their usage
requires casting in common code. Convert them to C syntax and provide
an empty dummy function. Remove the now unneeded (void) cast from two
munlock calls.
Signed-off-by: Olaf Hering <olaf@aepfle.de>
This patch sends global VIRQs to a domain designated as the VIRQ
handler
instead of sending all global VIRQ events to dom0. This is required in
order to run xenstored in a stubdom, because VIRQ_DOM_EXC must be sent
to xenstored for domain destruction to work properly.
This patch was inspired by the xenstored stubdomain patch series sent
to xen-devel by Alex Zeffertt in 2009.
Signed-off-by: Diego Ongaro <diego.ongaro@citrix.com>
Signed-off-by: Alex Zeffertt <alex.zeffertt@eu.citrix.com>
Signed-off-by: Daniel De Graaf <dgdegra@tycho.nsa.gov>
Acked-by: Ian Campbell <ian.campbell@citrix.com>
Committed-by: Keir Fraser <keir@xen.org>
If the first ioctl fails with ENOENT it means the command is known. If a
second attempt to map each gfn happens to fail then there is no need to
run the fallback code. Some gfns are paged and the fallback code would
not fix the failure. Instead return the EINVAL to the caller.
Signed-off-by: Olaf Hering <olaf@aepfle.de>
Acked-by: Andres Lagar-Cavilla <andres@lagarcavilla.org>
Committed-by: Ian Jackson <ian.jackson@eu.citrix.com>
Bring xc_mem_paging.c in line with other users of munlock in libxc.
Otherwise it trips over -Werror=unused-value
Signed-off-by: Tim Deegan <tim@xen.org>
Keir Fraser
Andrew Cooper
Wangzhenguo
Jan Beulich
Roger Pau Monne
Ian Campbell
Dario Faggioli
Ian Jackson
Xudong Hao
Stefano Stabellini
Shriram Rajagopalan
Andres Lagar-Cavilla
Aravindh Puthiyaparambil
Olaf Hering
Roger Pau Monne
Tim Deegan
Liu, Jinsong
David Vrabel
George Dunlap
Alex Zeffertt
Santosh Jodh
Boris Ostrovsky
Daniel De Graaf