Keir Fraser [Tue, 22 Jan 2008 17:20:39 +0000 (17:20 +0000)]
vmx: Disable vmxassist in the default build.
We now use x86 real-mode emulation by default.
Select old behaviour (for now) by specifying vmxassist=y at build time. Signed-off-by: Keir Fraser <keir.fraser@citrix.com>
Keir Fraser [Tue, 22 Jan 2008 17:18:51 +0000 (17:18 +0000)]
vmx realmode: Faster emulation of REP INS and REP OUTS.
This makes, in particular, reading files via the BIOS disk abstraction
much much much faster. Signed-off-by: Keir Fraser <keir.fraser@citrix.com>
Keir Fraser [Tue, 22 Jan 2008 09:51:49 +0000 (09:51 +0000)]
xend: On block-attach, remove device information when VmError occurs.
I tested xm block-attach command with a wrong
parameter(file:). Naturally a command error occurred. Then I retested
xm block-attach command with a correct parameter(phy:). But a command
error occurred again. The second command error occurred because Xend
did not remove device information from self.info when the first
command error occurred.
Keir Fraser [Tue, 22 Jan 2008 09:50:06 +0000 (09:50 +0000)]
hvm, ioemu: Fix the ide/cdrom emulation error on FC7 guest.
The new libata-eh in the Linux kernel is throwing a fit over the QEMU
cdrom device for two reasons:
1) DRQ can be set with ERR_STAT set. This is a violation of the ATAPI
state machine.
2) After a TEST_UNIT_READY ATAPI command is sent ERR_STAT is getting
set
which is correct. But, when the OS issues another ATAPI command
ERR_STAT is still set. Which is bad since the next expected command
from the OS is REQUEST_SENSE to find out why ERR_STAT is set.
bug this fixes: https://bugzilla.novell.com/show_bug.cgi?id=3D291775
Signed-off-by: Brandon Philips <bphilips@suse.de>
Porting this qemu fix to xen.
Keir Fraser [Tue, 22 Jan 2008 09:43:53 +0000 (09:43 +0000)]
Make tools/Makefile cross-friendly
Pass Xen cross-compilation options through to ioemu.
Export CROSS_SYS_ROOT for use by scripts. Unexport
IOEMU_DIR since it isn't used outside of the Makefile.
Keir Fraser [Mon, 21 Jan 2008 14:39:33 +0000 (14:39 +0000)]
hvm: Allow PCI high-memory resource to be decided by hvmloader at
HVM-guest boot time. Also make a few PCI0 resource ResourceProducer
type -- this seems to be specified by some randomly-sampled real ACPI
systems. Signed-off-by: Keir Fraser <keir.fraser@citrix.com>
Keir Fraser [Fri, 18 Jan 2008 15:28:59 +0000 (15:28 +0000)]
pv-on-hvm: fix/adjust pv driver's mkbuildtree
Allow invoking the script from other than the current directory (the
script now infers the tree is where the script lives). Print which
trees are actually being used.
Fix the linking of sources from Linux tree's drivers/xen/ - this was
needlessly invoking lndir on various sub-directories, in some cases
even twice.
Adjust the script to be able to deal with Linux 2.6.24's unified x86
include directory.
Keir Fraser [Fri, 18 Jan 2008 15:22:42 +0000 (15:22 +0000)]
minios: extend map_frames into being able to map a series of
contiguous frames, or the same frame several times, aligned, from
another domain, with specific protection, and with potential
failures.
Signed-off-by: Samuel Thibault <samuel.thibault@eu.citrix.com>
Keir Fraser [Fri, 18 Jan 2008 13:43:26 +0000 (13:43 +0000)]
Clarify 'xm mem-set' and 'xm vcpu-set' requirements in documentation.
From: Richard W.M. Jones <rjones@redhat.com> Signed-off-by: Keir Fraser <keir.fraser@citrix.com>
Keir Fraser [Thu, 17 Jan 2008 16:39:14 +0000 (16:39 +0000)]
hvm: Extboot support for Xen
This patch adds extboot to Xen. It should be pretty harmless as the
moment because it's never enabled. extboot allows arbitrary block
devices to be used to boot guests including SCSI and PV disks. I've
tested it with both Windows and Linux guests in QEMU.
Signed-off-by: Anthony Liguori <aliguori@us.ibm.com>
Keir Fraser [Thu, 17 Jan 2008 15:18:38 +0000 (15:18 +0000)]
tools: disable lomount and miniterm by default
lomount is a tool which reads and parses a partition table in a disk
image block device and then uses mount -o ...offset=... to mount it.
This is not an ideal approach. For example, if the intended
filesystem has corrupted metadata the kernel's filesystem driver may
start to write outside of the intended region. This might even be
exploitable in some perverse circumstances.
Nowadays people wanting to do this should use kpartx, which uses
devmapper to create appropriate range mappings. So lomount should be
disabled.
miniterm may well be useful but it is a clone-and-hack of an upstream
project and is currently built but not installed by default, partly
because it doesn't make sense to install on the dom0 which it might be
trying to debug.
It is probably useful to retain these two programs in the source tree
but IMO they should no longer be built by default.
The attached patch does these things:
* CONFIG_LOMOUNT and CONFIG_MINITERM in Config.mk
can enable and disable these programs
* They are disabled by default
* If CONFIG_MINITERM=y it is still built but not installed.
make -C tools/misc/miniterm install will install it.
Signed-off-by: Ian Jackson <ian.jackson@eu.citrix.com>
Keir Fraser [Thu, 17 Jan 2008 15:14:17 +0000 (15:14 +0000)]
network scripts: SUSE has sensible use of ifup nowadays
While wading through the patches made to Xen by OpenSUSE in their
[s]rpm, I found that they have removed various special casing for ifup
on ther distribution.
Signed-off-by: Ian Jackson <ian.jackson@eu.citrix.com>
Keir Fraser [Thu, 17 Jan 2008 15:13:40 +0000 (15:13 +0000)]
tools/docs: Fix example and default IP addresses.
In various places in documentation and code, IP addresses are provided
as examples, defaults, or dummy configuration. In general the
specific IP addresses used in Xen are not always appropriate. (For
example, 1.2.3.4 is used in a few places!)
The following addresses should be used:
* For examples and documentation, 192.0.2.0/24. (See RFC3330.)
* For defaults for private networks, a random network from RFC1918.
I have randomly selected 172.30.206.0/24 for this purpose and
documented this in at the only registry I know of,
www.ucam.org/cam-grin. This network should henceforth be used for
default configurations of local bridges, test networks, etc. in
Xen tools.
The following addresses should NOT be used:
* 10.0.*.*, 10.1.*.*, 192.168.0.*, 192.168.1.*, etc. Using these
addresses gives greatly increased likelihood of collision, as
ignorant network administrators and reckless middlebox vendors
often pick networks from the bottom of 10/8 and 192.168/16.
* 169.254.*.*. These are reserved for zeroconf (ad-hoc networking)
and should not be used for Xen private networks, bridges, etc.,
etc. Use of these addresses by Xen scripts causes trouble on hosts
(eg laptops) which find themselves in ad-hoc networking
environments. I think this is not hypothetical (!) since at least
one Linux distribution have specific code to detect this case and
cause Xen startup to fail iff the host already has an external
zeroconf address.
* 1.2.3.4. WTF !?
I have also used 127.0.255.255 in one place where apparently a dummy
address is needed (some Linux kernels won't accept a lack of an NFS
server address). If 127.0.255.255 is mistakenly used it is unlikely
to do any damage to real traffic even if it does escape into the
network at large.
Signed-off-by: Ian Jackson <ian.jackson@eu.citrix.com>
Keir Fraser [Thu, 17 Jan 2008 12:59:10 +0000 (12:59 +0000)]
hvm: Allocate an unmapped guard page to separate ordinary low RAM
from special pages (ioreq, xenstore, etc). This will stop some probe
functions (e.g., embedded in memtest) from stomping on the special
pages. Signed-off-by: Keir Fraser <keir.fraser@citrix.com>
Keir Fraser [Tue, 15 Jan 2008 14:22:50 +0000 (14:22 +0000)]
hvm, x86: Add QEMU BIOS menu for choosing boot device.
The attached patch implements a mini BIOS menu for choosing a
non-default
boot device. When a guest starts it'll display
'Press F10 to select boot device'
And wait 3 seconds, before continuing with the normal boot device. If
they press the F10 key, a menu is shown allowing a choice between
floppy, harddisk, cdrom and network (PXE).
I can't take credit for this originally - Jeremy Katz wrote it for
KVM, I merely re-diffed the patch to work against Xen's QEMU/BIOS code
tree. It has been tested in Fedora successfully against 3.1.x and 3.2.x
Signed-off-by: Daniel P. Berrange <berrange@redhat.com> CC: Jeremy Katz <katzj@redhat.com>
If more than one boot device has been selected, the currently selected
number displayed in the menu is wrong.
Keir Fraser [Fri, 11 Jan 2008 10:53:44 +0000 (10:53 +0000)]
x86 hvm: hvm_maybe_deassert_evtchn_irq() needs to be called early in
vmexit handling, before IO-APIC can be EOIed or unmasked and resample
the fake INTx line. Signed-off-by: Keir Fraser <keir.fraser@citrix.com>
Keir Fraser [Thu, 10 Jan 2008 22:53:43 +0000 (22:53 +0000)]
x86_emulate: Fix near CALL/JMP <reg,mem>. Broken by c/s 16491.
Thanks to AMD for narrowing this one down. Signed-off-by: Keir Fraser <keir.fraser@citrix.com>
Keir Fraser [Wed, 9 Jan 2008 22:21:28 +0000 (22:21 +0000)]
x86_emulate: Fix handling of RIP-relative addressing.
The handling of SHLD/SHRD introduced in 16696:b5b3e27f1af3f7 was
incorrect; thanks to Gary Grebus @ Virtual Iron for spotting it. Signed-off-by: Keir Fraser <keir.fraser@citrix.com>
Keir Fraser [Wed, 9 Jan 2008 10:35:52 +0000 (10:35 +0000)]
vt-d: Fix print_vtd_entries walk VTd mapping table.
DMA request to above guest physical memory will cause VTd fault, in
which print_vtd_entries() tries to walk VTd mapping table. However,
during walking, current Xen code didn't check if the PTE is valid and
may access to invalid memory address.
Signed-off-by: Xiaowei Yang <xiaowei.yang@intel.com>
Keir Fraser [Wed, 9 Jan 2008 10:32:13 +0000 (10:32 +0000)]
hvm: hpet: Clamp period to sane values to prevent excessive looping in
hpet_timer_fn(), or excessive numbers of set_timer() callbacks. Signed-off-by: Keir Fraser <keir.fraser@citrix.com>
projects / xen.git / log