Gianni Tedesco [Wed, 4 Aug 2010 13:23:29 +0000 (14:23 +0100)]
xl: PCI code cleanups
Get rid of scan_sys_pcidir() and open-code it inside
libxl_device_pci_list_assignable() since it's not a generically re-useable
function and we're not supporting pcistub driver now. Also use macros for sysfs
dirs in libxl_device_pci_reset
Ian Campbell [Tue, 3 Aug 2010 17:10:28 +0000 (18:10 +0100)]
xl: fix memory leaks in xl create
Found using "valgrind xl create -n ..." and "valgrind xl create -e ..."
freeing config_data solves:
==18276== 944 bytes in 1 blocks are definitely lost in loss record 12 of 12
==18276== at 0x4022F0A: malloc (vg_replace_malloc.c:236)
==18276== by 0x404AEC1: libxl_read_file_contents (libxl_utils.c:258)
==18276== by 0x8056865: create_domain (xl_cmdimpl.c:1314)
==18276== by 0x8057E2D: main_create (xl_cmdimpl.c:3135)
==18276== by 0x804B2FB: main (xl.c:76)
==18276==
Adding free_domain_config() solves the following (plus presumably others
which didn't trigger because I have no devices of that type).
d_config->disks:
==18276== 61 (32 direct, 29 indirect) bytes in 1 blocks are definitely lost in loss record 9 of 12
==18276== at 0x4022F0A: malloc (vg_replace_malloc.c:236)
==18276== by 0x4022F94: realloc (vg_replace_malloc.c:525)
==18276== by 0x804E2D3: parse_config_data (xl_cmdimpl.c:715)
==18276== by 0x8056A7C: create_domain (xl_cmdimpl.c:1347)
==18276== by 0x8057E2D: main_create (xl_cmdimpl.c:3135)
==18276== by 0x804B2FB: main (xl.c:76)
d_config->vifs:
==18276== 76 (48 direct, 28 indirect) bytes in 1 blocks are definitely lost in loss record 10 of 12
==18276== at 0x4022F0A: malloc (vg_replace_malloc.c:236)
==18276== by 0x4022F94: realloc (vg_replace_malloc.c:525)
==18276== by 0x804E665: parse_config_data (xl_cmdimpl.c:779)
==18276== by 0x8056A7C: create_domain (xl_cmdimpl.c:1347)
==18276== by 0x8057E2D: main_create (xl_cmdimpl.c:3135)
==18276== by 0x804B2FB: main (xl.c:76)
Signed-off-by: Ian Campbell <ian.campbell@citrix.com> Signed-off-by: Stefano Stabellini <stefano.stabellini@eu.citrix.com>
Ian Campbell [Tue, 3 Aug 2010 17:09:21 +0000 (18:09 +0100)]
libxl: fix memory leak in libxl_name_to_domid
Found with "valgrind xl destroy ...":
==16272== 53,248 bytes in 1 blocks are definitely lost in loss record 6 of 6
==16272== at 0x4022249: calloc (vg_replace_malloc.c:467)
==16272== by 0x403FD4A: libxl_list_domain (libxl.c:490)
==16272== by 0x404B901: libxl_name_to_domid (libxl_utils.c:65)
==16272== by 0x804B4D2: domain_qualifier_to_domid (xl_cmdimpl.c:181)
==16272== by 0x804B50F: find_domain (xl_cmdimpl.c:198)
==16272== by 0x804D70C: destroy_domain (xl_cmdimpl.c:2104)
==16272== by 0x8054E4C: main_destroy (xl_cmdimpl.c:2912)
==16272== by 0x804B2FB: main (xl.c:76)
Signed-off-by: Ian Campbell <ian.campbell@citrix.com> Signed-off-by: Stefano Stabellini <stefano.stabellini@eu.citrix.com>
Gianni Tedesco [Tue, 3 Aug 2010 16:34:08 +0000 (17:34 +0100)]
libblktapctl: fix use-after-free bug
This has not caused crashes because generally use after free is OK
provided nothing else is going on. However the patch makes things
correct. It also allows us to use heap poisoning feature of valgrind on
tools linking to libblktapctl.
Ian Campbell [Fri, 30 Jul 2010 15:20:48 +0000 (16:20 +0100)]
libxc: free thread specific hypercall buffer on xc_interface_close
The per-thread hypercall buffer is usually cleaned up on pthread_exit
by the destructor passed to pthread_key_create. However if the calling
application is not threaded then the destructor is never called.
This frees the data for the current thread only but that is OK since
any other threads will be cleaned up by the destructor.
Signed-off-by: Ian Campbell <ian.campbell@citrix.com> Acked-by: Ian Jackson <ian.jackson@eu.citrix.com> Acked-by: Stefano Stabellini <stefano.stabellini@eu.citrix.com> Committed-by: Ian Jackson <ian.jackson@eu.citrix.com>
Ian Campbell [Fri, 30 Jul 2010 14:15:24 +0000 (15:15 +0100)]
xl: return(N) from individual commands to top level instead of exit(N)
This allows the top level command dispatcher to cleanup some of its
own allocations.
This is a fairly mechanical conversion of exit(FOO) into return FOO
for the top-level command functions (i.e. main_*). There are still
code paths which will exit() further down the call chains which will
require actual thought.
At first glance not all the return codes are which one would normally
expect for process exit codes (e.g. some are negative) but I didn't
attempt to address that here.
Signed-off-by: Ian Campbell <ian.campbell@citrix.com> Committed-by: Ian Jackson <ian.jackson@eu.citrix.com>
Ian Jackson [Fri, 30 Jul 2010 14:12:56 +0000 (15:12 +0100)]
xl: support "xl list <domain>"
xm list takes an optional domain argument; make xl list do likewise.
This also gets rid of a small amount of code which was duplicated
between list_domains and list_domains_details.
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>
Ian Jackson [Fri, 30 Jul 2010 14:12:10 +0000 (15:12 +0100)]
libxl: const-correctness for libxl_uuid2string
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>
If the CPUID limit bit in MSR_IA32_MISC_ENABLE is set, clear it to
make all CPUID information available. This is required for some
features to work, such as MWAIT in cpuidle, get cpu topology, XSAVE,
etc.
Paul Durrant [Thu, 29 Jul 2010 17:58:19 +0000 (18:58 +0100)]
tools/misc: xen-hvmcrash: New tool to attempt to crash HVM guests
This tool reads the CPU save records, overwrites RIP with a bogus
value, and then restores them. This is, of course, not guaranteed
to crash the guest (since the CPUs may not be executing in kernel
at the time) but it's good for breaking into some tight loops that
would be hard to debug otherwise.
Signed-off-by: Paul Durrant <paul.durrant@citrix.com> Signed-off-by: Ian Jackson <ian.jackson@eu.citrix.com> Committed-by: Ian Jackson <ian.jackson@eu.citrix.com>
libxl: Check whether a PCI device is assignable before assigning it do a domU
Implement a new libxl function libxl_device_pci_list_assignable. This is
used to implement the xl list-assignable-pci-devices command and part of
the implementation is used to make sure that PCI devices are not multiply
assigned to one or more domU's before doing the passthrough assignment.
The function libxl_device_pci_list changes to libxl_device_pci_list_assigned
due to a parameter change for consistency with pci_list_assignable.
Signed-off-by: Gianni Tedesco <gianni.tedesco@citrix.com> Acked-by: Stefano Stabellini <stefano.stabellini@eu.citrix.com> Committed-by: Ian Jackson <ian.jackson@eu.citrix.com>
21866:e017930af272 causes SyntaxError on Python 2.4 or earlier, since
the syntax "true_value if condition else false_value" was introduced
by Python 2.5.
Signed-off-by: KUWAMURA Shin'ya <kuwa@jp.fujitsu.com> Committed-by: Ian Jackson <ian.jackson@eu.citrix.com>
Christoph Egger [Thu, 29 Jul 2010 15:12:50 +0000 (16:12 +0100)]
libxl: Portability fixes for pty handling, in libxl_osdeps.h
This is the same approach used in tools/console, although we have a
separate instance of it as it is difficult to share code between these
two places.
Also add a missing include of <termios.h>, and fix a literal use of
-lutil to refer to UTIL_LIBS.
Signed-off-by: Christoph Egger <Christoph.Egger@amd.com> Acked-by: Ian Jackson <ian.jackson@eu.citrix.com> Committed-by: Ian Jackson <ian.jackson@eu.citrix.com>
Ian Campbell [Thu, 29 Jul 2010 14:56:13 +0000 (15:56 +0100)]
Subject: xl: log unknown domain shutdown reason and default to destroy
Signed-off-by: Ian Campbell <ian.campbell@citrix.com> Acked-by: Christoph Egger <Christoph.Egger@amd.com> Committed-by: Ian Jackson <ian.jackson@eu.citrix.com>
There isn't seem to be a way to directly check for EPT, so instead
check for HAP and an Intel processor. If EPT isn't enabled, then
return an error to the tool.
There are a few places in Xen where we walk a domain's page lists
without holding the page_alloc lock. They race with updates to the
page lists, which are normally rare but can be quite common under PoD
when the domain is close to its memory limit and the PoD reclaimer is
busy. This patch protects those places by taking the page_alloc lock.
I think this is OK for the two debug-key printouts - they don't run
from irq context and look deadlock-free. The tboot change seems safe
too unless tboot shutdown functions are called from irq context or
with the page_alloc lock held. The p2m one is the scariest but there
are already code paths in PoD that take the page_alloc lock with the
p2m lock held so it's no worse than existing code.
Ian Campbell [Tue, 27 Jul 2010 15:58:51 +0000 (16:58 +0100)]
libxl: signal caller if domain already destroyed on domain death event
Currently libxl_event_get_domain_death_info returns 0 if the event was
not a domain death event and 1 if it was but does not infom the user
if someone else has already cleaned up the domain, which means the
caller must replicate some of the logic from within libxl.
Instead have the libxl_event_get_XXX_info functions required that the
event is of the right type (the caller must have recently switched on
event->type anyway).
This allows the return codes to be used in an event specific way and
we take advantage of this by returning an error from
libxl_event_get_domain_death_info if the domain is not dying.
Signed-off-by: Ian Campbell <ian.campbell@citrix.com> Committed-by: Ian Jackson <ian.jackson@eu.citrix.com>
xend (XenAPI): Error in Xend-API method VM_set_actions_after_crash
Xend-API defines the method VM_set_actions_after_crash with valid
action names coredump_and_destroy, coredump_and_restart,... . These
values have to be converted into internal representation
"coredump-destroy", "coredump-restart", ... otherwise start of the
domain is rejected. Same error occurs, if I try to create a VM using
the Xend-API with actions_after_crash set to coredump_and_destroy.
Could you please apply my patch to xen-4-0-testing, too.
Signed-off-by: Lutz Dube Lutz.Dube@ts.fujitsu.com Acked-by: Jim Fehlig <jfehlig@novell.com> Committed-by: Ian Jackson <ian.jackson@eu.citrix.com>
Olaf Hering [Mon, 26 Jul 2010 10:24:51 +0000 (11:24 +0100)]
libxenguest: correct order of zlib linker options to avoid link errors
The order of objects and linked libraries is important with recent
toolchain when --as-needed is used.
Introduce new variable COMPRESSION_LIBS and place it at the end of cc
commandline options. Without this change, symbols in libbzip2 are not
found when building with openSuSE 11.2 and newer.
Signed-off-by: Olaf Hering <olaf@aepfle.de> Committed-by: Ian Jackson <ian.jackson@eu.citrix.com>
Ian Jackson [Mon, 26 Jul 2010 10:15:40 +0000 (11:15 +0100)]
tools/hotplug etc.: remove Default-Enabled from LSB header
"Default-Enabled:" is not a LSB keyword.
Remove it from the LSB section.
This avoids an openSuSE buildcheck warning and we are not aware of any
reason why it should be kept.
From: Olaf Hering <olaf@aepfle.de> Signed-off-by: Olaf Hering <olaf@aepfle.de> Modified-by: Ian Jackson <ian.jackson@eu.citrix.com> Committed-by: Ian Jackson <ian.jackson@eu.citrix.com>
Revert df9d8319bd37 Fix blktap2 NetBSD build and also revert broken change e76befc7fe2d portability fixes from tools/console 24277e3237ca Fix linking error when creating the xl binary.
Ian Campbell [Fri, 23 Jul 2010 17:12:42 +0000 (18:12 +0100)]
xl: do not try and auto re-connect console on reboot
It is not possible to run the console client if we are rebooting a
guest via the backgrounded xl process so we may as well turn off
console autoconnect after the first boot.
Signed-off-by: Ian Campbell <ian.campbell@citrix.com> Committed-by: Ian Jackson <ian.jackson@eu.citrix.com>
Ian Campbell [Fri, 23 Jul 2010 17:11:03 +0000 (18:11 +0100)]
xl: use more descriptive names than info1 and info2
c_info and b_info appears to be the common idiom in other bits of xl.
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>
Olaf Hering [Fri, 23 Jul 2010 16:50:05 +0000 (17:50 +0100)]
tools and Makefiles: introduce and use new SYSCONFIG_DIR variable
On SuSE the configuration files in /etc/sysconfig/ will be updated in
a smart way from templates found in /var/adm/fillup-templates/
Introduce new variable SYSCONFIG_DIR for Linux installations. Now
only one place needs patching, or it can be specified as option to
make(1).
Signed-off-by: Olaf Hering <olaf@aepfle.de> Acked-by: Ian Jackson <ian.jackson@eu.citrix.com> Committed-by: Ian Jackson <ian.jackson@eu.citrix.com>
Ian Campbell [Fri, 23 Jul 2010 16:39:32 +0000 (17:39 +0100)]
libxl/xl: correctly report domain state.
In particular distinguish between domain shutdown and crash and the
blocked and running states.
Signed-off-by: Ian Campbell <ian.campbell@citrix.com> Acked-by: Stefano Stabellini <stefano.stabellini@eu.citrix.com> Acked-by: Ian Jackson <ian.jackson@eu.citrix.com> Committed-by: Ian Jackson <ian.jackson@eu.citrix.com>
Ian Campbell [Fri, 23 Jul 2010 16:35:59 +0000 (17:35 +0100)]
libxl: do not create /vss and /xapi paths in xenstore.
libxl doesn't know anything about these paths and there is no reason
it needs to manage them.
Signed-off-by: Ian Campbell <ian.campbell@citrix.com> Acked-by: Stefano Stabellini <stefano.stabellini@eu.citrix.com> Committed-by: Ian Jackson <ian.jackson@eu.citrix.com>
tools/xend: Fix performance of xend with more than 10000 FC device paths
On server startup xend start or a later xend restart needs approx. 30 min to
start/restart. Without attached FC devices xend start/restart needs only some
seconds.
server type: Fujitsu Primergy RX600-S5
The time gets lost in xen/xend/XendNode.py line 329 while calling
vscsi_util.get_all_scsi_device().
329 for pscsi_record in vscsi_util.get_all_scsi_devices():
330 scsi_id = pscsi_record['scsi_id']
331 if scsi_id:
332 saved_HBA_uuid = None
I think, in most cases we don't need all the PSCSI devices registered in
xend, but only a few of it.
So a good solution for this perforamce issue is to scan only the SCSI device
paths we need, controlled by a new option in xend-config.sxp.
I have made a patch to allow specification of scsi devices we need in xend
in the config file xend-config.sxp.
The new options pscsi-device-mask expects a list of device ids oder partial
device ids like the option of lsscsi, e.g.
(pscsi-device-mask ('<partial-dev-id1' 'partial-dev-id2' ...))
Without this option set in xend-config.sxp or if lsscsi is not support, all
device paths are process like today.
Signed-off-by: Lutz Dube Lutz.Dube@ts.fujitsu.com
Comment from Masaki Kanno <kanno.masaki@jp.fujitsu.com>: "Well done" Committed-by: Ian Jackson <ian.jackson@eu.citrix.com>
Lots of places in libxl return -1 as an error which is inconsistent with
libxl error codes since that is ERROR_VERSION. Also in other places the
xc_* function to implement a command is called but the return value is
either never checked or not passed on.
This patch makes callers of libxc functions use XL_LOG_ERRNO to log a
suitable message including the contents of errno, before returning
ERROR_FAIL, if an xc call fails. (Except for pci device removals and
domain destroys where release of resources ought to continue even if
an error occurs in an early step. In this case the error is still
logged at least.)
Also a bug is fixed where a caller depends on errno being set but is
cleared by cleanup code which calls in to library functions which modify
errno as a side-effect.
This is version 2 of the patch previously applied as 91c486918e02.
Changed since then: Removed mapping from libxc error codes to libxl -
this is redundant because libxl ought to return -1 with errno set to a
meaningful value.
Signed-off-by: Gianni Tedesco <gianni.tedesco@citrix.com> Signed-off-by: Ian Jackson <ian.jackson@eu.citrix.com> Committed-by: Ian Jackson <ian.jackson@eu.citrix.com>
Ian Campbell [Fri, 23 Jul 2010 16:06:16 +0000 (17:06 +0100)]
libxl: remove xenstore /local/domain/<domid> after reading /vm and /vss paths
On domain destroy libxl currently tries to find the /vm/<uuid> and
/vss/<uuid> paths by reading /local/domain/<domid>/{vm,vss} but does
so after removing /local/domain/<domid> which isn't much use.
Also construct the xenstore path in destroy the same way as in create. It's not
wrong but I had to look twice to check they were doing the same thing.
Signed-off-by: Ian Campbell <ian.campbell@citrix.com> Acked-by: Stefano Stabellini <stefano.stabellini@eu.citrix.com> Committed-by: Ian Jackson <ian.jackson@eu.citrix.com>
Signed-off-by: Christoph Egger <Christoph.Egger@amd.com> Acked-by: Stefano Stabellini <stefano.stabellini@eu.citrix.com> Committed-by: Ian Jackson <ian.jackson@eu.citrix.com>
libxc: Do not bother initialising shared-info page in HVM-domain builder.
It is already zeroed by the hypervisor, and setting the evtchn_mask
array is pointless since its location varies between 32- and 64-bit
guests, and the builder can't know which is the correct bitness.
Ian Jackson [Wed, 21 Jul 2010 15:50:15 +0000 (16:50 +0100)]
Revert 91c486918e02 "More consistent error handling in libxl"
I'm afraid this pattern is wrong for two reasons. Firstly,
xc_domain_pause and functions like it do not return XC_ERROR_* values.
They typically return -1 on error, and set errno.
Secondly, the error codes from libxc are not really all that useful.
They mostly serve to identify where the error originated.
Ian Campbell [Wed, 21 Jul 2010 15:36:19 +0000 (16:36 +0100)]
stubdom: correct handling DESTDIR during build
The stubdom/Makefile expects components to install themselves to
stubdom/$(CROSS_ROOT) and passes the appropriate configure and make
variables to make this happen.
However if the top-level build uses DESTDIR then this not only effects
the installation targets in stubdom/Makefile but is also propagated to
the individual components' build causing them to install into an
unexpected location and breaking the build.
In other words "make -C stubdom DESTDIR=/tmp/a-dest-dir" fails due to missing
headers because they had been installed into /tmp/a-dest-dir instead of
stubdom/$(CROSS_ROOT).
The existing attempt to clear DESTDIR for the newlib install phase was
insufficient.
Define a CROSS_MAKE variable which should be used to recurse into the
cross compiled components and which clears DESTDIR.
Also move the definition of the genpath target to after the first
"all" target so that the default target is all.
I have confirmed that with this fix only the Xen build system outputs end up in
/tmp/a-dest-dir.
Signed-off-by: Ian Campbell <ian.campbell@citrix.com> Committed-by: Ian Jackson <ian.jackson@eu.citrix.com>
Attached patch fixes linking error when creating the xl binary.
The xl binary compiles in xl_cmdimpl.c which calls code from
libxl_paths.c and libxl_bootloader.c.
Signed-off-by: Christoph Egger <Christoph.Egger@amd.com>
hvmloader: clear the xenbus event-channel when we're done with it.
Otherwise a later xenbus client that naively waits for the rising edge
could get stuck.
Lots of places in libxl return -1 as an error which is inconsistent with
libxl error codes since that is ERROR_VERSION. Also in other places the
xc_* function to implement a command is called but the return value is
either never checked or not passed on.
This patch introduces a new internal function libxl_xc_error() which
maps xc error codes to xl error codes. Callers of libxc functions are
converted to use this when returning error codes back to libxl callers.
Also a bug is fixed where a caller depends on errno being set but is
cleared by cleanup code which calls in to library functions which modify
errno as a side-effect.
This patch introduces libxl_primary_console_exec: a new libxl function
that finds the domid and console number corresponding to the primary
console of a given vm. The domid might be different from the domid of
the VM and the console number might not be 0 when using stubdoms.
The caller (xl_cmdimpl.c in this case) has to make sure that the stubdom
is already created before calling libxl_primary_console_exec in the hvm
case. In the PV case libxl_primary_console_exec has to be called before
libxl_run_bootloader.
x86: use cpuid vector 0xb when available for detecting cpu topology
cpuid leaf 0xb provides extended topology enumeration. This interface
provides the 32-bit x2APIC id of the logical processor and it also
provides a new mechanism to detect SMT and core siblings (which
provides increased addressability).
Signed-off-by: Weidong Han <weidong.han@intel.com>
All logical processors with APIC ID values of 255 and greater will
have their APIC reported through Processor X2APIC structure (type-9
entry type) and all logical processors with APIC ID less than 255 will
have their APIC reported through legacy Processor Local APIC (type-0
entry type) only. This is the same case even for NMI structure
reporting.
The Processor X2APIC Affinity structure provides the association
between the X2APIC ID of a logical processor and the proximity domain
to which the logical processor belongs.
This patch adds 2 new subtables to MADT and one new subtable to SRAT.
This patch also changes x86_acpiid_to_apicid from u8 to u32 for x2APIC
ID, and changes mp_register_lapic to accept 32-bit id. But there are
still some 8-bit apic id hardcode and assumptions in Xen code, it
needs to be fixed in future.
Signed-off-by: Weidong Han <weidong.han@intel.com>
xen: Send the debug VIRQ to guests after the rest of the domain dump is done.
Send the debug VIRQ to guests after the rest of the domain dump is
done. This stops all the 'q' debug-key output getting interleaved with
the debug-virq output from a pv-ops dom0 kernel.
projects / xen.git / log