The gnulib testsuite is relatively stable - the only times it is
likely to have a test change from pass to fail is on a gnulib
submodule update or a major system change (such as moving from
Fedora 18 to 19, or other large change to libc). While it is an
important test for end users on arbitrary machines (to make sure
that the portability glue works for their machine), it mostly
wastes time for development testing (as most developers aren't
making any of the major changes that would cause gnulib tests
to alter behavior). Thus, it pays to make the tests optional
at configure time, defaulting to off for development, on for
tarballs, with autobuilders requesting it to be on. It also
helps to allow a make-time override, via VIR_TEST_EXPENSIVE=[01]
(much the way automake sets up V=[01] for overriding the configure
time default of how verbose to be).
Automake has some pretty hard-coded magic with regards to the
TESTS variable; I had quite a job figuring out how to keep
'make distcheck' passing regardless of the configure option
setting in use, while still disabling the tests at runtime
when I did not configure them on and did not use the override
variable. Thankfully, we require GNU make, which lets me
hide some information from Automake's magic handling of TESTS.
* bootstrap.conf (bootstrap_epilogue): Munge gnulib test variable.
* configure.ac (--enable-expensive-tests): Add new enable switch.
(VIR_TEST_EXPENSIVE_DEFAULT, WITH_EXPENSIVE_TESTS): Set new
witnesses.
* gnulib/tests/Makefile.am (TESTS): Make tests conditional on
configure settings and the VIR_TEST_EXPENSIVE variable.
* tests/Makefile.am (TESTS_ENVIRONMENT): Expose VIR_TEST_EXPENSIVE
to all tests.
* autobuild.sh: Enable all tests during autobuilds.
* libvirt.spec.in (%configure): Likewise.
* mingw-libvirt.spec.in (%mingw_configure): Likewise.
* docs/hacking.html.in: Document the option.
* HACKING: Regenerate.
Signed-off-by: Eric Blake <eblake@redhat.com>
https://bugzilla.redhat.com/show_bug.cgi?id=951637
Newer gnutls uses nettle, rather than gcrypt, which is a lot nicer
regarding initialization. Yet we were unconditionally initializing
gcrypt even when gnutls wouldn't be using it, and having two crypto
libraries linked into libvirt.so is pointless, but mostly harmless
(it doesn't crash, but does interfere with certification efforts).
There are three distinct version ranges to worry about when
determining which crypto lib gnutls uses, per these gnutls mails:
2.12: http://lists.gnu.org/archive/html/gnutls-devel/2011-03/msg00034.html
3.0: http://lists.gnu.org/archive/html/gnutls-devel/2011-07/msg00035.html
If pkg-config can prove version numbers and/or list the crypto
library used for static linking, we have our proof; if not, it
is safer (even if pointless) to continue to use gcrypt ourselves.
* configure.ac (WITH_GNUTLS): Probe whether to add -lgcrypt, and
define a witness WITH_GNUTLS_GCRYPT.
* src/libvirt.c (virTLSMutexInit, virTLSMutexDestroy)
(virTLSMutexLock, virTLSMutexUnlock, virTLSThreadImpl)
(virGlobalInit): Honor the witness.
* libvirt.spec.in (BuildRequires): Make gcrypt usage conditional,
no longer needed in Fedora 19.
Signed-off-by: Eric Blake <eblake@redhat.com>
Commit d72ef888 introduced a bug in the libxl driver that will
segfault libvirtd if libxl reports an error message, e.g. when
attempting to initialize the driver on a non-Xen system. I
assumed it was valid to pass a NULL logger to libxl_ctx_alloc(),
but that is not the case since any errors associated with the ctx
that are emitted by libxl will dereference the logger and crash
libvirtd.
Errors associated with the libxl driver-wide ctx could be useful
for debugging anyway, so create a 'libxl-driver.log' to capture
these errors.
Recentish (2011) kernels introduced a new device called /dev/loop-control,
which causes libvirt's detection of loop devices to get confused
since it only checks for a prefix of 'loop'. Also check that the
next character is a digit
Signed-off-by: Daniel P. Berrange <berrange@redhat.com>
This adds two new pages to the website, acl.html describing
the general access control framework and permissions models,
and aclpolkit.html describing the use of polkit as an
access control driver.
page.xsl is modified to support a new syntax
<div id="include" filename="somefile.htmlinc"/>
which will cause the XSL transform to replace that <div>
with the contents of 'somefile.htmlinc'. We use this in
the acl.html.in file, to pull the table of permissions
for each libvirt object. This table is autogenerated
from the enums in src/access/viraccessperms.h by the
genaclperms.pl script.
newapi.xsl is modified so that the list of permissions
checks shown against each API will link to the description
of the permissions in acl.html
Signed-off-by: Daniel P. Berrange <berrange@redhat.com>
The gendispatch.pl script puts comments at the top of files
it creates, saying that it auto-generated them. Also include
the name of the source data file which it reads when doing
the auto-generation.
Signed-off-by: Daniel P. Berrange <berrange@redhat.com>
introduced by cs 4b9eec50fe ("libxl: implement per
NUMA node free memory reporting"). What was wrong was that
libxl_get_numainfo() put in nr_nodes the actual number of
host NUMA nodes, not the highest node ID (like libnuma's
numa_max_node() does instead).
While at it, turn the failure of libxl_get_numainfo() from
a simple warning to a proper error, as requested during the
review of another patch of the original series.
Signed-off-by: Dario Faggioli <dario.faggioli@citrix.com>
Cc: Daniel P. Berrange <berrange@redhat.com>
This is a second attempt at fixing the problem first attempted
in commit 2df8d99; basically undoing the fact that it was
reverted in commit 43cee32f, plus fixing two more issues: the
code in configure.ac has to EXACTLY match virnetdevbridge.c
with regards to declaring in6 types before using if_bridge.h,
and the fact that RHEL 5 has even more conflicts:
In file included from util/virnetdevbridge.c:49:
/usr/include/linux/in6.h:47: error: conflicting types for 'in6addr_any'
/usr/include/netinet/in.h:206: error: previous declaration of 'in6addr_any' was here
/usr/include/linux/in6.h:49: error: conflicting types for 'in6addr_loopback'
/usr/include/netinet/in.h:207: error: previous declaration of 'in6addr_loopback' was here
The rest of this commit message borrows from the original try
of 2df8d99:
A fresh checkout on a RHEL 6 machine with these packages:
kernel-headers-2.6.32-405.el6.x86_64
glibc-2.12-1.128.el6.x86_64
failed to configure with this message:
checking for linux/if_bridge.h... no
configure: error: You must install kernel-headers in order to compile libvirt with QEMU or LXC support
Digging in config.log, we see that the problem is identical to
what we fixed earlier in commit d12c2811:
configure:98831: checking for linux/if_bridge.h
configure:98853: gcc -std=gnu99 -c -g -O2 conftest.c >&5
In file included from /usr/include/linux/if_bridge.h:17,
from conftest.c:559:
/usr/include/linux/in6.h:31: error: redefinition of 'struct in6_addr'
/usr/include/linux/in6.h:48: error: redefinition of 'struct sockaddr_in6'
/usr/include/linux/in6.h:56: error: redefinition of 'struct ipv6_mreq'
configure:98860: $? = 1
I had not hit it earlier because I was using incremental builds,
where config.cache had shielded me from the kernel-headers breakage.
* configure.ac (if_bridge.h): Avoid conflicting type definitions.
* src/util/virnetdevbridge.c (includes): Also sanitize for RHEL 5.
Signed-off-by: Eric Blake <eblake@redhat.com>
Commit f1088c8 weakened a test, by not passing a value larger
than INT_MAX through an int slot. Make the fix in a different
way, using an explicit negative value. Suggested by Dan Berrange.
* tests/virdbustest.c (testMessageArray): Adjust previous fix.
(testMessageStruct): Use a negative number.
Signed-off-by: Eric Blake <eblake@redhat.com>
Currently, only one log file is created by the libxl driver, with
all output from libxl for all domains going to this one file.
Create a per-domain log file based on domain name, making sifting
through the logs a bit easier. This required deferring libxl_ctx
allocation until starting the domain, which is fine since the
ctx is not used when the domain is inactive.
Tested-by: Dario Faggioli <dario.faggioli@citrix.com>
The virtlockd daemon supports an /etc/libvirt/virtlockd.conf
config file, but we never installed a default config, nor
created any augeas scripts. This change addresses that omission.
Signed-off-by: Daniel P. Berrange <berrange@redhat.com>
Compiling with gcc 4.1.2 (RHEL 5) on a 32-bit platform complains:
virdbustest.c: In function 'testMessageSimple':
virdbustest.c:61: warning: integer constant is too large for 'long' type
virdbustest.c:62: warning: integer constant is too large for 'long' type
virdbustest.c: In function 'testMessageArray':
virdbustest.c:183: warning: this decimal constant is unsigned only in ISO C90
virdbustest.c: In function 'testMessageStruct':
virdbustest.c:239: warning: integer constant is too large for 'long' type
virdbustest.c:240: warning: integer constant is too large for 'long' type
* tests/virdbustest.c (testMessageSiple, testMessageArray)
(testMessageStruct): Don't violate C89 constant constraints.
Signed-off-by: Eric Blake <eblake@redhat.com>
The virLoginShellAllowedUser method must not free the 'groups'
parameter it is given, as that is owned by the caller.
The virLoginShellAllowedUser method should be checking
'!*ptr' (ie empty string) rather than '!ptr' (NULL string)
since the latter cannot be true.
Signed-off-by: Daniel P. Berrange <berrange@redhat.com>
Coverity complained about the usage of the uninitialized cacerts in the
event(s) that "access(certFile, R_OK)" and/or "access(cacertFile, R_OK)"
fail the for loop used to fill in the certs will have indeterminate data
as well as the possibility that both failures would result in the
gnutls_x509_crt_deinit() call having a similar fate.
Initializing cacerts only would resolve the issue; however, it still
would leave the indeterminate action, so rather add a parameter to
the virNetTLSContextLoadCACertListFromFile() to pass the max size rather
then overloading the returned count parameter. If the the call is never
made, then we won't go through the for loops referencing the empty
cacerts
Use a separate keyfile name for the two TLS test suites so that
they don't clash when running tests in parallel
Signed-off-by: Daniel P. Berrange <berrange@redhat.com>
Valgrind defects memory error:
==16759== 1 errors in context 1 of 8:
==16759== Invalid free() / delete / delete[] / realloc()
==16759== at 0x4A074C4: free (in /usr/lib64/valgrind/vgpreload_memcheck-amd64-linux.so)
==16759== by 0x83CD329: xdr_string (in /usr/lib64/libc-2.17.so)
==16759== by 0x4D93E4D: xdr_remote_nonnull_string (remote_protocol.c:31)
==16759== by 0x4D94350: xdr_remote_nonnull_domain (remote_protocol.c:58)
==16759== by 0x4D976C8: xdr_remote_domain_create_with_flags_ret (remote_protocol.c:1762)
==16759== by 0x83CC734: xdr_free (in /usr/lib64/libc-2.17.so)
==16759== by 0x4D7F1E0: remoteDomainCreateWithFlags (remote_driver.c:2441)
==16759== by 0x4D4BF17: virDomainCreateWithFlags (libvirt.c:9499)
==16759== by 0x13127A: cmdStart (virsh-domain.c:3376)
==16759== by 0x12BF83: vshCommandRun (virsh.c:1751)
==16759== by 0x126FFB: main (virsh.c:3205)
==16759== Address 0xe1394a0 is not stack'd, malloc'd or (recently) free'd
==16759== 1 errors in context 2 of 8:
==16759== Conditional jump or move depends on uninitialised value(s)
==16759== at 0x4A07477: free (in /usr/lib64/valgrind/vgpreload_memcheck-amd64-linux.so)
==16759== by 0x83CD329: xdr_string (in /usr/lib64/libc-2.17.so)
==16759== by 0x4D93E4D: xdr_remote_nonnull_string (remote_protocol.c:31)
==16759== by 0x4D94350: xdr_remote_nonnull_domain (remote_protocol.c:58)
==16759== by 0x4D976C8: xdr_remote_domain_create_with_flags_ret (remote_protocol.c:1762)
==16759== by 0x83CC734: xdr_free (in /usr/lib64/libc-2.17.so)
==16759== by 0x4D7F1E0: remoteDomainCreateWithFlags (remote_driver.c:2441)
==16759== by 0x4D4BF17: virDomainCreateWithFlags (libvirt.c:9499)
==16759== by 0x13127A: cmdStart (virsh-domain.c:3376)
==16759== by 0x12BF83: vshCommandRun (virsh.c:1751)
==16759== by 0x126FFB: main (virsh.c:3205)
==16759== Uninitialised value was created by a stack allocation
==16759== at 0x4D7F120: remoteDomainCreateWithFlags (remote_driver.c:2423)
How to reproduce?
# virsh start <domain> --paused
RHBZ: https://bugzilla.redhat.com/show_bug.cgi?id=994855
Signed-off-by: Alex Jia <ajia@redhat.com>
On RHEL 5, with dbus 1.1.2, compilation failed with:
virsystemdmock.c: In function 'dbus_connection_send_with_reply_and_block':
virsystemdmock.c:68: warning: implicit declaration of function 'dbus_message_set_serial'
Fix this by instead bypassing all attempts to use a dbus serial.
* tests/virsystemdmock.c (dbus_message_set_reply_serial): Add new
override.
(dbus_connection_send_with_reply_and_block): No longer bother with
the serial.
Signed-off-by: Eric Blake <eblake@redhat.com>
virt-login-shell.c was failing to compile with
CC virt_login_shell-virt-login-shell.o
virt-login-shell.c: In function 'main':
virt-login-shell.c:205:5: error: implicit declaration of function 'setlocale' [-Werror=implicit-function-declaration]
virt-login-shell.c:205:5: error: nested extern declaration of 'setlocale' [-Werror=nested-externs]
virt-login-shell.c:205:20: error: 'LC_ALL' undeclared (first use in this function)
It adds an empty space after the package version. Previously the error
message looked like:
"You must install the dbus-1 >= 1.0.0pkg-config module to compile libvirt"
Signed-off-by: Giuseppe Scrivano <gscrivan@redhat.com>
Add a virt-login-shell binary that can be set as a user's
shell, such that when they login, it causes them to enter
the LXC container with a name matching their user name.
Signed-off-by: Daniel P. Berrange <berrange@redhat.com>
If securityfs is available on the host, we should ensure to
mount it read-only in the container. This will avoid systemd
trying to mount it during startup causing SELinux AVCs.
Signed-off-by: Daniel P. Berrange <berrange@redhat.com>
Hotplugging a single SCSI device works, but adding additional ones
result in an error from QEMU:
[root@gpok197 ~]# virsh attach-device guest01 blah.xml
Device attached successfully
[root@gpok197 ~]# virsh attach-device guest01 blah2.xml
error: Failed to attach device from blah2.xml
error: internal error unable to execute QEMU command 'device_add': Duplicate ID 'hostdev0' for device
The hostdev ID that is created is always set to zero, regardless
of the contents of the XML. Changing the index in the hotplug case
to a negative one so the next available index is used.
Signed-off-by: Eric Farman <farman@linux.vnet.ibm.com>
Reviewed-by: Viktor Mihajlovski <mihajlov@linux.vnet.ibm.com>
So that app developers / admins know what access control checks
are performed for each API, this patch extends the API docs
generator to include details of the ACLs for each.
The gendispatch.pl script is extended so that it generates
a simple XML describing ACL rules, eg.
<aclinfo>
...
<api name='virConnectNumOfDomains'>
<check object='connect' perm='search_domains'/>
<filter object='domain' perm='getattr'/>
</api>
<api name='virDomainAttachDeviceFlags'>
<check object='domain' perm='write'/>
<check object='domain' perm='save' flags='!VIR_DOMAIN_AFFECT_CONFIG|VIR_DOMAIN_AFFECT_LIVE'/>
<check object='domain' perm='save' flags='VIR_DOMAIN_AFFECT_CONFIG'/>
</api>
...
</aclinfo>
The newapi.xsl template loads the XML files containing the ACL
rules and generates a short block of HTML for each API describing
the parameter checks and return value filters (if any).
Signed-off-by: Daniel P. Berrange <berrange@redhat.com>
The code added to validate CA certificates did not take into
account the possibility that the cacert.pem file can contain
multiple (concatenated) cert data blocks. Extend the code for
loading CA certs to use the gnutls APIs for loading cert lists.
Add test cases to check that multi-level trees of certs will
validate correctly.
Signed-off-by: Daniel P. Berrange <berrange@redhat.com>
Currently a 'struct testTLSCertReq' instance is passed into
the TLS test cases. This is not flexible enough to cope with
certificate chains, where one file now corresponds to multiple
certificates. Change the test cases so that we pass in filenames
instead.
Signed-off-by: Daniel P. Berrange <berrange@redhat.com>
Currently every test case in the TLS test suite generates the
certs fresh. This is a waste of time, since its parameters
don't change across test cases. Create certs once in main
method.
Signed-off-by: Daniel P. Berrange <berrange@redhat.com>
The virnettlscontexttest.c tests both virNetTLSContext
and virNetTLSSession functionality. Split into two
separate tests, to make the code size more manageable
Signed-off-by: Daniel P. Berrange <berrange@redhat.com>
Commit 3d0e3c1 reintroduced a problem previously squelched in
commit 7e5aa78. Add a syntax check this time around.
util/virutil.c: In function 'virGetGroupList':
util/virutil.c:1015: error: 'for' loop initial declaration used outside C99 mode
* cfg.mk (sc_prohibit_loop_var_decl): New rule.
* src/util/virutil.c (virGetGroupList): Fix offender.
Signed-off-by: Eric Blake <eblake@redhat.com>
https://bugzilla.redhat.com/show_bug.cgi?id=994589 complained that
even when using a cross-compiler not named 'gcc', the configure
output confusingly referred to gcc.
* m4/virt-compile-warnings.m4 (LIBVIRT_COMPILE_WARNINGS): Use a
more generic statement in configure output.
Signed-off-by: Eric Blake <eblake@redhat.com>
Before, missing attributes were only OK when adding entries;
modification and deletion required all of them.
Now, only deletion works with missing attributes, as long as
the host is uniquely identified.
Go through disks of guest, if one disk doesn't exist or its backing
chain is broken, with 'optional' startupPolicy, for CDROM and Floppy
we only discard its source path definition in xml, for disks we drop
it from disk list and free it.
Since iptables version 1.4.16 '-m state --state NEW' is converted to
'-m conntrack --ctstate NEW'. Therefore, when encountering this or later
versions of iptables use '-m conntrack --ctstate'.
Signed-off-by: Stefan Berger <stefanb@linux.vnet.ibm.com>
The change from initgroups to virGetGroupList/setgroups in
cab36cfe71ba83b71e536ba5c98e596f02b697b0 dropped the primary group from
processes group list iff the passed in group to virGetGroupList differs
from the user's primary group.
So always include the primary group to bring back the old behaviour.
Debian has the kvm group as primary group but uses
libvirt-qemu:libvirt-qemu as user:group to run the kvm process so
without this change the /dev/kvm is inaccessible.
A fresh checkout on a RHEL 6 machine with these packages:
kernel-headers-2.6.32-405.el6.x86_64
glibc-2.12-1.128.el6.x86_64
failed to configure with this message:
checking for linux/if_bridge.h... no
configure: error: You must install kernel-headers in order to compile libvirt with QEMU or LXC support
Digging in config.log, we see that the problem is identical to
what we fixed earlier in commit d12c2811:
configure:98831: checking for linux/if_bridge.h
configure:98853: gcc -std=gnu99 -c -g -O2 conftest.c >&5
In file included from /usr/include/linux/if_bridge.h:17,
from conftest.c:559:
/usr/include/linux/in6.h:31: error: redefinition of 'struct in6_addr'
/usr/include/linux/in6.h:48: error: redefinition of 'struct sockaddr_in6'
/usr/include/linux/in6.h:56: error: redefinition of 'struct ipv6_mreq'
configure:98860: $? = 1
I had not hit it earlier because I was using incremental builds,
where config.cache had shielded me from the kernel-headers breakage.
* configure.ac (if_bridge.h): Avoid conflicting type definitions.
Signed-off-by: Eric Blake <eblake@redhat.com>
Since commit 95e18efd most public interfaces (xenUnified...) obtain
a virDomainDefPtr via xenGetDomainDefFor...() which take the unified
lock.
This is already taken before calling xenDomainUsedCpus(), so we get
a deadlock for active guests. Avoid this by splitting up
xenUnifiedDomainGetVcpusFlags() and xenUnifiedDomainGetVcpus() into
public and private function calls (which get the virDomainDefPtr passed)
and use those in xenDomainUsedCpus().
xenDomainUsedCpus
...
nb_vcpu = xenUnifiedDomainGetMaxVcpus(dom);
return xenUnifiedDomainGetVcpusFlags(...)
...
if (!(def = xenGetDomainDefForDom(dom)))
return xenGetDomainDefForUUID(dom->conn, dom->uuid);
...
ret = xenHypervisorLookupDomainByUUID(conn, uuid);
...
xenUnifiedLock(priv);
name = xenStoreDomainGetName(conn, id);
xenUnifiedUnlock(priv);
...
if ((ncpus = xenUnifiedDomainGetVcpus(dom, cpuinfo, nb_vcpu,
...
if (!(def = xenGetDomainDefForDom(dom)))
[again like above]
Signed-off-by: Stefan Bader <stefan.bader@canonical.com>
This patch addresses two concerns with the error reporting when an
incompatible PCI address is specified for a device:
1) It wasn't always apparent which device had the problem. With this
patch applied, any error about an incompatible address will always
contain the full address as given in the config, so it will be easier
to determine which device's config aused the problem.
2) In some cases when the problem came from bad config, the error
message was erroneously classified as VIR_ERR_INTERNAL_ERROR. With
this patch applied, the same error message will be changed to indicate
either "internal" or "xml" error depending on whether the address came
from the config, or was automatically generated by libvirt.
Note that in the case of "internal" (due to bad auto-generation)
errors, the PCI address won't be of much use in finding the location
in config to change (because it was automatically generated). Of
course that makes perfect sense, but still the address could provide a
clue about a bug in libvirt attempting to use a type of pci bus that
doesn't have its flags set correctly (or something similar). In other
words, it's not perfect, but it is definitely better.
q35 machines have an implicit ahci (sata) controller at 00:1F.2 which
has no "id" associated with it. For this reason, we can't refer to it
as "ahci0". Instead, we don't give an id on the commandline, which
qemu interprets as "use the first ahci controller". We then need to
specify the unit with "unit=%d" rather than adding it onto the bus
arg.
https://bugzilla.redhat.com/show_bug.cgi?id=979477
Since 1.0.3 we are using the new way to copy non shared storage during
migration (the NBD way). However, whether the new or old way is used is
not controllable by user but unconditionally turned on if both sides of
migration support it. Moreover, the implementation is not complete: the
combination for VIR_MIGRATE_TUNNELLED flag is missing (as we need to
open new port on the destination) in which case we just error out. This
is a deadly combination: not letting users choose their destiny and
erroring out. We should not do that but VIR_WARN and turn the NBD off
instead.
We had been setting the device alias in the devinceinfo for pci
controllers to "pci%u", but then hardcoding "pci.%u" when creating the
device address for other devices using that pci bus. This all worked
just fine until we encountered the built-in "pcie.0" bus (the PCIe
root complex) in Q35 machines.
In order to create the correct commandline for this one case, this
patch:
1) sets the alias for PCI controllers correctly, to "pci.%u" (or
"pcie.%u" for the pcie-root controller)
2) eliminates the hardcoded "pci.%u" for pci controllers when
generatuing device address strings, and instead uses the controller's
alias.
3) plumbs a pointer to the virDomainDef all the way down to
qemuBuildDeviceAddressStr. This was necessary in order to make the
aliase of the controller *used by a device* available (previously
qemuBuildDeviceAddressStr only had the deviceinfo of the device
itself, *not* of the controller it was connecting to). This made for a
larger than desired diff, but at least in the future we won't have to
do it again, since all the information we could possibly ever need for
future enhancements is in the virDomainDef. (right?)
This should be done for *all* controllers, but for now we just do it
in the case of PCI controllers, to reduce the likelyhood of
regression.
This patch adds in special handling for a few devices that need to be
treated differently for q35 domains:
usb - there is no implicit/default usb controller for the q35
machinetype. This is done because normally the default usb controller
is added to a domain by just adding "-usb" to the qemu commandline,
and it's assumed that this will add a single piix3 usb1 controller at
slot 1 function 2. That's not what happens when the machinetype is
q35, though. Instead, adding -usb to the commandline adds 3 usb
(version 2) controllers to the domain at slot 0x1D.{1,2,7}. Rather
than having
<controller type='usb' index='0'/>
translate into 3 separate devices on the PCI bus, it's cleaner to not
automatically add a default usb device; one can always be added
explicitly if desired. Or we may decide that on q35 machines, 3 usb
controllers will be automatically added when none is given. But for
this initial commit, at least we aren't locking ourselves into
something we later won't want.
video - qemu always initializes the primary video device immediately
after any integrated devices for the machinetype. Unless instructed
otherwise (by using "-device vga..." instead of "-vga" which libvirt
uses in many cases to work around deficiencies and bugs in various
qemu versions) qemu will always pick the first unused slot. In the
case of the "pc" machinetype and its derivatives, this is always slot
2, but on q35 machinetypes, the first free slot is slot 1 (since the
q35's integrated peripheral devices are placed in other slots,
e.g. slot 0x1f). In order to make the PCI address of the video device
predictable, that slot (1 or 2, depending on machinetype) is reserved
even when no video device has been specified.
sata - a q35 machine always has a sata controller implicitly added at
slot 0x1F, function 2. There is no way to avoid this controller, so we
always add it. Note that the xml2xml tests for the pcie-root and q35
cases were changed to use DO_TEST_DIFFERENT() so that we can check for
the sata controller being automatically added. This is especially
important because we can't check for it in the xml2argv output (it has
no effect on that output since it's an implicit device).
ide - q35 has no ide controllers.
isa and smbus controllers - these two are always present in a q35 (at
slot 0x1F functions 0 and 3) but we have no way of modelling them in
our config. We do need to reserve those functions so that the user
doesn't attempt to put anything else there though. (note that the "pc"
machine type also has an ISA controller, which we also ignore).
This PCI controller, named "dmi-to-pci-bridge" in the libvirt config,
and implemented with qemu's "i82801b11-bridge" device, connects to a
PCI Express slot (e.g. one of the slots provided by the pcie-root
controller, aka "pcie.0" on the qemu commandline), and provides 31
*non-hot-pluggable* PCI (*not* PCIe) slots, numbered 1-31.
Any time a machine is defined which has a pcie-root controller
(i.e. any q35-based machinetype), libvirt will automatically add a
dmi-to-pci-bridge controller if one doesn't exist, and also add a
pci-bridge controller. The reasoning here is that any useful domain
will have either an immediate (startup time) or eventual (subsequent
hot-plug) need for a standard PCI slot; since the pcie-root controller
only provides PCIe slots, we need to connect a dmi-to-pci-bridge
controller to it in order to get a non-hot-plug PCI slot that we can
then use to connect a pci-bridge - the slots provided by the
pci-bridge will be both standard PCI and hot-pluggable.
Since pci-bridge devices themselves can not be hot-plugged into a
running system (although you can hot-plug other devices into a
pci-bridge's slots), any new pci-bridge controller that is added can
(and will) be plugged into the dmi-to-pci-bridge as long as it has
empty slots available.
This patch is also changing the qemuxml2xml-pcie test from a "DO_TEST"
to a "DO_DIFFERENT_TEST". This is so that the "before" xml can omit
the automatically added dmi-to-pci-bridge and pci-bridge devices, and
the "after" xml can include it - this way we are testing if libvirt is
properly adding these devices.
This controller is implicit on q35 machinetypes. It provides 31 PCIe
(*not* PCI) slots as controller 0.
Currently there are no devices that can connect to pcie-root, and no
implicit pci controller on a q35 machine, so q35 is still
unusable. For a usable q35 system, we need to add a
"dmi-to-pci-bridge" pci controller, which can connect to pcie-root,
and provides standard pci slots that can be used to connect other
devices.
Previous refactoring of the guest PCI address reservation/allocation
code allowed for slot types other than basic PCI (e.g. PCI express,
non-hotpluggable slots, etc) but would not auto-allocate a slot for a
device that required any type other than a basic hot-pluggable
PCI slot.
This patch refactors the code to be aware of different slot types
during auto-allocation of addresses as well - as long as there is an
empty slot of the required type, it will be found and used.
The piece that *wasn't* added is that we don't auto-create a new PCI
bus when needed for anything except basic PCI devices. This is because
there are multiple different types of controllers that can provide,
for example, a PCI express slot (in addition to the pcie-root
controller, these can also be found on a "root-port" or on a
"downstream-switch-port"). Since we currently don't support any PCIe
devices (except pending support for dmi-to-pci-bridge), we can defer
any decision on what to do about this.
