When an nwfilter rule sets the parameter CTRL_IP_LEARNING to "dhcp",
this turns on the "dhcpsnoop" thread, which uses libpcap to monitor
traffic on the domain's tap device and extract the IP address from the
DHCP response.
If libpcap on the host is built with HAVE_TPACKET3 defined (to enable
support for TPACKET_V3), the dhcpsnoop code's initialization of the
libpcap socket would fail with the following error:
virNWFilterSnoopDHCPOpen:1134 : internal error: pcap_setfilter: can't remove kernel filter: Bad file descriptor
It turns out that this was because TPACKET_V3 requires a larger buffer
size than libvirt was setting (we were setting it to 128k). Changing
the buffer size to 256k eliminates the error, and the dhcpsnoop thread
once again works properly.
A fuller explanation of why TPACKET_V3 requires such a large buffer,
for future git spelunkers:
libpcap calls setsockopt(... SOL_PACKET, PACKET_RX_RING...) to setup a
ring buffer for receiving packets; two of the attributes sent to this
API are called tp_frame_size, and tp_frame_nr. If libpcap was built
with HAVE_TPACKET3 defined, tp_trame_size is set to MAXIMUM_SNAPLEN
(defined in libpcap sources as 262144) and tp_frame_nr is set to:
[the buffer size we set, i.e. PCAP_BUFFERSIZE i.e. 262144] / tp_frame_size.
So if PCAP_BUFFERSIZE < MAXIMUM_SNAPLEN, then tp_frame_nr (the number
of frames in the ring buffer) is 0, which is nonsensical. This same
value is later used as a multiplier to determine the size for a call
to malloc() (which would also fail).
(NB: if HAVE_TPACKET3 is *not* defined, then tp_frame_size is set to
the snaplen set by the user (in our case 576) plus a small amount to
account for ethernet headers, so 256k is far more than adequate)
Since the TPACKET_V3 code in libpcap actually reads multiple packets
into each frame, it's not a problem to have only a single frame
(especially when we are monitoring such infrequent traffic), so it's
okay to set this relatively small buffer size (in comparison to the
default, which is 2MB), which is important since every guest using
dhcp snooping in a nwfilter rule will hold 2 of these buffers for the
entire life of the guest.
Thanks to Christian Ehrhardt for discovering that buffer size was the
problem (this was not at all obvious from the error that was logged!)
Resolves: https://bugzilla.redhat.com/1547237
Fixes: https://bugs.launchpad.net/libvirt/+bug/1758037
Signed-off-by: Laine Stump <laine@laine.org>
Reviewed-by: Christian Ehrhardt <christian.ehrhardt@canonical.com> (V1)
Reviewed-by: John Ferlan <jferlan@redhat.com>
Tested-by: Christian Ehrhardt <christian.ehrhardt@canonical.com>
(cherry picked from commit ce5aebeacd)
Sometimes we don't regenerate QEMU capabilities replies using QEMU
binary but we simply add a new entry manually. In that case you need
to manually fix all the replies ids. This helper will do that for you.
Reviewed-by: Andrea Bolognani <abologna@redhat.com>
Signed-off-by: Pavel Hrdina <phrdina@redhat.com>
The check was trying to use the shell variable $CC instead of
the make variable $(CC); it also interpreted grep's return code
wrong: 1 means the provided pattern was *not* matched. As a
result, pdwtags was never run, not even when building with gcc.
Signed-off-by: Andrea Bolognani <abologna@redhat.com>
Reviewed-by: Daniel P. Berrangé <berrange@redhat.com>
Commit id 'edae027c' blindly assumed that the passed @oldDev
parameter would not be NULL when calling virDomainDeviceGetInfo;
however, commit id 'b6a264e8' passed NULL for AttachDevice
callers under the premise that there wouldn't be a device
to check/update against.
Signed-off-by: John Ferlan <jferlan@redhat.com>
Enable testing of both the upstart and systemd init script handling.
We test a different one in each scenario. Even though trusty only
cares about upstart, it is fine for us to test rules that install
systemd, since we're not actually running these scripts for real.
Reviewed-by: Andrea Bolognani <abologna@redhat.com>
Signed-off-by: Daniel P. Berrangé <berrange@redhat.com>
We can't use "make distcheck" on macOS because many unit tests fail. We
can still get coverage of some of the things "distcheck" validates, by
running the "install" and "dist" targets. This is particularly useful
because many conditional features are disabled on macOS, and this helps
make sure we can still successfully install & dist when these bits are
disabled.
The default script is getting unreadable since it is all on one long
line. Rather than adding further conditional clauses to it, we make
use of the travis matrix config override for the script.
Reviewed-by: Andrea Bolognani <abologna@redhat.com>
Signed-off-by: Daniel P. Berrangé <berrange@redhat.com>
Running "make distcheck" includes the "make check", and "make dist"
targets. It ensures that we have CLEANFILES and uninstall rules setup
correctly, as well as validating VPATH builds succeed.
Reviewed-by: Andrea Bolognani <abologna@redhat.com>
Signed-off-by: Daniel P. Berrangé <berrange@redhat.com>
The precise distro is marked deprecated in travis and will be dropped
entirely in 2 months time.
Reviewed-by: Andrea Bolognani <abologna@redhat.com>
Signed-off-by: Daniel P. Berrangé <berrange@redhat.com>
When building with CLang the structs that are emitted by pdwtags appear
in a completely different order than with GCC, which causes the
comparison against expected data to fail.
Ideally the test would not be sensitive to the ordering, because even
future GCC could cause changes, but that's not easy to fix. So for now
just skip the test when using clang.
Reviewed-by: Andrea Bolognani <abologna@redhat.com>
Signed-off-by: Daniel P. Berrangé <berrange@redhat.com>
We have switched the docs to using the HTML5 doctype declaration in
commit b1c81567c7
Author: Daniel P. Berrange <berrange@redhat.com>
Date: Wed Jul 26 18:01:25 2017 +0100
docs: switch to using HTML5 doctype declaration
Signed-off-by: Daniel P. Berrangé <berrange@redhat.com>
Currently if cmd->skipChecks is set (done only from completers)
some basic checks are skipped because we're working over
partially parsed command. See a26ff63ae4 for more detailed
explanation. Anyway, the referenced commit was too aggressive in
disabling checks and effectively returned success even in clear
case of failure. For instance:
# domif-getlink --interface <TAB><TAB>
causes virshDomainInterfaceCompleter() to be called, which calls
virshDomainGetXML() which eventually calls
vshCommandOptStringReq(.., name = "domain"); The --domain
argument is required for the command and if not present -1 should
be returned to tell the caller the argument was not found. Well,
zero is returned meaning the argument was not found but it's not
required either.
Signed-off-by: Michal Privoznik <mprivozn@redhat.com>
The commands which requires a pool to perform any action for a volume is
throwing a segfault when you pass the volume name before a pool name or
without the argument '--pool'.
An example that works:
virsh # vol-list loops-pool
Name Path
-------------------------------------------------------------------
loop0 /mnt/loop0
virsh # vol-info --pool loops-pool lo<TAB>
An example that does not work:
virsh # vol-list loops-pool
Name Path
-------------------------------------------------------------------
loop0 /mnt/loop0
virsh # vol-info lo<TAB>
Segmentation Fault
The example 'vol-info' can be executed as 'vol-info loop0 --pool
loops-pool'. So, this commit fixes this problem when the arguments are
inverted and avoids the segfault.
Signed-off-by: Julio Faracco <jcfaracco@gmail.com>
Signed-off-by: Michal Privoznik <mprivozn@redhat.com>
12 bytes in 1 blocks are definitely lost in loss record 188 of 1,145
at 0x4C2B6CD: malloc (in /usr/lib/valgrind/vgpreload_memcheck-amd64-linux.so)
by 0x5D2CD77: xmlStrndup (in /lib/x86_64-linux-gnu/libxml2.so.2.7.8)
by 0x514E137: virXMLPropString (virxml.c:506)
by 0x234F51: qemuMigrationCookieNetworkXMLParse qemu_migration.c:1001)
by 0x235FF8: qemuMigrationCookieXMLParse (qemu_migration.c:1333)
by 0x236214: qemuMigrationCookieXMLParseStr (qemu_migration.c:1372)
by 0x2365D2: qemuMigrationEatCookie (qemu_migration.c:1456)
by 0x243DBA: qemuMigrationFinish (qemu_migration.c:6381)
by 0x204032: qemuDomainMigrateFinish3 (qemu_driver.c:13228)
by 0x521CCBB: virDomainMigrateFinish3 (libvirt-domain.c:4788)
by 0x1936DE: remoteDispatchDomainMigrateFinish3 (remote.c:4580)
by 0x16DBB1: remoteDispatchDomainMigrateFinish3Helper(remote_dispatch.h:7582)
Signed-off-by: ZhangZijian <zhang.zijian@h3c.com>
A problem encountered due to a bug in libpcap was reported to the
caller as:
An error occurred, but the cause is unknown
This was because the error had been logged in the DHCPSnoop
thread. The worker thread handling the API call to start a domain
spins up the DHCPSnoop thread which watches for dhcp packets with
libpcap, then uses virCondSignal() to notify the worker thread (which
has been waiting with virCondWait()). The worker thread knows that
there was an error (because threadStatus != THREAD_STATUS_OK), but the
error info had been stored in thread-specific storage for the other
thread, so the worker thread can only report that there was a failure,
but it doesn't know why.
The solution is to save the error that was logged (with
virErrorPreserveLast() into the object the is used to share info
between the threads, then we can set the error in the worker thread
using virErrorRestore().
In the case of the error I was looking at, this changed the "unknown"
message into:
internal error: pcap_setfilter: can't remove kernel filter:
Bad file descriptor
Signed-off-by: Laine Stump <laine@laine.org>
The libvirt_storage_backend_sheepdog_priv.la library depends on symbols
provided in the libvirt_driver_storage_impl.la library. As such the
latter must be listed 2nd when passed to the linker to avoid symbol
resolution problems. This mistake is being masked by the sheepdog
driver linking in a second copy of the storage driver code. Remove
this duplicate linkage of backend source and fix the test link order.
Signed-off-by: Daniel P. Berrangé <berrange@redhat.com>
A typo in the uninstall-data-extra rule expansion meant we just called
the install rule again, instead of the uninstall rule. While fixing
this, just inline the dependancy, since the intermediate
install-data-extra rule adds no value.
Signed-off-by: Daniel P. Berrangé <berrange@redhat.com>
Ensure all enum cases are listed in switch statements, or cast away
enum type in places where we don't wish to cover all cases.
Build is broken after 67966ad51 [1].
[1] m4: enforce that all enum cases are listed in switch statements
Reviewed-by: Daniel P. Berrangé <berrange@redhat.com>
Signed-off-by: Nikolay Shirokovskiy <nshirokovskiy@virtuozzo.com>
Build is broken by 5529b057 [1].
[1] cfg: forbid includes of headers in network and storage drivers again
Reviewed-by: Daniel P. Berrangé <berrange@redhat.com>
Signed-off-by: Nikolay Shirokovskiy <nshirokovskiy@virtuozzo.com>
This partially reverts 82592551cb.
When migrating a domain, qemuMigrationDstPrepareAny() is called
which eventually calls qemuProcessLaunch(conn = NULL, flags =
VIR_QEMU_PROCESS_START_AUTODESTROY); But the very first thing
that qemuProcessLaunch does is check if AUTODESTROY flag is set
and @conn is not NULL. Well, it is.
Signed-off-by: Michal Privoznik <mprivozn@redhat.com>
Reviewed-by: Daniel P. Berrangé <berrange@redhat.com>
https://bugzilla.redhat.com/show_bug.cgi?id=1494454
If a domain disk is stored on local filesystem (e.g. ext4) but is
not being migrated it is very likely that domain is not able to
run on destination. Regardless of share/cache mode.
Signed-off-by: Michal Privoznik <mprivozn@redhat.com>
Reviewed-by: Daniel P. Berrangé <berrange@redhat.com>
libxl requires the memory sizes to be rounded to 1MiB increments.
Attempting to start a domain that violates this requirement will
fail with the marginally helpful error
2018-02-22 01:55:32.921+0000: xc: panic: xc_dom_boot.c:141: xc_dom_boot_mem_init: can't allocate low memory for domain: Out of memory
2018-02-22 01:55:32.921+0000: libxl: libxl_dom.c:671:libxl__build_dom: xc_dom_boot_mem_init failed: No such file or directory
Round the maximum and current memory values to the next 1MiB
increment when generating the libxl_domain_config object.
Signed-off-by: Jim Fehlig <jfehlig@suse.com>
While libvirt-guests.sh is running cases can let guest_is_on fail which
causes check_guests_shutdown to print output.
That output shall not spill into the users of function
check_guests_shutdown which is therefore now returning values in a
variable like guest_is_on already did.
Original-Author: Christian Ehrhardt <christian.ehrhardt@canonical.com>
Modified-By: Jorge Niedbalski <niedbalski@ubuntu.com>
Signed-off-by: Michal Privoznik <mprivozn@redhat.com>
Commit v3.7.0-14-gc57f3fd2f8 prevented adding a <boot order='x'/>
element to an inactive domain with global <boot dev='...'/> element.
However, as a result of that change updating any device with boot order
would fail with 'boot order X is already used by another device', where
"another device" is in fact the device which is being updated.
To fix this we have to ignore the device which we're about to update
when checking for boot order conflicts.
https://bugzilla.redhat.com/show_bug.cgi?id=1546971
Signed-off-by: Jiri Denemark <jdenemar@redhat.com>
When calling virDomainDefCompatibleDevice to check a new device during
device update, we need to pass the original device which is going to be
updated in addition to the new device. Otherwise, the function can
report false conflicts.
The new argument is currently ignored by virDomainDefCompatibleDevice,
but this will change in the following patch.
https://bugzilla.redhat.com/show_bug.cgi?id=1546971
Signed-off-by: Jiri Denemark <jdenemar@redhat.com>
Checking the new device definition makes little sense when lxc driver
does not support live device update at all.
Signed-off-by: Jiri Denemark <jdenemar@redhat.com>
