The 'pattern' property has type string and format regex, so
it makes sense to allow Regexp objects to be used for it.
While check_type() doesn't know the format, Regexp objects
can be treated like strings anyway, including compared via
'eq' or matched via '=~', so we allow strings to generally
come from a Regexp object.
Commit de9a267 introduced vec() to optimize the generation
by using binary operations instead of converting back and
forth between hex and strings, but forgot to switch over to
the binary sha1 method. This resulted in only the first 6
hex digits of the output string making up the address.
Fix #132: hold a lock while setting up vlan bridges
It's possible for two simultaneous VM starts to try to
create vlan bridges on non-vlan-aware bridges
simultaneously, which causes one of them to fail with the
error "can't add bridge ...".
This essentially performas the task of systemd-run while
also waiting for the job to finish.
With the systemd-run version in jessie we run into a race
condition where the executed process can start forking child
processes before the systemd daemon is done setting up the
scope's cgroups, causing the children to NOT be included in
the cgroups. This means the child processes (in our case
qemu) will not adhere to the limits we want to apply to it
via cgroups.
enter_systemd_scope() performs the setup task of systemd-run
and waits for the job to finish, after this we can spawn the
qemu process without systemd-run.
Since we already allow this for container IP addresses it is
reasonable to assume the host might be using such a setup as
well. (You can use an additional route to reach the gateway
and then simply have no "LAN".) Some people seem to want
this...
Thomas Lamprecht [Wed, 25 May 2016 09:03:10 +0000 (11:03 +0200)]
run_cli_handler: add no_rpcenv parameter
This can be used if a CLIHandler child class does not needs or wants
RPCEnv. E.g. this is the case with pve-cluster, as it's a dependency
of the package where RPCEnv resides (access-control), so this can be
set to avoid a implicit circular build dependency.
Signed-off-by: Thomas Lamprecht <t.lamprecht@proxmox.com>
Emmanuel Kasper [Tue, 3 May 2016 10:43:12 +0000 (12:43 +0200)]
When starting in debug mode, starts the daemon without going through systemctl
This fixes the problem that debug options for daemons were ignored and sets
the following behaviour for the daemons when -debug 1 is given
* start on foreground with debug ouput on the console
* do not start a sytemd unit
Starting the systemd unit in debug mode makes less sense because:
* systemd will put the daemon in the background
* systemd ignores environment variables so the only way to start
the daemon with options would be through reading a config file
this would previously convert properties that matched
'^([a-z]+)(\d+)$' even if they were not part of an indexed
properties series (which always start with 0).
This fixes previously wrongly converted properties:
-smbios1
-server2
this would previously skip (i.e., ignore) properties that
matched '^([a-z]+)(\d+)$' even if they were not part of an
indexed properties series (which always start with 0).
This fixes previously missing API dump occurrences of:
-smbios1
-comments6
-netmask6
-gateway6
-address6
-dns1
-dns2
-dns3
-server1
-server2
tap_unplug() is only called in hotplugging functions which
perform a tap_plug() afterwards, and and in qemu-server's
bridge-down script.
The OVS port cleanup for non-firewall ports was done only in
tap_plug(), which means that in the case when qemu exits and
the tap device disappears, the OVS port for it stays.
With this patch the cleanup happens in both tap_unplug() and
tap_plug() to avoid problems with upgrades on systems
where there have already been ports leaked.
Replace 'string' with 'filepath' for mapped params
For the CLI man pages and documentation, mapped parameters
('content' strings in API2 which are loaded from file paths
in CLI) get special treatment and are displayed with the
type 'filepath'.
Handle string parameter to file content replacement
This is modelled after the way read_password() is used to
wrap -password parameters on the command line. If a mapping
for a certain API method and parameter is defined in the
sub class of CLIHandler.pm, the parameter is interpreted as
a file path on the command line and the parameter is
filled with the string contents of the referenced file.
This allows us to use the same API schema once in API2, but
overwrite the behaviour for individual parameters in the CLI
tools when desired.
Use case: networks for kvm use a <model>=<macaddr> scheme
where the model represents the network card. The schema
previously could not represent this, so we now introduce a
'group' key which works similar to an alias with the
difference that the data structure also gets an entry named
after the group filled with the name of the key that was
used to fill it.
Usage:
{
virtio => { group => 'model' },
e1000 => { group => 'model' },
model => {
type => 'string',
pattern => ... # pattern for mac address
...
}
}
Now the string 'virtio=aa:bb:cc:dd:ee:ff' gets parsed into:
{
model => 'virtio',
virtio => 'aa:bb:cc:dd:ee:ff'
}
Error examples:
With bad value:
virtio: value does not match the regex pattern
Missing group:
model: property is missing and it is not optional
parse_net() however used the 'macaddr' key for the mac
address, which can be achieved by aliasing 'model' to
'macaddr':
{
virtio => { group => 'model' },
e1000 => { group => 'model' },
model => { alias => 'macaddr' },
macaddr => {
type => 'string',
pattern => ... # pattern for mac address
...
}
}
Then the above string will be parsed into:
{
model => 'virtio',
macaddr => 'aa:bb:cc:dd:ee:ff'
}
The error output now always shows the 'macaddr' key:
Error examples:
With bad value:
macaddr: value does not match the regex pattern
Missing group:
macaddr: property is missing and it is not optional
In order to support specifying no mac address we can now set
model.default_key = 1 and macaddr.optional = 1.
That way `virtio,bridge=vmbr2` gets parsed correctly into
just a model with no macaddr. This works because default
keys as aliases have previously not been supported and would
not have been aliased accordingly. This case is now also
taken into account when printing default keys, which is now
skipped if it is also an alias.
Any action we take in tap_plug() when using OVS will undo
our rate limiting, which means any hotplug change must
restore the previous rate setting.
Since this means tap_plug() would always be followed by a
tap_rate_limit() call anyway we just include the rate
parameter here and let the caller decide whether the full
tap_plug() is required or the simple tap_rate_limit()
suffices.
projects / pve-common.git / log