| Commit message (Collapse) | Author | Age | Files | Lines |
|
|
|
|
| |
... Without appending the 'ro' to mount options of '/' mount point the
system will be still mounted as read-write, even when READONLY=yes.
|
|
|
|
|
|
|
|
| |
$" is a BASH extension to translate text.
Replace the shebang with /bin/bash to avoid parsing errors,
as even BASH has a POSIX compliant mode when run as sh.
Signed-off-by: Matteo Croce <mcroce@redhat.com>
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
network-functions uses awk to tranform the mac address to upper case.
Replace awk with tr which is two order of magnitude faster:
$ time awk '{ print toupper($0) }' </usr/share/dict/words >/dev/null
real 0m0,227s
user 0m0,224s
sys 0m0,004s
$ time tr '[a-z]' '[A-Z]' </usr/share/dict/words >/dev/null
real 0m0,005s
user 0m0,000s
sys 0m0,005s
Also use here-strings instead of spawning a subshell and a pipe.
Signed-off-by: Matteo Croce <mcroce@redhat.com>
|
|
|
|
|
|
|
|
|
|
| |
Instead of displaying messages without context, like this:
./network-functions: line 571: echo: write error: Invalid argument
./network-functions: line 598: echo: write error: Permission denied
We will now display pretty error messages (via net_log) about what has
actually failed...
|
|
|
|
|
|
|
|
|
|
| |
Previously, there were 2 additional conditions related to this part of
code, but they were removed in time, so this last condition has
remained an artifact and is de-facto obsolete.
However, the 'pidof -x dhclient' condition could cause the network
service to hang on stop/restart when processes were executed from a
network share... Removing that condition should hypothetically fix it.
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
* ifup-tunnel: Support 'external' tunnels
This is a newish feature upstream. You can now set the external flag on a
ip6_tunnel type interface (though not the primary one, ip6tnl0), and doing so
will allow it to decapsulate any packet, and assuming that the inner address is
the one on that interface, it'll drop it back on the stack.
This is useful for DSR vips. While v6-in-v6 was already supported, this allows
v4-in-v6 which is necessary to serve v4 traffic in a v6only infrastructure.
There's comments in ifup-tunnel that imply it was designed only for GRE tunnels,
but this still seems like the best place for this.
|
|
|
|
|
|
|
| |
Empty ${RESOLV_MODS} was causing a deletion of config files in
/etc/ppp/peers/*
Resolves: #155
|
| |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
This is complete rework of how we generate the /etc/resolv.conf.
Fixes:
* always update 'nameserver' & 'search' entries when DNS* & DOMAIN
options (respectively) were updated in ifcfg-* files
* always update the order of 'nameserver' entries when the order of
DNS* options was updated
* always keep /etc/resolv.conf.save as a proper backup - do not
overwrite it everytime when updating 'nameserver' or 'search'
Enhancements:
* added support for DNS3 option (equals to MAXNS value in GLIBC)
* added logic to process duplicate DNS* options
* added logic to process randomly omitted DNS* options (e.g. omitting
DNS1 while specifying DNS2 and/or DNS3 will still work now)
This work was based on these two RHEL-7.5 BZs:
https://bugzilla.redhat.com/show_bug.cgi?id=1364895
https://bugzilla.redhat.com/show_bug.cgi?id=1357658
|
| |
|
| |
|
| |
|
|
|
|
|
| |
To speed things up. Based on patch & comments from:
https://bugzilla.redhat.com/show_bug.cgi?id=1497759
|
|
|
|
|
|
|
|
|
|
|
| |
Using this package was very dangerous, and could lead to instability
of system, or even being unable to bootup at all, and value of this
functionality was at least questionable, and was not documented.
We decided to drop functionality completely, to make sure users do not
install it accidentally (e.g. by using globbing with 'dnf install').
Resolves: #69
|
|
|
|
|
| |
Users can still create the /etc/sysconfig/init file manually to
override the default values, which are now taken from init.d/functions.
|
|
|
|
|
| |
Otherwise obtaining IP from DHCP might fail. Based on patch from:
* Rich Alloway <richard.alloway@roguewave.com> | CentOS BZ #0011138
|
|
|
|
|
|
|
|
|
|
| |
And network-scripts/network-functions was patched to use convert2sec().
This function can be used to convert the value of its first parameter
to a number of seconds - based on the time unit (specified as the
second parameter).
This is mostly useful for converting values for use with the sleep(1).
|
|
|
|
| |
https://bugzilla.redhat.com/show_bug.cgi?id=1471489
|
|
|
|
|
|
|
|
|
| |
The ARPUPDATE option has been introduced. It defaults to 'yes'.
By setting the ARPUPDATE to 'no', administrator can disable
updating neighbouring computers with ARP information about current
NIC. This is especially needed when using LVS Load Balancing with
Direct routing enabled.
|
|
|
|
|
| |
The time units had to be manually converted from microseconds to
seconds. Hopefully this was done right.
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
The way DHCP_HOSTNAME is being sent to DHCP server has been completely
reworked...
DHCP_FQDN now allows user to specify fully qualified domain name to be
sent to DHCP server instead of just using simple hostname. This option
takes precedence before DHCP_HOSTNAME variable.
DHCP_SEND_HOSTNAME can be used to tell initscripts to not send either
DHCP_HOSTNAME or DHCP_FQDN to DHCP server.
Both of these options should behave same as they do in NetworkManager.
|
| |
|
| |
|
|
|
|
| |
... but it is stuck some time
|
|
|
|
|
| |
With quote marks all the parameters to dhclient are passed as one
Resolves: #1410429
|
|
|
|
|
|
|
|
| |
Instead of adding the IPv6 address when ifup is called, we replace the
address, because it might have been auto-assigned before by RA.
This is to assure that static NIC configuration will take effect if
the NIC was already dynamically configured by RA.
|
|
|
|
|
| |
Use dbus to refresh configuration files for NetworkManager in order
to make ifup and ifdown faster.
|
|
|
|
| |
Use dbus instead of nmcli to make ifup command faster.
|
|
|
|
|
|
|
| |
Previously, some administrators were unable to force initscripts to
not obtatin hostname from DHCP, even though they were using static
configuration of network. Righ now, setting 'NO_DHCP_HOSTNAME' to
'yes', 'true' or '1' in /etc/sysconfig/network will allow them to do so.
|
|
|
|
|
|
|
| |
Typo introduced in commit 432951d.
Related: https://github.com/fedora-sysv/initscripts/pull/2
Signed-off-by: Pavel Šimerda <pavlix@pavlix.net>
|
| |
|
| |
|
|
|
|
| |
Resolves: #1099026
|
|
|
|
| |
This has patch has been suggested in RHBZ #950453, by Konrad Moson.
|
| |
|
| |
|
|
|
|
|
|
|
|
|
|
|
|
| |
`ifup` can set _some_ route options. Today it's missing several and so, to set
something like `mtu` or `advmss` you have to do hacks like:
```
METRIC='10 mtu 1480'
```
which is sub-optimal. Rather than one-off add support for everything in the man
page, provide a hook for arbitrary extra route options.
|
|
|
|
|
|
| |
Thanks to shellcheck. :)
This is ontop of the whitespace PRs, those should go in first.
|
| |
|
| |
|
| |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
The network-scripts are a mix of tabs and spaces (!!) as well as
completely inconsistent indentation (some 2, some 4).
This is a series of 3 patches to move the entire of network-scripts to
2-space indentation for consistency so I can develop on the codebase
without my brain bleeding. I don't feel strongly about 2 vs 4... 2 was
just most common in the first file I picked, so I used that.
This set is in 3 sets for easier reviewing and easier reverting if
necessary:
* 1/3 - ipv6 files
* 2/3 - network-functions - it needed it's own
* 3/3 - ifup/ifdown files
|
|
|
|
|
|
|
| |
It's reasonable (and common) to assign static addresses while still
wanting to get your default gw from your router. The assumption that not
wanting SLAAC means not wanting RA is flawed, but I don't want to break
backwards compatability, so I'm adding an additional option.
|
|
|
|
|
|
|
|
|
|
|
| |
network-scripts makes odd assumptions, like if I don't want SLAAC (i.e.
if I turn off AUTOCONF) that I also don't want RA. This is not true,
it's common to have a static local address, but to get my gateway
through RA.
This is one solution. I will also be sending a more specific solution to
enable the above directly, but having a way to tell network-scripts
to not mess with systctls seems generally useful.
|
| |
|
|
|
|
| |
Resolves: #1336869
|
| |
|
| |
|
| |
|
| |
|