| Summary: | net-misc/minissdpd-1.3 - /etc/init.dminissdpd expects old ifconfig output style | ||
|---|---|---|---|
| Product: | Gentoo Linux | Reporter: | Christohper Harrington <ironiridis> |
| Component: | Current packages | Assignee: | Anthony Basile <blueness> |
| Status: | RESOLVED FIXED | ||
| Severity: | normal | Keywords: | PATCH |
| Priority: | Normal | ||
| Version: | unspecified | ||
| Hardware: | All | ||
| OS: | Linux | ||
| Whiteboard: | |||
| Package list: | Runtime testing required: | --- | |
| Attachments: | Update minissdpd init script to detect interfaces correctly | ||
|
Description
Christohper Harrington
2014-12-29 05:22:38 UTC
Created attachment 392614 [details, diff]
Update minissdpd init script to detect interfaces correctly
1. The minissdpd -i argument accepts either an interface OR an IP address, so it isn't neccessary for us to divine the IP before adding it to ARGS; we already have the interface. Pass the interface instead.
2. The invocation of ifconfig | grep | awk is counting on deprecated output formatting from ifconfig, and further that output isn't designed to be machine-parsable. The script is already breaking on output from ifconfig if not merged with the USE flag old-output. Instead, read /sys/class/net/[interface]/flags to determine the state of the interface.
3. The script again counts on the output format of route, which isn't designed to be machine-parsable. It also invokes route without the -n parameter, which may involve a reverse-DNS hit. Instead, parse /proc/net/route.
4. By default, the script will only configure one interface unless configured otherwise. The user may have two or more interfaces attached to their computer, especially possible in gateway scenarios. Instead, scan the local interface routes for local subnets, and match them against private subnet prefixes.
I thought I'd fixed this in bug #522562. See http://sources.gentoo.org/cgi-bin/viewvc.cgi/gentoo-x86/net-misc/minissdpd/files/minissdpd.initd-r1?r1=1.3&r2=1.4 Anyhow, your solution avoids ifconfig and route so I like it. Why not just remove checkconfig() rather than nop it? (In reply to Anthony Basile from comment #2) > Anyhow, your solution avoids ifconfig and route so I like it. Thanks! > Why not just remove checkconfig() rather than nop it? It seems convention to have a checkconfig() function, and there should likely be a way to enable ipv6 to the configuration file (passing -6 to minissdpd) down the road, which might need some sanity check from the init script, like whether ipv6 functionality is enabled in the kernel. checkconfig() may also be a good place to actually make sure /proc and /sys are really mounted. That said, of course, I'm not married to it. (In reply to Christohper Harrington from comment #3) > (In reply to Anthony Basile from comment #2) > > Anyhow, your solution avoids ifconfig and route so I like it. > Thanks! > > > Why not just remove checkconfig() rather than nop it? > It seems convention to have a checkconfig() function, and there should > likely be a way to enable ipv6 to the configuration file (passing -6 to > minissdpd) down the road, which might need some sanity check from the init > script, like whether ipv6 functionality is enabled in the kernel. > > checkconfig() may also be a good place to actually make sure /proc and /sys > are really mounted. > > That said, of course, I'm not married to it. I dropped checkconfig() in the commit. I see your point about a stub but its not necessary. Please test minissdpd-1.3-r1.ebuild. If it works for you I'll drop the older versions. Thanks for taking the time to think about this. You might want to pass a note upstream since it avoids ifconfig/route. Upstream uses that approach and my init script just mimicked it. The irony is, how do you think ifconfig/route work anyhow :) Tested. Thank you! |
