Kernel prepatch 4.1-rc1

Posted Apr 29, 2015 9:53 UTC (Wed) by karath (subscriber, #19025)
Parent article: Kernel prepatch 4.1-rc1

I'm generally in favour of the ideas behind systemd and, more recently, kdbus. I'm sure that some of the opposition to merging kdbus are motivated by resistance to the people (person?) behind systemd. However, watching some of the behaviour of some supporters of the current drive to merge, I see them doing themselves long-term harm.

The more thoughtful opponents to merging kdbus in its current form are raising issues that are not so easily and quickly dismissed. For example, on the issue of the meta-data: Andy Lutomirsky suggests that including capabilities in the metadata is a seriously bad idea while having previously given quite a bit of useful code review. On the other hand, Linus says that having the capabilities is not just a good idea, it is a necessity, but the process command line is a no-no. Additionally, Linus wanted attention paid to the translation of capabilities across user namespace boundaries. Clearly this an area where smart people can reasonably disagree and therefore could take some time to resolve.

Al Viro required documentation about the locking before he felt ready to review the code. Greg Kroah-Hartman acknowledged that this documentation was a valid requirement and, while Dave Hermann provided it, the timing was very late in the cycle. Likely Al had his own set of patches to manage, never mind that quality review may take quite a bit of time.

And now Linus is asking difficult questions about dbus and kdbus performance.

kdbus is not likely to go into the kernel in this release cycle, though Linus could override the convention as he occasionally has done in the past. However, some supporters seem willing to burn the trust and respect that the community has for them in their rush to get the immediate result.