Summary: | We can reply to the FAQ for RSBAC/SELinux in Gentoo Hardened SELinux FAQ | ||
---|---|---|---|
Product: | Documentation | Reporter: | kang <kang> |
Component: | Project-specific documentation | Assignee: | The Gentoo Linux Hardened Team <hardened> |
Status: | RESOLVED FIXED | ||
Severity: | normal | CC: | selinux |
Priority: | Normal | ||
Version: | unspecified | ||
Hardware: | All | ||
OS: | Linux | ||
URL: | http://www.gentoo.org/proj/en/hardened/selinux-faq.xml#rsbac | ||
Whiteboard: | |||
Package list: | Runtime testing required: | --- |
Description
kang
2011-07-12 18:28:21 UTC
thanks for the information! I'll update the FAQ accordingly Is the wording in http://goo.gl/uaaf4#rsbac good enough (I don't mention that RSBAC cannot be entirely disabled, I don't think it is that important since, if you don't want it, you don't ask the question of SELinux + RSBAC ;-) ? (In reply to comment #2) > Is the wording in http://goo.gl/uaaf4#rsbac good enough (I don't mention that > RSBAC cannot be entirely disabled, I don't think it is that important since, if > you don't want it, you don't ask the question of SELinux + RSBAC ;-) ? Its fine, however, I would word it in a way that makes it clear its not RSBAC that is slower, both LSM with SELinux and RSBAC impact performance. What you wrote is correct, I just fear people could understand it the wrong way, if we dont make it super clear. I hope thats not too nitpicking :) Some background info because I felt like it: They're both quite efficient but still the impact generally goes from 1 to 7% (in my benchs). When using both we're talking 2 to 14%, which starts to be an issue IMO. The thing is that LSM and RSBAC put many hooks (often at the same place) so we're checking twice every time. In fact RSBAC is generally slightly faster as the policy is loaded in ram at boot (single per module file attribute storage) and not on demand. RSBAC then caches file related decisions, while SELinux caches it all in the AVC instead, due to the fact that attributes are stored in xattrs. For LSM-supported modules, the hook happens only once no matter the number of modules (it does check for each module of course) RSBAC includes its own "LSM" so likewise, the hook happens only once, no matter the number of RSBAC modules. Using both will hook twice and the performance impact is larger than just the module check Ok, i hope I made sense :) Sure, this <http://goo.gl/uaaf4#rsbac> better? looks good to me, thanks! |