RSBAC Handbook
Releases
Patched kernels
Includes vanilla kernel with the RSBAC patch
Enhanced kernels
Combined patches with RSBAC and PaX, less well tested
GIT
RSBAC source code, can be unstable sometimes
Events
No events planned
Theses are benchmarks made with :
All tests were run 3 times and the average calculated in second.
We are compiling a vanilla 2.6.9 kernel from kernel.org
Four kernels have been tested:
Real represents the real time spent, in seconds User represents the total of CPU-seconds in user mode Sys represents the total of CPU-seconds in kernel mode Lower is better.
| real | user | sys | |
|---|---|---|---|
| Vanilla | 656 | 619 | 26 |
| RC+Auth | 661 | 619 | 34 |
| Gentoo | 664 | 620 | 44 |
| All On | 679 | 623 | 58 |
All theses values are percentages of performance decrease, based on Vanilla = 100 percent. Lower is better.
| real_p | user_p | sys_p | |
|---|---|---|---|
| Vanilla | 100 | 100 | 100 |
| RC+Auth | 100,7 | 100 | 130,7 |
| Gentoo | 101,2 | 100,1 | 169,23 |
| All On | 103,5 | 100,6 | 223,07 |
As you can see, only the kernel mode time increase significantly in RSBAC 1.2.3, but there are only very few actions done in this mode so it doesn’t really affect the real time very much.
To sum it up, we have a 1.2% difference of real time taken with a production kernel setup (Gentoo config), only 0.7% on a RC+Auth only enabled kernel, and a bigger 3.5% with every single option on.