grsecurity-1.9.7c-2.4.19.patch load increase

Discuss usability issues, general maintenance, and general support issues for a grsecurity-enabled system.

grsecurity-1.9.7c-2.4.19.patch load increase

Postby maynard » Sun Oct 13, 2002 12:51 am

Just upgraded from grsecurity 1.97 to 1.97c, but i it seems like my load usage increased after the upgrade. Am i going insane or has the new changes some effect on my load average. I use mrtg to plot the load usage and is has clearly gone up after the upgrade and i haven't changed anything else :)

Regards
Maynard
maynard
 
Posts: 6
Joined: Sat Aug 10, 2002 8:38 am

Re: grsecurity-1.9.7c-2.4.19.patch load increase

Postby PaX Team » Sun Oct 13, 2002 9:24 am

maynard wrote:Just upgraded from grsecurity 1.97 to 1.97c, but it seems like my load usage increased after the upgrade. Am i going insane or has the new changes some effect on my load average.

could you please disable SEGMEXEC (if you had it enabled at all of course) and see if the load changes back? also, can you give us some numbers on the load change?
PaX Team
 
Posts: 2310
Joined: Mon Mar 18, 2002 4:35 pm

Postby maynard » Mon Oct 14, 2002 9:21 am

sorry for the late reply but wanted to be sure that it actually was grsec 1.9.7c that was causing the problem. Have been trying different kernel's the last 24 hours, and came to the conclusion that it is grsecurity1.9.7c that is causing the load increase. But when running 1.9.7c and disabling SEGMEXEC seem to fix things.
The load increase isn't that much of a problem like a 0.10 increase but i just seemed it was odd. Please supply me with a email if you wish to have further details. Like a link to my stats page where it is actually possible to see what version of grsec is running by looking at the stats graph :)

Regards

Maynard
maynard
 
Posts: 6
Joined: Sat Aug 10, 2002 8:38 am

Postby PaX Team » Mon Oct 14, 2002 11:52 am

maynard wrote:But when running 1.9.7c and disabling SEGMEXEC seem to fix things.
The load increase isn't that much of a problem like a 0.10 increase but i just seemed it was odd.

hmm, well, it depends... the only change that would affect performance is the GDT switching code in the i386 context switch call. one would think a few more cycles don't make a difference, but as they do, i'll try to find a better implementation.
Please supply me with a email if you wish to have further details. Like a link to my stats page where it is actually possible to see what version of grsec is running by looking at the stats graph :)

pageexec at freemail.hu (it's on our web site as well btw ;-).
PaX Team
 
Posts: 2310
Joined: Mon Mar 18, 2002 4:35 pm


Return to grsecurity support