Please configure 11-current for an EXP-RUN containing the ASLR patches in review D473 and monitor build times and build failure of ports.
Take for exp-run , it will take around a week.
ok, which of those 3 do you want to compare:
1) D473 with options PAX_ASLR and options PAX_SYSCTLS
2) D473 without options PAX_ASLR and options PAX_SYSCTLS
3) vanilla kernel
Can you do all three? Unless the hardware is identical to the official package builder, in which case we already have a baseline.
All three would be best. Can you also monitor the build time of world+kernel with the patches? You need to build/install at least the kernel anyways, so getting metrics of the build would also kill two birds with one stone.
Created attachment 147971 [details]
Created attachment 147972 [details]
patched with ASLR on
Created attachment 147973 [details]
patched without ASLR
I attached the build times for the 3 scenarios, the precision is 1 second.
There were 2 new reproducible failures with ASLR on:
Those failure do not happen without ASLR.
(In reply to Antoine Brodin from comment #8)
> I attached the build times for the 3 scenarios, the precision is 1 second.
> There were 2 new reproducible failures with ASLR on:
> Failure logs:
> Those failure do not happen without ASLR.
Thank you! So that tells us that having ASLR on gives us better performance than without ASLR compiled in at all. A savings of 7984 seconds, just over two hours.
I don't think so, building phantomjs takes around 2 hours when it succeeds ;-)
Ah, so we're probably right in line with performance, then.
EXP-RUN done as requested. New bug request should be submitted if desired.
MARKED AS SPAM