On November 23, 2001 02:54 pm, Andrew Ebling wrote: > > > ii) disable CPU hardware cache for rogue processes > > > > This one needs more explanation > > OK... by example: imagine a badly/maliciously written program that > allocates a very large 2D array and accesses it the "wrong way around" > resulting in a low cache hit ratio. Not only does it run slowly itself > (but that is OK), but it can badly penalise other processes, even those > with a much higher priotity (NOT OK). Is this something we've actually seen, or is it something that has inspired countermeasures in existing kernels? > If the CPU hardware cache was disabled for all vm areas belonging to the > rogue process, other processes cache hit ratio and execution speed would > not be hit so hard. > > Certainly ix86 hardware could do this (just set the appropriate flag in > the page table), but I don't know about other archs. -- Daniel -- Kernelnewbies: Help each other learn about the Linux kernel. Archive: http://mail.nl.linux.org/kernelnewbies/ IRC Channel: irc.openprojects.net / #kernelnewbies Web Page: http://www.kernelnewbies.org/
