Brett Person wrote:
> Certainly, a user doesnt, or shouldnt need 90% of the total
> resources of a machine.
We at the kernel level cannot make such rash assumptions. Although
undesirable, it is not an error if a legitimate user process needs 100%
of physical ram. This is not uncommon in low-memory systems, university
lab-projects, large databases, and highly parallel kernel builds etc.
Yes, its a serious problem if a userspace program can bring down the
machine, but the correct behavior would be to deny requests for
resources beyond those available- no more... apparently this is easier
said than done at the moment.
> Lets say I do something stupid like write an infinite fork while I'm in
> userland. Why cant a daemon notice that user person is being a memory
> hog and kill off his biggest process?
Yes, we've done something like this. CS students have an affinity for
fork bombs- but you need to look for an excessive number of processes
for a single user, not his memory usage.... This is special case with a
potentially expensive solution.
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to email@example.com
Please read the FAQ at http://www.tux.org/lkml/