On Sat, 15 Nov 2003 07:01:09 EST, Bryan Allen <[EMAIL PROTECTED]> said: > Restarting sshd does not kill running ssh sessions, as they're all > forked and in memory.
What? You never said "Shit, I should have checked if I can su/login/ssh/whatever" *after* you closed your last root window? :) And yes, *your* ssh can still go away if something else does a runaway and runs the system out of swap space. Although many vendors have an out-of-memory handler that does the best thing most of the time, none of them do the best thing ALL the time. And even if it doesn't, the system may be simply thrashing itself to death but *not* actually killing anything.. What use is an open SSH window, Mr Anderson, if you have no character echo? I'm surprised that on a security list, so many people are so ready to take a "very unlikely to happen" condition and label it as a "*cant* happen". Come on guys - that sort of assumption is what often *creates* security holes: "Nobody could POSSIBLY unlink this file and replace it with a symlink between when we stat() it and when we open() it". Things rarely go wrong when everything is at nominal. Things usually go wrong in a failure cascade - when one sysadmin is installing software during a 2AM test window and he's tired and cranky because instead of getting some sleep, he had a big fight with his SO, and the software was built by another sysadmin who didn't get to actually finish those last few tests before she tripped and broke her ankle, while it's being installed onto a disk that's flaky and not always reporting write errors (and yes, I've actually seen all three of those happen in shops I was working in at the time, fortunately not all at once. On the flip side, each of the three was by itself enough to cause a failure....)
pgp00000.pgp
Description: PGP signature
