PJ wrote: > > Doesn't the blob have the privilege to do horrid things to > > a running kernel (crash, compromise, expose stuff?)
Raj Mathur <[EMAIL PROTECTED]> writes: > > > > I don't know if the above is true - I don't know kernel internals. > > But unverifiable code with kernel privileges is not a reassuring > > thought. > > That's the whole point, isn't it? The blob is not running as part of > your kernel. It has been offloaded to the card in question, and it is > the card's CPU that is running that code. Your kernel does nothing > with the code except to push it onto the card once at initialisation > time. Specifically, the kernel never executes it; heck, most of that > code probably won't even run on x86! Thanks for the clarification. I see that I am almost certainly wrong then in being that worried about the security and stability aspect for off-cpu blobs (I wonder if there are any off-cpu x86 cards around - anyone care to run a disassembler on the payloads and see what comes up? (hmmm... vaguely wondering about the legality of that in some jurisdictions)) Also, is there pretty much a standard hook that is used by all such off-cpu blobs to deliver the payload that is easy to follow for a reviewer? To avoid a trojan sneaking in the kernel "source" via an obfuscated delivery method? PJ _______________________________________________ ilugd mailinglist -- ilugd@lists.linux-delhi.org http://frodo.hserus.net/mailman/listinfo/ilugd Next Event: http://freed.in - February 22-24, 2008 Archives at: http://news.gmane.org/gmane.user-groups.linux.delhi http://www.mail-archive.com/ilugd@lists.linux-delhi.org/
