Re: pcp updates - yippee secure socket connections work

[fche@xxxxxxxxxx (Frank Ch. Eigler)]

Ken McDonell <kenj@xxxxxxxxxxxxxxxx> writes:

> [...]  Since this is ONLY at the core of EVERYTHING in PCP, it would
> be a good idea to have as may knowing eyes look at this particular
> change as possible and be ruthless in your reviewing.

It looks like it should cure the problem ... but ... now it puts a
blocking loop into the core pduread() function, which is not supposed
to block.  So it makes it possible for a remote attacker to open a 
connection, send just one byte down the pipe, and DoS the pmcd. 

Instead, how about inlining pduread() within __pmGetPDU(), where the
packet timeout may be observed during the incremental assembly of the
header *and* the payload.

(Also, the new code shouldn't ever use read(2) on the fd, but only
__pmRecv, methinks.)

- FChE