Kris Kennaway wrote: > I guess I was unclear: the stress2 test calls random syscalls with > random arguments, so the error condition is probably from insufficient > sanity checking of the input in aio_read(). Just a thought - does that include using random values of 0,1,2 for file descriptor?