Hi Theo, On Thu, Mar 02, 2023 at 09:06:17AM -0700, Theo de Raadt wrote: > While at it, maybe we need a /dev/bullshit linked to /dev/random?? > > This extra node looks like bullshit to me. > > It is encouraging a large application to do a round-trip through the > kernel, for a rare occurance.
The python case is a bad use of /dev/full. It's intended more for providing a convenient way of testing shell scripts against an ENOSPC condition than use by 'large applications'. But your point highlighting the potential for abuse of it is valid, and I can see why we shouldn't encourage that. If anybody ever wants it, the patch is in the archives. > What next? /dev/multiply -- you write two 32-bit values and then you can > read a new 32 bit value? Actually, that sounds quite cool :-).
