Propietary NVIDIA driver pwn3d

[soralx at cydem.org]

> > > > al least those can be promptly fixed...
> > > Wasn't this one promptly fixed too?
> > This bug was reported on 2004-12-21:
> > <https://bugs.freedesktop.org/show_bug.cgi?id=2129>.
> Well...sort of.  The *symptoms* are reported there but no-one
> identified it as a nvidia driver bug.

duh... could the fact that the bug was well hidden within the depths
of the binary blob play a role? Debugging _with_ a source code is a
hell of a way to waste time sometimes, let alone trying to fix
something when all you have is a binary module... a kernel module.

des has a good point, though. IMHO, what he wanted to say is this:
it doesn't help to be so ungrateful to nvidia. Bugs happen, especially
in things as comlicated as modern videocard drivers. Videocards these
days are far beyond the point when, given full documentation, a single
person could write a driver for. They're not like a NIC, a soundcard,
etc in that respect. Moreover, neither nvidia nor ati (or anyone else)
are allowed to release any docs on their chips without an NDA (I know
not what they're hiding, but, given how vast the windoze market is,
they have a lot to lose). So those companies who do release a working
driver for their videocards deserve a little respect for at least
_trying_ to support us OSS users. However, a little more cooperation
hardly would hurt, I think.

Now, what to do if you need to use the nvidia driver, but want to
avoid the security risk? Back in the good old days (for me, these
were the days of DOS and Windoze 98), I used to patch misbehaving
programs by disassembling and modifying the data... you know,
standard procedure: insert some instructions bytes in suitable empty
space, then trow in some form of jump, etc. I'm curious now, would
this type of approach work in FreeBSD kernel module?

> In fact there's no mention of nvidia at all until September 2006.
> 
> Kris

[SorAlx]  ridin' VN1500-B2