It makes no sense whatsoever to spend hours or days reworking a major
subsystem *just* so the swap device can do without a dev_t. I don't
really give a damn about /dev/drum -- I said before and I will say again
that we can leave it out. But we need to give the swap device its's
dev_t back to fix VN because no matter what you believe the *right*
thing is to do, (A) nobody has time to rework a major subsystem at this
time, and (B) VN was broken when the dev_t was removed, the dev_t should
be added back in until someone comes up with a better overall solution.
I also strongly believe that polluting *OTHER* subsystems (the buffer
chaining code) just to avoid creating a dev_t for SWAP makes even less
sense.
I do not have the inclination to spend huge amounts of my time reworking
the VOP or filesystem buffer subsystem because someone isn't willing to
spend a tiny amount of code to partially back out a patch that broke
something (VN) in the system, that someone also not themselves willing
to spend the time required to rework the system properly. It's that
simple. It makes no sense to impose such a requirement on someone else
to fix a problem that they did not create.
Now I would like to make a clarification: The commit we are talking
about was actually one Peter did with the approval of everyone, including
me. At the end of November. However, nobody including me realized that
the second part of his commit, which was an optimization to remove the
'cruft' of the SWAP device, broke VN. If we had known then that this
optimization would break VN it would *NOT* have gone in until the VN
problem was fixed.
-Matt
To Unsubscribe: send mail to [EMAIL PROTECTED]
with "unsubscribe freebsd-current" in the body of the message
