That's the canonical gcd algorithm, I think from Knuth's Art of Computer Programming

In D26646#594599, @emaste wrote:

uh, this should not have been closed

Yes do not list the review in a commit Differential Revision unless it should be closed. You can list the D# in the commit message and Phab will add a link between the commit and the review - e.g. something like "This commit is part of review D26646" or "Reported by: rene in review D26646"

@rene will you re-upload the diff to put it back in a known stte

In D26651#593797, @hselasky wrote:

@bz

I:vendor;I:device;I:#;I:#;I:#;I:#

Format should use "J" instead, I guess, because it allows -1 as ignore value.

In D26651#593796, @hselasky wrote:

Also sadly add subvendor, etc. fields as comments at the end

to the descriptor_string seem to prevent autoloading and that

Linux uses -1U for wildcard. Maybe devmatch needs support for that too!

In D26638#593511, @jhb wrote:

FYI, the versions I have are at https://github.com/bsdjhb/kdbg/blob/master/gdb/gdb6#L1783 They include a generic 'devwalk' command that invokes a callback on each device_t in a depth-first traversal that I use for some other places besides devinfo.

It would be better to just ignore this entry on errors... However, we don't do that for others, so that would be a stretch goal.

In D24240#592496, @mat wrote:

In D24240#568481, @mandree wrote:

No more complaints received in a month, can I commit?

We do not have more complains, but we still do not agree with the change, so the answer is still no.

All the other filesystems set this to the dev. UFS sets it to a time...

Update, per reviews from the PR

unleak the kenrel bits

Fix kyle's issue

make toggle work

Note: this fixes the logic error in the original code. So BHYVE now works.
The heuristic would be greatly aided by having the ability to read the actual PCI config space, but we have no wrappers for that and don't do it anywhere else so that will be a separate commit.

rebase

Ah, this comment was correct, restore after the move.

coment tweaks.

KASSERT, et al, now are a printf in the boot loader. We don't
have enough milage to turn it on just yet.

put nullability macro outside kernel ifdefs

This revision successfully builds and reboots, so I think it's right But since it's systm.h and we do some funky monkey business in there, I'd like more eyes on it to be sure.

rebase, again

rebase

Rebase (again!)

Rebase

Nice.

btw, the rest of the changes look totally fine. Sorry to be so nitpicky over this one...

This review is missing a lot of files (even prior versions). rtsx.c at the very least is missing
Does this driver support MMCCAM?

Non pmc bits look good to go