On 1/4/21 9:54 AM, Al Viro wrote: > On Mon, Jan 04, 2021 at 07:43:17AM -0700, Jens Axboe wrote: > >>> I've not put it into #for-next yet; yell if you see any problems with that >>> branch, or it'll end up there ;-) >> >> Thanks Al - but you picked out of v3, not v4. Not that there are huge >> changes between the two, from the posting of v4: >> >> - Rename LOOKUP_NONBLOCK -> LOOKUP_CACHED, and ditto for the RESOLVE_ >> flag. This better explains what the feature does, making it more self >> explanatory in terms of both code readability and for the user visible >> part. >> >> - Remove dead LOOKUP_NONBLOCK check after we've dropped LOOKUP_RCU >> already, spotted by Al. >> >> - Add O_TMPFILE to the checks upfront, so we can drop the checking in >> do_tmpfile(). >> >> and it sounds like you did the last two when merging yourself. > > Yes - back when I'd posted that review. Gotcha >> I do like >> LOOKUP_CACHED better than LOOKUP_NONBLOCK, mostly for the externally >> self-documenting feature of it. What do you think? > > Agreed, especially since _NONBLOCK would confuse users into assumption > that operation is actually non-blocking... > >> Here's the v4 posting, fwiw: >> >> https://lore.kernel.org/linux-fsdevel/20201217161911.743222-1-axboe@xxxxxxxxx/ > > Sorry, picked from the local branch that sat around since Mid-December ;-/ > Fixed. Another change: ..._child part in unlazy_child() is misleading - > it might as well be used for .. traversal, where dentry is usually the > _parent_ of nd->path.dentry. The real constraint here is that dentry/seq pair > had been valid next position at some point during the RCU walk. Renamed to > try_to_unlazy_next(), (hopefully) fixed the comment... > > Updated variant force-pushed. All good, looks good to me from a quick look. A diff between the two bases just show seems sane. I'll pull it in and rebase on top of that, and re-run the testing just to make sure that no ill effects snuck in there with the changes. -- Jens Axboe
