Al Viro <viro@xxxxxxxxxxxxxxxxxx> writes: > On Mon, Mar 26, 2012 at 12:54:23PM +0200, Miklos Szeredi wrote: >> This also eliminates the weird retry loop, that could, in theory, retry the >> cached lookup any number of times (very unlikely scenario: needs two parallel >> do_lookups and d_revalidate always returning zero). > > That really needs to be carved into much smaller pieces - the sucker is > convoluted as hell and there's a lot of codepaths in there with nearly > zero test coverage. I've split it up into provably equivalent > transformations, leading more or less to the state where yours ends up. > I _think_ I've reconstructed the sequence of changes more or less > close to what you were doing there, but the next time you have to do > something of that kind, do not collapse that into a single patch. It's > really easier to review step by step... Okay, but actually what I was doing there is looking at what the code actually does and realizing that it's equivalent to __lookup_hash(), so for me it was a single (albeit complex) single step. But I'll try to keep reviewability in mind. The do_last() reorganization in the atomic-open series needs to be split up, I realize. Do you have any other high level comments about that series? Thanks, Miklos -- To unsubscribe from this list: send the line "unsubscribe linux-fsdevel" in the body of a message to majordomo@xxxxxxxxxxxxxxx More majordomo info at http://vger.kernel.org/majordomo-info.html
