On Mon, 18 Mar 2019 at 16:15, Mark Syms <[email protected]> wrote: > [Mark Syms] That works better.
Okay, great. > We were wondering whether it might be a bit too aggressive though in that it > skips writing the inode entirely unless we have WB_SYNC_ALL whereas the patch > that Ross Lagerwall posted originally would use a trylock in the case where > we don't have WB_SYNC_ALL and then fail out. Yes, that may well be true. I'll have to think about it some more; maybe we can avoid the trylock somehow else. > Whether or not this should just silently return 0 or -EAGAIN as Ross' patch > does is a matter of style I guess. What are your thoughts on this? Well, write_inode is allowed to skip inodes in the WB_SYNC_NONE case. If we return an error here, that error might leak and create nasty surprises, so we should better not. Thanks, Andreas
