> On 02/16/2017 12:21 PM, Frank Filz wrote: > >> mdcache_readdir() checks !mdc_dircache_trusted(), and calls > >> mdcache_dirent_populate(), which calls mdcache_dirent_invalidate_all(). > > > > Ah, ok. Good. > > > > Path will be slightly re-worked with chunking, but will do that. > > > > I wonder if we should just always mark contents un-trusted and then in > > readdir, lookup, and create, dump the entries? > > > > Frank > > I don't have a strong opinion about that... It's this way because that's how > cache_inode did it, not because I chose to do it that way. > > Upcalls have the desired property, I think, of not introducing extra latency > into a thread owned by another system. I guess it might be good to remove > that latency from unlink/rename/etc.? But on the other hand, it frees > memory now that may never be freed otherwise, which can be beneficial...
Hmm, that might be a good point. With chunking at least we won't have the memory sitting forever since LRU will dispose of chunks that haven't been used in a while, combined with limit on number of chunks in play. Frank --- This email has been checked for viruses by Avast antivirus software. https://www.avast.com/antivirus ------------------------------------------------------------------------------ Check out the vibrant tech community on one of the world's most engaging tech sites, SlashDot.org! http://sdm.link/slashdot _______________________________________________ Nfs-ganesha-devel mailing list Nfs-ganesha-devel@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/nfs-ganesha-devel