On 9/8/25 3:40 PM, Jeff Layton wrote: > On Mon, 2025-09-08 at 11:38 +1000, NeilBrown wrote: >> This first of these patchs is part of my work to change how directory >> locking is managed. That will involve moving the lock as close as possible >> to the operation being locked, and using some standard interfaces >> which combine the lock and the lookup. Then changing the mechanics of >> taking a lock. >> >> nfsd4_list_rec_dir() currenty locks a direct and performs a lookup >> in a different function to where the lock and lookup results are needed, >> and does it even when those are not needed at all. So the first >> patch moves the lock and lookup to where it is needed. >> >> The second patch (arguably) improves the calling protocol for >> nfs4_client_to_reclaim(). If people don't like this second patch I'm >> happy for it to be dropped. It is the first patch which is particularly >> important to me. >> >> Thanks, >> NeilBrown >> >> >> [PATCH 1/2] nfsd: move name lookup out of nfsd4_list_rec_dir() >> [PATCH 2/2] nfsd: change nfs4_client_to_reclaim() to allocate data > > I'm fine with both of these, so: > > Reviewed-by: Jeff Layton <jlayton@xxxxxxxxxx> > > ...this does remind me though: > > Is it time to switch the default for CONFIG_NFSD_LEGACY_CLIENT_TRACKING > to N? It has been a little over a year since we added the Kconfig > option (and had it default to Y). <shrug> Send a patch? I'm not opposed. -- Chuck Lever
