On Fri, 2018-09-28 at 16:19 -0400, Olga Kornievskaia wrote: > On Fri, Sep 28, 2018 at 4:07 PM Trond Myklebust < > trondmy@xxxxxxxxxxxxxxx> wrote: > > > > On Fri, 2018-09-28 at 15:55 -0400, Olga Kornievskaia wrote: > > > On Fri, Sep 28, 2018 at 3:10 PM Olga Kornievskaia <aglo@xxxxxxxxx > > > > > > > wrote: > > > > > > > Wait, why are we suppose to reclaim the open state when we have a > > > valid open stateid? We don't have any cached opens that server > > > doesn't > > > know about. RFC7530 says "if the file has other open reference", > > > I > > > think the emphasis is on "other". I don't believe we need to be > > > sending anything besides the locks to the server. Then I'm back > > > to > > > square one. > > > > Holding a delegation does not imply that we hold an open stateid. > > Under > > Linux, the open stateid gets closed as soon as the application > > closes > > the file. > > > > The delegation, on the other hand, is retained until either it is > > recalled, or we see that the file has not been used for 2 lease > > periods. > > Ok I agree with all of it but I'm saying it doesn't need to be > reclaimed unconditionally or are you saying that's what the linux > client does? In this test case, the file hasn't been closed or > expired. I'm stating that the client has a valid open stateid and > should only be required to reclaim the locks (which with this patch > it > does). As I said earlier, the client is required to recover all _cached_ open and lock state. If it already holds an open stateid, then it should not need to reclaim the open modes that are covered by that stateid, however it may still need to reclaim those open modes that were not already subject to an explicit OPEN call. IOW: If the file was first opened with an open(O_RDRW) call by the application, but a second application then opened it using open(O_WRONLY), then we may already hold a stateid with a "SHARE_ACCESS_BOTH" open mode, however we will still need to send a reclaim for the cached SHARE_ACCESS_WRITE mode, so that a later OPEN_DOWNGRADE(SHARE_ACCESS_WRITE) can succeed. -- Trond Myklebust Linux NFS client maintainer, Hammerspace trond.myklebust@xxxxxxxxxxxxxxx
