1. 13 Feb, 2017 1 commit
  2. 10 Feb, 2017 9 commits
    • Chuck Lever's avatar
      sunrpc: Allow xprt->ops->timer method to sleep · b977b644
      Chuck Lever authored
      The transport lock is needed to protect the xprt_adjust_cwnd() call
      in xs_udp_timer, but it is not necessary for accessing the
      rq_reply_bytes_recvd or tk_status fields. It is correct to sublimate
      the lock into UDP's xs_udp_timer method, where it is required.
      
      The ->timer method has to take the transport lock if needed, but it
      can now sleep safely, or even call back into the RPC scheduler.
      
      This is more a clean-up than a fix, but the "issue" was introduced
      by my transport switch patches back in 2005.
      
      Fixes: 46c0ee8b ("RPC: separate xprt_timer implementations")
      Signed-off-by: default avatarChuck Lever <chuck.lever@oracle.com>
      Signed-off-by: default avatarAnna Schumaker <Anna.Schumaker@Netapp.com>
      b977b644
    • Chuck Lever's avatar
      xprtrdma: Refactor management of mw_list field · 9a5c63e9
      Chuck Lever authored
      Clean up some duplicate code.
      Signed-off-by: default avatarChuck Lever <chuck.lever@oracle.com>
      Signed-off-by: default avatarAnna Schumaker <Anna.Schumaker@Netapp.com>
      9a5c63e9
    • Chuck Lever's avatar
      xprtrdma: Handle stale connection rejection · 0a90487b
      Chuck Lever authored
      A server rejects a connection attempt with STALE_CONNECTION when a
      client attempts to connect to a working remote service, but uses a
      QPN and GUID that corresponds to an old connection that was
      abandoned. This might occur after a client crashes and restarts.
      
      Fix rpcrdma_conn_upcall() to distinguish between a normal rejection
      and rejection of stale connection parameters.
      
      As an additional clean-up, remove the code that retries the
      connection attempt with different ORD/IRD values. Code audit of
      other ULP initiators shows no similar special case handling of
      initiator_depth or responder_resources.
      Signed-off-by: default avatarChuck Lever <chuck.lever@oracle.com>
      Signed-off-by: default avatarAnna Schumaker <Anna.Schumaker@Netapp.com>
      0a90487b
    • Chuck Lever's avatar
      xprtrdma: Properly recover FRWRs with in-flight FASTREG WRs · 18c0fb31
      Chuck Lever authored
      Sriharsha (sriharsha.basavapatna@broadcom.com) reports an occasional
      double DMA unmap of an FRWR MR when a connection is lost. I see one
      way this can happen.
      
      When a request requires more than one segment or chunk,
      rpcrdma_marshal_req loops, invoking ->frwr_op_map for each segment
      (MR) in each chunk. Each call posts a FASTREG Work Request to
      register one MR.
      
      Now suppose that the transport connection is lost part-way through
      marshaling this request. As part of recovering and resetting that
      req, rpcrdma_marshal_req invokes ->frwr_op_unmap_safe, which hands
      all the req's registered FRWRs to the MR recovery thread.
      
      But note: FRWR registration is asynchronous. So it's possible that
      some of these "already registered" FRWRs are fully registered, and
      some are still waiting for their FASTREG WR to complete.
      
      When the connection is lost, the "already registered" frmrs are
      marked FRMR_IS_VALID, and the "still waiting" WRs flush. Then
      frwr_wc_fastreg marks these frmrs FRMR_FLUSHED_FR.
      
      But thanks to ->frwr_op_unmap_safe, the MR recovery thread is doing
      an unreg / alloc_mr, a DMA unmap, and marking each of these frwrs
      FRMR_IS_INVALID, at the same time frwr_wc_fastreg might be running.
      
      - If the recovery thread runs last, then the frmr is marked
      FRMR_IS_INVALID, and life continues.
      
      - If frwr_wc_fastreg runs last, the frmr is marked FRMR_FLUSHED_FR,
      but the recovery thread has already DMA unmapped that MR. When
      ->frwr_op_map later re-uses this frmr, it sees it is not marked
      FRMR_IS_INVALID, and tries to recover it before using it, resulting
      in a second DMA unmap of the same MR.
      
      The fix is to guarantee in-flight FASTREG WRs have flushed before MR
      recovery runs on those FRWRs. Thus we depend on ro_unmap_safe
      (called from xprt_rdma_send_request on retransmit, or from
      xprt_rdma_free) to clean up old registrations as needed.
      Reported-by: default avatarSriharsha Basavapatna <sriharsha.basavapatna@broadcom.com>
      Signed-off-by: default avatarChuck Lever <chuck.lever@oracle.com>
      Tested-by: default avatarSriharsha Basavapatna <sriharsha.basavapatna@broadcom.com>
      Signed-off-by: default avatarAnna Schumaker <Anna.Schumaker@Netapp.com>
      18c0fb31
    • Chuck Lever's avatar
      xprtrdma: Shrink send SGEs array · c6f5b47f
      Chuck Lever authored
      We no longer need to accommodate an xdr_buf whose pages start at an
      offset and cross extra page boundaries. If there are more partial or
      whole pages to send than there are available SGEs, the marshaling
      logic is now smart enough to use a Read chunk instead of failing.
      Signed-off-by: default avatarChuck Lever <chuck.lever@oracle.com>
      Signed-off-by: default avatarAnna Schumaker <Anna.Schumaker@Netapp.com>
      c6f5b47f
    • Chuck Lever's avatar
      xprtrdma: Reduce required number of send SGEs · 16f906d6
      Chuck Lever authored
      The MAX_SEND_SGES check introduced in commit 655fec69
      ("xprtrdma: Use gathered Send for large inline messages") fails
      for devices that have a small max_sge.
      
      Instead of checking for a large fixed maximum number of SGEs,
      check for a minimum small number. RPC-over-RDMA will switch to
      using a Read chunk if an xdr_buf has more pages than can fit in
      the device's max_sge limit. This is considerably better than
      failing all together to mount the server.
      
      This fix supports devices that have as few as three send SGEs
      available.
      Reported-by: default avatarSelvin Xavier <selvin.xavier@broadcom.com>
      Reported-by: default avatarDevesh Sharma <devesh.sharma@broadcom.com>
      Reported-by: default avatarHonggang Li <honli@redhat.com>
      Reported-by: default avatarRam Amrani <Ram.Amrani@cavium.com>
      Fixes: 655fec69 ("xprtrdma: Use gathered Send for large ...")
      Cc: stable@vger.kernel.org # v4.9+
      Tested-by: default avatarHonggang Li <honli@redhat.com>
      Tested-by: default avatarRam Amrani <Ram.Amrani@cavium.com>
      Tested-by: default avatarSteve Wise <swise@opengridcomputing.com>
      Reviewed-by: default avatarParav Pandit <parav@mellanox.com>
      Signed-off-by: default avatarChuck Lever <chuck.lever@oracle.com>
      Signed-off-by: default avatarAnna Schumaker <Anna.Schumaker@Netapp.com>
      16f906d6
    • Chuck Lever's avatar
      xprtrdma: Disable pad optimization by default · c95a3c6b
      Chuck Lever authored
      Commit d5440e27 ("xprtrdma: Enable pad optimization") made the
      Linux client omit XDR round-up padding in normal Read and Write
      chunks so that the client doesn't have to register and invalidate
      3-byte memory regions that contain no real data.
      
      Unfortunately, my cheery 2014 assessment that this optimization "is
      supported now by both Linux and Solaris servers" was premature.
      We've found bugs in Solaris in this area since commit d5440e27
      ("xprtrdma: Enable pad optimization") was merged (SYMLINK is the
      main offender).
      
      So for maximum interoperability, I'm disabling this optimization
      again. If a CM private message is exchanged when connecting, the
      client recognizes that the server is Linux, and enables the
      optimization for that connection.
      
      Until now the Solaris server bugs did not impact common operations,
      and were thus largely benign. Soon, less capable devices on Linux
      NFS/RDMA clients will make use of Read chunks more often, and these
      Solaris bugs will prevent interoperation in more cases.
      
      Fixes: 677eb17e ("xprtrdma: Fix XDR tail buffer marshalling")
      Cc: stable@vger.kernel.org # v4.9+
      Signed-off-by: default avatarChuck Lever <chuck.lever@oracle.com>
      Signed-off-by: default avatarAnna Schumaker <Anna.Schumaker@Netapp.com>
      c95a3c6b
    • Chuck Lever's avatar
      xprtrdma: Per-connection pad optimization · b5f0afbe
      Chuck Lever authored
      Pad optimization is changed by echoing into
      /proc/sys/sunrpc/rdma_pad_optimize. This is a global setting,
      affecting all RPC-over-RDMA connections to all servers.
      
      The marshaling code picks up that value and uses it for decisions
      about how to construct each RPC-over-RDMA frame. Having it change
      suddenly in mid-operation can result in unexpected failures. And
      some servers a client mounts might need chunk round-up, while
      others don't.
      
      So instead, copy the pad_optimize setting into each connection's
      rpcrdma_ia when the transport is created, and use the copy, which
      can't change during the life of the connection, instead.
      
      This also removes a hack: rpcrdma_convert_iovs was using
      the remote-invalidation-expected flag to predict when it could leave
      out Write chunk padding. This is because the Linux server handles
      implicit XDR padding on Write chunks correctly, and only Linux
      servers can set the connection's remote-invalidation-expected flag.
      
      It's more sensible to use the pad optimization setting instead.
      
      Fixes: 677eb17e ("xprtrdma: Fix XDR tail buffer marshalling")
      Cc: stable@vger.kernel.org # v4.9+
      Signed-off-by: default avatarChuck Lever <chuck.lever@oracle.com>
      Signed-off-by: default avatarAnna Schumaker <Anna.Schumaker@Netapp.com>
      b5f0afbe
    • Chuck Lever's avatar
      xprtrdma: Fix Read chunk padding · 24abdf1b
      Chuck Lever authored
      When pad optimization is disabled, rpcrdma_convert_iovs still
      does not add explicit XDR round-up padding to a Read chunk.
      
      Commit 677eb17e ("xprtrdma: Fix XDR tail buffer marshalling")
      incorrectly short-circuited the test for whether round-up padding
      is needed that appears later in rpcrdma_convert_iovs.
      
      However, if this is indeed a regular Read chunk (and not a
      Position-Zero Read chunk), the tail iovec _always_ contains the
      chunk's padding, and never anything else.
      
      So, it's easy to just skip the tail when padding optimization is
      enabled, and add the tail in a subsequent Read chunk segment, if
      disabled.
      
      Fixes: 677eb17e ("xprtrdma: Fix XDR tail buffer marshalling")
      Cc: stable@vger.kernel.org # v4.9+
      Signed-off-by: default avatarChuck Lever <chuck.lever@oracle.com>
      Signed-off-by: default avatarAnna Schumaker <Anna.Schumaker@Netapp.com>
      24abdf1b
  3. 09 Feb, 2017 4 commits
  4. 08 Feb, 2017 10 commits
  5. 30 Jan, 2017 16 commits