lwn.git/fs/nfs/nfs4filelayoutdev.c, branch v3.13.3

NFSv4.1: Ensure memory ordering between nfs4_ds_connect and nfs4_fl_prepare_ds

2013-09-29T19:58:35+00:00

We need to ensure that the initialisation of the data server nfs_client structure in nfs4_ds_connect is correctly ordered w.r.t. the read of ds->ds_clp in nfs4_fl_prepare_ds. Signed-off-by: Trond Myklebust

NFSv4.1: nfs4_fl_prepare_ds - fix bugs when the connect attempt fails

2013-09-29T19:56:35+00:00

- Fix an Oops when nfs4_ds_connect() returns an error. - Always check the device status after waiting for a connect to complete. Reported-by: Andy Adamson Reported-by: Jeff Layton Signed-off-by: Trond Myklebust Cc: # v3.10+

NFSv4.1 Fix gdia_maxcount calculation to fit in ca_maxresponsesize

2013-06-28T19:34:43+00:00

The GETDEVICEINFO gdia_maxcount represents all of the data being returned within the GETDEVICEINFO4resok structure and includes the XDR overhead. The CREATE_SESSION ca_maxresponsesize is the maximum reply and includes the RPC headers (including security flavor credentials and verifiers). Split out the struct pnfs_device field maxcount which is the gdia_maxcount from the pglen field which is the reply (the total) buffer length. Signed-off-by: Andy Adamson Signed-off-by: Trond Myklebust

NFSv4.1: Use layout credentials for get_deviceinfo calls

2013-06-06T20:24:37+00:00

This is not strictly needed, since get_deviceinfo is not allowed to return NFS4ERR_ACCESS or NFS4ERR_WRONG_CRED, but lets do it anyway for consistency with other pNFS operations. Signed-off-by: Trond Myklebust

NFS4.1 Fix data server connection race

2013-05-08T21:19:32+00:00

Unlike meta data server mounts which support multiple mount points to the same server via struct nfs_server, data servers support a single connection. Concurrent calls to setup the data server connection can race where the first call allocates the nfs_client struct, and before the cache struct nfs_client pointer can be set, a second call also tries to setup the connection, finds the already allocated nfs_client, bumps the reference count, re-initializes the session,etc. This results in a hanging data server session after umount. Signed-off-by: Andy Adamson Signed-off-by: Trond Myklebust

sunrpc: move address copy/cmp/convert routines and prototypes from clnt.h to addr.h

2013-02-05T14:41:14+00:00

These routines are used by server and client code, so having them in a separate header would be best. Signed-off-by: Jeff Layton Acked-by: Trond Myklebust Signed-off-by: J. Bruce Fields

NFSv4.1: Cleanup move session slot management to fs/nfs/nfs4session.c

2012-12-05T23:30:45+00:00

NFSv4.1 session management is getting complex enough to deserve a separate file. Signed-off-by: Trond Myklebust

NFSv4.1: Remove assertion BUG_ON()s from the files and generic layout code

2012-11-04T19:43:39+00:00

Signed-off-by: Trond Myklebust

NFSv4.1: Kill nfs4_ds_disconnect()

2012-10-15T14:49:42+00:00

There is nothing to prevent another thread from dereferencing ds->ds_clp during or after the call to nfs4_ds_disconnect(), and Oopsing due to the resulting NULL pointer. Instead, we should just rely on filelayout_mark_devid_invalid() to keep us out of trouble by avoiding that deviceid. Signed-off-by: Trond Myklebust

NFSv4.1: pNFS data servers may be temporarily offline

2012-09-28T20:03:09+00:00

In cases where the pNFS data server is just temporarily out of service, we want to mark it as such, and then try again later. Typically that will be in cases of network connection errors etc. This patch allows us to mark the devices as being "unavailable" for such transient errors, and will make them available for retries after a 2 minute timeout period. Signed-off-by: Trond Myklebust