glusterfs.git/xlators/protocol, branch v3.3.0qa23

core: utilize mempool for frame->local allocations

2012-02-21T10:42:09+00:00

in each translator, which uses 'frame->local', we are using
GF_CALLOC/GF_FREE, which would be costly considering the
number of allocation happening in a lifetime of 'fop'. It
would be good to utilize the mem pool framework for xlator's
local structures, so there is no allocation overhead.

Change-Id: Ida6e65039a24d9c219b380aa1c3559f36046dc94
Signed-off-by: Amar Tumballi 
BUG: 765336
Reviewed-on: http://review.gluster.com/2772
Tested-by: Gluster Build System 
Reviewed-by: Anand Avati

rpc/clnt: handle PARENT_DOWN event appropriately

2012-02-21T09:42:42+00:00

Change-Id: I4644e944bad4d240d16de47786b9fa277333dba4
BUG: 767862
Signed-off-by: Raghavendra G 
Signed-off-by: Amar Tumballi 
Reviewed-on: http://review.gluster.com/2735
Tested-by: Gluster Build System 
Reviewed-by: Vijay Bellur

protocol/client,server: fcntl lock self healing.

2012-02-20T12:45:31+00:00

Currently(with out this patch), on a disconnect the server cleans up
the transport which inturn closes the fd's and releases the locks acquired on
those fd's by that client. On a reconnect, client just reopens the fd's but
doesn't reacquire the locks. The application that had previously acquired
the locks still is under the assumption that it is the owner of those locks
which might have been granted to other clients(if they request) by the server
leading to data corruption.

This patch allows the client to reacquire the fcntl locks (held on the fd's)
during client-server handshake.

* The server identifies the client via process-uuid-xl (which is a combination
  of uuid and client-protocol name, it is assumed to be unique) and lk-version
  number.

* The client maintains a list of process-uuid-xl, lk-version pair for each
  accepted connection. On a connect, the server traverses the list for a
  matching pair, if a matching pair is not found the the server returns
  lk-version with value 0, else it returns the lk-version it has in store.

* On a disconnect, the server and client enter grace period, and on the
  completion of the grace period, the client bumps up its lk-version number
  (which means, it will reacquire the locks the next time) and the server will
  distroy the connection. If reconnection happens within the grace period, the
  server will find the matching (process-uuid-xl, lk-version) pair in its list
  which guarantees that the fd's and there corresponding locks are still valid
  for this client.

Configurable options:
  To set grace-timeout, the following options are
    option server.grace-timeout value
    option client.grace-timeout value

  To enable or disable the lk-heal,
    option lk-heal [on|off]

gluster volume set command can be used to configurable options
Change-Id: Id677ef1087b300d649f278b8b2aa0d94eae85ed2
BUG: 795386
Signed-off-by: Mohammed Junaid 
Reviewed-on: http://review.gluster.com/2766
Tested-by: Gluster Build System 
Reviewed-by: Vijay Bellur

protocol/client: delete locks only for non-anonymous fds

2012-02-20T08:49:44+00:00

delete_granted_lock_owners () is not called for anonymous
fds since they are not involved in locking

Change-Id: Icdc7818f98f5371232ba276ed442704ef69e6b0e
BUG: 787365
Signed-off-by: Rajesh Amaravathi 
Reviewed-on: http://review.gluster.com/2754
Tested-by: Gluster Build System 
Reviewed-by: Amar Tumballi 
Reviewed-by: Vijay Bellur

iobuf: use 'iobuf_get2()' to get variable sized buffers

2012-02-20T08:46:27+00:00

added 'TODO' in places where it is missing.

Change-Id: Ia802c94e3bb76930f7c88c990f078525be5459f5
Signed-off-by: Amar Tumballi 
BUG: 765264
Reviewed-on: http://review.gluster.com/388
Tested-by: Gluster Build System 
Reviewed-by: Vijay Bellur

glusterd: auth allow enhancements

2012-02-20T08:30:54+00:00

* PROBLEM:

  When address-based authentication is enabled on a volume,
  the gNfs server, self-heal daemon (shd), and other operations
  such as quota, rebalance, replace-brick and geo-replication
  either stop working or the services are not started if all
  the peers' ipv{4,6} addresses or hostnames are not added in
  the "set auth.allow" operation, breaking the functionality
  of several operations.

  E.g:
    volume vol in a cluster of two peers:
    /mnt/brick1 in 192.168.1.4
    /mnt/brick2 in 192.168.1.5

    option auth.allow 192.168.1.6
    (allow connection requests only from 192.168.1.6)

    This will disrupt the nfs servers on 192.168.1.{4,5}.
    brick server processes reject connection requests from both
    nfs servers (on 4,5), because the peer addresses are not in
    the auth.allow list.

    Same holds true for local mounts (on peer machines),
    self-heal daemon, and other operations which perform
    a glusterfs mount on one of the peers.

* SOLUTION:

  Login-based authentication (username/password pairs,
  henceforth referred to as "keys") for gluster services and
  operations.

  These *per-volume* keys can be used to by-pass the addr-based
  authentication, provided none of the peers' addresses are put
  in the auth.reject list, to enable gluster services like gNfs,
  self-heal daemon and internal operations on volumes when
  auth.allow option is exercised.

* IMPLEMENTATION:

  1. Glusterd generates keys for each volume and stores it in
     memory as well as in respective volfiles.
     A new TRUSTED-FUSE volfile is generated which is
     fuse volfile + keys in protocol/client,
     and is named trusted--fuse.vol.
     This is used by all local mounts. ANY local mount (on any peer)
     is granted the trusted-fuse volfile instead of fuse volfile
     via getspec. non-local mounts are NOT granted the trusted fuse
     volfile.

  2. The keys generated for the volume is written to each server
     volfile telling servers to allow users with these keys.

  3. NFS, self-heal daemon and replace-brick volfiles are updated
     with the volume's authentication keys.

  4. The keys are NOT written to fuse volfiles for obvious reasons.

  5. The ownership of volfiles and logfiles is restricted to root users.

  6. Merging two identical definitions of peer_info_t in auth/addr
     and rpc-lib, throwing away the one in auth/addr.

  7. Code cleanup in numerous places as appropriate.

* IMPORTANT NOTES:

  1. One SHOULD NOT put any of the peer addresses in the auth.reject
     list if one wants any of the glusterd services and features
     such as gNfs, self-heal, rebalance, geo-rep and quota.

  2. If one wants to use username/password based authentication
     to volumes, one shall append to the server, nfs and shd volfiles,
     the keys one wants to use for authentication, *while_retaining
     those_generated_by_glusterd*.
     See doc/authentication.txt file for details.

Change-Id: Ie0331d625ad000d63090e2d622fe1728fbfcc453
BUG: 789942
Signed-off-by: Rajesh Amaravathi 
Reviewed-on: http://review.gluster.com/2733
Tested-by: Gluster Build System 
Reviewed-by: Vijay Bellur

protocol: remove the 'path<>' from rename() and link()

2012-02-17T06:49:15+00:00

missed it in the previous round of cleanup, path is completely
useless in resolve function.

Change-Id: I1aef0f5276afb77dfacfcc0c337ac80b4fcacc55
Signed-off-by: Amar Tumballi 
BUG: 790298
Reviewed-on: http://review.gluster.com/2756
Tested-by: Gluster Build System 
Reviewed-by: Anand Avati

nfs: avoid logging invalid fdctx in case of anonymous fds

2012-02-14T18:25:30+00:00

if get_fd_ctx fails (as in case of anonymous fds), overwhelming
amount of entries are seen in the nfs log, causing dd and other
heavy i/o operations to become unresponsive. this patch logs an
invalid fdctx only if it is not an anonymous fd.

Change-Id: I4e917d150d6a053af77d47a94a2f1c2633acadb5
BUG: 787365
Signed-off-by: Rajesh Amaravathi 
Reviewed-on: http://review.gluster.com/2747
Tested-by: Gluster Build System 
Reviewed-by: Amar Tumballi

protocol xdr: remove 'path<>'

2012-02-14T12:34:06+00:00

client was sending 'path' on wire, which gets ignored on server
side, and also doesn't get freed up, which causes memory leak.

also with not having path on wire, the xdr size on wire most of the
time can remain constant, which helps in allocating RDMA buffers.

Change-Id: Ie0d36a670be60b02fd1e925c6f977b1a71def5cd
BUG: 790298
Signed-off-by: Amar Tumballi 
Reviewed-on: http://review.gluster.com/2744
Tested-by: Gluster Build System 
Reviewed-by: Anand Avati

core: add an extra flag to readv()/writev() API

2012-02-14T09:09:02+00:00

needed to implement a proper handling of open flag alterations
using fcntl() on fd.

Change-Id: Ic280d5db6f1dc0418d5c439abb8db1d3ac21ced0
Signed-off-by: Amar Tumballi 
BUG: 782265
Reviewed-on: http://review.gluster.com/2723
Tested-by: Gluster Build System 
Reviewed-by: Anand Avati