glusterfs.git/transport/ib-verbs/src, branch v3.0.4rc2

transport/ib-verbs: assign to qpreg before accessing it in __ib_verbs_lookup_peer.

2009-11-19T10:57:45+00:00

Signed-off-by: Raghavendra G 
Signed-off-by: Anand V. Avati 

BUG: 381 (glusterfs crash in ib-verbs)
URL: http://bugs.gluster.com/cgi-bin/bugzilla3/show_bug.cgi?id=381

transport/ib-verbs: fix race-condition resulting in freeing of transport while it was still being used.

2009-11-16T08:41:36+00:00

- while handling a failed work completion status, the transport is
    disconnected. But further down in the code, the transport is still
    used. There can be a possibility that transport might've been freed
    by the time control reaches this point. More detailed description of
    events leading to this situation can be found at
    http://bugs.gluster.com/cgi-bin/bugzilla3/show_bug.cgi?id=381

Signed-off-by: Raghavendra G 
Signed-off-by: Anand V. Avati 

BUG: 381 (glusterfs crash in ib-verbs)
URL: http://bugs.gluster.com/cgi-bin/bugzilla3/show_bug.cgi?id=381

transport/ib-verbs: synchronize ib_verbs_recv_completion_proc with ib_verbs_receive so that the former doesn't overwrite the pointer from which latter reads.

2009-11-16T08:41:32+00:00

- There can be a condition wherein,
    1. the thread executing ib_verbs_recv_completion_proc (thr 1) stores the
       buffer pointer and notifies the upper translators about a POLLIN event.
    2. the thread waiting for events on socket (thr 2) calls transport_receive,
       but in ib_verbs_receive it has not still read the data.
    3. thr 1 receives work completion event for another work request and
       overwrites the buffer pointer.
    4. thr 2 reads from the new pointer there by missing the data stored in
       buffer pointed by pointer which got overwritten.

Signed-off-by: Raghavendra G 
Signed-off-by: Anand V. Avati 

BUG: 382 (Data can be lost before it is read in ib_verbs_receive.)
URL: http://bugs.gluster.com/cgi-bin/bugzilla3/show_bug.cgi?id=382

transport/name.c: refine the address resolution logic when listen-host is not specified.

2009-11-03T14:48:58+00:00

- when listen-host is not specified and there are are no interfaces having
    adresses belonging to the address-family specified, listen at 0.0.0.0/::0.

  - this patch is necessary since with AI_ADDRCONFIG, getaddrinfo fails if there
    are no active interfaces for the address family specified and when
    listen-host is specified we still want the functionality provided with
    usage of AI_ADDRCONFIG.
Signed-off-by: Corentin Chary 
Signed-off-by: Anand V. Avati 

BUG: 339 (glusterfsd fails to start when there are no active interfaces having address in the address family configured.)
URL: http://bugs.gluster.com/cgi-bin/bugzilla3/show_bug.cgi?id=339

Changed occurrences of Z Research to Gluster.

2009-10-07T10:54:49+00:00

Signed-off-by: Anand V. Avati

transport/ib-verbs: initialize fini member of new-transports created during accepting client connections.

2009-09-15T06:33:04+00:00

- This bug used to cause a memory leak of 2 * sizeof(ib_verbs_private_t) for each new client connection.

Signed-off-by: Anand V. Avati 

BUG: 259 (Memory leak on server side when there are large number of disconnections from clients)
URL: http://bugs.gluster.com/cgi-bin/bugzilla3/show_bug.cgi?id=259

ib-verbs: handle failure of ibv_create_qp gracefully.

2009-09-02T17:30:15+00:00

Signed-off-by: Anand V. Avati 

BUG: 217 (crash in ib_verbs_create_qp)
URL: http://bugs.gluster.com/cgi-bin/bugzilla3/show_bug.cgi?id=217

increased ib-verbs buffer size

2009-08-05T00:07:04+00:00

Noticed that in few cases, where ib-verbs doesn't handle a bigger
sized buffer to be sent across, which happens without problem in
tcp. Caused frame losses in the case where server's reply msg was
bigger, hence the msg got dropped at the server end. With this patch
ib-verbs buffer size is fixed to 512KB. (4 x page-size)

Signed-off-by: Anand V. Avati 

BUG: 190 (missing frames due to larger reply message size.. (ib-verbs))
URL: http://bugs.gluster.com/cgi-bin/bugzilla3/show_bug.cgi?id=190

client-protocol: log messages at WARNING loglevel when client_bind on privileged port fails.

2009-07-31T20:11:32+00:00

Signed-off-by: Anand V. Avati 

BUG: 140 (use appropriate loglevel to log in case of failure to bind to privileged socket)
URL: http://bugs.gluster.com/cgi-bin/bugzilla3/show_bug.cgi?id=140

ib-verbs: don't append ioq_entry to pending_list if first attempt of writing to network fails

2009-07-16T06:15:04+00:00

* fix to bug #101 - When a buffer bigger than ib-verbs-work-request-send-size
    was attempted to write to network, ib-verbs returns ENOTCONN. Neverthless,
    the ioq_entry corresponding to the write was appended to the pending lists
    of ioq_entries waiting to be written to network. This resulted in double
    free of header, once in protocol_client_xfer and again during cleanup of
    transport. The transport cleanup happened due to the timeouts of
    subsequent operations, since ioq_entry corresponding to writev
    was blocking any other operations from reaching server.

Signed-off-by: Anand V. Avati 

BUG: 101 (ib-verbs config crashing while dd'ing with a big mtu size)
URL: http://bugs.gluster.com/cgi-bin/bugzilla3/show_bug.cgi?id=101