glusterfs.git/rpc/rpc-transport/socket/src, branch v3.7.0

socket: use OpenSSL multi-threading interfaces

2015-04-09T09:55:37+00:00

OpenSSL isn't thread-safe unless you register these locking and thread
ID functions.  Most often the crashes would occur around
X509_verify_cert, even though it's insane that the certificate parsing
functions wouldn't be thread-safe.  The bug for this was filed over
two years ago, but it didn't seem like a high priority because the bug
didn't bite anyone until it caused a spurious regression-test failure.
Ironically, that was on a test for a *different* spurious
regression-test failure, which I guess is just deserts[1] for leaving
this on the to-do list so long.

[1] Yes, it really is "deserts" in that phrase - not as in very dry
places, but from late Latin "deservire" meaning to serve well or
zealously.  Aren't commit messages educational?

Change-Id: I2a6c0e9b361abf54efa10ffbbbe071404f82b0d9
BUG: 906763
Signed-off-by: Jeff Darcy 
Reviewed-on: http://review.gluster.org/10075
Tested-by: Gluster Build System 
Reviewed-by: Kaleb KEITHLEY 
Reviewed-by: Vijay Bellur

socket: use TLS 1.2 instead of 1.0

2015-03-27T18:29:33+00:00

Change-Id: I96e9b37e4855f5e12b2dbecf1f0b0887b21ad5ad
Signed-off-by: Jeff Darcy 
Reviewed-on: http://review.gluster.org/9949
Reviewed-by: Kaleb KEITHLEY 
Tested-by: Gluster Build System 
Reviewed-by: Vijay Bellur

socket: use TCP_USER_TIMEOUT to detect client failures quicker

2015-03-17T12:10:17+00:00

Use the network.ping-timeout to set the TCP_USER_TIMEOUT socket option
(see 'man 7 tcp'). The option sets the transport.tcp-user-timeout option
that is handled in the rpc/socket layer on the protocol/server side.
This socket option makes detecting unclean disconnected clients more
reliable.

When the socket gets closed, any locks that the client held are been
released. This makes it possible to reduce the fail-over time for
applications that run on systems that became unreachable due to
a network partition or general system error client-side (kernel panic,
hang, ...).

It is not trivial to create a test-case for this at the moment. We need
a client that unclean disconnects and an other client that tries to take
over the lock from the disconnected client.

URL: http://supercolony.gluster.org/pipermail/gluster-devel/2014-May/040755.html
Change-Id: I5e5f540a49abfb5f398291f1818583a63a5f4bb4
BUG: 1129787
Signed-off-by: Niels de Vos 
Reviewed-on: http://review.gluster.org/8065
Tested-by: Gluster Build System 
Reviewed-by: soumya k 
Reviewed-by: Santosh Pradhan 
Reviewed-by: Kaleb KEITHLEY

epoll: Fix broken RPC throttling due to MT epoll

2015-03-02T06:50:07+00:00

The RPC throttle which kicks in by setting the poll-in event on a
socket to false, is broken with the MT epoll commit. This is due
to the event handler of poll-in attempting to read as much out of
the socket till it receives an EAGAIN. Which may never happen and
hence we would be processing far more RPCs that we want to.

This is being fixed by changing the epoll from ET to LT, and
reading request by request, so that we honor the throttle.

The downside is that we do not drain the socket, but go back to
epoll_wait before reading the next request, but when kicking in
throttle, we need to anyway and so a busy connection would degrade
to LT anyway to maintain the throttle. As a result this change
should not cause deviation in the performance much for busy
connections.

Change-Id: I522d284d2d0f40e1812ab4c1a453c8aec666464c
BUG: 1192114
Signed-off-by: Shyam 
Reviewed-on: http://review.gluster.org/9726
Tested-by: Gluster Build System 
Reviewed-by: Krishnan Parthasarathi 
Reviewed-by: Raghavendra G 
Tested-by: Raghavendra G

socket: allow only one epoll thread to read msg fragments

2015-02-28T05:16:11+00:00

__socket_read_reply function releases sock priv->lock briefly for
notifying higher layers of message's xid. This could result in other
epoll threads that are processing events on this socket to read further
fragments of the same message. This may lead to incorrect fragment
processing and result in a crash.

Change-Id: I915665b2e54ca16f2ad65970e51bf76c65d954a4
BUG: 1197118
Signed-off-by: Krishnan Parthasarathi 
Signed-off-by: Shyam 
Reviewed-on: http://review.gluster.org/9742
Tested-by: Gluster Build System 
Reviewed-by: Raghavendra G 
Tested-by: Raghavendra G

epoll: edge triggered and multi-threaded epoll

2015-02-07T21:17:30+00:00

- edge triggered (oneshot) polling with epoll
- pick one event to avoid multiple events getting picked up by same
  thread
  and so get better distribution of events against multiple threads
- wire support for multiple poll threads to epoll_wait in parallel
- evdata to store absolute index and not hint for epoll
- store index and gen of slot instead of fd and index hint
- perform fd close asynchronously inside event.c for multithread safety
- poll is still single threaded

Change-Id: I536851dda0ab224c5d5a1b130a571397c9cace8f
BUG: 1104462
Signed-off-by: Anand Avati 
Signed-off-by: Vijaikumar M 
Signed-off-by: Jeff Darcy 
Signed-off-by: Shyam 
Reviewed-on: http://review.gluster.org/3842
Tested-by: Gluster Build System 
Reviewed-by: Raghavendra G 
Reviewed-by: Vijay Bellur

socket: fix segfaults when TLS management connections fail

2015-01-27T14:03:58+00:00

Change-Id: I1fd085b04ad1ee68c982d3736b322c19dd12e071
BUG: 1160900
Signed-off-by: Jeff Darcy 
Reviewed-on: http://review.gluster.org/9059
Tested-by: Gluster Build System 
Reviewed-by: Harshavardhana 
Reviewed-by: Vijay Bellur

rpc/rpcsvc: add peername to log messages

2014-12-11T06:35:03+00:00

This would allow users/developers to associate rpc layer log messages
to the corresponding connection.

Change-Id: I040f79248dced7174a4364d9f995612ed3540dd4
Signed-off-by: Krishnan Parthasarathi 
Reviewed-on: http://review.gluster.org/8535
Tested-by: Gluster Build System 
Reviewed-by: Rajesh Joseph 
Reviewed-by: Raghavendra G 
Tested-by: Raghavendra G

socket: disallow CBC cipher modes

2014-10-27T11:40:55+00:00

This is related to CVE-2014-3566 a.k.a. POODLE.

	http://web.nvd.nist.gov/view/vuln/detail?vulnId=CVE-2014-3566

POODLE is specific to CBC cipher modes in SSLv3.  Because there is no
way to prevent SSLv3 fallback on a system with an unpatched version of
OpenSSL, users of such systems can only be protected by disallowing CBC
modes.  The default cipher-mode specification in our code has been
changed accordingly.  Users can still set their own cipher modes if they
wish.  To support them, the ssl-authz.t test script provides an example
of how to combine the CBC exclusion with other criteria in a script.

Change-Id: Ib1fa547082fbb7de9df94ffd182b1800d6e354e5
BUG: 1155328
Signed-off-by: Jeff Darcy 
Reviewed-on: http://review.gluster.org/8962
Tested-by: Gluster Build System 
Reviewed-by: Kaleb KEITHLEY 
Reviewed-by: Vijay Bellur

Sane default for SSL on OSX

2014-09-29T08:30:44+00:00

- /opt/local is not preferred anymore use /usr/local

Change-Id: I30cad4cbd28850063f26121cace05371e13bb314
BUG: 1129939
Signed-off-by: Harshavardhana 
Reviewed-on: http://review.gluster.org/8872
Tested-by: Gluster Build System 
Reviewed-by: Vijay Bellur