Fixes a problem where there are gaps in the recovery queue. Example my_aru = 5,
but there are messages at 7,8. 8 = my_high_seq_received which results
in data slots taken up in new message queue. What should really happen
is these last messages should be delivered after a transitional
configuration to maintain SAFE agreement. We don't have support for
SAFE atm, so it is probably safe just to throw these messages away. Without
this change, the new message queue on a new configuraton change is out of sync.
Signed-off-by: Steven Dake <sdake@redhat.com>
Tested-by: Tim Beale <tlbeale@gmail.com>
Reviewed-by: Jan Friesse <jfriesse@redhat.com>
Rather than curiously disable RDMA support by default in configure and
enable it by default in RPM builds, streamline the default
configuration to always turn RDMA support off. It can be enabled in
RPM builds with "--with rdma".
Signed-off-by: Florian Haas <florian.haas@linbit.com>
Reviewed-by: Steven Dake <sdake@redhat.com>
Having to force {ibverbs,rdmacm}_{LIBS,CFLAGS} looks positively odd;
so this may warrant further review. However, they are definitely not
needed if building without RDMA support.
Signed-off-by: Florian Haas <florian.haas@linbit.com>
Reviewed-by: Steven Dake <sdake@redhat.com>
Enable RDMA in RPM builds by default to maintain the previous behavior
(which always included --enable-rdma in the %configure invocation).
Signed-off-by: Florian Haas <florian.haas@linbit.com>
Reviewed-by: Steven Dake <sdake@redhat.com>
Failure to force "C" dates will have RPM et al. complain about invalid
dates and timestamps.
Signed-off-by: Florian Haas <florian.haas@linbit.com>
Reviewed-by: Steven Dake <sdake@redhat.com>
For the case where _POSIX_THREAD_PROCESS_SHARED < 1, the code doesn't compile
for corosync v1.3.1. And when it does compile, it crashes on our system - our
version of uClibc seems to always expect a 4th arg. The man pages suggests
the 4th arg is optional, but does say: 'For greater portability it is best to
always call semctl() with four arguments', which is what this patch does.
Also removed semop as it's an unused variable.
Signed-off-by: Tim Beale <tim.beale@alliedtelesis.co.nz>
Reviewed-by: Steven Dake <sdake@redhat.com>
On our system the expected buffer length is 256. This means calls to
getpwnam_r()/getgrnam_r() return ERANGE error and corosync fails to startup.
These 2 functions return ERANGE when insufficient buffer space is supplied.
Judging by the man page for getpwnam_r, the correct way to determine the
buffersize on any given system is to use sysconf().
Signed-off-by: Tim Beale <tim.beale@alliedtelesis.co.nz>
Reviewed-by: Steven Dake <sdake@redhat.com>
This patch automatically recovers redundant ring failures.
Please note that this patch introduced rrp_autorecovery_check_timeout
in totem config hence breaks internal ABI. The internal ABI users
of totem.h need to rebuild their binaries.
Signed-off-by: Jiaju Zhang <jjzhang@suse.de>
Signed-off-by: Steven Dake <sdake@redhat.com>
Tested-by: Jan Friesse <jfriesse@redhat.com>
Tested-by: Florian Haas <florian.haas@linbit.com>
Tested-by: Jiaju Zhang <jjzhang@suse.de>
Allow a make coverity target for those developers with coverity tools
available to them.
Signed-off-by: Steven Dake <sdake@redhat.com>
Reviewed-by: Angus Salkeld <asalkeld@redhat.com>
Spinlocks are now removed, because even spinlock can improve
speed is some special cases, in most cases it makes corosync CPU usage
much more intensive and less responsive then if only mutexes are used.
What we were doing is:
pthread_mutex_lock
pthread_spin_lock
pthread_spin_unlock
pthread_mutex_unlock
what is not safe.
Signed-off-by: Jan Friesse <jfriesse@redhat.com>
Reviewed-by: Steven Dake <sdake@redhat.com>
Test is disabled by default because it depends on SMP and about 2GB RAM.
It's also testing race, so test is unreliable.
Signed-off-by: Jan Friesse <jfriesse@redhat.com>
Reviewed-by: Steven Dake <sdake@redhat.com>
Following situation could happen:
- one thread is waiting for finish write operation (line 853), objdb is
locked
- flush (done in objdb_notify_dispatch) is called in main thread, but
this call will never appear because main thread is waiting for objdb
lock.
In this situation deadlock appears.
Commit solves this by:
- setting pipe to non-blocking mode
- pipe is used only as trigger for coropoll
- dispatch messages are stored in list
- main thread is processing messages from list
Signed-off-by: Jan Friesse <jfriesse@redhat.com>
Reviewed-by: Steven Dake <sdake@redhat.com>
Following situation could happen:
- process 1 thru confdb creates find handle
- calls find iteration once
- different process 2 deletes object pointed by process 1 iterator
- process 1 calls iteration again ->
object_find_instance->find_child_list is invalid pointer
-> segfault
Now object_find_create creates array of matching object handlers and
object_find_next uses that array together with check for name. This
prevents situation where between steps 2 and 3 new object is created
with different name but sadly with same handle.
Also good to note that this patch is more or less quick hack rather
then proper solution. Real proper solution is to not use pointers
and rather use handles everywhere. This is big TODO.
Signed-off-by: Jan Friesse <jfriesse@redhat.com>
Reviewed-by: Steven Dake <sdake@redhat.com>
Redundant ring has some problem in the UDP unicast mode. The problem
is the second ring has not been successfully initialized, that is, the
second time iface_changes happens, the member list for that interface
has not been added, which results in that ring cannot transmit normal
message. So the second ring cannot take over the work if the first
ring is down. This patch fixes this issue.
comments from review:
More work is needed probably in totemnet where totemnet maintains the
the of node list and an iterator for them, and totemudpu_member_add adds
state information to a context for the iteration.
In any regard, that is somewhat difficult to test, so I'll merge this
patch for now - keep in mind interface changes on the bindnetaddr will
cause problems with udpu after this patch has been commmitted.
Signed-off-by: Jiaju Zhang <jjzhang@suse.de>
Reviewed-by: Steven Dake <sdake@redhat.com>
According specification recvmsg can return 0, which means that
connection is closed. We had this check, but limited only for systems
other then Linux. recvmsg can return 0 even on Linux, so check is now
applied on all systems.
Signed-off-by: Jan Friesse <jfriesse@redhat.com>
Reviewed-by: Angus Salkeld <asalkeld@redhat.com>
in confdb_object_iter result of object_find_create is now properly
checked. object_find_create can return -1 if object doesn't exists.
Without this check, incorrect handle (memory garbage) was directly
passed to object_find_next.
Signed-off-by: Jan Friesse <jfriesse@redhat.com>
Reviewed-by: Angus Salkeld <asalkeld@redhat.com>
with bits set to 1023, buf of 256 bytes was filled by rng_get_bytes
up to 257 bytes. Buf is now 258 bytes so it's no longer problem.
Signed-off-by: Jan Friesse <jfriesse@redhat.com>
Reviewed-by: Steven Dake <sdake@redhat.com>
In this concrete case result is equivalent but makes coverity happy.
Signed-off-by: Jan Friesse <jfriesse@redhat.com>
Reviewed-by: Steven Dake <sdake@redhat.com>
In totemudp_mcast_thread_state_constructor memset to
sizeof(struct totemudp_mcast_thread_state) instead of size of
pointer.
Signed-off-by: Jan Friesse <jfriesse@redhat.com>
Reviewed-by: Steven Dake <sdake@redhat.com>
memory_map function internally limits maximum path size to
PATH_MAX but coroipcc_zcb_alloc passed smaller buffer.
Signed-off-by: Jan Friesse <jfriesse@redhat.com>
Reviewed-by: Steven Dake <sdake@redhat.com>
X86 processors are able to handle unaligned memory access. Improve
performance by using that feature on i386 and x86_64 compatible
processors, and use old aligning code on different processors.
Signed-off-by: Jan Friesse <jfriesse@redhat.com>
Reviewed-by: Steven Dake <sdake@redhat.com>
If one node is paused it can miss a config change and
thus report a larger old_members than expected.
The solution is to use the left_nodes field.
Master selection used to be "choose node with":
1) largest previous membership
2) (then as a tie-breaker) node with smallest nodeid
New selection:
1) largest (previous #nodes - #nodes know to have left)
2) (then as a tie-breaker) node with smallest nodeid
Signed-off-by: Angus Salkeld <asalkeld@redhat.com>
