using a similar approach to
43bead3645
"Send one confchg event per CPG group to CPG client"
which did the same for leave events on a network partition.
Signed-off-by: Fabian Grünbichler <f.gruenbichler@proxmox.com>
Reviewed-by: Jan Friesse <jfriesse@redhat.com>
Also add comment why return value is currently not used.
Signed-off-by: Jan Friesse <jfriesse@redhat.com>
Reviewed-by: Christine Caulfield <ccaulfie@redhat.com>
Check for NULL was invalid, because macaddr_str is ether defined in cmap
or set to "54:54:01:00:00:00".
Signed-off-by: Jan Friesse <jfriesse@redhat.com>
Reviewed-by: Christine Caulfield <ccaulfie@redhat.com>
ipaddr existence is checked for being not NULL by caller setup_nozzle.
Also ipaddr was passed to reparse_nozzle_ip_address function unchecked
so code would crash before reaching the actual check.
Signed-off-by: Jan Friesse <jfriesse@redhat.com>
Reviewed-by: Christine Caulfield <ccaulfie@redhat.com>
Setting res to -1 was not entirely following semantics of "equal"
operation. Set it to 0 and return it when families differs makes
compiler happy.
Signed-off-by: Jan Friesse <jfriesse@redhat.com>
Reviewed-by: Christine Caulfield <ccaulfie@redhat.com>
Compiler may have problem understanding relation between addr1p and
addrlen. Small change makes code a little more readable and compiler
happy.
Signed-off-by: Jan Friesse <jfriesse@redhat.com>
Reviewed-by: Christine Caulfield <ccaulfie@redhat.com>
This feature allows corosync to block packets received from unknown
nodes (nodes with IP address which is not in the nodelist). This is
mainly for situations when "forgotten" node is booted and tries to join
cluster which already removed such node from configuration. Another use
case is to allow atomic reconfiguration and rejoin of two separate
clusters.
Signed-off-by: Jan Friesse <jfriesse@redhat.com>
Reviewed-by: Christine Caulfield <ccaulfie@redhat.com>
It needs to be done at both reload and initialize time.
Also disable access lists if the config key is removed.
Signed-off-by: Christine Caulfield <ccaulfie@redhat.com>
Reviewed-by: Jan Friesse <jfriesse@redhat.com>
currently knet acl are only available on master
but they might be backported
to stable1 as they don´t break onwire protocol.
Signed-off-by: Fabio M. Di Nitto <fdinitto@redhat.com>
Reviewed-by: Christine Caulfield <ccaulfie@redhat.com>
Reviewed-by: Jan Friesse <jfriesse@redhat.com>
Make sure the retransmit timeout have the lowest limit
`MINIMUM_TIMEOUT`. So, the lowest limit of hold should be
recalculated.
Also token timeout and retransmits count should
keep a relational expression.
Signed-off-by: yuan ren <yren@suse.com>
Reviewed-by: Jan Friesse <jfriesse@redhat.com>
When adding a new link for the first time you will often see:
1) knet_link_set_ping_timers for nodeid 1, link 1 failed: Invalid
argument (22)
2) New config has different knet transport for link 1. Internal value
was NOT changed. To reconfigure an interface it must be deleted and
recreated. A working interface needs to be available to corosync at all
times
1) is caused by setting the ping timers twice, once in
totemknet_member_add() and once in totemknet_refresh_config().
The first time we don't know the value
so it's zero and thus display an error. For this we simply check
for the zero and skip the knet API call. It's not ideal, but
totemconfig needs a lot of reconfiguring itself before we can
make this more sane.
2) was caused by simply comparing an unconfigured link with
a configured one, so OF COURSE, they are going to be different!
Signed-off-by: Christine Caulfield <ccaulfie@redhat.com>
Reviewed-by: Jan Friesse <jfriesse@redhat.com>
When UDP is used as a transport, the error would occur
"Multicast address family does not match bind address family"
because there is no ipv6 in /etc/hosts specified but using the
totem.ip_version: ipv6-4. because
the mcastaddr generated (if not specified) only according to
the totem.ip_version.
Solution is to use bindnetaddr (configured or generated from
nodelist) addr family.
Signed-off-by: yuan ren <yren@suse.com>
Reviewed-by: Jan Friesse <jfriesse@redhat.com>
Checking whole structure is fine for IPv4, but IPv6 contains also scope
id, what may be problem for local address. It's possible to use a zone
index, but because it's not required when host name is used, it
shouldn't be needed when IPv6 address is used.
Example configuration snip which fails without patch:
...
nodelist {
node {
nodeid: 1
ring0_addr: fe80:🔢5678:9abc:def1
}
}
...
(example succeed when %eth0 is used).
With patch, zone index is not needed.
Signed-off-by: Jan Friesse <jfriesse@redhat.com>
Reviewed-by: Christine Caulfield <ccaulfie@redhat.com>
Previously the reason field for the member_list items
in cpg_totem_confchg_fn was unset what may be little confusing.
Solution is to add a special value CPG_REASON_UNDEFINED and use it for
the member_list items.
Signed-off-by: Jan Friesse <jfriesse@redhat.com>
Reviewed-by: Christine Caulfield <ccaulfie@redhat.com>
with the following semantics:
- default off
- implies crypto_hash SHA256 and crypto_cipher AES256
- crypto_* have higher precedence
- only applicable for knet, like crypto_*
this should make upgrading from Corosync 2.x less painful for users that
have an explicit secauth=on in their configuration.
Signed-off-by: Fabian Grünbichler <f.gruenbichler@proxmox.com>
Reviewed-by: Jan Friesse <jfriesse@redhat.com>
Triple DES is considered as a "weak cipher" since 2016 so there is
really no need to support it in the corosync. Thanks to bug in
Corosync/Knet/NSS which caused 3des to not work at all,
no matter what library was used, we can just remove support for 3des
without braking the compatibility.
Also fix coroparse so:
- totem.crypto_type is removed (this is 1.x construct which was not used
even in 2.x)
- Add checking of totem.crypto_model.
- Enumarate possible values for crypto_model, crypto_cipher and
crypto_hash error messages
Signed-off-by: Jan Friesse <jfriesse@redhat.com>
Reviewed-by: Christine Caulfield <ccaulfie@redhat.com>
so that we get the nice log message when attempting to modify them at
runtime, just like for totem.crypto_* and co.
Signed-off-by: Fabian Grünbichler <f.gruenbichler@proxmox.com>
Reviewed-by: Jan Friesse <jfriesse@redhat.com>
A nozzle device is a pseudo ethernet device that routes network
traffic through a channel on the corosync knet network (NOT cpg or any
corosync internal service) to other nodes in the cluster. It allows
applications to take advantage of knet features such as multipathing,
automatic failover, link switching etc.
Signed-off-by: Christine Caulfield <ccaulfie@redhat.com>
Reviewed-by: Jan Friesse <jfriesse@redhat.com>
AF_UNSPEC returns different results than AF_INET/AF_INET6, because of
nsswitch.conf search is in order and it stops asking other
modules once current module success.
Example of difference between previous and new code when ipv6-4 is used:
- /etc/hosts contains test_name with an ipv4
- previous code called AF_INET6 where /etc/hosts failed so other methods
were used which may return IPv6 addr -> result was ether fail or IPv6
address.
- new code calls AF_UNSPEC returning IPv4 defined in /etc/hosts ->
result is IPv4 address
New code behavior should solve problems caused by nss-myhostname.
Signed-off-by: Jan Friesse <jfriesse@redhat.com>
Reviewed-by: Fabio M. Di Nitto <fdinitto@redhat.com>
more changes are to be expected on this front as the API evolves in
knet master.
Signed-off-by: Fabio M. Di Nitto <fdinitto@redhat.com>
Reviewed-by: Jan Friesse <jfriesse@redhat.com>
Here we're very far from entering the main loop, even farther from
sending the READY notification to systemd. This sounded awkward:
systemd[1]: Starting Corosync Cluster Engine...
corosync[827]: [MAIN ] Corosync Cluster Engine ('2.99.5'):
started and ready to provide service.
corosync[827]: [MAIN ] Corosync built-in features: dbus monitoring
watchdog augeas systemd xmlconf snmp pie relro bindnow
corosync[827]: [MAIN ] parse error in config: No interfaces defined
corosync[827]: [MAIN ] Corosync Cluster Engine exiting with status 8
at main.c:1378.
systemd[1]: corosync.service: Main process exited, code=exited,
status=8/n/a
systemd[1]: corosync.service: Failed with result 'exit-code'.
systemd[1]: Failed to start Corosync Cluster Engine.
Signed-off-by: Ferenc Wágner <wferi@debian.org>
Reviewed-by: Jan Friesse <jfriesse@redhat.com>
Reviewed-by: Christine Caulfield <ccaulfie@redhat.com>
system.run_dir was a little bit unfortunate and confusing name. Rename
to state_dir makes more evident what is content of this directory. To
keep setting consistent with code, get_run_dir is changed to
get_state_dir.
Signed-off-by: Jan Friesse <jfriesse@redhat.com>
Reviewed-by: Christine Caulfield <ccaulfie@redhat.com>
Originally totem.ip_version was used to force ip version used by totem.
With Knet this variable didn't make too much sense so it was not used.
Sadly rely only on DNS resolver order doesn't always work (RFC is quite
complicated, but if IPv6 is not configured then IPv4 is preferred), what
we tried to solve by forcing IPv6 and only if that fails, use IPv4.
Sadly this collides with nss_myhostname which is able to return every
local address and today system usually have at least one autogenerated
link-local IPv6 address so it is able to "overwrite" /etc/hosts.
Solution is to enhance totem.ip_version and use it also for Knet.
totem.ip_version is now just a flag for resolver and can have four
states: ipv4 (only IPv4 is used), ipv6 (only IPv6 is used), ipv4-6 (ask
IPv4 first and if it fails ask for IPv6) and ipv6-4 (ask IPv6 first and
if it fails ask for IPv4). Default for Knet and UDPU transports is
ipv6-4, for UDP it's ipv4, because autogenerated mcast addr doesn't play
too well with ipv6-4.
So everywhere where nss_myhostname becomes problem, it's just possible
to set totem.ip_version to ipv4-6.
Signed-off-by: Jan Friesse <jfriesse@redhat.com>
Reviewed-by: Jan Friesse <jfriesse@redhat.com>
It's required to create TOTEM logsys subsys before totemip_parse is used
(so before totem_config_read). Logsys is not yet fully initialized, but
it's good enough.
Signed-off-by: Jan Friesse <jfriesse@redhat.com>
Reviewed-by: Christine Caulfield <ccaulfie@redhat.com>
Current practice is to let getaddrinfo() decide which address we get
but this is not necessarily deterministic as DNS servers won't
always return addresses in the same order if a node has
several. While this doesn't deal with node names that have
multiple IP addresses of the same family (that's an installation issue
IMHO) we can, at least, force a definite order for IPv6/IPv4 name
resolution.
I've chosen IPv6 then IPv4 as that's what happens on my test system (
using /etc/hosts) and it also seems more 'future proof'.
Signed-off-by: Christine Caulfield <ccaulfie@redhat.com>
Reviewed-by: Jan Friesse <jfriesse@redhat.com>
It didn't work anyway (the config system requires whole links
to be configured at once) and caused crashes.
Signed-off-by: Christine Caulfield <ccaulfie@redhat.com>
Reviewed-by: Jan Friesse <jfriesse@redhat.com>
Corosync used to just ignore parse errors so that un-resolved names
could cause silent failures. We now always check the result from
totemip_parse() and at least print something in syslog.
There's also a little get-out here that allows you to correct
a bad node address without having to destroy and recreate the
whole link. I'm being nice to you.
Signed-off-by: Christine Caulfield <ccaulfie@redhat.com>
Reviewed-by: Jan Friesse <jfriesse@redhat.com>
When cmap_track_delete was called to stats map (cmap created with
CMAP_MAP_STATS parameter) result was always ERR_BAD_HANDLE.
It turned out that corosync part of cmap is always calling icmap
function to get user data (where required hdb handle is stored)
instead of generalized map_fns.
After fixing this issue, valgrind showed error about jump depending on
unitialized data in stats_map_track_delete. Solution seems to be always
initialize tracker->events (so not only when track_type is add or
delete).
Signed-off-by: Jan Friesse <jfriesse@redhat.com>
Reviewed-by: Christine Caulfield <ccaulfie@redhat.com>
Remove last used environment variable (reasons similar to removal of
COROSYNC_MAIN_CONFIG_FILE).
This environment variable was never documented, so document it properly.
Signed-off-by: Jan Friesse <jfriesse@redhat.com>
Reviewed-by: Christine Caulfield <ccaulfie@redhat.com>
Remove another environment variable (reasons similar to removal of
COROSYNC_MAIN_CONFIG_FILE).
Also properly document both totem.keyfile and totem.key.
Signed-off-by: Jan Friesse <jfriesse@redhat.com>
Reviewed-by: Christine Caulfield <ccaulfie@redhat.com>
COROSYNC_MAIN_CONFIG_FILE environment variable was quite well hidden
and it was never used by init script. It also makes quite hard to debug
possible problems.
Replace it by -c option.
Also patch makes use of configuration file path as a base for uidgid.d
directory, so it's no longer needed to keep uidgid.d in sysconfdir.
Signed-off-by: Jan Friesse <jfriesse@redhat.com>
Reviewed-by: Christine Caulfield <ccaulfie@redhat.com>
The reason for this change is, that number of corosync CLI options
kind of exploded and scheduler based one are really beter to be kept in
config file.
Nice side-effect of this move is better "integration" with systemd,
because currently used EnvironmentFile should be really used for
environment and not that much for passing extra options to CLI.
Signed-off-by: Jan Friesse <jfriesse@redhat.com>
Reviewed-by: Christine Caulfield <ccaulfie@redhat.com>
The conversion to the new srp_addr format broke the feature where
UDP/UDPU nodes could get their nodeids generated from the IP address.
A big part of this was the removal of mandatory ring0_addr - it was used
as a placeholder when reading down the nodelist. I replaced this with
nodeid thinking that nodeid was now mandatory, forgetting this use case.
So the compare on "ring0_addr" or "nodeid" is now replaced with a more
robust check that we're only reading keys from the same node_pos once,
this was needed in votequorum.c as well as totemconfig.c
Another tidying side-effect of this patch is that the nodeid generation
is now all in a single routine in totemconfig.c and not shared between
it and totemip.c.
Signed-off-by: Christine Caulfield <ccaulfie@redhat.com>
Reviewed-by: Jan Friesse <jfriesse@redhat.com>
Feature depends on existence of libqb function qb_log_file_reopen.
New function call is added into CFG service API. This function is
used by corosync-cfgtool which now accepts -L parameter.
Finally, logrotate "postrotate" script is calling
corosync-cfgtool -L to notify corosync, instead of using
copytruncate option.
Signed-off-by: Jan Friesse <jfriesse@redhat.com>
Reviewed-by: Christine Caulfield <ccaulfie@redhat.com>
Formally not needed, because totemip_print should not return string
longer than INET6_ADDRSTRLEN, but static analysis tools are not capable
of such conclusion.
Signed-off-by: Jan Friesse <jfriesse@redhat.com>
Reviewed-by: Christine Caulfield <ccaulfie@redhat.com>
This is a bug I seem to have introduced in
429209f4aa where we compare links
for changes. if a new node was added on an existing link then it
was compared against a non-existant one in the previous configuration.
We now only compare nodes that are in both interfaces.
As I needed min() for this function, I moved it from individual
.c files into util.h so we only have one copy.
And the error message was fixed.
Signed-off-by: Christine Caulfield <ccaulfie@redhat.com>
Reviewed-by: Jan Friesse <jfriesse@redhat.com>
Instead of compiling totempg as a shared library, compile all totem code
directly into corosync binary.
Main idea of having totempg which may be
used in other projects was nice, but never really finished (and as far
as I know no project were ever really using it). So at the end of the
day, we've end with huge amount of problems (need to pass new arguments
thru X layers, hard debugging, ...) without any real benefit.
For a future version, we may consider to revisit idea of split totemsrp
into well tested library without unrelated bits like transports/ip/...
Signed-off-by: Jan Friesse <jfriesse@redhat.com>
Reviewed-by: Fabio M. Di Nitto <fdinitto@redhat.com>
Reviewed-by: Christine Caulfield <ccaulfie@redhat.com>
Complete removal of NSS from corosync tree. Most of the changes are
in build system and cpgverify had to be rewritten to use crc32 instead
of sha1.
Signed-off-by: Jan Friesse <jfriesse@redhat.com>
Reviewed-by: Fabio M. Di Nitto <fdinitto@redhat.com>
Reviewed-by: Christine Caulfield <ccaulfie@redhat.com>
It's just much easier to find out what is happening when message like
parser error: /etc/corosync/corosync.conf:39: Unexpected closing brace
is logged instead of
parser error: Unexpected closing brace
Signed-off-by: Jan Friesse <jfriesse@redhat.com>
Reviewed-by: Christine Caulfield <ccaulfie@redhat.com>
Corosync parser is not very clever, but it is able to detect more errors
without too much code.
1. Check if section name is not empty (just '{' character)
2. Check if there is no extra characters after opening bracket '{'
3. Check if there is no extra characters after or before closing bracket
'}'
4. Check if line is opening section, closing section or key/value
So following examples are reported as error:
totem {
version: 2
}}}}}}}}}}
Signed-off-by: Jan Friesse <jfriesse@redhat.com>
Reviewed-by: Christine Caulfield <ccaulfie@redhat.com>
When remove_whitespace function parameter is single character string
with whitespaces (like a:) then colon is not removed. Reason is end
condition end != start, which is valid for empty string, but invalid in
case described above. Solution is to check if *end is '\0'.
Signed-off-by: Jan Friesse <jfriesse@redhat.com>
Reviewed-by: Christine Caulfield <ccaulfie@redhat.com>
Libcgroup is deprecated and not shipping with new distributions
(OpenSuSE is one example). Solution is to have a partial implementation
of required functionality of libcgroup in the corosync code.
Patch uses hardcoded cgroup mount point, because most of the systems are
now systemd and systemd is also using hardcoded mountpoint (see
https://github.com/systemd/systemd/blob/master/src/core/mount-setup.c)
Configuration option --enable-cgroup is gone, because it's not needed
any longer.
Big thanks to Christine Caulfield <ccaulfie@redhat.com> for example of
simplified implementation of cgroup management code primitives.
Signed-off-by: Jan Friesse <jfriesse@redhat.com>
Reviewed-by: Christine Caulfield <ccaulfie@redhat.com>
Token_warning is used to present information about
when the token was last received.
Signed-off-by: Chris Walker <cwalker@cray.com>
Reviewed-by: Jan Friesse <jfriesse@redhat.com>
Add assert when there are no members in token_memb structure so
non-existing member is not accessed (token should always have
at least one member).
Signed-off-by: Jan Friesse <jfriesse@redhat.com>
Reviewed-by: Christine Caulfield <ccaulfie@redhat.com>
... so error_reason can be fully included into parse error message.
Signed-off-by: Jan Friesse <jfriesse@redhat.com>
Reviewed-by: Christine Caulfield <ccaulfie@redhat.com>
Trailing zero is always added so there is no need to have a warning
about unterminated destination string.
Signed-off-by: Jan Friesse <jfriesse@redhat.com>
Reviewed-by: Christine Caulfield <ccaulfie@redhat.com>
Trailing zero is always added so there is no need to have a warning
about unterminated destination string.
Signed-off-by: Jan Friesse <jfriesse@redhat.com>
Reviewed-by: Christine Caulfield <ccaulfie@redhat.com>
This patch intends to solve long time ifdown corosync problem. Idea is
to use local socket for sending both unicast and multicast messages if
interface is down.
Together with testing what is current bind state it's possible to keep
pretending existence of old IP address instead of rebinding to localhost
what breaks a lot things badly.
Heavilly based on Yu, Zou <zouyu@shiqichuban.com> work and it's
basically port of UDP patch created by
Jan Friesse <jfriesse@redhat.com>.
(ported from needle 96354fba72)
Signed-off-by: Bin Liu <bliu@suse.com>
Reviewed-by: Jan Friesse <jfriesse@redhat.com>
KNET requires that all links be full-mesh (this may change in the future
but almost certainly not before knet 2.0), so enforce this in the
config.
Also avoid a potential div-by-0 error if the local node is not fully
configured either.
Signed-off-by: Christine Caulfield <ccaulfie@redhat.com>
Reviewed-by: Jan Friesse <jfriesse@redhat.com>
If the local host does not have a 'name' attribute and the cluster
has more than one link then fail the validation test.
I'm open to the idea of checking all of the nodes in the nodelist
if necessary. It seems overkill as each node will check its own
entry though.
Signed-off-by: Christine Caulfield <ccaulfie@redhat.com>
Reviewed-by: Jan Friesse <jfriesse@redhat.com>
There are a few things in the interface that cannot be changed on the
fly. Warn about them and tell the user that these things need to be done
in two steps and why.
Signed-off-by: Christine Caulfield <ccaulfie@redhat.com>
Reviewed-by: Jan Friesse <jfriesse@redhat.com>
Compiler shows warnings about possible not large enough buffer, so check
snprintf return value properly.
Signed-off-by: Jan Friesse <jfriesse@redhat.com>
Reviewed-by: Christine Caulfield <ccaulfie@redhat.com>
It wasn't hardmful, but it generated an annoying message
Signed-off-by: Christine Caulfield <ccaulfie@redhat.com>
Reviewed-by: Jan Friesse <jfriesse@redhat.com>
knet sends log messages as struct knet_log_msg, not a string
of KNET_MAX_LOG_MSG_SIZE (which is only part of that structure).
So we were both losing and corrupting messages.
Signed-off-by: Christine Caulfield <ccaulfie@redhat.com>
Reviewed-by: Jan Friesse <jfriesse@redhat.com>
Patch tries to fix incorrect behaviour during following test-case:
- 3 nodes
- Node 1 is paused
- Node 2 and 3 detects node 1 as failed and informs CPG clients
- Node 1 is unpaused
- Node 1 clients are informed about new membership, but not about Node 1
being paused, so from Node 1 point-of-view, Node 2 and 3 failure
Solution is to:
- Remove downlist master choose and always choose local node downlist.
For Node 1 in example above, downlist contains Node 2 and 3.
- Keep code which informs clients about left nodes
- Use joinlist as a authoritative source of nodes/clients which exists
in membership
This patch doesn't break backwards compatibility.
I've walked thru all the patches which changed behavior of cpg to ensure
patch does not break CPG behavior. Most important were:
- 058f50314c - Base. Code was significantly
changed to handle double free by split group_info into two structures
cpg_pd (local node clients) and process_info (all clients). Joinlist
was
- 97c28ea756 - This patch removed
confchg_fn and made CPG sync correct
- feff0e8542 - I've tested described
behavior without any issues
- 6bbbfcb6b4 - Added idea of using
heuristics to choose same downlist on all nodes. Sadly this idea
was beginning of the problems described in
040fda8872,
ac1d79ea7c,
559d4083ed,
02c5dffa5b,
64d0e5ace0 and
b55f32fe2e
- 02c5dffa5b - Made joinlist as
authoritative source of nodes/clients but left downlist_master_choose
as a source of information about left nodes
Long story made short. This patch basically reverts
idea of using heuristics to choose same downlist on all nodes.
(ported from needle 9c2a97f4f9)
Signed-off-by: Jan Friesse <jfriesse@redhat.com>
Reviewed-by: Christine Caulfield <ccaulfie@redhat.com>
Leave message in totem is just join message where leaving member is
excluded from member list and included in fail list. It also contains
special nodeid in header.nodeid and system_from.nodeid fields.
Before "totem: Use nodeid ONLY in srp_addr" fix, most of the functions
were using system_from addresses and not nodeid, which was used only in
one specific case for memb_consensus_set function.
After the patch, addresses are gone and only nodeid is used. Result is,
that leaving node nodeid is not added into local fail list
(my_faillist) so node is unable to reach consensus till token timeout,
which starts new gather process.
Solution is to send valid leaving node nodeid in system_from.nodeid and
handle specific case for memb_consensus_set in memb_join_process.
Signed-off-by: Jan Friesse <jfriesse@redhat.com>
Reviewed-by: Christine Caulfield <ccaulfie@redhat.com>
These information are useful and with trace log level they should not be
too much irritating.
Signed-off-by: Jan Friesse <jfriesse@redhat.com>
Reviewed-by: Christine Caulfield <ccaulfie@redhat.com>
There is regression caused by "totem: Use nodeid ONLY in srp_addr" patch
in srp_addr_compare function. This function should be usable with qsort,
so it should return values less than, equal to or greater than zero. It
was however returning only zero or negation of a zero. Final results
were unable to reach consensus in following test case:
- 3 node cluster
- start nodes 1, 2, 3
- shutdown node 3
- start node 3
- shutdown node 2
- start node 2
- shutdown node 1
After this steps, node 2 and 3 were unable to reach consensus.
Signed-off-by: Jan Friesse <jfriesse@redhat.com>
Reviewed-by: Christine Caulfield <ccaulfie@redhat.com>
If number of proc_list, failed_list or active members is too high it
may be impossible to put them into message, which is allocated on the
stack what results in stack corruption.
Signed-off-by: Jan Friesse <jfriesse@redhat.com>
Reviewed-by: Christine Caulfield <ccaulfie@redhat.com>
Sanity checkers are used to prevent crashing because of
accessing unallocated memory.
Signed-off-by: Jan Friesse <jfriesse@redhat.com>
Reviewed-by: Christine Caulfield <ccaulfie@redhat.com>
To make finding victim of incompatible messages easier, IP of sender is
logged. Propagating IP in layers makes patch slightly larger.
Signed-off-by: Jan Friesse <jfriesse@redhat.com>
Reviewed-by: Christine Caulfield <ccaulfie@redhat.com>
Magic number (0xC070) together with version in every packet
is used for detecting that other node is really
Corosync 3.x.
Endian_detector field is removed and magic number is now
used instead.
If received packet magic number differs, guessing is used to show more
about the source (Corosync 2.3+, 2.2 are quite reliable, Knet and
unencrypted Corosync 2.1/2.0/1.x/OpenAIS are semi-reliable and encrypted
Corosync 2.1/2.0/1.x/OpenAIS are quite unreliable).
Signed-off-by: Jan Friesse <jfriesse@redhat.com>
Reviewed-by: Christine Caulfield <ccaulfie@redhat.com>
because totemknet always configures link0 as loopback even
if it's not known to corosync, we need to filter it
out when returning the link status, as things get misaligned
in cfg.
Signed-off-by: Christine Caulfield <ccaulfie@redhat.com>
Reviewed-by: Jan Friesse <jfriesse@redhat.com>
Even if it's not used for anything else.
Also, make cfgtool show the correct link ID when links are not
contiguous
Signed-off-by: Christine Caulfield <ccaulfie@redhat.com>
Reviewed-by: Jan Friesse <jfriesse@redhat.com>
Fix crash introduced a couple of commits ago in iface_get
Signed-off-by: Christine Caulfield <ccaulfie@redhat.com>
Reviewed-by: Jan Friesse <jfriesse@redhat.com>
Allow ring0_addr to be used in place of 'name' for
backwards compatibility
Signed-off-by: Christine Caulfield <ccaulfie@redhat.com>
Reviewed-by: Jan Friesse <jfriesse@redhat.com>
Make the message more representative of what's going on.
Signed-off-by: Christine Caulfield <ccaulfie@redhat.com>
Reviewed-by: Jan Friesse <jfriesse@redhat.com>
This shrinks the srp_addr (and consequently every packet sent by
corosync) so that instead of containing loads of IP addresses to
identify a node, it just sends the nodeid.
This then allows us to make ring0 optional and replaceable when running
knet.
It also means that we need some other way of identifying the local
node in corosync.conf, so the nodelist.node.name entry is now mandatory
and is mapped to the local host using the same algorithm as used in
cman.
This code needs LOTS of testing as it touches a huge amount of totemsrp
and totemconfig.
Signed-off-by: Christine Caulfield <ccaulfie@redhat.com>
Reviewed-by: Jan Friesse <jfriesse@redhat.com>
Commit 899cb29983 changed copy_len
to iovec[i].iov_len, assuming,
copy_len is always the same as iovec[i].iov_len under those
circumstances, but it missed the possability of small message being
partly put at the end of packet, which cuts this message in two parts
and therefore making copy_len not equal to iovec[i].iov_len.
This is revert of 899cb29983
Signed-off-by: Rytis Karpuška <rytisk@neurotechnology.com>
Reviewed-by: Jan Friesse <jfriesse@redhat.com>
To be more explicit that we are copying whole message.
Related to 0ebae6b47d.
Signed-off-by: Rytis Karpuška <rytisk@neurotechnology.com>
Reviewed-by: Jan Friesse <jfriesse@redhat.com>
The problem was that two or more messages were concatenated
together during fragmentation in mcast_msg() function. In specific case,
message of just short of 1MB was provided for mcast_msg() and it
happened so, that the remainder (212 bytes to be exact) left some free
space in packet, therefore branch
if ((copy_len + fragment_size) <
(max_packet_size - sizeof (unsigned short))) {
...
was selected and this was the last mesage in provided iovec.
Then, on the second call, came another big message (about 300KB ) and
during fragmentation mcast.fragmented was set to 1.
On the other end, while receiving messages, due to missing
mcast.fragmentation==0 those two messages were concatenated and
therefore assembly->data array overflowed overwriting linked list
pointers and offset (which happened to be set to 0 and that 300KB
message was being copied from the beginning again).
After whole 300KB message has been sent, mcast.fragmentation==0 arrived
and totempg_deliver_fn() tried to move assembly structure to
assembly_list_free list, but as linked list pointers has been overriden,
segfault occured.
Signed-off-by: Rytis Karpuška <rytisk@neurotechnology.com>
Reviewed-by: Jan Friesse <jfriesse@redhat.com>
