With gcc 5.3.1 20151207:
> log.c: In function 'cs_format':
> log.c:182:2: warning: function might be possible candidate for
> 'gnu_printf' format attribute [-Wsuggest-attribute=format]
> len = vsnprintf(str, QB_LOG_MAX_LEN, cs->format, ap_copy);
> ^
We certainly don't want to disable that warning globally so make use of
diagnostic pragmas for GCC instead in one instance that we cannot
annotate properly.
Signed-off-by: Jan Pokorný <jpokorny@redhat.com>
* remove pid/euid from qb_ipcc_connection
* use proper #elif defines
* return NULL instead of 0 for pointers
* return -ENOMEM when malloc fails
* remove redundant if check
* use -1 for uid to chown()
When creating a unix socket it's default gid is that of the parent
directory. If the SOCKETDIR is owned by root:wheel with 1777 mode
some of the pacemaker daemons end up unable to communicate with one
another due to having insufficient permissions on the sockets.
This can be fixed by setting the client sockets gid to the primary
group of the server socket owner it's attempting to connect to. And,
on the server side by setting the gid to the already captured gid
stored in the connection info. This ensures that regardless of who
owns the socket directory, as long as the applications have r/w
access to it they should work.
Currently the mmap files are created in LOCALSTATEDIR/run on non-Linux
platforms which can be problematic with pacemaker since it spawns processes
as the hacluster user, which by default doesn't have write permissions to
it. Using --with-socket-dir partially fixes the problem by allowing the
unix sockets to reside somewhere else but not the mmap files and this
patch puts them in the same directory.
When using sockets for IPC the file permissions default to whatever
the umask is. This isn't a problem on Linux since it uses abstract
namespace sockets which don't have any permissions, but on other
platforms this causes problems with pacemaker which spawns processes
under the hacluster user and ends up failing to connect.
In handle_new_connection(), connection_accept() could fail, which would
leave the state of the connection inactive. Previously, in this case,
the socket and the allocated qb_ipcs_connection would be leaked.
Some platforms (eg. FreeBSD 10+) don't support calling dlopen() from
within a dl_iterate_phdr() callback; so save all of the dlpi_names to
a list and iterate over them afterwards.
This fixes a deadlock which occurs in dlopen() when called from
within dl_iterate_phdr() on FreeBSD 10+ when linked to libthr.
Patch from dpejesh on github.
Reviewed-by: Christine Caulfield <ccaulfie@redhat.com>
Only one of the strndup calls should have subtracted 9 from
the name length. The other should have remained at 8.
Signed-off-by: Christine Caulfield <ccaulfie@redhat.com>
The current code closes the fd before passing it to the
provided delete function. In the default case this can generate
"epoll_ctl(del): Bad file descriptor" messages when debug logging
is enabled.
This patch swaps over the calls so that a valid FD is passed to
the delete function.
Signed-off-by: Christine Caulfield <ccaulfie@redhat.com>
Reviewed-By: Ken Gaillot <kgaillot@redhat.com>
The code removed the last 8 characters of a socket name
to get the 'base name', but 9 are needed.
Signed-off-by: Christine Caulfield <ccaulfie@redhat.com>
This revision switches libtool soname versioning from -version-number to
-version-info, and updates the travis configuration to use their newly
recommended infrastructure.
When both the corosync and pacemaker has PID larger or equal 10000
and when the file descriptor is larger or equal 10 the desciption
will have at least 17 characters.
If a log message contains a magic character, everything after that marker
will be considered "extended" information. Log targets default to showing
this extended information but qb_log_ctl() may be used to indicate that
the extended information should be stripped instead.
The use case is a log message that has some information of value to administrators
and other information that is only of value for debugging. Two targets can be
configured such that the administrator log is not cluttered with debug information,
but a separate debug log is still available with complete information.
The magic character is the bell (\a). If someone wants to log an actual bell
character, they can do so in the extended information or via a %c format
(only the first bell is magic and only in the format string).
libqb fails to build on the hppa architecture, because the built-in
testcases fail as can be seen here:
http://buildd.debian-ports.org/status/fetch.php?pkg=libqb&arch=hppa&ver=0.17.0-2&stamp=1409458262
I did analyzed why they fail, and the reason is that on hppa we have
somewhat more complicated requirements (e.g. alignments) which needs to
be followed in order to mmap shared pages between processes. It's
different than what can be done compared to ia64 and sparc.
The attached patch fixes libqb on the hppa architecture and with it all
testcases finish successful.
By the way, I fixed a small typo in configure.ac too where arm platforms
prints "ia64"...
Forwarded-From: https://bugs.debian.org/cgi-bin/bugreport.cgi?bug=760199
Signed-off-by: Christoph Berg <myon@debian.org>
In qb_ipcs_us_connect 4 files are created and bound.
I dont' know how this works for QB_LINUX or QB_CYGWIN.
But for the other OS the files are created and must be unlinked.
I use the same logic to construct the file names and unlink the files.
qb_ipcc_us_connect calls this
...
res = qb_ipc_dgram_sock_connect(r->response, "response", "request",
r->max_msg_size, &c->request.u.us.sock);
and qb_ipc_dgram_sock_connect calls
..
set_sock_addr
and in set_sock_addr the files are created if not Linux or Cygwin.
...
#if defined(QB_LINUX) || defined(QB_CYGWIN)
snprintf(address->sun_path + 1, UNIX_PATH_MAX - 1, "%s", socket_name);
#else
snprintf(address->sun_path, sizeof(address->sun_path), "%s/%s", SOCKETDIR,
socket_name);
#endif
...
If multiple epoll sources generate events simultaneously, it is possible
for more jobs to be added to particular priority level than will be
handled in one go by qb_loop_run_level(). If one of these epoll sources
gains a new event (say, by switching from "readable" to "readable and
writeable"), then this new event would be missed.
To fix this, merge new epoll events into revents regardless of whether
the job is on the joblist. When the job is dispatched, revents will be
cleared, or the epoll source will be deleted entirely.
qb_rb_open() expects the size field to represent the maximum chunk size.
It adds QB_RB_CHUNK_MARGIN + 1 and rounds up to the page size to
determine the ringbuffer's total size. When creating a ringbuffer from a
file we must compensate by subtracting this amount from the file's size.
