GitLab

We don't have any interrupt pipeline implementation for 3.x+ kernels
on NIOS2 currently, which is a requirement for running Xenomai 3, so
there is no point in keeping this architecture code in-tree.

We don't have any interrupt pipeline implementation for 3.x+ kernels
on SH4, currently which is a requirement for running Xenomai 3, so
there is no point in keeping this architecture code in-tree.

Signed-off-by: Jorge Ramirez-Ortiz <jorge.ramirez-ortiz@linaro.org>

Signed-off-by: Jorge Ramirez-Ortiz <jorge.ramirez-ortiz@linaro.org>

Signed-off-by: Jorge Ramirez-Ortiz <jorge.ramirez-ortiz@linaro.org>

The value passed to --mem-pool-size used to represent kilobytes, which
is at odds with the byte unit required by the mem_pool_size tunable.

--mem-pool-size now accepts a unit K|M|G suffix to values, defaulting
to bytes. For backward compatibility, any non-suffixed value below
65536 is interpreted as a count of kilobytes, not bytes,
e.g. --mem-pool-size=8192 will create a 8M heap for the
application. This conversion is noisy, to call for a parameter update
asap, unless --no-sanity is in effect.

If you need a tiny heap size below 64k, pass a kilobyte value with a 'K'
suffix.

We don't want the finalizer to run for the underlying sema4 object
when cleaning up upon error at creation. Call the uninit helper
instead.

We don't want the finalizer to run for the underlying event object
when cleaning up upon error at creation. Call the uninit helper
instead.

This one does not run the object finalizer, assuming a strictly
passive object (no waiters yet).

This one does not run the object finalizer, assuming a strictly
passive object (no waiters yet).

Allow users who belong to the same group to access the shared heap,
provided umask allows it too.

gdb injects software breakpoints as undefined instructions. Make sure we
detect them as well.
Signed-off-by: Jan Kiszka <jan.kiszka@siemens.com>

Trace and print the name of the relaxation reason whenever available.
Signed-off-by: Jan Kiszka <jan.kiszka@siemens.com>

We may want to start/stop a TP schedule without removing it from the
CPU entirely. To this end, this commit introduces the
sched_tp_start/stop requests, processed by sched_config_np().

sem_wait() must not pull any waiter from the sleep queue when undoing
do_trywait() upon XNBREAK.

At this chance, fold a few routines to avoid unnecessary hops,
improving readability too.

This makes the obscure rt_print_cleanup() call useless, and therefore
removed from the API.