GitLab

This enables threads running out-of-band to retrieve most of the raw
thread-related state information "evl ps" accesses through /sysfs
attributes.
Signed-off-by: Philippe Gerum <rpm@xenomai.org>

Signed-off-by: Philippe Gerum <rpm@xenomai.org>

Signed-off-by: Philippe Gerum <rpm@xenomai.org>

SCHED_FIFO is the most critical and frequently used scheduling policy
with EVL, so there is a net gain in inlining the small helpers
manipulating the thread list for this one.
Signed-off-by: Philippe Gerum <rpm@xenomai.org>

Signed-off-by: Philippe Gerum <rpm@xenomai.org>

A practical LART is to know beforehand when some out-of-band work is
about to run on a non-isolated CPU, so that the application may warn
the user about the potentially higher latency figures induced by
higher rates of cache and TLB misses which may be caused by heavy
in-band load running on the same core.

The control device now accepts the
[oob_]ioctl(EVL_CTLIOC_GET_CPUSTATE) request, which queries the
current state of a CPU:

- EVL_CPU_ISOL if it does not belong to housekeeping set of the
  in-band kernel, i.e. not mentioned in isolcpus= for the scheduling
  domain.

- EVL_CPU_OOB if it is part of the out-of-band set EVL manages.

- EVL_CPU_OFFLINE if currently off.
Signed-off-by: Philippe Gerum <rpm@xenomai.org>

The CPU the sampler runs on must be part of the out-of-band set.
Signed-off-by: Philippe Gerum <rpm@xenomai.org>

Signed-off-by: Philippe Gerum <rpm@xenomai.org>

Signed-off-by: Philippe Gerum <rpm@xenomai.org>

Although writing to a proxy is primarily intended for oob callers,
allowing inband tasks to use this channel too comes in handy as the
sequence of messages relayed by such proxy can be preserved regardless
of the execution stage they originate from, as they percolate through
the same buffer.

Inband writers can block until the output has drained, unless
O_NONBLOCK is set for the proxy.
Signed-off-by: Philippe Gerum <rpm@xenomai.org>

clock_sleep() only accepts absolute timespecs, so we have no use of a
pointer for collecting the remaining sleep time upon interrupt.
Signed-off-by: Philippe Gerum <rpm@xenomai.org>

Connecting a proxy with some special file may impose fixed-size writes
to the latter, e.g. an eventfd would require that 64bit values only be
written. These changes enable the application to set a fixed
granularity the proxy must abide by when writing to the target file,
so that such requirement is honored.

In addition:

- oob_write() may block until enough space is available in the
  transfer buffer to complete the sending, unless O_NONBLOCK is set on
  the proxy file.

- POLLOUT|POLLWRNORM can be monitored for waiting for the output to
drain.
Signed-off-by: Philippe Gerum <rpm@xenomai.org>

Concurrent reads or writes to an xbuf causing nested buffer space
reservations could end up in a miscalculation of the number of bytes
remaining to be read or written.
Signed-off-by: Philippe Gerum <rpm@xenomai.org>

Signed-off-by: Philippe Gerum <rpm@xenomai.org>

Signed-off-by: Philippe Gerum <rpm@xenomai.org>

Signed-off-by: Philippe Gerum <rpm@xenomai.org>

With this rework, we have two separate per-thread mode flags for
greater flexibility aimed at controlling the debug features on a
per-thread basis:

- T_WOSS can be set to trigger SIGDEBUG upon (unexpected) stage switch
  to in-band mode. This is strictly equivalent to the obsoleted T_WARN
  bit.

- T_WOLI enables/disables the detection of locking inconsistencies
  with mutexes via the EVL_THRIOC_{SET, CLEAR}_MODE interface. This
  combines the former static CONFIG_EVL_DEBUG_MUTEX_INBAND and
  CONFIG_EVL_DEBUG_MUTEX_SLEEP options.

Enabling CONFIG_EVL_DEBUG_WOLI turns on T_WOLI by default for every
new EVL thread running in userland, which can be opted out on a
per-thread basis using EVL_THRIOC_CLEAR_MODE.
Signed-off-by: Philippe Gerum <rpm@xenomai.org>

We want the policy to appear clearly in the name.
Signed-off-by: Philippe Gerum <rpm@xenomai.org>

The new hierachy of priority scales is as follows:

EVL_CORE_MIN_PRIO = EVL_WEAK_MIN_PRIO
...
	EVL_FIFO_MIN_PRIO ==
	     EVL_QUOTA_MIN_PRIO ==
	          EVL_TP_MIN_PRIO (== 1)
	...
	EVL_FIFO_MAX_PRIO =
	     EVL_QUOTA_MAX_PRIO ==
	          EVL_TP_MAX_PRIO ==
		       EVL_WEAK_MAX_PRIO (< MAX_USER_RT_PRIO)
...
EVL_CORE_MAX_PRIO (> MAX_RT_PRIO)

We reserve a couple of priority levels above the highest inband
kthread priority (MAX_RT_PRIO..MAX_RT_PRIO+1), which are guaranteed to
be higher than the highest inband user task priority
(MAX_USER_RT_PRIO-1) we use for SCHED_FIFO. Those extra levels can be
used for EVL kthreads which must top the priority of any userland
thread.

SCHED_EVL was dropped int the process, since userland is now
constrained to EVL_FIFO_MAX_PRIO by construction.
Signed-off-by: Philippe Gerum <rpm@xenomai.org>

Add missing policy accessors and init bits.
Signed-off-by: Philippe Gerum <rpm@xenomai.org>

When arming T_INBAND for the stage switching thread, we have to remove
it from the runqueue in the same move if T_READY is present in its
state flags.

Failing to do so creates a race with another CPU readying that thread
by calling evl_release_thread(), which leads to an inconsistent
scheduler state with both T_INBAND and T_READY set for the
thread. When this happens, evl_switch_inband() may pick the switching
thread from the runqueue for out-of-band scheduling in
__evl_schedule() despite being formally blocked by T_INBAND, instead
of waiting for the inband scheduler to do so for completing the
transition to inband context.

As a result, dovetail_resume_inband() spuriously runs from the
out-of-band stage eventually (caught by CONFIG_DEBUG_DOVETAIL), which
leads to a galactic mess.
Signed-off-by: Philippe Gerum <rpm@xenomai.org>

We create devices on the fly as part of the procedure for
instantiating new EVL elements, expecting them to appear dynamically
in the /dev filesystem hierarchy.
Signed-off-by: Philippe Gerum <rpm@xenomai.org>

Signed-off-by: Philippe Gerum <rpm@xenomai.org>

Signed-off-by: Philippe Gerum <rpm@xenomai.org>

Per-factory device types are introduced to control the ownership of
element devices. Those are created under the /dev/evl hierachy upon
receipt of the EVL_IOC_CLONE request by the clone device of their
parent factory.

With this change, the ownership of a clone device is inherited by all
element devices it instantiates (threads, xbuf, monitors and so
on). This is useful for enabling EVL services for non-privileged
users, which should only require to set the ownership and permissions
of the control and clone devices appropriately.

Caveat: for inherited non-default ownership to stick in presence of
udev/mdev, make sure to define a rule which prevents the default
root.root to be applied for these devices.
Signed-off-by: Philippe Gerum <rpm@xenomai.org>

Signed-off-by: Philippe Gerum <rpm@xenomai.org>

Signed-off-by: Philippe Gerum <rpm@xenomai.org>

Signed-off-by: Philippe Gerum <rpm@xenomai.org>

Signed-off-by: Philippe Gerum <rpm@xenomai.org>

Signed-off-by: Philippe Gerum <rpm@xenomai.org>

Requires a couple of ABI changes.
Signed-off-by: Philippe Gerum <rpm@xenomai.org>

Signed-off-by: Philippe Gerum <rpm@xenomai.org>

Signed-off-by: Philippe Gerum <rpm@xenomai.org>

Callers of the out-of-band polling interface are subject to spurious
wakeups by design, which means that they have to handle the case where
a signaled condition goes stale before execution resumes. On the other
hand, we certainly don't want to miss a wake up event in case the
condition is met again in the meantime (i.e. transient
true->false->true event), therefore making provisions for clearing
such event is asking for trouble.

Since the caller has to check for spurious wake ups, we should wake it
up as soon as a false->true transition is observed. Therefore,
evl_clear_poll_events() is deemed useless and even error-prone: drop
it.
Signed-off-by: Philippe Gerum <rpm@xenomai.org>

Signed-off-by: Philippe Gerum <rpm@xenomai.org>

Make sure the per-event gate_offset and gate pointer are reset
properly when the event is removed from the per-gate event tracking
queue.

This happens in two occasions:

- when waiters are unblocked on gate unlock because the event was
  signaled.

- upon return of an aborted wait operation
  (i.e. T_BREAK|T_TIMEO|T_RMID).
Signed-off-by: Philippe Gerum <rpm@xenomai.org>

Currently concerns SYS_NICE, IPC_LOCK and SYS_RAWIO. Capabilities
which have been granted are dropped when the thread detaches from the
core.
Signed-off-by: Philippe Gerum <rpm@xenomai.org>

We don't plan for switching CPU from out-of-band context anymore, so
T_MOVED is basically pointless now.
Signed-off-by: Philippe Gerum <rpm@xenomai.org>

The inband kernel does not attempt to restart an interrupted syscall
upon signal delivery, we should not either. Since signal delivery
requires switching inband, clear T_SYSRST in evl_switch_inband() to
prevent this.

This is a port of a Cobalt fix for the same issue by Jan Kiszka
<jan.kiszka@siemens.com>.
Signed-off-by: Philippe Gerum <rpm@xenomai.org>

This protocol operates the monitor value as a set of boolean event
flags forming a 32bit-wide event group. All bits from a group are
initially set to zero.

An event flag group is a lightweight notification mechanism. The
application can send bitmasks to raise individual bits from the group
(i.e. group_value |= bits), or wait for the group to have at least one
bit set for satisfying the request. In the latter case, the group
value is read then cleared atomically, and the collected bits are
returned to the thread heading the wait queue.
Signed-off-by: Philippe Gerum <rpm@xenomai.org>