1
0
mirror of git://projects.qi-hardware.com/wernermisc.git synced 2024-11-15 06:29:22 +02:00

m1/patches/rtems/: pending RTEMS patches (relative to latest CVS)

This commit is contained in:
Werner Almesberger 2011-11-12 09:22:48 -03:00
parent 20283fe9c2
commit aa0022d973
5 changed files with 226 additions and 0 deletions

View File

@ -0,0 +1,44 @@
This issue is under consideration:
http://www.rtems.org/pipermail/rtems-users/2011-November/009130.html
Doubly-linked lists ("chains") in RTEMS have a "control" block that
looks like the next/prev link pair in an element. The list elements
link both ways to this control block.
_Chain_Is_first and _Chain_Is_last only probed if the link to the
next element - which would be the control block - is non-NULL.
Telling by the function description and given that there are already
functions called _Chain_Is_head and _Chain_Is_tail (which could be
simplified), this is probably not the intended behaviour.
This also affects the aliases rtems_chain_is_first and
rtems_chain_is_last.
These functions are not used a lot and I haven't seen any immediate
effect on M1 after changing them, so I can't say whether this patch
may unearth other problems.
- Werner
Index: rtems/cpukit/score/inline/rtems/score/chain.inl
===================================================================
--- rtems.orig/cpukit/score/inline/rtems/score/chain.inl 2011-11-12 09:12:46.000000000 -0300
+++ rtems/cpukit/score/inline/rtems/score/chain.inl 2011-11-12 09:13:47.000000000 -0300
@@ -297,7 +297,7 @@
const Chain_Node *the_node
)
{
- return (the_node->previous == NULL);
+ return the_node->previous->previous == NULL;
}
/** @brief Is this the Last Node on the Chain
@@ -314,7 +314,7 @@
const Chain_Node *the_node
)
{
- return (the_node->next == NULL);
+ return the_node->next->next == NULL;
}
/** @brief Does this Chain have only One Node

View File

@ -0,0 +1,98 @@
This issue is under review:
https://www.rtems.org/bugzilla/show_bug.cgi?id=1961
If it's permissible to call rtems_message_queue_send from an
interrupt, then there is at least one race condition in the core
message subsystem.
This created the MIDI/mouse hang we love so much on M1.
The problem is as follows: RTEMS queues use pre-allocated message
buffers that are kept on an "inactive" (free) list. When enqueuing
a message, a buffer is first removed from the inactive list, data
it copied to it, and it is then added to the pending list.
The reverse happens when dequeuing. Besides these two queues, there
is also a counter called number_of_pending_messages keeping track,
as the name suggests, of the number of pending messages. It is
updated atomically together with changes to the pending buffers
list.
From the above it is clear that the counter will be out of sync with
the inactive list between the beginning and the end of an enqueue or
dequeue operation.
In order to minimize interrupt latency, RTEMS disables interrupts
only when adding and removing buffers from lists, but not throughout
the whole enqueuing/dequeuing operation. Instead, it disables the
scheduler during these operations, but this doesn't prevent
interrupts.
This means that the inconsistency between number_of_pending_messages
and the inactive list can be observed from an interrupt handler if
enqueuing or dequeuing is in progress.
_CORE_message_queue_Submit checks whether there is still room in the
queue by reading number_of_pending_messages. If there is room, it
then calls _CORE_message_queue_Allocate_message_buffer to obtain a
free buffer.
Given that number_of_pending_messages and the list of inactive
buffers can disagree, e.g., if _CORE_message_queue_Seize or another
_CORE_message_queue_Submit is executing concurrently,
_CORE_message_queue_Allocate_message_buffer may fail to obtain a
free buffer despite the prior test.
_CORE_message_queue_Allocate_message_buffer can detect a lack of
free buffers and indicates it by returning a NULL pointer. Checking
whether NULL has been returned instead of a buffer is optional and
depends on RTEMS_DEBUG.
If no check is performed, _CORE_message_queue_Submit will then try
to use the buffer. In the absence of hardware detecting the
de-referencing of NULL pointers, the wounded system will limp on a
little further until, at least in the case of M1, it finally hangs
somewhere.
The patch below avoids the problem in the scenario described above
by not using number_of_pending_messages as an indicator of whether
free buffers are available, but by simply trying to get a buffer,
and handling the result of failure.
This is similar to how _CORE_message_queue_Seize works.
Another possibility would be to make testing of the_message no
longer optional. But then, there would basically be two tests for
the same condition, which is ugly.
- Werner
Index: rtems/cpukit/score/src/coremsgsubmit.c
===================================================================
--- rtems.orig/cpukit/score/src/coremsgsubmit.c 2011-11-12 09:15:12.000000000 -0300
+++ rtems/cpukit/score/src/coremsgsubmit.c 2011-11-12 09:15:17.000000000 -0300
@@ -101,21 +101,9 @@
* No one waiting on the message queue at this time, so attempt to
* queue the message up for a future receive.
*/
- if ( the_message_queue->number_of_pending_messages <
- the_message_queue->maximum_pending_messages ) {
-
- the_message =
- _CORE_message_queue_Allocate_message_buffer( the_message_queue );
-
- #if defined(RTEMS_DEBUG)
- /*
- * NOTE: If the system is consistent, this error should never occur.
- */
-
- if ( !the_message )
- return CORE_MESSAGE_QUEUE_STATUS_UNSATISFIED;
- #endif
-
+ the_message =
+ _CORE_message_queue_Allocate_message_buffer( the_message_queue );
+ if ( the_message ) {
_CORE_message_queue_Copy_buffer(
buffer,
the_message->Contents.buffer,

View File

@ -0,0 +1,48 @@
This patch is under review:
https://www.rtems.org/bugzilla/show_bug.cgi?id=1956
The comments in cpukit/score/cpu/lm32/rtems/score/cpu.h state that
CPU_STACK_ALIGNMENT should either be 0 or >= CPU_ALIGNMENT. The
latter was 8, the former is 4.
Further investigation revealed that, contrary to what the comment
says, 0 is not a valid value for CPU_STACK_ALIGNMENT.
This patch sets CPU_ALIGNMENT to 4, since we don't have any machine
word larger than that. (doubles and long longs are handled by
software and either extremely slow already or rare.)
The patch also corrects the misleading comment before
CPU_STACK_ALIGNMENT.
I'm not sure if this fix has any real-life impact on M1 behaviour,
but I guess it can't hurt.
- Werner
Index: rtems/cpukit/score/cpu/lm32/rtems/score/cpu.h
===================================================================
--- rtems.orig/cpukit/score/cpu/lm32/rtems/score/cpu.h 2011-11-12 03:01:35.000000000 -0300
+++ rtems/cpukit/score/cpu/lm32/rtems/score/cpu.h 2011-11-12 03:03:46.000000000 -0300
@@ -637,7 +637,7 @@
*
* XXX document implementation including references if appropriate
*/
-#define CPU_ALIGNMENT 8
+#define CPU_ALIGNMENT 4
/**
* This number corresponds to the byte alignment requirement for the
@@ -687,9 +687,10 @@
* stack. This alignment requirement may be stricter than that for the
* data types alignment specified by @ref CPU_ALIGNMENT. If the
* @ref CPU_ALIGNMENT is strict enough for the stack, then this should be
- * set to 0.
+ * set to @ref CPU_ALIGNMENT.
*
- * @note This must be a power of 2 either 0 or greater than @ref CPU_ALIGNMENT.
+ * @note This must be a power of 2 either equal to or greater than
+ * @ref CPU_ALIGNMENT.
*
* Port Specific Information:
*

View File

@ -0,0 +1,32 @@
This issue is under consideration:
http://www.rtems.org/pipermail/rtems-users/2011-November/009097.html
There seem to be two issues in the original code:
- the "node" argument of the macro is not protected, which could lead
to very hard to find errors (this doesn't seem to cause any
immediate problems just now, but it's a lousy risk to take)
- more seriously, "offsetof" counts in bytes while arithmentic on the
"node" pointer counts in multiples of whatever size that object has
RTEMS with this patch applied runs well (on Milkymist), but I don't
know if the code in question actually is executed.
- Werner
Index: rtems/cpukit/score/include/rtems/score/rbtree.h
===================================================================
--- rtems.orig/cpukit/score/include/rtems/score/rbtree.h 2011-11-12 08:52:50.000000000 -0300
+++ rtems/cpukit/score/include/rtems/score/rbtree.h 2011-11-12 09:00:35.000000000 -0300
@@ -90,7 +90,9 @@
*
*/
#define _RBTree_Container_of(node,container_type, node_field_name) \
- ((container_type*) (node - offsetof(container_type,node_field_name)))
+ ((container_type*) ((void *) (node) - \
+ offsetof(container_type,node_field_name)))
+
/**
* This type indicates the direction.

4
m1/patches/rtems/series Normal file
View File

@ -0,0 +1,4 @@
lm32-stack-alignment.patch
rbtree-container-of.patch
chain-first-last.patch
coremsgsubmit-race.patch