[RTEMS Project] #1367: Malloc statistics wrong in R4.10

[RTEMS trac]

#1367: Malloc statistics wrong in R4.10
--------------------+------------------------------
 Reporter:  norume  |       Owner:  sebastian.huber
     Type:  defect  |      Status:  assigned
 Priority:  normal  |   Milestone:  4.11
Component:  cpukit  |     Version:  4.10
 Severity:  normal  |  Resolution:
 Keywords:          |
--------------------+------------------------------
Changes (by sebastian.huber):

 * milestone:  4.10 => 4.11


Old description:

> The malloc statistics space_available value is not being set properly on
> startup.
> Also the lifetime_allocated and lifetime_freed values are not being
> updated with the results of executive allocations when a unified pool is
> used.
> =============
> Here are some of Joel's comments from previous e-mails discussing this
> problem
>
> length is the parameter passed in.  So I am thinking there should
> be a "protected heap get heap size" routine which gives the
> full amount of memory.  This way it would work either way.
>
> Executive memory allocation requests go directly to the heap so the
> statistics are just counting
> malloc/free.  I am not even sure memalign() is getting counted
> at the moment.
>
> The statistics are at the malloc level not at the heap level
> so the workspace heap accesses bypass it.
>
> We would have to move the malloc statistics to the supercore
> heap to get this capability.

New description:

 The malloc statistics space_available value is not being set properly on
 startup.
 Also the lifetime_allocated and lifetime_freed values are not being
 updated with the results of executive allocations when a unified pool is
 used.
 =============
 Here are some of Joel's comments from previous e-mails discussing this
 problem

 length is the parameter passed in.  So I am thinking there should
 be a "protected heap get heap size" routine which gives the
 full amount of memory.  This way it would work either way.

 Executive memory allocation requests go directly to the heap so the
 statistics are just counting
 malloc/free.  I am not even sure memalign() is getting counted
 at the moment.

 The statistics are at the malloc level not at the heap level
 so the workspace heap accesses bypass it.

 We would have to move the malloc statistics to the supercore
 heap to get this capability.

--

Comment:

 The malloc statistics should get removed.  Instead the general heap
 handler statistics should be used and if necessary enhanced.

--
Ticket URL: <http://devel.rtems.org/ticket/1367#comment:4>
RTEMS Project <http://www.rtems.org/>
RTEMS Project