[Xenomai-core] [BUG] commit #1833 breaks rt_queue_create

[Xenomai-core] [BUG] commit #1833 breaks rt_queue_create

[Jan Kiszka]

[-- Attachment #1.1: Type: text/plain, Size: 455 bytes --]

As xnheap_size() changed, a sanity check in xnheap_mmap fails now, at
least for SVN head, maybe also 2.2.x.

I would suggest the following patch to fix this, though I'm not 100%
sure its correct in all cases (do we ever come across a heap of multiple
extents in xnheap_mmap?).

Jan


PS: I'll spent a beer (or two...) for the one who starts writing a
testsuite for the native skin - in user-space, please, to catch also
issues like this one.

[-- Attachment #1.2: fix-mapped-heap-size.patch --]
[-- Type: application/octet-stream, Size: 744 bytes --]

[-- Attachment #2: OpenPGP digital signature --]
[-- Type: application/pgp-signature, Size: 250 bytes --]

Re: [Xenomai-core] [BUG] commit #1833 breaks rt_queue_create

[Jan Kiszka]

[-- Attachment #1: Type: text/plain, Size: 313 bytes --]

Jan Kiszka wrote:
> As xnheap_size() changed, a sanity check in xnheap_mmap fails now, at
> least for SVN head, maybe also 2.2.x.
> 
> I would suggest the following patch to fix this,

Wasn't someone recently writing in the wiki, one should post plain-text
patches here...?

Hope this one is less octet-streamed.

[-- Attachment #2: fix-mapped-heap-size.patch --]
[-- Type: text/plain, Size: 736 bytes --]

Index: ksrc/nucleus/heap.c
===================================================================
--- ksrc/nucleus/heap.c	(Revision 1835)
+++ ksrc/nucleus/heap.c	(Arbeitskopie)
@@ -952,7 +952,7 @@ static int xnheap_mmap(struct file *file
 	size = vma->vm_end - vma->vm_start;
 	heap = (xnheap_t *)file->private_data;
 
-	if (size != heap->extentsize)
+	if (size != xnheap_size(heap))
 		return -ENXIO;	/* Doesn't match the heap size. */
 
 	vma->vm_ops = &xnheap_vmops;
@@ -1135,7 +1135,7 @@ int xnheap_destroy_mapped(xnheap_t *heap
 	xnlock_put_irqrestore(&nklock, s);
 
 	__unreserve_and_free_heap(heap->archdep.heapbase,
-				  heap->extentsize, heap->archdep.kmflags);
+				  xnheap_size(heap), heap->archdep.kmflags);
 	return 0;
 }
 

Re: [Xenomai-core] [BUG] commit #1833 breaks rt_queue_create

[Philippe Gerum]

On Wed, 2006-11-15 at 22:13 +0100, Jan Kiszka wrote:
> As xnheap_size() changed, a sanity check in xnheap_mmap fails now, at
> least for SVN head, maybe also 2.2.x.
> 
> I would suggest the following patch to fix this, though I'm not 100%
> sure its correct in all cases

It's correct. xnheap_size() is sent back to user-space when creating the
queue, which forwards this in turn to the mmap handler.

>  (do we ever come across a heap of multiple
> extents in xnheap_mmap?).

Yes we do, particularly for ppc which needs this for the system heap.
However, we should not allow mmaping multi-extent heaps. Merged fixes
and added sanity check to prevent the latter issue. Thanks.

> 
> Jan
> 
> 
> PS: I'll spent a beer (or two...) for the one who starts writing a
> testsuite for the native skin - in user-space, please, to catch also
> issues like this one.
> _______________________________________________
> Xenomai-core mailing list
> Xenomai-core@domain.hid
> https://mail.gna.org/listinfo/xenomai-core
-- 
Philippe.