[PATCH] mm: add a nopanic option for low bootmem

[PATCH] mm: add a nopanic option for low bootmem

[Jon Mason]

This patch adds a no panic option for low bootmem allocs.  This will
allow for a more graceful handling of "out of memory" for those
callers who wish to handle it.

Thanks,
Jon

Signed-off-by: Jon Mason <jdmason@us.ibm.com>

diff -r 307c5ac10c26 include/linux/bootmem.h
--- a/include/linux/bootmem.h	Sat Apr 22 12:21:45 2006
+++ b/include/linux/bootmem.h	Mon Apr 24 09:59:50 2006
@@ -46,6 +46,7 @@
 extern void __init free_bootmem (unsigned long addr, unsigned long size);
 extern void * __init __alloc_bootmem (unsigned long size, unsigned long align, unsigned long goal);
 extern void * __init __alloc_bootmem_nopanic (unsigned long size, unsigned long align, unsigned long goal);
+extern void * __init __alloc_bootmem_low_nopanic(unsigned long size, unsigned long align, unsigned long goal);
 extern void * __init __alloc_bootmem_low(unsigned long size,
 					 unsigned long align,
 					 unsigned long goal);
diff -r 307c5ac10c26 mm/bootmem.c
--- a/mm/bootmem.c	Sat Apr 22 12:21:45 2006
+++ b/mm/bootmem.c	Mon Apr 24 09:59:50 2006
@@ -463,3 +463,16 @@
 {
 	return __alloc_bootmem_core(pgdat->bdata, size, align, goal, LOW32LIMIT);
 }
+
+void * __init __alloc_bootmem_low_nopanic(unsigned long size, 
+					unsigned long align, unsigned long goal)
+{
+	bootmem_data_t *bdata;
+	void *ptr;
+
+	list_for_each_entry(bdata, &bdata_list, list)
+		if ((ptr = __alloc_bootmem_core(bdata, size, align, goal, 
+						LOW32LIMIT)))
+			return(ptr);
+	return NULL;
+}

Re: [PATCH] mm: add a nopanic option for low bootmem

[Andi Kleen]

On Monday 24 April 2006 23:44, Jon Mason wrote:
> This patch adds a no panic option for low bootmem allocs.  This will
> allow for a more graceful handling of "out of memory" for those
> callers who wish to handle it.

What's the user of it? Normally we don't merge such changes without
an user.

-Andi

Re: [PATCH] mm: add a nopanic option for low bootmem

[Muli Ben-Yehuda]

On Tue, Apr 25, 2006 at 12:43:38AM +0200, Andi Kleen wrote:
> On Monday 24 April 2006 23:44, Jon Mason wrote:
> > This patch adds a no panic option for low bootmem allocs.  This will
> > allow for a more graceful handling of "out of memory" for those
> > callers who wish to handle it.
> 
> What's the user of it? Normally we don't merge such changes without
> an user.

Calgary and swiotlb can use it to gracefully handle out of memory when
they allocate their aperture (swiotlb) or translation tables (Calgary)
on initialization. Either can fall back to no-iommu if initialization
fails so there's really no need to panic.

Jon, you should probably resubmit this with your swiotlb patches
making use of it.

Cheers,
Muli
-- 
Muli Ben-Yehuda
http://www.mulix.org | http://mulix.livejournal.com/

Re: [PATCH] mm: add a nopanic option for low bootmem

[Jon Mason]

On Tue, Apr 25, 2006 at 12:43:38AM +0200, Andi Kleen wrote:
> On Monday 24 April 2006 23:44, Jon Mason wrote:
> > This patch adds a no panic option for low bootmem allocs.  This will
> > allow for a more graceful handling of "out of memory" for those
> > callers who wish to handle it.
> 
> What's the user of it? Normally we don't merge such changes without
> an user.

Sorry.  Per your suggestion in the review of our Calgary IOMMU code, I
tried to use the alloc_bootmem_nopanic that you recently added.
Unfortunately, it needs low mem for our translation tables, so we needed
a new function to do this.  

Aside from Calgary, I have updated swiotlb to take advantage of this new
function (amongst other clean-ups and refinements).  I'll push those
swiotlb patches out today.

Thanks,
Jon

> 
> -Andi

Re: [PATCH] mm: add a nopanic option for low bootmem

[Andi Kleen]

On Tuesday 25 April 2006 16:08, Jon Mason wrote:
> On Tue, Apr 25, 2006 at 12:43:38AM +0200, Andi Kleen wrote:
> > On Monday 24 April 2006 23:44, Jon Mason wrote:
> > > This patch adds a no panic option for low bootmem allocs.  This will
> > > allow for a more graceful handling of "out of memory" for those
> > > callers who wish to handle it.
> > 
> > What's the user of it? Normally we don't merge such changes without
> > an user.
> 
> Sorry.  Per your suggestion in the review of our Calgary IOMMU code, I
> tried to use the alloc_bootmem_nopanic that you recently added.
> Unfortunately, it needs low mem for our translation tables, so we needed
> a new function to do this.  
> 
> Aside from Calgary, I have updated swiotlb to take advantage of this new
> function (amongst other clean-ups and refinements).  I'll push those
> swiotlb patches out today.


I will merge it together with the Calgary code then.
-Andi