.. since a failed allocation is being (initially) handled gracefully,
and panic()-ed upon failure explicitly in the function if retries with
smaller sizes failed.
Signed-off-by: Jan Beulich <[email protected]>
Cc: David Howells <[email protected]>
---
include/linux/bootmem.h | 4 ++++
mm/page_alloc.c | 2 +-
2 files changed, 5 insertions(+), 1 deletion(-)
--- linux-2.6.27-rc1/include/linux/bootmem.h 2008-07-31 17:08:35.000000000 +0200
+++ 2.6.27-rc1-alloc-large-system-hash/include/linux/bootmem.h 2008-07-31 17:56:18.000000000 +0200
@@ -97,10 +97,14 @@ extern void *__alloc_bootmem_low_node(pg
#ifndef CONFIG_HAVE_ARCH_BOOTMEM_NODE
#define alloc_bootmem(x) \
__alloc_bootmem(x, SMP_CACHE_BYTES, __pa(MAX_DMA_ADDRESS))
+#define alloc_bootmem_nopanic(x) \
+ __alloc_bootmem_nopanic(x, SMP_CACHE_BYTES, __pa(MAX_DMA_ADDRESS))
#define alloc_bootmem_low(x) \
__alloc_bootmem_low(x, SMP_CACHE_BYTES, 0)
#define alloc_bootmem_pages(x) \
__alloc_bootmem(x, PAGE_SIZE, __pa(MAX_DMA_ADDRESS))
+#define alloc_bootmem_pages_nopanic(x) \
+ __alloc_bootmem_nopanic(x, PAGE_SIZE, __pa(MAX_DMA_ADDRESS))
#define alloc_bootmem_low_pages(x) \
__alloc_bootmem_low(x, PAGE_SIZE, 0)
#define alloc_bootmem_node(pgdat, x) \
--- linux-2.6.27-rc1/mm/page_alloc.c 2008-07-31 17:08:36.000000000 +0200
+++ 2.6.27-rc1-alloc-large-system-hash/mm/page_alloc.c 2008-07-31 17:56:37.000000000 +0200
@@ -4454,7 +4454,7 @@ void *__init alloc_large_system_hash(con
do {
size = bucketsize << log2qty;
if (flags & HASH_EARLY)
- table = alloc_bootmem(size);
+ table = alloc_bootmem_nopanic(size);
else if (hashdist)
table = __vmalloc(size, GFP_ATOMIC, PAGE_KERNEL);
else {
Commit 74768ed833344b "page allocator: use no-panic variant of
alloc_bootmem() in alloc_large_system_hash()" introduced two new
_nopanic macros which are undefined for CONFIG_HAVE_ARCH_BOOTMEM_NODE.
Signed-off-by: Johannes Weiner <[email protected]>
---
These macros are a mess... :(
diff --git a/include/asm-x86/mmzone_32.h b/include/asm-x86/mmzone_32.h
index b2298a2..5862e64 100644
--- a/include/asm-x86/mmzone_32.h
+++ b/include/asm-x86/mmzone_32.h
@@ -97,10 +97,16 @@ static inline int pfn_valid(int pfn)
reserve_bootmem_node(NODE_DATA(0), (addr), (size), (flags))
#define alloc_bootmem(x) \
__alloc_bootmem_node(NODE_DATA(0), (x), SMP_CACHE_BYTES, __pa(MAX_DMA_ADDRESS))
+#define alloc_bootmem_nopanic(x) \
+ __alloc_bootmem_node_nopanic(NODE_DATA(0), (x), SMP_CACHE_BYTES, \
+ __pa(MAX_DMA_ADDRESS))
#define alloc_bootmem_low(x) \
__alloc_bootmem_node(NODE_DATA(0), (x), SMP_CACHE_BYTES, 0)
#define alloc_bootmem_pages(x) \
__alloc_bootmem_node(NODE_DATA(0), (x), PAGE_SIZE, __pa(MAX_DMA_ADDRESS))
+#define alloc_bootmem_pages_nopanic(x) \
+ __alloc_bootmem_node_nopanic(NODE_DATA(0), (x), PAGE_SIZE, \
+ __pa(MAX_DMA_ADDRESS))
#define alloc_bootmem_low_pages(x) \
__alloc_bootmem_node(NODE_DATA(0), (x), PAGE_SIZE, 0)
#define alloc_bootmem_node(pgdat, x) \
Sorry for not noticing this:
Acked-by: Jan Beulich <[email protected]>
>>> Johannes Weiner <[email protected]> 13.08.08 10:17 >>>
Commit 74768ed833344b "page allocator: use no-panic variant of
alloc_bootmem() in alloc_large_system_hash()" introduced two new
_nopanic macros which are undefined for CONFIG_HAVE_ARCH_BOOTMEM_NODE.
Signed-off-by: Johannes Weiner <[email protected]>
---
These macros are a mess... :(
diff --git a/include/asm-x86/mmzone_32.h b/include/asm-x86/mmzone_32.h
index b2298a2..5862e64 100644
--- a/include/asm-x86/mmzone_32.h
+++ b/include/asm-x86/mmzone_32.h
@@ -97,10 +97,16 @@ static inline int pfn_valid(int pfn)
reserve_bootmem_node(NODE_DATA(0), (addr), (size), (flags))
#define alloc_bootmem(x) \
__alloc_bootmem_node(NODE_DATA(0), (x), SMP_CACHE_BYTES, __pa(MAX_DMA_ADDRESS))
+#define alloc_bootmem_nopanic(x) \
+ __alloc_bootmem_node_nopanic(NODE_DATA(0), (x), SMP_CACHE_BYTES, \
+ __pa(MAX_DMA_ADDRESS))
#define alloc_bootmem_low(x) \
__alloc_bootmem_node(NODE_DATA(0), (x), SMP_CACHE_BYTES, 0)
#define alloc_bootmem_pages(x) \
__alloc_bootmem_node(NODE_DATA(0), (x), PAGE_SIZE, __pa(MAX_DMA_ADDRESS))
+#define alloc_bootmem_pages_nopanic(x) \
+ __alloc_bootmem_node_nopanic(NODE_DATA(0), (x), PAGE_SIZE, \
+ __pa(MAX_DMA_ADDRESS))
#define alloc_bootmem_low_pages(x) \
__alloc_bootmem_node(NODE_DATA(0), (x), PAGE_SIZE, 0)
#define alloc_bootmem_node(pgdat, x) \
* Johannes Weiner <[email protected]> wrote:
> Commit 74768ed833344b "page allocator: use no-panic variant of
> alloc_bootmem() in alloc_large_system_hash()" introduced two new
> _nopanic macros which are undefined for CONFIG_HAVE_ARCH_BOOTMEM_NODE.
the above bootmem commit from yesterday broke the -rc3 32-bit NUMA
build:
mm/page_alloc.c: In function ‘alloc_large_system_hash':
mm/page_alloc.c:4440: error: implicit declaration of function ‘alloc_bootmem_nopanic'
mm/page_alloc.c:4440: warning: assignment makes pointer from integer without a cast
i have queued your fix up in tip/x86/urgent, thanks Johannes.
Ingo
* Jan Beulich <[email protected]> wrote:
> Sorry for not noticing this:
> Acked-by: Jan Beulich <[email protected]>
>
> >>> Johannes Weiner <[email protected]> 13.08.08 10:17 >>>
> Commit 74768ed833344b "page allocator: use no-panic variant of
> alloc_bootmem() in alloc_large_system_hash()" introduced two new
> _nopanic macros which are undefined for CONFIG_HAVE_ARCH_BOOTMEM_NODE.
>
> Signed-off-by: Johannes Weiner <[email protected]>
btw., i dont think the fix is complete, a quick grep shows:
arch/avr32/Kconfig:config HAVE_ARCH_BOOTMEM_NODE
Ingo
Hi Ingo,
Ingo Molnar <[email protected]> writes:
> * Jan Beulich <[email protected]> wrote:
>
>> Sorry for not noticing this:
>> Acked-by: Jan Beulich <[email protected]>
>>
>> >>> Johannes Weiner <[email protected]> 13.08.08 10:17 >>>
>> Commit 74768ed833344b "page allocator: use no-panic variant of
>> alloc_bootmem() in alloc_large_system_hash()" introduced two new
>> _nopanic macros which are undefined for CONFIG_HAVE_ARCH_BOOTMEM_NODE.
>>
>> Signed-off-by: Johannes Weiner <[email protected]>
>
> btw., i dont think the fix is complete, a quick grep shows:
>
> arch/avr32/Kconfig:config HAVE_ARCH_BOOTMEM_NODE
Uhm, I think a fix for that would be removing the knob from the avr32
Kconfig. It defaults to n and I couldn't find a place where this gets
enabled.
Furthermore, there doesn't seem to be any arch-specific code associated
with it (grep -R ARCH_BOOTMEM_NODE arch/avr32).
Hannes
* Johannes Weiner <[email protected]> wrote:
> Hi Ingo,
>
> Ingo Molnar <[email protected]> writes:
>
> > * Jan Beulich <[email protected]> wrote:
> >
> >> Sorry for not noticing this:
> >> Acked-by: Jan Beulich <[email protected]>
> >>
> >> >>> Johannes Weiner <[email protected]> 13.08.08 10:17 >>>
> >> Commit 74768ed833344b "page allocator: use no-panic variant of
> >> alloc_bootmem() in alloc_large_system_hash()" introduced two new
> >> _nopanic macros which are undefined for CONFIG_HAVE_ARCH_BOOTMEM_NODE.
> >>
> >> Signed-off-by: Johannes Weiner <[email protected]>
> >
> > btw., i dont think the fix is complete, a quick grep shows:
> >
> > arch/avr32/Kconfig:config HAVE_ARCH_BOOTMEM_NODE
>
> Uhm, I think a fix for that would be removing the knob from the avr32
> Kconfig. It defaults to n and I couldn't find a place where this gets
> enabled.
>
> Furthermore, there doesn't seem to be any arch-specific code
> associated with it (grep -R ARCH_BOOTMEM_NODE arch/avr32).
ok, indeed. I dont think it should be removed: it makes sense for an
architecture to list all Kconfig knobs that _might_ matter - even if
they are disabled for an architecture.
That makes it easier to know whether a Kconfig knob that is disables is
disabled because the arch wants it disabled, because the arch wants it
enabled, or because it's a knew knob the arch does not know about yet.
So it's 3 states really.
Ingo