2007-05-16 06:15:32

by Christoph Lameter

[permalink] [raw]
Subject: Slab allocators: Define common size limitations

Currently we have a maze of configuration variables that determine
the maximum slab size. Worst of all it seems to vary between SLAB and SLUB.

So define a common maximum size for kmalloc. For conveniences sake
we use the maximum size ever supported which is 32 MB. We limit the maximum
size to a lower limit if MAX_ORDER does not allow such large allocations.

For many architectures this patch will have the effect of adding large
kmalloc sizes. x86_64 adds 5 new kmalloc sizes. So a small amount
of memory will be needed for these caches (contemporary SLAB has dynamically
sizeable node and cpu structure so the waste is less than in the past)

Most architectures will then be able to allocate object with sizes up to
MAX_ORDER. We have had repeated breakage (in fact whenever we doubled the
number of supported processors) on IA64 because one or the other struct
grew beyond what the slab allocators supported. This will avoid future
issues and f.e. avoid fixes for 2k and 4k cpu support.

CONFIG_LARGE_ALLOCS is no longer necessary so drop it.

Signed-off-by: Christoph Lameter <[email protected]>

---
arch/blackfin/Kconfig | 8 --------
arch/frv/Kconfig | 8 --------
arch/m68knommu/Kconfig | 8 --------
arch/v850/Kconfig | 8 --------
include/linux/kmalloc_sizes.h | 20 +++++++++++++++-----
include/linux/slab.h | 15 +++++++++++++++
include/linux/slub_def.h | 19 ++-----------------
mm/slab.c | 19 ++-----------------
8 files changed, 34 insertions(+), 71 deletions(-)

Index: slub/include/linux/slab.h
===================================================================
--- slub.orig/include/linux/slab.h 2007-05-15 21:17:15.000000000 -0700
+++ slub/include/linux/slab.h 2007-05-15 21:19:51.000000000 -0700
@@ -74,6 +74,21 @@ static inline void *kmem_cache_alloc_nod
#endif

/*
+ * The largest kmalloc size supported by the slab allocators is
+ * 32 megabyte (2^25) or the maximum allocatable page order if that is
+ * less than 32 MB.
+ *
+ * WARNING: Its not easy to increase this value since the allocators have
+ * to do various tricks to work around compiler limitations in order to
+ * ensure proper constant folding.
+ */
+#define KMALLOC_SHIFT_HIGH ((MAX_ORDER + PAGE_SHIFT) <= 25 ? \
+ (MAX_ORDER + PAGE_SHIFT) : 25)
+
+#define KMALLOC_MAX_SIZE (1UL << KMALLOC_SHIFT_HIGH)
+#define KMALLOC_MAX_ORDER (KMALLOC_SHIFT_HIGH - PAGE_SHIFT)
+
+/*
* Common kmalloc functions provided by all allocators
*/
void *__kmalloc(size_t, gfp_t);
Index: slub/include/linux/slub_def.h
===================================================================
--- slub.orig/include/linux/slub_def.h 2007-05-15 21:17:15.000000000 -0700
+++ slub/include/linux/slub_def.h 2007-05-15 21:21:27.000000000 -0700
@@ -58,17 +58,6 @@ struct kmem_cache {
*/
#define KMALLOC_SHIFT_LOW 3

-#ifdef CONFIG_LARGE_ALLOCS
-#define KMALLOC_SHIFT_HIGH ((MAX_ORDER + PAGE_SHIFT) =< 25 ? \
- (MAX_ORDER + PAGE_SHIFT - 1) : 25)
-#else
-#if !defined(CONFIG_MMU) || NR_CPUS > 512 || MAX_NUMNODES > 256
-#define KMALLOC_SHIFT_HIGH 20
-#else
-#define KMALLOC_SHIFT_HIGH 18
-#endif
-#endif
-
/*
* We keep the general caches in an array of slab caches that are used for
* 2^x bytes of allocations.
@@ -79,7 +68,7 @@ extern struct kmem_cache kmalloc_caches[
* Sorry that the following has to be that ugly but some versions of GCC
* have trouble with constant propagation and loops.
*/
-static inline int kmalloc_index(int size)
+static inline int kmalloc_index(size_t size)
{
/*
* We should return 0 if size == 0 but we use the smallest object
@@ -87,7 +76,7 @@ static inline int kmalloc_index(int size
*/
WARN_ON_ONCE(size == 0);

- if (size > (1 << KMALLOC_SHIFT_HIGH))
+ if (size > KMALLOC_MAX_SIZE)
return -1;

if (size > 64 && size <= 96)
@@ -110,17 +99,13 @@ static inline int kmalloc_index(int size
if (size <= 64 * 1024) return 16;
if (size <= 128 * 1024) return 17;
if (size <= 256 * 1024) return 18;
-#if KMALLOC_SHIFT_HIGH > 18
if (size <= 512 * 1024) return 19;
if (size <= 1024 * 1024) return 20;
-#endif
-#if KMALLOC_SHIFT_HIGH > 20
if (size <= 2 * 1024 * 1024) return 21;
if (size <= 4 * 1024 * 1024) return 22;
if (size <= 8 * 1024 * 1024) return 23;
if (size <= 16 * 1024 * 1024) return 24;
if (size <= 32 * 1024 * 1024) return 25;
-#endif
return -1;

/*
Index: slub/include/linux/kmalloc_sizes.h
===================================================================
--- slub.orig/include/linux/kmalloc_sizes.h 2007-05-12 18:45:56.000000000 -0700
+++ slub/include/linux/kmalloc_sizes.h 2007-05-15 21:19:51.000000000 -0700
@@ -19,17 +19,27 @@
CACHE(32768)
CACHE(65536)
CACHE(131072)
-#if (NR_CPUS > 512) || (MAX_NUMNODES > 256) || !defined(CONFIG_MMU)
+#if KMALLOC_MAX_SIZE >= 262144
CACHE(262144)
#endif
-#ifndef CONFIG_MMU
+#if KMALLOC_MAX_SIZE >= 524288
CACHE(524288)
+#endif
+#if KMALLOC_MAX_SIZE >= 1048576
CACHE(1048576)
-#ifdef CONFIG_LARGE_ALLOCS
+#endif
+#if KMALLOC_MAX_SIZE >= 2097152
CACHE(2097152)
+#endif
+#if KMALLOC_MAX_SIZE >= 4194304
CACHE(4194304)
+#endif
+#if KMALLOC_MAX_SIZE >= 8388608
CACHE(8388608)
+#endif
+#if KMALLOC_MAX_SIZE >= 16777216
CACHE(16777216)
+#endif
+#if KMALLOC_MAX_SIZE >= 33554432
CACHE(33554432)
-#endif /* CONFIG_LARGE_ALLOCS */
-#endif /* CONFIG_MMU */
+#endif
Index: slub/mm/slab.c
===================================================================
--- slub.orig/mm/slab.c 2007-05-15 21:17:15.000000000 -0700
+++ slub/mm/slab.c 2007-05-15 21:19:51.000000000 -0700
@@ -569,21 +569,6 @@ static void **dbg_userword(struct kmem_c
#endif

/*
- * Maximum size of an obj (in 2^order pages) and absolute limit for the gfp
- * order.
- */
-#if defined(CONFIG_LARGE_ALLOCS)
-#define MAX_OBJ_ORDER 13 /* up to 32Mb */
-#define MAX_GFP_ORDER 13 /* up to 32Mb */
-#elif defined(CONFIG_MMU)
-#define MAX_OBJ_ORDER 5 /* 32 pages */
-#define MAX_GFP_ORDER 5 /* 32 pages */
-#else
-#define MAX_OBJ_ORDER 8 /* up to 1Mb */
-#define MAX_GFP_ORDER 8 /* up to 1Mb */
-#endif
-
-/*
* Do not go above this order unless 0 objects fit into the slab.
*/
#define BREAK_GFP_ORDER_HI 1
@@ -2004,7 +1989,7 @@ static size_t calculate_slab_order(struc
size_t left_over = 0;
int gfporder;

- for (gfporder = 0; gfporder <= MAX_GFP_ORDER; gfporder++) {
+ for (gfporder = 0; gfporder <= KMALLOC_MAX_ORDER; gfporder++) {
unsigned int num;
size_t remainder;

@@ -2150,7 +2135,7 @@ kmem_cache_create (const char *name, siz
* Sanity checks... these are all serious usage bugs.
*/
if (!name || in_interrupt() || (size < BYTES_PER_WORD) ||
- (size > (1 << MAX_OBJ_ORDER) * PAGE_SIZE) || dtor) {
+ size > KMALLOC_MAX_SIZE || dtor) {
printk(KERN_ERR "%s: Early error in slab %s\n", __FUNCTION__,
name);
BUG();
Index: slub/arch/blackfin/Kconfig
===================================================================
--- slub.orig/arch/blackfin/Kconfig 2007-05-12 18:45:56.000000000 -0700
+++ slub/arch/blackfin/Kconfig 2007-05-15 21:19:51.000000000 -0700
@@ -560,14 +560,6 @@ endchoice

source "mm/Kconfig"

-config LARGE_ALLOCS
- bool "Allow allocating large blocks (> 1MB) of memory"
- help
- Allow the slab memory allocator to keep chains for very large
- memory sizes - upto 32MB. You may need this if your system has
- a lot of RAM, and you need to able to allocate very large
- contiguous chunks. If unsure, say N.
-
config BFIN_DMA_5XX
bool "Enable DMA Support"
depends on (BF533 || BF532 || BF531 || BF537 || BF536 || BF534 || BF561)
Index: slub/arch/frv/Kconfig
===================================================================
--- slub.orig/arch/frv/Kconfig 2007-05-12 18:45:56.000000000 -0700
+++ slub/arch/frv/Kconfig 2007-05-15 21:19:51.000000000 -0700
@@ -102,14 +102,6 @@ config HIGHPTE
with a lot of RAM, this can be wasteful of precious low memory.
Setting this option will put user-space page tables in high memory.

-config LARGE_ALLOCS
- bool "Allow allocating large blocks (> 1MB) of memory"
- help
- Allow the slab memory allocator to keep chains for very large memory
- sizes - up to 32MB. You may need this if your system has a lot of
- RAM, and you need to able to allocate very large contiguous chunks.
- If unsure, say N.
-
source "mm/Kconfig"

choice
Index: slub/arch/m68knommu/Kconfig
===================================================================
--- slub.orig/arch/m68knommu/Kconfig 2007-05-12 18:45:56.000000000 -0700
+++ slub/arch/m68knommu/Kconfig 2007-05-15 21:19:51.000000000 -0700
@@ -470,14 +470,6 @@ config AVNET
default y
depends on (AVNET5282)

-config LARGE_ALLOCS
- bool "Allow allocating large blocks (> 1MB) of memory"
- help
- Allow the slab memory allocator to keep chains for very large
- memory sizes - upto 32MB. You may need this if your system has
- a lot of RAM, and you need to able to allocate very large
- contiguous chunks. If unsure, say N.
-
config 4KSTACKS
bool "Use 4Kb for kernel stacks instead of 8Kb"
default y
Index: slub/arch/v850/Kconfig
===================================================================
--- slub.orig/arch/v850/Kconfig 2007-05-12 18:45:56.000000000 -0700
+++ slub/arch/v850/Kconfig 2007-05-15 21:19:51.000000000 -0700
@@ -240,14 +240,6 @@ menu "Processor type and features"
config RESET_GUARD
bool "Reset Guard"

- config LARGE_ALLOCS
- bool "Allow allocating large blocks (> 1MB) of memory"
- help
- Allow the slab memory allocator to keep chains for very large
- memory sizes - upto 32MB. You may need this if your system has
- a lot of RAM, and you need to able to allocate very large
- contiguous chunks. If unsure, say N.
-
source "mm/Kconfig"

endmenu


2007-05-16 06:34:55

by Andrew Morton

[permalink] [raw]
Subject: Re: Slab allocators: Define common size limitations

On Tue, 15 May 2007 23:15:24 -0700 (PDT) Christoph Lameter <[email protected]> wrote:

> Currently we have a maze of configuration variables that determine
> the maximum slab size. Worst of all it seems to vary between SLAB and SLUB.
>
> So define a common maximum size for kmalloc. For conveniences sake
> we use the maximum size ever supported which is 32 MB. We limit the maximum
> size to a lower limit if MAX_ORDER does not allow such large allocations.
>
> For many architectures this patch will have the effect of adding large
> kmalloc sizes. x86_64 adds 5 new kmalloc sizes. So a small amount
> of memory will be needed for these caches (contemporary SLAB has dynamically
> sizeable node and cpu structure so the waste is less than in the past)
>
> Most architectures will then be able to allocate object with sizes up to
> MAX_ORDER. We have had repeated breakage (in fact whenever we doubled the
> number of supported processors) on IA64 because one or the other struct
> grew beyond what the slab allocators supported. This will avoid future
> issues and f.e. avoid fixes for 2k and 4k cpu support.
>
> CONFIG_LARGE_ALLOCS is no longer necessary so drop it.
>
> Signed-off-by: Christoph Lameter <[email protected]>
>
> ---
> arch/blackfin/Kconfig | 8 --------
> arch/frv/Kconfig | 8 --------
> arch/m68knommu/Kconfig | 8 --------
> arch/v850/Kconfig | 8 --------
> include/linux/kmalloc_sizes.h | 20 +++++++++++++++-----
> include/linux/slab.h | 15 +++++++++++++++
> include/linux/slub_def.h | 19 ++-----------------
> mm/slab.c | 19 ++-----------------
> 8 files changed, 34 insertions(+), 71 deletions(-)

rofl. Really we shouldn't put this into 2.6.22, but it turfs out so much
crap that it's hard to justify holding it back.

2007-05-16 06:39:41

by David Miller

[permalink] [raw]
Subject: Re: Slab allocators: Define common size limitations

From: Andrew Morton <[email protected]>
Date: Tue, 15 May 2007 23:32:39 -0700

> rofl. Really we shouldn't put this into 2.6.22, but it turfs out so much
> crap that it's hard to justify holding it back.

If fixes sparc64 with SLAB for one thing. I was going to put
LARGE_ALLOCS back into sparc64/Kconfig but this is just soooo
much better.

2007-05-16 06:58:53

by Geert Uytterhoeven

[permalink] [raw]
Subject: Re: Slab allocators: Define common size limitations

On Tue, 15 May 2007, Christoph Lameter wrote:
> So define a common maximum size for kmalloc. For conveniences sake
> we use the maximum size ever supported which is 32 MB. We limit the maximum
> size to a lower limit if MAX_ORDER does not allow such large allocations.

What are the changes a large allocation will actually succeed?
Is there an alignment rule for large allocations?

E.g. for one of the PS3 drivers I need a physically contiguous 256 KiB-aligned
block of 256 KiB. Currently I'm using __alloc_bootmem() for that, but maybe
kmalloc() becomes a suitable alternative now?

Gr{oetje,eeting}s,

Geert

--
Geert Uytterhoeven -- Sony Network and Software Technology Center Europe (NSCE)
[email protected] ------- The Corporate Village, Da Vincilaan 7-D1
Voice +32-2-7008453 Fax +32-2-7008622 ---------------- B-1935 Zaventem, Belgium

2007-05-16 07:02:36

by David Miller

[permalink] [raw]
Subject: Re: Slab allocators: Define common size limitations

From: Geert Uytterhoeven <[email protected]>
Date: Wed, 16 May 2007 08:58:39 +0200 (CEST)

> E.g. for one of the PS3 drivers I need a physically contiguous 256
> KiB-aligned block of 256 KiB. Currently I'm using __alloc_bootmem()
> for that, but maybe kmalloc() becomes a suitable alternative now?

I'm allocating up to 1MB for per-process TLB hash tables
on sparc64. But I can gracefully handle failures and it's
just a performance tweak to use such large sized tables.

2007-05-16 17:41:59

by Christoph Lameter

[permalink] [raw]
Subject: Re: Slab allocators: Define common size limitations

On Wed, 16 May 2007, Geert Uytterhoeven wrote:

> On Tue, 15 May 2007, Christoph Lameter wrote:
> > So define a common maximum size for kmalloc. For conveniences sake
> > we use the maximum size ever supported which is 32 MB. We limit the maximum
> > size to a lower limit if MAX_ORDER does not allow such large allocations.
>
> What are the changes a large allocation will actually succeed?
> Is there an alignment rule for large allocations?
>
> E.g. for one of the PS3 drivers I need a physically contiguous 256 KiB-aligned
> block of 256 KiB. Currently I'm using __alloc_bootmem() for that, but maybe
> kmalloc() becomes a suitable alternative now?

The chance of succeeding drops with the time that the system has been
running. Typically these large allocs are used when the system is brought
up. Maybe we will be able to successfully allocate these even after
memory has gotten significant use when Mel's antifrag/defrag work has
progressed more.

2007-05-16 21:43:57

by Arnd Bergmann

[permalink] [raw]
Subject: Re: Slab allocators: Define common size limitations

On Wednesday 16 May 2007, Geert Uytterhoeven wrote:
> What are the changes a large allocation will actually succeed?
> Is there an alignment rule for large allocations?
>
> E.g. for one of the PS3 drivers I need a physically contiguous 256 KiB-aligned
> block of 256 KiB. Currently I'm using __alloc_bootmem() for that, but maybe
> kmalloc() becomes a suitable alternative now?

kmalloc is limited to 128KiB on most architectures. Normally there is no
need to use it anyway, just use __get_free_pages(). It will generally
succeed at early boot time, but not after the system has been running
for some time.

Arnd <><

2007-05-17 08:45:20

by Geert Uytterhoeven

[permalink] [raw]
Subject: Re: Slab allocators: Define common size limitations

On Wed, 16 May 2007, Arnd Bergmann wrote:
> On Wednesday 16 May 2007, Geert Uytterhoeven wrote:
> > What are the changes a large allocation will actually succeed?
> > Is there an alignment rule for large allocations?
> >
> > E.g. for one of the PS3 drivers I need a physically contiguous 256 KiB-aligned
> > block of 256 KiB. Currently I'm using __alloc_bootmem() for that, but maybe
> > kmalloc() becomes a suitable alternative now?
>
> kmalloc is limited to 128KiB on most architectures. Normally there is no
> need to use it anyway, just use __get_free_pages(). It will generally
> succeed at early boot time, but not after the system has been running
> for some time.

Exactly my understanding. And __get_free_pages() returns PAGE_SIZE-aligned
memory. So I'll keep the current code.

Gr{oetje,eeting}s,

Geert

--
Geert Uytterhoeven -- Sony Network and Software Technology Center Europe (NSCE)
[email protected] ------- The Corporate Village, Da Vincilaan 7-D1
Voice +32-2-7008453 Fax +32-2-7008622 ---------------- B-1935 Zaventem, Belgium