aboutsummaryrefslogtreecommitdiffstats
path: root/include/linux
diff options
context:
space:
mode:
authorAndi Kleen <ak@suse.de>2005-11-05 11:25:53 -0500
committerLinus Torvalds <torvalds@g5.osdl.org>2005-11-14 22:55:13 -0500
commita2f1b424900715ed9d1699c3bb88a434a2b42bc0 (patch)
tree8ef440f840656365166ff2d71aa445c224c53546 /include/linux
parent56720367cd89ef5265f39da2d674c5b92cd4cd87 (diff)
[PATCH] x86_64: Add 4GB DMA32 zone
Add a new 4GB GFP_DMA32 zone between the GFP_DMA and GFP_NORMAL zones. As a bit of historical background: when the x86-64 port was originally designed we had some discussion if we should use a 16MB DMA zone like i386 or a 4GB DMA zone like IA64 or both. Both was ruled out at this point because it was in early 2.4 when VM is still quite shakey and had bad troubles even dealing with one DMA zone. We settled on the 16MB DMA zone mainly because we worried about older soundcards and the floppy. But this has always caused problems since then because device drivers had trouble getting enough DMA able memory. These days the VM works much better and the wide use of NUMA has proven it can deal with many zones successfully. So this patch adds both zones. This helps drivers who need a lot of memory below 4GB because their hardware is not accessing more (graphic drivers - proprietary and free ones, video frame buffer drivers, sound drivers etc.). Previously they could only use IOMMU+16MB GFP_DMA, which was not enough memory. Another common problem is that hardware who has full memory addressing for >4GB misses it for some control structures in memory (like transmit rings or other metadata). They tended to allocate memory in the 16MB GFP_DMA or the IOMMU/swiotlb then using pci_alloc_consistent, but that can tie up a lot of precious 16MB GFPDMA/IOMMU/swiotlb memory (even on AMD systems the IOMMU tends to be quite small) especially if you have many devices. With the new zone pci_alloc_consistent can just put this stuff into memory below 4GB which works better. One argument was still if the zone should be 4GB or 2GB. The main motivation for 2GB would be an unnamed not so unpopular hardware raid controller (mostly found in older machines from a particular four letter company) who has a strange 2GB restriction in firmware. But that one works ok with swiotlb/IOMMU anyways, so it doesn't really need GFP_DMA32. I chose 4GB to be compatible with IA64 and because it seems to be the most common restriction. The new zone is so far added only for x86-64. For other architectures who don't set up this new zone nothing changes. Architectures can set a compatibility define in Kconfig CONFIG_DMA_IS_DMA32 that will define GFP_DMA32 as GFP_DMA. Otherwise it's a nop because on 32bit architectures it's normally not needed because GFP_NORMAL (=0) is DMA able enough. One problem is still that GFP_DMA means different things on different architectures. e.g. some drivers used to have #ifdef ia64 use GFP_DMA (trusting it to be 4GB) #elif __x86_64__ (use other hacks like the swiotlb because 16MB is not enough) ... . This was quite ugly and is now obsolete. These should be now converted to use GFP_DMA32 unconditionally. I haven't done this yet. Or best only use pci_alloc_consistent/dma_alloc_coherent which will use GFP_DMA32 transparently. Signed-off-by: Andi Kleen <ak@suse.de> Signed-off-by: Linus Torvalds <torvalds@osdl.org>
Diffstat (limited to 'include/linux')
-rw-r--r--include/linux/gfp.h11
-rw-r--r--include/linux/mmzone.h16
2 files changed, 20 insertions, 7 deletions
diff --git a/include/linux/gfp.h b/include/linux/gfp.h
index c3779432a723..4351e6bb5a79 100644
--- a/include/linux/gfp.h
+++ b/include/linux/gfp.h
@@ -14,6 +14,13 @@ struct vm_area_struct;
14/* Zone modifiers in GFP_ZONEMASK (see linux/mmzone.h - low two bits) */ 14/* Zone modifiers in GFP_ZONEMASK (see linux/mmzone.h - low two bits) */
15#define __GFP_DMA ((__force gfp_t)0x01u) 15#define __GFP_DMA ((__force gfp_t)0x01u)
16#define __GFP_HIGHMEM ((__force gfp_t)0x02u) 16#define __GFP_HIGHMEM ((__force gfp_t)0x02u)
17#ifdef CONFIG_DMA_IS_DMA32
18#define __GFP_DMA32 ((__force gfp_t)0x01) /* ZONE_DMA is ZONE_DMA32 */
19#elif BITS_PER_LONG < 64
20#define __GFP_DMA32 ((__force gfp_t)0x00) /* ZONE_NORMAL is ZONE_DMA32 */
21#else
22#define __GFP_DMA32 ((__force gfp_t)0x04) /* Has own ZONE_DMA32 */
23#endif
17 24
18/* 25/*
19 * Action modifiers - doesn't change the zoning 26 * Action modifiers - doesn't change the zoning
@@ -64,6 +71,10 @@ struct vm_area_struct;
64 71
65#define GFP_DMA __GFP_DMA 72#define GFP_DMA __GFP_DMA
66 73
74/* 4GB DMA on some platforms */
75#define GFP_DMA32 __GFP_DMA32
76
77
67#define gfp_zone(mask) ((__force int)((mask) & (__force gfp_t)GFP_ZONEMASK)) 78#define gfp_zone(mask) ((__force int)((mask) & (__force gfp_t)GFP_ZONEMASK))
68 79
69/* 80/*
diff --git a/include/linux/mmzone.h b/include/linux/mmzone.h
index f5fa3082fd6a..da7a829f8561 100644
--- a/include/linux/mmzone.h
+++ b/include/linux/mmzone.h
@@ -71,10 +71,11 @@ struct per_cpu_pageset {
71#endif 71#endif
72 72
73#define ZONE_DMA 0 73#define ZONE_DMA 0
74#define ZONE_NORMAL 1 74#define ZONE_DMA32 1
75#define ZONE_HIGHMEM 2 75#define ZONE_NORMAL 2
76#define ZONE_HIGHMEM 3
76 77
77#define MAX_NR_ZONES 3 /* Sync this with ZONES_SHIFT */ 78#define MAX_NR_ZONES 4 /* Sync this with ZONES_SHIFT */
78#define ZONES_SHIFT 2 /* ceil(log2(MAX_NR_ZONES)) */ 79#define ZONES_SHIFT 2 /* ceil(log2(MAX_NR_ZONES)) */
79 80
80 81
@@ -108,9 +109,10 @@ struct per_cpu_pageset {
108 109
109/* 110/*
110 * On machines where it is needed (eg PCs) we divide physical memory 111 * On machines where it is needed (eg PCs) we divide physical memory
111 * into multiple physical zones. On a PC we have 3 zones: 112 * into multiple physical zones. On a PC we have 4 zones:
112 * 113 *
113 * ZONE_DMA < 16 MB ISA DMA capable memory 114 * ZONE_DMA < 16 MB ISA DMA capable memory
115 * ZONE_DMA32 0 MB Empty
114 * ZONE_NORMAL 16-896 MB direct mapped by the kernel 116 * ZONE_NORMAL 16-896 MB direct mapped by the kernel
115 * ZONE_HIGHMEM > 896 MB only page cache and user processes 117 * ZONE_HIGHMEM > 896 MB only page cache and user processes
116 */ 118 */
@@ -455,10 +457,10 @@ extern struct pglist_data contig_page_data;
455 457
456#if BITS_PER_LONG == 32 || defined(ARCH_HAS_ATOMIC_UNSIGNED) 458#if BITS_PER_LONG == 32 || defined(ARCH_HAS_ATOMIC_UNSIGNED)
457/* 459/*
458 * with 32 bit page->flags field, we reserve 8 bits for node/zone info. 460 * with 32 bit page->flags field, we reserve 9 bits for node/zone info.
459 * there are 3 zones (2 bits) and this leaves 8-2=6 bits for nodes. 461 * there are 4 zones (3 bits) and this leaves 9-3=6 bits for nodes.
460 */ 462 */
461#define FLAGS_RESERVED 8 463#define FLAGS_RESERVED 9
462 464
463#elif BITS_PER_LONG == 64 465#elif BITS_PER_LONG == 64
464/* 466/*