aboutsummaryrefslogtreecommitdiffstats
path: root/arch
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 /arch
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 'arch')
-rw-r--r--arch/x86_64/mm/init.c65
-rw-r--r--arch/x86_64/mm/numa.c25
2 files changed, 47 insertions, 43 deletions
diff --git a/arch/x86_64/mm/init.c b/arch/x86_64/mm/init.c
index e60a1a848de8..a1ad4cc423a7 100644
--- a/arch/x86_64/mm/init.c
+++ b/arch/x86_64/mm/init.c
@@ -318,32 +318,51 @@ void zap_low_mappings(void)
318 flush_tlb_all(); 318 flush_tlb_all();
319} 319}
320 320
321/* Compute zone sizes for the DMA and DMA32 zones in a node. */
322__init void
323size_zones(unsigned long *z, unsigned long *h,
324 unsigned long start_pfn, unsigned long end_pfn)
325{
326 int i;
327 unsigned long w;
328
329 for (i = 0; i < MAX_NR_ZONES; i++)
330 z[i] = 0;
331
332 if (start_pfn < MAX_DMA_PFN)
333 z[ZONE_DMA] = MAX_DMA_PFN - start_pfn;
334 if (start_pfn < MAX_DMA32_PFN) {
335 unsigned long dma32_pfn = MAX_DMA32_PFN;
336 if (dma32_pfn > end_pfn)
337 dma32_pfn = end_pfn;
338 z[ZONE_DMA32] = dma32_pfn - start_pfn;
339 }
340 z[ZONE_NORMAL] = end_pfn - start_pfn;
341
342 /* Remove lower zones from higher ones. */
343 w = 0;
344 for (i = 0; i < MAX_NR_ZONES; i++) {
345 if (z[i])
346 z[i] -= w;
347 w += z[i];
348 }
349
350 /* Compute holes */
351 w = 0;
352 for (i = 0; i < MAX_NR_ZONES; i++) {
353 unsigned long s = w;
354 w += z[i];
355 h[i] = e820_hole_size(s, w);
356 }
357}
358
321#ifndef CONFIG_NUMA 359#ifndef CONFIG_NUMA
322void __init paging_init(void) 360void __init paging_init(void)
323{ 361{
324 { 362 unsigned long zones[MAX_NR_ZONES], holes[MAX_NR_ZONES];
325 unsigned long zones_size[MAX_NR_ZONES]; 363 size_zones(zones, holes, 0, end_pfn);
326 unsigned long holes[MAX_NR_ZONES]; 364 free_area_init_node(0, NODE_DATA(0), zones,
327 unsigned int max_dma; 365 __pa(PAGE_OFFSET) >> PAGE_SHIFT, holes);
328
329 memset(zones_size, 0, sizeof(zones_size));
330 memset(holes, 0, sizeof(holes));
331
332 max_dma = virt_to_phys((char *)MAX_DMA_ADDRESS) >> PAGE_SHIFT;
333
334 if (end_pfn < max_dma) {
335 zones_size[ZONE_DMA] = end_pfn;
336 holes[ZONE_DMA] = e820_hole_size(0, end_pfn);
337 } else {
338 zones_size[ZONE_DMA] = max_dma;
339 holes[ZONE_DMA] = e820_hole_size(0, max_dma);
340 zones_size[ZONE_NORMAL] = end_pfn - max_dma;
341 holes[ZONE_NORMAL] = e820_hole_size(max_dma, end_pfn);
342 }
343 free_area_init_node(0, NODE_DATA(0), zones_size,
344 __pa(PAGE_OFFSET) >> PAGE_SHIFT, holes);
345 }
346 return;
347} 366}
348#endif 367#endif
349 368
diff --git a/arch/x86_64/mm/numa.c b/arch/x86_64/mm/numa.c
index 214803821001..18e86e2eac2d 100644
--- a/arch/x86_64/mm/numa.c
+++ b/arch/x86_64/mm/numa.c
@@ -132,29 +132,14 @@ void __init setup_node_zones(int nodeid)
132 unsigned long start_pfn, end_pfn; 132 unsigned long start_pfn, end_pfn;
133 unsigned long zones[MAX_NR_ZONES]; 133 unsigned long zones[MAX_NR_ZONES];
134 unsigned long holes[MAX_NR_ZONES]; 134 unsigned long holes[MAX_NR_ZONES];
135 unsigned long dma_end_pfn;
136 135
137 memset(zones, 0, sizeof(unsigned long) * MAX_NR_ZONES); 136 start_pfn = node_start_pfn(nodeid);
138 memset(holes, 0, sizeof(unsigned long) * MAX_NR_ZONES); 137 end_pfn = node_end_pfn(nodeid);
139 138
140 start_pfn = node_start_pfn(nodeid); 139 Dprintk(KERN_INFO "setting up node %d %lx-%lx\n",
141 end_pfn = node_end_pfn(nodeid); 140 nodeid, start_pfn, end_pfn);
142 141
143 Dprintk(KERN_INFO "setting up node %d %lx-%lx\n", nodeid, start_pfn, end_pfn); 142 size_zones(zones, holes, start_pfn, end_pfn);
144
145 /* All nodes > 0 have a zero length zone DMA */
146 dma_end_pfn = __pa(MAX_DMA_ADDRESS) >> PAGE_SHIFT;
147 if (start_pfn < dma_end_pfn) {
148 zones[ZONE_DMA] = dma_end_pfn - start_pfn;
149 holes[ZONE_DMA] = e820_hole_size(start_pfn, dma_end_pfn);
150 zones[ZONE_NORMAL] = end_pfn - dma_end_pfn;
151 holes[ZONE_NORMAL] = e820_hole_size(dma_end_pfn, end_pfn);
152
153 } else {
154 zones[ZONE_NORMAL] = end_pfn - start_pfn;
155 holes[ZONE_NORMAL] = e820_hole_size(start_pfn, end_pfn);
156 }
157
158 free_area_init_node(nodeid, NODE_DATA(nodeid), zones, 143 free_area_init_node(nodeid, NODE_DATA(nodeid), zones,
159 start_pfn, holes); 144 start_pfn, holes);
160} 145}