From 68a3a7feb08f960095072f28ec20f7900793c506 Mon Sep 17 00:00:00 2001 From: Andi Kleen Date: Fri, 7 Apr 2006 19:49:18 +0200 Subject: [PATCH] [PATCH] x86_64: Reserve SRAT hotadd memory on x86-64 From: Keith Mannthey, Andi Kleen Implement memory hotadd without sparsemem. The memory in the SRAT hotadd area is just preserved instead and can be activated later. There are a few restrictions: - Only one continuous hotadd area allowed per node The main problem is dealing with the many buggy SRAT tables that are out there. The strategy here is to reject anything suspicious. Originally from Keith Mannthey, with several hacks and changes by AK and also contributions from Andrew Morton [ TBD: Problems pointed out by KAMEZAWA Hiroyuki : 1) Goto's rebuild_zonelist patch will not work if CONFIG_MEMORY_HOTPLUG=n. Rebuilding zonelist is necessary when the system has just memory < 4G at boot, and hot add memory > 4G. because x86_64 has DMA32, ZONE_NORAML is not included into zonelist at boot time if system doesn't have memory >4G at boot. [AK: should just force the higher zones at boot time when SRAT tells us] 2) zone and node's spanned_pages and present_pages are not incremented. They should be. For example, our server (ia64/Fujitsu PrimeQuest) can equip memory from 4G to 1T(maybe 2T in future), and SRAT will *always* say we have possible 1T +memory. (Microsoft requires "write all possible memory in SRAT") When we reserve memmap for possible 1T memory, Linux will not work well in +minimum 4G configuraion ;) [AK: needs limiting to 5-10% of max memory] ] Signed-off-by: Andi Kleen Signed-off-by: Linus Torvalds --- Documentation/x86_64/boot-options.txt | 5 + arch/x86_64/mm/init.c | 3 +- arch/x86_64/mm/numa.c | 5 + arch/x86_64/mm/srat.c | 164 +++++++++++++++++++++++++- include/asm-x86_64/numa.h | 2 + 5 files changed, 171 insertions(+), 8 deletions(-) diff --git a/Documentation/x86_64/boot-options.txt b/Documentation/x86_64/boot-options.txt index 1921353259ae..f2cd6ef53ff3 100644 --- a/Documentation/x86_64/boot-options.txt +++ b/Documentation/x86_64/boot-options.txt @@ -151,6 +151,11 @@ NUMA numa=fake=X Fake X nodes and ignore NUMA setup of the actual machine. + numa=hotadd=percent + Only allow hotadd memory to preallocate page structures upto + percent of already available memory. + numa=hotadd=0 will disable hotadd memory. + ACPI acpi=off Don't enable ACPI diff --git a/arch/x86_64/mm/init.c b/arch/x86_64/mm/init.c index 492161168402..dff870534199 100644 --- a/arch/x86_64/mm/init.c +++ b/arch/x86_64/mm/init.c @@ -530,8 +530,7 @@ int __add_pages(struct zone *z, unsigned long start_pfn, unsigned long nr_pages) unsigned long pfn; unsigned long total = 0, mem = 0; for (pfn = start_pfn; pfn < start_pfn + nr_pages; pfn++) { - unsigned long addr = pfn << PAGE_SHIFT; - if (pfn_valid(pfn) && e820_mapped(addr, addr+1, E820_RAM)) { + if (pfn_valid(pfn)) { online_page(pfn_to_page(pfn)); err = 0; mem++; diff --git a/arch/x86_64/mm/numa.c b/arch/x86_64/mm/numa.c index 4be82d6e2b48..779132af29a7 100644 --- a/arch/x86_64/mm/numa.c +++ b/arch/x86_64/mm/numa.c @@ -142,6 +142,9 @@ void __init setup_node_bootmem(int nodeid, unsigned long start, unsigned long en reserve_bootmem_node(NODE_DATA(nodeid), nodedata_phys, pgdat_size); reserve_bootmem_node(NODE_DATA(nodeid), bootmap_start, bootmap_pages<> PAGE_SHIFT; + unsigned long e_pfn = end >> PAGE_SHIFT; + int changed = 0; + struct bootnode *nd = &nodes_add[node]; + + /* I had some trouble with strange memory hotadd regions breaking + the boot. Be very strict here and reject anything unexpected. + If you want working memory hotadd write correct SRATs. + + The node size check is a basic sanity check to guard against + mistakes */ + if ((signed long)(end - start) < NODE_MIN_SIZE) { + printk(KERN_ERR "SRAT: Hotplug area too small\n"); + return -1; + } + + /* This check might be a bit too strict, but I'm keeping it for now. */ + if (e820_hole_size(s_pfn, e_pfn) != e_pfn - s_pfn) { + printk(KERN_ERR "SRAT: Hotplug area has existing memory\n"); + return -1; + } + + if (!hotadd_enough_memory(&nodes_add[node])) { + printk(KERN_ERR "SRAT: Hotplug area too large\n"); + return -1; + } + + /* Looks good */ + + found_add_area = 1; + if (nd->start == nd->end) { + nd->start = start; + nd->end = end; + changed = 1; + } else { + if (nd->start == end) { + nd->start = start; + changed = 1; + } + if (nd->end == start) { + nd->end = end; + changed = 1; + } + if (!changed) + printk(KERN_ERR "SRAT: Hotplug zone not continuous. Partly ignored\n"); + } + + if ((nd->end >> PAGE_SHIFT) > end_pfn) + end_pfn = nd->end >> PAGE_SHIFT; + + if (changed) + printk(KERN_INFO "SRAT: hot plug zone found %Lx - %Lx\n", nd->start, nd->end); + return 0; +} +#endif + /* Callback for parsing of the Proximity Domain <-> Memory Area mappings */ void __init acpi_numa_memory_affinity_init(struct acpi_table_memory_affinity *ma) { - struct bootnode *nd; + struct bootnode *nd, oldnode; unsigned long start, end; int node, pxm; int i; @@ -172,6 +292,8 @@ acpi_numa_memory_affinity_init(struct acpi_table_memory_affinity *ma) } if (ma->flags.enabled == 0) return; + if (ma->flags.hot_pluggable && hotadd_percent == 0) + return; start = ma->base_addr_lo | ((u64)ma->base_addr_hi << 32); end = start + (ma->length_lo | ((u64)ma->length_hi << 32)); pxm = ma->proximity_domain; @@ -181,10 +303,6 @@ acpi_numa_memory_affinity_init(struct acpi_table_memory_affinity *ma) bad_srat(); return; } - /* It is fine to add this area to the nodes data it will be used later*/ - if (ma->flags.hot_pluggable == 1) - printk(KERN_INFO "SRAT: hot plug zone found %lx - %lx \n", - start, end); i = conflicting_nodes(start, end); if (i == node) { printk(KERN_WARNING @@ -199,6 +317,7 @@ acpi_numa_memory_affinity_init(struct acpi_table_memory_affinity *ma) return; } nd = &nodes[node]; + oldnode = *nd; if (!node_test_and_set(node, nodes_parsed)) { nd->start = start; nd->end = end; @@ -208,8 +327,19 @@ acpi_numa_memory_affinity_init(struct acpi_table_memory_affinity *ma) if (nd->end < end) nd->end = end; } + printk(KERN_INFO "SRAT: Node %u PXM %u %Lx-%Lx\n", node, pxm, nd->start, nd->end); + +#ifdef RESERVE_HOTADD + if (ma->flags.hot_pluggable && reserve_hotadd(node, start, end) < 0) { + /* Ignore hotadd region. Undo damage */ + printk(KERN_NOTICE "SRAT: Hotplug region ignored\n"); + *nd = oldnode; + if ((nd->start | nd->end) == 0) + node_clear(node, nodes_parsed); + } +#endif } /* Sanity check to catch more bad SRATs (they are amazingly common). @@ -225,6 +355,9 @@ static int nodes_cover_memory(void) unsigned long e = nodes[i].end >> PAGE_SHIFT; pxmram += e - s; pxmram -= e820_hole_size(s, e); + pxmram -= nodes_add[i].end - nodes_add[i].start; + if ((long)pxmram < 0) + pxmram = 0; } e820ram = end_pfn - e820_hole_size(0, end_pfn); @@ -258,7 +391,7 @@ int __init acpi_scan_nodes(unsigned long start, unsigned long end) /* First clean up the node list */ for (i = 0; i < MAX_NUMNODES; i++) { - cutoff_node(i, start, end); + cutoff_node(i, start, end); if ((nodes[i].end - nodes[i].start) < NODE_MIN_SIZE) unparse_node(i); } @@ -303,6 +436,25 @@ static int node_to_pxm(int n) return 0; } +void __init srat_reserve_add_area(int nodeid) +{ + if (found_add_area && nodes_add[nodeid].end) { + u64 total_mb; + + printk(KERN_INFO "SRAT: Reserving hot-add memory space " + "for node %d at %Lx-%Lx\n", + nodeid, nodes_add[nodeid].start, nodes_add[nodeid].end); + total_mb = (nodes_add[nodeid].end - nodes_add[nodeid].start) + >> PAGE_SHIFT; + total_mb *= sizeof(struct page); + total_mb >>= 20; + printk(KERN_INFO "SRAT: This will cost you %Lu MB of " + "pre-allocated memory.\n", (unsigned long long)total_mb); + reserve_bootmem_node(NODE_DATA(nodeid), nodes_add[nodeid].start, + nodes_add[nodeid].end - nodes_add[nodeid].start); + } +} + int __node_distance(int a, int b) { int index; diff --git a/include/asm-x86_64/numa.h b/include/asm-x86_64/numa.h index f6cbb4cbb5a3..f0ba4d984bdf 100644 --- a/include/asm-x86_64/numa.h +++ b/include/asm-x86_64/numa.h @@ -18,6 +18,8 @@ extern void numa_init_array(void); extern int numa_off; extern void numa_set_node(int cpu, int node); +extern void srat_reserve_add_area(int nodeid); +extern int hotadd_percent; extern unsigned char apicid_to_node[256]; #ifdef CONFIG_NUMA