linux/arch/x86/mm/numa_64.c

902 lines
23 KiB
C
Raw Normal View History

/*
* Generic VM initialization for x86-64 NUMA setups.
* Copyright 2002,2003 Andi Kleen, SuSE Labs.
*/
#include <linux/kernel.h>
#include <linux/mm.h>
#include <linux/string.h>
#include <linux/init.h>
#include <linux/bootmem.h>
#include <linux/mmzone.h>
#include <linux/ctype.h>
#include <linux/module.h>
#include <linux/nodemask.h>
#include <linux/sched.h>
#include <asm/e820.h>
#include <asm/proto.h>
#include <asm/dma.h>
#include <asm/numa.h>
#include <asm/acpi.h>
#include <asm/k8.h>
struct pglist_data *node_data[MAX_NUMNODES] __read_mostly;
EXPORT_SYMBOL(node_data);
struct memnode memnode;
s16 apicid_to_node[MAX_LOCAL_APIC] __cpuinitdata = {
[0 ... MAX_LOCAL_APIC-1] = NUMA_NO_NODE
};
int numa_off __initdata;
static unsigned long __initdata nodemap_addr;
static unsigned long __initdata nodemap_size;
/*
* Map cpu index to node index
*/
DEFINE_EARLY_PER_CPU(int, x86_cpu_to_node_map, NUMA_NO_NODE);
EXPORT_EARLY_PER_CPU_SYMBOL(x86_cpu_to_node_map);
/*
* Given a shift value, try to populate memnodemap[]
* Returns :
* 1 if OK
* 0 if memnodmap[] too small (of shift too small)
* -1 if node overlap or lost ram (shift too big)
*/
static int __init populate_memnodemap(const struct bootnode *nodes,
int numnodes, int shift, int *nodeids)
{
unsigned long addr, end;
int i, res = -1;
memset(memnodemap, 0xff, sizeof(s16)*memnodemapsize);
for (i = 0; i < numnodes; i++) {
addr = nodes[i].start;
end = nodes[i].end;
if (addr >= end)
continue;
if ((end >> shift) >= memnodemapsize)
return 0;
do {
if (memnodemap[addr >> shift] != NUMA_NO_NODE)
return -1;
if (!nodeids)
memnodemap[addr >> shift] = i;
else
memnodemap[addr >> shift] = nodeids[i];
addr += (1UL << shift);
} while (addr < end);
res = 1;
}
return res;
}
static int __init allocate_cachealigned_memnodemap(void)
{
x86_64: make bootmap_start page align v6 boot oopses when a system has 64 or 128 GB of RAM installed: Calling initcall 0xffffffff80bc33b6: sctp_init+0x0/0x711() BUG: unable to handle kernel NULL pointer dereference at 000000000000005f IP: [<ffffffff802bfe55>] proc_register+0xe7/0x10f PGD 0 Oops: 0000 [1] SMP CPU 0 Modules linked in: Pid: 1, comm: swapper Not tainted 2.6.24-smp-g5a514e21-dirty #6 RIP: 0010:[<ffffffff802bfe55>] [<ffffffff802bfe55>] proc_register+0xe7/0x10f RSP: 0000:ffff810824c57e60 EFLAGS: 00010246 RAX: 000000000000d7d7 RBX: ffff811024c5fa80 RCX: ffff810824c57e08 RDX: 0000000000000000 RSI: 0000000000000195 RDI: ffffffff80cc2460 RBP: ffffffffffffffff R08: 0000000000000000 R09: ffff811024c5fa80 R10: 0000000000000000 R11: 0000000000000002 R12: ffff810824c57e6c R13: 0000000000000000 R14: ffff810824c57ee0 R15: 00000006abd25bee FS: 0000000000000000(0000) GS:ffffffff80b4d000(0000) knlGS:0000000000000000 CS: 0010 DS: 0018 ES: 0018 CR0: 000000008005003b CR2: 000000000000005f CR3: 0000000000201000 CR4: 00000000000006e0 DR0: 0000000000000000 DR1: 0000000000000000 DR2: 0000000000000000 DR3: 0000000000000000 DR6: 00000000ffff0ff0 DR7: 0000000000000400 Process swapper (pid: 1, threadinfo ffff810824c56000, task ffff812024c52000) Stack: ffffffff80a57348 0000019500000000 ffff811024c5fa80 0000000000000000 00000000ffffff97 ffffffff802bfef0 0000000000000000 ffffffffffffffff 0000000000000000 ffffffff80bc3b4b ffff810824c57ee0 ffffffff80bc34a5 Call Trace: [<ffffffff802bfef0>] ? create_proc_entry+0x73/0x8a [<ffffffff80bc3b4b>] ? sctp_snmp_proc_init+0x1c/0x34 [<ffffffff80bc34a5>] ? sctp_init+0xef/0x711 [<ffffffff80b976e3>] ? kernel_init+0x175/0x2e1 [<ffffffff8020ccf8>] ? child_rip+0xa/0x12 [<ffffffff80b9756e>] ? kernel_init+0x0/0x2e1 [<ffffffff8020ccee>] ? child_rip+0x0/0x12 Code: 1e 48 83 7b 38 00 75 08 48 c7 43 38 f0 e8 82 80 48 83 7b 30 00 75 08 48 c7 43 30 d0 e9 82 80 48 c7 c7 60 24 cc 80 e8 bd 5a 54 00 <48> 8b 45 60 48 89 6b 58 48 89 5d 60 48 89 43 50 fe 05 f5 25 a0 RIP [<ffffffff802bfe55>] proc_register+0xe7/0x10f RSP <ffff810824c57e60> CR2: 000000000000005f ---[ end trace 02c2d78def82877a ]--- Kernel panic - not syncing: Attempted to kill init! it turns out some variables near end of bss are corrupted already. in System.map we have ffffffff80d40420 b rsi_table ffffffff80d40620 B krb5_seq_lock ffffffff80d40628 b i.20437 ffffffff80d40630 b xprt_rdma_inline_write_padding ffffffff80d40638 b sunrpc_table_header ffffffff80d40640 b zero ffffffff80d40644 b min_memreg ffffffff80d40648 b rpcrdma_tk_lock_g ffffffff80d40650 B sctp_assocs_id_lock ffffffff80d40658 B proc_net_sctp ffffffff80d40660 B sctp_assocs_id ffffffff80d40680 B sysctl_sctp_mem ffffffff80d40690 B sysctl_sctp_rmem ffffffff80d406a0 B sysctl_sctp_wmem ffffffff80d406b0 b sctp_ctl_socket ffffffff80d406b8 b sctp_pf_inet6_specific ffffffff80d406c0 b sctp_pf_inet_specific ffffffff80d406c8 b sctp_af_v4_specific ffffffff80d406d0 b sctp_af_v6_specific ffffffff80d406d8 b sctp_rand.33270 ffffffff80d406dc b sctp_memory_pressure ffffffff80d406e0 b sctp_sockets_allocated ffffffff80d406e4 b sctp_memory_allocated ffffffff80d406e8 b sctp_sysctl_header ffffffff80d406f0 b zero ffffffff80d406f4 A __bss_stop ffffffff80d406f4 A _end and setup_node_bootmem() will use that page 0xd40000 for bootmap Bootmem setup node 0 0000000000000000-0000000828000000 NODE_DATA [000000000008a485 - 0000000000091484] bootmap [0000000000d406f4 - 0000000000e456f3] pages 105 Bootmem setup node 1 0000000828000000-0000001028000000 NODE_DATA [0000000828000000 - 0000000828006fff] bootmap [0000000828007000 - 0000000828106fff] pages 100 Bootmem setup node 2 0000001028000000-0000001828000000 NODE_DATA [0000001028000000 - 0000001028006fff] bootmap [0000001028007000 - 0000001028106fff] pages 100 Bootmem setup node 3 0000001828000000-0000002028000000 NODE_DATA [0000001828000000 - 0000001828006fff] bootmap [0000001828007000 - 0000001828106fff] pages 100 setup_node_bootmem() makes NODE_DATA cacheline aligned, and bootmap is page-aligned. the patch updates find_e820_area() to make sure we can meet the alignment constraints. Signed-off-by: Yinghai Lu <yinghai.lu@sun.com> Signed-off-by: Ingo Molnar <mingo@elte.hu>
2008-02-01 16:49:41 +00:00
unsigned long addr;
memnodemap = memnode.embedded_map;
if (memnodemapsize <= ARRAY_SIZE(memnode.embedded_map))
return 0;
x86_64: make bootmap_start page align v6 boot oopses when a system has 64 or 128 GB of RAM installed: Calling initcall 0xffffffff80bc33b6: sctp_init+0x0/0x711() BUG: unable to handle kernel NULL pointer dereference at 000000000000005f IP: [<ffffffff802bfe55>] proc_register+0xe7/0x10f PGD 0 Oops: 0000 [1] SMP CPU 0 Modules linked in: Pid: 1, comm: swapper Not tainted 2.6.24-smp-g5a514e21-dirty #6 RIP: 0010:[<ffffffff802bfe55>] [<ffffffff802bfe55>] proc_register+0xe7/0x10f RSP: 0000:ffff810824c57e60 EFLAGS: 00010246 RAX: 000000000000d7d7 RBX: ffff811024c5fa80 RCX: ffff810824c57e08 RDX: 0000000000000000 RSI: 0000000000000195 RDI: ffffffff80cc2460 RBP: ffffffffffffffff R08: 0000000000000000 R09: ffff811024c5fa80 R10: 0000000000000000 R11: 0000000000000002 R12: ffff810824c57e6c R13: 0000000000000000 R14: ffff810824c57ee0 R15: 00000006abd25bee FS: 0000000000000000(0000) GS:ffffffff80b4d000(0000) knlGS:0000000000000000 CS: 0010 DS: 0018 ES: 0018 CR0: 000000008005003b CR2: 000000000000005f CR3: 0000000000201000 CR4: 00000000000006e0 DR0: 0000000000000000 DR1: 0000000000000000 DR2: 0000000000000000 DR3: 0000000000000000 DR6: 00000000ffff0ff0 DR7: 0000000000000400 Process swapper (pid: 1, threadinfo ffff810824c56000, task ffff812024c52000) Stack: ffffffff80a57348 0000019500000000 ffff811024c5fa80 0000000000000000 00000000ffffff97 ffffffff802bfef0 0000000000000000 ffffffffffffffff 0000000000000000 ffffffff80bc3b4b ffff810824c57ee0 ffffffff80bc34a5 Call Trace: [<ffffffff802bfef0>] ? create_proc_entry+0x73/0x8a [<ffffffff80bc3b4b>] ? sctp_snmp_proc_init+0x1c/0x34 [<ffffffff80bc34a5>] ? sctp_init+0xef/0x711 [<ffffffff80b976e3>] ? kernel_init+0x175/0x2e1 [<ffffffff8020ccf8>] ? child_rip+0xa/0x12 [<ffffffff80b9756e>] ? kernel_init+0x0/0x2e1 [<ffffffff8020ccee>] ? child_rip+0x0/0x12 Code: 1e 48 83 7b 38 00 75 08 48 c7 43 38 f0 e8 82 80 48 83 7b 30 00 75 08 48 c7 43 30 d0 e9 82 80 48 c7 c7 60 24 cc 80 e8 bd 5a 54 00 <48> 8b 45 60 48 89 6b 58 48 89 5d 60 48 89 43 50 fe 05 f5 25 a0 RIP [<ffffffff802bfe55>] proc_register+0xe7/0x10f RSP <ffff810824c57e60> CR2: 000000000000005f ---[ end trace 02c2d78def82877a ]--- Kernel panic - not syncing: Attempted to kill init! it turns out some variables near end of bss are corrupted already. in System.map we have ffffffff80d40420 b rsi_table ffffffff80d40620 B krb5_seq_lock ffffffff80d40628 b i.20437 ffffffff80d40630 b xprt_rdma_inline_write_padding ffffffff80d40638 b sunrpc_table_header ffffffff80d40640 b zero ffffffff80d40644 b min_memreg ffffffff80d40648 b rpcrdma_tk_lock_g ffffffff80d40650 B sctp_assocs_id_lock ffffffff80d40658 B proc_net_sctp ffffffff80d40660 B sctp_assocs_id ffffffff80d40680 B sysctl_sctp_mem ffffffff80d40690 B sysctl_sctp_rmem ffffffff80d406a0 B sysctl_sctp_wmem ffffffff80d406b0 b sctp_ctl_socket ffffffff80d406b8 b sctp_pf_inet6_specific ffffffff80d406c0 b sctp_pf_inet_specific ffffffff80d406c8 b sctp_af_v4_specific ffffffff80d406d0 b sctp_af_v6_specific ffffffff80d406d8 b sctp_rand.33270 ffffffff80d406dc b sctp_memory_pressure ffffffff80d406e0 b sctp_sockets_allocated ffffffff80d406e4 b sctp_memory_allocated ffffffff80d406e8 b sctp_sysctl_header ffffffff80d406f0 b zero ffffffff80d406f4 A __bss_stop ffffffff80d406f4 A _end and setup_node_bootmem() will use that page 0xd40000 for bootmap Bootmem setup node 0 0000000000000000-0000000828000000 NODE_DATA [000000000008a485 - 0000000000091484] bootmap [0000000000d406f4 - 0000000000e456f3] pages 105 Bootmem setup node 1 0000000828000000-0000001028000000 NODE_DATA [0000000828000000 - 0000000828006fff] bootmap [0000000828007000 - 0000000828106fff] pages 100 Bootmem setup node 2 0000001028000000-0000001828000000 NODE_DATA [0000001028000000 - 0000001028006fff] bootmap [0000001028007000 - 0000001028106fff] pages 100 Bootmem setup node 3 0000001828000000-0000002028000000 NODE_DATA [0000001828000000 - 0000001828006fff] bootmap [0000001828007000 - 0000001828106fff] pages 100 setup_node_bootmem() makes NODE_DATA cacheline aligned, and bootmap is page-aligned. the patch updates find_e820_area() to make sure we can meet the alignment constraints. Signed-off-by: Yinghai Lu <yinghai.lu@sun.com> Signed-off-by: Ingo Molnar <mingo@elte.hu>
2008-02-01 16:49:41 +00:00
addr = 0x8000;
nodemap_size = roundup(sizeof(s16) * memnodemapsize, L1_CACHE_BYTES);
nodemap_addr = find_e820_area(addr, max_pfn<<PAGE_SHIFT,
x86_64: make bootmap_start page align v6 boot oopses when a system has 64 or 128 GB of RAM installed: Calling initcall 0xffffffff80bc33b6: sctp_init+0x0/0x711() BUG: unable to handle kernel NULL pointer dereference at 000000000000005f IP: [<ffffffff802bfe55>] proc_register+0xe7/0x10f PGD 0 Oops: 0000 [1] SMP CPU 0 Modules linked in: Pid: 1, comm: swapper Not tainted 2.6.24-smp-g5a514e21-dirty #6 RIP: 0010:[<ffffffff802bfe55>] [<ffffffff802bfe55>] proc_register+0xe7/0x10f RSP: 0000:ffff810824c57e60 EFLAGS: 00010246 RAX: 000000000000d7d7 RBX: ffff811024c5fa80 RCX: ffff810824c57e08 RDX: 0000000000000000 RSI: 0000000000000195 RDI: ffffffff80cc2460 RBP: ffffffffffffffff R08: 0000000000000000 R09: ffff811024c5fa80 R10: 0000000000000000 R11: 0000000000000002 R12: ffff810824c57e6c R13: 0000000000000000 R14: ffff810824c57ee0 R15: 00000006abd25bee FS: 0000000000000000(0000) GS:ffffffff80b4d000(0000) knlGS:0000000000000000 CS: 0010 DS: 0018 ES: 0018 CR0: 000000008005003b CR2: 000000000000005f CR3: 0000000000201000 CR4: 00000000000006e0 DR0: 0000000000000000 DR1: 0000000000000000 DR2: 0000000000000000 DR3: 0000000000000000 DR6: 00000000ffff0ff0 DR7: 0000000000000400 Process swapper (pid: 1, threadinfo ffff810824c56000, task ffff812024c52000) Stack: ffffffff80a57348 0000019500000000 ffff811024c5fa80 0000000000000000 00000000ffffff97 ffffffff802bfef0 0000000000000000 ffffffffffffffff 0000000000000000 ffffffff80bc3b4b ffff810824c57ee0 ffffffff80bc34a5 Call Trace: [<ffffffff802bfef0>] ? create_proc_entry+0x73/0x8a [<ffffffff80bc3b4b>] ? sctp_snmp_proc_init+0x1c/0x34 [<ffffffff80bc34a5>] ? sctp_init+0xef/0x711 [<ffffffff80b976e3>] ? kernel_init+0x175/0x2e1 [<ffffffff8020ccf8>] ? child_rip+0xa/0x12 [<ffffffff80b9756e>] ? kernel_init+0x0/0x2e1 [<ffffffff8020ccee>] ? child_rip+0x0/0x12 Code: 1e 48 83 7b 38 00 75 08 48 c7 43 38 f0 e8 82 80 48 83 7b 30 00 75 08 48 c7 43 30 d0 e9 82 80 48 c7 c7 60 24 cc 80 e8 bd 5a 54 00 <48> 8b 45 60 48 89 6b 58 48 89 5d 60 48 89 43 50 fe 05 f5 25 a0 RIP [<ffffffff802bfe55>] proc_register+0xe7/0x10f RSP <ffff810824c57e60> CR2: 000000000000005f ---[ end trace 02c2d78def82877a ]--- Kernel panic - not syncing: Attempted to kill init! it turns out some variables near end of bss are corrupted already. in System.map we have ffffffff80d40420 b rsi_table ffffffff80d40620 B krb5_seq_lock ffffffff80d40628 b i.20437 ffffffff80d40630 b xprt_rdma_inline_write_padding ffffffff80d40638 b sunrpc_table_header ffffffff80d40640 b zero ffffffff80d40644 b min_memreg ffffffff80d40648 b rpcrdma_tk_lock_g ffffffff80d40650 B sctp_assocs_id_lock ffffffff80d40658 B proc_net_sctp ffffffff80d40660 B sctp_assocs_id ffffffff80d40680 B sysctl_sctp_mem ffffffff80d40690 B sysctl_sctp_rmem ffffffff80d406a0 B sysctl_sctp_wmem ffffffff80d406b0 b sctp_ctl_socket ffffffff80d406b8 b sctp_pf_inet6_specific ffffffff80d406c0 b sctp_pf_inet_specific ffffffff80d406c8 b sctp_af_v4_specific ffffffff80d406d0 b sctp_af_v6_specific ffffffff80d406d8 b sctp_rand.33270 ffffffff80d406dc b sctp_memory_pressure ffffffff80d406e0 b sctp_sockets_allocated ffffffff80d406e4 b sctp_memory_allocated ffffffff80d406e8 b sctp_sysctl_header ffffffff80d406f0 b zero ffffffff80d406f4 A __bss_stop ffffffff80d406f4 A _end and setup_node_bootmem() will use that page 0xd40000 for bootmap Bootmem setup node 0 0000000000000000-0000000828000000 NODE_DATA [000000000008a485 - 0000000000091484] bootmap [0000000000d406f4 - 0000000000e456f3] pages 105 Bootmem setup node 1 0000000828000000-0000001028000000 NODE_DATA [0000000828000000 - 0000000828006fff] bootmap [0000000828007000 - 0000000828106fff] pages 100 Bootmem setup node 2 0000001028000000-0000001828000000 NODE_DATA [0000001028000000 - 0000001028006fff] bootmap [0000001028007000 - 0000001028106fff] pages 100 Bootmem setup node 3 0000001828000000-0000002028000000 NODE_DATA [0000001828000000 - 0000001828006fff] bootmap [0000001828007000 - 0000001828106fff] pages 100 setup_node_bootmem() makes NODE_DATA cacheline aligned, and bootmap is page-aligned. the patch updates find_e820_area() to make sure we can meet the alignment constraints. Signed-off-by: Yinghai Lu <yinghai.lu@sun.com> Signed-off-by: Ingo Molnar <mingo@elte.hu>
2008-02-01 16:49:41 +00:00
nodemap_size, L1_CACHE_BYTES);
if (nodemap_addr == -1UL) {
printk(KERN_ERR
"NUMA: Unable to allocate Memory to Node hash map\n");
nodemap_addr = nodemap_size = 0;
return -1;
}
x86_64: make bootmap_start page align v6 boot oopses when a system has 64 or 128 GB of RAM installed: Calling initcall 0xffffffff80bc33b6: sctp_init+0x0/0x711() BUG: unable to handle kernel NULL pointer dereference at 000000000000005f IP: [<ffffffff802bfe55>] proc_register+0xe7/0x10f PGD 0 Oops: 0000 [1] SMP CPU 0 Modules linked in: Pid: 1, comm: swapper Not tainted 2.6.24-smp-g5a514e21-dirty #6 RIP: 0010:[<ffffffff802bfe55>] [<ffffffff802bfe55>] proc_register+0xe7/0x10f RSP: 0000:ffff810824c57e60 EFLAGS: 00010246 RAX: 000000000000d7d7 RBX: ffff811024c5fa80 RCX: ffff810824c57e08 RDX: 0000000000000000 RSI: 0000000000000195 RDI: ffffffff80cc2460 RBP: ffffffffffffffff R08: 0000000000000000 R09: ffff811024c5fa80 R10: 0000000000000000 R11: 0000000000000002 R12: ffff810824c57e6c R13: 0000000000000000 R14: ffff810824c57ee0 R15: 00000006abd25bee FS: 0000000000000000(0000) GS:ffffffff80b4d000(0000) knlGS:0000000000000000 CS: 0010 DS: 0018 ES: 0018 CR0: 000000008005003b CR2: 000000000000005f CR3: 0000000000201000 CR4: 00000000000006e0 DR0: 0000000000000000 DR1: 0000000000000000 DR2: 0000000000000000 DR3: 0000000000000000 DR6: 00000000ffff0ff0 DR7: 0000000000000400 Process swapper (pid: 1, threadinfo ffff810824c56000, task ffff812024c52000) Stack: ffffffff80a57348 0000019500000000 ffff811024c5fa80 0000000000000000 00000000ffffff97 ffffffff802bfef0 0000000000000000 ffffffffffffffff 0000000000000000 ffffffff80bc3b4b ffff810824c57ee0 ffffffff80bc34a5 Call Trace: [<ffffffff802bfef0>] ? create_proc_entry+0x73/0x8a [<ffffffff80bc3b4b>] ? sctp_snmp_proc_init+0x1c/0x34 [<ffffffff80bc34a5>] ? sctp_init+0xef/0x711 [<ffffffff80b976e3>] ? kernel_init+0x175/0x2e1 [<ffffffff8020ccf8>] ? child_rip+0xa/0x12 [<ffffffff80b9756e>] ? kernel_init+0x0/0x2e1 [<ffffffff8020ccee>] ? child_rip+0x0/0x12 Code: 1e 48 83 7b 38 00 75 08 48 c7 43 38 f0 e8 82 80 48 83 7b 30 00 75 08 48 c7 43 30 d0 e9 82 80 48 c7 c7 60 24 cc 80 e8 bd 5a 54 00 <48> 8b 45 60 48 89 6b 58 48 89 5d 60 48 89 43 50 fe 05 f5 25 a0 RIP [<ffffffff802bfe55>] proc_register+0xe7/0x10f RSP <ffff810824c57e60> CR2: 000000000000005f ---[ end trace 02c2d78def82877a ]--- Kernel panic - not syncing: Attempted to kill init! it turns out some variables near end of bss are corrupted already. in System.map we have ffffffff80d40420 b rsi_table ffffffff80d40620 B krb5_seq_lock ffffffff80d40628 b i.20437 ffffffff80d40630 b xprt_rdma_inline_write_padding ffffffff80d40638 b sunrpc_table_header ffffffff80d40640 b zero ffffffff80d40644 b min_memreg ffffffff80d40648 b rpcrdma_tk_lock_g ffffffff80d40650 B sctp_assocs_id_lock ffffffff80d40658 B proc_net_sctp ffffffff80d40660 B sctp_assocs_id ffffffff80d40680 B sysctl_sctp_mem ffffffff80d40690 B sysctl_sctp_rmem ffffffff80d406a0 B sysctl_sctp_wmem ffffffff80d406b0 b sctp_ctl_socket ffffffff80d406b8 b sctp_pf_inet6_specific ffffffff80d406c0 b sctp_pf_inet_specific ffffffff80d406c8 b sctp_af_v4_specific ffffffff80d406d0 b sctp_af_v6_specific ffffffff80d406d8 b sctp_rand.33270 ffffffff80d406dc b sctp_memory_pressure ffffffff80d406e0 b sctp_sockets_allocated ffffffff80d406e4 b sctp_memory_allocated ffffffff80d406e8 b sctp_sysctl_header ffffffff80d406f0 b zero ffffffff80d406f4 A __bss_stop ffffffff80d406f4 A _end and setup_node_bootmem() will use that page 0xd40000 for bootmap Bootmem setup node 0 0000000000000000-0000000828000000 NODE_DATA [000000000008a485 - 0000000000091484] bootmap [0000000000d406f4 - 0000000000e456f3] pages 105 Bootmem setup node 1 0000000828000000-0000001028000000 NODE_DATA [0000000828000000 - 0000000828006fff] bootmap [0000000828007000 - 0000000828106fff] pages 100 Bootmem setup node 2 0000001028000000-0000001828000000 NODE_DATA [0000001028000000 - 0000001028006fff] bootmap [0000001028007000 - 0000001028106fff] pages 100 Bootmem setup node 3 0000001828000000-0000002028000000 NODE_DATA [0000001828000000 - 0000001828006fff] bootmap [0000001828007000 - 0000001828106fff] pages 100 setup_node_bootmem() makes NODE_DATA cacheline aligned, and bootmap is page-aligned. the patch updates find_e820_area() to make sure we can meet the alignment constraints. Signed-off-by: Yinghai Lu <yinghai.lu@sun.com> Signed-off-by: Ingo Molnar <mingo@elte.hu>
2008-02-01 16:49:41 +00:00
memnodemap = phys_to_virt(nodemap_addr);
reserve_early(nodemap_addr, nodemap_addr + nodemap_size, "MEMNODEMAP");
printk(KERN_DEBUG "NUMA: Allocated memnodemap from %lx - %lx\n",
nodemap_addr, nodemap_addr + nodemap_size);
return 0;
}
/*
* The LSB of all start and end addresses in the node map is the value of the
* maximum possible shift.
*/
static int __init extract_lsb_from_nodes(const struct bootnode *nodes,
int numnodes)
{
int i, nodes_used = 0;
unsigned long start, end;
unsigned long bitfield = 0, memtop = 0;
for (i = 0; i < numnodes; i++) {
start = nodes[i].start;
end = nodes[i].end;
if (start >= end)
continue;
bitfield |= start;
nodes_used++;
if (end > memtop)
memtop = end;
}
if (nodes_used <= 1)
i = 63;
else
i = find_first_bit(&bitfield, sizeof(unsigned long)*8);
memnodemapsize = (memtop >> i)+1;
return i;
}
int __init compute_hash_shift(struct bootnode *nodes, int numnodes,
int *nodeids)
{
int shift;
shift = extract_lsb_from_nodes(nodes, numnodes);
if (allocate_cachealigned_memnodemap())
return -1;
printk(KERN_DEBUG "NUMA: Using %d for the hash shift.\n",
shift);
if (populate_memnodemap(nodes, numnodes, shift, nodeids) != 1) {
printk(KERN_INFO "Your memory is not aligned you need to "
"rebuild your kernel with a bigger NODEMAPSIZE "
"shift=%d\n", shift);
return -1;
}
return shift;
}
mm: clean up for early_pfn_to_nid() What's happening is that the assertion in mm/page_alloc.c:move_freepages() is triggering: BUG_ON(page_zone(start_page) != page_zone(end_page)); Once I knew this is what was happening, I added some annotations: if (unlikely(page_zone(start_page) != page_zone(end_page))) { printk(KERN_ERR "move_freepages: Bogus zones: " "start_page[%p] end_page[%p] zone[%p]\n", start_page, end_page, zone); printk(KERN_ERR "move_freepages: " "start_zone[%p] end_zone[%p]\n", page_zone(start_page), page_zone(end_page)); printk(KERN_ERR "move_freepages: " "start_pfn[0x%lx] end_pfn[0x%lx]\n", page_to_pfn(start_page), page_to_pfn(end_page)); printk(KERN_ERR "move_freepages: " "start_nid[%d] end_nid[%d]\n", page_to_nid(start_page), page_to_nid(end_page)); ... And here's what I got: move_freepages: Bogus zones: start_page[2207d0000] end_page[2207dffc0] zone[fffff8103effcb00] move_freepages: start_zone[fffff8103effcb00] end_zone[fffff8003fffeb00] move_freepages: start_pfn[0x81f600] end_pfn[0x81f7ff] move_freepages: start_nid[1] end_nid[0] My memory layout on this box is: [ 0.000000] Zone PFN ranges: [ 0.000000] Normal 0x00000000 -> 0x0081ff5d [ 0.000000] Movable zone start PFN for each node [ 0.000000] early_node_map[8] active PFN ranges [ 0.000000] 0: 0x00000000 -> 0x00020000 [ 0.000000] 1: 0x00800000 -> 0x0081f7ff [ 0.000000] 1: 0x0081f800 -> 0x0081fe50 [ 0.000000] 1: 0x0081fed1 -> 0x0081fed8 [ 0.000000] 1: 0x0081feda -> 0x0081fedb [ 0.000000] 1: 0x0081fedd -> 0x0081fee5 [ 0.000000] 1: 0x0081fee7 -> 0x0081ff51 [ 0.000000] 1: 0x0081ff59 -> 0x0081ff5d So it's a block move in that 0x81f600-->0x81f7ff region which triggers the problem. This patch: Declaration of early_pfn_to_nid() is scattered over per-arch include files, and it seems it's complicated to know when the declaration is used. I think it makes fix-for-memmap-init not easy. This patch moves all declaration to include/linux/mm.h After this, if !CONFIG_NODES_POPULATES_NODE_MAP && !CONFIG_HAVE_ARCH_EARLY_PFN_TO_NID -> Use static definition in include/linux/mm.h else if !CONFIG_HAVE_ARCH_EARLY_PFN_TO_NID -> Use generic definition in mm/page_alloc.c else -> per-arch back end function will be called. Signed-off-by: KAMEZAWA Hiroyuki <kamezawa.hiroyu@jp.fujitsu.com> Tested-by: KOSAKI Motohiro <kosaki.motohiro@jp.fujitsu.com> Reported-by: David Miller <davem@davemlloft.net> Cc: Mel Gorman <mel@csn.ul.ie> Cc: Heiko Carstens <heiko.carstens@de.ibm.com> Cc: <stable@kernel.org> [2.6.25.x, 2.6.26.x, 2.6.27.x, 2.6.28.x] Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2009-02-18 22:48:32 +00:00
int __meminit __early_pfn_to_nid(unsigned long pfn)
{
return phys_to_nid(pfn << PAGE_SHIFT);
}
static void * __init early_node_mem(int nodeid, unsigned long start,
x86_64: make bootmap_start page align v6 boot oopses when a system has 64 or 128 GB of RAM installed: Calling initcall 0xffffffff80bc33b6: sctp_init+0x0/0x711() BUG: unable to handle kernel NULL pointer dereference at 000000000000005f IP: [<ffffffff802bfe55>] proc_register+0xe7/0x10f PGD 0 Oops: 0000 [1] SMP CPU 0 Modules linked in: Pid: 1, comm: swapper Not tainted 2.6.24-smp-g5a514e21-dirty #6 RIP: 0010:[<ffffffff802bfe55>] [<ffffffff802bfe55>] proc_register+0xe7/0x10f RSP: 0000:ffff810824c57e60 EFLAGS: 00010246 RAX: 000000000000d7d7 RBX: ffff811024c5fa80 RCX: ffff810824c57e08 RDX: 0000000000000000 RSI: 0000000000000195 RDI: ffffffff80cc2460 RBP: ffffffffffffffff R08: 0000000000000000 R09: ffff811024c5fa80 R10: 0000000000000000 R11: 0000000000000002 R12: ffff810824c57e6c R13: 0000000000000000 R14: ffff810824c57ee0 R15: 00000006abd25bee FS: 0000000000000000(0000) GS:ffffffff80b4d000(0000) knlGS:0000000000000000 CS: 0010 DS: 0018 ES: 0018 CR0: 000000008005003b CR2: 000000000000005f CR3: 0000000000201000 CR4: 00000000000006e0 DR0: 0000000000000000 DR1: 0000000000000000 DR2: 0000000000000000 DR3: 0000000000000000 DR6: 00000000ffff0ff0 DR7: 0000000000000400 Process swapper (pid: 1, threadinfo ffff810824c56000, task ffff812024c52000) Stack: ffffffff80a57348 0000019500000000 ffff811024c5fa80 0000000000000000 00000000ffffff97 ffffffff802bfef0 0000000000000000 ffffffffffffffff 0000000000000000 ffffffff80bc3b4b ffff810824c57ee0 ffffffff80bc34a5 Call Trace: [<ffffffff802bfef0>] ? create_proc_entry+0x73/0x8a [<ffffffff80bc3b4b>] ? sctp_snmp_proc_init+0x1c/0x34 [<ffffffff80bc34a5>] ? sctp_init+0xef/0x711 [<ffffffff80b976e3>] ? kernel_init+0x175/0x2e1 [<ffffffff8020ccf8>] ? child_rip+0xa/0x12 [<ffffffff80b9756e>] ? kernel_init+0x0/0x2e1 [<ffffffff8020ccee>] ? child_rip+0x0/0x12 Code: 1e 48 83 7b 38 00 75 08 48 c7 43 38 f0 e8 82 80 48 83 7b 30 00 75 08 48 c7 43 30 d0 e9 82 80 48 c7 c7 60 24 cc 80 e8 bd 5a 54 00 <48> 8b 45 60 48 89 6b 58 48 89 5d 60 48 89 43 50 fe 05 f5 25 a0 RIP [<ffffffff802bfe55>] proc_register+0xe7/0x10f RSP <ffff810824c57e60> CR2: 000000000000005f ---[ end trace 02c2d78def82877a ]--- Kernel panic - not syncing: Attempted to kill init! it turns out some variables near end of bss are corrupted already. in System.map we have ffffffff80d40420 b rsi_table ffffffff80d40620 B krb5_seq_lock ffffffff80d40628 b i.20437 ffffffff80d40630 b xprt_rdma_inline_write_padding ffffffff80d40638 b sunrpc_table_header ffffffff80d40640 b zero ffffffff80d40644 b min_memreg ffffffff80d40648 b rpcrdma_tk_lock_g ffffffff80d40650 B sctp_assocs_id_lock ffffffff80d40658 B proc_net_sctp ffffffff80d40660 B sctp_assocs_id ffffffff80d40680 B sysctl_sctp_mem ffffffff80d40690 B sysctl_sctp_rmem ffffffff80d406a0 B sysctl_sctp_wmem ffffffff80d406b0 b sctp_ctl_socket ffffffff80d406b8 b sctp_pf_inet6_specific ffffffff80d406c0 b sctp_pf_inet_specific ffffffff80d406c8 b sctp_af_v4_specific ffffffff80d406d0 b sctp_af_v6_specific ffffffff80d406d8 b sctp_rand.33270 ffffffff80d406dc b sctp_memory_pressure ffffffff80d406e0 b sctp_sockets_allocated ffffffff80d406e4 b sctp_memory_allocated ffffffff80d406e8 b sctp_sysctl_header ffffffff80d406f0 b zero ffffffff80d406f4 A __bss_stop ffffffff80d406f4 A _end and setup_node_bootmem() will use that page 0xd40000 for bootmap Bootmem setup node 0 0000000000000000-0000000828000000 NODE_DATA [000000000008a485 - 0000000000091484] bootmap [0000000000d406f4 - 0000000000e456f3] pages 105 Bootmem setup node 1 0000000828000000-0000001028000000 NODE_DATA [0000000828000000 - 0000000828006fff] bootmap [0000000828007000 - 0000000828106fff] pages 100 Bootmem setup node 2 0000001028000000-0000001828000000 NODE_DATA [0000001028000000 - 0000001028006fff] bootmap [0000001028007000 - 0000001028106fff] pages 100 Bootmem setup node 3 0000001828000000-0000002028000000 NODE_DATA [0000001828000000 - 0000001828006fff] bootmap [0000001828007000 - 0000001828106fff] pages 100 setup_node_bootmem() makes NODE_DATA cacheline aligned, and bootmap is page-aligned. the patch updates find_e820_area() to make sure we can meet the alignment constraints. Signed-off-by: Yinghai Lu <yinghai.lu@sun.com> Signed-off-by: Ingo Molnar <mingo@elte.hu>
2008-02-01 16:49:41 +00:00
unsigned long end, unsigned long size,
unsigned long align)
{
x86: Make early_node_mem get mem > 4 GB if possible So we could put pgdata for the node high, and later sparse vmmap will get the section nr that need. With this patch will make <4 GB ram not use a sparse vmmap. before this patch, will get, before swiotlb try get bootmem [ 0.000000] nid=1 start=0 end=2080000 aligned=1 [ 0.000000] free [10 - 96] [ 0.000000] free [b12 - 1000] [ 0.000000] free [359f - 38a3] [ 0.000000] free [38b5 - 3a00] [ 0.000000] free [41e01 - 42000] [ 0.000000] free [73dde - 73e00] [ 0.000000] free [73fdd - 74000] [ 0.000000] free [741dd - 74200] [ 0.000000] free [743dd - 74400] [ 0.000000] free [745dd - 74600] [ 0.000000] free [747dd - 74800] [ 0.000000] free [749dd - 74a00] [ 0.000000] free [74bdd - 74c00] [ 0.000000] free [74ddd - 74e00] [ 0.000000] free [74fdd - 75000] [ 0.000000] free [751dd - 75200] [ 0.000000] free [753dd - 75400] [ 0.000000] free [755dd - 75600] [ 0.000000] free [757dd - 75800] [ 0.000000] free [759dd - 75a00] [ 0.000000] free [75bdd - 7bf5f] [ 0.000000] free [7f730 - 7f750] [ 0.000000] free [100000 - 2080000] [ 0.000000] total free 1f87170 [ 93.301474] Placing 64MB software IO TLB between ffff880075bdd000 - ffff880079bdd000 [ 93.311814] software IO TLB at phys 0x75bdd000 - 0x79bdd000 with this patch will get: before swiotlb try get bootmem [ 0.000000] nid=1 start=0 end=2080000 aligned=1 [ 0.000000] free [a - 96] [ 0.000000] free [702 - 1000] [ 0.000000] free [359f - 3600] [ 0.000000] free [37de - 3800] [ 0.000000] free [39dd - 3a00] [ 0.000000] free [3bdd - 3c00] [ 0.000000] free [3ddd - 3e00] [ 0.000000] free [3fdd - 4000] [ 0.000000] free [41dd - 4200] [ 0.000000] free [43dd - 4400] [ 0.000000] free [45dd - 4600] [ 0.000000] free [47dd - 4800] [ 0.000000] free [49dd - 4a00] [ 0.000000] free [4bdd - 4c00] [ 0.000000] free [4ddd - 4e00] [ 0.000000] free [4fdd - 5000] [ 0.000000] free [51dd - 5200] [ 0.000000] free [53dd - 5400] [ 0.000000] free [55dd - 7bf5f] [ 0.000000] free [7f730 - 7f750] [ 0.000000] free [100428 - 100600] [ 0.000000] free [13ea01 - 13ec00] [ 0.000000] free [170800 - 2080000] [ 0.000000] total free 1f87170 [ 92.689485] PCI-DMA: Using software bounce buffering for IO (SWIOTLB) [ 92.699799] Placing 64MB software IO TLB between ffff8800055dd000 - ffff8800095dd000 [ 92.710916] software IO TLB at phys 0x55dd000 - 0x95dd000 so will get enough space below 4G, aka pfn 0x100000 Signed-off-by: Yinghai Lu <yinghai@kernel.org> LKML-Reference: <1265793639-15071-15-git-send-email-yinghai@kernel.org> Signed-off-by: H. Peter Anvin <hpa@zytor.com>
2010-02-10 09:20:18 +00:00
unsigned long mem;
x86: Make early_node_mem get mem > 4 GB if possible So we could put pgdata for the node high, and later sparse vmmap will get the section nr that need. With this patch will make <4 GB ram not use a sparse vmmap. before this patch, will get, before swiotlb try get bootmem [ 0.000000] nid=1 start=0 end=2080000 aligned=1 [ 0.000000] free [10 - 96] [ 0.000000] free [b12 - 1000] [ 0.000000] free [359f - 38a3] [ 0.000000] free [38b5 - 3a00] [ 0.000000] free [41e01 - 42000] [ 0.000000] free [73dde - 73e00] [ 0.000000] free [73fdd - 74000] [ 0.000000] free [741dd - 74200] [ 0.000000] free [743dd - 74400] [ 0.000000] free [745dd - 74600] [ 0.000000] free [747dd - 74800] [ 0.000000] free [749dd - 74a00] [ 0.000000] free [74bdd - 74c00] [ 0.000000] free [74ddd - 74e00] [ 0.000000] free [74fdd - 75000] [ 0.000000] free [751dd - 75200] [ 0.000000] free [753dd - 75400] [ 0.000000] free [755dd - 75600] [ 0.000000] free [757dd - 75800] [ 0.000000] free [759dd - 75a00] [ 0.000000] free [75bdd - 7bf5f] [ 0.000000] free [7f730 - 7f750] [ 0.000000] free [100000 - 2080000] [ 0.000000] total free 1f87170 [ 93.301474] Placing 64MB software IO TLB between ffff880075bdd000 - ffff880079bdd000 [ 93.311814] software IO TLB at phys 0x75bdd000 - 0x79bdd000 with this patch will get: before swiotlb try get bootmem [ 0.000000] nid=1 start=0 end=2080000 aligned=1 [ 0.000000] free [a - 96] [ 0.000000] free [702 - 1000] [ 0.000000] free [359f - 3600] [ 0.000000] free [37de - 3800] [ 0.000000] free [39dd - 3a00] [ 0.000000] free [3bdd - 3c00] [ 0.000000] free [3ddd - 3e00] [ 0.000000] free [3fdd - 4000] [ 0.000000] free [41dd - 4200] [ 0.000000] free [43dd - 4400] [ 0.000000] free [45dd - 4600] [ 0.000000] free [47dd - 4800] [ 0.000000] free [49dd - 4a00] [ 0.000000] free [4bdd - 4c00] [ 0.000000] free [4ddd - 4e00] [ 0.000000] free [4fdd - 5000] [ 0.000000] free [51dd - 5200] [ 0.000000] free [53dd - 5400] [ 0.000000] free [55dd - 7bf5f] [ 0.000000] free [7f730 - 7f750] [ 0.000000] free [100428 - 100600] [ 0.000000] free [13ea01 - 13ec00] [ 0.000000] free [170800 - 2080000] [ 0.000000] total free 1f87170 [ 92.689485] PCI-DMA: Using software bounce buffering for IO (SWIOTLB) [ 92.699799] Placing 64MB software IO TLB between ffff8800055dd000 - ffff8800095dd000 [ 92.710916] software IO TLB at phys 0x55dd000 - 0x95dd000 so will get enough space below 4G, aka pfn 0x100000 Signed-off-by: Yinghai Lu <yinghai@kernel.org> LKML-Reference: <1265793639-15071-15-git-send-email-yinghai@kernel.org> Signed-off-by: H. Peter Anvin <hpa@zytor.com>
2010-02-10 09:20:18 +00:00
/*
* put it on high as possible
* something will go with NODE_DATA
*/
if (start < (MAX_DMA_PFN<<PAGE_SHIFT))
start = MAX_DMA_PFN<<PAGE_SHIFT;
if (start < (MAX_DMA32_PFN<<PAGE_SHIFT) &&
end > (MAX_DMA32_PFN<<PAGE_SHIFT))
start = MAX_DMA32_PFN<<PAGE_SHIFT;
mem = find_e820_area(start, end, size, align);
if (mem != -1L)
return __va(mem);
x86: Make early_node_mem get mem > 4 GB if possible So we could put pgdata for the node high, and later sparse vmmap will get the section nr that need. With this patch will make <4 GB ram not use a sparse vmmap. before this patch, will get, before swiotlb try get bootmem [ 0.000000] nid=1 start=0 end=2080000 aligned=1 [ 0.000000] free [10 - 96] [ 0.000000] free [b12 - 1000] [ 0.000000] free [359f - 38a3] [ 0.000000] free [38b5 - 3a00] [ 0.000000] free [41e01 - 42000] [ 0.000000] free [73dde - 73e00] [ 0.000000] free [73fdd - 74000] [ 0.000000] free [741dd - 74200] [ 0.000000] free [743dd - 74400] [ 0.000000] free [745dd - 74600] [ 0.000000] free [747dd - 74800] [ 0.000000] free [749dd - 74a00] [ 0.000000] free [74bdd - 74c00] [ 0.000000] free [74ddd - 74e00] [ 0.000000] free [74fdd - 75000] [ 0.000000] free [751dd - 75200] [ 0.000000] free [753dd - 75400] [ 0.000000] free [755dd - 75600] [ 0.000000] free [757dd - 75800] [ 0.000000] free [759dd - 75a00] [ 0.000000] free [75bdd - 7bf5f] [ 0.000000] free [7f730 - 7f750] [ 0.000000] free [100000 - 2080000] [ 0.000000] total free 1f87170 [ 93.301474] Placing 64MB software IO TLB between ffff880075bdd000 - ffff880079bdd000 [ 93.311814] software IO TLB at phys 0x75bdd000 - 0x79bdd000 with this patch will get: before swiotlb try get bootmem [ 0.000000] nid=1 start=0 end=2080000 aligned=1 [ 0.000000] free [a - 96] [ 0.000000] free [702 - 1000] [ 0.000000] free [359f - 3600] [ 0.000000] free [37de - 3800] [ 0.000000] free [39dd - 3a00] [ 0.000000] free [3bdd - 3c00] [ 0.000000] free [3ddd - 3e00] [ 0.000000] free [3fdd - 4000] [ 0.000000] free [41dd - 4200] [ 0.000000] free [43dd - 4400] [ 0.000000] free [45dd - 4600] [ 0.000000] free [47dd - 4800] [ 0.000000] free [49dd - 4a00] [ 0.000000] free [4bdd - 4c00] [ 0.000000] free [4ddd - 4e00] [ 0.000000] free [4fdd - 5000] [ 0.000000] free [51dd - 5200] [ 0.000000] free [53dd - 5400] [ 0.000000] free [55dd - 7bf5f] [ 0.000000] free [7f730 - 7f750] [ 0.000000] free [100428 - 100600] [ 0.000000] free [13ea01 - 13ec00] [ 0.000000] free [170800 - 2080000] [ 0.000000] total free 1f87170 [ 92.689485] PCI-DMA: Using software bounce buffering for IO (SWIOTLB) [ 92.699799] Placing 64MB software IO TLB between ffff8800055dd000 - ffff8800095dd000 [ 92.710916] software IO TLB at phys 0x55dd000 - 0x95dd000 so will get enough space below 4G, aka pfn 0x100000 Signed-off-by: Yinghai Lu <yinghai@kernel.org> LKML-Reference: <1265793639-15071-15-git-send-email-yinghai@kernel.org> Signed-off-by: H. Peter Anvin <hpa@zytor.com>
2010-02-10 09:20:18 +00:00
/* extend the search scope */
end = max_pfn_mapped << PAGE_SHIFT;
if (end > (MAX_DMA32_PFN<<PAGE_SHIFT))
start = MAX_DMA32_PFN<<PAGE_SHIFT;
else
start = MAX_DMA_PFN<<PAGE_SHIFT;
mem = find_e820_area(start, end, size, align);
if (mem != -1L)
return __va(mem);
printk(KERN_ERR "Cannot find %lu bytes in node %d\n",
size, nodeid);
return NULL;
}
/* Initialize bootmem allocator for a node */
void __init
setup_node_bootmem(int nodeid, unsigned long start, unsigned long end)
{
unsigned long start_pfn, last_pfn, nodedata_phys;
const int pgdat_size = roundup(sizeof(pg_data_t), PAGE_SIZE);
int nid;
#ifndef CONFIG_NO_BOOTMEM
unsigned long bootmap_start, bootmap_pages, bootmap_size;
void *bootmap;
#endif
if (!end)
return;
/*
* Don't confuse VM with a node that doesn't have the
* minimum amount of memory:
*/
if (end && (end - start) < NODE_MIN_SIZE)
return;
start = roundup(start, ZONE_ALIGN);
printk(KERN_INFO "Initmem setup node %d %016lx-%016lx\n", nodeid,
start, end);
start_pfn = start >> PAGE_SHIFT;
last_pfn = end >> PAGE_SHIFT;
x86_64: make bootmap_start page align v6 boot oopses when a system has 64 or 128 GB of RAM installed: Calling initcall 0xffffffff80bc33b6: sctp_init+0x0/0x711() BUG: unable to handle kernel NULL pointer dereference at 000000000000005f IP: [<ffffffff802bfe55>] proc_register+0xe7/0x10f PGD 0 Oops: 0000 [1] SMP CPU 0 Modules linked in: Pid: 1, comm: swapper Not tainted 2.6.24-smp-g5a514e21-dirty #6 RIP: 0010:[<ffffffff802bfe55>] [<ffffffff802bfe55>] proc_register+0xe7/0x10f RSP: 0000:ffff810824c57e60 EFLAGS: 00010246 RAX: 000000000000d7d7 RBX: ffff811024c5fa80 RCX: ffff810824c57e08 RDX: 0000000000000000 RSI: 0000000000000195 RDI: ffffffff80cc2460 RBP: ffffffffffffffff R08: 0000000000000000 R09: ffff811024c5fa80 R10: 0000000000000000 R11: 0000000000000002 R12: ffff810824c57e6c R13: 0000000000000000 R14: ffff810824c57ee0 R15: 00000006abd25bee FS: 0000000000000000(0000) GS:ffffffff80b4d000(0000) knlGS:0000000000000000 CS: 0010 DS: 0018 ES: 0018 CR0: 000000008005003b CR2: 000000000000005f CR3: 0000000000201000 CR4: 00000000000006e0 DR0: 0000000000000000 DR1: 0000000000000000 DR2: 0000000000000000 DR3: 0000000000000000 DR6: 00000000ffff0ff0 DR7: 0000000000000400 Process swapper (pid: 1, threadinfo ffff810824c56000, task ffff812024c52000) Stack: ffffffff80a57348 0000019500000000 ffff811024c5fa80 0000000000000000 00000000ffffff97 ffffffff802bfef0 0000000000000000 ffffffffffffffff 0000000000000000 ffffffff80bc3b4b ffff810824c57ee0 ffffffff80bc34a5 Call Trace: [<ffffffff802bfef0>] ? create_proc_entry+0x73/0x8a [<ffffffff80bc3b4b>] ? sctp_snmp_proc_init+0x1c/0x34 [<ffffffff80bc34a5>] ? sctp_init+0xef/0x711 [<ffffffff80b976e3>] ? kernel_init+0x175/0x2e1 [<ffffffff8020ccf8>] ? child_rip+0xa/0x12 [<ffffffff80b9756e>] ? kernel_init+0x0/0x2e1 [<ffffffff8020ccee>] ? child_rip+0x0/0x12 Code: 1e 48 83 7b 38 00 75 08 48 c7 43 38 f0 e8 82 80 48 83 7b 30 00 75 08 48 c7 43 30 d0 e9 82 80 48 c7 c7 60 24 cc 80 e8 bd 5a 54 00 <48> 8b 45 60 48 89 6b 58 48 89 5d 60 48 89 43 50 fe 05 f5 25 a0 RIP [<ffffffff802bfe55>] proc_register+0xe7/0x10f RSP <ffff810824c57e60> CR2: 000000000000005f ---[ end trace 02c2d78def82877a ]--- Kernel panic - not syncing: Attempted to kill init! it turns out some variables near end of bss are corrupted already. in System.map we have ffffffff80d40420 b rsi_table ffffffff80d40620 B krb5_seq_lock ffffffff80d40628 b i.20437 ffffffff80d40630 b xprt_rdma_inline_write_padding ffffffff80d40638 b sunrpc_table_header ffffffff80d40640 b zero ffffffff80d40644 b min_memreg ffffffff80d40648 b rpcrdma_tk_lock_g ffffffff80d40650 B sctp_assocs_id_lock ffffffff80d40658 B proc_net_sctp ffffffff80d40660 B sctp_assocs_id ffffffff80d40680 B sysctl_sctp_mem ffffffff80d40690 B sysctl_sctp_rmem ffffffff80d406a0 B sysctl_sctp_wmem ffffffff80d406b0 b sctp_ctl_socket ffffffff80d406b8 b sctp_pf_inet6_specific ffffffff80d406c0 b sctp_pf_inet_specific ffffffff80d406c8 b sctp_af_v4_specific ffffffff80d406d0 b sctp_af_v6_specific ffffffff80d406d8 b sctp_rand.33270 ffffffff80d406dc b sctp_memory_pressure ffffffff80d406e0 b sctp_sockets_allocated ffffffff80d406e4 b sctp_memory_allocated ffffffff80d406e8 b sctp_sysctl_header ffffffff80d406f0 b zero ffffffff80d406f4 A __bss_stop ffffffff80d406f4 A _end and setup_node_bootmem() will use that page 0xd40000 for bootmap Bootmem setup node 0 0000000000000000-0000000828000000 NODE_DATA [000000000008a485 - 0000000000091484] bootmap [0000000000d406f4 - 0000000000e456f3] pages 105 Bootmem setup node 1 0000000828000000-0000001028000000 NODE_DATA [0000000828000000 - 0000000828006fff] bootmap [0000000828007000 - 0000000828106fff] pages 100 Bootmem setup node 2 0000001028000000-0000001828000000 NODE_DATA [0000001028000000 - 0000001028006fff] bootmap [0000001028007000 - 0000001028106fff] pages 100 Bootmem setup node 3 0000001828000000-0000002028000000 NODE_DATA [0000001828000000 - 0000001828006fff] bootmap [0000001828007000 - 0000001828106fff] pages 100 setup_node_bootmem() makes NODE_DATA cacheline aligned, and bootmap is page-aligned. the patch updates find_e820_area() to make sure we can meet the alignment constraints. Signed-off-by: Yinghai Lu <yinghai.lu@sun.com> Signed-off-by: Ingo Molnar <mingo@elte.hu>
2008-02-01 16:49:41 +00:00
node_data[nodeid] = early_node_mem(nodeid, start, end, pgdat_size,
SMP_CACHE_BYTES);
if (node_data[nodeid] == NULL)
return;
nodedata_phys = __pa(node_data[nodeid]);
reserve_early(nodedata_phys, nodedata_phys + pgdat_size, "NODE_DATA");
printk(KERN_INFO " NODE_DATA [%016lx - %016lx]\n", nodedata_phys,
nodedata_phys + pgdat_size - 1);
nid = phys_to_nid(nodedata_phys);
if (nid != nodeid)
printk(KERN_INFO " NODE_DATA(%d) on node %d\n", nodeid, nid);
memset(NODE_DATA(nodeid), 0, sizeof(pg_data_t));
NODE_DATA(nodeid)->node_id = nodeid;
NODE_DATA(nodeid)->node_start_pfn = start_pfn;
NODE_DATA(nodeid)->node_spanned_pages = last_pfn - start_pfn;
#ifndef CONFIG_NO_BOOTMEM
NODE_DATA(nodeid)->bdata = &bootmem_node_data[nodeid];
/*
* Find a place for the bootmem map
* nodedata_phys could be on other nodes by alloc_bootmem,
* so need to sure bootmap_start not to be small, otherwise
* early_node_mem will get that with find_e820_area instead
* of alloc_bootmem, that could clash with reserved range
*/
bootmap_pages = bootmem_bootmap_pages(last_pfn - start_pfn);
bootmap_start = roundup(nodedata_phys + pgdat_size, PAGE_SIZE);
x86_64: make bootmap_start page align v6 boot oopses when a system has 64 or 128 GB of RAM installed: Calling initcall 0xffffffff80bc33b6: sctp_init+0x0/0x711() BUG: unable to handle kernel NULL pointer dereference at 000000000000005f IP: [<ffffffff802bfe55>] proc_register+0xe7/0x10f PGD 0 Oops: 0000 [1] SMP CPU 0 Modules linked in: Pid: 1, comm: swapper Not tainted 2.6.24-smp-g5a514e21-dirty #6 RIP: 0010:[<ffffffff802bfe55>] [<ffffffff802bfe55>] proc_register+0xe7/0x10f RSP: 0000:ffff810824c57e60 EFLAGS: 00010246 RAX: 000000000000d7d7 RBX: ffff811024c5fa80 RCX: ffff810824c57e08 RDX: 0000000000000000 RSI: 0000000000000195 RDI: ffffffff80cc2460 RBP: ffffffffffffffff R08: 0000000000000000 R09: ffff811024c5fa80 R10: 0000000000000000 R11: 0000000000000002 R12: ffff810824c57e6c R13: 0000000000000000 R14: ffff810824c57ee0 R15: 00000006abd25bee FS: 0000000000000000(0000) GS:ffffffff80b4d000(0000) knlGS:0000000000000000 CS: 0010 DS: 0018 ES: 0018 CR0: 000000008005003b CR2: 000000000000005f CR3: 0000000000201000 CR4: 00000000000006e0 DR0: 0000000000000000 DR1: 0000000000000000 DR2: 0000000000000000 DR3: 0000000000000000 DR6: 00000000ffff0ff0 DR7: 0000000000000400 Process swapper (pid: 1, threadinfo ffff810824c56000, task ffff812024c52000) Stack: ffffffff80a57348 0000019500000000 ffff811024c5fa80 0000000000000000 00000000ffffff97 ffffffff802bfef0 0000000000000000 ffffffffffffffff 0000000000000000 ffffffff80bc3b4b ffff810824c57ee0 ffffffff80bc34a5 Call Trace: [<ffffffff802bfef0>] ? create_proc_entry+0x73/0x8a [<ffffffff80bc3b4b>] ? sctp_snmp_proc_init+0x1c/0x34 [<ffffffff80bc34a5>] ? sctp_init+0xef/0x711 [<ffffffff80b976e3>] ? kernel_init+0x175/0x2e1 [<ffffffff8020ccf8>] ? child_rip+0xa/0x12 [<ffffffff80b9756e>] ? kernel_init+0x0/0x2e1 [<ffffffff8020ccee>] ? child_rip+0x0/0x12 Code: 1e 48 83 7b 38 00 75 08 48 c7 43 38 f0 e8 82 80 48 83 7b 30 00 75 08 48 c7 43 30 d0 e9 82 80 48 c7 c7 60 24 cc 80 e8 bd 5a 54 00 <48> 8b 45 60 48 89 6b 58 48 89 5d 60 48 89 43 50 fe 05 f5 25 a0 RIP [<ffffffff802bfe55>] proc_register+0xe7/0x10f RSP <ffff810824c57e60> CR2: 000000000000005f ---[ end trace 02c2d78def82877a ]--- Kernel panic - not syncing: Attempted to kill init! it turns out some variables near end of bss are corrupted already. in System.map we have ffffffff80d40420 b rsi_table ffffffff80d40620 B krb5_seq_lock ffffffff80d40628 b i.20437 ffffffff80d40630 b xprt_rdma_inline_write_padding ffffffff80d40638 b sunrpc_table_header ffffffff80d40640 b zero ffffffff80d40644 b min_memreg ffffffff80d40648 b rpcrdma_tk_lock_g ffffffff80d40650 B sctp_assocs_id_lock ffffffff80d40658 B proc_net_sctp ffffffff80d40660 B sctp_assocs_id ffffffff80d40680 B sysctl_sctp_mem ffffffff80d40690 B sysctl_sctp_rmem ffffffff80d406a0 B sysctl_sctp_wmem ffffffff80d406b0 b sctp_ctl_socket ffffffff80d406b8 b sctp_pf_inet6_specific ffffffff80d406c0 b sctp_pf_inet_specific ffffffff80d406c8 b sctp_af_v4_specific ffffffff80d406d0 b sctp_af_v6_specific ffffffff80d406d8 b sctp_rand.33270 ffffffff80d406dc b sctp_memory_pressure ffffffff80d406e0 b sctp_sockets_allocated ffffffff80d406e4 b sctp_memory_allocated ffffffff80d406e8 b sctp_sysctl_header ffffffff80d406f0 b zero ffffffff80d406f4 A __bss_stop ffffffff80d406f4 A _end and setup_node_bootmem() will use that page 0xd40000 for bootmap Bootmem setup node 0 0000000000000000-0000000828000000 NODE_DATA [000000000008a485 - 0000000000091484] bootmap [0000000000d406f4 - 0000000000e456f3] pages 105 Bootmem setup node 1 0000000828000000-0000001028000000 NODE_DATA [0000000828000000 - 0000000828006fff] bootmap [0000000828007000 - 0000000828106fff] pages 100 Bootmem setup node 2 0000001028000000-0000001828000000 NODE_DATA [0000001028000000 - 0000001028006fff] bootmap [0000001028007000 - 0000001028106fff] pages 100 Bootmem setup node 3 0000001828000000-0000002028000000 NODE_DATA [0000001828000000 - 0000001828006fff] bootmap [0000001828007000 - 0000001828106fff] pages 100 setup_node_bootmem() makes NODE_DATA cacheline aligned, and bootmap is page-aligned. the patch updates find_e820_area() to make sure we can meet the alignment constraints. Signed-off-by: Yinghai Lu <yinghai.lu@sun.com> Signed-off-by: Ingo Molnar <mingo@elte.hu>
2008-02-01 16:49:41 +00:00
/*
* SMP_CACHE_BYTES could be enough, but init_bootmem_node like
x86_64: make bootmap_start page align v6 boot oopses when a system has 64 or 128 GB of RAM installed: Calling initcall 0xffffffff80bc33b6: sctp_init+0x0/0x711() BUG: unable to handle kernel NULL pointer dereference at 000000000000005f IP: [<ffffffff802bfe55>] proc_register+0xe7/0x10f PGD 0 Oops: 0000 [1] SMP CPU 0 Modules linked in: Pid: 1, comm: swapper Not tainted 2.6.24-smp-g5a514e21-dirty #6 RIP: 0010:[<ffffffff802bfe55>] [<ffffffff802bfe55>] proc_register+0xe7/0x10f RSP: 0000:ffff810824c57e60 EFLAGS: 00010246 RAX: 000000000000d7d7 RBX: ffff811024c5fa80 RCX: ffff810824c57e08 RDX: 0000000000000000 RSI: 0000000000000195 RDI: ffffffff80cc2460 RBP: ffffffffffffffff R08: 0000000000000000 R09: ffff811024c5fa80 R10: 0000000000000000 R11: 0000000000000002 R12: ffff810824c57e6c R13: 0000000000000000 R14: ffff810824c57ee0 R15: 00000006abd25bee FS: 0000000000000000(0000) GS:ffffffff80b4d000(0000) knlGS:0000000000000000 CS: 0010 DS: 0018 ES: 0018 CR0: 000000008005003b CR2: 000000000000005f CR3: 0000000000201000 CR4: 00000000000006e0 DR0: 0000000000000000 DR1: 0000000000000000 DR2: 0000000000000000 DR3: 0000000000000000 DR6: 00000000ffff0ff0 DR7: 0000000000000400 Process swapper (pid: 1, threadinfo ffff810824c56000, task ffff812024c52000) Stack: ffffffff80a57348 0000019500000000 ffff811024c5fa80 0000000000000000 00000000ffffff97 ffffffff802bfef0 0000000000000000 ffffffffffffffff 0000000000000000 ffffffff80bc3b4b ffff810824c57ee0 ffffffff80bc34a5 Call Trace: [<ffffffff802bfef0>] ? create_proc_entry+0x73/0x8a [<ffffffff80bc3b4b>] ? sctp_snmp_proc_init+0x1c/0x34 [<ffffffff80bc34a5>] ? sctp_init+0xef/0x711 [<ffffffff80b976e3>] ? kernel_init+0x175/0x2e1 [<ffffffff8020ccf8>] ? child_rip+0xa/0x12 [<ffffffff80b9756e>] ? kernel_init+0x0/0x2e1 [<ffffffff8020ccee>] ? child_rip+0x0/0x12 Code: 1e 48 83 7b 38 00 75 08 48 c7 43 38 f0 e8 82 80 48 83 7b 30 00 75 08 48 c7 43 30 d0 e9 82 80 48 c7 c7 60 24 cc 80 e8 bd 5a 54 00 <48> 8b 45 60 48 89 6b 58 48 89 5d 60 48 89 43 50 fe 05 f5 25 a0 RIP [<ffffffff802bfe55>] proc_register+0xe7/0x10f RSP <ffff810824c57e60> CR2: 000000000000005f ---[ end trace 02c2d78def82877a ]--- Kernel panic - not syncing: Attempted to kill init! it turns out some variables near end of bss are corrupted already. in System.map we have ffffffff80d40420 b rsi_table ffffffff80d40620 B krb5_seq_lock ffffffff80d40628 b i.20437 ffffffff80d40630 b xprt_rdma_inline_write_padding ffffffff80d40638 b sunrpc_table_header ffffffff80d40640 b zero ffffffff80d40644 b min_memreg ffffffff80d40648 b rpcrdma_tk_lock_g ffffffff80d40650 B sctp_assocs_id_lock ffffffff80d40658 B proc_net_sctp ffffffff80d40660 B sctp_assocs_id ffffffff80d40680 B sysctl_sctp_mem ffffffff80d40690 B sysctl_sctp_rmem ffffffff80d406a0 B sysctl_sctp_wmem ffffffff80d406b0 b sctp_ctl_socket ffffffff80d406b8 b sctp_pf_inet6_specific ffffffff80d406c0 b sctp_pf_inet_specific ffffffff80d406c8 b sctp_af_v4_specific ffffffff80d406d0 b sctp_af_v6_specific ffffffff80d406d8 b sctp_rand.33270 ffffffff80d406dc b sctp_memory_pressure ffffffff80d406e0 b sctp_sockets_allocated ffffffff80d406e4 b sctp_memory_allocated ffffffff80d406e8 b sctp_sysctl_header ffffffff80d406f0 b zero ffffffff80d406f4 A __bss_stop ffffffff80d406f4 A _end and setup_node_bootmem() will use that page 0xd40000 for bootmap Bootmem setup node 0 0000000000000000-0000000828000000 NODE_DATA [000000000008a485 - 0000000000091484] bootmap [0000000000d406f4 - 0000000000e456f3] pages 105 Bootmem setup node 1 0000000828000000-0000001028000000 NODE_DATA [0000000828000000 - 0000000828006fff] bootmap [0000000828007000 - 0000000828106fff] pages 100 Bootmem setup node 2 0000001028000000-0000001828000000 NODE_DATA [0000001028000000 - 0000001028006fff] bootmap [0000001028007000 - 0000001028106fff] pages 100 Bootmem setup node 3 0000001828000000-0000002028000000 NODE_DATA [0000001828000000 - 0000001828006fff] bootmap [0000001828007000 - 0000001828106fff] pages 100 setup_node_bootmem() makes NODE_DATA cacheline aligned, and bootmap is page-aligned. the patch updates find_e820_area() to make sure we can meet the alignment constraints. Signed-off-by: Yinghai Lu <yinghai.lu@sun.com> Signed-off-by: Ingo Molnar <mingo@elte.hu>
2008-02-01 16:49:41 +00:00
* to use that to align to PAGE_SIZE
*/
bootmap = early_node_mem(nodeid, bootmap_start, end,
x86_64: make bootmap_start page align v6 boot oopses when a system has 64 or 128 GB of RAM installed: Calling initcall 0xffffffff80bc33b6: sctp_init+0x0/0x711() BUG: unable to handle kernel NULL pointer dereference at 000000000000005f IP: [<ffffffff802bfe55>] proc_register+0xe7/0x10f PGD 0 Oops: 0000 [1] SMP CPU 0 Modules linked in: Pid: 1, comm: swapper Not tainted 2.6.24-smp-g5a514e21-dirty #6 RIP: 0010:[<ffffffff802bfe55>] [<ffffffff802bfe55>] proc_register+0xe7/0x10f RSP: 0000:ffff810824c57e60 EFLAGS: 00010246 RAX: 000000000000d7d7 RBX: ffff811024c5fa80 RCX: ffff810824c57e08 RDX: 0000000000000000 RSI: 0000000000000195 RDI: ffffffff80cc2460 RBP: ffffffffffffffff R08: 0000000000000000 R09: ffff811024c5fa80 R10: 0000000000000000 R11: 0000000000000002 R12: ffff810824c57e6c R13: 0000000000000000 R14: ffff810824c57ee0 R15: 00000006abd25bee FS: 0000000000000000(0000) GS:ffffffff80b4d000(0000) knlGS:0000000000000000 CS: 0010 DS: 0018 ES: 0018 CR0: 000000008005003b CR2: 000000000000005f CR3: 0000000000201000 CR4: 00000000000006e0 DR0: 0000000000000000 DR1: 0000000000000000 DR2: 0000000000000000 DR3: 0000000000000000 DR6: 00000000ffff0ff0 DR7: 0000000000000400 Process swapper (pid: 1, threadinfo ffff810824c56000, task ffff812024c52000) Stack: ffffffff80a57348 0000019500000000 ffff811024c5fa80 0000000000000000 00000000ffffff97 ffffffff802bfef0 0000000000000000 ffffffffffffffff 0000000000000000 ffffffff80bc3b4b ffff810824c57ee0 ffffffff80bc34a5 Call Trace: [<ffffffff802bfef0>] ? create_proc_entry+0x73/0x8a [<ffffffff80bc3b4b>] ? sctp_snmp_proc_init+0x1c/0x34 [<ffffffff80bc34a5>] ? sctp_init+0xef/0x711 [<ffffffff80b976e3>] ? kernel_init+0x175/0x2e1 [<ffffffff8020ccf8>] ? child_rip+0xa/0x12 [<ffffffff80b9756e>] ? kernel_init+0x0/0x2e1 [<ffffffff8020ccee>] ? child_rip+0x0/0x12 Code: 1e 48 83 7b 38 00 75 08 48 c7 43 38 f0 e8 82 80 48 83 7b 30 00 75 08 48 c7 43 30 d0 e9 82 80 48 c7 c7 60 24 cc 80 e8 bd 5a 54 00 <48> 8b 45 60 48 89 6b 58 48 89 5d 60 48 89 43 50 fe 05 f5 25 a0 RIP [<ffffffff802bfe55>] proc_register+0xe7/0x10f RSP <ffff810824c57e60> CR2: 000000000000005f ---[ end trace 02c2d78def82877a ]--- Kernel panic - not syncing: Attempted to kill init! it turns out some variables near end of bss are corrupted already. in System.map we have ffffffff80d40420 b rsi_table ffffffff80d40620 B krb5_seq_lock ffffffff80d40628 b i.20437 ffffffff80d40630 b xprt_rdma_inline_write_padding ffffffff80d40638 b sunrpc_table_header ffffffff80d40640 b zero ffffffff80d40644 b min_memreg ffffffff80d40648 b rpcrdma_tk_lock_g ffffffff80d40650 B sctp_assocs_id_lock ffffffff80d40658 B proc_net_sctp ffffffff80d40660 B sctp_assocs_id ffffffff80d40680 B sysctl_sctp_mem ffffffff80d40690 B sysctl_sctp_rmem ffffffff80d406a0 B sysctl_sctp_wmem ffffffff80d406b0 b sctp_ctl_socket ffffffff80d406b8 b sctp_pf_inet6_specific ffffffff80d406c0 b sctp_pf_inet_specific ffffffff80d406c8 b sctp_af_v4_specific ffffffff80d406d0 b sctp_af_v6_specific ffffffff80d406d8 b sctp_rand.33270 ffffffff80d406dc b sctp_memory_pressure ffffffff80d406e0 b sctp_sockets_allocated ffffffff80d406e4 b sctp_memory_allocated ffffffff80d406e8 b sctp_sysctl_header ffffffff80d406f0 b zero ffffffff80d406f4 A __bss_stop ffffffff80d406f4 A _end and setup_node_bootmem() will use that page 0xd40000 for bootmap Bootmem setup node 0 0000000000000000-0000000828000000 NODE_DATA [000000000008a485 - 0000000000091484] bootmap [0000000000d406f4 - 0000000000e456f3] pages 105 Bootmem setup node 1 0000000828000000-0000001028000000 NODE_DATA [0000000828000000 - 0000000828006fff] bootmap [0000000828007000 - 0000000828106fff] pages 100 Bootmem setup node 2 0000001028000000-0000001828000000 NODE_DATA [0000001028000000 - 0000001028006fff] bootmap [0000001028007000 - 0000001028106fff] pages 100 Bootmem setup node 3 0000001828000000-0000002028000000 NODE_DATA [0000001828000000 - 0000001828006fff] bootmap [0000001828007000 - 0000001828106fff] pages 100 setup_node_bootmem() makes NODE_DATA cacheline aligned, and bootmap is page-aligned. the patch updates find_e820_area() to make sure we can meet the alignment constraints. Signed-off-by: Yinghai Lu <yinghai.lu@sun.com> Signed-off-by: Ingo Molnar <mingo@elte.hu>
2008-02-01 16:49:41 +00:00
bootmap_pages<<PAGE_SHIFT, PAGE_SIZE);
if (bootmap == NULL) {
free_early(nodedata_phys, nodedata_phys + pgdat_size);
node_data[nodeid] = NULL;
return;
}
bootmap_start = __pa(bootmap);
reserve_early(bootmap_start, bootmap_start+(bootmap_pages<<PAGE_SHIFT),
"BOOTMAP");
bootmap_size = init_bootmem_node(NODE_DATA(nodeid),
bootmap_start >> PAGE_SHIFT,
start_pfn, last_pfn);
printk(KERN_INFO " bootmap [%016lx - %016lx] pages %lx\n",
bootmap_start, bootmap_start + bootmap_size - 1,
bootmap_pages);
nid = phys_to_nid(bootmap_start);
if (nid != nodeid)
printk(KERN_INFO " bootmap(%d) on node %d\n", nodeid, nid);
free_bootmem_with_active_regions(nodeid, end);
#endif
node_set_online(nodeid);
}
/*
* There are unfortunately some poorly designed mainboards around that
* only connect memory to a single CPU. This breaks the 1:1 cpu->node
* mapping. To avoid this fill in the mapping for all possible CPUs,
* as the number of CPUs is not known yet. We round robin the existing
* nodes.
*/
void __init numa_init_array(void)
{
int rr, i;
rr = first_node(node_online_map);
for (i = 0; i < nr_cpu_ids; i++) {
if (early_cpu_to_node(i) != NUMA_NO_NODE)
continue;
numa_set_node(i, rr);
rr = next_node(rr, node_online_map);
if (rr == MAX_NUMNODES)
rr = first_node(node_online_map);
}
}
#ifdef CONFIG_NUMA_EMU
/* Numa emulation */
x86: Interleave emulated nodes over physical nodes Add interleaved NUMA emulation support This patch interleaves emulated nodes over the system's physical nodes. This is required for interleave optimizations since mempolicies, for example, operate by iterating over a nodemask and act without knowledge of node distances. It can also be used for testing memory latencies and NUMA bugs in the kernel. There're a couple of ways to do this: - divide the number of emulated nodes by the number of physical nodes and allocate the result on each physical node, or - allocate each successive emulated node on a different physical node until all memory is exhausted. The disadvantage of the first option is, depending on the asymmetry in node capacities of each physical node, emulated nodes may substantially differ in size on a particular physical node compared to another. The disadvantage of the second option is, also depending on the asymmetry in node capacities of each physical node, there may be more emulated nodes allocated on a single physical node as another. This patch implements the second option; we sacrifice the possibility that we may have slightly more emulated nodes on a particular physical node compared to another in lieu of node size asymmetry. [ Note that "node capacity" of a physical node is not only a function of its addressable range, but also is affected by subtracting out the amount of reserved memory over that range. NUMA emulation only deals with available, non-reserved memory quantities. ] We ensure there is at least a minimal amount of available memory allocated to each node. We also make sure that at least this amount of available memory is available in ZONE_DMA32 for any node that includes both ZONE_DMA32 and ZONE_NORMAL. This patch also cleans the emulation code up by no longer passing the statically allocated struct bootnode array among the various functions. This init.data array is not allocated on the stack since it may be very large and thus it may be accessed at file scope. The WARN_ON() for nodes_cover_memory() when faking proximity domains is removed since it relies on successive nodes always having greater start addresses than previous nodes; with interleaving this is no longer always true. Signed-off-by: David Rientjes <rientjes@google.com> Cc: Linus Torvalds <torvalds@linux-foundation.org> Cc: Andreas Herrmann <andreas.herrmann3@amd.com> Cc: Yinghai Lu <yinghai@kernel.org> Cc: Balbir Singh <balbir@linux.vnet.ibm.com> Cc: Ankita Garg <ankita@in.ibm.com> Cc: Len Brown <len.brown@intel.com> LKML-Reference: <alpine.DEB.1.00.0909251519150.14754@chino.kir.corp.google.com> Signed-off-by: Ingo Molnar <mingo@elte.hu>
2009-09-25 22:20:09 +00:00
static struct bootnode nodes[MAX_NUMNODES] __initdata;
static struct bootnode physnodes[MAX_NUMNODES] __initdata;
static char *cmdline __initdata;
x86: Interleave emulated nodes over physical nodes Add interleaved NUMA emulation support This patch interleaves emulated nodes over the system's physical nodes. This is required for interleave optimizations since mempolicies, for example, operate by iterating over a nodemask and act without knowledge of node distances. It can also be used for testing memory latencies and NUMA bugs in the kernel. There're a couple of ways to do this: - divide the number of emulated nodes by the number of physical nodes and allocate the result on each physical node, or - allocate each successive emulated node on a different physical node until all memory is exhausted. The disadvantage of the first option is, depending on the asymmetry in node capacities of each physical node, emulated nodes may substantially differ in size on a particular physical node compared to another. The disadvantage of the second option is, also depending on the asymmetry in node capacities of each physical node, there may be more emulated nodes allocated on a single physical node as another. This patch implements the second option; we sacrifice the possibility that we may have slightly more emulated nodes on a particular physical node compared to another in lieu of node size asymmetry. [ Note that "node capacity" of a physical node is not only a function of its addressable range, but also is affected by subtracting out the amount of reserved memory over that range. NUMA emulation only deals with available, non-reserved memory quantities. ] We ensure there is at least a minimal amount of available memory allocated to each node. We also make sure that at least this amount of available memory is available in ZONE_DMA32 for any node that includes both ZONE_DMA32 and ZONE_NORMAL. This patch also cleans the emulation code up by no longer passing the statically allocated struct bootnode array among the various functions. This init.data array is not allocated on the stack since it may be very large and thus it may be accessed at file scope. The WARN_ON() for nodes_cover_memory() when faking proximity domains is removed since it relies on successive nodes always having greater start addresses than previous nodes; with interleaving this is no longer always true. Signed-off-by: David Rientjes <rientjes@google.com> Cc: Linus Torvalds <torvalds@linux-foundation.org> Cc: Andreas Herrmann <andreas.herrmann3@amd.com> Cc: Yinghai Lu <yinghai@kernel.org> Cc: Balbir Singh <balbir@linux.vnet.ibm.com> Cc: Ankita Garg <ankita@in.ibm.com> Cc: Len Brown <len.brown@intel.com> LKML-Reference: <alpine.DEB.1.00.0909251519150.14754@chino.kir.corp.google.com> Signed-off-by: Ingo Molnar <mingo@elte.hu>
2009-09-25 22:20:09 +00:00
static int __init setup_physnodes(unsigned long start, unsigned long end,
int acpi, int k8)
{
int nr_nodes = 0;
int ret = 0;
int i;
#ifdef CONFIG_ACPI_NUMA
if (acpi)
nr_nodes = acpi_get_nodes(physnodes);
#endif
#ifdef CONFIG_K8_NUMA
if (k8)
nr_nodes = k8_get_nodes(physnodes);
#endif
/*
* Basic sanity checking on the physical node map: there may be errors
* if the SRAT or K8 incorrectly reported the topology or the mem=
* kernel parameter is used.
*/
for (i = 0; i < nr_nodes; i++) {
if (physnodes[i].start == physnodes[i].end)
continue;
if (physnodes[i].start > end) {
physnodes[i].end = physnodes[i].start;
continue;
}
if (physnodes[i].end < start) {
physnodes[i].start = physnodes[i].end;
continue;
}
if (physnodes[i].start < start)
physnodes[i].start = start;
if (physnodes[i].end > end)
physnodes[i].end = end;
}
/*
* Remove all nodes that have no memory or were truncated because of the
* limited address range.
*/
for (i = 0; i < nr_nodes; i++) {
if (physnodes[i].start == physnodes[i].end)
continue;
physnodes[ret].start = physnodes[i].start;
physnodes[ret].end = physnodes[i].end;
ret++;
}
/*
* If no physical topology was detected, a single node is faked to cover
* the entire address space.
*/
if (!ret) {
physnodes[ret].start = start;
physnodes[ret].end = end;
ret = 1;
}
return ret;
}
/*
* Setups up nid to range from addr to addr + size. If the end
* boundary is greater than max_addr, then max_addr is used instead.
* The return value is 0 if there is additional memory left for
* allocation past addr and -1 otherwise. addr is adjusted to be at
* the end of the node.
*/
x86: Interleave emulated nodes over physical nodes Add interleaved NUMA emulation support This patch interleaves emulated nodes over the system's physical nodes. This is required for interleave optimizations since mempolicies, for example, operate by iterating over a nodemask and act without knowledge of node distances. It can also be used for testing memory latencies and NUMA bugs in the kernel. There're a couple of ways to do this: - divide the number of emulated nodes by the number of physical nodes and allocate the result on each physical node, or - allocate each successive emulated node on a different physical node until all memory is exhausted. The disadvantage of the first option is, depending on the asymmetry in node capacities of each physical node, emulated nodes may substantially differ in size on a particular physical node compared to another. The disadvantage of the second option is, also depending on the asymmetry in node capacities of each physical node, there may be more emulated nodes allocated on a single physical node as another. This patch implements the second option; we sacrifice the possibility that we may have slightly more emulated nodes on a particular physical node compared to another in lieu of node size asymmetry. [ Note that "node capacity" of a physical node is not only a function of its addressable range, but also is affected by subtracting out the amount of reserved memory over that range. NUMA emulation only deals with available, non-reserved memory quantities. ] We ensure there is at least a minimal amount of available memory allocated to each node. We also make sure that at least this amount of available memory is available in ZONE_DMA32 for any node that includes both ZONE_DMA32 and ZONE_NORMAL. This patch also cleans the emulation code up by no longer passing the statically allocated struct bootnode array among the various functions. This init.data array is not allocated on the stack since it may be very large and thus it may be accessed at file scope. The WARN_ON() for nodes_cover_memory() when faking proximity domains is removed since it relies on successive nodes always having greater start addresses than previous nodes; with interleaving this is no longer always true. Signed-off-by: David Rientjes <rientjes@google.com> Cc: Linus Torvalds <torvalds@linux-foundation.org> Cc: Andreas Herrmann <andreas.herrmann3@amd.com> Cc: Yinghai Lu <yinghai@kernel.org> Cc: Balbir Singh <balbir@linux.vnet.ibm.com> Cc: Ankita Garg <ankita@in.ibm.com> Cc: Len Brown <len.brown@intel.com> LKML-Reference: <alpine.DEB.1.00.0909251519150.14754@chino.kir.corp.google.com> Signed-off-by: Ingo Molnar <mingo@elte.hu>
2009-09-25 22:20:09 +00:00
static int __init setup_node_range(int nid, u64 *addr, u64 size, u64 max_addr)
{
[PATCH] x86-64: configurable fake numa node sizes Extends the numa=fake x86_64 command-line option to allow for configurable node sizes. These nodes can be used in conjunction with cpusets for coarse memory resource management. The old command-line option is still supported: numa=fake=32 gives 32 fake NUMA nodes, ignoring the NUMA setup of the actual machine. But now you may configure your system for the node sizes of your choice: numa=fake=2*512,1024,2*256 gives two 512M nodes, one 1024M node, two 256M nodes, and the rest of system memory to a sixth node. The existing hash function is maintained to support the various node sizes that are possible with this implementation. Each node of the same size receives roughly the same amount of available pages, regardless of any reserved memory with its address range. The total available pages on the system is calculated and divided by the number of equal nodes to allocate. These nodes are then dynamically allocated and their borders extended until such time as their number of available pages reaches the required size. Configurable node sizes are recommended when used in conjunction with cpusets for memory control because it eliminates the overhead associated with scanning the zonelists of many smaller full nodes on page_alloc(). Cc: Andi Kleen <ak@suse.de> Signed-off-by: David Rientjes <rientjes@google.com> Signed-off-by: Andi Kleen <ak@suse.de> Cc: Paul Jackson <pj@sgi.com> Cc: Christoph Lameter <clameter@engr.sgi.com> Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
2007-05-02 17:27:09 +00:00
int ret = 0;
nodes[nid].start = *addr;
*addr += size;
if (*addr >= max_addr) {
*addr = max_addr;
ret = -1;
}
nodes[nid].end = *addr;
node_set(nid, node_possible_map);
[PATCH] x86-64: configurable fake numa node sizes Extends the numa=fake x86_64 command-line option to allow for configurable node sizes. These nodes can be used in conjunction with cpusets for coarse memory resource management. The old command-line option is still supported: numa=fake=32 gives 32 fake NUMA nodes, ignoring the NUMA setup of the actual machine. But now you may configure your system for the node sizes of your choice: numa=fake=2*512,1024,2*256 gives two 512M nodes, one 1024M node, two 256M nodes, and the rest of system memory to a sixth node. The existing hash function is maintained to support the various node sizes that are possible with this implementation. Each node of the same size receives roughly the same amount of available pages, regardless of any reserved memory with its address range. The total available pages on the system is calculated and divided by the number of equal nodes to allocate. These nodes are then dynamically allocated and their borders extended until such time as their number of available pages reaches the required size. Configurable node sizes are recommended when used in conjunction with cpusets for memory control because it eliminates the overhead associated with scanning the zonelists of many smaller full nodes on page_alloc(). Cc: Andi Kleen <ak@suse.de> Signed-off-by: David Rientjes <rientjes@google.com> Signed-off-by: Andi Kleen <ak@suse.de> Cc: Paul Jackson <pj@sgi.com> Cc: Christoph Lameter <clameter@engr.sgi.com> Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
2007-05-02 17:27:09 +00:00
printk(KERN_INFO "Faking node %d at %016Lx-%016Lx (%LuMB)\n", nid,
nodes[nid].start, nodes[nid].end,
(nodes[nid].end - nodes[nid].start) >> 20);
return ret;
}
x86: Interleave emulated nodes over physical nodes Add interleaved NUMA emulation support This patch interleaves emulated nodes over the system's physical nodes. This is required for interleave optimizations since mempolicies, for example, operate by iterating over a nodemask and act without knowledge of node distances. It can also be used for testing memory latencies and NUMA bugs in the kernel. There're a couple of ways to do this: - divide the number of emulated nodes by the number of physical nodes and allocate the result on each physical node, or - allocate each successive emulated node on a different physical node until all memory is exhausted. The disadvantage of the first option is, depending on the asymmetry in node capacities of each physical node, emulated nodes may substantially differ in size on a particular physical node compared to another. The disadvantage of the second option is, also depending on the asymmetry in node capacities of each physical node, there may be more emulated nodes allocated on a single physical node as another. This patch implements the second option; we sacrifice the possibility that we may have slightly more emulated nodes on a particular physical node compared to another in lieu of node size asymmetry. [ Note that "node capacity" of a physical node is not only a function of its addressable range, but also is affected by subtracting out the amount of reserved memory over that range. NUMA emulation only deals with available, non-reserved memory quantities. ] We ensure there is at least a minimal amount of available memory allocated to each node. We also make sure that at least this amount of available memory is available in ZONE_DMA32 for any node that includes both ZONE_DMA32 and ZONE_NORMAL. This patch also cleans the emulation code up by no longer passing the statically allocated struct bootnode array among the various functions. This init.data array is not allocated on the stack since it may be very large and thus it may be accessed at file scope. The WARN_ON() for nodes_cover_memory() when faking proximity domains is removed since it relies on successive nodes always having greater start addresses than previous nodes; with interleaving this is no longer always true. Signed-off-by: David Rientjes <rientjes@google.com> Cc: Linus Torvalds <torvalds@linux-foundation.org> Cc: Andreas Herrmann <andreas.herrmann3@amd.com> Cc: Yinghai Lu <yinghai@kernel.org> Cc: Balbir Singh <balbir@linux.vnet.ibm.com> Cc: Ankita Garg <ankita@in.ibm.com> Cc: Len Brown <len.brown@intel.com> LKML-Reference: <alpine.DEB.1.00.0909251519150.14754@chino.kir.corp.google.com> Signed-off-by: Ingo Molnar <mingo@elte.hu>
2009-09-25 22:20:09 +00:00
/*
* Sets up nr_nodes fake nodes interleaved over physical nodes ranging from addr
* to max_addr. The return value is the number of nodes allocated.
*/
static int __init split_nodes_interleave(u64 addr, u64 max_addr,
int nr_phys_nodes, int nr_nodes)
{
nodemask_t physnode_mask = NODE_MASK_NONE;
u64 size;
int big;
int ret = 0;
int i;
if (nr_nodes <= 0)
return -1;
if (nr_nodes > MAX_NUMNODES) {
pr_info("numa=fake=%d too large, reducing to %d\n",
nr_nodes, MAX_NUMNODES);
nr_nodes = MAX_NUMNODES;
}
size = (max_addr - addr - e820_hole_size(addr, max_addr)) / nr_nodes;
/*
* Calculate the number of big nodes that can be allocated as a result
* of consolidating the remainder.
*/
big = ((size & ~FAKE_NODE_MIN_HASH_MASK) * nr_nodes) /
x86: Interleave emulated nodes over physical nodes Add interleaved NUMA emulation support This patch interleaves emulated nodes over the system's physical nodes. This is required for interleave optimizations since mempolicies, for example, operate by iterating over a nodemask and act without knowledge of node distances. It can also be used for testing memory latencies and NUMA bugs in the kernel. There're a couple of ways to do this: - divide the number of emulated nodes by the number of physical nodes and allocate the result on each physical node, or - allocate each successive emulated node on a different physical node until all memory is exhausted. The disadvantage of the first option is, depending on the asymmetry in node capacities of each physical node, emulated nodes may substantially differ in size on a particular physical node compared to another. The disadvantage of the second option is, also depending on the asymmetry in node capacities of each physical node, there may be more emulated nodes allocated on a single physical node as another. This patch implements the second option; we sacrifice the possibility that we may have slightly more emulated nodes on a particular physical node compared to another in lieu of node size asymmetry. [ Note that "node capacity" of a physical node is not only a function of its addressable range, but also is affected by subtracting out the amount of reserved memory over that range. NUMA emulation only deals with available, non-reserved memory quantities. ] We ensure there is at least a minimal amount of available memory allocated to each node. We also make sure that at least this amount of available memory is available in ZONE_DMA32 for any node that includes both ZONE_DMA32 and ZONE_NORMAL. This patch also cleans the emulation code up by no longer passing the statically allocated struct bootnode array among the various functions. This init.data array is not allocated on the stack since it may be very large and thus it may be accessed at file scope. The WARN_ON() for nodes_cover_memory() when faking proximity domains is removed since it relies on successive nodes always having greater start addresses than previous nodes; with interleaving this is no longer always true. Signed-off-by: David Rientjes <rientjes@google.com> Cc: Linus Torvalds <torvalds@linux-foundation.org> Cc: Andreas Herrmann <andreas.herrmann3@amd.com> Cc: Yinghai Lu <yinghai@kernel.org> Cc: Balbir Singh <balbir@linux.vnet.ibm.com> Cc: Ankita Garg <ankita@in.ibm.com> Cc: Len Brown <len.brown@intel.com> LKML-Reference: <alpine.DEB.1.00.0909251519150.14754@chino.kir.corp.google.com> Signed-off-by: Ingo Molnar <mingo@elte.hu>
2009-09-25 22:20:09 +00:00
FAKE_NODE_MIN_SIZE;
size &= FAKE_NODE_MIN_HASH_MASK;
if (!size) {
pr_err("Not enough memory for each node. "
"NUMA emulation disabled.\n");
return -1;
}
for (i = 0; i < nr_phys_nodes; i++)
if (physnodes[i].start != physnodes[i].end)
node_set(i, physnode_mask);
/*
* Continue to fill physical nodes with fake nodes until there is no
* memory left on any of them.
*/
while (nodes_weight(physnode_mask)) {
for_each_node_mask(i, physnode_mask) {
u64 end = physnodes[i].start + size;
u64 dma32_end = PFN_PHYS(MAX_DMA32_PFN);
if (ret < big)
end += FAKE_NODE_MIN_SIZE;
/*
* Continue to add memory to this fake node if its
* non-reserved memory is less than the per-node size.
*/
while (end - physnodes[i].start -
e820_hole_size(physnodes[i].start, end) < size) {
end += FAKE_NODE_MIN_SIZE;
if (end > physnodes[i].end) {
end = physnodes[i].end;
break;
}
}
/*
* If there won't be at least FAKE_NODE_MIN_SIZE of
* non-reserved memory in ZONE_DMA32 for the next node,
* this one must extend to the boundary.
*/
if (end < dma32_end && dma32_end - end -
e820_hole_size(end, dma32_end) < FAKE_NODE_MIN_SIZE)
end = dma32_end;
/*
* If there won't be enough non-reserved memory for the
* next node, this one must extend to the end of the
* physical node.
*/
if (physnodes[i].end - end -
e820_hole_size(end, physnodes[i].end) < size)
end = physnodes[i].end;
/*
* Avoid allocating more nodes than requested, which can
* happen as a result of rounding down each node's size
* to FAKE_NODE_MIN_SIZE.
*/
if (nodes_weight(physnode_mask) + ret >= nr_nodes)
end = physnodes[i].end;
if (setup_node_range(ret++, &physnodes[i].start,
end - physnodes[i].start,
physnodes[i].end) < 0)
node_clear(i, physnode_mask);
}
}
return ret;
}
/*
* Returns the end address of a node so that there is at least `size' amount of
* non-reserved memory or `max_addr' is reached.
*/
static u64 __init find_end_of_node(u64 start, u64 max_addr, u64 size)
{
u64 end = start + size;
while (end - start - e820_hole_size(start, end) < size) {
end += FAKE_NODE_MIN_SIZE;
if (end > max_addr) {
end = max_addr;
break;
}
}
return end;
}
/*
* Sets up fake nodes of `size' interleaved over physical nodes ranging from
* `addr' to `max_addr'. The return value is the number of nodes allocated.
*/
static int __init split_nodes_size_interleave(u64 addr, u64 max_addr, u64 size)
{
nodemask_t physnode_mask = NODE_MASK_NONE;
u64 min_size;
int ret = 0;
int i;
if (!size)
return -1;
/*
* The limit on emulated nodes is MAX_NUMNODES, so the size per node is
* increased accordingly if the requested size is too small. This
* creates a uniform distribution of node sizes across the entire
* machine (but not necessarily over physical nodes).
*/
min_size = (max_addr - addr - e820_hole_size(addr, max_addr)) /
MAX_NUMNODES;
min_size = max(min_size, FAKE_NODE_MIN_SIZE);
if ((min_size & FAKE_NODE_MIN_HASH_MASK) < min_size)
min_size = (min_size + FAKE_NODE_MIN_SIZE) &
FAKE_NODE_MIN_HASH_MASK;
if (size < min_size) {
pr_err("Fake node size %LuMB too small, increasing to %LuMB\n",
size >> 20, min_size >> 20);
size = min_size;
}
size &= FAKE_NODE_MIN_HASH_MASK;
for (i = 0; i < MAX_NUMNODES; i++)
if (physnodes[i].start != physnodes[i].end)
node_set(i, physnode_mask);
/*
* Fill physical nodes with fake nodes of size until there is no memory
* left on any of them.
*/
while (nodes_weight(physnode_mask)) {
for_each_node_mask(i, physnode_mask) {
u64 dma32_end = MAX_DMA32_PFN << PAGE_SHIFT;
u64 end;
end = find_end_of_node(physnodes[i].start,
physnodes[i].end, size);
/*
* If there won't be at least FAKE_NODE_MIN_SIZE of
* non-reserved memory in ZONE_DMA32 for the next node,
* this one must extend to the boundary.
*/
if (end < dma32_end && dma32_end - end -
e820_hole_size(end, dma32_end) < FAKE_NODE_MIN_SIZE)
end = dma32_end;
/*
* If there won't be enough non-reserved memory for the
* next node, this one must extend to the end of the
* physical node.
*/
if (physnodes[i].end - end -
e820_hole_size(end, physnodes[i].end) < size)
end = physnodes[i].end;
/*
* Setup the fake node that will be allocated as bootmem
* later. If setup_node_range() returns non-zero, there
* is no more memory available on this physical node.
*/
if (setup_node_range(ret++, &physnodes[i].start,
end - physnodes[i].start,
physnodes[i].end) < 0)
node_clear(i, physnode_mask);
}
}
return ret;
}
[PATCH] x86-64: configurable fake numa node sizes Extends the numa=fake x86_64 command-line option to allow for configurable node sizes. These nodes can be used in conjunction with cpusets for coarse memory resource management. The old command-line option is still supported: numa=fake=32 gives 32 fake NUMA nodes, ignoring the NUMA setup of the actual machine. But now you may configure your system for the node sizes of your choice: numa=fake=2*512,1024,2*256 gives two 512M nodes, one 1024M node, two 256M nodes, and the rest of system memory to a sixth node. The existing hash function is maintained to support the various node sizes that are possible with this implementation. Each node of the same size receives roughly the same amount of available pages, regardless of any reserved memory with its address range. The total available pages on the system is calculated and divided by the number of equal nodes to allocate. These nodes are then dynamically allocated and their borders extended until such time as their number of available pages reaches the required size. Configurable node sizes are recommended when used in conjunction with cpusets for memory control because it eliminates the overhead associated with scanning the zonelists of many smaller full nodes on page_alloc(). Cc: Andi Kleen <ak@suse.de> Signed-off-by: David Rientjes <rientjes@google.com> Signed-off-by: Andi Kleen <ak@suse.de> Cc: Paul Jackson <pj@sgi.com> Cc: Christoph Lameter <clameter@engr.sgi.com> Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
2007-05-02 17:27:09 +00:00
/*
* Sets up the system RAM area from start_pfn to last_pfn according to the
[PATCH] x86-64: configurable fake numa node sizes Extends the numa=fake x86_64 command-line option to allow for configurable node sizes. These nodes can be used in conjunction with cpusets for coarse memory resource management. The old command-line option is still supported: numa=fake=32 gives 32 fake NUMA nodes, ignoring the NUMA setup of the actual machine. But now you may configure your system for the node sizes of your choice: numa=fake=2*512,1024,2*256 gives two 512M nodes, one 1024M node, two 256M nodes, and the rest of system memory to a sixth node. The existing hash function is maintained to support the various node sizes that are possible with this implementation. Each node of the same size receives roughly the same amount of available pages, regardless of any reserved memory with its address range. The total available pages on the system is calculated and divided by the number of equal nodes to allocate. These nodes are then dynamically allocated and their borders extended until such time as their number of available pages reaches the required size. Configurable node sizes are recommended when used in conjunction with cpusets for memory control because it eliminates the overhead associated with scanning the zonelists of many smaller full nodes on page_alloc(). Cc: Andi Kleen <ak@suse.de> Signed-off-by: David Rientjes <rientjes@google.com> Signed-off-by: Andi Kleen <ak@suse.de> Cc: Paul Jackson <pj@sgi.com> Cc: Christoph Lameter <clameter@engr.sgi.com> Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
2007-05-02 17:27:09 +00:00
* numa=fake command-line option.
*/
x86: Interleave emulated nodes over physical nodes Add interleaved NUMA emulation support This patch interleaves emulated nodes over the system's physical nodes. This is required for interleave optimizations since mempolicies, for example, operate by iterating over a nodemask and act without knowledge of node distances. It can also be used for testing memory latencies and NUMA bugs in the kernel. There're a couple of ways to do this: - divide the number of emulated nodes by the number of physical nodes and allocate the result on each physical node, or - allocate each successive emulated node on a different physical node until all memory is exhausted. The disadvantage of the first option is, depending on the asymmetry in node capacities of each physical node, emulated nodes may substantially differ in size on a particular physical node compared to another. The disadvantage of the second option is, also depending on the asymmetry in node capacities of each physical node, there may be more emulated nodes allocated on a single physical node as another. This patch implements the second option; we sacrifice the possibility that we may have slightly more emulated nodes on a particular physical node compared to another in lieu of node size asymmetry. [ Note that "node capacity" of a physical node is not only a function of its addressable range, but also is affected by subtracting out the amount of reserved memory over that range. NUMA emulation only deals with available, non-reserved memory quantities. ] We ensure there is at least a minimal amount of available memory allocated to each node. We also make sure that at least this amount of available memory is available in ZONE_DMA32 for any node that includes both ZONE_DMA32 and ZONE_NORMAL. This patch also cleans the emulation code up by no longer passing the statically allocated struct bootnode array among the various functions. This init.data array is not allocated on the stack since it may be very large and thus it may be accessed at file scope. The WARN_ON() for nodes_cover_memory() when faking proximity domains is removed since it relies on successive nodes always having greater start addresses than previous nodes; with interleaving this is no longer always true. Signed-off-by: David Rientjes <rientjes@google.com> Cc: Linus Torvalds <torvalds@linux-foundation.org> Cc: Andreas Herrmann <andreas.herrmann3@amd.com> Cc: Yinghai Lu <yinghai@kernel.org> Cc: Balbir Singh <balbir@linux.vnet.ibm.com> Cc: Ankita Garg <ankita@in.ibm.com> Cc: Len Brown <len.brown@intel.com> LKML-Reference: <alpine.DEB.1.00.0909251519150.14754@chino.kir.corp.google.com> Signed-off-by: Ingo Molnar <mingo@elte.hu>
2009-09-25 22:20:09 +00:00
static int __init numa_emulation(unsigned long start_pfn,
unsigned long last_pfn, int acpi, int k8)
[PATCH] x86-64: configurable fake numa node sizes Extends the numa=fake x86_64 command-line option to allow for configurable node sizes. These nodes can be used in conjunction with cpusets for coarse memory resource management. The old command-line option is still supported: numa=fake=32 gives 32 fake NUMA nodes, ignoring the NUMA setup of the actual machine. But now you may configure your system for the node sizes of your choice: numa=fake=2*512,1024,2*256 gives two 512M nodes, one 1024M node, two 256M nodes, and the rest of system memory to a sixth node. The existing hash function is maintained to support the various node sizes that are possible with this implementation. Each node of the same size receives roughly the same amount of available pages, regardless of any reserved memory with its address range. The total available pages on the system is calculated and divided by the number of equal nodes to allocate. These nodes are then dynamically allocated and their borders extended until such time as their number of available pages reaches the required size. Configurable node sizes are recommended when used in conjunction with cpusets for memory control because it eliminates the overhead associated with scanning the zonelists of many smaller full nodes on page_alloc(). Cc: Andi Kleen <ak@suse.de> Signed-off-by: David Rientjes <rientjes@google.com> Signed-off-by: Andi Kleen <ak@suse.de> Cc: Paul Jackson <pj@sgi.com> Cc: Christoph Lameter <clameter@engr.sgi.com> Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
2007-05-02 17:27:09 +00:00
{
u64 addr = start_pfn << PAGE_SHIFT;
u64 max_addr = last_pfn << PAGE_SHIFT;
x86: Interleave emulated nodes over physical nodes Add interleaved NUMA emulation support This patch interleaves emulated nodes over the system's physical nodes. This is required for interleave optimizations since mempolicies, for example, operate by iterating over a nodemask and act without knowledge of node distances. It can also be used for testing memory latencies and NUMA bugs in the kernel. There're a couple of ways to do this: - divide the number of emulated nodes by the number of physical nodes and allocate the result on each physical node, or - allocate each successive emulated node on a different physical node until all memory is exhausted. The disadvantage of the first option is, depending on the asymmetry in node capacities of each physical node, emulated nodes may substantially differ in size on a particular physical node compared to another. The disadvantage of the second option is, also depending on the asymmetry in node capacities of each physical node, there may be more emulated nodes allocated on a single physical node as another. This patch implements the second option; we sacrifice the possibility that we may have slightly more emulated nodes on a particular physical node compared to another in lieu of node size asymmetry. [ Note that "node capacity" of a physical node is not only a function of its addressable range, but also is affected by subtracting out the amount of reserved memory over that range. NUMA emulation only deals with available, non-reserved memory quantities. ] We ensure there is at least a minimal amount of available memory allocated to each node. We also make sure that at least this amount of available memory is available in ZONE_DMA32 for any node that includes both ZONE_DMA32 and ZONE_NORMAL. This patch also cleans the emulation code up by no longer passing the statically allocated struct bootnode array among the various functions. This init.data array is not allocated on the stack since it may be very large and thus it may be accessed at file scope. The WARN_ON() for nodes_cover_memory() when faking proximity domains is removed since it relies on successive nodes always having greater start addresses than previous nodes; with interleaving this is no longer always true. Signed-off-by: David Rientjes <rientjes@google.com> Cc: Linus Torvalds <torvalds@linux-foundation.org> Cc: Andreas Herrmann <andreas.herrmann3@amd.com> Cc: Yinghai Lu <yinghai@kernel.org> Cc: Balbir Singh <balbir@linux.vnet.ibm.com> Cc: Ankita Garg <ankita@in.ibm.com> Cc: Len Brown <len.brown@intel.com> LKML-Reference: <alpine.DEB.1.00.0909251519150.14754@chino.kir.corp.google.com> Signed-off-by: Ingo Molnar <mingo@elte.hu>
2009-09-25 22:20:09 +00:00
int num_phys_nodes;
int num_nodes;
int i;
[PATCH] x86-64: configurable fake numa node sizes Extends the numa=fake x86_64 command-line option to allow for configurable node sizes. These nodes can be used in conjunction with cpusets for coarse memory resource management. The old command-line option is still supported: numa=fake=32 gives 32 fake NUMA nodes, ignoring the NUMA setup of the actual machine. But now you may configure your system for the node sizes of your choice: numa=fake=2*512,1024,2*256 gives two 512M nodes, one 1024M node, two 256M nodes, and the rest of system memory to a sixth node. The existing hash function is maintained to support the various node sizes that are possible with this implementation. Each node of the same size receives roughly the same amount of available pages, regardless of any reserved memory with its address range. The total available pages on the system is calculated and divided by the number of equal nodes to allocate. These nodes are then dynamically allocated and their borders extended until such time as their number of available pages reaches the required size. Configurable node sizes are recommended when used in conjunction with cpusets for memory control because it eliminates the overhead associated with scanning the zonelists of many smaller full nodes on page_alloc(). Cc: Andi Kleen <ak@suse.de> Signed-off-by: David Rientjes <rientjes@google.com> Signed-off-by: Andi Kleen <ak@suse.de> Cc: Paul Jackson <pj@sgi.com> Cc: Christoph Lameter <clameter@engr.sgi.com> Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
2007-05-02 17:27:09 +00:00
x86: Interleave emulated nodes over physical nodes Add interleaved NUMA emulation support This patch interleaves emulated nodes over the system's physical nodes. This is required for interleave optimizations since mempolicies, for example, operate by iterating over a nodemask and act without knowledge of node distances. It can also be used for testing memory latencies and NUMA bugs in the kernel. There're a couple of ways to do this: - divide the number of emulated nodes by the number of physical nodes and allocate the result on each physical node, or - allocate each successive emulated node on a different physical node until all memory is exhausted. The disadvantage of the first option is, depending on the asymmetry in node capacities of each physical node, emulated nodes may substantially differ in size on a particular physical node compared to another. The disadvantage of the second option is, also depending on the asymmetry in node capacities of each physical node, there may be more emulated nodes allocated on a single physical node as another. This patch implements the second option; we sacrifice the possibility that we may have slightly more emulated nodes on a particular physical node compared to another in lieu of node size asymmetry. [ Note that "node capacity" of a physical node is not only a function of its addressable range, but also is affected by subtracting out the amount of reserved memory over that range. NUMA emulation only deals with available, non-reserved memory quantities. ] We ensure there is at least a minimal amount of available memory allocated to each node. We also make sure that at least this amount of available memory is available in ZONE_DMA32 for any node that includes both ZONE_DMA32 and ZONE_NORMAL. This patch also cleans the emulation code up by no longer passing the statically allocated struct bootnode array among the various functions. This init.data array is not allocated on the stack since it may be very large and thus it may be accessed at file scope. The WARN_ON() for nodes_cover_memory() when faking proximity domains is removed since it relies on successive nodes always having greater start addresses than previous nodes; with interleaving this is no longer always true. Signed-off-by: David Rientjes <rientjes@google.com> Cc: Linus Torvalds <torvalds@linux-foundation.org> Cc: Andreas Herrmann <andreas.herrmann3@amd.com> Cc: Yinghai Lu <yinghai@kernel.org> Cc: Balbir Singh <balbir@linux.vnet.ibm.com> Cc: Ankita Garg <ankita@in.ibm.com> Cc: Len Brown <len.brown@intel.com> LKML-Reference: <alpine.DEB.1.00.0909251519150.14754@chino.kir.corp.google.com> Signed-off-by: Ingo Molnar <mingo@elte.hu>
2009-09-25 22:20:09 +00:00
num_phys_nodes = setup_physnodes(addr, max_addr, acpi, k8);
/*
* If the numa=fake command-line contains a 'M' or 'G', it represents
* the fixed node size. Otherwise, if it is just a single number N,
* split the system RAM into N fake nodes.
*/
if (strchr(cmdline, 'M') || strchr(cmdline, 'G')) {
u64 size;
size = memparse(cmdline, &cmdline);
num_nodes = split_nodes_size_interleave(addr, max_addr, size);
} else {
unsigned long n;
n = simple_strtoul(cmdline, NULL, 0);
num_nodes = split_nodes_interleave(addr, max_addr, num_phys_nodes, n);
[PATCH] x86-64: configurable fake numa node sizes Extends the numa=fake x86_64 command-line option to allow for configurable node sizes. These nodes can be used in conjunction with cpusets for coarse memory resource management. The old command-line option is still supported: numa=fake=32 gives 32 fake NUMA nodes, ignoring the NUMA setup of the actual machine. But now you may configure your system for the node sizes of your choice: numa=fake=2*512,1024,2*256 gives two 512M nodes, one 1024M node, two 256M nodes, and the rest of system memory to a sixth node. The existing hash function is maintained to support the various node sizes that are possible with this implementation. Each node of the same size receives roughly the same amount of available pages, regardless of any reserved memory with its address range. The total available pages on the system is calculated and divided by the number of equal nodes to allocate. These nodes are then dynamically allocated and their borders extended until such time as their number of available pages reaches the required size. Configurable node sizes are recommended when used in conjunction with cpusets for memory control because it eliminates the overhead associated with scanning the zonelists of many smaller full nodes on page_alloc(). Cc: Andi Kleen <ak@suse.de> Signed-off-by: David Rientjes <rientjes@google.com> Signed-off-by: Andi Kleen <ak@suse.de> Cc: Paul Jackson <pj@sgi.com> Cc: Christoph Lameter <clameter@engr.sgi.com> Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
2007-05-02 17:27:09 +00:00
}
if (num_nodes < 0)
return num_nodes;
memnode_shift = compute_hash_shift(nodes, num_nodes, NULL);
[PATCH] x86-64: configurable fake numa node sizes Extends the numa=fake x86_64 command-line option to allow for configurable node sizes. These nodes can be used in conjunction with cpusets for coarse memory resource management. The old command-line option is still supported: numa=fake=32 gives 32 fake NUMA nodes, ignoring the NUMA setup of the actual machine. But now you may configure your system for the node sizes of your choice: numa=fake=2*512,1024,2*256 gives two 512M nodes, one 1024M node, two 256M nodes, and the rest of system memory to a sixth node. The existing hash function is maintained to support the various node sizes that are possible with this implementation. Each node of the same size receives roughly the same amount of available pages, regardless of any reserved memory with its address range. The total available pages on the system is calculated and divided by the number of equal nodes to allocate. These nodes are then dynamically allocated and their borders extended until such time as their number of available pages reaches the required size. Configurable node sizes are recommended when used in conjunction with cpusets for memory control because it eliminates the overhead associated with scanning the zonelists of many smaller full nodes on page_alloc(). Cc: Andi Kleen <ak@suse.de> Signed-off-by: David Rientjes <rientjes@google.com> Signed-off-by: Andi Kleen <ak@suse.de> Cc: Paul Jackson <pj@sgi.com> Cc: Christoph Lameter <clameter@engr.sgi.com> Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
2007-05-02 17:27:09 +00:00
if (memnode_shift < 0) {
memnode_shift = 0;
printk(KERN_ERR "No NUMA hash function found. NUMA emulation "
"disabled.\n");
return -1;
}
/*
x86: Interleave emulated nodes over physical nodes Add interleaved NUMA emulation support This patch interleaves emulated nodes over the system's physical nodes. This is required for interleave optimizations since mempolicies, for example, operate by iterating over a nodemask and act without knowledge of node distances. It can also be used for testing memory latencies and NUMA bugs in the kernel. There're a couple of ways to do this: - divide the number of emulated nodes by the number of physical nodes and allocate the result on each physical node, or - allocate each successive emulated node on a different physical node until all memory is exhausted. The disadvantage of the first option is, depending on the asymmetry in node capacities of each physical node, emulated nodes may substantially differ in size on a particular physical node compared to another. The disadvantage of the second option is, also depending on the asymmetry in node capacities of each physical node, there may be more emulated nodes allocated on a single physical node as another. This patch implements the second option; we sacrifice the possibility that we may have slightly more emulated nodes on a particular physical node compared to another in lieu of node size asymmetry. [ Note that "node capacity" of a physical node is not only a function of its addressable range, but also is affected by subtracting out the amount of reserved memory over that range. NUMA emulation only deals with available, non-reserved memory quantities. ] We ensure there is at least a minimal amount of available memory allocated to each node. We also make sure that at least this amount of available memory is available in ZONE_DMA32 for any node that includes both ZONE_DMA32 and ZONE_NORMAL. This patch also cleans the emulation code up by no longer passing the statically allocated struct bootnode array among the various functions. This init.data array is not allocated on the stack since it may be very large and thus it may be accessed at file scope. The WARN_ON() for nodes_cover_memory() when faking proximity domains is removed since it relies on successive nodes always having greater start addresses than previous nodes; with interleaving this is no longer always true. Signed-off-by: David Rientjes <rientjes@google.com> Cc: Linus Torvalds <torvalds@linux-foundation.org> Cc: Andreas Herrmann <andreas.herrmann3@amd.com> Cc: Yinghai Lu <yinghai@kernel.org> Cc: Balbir Singh <balbir@linux.vnet.ibm.com> Cc: Ankita Garg <ankita@in.ibm.com> Cc: Len Brown <len.brown@intel.com> LKML-Reference: <alpine.DEB.1.00.0909251519150.14754@chino.kir.corp.google.com> Signed-off-by: Ingo Molnar <mingo@elte.hu>
2009-09-25 22:20:09 +00:00
* We need to vacate all active ranges that may have been registered for
* the e820 memory map.
[PATCH] x86-64: configurable fake numa node sizes Extends the numa=fake x86_64 command-line option to allow for configurable node sizes. These nodes can be used in conjunction with cpusets for coarse memory resource management. The old command-line option is still supported: numa=fake=32 gives 32 fake NUMA nodes, ignoring the NUMA setup of the actual machine. But now you may configure your system for the node sizes of your choice: numa=fake=2*512,1024,2*256 gives two 512M nodes, one 1024M node, two 256M nodes, and the rest of system memory to a sixth node. The existing hash function is maintained to support the various node sizes that are possible with this implementation. Each node of the same size receives roughly the same amount of available pages, regardless of any reserved memory with its address range. The total available pages on the system is calculated and divided by the number of equal nodes to allocate. These nodes are then dynamically allocated and their borders extended until such time as their number of available pages reaches the required size. Configurable node sizes are recommended when used in conjunction with cpusets for memory control because it eliminates the overhead associated with scanning the zonelists of many smaller full nodes on page_alloc(). Cc: Andi Kleen <ak@suse.de> Signed-off-by: David Rientjes <rientjes@google.com> Signed-off-by: Andi Kleen <ak@suse.de> Cc: Paul Jackson <pj@sgi.com> Cc: Christoph Lameter <clameter@engr.sgi.com> Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
2007-05-02 17:27:09 +00:00
*/
remove_all_active_ranges();
for_each_node_mask(i, node_possible_map) {
e820_register_active_regions(i, nodes[i].start >> PAGE_SHIFT,
nodes[i].end >> PAGE_SHIFT);
setup_node_bootmem(i, nodes[i].start, nodes[i].end);
}
acpi_fake_nodes(nodes, num_nodes);
numa_init_array();
return 0;
}
[PATCH] x86-64: configurable fake numa node sizes Extends the numa=fake x86_64 command-line option to allow for configurable node sizes. These nodes can be used in conjunction with cpusets for coarse memory resource management. The old command-line option is still supported: numa=fake=32 gives 32 fake NUMA nodes, ignoring the NUMA setup of the actual machine. But now you may configure your system for the node sizes of your choice: numa=fake=2*512,1024,2*256 gives two 512M nodes, one 1024M node, two 256M nodes, and the rest of system memory to a sixth node. The existing hash function is maintained to support the various node sizes that are possible with this implementation. Each node of the same size receives roughly the same amount of available pages, regardless of any reserved memory with its address range. The total available pages on the system is calculated and divided by the number of equal nodes to allocate. These nodes are then dynamically allocated and their borders extended until such time as their number of available pages reaches the required size. Configurable node sizes are recommended when used in conjunction with cpusets for memory control because it eliminates the overhead associated with scanning the zonelists of many smaller full nodes on page_alloc(). Cc: Andi Kleen <ak@suse.de> Signed-off-by: David Rientjes <rientjes@google.com> Signed-off-by: Andi Kleen <ak@suse.de> Cc: Paul Jackson <pj@sgi.com> Cc: Christoph Lameter <clameter@engr.sgi.com> Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
2007-05-02 17:27:09 +00:00
#endif /* CONFIG_NUMA_EMU */
x86: Export k8 physical topology To eventually interleave emulated nodes over physical nodes, we need to know the physical topology of the machine without actually registering it. This does the k8 node setup in two parts: detection and registration. NUMA emulation can then used the physical topology detected to setup the address ranges of emulated nodes accordingly. If emulation isn't used, the k8 nodes are registered as normal. Two formals are added to the x86 NUMA setup functions: `acpi' and `k8'. These represent whether ACPI or K8 NUMA has been detected; both cannot be true at the same time. This specifies to the NUMA emulation code whether an underlying physical NUMA topology exists and which interface to use. This patch deals solely with separating the k8 setup path into Northbridge detection and registration steps and leaves the ACPI changes for a subsequent patch. The `acpi' formal is added here, however, to avoid touching all the header files again in the next patch. This approach also ensures emulated nodes will not span physical nodes so the true memory latency is not misrepresented. k8_get_nodes() may now be used to export the k8 physical topology of the machine for NUMA emulation. Signed-off-by: David Rientjes <rientjes@google.com> Cc: Andreas Herrmann <andreas.herrmann3@amd.com> Cc: Yinghai Lu <yinghai@kernel.org> Cc: Balbir Singh <balbir@linux.vnet.ibm.com> Cc: Ankita Garg <ankita@in.ibm.com> Cc: Len Brown <len.brown@intel.com> LKML-Reference: <alpine.DEB.1.00.0909251518400.14754@chino.kir.corp.google.com> Signed-off-by: Ingo Molnar <mingo@elte.hu>
2009-09-25 22:20:00 +00:00
void __init initmem_init(unsigned long start_pfn, unsigned long last_pfn,
int acpi, int k8)
{
int i;
nodes_clear(node_possible_map);
x86: reenable support for system without on node0 One system doesn't have RAM for node0 installed. SRAT: PXM 0 -> APIC 0 -> Node 0 SRAT: PXM 0 -> APIC 1 -> Node 0 SRAT: PXM 1 -> APIC 2 -> Node 1 SRAT: PXM 1 -> APIC 3 -> Node 1 SRAT: Node 1 PXM 1 0-a0000 SRAT: Node 1 PXM 1 0-dd000000 SRAT: Node 1 PXM 1 0-123000000 ACPI: SLIT: nodes = 2 10 13 13 10 mapped APIC to ffffffffff5fb000 ( fee00000) Bootmem setup node 1 0000000000000000-0000000123000000 NODE_DATA [000000000000e000 - 0000000000014fff] bootmap [0000000000015000 - 00000000000395ff] pages 25 Could not find start_pfn for node 0 Pid: 0, comm: swapper Not tainted 2.6.24-smp-g5a514e21-dirty #14 Call Trace: [<ffffffff80bab498>] free_area_init_node+0x22/0x381 [<ffffffff8045ffc5>] generic_swap+0x0/0x17 [<ffffffff80bab0cc>] find_zone_movable_pfns_for_nodes+0x54/0x271 [<ffffffff80baba5f>] free_area_init_nodes+0x239/0x287 [<ffffffff80ba6311>] paging_init+0x46/0x4c [<ffffffff80b9dda5>] setup_arch+0x3c3/0x44e [<ffffffff80b978be>] start_kernel+0x6f/0x2c7 [<ffffffff80b971cc>] _sinittext+0x1cc/0x1d3 This happens because node 0 is not online, but the node state in mm/page_alloc.c has node 0 set. nodemask_t node_states[NR_NODE_STATES] __read_mostly = { [N_POSSIBLE] = NODE_MASK_ALL, [N_ONLINE] = { { [0] = 1UL } }, So we need to clear node_online_map before initializing the memory. Signed-off-by: Yinghai Lu <yinghai.lu@sun.com> Cc: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Ingo Molnar <mingo@elte.hu> Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
2008-02-17 10:02:21 +00:00
nodes_clear(node_online_map);
#ifdef CONFIG_NUMA_EMU
x86: Interleave emulated nodes over physical nodes Add interleaved NUMA emulation support This patch interleaves emulated nodes over the system's physical nodes. This is required for interleave optimizations since mempolicies, for example, operate by iterating over a nodemask and act without knowledge of node distances. It can also be used for testing memory latencies and NUMA bugs in the kernel. There're a couple of ways to do this: - divide the number of emulated nodes by the number of physical nodes and allocate the result on each physical node, or - allocate each successive emulated node on a different physical node until all memory is exhausted. The disadvantage of the first option is, depending on the asymmetry in node capacities of each physical node, emulated nodes may substantially differ in size on a particular physical node compared to another. The disadvantage of the second option is, also depending on the asymmetry in node capacities of each physical node, there may be more emulated nodes allocated on a single physical node as another. This patch implements the second option; we sacrifice the possibility that we may have slightly more emulated nodes on a particular physical node compared to another in lieu of node size asymmetry. [ Note that "node capacity" of a physical node is not only a function of its addressable range, but also is affected by subtracting out the amount of reserved memory over that range. NUMA emulation only deals with available, non-reserved memory quantities. ] We ensure there is at least a minimal amount of available memory allocated to each node. We also make sure that at least this amount of available memory is available in ZONE_DMA32 for any node that includes both ZONE_DMA32 and ZONE_NORMAL. This patch also cleans the emulation code up by no longer passing the statically allocated struct bootnode array among the various functions. This init.data array is not allocated on the stack since it may be very large and thus it may be accessed at file scope. The WARN_ON() for nodes_cover_memory() when faking proximity domains is removed since it relies on successive nodes always having greater start addresses than previous nodes; with interleaving this is no longer always true. Signed-off-by: David Rientjes <rientjes@google.com> Cc: Linus Torvalds <torvalds@linux-foundation.org> Cc: Andreas Herrmann <andreas.herrmann3@amd.com> Cc: Yinghai Lu <yinghai@kernel.org> Cc: Balbir Singh <balbir@linux.vnet.ibm.com> Cc: Ankita Garg <ankita@in.ibm.com> Cc: Len Brown <len.brown@intel.com> LKML-Reference: <alpine.DEB.1.00.0909251519150.14754@chino.kir.corp.google.com> Signed-off-by: Ingo Molnar <mingo@elte.hu>
2009-09-25 22:20:09 +00:00
if (cmdline && !numa_emulation(start_pfn, last_pfn, acpi, k8))
return;
nodes_clear(node_possible_map);
x86: reenable support for system without on node0 One system doesn't have RAM for node0 installed. SRAT: PXM 0 -> APIC 0 -> Node 0 SRAT: PXM 0 -> APIC 1 -> Node 0 SRAT: PXM 1 -> APIC 2 -> Node 1 SRAT: PXM 1 -> APIC 3 -> Node 1 SRAT: Node 1 PXM 1 0-a0000 SRAT: Node 1 PXM 1 0-dd000000 SRAT: Node 1 PXM 1 0-123000000 ACPI: SLIT: nodes = 2 10 13 13 10 mapped APIC to ffffffffff5fb000 ( fee00000) Bootmem setup node 1 0000000000000000-0000000123000000 NODE_DATA [000000000000e000 - 0000000000014fff] bootmap [0000000000015000 - 00000000000395ff] pages 25 Could not find start_pfn for node 0 Pid: 0, comm: swapper Not tainted 2.6.24-smp-g5a514e21-dirty #14 Call Trace: [<ffffffff80bab498>] free_area_init_node+0x22/0x381 [<ffffffff8045ffc5>] generic_swap+0x0/0x17 [<ffffffff80bab0cc>] find_zone_movable_pfns_for_nodes+0x54/0x271 [<ffffffff80baba5f>] free_area_init_nodes+0x239/0x287 [<ffffffff80ba6311>] paging_init+0x46/0x4c [<ffffffff80b9dda5>] setup_arch+0x3c3/0x44e [<ffffffff80b978be>] start_kernel+0x6f/0x2c7 [<ffffffff80b971cc>] _sinittext+0x1cc/0x1d3 This happens because node 0 is not online, but the node state in mm/page_alloc.c has node 0 set. nodemask_t node_states[NR_NODE_STATES] __read_mostly = { [N_POSSIBLE] = NODE_MASK_ALL, [N_ONLINE] = { { [0] = 1UL } }, So we need to clear node_online_map before initializing the memory. Signed-off-by: Yinghai Lu <yinghai.lu@sun.com> Cc: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Ingo Molnar <mingo@elte.hu> Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
2008-02-17 10:02:21 +00:00
nodes_clear(node_online_map);
#endif
#ifdef CONFIG_ACPI_NUMA
if (!numa_off && acpi && !acpi_scan_nodes(start_pfn << PAGE_SHIFT,
last_pfn << PAGE_SHIFT))
return;
nodes_clear(node_possible_map);
x86: reenable support for system without on node0 One system doesn't have RAM for node0 installed. SRAT: PXM 0 -> APIC 0 -> Node 0 SRAT: PXM 0 -> APIC 1 -> Node 0 SRAT: PXM 1 -> APIC 2 -> Node 1 SRAT: PXM 1 -> APIC 3 -> Node 1 SRAT: Node 1 PXM 1 0-a0000 SRAT: Node 1 PXM 1 0-dd000000 SRAT: Node 1 PXM 1 0-123000000 ACPI: SLIT: nodes = 2 10 13 13 10 mapped APIC to ffffffffff5fb000 ( fee00000) Bootmem setup node 1 0000000000000000-0000000123000000 NODE_DATA [000000000000e000 - 0000000000014fff] bootmap [0000000000015000 - 00000000000395ff] pages 25 Could not find start_pfn for node 0 Pid: 0, comm: swapper Not tainted 2.6.24-smp-g5a514e21-dirty #14 Call Trace: [<ffffffff80bab498>] free_area_init_node+0x22/0x381 [<ffffffff8045ffc5>] generic_swap+0x0/0x17 [<ffffffff80bab0cc>] find_zone_movable_pfns_for_nodes+0x54/0x271 [<ffffffff80baba5f>] free_area_init_nodes+0x239/0x287 [<ffffffff80ba6311>] paging_init+0x46/0x4c [<ffffffff80b9dda5>] setup_arch+0x3c3/0x44e [<ffffffff80b978be>] start_kernel+0x6f/0x2c7 [<ffffffff80b971cc>] _sinittext+0x1cc/0x1d3 This happens because node 0 is not online, but the node state in mm/page_alloc.c has node 0 set. nodemask_t node_states[NR_NODE_STATES] __read_mostly = { [N_POSSIBLE] = NODE_MASK_ALL, [N_ONLINE] = { { [0] = 1UL } }, So we need to clear node_online_map before initializing the memory. Signed-off-by: Yinghai Lu <yinghai.lu@sun.com> Cc: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Ingo Molnar <mingo@elte.hu> Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
2008-02-17 10:02:21 +00:00
nodes_clear(node_online_map);
#endif
#ifdef CONFIG_K8_NUMA
x86: Export k8 physical topology To eventually interleave emulated nodes over physical nodes, we need to know the physical topology of the machine without actually registering it. This does the k8 node setup in two parts: detection and registration. NUMA emulation can then used the physical topology detected to setup the address ranges of emulated nodes accordingly. If emulation isn't used, the k8 nodes are registered as normal. Two formals are added to the x86 NUMA setup functions: `acpi' and `k8'. These represent whether ACPI or K8 NUMA has been detected; both cannot be true at the same time. This specifies to the NUMA emulation code whether an underlying physical NUMA topology exists and which interface to use. This patch deals solely with separating the k8 setup path into Northbridge detection and registration steps and leaves the ACPI changes for a subsequent patch. The `acpi' formal is added here, however, to avoid touching all the header files again in the next patch. This approach also ensures emulated nodes will not span physical nodes so the true memory latency is not misrepresented. k8_get_nodes() may now be used to export the k8 physical topology of the machine for NUMA emulation. Signed-off-by: David Rientjes <rientjes@google.com> Cc: Andreas Herrmann <andreas.herrmann3@amd.com> Cc: Yinghai Lu <yinghai@kernel.org> Cc: Balbir Singh <balbir@linux.vnet.ibm.com> Cc: Ankita Garg <ankita@in.ibm.com> Cc: Len Brown <len.brown@intel.com> LKML-Reference: <alpine.DEB.1.00.0909251518400.14754@chino.kir.corp.google.com> Signed-off-by: Ingo Molnar <mingo@elte.hu>
2009-09-25 22:20:00 +00:00
if (!numa_off && k8 && !k8_scan_nodes())
return;
nodes_clear(node_possible_map);
x86: reenable support for system without on node0 One system doesn't have RAM for node0 installed. SRAT: PXM 0 -> APIC 0 -> Node 0 SRAT: PXM 0 -> APIC 1 -> Node 0 SRAT: PXM 1 -> APIC 2 -> Node 1 SRAT: PXM 1 -> APIC 3 -> Node 1 SRAT: Node 1 PXM 1 0-a0000 SRAT: Node 1 PXM 1 0-dd000000 SRAT: Node 1 PXM 1 0-123000000 ACPI: SLIT: nodes = 2 10 13 13 10 mapped APIC to ffffffffff5fb000 ( fee00000) Bootmem setup node 1 0000000000000000-0000000123000000 NODE_DATA [000000000000e000 - 0000000000014fff] bootmap [0000000000015000 - 00000000000395ff] pages 25 Could not find start_pfn for node 0 Pid: 0, comm: swapper Not tainted 2.6.24-smp-g5a514e21-dirty #14 Call Trace: [<ffffffff80bab498>] free_area_init_node+0x22/0x381 [<ffffffff8045ffc5>] generic_swap+0x0/0x17 [<ffffffff80bab0cc>] find_zone_movable_pfns_for_nodes+0x54/0x271 [<ffffffff80baba5f>] free_area_init_nodes+0x239/0x287 [<ffffffff80ba6311>] paging_init+0x46/0x4c [<ffffffff80b9dda5>] setup_arch+0x3c3/0x44e [<ffffffff80b978be>] start_kernel+0x6f/0x2c7 [<ffffffff80b971cc>] _sinittext+0x1cc/0x1d3 This happens because node 0 is not online, but the node state in mm/page_alloc.c has node 0 set. nodemask_t node_states[NR_NODE_STATES] __read_mostly = { [N_POSSIBLE] = NODE_MASK_ALL, [N_ONLINE] = { { [0] = 1UL } }, So we need to clear node_online_map before initializing the memory. Signed-off-by: Yinghai Lu <yinghai.lu@sun.com> Cc: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Ingo Molnar <mingo@elte.hu> Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
2008-02-17 10:02:21 +00:00
nodes_clear(node_online_map);
#endif
printk(KERN_INFO "%s\n",
numa_off ? "NUMA turned off" : "No NUMA configuration found");
printk(KERN_INFO "Faking a node at %016lx-%016lx\n",
start_pfn << PAGE_SHIFT,
last_pfn << PAGE_SHIFT);
/* setup dummy node covering all memory */
memnode_shift = 63;
memnodemap = memnode.embedded_map;
memnodemap[0] = 0;
node_set_online(0);
node_set(0, node_possible_map);
for (i = 0; i < nr_cpu_ids; i++)
numa_set_node(i, 0);
e820_register_active_regions(0, start_pfn, last_pfn);
setup_node_bootmem(0, start_pfn << PAGE_SHIFT, last_pfn << PAGE_SHIFT);
}
unsigned long __init numa_free_all_bootmem(void)
{
unsigned long pages = 0;
int i;
for_each_online_node(i)
pages += free_all_bootmem_node(NODE_DATA(i));
#ifdef CONFIG_NO_BOOTMEM
pages += free_all_memory_core_early(MAX_NUMNODES);
#endif
return pages;
}
static __init int numa_setup(char *opt)
{
if (!opt)
return -EINVAL;
if (!strncmp(opt, "off", 3))
numa_off = 1;
#ifdef CONFIG_NUMA_EMU
[PATCH] x86-64: configurable fake numa node sizes Extends the numa=fake x86_64 command-line option to allow for configurable node sizes. These nodes can be used in conjunction with cpusets for coarse memory resource management. The old command-line option is still supported: numa=fake=32 gives 32 fake NUMA nodes, ignoring the NUMA setup of the actual machine. But now you may configure your system for the node sizes of your choice: numa=fake=2*512,1024,2*256 gives two 512M nodes, one 1024M node, two 256M nodes, and the rest of system memory to a sixth node. The existing hash function is maintained to support the various node sizes that are possible with this implementation. Each node of the same size receives roughly the same amount of available pages, regardless of any reserved memory with its address range. The total available pages on the system is calculated and divided by the number of equal nodes to allocate. These nodes are then dynamically allocated and their borders extended until such time as their number of available pages reaches the required size. Configurable node sizes are recommended when used in conjunction with cpusets for memory control because it eliminates the overhead associated with scanning the zonelists of many smaller full nodes on page_alloc(). Cc: Andi Kleen <ak@suse.de> Signed-off-by: David Rientjes <rientjes@google.com> Signed-off-by: Andi Kleen <ak@suse.de> Cc: Paul Jackson <pj@sgi.com> Cc: Christoph Lameter <clameter@engr.sgi.com> Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
2007-05-02 17:27:09 +00:00
if (!strncmp(opt, "fake=", 5))
cmdline = opt + 5;
#endif
#ifdef CONFIG_ACPI_NUMA
if (!strncmp(opt, "noacpi", 6))
acpi_numa = -1;
#endif
return 0;
}
early_param("numa", numa_setup);
x86: cleanup early per cpu variables/accesses v4 * Introduce a new PER_CPU macro called "EARLY_PER_CPU". This is used by some per_cpu variables that are initialized and accessed before there are per_cpu areas allocated. ["Early" in respect to per_cpu variables is "earlier than the per_cpu areas have been setup".] This patchset adds these new macros: DEFINE_EARLY_PER_CPU(_type, _name, _initvalue) EXPORT_EARLY_PER_CPU_SYMBOL(_name) DECLARE_EARLY_PER_CPU(_type, _name) early_per_cpu_ptr(_name) early_per_cpu_map(_name, _idx) early_per_cpu(_name, _cpu) The DEFINE macro defines the per_cpu variable as well as the early map and pointer. It also initializes the per_cpu variable and map elements to "_initvalue". The early_* macros provide access to the initial map (usually setup during system init) and the early pointer. This pointer is initialized to point to the early map but is then NULL'ed when the actual per_cpu areas are setup. After that the per_cpu variable is the correct access to the variable. The early_per_cpu() macro is not very efficient but does show how to access the variable if you have a function that can be called both "early" and "late". It tests the early ptr to be NULL, and if not then it's still valid. Otherwise, the per_cpu variable is used instead: #define early_per_cpu(_name, _cpu) \ (early_per_cpu_ptr(_name) ? \ early_per_cpu_ptr(_name)[_cpu] : \ per_cpu(_name, _cpu)) A better method is to actually check the pointer manually. In the case below, numa_set_node can be called both "early" and "late": void __cpuinit numa_set_node(int cpu, int node) { int *cpu_to_node_map = early_per_cpu_ptr(x86_cpu_to_node_map); if (cpu_to_node_map) cpu_to_node_map[cpu] = node; else per_cpu(x86_cpu_to_node_map, cpu) = node; } * Add a flag "arch_provides_topology_pointers" that indicates pointers to topology cpumask_t maps are available. Otherwise, use the function returning the cpumask_t value. This is useful if cpumask_t set size is very large to avoid copying data on to/off of the stack. * The coverage of CONFIG_DEBUG_PER_CPU_MAPS has been increased while the non-debug case has been optimized a bit. * Remove an unreferenced compiler warning in drivers/base/topology.c * Clean up #ifdef in setup.c For inclusion into sched-devel/latest tree. Based on: git://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux-2.6.git + sched-devel/latest .../mingo/linux-2.6-sched-devel.git Signed-off-by: Mike Travis <travis@sgi.com> Signed-off-by: Ingo Molnar <mingo@elte.hu> Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
2008-05-12 19:21:12 +00:00
#ifdef CONFIG_NUMA
static __init int find_near_online_node(int node)
{
int n, val;
int min_val = INT_MAX;
int best_node = -1;
for_each_online_node(n) {
val = node_distance(node, n);
if (val < min_val) {
min_val = val;
best_node = n;
}
}
return best_node;
}
/*
* Setup early cpu_to_node.
*
* Populate cpu_to_node[] only if x86_cpu_to_apicid[],
* and apicid_to_node[] tables have valid entries for a CPU.
* This means we skip cpu_to_node[] initialisation for NUMA
* emulation and faking node case (when running a kernel compiled
* for NUMA on a non NUMA box), which is OK as cpu_to_node[]
* is already initialized in a round robin manner at numa_init_array,
* prior to this call, and this initialization is good enough
* for the fake NUMA cases.
x86: cleanup early per cpu variables/accesses v4 * Introduce a new PER_CPU macro called "EARLY_PER_CPU". This is used by some per_cpu variables that are initialized and accessed before there are per_cpu areas allocated. ["Early" in respect to per_cpu variables is "earlier than the per_cpu areas have been setup".] This patchset adds these new macros: DEFINE_EARLY_PER_CPU(_type, _name, _initvalue) EXPORT_EARLY_PER_CPU_SYMBOL(_name) DECLARE_EARLY_PER_CPU(_type, _name) early_per_cpu_ptr(_name) early_per_cpu_map(_name, _idx) early_per_cpu(_name, _cpu) The DEFINE macro defines the per_cpu variable as well as the early map and pointer. It also initializes the per_cpu variable and map elements to "_initvalue". The early_* macros provide access to the initial map (usually setup during system init) and the early pointer. This pointer is initialized to point to the early map but is then NULL'ed when the actual per_cpu areas are setup. After that the per_cpu variable is the correct access to the variable. The early_per_cpu() macro is not very efficient but does show how to access the variable if you have a function that can be called both "early" and "late". It tests the early ptr to be NULL, and if not then it's still valid. Otherwise, the per_cpu variable is used instead: #define early_per_cpu(_name, _cpu) \ (early_per_cpu_ptr(_name) ? \ early_per_cpu_ptr(_name)[_cpu] : \ per_cpu(_name, _cpu)) A better method is to actually check the pointer manually. In the case below, numa_set_node can be called both "early" and "late": void __cpuinit numa_set_node(int cpu, int node) { int *cpu_to_node_map = early_per_cpu_ptr(x86_cpu_to_node_map); if (cpu_to_node_map) cpu_to_node_map[cpu] = node; else per_cpu(x86_cpu_to_node_map, cpu) = node; } * Add a flag "arch_provides_topology_pointers" that indicates pointers to topology cpumask_t maps are available. Otherwise, use the function returning the cpumask_t value. This is useful if cpumask_t set size is very large to avoid copying data on to/off of the stack. * The coverage of CONFIG_DEBUG_PER_CPU_MAPS has been increased while the non-debug case has been optimized a bit. * Remove an unreferenced compiler warning in drivers/base/topology.c * Clean up #ifdef in setup.c For inclusion into sched-devel/latest tree. Based on: git://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux-2.6.git + sched-devel/latest .../mingo/linux-2.6-sched-devel.git Signed-off-by: Mike Travis <travis@sgi.com> Signed-off-by: Ingo Molnar <mingo@elte.hu> Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
2008-05-12 19:21:12 +00:00
*
* Called before the per_cpu areas are setup.
*/
void __init init_cpu_to_node(void)
{
x86: cleanup early per cpu variables/accesses v4 * Introduce a new PER_CPU macro called "EARLY_PER_CPU". This is used by some per_cpu variables that are initialized and accessed before there are per_cpu areas allocated. ["Early" in respect to per_cpu variables is "earlier than the per_cpu areas have been setup".] This patchset adds these new macros: DEFINE_EARLY_PER_CPU(_type, _name, _initvalue) EXPORT_EARLY_PER_CPU_SYMBOL(_name) DECLARE_EARLY_PER_CPU(_type, _name) early_per_cpu_ptr(_name) early_per_cpu_map(_name, _idx) early_per_cpu(_name, _cpu) The DEFINE macro defines the per_cpu variable as well as the early map and pointer. It also initializes the per_cpu variable and map elements to "_initvalue". The early_* macros provide access to the initial map (usually setup during system init) and the early pointer. This pointer is initialized to point to the early map but is then NULL'ed when the actual per_cpu areas are setup. After that the per_cpu variable is the correct access to the variable. The early_per_cpu() macro is not very efficient but does show how to access the variable if you have a function that can be called both "early" and "late". It tests the early ptr to be NULL, and if not then it's still valid. Otherwise, the per_cpu variable is used instead: #define early_per_cpu(_name, _cpu) \ (early_per_cpu_ptr(_name) ? \ early_per_cpu_ptr(_name)[_cpu] : \ per_cpu(_name, _cpu)) A better method is to actually check the pointer manually. In the case below, numa_set_node can be called both "early" and "late": void __cpuinit numa_set_node(int cpu, int node) { int *cpu_to_node_map = early_per_cpu_ptr(x86_cpu_to_node_map); if (cpu_to_node_map) cpu_to_node_map[cpu] = node; else per_cpu(x86_cpu_to_node_map, cpu) = node; } * Add a flag "arch_provides_topology_pointers" that indicates pointers to topology cpumask_t maps are available. Otherwise, use the function returning the cpumask_t value. This is useful if cpumask_t set size is very large to avoid copying data on to/off of the stack. * The coverage of CONFIG_DEBUG_PER_CPU_MAPS has been increased while the non-debug case has been optimized a bit. * Remove an unreferenced compiler warning in drivers/base/topology.c * Clean up #ifdef in setup.c For inclusion into sched-devel/latest tree. Based on: git://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux-2.6.git + sched-devel/latest .../mingo/linux-2.6-sched-devel.git Signed-off-by: Mike Travis <travis@sgi.com> Signed-off-by: Ingo Molnar <mingo@elte.hu> Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
2008-05-12 19:21:12 +00:00
int cpu;
u16 *cpu_to_apicid = early_per_cpu_ptr(x86_cpu_to_apicid);
x86: cleanup early per cpu variables/accesses v4 * Introduce a new PER_CPU macro called "EARLY_PER_CPU". This is used by some per_cpu variables that are initialized and accessed before there are per_cpu areas allocated. ["Early" in respect to per_cpu variables is "earlier than the per_cpu areas have been setup".] This patchset adds these new macros: DEFINE_EARLY_PER_CPU(_type, _name, _initvalue) EXPORT_EARLY_PER_CPU_SYMBOL(_name) DECLARE_EARLY_PER_CPU(_type, _name) early_per_cpu_ptr(_name) early_per_cpu_map(_name, _idx) early_per_cpu(_name, _cpu) The DEFINE macro defines the per_cpu variable as well as the early map and pointer. It also initializes the per_cpu variable and map elements to "_initvalue". The early_* macros provide access to the initial map (usually setup during system init) and the early pointer. This pointer is initialized to point to the early map but is then NULL'ed when the actual per_cpu areas are setup. After that the per_cpu variable is the correct access to the variable. The early_per_cpu() macro is not very efficient but does show how to access the variable if you have a function that can be called both "early" and "late". It tests the early ptr to be NULL, and if not then it's still valid. Otherwise, the per_cpu variable is used instead: #define early_per_cpu(_name, _cpu) \ (early_per_cpu_ptr(_name) ? \ early_per_cpu_ptr(_name)[_cpu] : \ per_cpu(_name, _cpu)) A better method is to actually check the pointer manually. In the case below, numa_set_node can be called both "early" and "late": void __cpuinit numa_set_node(int cpu, int node) { int *cpu_to_node_map = early_per_cpu_ptr(x86_cpu_to_node_map); if (cpu_to_node_map) cpu_to_node_map[cpu] = node; else per_cpu(x86_cpu_to_node_map, cpu) = node; } * Add a flag "arch_provides_topology_pointers" that indicates pointers to topology cpumask_t maps are available. Otherwise, use the function returning the cpumask_t value. This is useful if cpumask_t set size is very large to avoid copying data on to/off of the stack. * The coverage of CONFIG_DEBUG_PER_CPU_MAPS has been increased while the non-debug case has been optimized a bit. * Remove an unreferenced compiler warning in drivers/base/topology.c * Clean up #ifdef in setup.c For inclusion into sched-devel/latest tree. Based on: git://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux-2.6.git + sched-devel/latest .../mingo/linux-2.6-sched-devel.git Signed-off-by: Mike Travis <travis@sgi.com> Signed-off-by: Ingo Molnar <mingo@elte.hu> Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
2008-05-12 19:21:12 +00:00
BUG_ON(cpu_to_apicid == NULL);
for_each_possible_cpu(cpu) {
int node;
x86: cleanup early per cpu variables/accesses v4 * Introduce a new PER_CPU macro called "EARLY_PER_CPU". This is used by some per_cpu variables that are initialized and accessed before there are per_cpu areas allocated. ["Early" in respect to per_cpu variables is "earlier than the per_cpu areas have been setup".] This patchset adds these new macros: DEFINE_EARLY_PER_CPU(_type, _name, _initvalue) EXPORT_EARLY_PER_CPU_SYMBOL(_name) DECLARE_EARLY_PER_CPU(_type, _name) early_per_cpu_ptr(_name) early_per_cpu_map(_name, _idx) early_per_cpu(_name, _cpu) The DEFINE macro defines the per_cpu variable as well as the early map and pointer. It also initializes the per_cpu variable and map elements to "_initvalue". The early_* macros provide access to the initial map (usually setup during system init) and the early pointer. This pointer is initialized to point to the early map but is then NULL'ed when the actual per_cpu areas are setup. After that the per_cpu variable is the correct access to the variable. The early_per_cpu() macro is not very efficient but does show how to access the variable if you have a function that can be called both "early" and "late". It tests the early ptr to be NULL, and if not then it's still valid. Otherwise, the per_cpu variable is used instead: #define early_per_cpu(_name, _cpu) \ (early_per_cpu_ptr(_name) ? \ early_per_cpu_ptr(_name)[_cpu] : \ per_cpu(_name, _cpu)) A better method is to actually check the pointer manually. In the case below, numa_set_node can be called both "early" and "late": void __cpuinit numa_set_node(int cpu, int node) { int *cpu_to_node_map = early_per_cpu_ptr(x86_cpu_to_node_map); if (cpu_to_node_map) cpu_to_node_map[cpu] = node; else per_cpu(x86_cpu_to_node_map, cpu) = node; } * Add a flag "arch_provides_topology_pointers" that indicates pointers to topology cpumask_t maps are available. Otherwise, use the function returning the cpumask_t value. This is useful if cpumask_t set size is very large to avoid copying data on to/off of the stack. * The coverage of CONFIG_DEBUG_PER_CPU_MAPS has been increased while the non-debug case has been optimized a bit. * Remove an unreferenced compiler warning in drivers/base/topology.c * Clean up #ifdef in setup.c For inclusion into sched-devel/latest tree. Based on: git://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux-2.6.git + sched-devel/latest .../mingo/linux-2.6-sched-devel.git Signed-off-by: Mike Travis <travis@sgi.com> Signed-off-by: Ingo Molnar <mingo@elte.hu> Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
2008-05-12 19:21:12 +00:00
u16 apicid = cpu_to_apicid[cpu];
if (apicid == BAD_APICID)
continue;
node = apicid_to_node[apicid];
if (node == NUMA_NO_NODE)
continue;
if (!node_online(node))
node = find_near_online_node(node);
x86: cleanup early per cpu variables/accesses v4 * Introduce a new PER_CPU macro called "EARLY_PER_CPU". This is used by some per_cpu variables that are initialized and accessed before there are per_cpu areas allocated. ["Early" in respect to per_cpu variables is "earlier than the per_cpu areas have been setup".] This patchset adds these new macros: DEFINE_EARLY_PER_CPU(_type, _name, _initvalue) EXPORT_EARLY_PER_CPU_SYMBOL(_name) DECLARE_EARLY_PER_CPU(_type, _name) early_per_cpu_ptr(_name) early_per_cpu_map(_name, _idx) early_per_cpu(_name, _cpu) The DEFINE macro defines the per_cpu variable as well as the early map and pointer. It also initializes the per_cpu variable and map elements to "_initvalue". The early_* macros provide access to the initial map (usually setup during system init) and the early pointer. This pointer is initialized to point to the early map but is then NULL'ed when the actual per_cpu areas are setup. After that the per_cpu variable is the correct access to the variable. The early_per_cpu() macro is not very efficient but does show how to access the variable if you have a function that can be called both "early" and "late". It tests the early ptr to be NULL, and if not then it's still valid. Otherwise, the per_cpu variable is used instead: #define early_per_cpu(_name, _cpu) \ (early_per_cpu_ptr(_name) ? \ early_per_cpu_ptr(_name)[_cpu] : \ per_cpu(_name, _cpu)) A better method is to actually check the pointer manually. In the case below, numa_set_node can be called both "early" and "late": void __cpuinit numa_set_node(int cpu, int node) { int *cpu_to_node_map = early_per_cpu_ptr(x86_cpu_to_node_map); if (cpu_to_node_map) cpu_to_node_map[cpu] = node; else per_cpu(x86_cpu_to_node_map, cpu) = node; } * Add a flag "arch_provides_topology_pointers" that indicates pointers to topology cpumask_t maps are available. Otherwise, use the function returning the cpumask_t value. This is useful if cpumask_t set size is very large to avoid copying data on to/off of the stack. * The coverage of CONFIG_DEBUG_PER_CPU_MAPS has been increased while the non-debug case has been optimized a bit. * Remove an unreferenced compiler warning in drivers/base/topology.c * Clean up #ifdef in setup.c For inclusion into sched-devel/latest tree. Based on: git://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux-2.6.git + sched-devel/latest .../mingo/linux-2.6-sched-devel.git Signed-off-by: Mike Travis <travis@sgi.com> Signed-off-by: Ingo Molnar <mingo@elte.hu> Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
2008-05-12 19:21:12 +00:00
numa_set_node(cpu, node);
}
}
x86: cleanup early per cpu variables/accesses v4 * Introduce a new PER_CPU macro called "EARLY_PER_CPU". This is used by some per_cpu variables that are initialized and accessed before there are per_cpu areas allocated. ["Early" in respect to per_cpu variables is "earlier than the per_cpu areas have been setup".] This patchset adds these new macros: DEFINE_EARLY_PER_CPU(_type, _name, _initvalue) EXPORT_EARLY_PER_CPU_SYMBOL(_name) DECLARE_EARLY_PER_CPU(_type, _name) early_per_cpu_ptr(_name) early_per_cpu_map(_name, _idx) early_per_cpu(_name, _cpu) The DEFINE macro defines the per_cpu variable as well as the early map and pointer. It also initializes the per_cpu variable and map elements to "_initvalue". The early_* macros provide access to the initial map (usually setup during system init) and the early pointer. This pointer is initialized to point to the early map but is then NULL'ed when the actual per_cpu areas are setup. After that the per_cpu variable is the correct access to the variable. The early_per_cpu() macro is not very efficient but does show how to access the variable if you have a function that can be called both "early" and "late". It tests the early ptr to be NULL, and if not then it's still valid. Otherwise, the per_cpu variable is used instead: #define early_per_cpu(_name, _cpu) \ (early_per_cpu_ptr(_name) ? \ early_per_cpu_ptr(_name)[_cpu] : \ per_cpu(_name, _cpu)) A better method is to actually check the pointer manually. In the case below, numa_set_node can be called both "early" and "late": void __cpuinit numa_set_node(int cpu, int node) { int *cpu_to_node_map = early_per_cpu_ptr(x86_cpu_to_node_map); if (cpu_to_node_map) cpu_to_node_map[cpu] = node; else per_cpu(x86_cpu_to_node_map, cpu) = node; } * Add a flag "arch_provides_topology_pointers" that indicates pointers to topology cpumask_t maps are available. Otherwise, use the function returning the cpumask_t value. This is useful if cpumask_t set size is very large to avoid copying data on to/off of the stack. * The coverage of CONFIG_DEBUG_PER_CPU_MAPS has been increased while the non-debug case has been optimized a bit. * Remove an unreferenced compiler warning in drivers/base/topology.c * Clean up #ifdef in setup.c For inclusion into sched-devel/latest tree. Based on: git://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux-2.6.git + sched-devel/latest .../mingo/linux-2.6-sched-devel.git Signed-off-by: Mike Travis <travis@sgi.com> Signed-off-by: Ingo Molnar <mingo@elte.hu> Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
2008-05-12 19:21:12 +00:00
#endif
void __cpuinit numa_set_node(int cpu, int node)
{
int *cpu_to_node_map = early_per_cpu_ptr(x86_cpu_to_node_map);
/* early setting, no percpu area yet */
if (cpu_to_node_map) {
cpu_to_node_map[cpu] = node;
return;
}
#ifdef CONFIG_DEBUG_PER_CPU_MAPS
if (cpu >= nr_cpu_ids || !cpu_possible(cpu)) {
printk(KERN_ERR "numa_set_node: invalid cpu# (%d)\n", cpu);
dump_stack();
return;
}
#endif
per_cpu(x86_cpu_to_node_map, cpu) = node;
if (node != NUMA_NO_NODE)
set_cpu_numa_node(cpu, node);
}
void __cpuinit numa_clear_node(int cpu)
{
numa_set_node(cpu, NUMA_NO_NODE);
}
#ifndef CONFIG_DEBUG_PER_CPU_MAPS
void __cpuinit numa_add_cpu(int cpu)
{
cpumask_set_cpu(cpu, node_to_cpumask_map[early_cpu_to_node(cpu)]);
}
void __cpuinit numa_remove_cpu(int cpu)
{
cpumask_clear_cpu(cpu, node_to_cpumask_map[early_cpu_to_node(cpu)]);
}
#else /* CONFIG_DEBUG_PER_CPU_MAPS */
/*
* --------- debug versions of the numa functions ---------
*/
static void __cpuinit numa_set_cpumask(int cpu, int enable)
{
int node = early_cpu_to_node(cpu);
struct cpumask *mask;
char buf[64];
mask = node_to_cpumask_map[node];
if (mask == NULL) {
printk(KERN_ERR "node_to_cpumask_map[%i] NULL\n", node);
dump_stack();
return;
}
if (enable)
cpumask_set_cpu(cpu, mask);
else
cpumask_clear_cpu(cpu, mask);
cpulist_scnprintf(buf, sizeof(buf), mask);
printk(KERN_DEBUG "%s cpu %d node %d: mask now %s\n",
enable ? "numa_add_cpu" : "numa_remove_cpu", cpu, node, buf);
}
void __cpuinit numa_add_cpu(int cpu)
{
numa_set_cpumask(cpu, 1);
}
void __cpuinit numa_remove_cpu(int cpu)
{
numa_set_cpumask(cpu, 0);
}
int __cpu_to_node(int cpu)
{
if (early_per_cpu_ptr(x86_cpu_to_node_map)) {
printk(KERN_WARNING
"cpu_to_node(%d): usage too early!\n", cpu);
dump_stack();
return early_per_cpu_ptr(x86_cpu_to_node_map)[cpu];
}
return per_cpu(x86_cpu_to_node_map, cpu);
}
EXPORT_SYMBOL(__cpu_to_node);
/*
* Same function as cpu_to_node() but used if called before the
* per_cpu areas are setup.
*/
int early_cpu_to_node(int cpu)
{
if (early_per_cpu_ptr(x86_cpu_to_node_map))
return early_per_cpu_ptr(x86_cpu_to_node_map)[cpu];
if (!cpu_possible(cpu)) {
printk(KERN_WARNING
"early_cpu_to_node(%d): no per_cpu area!\n", cpu);
dump_stack();
return NUMA_NO_NODE;
}
return per_cpu(x86_cpu_to_node_map, cpu);
}
/*
* --------- end of debug versions of the numa functions ---------
*/
#endif /* CONFIG_DEBUG_PER_CPU_MAPS */