i86pc/os/lgrpplat.c

96  *					maximum possible number of NUMA nodes
112  * above as long as the proximity domains IDs are numbered from 0 to <number of
114  * from 0 to <number of nodes - 1>.  Then proximity ID N will hash into node ID
117  * from 0 to <number of nodes - 1>, then hashing the proximity domain IDs into
121  * <number of nodes - 1> to minimize cost of searching and potentially space.
196 #define	LGRP_PLAT_PROBE_NREADS		256	/* number of vendor ID reads */
206  * Hash proximity domain ID into node to domain mapping table "mod" number of
282  * CPU to node ID mapping table (only used for SRAT) and its max number of
357  * Maximum possible number of nodes in system
1042  * This latency number can only be used for relative comparison
1046  * number was at least proportional to make comparisons more meaningful though.
1103  * Return the maximum number of lgrps supported by the platform.
1104  * Before lgrp topology is known it returns an estimate based on the number of
1106  * 1) the actual maximim number of lgrps created if CPU/memory DR operations
1108  * 2) the maximum possible number of lgrps if CPU/memory DR operations are
1124  * Count number of memory pages (_t) based on mnode id (_n) and query type (_t).
1144  * Return the number of free pages in an lgroup.
1146  * For query of LGRP_MEM_SIZE_FREE, return the number of base pagesize
1148  * number of allocatable base pagesize pages corresponding to the
1185  * corresponding to the given page frame number
1213  * nodes probing from each node to each of the other nodes some number of
1214  * times.  Furthermore, each node is probed some number of times before moving
1358 	 * Get node number for proximity domain
1405 	 * to user limits number of CPUs powered on at boot by specifying the
1451  * Return node number for given proximity domain/system locality
1488 	 * determine number of CPUs
1493 	 * Determine which CPUs and memory are local to each other and number
1653 	 * than number of nodes on a system with 4 or more
2120  * Return the number of free, allocatable, or installed
2193 	 * Ran out of supported number of entries which shouldn't happen....
2212 	 * Get node number for proximity domain
2356 	 * Just use bubble sort since number of nodes is small
2483 	 * and take specified number of samples
2550  * and return number of CPU APIC IDs.
2575 	 * Calculate number of entries in array and return when the system is
2590 	 * Just return number of CPU APIC IDs if CPU to node mapping table is
2615 	 * Return number of CPUs based on number of APIC IDs
2646 	localities = tp->number;
2794  * and memory are local to each other in the same NUMA node and return number
2824 		 * Determine number of nodes by counting number of proximity
2830 	 * Return if number of nodes is 1 or less since don't need to read SRAT.
3003  * Return number of proximity domains given in ACPI SRAT
3268  * machine are the node ID register for the number of NUMA nodes and the DRAM
3318  * number of nodes in system, etc. for Opteron K8.  The following
3386 #define	OPT_PCS_DEV_NODE0		24	/* device number for node 0 */
3476 	 * If number of nodes is more than maximum supported, then set node
3543 		 * Increment device number to next node and register offsets