[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index] [PATCH v3 09/17] xen/arm: introduce a helper to parse device tree NUMA distance map
From: Wei Chen <wei.chen@xxxxxxx> A NUMA aware device tree will provide a "distance-map" node to describe distance between any two nodes. This patch introduce a new helper to parse this distance map. Signed-off-by: Wei Chen <wei.chen@xxxxxxx> Signed-off-by: Henry Wang <Henry.Wang@xxxxxxx> --- v2 -> v3: 1. No change. v1 -> v2: 1. Get rid of useless braces. 2. Use new NUMA status helper. 3. Use PRIu32 to replace u in print messages. 4. Fix opposite = __node_distance(to, from). 5. disable dtb numa info table when we find an invalid data in dtb. --- xen/arch/arm/numa_device_tree.c | 107 ++++++++++++++++++++++++++++++++ 1 file changed, 107 insertions(+) diff --git a/xen/arch/arm/numa_device_tree.c b/xen/arch/arm/numa_device_tree.c index 9796490747..718afca826 100644 --- a/xen/arch/arm/numa_device_tree.c +++ b/xen/arch/arm/numa_device_tree.c @@ -151,3 +151,110 @@ invalid_data: numa_fw_bad(); return -EINVAL; } + +/* Parse NUMA distance map v1 */ +static int __init fdt_parse_numa_distance_map_v1(const void *fdt, int node) +{ + const struct fdt_property *prop; + const __be32 *matrix; + unsigned int i, entry_count; + int len; + + printk(XENLOG_INFO "NUMA: parsing numa-distance-map\n"); + + prop = fdt_get_property(fdt, node, "distance-matrix", &len); + if ( !prop ) + { + printk(XENLOG_WARNING + "NUMA: No distance-matrix property in distance-map\n"); + goto invalid_data; + } + + if ( len % sizeof(__be32) != 0 ) + { + printk(XENLOG_WARNING + "distance-matrix in node is not a multiple of u32\n"); + goto invalid_data; + } + + entry_count = len / sizeof(__be32); + if ( entry_count == 0 ) + { + printk(XENLOG_WARNING "NUMA: Invalid distance-matrix\n"); + goto invalid_data; + } + + matrix = (const __be32 *)prop->data; + for ( i = 0; i + 2 < entry_count; i += 3 ) + { + unsigned int from, to, distance, opposite; + + from = dt_next_cell(1, &matrix); + to = dt_next_cell(1, &matrix); + distance = dt_next_cell(1, &matrix); + if ( (from == to && distance != NUMA_LOCAL_DISTANCE) || + (from != to && distance <= NUMA_LOCAL_DISTANCE) ) + { + printk(XENLOG_WARNING + "NUMA: Invalid distance: NODE#%"PRIu32"->NODE#%"PRIu32":%"PRIu32"\n", + from, to, distance); + goto invalid_data; + } + + printk(XENLOG_INFO "NUMA: distance: NODE#%"PRIu32"->NODE#%"PRIu32":%"PRIu32"\n", + from, to, distance); + + /* Get opposite way distance */ + opposite = __node_distance(to, from); + if ( opposite == 0 ) + { + /* Bi-directions are not set, set both */ + numa_set_distance(from, to, distance); + numa_set_distance(to, from, distance); + } + else + { + /* + * Opposite way distance has been set to a different value. + * It may be a firmware device tree bug? + */ + if ( opposite != distance ) + { + /* + * In device tree NUMA distance-matrix binding: + * https://www.kernel.org/doc/Documentation/devicetree/bindings/numa.txt + * There is a notes mentions: + * "Each entry represents distance from first node to + * second node. The distances are equal in either + * direction." + * + * That means device tree doesn't permit this case. + * But in ACPI spec, it cares to specifically permit this + * case: + * "Except for the relative distance from a System Locality + * to itself, each relative distance is stored twice in the + * matrix. This provides the capability to describe the + * scenario where the relative distances for the two + * directions between System Localities is different." + * + * That means a real machine allows such NUMA configuration. + * So, place a WARNING here to notice system administrators, + * is it the specail case that they hijack the device tree + * to support their rare machines? + */ + printk(XENLOG_WARNING + "Un-matched bi-direction! NODE#%"PRIu32"->NODE#%"PRIu32":%"PRIu32", NODE#%"PRIu32"->NODE#%"PRIu32":%"PRIu32"\n", + from, to, distance, to, from, opposite); + } + + /* Opposite way distance has been set, just set this way */ + numa_set_distance(from, to, distance); + } + } + + return 0; + +invalid_data: + numa_fw_bad(); + return -EINVAL; +} -- 2.25.1
|
Lists.xenproject.org is hosted with RackSpace, monitoring our |