summaryrefslogtreecommitdiff
path: root/drivers/gpu/drm/amd/amdkfd/kfd_topology.c
diff options
context:
space:
mode:
Diffstat (limited to 'drivers/gpu/drm/amd/amdkfd/kfd_topology.c')
-rw-r--r--drivers/gpu/drm/amd/amdkfd/kfd_topology.c77
1 files changed, 53 insertions, 24 deletions
diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_topology.c b/drivers/gpu/drm/amd/amdkfd/kfd_topology.c
index c51f131eaa2f..6f89b06f89d3 100644
--- a/drivers/gpu/drm/amd/amdkfd/kfd_topology.c
+++ b/drivers/gpu/drm/amd/amdkfd/kfd_topology.c
@@ -31,6 +31,7 @@
#include <linux/log2.h>
#include <linux/dmi.h>
#include <linux/atomic.h>
+#include <linux/crc16.h>
#include "kfd_priv.h"
#include "kfd_crat.h"
@@ -958,8 +959,7 @@ static void kfd_update_system_properties(void)
dev = list_last_entry(&topology_device_list,
struct kfd_topology_device, list);
if (dev) {
- sys_props.platform_id =
- (*((uint64_t *)dev->oem_id)) & CRAT_OEMID_64BIT_MASK;
+ sys_props.platform_id = dev->oem_id64;
sys_props.platform_oem = *((uint64_t *)dev->oem_table_id);
sys_props.platform_rev = dev->oem_revision;
}
@@ -1092,14 +1092,17 @@ void kfd_topology_shutdown(void)
static uint32_t kfd_generate_gpu_id(struct kfd_node *gpu)
{
- uint32_t hashout;
+ uint32_t gpu_id;
uint32_t buf[8];
uint64_t local_mem_size;
- int i;
+ struct kfd_topology_device *dev;
+ bool is_unique;
+ uint8_t *crc_buf;
if (!gpu)
return 0;
+ crc_buf = (uint8_t *)&buf;
local_mem_size = gpu->local_mem_info.local_mem_size_private +
gpu->local_mem_info.local_mem_size_public;
buf[0] = gpu->adev->pdev->devfn;
@@ -1112,10 +1115,34 @@ static uint32_t kfd_generate_gpu_id(struct kfd_node *gpu)
buf[6] = upper_32_bits(local_mem_size);
buf[7] = (ffs(gpu->xcc_mask) - 1) | (NUM_XCC(gpu->xcc_mask) << 16);
- for (i = 0, hashout = 0; i < 8; i++)
- hashout ^= hash_32(buf[i], KFD_GPU_ID_HASH_WIDTH);
+ gpu_id = crc16(0, crc_buf, sizeof(buf)) &
+ ((1 << KFD_GPU_ID_HASH_WIDTH) - 1);
+
+ /* There is a very small possibility when generating a
+ * 16 (KFD_GPU_ID_HASH_WIDTH) bit value from 8 word buffer
+ * that the value could be 0 or non-unique. So, check if
+ * it is unique and non-zero. If not unique increment till
+ * unique one is found. In case of overflow, restart from 1
+ */
- return hashout;
+ down_read(&topology_lock);
+ do {
+ is_unique = true;
+ if (!gpu_id)
+ gpu_id = 1;
+ list_for_each_entry(dev, &topology_device_list, list) {
+ if (dev->gpu && dev->gpu_id == gpu_id) {
+ is_unique = false;
+ break;
+ }
+ }
+ if (unlikely(!is_unique))
+ gpu_id = (gpu_id + 1) &
+ ((1 << KFD_GPU_ID_HASH_WIDTH) - 1);
+ } while (!is_unique);
+ up_read(&topology_lock);
+
+ return gpu_id;
}
/* kfd_assign_gpu - Attach @gpu to the correct kfd topology device. If
* the GPU device is not already present in the topology device
@@ -1635,7 +1662,8 @@ static int fill_in_l2_l3_pcache(struct kfd_cache_properties **props_ext,
pcache->cache_level = pcache_info[cache_type].cache_level;
pcache->cacheline_size = pcache_info[cache_type].cache_line_size;
- if (KFD_GC_VERSION(knode) == IP_VERSION(9, 4, 3))
+ if (KFD_GC_VERSION(knode) == IP_VERSION(9, 4, 3) ||
+ KFD_GC_VERSION(knode) == IP_VERSION(9, 4, 4))
mode = adev->gmc.gmc_funcs->query_mem_partition_mode(adev);
else
mode = UNKNOWN_MEMORY_PARTITION_MODE;
@@ -1772,7 +1800,7 @@ static void kfd_fill_cache_non_crat_info(struct kfd_topology_device *dev, struct
pr_debug("Added [%d] GPU cache entries\n", num_of_entries);
}
-static int kfd_topology_add_device_locked(struct kfd_node *gpu, uint32_t gpu_id,
+static int kfd_topology_add_device_locked(struct kfd_node *gpu,
struct kfd_topology_device **dev)
{
int proximity_domain = ++topology_crat_proximity_domain;
@@ -1785,8 +1813,7 @@ static int kfd_topology_add_device_locked(struct kfd_node *gpu, uint32_t gpu_id,
COMPUTE_UNIT_GPU, gpu,
proximity_domain);
if (res) {
- pr_err("Error creating VCRAT for GPU (ID: 0x%x)\n",
- gpu_id);
+ dev_err(gpu->adev->dev, "Error creating VCRAT\n");
topology_crat_proximity_domain--;
goto err;
}
@@ -1797,8 +1824,7 @@ static int kfd_topology_add_device_locked(struct kfd_node *gpu, uint32_t gpu_id,
&temp_topology_device_list,
proximity_domain);
if (res) {
- pr_err("Error parsing VCRAT for GPU (ID: 0x%x)\n",
- gpu_id);
+ dev_err(gpu->adev->dev, "Error parsing VCRAT\n");
topology_crat_proximity_domain--;
goto err;
}
@@ -1824,8 +1850,8 @@ static int kfd_topology_add_device_locked(struct kfd_node *gpu, uint32_t gpu_id,
if (!res)
sys_props.generation_count++;
else
- pr_err("Failed to update GPU (ID: 0x%x) to sysfs topology. res=%d\n",
- gpu_id, res);
+ dev_err(gpu->adev->dev, "Failed to update GPU to sysfs topology. res=%d\n",
+ res);
err:
kfd_destroy_crat_image(crat_image);
@@ -1908,7 +1934,8 @@ static void kfd_topology_set_capabilities(struct kfd_topology_device *dev)
dev->node_props.debug_prop |= HSA_DBG_DISPATCH_INFO_ALWAYS_VALID;
if (KFD_GC_VERSION(dev->gpu) < IP_VERSION(10, 0, 0)) {
- if (KFD_GC_VERSION(dev->gpu) == IP_VERSION(9, 4, 3))
+ if (KFD_GC_VERSION(dev->gpu) == IP_VERSION(9, 4, 3) ||
+ KFD_GC_VERSION(dev->gpu) == IP_VERSION(9, 4, 4))
dev->node_props.debug_prop |=
HSA_DBG_WATCH_ADDR_MASK_LO_BIT_GFX9_4_3 |
HSA_DBG_WATCH_ADDR_MASK_HI_BIT_GFX9_4_3;
@@ -1927,6 +1954,10 @@ static void kfd_topology_set_capabilities(struct kfd_topology_device *dev)
if (KFD_GC_VERSION(dev->gpu) >= IP_VERSION(11, 0, 0))
dev->node_props.capability |=
HSA_CAP_TRAP_DEBUG_PRECISE_MEMORY_OPERATIONS_SUPPORTED;
+
+ if (KFD_GC_VERSION(dev->gpu) >= IP_VERSION(12, 0, 0))
+ dev->node_props.capability |=
+ HSA_CAP_TRAP_DEBUG_PRECISE_ALU_OPERATIONS_SUPPORTED;
}
kfd_topology_set_dbg_firmware_support(dev);
@@ -1942,14 +1973,12 @@ int kfd_topology_add_device(struct kfd_node *gpu)
struct amdgpu_gfx_config *gfx_info = &gpu->adev->gfx.config;
struct amdgpu_cu_info *cu_info = &gpu->adev->gfx.cu_info;
- gpu_id = kfd_generate_gpu_id(gpu);
if (gpu->xcp && !gpu->xcp->ddev) {
dev_warn(gpu->adev->dev,
- "Won't add GPU (ID: 0x%x) to topology since it has no drm node assigned.",
- gpu_id);
+ "Won't add GPU to topology since it has no drm node assigned.");
return 0;
} else {
- pr_debug("Adding new GPU (ID: 0x%x) to topology\n", gpu_id);
+ dev_dbg(gpu->adev->dev, "Adding new GPU to topology\n");
}
/* Check to see if this gpu device exists in the topology_device_list.
@@ -1961,11 +1990,12 @@ int kfd_topology_add_device(struct kfd_node *gpu)
down_write(&topology_lock);
dev = kfd_assign_gpu(gpu);
if (!dev)
- res = kfd_topology_add_device_locked(gpu, gpu_id, &dev);
+ res = kfd_topology_add_device_locked(gpu, &dev);
up_write(&topology_lock);
if (res)
return res;
+ gpu_id = kfd_generate_gpu_id(gpu);
dev->gpu_id = gpu_id;
gpu->id = gpu_id;
@@ -1997,9 +2027,8 @@ int kfd_topology_add_device(struct kfd_node *gpu)
HSA_CAP_ASIC_REVISION_MASK);
dev->node_props.location_id = pci_dev_id(gpu->adev->pdev);
- /* On multi-partition nodes, node id = location_id[31:28] */
- if (gpu->kfd->num_nodes > 1)
- dev->node_props.location_id |= (dev->gpu->node_id << 28);
+ if (KFD_GC_VERSION(dev->gpu->kfd) == IP_VERSION(9, 4, 3))
+ dev->node_props.location_id |= dev->gpu->node_id;
dev->node_props.domain = pci_domain_nr(gpu->adev->pdev->bus);
dev->node_props.max_engine_clk_fcompute =
generated by cgit (git 2.34.1) at 2025-06-03 22:01:18 +0000