Commit 851a645e authored by Felix Kuehling's avatar Felix Kuehling Committed by Oded Gabbay

drm/amdkfd: Add debugfs support to KFD

This commit adds several debugfs entries for kfd:

kfd/hqds: dumps all HQDs on all GPUs for KFD-controlled compute and
    SDMA RLC queues

kfd/mqds: dumps all MQDs of all KFD processes on all GPUs

kfd/rls: dumps HWS runlists on all GPUs
Signed-off-by: default avatarYong Zhao <yong.zhao@amd.com>
Signed-off-by: default avatarFelix Kuehling <Felix.Kuehling@amd.com>
Reviewed-by: default avatarOded Gabbay <oded.gabbay@gmail.com>
Signed-off-by: default avatarOded Gabbay <oded.gabbay@gmail.com>
parent 80c195f5
...@@ -37,4 +37,6 @@ amdkfd-y := kfd_module.o kfd_device.o kfd_chardev.o kfd_topology.o \ ...@@ -37,4 +37,6 @@ amdkfd-y := kfd_module.o kfd_device.o kfd_chardev.o kfd_topology.o \
kfd_interrupt.o kfd_events.o cik_event_interrupt.o \ kfd_interrupt.o kfd_events.o cik_event_interrupt.o \
kfd_dbgdev.o kfd_dbgmgr.o kfd_dbgdev.o kfd_dbgmgr.o
amdkfd-$(CONFIG_DEBUG_FS) += kfd_debugfs.o
obj-$(CONFIG_HSA_AMD) += amdkfd.o obj-$(CONFIG_HSA_AMD) += amdkfd.o
/*
* Copyright 2016-2017 Advanced Micro Devices, Inc.
*
* Permission is hereby granted, free of charge, to any person obtaining a
* copy of this software and associated documentation files (the "Software"),
* to deal in the Software without restriction, including without limitation
* the rights to use, copy, modify, merge, publish, distribute, sublicense,
* and/or sell copies of the Software, and to permit persons to whom the
* Software is furnished to do so, subject to the following conditions:
*
* The above copyright notice and this permission notice shall be included in
* all copies or substantial portions of the Software.
*
* THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
* IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
* FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL
* THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR
* OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE,
* ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR
* OTHER DEALINGS IN THE SOFTWARE.
*/
#include <linux/debugfs.h>
#include "kfd_priv.h"
static struct dentry *debugfs_root;
static int kfd_debugfs_open(struct inode *inode, struct file *file)
{
int (*show)(struct seq_file *, void *) = inode->i_private;
return single_open(file, show, NULL);
}
static const struct file_operations kfd_debugfs_fops = {
.owner = THIS_MODULE,
.open = kfd_debugfs_open,
.read = seq_read,
.llseek = seq_lseek,
.release = single_release,
};
void kfd_debugfs_init(void)
{
struct dentry *ent;
debugfs_root = debugfs_create_dir("kfd", NULL);
if (!debugfs_root || debugfs_root == ERR_PTR(-ENODEV)) {
pr_warn("Failed to create kfd debugfs dir\n");
return;
}
ent = debugfs_create_file("mqds", S_IFREG | 0444, debugfs_root,
kfd_debugfs_mqds_by_process,
&kfd_debugfs_fops);
if (!ent)
pr_warn("Failed to create mqds in kfd debugfs\n");
ent = debugfs_create_file("hqds", S_IFREG | 0444, debugfs_root,
kfd_debugfs_hqds_by_device,
&kfd_debugfs_fops);
if (!ent)
pr_warn("Failed to create hqds in kfd debugfs\n");
ent = debugfs_create_file("rls", S_IFREG | 0444, debugfs_root,
kfd_debugfs_rls_by_device,
&kfd_debugfs_fops);
if (!ent)
pr_warn("Failed to create rls in kfd debugfs\n");
}
void kfd_debugfs_fini(void)
{
debugfs_remove_recursive(debugfs_root);
}
...@@ -1311,3 +1311,74 @@ void device_queue_manager_uninit(struct device_queue_manager *dqm) ...@@ -1311,3 +1311,74 @@ void device_queue_manager_uninit(struct device_queue_manager *dqm)
dqm->ops.uninitialize(dqm); dqm->ops.uninitialize(dqm);
kfree(dqm); kfree(dqm);
} }
#if defined(CONFIG_DEBUG_FS)
static void seq_reg_dump(struct seq_file *m,
uint32_t (*dump)[2], uint32_t n_regs)
{
uint32_t i, count;
for (i = 0, count = 0; i < n_regs; i++) {
if (count == 0 ||
dump[i-1][0] + sizeof(uint32_t) != dump[i][0]) {
seq_printf(m, "%s %08x: %08x",
i ? "\n" : "",
dump[i][0], dump[i][1]);
count = 7;
} else {
seq_printf(m, " %08x", dump[i][1]);
count--;
}
}
seq_puts(m, "\n");
}
int dqm_debugfs_hqds(struct seq_file *m, void *data)
{
struct device_queue_manager *dqm = data;
uint32_t (*dump)[2], n_regs;
int pipe, queue;
int r = 0;
for (pipe = 0; pipe < get_pipes_per_mec(dqm); pipe++) {
int pipe_offset = pipe * get_queues_per_pipe(dqm);
for (queue = 0; queue < get_queues_per_pipe(dqm); queue++) {
if (!test_bit(pipe_offset + queue,
dqm->dev->shared_resources.queue_bitmap))
continue;
r = dqm->dev->kfd2kgd->hqd_dump(
dqm->dev->kgd, pipe, queue, &dump, &n_regs);
if (r)
break;
seq_printf(m, " CP Pipe %d, Queue %d\n",
pipe, queue);
seq_reg_dump(m, dump, n_regs);
kfree(dump);
}
}
for (pipe = 0; pipe < CIK_SDMA_ENGINE_NUM; pipe++) {
for (queue = 0; queue < CIK_SDMA_QUEUES_PER_ENGINE; queue++) {
r = dqm->dev->kfd2kgd->hqd_sdma_dump(
dqm->dev->kgd, pipe, queue, &dump, &n_regs);
if (r)
break;
seq_printf(m, " SDMA Engine %d, RLC %d\n",
pipe, queue);
seq_reg_dump(m, dump, n_regs);
kfree(dump);
}
}
return r;
}
#endif
...@@ -123,6 +123,8 @@ static int __init kfd_module_init(void) ...@@ -123,6 +123,8 @@ static int __init kfd_module_init(void)
kfd_process_create_wq(); kfd_process_create_wq();
kfd_debugfs_init();
amdkfd_init_completed = 1; amdkfd_init_completed = 1;
dev_info(kfd_device, "Initialized module\n"); dev_info(kfd_device, "Initialized module\n");
...@@ -139,6 +141,7 @@ static void __exit kfd_module_exit(void) ...@@ -139,6 +141,7 @@ static void __exit kfd_module_exit(void)
{ {
amdkfd_init_completed = 0; amdkfd_init_completed = 0;
kfd_debugfs_fini();
kfd_process_destroy_wq(); kfd_process_destroy_wq();
kfd_topology_shutdown(); kfd_topology_shutdown();
kfd_chardev_exit(); kfd_chardev_exit();
......
...@@ -85,6 +85,10 @@ struct mqd_manager { ...@@ -85,6 +85,10 @@ struct mqd_manager {
uint64_t queue_address, uint32_t pipe_id, uint64_t queue_address, uint32_t pipe_id,
uint32_t queue_id); uint32_t queue_id);
#if defined(CONFIG_DEBUG_FS)
int (*debugfs_show_mqd)(struct seq_file *m, void *data);
#endif
struct mutex mqd_mutex; struct mutex mqd_mutex;
struct kfd_dev *dev; struct kfd_dev *dev;
}; };
......
...@@ -365,6 +365,24 @@ static int update_mqd_hiq(struct mqd_manager *mm, void *mqd, ...@@ -365,6 +365,24 @@ static int update_mqd_hiq(struct mqd_manager *mm, void *mqd,
return 0; return 0;
} }
#if defined(CONFIG_DEBUG_FS)
static int debugfs_show_mqd(struct seq_file *m, void *data)
{
seq_hex_dump(m, " ", DUMP_PREFIX_OFFSET, 32, 4,
data, sizeof(struct cik_mqd), false);
return 0;
}
static int debugfs_show_mqd_sdma(struct seq_file *m, void *data)
{
seq_hex_dump(m, " ", DUMP_PREFIX_OFFSET, 32, 4,
data, sizeof(struct cik_sdma_rlc_registers), false);
return 0;
}
#endif
struct mqd_manager *mqd_manager_init_cik(enum KFD_MQD_TYPE type, struct mqd_manager *mqd_manager_init_cik(enum KFD_MQD_TYPE type,
struct kfd_dev *dev) struct kfd_dev *dev)
...@@ -389,6 +407,9 @@ struct mqd_manager *mqd_manager_init_cik(enum KFD_MQD_TYPE type, ...@@ -389,6 +407,9 @@ struct mqd_manager *mqd_manager_init_cik(enum KFD_MQD_TYPE type,
mqd->update_mqd = update_mqd; mqd->update_mqd = update_mqd;
mqd->destroy_mqd = destroy_mqd; mqd->destroy_mqd = destroy_mqd;
mqd->is_occupied = is_occupied; mqd->is_occupied = is_occupied;
#if defined(CONFIG_DEBUG_FS)
mqd->debugfs_show_mqd = debugfs_show_mqd;
#endif
break; break;
case KFD_MQD_TYPE_HIQ: case KFD_MQD_TYPE_HIQ:
mqd->init_mqd = init_mqd_hiq; mqd->init_mqd = init_mqd_hiq;
...@@ -397,6 +418,9 @@ struct mqd_manager *mqd_manager_init_cik(enum KFD_MQD_TYPE type, ...@@ -397,6 +418,9 @@ struct mqd_manager *mqd_manager_init_cik(enum KFD_MQD_TYPE type,
mqd->update_mqd = update_mqd_hiq; mqd->update_mqd = update_mqd_hiq;
mqd->destroy_mqd = destroy_mqd; mqd->destroy_mqd = destroy_mqd;
mqd->is_occupied = is_occupied; mqd->is_occupied = is_occupied;
#if defined(CONFIG_DEBUG_FS)
mqd->debugfs_show_mqd = debugfs_show_mqd;
#endif
break; break;
case KFD_MQD_TYPE_SDMA: case KFD_MQD_TYPE_SDMA:
mqd->init_mqd = init_mqd_sdma; mqd->init_mqd = init_mqd_sdma;
...@@ -405,6 +429,9 @@ struct mqd_manager *mqd_manager_init_cik(enum KFD_MQD_TYPE type, ...@@ -405,6 +429,9 @@ struct mqd_manager *mqd_manager_init_cik(enum KFD_MQD_TYPE type,
mqd->update_mqd = update_mqd_sdma; mqd->update_mqd = update_mqd_sdma;
mqd->destroy_mqd = destroy_mqd_sdma; mqd->destroy_mqd = destroy_mqd_sdma;
mqd->is_occupied = is_occupied_sdma; mqd->is_occupied = is_occupied_sdma;
#if defined(CONFIG_DEBUG_FS)
mqd->debugfs_show_mqd = debugfs_show_mqd_sdma;
#endif
break; break;
default: default:
kfree(mqd); kfree(mqd);
......
...@@ -358,7 +358,23 @@ static bool is_occupied_sdma(struct mqd_manager *mm, void *mqd, ...@@ -358,7 +358,23 @@ static bool is_occupied_sdma(struct mqd_manager *mm, void *mqd,
return mm->dev->kfd2kgd->hqd_sdma_is_occupied(mm->dev->kgd, mqd); return mm->dev->kfd2kgd->hqd_sdma_is_occupied(mm->dev->kgd, mqd);
} }
#if defined(CONFIG_DEBUG_FS)
static int debugfs_show_mqd(struct seq_file *m, void *data)
{
seq_hex_dump(m, " ", DUMP_PREFIX_OFFSET, 32, 4,
data, sizeof(struct vi_mqd), false);
return 0;
}
static int debugfs_show_mqd_sdma(struct seq_file *m, void *data)
{
seq_hex_dump(m, " ", DUMP_PREFIX_OFFSET, 32, 4,
data, sizeof(struct vi_sdma_mqd), false);
return 0;
}
#endif
struct mqd_manager *mqd_manager_init_vi(enum KFD_MQD_TYPE type, struct mqd_manager *mqd_manager_init_vi(enum KFD_MQD_TYPE type,
struct kfd_dev *dev) struct kfd_dev *dev)
...@@ -383,6 +399,9 @@ struct mqd_manager *mqd_manager_init_vi(enum KFD_MQD_TYPE type, ...@@ -383,6 +399,9 @@ struct mqd_manager *mqd_manager_init_vi(enum KFD_MQD_TYPE type,
mqd->update_mqd = update_mqd; mqd->update_mqd = update_mqd;
mqd->destroy_mqd = destroy_mqd; mqd->destroy_mqd = destroy_mqd;
mqd->is_occupied = is_occupied; mqd->is_occupied = is_occupied;
#if defined(CONFIG_DEBUG_FS)
mqd->debugfs_show_mqd = debugfs_show_mqd;
#endif
break; break;
case KFD_MQD_TYPE_HIQ: case KFD_MQD_TYPE_HIQ:
mqd->init_mqd = init_mqd_hiq; mqd->init_mqd = init_mqd_hiq;
...@@ -391,6 +410,9 @@ struct mqd_manager *mqd_manager_init_vi(enum KFD_MQD_TYPE type, ...@@ -391,6 +410,9 @@ struct mqd_manager *mqd_manager_init_vi(enum KFD_MQD_TYPE type,
mqd->update_mqd = update_mqd_hiq; mqd->update_mqd = update_mqd_hiq;
mqd->destroy_mqd = destroy_mqd; mqd->destroy_mqd = destroy_mqd;
mqd->is_occupied = is_occupied; mqd->is_occupied = is_occupied;
#if defined(CONFIG_DEBUG_FS)
mqd->debugfs_show_mqd = debugfs_show_mqd;
#endif
break; break;
case KFD_MQD_TYPE_SDMA: case KFD_MQD_TYPE_SDMA:
mqd->init_mqd = init_mqd_sdma; mqd->init_mqd = init_mqd_sdma;
...@@ -399,6 +421,9 @@ struct mqd_manager *mqd_manager_init_vi(enum KFD_MQD_TYPE type, ...@@ -399,6 +421,9 @@ struct mqd_manager *mqd_manager_init_vi(enum KFD_MQD_TYPE type,
mqd->update_mqd = update_mqd_sdma; mqd->update_mqd = update_mqd_sdma;
mqd->destroy_mqd = destroy_mqd_sdma; mqd->destroy_mqd = destroy_mqd_sdma;
mqd->is_occupied = is_occupied_sdma; mqd->is_occupied = is_occupied_sdma;
#if defined(CONFIG_DEBUG_FS)
mqd->debugfs_show_mqd = debugfs_show_mqd_sdma;
#endif
break; break;
default: default:
kfree(mqd); kfree(mqd);
......
...@@ -278,6 +278,7 @@ static int pm_create_runlist_ib(struct packet_manager *pm, ...@@ -278,6 +278,7 @@ static int pm_create_runlist_ib(struct packet_manager *pm,
return retval; return retval;
*rl_size_bytes = alloc_size_bytes; *rl_size_bytes = alloc_size_bytes;
pm->ib_size_bytes = alloc_size_bytes;
pr_debug("Building runlist ib process count: %d queues count %d\n", pr_debug("Building runlist ib process count: %d queues count %d\n",
pm->dqm->processes_count, pm->dqm->queue_count); pm->dqm->processes_count, pm->dqm->queue_count);
...@@ -591,3 +592,26 @@ void pm_release_ib(struct packet_manager *pm) ...@@ -591,3 +592,26 @@ void pm_release_ib(struct packet_manager *pm)
} }
mutex_unlock(&pm->lock); mutex_unlock(&pm->lock);
} }
#if defined(CONFIG_DEBUG_FS)
int pm_debugfs_runlist(struct seq_file *m, void *data)
{
struct packet_manager *pm = data;
mutex_lock(&pm->lock);
if (!pm->allocated) {
seq_puts(m, " No active runlist\n");
goto out;
}
seq_hex_dump(m, " ", DUMP_PREFIX_OFFSET, 32, 4,
pm->ib_buffer_obj->cpu_ptr, pm->ib_size_bytes, false);
out:
mutex_unlock(&pm->lock);
return 0;
}
#endif
...@@ -33,6 +33,7 @@ ...@@ -33,6 +33,7 @@
#include <linux/kfd_ioctl.h> #include <linux/kfd_ioctl.h>
#include <linux/idr.h> #include <linux/idr.h>
#include <linux/kfifo.h> #include <linux/kfifo.h>
#include <linux/seq_file.h>
#include <kgd_kfd_interface.h> #include <kgd_kfd_interface.h>
#include "amd_shared.h" #include "amd_shared.h"
...@@ -735,6 +736,7 @@ struct packet_manager { ...@@ -735,6 +736,7 @@ struct packet_manager {
struct mutex lock; struct mutex lock;
bool allocated; bool allocated;
struct kfd_mem_obj *ib_buffer_obj; struct kfd_mem_obj *ib_buffer_obj;
unsigned int ib_size_bytes;
}; };
int pm_init(struct packet_manager *pm, struct device_queue_manager *dqm); int pm_init(struct packet_manager *pm, struct device_queue_manager *dqm);
...@@ -781,4 +783,23 @@ int kfd_event_destroy(struct kfd_process *p, uint32_t event_id); ...@@ -781,4 +783,23 @@ int kfd_event_destroy(struct kfd_process *p, uint32_t event_id);
int dbgdev_wave_reset_wavefronts(struct kfd_dev *dev, struct kfd_process *p); int dbgdev_wave_reset_wavefronts(struct kfd_dev *dev, struct kfd_process *p);
/* Debugfs */
#if defined(CONFIG_DEBUG_FS)
void kfd_debugfs_init(void);
void kfd_debugfs_fini(void);
int kfd_debugfs_mqds_by_process(struct seq_file *m, void *data);
int pqm_debugfs_mqds(struct seq_file *m, void *data);
int kfd_debugfs_hqds_by_device(struct seq_file *m, void *data);
int dqm_debugfs_hqds(struct seq_file *m, void *data);
int kfd_debugfs_rls_by_device(struct seq_file *m, void *data);
int pm_debugfs_runlist(struct seq_file *m, void *data);
#else
static inline void kfd_debugfs_init(void) {}
static inline void kfd_debugfs_fini(void) {}
#endif
#endif #endif
...@@ -620,3 +620,32 @@ int kfd_reserved_mem_mmap(struct kfd_process *process, ...@@ -620,3 +620,32 @@ int kfd_reserved_mem_mmap(struct kfd_process *process,
PFN_DOWN(__pa(qpd->cwsr_kaddr)), PFN_DOWN(__pa(qpd->cwsr_kaddr)),
KFD_CWSR_TBA_TMA_SIZE, vma->vm_page_prot); KFD_CWSR_TBA_TMA_SIZE, vma->vm_page_prot);
} }
#if defined(CONFIG_DEBUG_FS)
int kfd_debugfs_mqds_by_process(struct seq_file *m, void *data)
{
struct kfd_process *p;
unsigned int temp;
int r = 0;
int idx = srcu_read_lock(&kfd_processes_srcu);
hash_for_each_rcu(kfd_processes_table, temp, p, kfd_processes) {
seq_printf(m, "Process %d PASID %d:\n",
p->lead_thread->tgid, p->pasid);
mutex_lock(&p->mutex);
r = pqm_debugfs_mqds(m, &p->pqm);
mutex_unlock(&p->mutex);
if (r)
break;
}
srcu_read_unlock(&kfd_processes_srcu, idx);
return r;
}
#endif
...@@ -368,4 +368,67 @@ struct kernel_queue *pqm_get_kernel_queue( ...@@ -368,4 +368,67 @@ struct kernel_queue *pqm_get_kernel_queue(
return NULL; return NULL;
} }
#if defined(CONFIG_DEBUG_FS)
int pqm_debugfs_mqds(struct seq_file *m, void *data)
{
struct process_queue_manager *pqm = data;
struct process_queue_node *pqn;
struct queue *q;
enum KFD_MQD_TYPE mqd_type;
struct mqd_manager *mqd_manager;
int r = 0;
list_for_each_entry(pqn, &pqm->queues, process_queue_list) {
if (pqn->q) {
q = pqn->q;
switch (q->properties.type) {
case KFD_QUEUE_TYPE_SDMA:
seq_printf(m, " SDMA queue on device %x\n",
q->device->id);
mqd_type = KFD_MQD_TYPE_SDMA;
break;
case KFD_QUEUE_TYPE_COMPUTE:
seq_printf(m, " Compute queue on device %x\n",
q->device->id);
mqd_type = KFD_MQD_TYPE_CP;
break;
default:
seq_printf(m,
" Bad user queue type %d on device %x\n",
q->properties.type, q->device->id);
continue;
}
mqd_manager = q->device->dqm->ops.get_mqd_manager(
q->device->dqm, mqd_type);
} else if (pqn->kq) {
q = pqn->kq->queue;
mqd_manager = pqn->kq->mqd;
switch (q->properties.type) {
case KFD_QUEUE_TYPE_DIQ:
seq_printf(m, " DIQ on device %x\n",
pqn->kq->dev->id);
mqd_type = KFD_MQD_TYPE_HIQ;
break;
default:
seq_printf(m,
" Bad kernel queue type %d on device %x\n",
q->properties.type,
pqn->kq->dev->id);
continue;
}
} else {
seq_printf(m,
" Weird: Queue node with neither kernel nor user queue\n");
continue;
}
r = mqd_manager->debugfs_show_mqd(m, q->mqd);
if (r != 0)
break;
}
return r;
}
#endif
...@@ -32,6 +32,7 @@ ...@@ -32,6 +32,7 @@
#include "kfd_priv.h" #include "kfd_priv.h"
#include "kfd_crat.h" #include "kfd_crat.h"
#include "kfd_topology.h" #include "kfd_topology.h"
#include "kfd_device_queue_manager.h"
static struct list_head topology_device_list; static struct list_head topology_device_list;
static int topology_crat_parsed; static int topology_crat_parsed;
...@@ -1233,3 +1234,57 @@ struct kfd_dev *kfd_topology_enum_kfd_devices(uint8_t idx) ...@@ -1233,3 +1234,57 @@ struct kfd_dev *kfd_topology_enum_kfd_devices(uint8_t idx)
return device; return device;
} }
#if defined(CONFIG_DEBUG_FS)
int kfd_debugfs_hqds_by_device(struct seq_file *m, void *data)
{
struct kfd_topology_device *dev;
unsigned int i = 0;
int r = 0;
down_read(&topology_lock);
list_for_each_entry(dev, &topology_device_list, list) {
if (!dev->gpu) {
i++;
continue;
}
seq_printf(m, "Node %u, gpu_id %x:\n", i++, dev->gpu->id);
r = dqm_debugfs_hqds(m, dev->gpu->dqm);
if (r)
break;
}
up_read(&topology_lock);
return r;
}
int kfd_debugfs_rls_by_device(struct seq_file *m, void *data)
{
struct kfd_topology_device *dev;
unsigned int i = 0;
int r = 0;
down_read(&topology_lock);
list_for_each_entry(dev, &topology_device_list, list) {
if (!dev->gpu) {
i++;
continue;
}
seq_printf(m, "Node %u, gpu_id %x:\n", i++, dev->gpu->id);
r = pm_debugfs_runlist(m, &dev->gpu->dqm->packets);
if (r)
break;
}
up_read(&topology_lock);
return r;
}
#endif
Markdown is supported
0%
or
You are about to add 0 people to the discussion. Proceed with caution.
Finish editing this message first!
Please register or to comment