From 4268b9df5aef5df448e428d10f8aeb0b903ebd03 Mon Sep 17 00:00:00 2001
From: Oak Zeng <Oak.Zeng@amd.com>
Date: Thu, 15 Jul 2021 18:34:25 -0500
Subject: drm/amdkfd: Fix a concurrency issue during kfd recovery
Git-commit: 4f942aaeb19dbf2135931120cc806d459add4788
Patch-mainline: v5.15-rc1
References: jsc#PED-1166 jsc#PED-1168 jsc#PED-1170 jsc#PED-1218 jsc#PED-1220 jsc#PED-1222 jsc#PED-1223 jsc#PED-1225

start_cpsch and stop_cpsch can be called during kfd device
initialization or during gpu reset/recovery. So they can
run concurrently. Currently in start_cpsch and stop_cpsch,
pm_init and pm_uninit is not protected by the dpm lock.
Imagine such a case that user use packet manager's function
to submit a pm4 packet to hang hws (ie through command
cat /sys/class/kfd/kfd/topology/nodes/1/gpu_id | sudo tee
/sys/kernel/debug/kfd/hang_hws), while kfd device is under
device reset/recovery so packet manager can be not initialized.
There will be unpredictable protection fault in such case.

This patch moves pm_init/uninit inside the dpm lock and check
packet manager is initialized before using packet manager
function.

Signed-off-by: Oak Zeng <Oak.Zeng@amd.com>
Acked-by: Christian Konig <christian.koenig@amd.com>
Reviewed-by: Felix Kuehling <Felix.Kuehling@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Acked-by: Patrik Jakobsson <pjakobsson@suse.de>
---
 drivers/gpu/drm/amd/amdkfd/kfd_device.c              |  8 +-------
 .../gpu/drm/amd/amdkfd/kfd_device_queue_manager.c    | 12 +++++++++---
 drivers/gpu/drm/amd/amdkfd/kfd_packet_manager.c      |  3 +++
 drivers/gpu/drm/amd/amdkfd/kfd_priv.h                |  2 +-
 4 files changed, 14 insertions(+), 11 deletions(-)

diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_device.c b/drivers/gpu/drm/amd/amdkfd/kfd_device.c
index 73132c90791b..7fc7fe9d2b45 100644
--- a/drivers/gpu/drm/amd/amdkfd/kfd_device.c
+++ b/drivers/gpu/drm/amd/amdkfd/kfd_device.c
@@ -1382,18 +1382,12 @@ void kgd2kfd_smi_event_throttle(struct kfd_dev *kfd, uint32_t throttle_bitmask)
  */
 int kfd_debugfs_hang_hws(struct kfd_dev *dev)
 {
-	int r = 0;
-
 	if (dev->dqm->sched_policy != KFD_SCHED_POLICY_HWS) {
 		pr_err("HWS is not enabled");
 		return -EINVAL;
 	}
 
-	r = pm_debugfs_hang_hws(&dev->dqm->packet_mgr);
-	if (!r)
-		r = dqm_debugfs_execute_queues(dev->dqm);
-
-	return r;
+	return dqm_debugfs_hang_hws(dev->dqm);
 }
 
 #endif
diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_device_queue_manager.c b/drivers/gpu/drm/amd/amdkfd/kfd_device_queue_manager.c
index 6b2f5940c1f6..6b89ca6ddc65 100644
--- a/drivers/gpu/drm/amd/amdkfd/kfd_device_queue_manager.c
+++ b/drivers/gpu/drm/amd/amdkfd/kfd_device_queue_manager.c
@@ -1164,6 +1164,7 @@ static int start_cpsch(struct device_queue_manager *dqm)
 
 	retval = 0;
 
+	dqm_lock(dqm);
 	retval = pm_init(&dqm->packet_mgr, dqm);
 	if (retval)
 		goto fail_packet_manager_init;
@@ -1186,7 +1187,6 @@ static int start_cpsch(struct device_queue_manager *dqm)
 
 	init_interrupts(dqm);
 
-	dqm_lock(dqm);
 	/* clear hang status when driver try to start the hw scheduler */
 	dqm->is_hws_hang = false;
 	dqm->is_resetting = false;
@@ -1199,6 +1199,7 @@ static int start_cpsch(struct device_queue_manager *dqm)
 fail_set_sched_resources:
 	pm_uninit(&dqm->packet_mgr, false);
 fail_packet_manager_init:
+	dqm_unlock(dqm);
 	return retval;
 }
 
@@ -1211,12 +1212,12 @@ static int stop_cpsch(struct device_queue_manager *dqm)
 		unmap_queues_cpsch(dqm, KFD_UNMAP_QUEUES_FILTER_ALL_QUEUES, 0);
 	hanging = dqm->is_hws_hang || dqm->is_resetting;
 	dqm->sched_running = false;
-	dqm_unlock(dqm);
 
 	pm_release_ib(&dqm->packet_mgr);
 
 	kfd_gtt_sa_free(dqm->dev, dqm->fence_mem);
 	pm_uninit(&dqm->packet_mgr, hanging);
+	dqm_unlock(dqm);
 
 	return 0;
 }
@@ -2099,11 +2100,16 @@ int dqm_debugfs_hqds(struct seq_file *m, void *data)
 	return r;
 }
 
-int dqm_debugfs_execute_queues(struct device_queue_manager *dqm)
+int dqm_debugfs_hang_hws(struct device_queue_manager *dqm)
 {
 	int r = 0;
 
 	dqm_lock(dqm);
+	r = pm_debugfs_hang_hws(&dqm->packet_mgr);
+	if (r) {
+		dqm_unlock(dqm);
+		return r;
+	}
 	dqm->active_runlist = true;
 	r = execute_queues_cpsch(dqm, KFD_UNMAP_QUEUES_FILTER_ALL_QUEUES, 0);
 	dqm_unlock(dqm);
diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_packet_manager.c b/drivers/gpu/drm/amd/amdkfd/kfd_packet_manager.c
index b130cc0a25b3..b33ebe81cd95 100644
--- a/drivers/gpu/drm/amd/amdkfd/kfd_packet_manager.c
+++ b/drivers/gpu/drm/amd/amdkfd/kfd_packet_manager.c
@@ -448,6 +448,9 @@ int pm_debugfs_hang_hws(struct packet_manager *pm)
 	uint32_t *buffer, size;
 	int r = 0;
 
+	if (!pm->priv_queue)
+		return -EAGAIN;
+
 	size = pm->pmf->query_status_size;
 	mutex_lock(&pm->lock);
 	kq_acquire_packet_buffer(pm->priv_queue,
diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_priv.h b/drivers/gpu/drm/amd/amdkfd/kfd_priv.h
index 3426743ed228..8a5dfda224bf 100644
--- a/drivers/gpu/drm/amd/amdkfd/kfd_priv.h
+++ b/drivers/gpu/drm/amd/amdkfd/kfd_priv.h
@@ -1194,7 +1194,7 @@ int pm_debugfs_runlist(struct seq_file *m, void *data);
 
 int kfd_debugfs_hang_hws(struct kfd_dev *dev);
 int pm_debugfs_hang_hws(struct packet_manager *pm);
-int dqm_debugfs_execute_queues(struct device_queue_manager *dqm);
+int dqm_debugfs_hang_hws(struct device_queue_manager *dqm);
 
 #else
 
-- 
2.38.1