drm/amdkfd: Remove sync_with_hw() from amdkfd

This patch completely removes the sync_with_hw() because it was broken and
actually there is no point of using it.

This function was used to:

- Make sure that the submitted packet to the HIQ (which is a kernel queue) was
  read by the CP. However, it was discovered that the method this function used
  to do that (checking wptr == rptr) is not consistent with how the actual CP
  firmware works in all cases.

- Make sure that the queue is empty before issuing the next packet. To achieve
  that, the function blocked amdkfd from continuing until the recently
  submitted packet was consumed. However, the acquire_packet_buffer() already
  checks if there is enough room for a new packet so calling sync_with_hw() is
  redundant.

Signed-off-by: Oded Gabbay <oded.gabbay@amd.com>
Acked-by: Alex Deucher <alexander.deucher@amd.com>

diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_kernel_queue.c b/drivers/gpu/drm/amd/amdkfd/kfd_kernel_queue.c
index 75950ed7a1bce5de813b0adab1c8203f3f219e9b..c04b1ac60bd9e7f9f8e56e30e3b34a811479511c 100644 (file)
--- a/drivers/gpu/drm/amd/amdkfd/kfd_kernel_queue.c
+++ b/drivers/gpu/drm/amd/amdkfd/kfd_kernel_queue.c
@@ -262,28 +262,6 @@ static void submit_packet(struct kernel_queue *kq)
                                kq->pending_wptr);
 }
 
-static int sync_with_hw(struct kernel_queue *kq, unsigned long timeout_ms)
-{
-       unsigned long org_timeout_ms;
-
-       BUG_ON(!kq);
-
-       org_timeout_ms = timeout_ms;
-       timeout_ms += jiffies * 1000 / HZ;
-       while (*kq->wptr_kernel != *kq->rptr_kernel) {
-               if (time_after(jiffies * 1000 / HZ, timeout_ms)) {
-                       pr_err("kfd: kernel_queue %s timeout expired %lu\n",
-                               __func__, org_timeout_ms);
-                       pr_err("kfd: wptr: %d rptr: %d\n",
-                               *kq->wptr_kernel, *kq->rptr_kernel);
-                       return -ETIME;
-               }
-               schedule();
-       }
-
-       return 0;
-}
-
 static void rollback_packet(struct kernel_queue *kq)
 {
        BUG_ON(!kq);
@@ -305,7 +283,6 @@ struct kernel_queue *kernel_queue_init(struct kfd_dev *dev,
        kq->ops.uninitialize = uninitialize;
        kq->ops.acquire_packet_buffer = acquire_packet_buffer;
        kq->ops.submit_packet = submit_packet;
-       kq->ops.sync_with_hw = sync_with_hw;
        kq->ops.rollback_packet = rollback_packet;
 
        switch (dev->device_info->asic_family) {
@@ -349,7 +326,6 @@ static __attribute__((unused)) void test_kq(struct kfd_dev *dev)
        for (i = 0; i < 5; i++)
                buffer[i] = kq->nop_packet;
        kq->ops.submit_packet(kq);
-       kq->ops.sync_with_hw(kq, 1000);
 
        pr_err("kfd: ending kernel queue test\n");
 }

diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_kernel_queue.h b/drivers/gpu/drm/amd/amdkfd/kfd_kernel_queue.h
index 2659d936ba732146809ce699c54e0af10ee35668..594053136ee4d44812784c6ac402ddef0cd66b65 100644 (file)
--- a/drivers/gpu/drm/amd/amdkfd/kfd_kernel_queue.h
+++ b/drivers/gpu/drm/amd/amdkfd/kfd_kernel_queue.h
@@ -61,8 +61,6 @@ struct kernel_queue_ops {
                                        unsigned int **buffer_ptr);
 
        void    (*submit_packet)(struct kernel_queue *kq);
-       int     (*sync_with_hw)(struct kernel_queue *kq,
-                               unsigned long timeout_ms);
        void    (*rollback_packet)(struct kernel_queue *kq);
 };
 

diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_packet_manager.c b/drivers/gpu/drm/amd/amdkfd/kfd_packet_manager.c
index 5fb5c032d5d8d543682ab5617a4efdc8f82f3375..e2533d875f4311e2c6fa426b4ea606c981b1675b 100644 (file)
--- a/drivers/gpu/drm/amd/amdkfd/kfd_packet_manager.c
+++ b/drivers/gpu/drm/amd/amdkfd/kfd_packet_manager.c
@@ -376,7 +376,6 @@ int pm_send_set_resources(struct packet_manager *pm,
        packet->queue_mask_hi = upper_32_bits(res->queue_mask);
 
        pm->priv_queue->ops.submit_packet(pm->priv_queue);
-       pm->priv_queue->ops.sync_with_hw(pm->priv_queue, KFD_HIQ_TIMEOUT);
 
        mutex_unlock(&pm->lock);
 
@@ -413,7 +412,6 @@ int pm_send_runlist(struct packet_manager *pm, struct list_head *dqm_queues)
                goto fail_create_runlist;
 
        pm->priv_queue->ops.submit_packet(pm->priv_queue);
-       pm->priv_queue->ops.sync_with_hw(pm->priv_queue, KFD_HIQ_TIMEOUT);
 
        mutex_unlock(&pm->lock);
 
@@ -460,7 +458,6 @@ int pm_send_query_status(struct packet_manager *pm, uint64_t fence_address,
        packet->data_lo = lower_32_bits((uint64_t)fence_value);
 
        pm->priv_queue->ops.submit_packet(pm->priv_queue);
-       pm->priv_queue->ops.sync_with_hw(pm->priv_queue, KFD_HIQ_TIMEOUT);
        mutex_unlock(&pm->lock);
 
        return 0;
@@ -538,7 +535,6 @@ int pm_send_unmap_queue(struct packet_manager *pm, enum kfd_queue_type type,
        };
 
        pm->priv_queue->ops.submit_packet(pm->priv_queue);
-       pm->priv_queue->ops.sync_with_hw(pm->priv_queue, KFD_HIQ_TIMEOUT);
 
        mutex_unlock(&pm->lock);
        return 0;