Merge tag 'drm-habanalabs-next-2023-06-08' of https://git.kernel.org/pub/scm/linux/kernel/git/ogabbay/linux into drm-next

This tag contains additional habanalabs driver changes for v6.5:

- uAPI changes:
  - Return 0 when user queries if there was a h/w or f/w error and no such error happened.
    Previously we returned an error in such case.

- New features and improvements:
  - Add pci health check when we lose connection with the firmware. This can be used to
    distinguish between pci link down and firmware getting stuck.
  - Add more info to the error print when TPC interrupt occur.
  - Reduce amount of code under mutex in the command submission of signal event.

- Firmware related fixes:
  - Fixes to the handshake protocol during f/w initialization.
  - Display information that the f/w sends us when encountering a DMA error.
  - Do soft-reset using a message sent to firmware instead of writing to MMIO.
  - Prepare generic code to extract f/w version numbers.

- Bug fixes and code cleanups. Notable fixes are:
  - Unsecure certain TPC registers that the user should access.
  - Fix handling of QMAN errors
  - Fix memory leak when recording errors (to later pass them to the user)
  - Multiple fixes to razwi interrupt handling code

Signed-off-by: Dave Airlie <airlied@redhat.com>

From: Oded Gabbay <ogabbay@kernel.org>
Link: https://patchwork.freedesktop.org/patch/msgid/20230608103043.GA2699019@ogabbay-vm-u20.habana-labs.com

Author: airlied

drivers/accel/habanalabs/common/command_buffer.c

@@ -27,12 +27,6 @@ static int cb_map_mem(struct hl_ctx *ctx, struct hl_cb *cb)
 		return -EINVAL;
 	}
 
-	if (!hdev->mmu_enable) {
-		dev_err_ratelimited(hdev->dev,
-				"Cannot map CB because MMU is disabled\n");
-		return -EINVAL;
-	}
-
 	if (cb->is_mmu_mapped)
 		return 0;
 

drivers/accel/habanalabs/common/command_submission.c

@@ -280,14 +280,8 @@ bool cs_needs_timeout(struct hl_cs *cs)
 
 static bool is_cb_patched(struct hl_device *hdev, struct hl_cs_job *job)
 {
-	/*
-	 * Patched CB is created for external queues jobs, and for H/W queues
-	 * jobs if the user CB was allocated by driver and MMU is disabled.
-	 */
-	return (job->queue_type == QUEUE_TYPE_EXT ||
-			(job->queue_type == QUEUE_TYPE_HW &&
-					job->is_kernel_allocated_cb &&
-					!hdev->mmu_enable));
+	/* Patched CB is created for external queues jobs */
+	return (job->queue_type == QUEUE_TYPE_EXT);
 }
 
 /*
@@ -363,14 +357,13 @@ static void hl_complete_job(struct hl_device *hdev, struct hl_cs_job *job)
 		}
 	}
 
-	/* For H/W queue jobs, if a user CB was allocated by driver and MMU is
-	 * enabled, the user CB isn't released in cs_parser() and thus should be
+	/* For H/W queue jobs, if a user CB was allocated by driver,
+	 * the user CB isn't released in cs_parser() and thus should be
 	 * released here. This is also true for INT queues jobs which were
 	 * allocated by driver.
 	 */
-	if ((job->is_kernel_allocated_cb &&
-		((job->queue_type == QUEUE_TYPE_HW && hdev->mmu_enable) ||
-				job->queue_type == QUEUE_TYPE_INT))) {
+	if (job->is_kernel_allocated_cb &&
+			(job->queue_type == QUEUE_TYPE_HW || job->queue_type == QUEUE_TYPE_INT)) {
 		atomic_dec(&job->user_cb->cs_cnt);
 		hl_cb_put(job->user_cb);
 	}
@@ -804,12 +797,14 @@ static void cs_do_release(struct kref *ref)
 
 static void cs_timedout(struct work_struct *work)
 {
+	struct hl_cs *cs = container_of(work, struct hl_cs, work_tdr.work);
+	bool skip_reset_on_timeout, device_reset = false;
 	struct hl_device *hdev;
 	u64 event_mask = 0x0;
+	uint timeout_sec;
 	int rc;
-	struct hl_cs *cs = container_of(work, struct hl_cs,
-						 work_tdr.work);
-	bool skip_reset_on_timeout = cs->skip_reset_on_timeout, device_reset = false;
+
+	skip_reset_on_timeout = cs->skip_reset_on_timeout;
 
 	rc = cs_get_unless_zero(cs);
 	if (!rc)
@@ -840,29 +835,31 @@ static void cs_timedout(struct work_struct *work)
 		event_mask |= HL_NOTIFIER_EVENT_CS_TIMEOUT;
 	}
 
+	timeout_sec = jiffies_to_msecs(hdev->timeout_jiffies) / 1000;
+
 	switch (cs->type) {
 	case CS_TYPE_SIGNAL:
 		dev_err(hdev->dev,
-			"Signal command submission %llu has not finished in time!\n",
-			cs->sequence);
+			"Signal command submission %llu has not finished in %u seconds!\n",
+			cs->sequence, timeout_sec);
 		break;
 
 	case CS_TYPE_WAIT:
 		dev_err(hdev->dev,
-			"Wait command submission %llu has not finished in time!\n",
-			cs->sequence);
+			"Wait command submission %llu has not finished in %u seconds!\n",
+			cs->sequence, timeout_sec);
 		break;
 
 	case CS_TYPE_COLLECTIVE_WAIT:
 		dev_err(hdev->dev,
-			"Collective Wait command submission %llu has not finished in time!\n",
-			cs->sequence);
+			"Collective Wait command submission %llu has not finished in %u seconds!\n",
+			cs->sequence, timeout_sec);
 		break;
 
 	default:
 		dev_err(hdev->dev,
-			"Command submission %llu has not finished in time!\n",
-			cs->sequence);
+			"Command submission %llu has not finished in %u seconds!\n",
+			cs->sequence, timeout_sec);
 		break;
 	}
 
@@ -1139,11 +1136,10 @@ static void force_complete_cs(struct hl_device *hdev)
 	spin_unlock(&hdev->cs_mirror_lock);
 }
 
-void hl_abort_waitings_for_completion(struct hl_device *hdev)
+void hl_abort_waiting_for_cs_completions(struct hl_device *hdev)
 {
 	force_complete_cs(hdev);
 	force_complete_multi_cs(hdev);
-	hl_release_pending_user_interrupts(hdev);
 }
 
 static void job_wq_completion(struct work_struct *work)
@@ -1948,8 +1944,7 @@ static int cs_ioctl_signal_wait_create_jobs(struct hl_device *hdev,
 	else
 		cb_size = hdev->asic_funcs->get_signal_cb_size(hdev);
 
-	cb = hl_cb_kernel_create(hdev, cb_size,
-				q_type == QUEUE_TYPE_HW && hdev->mmu_enable);
+	cb = hl_cb_kernel_create(hdev, cb_size, q_type == QUEUE_TYPE_HW);
 	if (!cb) {
 		atomic64_inc(&ctx->cs_counters.out_of_mem_drop_cnt);
 		atomic64_inc(&cntr->out_of_mem_drop_cnt);
@@ -2152,7 +2147,7 @@ static int cs_ioctl_unreserve_signals(struct hl_fpriv *hpriv, u32 handle_id)
 
 			hdev->asic_funcs->hw_queues_unlock(hdev);
 			rc = -EINVAL;
-			goto out;
+			goto out_unlock;
 		}
 
 		/*
@@ -2167,15 +2162,21 @@ static int cs_ioctl_unreserve_signals(struct hl_fpriv *hpriv, u32 handle_id)
 
 		/* Release the id and free allocated memory of the handle */
 		idr_remove(&mgr->handles, handle_id);
+
+		/* unlock before calling ctx_put, where we might sleep */
+		spin_unlock(&mgr->lock);
 		hl_ctx_put(encaps_sig_hdl->ctx);
 		kfree(encaps_sig_hdl);
+		goto out;
 	} else {
 		rc = -EINVAL;
 		dev_err(hdev->dev, "failed to unreserve signals, cannot find handler\n");
 	}
-out:
+
+out_unlock:
 	spin_unlock(&mgr->lock);
 
+out:
 	return rc;
 }
 

drivers/accel/habanalabs/common/debugfs.c

@@ -255,9 +255,6 @@ static int vm_show(struct seq_file *s, void *data)
 	u64 j;
 	int i;
 
-	if (!dev_entry->hdev->mmu_enable)
-		return 0;
-
 	mutex_lock(&dev_entry->ctx_mem_hash_mutex);
 
 	list_for_each_entry(ctx, &dev_entry->ctx_mem_hash_list, debugfs_list) {
@@ -436,9 +433,6 @@ static int mmu_show(struct seq_file *s, void *data)
 	u64 virt_addr = dev_entry->mmu_addr, phys_addr;
 	int i;
 
-	if (!hdev->mmu_enable)
-		return 0;
-
 	if (dev_entry->mmu_asid == HL_KERNEL_ASID_ID)
 		ctx = hdev->kernel_ctx;
 	else
@@ -496,9 +490,6 @@ static ssize_t mmu_asid_va_write(struct file *file, const char __user *buf,
 	char *c;
 	ssize_t rc;
 
-	if (!hdev->mmu_enable)
-		return count;
-
 	if (count > sizeof(kbuf) - 1)
 		goto err;
 	if (copy_from_user(kbuf, buf, count))
@@ -535,9 +526,6 @@ static int mmu_ack_error(struct seq_file *s, void *data)
 	struct hl_device *hdev = dev_entry->hdev;
 	int rc;
 
-	if (!hdev->mmu_enable)
-		return 0;
-
 	if (!dev_entry->mmu_cap_mask) {
 		dev_err(hdev->dev, "mmu_cap_mask is not set\n");
 		goto err;
@@ -563,9 +551,6 @@ static ssize_t mmu_ack_error_value_write(struct file *file,
 	char kbuf[MMU_KBUF_SIZE];
 	ssize_t rc;
 
-	if (!hdev->mmu_enable)
-		return count;
-
 	if (count > sizeof(kbuf) - 1)
 		goto err;
 
@@ -661,9 +646,6 @@ static bool hl_is_device_va(struct hl_device *hdev, u64 addr)
 {
 	struct asic_fixed_properties *prop = &hdev->asic_prop;
 
-	if (!hdev->mmu_enable)
-		goto out;
-
 	if (prop->dram_supports_virtual_memory &&
 		(addr >= prop->dmmu.start_addr && addr < prop->dmmu.end_addr))
 		return true;
@@ -675,7 +657,7 @@ static bool hl_is_device_va(struct hl_device *hdev, u64 addr)
 	if (addr >= prop->pmmu_huge.start_addr &&
 		addr < prop->pmmu_huge.end_addr)
 		return true;
-out:
+
 	return false;
 }
 
@@ -685,9 +667,6 @@ static bool hl_is_device_internal_memory_va(struct hl_device *hdev, u64 addr,
 	struct asic_fixed_properties *prop = &hdev->asic_prop;
 	u64 dram_start_addr, dram_end_addr;
 
-	if (!hdev->mmu_enable)
-		return false;
-
 	if (prop->dram_supports_virtual_memory) {
 		dram_start_addr = prop->dmmu.start_addr;
 		dram_end_addr = prop->dmmu.end_addr;
@@ -1756,17 +1735,15 @@ static void add_files_to_device(struct hl_device *hdev, struct hl_dbg_device_ent
 	}
 }
 
-void hl_debugfs_add_device(struct hl_device *hdev)
+int hl_debugfs_device_init(struct hl_device *hdev)
 {
 	struct hl_dbg_device_entry *dev_entry = &hdev->hl_debugfs;
 	int count = ARRAY_SIZE(hl_debugfs_list);
 
 	dev_entry->hdev = hdev;
-	dev_entry->entry_arr = kmalloc_array(count,
-					sizeof(struct hl_debugfs_entry),
-					GFP_KERNEL);
+	dev_entry->entry_arr = kmalloc_array(count, sizeof(struct hl_debugfs_entry), GFP_KERNEL);
 	if (!dev_entry->entry_arr)
-		return;
+		return -ENOMEM;
 
 	dev_entry->data_dma_blob_desc.size = 0;
 	dev_entry->data_dma_blob_desc.data = NULL;
@@ -1787,21 +1764,14 @@ void hl_debugfs_add_device(struct hl_device *hdev)
 	spin_lock_init(&dev_entry->userptr_spinlock);
 	mutex_init(&dev_entry->ctx_mem_hash_mutex);
 
-	dev_entry->root = debugfs_create_dir(dev_name(hdev->dev),
-						hl_debug_root);
-
-	add_files_to_device(hdev, dev_entry, dev_entry->root);
-	if (!hdev->asic_prop.fw_security_enabled)
-		add_secured_nodes(dev_entry, dev_entry->root);
+	return 0;
 }
 
-void hl_debugfs_remove_device(struct hl_device *hdev)
+void hl_debugfs_device_fini(struct hl_device *hdev)
 {
 	struct hl_dbg_device_entry *entry = &hdev->hl_debugfs;
 	int i;
 
-	debugfs_remove_recursive(entry->root);
-
 	mutex_destroy(&entry->ctx_mem_hash_mutex);
 	mutex_destroy(&entry->file_mutex);
 
@@ -1814,6 +1784,24 @@ void hl_debugfs_remove_device(struct hl_device *hdev)
 	kfree(entry->entry_arr);
 }
 
+void hl_debugfs_add_device(struct hl_device *hdev)
+{
+	struct hl_dbg_device_entry *dev_entry = &hdev->hl_debugfs;
+
+	dev_entry->root = debugfs_create_dir(dev_name(hdev->dev), hl_debug_root);
+
+	add_files_to_device(hdev, dev_entry, dev_entry->root);
+	if (!hdev->asic_prop.fw_security_enabled)
+		add_secured_nodes(dev_entry, dev_entry->root);
+}
+
+void hl_debugfs_remove_device(struct hl_device *hdev)
+{
+	struct hl_dbg_device_entry *entry = &hdev->hl_debugfs;
+
+	debugfs_remove_recursive(entry->root);
+}
+
 void hl_debugfs_add_file(struct hl_fpriv *hpriv)
 {
 	struct hl_dbg_device_entry *dev_entry = &hpriv->hdev->hl_debugfs;