Re: [RFC PATCH v7 12/12] memory: RAS2: Add memory RAS2 driver

[Daniel Ferguson <danielf@os.amperecomputing.com>]

> +/*
> + * The below functions are exposed to OSPM, to query, configure and
> + * initiate memory patrol scrub.
> + */
> +static int ras2_is_patrol_scrub_support(struct ras2_context *ras2_ctx)
> +{
> +	int ret;
> +	struct acpi_ras2_shared_memory  __iomem *generic_comm_base;
> +
> +	if (!ras2_ctx || !ras2_ctx->pcc_comm_addr)
> +		return -EFAULT;
> +
> +	generic_comm_base = ras2_ctx->pcc_comm_addr;
> +	guard(spinlock_irqsave)(&ras2_ctx->spinlock);
> +	generic_comm_base->set_capabilities[0] = 0;
> +
> +	/* send command for reading RAS2 capabilities */
> +	ret = ras2_send_pcc_cmd(ras2_ctx, RAS2_PCC_CMD_EXEC);
> +	if (ret) {
> +		dev_err(ras2_ctx->dev,
> +			"%s: ras2_send_pcc_cmd failed\n", __func__);
> +		return ret;
> +	}
> +
> +	return generic_comm_base->features[0] & RAS2_SUPPORT_HW_PARTOL_SCRUB;

Since firmware populates the feature bitmask on initialization, it would 
seem
that we do not need to send a PCC CMD EXEC to read RAS2 capabilities.
> +}
> +
> +static int ras2_get_patrol_scrub_params(struct ras2_context *ras2_ctx,
> +					struct ras2_scrub_params *params)
> +{
> +	int ret = 0;
> +	u8  min_supp_scrub_rate, max_supp_scrub_rate;
> +	struct acpi_ras2_shared_memory  __iomem *generic_comm_base;
> +	struct acpi_ras2_patrol_scrub_parameter __iomem *patrol_scrub_params;
> +
> +	if (!ras2_ctx || !ras2_ctx->pcc_comm_addr)
> +		return -EFAULT;
> +
> +	generic_comm_base = ras2_ctx->pcc_comm_addr;
> +	patrol_scrub_params = ras2_ctx->pcc_comm_addr + sizeof(*generic_comm_base);
> +
> +	guard(spinlock_irqsave)(&ras2_ctx->spinlock);
> +	generic_comm_base->set_capabilities[0] = RAS2_SUPPORT_HW_PARTOL_SCRUB;
> +	/* send command for reading RAS2 capabilities */
> +	ret = ras2_send_pcc_cmd(ras2_ctx, RAS2_PCC_CMD_EXEC);
> +	if (ret) {
> +		dev_err(ras2_ctx->dev,
> +			"%s: ras2_send_pcc_cmd failed\n", __func__);
> +		return ret;
> +	}

Similarly, since firmware populates the feature bitmask on
initialization, it would seem that we do not need to send
a PCC CMD EXEC to read RAS2 capabilities.
> +
> +	if (!(generic_comm_base->features[0] & RAS2_SUPPORT_HW_PARTOL_SCRUB) ||
> +	    !(generic_comm_base->num_parameter_blocks)) {
> +		dev_err(ras2_ctx->dev,
> +			"%s: Platform does not support HW Patrol Scrubber\n", __func__);
> +		return -EOPNOTSUPP;
> +	}
> +
> +	if (!patrol_scrub_params->requested_address_range[1]) {
> +		dev_err(ras2_ctx->dev,
> +			"%s: Invalid requested address range, \
> +			requested_address_range[0]=0x%llx \
> +			requested_address_range[1]=0x%llx\n",
> +			__func__,
> +			patrol_scrub_params->requested_address_range[0],
> +			patrol_scrub_params->requested_address_range[1]);
> +		return -EOPNOTSUPP;
> +	}
> +
> +	generic_comm_base->set_capabilities[0] = RAS2_SUPPORT_HW_PARTOL_SCRUB;
> +	patrol_scrub_params->header.type = RAS2_TYPE_PATROL_SCRUB;

header.type should already be populated by firmware. Is assigning it
here necessary?
> +	patrol_scrub_params->patrol_scrub_command = RAS2_GET_PATROL_PARAMETERS;
> +
> +	/* send command for reading the HW patrol scrub parameters */
> +	ret = ras2_send_pcc_cmd(ras2_ctx, RAS2_PCC_CMD_EXEC);
> +	if (ret) {
> +		dev_err(ras2_ctx->dev,
> +			"%s: failed to read HW patrol scrub parameters\n",
> +			__func__);
> +		return ret;
> +	}
> +
> +	/* copy output scrub parameters */
> +	params->addr_base = patrol_scrub_params->actual_address_range[0];
> +	params->addr_size = patrol_scrub_params->actual_address_range[1];
> +	params->flags = patrol_scrub_params->flags;
> +	params->rate = FIELD_GET(RAS2_PATROL_SCRUB_RATE_OUT_MASK,
> +				 patrol_scrub_params->scrub_params_out);
> +	min_supp_scrub_rate = FIELD_GET(RAS2_PATROL_SCRUB_MIN_RATE_OUT_MASK,
> +					patrol_scrub_params->scrub_params_out);
> +	max_supp_scrub_rate = FIELD_GET(RAS2_PATROL_SCRUB_MAX_RATE_OUT_MASK,
> +					patrol_scrub_params->scrub_params_out);
> +	snprintf(params->rate_avail, RAS2_MAX_RATE_RANGE_LENGTH,
> +		 "%d-%d", min_supp_scrub_rate, max_supp_scrub_rate);
> +
> +	return 0;
> +}
> +
> +static int ras2_enable_patrol_scrub(struct ras2_context *ras2_ctx, bool enable)
> +{
> +	int ret = 0;
> +	struct ras2_scrub_params params;
> +	struct acpi_ras2_shared_memory  __iomem *generic_comm_base;
> +	u8 scrub_rate_to_set, min_supp_scrub_rate, max_supp_scrub_rate;
> +	struct acpi_ras2_patrol_scrub_parameter __iomem *patrol_scrub_params;
> +
> +	if (!ras2_ctx || !ras2_ctx->pcc_comm_addr)
> +		return -EFAULT;
> +
> +	generic_comm_base = ras2_ctx->pcc_comm_addr;
> +	patrol_scrub_params = ras2_ctx->pcc_comm_addr + sizeof(*generic_comm_base);
> +
> +	if (enable) {
> +		ret = ras2_get_patrol_scrub_params(ras2_ctx, &params);
> +		if (ret)
> +			return ret;
> +	}
> +
> +	guard(spinlock_irqsave)(&ras2_ctx->spinlock);
> +	generic_comm_base->set_capabilities[0] = RAS2_SUPPORT_HW_PARTOL_SCRUB;
> +	patrol_scrub_params->header.type = RAS2_TYPE_PATROL_SCRUB;

header.type should already be populated by firmware. Is assigning it
here necessary?
> +
> +	if (enable) {
> +		patrol_scrub_params->patrol_scrub_command = RAS2_START_PATROL_SCRUBBER;
> +		patrol_scrub_params->requested_address_range[0] = params.addr_base;
> +		patrol_scrub_params->requested_address_range[1] = params.addr_size;
> +
> +		scrub_rate_to_set = FIELD_GET(RAS2_PATROL_SCRUB_RATE_IN_MASK,
> +					      patrol_scrub_params->scrub_params_in);
> +		min_supp_scrub_rate = FIELD_GET(RAS2_PATROL_SCRUB_MIN_RATE_OUT_MASK,
> +						patrol_scrub_params->scrub_params_out);
> +		max_supp_scrub_rate = FIELD_GET(RAS2_PATROL_SCRUB_MAX_RATE_OUT_MASK,
> +						patrol_scrub_params->scrub_params_out);
> +		if (scrub_rate_to_set < min_supp_scrub_rate ||
> +		    scrub_rate_to_set > max_supp_scrub_rate) {
> +			dev_warn(ras2_ctx->dev,
> +				 "patrol scrub rate to set is out of the supported range\n");
> +			dev_warn(ras2_ctx->dev,
> +				 "min_supp_scrub_rate=%d max_supp_scrub_rate=%d\n",
> +				 min_supp_scrub_rate, max_supp_scrub_rate);
> +			return -EINVAL;
> +		}
> +	} else {
> +		patrol_scrub_params->patrol_scrub_command = RAS2_STOP_PATROL_SCRUBBER;
> +	}
> +
> +	/* send command for enable/disable HW patrol scrub */
> +	ret = ras2_send_pcc_cmd(ras2_ctx, RAS2_PCC_CMD_EXEC);
> +	if (ret) {
> +		pr_err("%s: failed to enable/disable the HW patrol scrub\n", __func__);
> +		return ret;
> +	}
> +
> +	return 0;
> +}
> +
> +static int ras2_enable_background_scrub(struct ras2_context *ras2_ctx, bool enable)
> +{
> +	int ret;
> +	struct acpi_ras2_shared_memory  __iomem *generic_comm_base;
> +	struct acpi_ras2_patrol_scrub_parameter __iomem *patrol_scrub_params;
> +
> +	if (!ras2_ctx || !ras2_ctx->pcc_comm_addr)
> +		return -EFAULT;
> +
> +	generic_comm_base = ras2_ctx->pcc_comm_addr;
> +	patrol_scrub_params = ras2_ctx->pcc_comm_addr + sizeof(*generic_comm_base);
> +
> +	guard(spinlock_irqsave)(&ras2_ctx->spinlock);
> +	generic_comm_base->set_capabilities[0] = RAS2_SUPPORT_HW_PARTOL_SCRUB;
> +	patrol_scrub_params->header.type = RAS2_TYPE_PATROL_SCRUB;

header.type should already be populated by firmware. Is assigning it
here necessary?
> +	patrol_scrub_params->patrol_scrub_command = RAS2_START_PATROL_SCRUBBER;
> +
> +	patrol_scrub_params->scrub_params_in &= ~RAS2_PATROL_SCRUB_EN_BACKGROUND;
> +	patrol_scrub_params->scrub_params_in |= FIELD_PREP(RAS2_PATROL_SCRUB_EN_BACKGROUND,
> +							   enable);
> +
> +	/* send command for enable/disable HW patrol scrub */
> +	ret = ras2_send_pcc_cmd(ras2_ctx, RAS2_PCC_CMD_EXEC);
> +	if (ret) {
> +		dev_err(ras2_ctx->dev,
> +			"%s: failed to enable/disable background patrol scrubbing\n",
> +			__func__);
> +		return ret;
> +	}
> +
> +	return 0;
> +}
> +static int ras2_set_patrol_scrub_params(struct ras2_context *ras2_ctx,
> +					struct ras2_scrub_params *params, u8 param_type)
> +{
> +	struct acpi_ras2_shared_memory  __iomem *generic_comm_base;
> +	struct acpi_ras2_patrol_scrub_parameter __iomem *patrol_scrub_params;
> +
> +	if (!ras2_ctx || !ras2_ctx->pcc_comm_addr)
> +		return -EFAULT;
> +
> +	generic_comm_base = ras2_ctx->pcc_comm_addr;
> +	patrol_scrub_params = ras2_ctx->pcc_comm_addr + sizeof(*generic_comm_base);
> +
> +	guard(spinlock_irqsave)(&ras2_ctx->spinlock);
> +	patrol_scrub_params->header.type = RAS2_TYPE_PATROL_SCRUB;
> +	if (param_type == RAS2_MEM_SCRUB_PARAM_ADDR_BASE && params->addr_base) {
> +		patrol_scrub_params->requested_address_range[0] = params->addr_base;
> +	} else if (param_type == RAS2_MEM_SCRUB_PARAM_ADDR_SIZE && params->addr_size) {
> +		patrol_scrub_params->requested_address_range[1] = params->addr_size;
> +	} else if (param_type == RAS2_MEM_SCRUB_PARAM_RATE) {
> +		patrol_scrub_params->scrub_params_in &= ~RAS2_PATROL_SCRUB_RATE_IN_MASK;
> +		patrol_scrub_params->scrub_params_in |= FIELD_PREP(RAS2_PATROL_SCRUB_RATE_IN_MASK,
> +								   params->rate);
> +	} else {
> +		dev_err(ras2_ctx->dev, "Invalid patrol scrub parameter to set\n");
> +		return -EINVAL;
> +	}
> +
> +	return 0;
> +}
> +
> +static const struct ras2_hw_scrub_ops ras2_hw_ops = {
> +	.enable_scrub = ras2_enable_patrol_scrub,
> +	.enable_background_scrub = ras2_enable_background_scrub,
> +	.get_scrub_params = ras2_get_patrol_scrub_params,
> +	.set_scrub_params = ras2_set_patrol_scrub_params,
> +};
> +
> +static const struct scrub_ops ras2_scrub_ops = {
> +	.is_visible = ras2_hw_scrub_is_visible,
> +	.read = ras2_hw_scrub_read,
> +	.write = ras2_hw_scrub_write,
> +	.read_string = ras2_hw_scrub_read_strings,
> +};
> +
> +static DEFINE_IDA(ras2_ida);
> +
> +static void devm_ras2_release(void *ctx)
> +{
> +	struct ras2_context *ras2_ctx = ctx;
> +
> +	ida_free(&ras2_ida, ras2_ctx->id);
> +	ras2_unregister_pcc_channel(ras2_ctx);
> +}
> +
> +static int ras2_probe(struct platform_device *pdev)
> +{
> +	int ret, id;
> +	struct mbox_client *cl;
> +	struct device *hw_scrub_dev;
> +	struct ras2_context *ras2_ctx;
> +	char scrub_name[RAS2_MAX_NAME_LENGTH];
> +
> +	ras2_ctx = devm_kzalloc(&pdev->dev, sizeof(*ras2_ctx), GFP_KERNEL);
> +	if (!ras2_ctx)
> +		return -ENOMEM;
> +
> +	ras2_ctx->dev = &pdev->dev;
> +	ras2_ctx->ops = &ras2_hw_ops;
> +	spin_lock_init(&ras2_ctx->spinlock);
> +	platform_set_drvdata(pdev, ras2_ctx);
> +
> +	cl = &ras2_ctx->mbox_client;
> +	/* Request mailbox channel */
> +	cl->dev = &pdev->dev;
> +	cl->tx_done = ras2_tx_done;
> +	cl->knows_txdone = true;
> +	ras2_ctx->pcc_subspace_idx = *((int *)pdev->dev.platform_data);
> +	dev_dbg(&pdev->dev, "pcc-subspace-id=%d\n", ras2_ctx->pcc_subspace_idx);
> +	ret = ras2_register_pcc_channel(ras2_ctx);

In our enabling activities, we have found a challenge here.
Our hardware has a single PCC channel corresponding to a single
platform-wide scrub interface. This driver, following the ACPI spec,
will create a new scrub node for each NUMA node. However, for us,
this means that each scrub device will try to map the same PCC channel,
and this causes an error.
> +	if (ret < 0)
> +		return ret;
> +
> +	ret = devm_add_action_or_reset(&pdev->dev, devm_ras2_release, ras2_ctx);
> +	if (ret < 0)
> +		return ret;
> +
> +	if (ras2_is_patrol_scrub_support(ras2_ctx)) {
> +		id = ida_alloc(&ras2_ida, GFP_KERNEL);
> +		if (id < 0)
> +			return id;
> +		ras2_ctx->id = id;
> +		snprintf(scrub_name, sizeof(scrub_name), "%s%d", RAS2_SCRUB, id);
> +		dev_set_name(&pdev->dev, RAS2_ID_FORMAT, id);
> +		hw_scrub_dev = devm_scrub_device_register(&pdev->dev, scrub_name,
> +							  ras2_ctx, &ras2_scrub_ops,
> +							  0, NULL);
> +		if (PTR_ERR_OR_ZERO(hw_scrub_dev))
> +			return PTR_ERR_OR_ZERO(hw_scrub_dev);
> +	}
> +	ras2_ctx->scrub_dev = hw_scrub_dev;
> +
> +	return 0;
> +}