Re: [PATCH v3 1/8] cxl/memdev: Add support for the Inject Poison mailbox command

[Jonathan Cameron <Jonathan.Cameron@Huawei.com>]

On Wed,  1 Mar 2023 13:36:26 -0800
alison.schofield@intel.com wrote:

> From: Alison Schofield <alison.schofield@intel.com>
> 
> CXL devices optionally support the INJECT POISON mailbox command. Add
> a sysfs attribute and memdev driver support for injecting poison. The
> attribute is only visible for devices supporting the capability when
> the kernel is built with CONFIG_CXL_POISON_INJECT.
> 
> When a Device Physical Address (DPA) is written to the inject_poison
> sysfs attribute, send an inject poison command to the device for the
> specified address.
> 
> Per the CXL Specification (3.0 8.2.9.8.4.2), after receiving a valid
> inject poison request, the device will return poison when the address
> is accessed through the CXL.mem bus. Injecting poison adds the address
> to the device's Poison List and the error source is set to Injected.
> In addition, the device adds a poison creation event to its internal
> Informational Event log, updates the Event Status register, and if
> configured, interrupts the host.
> 
> Also, per the CXL Specification, it is not an error to inject poison
> into an address that already has poison present and no error is
> returned from the device.
> 
> Signed-off-by: Alison Schofield <alison.schofield@intel.com>
> Reviewed-by: Jonathan Cameron <Jonathan.Cameron@huawei.com>

One query inline about the protection against DPAs that are mapped.
I'm not sure the checks won't rule out injection on a DPA that is
a configured but not yet committed HDM decoder range.
I think we should allow injection into such a memory address as it's
not in use.

> ---
>  Documentation/ABI/testing/sysfs-bus-cxl | 22 ++++++
>  drivers/cxl/Kconfig                     | 11 +++
>  drivers/cxl/core/memdev.c               | 97 +++++++++++++++++++++++++
>  drivers/cxl/cxlmem.h                    |  5 ++
>  4 files changed, 135 insertions(+)
> 
> diff --git a/Documentation/ABI/testing/sysfs-bus-cxl b/Documentation/ABI/testing/sysfs-bus-cxl
> index d9421c965a3b..e19d1020f30a 100644
> --- a/Documentation/ABI/testing/sysfs-bus-cxl
> +++ b/Documentation/ABI/testing/sysfs-bus-cxl
> @@ -429,3 +429,25 @@ Description:
>  		attribute is only visible for devices supporting the
>  		capability. The retrieved errors are logged as kernel
>  		trace events with the label 'cxl_poison'.
> +
> +
> +What:		/sys/bus/cxl/devices/memX/inject_poison
> +Date:		January, 2023
> +KernelVersion:	v6.3
> +Contact:	linux-cxl@vger.kernel.org
> +Description:
> +		(WO) When a Device Physical Address (DPA) is written to this
> +		attribute, the memdev driver sends an inject poison command to
> +		the device for the specified address. The DPA must be 64-byte
> +		aligned and the length of the injected poison is 64-bytes. If
> +		successful, the device returns poison when the address is
> +		accessed through the CXL.mem bus. Injecting poison adds the
> +		address to the device's Poison List and the error source is set
> +		to Injected. In addition, the device adds a poison creation
> +		event to its internal Informational Event log, updates the
> +		Event Status register, and if configured, interrupts the host.
> +		It is not an error to inject poison into an address that
> +		already has poison present and no error is returned. The
> +		inject_poison attribute is only visible for devices supporting
> +		the capability. Kconfig option CXL_POISON_INJECT must be on
> +		to enable this option. The default is off.
> diff --git a/drivers/cxl/Kconfig b/drivers/cxl/Kconfig
> index ff4e78117b31..a7ca0bbb8475 100644
> --- a/drivers/cxl/Kconfig
> +++ b/drivers/cxl/Kconfig
> @@ -139,4 +139,15 @@ config CXL_REGION_INVALIDATION_TEST
>  	  If unsure, or if this kernel is meant for production environments,
>  	  say N.
>  
> +config CXL_POISON_INJECT
> +	bool "CXL: Support CXL Memory Device Poison Inject"
> +	depends on CXL_MEM
> +	help
> +	  Selecting this option creates the sysfs attributes inject_poison
> +	  and clear_poison for CXL memory devices supporting the capability.
> +	  This option is intended for debug scenarios only and is disabled
> +	  by default. See Documentation/ABI/testing/sysfs-bus-cxl.

Default for almost everything in Kconfig is off.  Is it useful to call that out
for this one?  The unsure bit below is enough for me.

> +
> +	  If unsure, say N.
> +
>  endif
> diff --git a/drivers/cxl/core/memdev.c b/drivers/cxl/core/memdev.c
> index c11b7bc253b4..82e09b81e9c6 100644
> --- a/drivers/cxl/core/memdev.c
> +++ b/drivers/cxl/core/memdev.c
> @@ -165,6 +165,92 @@ static ssize_t trigger_poison_list_store(struct device *dev,
>  }
>  static DEVICE_ATTR_WO(trigger_poison_list);
>  
> +static int cxl_dpa_mapped(struct device *dev, void *data)
> +{
> +	struct cxl_endpoint_decoder *cxled;
> +	u64 *dpa = data;
> +
> +	if (!is_endpoint_decoder(dev))
> +		return 0;
> +
> +	cxled = to_cxl_endpoint_decoder(dev);
> +	if (!cxled->dpa_res || !resource_size(cxled->dpa_res))
> +		return 0;
> +
> +	if (*dpa <= cxled->dpa_res->end && *dpa >= cxled->dpa_res->start) {

I haven't chased it all the way through, but are we guaranteed that this
particular decoder is committed if we reach here? 
I think the resource is setup when we set size not at point of committing
so checking that may not be sufficient.

> +		dev_dbg(dev, "dpa:0x%llx mapped in region:%s\n",
> +			*dpa, dev_name(&cxled->cxld.region->dev));
> +		return 1;
> +	}
> +
> +	return 0;
> +}
> +
> +static int cxl_validate_poison_dpa(struct cxl_memdev *cxlmd, u64 dpa)
> +{
> +	struct cxl_dev_state *cxlds = cxlmd->cxlds;
> +	struct cxl_port *port;
> +	int rc;
> +
> +	if (!resource_size(&cxlds->dpa_res)) {
> +		dev_dbg(cxlds->dev, "device has no dpa resource\n");
> +		return -EINVAL;
> +	}
> +	if (dpa < cxlds->dpa_res.start || dpa > cxlds->dpa_res.end) {
> +		dev_dbg(cxlds->dev, "dpa:0x%llx not in resource:%pR\n",
> +			dpa, &cxlds->dpa_res);
> +		return -EINVAL;
> +	}
> +	if (!IS_ALIGNED(dpa, 64)) {
> +		dev_dbg(cxlds->dev, "dpa:0x%llx is not 64-byte aligned\n", dpa);
> +		return -EINVAL;
> +	}
> +	port = dev_get_drvdata(&cxlmd->dev);
> +	if (port && is_cxl_endpoint(port) && port->commit_end != -1) {
> +		rc = device_for_each_child(&port->dev, &dpa, cxl_dpa_mapped);
> +		if (rc)
> +			return -EINVAL;
> +	}
> +
> +	return 0;
> +}
> +
> +static ssize_t inject_poison_store(struct device *dev,
> +				   struct device_attribute *attr,
> +				   const char *buf, size_t len)
> +{
> +	struct cxl_memdev *cxlmd = to_cxl_memdev(dev);
> +	struct cxl_mbox_inject_poison inject;
> +	struct cxl_mbox_cmd mbox_cmd;
> +	u64 dpa;
> +	int rc;
> +
> +	rc = kstrtou64(buf, 0, &dpa);
> +	if (rc)
> +		return rc;
> +
> +	down_read(&cxl_dpa_rwsem);
> +	rc = cxl_validate_poison_dpa(cxlmd, dpa);
> +	if (rc) {
> +		up_read(&cxl_dpa_rwsem);
> +		return rc;
> +	}
> +
> +	inject = (struct cxl_mbox_inject_poison) {
> +		.address = cpu_to_le64(dpa)
> +	};
> +	mbox_cmd = (struct cxl_mbox_cmd) {
> +		.opcode = CXL_MBOX_OP_INJECT_POISON,
> +		.size_in = sizeof(inject),
> +		.payload_in = &inject,
> +	};
> +	rc = cxl_internal_send_cmd(cxlmd->cxlds, &mbox_cmd);
> +
> +	up_read(&cxl_dpa_rwsem);
> +	return rc ? rc : len;
> +}
> +static DEVICE_ATTR_WO(inject_poison);
> +
>  static struct attribute *cxl_memdev_attributes[] = {
>  	&dev_attr_serial.attr,
>  	&dev_attr_firmware_version.attr,
> @@ -172,6 +258,7 @@ static struct attribute *cxl_memdev_attributes[] = {
>  	&dev_attr_label_storage_size.attr,
>  	&dev_attr_numa_node.attr,
>  	&dev_attr_trigger_poison_list.attr,
> +	&dev_attr_inject_poison.attr,
>  	NULL,
>  };
>  
> @@ -198,6 +285,16 @@ static umode_t cxl_memdev_visible(struct kobject *kobj, struct attribute *a,
>  			      to_cxl_memdev(dev)->cxlds->enabled_cmds))
>  			return 0;
>  	}
> +	if (a == &dev_attr_inject_poison.attr) {
> +		struct device *dev = kobj_to_dev(kobj);
> +
> +		if (!IS_ENABLED(CONFIG_CXL_POISON_INJECT))
> +			return 0;
> +
> +		if (!test_bit(CXL_MEM_COMMAND_ID_INJECT_POISON,
> +			      to_cxl_memdev(dev)->cxlds->enabled_cmds))
> +			return 0;
> +	}
>  	return a->mode;
>  }
>  
> diff --git a/drivers/cxl/cxlmem.h b/drivers/cxl/cxlmem.h
> index a6eb1b42eb88..01d27f362cd6 100644
> --- a/drivers/cxl/cxlmem.h
> +++ b/drivers/cxl/cxlmem.h
> @@ -602,6 +602,11 @@ struct cxl_mbox_poison_payload_out {
>  #define CXL_POISON_SOURCE_INJECTED	3
>  #define CXL_POISON_SOURCE_VENDOR	7
>  
> +/* Inject & Clear Poison  CXL 3.0 Spec 8.2.9.8.4.2/3 */
> +struct cxl_mbox_inject_poison {
> +	__le64 address;
> +};
> +
>  /**
>   * struct cxl_mem_command - Driver representation of a memory device command
>   * @info: Command information as it exists for the UAPI