Re: [PATCH v9 4/5] cxl/core, EINJ: Add EINJ CXL debugfs files and EINJ helper functions

[Ben Cheatham <benjamin.cheatham@amd.com>]

On 1/16/24 6:06 PM, Dan Williams wrote:
> Ben Cheatham wrote:
>> Implement CXL helper functions in the EINJ module for getting/injecting
>> available CXL protocol error types and export them to sysfs under
>> kernel/debug/cxl.
>>
>> The kernel/debug/cxl/einj_types file will print the available CXL
>> protocol errors in the same format as the available_error_types
>> file provided by the EINJ module. The
>> kernel/debug/cxl/$dport_dev/einj_inject is functionally the same as the
>> error_type and error_inject files provided by the EINJ module, i.e.:
>> writing an error type into $dport_dev/einj_inject will inject said error
>> type into the CXL dport represented by $dport_dev.
>>
>> Signed-off-by: Ben Cheatham <Benjamin.Cheatham@amd.com>
>> ---
>>  Documentation/ABI/testing/debugfs-cxl |  22 ++++
>>  drivers/acpi/apei/einj.c              | 144 ++++++++++++++++++++++++--
>>  drivers/cxl/core/port.c               |  39 +++++++
>>  include/linux/einj-cxl.h              |  42 ++++++++
>>  4 files changed, 237 insertions(+), 10 deletions(-)
>>  create mode 100644 include/linux/einj-cxl.h
>>
>> diff --git a/Documentation/ABI/testing/debugfs-cxl b/Documentation/ABI/testing/debugfs-cxl
>> index fe61d372e3fa..bcd985cca66a 100644
>> --- a/Documentation/ABI/testing/debugfs-cxl
>> +++ b/Documentation/ABI/testing/debugfs-cxl
>> @@ -33,3 +33,25 @@ Description:
>>  		device cannot clear poison from the address, -ENXIO is returned.
>>  		The clear_poison attribute is only visible for devices
>>  		supporting the capability.
>> +
>> +What:		/sys/kernel/debug/cxl/einj_types
>> +Date:		January, 2024
>> +KernelVersion:	v6.9
>> +Contact:	linux-cxl@vger.kernel.org
>> +Description:
>> +		(RO) Prints the CXL protocol error types made available by
>> +		the platform in the format "0x<error number>	<error type>".
>> +		The <error number> can be written to einj_inject to inject
>> +		<error type> into a chosen dport.
>> +
>> +What:		/sys/kernel/debug/cxl/$dport_dev/einj_inject
>> +Date:		January, 2024
>> +KernelVersion:	v6.9
>> +Contact:	linux-cxl@vger.kernel.org
>> +Description:
>> +		(WO) Writing an integer to this file injects the corresponding
>> +		CXL protocol error into $dport_dev ($dport_dev will be a device
>> +		name from /sys/bus/pci/devices). The integer to type mapping for
>> +		injection can be found by reading from einj_types. If the dport
>> +		was enumerated in RCH mode, a CXL 1.1 error is injected, otherwise
>> +		a CXL 2.0 error is injected.
>> diff --git a/drivers/acpi/apei/einj.c b/drivers/acpi/apei/einj.c
>> index 10d51cd73fa4..c3ec03583946 100644
>> --- a/drivers/acpi/apei/einj.c
>> +++ b/drivers/acpi/apei/einj.c
>> @@ -21,6 +21,7 @@
>>  #include <linux/nmi.h>
>>  #include <linux/delay.h>
>>  #include <linux/mm.h>
>> +#include <linux/einj-cxl.h>
>>  #include <linux/platform_device.h>
>>  #include <asm/unaligned.h>
>>  
>> @@ -37,6 +38,12 @@
>>  #define MEM_ERROR_MASK		(ACPI_EINJ_MEMORY_CORRECTABLE | \
>>  				ACPI_EINJ_MEMORY_UNCORRECTABLE | \
>>  				ACPI_EINJ_MEMORY_FATAL)
>> +#define CXL_ERROR_MASK		(ACPI_EINJ_CXL_CACHE_CORRECTABLE | \
>> +				ACPI_EINJ_CXL_CACHE_UNCORRECTABLE | \
>> +				ACPI_EINJ_CXL_CACHE_FATAL | \
>> +				ACPI_EINJ_CXL_MEM_CORRECTABLE | \
>> +				ACPI_EINJ_CXL_MEM_UNCORRECTABLE | \
>> +				ACPI_EINJ_CXL_MEM_FATAL)
>>  
>>  /*
>>   * ACPI version 5 provides a SET_ERROR_TYPE_WITH_ADDRESS action.
>> @@ -544,8 +551,11 @@ static int einj_error_inject(u32 type, u32 flags, u64 param1, u64 param2,
>>  	if (type & ACPI5_VENDOR_BIT) {
>>  		if (vendor_flags != SETWA_FLAGS_MEM)
>>  			goto inject;
>> -	} else if (!(type & MEM_ERROR_MASK) && !(flags & SETWA_FLAGS_MEM))
>> +	} else if (!(type & MEM_ERROR_MASK) && !(flags & SETWA_FLAGS_MEM)) {
>>  		goto inject;
>> +	} else if ((type & CXL_ERROR_MASK) && (flags & SETWA_FLAGS_MEM)) {
>> +		goto inject;
>> +	}
>>  
>>  	/*
>>  	 * Disallow crazy address masks that give BIOS leeway to pick
>> @@ -597,6 +607,9 @@ static const char * const einj_error_type_string[] = {
>>  	"0x00000200\tPlatform Correctable\n",
>>  	"0x00000400\tPlatform Uncorrectable non-fatal\n",
>>  	"0x00000800\tPlatform Uncorrectable fatal\n",
>> +};
>> +
>> +static const char * const einj_cxl_error_type_string[] = {
>>  	"0x00001000\tCXL.cache Protocol Correctable\n",
>>  	"0x00002000\tCXL.cache Protocol Uncorrectable non-fatal\n",
>>  	"0x00004000\tCXL.cache Protocol Uncorrectable fatal\n",
>> @@ -622,29 +635,44 @@ static int available_error_type_show(struct seq_file *m, void *v)
>>  
>>  DEFINE_SHOW_ATTRIBUTE(available_error_type);
>>  
>> -static int error_type_get(void *data, u64 *val)
>> +int einj_cxl_available_error_type_show(struct seq_file *m, void *v)
>>  {
>> -	*val = error_type;
>> +	int cxl_err, rc;
>> +	u32 available_error_type = 0;
>> +
>> +	if (!einj_initialized)
>> +		return -ENXIO;
>> +
>> +	rc = einj_get_available_error_type(&available_error_type);
>> +	if (rc)
>> +		return rc;
>> +
>> +	for (int pos = 0; pos < ARRAY_SIZE(einj_cxl_error_type_string); pos++) {
>> +		cxl_err = ACPI_EINJ_CXL_CACHE_CORRECTABLE << pos;
>> +
>> +		if (available_error_type & cxl_err)
>> +			seq_puts(m, einj_cxl_error_type_string[pos]);
>> +	}
>>  
>>  	return 0;
>>  }
>> +EXPORT_SYMBOL_NS_GPL(einj_cxl_available_error_type_show, CXL);
>>  
>> -static int error_type_set(void *data, u64 val)
>> +static int validate_error_type(u64 type)
>>  {
>> +	u32 tval, vendor, available_error_type = 0;
>>  	int rc;
>> -	u32 available_error_type = 0;
>> -	u32 tval, vendor;
>>  
>>  	/* Only low 32 bits for error type are valid */
>> -	if (val & GENMASK_ULL(63, 32))
>> +	if (type & GENMASK_ULL(63, 32))
>>  		return -EINVAL;
>>  
>>  	/*
>>  	 * Vendor defined types have 0x80000000 bit set, and
>>  	 * are not enumerated by ACPI_EINJ_GET_ERROR_TYPE
>>  	 */
>> -	vendor = val & ACPI5_VENDOR_BIT;
>> -	tval = val & 0x7fffffff;
>> +	vendor = type & ACPI5_VENDOR_BIT;
>> +	tval = type & 0x7fffffff;
>>  
>>  	/* Only one error type can be specified */
>>  	if (tval & (tval - 1))
>> @@ -653,9 +681,105 @@ static int error_type_set(void *data, u64 val)
>>  		rc = einj_get_available_error_type(&available_error_type);
>>  		if (rc)
>>  			return rc;
>> -		if (!(val & available_error_type))
>> +		if (!(type & available_error_type))
>>  			return -EINVAL;
>>  	}
>> +
>> +	return 0;
>> +}
>> +
>> +static int cxl_dport_get_sbdf(struct pci_dev *dport_dev, u64 *sbdf)
>> +{
>> +	struct pci_bus *pbus;
>> +	struct pci_host_bridge *bridge;
>> +	u64 seg = 0, bus;
>> +
>> +	pbus = dport_dev->bus;
>> +	bridge = pci_find_host_bridge(pbus);
>> +
>> +	if (!bridge)
>> +		return -ENODEV;
>> +
>> +	if (bridge->domain_nr != PCI_DOMAIN_NR_NOT_SET)
>> +		seg = bridge->domain_nr << 24;
>> +
>> +	bus = pbus->number << 16;
>> +	*sbdf = seg | bus | dport_dev->devfn;
>> +
>> +	return 0;
>> +}
>> +
>> +int einj_cxl_inject_rch_error(u64 rcrb, u64 type)
>> +{
>> +	u64 param1 = 0, param2 = 0, param4 = 0;
>> +	u32 flags;
>> +	int rc;
>> +
>> +	if (!einj_initialized)
>> +		return -ENXIO;
>> +
>> +	/* Only CXL error types can be specified */
>> +	if (type & ~CXL_ERROR_MASK || (type & ACPI5_VENDOR_BIT))
>> +		return -EINVAL;
>> +
>> +	rc = validate_error_type(type);
>> +	if (rc)
>> +		return rc;
>> +
>> +	param1 = (u64) rcrb;
>> +	param2 = 0xfffffffffffff000;
>> +	flags = 0x2;
>> +
>> +	return einj_error_inject(type, flags, param1, param2, 0, param4);
>> +}
>> +EXPORT_SYMBOL_NS_GPL(einj_cxl_inject_rch_error, CXL);
>> +
>> +int einj_cxl_inject_error(struct pci_dev *dport, u64 type)
>> +{
>> +	u64 param1 = 0, param2 = 0, param4 = 0;
>> +	u32 flags;
>> +	int rc;
>> +
>> +	if (!einj_initialized)
>> +		return -ENXIO;
>> +
>> +	/* Only CXL error types can be specified */
>> +	if (type & ~CXL_ERROR_MASK || (type & ACPI5_VENDOR_BIT))
>> +		return -EINVAL;
>> +
>> +	rc = validate_error_type(type);
>> +	if (rc)
>> +		return rc;
>> +
>> +	rc = cxl_dport_get_sbdf(dport, &param4);
>> +	if (rc)
>> +		return rc;
>> +
>> +	flags = 0x4;
>> +
>> +	return einj_error_inject(type, flags, param1, param2, 0, param4);
>> +}
>> +EXPORT_SYMBOL_NS_GPL(einj_cxl_inject_error, CXL);
>> +
>> +static int error_type_get(void *data, u64 *val)
>> +{
>> +	*val = error_type;
>> +
>> +	return 0;
>> +}
>> +
>> +static int error_type_set(void *data, u64 val)
>> +{
>> +	int rc;
>> +
>> +	/* CXL error types have to be injected from cxl debugfs */
>> +	if (val & CXL_ERROR_MASK && !(val & ACPI5_VENDOR_BIT))
>> +		return -EINVAL;
>> +
>> +	rc = validate_error_type(val);
>> +	if (rc)
>> +		return rc;
>> +
>>  	error_type = val;
>>  
>>  	return 0;
>> diff --git a/drivers/cxl/core/port.c b/drivers/cxl/core/port.c
>> index 8c00fd6be730..a41618ce380b 100644
>> --- a/drivers/cxl/core/port.c
>> +++ b/drivers/cxl/core/port.c
>> @@ -3,6 +3,7 @@
>>  #include <linux/platform_device.h>
>>  #include <linux/memregion.h>
>>  #include <linux/workqueue.h>
>> +#include <linux/einj-cxl.h>
>>  #include <linux/debugfs.h>
>>  #include <linux/device.h>
>>  #include <linux/module.h>
>> @@ -797,6 +798,37 @@ static int cxl_dport_setup_regs(struct device *host, struct cxl_dport *dport,
>>  	return rc;
>>  }
>>  
>> +DEFINE_SHOW_ATTRIBUTE(einj_cxl_available_error_type);
>> +
>> +static int cxl_einj_inject(void *data, u64 type)
>> +{
>> +	struct cxl_dport *dport = data;
>> +
>> +	if (dport->rch)
>> +		return einj_cxl_inject_rch_error(dport->rcrb.base, type);
>> +
>> +	return einj_cxl_inject_error(to_pci_dev(dport->dport_dev), type);
>> +}
>> +DEFINE_DEBUGFS_ATTRIBUTE(cxl_einj_inject_fops, NULL, cxl_einj_inject, "%llx\n");
>> +
>> +static void cxl_debugfs_create_dport_dir(struct cxl_dport *dport)
>> +{
>> +	struct dentry *dir;
>> +
>> +	/*
>> +	 * dport_dev needs to be a PCIe port for CXL 2.0+ ports because
>> +	 * EINJ expects a dport SBDF to be specified for 2.0 error injection.
>> +	 */
>> +	if (!IS_ENABLED(CONFIG_CXL_EINJ) ||
>> +	    (!dport->rch && !dev_is_pci(dport->dport_dev)))
>> +		return;
> 
> Perhaps this IS_ENABLED(CONFIG_CXL_EINJ) and the one below should be
> replace with a helper that also queries einj_initialized? Such a helper
> would naturally have a static inline stub that fails in the
> CONFIG_CXL_EINJ=n case, and it otherwise behaves the same as the base
> ACPI case where the debugfs files do not appear if init fails.

Ok, I'll add that in.