OpenCloudOS-Kernel/drivers/acpi/nfit/nfit.h

353 lines
9.6 KiB
C
Raw Normal View History

/* SPDX-License-Identifier: GPL-2.0-only */
/*
* NVDIMM Firmware Interface Table - NFIT
*
* Copyright(c) 2013-2015 Intel Corporation. All rights reserved.
*/
#ifndef __NFIT_H__
#define __NFIT_H__
#include <linux/workqueue.h>
#include <linux/libnvdimm.h>
#include <linux/ndctl.h>
#include <linux/types.h>
#include <linux/acpi.h>
#include <acpi/acuuid.h>
nfit, libnvdimm: limited/whitelisted dimm command marshaling mechanism There are currently 4 known similar but incompatible definitions of the command sets that can be sent to an NVDIMM through ACPI. It is also clear that future platform generations (ACPI or not) will continue to revise and extend the DIMM command set as new devices and use cases arrive. It is obviously untenable to continue to proliferate divergence of these command definitions, and to that end a standardization process has begun to provide for a unified specification. However, that leaves a problem about what to do with this first generation where vendors are already shipping divergence. The Linux kernel can support these initial diverged platforms without giving platform-firmware free reign to continue to diverge and compound kernel maintenance overhead. The kernel implementation can encourage standardization in two ways: 1/ Require that any function code that userspace wants to send be explicitly white-listed in the implementation. For ACPI this means function codes marked as supported by acpi_check_dsm() may only be invoked if they appear in the white-list. A function must be publicly documented before it is added to the white-list. 2/ The above restrictions can be trivially bypassed by using the "vendor-specific" payload command. However, since vendor-specific commands are by definition not publicly documented and have the potential to corrupt the kernel's view of the dimm state, we provide a toggle to disable vendor-specific operations. Enabling undefined behavior is a policy decision that can be made by the platform owner and encourages firmware implementations to choose public over private command implementations. Based on an initial patch from Jerry Hoemann Cc: Jerry Hoemann <jerry.hoemann@hpe.com> Cc: Christoph Hellwig <hch@infradead.org> Signed-off-by: Dan Williams <dan.j.williams@intel.com>
2016-04-29 07:23:43 +08:00
/* ACPI 6.1 */
#define UUID_NFIT_BUS "2f10e7a4-9e91-11e4-89d3-123b93f75cba"
nfit, libnvdimm: limited/whitelisted dimm command marshaling mechanism There are currently 4 known similar but incompatible definitions of the command sets that can be sent to an NVDIMM through ACPI. It is also clear that future platform generations (ACPI or not) will continue to revise and extend the DIMM command set as new devices and use cases arrive. It is obviously untenable to continue to proliferate divergence of these command definitions, and to that end a standardization process has begun to provide for a unified specification. However, that leaves a problem about what to do with this first generation where vendors are already shipping divergence. The Linux kernel can support these initial diverged platforms without giving platform-firmware free reign to continue to diverge and compound kernel maintenance overhead. The kernel implementation can encourage standardization in two ways: 1/ Require that any function code that userspace wants to send be explicitly white-listed in the implementation. For ACPI this means function codes marked as supported by acpi_check_dsm() may only be invoked if they appear in the white-list. A function must be publicly documented before it is added to the white-list. 2/ The above restrictions can be trivially bypassed by using the "vendor-specific" payload command. However, since vendor-specific commands are by definition not publicly documented and have the potential to corrupt the kernel's view of the dimm state, we provide a toggle to disable vendor-specific operations. Enabling undefined behavior is a policy decision that can be made by the platform owner and encourages firmware implementations to choose public over private command implementations. Based on an initial patch from Jerry Hoemann Cc: Jerry Hoemann <jerry.hoemann@hpe.com> Cc: Christoph Hellwig <hch@infradead.org> Signed-off-by: Dan Williams <dan.j.williams@intel.com>
2016-04-29 07:23:43 +08:00
/* https://pmem.io/documents/NVDIMM_DSM_Interface-V1.6.pdf */
#define UUID_NFIT_DIMM "4309ac30-0d11-11e4-9191-0800200c9a66"
#define UUID_INTEL_BUS "c7d8acd4-2df8-4b82-9f65-a325335af149"
nfit, libnvdimm: limited/whitelisted dimm command marshaling mechanism There are currently 4 known similar but incompatible definitions of the command sets that can be sent to an NVDIMM through ACPI. It is also clear that future platform generations (ACPI or not) will continue to revise and extend the DIMM command set as new devices and use cases arrive. It is obviously untenable to continue to proliferate divergence of these command definitions, and to that end a standardization process has begun to provide for a unified specification. However, that leaves a problem about what to do with this first generation where vendors are already shipping divergence. The Linux kernel can support these initial diverged platforms without giving platform-firmware free reign to continue to diverge and compound kernel maintenance overhead. The kernel implementation can encourage standardization in two ways: 1/ Require that any function code that userspace wants to send be explicitly white-listed in the implementation. For ACPI this means function codes marked as supported by acpi_check_dsm() may only be invoked if they appear in the white-list. A function must be publicly documented before it is added to the white-list. 2/ The above restrictions can be trivially bypassed by using the "vendor-specific" payload command. However, since vendor-specific commands are by definition not publicly documented and have the potential to corrupt the kernel's view of the dimm state, we provide a toggle to disable vendor-specific operations. Enabling undefined behavior is a policy decision that can be made by the platform owner and encourages firmware implementations to choose public over private command implementations. Based on an initial patch from Jerry Hoemann Cc: Jerry Hoemann <jerry.hoemann@hpe.com> Cc: Christoph Hellwig <hch@infradead.org> Signed-off-by: Dan Williams <dan.j.williams@intel.com>
2016-04-29 07:23:43 +08:00
/* https://github.com/HewlettPackard/hpe-nvm/blob/master/Documentation/ */
#define UUID_NFIT_DIMM_N_HPE1 "9002c334-acf3-4c0e-9642-a235f0d53bc6"
#define UUID_NFIT_DIMM_N_HPE2 "5008664b-b758-41a0-a03c-27c2f2d04f7e"
/* https://msdn.microsoft.com/library/windows/hardware/mt604741 */
#define UUID_NFIT_DIMM_N_MSFT "1ee68b36-d4bd-4a1a-9a16-4f8e53d46e05"
/* http://www.uefi.org/RFIC_LIST (see "Virtual NVDIMM 0x1901") */
#define UUID_NFIT_DIMM_N_HYPERV "5746c5f2-a9a2-4264-ad0e-e4ddc9e09e80"
#define ACPI_NFIT_MEM_FAILED_MASK (ACPI_NFIT_MEM_SAVE_FAILED \
| ACPI_NFIT_MEM_RESTORE_FAILED | ACPI_NFIT_MEM_FLUSH_FAILED \
| ACPI_NFIT_MEM_NOT_ARMED | ACPI_NFIT_MEM_MAP_FAILED)
#define NVDIMM_CMD_MAX 31
#define NVDIMM_STANDARD_CMDMASK \
(1 << ND_CMD_SMART | 1 << ND_CMD_SMART_THRESHOLD | 1 << ND_CMD_DIMM_FLAGS \
| 1 << ND_CMD_GET_CONFIG_SIZE | 1 << ND_CMD_GET_CONFIG_DATA \
| 1 << ND_CMD_SET_CONFIG_DATA | 1 << ND_CMD_VENDOR_EFFECT_LOG_SIZE \
| 1 << ND_CMD_VENDOR_EFFECT_LOG | 1 << ND_CMD_VENDOR)
/*
* Command numbers that the kernel needs to know about to handle
* non-default DSM revision ids
*/
enum nvdimm_family_cmds {
NVDIMM_INTEL_LATCH_SHUTDOWN = 10,
NVDIMM_INTEL_GET_MODES = 11,
NVDIMM_INTEL_GET_FWINFO = 12,
NVDIMM_INTEL_START_FWUPDATE = 13,
NVDIMM_INTEL_SEND_FWUPDATE = 14,
NVDIMM_INTEL_FINISH_FWUPDATE = 15,
NVDIMM_INTEL_QUERY_FWUPDATE = 16,
NVDIMM_INTEL_SET_THRESHOLD = 17,
NVDIMM_INTEL_INJECT_ERROR = 18,
NVDIMM_INTEL_GET_SECURITY_STATE = 19,
NVDIMM_INTEL_SET_PASSPHRASE = 20,
NVDIMM_INTEL_DISABLE_PASSPHRASE = 21,
NVDIMM_INTEL_UNLOCK_UNIT = 22,
NVDIMM_INTEL_FREEZE_LOCK = 23,
NVDIMM_INTEL_SECURE_ERASE = 24,
NVDIMM_INTEL_OVERWRITE = 25,
NVDIMM_INTEL_QUERY_OVERWRITE = 26,
NVDIMM_INTEL_SET_MASTER_PASSPHRASE = 27,
NVDIMM_INTEL_MASTER_SECURE_ERASE = 28,
NVDIMM_INTEL_FW_ACTIVATE_DIMMINFO = 29,
NVDIMM_INTEL_FW_ACTIVATE_ARM = 30,
};
enum nvdimm_bus_family_cmds {
NVDIMM_BUS_INTEL_FW_ACTIVATE_BUSINFO = 1,
NVDIMM_BUS_INTEL_FW_ACTIVATE = 2,
};
#define NVDIMM_INTEL_SECURITY_CMDMASK \
(1 << NVDIMM_INTEL_GET_SECURITY_STATE | 1 << NVDIMM_INTEL_SET_PASSPHRASE \
| 1 << NVDIMM_INTEL_DISABLE_PASSPHRASE | 1 << NVDIMM_INTEL_UNLOCK_UNIT \
| 1 << NVDIMM_INTEL_FREEZE_LOCK | 1 << NVDIMM_INTEL_SECURE_ERASE \
| 1 << NVDIMM_INTEL_OVERWRITE | 1 << NVDIMM_INTEL_QUERY_OVERWRITE \
| 1 << NVDIMM_INTEL_SET_MASTER_PASSPHRASE \
| 1 << NVDIMM_INTEL_MASTER_SECURE_ERASE)
#define NVDIMM_INTEL_FW_ACTIVATE_CMDMASK \
(1 << NVDIMM_INTEL_FW_ACTIVATE_DIMMINFO | 1 << NVDIMM_INTEL_FW_ACTIVATE_ARM)
#define NVDIMM_BUS_INTEL_FW_ACTIVATE_CMDMASK \
(1 << NVDIMM_BUS_INTEL_FW_ACTIVATE_BUSINFO | 1 << NVDIMM_BUS_INTEL_FW_ACTIVATE)
#define NVDIMM_INTEL_CMDMASK \
(NVDIMM_STANDARD_CMDMASK | 1 << NVDIMM_INTEL_GET_MODES \
| 1 << NVDIMM_INTEL_GET_FWINFO | 1 << NVDIMM_INTEL_START_FWUPDATE \
| 1 << NVDIMM_INTEL_SEND_FWUPDATE | 1 << NVDIMM_INTEL_FINISH_FWUPDATE \
| 1 << NVDIMM_INTEL_QUERY_FWUPDATE | 1 << NVDIMM_INTEL_SET_THRESHOLD \
| 1 << NVDIMM_INTEL_INJECT_ERROR | 1 << NVDIMM_INTEL_LATCH_SHUTDOWN \
| NVDIMM_INTEL_SECURITY_CMDMASK | NVDIMM_INTEL_FW_ACTIVATE_CMDMASK)
#define NVDIMM_INTEL_DENY_CMDMASK \
(NVDIMM_INTEL_SECURITY_CMDMASK | NVDIMM_INTEL_FW_ACTIVATE_CMDMASK)
enum nfit_uuids {
nfit, libnvdimm: limited/whitelisted dimm command marshaling mechanism There are currently 4 known similar but incompatible definitions of the command sets that can be sent to an NVDIMM through ACPI. It is also clear that future platform generations (ACPI or not) will continue to revise and extend the DIMM command set as new devices and use cases arrive. It is obviously untenable to continue to proliferate divergence of these command definitions, and to that end a standardization process has begun to provide for a unified specification. However, that leaves a problem about what to do with this first generation where vendors are already shipping divergence. The Linux kernel can support these initial diverged platforms without giving platform-firmware free reign to continue to diverge and compound kernel maintenance overhead. The kernel implementation can encourage standardization in two ways: 1/ Require that any function code that userspace wants to send be explicitly white-listed in the implementation. For ACPI this means function codes marked as supported by acpi_check_dsm() may only be invoked if they appear in the white-list. A function must be publicly documented before it is added to the white-list. 2/ The above restrictions can be trivially bypassed by using the "vendor-specific" payload command. However, since vendor-specific commands are by definition not publicly documented and have the potential to corrupt the kernel's view of the dimm state, we provide a toggle to disable vendor-specific operations. Enabling undefined behavior is a policy decision that can be made by the platform owner and encourages firmware implementations to choose public over private command implementations. Based on an initial patch from Jerry Hoemann Cc: Jerry Hoemann <jerry.hoemann@hpe.com> Cc: Christoph Hellwig <hch@infradead.org> Signed-off-by: Dan Williams <dan.j.williams@intel.com>
2016-04-29 07:23:43 +08:00
/* for simplicity alias the uuid index with the family id */
NFIT_DEV_DIMM = NVDIMM_FAMILY_INTEL,
NFIT_DEV_DIMM_N_HPE1 = NVDIMM_FAMILY_HPE1,
NFIT_DEV_DIMM_N_HPE2 = NVDIMM_FAMILY_HPE2,
NFIT_DEV_DIMM_N_MSFT = NVDIMM_FAMILY_MSFT,
NFIT_DEV_DIMM_N_HYPERV = NVDIMM_FAMILY_HYPERV,
/*
* to_nfit_bus_uuid() expects to translate bus uuid family ids
* to a UUID index using NVDIMM_FAMILY_MAX as an offset
*/
NFIT_BUS_INTEL = NVDIMM_FAMILY_MAX + NVDIMM_BUS_FAMILY_INTEL,
NFIT_SPA_VOLATILE,
NFIT_SPA_PM,
NFIT_SPA_DCR,
NFIT_SPA_BDW,
NFIT_SPA_VDISK,
NFIT_SPA_VCD,
NFIT_SPA_PDISK,
NFIT_SPA_PCD,
NFIT_DEV_BUS,
NFIT_UUID_MAX,
};
/*
* Region format interface codes are stored with the interface as the
* LSB and the function as the MSB.
*/
#define NFIT_FIC_BYTE cpu_to_le16(0x101) /* byte-addressable energy backed */
#define NFIT_FIC_BLK cpu_to_le16(0x201) /* block-addressable non-energy backed */
#define NFIT_FIC_BYTEN cpu_to_le16(0x301) /* byte-addressable non-energy backed */
enum {
NFIT_BLK_READ_FLUSH = 1,
NFIT_BLK_DCR_LATCH = 2,
NFIT_ARS_STATUS_DONE = 0,
NFIT_ARS_STATUS_BUSY = 1 << 16,
NFIT_ARS_STATUS_NONE = 2 << 16,
NFIT_ARS_STATUS_INTR = 3 << 16,
NFIT_ARS_START_BUSY = 6,
NFIT_ARS_CAP_NONE = 1,
NFIT_ARS_F_OVERFLOW = 1,
NFIT_ARS_TIMEOUT = 90,
};
enum nfit_root_notifiers {
NFIT_NOTIFY_UPDATE = 0x80,
NFIT_NOTIFY_UC_MEMORY_ERROR = 0x81,
};
enum nfit_dimm_notifiers {
NFIT_NOTIFY_DIMM_HEALTH = 0x81,
};
enum nfit_ars_state {
acpi, nfit: Fix Address Range Scrub completion tracking The Address Range Scrub implementation tried to skip running scrubs against ranges that were already scrubbed by the BIOS. Unfortunately that support also resulted in early scrub completions as evidenced by this debug output from nfit_test: nd_region region9: ARS: range 1 short complete nd_region region3: ARS: range 1 short complete nd_region region4: ARS: range 2 ARS start (0) nd_region region4: ARS: range 2 short complete ...i.e. completions without any indications that the scrub was started. This state of affairs was hard to see in the code due to the proliferation of state bits and mistakenly trying to track done state per-range when the completion is a global property of the bus. So, kill the four ARS state bits (ARS_REQ, ARS_REQ_REDO, ARS_DONE, and ARS_SHORT), and replace them with just 2 request flags ARS_REQ_SHORT and ARS_REQ_LONG. The implementation will still complete and reap the results of BIOS initiated ARS, but it will not attempt to use that information to affect the completion status of scrubbing the ranges from a Linux perspective. Instead, try to synchronously run a short ARS per range at init time and schedule a long scrub in the background. If ARS is busy with an ARS request, schedule both a short and a long scrub for when ARS returns to idle. This logic also satisfies the intent of what ARS_REQ_REDO was trying to achieve. The new rule is that the REQ flag stays set until the next successful ars_start() for that range. With the new policy that the REQ flags are not cleared until the next start, the implementation no longer loses requests as can be seen from the following log: nd_region region3: ARS: range 1 ARS start short (0) nd_region region9: ARS: range 1 ARS start short (0) nd_region region3: ARS: range 1 complete nd_region region4: ARS: range 2 ARS start short (0) nd_region region9: ARS: range 1 complete nd_region region9: ARS: range 1 ARS start long (0) nd_region region4: ARS: range 2 complete nd_region region3: ARS: range 1 ARS start long (0) nd_region region9: ARS: range 1 complete nd_region region3: ARS: range 1 complete nd_region region4: ARS: range 2 ARS start long (0) nd_region region4: ARS: range 2 complete ...note that the nfit_test emulated driver provides 2 buses, that is why some of the range indices are duplicated. Notice that each range now successfully completes a short and long scrub. Cc: <stable@vger.kernel.org> Fixes: 14c73f997a5e ("nfit, address-range-scrub: introduce nfit_spa->ars_state") Fixes: cc3d3458d46f ("acpi/nfit: queue issuing of ars when an uc error...") Reported-by: Jacek Zloch <jacek.zloch@intel.com> Reported-by: Krzysztof Rusocki <krzysztof.rusocki@intel.com> Reviewed-by: Dave Jiang <dave.jiang@intel.com> Signed-off-by: Dan Williams <dan.j.williams@intel.com>
2018-10-14 11:32:17 +08:00
ARS_REQ_SHORT,
ARS_REQ_LONG,
ARS_FAILED,
};
struct nfit_spa {
struct list_head list;
struct nd_region *nd_region;
unsigned long ars_state;
u32 clear_err_unit;
u32 max_ars;
struct acpi_nfit_system_address spa[];
};
struct nfit_dcr {
struct list_head list;
struct acpi_nfit_control_region dcr[];
};
struct nfit_bdw {
struct list_head list;
struct acpi_nfit_data_region bdw[];
};
struct nfit_idt {
struct list_head list;
struct acpi_nfit_interleave idt[];
};
struct nfit_flush {
struct list_head list;
struct acpi_nfit_flush_address flush[];
};
struct nfit_memdev {
struct list_head list;
struct acpi_nfit_memory_map memdev[];
};
enum nfit_mem_flags {
NFIT_MEM_LSR,
NFIT_MEM_LSW,
NFIT_MEM_DIRTY,
NFIT_MEM_DIRTY_COUNT,
};
#define NFIT_DIMM_ID_LEN 22
/* assembled tables for a given dimm/memory-device */
struct nfit_mem {
struct nvdimm *nvdimm;
struct acpi_nfit_memory_map *memdev_dcr;
struct acpi_nfit_memory_map *memdev_pmem;
struct acpi_nfit_control_region *dcr;
struct acpi_nfit_system_address *spa_dcr;
struct acpi_nfit_interleave *idt_dcr;
struct kernfs_node *flags_attr;
struct nfit_flush *nfit_flush;
struct list_head list;
2015-06-09 02:27:06 +08:00
struct acpi_device *adev;
struct acpi_nfit_desc *acpi_desc;
enum nvdimm_fwa_state fwa_state;
enum nvdimm_fwa_result fwa_result;
int fwa_count;
char id[NFIT_DIMM_ID_LEN+1];
struct resource *flush_wpq;
2015-06-09 02:27:06 +08:00
unsigned long dsm_mask;
unsigned long flags;
u32 dirty_shutdown;
nfit, libnvdimm: limited/whitelisted dimm command marshaling mechanism There are currently 4 known similar but incompatible definitions of the command sets that can be sent to an NVDIMM through ACPI. It is also clear that future platform generations (ACPI or not) will continue to revise and extend the DIMM command set as new devices and use cases arrive. It is obviously untenable to continue to proliferate divergence of these command definitions, and to that end a standardization process has begun to provide for a unified specification. However, that leaves a problem about what to do with this first generation where vendors are already shipping divergence. The Linux kernel can support these initial diverged platforms without giving platform-firmware free reign to continue to diverge and compound kernel maintenance overhead. The kernel implementation can encourage standardization in two ways: 1/ Require that any function code that userspace wants to send be explicitly white-listed in the implementation. For ACPI this means function codes marked as supported by acpi_check_dsm() may only be invoked if they appear in the white-list. A function must be publicly documented before it is added to the white-list. 2/ The above restrictions can be trivially bypassed by using the "vendor-specific" payload command. However, since vendor-specific commands are by definition not publicly documented and have the potential to corrupt the kernel's view of the dimm state, we provide a toggle to disable vendor-specific operations. Enabling undefined behavior is a policy decision that can be made by the platform owner and encourages firmware implementations to choose public over private command implementations. Based on an initial patch from Jerry Hoemann Cc: Jerry Hoemann <jerry.hoemann@hpe.com> Cc: Christoph Hellwig <hch@infradead.org> Signed-off-by: Dan Williams <dan.j.williams@intel.com>
2016-04-29 07:23:43 +08:00
int family;
};
enum scrub_flags {
ARS_BUSY,
ARS_CANCEL,
ARS_VALID,
ARS_POLL,
};
struct acpi_nfit_desc {
struct nvdimm_bus_descriptor nd_desc;
struct acpi_table_header acpi_header;
struct mutex init_mutex;
struct list_head memdevs;
struct list_head flushes;
struct list_head dimms;
struct list_head spas;
struct list_head dcrs;
struct list_head bdws;
struct list_head idts;
struct nvdimm_bus *nvdimm_bus;
struct device *dev;
struct nd_cmd_ars_status *ars_status;
acpi, nfit: Fix Address Range Scrub completion tracking The Address Range Scrub implementation tried to skip running scrubs against ranges that were already scrubbed by the BIOS. Unfortunately that support also resulted in early scrub completions as evidenced by this debug output from nfit_test: nd_region region9: ARS: range 1 short complete nd_region region3: ARS: range 1 short complete nd_region region4: ARS: range 2 ARS start (0) nd_region region4: ARS: range 2 short complete ...i.e. completions without any indications that the scrub was started. This state of affairs was hard to see in the code due to the proliferation of state bits and mistakenly trying to track done state per-range when the completion is a global property of the bus. So, kill the four ARS state bits (ARS_REQ, ARS_REQ_REDO, ARS_DONE, and ARS_SHORT), and replace them with just 2 request flags ARS_REQ_SHORT and ARS_REQ_LONG. The implementation will still complete and reap the results of BIOS initiated ARS, but it will not attempt to use that information to affect the completion status of scrubbing the ranges from a Linux perspective. Instead, try to synchronously run a short ARS per range at init time and schedule a long scrub in the background. If ARS is busy with an ARS request, schedule both a short and a long scrub for when ARS returns to idle. This logic also satisfies the intent of what ARS_REQ_REDO was trying to achieve. The new rule is that the REQ flag stays set until the next successful ars_start() for that range. With the new policy that the REQ flags are not cleared until the next start, the implementation no longer loses requests as can be seen from the following log: nd_region region3: ARS: range 1 ARS start short (0) nd_region region9: ARS: range 1 ARS start short (0) nd_region region3: ARS: range 1 complete nd_region region4: ARS: range 2 ARS start short (0) nd_region region9: ARS: range 1 complete nd_region region9: ARS: range 1 ARS start long (0) nd_region region4: ARS: range 2 complete nd_region region3: ARS: range 1 ARS start long (0) nd_region region9: ARS: range 1 complete nd_region region3: ARS: range 1 complete nd_region region4: ARS: range 2 ARS start long (0) nd_region region4: ARS: range 2 complete ...note that the nfit_test emulated driver provides 2 buses, that is why some of the range indices are duplicated. Notice that each range now successfully completes a short and long scrub. Cc: <stable@vger.kernel.org> Fixes: 14c73f997a5e ("nfit, address-range-scrub: introduce nfit_spa->ars_state") Fixes: cc3d3458d46f ("acpi/nfit: queue issuing of ars when an uc error...") Reported-by: Jacek Zloch <jacek.zloch@intel.com> Reported-by: Krzysztof Rusocki <krzysztof.rusocki@intel.com> Reviewed-by: Dave Jiang <dave.jiang@intel.com> Signed-off-by: Dan Williams <dan.j.williams@intel.com>
2018-10-14 11:32:17 +08:00
struct nfit_spa *scrub_spa;
struct delayed_work dwork;
struct list_head list;
struct kernfs_node *scrub_count_state;
unsigned int max_ars;
unsigned int scrub_count;
unsigned int scrub_mode;
unsigned long scrub_flags;
unsigned long dimm_cmd_force_en;
unsigned long bus_cmd_force_en;
unsigned long bus_dsm_mask;
unsigned long family_dsm_mask[NVDIMM_BUS_FAMILY_MAX + 1];
unsigned int platform_cap;
unsigned int scrub_tmo;
enum nvdimm_fwa_state fwa_state;
enum nvdimm_fwa_capability fwa_cap;
int fwa_count;
bool fwa_noidle;
bool fwa_nosuspend;
};
enum scrub_mode {
HW_ERROR_SCRUB_OFF,
HW_ERROR_SCRUB_ON,
};
enum nd_blk_mmio_selector {
BDW,
DCR,
};
nd_blk: change aperture mapping from WC to WB This should result in a pretty sizeable performance gain for reads. For rough comparison I did some simple read testing using PMEM to compare reads of write combining (WC) mappings vs write-back (WB). This was done on a random lab machine. PMEM reads from a write combining mapping: # dd of=/dev/null if=/dev/pmem0 bs=4096 count=100000 100000+0 records in 100000+0 records out 409600000 bytes (410 MB) copied, 9.2855 s, 44.1 MB/s PMEM reads from a write-back mapping: # dd of=/dev/null if=/dev/pmem0 bs=4096 count=1000000 1000000+0 records in 1000000+0 records out 4096000000 bytes (4.1 GB) copied, 3.44034 s, 1.2 GB/s To be able to safely support a write-back aperture I needed to add support for the "read flush" _DSM flag, as outlined in the DSM spec: http://pmem.io/documents/NVDIMM_DSM_Interface_Example.pdf This flag tells the ND BLK driver that it needs to flush the cache lines associated with the aperture after the aperture is moved but before any new data is read. This ensures that any stale cache lines from the previous contents of the aperture will be discarded from the processor cache, and the new data will be read properly from the DIMM. We know that the cache lines are clean and will be discarded without any writeback because either a) the previous aperture operation was a read, and we never modified the contents of the aperture, or b) the previous aperture operation was a write and we must have written back the dirtied contents of the aperture to the DIMM before the I/O was completed. In order to add support for the "read flush" flag I needed to add a generic routine to invalidate cache lines, mmio_flush_range(). This is protected by the ARCH_HAS_MMIO_FLUSH Kconfig variable, and is currently only supported on x86. Signed-off-by: Ross Zwisler <ross.zwisler@linux.intel.com> Signed-off-by: Dan Williams <dan.j.williams@intel.com>
2015-08-28 03:14:20 +08:00
struct nd_blk_addr {
union {
void __iomem *base;
void *aperture;
nd_blk: change aperture mapping from WC to WB This should result in a pretty sizeable performance gain for reads. For rough comparison I did some simple read testing using PMEM to compare reads of write combining (WC) mappings vs write-back (WB). This was done on a random lab machine. PMEM reads from a write combining mapping: # dd of=/dev/null if=/dev/pmem0 bs=4096 count=100000 100000+0 records in 100000+0 records out 409600000 bytes (410 MB) copied, 9.2855 s, 44.1 MB/s PMEM reads from a write-back mapping: # dd of=/dev/null if=/dev/pmem0 bs=4096 count=1000000 1000000+0 records in 1000000+0 records out 4096000000 bytes (4.1 GB) copied, 3.44034 s, 1.2 GB/s To be able to safely support a write-back aperture I needed to add support for the "read flush" _DSM flag, as outlined in the DSM spec: http://pmem.io/documents/NVDIMM_DSM_Interface_Example.pdf This flag tells the ND BLK driver that it needs to flush the cache lines associated with the aperture after the aperture is moved but before any new data is read. This ensures that any stale cache lines from the previous contents of the aperture will be discarded from the processor cache, and the new data will be read properly from the DIMM. We know that the cache lines are clean and will be discarded without any writeback because either a) the previous aperture operation was a read, and we never modified the contents of the aperture, or b) the previous aperture operation was a write and we must have written back the dirtied contents of the aperture to the DIMM before the I/O was completed. In order to add support for the "read flush" flag I needed to add a generic routine to invalidate cache lines, mmio_flush_range(). This is protected by the ARCH_HAS_MMIO_FLUSH Kconfig variable, and is currently only supported on x86. Signed-off-by: Ross Zwisler <ross.zwisler@linux.intel.com> Signed-off-by: Dan Williams <dan.j.williams@intel.com>
2015-08-28 03:14:20 +08:00
};
};
struct nfit_blk {
struct nfit_blk_mmio {
nd_blk: change aperture mapping from WC to WB This should result in a pretty sizeable performance gain for reads. For rough comparison I did some simple read testing using PMEM to compare reads of write combining (WC) mappings vs write-back (WB). This was done on a random lab machine. PMEM reads from a write combining mapping: # dd of=/dev/null if=/dev/pmem0 bs=4096 count=100000 100000+0 records in 100000+0 records out 409600000 bytes (410 MB) copied, 9.2855 s, 44.1 MB/s PMEM reads from a write-back mapping: # dd of=/dev/null if=/dev/pmem0 bs=4096 count=1000000 1000000+0 records in 1000000+0 records out 4096000000 bytes (4.1 GB) copied, 3.44034 s, 1.2 GB/s To be able to safely support a write-back aperture I needed to add support for the "read flush" _DSM flag, as outlined in the DSM spec: http://pmem.io/documents/NVDIMM_DSM_Interface_Example.pdf This flag tells the ND BLK driver that it needs to flush the cache lines associated with the aperture after the aperture is moved but before any new data is read. This ensures that any stale cache lines from the previous contents of the aperture will be discarded from the processor cache, and the new data will be read properly from the DIMM. We know that the cache lines are clean and will be discarded without any writeback because either a) the previous aperture operation was a read, and we never modified the contents of the aperture, or b) the previous aperture operation was a write and we must have written back the dirtied contents of the aperture to the DIMM before the I/O was completed. In order to add support for the "read flush" flag I needed to add a generic routine to invalidate cache lines, mmio_flush_range(). This is protected by the ARCH_HAS_MMIO_FLUSH Kconfig variable, and is currently only supported on x86. Signed-off-by: Ross Zwisler <ross.zwisler@linux.intel.com> Signed-off-by: Dan Williams <dan.j.williams@intel.com>
2015-08-28 03:14:20 +08:00
struct nd_blk_addr addr;
u64 size;
u64 base_offset;
u32 line_size;
u32 num_lines;
u32 table_size;
struct acpi_nfit_interleave *idt;
struct acpi_nfit_system_address *spa;
} mmio[2];
struct nd_region *nd_region;
u64 bdw_offset; /* post interleave offset */
u64 stat_offset;
u64 cmd_offset;
u32 dimm_flags;
};
extern struct list_head acpi_descs;
extern struct mutex acpi_desc_lock;
acpi, nfit: Fix Address Range Scrub completion tracking The Address Range Scrub implementation tried to skip running scrubs against ranges that were already scrubbed by the BIOS. Unfortunately that support also resulted in early scrub completions as evidenced by this debug output from nfit_test: nd_region region9: ARS: range 1 short complete nd_region region3: ARS: range 1 short complete nd_region region4: ARS: range 2 ARS start (0) nd_region region4: ARS: range 2 short complete ...i.e. completions without any indications that the scrub was started. This state of affairs was hard to see in the code due to the proliferation of state bits and mistakenly trying to track done state per-range when the completion is a global property of the bus. So, kill the four ARS state bits (ARS_REQ, ARS_REQ_REDO, ARS_DONE, and ARS_SHORT), and replace them with just 2 request flags ARS_REQ_SHORT and ARS_REQ_LONG. The implementation will still complete and reap the results of BIOS initiated ARS, but it will not attempt to use that information to affect the completion status of scrubbing the ranges from a Linux perspective. Instead, try to synchronously run a short ARS per range at init time and schedule a long scrub in the background. If ARS is busy with an ARS request, schedule both a short and a long scrub for when ARS returns to idle. This logic also satisfies the intent of what ARS_REQ_REDO was trying to achieve. The new rule is that the REQ flag stays set until the next successful ars_start() for that range. With the new policy that the REQ flags are not cleared until the next start, the implementation no longer loses requests as can be seen from the following log: nd_region region3: ARS: range 1 ARS start short (0) nd_region region9: ARS: range 1 ARS start short (0) nd_region region3: ARS: range 1 complete nd_region region4: ARS: range 2 ARS start short (0) nd_region region9: ARS: range 1 complete nd_region region9: ARS: range 1 ARS start long (0) nd_region region4: ARS: range 2 complete nd_region region3: ARS: range 1 ARS start long (0) nd_region region9: ARS: range 1 complete nd_region region3: ARS: range 1 complete nd_region region4: ARS: range 2 ARS start long (0) nd_region region4: ARS: range 2 complete ...note that the nfit_test emulated driver provides 2 buses, that is why some of the range indices are duplicated. Notice that each range now successfully completes a short and long scrub. Cc: <stable@vger.kernel.org> Fixes: 14c73f997a5e ("nfit, address-range-scrub: introduce nfit_spa->ars_state") Fixes: cc3d3458d46f ("acpi/nfit: queue issuing of ars when an uc error...") Reported-by: Jacek Zloch <jacek.zloch@intel.com> Reported-by: Krzysztof Rusocki <krzysztof.rusocki@intel.com> Reviewed-by: Dave Jiang <dave.jiang@intel.com> Signed-off-by: Dan Williams <dan.j.williams@intel.com>
2018-10-14 11:32:17 +08:00
int acpi_nfit_ars_rescan(struct acpi_nfit_desc *acpi_desc,
enum nfit_ars_state req_type);
#ifdef CONFIG_X86_MCE
void nfit_mce_register(void);
void nfit_mce_unregister(void);
#else
static inline void nfit_mce_register(void)
{
}
static inline void nfit_mce_unregister(void)
{
}
#endif
int nfit_spa_type(struct acpi_nfit_system_address *spa);
static inline struct acpi_nfit_memory_map *__to_nfit_memdev(
struct nfit_mem *nfit_mem)
{
if (nfit_mem->memdev_dcr)
return nfit_mem->memdev_dcr;
return nfit_mem->memdev_pmem;
}
static inline struct acpi_nfit_desc *to_acpi_desc(
struct nvdimm_bus_descriptor *nd_desc)
{
return container_of(nd_desc, struct acpi_nfit_desc, nd_desc);
}
tools/testing/nvdimm: libnvdimm unit test infrastructure 'libnvdimm' is the first driver sub-system in the kernel to implement mocking for unit test coverage. The nfit_test module gets built as an external module and arranges for external module replacements of nfit, libnvdimm, nd_pmem, and nd_blk. These replacements use the linker --wrap option to redirect calls to ioremap() + request_mem_region() to custom defined unit test resources. The end result is a fully functional nvdimm_bus, as far as userspace is concerned, but with the capability to perform otherwise destructive tests on emulated resources. Q: Why not use QEMU for this emulation? QEMU is not suitable for unit testing. QEMU's role is to faithfully emulate the platform. A unit test's role is to unfaithfully implement the platform with the goal of triggering bugs in the corners of the sub-system implementation. As bugs are discovered in platforms, or the sub-system itself, the unit tests are extended to backstop a fix with a reproducer unit test. Another problem with QEMU is that it would require coordination of 3 software projects instead of 2 (kernel + libndctl [1]) to maintain and execute the tests. The chances for bit rot and the difficulty of getting the tests running goes up non-linearly the more components involved. Q: Why submit this to the kernel tree instead of external modules in libndctl? Simple, to alleviate the same risk that out-of-tree external modules face. Updates to drivers/nvdimm/ can be immediately evaluated to see if they have any impact on tools/testing/nvdimm/. Q: What are the negative implications of merging this? It is a unique maintenance burden because the purpose of mocking an interface to enable a unit test is to purposefully short circuit the semantics of a routine to enable testing. For example __wrap_ioremap_cache() fakes the pmem driver into "ioremap()'ing" a test resource buffer allocated by dma_alloc_coherent(). The future maintenance burden hits when someone changes the semantics of ioremap_cache() and wonders what the implications are for the unit test. [1]: https://github.com/pmem/ndctl Cc: <linux-acpi@vger.kernel.org> Cc: Lv Zheng <lv.zheng@intel.com> Cc: Robert Moore <robert.moore@intel.com> Cc: Rafael J. Wysocki <rafael.j.wysocki@intel.com> Cc: Christoph Hellwig <hch@lst.de> Signed-off-by: Dan Williams <dan.j.williams@intel.com>
2015-06-18 05:23:32 +08:00
const guid_t *to_nfit_uuid(enum nfit_uuids id);
int acpi_nfit_init(struct acpi_nfit_desc *acpi_desc, void *nfit, acpi_size sz);
void acpi_nfit_shutdown(void *data);
void __acpi_nfit_notify(struct device *dev, acpi_handle handle, u32 event);
void __acpi_nvdimm_notify(struct device *dev, u32 event);
int acpi_nfit_ctl(struct nvdimm_bus_descriptor *nd_desc, struct nvdimm *nvdimm,
unsigned int cmd, void *buf, unsigned int buf_len, int *cmd_rc);
void acpi_nfit_desc_init(struct acpi_nfit_desc *acpi_desc, struct device *dev);
bool intel_fwa_supported(struct nvdimm_bus *nvdimm_bus);
extern struct device_attribute dev_attr_firmware_activate_noidle;
void nfit_intel_shutdown_status(struct nfit_mem *nfit_mem);
#endif /* __NFIT_H__ */