Linux kernel mirror (for testing)
git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git
tjh.dev
kernel
os
linux
1# SPDX-License-Identifier: GPL-2.0-only
2menuconfig LIBNVDIMM
3 tristate "NVDIMM (Non-Volatile Memory Device) Support"
4 depends on PHYS_ADDR_T_64BIT
5 depends on HAS_IOMEM
6 depends on BLK_DEV
7 select MEMREGION
8 help
9 Generic support for non-volatile memory devices including
10 ACPI-6-NFIT defined resources. On platforms that define an
11 NFIT, or otherwise can discover NVDIMM resources, a libnvdimm
12 bus is registered to advertise PMEM (persistent memory)
13 namespaces (/dev/pmemX). A PMEM namespace refers to a
14 memory resource that may span multiple DIMMs and support DAX
15 (see CONFIG_DAX).
16
17if LIBNVDIMM
18
19config BLK_DEV_PMEM
20 tristate "PMEM: Persistent memory block device support"
21 default LIBNVDIMM
22 select DAX
23 select ND_BTT if BTT
24 select ND_PFN if NVDIMM_PFN
25 help
26 Memory ranges for PMEM are described by either an NFIT
27 (NVDIMM Firmware Interface Table, see CONFIG_ACPI_NFIT), a
28 non-standard OEM-specific E820 memory type (type-12, see
29 CONFIG_X86_PMEM_LEGACY), or it is manually specified by the
30 'memmap=nn[KMG]!ss[KMG]' kernel command line (see
31 Documentation/admin-guide/kernel-parameters.rst). This driver converts
32 these persistent memory ranges into block devices that are
33 capable of DAX (direct-access) file system mappings. See
34 Documentation/driver-api/nvdimm/nvdimm.rst for more details.
35
36 Say Y if you want to use an NVDIMM
37
38config ND_CLAIM
39 bool
40
41config ND_BTT
42 tristate
43
44config BTT
45 bool "BTT: Block Translation Table (atomic sector updates)"
46 default y if LIBNVDIMM
47 select ND_CLAIM
48 help
49 The Block Translation Table (BTT) provides atomic sector
50 update semantics for persistent memory devices, so that
51 applications that rely on sector writes not being torn (a
52 guarantee that typical disks provide) can continue to do so.
53 The BTT manifests itself as an alternate personality for an
54 NVDIMM namespace, i.e. a namespace can be in raw mode pmemX,
55 or 'sectored' mode.
56
57 Select Y if unsure
58
59config ND_PFN
60 tristate
61
62config NVDIMM_PFN
63 bool "PFN: Map persistent (device) memory"
64 default LIBNVDIMM
65 depends on ZONE_DEVICE
66 select ND_CLAIM
67 help
68 Map persistent memory, i.e. advertise it to the memory
69 management sub-system. By default persistent memory does
70 not support direct I/O, RDMA, or any other usage that
71 requires a 'struct page' to mediate an I/O request. This
72 driver allocates and initializes the infrastructure needed
73 to support those use cases.
74
75 Select Y if unsure
76
77config NVDIMM_DAX
78 bool "NVDIMM DAX: Raw access to persistent memory"
79 default LIBNVDIMM
80 depends on NVDIMM_PFN
81 help
82 Support raw device dax access to a persistent memory
83 namespace. For environments that want to hard partition
84 persistent memory, this capability provides a mechanism to
85 sub-divide a namespace into character devices that can only be
86 accessed via DAX (mmap(2)).
87
88 Select Y if unsure
89
90config OF_PMEM
91 tristate "Device-tree support for persistent memory regions"
92 depends on OF
93 default LIBNVDIMM
94 help
95 Allows regions of persistent memory to be described in the
96 device-tree.
97
98 Select Y if unsure.
99
100config RAMDAX
101 tristate "Support persistent memory interfaces on RAM carveouts"
102 depends on X86_PMEM_LEGACY || OF || COMPILE_TEST
103 default LIBNVDIMM
104 help
105 Allows creation of DAX devices on RAM carveouts.
106
107 Memory ranges that are manually specified by the
108 'memmap=nn[KMG]!ss[KMG]' kernel command line or defined by dummy
109 pmem-region device tree nodes would be managed by this driver as DIMM
110 devices with support for dynamic layout of namespaces.
111 The driver steals 128K in the end of the memmap range for the
112 namespace management. This allows supporting up to 509 namespaces
113 (see 'ndctl create-namespace --help').
114 The driver should be force bound to e820_pmem or pmem-region platform
115 devices using 'driver_override' device attribute.
116
117 Select N if unsure.
118
119config NVDIMM_KEYS
120 def_bool y
121 depends on ENCRYPTED_KEYS
122 depends on (LIBNVDIMM=ENCRYPTED_KEYS) || LIBNVDIMM=m
123
124config NVDIMM_KMSAN
125 bool
126 depends on KMSAN
127 help
128 KMSAN, and other memory debug facilities, increase the size of
129 'struct page' to contain extra metadata. This collides with
130 the NVDIMM capability to store a potentially
131 larger-than-"System RAM" size 'struct page' array in a
132 reservation of persistent memory rather than limited /
133 precious DRAM. However, that reservation needs to persist for
134 the life of the given NVDIMM namespace. If you are using KMSAN
135 to debug an issue unrelated to NVDIMMs or DAX then say N to this
136 option. Otherwise, say Y but understand that any namespaces
137 (with the page array stored pmem) created with this build of
138 the kernel will permanently reserve and strand excess
139 capacity compared to the CONFIG_KMSAN=n case.
140
141 Select N if unsure.
142
143config NVDIMM_TEST_BUILD
144 tristate "Build the unit test core"
145 depends on m
146 depends on COMPILE_TEST && X86_64
147 default m if COMPILE_TEST
148 help
149 Build the core of the unit test infrastructure. The result of
150 this build is non-functional for unit test execution, but it
151 otherwise helps catch build errors induced by changes to the
152 core devm_memremap_pages() implementation and other
153 infrastructure.
154
155config NVDIMM_SECURITY_TEST
156 bool "Enable NVDIMM security unit tests"
157 depends on NVDIMM_KEYS
158 help
159 The NVDIMM and CXL subsystems support unit testing of their device
160 security state machines. The NVDIMM_SECURITY_TEST option disables CPU
161 cache maintenance operations around events like secure erase and
162 overwrite. Also, when enabled, the NVDIMM subsystem core helps the unit
163 test implement a mock state machine.
164
165 Select N if unsure.
166
167endif