1 ============================================== 1 ===============================================
2 The irq_domain interrupt number mapping librar 2 The irq_domain interrupt number mapping library
3 ============================================== 3 ===============================================
4 4
5 The current design of the Linux kernel uses a 5 The current design of the Linux kernel uses a single large number
6 space where each separate IRQ source is assign 6 space where each separate IRQ source is assigned a different number.
7 This is simple when there is only one interrup 7 This is simple when there is only one interrupt controller, but in
8 systems with multiple interrupt controllers th 8 systems with multiple interrupt controllers the kernel must ensure
9 that each one gets assigned non-overlapping al 9 that each one gets assigned non-overlapping allocations of Linux
10 IRQ numbers. 10 IRQ numbers.
11 11
12 The number of interrupt controllers registered 12 The number of interrupt controllers registered as unique irqchips
13 show a rising tendency: for example subdrivers 13 show a rising tendency: for example subdrivers of different kinds
14 such as GPIO controllers avoid reimplementing 14 such as GPIO controllers avoid reimplementing identical callback
15 mechanisms as the IRQ core system by modelling 15 mechanisms as the IRQ core system by modelling their interrupt
16 handlers as irqchips, i.e. in effect cascading 16 handlers as irqchips, i.e. in effect cascading interrupt controllers.
17 17
18 Here the interrupt number loose all kind of co 18 Here the interrupt number loose all kind of correspondence to
19 hardware interrupt numbers: whereas in the pas 19 hardware interrupt numbers: whereas in the past, IRQ numbers could
20 be chosen so they matched the hardware IRQ lin 20 be chosen so they matched the hardware IRQ line into the root
21 interrupt controller (i.e. the component actua 21 interrupt controller (i.e. the component actually fireing the
22 interrupt line to the CPU) nowadays this numbe 22 interrupt line to the CPU) nowadays this number is just a number.
23 23
24 For this reason we need a mechanism to separat 24 For this reason we need a mechanism to separate controller-local
25 interrupt numbers, called hardware irq's, from 25 interrupt numbers, called hardware irq's, from Linux IRQ numbers.
26 26
27 The irq_alloc_desc*() and irq_free_desc*() API 27 The irq_alloc_desc*() and irq_free_desc*() APIs provide allocation of
28 irq numbers, but they don't provide any suppor 28 irq numbers, but they don't provide any support for reverse mapping of
29 the controller-local IRQ (hwirq) number into t 29 the controller-local IRQ (hwirq) number into the Linux IRQ number
30 space. 30 space.
31 31
32 The irq_domain library adds mapping between hw 32 The irq_domain library adds mapping between hwirq and IRQ numbers on
33 top of the irq_alloc_desc*() API. An irq_doma 33 top of the irq_alloc_desc*() API. An irq_domain to manage mapping is
34 preferred over interrupt controller drivers op 34 preferred over interrupt controller drivers open coding their own
35 reverse mapping scheme. 35 reverse mapping scheme.
36 36
37 irq_domain also implements translation from an 37 irq_domain also implements translation from an abstract irq_fwspec
38 structure to hwirq numbers (Device Tree and AC 38 structure to hwirq numbers (Device Tree and ACPI GSI so far), and can
39 be easily extended to support other IRQ topolo 39 be easily extended to support other IRQ topology data sources.
40 40
41 irq_domain usage 41 irq_domain usage
42 ================ 42 ================
43 43
44 An interrupt controller driver creates and reg 44 An interrupt controller driver creates and registers an irq_domain by
45 calling one of the irq_domain_add_*() or irq_d 45 calling one of the irq_domain_add_*() or irq_domain_create_*() functions
46 (each mapping method has a different allocator 46 (each mapping method has a different allocator function, more on that later).
47 The function will return a pointer to the irq_ 47 The function will return a pointer to the irq_domain on success. The caller
48 must provide the allocator function with an ir 48 must provide the allocator function with an irq_domain_ops structure.
49 49
50 In most cases, the irq_domain will begin empty 50 In most cases, the irq_domain will begin empty without any mappings
51 between hwirq and IRQ numbers. Mappings are a 51 between hwirq and IRQ numbers. Mappings are added to the irq_domain
52 by calling irq_create_mapping() which accepts 52 by calling irq_create_mapping() which accepts the irq_domain and a
53 hwirq number as arguments. If a mapping for t 53 hwirq number as arguments. If a mapping for the hwirq doesn't already
54 exist then it will allocate a new Linux irq_de 54 exist then it will allocate a new Linux irq_desc, associate it with
55 the hwirq, and call the .map() callback so the 55 the hwirq, and call the .map() callback so the driver can perform any
56 required hardware setup. 56 required hardware setup.
57 57
58 Once a mapping has been established, it can be 58 Once a mapping has been established, it can be retrieved or used via a
59 variety of methods: 59 variety of methods:
60 60
61 - irq_resolve_mapping() returns a pointer to t 61 - irq_resolve_mapping() returns a pointer to the irq_desc structure
62 for a given domain and hwirq number, and NUL 62 for a given domain and hwirq number, and NULL if there was no
63 mapping. 63 mapping.
64 - irq_find_mapping() returns a Linux IRQ numbe 64 - irq_find_mapping() returns a Linux IRQ number for a given domain and
65 hwirq number, and 0 if there was no mapping 65 hwirq number, and 0 if there was no mapping
66 - irq_linear_revmap() is now identical to irq_ 66 - irq_linear_revmap() is now identical to irq_find_mapping(), and is
67 deprecated 67 deprecated
68 - generic_handle_domain_irq() handles an inter 68 - generic_handle_domain_irq() handles an interrupt described by a
69 domain and a hwirq number 69 domain and a hwirq number
70 70
71 Note that irq domain lookups must happen in co 71 Note that irq domain lookups must happen in contexts that are
72 compatible with a RCU read-side critical secti 72 compatible with a RCU read-side critical section.
73 73
74 The irq_create_mapping() function must be call 74 The irq_create_mapping() function must be called *at least once*
75 before any call to irq_find_mapping(), lest th 75 before any call to irq_find_mapping(), lest the descriptor will not
76 be allocated. 76 be allocated.
77 77
78 If the driver has the Linux IRQ number or the 78 If the driver has the Linux IRQ number or the irq_data pointer, and
79 needs to know the associated hwirq number (suc 79 needs to know the associated hwirq number (such as in the irq_chip
80 callbacks) then it can be directly obtained fr 80 callbacks) then it can be directly obtained from irq_data->hwirq.
81 81
82 Types of irq_domain mappings 82 Types of irq_domain mappings
83 ============================ 83 ============================
84 84
85 There are several mechanisms available for rev 85 There are several mechanisms available for reverse mapping from hwirq
86 to Linux irq, and each mechanism uses a differ 86 to Linux irq, and each mechanism uses a different allocation function.
87 Which reverse map type should be used depends 87 Which reverse map type should be used depends on the use case. Each
88 of the reverse map types are described below: 88 of the reverse map types are described below:
89 89
90 Linear 90 Linear
91 ------ 91 ------
92 92
93 :: 93 ::
94 94
95 irq_domain_add_linear() 95 irq_domain_add_linear()
96 irq_domain_create_linear() 96 irq_domain_create_linear()
97 97
98 The linear reverse map maintains a fixed size 98 The linear reverse map maintains a fixed size table indexed by the
99 hwirq number. When a hwirq is mapped, an irq_ 99 hwirq number. When a hwirq is mapped, an irq_desc is allocated for
100 the hwirq, and the IRQ number is stored in the 100 the hwirq, and the IRQ number is stored in the table.
101 101
102 The Linear map is a good choice when the maxim 102 The Linear map is a good choice when the maximum number of hwirqs is
103 fixed and a relatively small number (~ < 256). 103 fixed and a relatively small number (~ < 256). The advantages of this
104 map are fixed time lookup for IRQ numbers, and 104 map are fixed time lookup for IRQ numbers, and irq_descs are only
105 allocated for in-use IRQs. The disadvantage i 105 allocated for in-use IRQs. The disadvantage is that the table must be
106 as large as the largest possible hwirq number. 106 as large as the largest possible hwirq number.
107 107
108 irq_domain_add_linear() and irq_domain_create_ 108 irq_domain_add_linear() and irq_domain_create_linear() are functionally
109 equivalent, except for the first argument is d 109 equivalent, except for the first argument is different - the former
110 accepts an Open Firmware specific 'struct devi 110 accepts an Open Firmware specific 'struct device_node', while the latter
111 accepts a more general abstraction 'struct fwn 111 accepts a more general abstraction 'struct fwnode_handle'.
112 112
113 The majority of drivers should use the linear 113 The majority of drivers should use the linear map.
114 114
115 Tree 115 Tree
116 ---- 116 ----
117 117
118 :: 118 ::
119 119
120 irq_domain_add_tree() 120 irq_domain_add_tree()
121 irq_domain_create_tree() 121 irq_domain_create_tree()
122 122
123 The irq_domain maintains a radix tree map from 123 The irq_domain maintains a radix tree map from hwirq numbers to Linux
124 IRQs. When an hwirq is mapped, an irq_desc is 124 IRQs. When an hwirq is mapped, an irq_desc is allocated and the
125 hwirq is used as the lookup key for the radix 125 hwirq is used as the lookup key for the radix tree.
126 126
127 The tree map is a good choice if the hwirq num 127 The tree map is a good choice if the hwirq number can be very large
128 since it doesn't need to allocate a table as l 128 since it doesn't need to allocate a table as large as the largest
129 hwirq number. The disadvantage is that hwirq 129 hwirq number. The disadvantage is that hwirq to IRQ number lookup is
130 dependent on how many entries are in the table 130 dependent on how many entries are in the table.
131 131
132 irq_domain_add_tree() and irq_domain_create_tr 132 irq_domain_add_tree() and irq_domain_create_tree() are functionally
133 equivalent, except for the first argument is d 133 equivalent, except for the first argument is different - the former
134 accepts an Open Firmware specific 'struct devi 134 accepts an Open Firmware specific 'struct device_node', while the latter
135 accepts a more general abstraction 'struct fwn 135 accepts a more general abstraction 'struct fwnode_handle'.
136 136
137 Very few drivers should need this mapping. 137 Very few drivers should need this mapping.
138 138
139 No Map 139 No Map
140 ------ 140 ------
141 141
142 :: 142 ::
143 143
144 irq_domain_add_nomap() 144 irq_domain_add_nomap()
145 145
146 The No Map mapping is to be used when the hwir 146 The No Map mapping is to be used when the hwirq number is
147 programmable in the hardware. In this case it 147 programmable in the hardware. In this case it is best to program the
148 Linux IRQ number into the hardware itself so t 148 Linux IRQ number into the hardware itself so that no mapping is
149 required. Calling irq_create_direct_mapping() 149 required. Calling irq_create_direct_mapping() will allocate a Linux
150 IRQ number and call the .map() callback so tha 150 IRQ number and call the .map() callback so that driver can program the
151 Linux IRQ number into the hardware. 151 Linux IRQ number into the hardware.
152 152
153 Most drivers cannot use this mapping, and it i 153 Most drivers cannot use this mapping, and it is now gated on the
154 CONFIG_IRQ_DOMAIN_NOMAP option. Please refrain 154 CONFIG_IRQ_DOMAIN_NOMAP option. Please refrain from introducing new
155 users of this API. 155 users of this API.
156 156
157 Legacy 157 Legacy
158 ------ 158 ------
159 159
160 :: 160 ::
161 161
162 irq_domain_add_simple() 162 irq_domain_add_simple()
163 irq_domain_add_legacy() 163 irq_domain_add_legacy()
164 irq_domain_create_simple() 164 irq_domain_create_simple()
165 irq_domain_create_legacy() 165 irq_domain_create_legacy()
166 166
167 The Legacy mapping is a special case for drive 167 The Legacy mapping is a special case for drivers that already have a
168 range of irq_descs allocated for the hwirqs. 168 range of irq_descs allocated for the hwirqs. It is used when the
169 driver cannot be immediately converted to use 169 driver cannot be immediately converted to use the linear mapping. For
170 example, many embedded system board support fi 170 example, many embedded system board support files use a set of #defines
171 for IRQ numbers that are passed to struct devi 171 for IRQ numbers that are passed to struct device registrations. In that
172 case the Linux IRQ numbers cannot be dynamical 172 case the Linux IRQ numbers cannot be dynamically assigned and the legacy
173 mapping should be used. 173 mapping should be used.
174 174
175 As the name implies, the \*_legacy() functions 175 As the name implies, the \*_legacy() functions are deprecated and only
176 exist to ease the support of ancient platforms 176 exist to ease the support of ancient platforms. No new users should be
177 added. Same goes for the \*_simple() functions 177 added. Same goes for the \*_simple() functions when their use results
178 in the legacy behaviour. 178 in the legacy behaviour.
179 179
180 The legacy map assumes a contiguous range of I 180 The legacy map assumes a contiguous range of IRQ numbers has already
181 been allocated for the controller and that the 181 been allocated for the controller and that the IRQ number can be
182 calculated by adding a fixed offset to the hwi 182 calculated by adding a fixed offset to the hwirq number, and
183 visa-versa. The disadvantage is that it requi 183 visa-versa. The disadvantage is that it requires the interrupt
184 controller to manage IRQ allocations and it re 184 controller to manage IRQ allocations and it requires an irq_desc to be
185 allocated for every hwirq, even if it is unuse 185 allocated for every hwirq, even if it is unused.
186 186
187 The legacy map should only be used if fixed IR 187 The legacy map should only be used if fixed IRQ mappings must be
188 supported. For example, ISA controllers would 188 supported. For example, ISA controllers would use the legacy map for
189 mapping Linux IRQs 0-15 so that existing ISA d 189 mapping Linux IRQs 0-15 so that existing ISA drivers get the correct IRQ
190 numbers. 190 numbers.
191 191
192 Most users of legacy mappings should use irq_d 192 Most users of legacy mappings should use irq_domain_add_simple() or
193 irq_domain_create_simple() which will use a le 193 irq_domain_create_simple() which will use a legacy domain only if an IRQ range
194 is supplied by the system and will otherwise u 194 is supplied by the system and will otherwise use a linear domain mapping.
195 The semantics of this call are such that if an 195 The semantics of this call are such that if an IRQ range is specified then
196 descriptors will be allocated on-the-fly for i 196 descriptors will be allocated on-the-fly for it, and if no range is
197 specified it will fall through to irq_domain_a 197 specified it will fall through to irq_domain_add_linear() or
198 irq_domain_create_linear() which means *no* ir 198 irq_domain_create_linear() which means *no* irq descriptors will be allocated.
199 199
200 A typical use case for simple domains is where 200 A typical use case for simple domains is where an irqchip provider
201 is supporting both dynamic and static IRQ assi 201 is supporting both dynamic and static IRQ assignments.
202 202
203 In order to avoid ending up in a situation whe 203 In order to avoid ending up in a situation where a linear domain is
204 used and no descriptor gets allocated it is ve 204 used and no descriptor gets allocated it is very important to make sure
205 that the driver using the simple domain call i 205 that the driver using the simple domain call irq_create_mapping()
206 before any irq_find_mapping() since the latter 206 before any irq_find_mapping() since the latter will actually work
207 for the static IRQ assignment case. 207 for the static IRQ assignment case.
208 208
209 irq_domain_add_simple() and irq_domain_create_ 209 irq_domain_add_simple() and irq_domain_create_simple() as well as
210 irq_domain_add_legacy() and irq_domain_create_ 210 irq_domain_add_legacy() and irq_domain_create_legacy() are functionally
211 equivalent, except for the first argument is d 211 equivalent, except for the first argument is different - the former
212 accepts an Open Firmware specific 'struct devi 212 accepts an Open Firmware specific 'struct device_node', while the latter
213 accepts a more general abstraction 'struct fwn 213 accepts a more general abstraction 'struct fwnode_handle'.
214 214
215 Hierarchy IRQ domain 215 Hierarchy IRQ domain
216 -------------------- 216 --------------------
217 217
218 On some architectures, there may be multiple i 218 On some architectures, there may be multiple interrupt controllers
219 involved in delivering an interrupt from the d 219 involved in delivering an interrupt from the device to the target CPU.
220 Let's look at a typical interrupt delivering p 220 Let's look at a typical interrupt delivering path on x86 platforms::
221 221
222 Device --> IOAPIC -> Interrupt remapping Con 222 Device --> IOAPIC -> Interrupt remapping Controller -> Local APIC -> CPU
223 223
224 There are three interrupt controllers involved 224 There are three interrupt controllers involved:
225 225
226 1) IOAPIC controller 226 1) IOAPIC controller
227 2) Interrupt remapping controller 227 2) Interrupt remapping controller
228 3) Local APIC controller 228 3) Local APIC controller
229 229
230 To support such a hardware topology and make s 230 To support such a hardware topology and make software architecture match
231 hardware architecture, an irq_domain data stru 231 hardware architecture, an irq_domain data structure is built for each
232 interrupt controller and those irq_domains are 232 interrupt controller and those irq_domains are organized into hierarchy.
233 When building irq_domain hierarchy, the irq_do 233 When building irq_domain hierarchy, the irq_domain near to the device is
234 child and the irq_domain near to CPU is parent 234 child and the irq_domain near to CPU is parent. So a hierarchy structure
235 as below will be built for the example above:: 235 as below will be built for the example above::
236 236
237 CPU Vector irq_domain (root irq_domain 237 CPU Vector irq_domain (root irq_domain to manage CPU vectors)
238 ^ 238 ^
239 | 239 |
240 Interrupt Remapping irq_domain (manage 240 Interrupt Remapping irq_domain (manage irq_remapping entries)
241 ^ 241 ^
242 | 242 |
243 IOAPIC irq_domain (manage IOAPIC deliv 243 IOAPIC irq_domain (manage IOAPIC delivery entries/pins)
244 244
245 There are four major interfaces to use hierarc 245 There are four major interfaces to use hierarchy irq_domain:
246 246
247 1) irq_domain_alloc_irqs(): allocate IRQ descr 247 1) irq_domain_alloc_irqs(): allocate IRQ descriptors and interrupt
248 controller related resources to deliver the 248 controller related resources to deliver these interrupts.
249 2) irq_domain_free_irqs(): free IRQ descriptor 249 2) irq_domain_free_irqs(): free IRQ descriptors and interrupt controller
250 related resources associated with these int 250 related resources associated with these interrupts.
251 3) irq_domain_activate_irq(): activate interru 251 3) irq_domain_activate_irq(): activate interrupt controller hardware to
252 deliver the interrupt. 252 deliver the interrupt.
253 4) irq_domain_deactivate_irq(): deactivate int 253 4) irq_domain_deactivate_irq(): deactivate interrupt controller hardware
254 to stop delivering the interrupt. 254 to stop delivering the interrupt.
255 255
256 Following changes are needed to support hierar 256 Following changes are needed to support hierarchy irq_domain:
257 257
258 1) a new field 'parent' is added to struct irq 258 1) a new field 'parent' is added to struct irq_domain; it's used to
259 maintain irq_domain hierarchy information. 259 maintain irq_domain hierarchy information.
260 2) a new field 'parent_data' is added to struc 260 2) a new field 'parent_data' is added to struct irq_data; it's used to
261 build hierarchy irq_data to match hierarchy 261 build hierarchy irq_data to match hierarchy irq_domains. The irq_data
262 is used to store irq_domain pointer and har 262 is used to store irq_domain pointer and hardware irq number.
263 3) new callbacks are added to struct irq_domai 263 3) new callbacks are added to struct irq_domain_ops to support hierarchy
264 irq_domain operations. 264 irq_domain operations.
265 265
266 With support of hierarchy irq_domain and hiera 266 With support of hierarchy irq_domain and hierarchy irq_data ready, an
267 irq_domain structure is built for each interru 267 irq_domain structure is built for each interrupt controller, and an
268 irq_data structure is allocated for each irq_d 268 irq_data structure is allocated for each irq_domain associated with an
269 IRQ. Now we could go one step further to suppo 269 IRQ. Now we could go one step further to support stacked(hierarchy)
270 irq_chip. That is, an irq_chip is associated w 270 irq_chip. That is, an irq_chip is associated with each irq_data along
271 the hierarchy. A child irq_chip may implement 271 the hierarchy. A child irq_chip may implement a required action by
272 itself or by cooperating with its parent irq_c 272 itself or by cooperating with its parent irq_chip.
273 273
274 With stacked irq_chip, interrupt controller dr 274 With stacked irq_chip, interrupt controller driver only needs to deal
275 with the hardware managed by itself and may as 275 with the hardware managed by itself and may ask for services from its
276 parent irq_chip when needed. So we could achie 276 parent irq_chip when needed. So we could achieve a much cleaner
277 software architecture. 277 software architecture.
278 278
279 For an interrupt controller driver to support 279 For an interrupt controller driver to support hierarchy irq_domain, it
280 needs to: 280 needs to:
281 281
282 1) Implement irq_domain_ops.alloc and irq_doma 282 1) Implement irq_domain_ops.alloc and irq_domain_ops.free
283 2) Optionally implement irq_domain_ops.activat 283 2) Optionally implement irq_domain_ops.activate and
284 irq_domain_ops.deactivate. 284 irq_domain_ops.deactivate.
285 3) Optionally implement an irq_chip to manage 285 3) Optionally implement an irq_chip to manage the interrupt controller
286 hardware. 286 hardware.
287 4) No need to implement irq_domain_ops.map and 287 4) No need to implement irq_domain_ops.map and irq_domain_ops.unmap,
288 they are unused with hierarchy irq_domain. 288 they are unused with hierarchy irq_domain.
289 289
290 Hierarchy irq_domain is in no way x86 specific 290 Hierarchy irq_domain is in no way x86 specific, and is heavily used to
291 support other architectures, such as ARM, ARM6 291 support other architectures, such as ARM, ARM64 etc.
292 292
293 Debugging 293 Debugging
294 ========= 294 =========
295 295
296 Most of the internals of the IRQ subsystem are 296 Most of the internals of the IRQ subsystem are exposed in debugfs by
297 turning CONFIG_GENERIC_IRQ_DEBUGFS on. 297 turning CONFIG_GENERIC_IRQ_DEBUGFS on.
Linux® is a registered trademark of Linus Torvalds in the United States and other countries.
TOMOYO® is a registered trademark of NTT DATA CORPORATION.