Loading...
1===============================================
2The irq_domain interrupt number mapping library
3===============================================
4
5The current design of the Linux kernel uses a single large number
6space where each separate IRQ source is assigned a different number.
7This is simple when there is only one interrupt controller, but in
8systems with multiple interrupt controllers the kernel must ensure
9that each one gets assigned non-overlapping allocations of Linux
10IRQ numbers.
11
12The number of interrupt controllers registered as unique irqchips
13show a rising tendency: for example subdrivers of different kinds
14such as GPIO controllers avoid reimplementing identical callback
15mechanisms as the IRQ core system by modelling their interrupt
16handlers as irqchips, i.e. in effect cascading interrupt controllers.
17
18Here the interrupt number loose all kind of correspondence to
19hardware interrupt numbers: whereas in the past, IRQ numbers could
20be chosen so they matched the hardware IRQ line into the root
21interrupt controller (i.e. the component actually fireing the
22interrupt line to the CPU) nowadays this number is just a number.
23
24For this reason we need a mechanism to separate controller-local
25interrupt numbers, called hardware irq's, from Linux IRQ numbers.
26
27The irq_alloc_desc*() and irq_free_desc*() APIs provide allocation of
28irq numbers, but they don't provide any support for reverse mapping of
29the controller-local IRQ (hwirq) number into the Linux IRQ number
30space.
31
32The irq_domain library adds mapping between hwirq and IRQ numbers on
33top of the irq_alloc_desc*() API. An irq_domain to manage mapping is
34preferred over interrupt controller drivers open coding their own
35reverse mapping scheme.
36
37irq_domain also implements translation from an abstract irq_fwspec
38structure to hwirq numbers (Device Tree and ACPI GSI so far), and can
39be easily extended to support other IRQ topology data sources.
40
41irq_domain usage
42================
43
44An interrupt controller driver creates and registers an irq_domain by
45calling one of the irq_domain_add_*() functions (each mapping method
46has a different allocator function, more on that later). The function
47will return a pointer to the irq_domain on success. The caller must
48provide the allocator function with an irq_domain_ops structure.
49
50In most cases, the irq_domain will begin empty without any mappings
51between hwirq and IRQ numbers. Mappings are added to the irq_domain
52by calling irq_create_mapping() which accepts the irq_domain and a
53hwirq number as arguments. If a mapping for the hwirq doesn't already
54exist then it will allocate a new Linux irq_desc, associate it with
55the hwirq, and call the .map() callback so the driver can perform any
56required hardware setup.
57
58When an interrupt is received, irq_find_mapping() function should
59be used to find the Linux IRQ number from the hwirq number.
60
61The irq_create_mapping() function must be called *atleast once*
62before any call to irq_find_mapping(), lest the descriptor will not
63be allocated.
64
65If the driver has the Linux IRQ number or the irq_data pointer, and
66needs to know the associated hwirq number (such as in the irq_chip
67callbacks) then it can be directly obtained from irq_data->hwirq.
68
69Types of irq_domain mappings
70============================
71
72There are several mechanisms available for reverse mapping from hwirq
73to Linux irq, and each mechanism uses a different allocation function.
74Which reverse map type should be used depends on the use case. Each
75of the reverse map types are described below:
76
77Linear
78------
79
80::
81
82 irq_domain_add_linear()
83 irq_domain_create_linear()
84
85The linear reverse map maintains a fixed size table indexed by the
86hwirq number. When a hwirq is mapped, an irq_desc is allocated for
87the hwirq, and the IRQ number is stored in the table.
88
89The Linear map is a good choice when the maximum number of hwirqs is
90fixed and a relatively small number (~ < 256). The advantages of this
91map are fixed time lookup for IRQ numbers, and irq_descs are only
92allocated for in-use IRQs. The disadvantage is that the table must be
93as large as the largest possible hwirq number.
94
95irq_domain_add_linear() and irq_domain_create_linear() are functionally
96equivalent, except for the first argument is different - the former
97accepts an Open Firmware specific 'struct device_node', while the latter
98accepts a more general abstraction 'struct fwnode_handle'.
99
100The majority of drivers should use the linear map.
101
102Tree
103----
104
105::
106
107 irq_domain_add_tree()
108 irq_domain_create_tree()
109
110The irq_domain maintains a radix tree map from hwirq numbers to Linux
111IRQs. When an hwirq is mapped, an irq_desc is allocated and the
112hwirq is used as the lookup key for the radix tree.
113
114The tree map is a good choice if the hwirq number can be very large
115since it doesn't need to allocate a table as large as the largest
116hwirq number. The disadvantage is that hwirq to IRQ number lookup is
117dependent on how many entries are in the table.
118
119irq_domain_add_tree() and irq_domain_create_tree() are functionally
120equivalent, except for the first argument is different - the former
121accepts an Open Firmware specific 'struct device_node', while the latter
122accepts a more general abstraction 'struct fwnode_handle'.
123
124Very few drivers should need this mapping.
125
126No Map
127------
128
129::
130
131 irq_domain_add_nomap()
132
133The No Map mapping is to be used when the hwirq number is
134programmable in the hardware. In this case it is best to program the
135Linux IRQ number into the hardware itself so that no mapping is
136required. Calling irq_create_direct_mapping() will allocate a Linux
137IRQ number and call the .map() callback so that driver can program the
138Linux IRQ number into the hardware.
139
140Most drivers cannot use this mapping.
141
142Legacy
143------
144
145::
146
147 irq_domain_add_simple()
148 irq_domain_add_legacy()
149 irq_domain_add_legacy_isa()
150
151The Legacy mapping is a special case for drivers that already have a
152range of irq_descs allocated for the hwirqs. It is used when the
153driver cannot be immediately converted to use the linear mapping. For
154example, many embedded system board support files use a set of #defines
155for IRQ numbers that are passed to struct device registrations. In that
156case the Linux IRQ numbers cannot be dynamically assigned and the legacy
157mapping should be used.
158
159The legacy map assumes a contiguous range of IRQ numbers has already
160been allocated for the controller and that the IRQ number can be
161calculated by adding a fixed offset to the hwirq number, and
162visa-versa. The disadvantage is that it requires the interrupt
163controller to manage IRQ allocations and it requires an irq_desc to be
164allocated for every hwirq, even if it is unused.
165
166The legacy map should only be used if fixed IRQ mappings must be
167supported. For example, ISA controllers would use the legacy map for
168mapping Linux IRQs 0-15 so that existing ISA drivers get the correct IRQ
169numbers.
170
171Most users of legacy mappings should use irq_domain_add_simple() which
172will use a legacy domain only if an IRQ range is supplied by the
173system and will otherwise use a linear domain mapping. The semantics
174of this call are such that if an IRQ range is specified then
175descriptors will be allocated on-the-fly for it, and if no range is
176specified it will fall through to irq_domain_add_linear() which means
177*no* irq descriptors will be allocated.
178
179A typical use case for simple domains is where an irqchip provider
180is supporting both dynamic and static IRQ assignments.
181
182In order to avoid ending up in a situation where a linear domain is
183used and no descriptor gets allocated it is very important to make sure
184that the driver using the simple domain call irq_create_mapping()
185before any irq_find_mapping() since the latter will actually work
186for the static IRQ assignment case.
187
188Hierarchy IRQ domain
189--------------------
190
191On some architectures, there may be multiple interrupt controllers
192involved in delivering an interrupt from the device to the target CPU.
193Let's look at a typical interrupt delivering path on x86 platforms::
194
195 Device --> IOAPIC -> Interrupt remapping Controller -> Local APIC -> CPU
196
197There are three interrupt controllers involved:
198
1991) IOAPIC controller
2002) Interrupt remapping controller
2013) Local APIC controller
202
203To support such a hardware topology and make software architecture match
204hardware architecture, an irq_domain data structure is built for each
205interrupt controller and those irq_domains are organized into hierarchy.
206When building irq_domain hierarchy, the irq_domain near to the device is
207child and the irq_domain near to CPU is parent. So a hierarchy structure
208as below will be built for the example above::
209
210 CPU Vector irq_domain (root irq_domain to manage CPU vectors)
211 ^
212 |
213 Interrupt Remapping irq_domain (manage irq_remapping entries)
214 ^
215 |
216 IOAPIC irq_domain (manage IOAPIC delivery entries/pins)
217
218There are four major interfaces to use hierarchy irq_domain:
219
2201) irq_domain_alloc_irqs(): allocate IRQ descriptors and interrupt
221 controller related resources to deliver these interrupts.
2222) irq_domain_free_irqs(): free IRQ descriptors and interrupt controller
223 related resources associated with these interrupts.
2243) irq_domain_activate_irq(): activate interrupt controller hardware to
225 deliver the interrupt.
2264) irq_domain_deactivate_irq(): deactivate interrupt controller hardware
227 to stop delivering the interrupt.
228
229Following changes are needed to support hierarchy irq_domain:
230
2311) a new field 'parent' is added to struct irq_domain; it's used to
232 maintain irq_domain hierarchy information.
2332) a new field 'parent_data' is added to struct irq_data; it's used to
234 build hierarchy irq_data to match hierarchy irq_domains. The irq_data
235 is used to store irq_domain pointer and hardware irq number.
2363) new callbacks are added to struct irq_domain_ops to support hierarchy
237 irq_domain operations.
238
239With support of hierarchy irq_domain and hierarchy irq_data ready, an
240irq_domain structure is built for each interrupt controller, and an
241irq_data structure is allocated for each irq_domain associated with an
242IRQ. Now we could go one step further to support stacked(hierarchy)
243irq_chip. That is, an irq_chip is associated with each irq_data along
244the hierarchy. A child irq_chip may implement a required action by
245itself or by cooperating with its parent irq_chip.
246
247With stacked irq_chip, interrupt controller driver only needs to deal
248with the hardware managed by itself and may ask for services from its
249parent irq_chip when needed. So we could achieve a much cleaner
250software architecture.
251
252For an interrupt controller driver to support hierarchy irq_domain, it
253needs to:
254
2551) Implement irq_domain_ops.alloc and irq_domain_ops.free
2562) Optionally implement irq_domain_ops.activate and
257 irq_domain_ops.deactivate.
2583) Optionally implement an irq_chip to manage the interrupt controller
259 hardware.
2604) No need to implement irq_domain_ops.map and irq_domain_ops.unmap,
261 they are unused with hierarchy irq_domain.
262
263Hierarchy irq_domain is in no way x86 specific, and is heavily used to
264support other architectures, such as ARM, ARM64 etc.
265
266Debugging
267=========
268
269Most of the internals of the IRQ subsystem are exposed in debugfs by
270turning CONFIG_GENERIC_IRQ_DEBUGFS on.
1===============================================
2The irq_domain interrupt number mapping library
3===============================================
4
5The current design of the Linux kernel uses a single large number
6space where each separate IRQ source is assigned a different number.
7This is simple when there is only one interrupt controller, but in
8systems with multiple interrupt controllers the kernel must ensure
9that each one gets assigned non-overlapping allocations of Linux
10IRQ numbers.
11
12The number of interrupt controllers registered as unique irqchips
13show a rising tendency: for example subdrivers of different kinds
14such as GPIO controllers avoid reimplementing identical callback
15mechanisms as the IRQ core system by modelling their interrupt
16handlers as irqchips, i.e. in effect cascading interrupt controllers.
17
18Here the interrupt number loose all kind of correspondence to
19hardware interrupt numbers: whereas in the past, IRQ numbers could
20be chosen so they matched the hardware IRQ line into the root
21interrupt controller (i.e. the component actually fireing the
22interrupt line to the CPU) nowadays this number is just a number.
23
24For this reason we need a mechanism to separate controller-local
25interrupt numbers, called hardware irq's, from Linux IRQ numbers.
26
27The irq_alloc_desc*() and irq_free_desc*() APIs provide allocation of
28irq numbers, but they don't provide any support for reverse mapping of
29the controller-local IRQ (hwirq) number into the Linux IRQ number
30space.
31
32The irq_domain library adds mapping between hwirq and IRQ numbers on
33top of the irq_alloc_desc*() API. An irq_domain to manage mapping is
34preferred over interrupt controller drivers open coding their own
35reverse mapping scheme.
36
37irq_domain also implements translation from an abstract irq_fwspec
38structure to hwirq numbers (Device Tree and ACPI GSI so far), and can
39be easily extended to support other IRQ topology data sources.
40
41irq_domain usage
42================
43
44An interrupt controller driver creates and registers an irq_domain by
45calling one of the irq_domain_add_*() or irq_domain_create_*() functions
46(each mapping method has a different allocator function, more on that later).
47The function will return a pointer to the irq_domain on success. The caller
48must provide the allocator function with an irq_domain_ops structure.
49
50In most cases, the irq_domain will begin empty without any mappings
51between hwirq and IRQ numbers. Mappings are added to the irq_domain
52by calling irq_create_mapping() which accepts the irq_domain and a
53hwirq number as arguments. If a mapping for the hwirq doesn't already
54exist then it will allocate a new Linux irq_desc, associate it with
55the hwirq, and call the .map() callback so the driver can perform any
56required hardware setup.
57
58Once a mapping has been established, it can be retrieved or used via a
59variety of methods:
60
61- irq_resolve_mapping() returns a pointer to the irq_desc structure
62 for a given domain and hwirq number, and NULL if there was no
63 mapping.
64- irq_find_mapping() returns a Linux IRQ number for a given domain and
65 hwirq number, and 0 if there was no mapping
66- irq_linear_revmap() is now identical to irq_find_mapping(), and is
67 deprecated
68- generic_handle_domain_irq() handles an interrupt described by a
69 domain and a hwirq number
70
71Note that irq domain lookups must happen in contexts that are
72compatible with a RCU read-side critical section.
73
74The irq_create_mapping() function must be called *at least once*
75before any call to irq_find_mapping(), lest the descriptor will not
76be allocated.
77
78If the driver has the Linux IRQ number or the irq_data pointer, and
79needs to know the associated hwirq number (such as in the irq_chip
80callbacks) then it can be directly obtained from irq_data->hwirq.
81
82Types of irq_domain mappings
83============================
84
85There are several mechanisms available for reverse mapping from hwirq
86to Linux irq, and each mechanism uses a different allocation function.
87Which reverse map type should be used depends on the use case. Each
88of the reverse map types are described below:
89
90Linear
91------
92
93::
94
95 irq_domain_add_linear()
96 irq_domain_create_linear()
97
98The linear reverse map maintains a fixed size table indexed by the
99hwirq number. When a hwirq is mapped, an irq_desc is allocated for
100the hwirq, and the IRQ number is stored in the table.
101
102The Linear map is a good choice when the maximum number of hwirqs is
103fixed and a relatively small number (~ < 256). The advantages of this
104map are fixed time lookup for IRQ numbers, and irq_descs are only
105allocated for in-use IRQs. The disadvantage is that the table must be
106as large as the largest possible hwirq number.
107
108irq_domain_add_linear() and irq_domain_create_linear() are functionally
109equivalent, except for the first argument is different - the former
110accepts an Open Firmware specific 'struct device_node', while the latter
111accepts a more general abstraction 'struct fwnode_handle'.
112
113The majority of drivers should use the linear map.
114
115Tree
116----
117
118::
119
120 irq_domain_add_tree()
121 irq_domain_create_tree()
122
123The irq_domain maintains a radix tree map from hwirq numbers to Linux
124IRQs. When an hwirq is mapped, an irq_desc is allocated and the
125hwirq is used as the lookup key for the radix tree.
126
127The tree map is a good choice if the hwirq number can be very large
128since it doesn't need to allocate a table as large as the largest
129hwirq number. The disadvantage is that hwirq to IRQ number lookup is
130dependent on how many entries are in the table.
131
132irq_domain_add_tree() and irq_domain_create_tree() are functionally
133equivalent, except for the first argument is different - the former
134accepts an Open Firmware specific 'struct device_node', while the latter
135accepts a more general abstraction 'struct fwnode_handle'.
136
137Very few drivers should need this mapping.
138
139No Map
140------
141
142::
143
144 irq_domain_add_nomap()
145
146The No Map mapping is to be used when the hwirq number is
147programmable in the hardware. In this case it is best to program the
148Linux IRQ number into the hardware itself so that no mapping is
149required. Calling irq_create_direct_mapping() will allocate a Linux
150IRQ number and call the .map() callback so that driver can program the
151Linux IRQ number into the hardware.
152
153Most drivers cannot use this mapping, and it is now gated on the
154CONFIG_IRQ_DOMAIN_NOMAP option. Please refrain from introducing new
155users of this API.
156
157Legacy
158------
159
160::
161
162 irq_domain_add_simple()
163 irq_domain_add_legacy()
164 irq_domain_create_simple()
165 irq_domain_create_legacy()
166
167The Legacy mapping is a special case for drivers that already have a
168range of irq_descs allocated for the hwirqs. It is used when the
169driver cannot be immediately converted to use the linear mapping. For
170example, many embedded system board support files use a set of #defines
171for IRQ numbers that are passed to struct device registrations. In that
172case the Linux IRQ numbers cannot be dynamically assigned and the legacy
173mapping should be used.
174
175As the name implies, the \*_legacy() functions are deprecated and only
176exist to ease the support of ancient platforms. No new users should be
177added. Same goes for the \*_simple() functions when their use results
178in the legacy behaviour.
179
180The legacy map assumes a contiguous range of IRQ numbers has already
181been allocated for the controller and that the IRQ number can be
182calculated by adding a fixed offset to the hwirq number, and
183visa-versa. The disadvantage is that it requires the interrupt
184controller to manage IRQ allocations and it requires an irq_desc to be
185allocated for every hwirq, even if it is unused.
186
187The legacy map should only be used if fixed IRQ mappings must be
188supported. For example, ISA controllers would use the legacy map for
189mapping Linux IRQs 0-15 so that existing ISA drivers get the correct IRQ
190numbers.
191
192Most users of legacy mappings should use irq_domain_add_simple() or
193irq_domain_create_simple() which will use a legacy domain only if an IRQ range
194is supplied by the system and will otherwise use a linear domain mapping.
195The semantics of this call are such that if an IRQ range is specified then
196descriptors will be allocated on-the-fly for it, and if no range is
197specified it will fall through to irq_domain_add_linear() or
198irq_domain_create_linear() which means *no* irq descriptors will be allocated.
199
200A typical use case for simple domains is where an irqchip provider
201is supporting both dynamic and static IRQ assignments.
202
203In order to avoid ending up in a situation where a linear domain is
204used and no descriptor gets allocated it is very important to make sure
205that the driver using the simple domain call irq_create_mapping()
206before any irq_find_mapping() since the latter will actually work
207for the static IRQ assignment case.
208
209irq_domain_add_simple() and irq_domain_create_simple() as well as
210irq_domain_add_legacy() and irq_domain_create_legacy() are functionally
211equivalent, except for the first argument is different - the former
212accepts an Open Firmware specific 'struct device_node', while the latter
213accepts a more general abstraction 'struct fwnode_handle'.
214
215Hierarchy IRQ domain
216--------------------
217
218On some architectures, there may be multiple interrupt controllers
219involved in delivering an interrupt from the device to the target CPU.
220Let's look at a typical interrupt delivering path on x86 platforms::
221
222 Device --> IOAPIC -> Interrupt remapping Controller -> Local APIC -> CPU
223
224There are three interrupt controllers involved:
225
2261) IOAPIC controller
2272) Interrupt remapping controller
2283) Local APIC controller
229
230To support such a hardware topology and make software architecture match
231hardware architecture, an irq_domain data structure is built for each
232interrupt controller and those irq_domains are organized into hierarchy.
233When building irq_domain hierarchy, the irq_domain near to the device is
234child and the irq_domain near to CPU is parent. So a hierarchy structure
235as below will be built for the example above::
236
237 CPU Vector irq_domain (root irq_domain to manage CPU vectors)
238 ^
239 |
240 Interrupt Remapping irq_domain (manage irq_remapping entries)
241 ^
242 |
243 IOAPIC irq_domain (manage IOAPIC delivery entries/pins)
244
245There are four major interfaces to use hierarchy irq_domain:
246
2471) irq_domain_alloc_irqs(): allocate IRQ descriptors and interrupt
248 controller related resources to deliver these interrupts.
2492) irq_domain_free_irqs(): free IRQ descriptors and interrupt controller
250 related resources associated with these interrupts.
2513) irq_domain_activate_irq(): activate interrupt controller hardware to
252 deliver the interrupt.
2534) irq_domain_deactivate_irq(): deactivate interrupt controller hardware
254 to stop delivering the interrupt.
255
256Following changes are needed to support hierarchy irq_domain:
257
2581) a new field 'parent' is added to struct irq_domain; it's used to
259 maintain irq_domain hierarchy information.
2602) a new field 'parent_data' is added to struct irq_data; it's used to
261 build hierarchy irq_data to match hierarchy irq_domains. The irq_data
262 is used to store irq_domain pointer and hardware irq number.
2633) new callbacks are added to struct irq_domain_ops to support hierarchy
264 irq_domain operations.
265
266With support of hierarchy irq_domain and hierarchy irq_data ready, an
267irq_domain structure is built for each interrupt controller, and an
268irq_data structure is allocated for each irq_domain associated with an
269IRQ. Now we could go one step further to support stacked(hierarchy)
270irq_chip. That is, an irq_chip is associated with each irq_data along
271the hierarchy. A child irq_chip may implement a required action by
272itself or by cooperating with its parent irq_chip.
273
274With stacked irq_chip, interrupt controller driver only needs to deal
275with the hardware managed by itself and may ask for services from its
276parent irq_chip when needed. So we could achieve a much cleaner
277software architecture.
278
279For an interrupt controller driver to support hierarchy irq_domain, it
280needs to:
281
2821) Implement irq_domain_ops.alloc and irq_domain_ops.free
2832) Optionally implement irq_domain_ops.activate and
284 irq_domain_ops.deactivate.
2853) Optionally implement an irq_chip to manage the interrupt controller
286 hardware.
2874) No need to implement irq_domain_ops.map and irq_domain_ops.unmap,
288 they are unused with hierarchy irq_domain.
289
290Hierarchy irq_domain is in no way x86 specific, and is heavily used to
291support other architectures, such as ARM, ARM64 etc.
292
293Debugging
294=========
295
296Most of the internals of the IRQ subsystem are exposed in debugfs by
297turning CONFIG_GENERIC_IRQ_DEBUGFS on.