]> git.proxmox.com Git - mirror_qemu.git/blob - docs/specs/ppc-xive.rst
docs/devel/testing: Fix typo in dockerfile path
[mirror_qemu.git] / docs / specs / ppc-xive.rst
1 ================================
2 POWER9 XIVE interrupt controller
3 ================================
4
5 The POWER9 processor comes with a new interrupt controller
6 architecture, called XIVE as "eXternal Interrupt Virtualization
7 Engine".
8
9 Compared to the previous architecture, the main characteristics of
10 XIVE are to support a larger number of interrupt sources and to
11 deliver interrupts directly to virtual processors without hypervisor
12 assistance. This removes the context switches required for the
13 delivery process.
14
15
16 XIVE architecture
17 =================
18
19 The XIVE IC is composed of three sub-engines, each taking care of a
20 processing layer of external interrupts:
21
22 - Interrupt Virtualization Source Engine (IVSE), or Source Controller
23 (SC). These are found in PCI PHBs, in the PSI host bridge
24 controller, but also inside the main controller for the core IPIs
25 and other sub-chips (NX, CAP, NPU) of the chip/processor. They are
26 configured to feed the IVRE with events.
27 - Interrupt Virtualization Routing Engine (IVRE) or Virtualization
28 Controller (VC). It handles event coalescing and perform interrupt
29 routing by matching an event source number with an Event
30 Notification Descriptor (END).
31 - Interrupt Virtualization Presentation Engine (IVPE) or Presentation
32 Controller (PC). It maintains the interrupt context state of each
33 thread and handles the delivery of the external interrupt to the
34 thread.
35
36 ::
37
38 XIVE Interrupt Controller
39 +------------------------------------+ IPIs
40 | +---------+ +---------+ +--------+ | +-------+
41 | |IVRE | |Common Q | |IVPE |----> | CORES |
42 | | esb | | | | |----> | |
43 | | eas | | Bridge | | tctx |----> | |
44 | |SC end | | | | nvt | | | |
45 +------+ | +---------+ +----+----+ +--------+ | +-+-+-+-+
46 | RAM | +------------------|-----------------+ | | |
47 | | | | | |
48 | | | | | |
49 | | +--------------------v------------------------v-v-v--+ other
50 | <--+ Power Bus +--> chips
51 | esb | +---------+-----------------------+------------------+
52 | eas | | |
53 | end | +--|------+ |
54 | nvt | +----+----+ | +----+----+
55 +------+ |IVSE | | |IVSE |
56 | | | | |
57 | PQ-bits | | | PQ-bits |
58 | local |-+ | in VC |
59 +---------+ +---------+
60 PCIe NX,NPU,CAPI
61
62
63 PQ-bits: 2 bits source state machine (P:pending Q:queued)
64 esb: Event State Buffer (Array of PQ bits in an IVSE)
65 eas: Event Assignment Structure
66 end: Event Notification Descriptor
67 nvt: Notification Virtual Target
68 tctx: Thread interrupt Context registers
69
70
71
72 XIVE internal tables
73 --------------------
74
75 Each of the sub-engines uses a set of tables to redirect interrupts
76 from event sources to CPU threads.
77
78 ::
79
80 +-------+
81 User or O/S | EQ |
82 or +------>|entries|
83 Hypervisor | | .. |
84 Memory | +-------+
85 | ^
86 | |
87 +-------------------------------------------------+
88 | |
89 Hypervisor +------+ +---+--+ +---+--+ +------+
90 Memory | ESB | | EAT | | ENDT | | NVTT |
91 (skiboot) +----+-+ +----+-+ +----+-+ +------+
92 ^ | ^ | ^ | ^
93 | | | | | | |
94 +-------------------------------------------------+
95 | | | | | | |
96 | | | | | | |
97 +----|--|--------|--|--------|--|-+ +-|-----+ +------+
98 | | | | | | | | | | tctx| |Thread|
99 IPI or ---+ + v + v + v |---| + .. |-----> |
100 HW events | | | | | |
101 | IVRE | | IVPE | +------+
102 +---------------------------------+ +-------+
103
104
105 The IVSE have a 2-bits state machine, P for pending and Q for queued,
106 for each source that allows events to be triggered. They are stored in
107 an Event State Buffer (ESB) array and can be controlled by MMIOs.
108
109 If the event is let through, the IVRE looks up in the Event Assignment
110 Structure (EAS) table for an Event Notification Descriptor (END)
111 configured for the source. Each Event Notification Descriptor defines
112 a notification path to a CPU and an in-memory Event Queue, in which
113 will be enqueued an EQ data for the O/S to pull.
114
115 The IVPE determines if a Notification Virtual Target (NVT) can handle
116 the event by scanning the thread contexts of the VCPUs dispatched on
117 the processor HW threads. It maintains the interrupt context state of
118 each thread in a NVT table.
119
120 XIVE thread interrupt context
121 -----------------------------
122
123 The XIVE presenter can generate four different exceptions to its
124 HW threads:
125
126 - hypervisor exception
127 - O/S exception
128 - Event-Based Branch (user level)
129 - msgsnd (doorbell)
130
131 Each exception has a state independent from the others called a Thread
132 Interrupt Management context. This context is a set of registers which
133 lets the thread handle priority management and interrupt
134 acknowledgment among other things. The most important ones being :
135
136 - Interrupt Priority Register (PIPR)
137 - Interrupt Pending Buffer (IPB)
138 - Current Processor Priority (CPPR)
139 - Notification Source Register (NSR)
140
141 TIMA
142 ~~~~
143
144 The Thread Interrupt Management registers are accessible through a
145 specific MMIO region, called the Thread Interrupt Management Area
146 (TIMA), four aligned pages, each exposing a different view of the
147 registers. First page (page address ending in ``0b00``) gives access
148 to the entire context and is reserved for the ring 0 view for the
149 physical thread context. The second (page address ending in ``0b01``)
150 is for the hypervisor, ring 1 view. The third (page address ending in
151 ``0b10``) is for the operating system, ring 2 view. The fourth (page
152 address ending in ``0b11``) is for user level, ring 3 view.
153
154 Interrupt flow from an O/S perspective
155 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
156
157 After an event data has been enqueued in the O/S Event Queue, the IVPE
158 raises the bit corresponding to the priority of the pending interrupt
159 in the register IBP (Interrupt Pending Buffer) to indicate that an
160 event is pending in one of the 8 priority queues. The Pending
161 Interrupt Priority Register (PIPR) is also updated using the IPB. This
162 register represent the priority of the most favored pending
163 notification.
164
165 The PIPR is then compared to the the Current Processor Priority
166 Register (CPPR). If it is more favored (numerically less than), the
167 CPU interrupt line is raised and the EO bit of the Notification Source
168 Register (NSR) is updated to notify the presence of an exception for
169 the O/S. The O/S acknowledges the interrupt with a special load in the
170 Thread Interrupt Management Area.
171
172 The O/S handles the interrupt and when done, performs an EOI using a
173 MMIO operation on the ESB management page of the associate source.
174
175 Overview of the QEMU models for XIVE
176 ====================================
177
178 The XiveSource models the IVSE in general, internal and external. It
179 handles the source ESBs and the MMIO interface to control them.
180
181 The XiveNotifier is a small helper interface interconnecting the
182 XiveSource to the XiveRouter.
183
184 The XiveRouter is an abstract model acting as a combined IVRE and
185 IVPE. It routes event notifications using the EAS and END tables to
186 the IVPE sub-engine which does a CAM scan to find a CPU to deliver the
187 exception. Storage should be provided by the inheriting classes.
188
189 XiveEnDSource is a special source object. It exposes the END ESB MMIOs
190 of the Event Queues which are used for coalescing event notifications
191 and for escalation. Not used on the field, only to sync the EQ cache
192 in OPAL.
193
194 Finally, the XiveTCTX contains the interrupt state context of a thread,
195 four sets of registers, one for each exception that can be delivered
196 to a CPU. These contexts are scanned by the IVPE to find a matching VP
197 when a notification is triggered. It also models the Thread Interrupt
198 Management Area (TIMA), which exposes the thread context registers to
199 the CPU for interrupt management.