qemu/include/hw/ppc/xive.h
<<
>>
Prefs
   1/*
   2 * QEMU PowerPC XIVE interrupt controller model
   3 *
   4 *
   5 * The POWER9 processor comes with a new interrupt controller, called
   6 * XIVE as "eXternal Interrupt Virtualization Engine".
   7 *
   8 * = Overall architecture
   9 *
  10 *
  11 *              XIVE Interrupt Controller
  12 *              +------------------------------------+      IPIs
  13 *              | +---------+ +---------+ +--------+ |    +-------+
  14 *              | |VC       | |CQ       | |PC      |----> | CORES |
  15 *              | |     esb | |         | |        |----> |       |
  16 *              | |     eas | |  Bridge | |   tctx |----> |       |
  17 *              | |SC   end | |         | |    nvt | |    |       |
  18 *  +------+    | +---------+ +----+----+ +--------+ |    +-+-+-+-+
  19 *  | RAM  |    +------------------|-----------------+      | | |
  20 *  |      |                       |                        | | |
  21 *  |      |                       |                        | | |
  22 *  |      |  +--------------------v------------------------v-v-v--+    other
  23 *  |      <--+                     Power Bus                      +--> chips
  24 *  |  esb |  +---------+-----------------------+------------------+
  25 *  |  eas |            |                       |
  26 *  |  end |         +--|------+                |
  27 *  |  nvt |       +----+----+ |           +----+----+
  28 *  +------+       |SC       | |           |SC       |
  29 *                 |         | |           |         |
  30 *                 | PQ-bits | |           | PQ-bits |
  31 *                 | local   |-+           |  in VC  |
  32 *                 +---------+             +---------+
  33 *                    PCIe                 NX,NPU,CAPI
  34 *
  35 *                   SC: Source Controller (aka. IVSE)
  36 *                   VC: Virtualization Controller (aka. IVRE)
  37 *                   PC: Presentation Controller (aka. IVPE)
  38 *                   CQ: Common Queue (Bridge)
  39 *
  40 *              PQ-bits: 2 bits source state machine (P:pending Q:queued)
  41 *                  esb: Event State Buffer (Array of PQ bits in an IVSE)
  42 *                  eas: Event Assignment Structure
  43 *                  end: Event Notification Descriptor
  44 *                  nvt: Notification Virtual Target
  45 *                 tctx: Thread interrupt Context
  46 *
  47 *
  48 * The XIVE IC is composed of three sub-engines :
  49 *
  50 * - Interrupt Virtualization Source Engine (IVSE), or Source
  51 *   Controller (SC). These are found in PCI PHBs, in the PSI host
  52 *   bridge controller, but also inside the main controller for the
  53 *   core IPIs and other sub-chips (NX, CAP, NPU) of the
  54 *   chip/processor. They are configured to feed the IVRE with events.
  55 *
  56 * - Interrupt Virtualization Routing Engine (IVRE) or Virtualization
  57 *   Controller (VC). Its job is to match an event source with an
  58 *   Event Notification Descriptor (END).
  59 *
  60 * - Interrupt Virtualization Presentation Engine (IVPE) or
  61 *   Presentation Controller (PC). It maintains the interrupt context
  62 *   state of each thread and handles the delivery of the external
  63 *   exception to the thread.
  64 *
  65 * In XIVE 1.0, the sub-engines used to be referred as:
  66 *
  67 *   SC     Source Controller
  68 *   VC     Virtualization Controller
  69 *   PC     Presentation Controller
  70 *   CQ     Common Queue (PowerBUS Bridge)
  71 *
  72 *
  73 * = XIVE internal tables
  74 *
  75 * Each of the sub-engines uses a set of tables to redirect exceptions
  76 * from event sources to CPU threads.
  77 *
  78 *                                           +-------+
  79 *   User or OS                              |  EQ   |
  80 *       or                          +------>|entries|
  81 *   Hypervisor                      |       |  ..   |
  82 *     Memory                        |       +-------+
  83 *                                   |           ^
  84 *                                   |           |
  85 *              +-------------------------------------------------+
  86 *                                   |           |
  87 *   Hypervisor      +------+    +---+--+    +---+--+   +------+
  88 *     Memory        | ESB  |    | EAT  |    | ENDT |   | NVTT |
  89 *    (skiboot)      +----+-+    +----+-+    +----+-+   +------+
  90 *                     ^  |        ^  |        ^  |       ^
  91 *                     |  |        |  |        |  |       |
  92 *              +-------------------------------------------------+
  93 *                     |  |        |  |        |  |       |
  94 *                     |  |        |  |        |  |       |
  95 *                +----|--|--------|--|--------|--|-+   +-|-----+    +------+
  96 *                |    |  |        |  |        |  | |   | | tctx|    |Thread|
  97 *   IPI or   --> |    +  v        +  v        +  v |---| +  .. |----->     |
  98 *  HW events --> |                                 |   |       |    |      |
  99 *    IVSE        |             IVRE                |   | IVPE  |    +------+
 100 *                +---------------------------------+   +-------+
 101 *
 102 *
 103 *
 104 * The IVSE have a 2-bits state machine, P for pending and Q for queued,
 105 * for each source that allows events to be triggered. They are stored in
 106 * an Event State Buffer (ESB) array and can be controlled by MMIOs.
 107 *
 108 * If the event is let through, the IVRE looks up in the Event Assignment
 109 * Structure (EAS) table for an Event Notification Descriptor (END)
 110 * configured for the source. Each Event Notification Descriptor defines
 111 * a notification path to a CPU and an in-memory Event Queue, in which
 112 * will be enqueued an EQ data for the OS to pull.
 113 *
 114 * The IVPE determines if a Notification Virtual Target (NVT) can
 115 * handle the event by scanning the thread contexts of the VCPUs
 116 * dispatched on the processor HW threads. It maintains the state of
 117 * the thread interrupt context (TCTX) of each thread in a NVT table.
 118 *
 119 * = Acronyms
 120 *
 121 *          Description                     In XIVE 1.0, used to be referred as
 122 *
 123 *   EAS    Event Assignment Structure      IVE   Interrupt Virt. Entry
 124 *   EAT    Event Assignment Table          IVT   Interrupt Virt. Table
 125 *   ENDT   Event Notif. Descriptor Table   EQDT  Event Queue Desc. Table
 126 *   EQ     Event Queue                     same
 127 *   ESB    Event State Buffer              SBE   State Bit Entry
 128 *   NVT    Notif. Virtual Target           VPD   Virtual Processor Desc.
 129 *   NVTT   Notif. Virtual Target Table     VPDT  Virtual Processor Desc. Table
 130 *   TCTX   Thread interrupt Context
 131 *
 132 *
 133 * Copyright (c) 2017-2018, IBM Corporation.
 134 *
 135 * This code is licensed under the GPL version 2 or later. See the
 136 * COPYING file in the top-level directory.
 137 *
 138 */
 139
 140#ifndef PPC_XIVE_H
 141#define PPC_XIVE_H
 142
 143#include "sysemu/kvm.h"
 144#include "hw/qdev-core.h"
 145#include "hw/sysbus.h"
 146#include "hw/ppc/xive_regs.h"
 147
 148/*
 149 * XIVE Notifier (Interface between Source and Router)
 150 */
 151
 152typedef struct XiveNotifier {
 153    Object parent;
 154} XiveNotifier;
 155
 156#define TYPE_XIVE_NOTIFIER "xive-notifier"
 157#define XIVE_NOTIFIER(obj)                                     \
 158    OBJECT_CHECK(XiveNotifier, (obj), TYPE_XIVE_NOTIFIER)
 159#define XIVE_NOTIFIER_CLASS(klass)                                     \
 160    OBJECT_CLASS_CHECK(XiveNotifierClass, (klass), TYPE_XIVE_NOTIFIER)
 161#define XIVE_NOTIFIER_GET_CLASS(obj)                                   \
 162    OBJECT_GET_CLASS(XiveNotifierClass, (obj), TYPE_XIVE_NOTIFIER)
 163
 164typedef struct XiveNotifierClass {
 165    InterfaceClass parent;
 166    void (*notify)(XiveNotifier *xn, uint32_t lisn);
 167} XiveNotifierClass;
 168
 169/*
 170 * XIVE Interrupt Source
 171 */
 172
 173#define TYPE_XIVE_SOURCE "xive-source"
 174#define XIVE_SOURCE(obj) OBJECT_CHECK(XiveSource, (obj), TYPE_XIVE_SOURCE)
 175
 176/*
 177 * XIVE Interrupt Source characteristics, which define how the ESB are
 178 * controlled.
 179 */
 180#define XIVE_SRC_H_INT_ESB     0x1 /* ESB managed with hcall H_INT_ESB */
 181#define XIVE_SRC_STORE_EOI     0x2 /* Store EOI supported */
 182
 183typedef struct XiveSource {
 184    DeviceState parent;
 185
 186    /* IRQs */
 187    uint32_t        nr_irqs;
 188    unsigned long   *lsi_map;
 189
 190    /* PQ bits and LSI assertion bit */
 191    uint8_t         *status;
 192
 193    /* ESB memory region */
 194    uint64_t        esb_flags;
 195    uint32_t        esb_shift;
 196    MemoryRegion    esb_mmio;
 197
 198    /* KVM support */
 199    void            *esb_mmap;
 200    MemoryRegion    esb_mmio_kvm;
 201
 202    XiveNotifier    *xive;
 203} XiveSource;
 204
 205/*
 206 * ESB MMIO setting. Can be one page, for both source triggering and
 207 * source management, or two different pages. See below for magic
 208 * values.
 209 */
 210#define XIVE_ESB_4K          12 /* PSI HB only */
 211#define XIVE_ESB_4K_2PAGE    13
 212#define XIVE_ESB_64K         16
 213#define XIVE_ESB_64K_2PAGE   17
 214
 215static inline bool xive_source_esb_has_2page(XiveSource *xsrc)
 216{
 217    return xsrc->esb_shift == XIVE_ESB_64K_2PAGE ||
 218        xsrc->esb_shift == XIVE_ESB_4K_2PAGE;
 219}
 220
 221/* The trigger page is always the first/even page */
 222static inline hwaddr xive_source_esb_page(XiveSource *xsrc, uint32_t srcno)
 223{
 224    assert(srcno < xsrc->nr_irqs);
 225    return (1ull << xsrc->esb_shift) * srcno;
 226}
 227
 228/* In a two pages ESB MMIO setting, the odd page is for management */
 229static inline hwaddr xive_source_esb_mgmt(XiveSource *xsrc, int srcno)
 230{
 231    hwaddr addr = xive_source_esb_page(xsrc, srcno);
 232
 233    if (xive_source_esb_has_2page(xsrc)) {
 234        addr += (1 << (xsrc->esb_shift - 1));
 235    }
 236
 237    return addr;
 238}
 239
 240/*
 241 * Each interrupt source has a 2-bit state machine which can be
 242 * controlled by MMIO. P indicates that an interrupt is pending (has
 243 * been sent to a queue and is waiting for an EOI). Q indicates that
 244 * the interrupt has been triggered while pending.
 245 *
 246 * This acts as a coalescing mechanism in order to guarantee that a
 247 * given interrupt only occurs at most once in a queue.
 248 *
 249 * When doing an EOI, the Q bit will indicate if the interrupt
 250 * needs to be re-triggered.
 251 */
 252#define XIVE_STATUS_ASSERTED  0x4  /* Extra bit for LSI */
 253#define XIVE_ESB_VAL_P        0x2
 254#define XIVE_ESB_VAL_Q        0x1
 255
 256#define XIVE_ESB_RESET        0x0
 257#define XIVE_ESB_PENDING      XIVE_ESB_VAL_P
 258#define XIVE_ESB_QUEUED       (XIVE_ESB_VAL_P | XIVE_ESB_VAL_Q)
 259#define XIVE_ESB_OFF          XIVE_ESB_VAL_Q
 260
 261/*
 262 * "magic" Event State Buffer (ESB) MMIO offsets.
 263 *
 264 * The following offsets into the ESB MMIO allow to read or manipulate
 265 * the PQ bits. They must be used with an 8-byte load instruction.
 266 * They all return the previous state of the interrupt (atomically).
 267 *
 268 * Additionally, some ESB pages support doing an EOI via a store and
 269 * some ESBs support doing a trigger via a separate trigger page.
 270 */
 271#define XIVE_ESB_STORE_EOI      0x400 /* Store */
 272#define XIVE_ESB_LOAD_EOI       0x000 /* Load */
 273#define XIVE_ESB_GET            0x800 /* Load */
 274#define XIVE_ESB_SET_PQ_00      0xc00 /* Load */
 275#define XIVE_ESB_SET_PQ_01      0xd00 /* Load */
 276#define XIVE_ESB_SET_PQ_10      0xe00 /* Load */
 277#define XIVE_ESB_SET_PQ_11      0xf00 /* Load */
 278
 279uint8_t xive_source_esb_get(XiveSource *xsrc, uint32_t srcno);
 280uint8_t xive_source_esb_set(XiveSource *xsrc, uint32_t srcno, uint8_t pq);
 281
 282void xive_source_pic_print_info(XiveSource *xsrc, uint32_t offset,
 283                                Monitor *mon);
 284
 285static inline bool xive_source_irq_is_lsi(XiveSource *xsrc, uint32_t srcno)
 286{
 287    assert(srcno < xsrc->nr_irqs);
 288    return test_bit(srcno, xsrc->lsi_map);
 289}
 290
 291static inline void xive_source_irq_set_lsi(XiveSource *xsrc, uint32_t srcno)
 292{
 293    assert(srcno < xsrc->nr_irqs);
 294    bitmap_set(xsrc->lsi_map, srcno, 1);
 295}
 296
 297void xive_source_set_irq(void *opaque, int srcno, int val);
 298
 299/*
 300 * XIVE Thread interrupt Management (TM) context
 301 */
 302
 303#define TYPE_XIVE_TCTX "xive-tctx"
 304#define XIVE_TCTX(obj) OBJECT_CHECK(XiveTCTX, (obj), TYPE_XIVE_TCTX)
 305
 306/*
 307 * XIVE Thread interrupt Management register rings :
 308 *
 309 *   QW-0  User       event-based exception state
 310 *   QW-1  O/S        OS context for priority management, interrupt acks
 311 *   QW-2  Pool       hypervisor pool context for virtual processors dispatched
 312 *   QW-3  Physical   physical thread context and security context
 313 */
 314#define XIVE_TM_RING_COUNT      4
 315#define XIVE_TM_RING_SIZE       0x10
 316
 317typedef struct XiveTCTX {
 318    DeviceState parent_obj;
 319
 320    CPUState    *cs;
 321    qemu_irq    hv_output;
 322    qemu_irq    os_output;
 323
 324    uint8_t     regs[XIVE_TM_RING_COUNT * XIVE_TM_RING_SIZE];
 325} XiveTCTX;
 326
 327/*
 328 * XIVE Router
 329 */
 330
 331typedef struct XiveRouter {
 332    SysBusDevice    parent;
 333} XiveRouter;
 334
 335#define TYPE_XIVE_ROUTER "xive-router"
 336#define XIVE_ROUTER(obj)                                \
 337    OBJECT_CHECK(XiveRouter, (obj), TYPE_XIVE_ROUTER)
 338#define XIVE_ROUTER_CLASS(klass)                                        \
 339    OBJECT_CLASS_CHECK(XiveRouterClass, (klass), TYPE_XIVE_ROUTER)
 340#define XIVE_ROUTER_GET_CLASS(obj)                              \
 341    OBJECT_GET_CLASS(XiveRouterClass, (obj), TYPE_XIVE_ROUTER)
 342
 343typedef struct XiveRouterClass {
 344    SysBusDeviceClass parent;
 345
 346    /* XIVE table accessors */
 347    int (*get_eas)(XiveRouter *xrtr, uint8_t eas_blk, uint32_t eas_idx,
 348                   XiveEAS *eas);
 349    int (*get_end)(XiveRouter *xrtr, uint8_t end_blk, uint32_t end_idx,
 350                   XiveEND *end);
 351    int (*write_end)(XiveRouter *xrtr, uint8_t end_blk, uint32_t end_idx,
 352                     XiveEND *end, uint8_t word_number);
 353    int (*get_nvt)(XiveRouter *xrtr, uint8_t nvt_blk, uint32_t nvt_idx,
 354                   XiveNVT *nvt);
 355    int (*write_nvt)(XiveRouter *xrtr, uint8_t nvt_blk, uint32_t nvt_idx,
 356                     XiveNVT *nvt, uint8_t word_number);
 357    XiveTCTX *(*get_tctx)(XiveRouter *xrtr, CPUState *cs);
 358} XiveRouterClass;
 359
 360void xive_eas_pic_print_info(XiveEAS *eas, uint32_t lisn, Monitor *mon);
 361
 362int xive_router_get_eas(XiveRouter *xrtr, uint8_t eas_blk, uint32_t eas_idx,
 363                        XiveEAS *eas);
 364int xive_router_get_end(XiveRouter *xrtr, uint8_t end_blk, uint32_t end_idx,
 365                        XiveEND *end);
 366int xive_router_write_end(XiveRouter *xrtr, uint8_t end_blk, uint32_t end_idx,
 367                          XiveEND *end, uint8_t word_number);
 368int xive_router_get_nvt(XiveRouter *xrtr, uint8_t nvt_blk, uint32_t nvt_idx,
 369                        XiveNVT *nvt);
 370int xive_router_write_nvt(XiveRouter *xrtr, uint8_t nvt_blk, uint32_t nvt_idx,
 371                          XiveNVT *nvt, uint8_t word_number);
 372XiveTCTX *xive_router_get_tctx(XiveRouter *xrtr, CPUState *cs);
 373void xive_router_notify(XiveNotifier *xn, uint32_t lisn);
 374
 375/*
 376 * XIVE END ESBs
 377 */
 378
 379#define TYPE_XIVE_END_SOURCE "xive-end-source"
 380#define XIVE_END_SOURCE(obj) \
 381    OBJECT_CHECK(XiveENDSource, (obj), TYPE_XIVE_END_SOURCE)
 382
 383typedef struct XiveENDSource {
 384    DeviceState parent;
 385
 386    uint32_t        nr_ends;
 387    uint8_t         block_id;
 388
 389    /* ESB memory region */
 390    uint32_t        esb_shift;
 391    MemoryRegion    esb_mmio;
 392
 393    XiveRouter      *xrtr;
 394} XiveENDSource;
 395
 396/*
 397 * For legacy compatibility, the exceptions define up to 256 different
 398 * priorities. P9 implements only 9 levels : 8 active levels [0 - 7]
 399 * and the least favored level 0xFF.
 400 */
 401#define XIVE_PRIORITY_MAX  7
 402
 403void xive_end_pic_print_info(XiveEND *end, uint32_t end_idx, Monitor *mon);
 404void xive_end_queue_pic_print_info(XiveEND *end, uint32_t width, Monitor *mon);
 405
 406/*
 407 * XIVE Thread Interrupt Management Aera (TIMA)
 408 *
 409 * This region gives access to the registers of the thread interrupt
 410 * management context. It is four page wide, each page providing a
 411 * different view of the registers. The page with the lower offset is
 412 * the most privileged and gives access to the entire context.
 413 */
 414#define XIVE_TM_HW_PAGE         0x0
 415#define XIVE_TM_HV_PAGE         0x1
 416#define XIVE_TM_OS_PAGE         0x2
 417#define XIVE_TM_USER_PAGE       0x3
 418
 419extern const MemoryRegionOps xive_tm_ops;
 420void xive_tctx_tm_write(XiveTCTX *tctx, hwaddr offset, uint64_t value,
 421                        unsigned size);
 422uint64_t xive_tctx_tm_read(XiveTCTX *tctx, hwaddr offset, unsigned size);
 423
 424void xive_tctx_pic_print_info(XiveTCTX *tctx, Monitor *mon);
 425Object *xive_tctx_create(Object *cpu, XiveRouter *xrtr, Error **errp);
 426
 427static inline uint32_t xive_nvt_cam_line(uint8_t nvt_blk, uint32_t nvt_idx)
 428{
 429    return (nvt_blk << 19) | nvt_idx;
 430}
 431
 432/*
 433 * KVM XIVE device helpers
 434 */
 435
 436void kvmppc_xive_source_reset_one(XiveSource *xsrc, int srcno, Error **errp);
 437void kvmppc_xive_source_set_irq(void *opaque, int srcno, int val);
 438void kvmppc_xive_cpu_connect(XiveTCTX *tctx, Error **errp);
 439void kvmppc_xive_cpu_synchronize_state(XiveTCTX *tctx, Error **errp);
 440void kvmppc_xive_cpu_get_state(XiveTCTX *tctx, Error **errp);
 441void kvmppc_xive_cpu_set_state(XiveTCTX *tctx, Error **errp);
 442
 443#endif /* PPC_XIVE_H */
 444