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/sysbus.h"
 145#include "hw/ppc/xive_regs.h"
 146#include "qom/object.h"
 147
 148/*
 149 * XIVE Notifier (Interface between Source and Router)
 150 */
 151
 152typedef struct XiveNotifier XiveNotifier;
 153
 154#define TYPE_XIVE_NOTIFIER "xive-notifier"
 155#define XIVE_NOTIFIER(obj)                                     \
 156    INTERFACE_CHECK(XiveNotifier, (obj), TYPE_XIVE_NOTIFIER)
 157typedef struct XiveNotifierClass XiveNotifierClass;
 158DECLARE_CLASS_CHECKERS(XiveNotifierClass, XIVE_NOTIFIER,
 159                       TYPE_XIVE_NOTIFIER)
 160
 161struct XiveNotifierClass {
 162    InterfaceClass parent;
 163    void (*notify)(XiveNotifier *xn, uint32_t lisn);
 164};
 165
 166/*
 167 * XIVE Interrupt Source
 168 */
 169
 170#define TYPE_XIVE_SOURCE "xive-source"
 171OBJECT_DECLARE_SIMPLE_TYPE(XiveSource, XIVE_SOURCE)
 172
 173/*
 174 * XIVE Interrupt Source characteristics, which define how the ESB are
 175 * controlled.
 176 */
 177#define XIVE_SRC_H_INT_ESB     0x1 /* ESB managed with hcall H_INT_ESB */
 178#define XIVE_SRC_STORE_EOI     0x2 /* Store EOI supported */
 179
 180struct XiveSource {
 181    DeviceState parent;
 182
 183    /* IRQs */
 184    uint32_t        nr_irqs;
 185    unsigned long   *lsi_map;
 186
 187    /* PQ bits and LSI assertion bit */
 188    uint8_t         *status;
 189
 190    /* ESB memory region */
 191    uint64_t        esb_flags;
 192    uint32_t        esb_shift;
 193    MemoryRegion    esb_mmio;
 194    MemoryRegion    esb_mmio_emulated;
 195
 196    /* KVM support */
 197    void            *esb_mmap;
 198    MemoryRegion    esb_mmio_kvm;
 199
 200    XiveNotifier    *xive;
 201};
 202
 203/*
 204 * ESB MMIO setting. Can be one page, for both source triggering and
 205 * source management, or two different pages. See below for magic
 206 * values.
 207 */
 208#define XIVE_ESB_4K          12 /* PSI HB only */
 209#define XIVE_ESB_4K_2PAGE    13
 210#define XIVE_ESB_64K         16
 211#define XIVE_ESB_64K_2PAGE   17
 212
 213static inline bool xive_source_esb_has_2page(XiveSource *xsrc)
 214{
 215    return xsrc->esb_shift == XIVE_ESB_64K_2PAGE ||
 216        xsrc->esb_shift == XIVE_ESB_4K_2PAGE;
 217}
 218
 219static inline size_t xive_source_esb_len(XiveSource *xsrc)
 220{
 221    return (1ull << xsrc->esb_shift) * xsrc->nr_irqs;
 222}
 223
 224/* The trigger page is always the first/even page */
 225static inline hwaddr xive_source_esb_page(XiveSource *xsrc, uint32_t srcno)
 226{
 227    assert(srcno < xsrc->nr_irqs);
 228    return (1ull << xsrc->esb_shift) * srcno;
 229}
 230
 231/* In a two pages ESB MMIO setting, the odd page is for management */
 232static inline hwaddr xive_source_esb_mgmt(XiveSource *xsrc, int srcno)
 233{
 234    hwaddr addr = xive_source_esb_page(xsrc, srcno);
 235
 236    if (xive_source_esb_has_2page(xsrc)) {
 237        addr += (1 << (xsrc->esb_shift - 1));
 238    }
 239
 240    return addr;
 241}
 242
 243/*
 244 * Each interrupt source has a 2-bit state machine which can be
 245 * controlled by MMIO. P indicates that an interrupt is pending (has
 246 * been sent to a queue and is waiting for an EOI). Q indicates that
 247 * the interrupt has been triggered while pending.
 248 *
 249 * This acts as a coalescing mechanism in order to guarantee that a
 250 * given interrupt only occurs at most once in a queue.
 251 *
 252 * When doing an EOI, the Q bit will indicate if the interrupt
 253 * needs to be re-triggered.
 254 */
 255#define XIVE_STATUS_ASSERTED  0x4  /* Extra bit for LSI */
 256#define XIVE_ESB_VAL_P        0x2
 257#define XIVE_ESB_VAL_Q        0x1
 258
 259#define XIVE_ESB_RESET        0x0
 260#define XIVE_ESB_PENDING      XIVE_ESB_VAL_P
 261#define XIVE_ESB_QUEUED       (XIVE_ESB_VAL_P | XIVE_ESB_VAL_Q)
 262#define XIVE_ESB_OFF          XIVE_ESB_VAL_Q
 263
 264bool xive_esb_trigger(uint8_t *pq);
 265bool xive_esb_eoi(uint8_t *pq);
 266uint8_t xive_esb_set(uint8_t *pq, uint8_t value);
 267
 268/*
 269 * "magic" Event State Buffer (ESB) MMIO offsets.
 270 *
 271 * The following offsets into the ESB MMIO allow to read or manipulate
 272 * the PQ bits. They must be used with an 8-byte load instruction.
 273 * They all return the previous state of the interrupt (atomically).
 274 *
 275 * Additionally, some ESB pages support doing an EOI via a store and
 276 * some ESBs support doing a trigger via a separate trigger page.
 277 */
 278#define XIVE_ESB_STORE_EOI      0x400 /* Store */
 279#define XIVE_ESB_LOAD_EOI       0x000 /* Load */
 280#define XIVE_ESB_GET            0x800 /* Load */
 281#define XIVE_ESB_SET_PQ_00      0xc00 /* Load */
 282#define XIVE_ESB_SET_PQ_01      0xd00 /* Load */
 283#define XIVE_ESB_SET_PQ_10      0xe00 /* Load */
 284#define XIVE_ESB_SET_PQ_11      0xf00 /* Load */
 285
 286uint8_t xive_source_esb_get(XiveSource *xsrc, uint32_t srcno);
 287uint8_t xive_source_esb_set(XiveSource *xsrc, uint32_t srcno, uint8_t pq);
 288
 289/*
 290 * Source status helpers
 291 */
 292static inline void xive_source_set_status(XiveSource *xsrc, uint32_t srcno,
 293                                          uint8_t status, bool enable)
 294{
 295    if (enable) {
 296        xsrc->status[srcno] |= status;
 297    } else {
 298        xsrc->status[srcno] &= ~status;
 299    }
 300}
 301
 302static inline void xive_source_set_asserted(XiveSource *xsrc, uint32_t srcno,
 303                                            bool enable)
 304{
 305    xive_source_set_status(xsrc, srcno, XIVE_STATUS_ASSERTED, enable);
 306}
 307
 308static inline bool xive_source_is_asserted(XiveSource *xsrc, uint32_t srcno)
 309{
 310    return xsrc->status[srcno] & XIVE_STATUS_ASSERTED;
 311}
 312
 313void xive_source_pic_print_info(XiveSource *xsrc, uint32_t offset,
 314                                Monitor *mon);
 315
 316static inline bool xive_source_irq_is_lsi(XiveSource *xsrc, uint32_t srcno)
 317{
 318    assert(srcno < xsrc->nr_irqs);
 319    return test_bit(srcno, xsrc->lsi_map);
 320}
 321
 322static inline void xive_source_irq_set_lsi(XiveSource *xsrc, uint32_t srcno)
 323{
 324    assert(srcno < xsrc->nr_irqs);
 325    bitmap_set(xsrc->lsi_map, srcno, 1);
 326}
 327
 328void xive_source_set_irq(void *opaque, int srcno, int val);
 329
 330/*
 331 * XIVE Thread interrupt Management (TM) context
 332 */
 333
 334#define TYPE_XIVE_TCTX "xive-tctx"
 335OBJECT_DECLARE_SIMPLE_TYPE(XiveTCTX, XIVE_TCTX)
 336
 337/*
 338 * XIVE Thread interrupt Management register rings :
 339 *
 340 *   QW-0  User       event-based exception state
 341 *   QW-1  O/S        OS context for priority management, interrupt acks
 342 *   QW-2  Pool       hypervisor pool context for virtual processors dispatched
 343 *   QW-3  Physical   physical thread context and security context
 344 */
 345#define XIVE_TM_RING_COUNT      4
 346#define XIVE_TM_RING_SIZE       0x10
 347
 348typedef struct XivePresenter XivePresenter;
 349
 350struct XiveTCTX {
 351    DeviceState parent_obj;
 352
 353    CPUState    *cs;
 354    qemu_irq    hv_output;
 355    qemu_irq    os_output;
 356
 357    uint8_t     regs[XIVE_TM_RING_COUNT * XIVE_TM_RING_SIZE];
 358
 359    XivePresenter *xptr;
 360};
 361
 362static inline uint32_t xive_tctx_word2(uint8_t *ring)
 363{
 364    return *((uint32_t *) &ring[TM_WORD2]);
 365}
 366
 367/*
 368 * XIVE Router
 369 */
 370typedef struct XiveFabric XiveFabric;
 371
 372struct XiveRouter {
 373    SysBusDevice    parent;
 374
 375    XiveFabric *xfb;
 376};
 377
 378#define TYPE_XIVE_ROUTER "xive-router"
 379OBJECT_DECLARE_TYPE(XiveRouter, XiveRouterClass,
 380                    XIVE_ROUTER)
 381
 382struct XiveRouterClass {
 383    SysBusDeviceClass parent;
 384
 385    /* XIVE table accessors */
 386    int (*get_eas)(XiveRouter *xrtr, uint8_t eas_blk, uint32_t eas_idx,
 387                   XiveEAS *eas);
 388    int (*get_end)(XiveRouter *xrtr, uint8_t end_blk, uint32_t end_idx,
 389                   XiveEND *end);
 390    int (*write_end)(XiveRouter *xrtr, uint8_t end_blk, uint32_t end_idx,
 391                     XiveEND *end, uint8_t word_number);
 392    int (*get_nvt)(XiveRouter *xrtr, uint8_t nvt_blk, uint32_t nvt_idx,
 393                   XiveNVT *nvt);
 394    int (*write_nvt)(XiveRouter *xrtr, uint8_t nvt_blk, uint32_t nvt_idx,
 395                     XiveNVT *nvt, uint8_t word_number);
 396    uint8_t (*get_block_id)(XiveRouter *xrtr);
 397};
 398
 399int xive_router_get_eas(XiveRouter *xrtr, uint8_t eas_blk, uint32_t eas_idx,
 400                        XiveEAS *eas);
 401int xive_router_get_end(XiveRouter *xrtr, uint8_t end_blk, uint32_t end_idx,
 402                        XiveEND *end);
 403int xive_router_write_end(XiveRouter *xrtr, uint8_t end_blk, uint32_t end_idx,
 404                          XiveEND *end, uint8_t word_number);
 405int xive_router_get_nvt(XiveRouter *xrtr, uint8_t nvt_blk, uint32_t nvt_idx,
 406                        XiveNVT *nvt);
 407int xive_router_write_nvt(XiveRouter *xrtr, uint8_t nvt_blk, uint32_t nvt_idx,
 408                          XiveNVT *nvt, uint8_t word_number);
 409void xive_router_notify(XiveNotifier *xn, uint32_t lisn);
 410
 411/*
 412 * XIVE Presenter
 413 */
 414
 415typedef struct XiveTCTXMatch {
 416    XiveTCTX *tctx;
 417    uint8_t ring;
 418} XiveTCTXMatch;
 419
 420#define TYPE_XIVE_PRESENTER "xive-presenter"
 421#define XIVE_PRESENTER(obj)                                     \
 422    INTERFACE_CHECK(XivePresenter, (obj), TYPE_XIVE_PRESENTER)
 423typedef struct XivePresenterClass XivePresenterClass;
 424DECLARE_CLASS_CHECKERS(XivePresenterClass, XIVE_PRESENTER,
 425                       TYPE_XIVE_PRESENTER)
 426
 427struct XivePresenterClass {
 428    InterfaceClass parent;
 429    int (*match_nvt)(XivePresenter *xptr, uint8_t format,
 430                     uint8_t nvt_blk, uint32_t nvt_idx,
 431                     bool cam_ignore, uint8_t priority,
 432                     uint32_t logic_serv, XiveTCTXMatch *match);
 433    bool (*in_kernel)(const XivePresenter *xptr);
 434};
 435
 436int xive_presenter_tctx_match(XivePresenter *xptr, XiveTCTX *tctx,
 437                              uint8_t format,
 438                              uint8_t nvt_blk, uint32_t nvt_idx,
 439                              bool cam_ignore, uint32_t logic_serv);
 440bool xive_presenter_notify(XiveFabric *xfb, uint8_t format,
 441                           uint8_t nvt_blk, uint32_t nvt_idx,
 442                           bool cam_ignore, uint8_t priority,
 443                           uint32_t logic_serv);
 444
 445/*
 446 * XIVE Fabric (Interface between Interrupt Controller and Machine)
 447 */
 448
 449#define TYPE_XIVE_FABRIC "xive-fabric"
 450#define XIVE_FABRIC(obj)                                     \
 451    INTERFACE_CHECK(XiveFabric, (obj), TYPE_XIVE_FABRIC)
 452typedef struct XiveFabricClass XiveFabricClass;
 453DECLARE_CLASS_CHECKERS(XiveFabricClass, XIVE_FABRIC,
 454                       TYPE_XIVE_FABRIC)
 455
 456struct XiveFabricClass {
 457    InterfaceClass parent;
 458    int (*match_nvt)(XiveFabric *xfb, uint8_t format,
 459                     uint8_t nvt_blk, uint32_t nvt_idx,
 460                     bool cam_ignore, uint8_t priority,
 461                     uint32_t logic_serv, XiveTCTXMatch *match);
 462};
 463
 464/*
 465 * XIVE END ESBs
 466 */
 467
 468#define TYPE_XIVE_END_SOURCE "xive-end-source"
 469OBJECT_DECLARE_SIMPLE_TYPE(XiveENDSource, XIVE_END_SOURCE)
 470
 471struct XiveENDSource {
 472    DeviceState parent;
 473
 474    uint32_t        nr_ends;
 475
 476    /* ESB memory region */
 477    uint32_t        esb_shift;
 478    MemoryRegion    esb_mmio;
 479
 480    XiveRouter      *xrtr;
 481};
 482
 483/*
 484 * For legacy compatibility, the exceptions define up to 256 different
 485 * priorities. P9 implements only 9 levels : 8 active levels [0 - 7]
 486 * and the least favored level 0xFF.
 487 */
 488#define XIVE_PRIORITY_MAX  7
 489
 490/*
 491 * Convert a priority number to an Interrupt Pending Buffer (IPB)
 492 * register, which indicates a pending interrupt at the priority
 493 * corresponding to the bit number
 494 */
 495static inline uint8_t xive_priority_to_ipb(uint8_t priority)
 496{
 497    return priority > XIVE_PRIORITY_MAX ?
 498        0 : 1 << (XIVE_PRIORITY_MAX - priority);
 499}
 500
 501/*
 502 * XIVE Thread Interrupt Management Aera (TIMA)
 503 *
 504 * This region gives access to the registers of the thread interrupt
 505 * management context. It is four page wide, each page providing a
 506 * different view of the registers. The page with the lower offset is
 507 * the most privileged and gives access to the entire context.
 508 */
 509#define XIVE_TM_HW_PAGE         0x0
 510#define XIVE_TM_HV_PAGE         0x1
 511#define XIVE_TM_OS_PAGE         0x2
 512#define XIVE_TM_USER_PAGE       0x3
 513
 514void xive_tctx_tm_write(XivePresenter *xptr, XiveTCTX *tctx, hwaddr offset,
 515                        uint64_t value, unsigned size);
 516uint64_t xive_tctx_tm_read(XivePresenter *xptr, XiveTCTX *tctx, hwaddr offset,
 517                           unsigned size);
 518
 519void xive_tctx_pic_print_info(XiveTCTX *tctx, Monitor *mon);
 520Object *xive_tctx_create(Object *cpu, XivePresenter *xptr, Error **errp);
 521void xive_tctx_reset(XiveTCTX *tctx);
 522void xive_tctx_destroy(XiveTCTX *tctx);
 523void xive_tctx_ipb_update(XiveTCTX *tctx, uint8_t ring, uint8_t ipb);
 524
 525/*
 526 * KVM XIVE device helpers
 527 */
 528
 529int kvmppc_xive_source_reset_one(XiveSource *xsrc, int srcno, Error **errp);
 530void kvmppc_xive_source_set_irq(void *opaque, int srcno, int val);
 531int kvmppc_xive_cpu_connect(XiveTCTX *tctx, Error **errp);
 532int kvmppc_xive_cpu_synchronize_state(XiveTCTX *tctx, Error **errp);
 533int kvmppc_xive_cpu_get_state(XiveTCTX *tctx, Error **errp);
 534int kvmppc_xive_cpu_set_state(XiveTCTX *tctx, Error **errp);
 535
 536#endif /* PPC_XIVE_H */
 537