docs/specs/ppc-spapr-hotplug.txt

   1 = sPAPR Dynamic Reconfiguration =
   2
   3 sPAPR/"pseries" guests make use of a facility called dynamic-reconfiguration
   4 to handle hotplugging of dynamic "physical" resources like PCI cards, or
   5 "logical"/paravirtual resources like memory, CPUs, and "physical"
   6 host-bridges, which are generally managed by the host/hypervisor and provided
   7 to guests as virtualized resources. The specifics of dynamic-reconfiguration
   8 are documented extensively in PAPR+ v2.7, Section 13.1. This document
   9 provides a summary of that information as it applies to the implementation
  10 within QEMU.
  11
  12 == Dynamic-reconfiguration Connectors ==
  13
  14 To manage hotplug/unplug of these resources, a firmware abstraction known as
  15 a Dynamic Resource Connector (DRC) is used to assign a particular dynamic
  16 resource to the guest, and provide an interface for the guest to manage
  17 configuration/removal of the resource associated with it.
  18
  19 == Device-tree description of DRCs ==
  20
  21 A set of 4 Open Firmware device tree array properties are used to describe
  22 the name/index/power-domain/type of each DRC allocated to a guest at
  23 boot-time. There may be multiple sets of these arrays, rooted at different
  24 paths in the device tree depending on the type of resource the DRCs manage.
  25
  26 In some cases, the DRCs themselves may be provided by a dynamic resource,
  27 such as the DRCs managing PCI slots on a hotplugged PHB. In this case the
  28 arrays would be fetched as part of the device tree retrieval interfaces
  29 for hotplugged resources described under "Guest->Host interface".
  30
  31 The array properties are described below. Each entry/element in an array
  32 describes the DRC identified by the element in the corresponding position
  33 of ibm,drc-indexes:
  34
  35 ibm,drc-names:
  36   first 4-bytes: BE-encoded integer denoting the number of entries
  37   each entry: a NULL-terminated <name> string encoded as a byte array
  38
  39   <name> values for logical/virtual resources are defined in PAPR+ v2.7,
  40   Section 13.5.2.4, and basically consist of the type of the resource
  41   followed by a space and a numerical value that's unique across resources
  42   of that type.
  43
  44   <name> values for "physical" resources such as PCI or VIO devices are
  45   defined as being "location codes", which are the "location labels" of
  46   each encapsulating device, starting from the chassis down to the
  47   individual slot for the device, concatenated by a hyphen. This provides
  48   a mapping of resources to a physical location in a chassis for debugging
  49   purposes. For QEMU, this mapping is less important, so we assign a
  50   location code that conforms to naming specifications, but is simply a
  51   location label for the slot by itself to simplify the implementation.
  52   The naming convention for location labels is documented in detail in
  53   PAPR+ v2.7, Section 12.3.1.5, and in our case amounts to using "C<n>"
  54   for PCI/VIO device slots, where <n> is unique across all PCI/VIO
  55   device slots.
  56
  57 ibm,drc-indexes:
  58   first 4-bytes: BE-encoded integer denoting the number of entries
  59   each 4-byte entry: BE-encoded <index> integer that is unique across all DRCs
  60     in the machine
  61
  62   <index> is arbitrary, but in the case of QEMU we try to maintain the
  63   convention used to assign them to pSeries guests on pHyp:
  64
  65     bit[31:28]: integer encoding of <type>, where <type> is:
  66                   1 for CPU resource
  67                   2 for PHB resource
  68                   3 for VIO resource
  69                   4 for PCI resource
  70                   8 for Memory resource
  71     bit[27:0]: integer encoding of <id>, where <id> is unique across
  72                  all resources of specified type
  73
  74 ibm,drc-power-domains:
  75   first 4-bytes: BE-encoded integer denoting the number of entries
  76   each 4-byte entry: 32-bit, BE-encoded <index> integer that specifies the
  77     power domain the resource will be assigned to. In the case of QEMU
  78     we associated all resources with a "live insertion" domain, where the
  79     power is assumed to be managed automatically. The integer value for
  80     this domain is a special value of -1.
  81
  82
  83 ibm,drc-types:
  84   first 4-bytes: BE-encoded integer denoting the number of entries
  85   each entry: a NULL-terminated <type> string encoded as a byte array
  86
  87   <type> is assigned as follows:
  88     "CPU" for a CPU
  89     "PHB" for a physical host-bridge
  90     "SLOT" for a VIO slot
  91     "28" for a PCI slot
  92     "MEM" for memory resource
  93
  94 == Guest->Host interface to manage dynamic resources ==
  95
  96 Each DRC is given a globally unique DRC Index, and resources associated with
  97 a particular DRC are configured/managed by the guest via a number of RTAS
  98 calls which reference individual DRCs based on the DRC index. This can be
  99 considered the guest->host interface.
 100
 101 rtas-set-power-level:
 102   arg[0]: integer identifying power domain
 103   arg[1]: new power level for the domain, 0-100
 104   output[0]: status, 0 on success
 105   output[1]: power level after command
 106
 107   Set the power level for a specified power domain
 108
 109 rtas-get-power-level:
 110   arg[0]: integer identifying power domain
 111   output[0]: status, 0 on success
 112   output[1]: current power level
 113
 114   Get the power level for a specified power domain
 115
 116 rtas-set-indicator:
 117   arg[0]: integer identifying sensor/indicator type
 118   arg[1]: index of sensor, for DR-related sensors this is generally the
 119           DRC index
 120   arg[2]: desired sensor value
 121   output[0]: status, 0 on success
 122
 123   Set the state of an indicator or sensor. For the purpose of this document we
 124   focus on the indicator/sensor types associated with a DRC. The types are:
 125
 126     9001: isolation-state, controls/indicates whether a device has been made
 127           accessible to a guest
 128
 129           supported sensor values:
 130             0: isolate, device is made unaccessible by guest OS
 131             1: unisolate, device is made available to guest OS
 132
 133     9002: dr-indicator, controls "visual" indicator associated with device
 134
 135           supported sensor values:
 136             0: inactive, resource may be safely removed
 137             1: active, resource is in use and cannot be safely removed
 138             2: identify, used to visually identify slot for interactive hotplug
 139             3: action, in most cases, used in the same manner as identify
 140
 141     9003: allocation-state, generally only used for "logical" DR resources to
 142           request the allocation/deallocation of a resource prior to acquiring
 143           it via isolation-state->unisolate, or after releasing it via
 144           isolation-state->isolate, respectively. for "physical" DR (like PCI
 145           hotplug/unplug) the pre-allocation of the resource is implied and
 146           this sensor is unused.
 147
 148           supported sensor values:
 149             0: unusable, tell firmware/system the resource can be
 150                unallocated/reclaimed and added back to the system resource pool
 151             1: usable, request the resource be allocated/reserved for use by
 152                guest OS
 153             2: exchange, used to allocate a spare resource to use for fail-over
 154                in certain situations. unused in QEMU
 155             3: recover, used to reclaim a previously allocated resource that's
 156                not currently allocated to the guest OS. unused in QEMU
 157
 158 rtas-get-sensor-state:
 159   arg[0]: integer identifying sensor/indicator type
 160   arg[1]: index of sensor, for DR-related sensors this is generally the
 161           DRC index
 162   output[0]: status, 0 on success
 163
 164   Used to read an indicator or sensor value.
 165
 166   For DR-related operations, the only noteworthy sensor is dr-entity-sense,
 167   which has a type value of 9003, as allocation-state does in the case of
 168   rtas-set-indicator. The semantics/encodings of the sensor values are distinct
 169   however:
 170
 171   supported sensor values for dr-entity-sense (9003) sensor:
 172     0: empty,
 173          for physical resources: DRC/slot is empty
 174          for logical resources: unused
 175     1: present,
 176          for physical resources: DRC/slot is populated with a device/resource
 177          for logical resources: resource has been allocated to the DRC
 178     2: unusable,
 179          for physical resources: unused
 180          for logical resources: DRC has no resource allocated to it
 181     3: exchange,
 182          for physical resources: unused
 183          for logical resources: resource available for exchange (see
 184            allocation-state sensor semantics above)
 185     4: recovery,
 186          for physical resources: unused
 187          for logical resources: resource available for recovery (see
 188            allocation-state sensor semantics above)
 189
 190 rtas-ibm-configure-connector:
 191   arg[0]: guest physical address of 4096-byte work area buffer
 192   arg[1]: 0, or address of additional 4096-byte work area buffer. only non-zero
 193           if a prior RTAS response indicated a need for additional memory
 194   output[0]: status:
 195                0: completed transmittal of device-tree node
 196                1: instruct guest to prepare for next DT sibling node
 197                2: instruct guest to prepare for next DT child node
 198                3: instruct guest to prepare for next DT property
 199                4: instruct guest to ascend to parent DT node
 200                5: instruct guest to provide additional work-area buffer
 201                   via arg[1]
 202             990x: instruct guest that operation took too long and to try
 203                   again later
 204
 205   Used to fetch an OF device-tree description of the resource associated with
 206   a particular DRC. The DRC index is encoded in the first 4-bytes of the first
 207   work area buffer.
 208
 209   Work area layout, using 4-byte offsets:
 210     wa[0]: DRC index of the DRC to fetch device-tree nodes from
 211     wa[1]: 0 (hard-coded)
 212     wa[2]: for next-sibling/next-child response:
 213              wa offset of null-terminated string denoting the new node's name
 214            for next-property response:
 215              wa offset of null-terminated string denoting new property's name
 216     wa[3]: for next-property response (unused otherwise):
 217              byte-length of new property's value
 218     wa[4]: for next-property response (unused otherwise):
 219              new property's value, encoded as an OFDT-compatible byte array
 220
 221 == hotplug/unplug events ==
 222
 223 For most DR operations, the hypervisor will issue host->guest add/remove events
 224 using the EPOW/check-exception notification framework, where the host issues a
 225 check-exception interrupt, then provides an RTAS event log via an
 226 rtas-check-exception call issued by the guest in response. This framework is
 227 documented by PAPR+ v2.7, and already use in by QEMU for generating powerdown
 228 requests via EPOW events.
 229
 230 For DR, this framework has been extended to include hotplug events, which were
 231 previously unneeded due to direct manipulation of DR-related guest userspace
 232 tools by host-level management such as an HMC. This level of management is not
 233 applicable to PowerKVM, hence the reason for extending the notification
 234 framework to support hotplug events.
 235
 236 The format for these EPOW-signalled events is described below under
 237 "hotplug/unplug event structure". Note that these events are not
 238 formally part of the PAPR+ specification, and have been superseded by a
 239 newer format, also described below under "hotplug/unplug event structure",
 240 and so are now deemed a "legacy" format. The formats are similar, but the
 241 "modern" format contains additional fields/flags, which are denoted for the
 242 purposes of this documentation with "#ifdef GUEST_SUPPORTS_MODERN" guards.
 243
 244 QEMU should assume support only for "legacy" fields/flags unless the guest
 245 advertises support for the "modern" format via ibm,client-architecture-support
 246 hcall by setting byte 5, bit 6 of it's ibm,architecture-vec-5 option vector
 247 structure (as described by LoPAPR v11, B.6.2.3). As with "legacy" format events,
 248 "modern" format events are surfaced to the guest via check-exception RTAS calls,
 249 but use a dedicated event source to signal the guest. This event source is
 250 advertised to the guest by the addition of a "hot-plug-events" node under
 251 "/event-sources" node of the guest's device tree using the standard format
 252 described in LoPAPR v11, B.6.12.1.
 253
 254 == hotplug/unplug event structure ==
 255
 256 The hotplug-specific payload in QEMU is implemented as follows (with all values
 257 encoded in big-endian format):
 258
 259 struct rtas_event_log_v6_hp {
 260 #define SECTION_ID_HOTPLUG              0x4850 /* HP */
 261     struct section_header {
 262         uint16_t section_id;            /* set to SECTION_ID_HOTPLUG */
 263         uint16_t section_length;        /* sizeof(rtas_event_log_v6_hp),
 264                                          * plus the length of the DRC name
 265                                          * if a DRC name identifier is
 266                                          * specified for hotplug_identifier
 267                                          */
 268         uint8_t section_version;        /* version 1 */
 269         uint8_t section_subtype;        /* unused */
 270         uint16_t creator_component_id;  /* unused */
 271     } hdr;
 272 #define RTAS_LOG_V6_HP_TYPE_CPU         1
 273 #define RTAS_LOG_V6_HP_TYPE_MEMORY      2
 274 #define RTAS_LOG_V6_HP_TYPE_SLOT        3
 275 #define RTAS_LOG_V6_HP_TYPE_PHB         4
 276 #define RTAS_LOG_V6_HP_TYPE_PCI         5
 277     uint8_t hotplug_type;               /* type of resource/device */
 278 #define RTAS_LOG_V6_HP_ACTION_ADD       1
 279 #define RTAS_LOG_V6_HP_ACTION_REMOVE    2
 280     uint8_t hotplug_action;             /* action (add/remove) */
 281 #define RTAS_LOG_V6_HP_ID_DRC_NAME          1
 282 #define RTAS_LOG_V6_HP_ID_DRC_INDEX         2
 283 #define RTAS_LOG_V6_HP_ID_DRC_COUNT         3
 284 #ifdef GUEST_SUPPORTS_MODERN
 285 #define RTAS_LOG_V6_HP_ID_DRC_COUNT_INDEXED 4
 286 #endif
 287     uint8_t hotplug_identifier;         /* type of the resource identifier,
 288                                          * which serves as the discriminator
 289                                          * for the 'drc' union field below
 290                                          */
 291 #ifdef GUEST_SUPPORTS_MODERN
 292     uint8_t capabilities;               /* capability flags, currently unused
 293                                          * by QEMU
 294                                          */
 295 #else
 296     uint8_t reserved;
 297 #endif
 298     union {
 299         uint32_t index;                 /* DRC index of resource to take action
 300                                          * on
 301                                          */
 302         uint32_t count;                 /* number of DR resources to take
 303                                          * action on (guest chooses which)
 304                                          */
 305 #ifdef GUEST_SUPPORTS_MODERN
 306         struct {
 307             uint32_t count;             /* number of DR resources to take
 308                                          * action on
 309                                          */
 310             uint32_t index;             /* DRC index of first resource to take
 311                                          * action on. guest will take action
 312                                          * on DRC index <index> through
 313                                          * DRC index <index + count - 1> in
 314                                          * sequential order
 315                                          */
 316         } count_indexed;
 317 #endif
 318         char name[1];                   /* string representing the name of the
 319                                          * DRC to take action on
 320                                          */
 321     } drc;
 322 } QEMU_PACKED;
 323
 324 == ibm,lrdr-capacity ==
 325
 326 ibm,lrdr-capacity is a property in the /rtas device tree node that identifies
 327 the dynamic reconfiguration capabilities of the guest. It consists of a triple
 328 consisting of <phys>, <size> and <maxcpus>.
 329
 330   <phys>, encoded in BE format represents the maximum address in bytes and
 331   hence the maximum memory that can be allocated to the guest.
 332
 333   <size>, encoded in BE format represents the size increments in which
 334   memory can be hot-plugged to the guest.
 335
 336   <maxcpus>, a BE-encoded integer, represents the maximum number of
 337   processors that the guest can have.
 338
 339 pseries guests use this property to note the maximum allowed CPUs for the
 340 guest.
 341
 342 == ibm,dynamic-reconfiguration-memory ==
 343
 344 ibm,dynamic-reconfiguration-memory is a device tree node that represents
 345 dynamically reconfigurable logical memory blocks (LMB). This node
 346 is generated only when the guest advertises the support for it via
 347 ibm,client-architecture-support call. Memory that is not dynamically
 348 reconfigurable is represented by /memory nodes. The properties of this
 349 node that are of interest to the sPAPR memory hotplug implementation
 350 in QEMU are described here.
 351
 352 ibm,lmb-size
 353
 354 This 64bit integer defines the size of each dynamically reconfigurable LMB.
 355
 356 ibm,associativity-lookup-arrays
 357
 358 This property defines a lookup array in which the NUMA associativity
 359 information for each LMB can be found. It is a property encoded array
 360 that begins with an integer M, the number of associativity lists followed
 361 by an integer N, the number of entries per associativity list and terminated
 362 by M associativity lists each of length N integers.
 363
 364 This property provides the same information as given by ibm,associativity
 365 property in a /memory node. Each assigned LMB has an index value between
 366 0 and M-1 which is used as an index into this table to select which
 367 associativity list to use for the LMB. This index value for each LMB
 368 is defined in ibm,dynamic-memory property.
 369
 370 ibm,dynamic-memory
 371
 372 This property describes the dynamically reconfigurable memory. It is a
 373 property encoded array that has an integer N, the number of LMBs followed
 374 by N LMB list entires.
 375
 376 Each LMB list entry consists of the following elements:
 377
 378 - Logical address of the start of the LMB encoded as a 64bit integer. This
 379   corresponds to reg property in /memory node.
 380 - DRC index of the LMB that corresponds to ibm,my-drc-index property
 381   in a /memory node.
 382 - Four bytes reserved for expansion.
 383 - Associativity list index for the LMB that is used as an index into
 384   ibm,associativity-lookup-arrays property described earlier. This
 385   is used to retrieve the right associativity list to be used for this
 386   LMB.
 387 - A 32bit flags word. The bit at bit position 0x00000008 defines whether
 388   the LMB is assigned to the the partition as of boot time.
 389
 390 [1] http://thread.gmane.org/gmane.linux.ports.ppc.embedded/75350/focus=106867