1/****************************************************************************** 2 * xen.h 3 * 4 * Guest OS interface to Xen. 5 * 6 * Permission is hereby granted, free of charge, to any person obtaining a copy 7 * of this software and associated documentation files (the "Software"), to 8 * deal in the Software without restriction, including without limitation the 9 * rights to use, copy, modify, merge, publish, distribute, sublicense, and/or 10 * sell copies of the Software, and to permit persons to whom the Software is 11 * furnished to do so, subject to the following conditions: 12 * 13 * The above copyright notice and this permission notice shall be included in 14 * all copies or substantial portions of the Software. 15 * 16 * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR 17 * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, 18 * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE 19 * AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER 20 * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING 21 * FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER 22 * DEALINGS IN THE SOFTWARE. 23 * 24 * Copyright (c) 2004, K A Fraser 25 */ 26 27#ifndef __XEN_PUBLIC_XEN_H__ 28#define __XEN_PUBLIC_XEN_H__ 29 30#include "xen-compat.h" 31 32#if defined(__i386__) || defined(__x86_64__) 33#include "arch-x86/xen.h" 34#elif defined(__arm__) || defined (__aarch64__) 35#include "arch-arm.h" 36#else 37#error "Unsupported architecture" 38#endif 39 40#ifndef __ASSEMBLY__ 41/* Guest handles for primitive C types. */ 42DEFINE_XEN_GUEST_HANDLE(char); 43__DEFINE_XEN_GUEST_HANDLE(uchar, unsigned char); 44DEFINE_XEN_GUEST_HANDLE(int); 45__DEFINE_XEN_GUEST_HANDLE(uint, unsigned int); 46#if __XEN_INTERFACE_VERSION__ < 0x00040300 47DEFINE_XEN_GUEST_HANDLE(long); 48__DEFINE_XEN_GUEST_HANDLE(ulong, unsigned long); 49#endif 50DEFINE_XEN_GUEST_HANDLE(void); 51 52DEFINE_XEN_GUEST_HANDLE(uint64_t); 53DEFINE_XEN_GUEST_HANDLE(xen_pfn_t); 54DEFINE_XEN_GUEST_HANDLE(xen_ulong_t); 55#endif 56 57/* 58 * HYPERCALLS 59 */ 60 61/* `incontents 100 hcalls List of hypercalls 62 * ` enum hypercall_num { // __HYPERVISOR_* => HYPERVISOR_*() 63 */ 64 65#define __HYPERVISOR_set_trap_table 0 66#define __HYPERVISOR_mmu_update 1 67#define __HYPERVISOR_set_gdt 2 68#define __HYPERVISOR_stack_switch 3 69#define __HYPERVISOR_set_callbacks 4 70#define __HYPERVISOR_fpu_taskswitch 5 71#define __HYPERVISOR_sched_op_compat 6 /* compat since 0x00030101 */ 72#define __HYPERVISOR_platform_op 7 73#define __HYPERVISOR_set_debugreg 8 74#define __HYPERVISOR_get_debugreg 9 75#define __HYPERVISOR_update_descriptor 10 76#define __HYPERVISOR_memory_op 12 77#define __HYPERVISOR_multicall 13 78#define __HYPERVISOR_update_va_mapping 14 79#define __HYPERVISOR_set_timer_op 15 80#define __HYPERVISOR_event_channel_op_compat 16 /* compat since 0x00030202 */ 81#define __HYPERVISOR_xen_version 17 82#define __HYPERVISOR_console_io 18 83#define __HYPERVISOR_physdev_op_compat 19 /* compat since 0x00030202 */ 84#define __HYPERVISOR_grant_table_op 20 85#define __HYPERVISOR_vm_assist 21 86#define __HYPERVISOR_update_va_mapping_otherdomain 22 87#define __HYPERVISOR_iret 23 /* x86 only */ 88#define __HYPERVISOR_vcpu_op 24 89#define __HYPERVISOR_set_segment_base 25 /* x86/64 only */ 90#define __HYPERVISOR_mmuext_op 26 91#define __HYPERVISOR_xsm_op 27 92#define __HYPERVISOR_nmi_op 28 93#define __HYPERVISOR_sched_op 29 94#define __HYPERVISOR_callback_op 30 95#define __HYPERVISOR_xenoprof_op 31 96#define __HYPERVISOR_event_channel_op 32 97#define __HYPERVISOR_physdev_op 33 98#define __HYPERVISOR_hvm_op 34 99#define __HYPERVISOR_sysctl 35 100#define __HYPERVISOR_domctl 36 101#define __HYPERVISOR_kexec_op 37 102#define __HYPERVISOR_tmem_op 38 103#define __HYPERVISOR_xc_reserved_op 39 /* reserved for XenClient */ 104#define __HYPERVISOR_xenpmu_op 40 105 106/* Architecture-specific hypercall definitions. */ 107#define __HYPERVISOR_arch_0 48 108#define __HYPERVISOR_arch_1 49 109#define __HYPERVISOR_arch_2 50 110#define __HYPERVISOR_arch_3 51 111#define __HYPERVISOR_arch_4 52 112#define __HYPERVISOR_arch_5 53 113#define __HYPERVISOR_arch_6 54 114#define __HYPERVISOR_arch_7 55 115 116/* ` } */ 117 118/* 119 * HYPERCALL COMPATIBILITY. 120 */ 121 122/* New sched_op hypercall introduced in 0x00030101. */ 123#if __XEN_INTERFACE_VERSION__ < 0x00030101 124#undef __HYPERVISOR_sched_op 125#define __HYPERVISOR_sched_op __HYPERVISOR_sched_op_compat 126#endif 127 128/* New event-channel and physdev hypercalls introduced in 0x00030202. */ 129#if __XEN_INTERFACE_VERSION__ < 0x00030202 130#undef __HYPERVISOR_event_channel_op 131#define __HYPERVISOR_event_channel_op __HYPERVISOR_event_channel_op_compat 132#undef __HYPERVISOR_physdev_op 133#define __HYPERVISOR_physdev_op __HYPERVISOR_physdev_op_compat 134#endif 135 136/* New platform_op hypercall introduced in 0x00030204. */ 137#if __XEN_INTERFACE_VERSION__ < 0x00030204 138#define __HYPERVISOR_dom0_op __HYPERVISOR_platform_op 139#endif 140 141/* 142 * VIRTUAL INTERRUPTS 143 * 144 * Virtual interrupts that a guest OS may receive from Xen. 145 * 146 * In the side comments, 'V.' denotes a per-VCPU VIRQ while 'G.' denotes a 147 * global VIRQ. The former can be bound once per VCPU and cannot be re-bound. 148 * The latter can be allocated only once per guest: they must initially be 149 * allocated to VCPU0 but can subsequently be re-bound. 150 */ 151/* ` enum virq { */ 152#define VIRQ_TIMER 0 /* V. Timebase update, and/or requested timeout. */ 153#define VIRQ_DEBUG 1 /* V. Request guest to dump debug info. */ 154#define VIRQ_CONSOLE 2 /* G. (DOM0) Bytes received on emergency console. */ 155#define VIRQ_DOM_EXC 3 /* G. (DOM0) Exceptional event for some domain. */ 156#define VIRQ_TBUF 4 /* G. (DOM0) Trace buffer has records available. */ 157#define VIRQ_DEBUGGER 6 /* G. (DOM0) A domain has paused for debugging. */ 158#define VIRQ_XENOPROF 7 /* V. XenOprofile interrupt: new sample available */ 159#define VIRQ_CON_RING 8 /* G. (DOM0) Bytes received on console */ 160#define VIRQ_PCPU_STATE 9 /* G. (DOM0) PCPU state changed */ 161#define VIRQ_MEM_EVENT 10 /* G. (DOM0) A memory event has occured */ 162#define VIRQ_XC_RESERVED 11 /* G. Reserved for XenClient */ 163#define VIRQ_ENOMEM 12 /* G. (DOM0) Low on heap memory */ 164#define VIRQ_XENPMU 13 /* V. PMC interrupt */ 165 166/* Architecture-specific VIRQ definitions. */ 167#define VIRQ_ARCH_0 16 168#define VIRQ_ARCH_1 17 169#define VIRQ_ARCH_2 18 170#define VIRQ_ARCH_3 19 171#define VIRQ_ARCH_4 20 172#define VIRQ_ARCH_5 21 173#define VIRQ_ARCH_6 22 174#define VIRQ_ARCH_7 23 175/* ` } */ 176 177#define NR_VIRQS 24 178 179/* 180 * ` enum neg_errnoval 181 * ` HYPERVISOR_mmu_update(const struct mmu_update reqs[], 182 * ` unsigned count, unsigned *done_out, 183 * ` unsigned foreigndom) 184 * ` 185 * @reqs is an array of mmu_update_t structures ((ptr, val) pairs). 186 * @count is the length of the above array. 187 * @pdone is an output parameter indicating number of completed operations 188 * @foreigndom[15:0]: FD, the expected owner of data pages referenced in this 189 * hypercall invocation. Can be DOMID_SELF. 190 * @foreigndom[31:16]: PFD, the expected owner of pagetable pages referenced 191 * in this hypercall invocation. The value of this field 192 * (x) encodes the PFD as follows: 193 * x == 0 => PFD == DOMID_SELF 194 * x != 0 => PFD == x - 1 195 * 196 * Sub-commands: ptr[1:0] specifies the appropriate MMU_* command. 197 * ------------- 198 * ptr[1:0] == MMU_NORMAL_PT_UPDATE: 199 * Updates an entry in a page table belonging to PFD. If updating an L1 table, 200 * and the new table entry is valid/present, the mapped frame must belong to 201 * FD. If attempting to map an I/O page then the caller assumes the privilege 202 * of the FD. 203 * FD == DOMID_IO: Permit /only/ I/O mappings, at the priv level of the caller. 204 * FD == DOMID_XEN: Map restricted areas of Xen's heap space. 205 * ptr[:2] -- Machine address of the page-table entry to modify. 206 * val -- Value to write. 207 * 208 * There also certain implicit requirements when using this hypercall. The 209 * pages that make up a pagetable must be mapped read-only in the guest. 210 * This prevents uncontrolled guest updates to the pagetable. Xen strictly 211 * enforces this, and will disallow any pagetable update which will end up 212 * mapping pagetable page RW, and will disallow using any writable page as a 213 * pagetable. In practice it means that when constructing a page table for a 214 * process, thread, etc, we MUST be very dilligient in following these rules: 215 * 1). Start with top-level page (PGD or in Xen language: L4). Fill out 216 * the entries. 217 * 2). Keep on going, filling out the upper (PUD or L3), and middle (PMD 218 * or L2). 219 * 3). Start filling out the PTE table (L1) with the PTE entries. Once 220 * done, make sure to set each of those entries to RO (so writeable bit 221 * is unset). Once that has been completed, set the PMD (L2) for this 222 * PTE table as RO. 223 * 4). When completed with all of the PMD (L2) entries, and all of them have 224 * been set to RO, make sure to set RO the PUD (L3). Do the same 225 * operation on PGD (L4) pagetable entries that have a PUD (L3) entry. 226 * 5). Now before you can use those pages (so setting the cr3), you MUST also 227 * pin them so that the hypervisor can verify the entries. This is done 228 * via the HYPERVISOR_mmuext_op(MMUEXT_PIN_L4_TABLE, guest physical frame 229 * number of the PGD (L4)). And this point the HYPERVISOR_mmuext_op( 230 * MMUEXT_NEW_BASEPTR, guest physical frame number of the PGD (L4)) can be 231 * issued. 232 * For 32-bit guests, the L4 is not used (as there is less pagetables), so 233 * instead use L3. 234 * At this point the pagetables can be modified using the MMU_NORMAL_PT_UPDATE 235 * hypercall. Also if so desired the OS can also try to write to the PTE 236 * and be trapped by the hypervisor (as the PTE entry is RO). 237 * 238 * To deallocate the pages, the operations are the reverse of the steps 239 * mentioned above. The argument is MMUEXT_UNPIN_TABLE for all levels and the 240 * pagetable MUST not be in use (meaning that the cr3 is not set to it). 241 * 242 * ptr[1:0] == MMU_MACHPHYS_UPDATE: 243 * Updates an entry in the machine->pseudo-physical mapping table. 244 * ptr[:2] -- Machine address within the frame whose mapping to modify. 245 * The frame must belong to the FD, if one is specified. 246 * val -- Value to write into the mapping entry. 247 * 248 * ptr[1:0] == MMU_PT_UPDATE_PRESERVE_AD: 249 * As MMU_NORMAL_PT_UPDATE above, but A/D bits currently in the PTE are ORed 250 * with those in @val. 251 * 252 * @val is usually the machine frame number along with some attributes. 253 * The attributes by default follow the architecture defined bits. Meaning that 254 * if this is a X86_64 machine and four page table layout is used, the layout 255 * of val is: 256 * - 63 if set means No execute (NX) 257 * - 46-13 the machine frame number 258 * - 12 available for guest 259 * - 11 available for guest 260 * - 10 available for guest 261 * - 9 available for guest 262 * - 8 global 263 * - 7 PAT (PSE is disabled, must use hypercall to make 4MB or 2MB pages) 264 * - 6 dirty 265 * - 5 accessed 266 * - 4 page cached disabled 267 * - 3 page write through 268 * - 2 userspace accessible 269 * - 1 writeable 270 * - 0 present 271 * 272 * The one bits that does not fit with the default layout is the PAGE_PSE 273 * also called PAGE_PAT). The MMUEXT_[UN]MARK_SUPER arguments to the 274 * HYPERVISOR_mmuext_op serve as mechanism to set a pagetable to be 4MB 275 * (or 2MB) instead of using the PAGE_PSE bit. 276 * 277 * The reason that the PAGE_PSE (bit 7) is not being utilized is due to Xen 278 * using it as the Page Attribute Table (PAT) bit - for details on it please 279 * refer to Intel SDM 10.12. The PAT allows to set the caching attributes of 280 * pages instead of using MTRRs. 281 * 282 * The PAT MSR is as follows (it is a 64-bit value, each entry is 8 bits): 283 * PAT4 PAT0 284 * +-----+-----+----+----+----+-----+----+----+ 285 * | UC | UC- | WC | WB | UC | UC- | WC | WB | <= Linux 286 * +-----+-----+----+----+----+-----+----+----+ 287 * | UC | UC- | WT | WB | UC | UC- | WT | WB | <= BIOS (default when machine boots) 288 * +-----+-----+----+----+----+-----+----+----+ 289 * | rsv | rsv | WP | WC | UC | UC- | WT | WB | <= Xen 290 * +-----+-----+----+----+----+-----+----+----+ 291 * 292 * The lookup of this index table translates to looking up 293 * Bit 7, Bit 4, and Bit 3 of val entry: 294 * 295 * PAT/PSE (bit 7) ... PCD (bit 4) .. PWT (bit 3). 296 * 297 * If all bits are off, then we are using PAT0. If bit 3 turned on, 298 * then we are using PAT1, if bit 3 and bit 4, then PAT2.. 299 * 300 * As you can see, the Linux PAT1 translates to PAT4 under Xen. Which means 301 * that if a guest that follows Linux's PAT setup and would like to set Write 302 * Combined on pages it MUST use PAT4 entry. Meaning that Bit 7 (PAGE_PAT) is 303 * set. For example, under Linux it only uses PAT0, PAT1, and PAT2 for the 304 * caching as: 305 * 306 * WB = none (so PAT0) 307 * WC = PWT (bit 3 on) 308 * UC = PWT | PCD (bit 3 and 4 are on). 309 * 310 * To make it work with Xen, it needs to translate the WC bit as so: 311 * 312 * PWT (so bit 3 on) --> PAT (so bit 7 is on) and clear bit 3 313 * 314 * And to translate back it would: 315 * 316 * PAT (bit 7 on) --> PWT (bit 3 on) and clear bit 7. 317 */ 318#define MMU_NORMAL_PT_UPDATE 0 /* checked '*ptr = val'. ptr is MA. */ 319#define MMU_MACHPHYS_UPDATE 1 /* ptr = MA of frame to modify entry for */ 320#define MMU_PT_UPDATE_PRESERVE_AD 2 /* atomically: *ptr = val | (*ptr&(A|D)) */ 321 322/* 323 * MMU EXTENDED OPERATIONS 324 * 325 * ` enum neg_errnoval 326 * ` HYPERVISOR_mmuext_op(mmuext_op_t uops[], 327 * ` unsigned int count, 328 * ` unsigned int *pdone, 329 * ` unsigned int foreigndom) 330 */ 331/* HYPERVISOR_mmuext_op() accepts a list of mmuext_op structures. 332 * A foreigndom (FD) can be specified (or DOMID_SELF for none). 333 * Where the FD has some effect, it is described below. 334 * 335 * cmd: MMUEXT_(UN)PIN_*_TABLE 336 * mfn: Machine frame number to be (un)pinned as a p.t. page. 337 * The frame must belong to the FD, if one is specified. 338 * 339 * cmd: MMUEXT_NEW_BASEPTR 340 * mfn: Machine frame number of new page-table base to install in MMU. 341 * 342 * cmd: MMUEXT_NEW_USER_BASEPTR [x86/64 only] 343 * mfn: Machine frame number of new page-table base to install in MMU 344 * when in user space. 345 * 346 * cmd: MMUEXT_TLB_FLUSH_LOCAL 347 * No additional arguments. Flushes local TLB. 348 * 349 * cmd: MMUEXT_INVLPG_LOCAL 350 * linear_addr: Linear address to be flushed from the local TLB. 351 * 352 * cmd: MMUEXT_TLB_FLUSH_MULTI 353 * vcpumask: Pointer to bitmap of VCPUs to be flushed. 354 * 355 * cmd: MMUEXT_INVLPG_MULTI 356 * linear_addr: Linear address to be flushed. 357 * vcpumask: Pointer to bitmap of VCPUs to be flushed. 358 * 359 * cmd: MMUEXT_TLB_FLUSH_ALL 360 * No additional arguments. Flushes all VCPUs' TLBs. 361 * 362 * cmd: MMUEXT_INVLPG_ALL 363 * linear_addr: Linear address to be flushed from all VCPUs' TLBs. 364 * 365 * cmd: MMUEXT_FLUSH_CACHE 366 * No additional arguments. Writes back and flushes cache contents. 367 * 368 * cmd: MMUEXT_FLUSH_CACHE_GLOBAL 369 * No additional arguments. Writes back and flushes cache contents 370 * on all CPUs in the system. 371 * 372 * cmd: MMUEXT_SET_LDT 373 * linear_addr: Linear address of LDT base (NB. must be page-aligned). 374 * nr_ents: Number of entries in LDT. 375 * 376 * cmd: MMUEXT_CLEAR_PAGE 377 * mfn: Machine frame number to be cleared. 378 * 379 * cmd: MMUEXT_COPY_PAGE 380 * mfn: Machine frame number of the destination page. 381 * src_mfn: Machine frame number of the source page. 382 * 383 * cmd: MMUEXT_[UN]MARK_SUPER 384 * mfn: Machine frame number of head of superpage to be [un]marked. 385 */ 386/* ` enum mmuext_cmd { */ 387#define MMUEXT_PIN_L1_TABLE 0 388#define MMUEXT_PIN_L2_TABLE 1 389#define MMUEXT_PIN_L3_TABLE 2 390#define MMUEXT_PIN_L4_TABLE 3 391#define MMUEXT_UNPIN_TABLE 4 392#define MMUEXT_NEW_BASEPTR 5 393#define MMUEXT_TLB_FLUSH_LOCAL 6 394#define MMUEXT_INVLPG_LOCAL 7 395#define MMUEXT_TLB_FLUSH_MULTI 8 396#define MMUEXT_INVLPG_MULTI 9 397#define MMUEXT_TLB_FLUSH_ALL 10 398#define MMUEXT_INVLPG_ALL 11 399#define MMUEXT_FLUSH_CACHE 12 400#define MMUEXT_SET_LDT 13 401#define MMUEXT_NEW_USER_BASEPTR 15 402#define MMUEXT_CLEAR_PAGE 16 403#define MMUEXT_COPY_PAGE 17 404#define MMUEXT_FLUSH_CACHE_GLOBAL 18 405#define MMUEXT_MARK_SUPER 19 406#define MMUEXT_UNMARK_SUPER 20 407/* ` } */ 408 409#ifndef __ASSEMBLY__ 410struct mmuext_op { 411 unsigned int cmd; /* => enum mmuext_cmd */ 412 union { 413 /* [UN]PIN_TABLE, NEW_BASEPTR, NEW_USER_BASEPTR 414 * CLEAR_PAGE, COPY_PAGE, [UN]MARK_SUPER */ 415 xen_pfn_t mfn; 416 /* INVLPG_LOCAL, INVLPG_ALL, SET_LDT */ 417 unsigned long linear_addr; 418 } arg1; 419 union { 420 /* SET_LDT */ 421 unsigned int nr_ents; 422 /* TLB_FLUSH_MULTI, INVLPG_MULTI */ 423#if __XEN_INTERFACE_VERSION__ >= 0x00030205 424 XEN_GUEST_HANDLE(const_void) vcpumask; 425#else 426 const void *vcpumask; 427#endif 428 /* COPY_PAGE */ 429 xen_pfn_t src_mfn; 430 } arg2; 431}; 432typedef struct mmuext_op mmuext_op_t; 433DEFINE_XEN_GUEST_HANDLE(mmuext_op_t); 434#endif 435 436/* 437 * ` enum neg_errnoval 438 * ` HYPERVISOR_update_va_mapping(unsigned long va, u64 val, 439 * ` enum uvm_flags flags) 440 * ` 441 * ` enum neg_errnoval 442 * ` HYPERVISOR_update_va_mapping_otherdomain(unsigned long va, u64 val, 443 * ` enum uvm_flags flags, 444 * ` domid_t domid) 445 * ` 446 * ` @va: The virtual address whose mapping we want to change 447 * ` @val: The new page table entry, must contain a machine address 448 * ` @flags: Control TLB flushes 449 */ 450/* These are passed as 'flags' to update_va_mapping. They can be ORed. */ 451/* When specifying UVMF_MULTI, also OR in a pointer to a CPU bitmap. */ 452/* UVMF_LOCAL is merely UVMF_MULTI with a NULL bitmap pointer. */ 453/* ` enum uvm_flags { */ 454#define UVMF_NONE (0UL<<0) /* No flushing at all. */ 455#define UVMF_TLB_FLUSH (1UL<<0) /* Flush entire TLB(s). */ 456#define UVMF_INVLPG (2UL<<0) /* Flush only one entry. */ 457#define UVMF_FLUSHTYPE_MASK (3UL<<0) 458#define UVMF_MULTI (0UL<<2) /* Flush subset of TLBs. */ 459#define UVMF_LOCAL (0UL<<2) /* Flush local TLB. */ 460#define UVMF_ALL (1UL<<2) /* Flush all TLBs. */ 461/* ` } */ 462 463/* 464 * Commands to HYPERVISOR_console_io(). 465 */ 466#define CONSOLEIO_write 0 467#define CONSOLEIO_read 1 468 469/* 470 * Commands to HYPERVISOR_vm_assist(). 471 */ 472#define VMASST_CMD_enable 0 473#define VMASST_CMD_disable 1 474 475/* x86/32 guests: simulate full 4GB segment limits. */ 476#define VMASST_TYPE_4gb_segments 0 477 478/* x86/32 guests: trap (vector 15) whenever above vmassist is used. */ 479#define VMASST_TYPE_4gb_segments_notify 1 480 481/* 482 * x86 guests: support writes to bottom-level PTEs. 483 * NB1. Page-directory entries cannot be written. 484 * NB2. Guest must continue to remove all writable mappings of PTEs. 485 */ 486#define VMASST_TYPE_writable_pagetables 2 487 488/* x86/PAE guests: support PDPTs above 4GB. */ 489#define VMASST_TYPE_pae_extended_cr3 3 490 491/* 492 * x86/64 guests: strictly hide M2P from user mode. 493 * This allows the guest to control respective hypervisor behavior: 494 * - when not set, L4 tables get created with the respective slot blank, 495 * and whenever the L4 table gets used as a kernel one the missing 496 * mapping gets inserted, 497 * - when set, L4 tables get created with the respective slot initialized 498 * as before, and whenever the L4 table gets used as a user one the 499 * mapping gets zapped. 500 */ 501#define VMASST_TYPE_m2p_strict 32 502 503#if __XEN_INTERFACE_VERSION__ < 0x00040600 504#define MAX_VMASST_TYPE 3 505#endif 506 507#ifndef __ASSEMBLY__ 508 509typedef uint16_t domid_t; 510 511/* Domain ids >= DOMID_FIRST_RESERVED cannot be used for ordinary domains. */ 512#define DOMID_FIRST_RESERVED (0x7FF0U) 513 514/* DOMID_SELF is used in certain contexts to refer to oneself. */ 515#define DOMID_SELF (0x7FF0U) 516 517/* 518 * DOMID_IO is used to restrict page-table updates to mapping I/O memory. 519 * Although no Foreign Domain need be specified to map I/O pages, DOMID_IO 520 * is useful to ensure that no mappings to the OS's own heap are accidentally 521 * installed. (e.g., in Linux this could cause havoc as reference counts 522 * aren't adjusted on the I/O-mapping code path). 523 * This only makes sense in MMUEXT_SET_FOREIGNDOM, but in that context can 524 * be specified by any calling domain. 525 */ 526#define DOMID_IO (0x7FF1U) 527 528/* 529 * DOMID_XEN is used to allow privileged domains to map restricted parts of 530 * Xen's heap space (e.g., the machine_to_phys table). 531 * This only makes sense in MMUEXT_SET_FOREIGNDOM, and is only permitted if 532 * the caller is privileged. 533 */ 534#define DOMID_XEN (0x7FF2U) 535 536/* 537 * DOMID_COW is used as the owner of sharable pages */ 538#define DOMID_COW (0x7FF3U) 539 540/* DOMID_INVALID is used to identify pages with unknown owner. */ 541#define DOMID_INVALID (0x7FF4U) 542 543/* Idle domain. */ 544#define DOMID_IDLE (0x7FFFU) 545 546/* 547 * Send an array of these to HYPERVISOR_mmu_update(). 548 * NB. The fields are natural pointer/address size for this architecture. 549 */ 550struct mmu_update { 551 uint64_t ptr; /* Machine address of PTE. */ 552 uint64_t val; /* New contents of PTE. */ 553}; 554typedef struct mmu_update mmu_update_t; 555DEFINE_XEN_GUEST_HANDLE(mmu_update_t); 556 557/* 558 * ` enum neg_errnoval 559 * ` HYPERVISOR_multicall(multicall_entry_t call_list[], 560 * ` uint32_t nr_calls); 561 * 562 * NB. The fields are logically the natural register size for this 563 * architecture. In cases where xen_ulong_t is larger than this then 564 * any unused bits in the upper portion must be zero. 565 */ 566struct multicall_entry { 567 xen_ulong_t op, result; 568 xen_ulong_t args[6]; 569}; 570typedef struct multicall_entry multicall_entry_t; 571DEFINE_XEN_GUEST_HANDLE(multicall_entry_t); 572 573#if __XEN_INTERFACE_VERSION__ < 0x00040400 574/* 575 * Event channel endpoints per domain (when using the 2-level ABI): 576 * 1024 if a long is 32 bits; 4096 if a long is 64 bits. 577 */ 578#define NR_EVENT_CHANNELS EVTCHN_2L_NR_CHANNELS 579#endif 580 581struct vcpu_time_info { 582 /* 583 * Updates to the following values are preceded and followed by an 584 * increment of 'version'. The guest can therefore detect updates by 585 * looking for changes to 'version'. If the least-significant bit of 586 * the version number is set then an update is in progress and the guest 587 * must wait to read a consistent set of values. 588 * The correct way to interact with the version number is similar to 589 * Linux's seqlock: see the implementations of read_seqbegin/read_seqretry. 590 */ 591 uint32_t version; 592 uint32_t pad0; 593 uint64_t tsc_timestamp; /* TSC at last update of time vals. */ 594 uint64_t system_time; /* Time, in nanosecs, since boot. */ 595 /* 596 * Current system time: 597 * system_time + 598 * ((((tsc - tsc_timestamp) << tsc_shift) * tsc_to_system_mul) >> 32) 599 * CPU frequency (Hz): 600 * ((10^9 << 32) / tsc_to_system_mul) >> tsc_shift 601 */ 602 uint32_t tsc_to_system_mul; 603 int8_t tsc_shift; 604 int8_t pad1[3]; 605}; /* 32 bytes */ 606typedef struct vcpu_time_info vcpu_time_info_t; 607 608struct vcpu_info { 609 /* 610 * 'evtchn_upcall_pending' is written non-zero by Xen to indicate 611 * a pending notification for a particular VCPU. It is then cleared 612 * by the guest OS /before/ checking for pending work, thus avoiding 613 * a set-and-check race. Note that the mask is only accessed by Xen 614 * on the CPU that is currently hosting the VCPU. This means that the 615 * pending and mask flags can be updated by the guest without special 616 * synchronisation (i.e., no need for the x86 LOCK prefix). 617 * This may seem suboptimal because if the pending flag is set by 618 * a different CPU then an IPI may be scheduled even when the mask 619 * is set. However, note: 620 * 1. The task of 'interrupt holdoff' is covered by the per-event- 621 * channel mask bits. A 'noisy' event that is continually being 622 * triggered can be masked at source at this very precise 623 * granularity. 624 * 2. The main purpose of the per-VCPU mask is therefore to restrict 625 * reentrant execution: whether for concurrency control, or to 626 * prevent unbounded stack usage. Whatever the purpose, we expect 627 * that the mask will be asserted only for short periods at a time, 628 * and so the likelihood of a 'spurious' IPI is suitably small. 629 * The mask is read before making an event upcall to the guest: a 630 * non-zero mask therefore guarantees that the VCPU will not receive 631 * an upcall activation. The mask is cleared when the VCPU requests 632 * to block: this avoids wakeup-waiting races. 633 */ 634 uint8_t evtchn_upcall_pending; 635#ifdef XEN_HAVE_PV_UPCALL_MASK 636 uint8_t evtchn_upcall_mask; 637#else /* XEN_HAVE_PV_UPCALL_MASK */ 638 uint8_t pad0; 639#endif /* XEN_HAVE_PV_UPCALL_MASK */ 640 xen_ulong_t evtchn_pending_sel; 641 struct arch_vcpu_info arch; 642 struct vcpu_time_info time; 643}; /* 64 bytes (x86) */ 644#ifndef __XEN__ 645typedef struct vcpu_info vcpu_info_t; 646#endif 647 648/* 649 * `incontents 200 startofday_shared Start-of-day shared data structure 650 * Xen/kernel shared data -- pointer provided in start_info. 651 * 652 * This structure is defined to be both smaller than a page, and the 653 * only data on the shared page, but may vary in actual size even within 654 * compatible Xen versions; guests should not rely on the size 655 * of this structure remaining constant. 656 */ 657struct shared_info { 658 struct vcpu_info vcpu_info[XEN_LEGACY_MAX_VCPUS]; 659 660 /* 661 * A domain can create "event channels" on which it can send and receive 662 * asynchronous event notifications. There are three classes of event that 663 * are delivered by this mechanism: 664 * 1. Bi-directional inter- and intra-domain connections. Domains must 665 * arrange out-of-band to set up a connection (usually by allocating 666 * an unbound 'listener' port and avertising that via a storage service 667 * such as xenstore). 668 * 2. Physical interrupts. A domain with suitable hardware-access 669 * privileges can bind an event-channel port to a physical interrupt 670 * source. 671 * 3. Virtual interrupts ('events'). A domain can bind an event-channel 672 * port to a virtual interrupt source, such as the virtual-timer 673 * device or the emergency console. 674 * 675 * Event channels are addressed by a "port index". Each channel is 676 * associated with two bits of information: 677 * 1. PENDING -- notifies the domain that there is a pending notification 678 * to be processed. This bit is cleared by the guest. 679 * 2. MASK -- if this bit is clear then a 0->1 transition of PENDING 680 * will cause an asynchronous upcall to be scheduled. This bit is only 681 * updated by the guest. It is read-only within Xen. If a channel 682 * becomes pending while the channel is masked then the 'edge' is lost 683 * (i.e., when the channel is unmasked, the guest must manually handle 684 * pending notifications as no upcall will be scheduled by Xen). 685 * 686 * To expedite scanning of pending notifications, any 0->1 pending 687 * transition on an unmasked channel causes a corresponding bit in a 688 * per-vcpu selector word to be set. Each bit in the selector covers a 689 * 'C long' in the PENDING bitfield array. 690 */ 691 xen_ulong_t evtchn_pending[sizeof(xen_ulong_t) * 8]; 692 xen_ulong_t evtchn_mask[sizeof(xen_ulong_t) * 8]; 693 694 /* 695 * Wallclock time: updated only by control software. Guests should base 696 * their gettimeofday() syscall on this wallclock-base value. 697 */ 698 uint32_t wc_version; /* Version counter: see vcpu_time_info_t. */ 699 uint32_t wc_sec; /* Secs 00:00:00 UTC, Jan 1, 1970. */ 700 uint32_t wc_nsec; /* Nsecs 00:00:00 UTC, Jan 1, 1970. */ 701#if !defined(__i386__) 702 uint32_t wc_sec_hi; 703# define xen_wc_sec_hi wc_sec_hi 704#elif !defined(__XEN__) && !defined(__XEN_TOOLS__) 705# define xen_wc_sec_hi arch.wc_sec_hi 706#endif 707 708 struct arch_shared_info arch; 709 710}; 711#ifndef __XEN__ 712typedef struct shared_info shared_info_t; 713#endif 714 715/* 716 * `incontents 200 startofday Start-of-day memory layout 717 * 718 * 1. The domain is started within contiguous virtual-memory region. 719 * 2. The contiguous region ends on an aligned 4MB boundary. 720 * 3. This the order of bootstrap elements in the initial virtual region: 721 * a. relocated kernel image 722 * b. initial ram disk [mod_start, mod_len] 723 * (may be omitted) 724 * c. list of allocated page frames [mfn_list, nr_pages] 725 * (unless relocated due to XEN_ELFNOTE_INIT_P2M) 726 * d. start_info_t structure [register ESI (x86)] 727 * in case of dom0 this page contains the console info, too 728 * e. unless dom0: xenstore ring page 729 * f. unless dom0: console ring page 730 * g. bootstrap page tables [pt_base and CR3 (x86)] 731 * h. bootstrap stack [register ESP (x86)] 732 * 4. Bootstrap elements are packed together, but each is 4kB-aligned. 733 * 5. The list of page frames forms a contiguous 'pseudo-physical' memory 734 * layout for the domain. In particular, the bootstrap virtual-memory 735 * region is a 1:1 mapping to the first section of the pseudo-physical map. 736 * 6. All bootstrap elements are mapped read-writable for the guest OS. The 737 * only exception is the bootstrap page table, which is mapped read-only. 738 * 7. There is guaranteed to be at least 512kB padding after the final 739 * bootstrap element. If necessary, the bootstrap virtual region is 740 * extended by an extra 4MB to ensure this. 741 * 742 * Note: Prior to 25833:bb85bbccb1c9. ("x86/32-on-64 adjust Dom0 initial page 743 * table layout") a bug caused the pt_base (3.g above) and cr3 to not point 744 * to the start of the guest page tables (it was offset by two pages). 745 * This only manifested itself on 32-on-64 dom0 kernels and not 32-on-64 domU 746 * or 64-bit kernels of any colour. The page tables for a 32-on-64 dom0 got 747 * allocated in the order: 'first L1','first L2', 'first L3', so the offset 748 * to the page table base is by two pages back. The initial domain if it is 749 * 32-bit and runs under a 64-bit hypervisor should _NOT_ use two of the 750 * pages preceding pt_base and mark them as reserved/unused. 751 */ 752#ifdef XEN_HAVE_PV_GUEST_ENTRY 753struct start_info { 754 /* THE FOLLOWING ARE FILLED IN BOTH ON INITIAL BOOT AND ON RESUME. */ 755 char magic[32]; /* "xen-<version>-<platform>". */ 756 unsigned long nr_pages; /* Total pages allocated to this domain. */ 757 unsigned long shared_info; /* MACHINE address of shared info struct. */ 758 uint32_t flags; /* SIF_xxx flags. */ 759 xen_pfn_t store_mfn; /* MACHINE page number of shared page. */ 760 uint32_t store_evtchn; /* Event channel for store communication. */ 761 union { 762 struct { 763 xen_pfn_t mfn; /* MACHINE page number of console page. */ 764 uint32_t evtchn; /* Event channel for console page. */ 765 } domU; 766 struct { 767 uint32_t info_off; /* Offset of console_info struct. */ 768 uint32_t info_size; /* Size of console_info struct from start.*/ 769 } dom0; 770 } console; 771 /* THE FOLLOWING ARE ONLY FILLED IN ON INITIAL BOOT (NOT RESUME). */ 772 unsigned long pt_base; /* VIRTUAL address of page directory. */ 773 unsigned long nr_pt_frames; /* Number of bootstrap p.t. frames. */ 774 unsigned long mfn_list; /* VIRTUAL address of page-frame list. */ 775 unsigned long mod_start; /* VIRTUAL address of pre-loaded module */ 776 /* (PFN of pre-loaded module if */ 777 /* SIF_MOD_START_PFN set in flags). */ 778 unsigned long mod_len; /* Size (bytes) of pre-loaded module. */ 779#define MAX_GUEST_CMDLINE 1024 780 int8_t cmd_line[MAX_GUEST_CMDLINE]; 781 /* The pfn range here covers both page table and p->m table frames. */ 782 unsigned long first_p2m_pfn;/* 1st pfn forming initial P->M table. */ 783 unsigned long nr_p2m_frames;/* # of pfns forming initial P->M table. */ 784}; 785typedef struct start_info start_info_t; 786 787/* New console union for dom0 introduced in 0x00030203. */ 788#if __XEN_INTERFACE_VERSION__ < 0x00030203 789#define console_mfn console.domU.mfn 790#define console_evtchn console.domU.evtchn 791#endif 792#endif /* XEN_HAVE_PV_GUEST_ENTRY */ 793 794/* These flags are passed in the 'flags' field of start_info_t. */ 795#define SIF_PRIVILEGED (1<<0) /* Is the domain privileged? */ 796#define SIF_INITDOMAIN (1<<1) /* Is this the initial control domain? */ 797#define SIF_MULTIBOOT_MOD (1<<2) /* Is mod_start a multiboot module? */ 798#define SIF_MOD_START_PFN (1<<3) /* Is mod_start a PFN? */ 799#define SIF_VIRT_P2M_4TOOLS (1<<4) /* Do Xen tools understand a virt. mapped */ 800 /* P->M making the 3 level tree obsolete? */ 801#define SIF_PM_MASK (0xFF<<8) /* reserve 1 byte for xen-pm options */ 802 803/* 804 * A multiboot module is a package containing modules very similar to a 805 * multiboot module array. The only differences are: 806 * - the array of module descriptors is by convention simply at the beginning 807 * of the multiboot module, 808 * - addresses in the module descriptors are based on the beginning of the 809 * multiboot module, 810 * - the number of modules is determined by a termination descriptor that has 811 * mod_start == 0. 812 * 813 * This permits to both build it statically and reference it in a configuration 814 * file, and let the PV guest easily rebase the addresses to virtual addresses 815 * and at the same time count the number of modules. 816 */ 817struct xen_multiboot_mod_list 818{ 819 /* Address of first byte of the module */ 820 uint32_t mod_start; 821 /* Address of last byte of the module (inclusive) */ 822 uint32_t mod_end; 823 /* Address of zero-terminated command line */ 824 uint32_t cmdline; 825 /* Unused, must be zero */ 826 uint32_t pad; 827}; 828/* 829 * `incontents 200 startofday_dom0_console Dom0_console 830 * 831 * The console structure in start_info.console.dom0 832 * 833 * This structure includes a variety of information required to 834 * have a working VGA/VESA console. 835 */ 836typedef struct dom0_vga_console_info { 837 uint8_t video_type; /* DOM0_VGA_CONSOLE_??? */ 838#define XEN_VGATYPE_TEXT_MODE_3 0x03 839#define XEN_VGATYPE_VESA_LFB 0x23 840#define XEN_VGATYPE_EFI_LFB 0x70 841 842 union { 843 struct { 844 /* Font height, in pixels. */ 845 uint16_t font_height; 846 /* Cursor location (column, row). */ 847 uint16_t cursor_x, cursor_y; 848 /* Number of rows and columns (dimensions in characters). */ 849 uint16_t rows, columns; 850 } text_mode_3; 851 852 struct { 853 /* Width and height, in pixels. */ 854 uint16_t width, height; 855 /* Bytes per scan line. */ 856 uint16_t bytes_per_line; 857 /* Bits per pixel. */ 858 uint16_t bits_per_pixel; 859 /* LFB physical address, and size (in units of 64kB). */ 860 uint32_t lfb_base; 861 uint32_t lfb_size; 862 /* RGB mask offsets and sizes, as defined by VBE 1.2+ */ 863 uint8_t red_pos, red_size; 864 uint8_t green_pos, green_size; 865 uint8_t blue_pos, blue_size; 866 uint8_t rsvd_pos, rsvd_size; 867#if __XEN_INTERFACE_VERSION__ >= 0x00030206 868 /* VESA capabilities (offset 0xa, VESA command 0x4f00). */ 869 uint32_t gbl_caps; 870 /* Mode attributes (offset 0x0, VESA command 0x4f01). */ 871 uint16_t mode_attrs; 872#endif 873 } vesa_lfb; 874 } u; 875} dom0_vga_console_info_t; 876#define xen_vga_console_info dom0_vga_console_info 877#define xen_vga_console_info_t dom0_vga_console_info_t 878 879typedef uint8_t xen_domain_handle_t[16]; 880 881/* Turn a plain number into a C unsigned long constant. */ 882#define __mk_unsigned_long(x) x ## UL 883#define mk_unsigned_long(x) __mk_unsigned_long(x) 884 885__DEFINE_XEN_GUEST_HANDLE(uint8, uint8_t); 886__DEFINE_XEN_GUEST_HANDLE(uint16, uint16_t); 887__DEFINE_XEN_GUEST_HANDLE(uint32, uint32_t); 888__DEFINE_XEN_GUEST_HANDLE(uint64, uint64_t); 889 890#else /* __ASSEMBLY__ */ 891 892/* In assembly code we cannot use C numeric constant suffixes. */ 893#define mk_unsigned_long(x) x 894 895#endif /* !__ASSEMBLY__ */ 896 897/* Default definitions for macros used by domctl/sysctl. */ 898#if defined(__XEN__) || defined(__XEN_TOOLS__) 899 900#ifndef int64_aligned_t 901#define int64_aligned_t int64_t 902#endif 903#ifndef uint64_aligned_t 904#define uint64_aligned_t uint64_t 905#endif 906#ifndef XEN_GUEST_HANDLE_64 907#define XEN_GUEST_HANDLE_64(name) XEN_GUEST_HANDLE(name) 908#endif 909 910#ifndef __ASSEMBLY__ 911struct xenctl_bitmap { 912 XEN_GUEST_HANDLE_64(uint8) bitmap; 913 uint32_t nr_bits; 914}; 915#endif 916 917#endif /* defined(__XEN__) || defined(__XEN_TOOLS__) */ 918 919#endif /* __XEN_PUBLIC_XEN_H__ */ 920 921/* 922 * Local variables: 923 * mode: C 924 * c-file-style: "BSD" 925 * c-basic-offset: 4 926 * tab-width: 4 927 * indent-tabs-mode: nil 928 * End: 929 */ 930