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. */ 42 DEFINE_XEN_GUEST_HANDLE(char); 43 __DEFINE_XEN_GUEST_HANDLE(uchar, unsigned char); 44 DEFINE_XEN_GUEST_HANDLE(int); 45 __DEFINE_XEN_GUEST_HANDLE(uint, unsigned int); 46 #if __XEN_INTERFACE_VERSION__ < 0x00040300 47 DEFINE_XEN_GUEST_HANDLE(long); 48 __DEFINE_XEN_GUEST_HANDLE(ulong, unsigned long); 49 #endif 50 DEFINE_XEN_GUEST_HANDLE(void); 51 52 DEFINE_XEN_GUEST_HANDLE(uint64_t); 53 DEFINE_XEN_GUEST_HANDLE(xen_pfn_t); 54 DEFINE_XEN_GUEST_HANDLE(xen_ulong_t); 55 56 /* Turn a plain number into a C unsigned (long (long)) constant. */ 57 #define __xen_mk_uint(x) x ## U 58 #define __xen_mk_ulong(x) x ## UL 59 #ifndef __xen_mk_ullong 60 # define __xen_mk_ullong(x) x ## ULL 61 #endif 62 #define xen_mk_uint(x) __xen_mk_uint(x) 63 #define xen_mk_ulong(x) __xen_mk_ulong(x) 64 #define xen_mk_ullong(x) __xen_mk_ullong(x) 65 66 #else 67 68 /* In assembly code we cannot use C numeric constant suffixes. */ 69 #define xen_mk_uint(x) x 70 #define xen_mk_ulong(x) x 71 #define xen_mk_ullong(x) x 72 73 #endif 74 75 /* 76 * HYPERCALLS 77 */ 78 79 /* `incontents 100 hcalls List of hypercalls 80 * ` enum hypercall_num { // __HYPERVISOR_* => HYPERVISOR_*() 81 */ 82 83 #define __HYPERVISOR_set_trap_table 0 84 #define __HYPERVISOR_mmu_update 1 85 #define __HYPERVISOR_set_gdt 2 86 #define __HYPERVISOR_stack_switch 3 87 #define __HYPERVISOR_set_callbacks 4 88 #define __HYPERVISOR_fpu_taskswitch 5 89 #define __HYPERVISOR_sched_op_compat 6 /* compat since 0x00030101 */ 90 #define __HYPERVISOR_platform_op 7 91 #define __HYPERVISOR_set_debugreg 8 92 #define __HYPERVISOR_get_debugreg 9 93 #define __HYPERVISOR_update_descriptor 10 94 #define __HYPERVISOR_memory_op 12 95 #define __HYPERVISOR_multicall 13 96 #define __HYPERVISOR_update_va_mapping 14 97 #define __HYPERVISOR_set_timer_op 15 98 #define __HYPERVISOR_event_channel_op_compat 16 /* compat since 0x00030202 */ 99 #define __HYPERVISOR_xen_version 17 100 #define __HYPERVISOR_console_io 18 101 #define __HYPERVISOR_physdev_op_compat 19 /* compat since 0x00030202 */ 102 #define __HYPERVISOR_grant_table_op 20 103 #define __HYPERVISOR_vm_assist 21 104 #define __HYPERVISOR_update_va_mapping_otherdomain 22 105 #define __HYPERVISOR_iret 23 /* x86 only */ 106 #define __HYPERVISOR_vcpu_op 24 107 #define __HYPERVISOR_set_segment_base 25 /* x86/64 only */ 108 #define __HYPERVISOR_mmuext_op 26 109 #define __HYPERVISOR_xsm_op 27 110 #define __HYPERVISOR_nmi_op 28 111 #define __HYPERVISOR_sched_op 29 112 #define __HYPERVISOR_callback_op 30 113 #define __HYPERVISOR_xenoprof_op 31 114 #define __HYPERVISOR_event_channel_op 32 115 #define __HYPERVISOR_physdev_op 33 116 #define __HYPERVISOR_hvm_op 34 117 #define __HYPERVISOR_sysctl 35 118 #define __HYPERVISOR_domctl 36 119 #define __HYPERVISOR_kexec_op 37 120 #define __HYPERVISOR_tmem_op 38 121 #define __HYPERVISOR_xc_reserved_op 39 /* reserved for XenClient */ 122 #define __HYPERVISOR_xenpmu_op 40 123 #define __HYPERVISOR_dm_op 41 124 125 /* Architecture-specific hypercall definitions. */ 126 #define __HYPERVISOR_arch_0 48 127 #define __HYPERVISOR_arch_1 49 128 #define __HYPERVISOR_arch_2 50 129 #define __HYPERVISOR_arch_3 51 130 #define __HYPERVISOR_arch_4 52 131 #define __HYPERVISOR_arch_5 53 132 #define __HYPERVISOR_arch_6 54 133 #define __HYPERVISOR_arch_7 55 134 135 /* ` } */ 136 137 /* 138 * HYPERCALL COMPATIBILITY. 139 */ 140 141 /* New sched_op hypercall introduced in 0x00030101. */ 142 #if __XEN_INTERFACE_VERSION__ < 0x00030101 143 #undef __HYPERVISOR_sched_op 144 #define __HYPERVISOR_sched_op __HYPERVISOR_sched_op_compat 145 #endif 146 147 /* New event-channel and physdev hypercalls introduced in 0x00030202. */ 148 #if __XEN_INTERFACE_VERSION__ < 0x00030202 149 #undef __HYPERVISOR_event_channel_op 150 #define __HYPERVISOR_event_channel_op __HYPERVISOR_event_channel_op_compat 151 #undef __HYPERVISOR_physdev_op 152 #define __HYPERVISOR_physdev_op __HYPERVISOR_physdev_op_compat 153 #endif 154 155 /* New platform_op hypercall introduced in 0x00030204. */ 156 #if __XEN_INTERFACE_VERSION__ < 0x00030204 157 #define __HYPERVISOR_dom0_op __HYPERVISOR_platform_op 158 #endif 159 160 /* 161 * VIRTUAL INTERRUPTS 162 * 163 * Virtual interrupts that a guest OS may receive from Xen. 164 * 165 * In the side comments, 'V.' denotes a per-VCPU VIRQ while 'G.' denotes a 166 * global VIRQ. The former can be bound once per VCPU and cannot be re-bound. 167 * The latter can be allocated only once per guest: they must initially be 168 * allocated to VCPU0 but can subsequently be re-bound. 169 */ 170 /* ` enum virq { */ 171 #define VIRQ_TIMER 0 /* V. Timebase update, and/or requested timeout. */ 172 #define VIRQ_DEBUG 1 /* V. Request guest to dump debug info. */ 173 #define VIRQ_CONSOLE 2 /* G. (DOM0) Bytes received on emergency console. */ 174 #define VIRQ_DOM_EXC 3 /* G. (DOM0) Exceptional event for some domain. */ 175 #define VIRQ_TBUF 4 /* G. (DOM0) Trace buffer has records available. */ 176 #define VIRQ_DEBUGGER 6 /* G. (DOM0) A domain has paused for debugging. */ 177 #define VIRQ_XENOPROF 7 /* V. XenOprofile interrupt: new sample available */ 178 #define VIRQ_CON_RING 8 /* G. (DOM0) Bytes received on console */ 179 #define VIRQ_PCPU_STATE 9 /* G. (DOM0) PCPU state changed */ 180 #define VIRQ_MEM_EVENT 10 /* G. (DOM0) A memory event has occured */ 181 #define VIRQ_XC_RESERVED 11 /* G. Reserved for XenClient */ 182 #define VIRQ_ENOMEM 12 /* G. (DOM0) Low on heap memory */ 183 #define VIRQ_XENPMU 13 /* V. PMC interrupt */ 184 185 /* Architecture-specific VIRQ definitions. */ 186 #define VIRQ_ARCH_0 16 187 #define VIRQ_ARCH_1 17 188 #define VIRQ_ARCH_2 18 189 #define VIRQ_ARCH_3 19 190 #define VIRQ_ARCH_4 20 191 #define VIRQ_ARCH_5 21 192 #define VIRQ_ARCH_6 22 193 #define VIRQ_ARCH_7 23 194 /* ` } */ 195 196 #define NR_VIRQS 24 197 198 /* 199 * ` enum neg_errnoval 200 * ` HYPERVISOR_mmu_update(const struct mmu_update reqs[], 201 * ` unsigned count, unsigned *done_out, 202 * ` unsigned foreigndom) 203 * ` 204 * @reqs is an array of mmu_update_t structures ((ptr, val) pairs). 205 * @count is the length of the above array. 206 * @pdone is an output parameter indicating number of completed operations 207 * @foreigndom[15:0]: FD, the expected owner of data pages referenced in this 208 * hypercall invocation. Can be DOMID_SELF. 209 * @foreigndom[31:16]: PFD, the expected owner of pagetable pages referenced 210 * in this hypercall invocation. The value of this field 211 * (x) encodes the PFD as follows: 212 * x == 0 => PFD == DOMID_SELF 213 * x != 0 => PFD == x - 1 214 * 215 * Sub-commands: ptr[1:0] specifies the appropriate MMU_* command. 216 * ------------- 217 * ptr[1:0] == MMU_NORMAL_PT_UPDATE: 218 * Updates an entry in a page table belonging to PFD. If updating an L1 table, 219 * and the new table entry is valid/present, the mapped frame must belong to 220 * FD. If attempting to map an I/O page then the caller assumes the privilege 221 * of the FD. 222 * FD == DOMID_IO: Permit /only/ I/O mappings, at the priv level of the caller. 223 * FD == DOMID_XEN: Map restricted areas of Xen's heap space. 224 * ptr[:2] -- Machine address of the page-table entry to modify. 225 * val -- Value to write. 226 * 227 * There also certain implicit requirements when using this hypercall. The 228 * pages that make up a pagetable must be mapped read-only in the guest. 229 * This prevents uncontrolled guest updates to the pagetable. Xen strictly 230 * enforces this, and will disallow any pagetable update which will end up 231 * mapping pagetable page RW, and will disallow using any writable page as a 232 * pagetable. In practice it means that when constructing a page table for a 233 * process, thread, etc, we MUST be very dilligient in following these rules: 234 * 1). Start with top-level page (PGD or in Xen language: L4). Fill out 235 * the entries. 236 * 2). Keep on going, filling out the upper (PUD or L3), and middle (PMD 237 * or L2). 238 * 3). Start filling out the PTE table (L1) with the PTE entries. Once 239 * done, make sure to set each of those entries to RO (so writeable bit 240 * is unset). Once that has been completed, set the PMD (L2) for this 241 * PTE table as RO. 242 * 4). When completed with all of the PMD (L2) entries, and all of them have 243 * been set to RO, make sure to set RO the PUD (L3). Do the same 244 * operation on PGD (L4) pagetable entries that have a PUD (L3) entry. 245 * 5). Now before you can use those pages (so setting the cr3), you MUST also 246 * pin them so that the hypervisor can verify the entries. This is done 247 * via the HYPERVISOR_mmuext_op(MMUEXT_PIN_L4_TABLE, guest physical frame 248 * number of the PGD (L4)). And this point the HYPERVISOR_mmuext_op( 249 * MMUEXT_NEW_BASEPTR, guest physical frame number of the PGD (L4)) can be 250 * issued. 251 * For 32-bit guests, the L4 is not used (as there is less pagetables), so 252 * instead use L3. 253 * At this point the pagetables can be modified using the MMU_NORMAL_PT_UPDATE 254 * hypercall. Also if so desired the OS can also try to write to the PTE 255 * and be trapped by the hypervisor (as the PTE entry is RO). 256 * 257 * To deallocate the pages, the operations are the reverse of the steps 258 * mentioned above. The argument is MMUEXT_UNPIN_TABLE for all levels and the 259 * pagetable MUST not be in use (meaning that the cr3 is not set to it). 260 * 261 * ptr[1:0] == MMU_MACHPHYS_UPDATE: 262 * Updates an entry in the machine->pseudo-physical mapping table. 263 * ptr[:2] -- Machine address within the frame whose mapping to modify. 264 * The frame must belong to the FD, if one is specified. 265 * val -- Value to write into the mapping entry. 266 * 267 * ptr[1:0] == MMU_PT_UPDATE_PRESERVE_AD: 268 * As MMU_NORMAL_PT_UPDATE above, but A/D bits currently in the PTE are ORed 269 * with those in @val. 270 * 271 * @val is usually the machine frame number along with some attributes. 272 * The attributes by default follow the architecture defined bits. Meaning that 273 * if this is a X86_64 machine and four page table layout is used, the layout 274 * of val is: 275 * - 63 if set means No execute (NX) 276 * - 46-13 the machine frame number 277 * - 12 available for guest 278 * - 11 available for guest 279 * - 10 available for guest 280 * - 9 available for guest 281 * - 8 global 282 * - 7 PAT (PSE is disabled, must use hypercall to make 4MB or 2MB pages) 283 * - 6 dirty 284 * - 5 accessed 285 * - 4 page cached disabled 286 * - 3 page write through 287 * - 2 userspace accessible 288 * - 1 writeable 289 * - 0 present 290 * 291 * The one bits that does not fit with the default layout is the PAGE_PSE 292 * also called PAGE_PAT). The MMUEXT_[UN]MARK_SUPER arguments to the 293 * HYPERVISOR_mmuext_op serve as mechanism to set a pagetable to be 4MB 294 * (or 2MB) instead of using the PAGE_PSE bit. 295 * 296 * The reason that the PAGE_PSE (bit 7) is not being utilized is due to Xen 297 * using it as the Page Attribute Table (PAT) bit - for details on it please 298 * refer to Intel SDM 10.12. The PAT allows to set the caching attributes of 299 * pages instead of using MTRRs. 300 * 301 * The PAT MSR is as follows (it is a 64-bit value, each entry is 8 bits): 302 * PAT4 PAT0 303 * +-----+-----+----+----+----+-----+----+----+ 304 * | UC | UC- | WC | WB | UC | UC- | WC | WB | <= Linux 305 * +-----+-----+----+----+----+-----+----+----+ 306 * | UC | UC- | WT | WB | UC | UC- | WT | WB | <= BIOS (default when machine boots) 307 * +-----+-----+----+----+----+-----+----+----+ 308 * | rsv | rsv | WP | WC | UC | UC- | WT | WB | <= Xen 309 * +-----+-----+----+----+----+-----+----+----+ 310 * 311 * The lookup of this index table translates to looking up 312 * Bit 7, Bit 4, and Bit 3 of val entry: 313 * 314 * PAT/PSE (bit 7) ... PCD (bit 4) .. PWT (bit 3). 315 * 316 * If all bits are off, then we are using PAT0. If bit 3 turned on, 317 * then we are using PAT1, if bit 3 and bit 4, then PAT2.. 318 * 319 * As you can see, the Linux PAT1 translates to PAT4 under Xen. Which means 320 * that if a guest that follows Linux's PAT setup and would like to set Write 321 * Combined on pages it MUST use PAT4 entry. Meaning that Bit 7 (PAGE_PAT) is 322 * set. For example, under Linux it only uses PAT0, PAT1, and PAT2 for the 323 * caching as: 324 * 325 * WB = none (so PAT0) 326 * WC = PWT (bit 3 on) 327 * UC = PWT | PCD (bit 3 and 4 are on). 328 * 329 * To make it work with Xen, it needs to translate the WC bit as so: 330 * 331 * PWT (so bit 3 on) --> PAT (so bit 7 is on) and clear bit 3 332 * 333 * And to translate back it would: 334 * 335 * PAT (bit 7 on) --> PWT (bit 3 on) and clear bit 7. 336 */ 337 #define MMU_NORMAL_PT_UPDATE 0 /* checked '*ptr = val'. ptr is MA. */ 338 #define MMU_MACHPHYS_UPDATE 1 /* ptr = MA of frame to modify entry for */ 339 #define MMU_PT_UPDATE_PRESERVE_AD 2 /* atomically: *ptr = val | (*ptr&(A|D)) */ 340 341 /* 342 * MMU EXTENDED OPERATIONS 343 * 344 * ` enum neg_errnoval 345 * ` HYPERVISOR_mmuext_op(mmuext_op_t uops[], 346 * ` unsigned int count, 347 * ` unsigned int *pdone, 348 * ` unsigned int foreigndom) 349 */ 350 /* HYPERVISOR_mmuext_op() accepts a list of mmuext_op structures. 351 * A foreigndom (FD) can be specified (or DOMID_SELF for none). 352 * Where the FD has some effect, it is described below. 353 * 354 * cmd: MMUEXT_(UN)PIN_*_TABLE 355 * mfn: Machine frame number to be (un)pinned as a p.t. page. 356 * The frame must belong to the FD, if one is specified. 357 * 358 * cmd: MMUEXT_NEW_BASEPTR 359 * mfn: Machine frame number of new page-table base to install in MMU. 360 * 361 * cmd: MMUEXT_NEW_USER_BASEPTR [x86/64 only] 362 * mfn: Machine frame number of new page-table base to install in MMU 363 * when in user space. 364 * 365 * cmd: MMUEXT_TLB_FLUSH_LOCAL 366 * No additional arguments. Flushes local TLB. 367 * 368 * cmd: MMUEXT_INVLPG_LOCAL 369 * linear_addr: Linear address to be flushed from the local TLB. 370 * 371 * cmd: MMUEXT_TLB_FLUSH_MULTI 372 * vcpumask: Pointer to bitmap of VCPUs to be flushed. 373 * 374 * cmd: MMUEXT_INVLPG_MULTI 375 * linear_addr: Linear address to be flushed. 376 * vcpumask: Pointer to bitmap of VCPUs to be flushed. 377 * 378 * cmd: MMUEXT_TLB_FLUSH_ALL 379 * No additional arguments. Flushes all VCPUs' TLBs. 380 * 381 * cmd: MMUEXT_INVLPG_ALL 382 * linear_addr: Linear address to be flushed from all VCPUs' TLBs. 383 * 384 * cmd: MMUEXT_FLUSH_CACHE 385 * No additional arguments. Writes back and flushes cache contents. 386 * 387 * cmd: MMUEXT_FLUSH_CACHE_GLOBAL 388 * No additional arguments. Writes back and flushes cache contents 389 * on all CPUs in the system. 390 * 391 * cmd: MMUEXT_SET_LDT 392 * linear_addr: Linear address of LDT base (NB. must be page-aligned). 393 * nr_ents: Number of entries in LDT. 394 * 395 * cmd: MMUEXT_CLEAR_PAGE 396 * mfn: Machine frame number to be cleared. 397 * 398 * cmd: MMUEXT_COPY_PAGE 399 * mfn: Machine frame number of the destination page. 400 * src_mfn: Machine frame number of the source page. 401 * 402 * cmd: MMUEXT_[UN]MARK_SUPER 403 * mfn: Machine frame number of head of superpage to be [un]marked. 404 */ 405 /* ` enum mmuext_cmd { */ 406 #define MMUEXT_PIN_L1_TABLE 0 407 #define MMUEXT_PIN_L2_TABLE 1 408 #define MMUEXT_PIN_L3_TABLE 2 409 #define MMUEXT_PIN_L4_TABLE 3 410 #define MMUEXT_UNPIN_TABLE 4 411 #define MMUEXT_NEW_BASEPTR 5 412 #define MMUEXT_TLB_FLUSH_LOCAL 6 413 #define MMUEXT_INVLPG_LOCAL 7 414 #define MMUEXT_TLB_FLUSH_MULTI 8 415 #define MMUEXT_INVLPG_MULTI 9 416 #define MMUEXT_TLB_FLUSH_ALL 10 417 #define MMUEXT_INVLPG_ALL 11 418 #define MMUEXT_FLUSH_CACHE 12 419 #define MMUEXT_SET_LDT 13 420 #define MMUEXT_NEW_USER_BASEPTR 15 421 #define MMUEXT_CLEAR_PAGE 16 422 #define MMUEXT_COPY_PAGE 17 423 #define MMUEXT_FLUSH_CACHE_GLOBAL 18 424 #define MMUEXT_MARK_SUPER 19 425 #define MMUEXT_UNMARK_SUPER 20 426 /* ` } */ 427 428 #ifndef __ASSEMBLY__ 429 struct mmuext_op { 430 unsigned int cmd; /* => enum mmuext_cmd */ 431 union { 432 /* [UN]PIN_TABLE, NEW_BASEPTR, NEW_USER_BASEPTR 433 * CLEAR_PAGE, COPY_PAGE, [UN]MARK_SUPER */ 434 xen_pfn_t mfn; 435 /* INVLPG_LOCAL, INVLPG_ALL, SET_LDT */ 436 unsigned long linear_addr; 437 } arg1; 438 union { 439 /* SET_LDT */ 440 unsigned int nr_ents; 441 /* TLB_FLUSH_MULTI, INVLPG_MULTI */ 442 #if __XEN_INTERFACE_VERSION__ >= 0x00030205 443 XEN_GUEST_HANDLE(const_void) vcpumask; 444 #else 445 const void *vcpumask; 446 #endif 447 /* COPY_PAGE */ 448 xen_pfn_t src_mfn; 449 } arg2; 450 }; 451 typedef struct mmuext_op mmuext_op_t; 452 DEFINE_XEN_GUEST_HANDLE(mmuext_op_t); 453 #endif 454 455 /* 456 * ` enum neg_errnoval 457 * ` HYPERVISOR_update_va_mapping(unsigned long va, u64 val, 458 * ` enum uvm_flags flags) 459 * ` 460 * ` enum neg_errnoval 461 * ` HYPERVISOR_update_va_mapping_otherdomain(unsigned long va, u64 val, 462 * ` enum uvm_flags flags, 463 * ` domid_t domid) 464 * ` 465 * ` @va: The virtual address whose mapping we want to change 466 * ` @val: The new page table entry, must contain a machine address 467 * ` @flags: Control TLB flushes 468 */ 469 /* These are passed as 'flags' to update_va_mapping. They can be ORed. */ 470 /* When specifying UVMF_MULTI, also OR in a pointer to a CPU bitmap. */ 471 /* UVMF_LOCAL is merely UVMF_MULTI with a NULL bitmap pointer. */ 472 /* ` enum uvm_flags { */ 473 #define UVMF_NONE (xen_mk_ulong(0)<<0) /* No flushing at all. */ 474 #define UVMF_TLB_FLUSH (xen_mk_ulong(1)<<0) /* Flush entire TLB(s). */ 475 #define UVMF_INVLPG (xen_mk_ulong(2)<<0) /* Flush only one entry. */ 476 #define UVMF_FLUSHTYPE_MASK (xen_mk_ulong(3)<<0) 477 #define UVMF_MULTI (xen_mk_ulong(0)<<2) /* Flush subset of TLBs. */ 478 #define UVMF_LOCAL (xen_mk_ulong(0)<<2) /* Flush local TLB. */ 479 #define UVMF_ALL (xen_mk_ulong(1)<<2) /* Flush all TLBs. */ 480 /* ` } */ 481 482 /* 483 * Commands to HYPERVISOR_console_io(). 484 */ 485 #define CONSOLEIO_write 0 486 #define CONSOLEIO_read 1 487 488 /* 489 * Commands to HYPERVISOR_vm_assist(). 490 */ 491 #define VMASST_CMD_enable 0 492 #define VMASST_CMD_disable 1 493 494 /* x86/32 guests: simulate full 4GB segment limits. */ 495 #define VMASST_TYPE_4gb_segments 0 496 497 /* x86/32 guests: trap (vector 15) whenever above vmassist is used. */ 498 #define VMASST_TYPE_4gb_segments_notify 1 499 500 /* 501 * x86 guests: support writes to bottom-level PTEs. 502 * NB1. Page-directory entries cannot be written. 503 * NB2. Guest must continue to remove all writable mappings of PTEs. 504 */ 505 #define VMASST_TYPE_writable_pagetables 2 506 507 /* x86/PAE guests: support PDPTs above 4GB. */ 508 #define VMASST_TYPE_pae_extended_cr3 3 509 510 /* 511 * x86 guests: Sane behaviour for virtual iopl 512 * - virtual iopl updated from do_iret() hypercalls. 513 * - virtual iopl reported in bounce frames. 514 * - guest kernels assumed to be level 0 for the purpose of iopl checks. 515 */ 516 #define VMASST_TYPE_architectural_iopl 4 517 518 /* 519 * All guests: activate update indicator in vcpu_runstate_info 520 * Enable setting the XEN_RUNSTATE_UPDATE flag in guest memory mapped 521 * vcpu_runstate_info during updates of the runstate information. 522 */ 523 #define VMASST_TYPE_runstate_update_flag 5 524 525 /* 526 * x86/64 guests: strictly hide M2P from user mode. 527 * This allows the guest to control respective hypervisor behavior: 528 * - when not set, L4 tables get created with the respective slot blank, 529 * and whenever the L4 table gets used as a kernel one the missing 530 * mapping gets inserted, 531 * - when set, L4 tables get created with the respective slot initialized 532 * as before, and whenever the L4 table gets used as a user one the 533 * mapping gets zapped. 534 */ 535 #define VMASST_TYPE_m2p_strict 32 536 537 #if __XEN_INTERFACE_VERSION__ < 0x00040600 538 #define MAX_VMASST_TYPE 3 539 #endif 540 541 /* Domain ids >= DOMID_FIRST_RESERVED cannot be used for ordinary domains. */ 542 #define DOMID_FIRST_RESERVED xen_mk_uint(0x7FF0) 543 544 /* DOMID_SELF is used in certain contexts to refer to oneself. */ 545 #define DOMID_SELF xen_mk_uint(0x7FF0) 546 547 /* 548 * DOMID_IO is used to restrict page-table updates to mapping I/O memory. 549 * Although no Foreign Domain need be specified to map I/O pages, DOMID_IO 550 * is useful to ensure that no mappings to the OS's own heap are accidentally 551 * installed. (e.g., in Linux this could cause havoc as reference counts 552 * aren't adjusted on the I/O-mapping code path). 553 * This only makes sense as HYPERVISOR_mmu_update()'s and 554 * HYPERVISOR_update_va_mapping_otherdomain()'s "foreigndom" argument. For 555 * HYPERVISOR_mmu_update() context it can be specified by any calling domain, 556 * otherwise it's only permitted if the caller is privileged. 557 */ 558 #define DOMID_IO xen_mk_uint(0x7FF1) 559 560 /* 561 * DOMID_XEN is used to allow privileged domains to map restricted parts of 562 * Xen's heap space (e.g., the machine_to_phys table). 563 * This only makes sense as 564 * - HYPERVISOR_mmu_update()'s, HYPERVISOR_mmuext_op()'s, or 565 * HYPERVISOR_update_va_mapping_otherdomain()'s "foreigndom" argument, 566 * - with XENMAPSPACE_gmfn_foreign, 567 * and is only permitted if the caller is privileged. 568 */ 569 #define DOMID_XEN xen_mk_uint(0x7FF2) 570 571 /* 572 * DOMID_COW is used as the owner of sharable pages */ 573 #define DOMID_COW xen_mk_uint(0x7FF3) 574 575 /* DOMID_INVALID is used to identify pages with unknown owner. */ 576 #define DOMID_INVALID xen_mk_uint(0x7FF4) 577 578 /* Idle domain. */ 579 #define DOMID_IDLE xen_mk_uint(0x7FFF) 580 581 #ifndef __ASSEMBLY__ 582 583 typedef uint16_t domid_t; 584 585 /* 586 * Send an array of these to HYPERVISOR_mmu_update(). 587 * NB. The fields are natural pointer/address size for this architecture. 588 */ 589 struct mmu_update { 590 uint64_t ptr; /* Machine address of PTE. */ 591 uint64_t val; /* New contents of PTE. */ 592 }; 593 typedef struct mmu_update mmu_update_t; 594 DEFINE_XEN_GUEST_HANDLE(mmu_update_t); 595 596 /* 597 * ` enum neg_errnoval 598 * ` HYPERVISOR_multicall(multicall_entry_t call_list[], 599 * ` uint32_t nr_calls); 600 * 601 * NB. The fields are logically the natural register size for this 602 * architecture. In cases where xen_ulong_t is larger than this then 603 * any unused bits in the upper portion must be zero. 604 */ 605 struct multicall_entry { 606 xen_ulong_t op, result; 607 xen_ulong_t args[6]; 608 }; 609 typedef struct multicall_entry multicall_entry_t; 610 DEFINE_XEN_GUEST_HANDLE(multicall_entry_t); 611 612 #if __XEN_INTERFACE_VERSION__ < 0x00040400 613 /* 614 * Event channel endpoints per domain (when using the 2-level ABI): 615 * 1024 if a long is 32 bits; 4096 if a long is 64 bits. 616 */ 617 #define NR_EVENT_CHANNELS EVTCHN_2L_NR_CHANNELS 618 #endif 619 620 struct vcpu_time_info { 621 /* 622 * Updates to the following values are preceded and followed by an 623 * increment of 'version'. The guest can therefore detect updates by 624 * looking for changes to 'version'. If the least-significant bit of 625 * the version number is set then an update is in progress and the guest 626 * must wait to read a consistent set of values. 627 * The correct way to interact with the version number is similar to 628 * Linux's seqlock: see the implementations of read_seqbegin/read_seqretry. 629 */ 630 uint32_t version; 631 uint32_t pad0; 632 uint64_t tsc_timestamp; /* TSC at last update of time vals. */ 633 uint64_t system_time; /* Time, in nanosecs, since boot. */ 634 /* 635 * Current system time: 636 * system_time + 637 * ((((tsc - tsc_timestamp) << tsc_shift) * tsc_to_system_mul) >> 32) 638 * CPU frequency (Hz): 639 * ((10^9 << 32) / tsc_to_system_mul) >> tsc_shift 640 */ 641 uint32_t tsc_to_system_mul; 642 int8_t tsc_shift; 643 #if __XEN_INTERFACE_VERSION__ > 0x040600 644 uint8_t flags; 645 uint8_t pad1[2]; 646 #else 647 int8_t pad1[3]; 648 #endif 649 }; /* 32 bytes */ 650 typedef struct vcpu_time_info vcpu_time_info_t; 651 652 #define XEN_PVCLOCK_TSC_STABLE_BIT (1 << 0) 653 #define XEN_PVCLOCK_GUEST_STOPPED (1 << 1) 654 655 struct vcpu_info { 656 /* 657 * 'evtchn_upcall_pending' is written non-zero by Xen to indicate 658 * a pending notification for a particular VCPU. It is then cleared 659 * by the guest OS /before/ checking for pending work, thus avoiding 660 * a set-and-check race. Note that the mask is only accessed by Xen 661 * on the CPU that is currently hosting the VCPU. This means that the 662 * pending and mask flags can be updated by the guest without special 663 * synchronisation (i.e., no need for the x86 LOCK prefix). 664 * This may seem suboptimal because if the pending flag is set by 665 * a different CPU then an IPI may be scheduled even when the mask 666 * is set. However, note: 667 * 1. The task of 'interrupt holdoff' is covered by the per-event- 668 * channel mask bits. A 'noisy' event that is continually being 669 * triggered can be masked at source at this very precise 670 * granularity. 671 * 2. The main purpose of the per-VCPU mask is therefore to restrict 672 * reentrant execution: whether for concurrency control, or to 673 * prevent unbounded stack usage. Whatever the purpose, we expect 674 * that the mask will be asserted only for short periods at a time, 675 * and so the likelihood of a 'spurious' IPI is suitably small. 676 * The mask is read before making an event upcall to the guest: a 677 * non-zero mask therefore guarantees that the VCPU will not receive 678 * an upcall activation. The mask is cleared when the VCPU requests 679 * to block: this avoids wakeup-waiting races. 680 */ 681 uint8_t evtchn_upcall_pending; 682 #ifdef XEN_HAVE_PV_UPCALL_MASK 683 uint8_t evtchn_upcall_mask; 684 #else /* XEN_HAVE_PV_UPCALL_MASK */ 685 uint8_t pad0; 686 #endif /* XEN_HAVE_PV_UPCALL_MASK */ 687 xen_ulong_t evtchn_pending_sel; 688 struct arch_vcpu_info arch; 689 struct vcpu_time_info time; 690 }; /* 64 bytes (x86) */ 691 #ifndef __XEN__ 692 typedef struct vcpu_info vcpu_info_t; 693 #endif 694 695 /* 696 * `incontents 200 startofday_shared Start-of-day shared data structure 697 * Xen/kernel shared data -- pointer provided in start_info. 698 * 699 * This structure is defined to be both smaller than a page, and the 700 * only data on the shared page, but may vary in actual size even within 701 * compatible Xen versions; guests should not rely on the size 702 * of this structure remaining constant. 703 */ 704 struct shared_info { 705 struct vcpu_info vcpu_info[XEN_LEGACY_MAX_VCPUS]; 706 707 /* 708 * A domain can create "event channels" on which it can send and receive 709 * asynchronous event notifications. There are three classes of event that 710 * are delivered by this mechanism: 711 * 1. Bi-directional inter- and intra-domain connections. Domains must 712 * arrange out-of-band to set up a connection (usually by allocating 713 * an unbound 'listener' port and avertising that via a storage service 714 * such as xenstore). 715 * 2. Physical interrupts. A domain with suitable hardware-access 716 * privileges can bind an event-channel port to a physical interrupt 717 * source. 718 * 3. Virtual interrupts ('events'). A domain can bind an event-channel 719 * port to a virtual interrupt source, such as the virtual-timer 720 * device or the emergency console. 721 * 722 * Event channels are addressed by a "port index". Each channel is 723 * associated with two bits of information: 724 * 1. PENDING -- notifies the domain that there is a pending notification 725 * to be processed. This bit is cleared by the guest. 726 * 2. MASK -- if this bit is clear then a 0->1 transition of PENDING 727 * will cause an asynchronous upcall to be scheduled. This bit is only 728 * updated by the guest. It is read-only within Xen. If a channel 729 * becomes pending while the channel is masked then the 'edge' is lost 730 * (i.e., when the channel is unmasked, the guest must manually handle 731 * pending notifications as no upcall will be scheduled by Xen). 732 * 733 * To expedite scanning of pending notifications, any 0->1 pending 734 * transition on an unmasked channel causes a corresponding bit in a 735 * per-vcpu selector word to be set. Each bit in the selector covers a 736 * 'C long' in the PENDING bitfield array. 737 */ 738 xen_ulong_t evtchn_pending[sizeof(xen_ulong_t) * 8]; 739 xen_ulong_t evtchn_mask[sizeof(xen_ulong_t) * 8]; 740 741 /* 742 * Wallclock time: updated only by control software. Guests should base 743 * their gettimeofday() syscall on this wallclock-base value. 744 */ 745 uint32_t wc_version; /* Version counter: see vcpu_time_info_t. */ 746 uint32_t wc_sec; /* Secs 00:00:00 UTC, Jan 1, 1970. */ 747 uint32_t wc_nsec; /* Nsecs 00:00:00 UTC, Jan 1, 1970. */ 748 #if !defined(__i386__) 749 uint32_t wc_sec_hi; 750 # define xen_wc_sec_hi wc_sec_hi 751 #elif !defined(__XEN__) && !defined(__XEN_TOOLS__) 752 # define xen_wc_sec_hi arch.wc_sec_hi 753 #endif 754 755 struct arch_shared_info arch; 756 757 }; 758 #ifndef __XEN__ 759 typedef struct shared_info shared_info_t; 760 #endif 761 762 /* 763 * `incontents 200 startofday Start-of-day memory layout 764 * 765 * 1. The domain is started within contiguous virtual-memory region. 766 * 2. The contiguous region ends on an aligned 4MB boundary. 767 * 3. This the order of bootstrap elements in the initial virtual region: 768 * a. relocated kernel image 769 * b. initial ram disk [mod_start, mod_len] 770 * (may be omitted) 771 * c. list of allocated page frames [mfn_list, nr_pages] 772 * (unless relocated due to XEN_ELFNOTE_INIT_P2M) 773 * d. start_info_t structure [register rSI (x86)] 774 * in case of dom0 this page contains the console info, too 775 * e. unless dom0: xenstore ring page 776 * f. unless dom0: console ring page 777 * g. bootstrap page tables [pt_base and CR3 (x86)] 778 * h. bootstrap stack [register ESP (x86)] 779 * 4. Bootstrap elements are packed together, but each is 4kB-aligned. 780 * 5. The list of page frames forms a contiguous 'pseudo-physical' memory 781 * layout for the domain. In particular, the bootstrap virtual-memory 782 * region is a 1:1 mapping to the first section of the pseudo-physical map. 783 * 6. All bootstrap elements are mapped read-writable for the guest OS. The 784 * only exception is the bootstrap page table, which is mapped read-only. 785 * 7. There is guaranteed to be at least 512kB padding after the final 786 * bootstrap element. If necessary, the bootstrap virtual region is 787 * extended by an extra 4MB to ensure this. 788 * 789 * Note: Prior to 25833:bb85bbccb1c9. ("x86/32-on-64 adjust Dom0 initial page 790 * table layout") a bug caused the pt_base (3.g above) and cr3 to not point 791 * to the start of the guest page tables (it was offset by two pages). 792 * This only manifested itself on 32-on-64 dom0 kernels and not 32-on-64 domU 793 * or 64-bit kernels of any colour. The page tables for a 32-on-64 dom0 got 794 * allocated in the order: 'first L1','first L2', 'first L3', so the offset 795 * to the page table base is by two pages back. The initial domain if it is 796 * 32-bit and runs under a 64-bit hypervisor should _NOT_ use two of the 797 * pages preceding pt_base and mark them as reserved/unused. 798 */ 799 #ifdef XEN_HAVE_PV_GUEST_ENTRY 800 struct start_info { 801 /* THE FOLLOWING ARE FILLED IN BOTH ON INITIAL BOOT AND ON RESUME. */ 802 char magic[32]; /* "xen-<version>-<platform>". */ 803 unsigned long nr_pages; /* Total pages allocated to this domain. */ 804 unsigned long shared_info; /* MACHINE address of shared info struct. */ 805 uint32_t flags; /* SIF_xxx flags. */ 806 xen_pfn_t store_mfn; /* MACHINE page number of shared page. */ 807 uint32_t store_evtchn; /* Event channel for store communication. */ 808 union { 809 struct { 810 xen_pfn_t mfn; /* MACHINE page number of console page. */ 811 uint32_t evtchn; /* Event channel for console page. */ 812 } domU; 813 struct { 814 uint32_t info_off; /* Offset of console_info struct. */ 815 uint32_t info_size; /* Size of console_info struct from start.*/ 816 } dom0; 817 } console; 818 /* THE FOLLOWING ARE ONLY FILLED IN ON INITIAL BOOT (NOT RESUME). */ 819 unsigned long pt_base; /* VIRTUAL address of page directory. */ 820 unsigned long nr_pt_frames; /* Number of bootstrap p.t. frames. */ 821 unsigned long mfn_list; /* VIRTUAL address of page-frame list. */ 822 unsigned long mod_start; /* VIRTUAL address of pre-loaded module */ 823 /* (PFN of pre-loaded module if */ 824 /* SIF_MOD_START_PFN set in flags). */ 825 unsigned long mod_len; /* Size (bytes) of pre-loaded module. */ 826 #define MAX_GUEST_CMDLINE 1024 827 int8_t cmd_line[MAX_GUEST_CMDLINE]; 828 /* The pfn range here covers both page table and p->m table frames. */ 829 unsigned long first_p2m_pfn;/* 1st pfn forming initial P->M table. */ 830 unsigned long nr_p2m_frames;/* # of pfns forming initial P->M table. */ 831 }; 832 typedef struct start_info start_info_t; 833 834 /* New console union for dom0 introduced in 0x00030203. */ 835 #if __XEN_INTERFACE_VERSION__ < 0x00030203 836 #define console_mfn console.domU.mfn 837 #define console_evtchn console.domU.evtchn 838 #endif 839 #endif /* XEN_HAVE_PV_GUEST_ENTRY */ 840 841 /* These flags are passed in the 'flags' field of start_info_t. */ 842 #define SIF_PRIVILEGED (1<<0) /* Is the domain privileged? */ 843 #define SIF_INITDOMAIN (1<<1) /* Is this the initial control domain? */ 844 #define SIF_MULTIBOOT_MOD (1<<2) /* Is mod_start a multiboot module? */ 845 #define SIF_MOD_START_PFN (1<<3) /* Is mod_start a PFN? */ 846 #define SIF_VIRT_P2M_4TOOLS (1<<4) /* Do Xen tools understand a virt. mapped */ 847 /* P->M making the 3 level tree obsolete? */ 848 #define SIF_PM_MASK (0xFF<<8) /* reserve 1 byte for xen-pm options */ 849 850 /* 851 * A multiboot module is a package containing modules very similar to a 852 * multiboot module array. The only differences are: 853 * - the array of module descriptors is by convention simply at the beginning 854 * of the multiboot module, 855 * - addresses in the module descriptors are based on the beginning of the 856 * multiboot module, 857 * - the number of modules is determined by a termination descriptor that has 858 * mod_start == 0. 859 * 860 * This permits to both build it statically and reference it in a configuration 861 * file, and let the PV guest easily rebase the addresses to virtual addresses 862 * and at the same time count the number of modules. 863 */ 864 struct xen_multiboot_mod_list 865 { 866 /* Address of first byte of the module */ 867 uint32_t mod_start; 868 /* Address of last byte of the module (inclusive) */ 869 uint32_t mod_end; 870 /* Address of zero-terminated command line */ 871 uint32_t cmdline; 872 /* Unused, must be zero */ 873 uint32_t pad; 874 }; 875 /* 876 * `incontents 200 startofday_dom0_console Dom0_console 877 * 878 * The console structure in start_info.console.dom0 879 * 880 * This structure includes a variety of information required to 881 * have a working VGA/VESA console. 882 */ 883 typedef struct dom0_vga_console_info { 884 uint8_t video_type; /* DOM0_VGA_CONSOLE_??? */ 885 #define XEN_VGATYPE_TEXT_MODE_3 0x03 886 #define XEN_VGATYPE_VESA_LFB 0x23 887 #define XEN_VGATYPE_EFI_LFB 0x70 888 889 union { 890 struct { 891 /* Font height, in pixels. */ 892 uint16_t font_height; 893 /* Cursor location (column, row). */ 894 uint16_t cursor_x, cursor_y; 895 /* Number of rows and columns (dimensions in characters). */ 896 uint16_t rows, columns; 897 } text_mode_3; 898 899 struct { 900 /* Width and height, in pixels. */ 901 uint16_t width, height; 902 /* Bytes per scan line. */ 903 uint16_t bytes_per_line; 904 /* Bits per pixel. */ 905 uint16_t bits_per_pixel; 906 /* LFB physical address, and size (in units of 64kB). */ 907 uint32_t lfb_base; 908 uint32_t lfb_size; 909 /* RGB mask offsets and sizes, as defined by VBE 1.2+ */ 910 uint8_t red_pos, red_size; 911 uint8_t green_pos, green_size; 912 uint8_t blue_pos, blue_size; 913 uint8_t rsvd_pos, rsvd_size; 914 #if __XEN_INTERFACE_VERSION__ >= 0x00030206 915 /* VESA capabilities (offset 0xa, VESA command 0x4f00). */ 916 uint32_t gbl_caps; 917 /* Mode attributes (offset 0x0, VESA command 0x4f01). */ 918 uint16_t mode_attrs; 919 #endif 920 } vesa_lfb; 921 } u; 922 } dom0_vga_console_info_t; 923 #define xen_vga_console_info dom0_vga_console_info 924 #define xen_vga_console_info_t dom0_vga_console_info_t 925 926 typedef uint8_t xen_domain_handle_t[16]; 927 928 __DEFINE_XEN_GUEST_HANDLE(uint8, uint8_t); 929 __DEFINE_XEN_GUEST_HANDLE(uint16, uint16_t); 930 __DEFINE_XEN_GUEST_HANDLE(uint32, uint32_t); 931 __DEFINE_XEN_GUEST_HANDLE(uint64, uint64_t); 932 933 typedef struct { 934 uint8_t a[16]; 935 } xen_uuid_t; 936 937 /* 938 * XEN_DEFINE_UUID(0x00112233, 0x4455, 0x6677, 0x8899, 939 * 0xaa, 0xbb, 0xcc, 0xdd, 0xee, 0xff) 940 * will construct UUID 00112233-4455-6677-8899-aabbccddeeff presented as 941 * {0x00, 0x11, 0x22, 0x33, 0x44, 0x55, 0x66, 0x77, 0x88, 942 * 0x99, 0xaa, 0xbb, 0xcc, 0xdd, 0xee, 0xff}; 943 * 944 * NB: This is compatible with Linux kernel and with libuuid, but it is not 945 * compatible with Microsoft, as they use mixed-endian encoding (some 946 * components are little-endian, some are big-endian). 947 */ 948 #define XEN_DEFINE_UUID_(a, b, c, d, e1, e2, e3, e4, e5, e6) \ 949 {{((a) >> 24) & 0xFF, ((a) >> 16) & 0xFF, \ 950 ((a) >> 8) & 0xFF, ((a) >> 0) & 0xFF, \ 951 ((b) >> 8) & 0xFF, ((b) >> 0) & 0xFF, \ 952 ((c) >> 8) & 0xFF, ((c) >> 0) & 0xFF, \ 953 ((d) >> 8) & 0xFF, ((d) >> 0) & 0xFF, \ 954 e1, e2, e3, e4, e5, e6}} 955 956 #if defined(__STDC_VERSION__) ? __STDC_VERSION__ >= 199901L : defined(__GNUC__) 957 #define XEN_DEFINE_UUID(a, b, c, d, e1, e2, e3, e4, e5, e6) \ 958 ((xen_uuid_t)XEN_DEFINE_UUID_(a, b, c, d, e1, e2, e3, e4, e5, e6)) 959 #else 960 #define XEN_DEFINE_UUID(a, b, c, d, e1, e2, e3, e4, e5, e6) \ 961 XEN_DEFINE_UUID_(a, b, c, d, e1, e2, e3, e4, e5, e6) 962 #endif /* __STDC_VERSION__ / __GNUC__ */ 963 964 #endif /* !__ASSEMBLY__ */ 965 966 /* Default definitions for macros used by domctl/sysctl. */ 967 #if defined(__XEN__) || defined(__XEN_TOOLS__) 968 969 #ifndef int64_aligned_t 970 #define int64_aligned_t int64_t 971 #endif 972 #ifndef uint64_aligned_t 973 #define uint64_aligned_t uint64_t 974 #endif 975 #ifndef XEN_GUEST_HANDLE_64 976 #define XEN_GUEST_HANDLE_64(name) XEN_GUEST_HANDLE(name) 977 #endif 978 979 #ifndef __ASSEMBLY__ 980 struct xenctl_bitmap { 981 XEN_GUEST_HANDLE_64(uint8) bitmap; 982 uint32_t nr_bits; 983 }; 984 #endif 985 986 #endif /* defined(__XEN__) || defined(__XEN_TOOLS__) */ 987 988 #endif /* __XEN_PUBLIC_XEN_H__ */ 989 990 /* 991 * Local variables: 992 * mode: C 993 * c-file-style: "BSD" 994 * c-basic-offset: 4 995 * tab-width: 4 996 * indent-tabs-mode: nil 997 * End: 998 */ 999