xref: /openbmc/linux/include/xen/interface/xen.h (revision bc5aa3a0)
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 <asm/xen/interface.h>
31 
32 /*
33  * XEN "SYSTEM CALLS" (a.k.a. HYPERCALLS).
34  */
35 
36 /*
37  * x86_32: EAX = vector; EBX, ECX, EDX, ESI, EDI = args 1, 2, 3, 4, 5.
38  *         EAX = return value
39  *         (argument registers may be clobbered on return)
40  * x86_64: RAX = vector; RDI, RSI, RDX, R10, R8, R9 = args 1, 2, 3, 4, 5, 6.
41  *         RAX = return value
42  *         (argument registers not clobbered on return; RCX, R11 are)
43  */
44 #define __HYPERVISOR_set_trap_table        0
45 #define __HYPERVISOR_mmu_update            1
46 #define __HYPERVISOR_set_gdt               2
47 #define __HYPERVISOR_stack_switch          3
48 #define __HYPERVISOR_set_callbacks         4
49 #define __HYPERVISOR_fpu_taskswitch        5
50 #define __HYPERVISOR_sched_op_compat       6
51 #define __HYPERVISOR_platform_op           7
52 #define __HYPERVISOR_set_debugreg          8
53 #define __HYPERVISOR_get_debugreg          9
54 #define __HYPERVISOR_update_descriptor    10
55 #define __HYPERVISOR_memory_op            12
56 #define __HYPERVISOR_multicall            13
57 #define __HYPERVISOR_update_va_mapping    14
58 #define __HYPERVISOR_set_timer_op         15
59 #define __HYPERVISOR_event_channel_op_compat 16
60 #define __HYPERVISOR_xen_version          17
61 #define __HYPERVISOR_console_io           18
62 #define __HYPERVISOR_physdev_op_compat    19
63 #define __HYPERVISOR_grant_table_op       20
64 #define __HYPERVISOR_vm_assist            21
65 #define __HYPERVISOR_update_va_mapping_otherdomain 22
66 #define __HYPERVISOR_iret                 23 /* x86 only */
67 #define __HYPERVISOR_vcpu_op              24
68 #define __HYPERVISOR_set_segment_base     25 /* x86/64 only */
69 #define __HYPERVISOR_mmuext_op            26
70 #define __HYPERVISOR_xsm_op               27
71 #define __HYPERVISOR_nmi_op               28
72 #define __HYPERVISOR_sched_op             29
73 #define __HYPERVISOR_callback_op          30
74 #define __HYPERVISOR_xenoprof_op          31
75 #define __HYPERVISOR_event_channel_op     32
76 #define __HYPERVISOR_physdev_op           33
77 #define __HYPERVISOR_hvm_op               34
78 #define __HYPERVISOR_sysctl               35
79 #define __HYPERVISOR_domctl               36
80 #define __HYPERVISOR_kexec_op             37
81 #define __HYPERVISOR_tmem_op              38
82 #define __HYPERVISOR_xc_reserved_op       39 /* reserved for XenClient */
83 #define __HYPERVISOR_xenpmu_op            40
84 
85 /* Architecture-specific hypercall definitions. */
86 #define __HYPERVISOR_arch_0               48
87 #define __HYPERVISOR_arch_1               49
88 #define __HYPERVISOR_arch_2               50
89 #define __HYPERVISOR_arch_3               51
90 #define __HYPERVISOR_arch_4               52
91 #define __HYPERVISOR_arch_5               53
92 #define __HYPERVISOR_arch_6               54
93 #define __HYPERVISOR_arch_7               55
94 
95 /*
96  * VIRTUAL INTERRUPTS
97  *
98  * Virtual interrupts that a guest OS may receive from Xen.
99  * In the side comments, 'V.' denotes a per-VCPU VIRQ while 'G.' denotes a
100  * global VIRQ. The former can be bound once per VCPU and cannot be re-bound.
101  * The latter can be allocated only once per guest: they must initially be
102  * allocated to VCPU0 but can subsequently be re-bound.
103  */
104 #define VIRQ_TIMER      0  /* V. Timebase update, and/or requested timeout.  */
105 #define VIRQ_DEBUG      1  /* V. Request guest to dump debug info.           */
106 #define VIRQ_CONSOLE    2  /* G. (DOM0) Bytes received on emergency console. */
107 #define VIRQ_DOM_EXC    3  /* G. (DOM0) Exceptional event for some domain.   */
108 #define VIRQ_TBUF       4  /* G. (DOM0) Trace buffer has records available.  */
109 #define VIRQ_DEBUGGER   6  /* G. (DOM0) A domain has paused for debugging.   */
110 #define VIRQ_XENOPROF   7  /* V. XenOprofile interrupt: new sample available */
111 #define VIRQ_CON_RING   8  /* G. (DOM0) Bytes received on console            */
112 #define VIRQ_PCPU_STATE 9  /* G. (DOM0) PCPU state changed                   */
113 #define VIRQ_MEM_EVENT  10 /* G. (DOM0) A memory event has occured           */
114 #define VIRQ_XC_RESERVED 11 /* G. Reserved for XenClient                     */
115 #define VIRQ_ENOMEM     12 /* G. (DOM0) Low on heap memory       */
116 #define VIRQ_XENPMU     13  /* PMC interrupt                                 */
117 
118 /* Architecture-specific VIRQ definitions. */
119 #define VIRQ_ARCH_0    16
120 #define VIRQ_ARCH_1    17
121 #define VIRQ_ARCH_2    18
122 #define VIRQ_ARCH_3    19
123 #define VIRQ_ARCH_4    20
124 #define VIRQ_ARCH_5    21
125 #define VIRQ_ARCH_6    22
126 #define VIRQ_ARCH_7    23
127 
128 #define NR_VIRQS       24
129 
130 /*
131  * enum neg_errnoval HYPERVISOR_mmu_update(const struct mmu_update reqs[],
132  *                                         unsigned count, unsigned *done_out,
133  *                                         unsigned foreigndom)
134  * @reqs is an array of mmu_update_t structures ((ptr, val) pairs).
135  * @count is the length of the above array.
136  * @pdone is an output parameter indicating number of completed operations
137  * @foreigndom[15:0]: FD, the expected owner of data pages referenced in this
138  *                    hypercall invocation. Can be DOMID_SELF.
139  * @foreigndom[31:16]: PFD, the expected owner of pagetable pages referenced
140  *                     in this hypercall invocation. The value of this field
141  *                     (x) encodes the PFD as follows:
142  *                     x == 0 => PFD == DOMID_SELF
143  *                     x != 0 => PFD == x - 1
144  *
145  * Sub-commands: ptr[1:0] specifies the appropriate MMU_* command.
146  * -------------
147  * ptr[1:0] == MMU_NORMAL_PT_UPDATE:
148  * Updates an entry in a page table belonging to PFD. If updating an L1 table,
149  * and the new table entry is valid/present, the mapped frame must belong to
150  * FD. If attempting to map an I/O page then the caller assumes the privilege
151  * of the FD.
152  * FD == DOMID_IO: Permit /only/ I/O mappings, at the priv level of the caller.
153  * FD == DOMID_XEN: Map restricted areas of Xen's heap space.
154  * ptr[:2]  -- Machine address of the page-table entry to modify.
155  * val      -- Value to write.
156  *
157  * There also certain implicit requirements when using this hypercall. The
158  * pages that make up a pagetable must be mapped read-only in the guest.
159  * This prevents uncontrolled guest updates to the pagetable. Xen strictly
160  * enforces this, and will disallow any pagetable update which will end up
161  * mapping pagetable page RW, and will disallow using any writable page as a
162  * pagetable. In practice it means that when constructing a page table for a
163  * process, thread, etc, we MUST be very dilligient in following these rules:
164  *  1). Start with top-level page (PGD or in Xen language: L4). Fill out
165  *      the entries.
166  *  2). Keep on going, filling out the upper (PUD or L3), and middle (PMD
167  *      or L2).
168  *  3). Start filling out the PTE table (L1) with the PTE entries. Once
169  *      done, make sure to set each of those entries to RO (so writeable bit
170  *      is unset). Once that has been completed, set the PMD (L2) for this
171  *      PTE table as RO.
172  *  4). When completed with all of the PMD (L2) entries, and all of them have
173  *      been set to RO, make sure to set RO the PUD (L3). Do the same
174  *      operation on PGD (L4) pagetable entries that have a PUD (L3) entry.
175  *  5). Now before you can use those pages (so setting the cr3), you MUST also
176  *      pin them so that the hypervisor can verify the entries. This is done
177  *      via the HYPERVISOR_mmuext_op(MMUEXT_PIN_L4_TABLE, guest physical frame
178  *      number of the PGD (L4)). And this point the HYPERVISOR_mmuext_op(
179  *      MMUEXT_NEW_BASEPTR, guest physical frame number of the PGD (L4)) can be
180  *      issued.
181  * For 32-bit guests, the L4 is not used (as there is less pagetables), so
182  * instead use L3.
183  * At this point the pagetables can be modified using the MMU_NORMAL_PT_UPDATE
184  * hypercall. Also if so desired the OS can also try to write to the PTE
185  * and be trapped by the hypervisor (as the PTE entry is RO).
186  *
187  * To deallocate the pages, the operations are the reverse of the steps
188  * mentioned above. The argument is MMUEXT_UNPIN_TABLE for all levels and the
189  * pagetable MUST not be in use (meaning that the cr3 is not set to it).
190  *
191  * ptr[1:0] == MMU_MACHPHYS_UPDATE:
192  * Updates an entry in the machine->pseudo-physical mapping table.
193  * ptr[:2]  -- Machine address within the frame whose mapping to modify.
194  *             The frame must belong to the FD, if one is specified.
195  * val      -- Value to write into the mapping entry.
196  *
197  * ptr[1:0] == MMU_PT_UPDATE_PRESERVE_AD:
198  * As MMU_NORMAL_PT_UPDATE above, but A/D bits currently in the PTE are ORed
199  * with those in @val.
200  *
201  * @val is usually the machine frame number along with some attributes.
202  * The attributes by default follow the architecture defined bits. Meaning that
203  * if this is a X86_64 machine and four page table layout is used, the layout
204  * of val is:
205  *  - 63 if set means No execute (NX)
206  *  - 46-13 the machine frame number
207  *  - 12 available for guest
208  *  - 11 available for guest
209  *  - 10 available for guest
210  *  - 9 available for guest
211  *  - 8 global
212  *  - 7 PAT (PSE is disabled, must use hypercall to make 4MB or 2MB pages)
213  *  - 6 dirty
214  *  - 5 accessed
215  *  - 4 page cached disabled
216  *  - 3 page write through
217  *  - 2 userspace accessible
218  *  - 1 writeable
219  *  - 0 present
220  *
221  *  The one bits that does not fit with the default layout is the PAGE_PSE
222  *  also called PAGE_PAT). The MMUEXT_[UN]MARK_SUPER arguments to the
223  *  HYPERVISOR_mmuext_op serve as mechanism to set a pagetable to be 4MB
224  *  (or 2MB) instead of using the PAGE_PSE bit.
225  *
226  *  The reason that the PAGE_PSE (bit 7) is not being utilized is due to Xen
227  *  using it as the Page Attribute Table (PAT) bit - for details on it please
228  *  refer to Intel SDM 10.12. The PAT allows to set the caching attributes of
229  *  pages instead of using MTRRs.
230  *
231  *  The PAT MSR is as follows (it is a 64-bit value, each entry is 8 bits):
232  *                    PAT4                 PAT0
233  *  +-----+-----+----+----+----+-----+----+----+
234  *  | UC  | UC- | WC | WB | UC | UC- | WC | WB |  <= Linux
235  *  +-----+-----+----+----+----+-----+----+----+
236  *  | UC  | UC- | WT | WB | UC | UC- | WT | WB |  <= BIOS (default when machine boots)
237  *  +-----+-----+----+----+----+-----+----+----+
238  *  | rsv | rsv | WP | WC | UC | UC- | WT | WB |  <= Xen
239  *  +-----+-----+----+----+----+-----+----+----+
240  *
241  *  The lookup of this index table translates to looking up
242  *  Bit 7, Bit 4, and Bit 3 of val entry:
243  *
244  *  PAT/PSE (bit 7) ... PCD (bit 4) .. PWT (bit 3).
245  *
246  *  If all bits are off, then we are using PAT0. If bit 3 turned on,
247  *  then we are using PAT1, if bit 3 and bit 4, then PAT2..
248  *
249  *  As you can see, the Linux PAT1 translates to PAT4 under Xen. Which means
250  *  that if a guest that follows Linux's PAT setup and would like to set Write
251  *  Combined on pages it MUST use PAT4 entry. Meaning that Bit 7 (PAGE_PAT) is
252  *  set. For example, under Linux it only uses PAT0, PAT1, and PAT2 for the
253  *  caching as:
254  *
255  *   WB = none (so PAT0)
256  *   WC = PWT (bit 3 on)
257  *   UC = PWT | PCD (bit 3 and 4 are on).
258  *
259  * To make it work with Xen, it needs to translate the WC bit as so:
260  *
261  *  PWT (so bit 3 on) --> PAT (so bit 7 is on) and clear bit 3
262  *
263  * And to translate back it would:
264  *
265  * PAT (bit 7 on) --> PWT (bit 3 on) and clear bit 7.
266  */
267 #define MMU_NORMAL_PT_UPDATE      0 /* checked '*ptr = val'. ptr is MA.       */
268 #define MMU_MACHPHYS_UPDATE       1 /* ptr = MA of frame to modify entry for  */
269 #define MMU_PT_UPDATE_PRESERVE_AD 2 /* atomically: *ptr = val | (*ptr&(A|D)) */
270 
271 /*
272  * MMU EXTENDED OPERATIONS
273  *
274  * enum neg_errnoval HYPERVISOR_mmuext_op(mmuext_op_t uops[],
275  *                                        unsigned int count,
276  *                                        unsigned int *pdone,
277  *                                        unsigned int foreigndom)
278  */
279 /* HYPERVISOR_mmuext_op() accepts a list of mmuext_op structures.
280  * A foreigndom (FD) can be specified (or DOMID_SELF for none).
281  * Where the FD has some effect, it is described below.
282  *
283  * cmd: MMUEXT_(UN)PIN_*_TABLE
284  * mfn: Machine frame number to be (un)pinned as a p.t. page.
285  *      The frame must belong to the FD, if one is specified.
286  *
287  * cmd: MMUEXT_NEW_BASEPTR
288  * mfn: Machine frame number of new page-table base to install in MMU.
289  *
290  * cmd: MMUEXT_NEW_USER_BASEPTR [x86/64 only]
291  * mfn: Machine frame number of new page-table base to install in MMU
292  *      when in user space.
293  *
294  * cmd: MMUEXT_TLB_FLUSH_LOCAL
295  * No additional arguments. Flushes local TLB.
296  *
297  * cmd: MMUEXT_INVLPG_LOCAL
298  * linear_addr: Linear address to be flushed from the local TLB.
299  *
300  * cmd: MMUEXT_TLB_FLUSH_MULTI
301  * vcpumask: Pointer to bitmap of VCPUs to be flushed.
302  *
303  * cmd: MMUEXT_INVLPG_MULTI
304  * linear_addr: Linear address to be flushed.
305  * vcpumask: Pointer to bitmap of VCPUs to be flushed.
306  *
307  * cmd: MMUEXT_TLB_FLUSH_ALL
308  * No additional arguments. Flushes all VCPUs' TLBs.
309  *
310  * cmd: MMUEXT_INVLPG_ALL
311  * linear_addr: Linear address to be flushed from all VCPUs' TLBs.
312  *
313  * cmd: MMUEXT_FLUSH_CACHE
314  * No additional arguments. Writes back and flushes cache contents.
315  *
316  * cmd: MMUEXT_FLUSH_CACHE_GLOBAL
317  * No additional arguments. Writes back and flushes cache contents
318  * on all CPUs in the system.
319  *
320  * cmd: MMUEXT_SET_LDT
321  * linear_addr: Linear address of LDT base (NB. must be page-aligned).
322  * nr_ents: Number of entries in LDT.
323  *
324  * cmd: MMUEXT_CLEAR_PAGE
325  * mfn: Machine frame number to be cleared.
326  *
327  * cmd: MMUEXT_COPY_PAGE
328  * mfn: Machine frame number of the destination page.
329  * src_mfn: Machine frame number of the source page.
330  *
331  * cmd: MMUEXT_[UN]MARK_SUPER
332  * mfn: Machine frame number of head of superpage to be [un]marked.
333  */
334 #define MMUEXT_PIN_L1_TABLE      0
335 #define MMUEXT_PIN_L2_TABLE      1
336 #define MMUEXT_PIN_L3_TABLE      2
337 #define MMUEXT_PIN_L4_TABLE      3
338 #define MMUEXT_UNPIN_TABLE       4
339 #define MMUEXT_NEW_BASEPTR       5
340 #define MMUEXT_TLB_FLUSH_LOCAL   6
341 #define MMUEXT_INVLPG_LOCAL      7
342 #define MMUEXT_TLB_FLUSH_MULTI   8
343 #define MMUEXT_INVLPG_MULTI      9
344 #define MMUEXT_TLB_FLUSH_ALL    10
345 #define MMUEXT_INVLPG_ALL       11
346 #define MMUEXT_FLUSH_CACHE      12
347 #define MMUEXT_SET_LDT          13
348 #define MMUEXT_NEW_USER_BASEPTR 15
349 #define MMUEXT_CLEAR_PAGE       16
350 #define MMUEXT_COPY_PAGE        17
351 #define MMUEXT_FLUSH_CACHE_GLOBAL 18
352 #define MMUEXT_MARK_SUPER       19
353 #define MMUEXT_UNMARK_SUPER     20
354 
355 #ifndef __ASSEMBLY__
356 struct mmuext_op {
357 	unsigned int cmd;
358 	union {
359 		/* [UN]PIN_TABLE, NEW_BASEPTR, NEW_USER_BASEPTR
360 		 * CLEAR_PAGE, COPY_PAGE, [UN]MARK_SUPER */
361 		xen_pfn_t mfn;
362 		/* INVLPG_LOCAL, INVLPG_ALL, SET_LDT */
363 		unsigned long linear_addr;
364 	} arg1;
365 	union {
366 		/* SET_LDT */
367 		unsigned int nr_ents;
368 		/* TLB_FLUSH_MULTI, INVLPG_MULTI */
369 		void *vcpumask;
370 		/* COPY_PAGE */
371 		xen_pfn_t src_mfn;
372 	} arg2;
373 };
374 DEFINE_GUEST_HANDLE_STRUCT(mmuext_op);
375 #endif
376 
377 /* These are passed as 'flags' to update_va_mapping. They can be ORed. */
378 /* When specifying UVMF_MULTI, also OR in a pointer to a CPU bitmap.   */
379 /* UVMF_LOCAL is merely UVMF_MULTI with a NULL bitmap pointer.         */
380 #define UVMF_NONE               (0UL<<0) /* No flushing at all.   */
381 #define UVMF_TLB_FLUSH          (1UL<<0) /* Flush entire TLB(s).  */
382 #define UVMF_INVLPG             (2UL<<0) /* Flush only one entry. */
383 #define UVMF_FLUSHTYPE_MASK     (3UL<<0)
384 #define UVMF_MULTI              (0UL<<2) /* Flush subset of TLBs. */
385 #define UVMF_LOCAL              (0UL<<2) /* Flush local TLB.      */
386 #define UVMF_ALL                (1UL<<2) /* Flush all TLBs.       */
387 
388 /*
389  * Commands to HYPERVISOR_console_io().
390  */
391 #define CONSOLEIO_write         0
392 #define CONSOLEIO_read          1
393 
394 /*
395  * Commands to HYPERVISOR_vm_assist().
396  */
397 #define VMASST_CMD_enable                0
398 #define VMASST_CMD_disable               1
399 
400 /* x86/32 guests: simulate full 4GB segment limits. */
401 #define VMASST_TYPE_4gb_segments         0
402 
403 /* x86/32 guests: trap (vector 15) whenever above vmassist is used. */
404 #define VMASST_TYPE_4gb_segments_notify  1
405 
406 /*
407  * x86 guests: support writes to bottom-level PTEs.
408  * NB1. Page-directory entries cannot be written.
409  * NB2. Guest must continue to remove all writable mappings of PTEs.
410  */
411 #define VMASST_TYPE_writable_pagetables  2
412 
413 /* x86/PAE guests: support PDPTs above 4GB. */
414 #define VMASST_TYPE_pae_extended_cr3     3
415 
416 /*
417  * x86 guests: Sane behaviour for virtual iopl
418  *  - virtual iopl updated from do_iret() hypercalls.
419  *  - virtual iopl reported in bounce frames.
420  *  - guest kernels assumed to be level 0 for the purpose of iopl checks.
421  */
422 #define VMASST_TYPE_architectural_iopl   4
423 
424 /*
425  * All guests: activate update indicator in vcpu_runstate_info
426  * Enable setting the XEN_RUNSTATE_UPDATE flag in guest memory mapped
427  * vcpu_runstate_info during updates of the runstate information.
428  */
429 #define VMASST_TYPE_runstate_update_flag 5
430 
431 #define MAX_VMASST_TYPE 5
432 
433 #ifndef __ASSEMBLY__
434 
435 typedef uint16_t domid_t;
436 
437 /* Domain ids >= DOMID_FIRST_RESERVED cannot be used for ordinary domains. */
438 #define DOMID_FIRST_RESERVED (0x7FF0U)
439 
440 /* DOMID_SELF is used in certain contexts to refer to oneself. */
441 #define DOMID_SELF (0x7FF0U)
442 
443 /*
444  * DOMID_IO is used to restrict page-table updates to mapping I/O memory.
445  * Although no Foreign Domain need be specified to map I/O pages, DOMID_IO
446  * is useful to ensure that no mappings to the OS's own heap are accidentally
447  * installed. (e.g., in Linux this could cause havoc as reference counts
448  * aren't adjusted on the I/O-mapping code path).
449  * This only makes sense in MMUEXT_SET_FOREIGNDOM, but in that context can
450  * be specified by any calling domain.
451  */
452 #define DOMID_IO   (0x7FF1U)
453 
454 /*
455  * DOMID_XEN is used to allow privileged domains to map restricted parts of
456  * Xen's heap space (e.g., the machine_to_phys table).
457  * This only makes sense in MMUEXT_SET_FOREIGNDOM, and is only permitted if
458  * the caller is privileged.
459  */
460 #define DOMID_XEN  (0x7FF2U)
461 
462 /* DOMID_COW is used as the owner of sharable pages */
463 #define DOMID_COW  (0x7FF3U)
464 
465 /* DOMID_INVALID is used to identify pages with unknown owner. */
466 #define DOMID_INVALID (0x7FF4U)
467 
468 /* Idle domain. */
469 #define DOMID_IDLE (0x7FFFU)
470 
471 /*
472  * Send an array of these to HYPERVISOR_mmu_update().
473  * NB. The fields are natural pointer/address size for this architecture.
474  */
475 struct mmu_update {
476     uint64_t ptr;       /* Machine address of PTE. */
477     uint64_t val;       /* New contents of PTE.    */
478 };
479 DEFINE_GUEST_HANDLE_STRUCT(mmu_update);
480 
481 /*
482  * Send an array of these to HYPERVISOR_multicall().
483  * NB. The fields are logically the natural register size for this
484  * architecture. In cases where xen_ulong_t is larger than this then
485  * any unused bits in the upper portion must be zero.
486  */
487 struct multicall_entry {
488     xen_ulong_t op;
489     xen_long_t result;
490     xen_ulong_t args[6];
491 };
492 DEFINE_GUEST_HANDLE_STRUCT(multicall_entry);
493 
494 struct vcpu_time_info {
495 	/*
496 	 * Updates to the following values are preceded and followed
497 	 * by an increment of 'version'. The guest can therefore
498 	 * detect updates by looking for changes to 'version'. If the
499 	 * least-significant bit of the version number is set then an
500 	 * update is in progress and the guest must wait to read a
501 	 * consistent set of values.  The correct way to interact with
502 	 * the version number is similar to Linux's seqlock: see the
503 	 * implementations of read_seqbegin/read_seqretry.
504 	 */
505 	uint32_t version;
506 	uint32_t pad0;
507 	uint64_t tsc_timestamp;   /* TSC at last update of time vals.  */
508 	uint64_t system_time;     /* Time, in nanosecs, since boot.    */
509 	/*
510 	 * Current system time:
511 	 *   system_time + ((tsc - tsc_timestamp) << tsc_shift) * tsc_to_system_mul
512 	 * CPU frequency (Hz):
513 	 *   ((10^9 << 32) / tsc_to_system_mul) >> tsc_shift
514 	 */
515 	uint32_t tsc_to_system_mul;
516 	int8_t   tsc_shift;
517 	int8_t   pad1[3];
518 }; /* 32 bytes */
519 
520 struct vcpu_info {
521 	/*
522 	 * 'evtchn_upcall_pending' is written non-zero by Xen to indicate
523 	 * a pending notification for a particular VCPU. It is then cleared
524 	 * by the guest OS /before/ checking for pending work, thus avoiding
525 	 * a set-and-check race. Note that the mask is only accessed by Xen
526 	 * on the CPU that is currently hosting the VCPU. This means that the
527 	 * pending and mask flags can be updated by the guest without special
528 	 * synchronisation (i.e., no need for the x86 LOCK prefix).
529 	 * This may seem suboptimal because if the pending flag is set by
530 	 * a different CPU then an IPI may be scheduled even when the mask
531 	 * is set. However, note:
532 	 *  1. The task of 'interrupt holdoff' is covered by the per-event-
533 	 *     channel mask bits. A 'noisy' event that is continually being
534 	 *     triggered can be masked at source at this very precise
535 	 *     granularity.
536 	 *  2. The main purpose of the per-VCPU mask is therefore to restrict
537 	 *     reentrant execution: whether for concurrency control, or to
538 	 *     prevent unbounded stack usage. Whatever the purpose, we expect
539 	 *     that the mask will be asserted only for short periods at a time,
540 	 *     and so the likelihood of a 'spurious' IPI is suitably small.
541 	 * The mask is read before making an event upcall to the guest: a
542 	 * non-zero mask therefore guarantees that the VCPU will not receive
543 	 * an upcall activation. The mask is cleared when the VCPU requests
544 	 * to block: this avoids wakeup-waiting races.
545 	 */
546 	uint8_t evtchn_upcall_pending;
547 	uint8_t evtchn_upcall_mask;
548 	xen_ulong_t evtchn_pending_sel;
549 	struct arch_vcpu_info arch;
550 	struct pvclock_vcpu_time_info time;
551 }; /* 64 bytes (x86) */
552 
553 /*
554  * Xen/kernel shared data -- pointer provided in start_info.
555  * NB. We expect that this struct is smaller than a page.
556  */
557 struct shared_info {
558 	struct vcpu_info vcpu_info[MAX_VIRT_CPUS];
559 
560 	/*
561 	 * A domain can create "event channels" on which it can send and receive
562 	 * asynchronous event notifications. There are three classes of event that
563 	 * are delivered by this mechanism:
564 	 *  1. Bi-directional inter- and intra-domain connections. Domains must
565 	 *     arrange out-of-band to set up a connection (usually by allocating
566 	 *     an unbound 'listener' port and avertising that via a storage service
567 	 *     such as xenstore).
568 	 *  2. Physical interrupts. A domain with suitable hardware-access
569 	 *     privileges can bind an event-channel port to a physical interrupt
570 	 *     source.
571 	 *  3. Virtual interrupts ('events'). A domain can bind an event-channel
572 	 *     port to a virtual interrupt source, such as the virtual-timer
573 	 *     device or the emergency console.
574 	 *
575 	 * Event channels are addressed by a "port index". Each channel is
576 	 * associated with two bits of information:
577 	 *  1. PENDING -- notifies the domain that there is a pending notification
578 	 *     to be processed. This bit is cleared by the guest.
579 	 *  2. MASK -- if this bit is clear then a 0->1 transition of PENDING
580 	 *     will cause an asynchronous upcall to be scheduled. This bit is only
581 	 *     updated by the guest. It is read-only within Xen. If a channel
582 	 *     becomes pending while the channel is masked then the 'edge' is lost
583 	 *     (i.e., when the channel is unmasked, the guest must manually handle
584 	 *     pending notifications as no upcall will be scheduled by Xen).
585 	 *
586 	 * To expedite scanning of pending notifications, any 0->1 pending
587 	 * transition on an unmasked channel causes a corresponding bit in a
588 	 * per-vcpu selector word to be set. Each bit in the selector covers a
589 	 * 'C long' in the PENDING bitfield array.
590 	 */
591 	xen_ulong_t evtchn_pending[sizeof(xen_ulong_t) * 8];
592 	xen_ulong_t evtchn_mask[sizeof(xen_ulong_t) * 8];
593 
594 	/*
595 	 * Wallclock time: updated only by control software. Guests should base
596 	 * their gettimeofday() syscall on this wallclock-base value.
597 	 */
598 	struct pvclock_wall_clock wc;
599 
600 	struct arch_shared_info arch;
601 
602 };
603 
604 /*
605  * Start-of-day memory layout
606  *
607  *  1. The domain is started within contiguous virtual-memory region.
608  *  2. The contiguous region begins and ends on an aligned 4MB boundary.
609  *  3. This the order of bootstrap elements in the initial virtual region:
610  *      a. relocated kernel image
611  *      b. initial ram disk              [mod_start, mod_len]
612  *         (may be omitted)
613  *      c. list of allocated page frames [mfn_list, nr_pages]
614  *         (unless relocated due to XEN_ELFNOTE_INIT_P2M)
615  *      d. start_info_t structure        [register ESI (x86)]
616  *         in case of dom0 this page contains the console info, too
617  *      e. unless dom0: xenstore ring page
618  *      f. unless dom0: console ring page
619  *      g. bootstrap page tables         [pt_base, CR3 (x86)]
620  *      h. bootstrap stack               [register ESP (x86)]
621  *  4. Bootstrap elements are packed together, but each is 4kB-aligned.
622  *  5. The list of page frames forms a contiguous 'pseudo-physical' memory
623  *     layout for the domain. In particular, the bootstrap virtual-memory
624  *     region is a 1:1 mapping to the first section of the pseudo-physical map.
625  *  6. All bootstrap elements are mapped read-writable for the guest OS. The
626  *     only exception is the bootstrap page table, which is mapped read-only.
627  *  7. There is guaranteed to be at least 512kB padding after the final
628  *     bootstrap element. If necessary, the bootstrap virtual region is
629  *     extended by an extra 4MB to ensure this.
630  */
631 
632 #define MAX_GUEST_CMDLINE 1024
633 struct start_info {
634 	/* THE FOLLOWING ARE FILLED IN BOTH ON INITIAL BOOT AND ON RESUME.    */
635 	char magic[32];             /* "xen-<version>-<platform>".            */
636 	unsigned long nr_pages;     /* Total pages allocated to this domain.  */
637 	unsigned long shared_info;  /* MACHINE address of shared info struct. */
638 	uint32_t flags;             /* SIF_xxx flags.                         */
639 	xen_pfn_t store_mfn;        /* MACHINE page number of shared page.    */
640 	uint32_t store_evtchn;      /* Event channel for store communication. */
641 	union {
642 		struct {
643 			xen_pfn_t mfn;      /* MACHINE page number of console page.   */
644 			uint32_t  evtchn;   /* Event channel for console page.        */
645 		} domU;
646 		struct {
647 			uint32_t info_off;  /* Offset of console_info struct.         */
648 			uint32_t info_size; /* Size of console_info struct from start.*/
649 		} dom0;
650 	} console;
651 	/* THE FOLLOWING ARE ONLY FILLED IN ON INITIAL BOOT (NOT RESUME).     */
652 	unsigned long pt_base;      /* VIRTUAL address of page directory.     */
653 	unsigned long nr_pt_frames; /* Number of bootstrap p.t. frames.       */
654 	unsigned long mfn_list;     /* VIRTUAL address of page-frame list.    */
655 	unsigned long mod_start;    /* VIRTUAL address of pre-loaded module.  */
656 	unsigned long mod_len;      /* Size (bytes) of pre-loaded module.     */
657 	int8_t cmd_line[MAX_GUEST_CMDLINE];
658 	/* The pfn range here covers both page table and p->m table frames.   */
659 	unsigned long first_p2m_pfn;/* 1st pfn forming initial P->M table.    */
660 	unsigned long nr_p2m_frames;/* # of pfns forming initial P->M table.  */
661 };
662 
663 /* These flags are passed in the 'flags' field of start_info_t. */
664 #define SIF_PRIVILEGED      (1<<0)  /* Is the domain privileged? */
665 #define SIF_INITDOMAIN      (1<<1)  /* Is this the initial control domain? */
666 #define SIF_MULTIBOOT_MOD   (1<<2)  /* Is mod_start a multiboot module? */
667 #define SIF_MOD_START_PFN   (1<<3)  /* Is mod_start a PFN? */
668 #define SIF_VIRT_P2M_4TOOLS (1<<4)  /* Do Xen tools understand a virt. mapped */
669 				    /* P->M making the 3 level tree obsolete? */
670 #define SIF_PM_MASK       (0xFF<<8) /* reserve 1 byte for xen-pm options */
671 
672 /*
673  * A multiboot module is a package containing modules very similar to a
674  * multiboot module array. The only differences are:
675  * - the array of module descriptors is by convention simply at the beginning
676  *   of the multiboot module,
677  * - addresses in the module descriptors are based on the beginning of the
678  *   multiboot module,
679  * - the number of modules is determined by a termination descriptor that has
680  *   mod_start == 0.
681  *
682  * This permits to both build it statically and reference it in a configuration
683  * file, and let the PV guest easily rebase the addresses to virtual addresses
684  * and at the same time count the number of modules.
685  */
686 struct xen_multiboot_mod_list {
687 	/* Address of first byte of the module */
688 	uint32_t mod_start;
689 	/* Address of last byte of the module (inclusive) */
690 	uint32_t mod_end;
691 	/* Address of zero-terminated command line */
692 	uint32_t cmdline;
693 	/* Unused, must be zero */
694 	uint32_t pad;
695 };
696 /*
697  * The console structure in start_info.console.dom0
698  *
699  * This structure includes a variety of information required to
700  * have a working VGA/VESA console.
701  */
702 struct dom0_vga_console_info {
703 	uint8_t video_type;
704 #define XEN_VGATYPE_TEXT_MODE_3 0x03
705 #define XEN_VGATYPE_VESA_LFB    0x23
706 #define XEN_VGATYPE_EFI_LFB     0x70
707 
708 	union {
709 		struct {
710 			/* Font height, in pixels. */
711 			uint16_t font_height;
712 			/* Cursor location (column, row). */
713 			uint16_t cursor_x, cursor_y;
714 			/* Number of rows and columns (dimensions in characters). */
715 			uint16_t rows, columns;
716 		} text_mode_3;
717 
718 		struct {
719 			/* Width and height, in pixels. */
720 			uint16_t width, height;
721 			/* Bytes per scan line. */
722 			uint16_t bytes_per_line;
723 			/* Bits per pixel. */
724 			uint16_t bits_per_pixel;
725 			/* LFB physical address, and size (in units of 64kB). */
726 			uint32_t lfb_base;
727 			uint32_t lfb_size;
728 			/* RGB mask offsets and sizes, as defined by VBE 1.2+ */
729 			uint8_t  red_pos, red_size;
730 			uint8_t  green_pos, green_size;
731 			uint8_t  blue_pos, blue_size;
732 			uint8_t  rsvd_pos, rsvd_size;
733 
734 			/* VESA capabilities (offset 0xa, VESA command 0x4f00). */
735 			uint32_t gbl_caps;
736 			/* Mode attributes (offset 0x0, VESA command 0x4f01). */
737 			uint16_t mode_attrs;
738 		} vesa_lfb;
739 	} u;
740 };
741 
742 typedef uint64_t cpumap_t;
743 
744 typedef uint8_t xen_domain_handle_t[16];
745 
746 /* Turn a plain number into a C unsigned long constant. */
747 #define __mk_unsigned_long(x) x ## UL
748 #define mk_unsigned_long(x) __mk_unsigned_long(x)
749 
750 #define TMEM_SPEC_VERSION 1
751 
752 struct tmem_op {
753 	uint32_t cmd;
754 	int32_t pool_id;
755 	union {
756 		struct {  /* for cmd == TMEM_NEW_POOL */
757 			uint64_t uuid[2];
758 			uint32_t flags;
759 		} new;
760 		struct {
761 			uint64_t oid[3];
762 			uint32_t index;
763 			uint32_t tmem_offset;
764 			uint32_t pfn_offset;
765 			uint32_t len;
766 			GUEST_HANDLE(void) gmfn; /* guest machine page frame */
767 		} gen;
768 	} u;
769 };
770 
771 DEFINE_GUEST_HANDLE(u64);
772 
773 #else /* __ASSEMBLY__ */
774 
775 /* In assembly code we cannot use C numeric constant suffixes. */
776 #define mk_unsigned_long(x) x
777 
778 #endif /* !__ASSEMBLY__ */
779 
780 #endif /* __XEN_PUBLIC_XEN_H__ */
781