ia64/xen-unstable

view xen/include/asm-ia64/mm.h @ 15826:7e79e7f01f3d

Implement ia64 continuable domain destroy.
Signed-off-by: Isaku Yamahata <yamahata@valinux.co.jp>
author kfraser@localhost.localdomain
date Fri Aug 31 15:46:37 2007 +0100 (2007-08-31)
parents cb3c7f006077
children d956779d8d47
line source
1 /*
2 * Copyright (c) 2006 Isaku Yamahata <yamahata at valinux co jp>
3 * VA Linux Systems Japan K.K.
4 * dom0 vp model support
5 */
6 #ifndef __ASM_IA64_MM_H__
7 #define __ASM_IA64_MM_H__
9 #include <xen/config.h>
10 #ifdef LINUX_2_6
11 #include <linux/gfp.h>
12 #endif
13 #include <xen/list.h>
14 #include <xen/spinlock.h>
15 #include <xen/perfc.h>
16 #include <xen/sched.h>
18 #include <asm/processor.h>
19 #include <asm/atomic.h>
20 #include <asm/tlbflush.h>
21 #include <asm/flushtlb.h>
22 #include <asm/io.h>
24 #include <public/xen.h>
26 /*
27 * The following is for page_alloc.c.
28 */
30 typedef unsigned long page_flags_t;
32 /*
33 * Per-page-frame information.
34 *
35 * Every architecture must ensure the following:
36 * 1. 'struct page_info' contains a 'struct list_head list'.
37 * 2. Provide a PFN_ORDER() macro for accessing the order of a free page.
38 */
39 #define PFN_ORDER(_pfn) ((_pfn)->u.free.order)
41 #define PRtype_info "016lx"
43 struct page_info
44 {
45 /* Each frame can be threaded onto a doubly-linked list. */
46 struct list_head list;
48 /* Reference count and various PGC_xxx flags and fields. */
49 u32 count_info;
51 /* Context-dependent fields follow... */
52 union {
54 /* Page is in use: ((count_info & PGC_count_mask) != 0). */
55 struct {
56 /* Owner of this page (NULL if page is anonymous). */
57 u32 _domain; /* pickled format */
58 /* Type reference count and various PGT_xxx flags and fields. */
59 unsigned long type_info;
60 } __attribute__ ((packed)) inuse;
62 /* Page is on a free list: ((count_info & PGC_count_mask) == 0). */
63 struct {
64 /* Order-size of the free chunk this page is the head of. */
65 u32 order;
66 /* Mask of possibly-tainted TLBs. */
67 cpumask_t cpumask;
68 } __attribute__ ((packed)) free;
70 } u;
72 /* Timestamp from 'TLB clock', used to reduce need for safety flushes. */
73 u32 tlbflush_timestamp;
75 #if 0
76 // following added for Linux compiling
77 page_flags_t flags;
78 atomic_t _count;
79 struct list_head lru; // is this the same as above "list"?
80 #endif
81 };
83 #define set_page_count(p,v) atomic_set(&(p)->_count, v - 1)
85 /*
86 * Still small set of flags defined by far on IA-64.
87 * IA-64 should make it a definition same as x86_64.
88 */
89 /* The following page types are MUTUALLY EXCLUSIVE. */
90 #define PGT_none (0UL<<29) /* no special uses of this page */
91 #define PGT_l1_page_table (1UL<<29) /* using this page as an L1 page table? */
92 #define PGT_l2_page_table (2UL<<29) /* using this page as an L2 page table? */
93 #define PGT_l3_page_table (3UL<<29) /* using this page as an L3 page table? */
94 #define PGT_l4_page_table (4UL<<29) /* using this page as an L4 page table? */
95 /* Value 5 reserved. See asm-x86/mm.h */
96 /* Value 6 reserved. See asm-x86/mm.h */
97 #define PGT_writable_page (7UL<<29) /* has writable mappings of this page? */
98 #define PGT_type_mask (7UL<<29) /* Bits 29-31. */
100 /* Has this page been validated for use as its current type? */
101 #define _PGT_validated 28
102 #define PGT_validated (1UL<<_PGT_validated)
103 /* Owning guest has pinned this page to its current type? */
104 #define _PGT_pinned 27
105 #define PGT_pinned (1UL<<_PGT_pinned)
107 /* 16-bit count of uses of this frame as its current type. */
108 #define PGT_count_mask ((1UL<<16)-1)
110 /* Cleared when the owning guest 'frees' this page. */
111 #define _PGC_allocated 31
112 #define PGC_allocated (1UL<<_PGC_allocated)
113 /* Bit 30 reserved. See asm-x86/mm.h */
114 /* Bit 29 reserved. See asm-x86/mm.h */
115 /* 29-bit count of references to this frame. */
116 #define PGC_count_mask ((1UL<<29)-1)
118 #define is_xen_heap_frame(pfn) ((page_to_maddr(pfn) < xenheap_phys_end) \
119 && (page_to_maddr(pfn) >= xen_pstart))
121 extern void* xen_pickle_offset;
122 #define __pickle(a) ((unsigned long)a - (unsigned long)xen_pickle_offset)
123 #define __unpickle(a) (void *)(a + xen_pickle_offset)
125 static inline struct domain *unpickle_domptr(u64 _d)
126 { return (_d == 0) ? NULL : __unpickle(_d); }
127 static inline u32 pickle_domptr(struct domain *_d)
128 { return (_d == NULL) ? 0 : (u32)__pickle(_d); }
130 #define page_get_owner(_p) (unpickle_domptr((_p)->u.inuse._domain))
131 #define page_set_owner(_p, _d) ((_p)->u.inuse._domain = pickle_domptr(_d))
133 #define XENSHARE_writable 0
134 #define XENSHARE_readonly 1
135 void share_xen_page_with_guest(struct page_info *page,
136 struct domain *d, int readonly);
137 void share_xen_page_with_privileged_guests(struct page_info *page,
138 int readonly);
140 extern struct page_info *frame_table;
141 extern unsigned long frame_table_size;
142 extern struct list_head free_list;
143 extern spinlock_t free_list_lock;
144 extern unsigned int free_pfns;
145 extern unsigned long max_page;
147 extern void __init init_frametable(void);
148 void add_to_domain_alloc_list(unsigned long ps, unsigned long pe);
150 static inline void put_page(struct page_info *page)
151 {
152 u32 nx, x, y = page->count_info;
154 do {
155 x = y;
156 nx = x - 1;
157 }
158 while (unlikely((y = cmpxchg_rel(&page->count_info, x, nx)) != x));
160 if (unlikely((nx & PGC_count_mask) == 0))
161 free_domheap_page(page);
162 }
164 /* count_info and ownership are checked atomically. */
165 static inline int get_page(struct page_info *page,
166 struct domain *domain)
167 {
168 u64 x, nx, y = *((u64*)&page->count_info);
169 u32 _domain = pickle_domptr(domain);
171 do {
172 x = y;
173 nx = x + 1;
174 if (unlikely((x & PGC_count_mask) == 0) || /* Not allocated? */
175 unlikely((nx & PGC_count_mask) == 0) || /* Count overflow? */
176 unlikely((x >> 32) != _domain)) { /* Wrong owner? */
178 gdprintk(XENLOG_INFO, "Error pfn %lx: rd=%p, od=%p, caf=%016lx, taf=%"
179 PRtype_info "\n", page_to_mfn(page), domain,
180 unpickle_domptr(x >> 32), x, page->u.inuse.type_info);
181 return 0;
182 }
183 }
184 while(unlikely((y = cmpxchg_acq((u64*)&page->count_info, x, nx)) != x));
185 return 1;
186 }
188 extern void put_page_type(struct page_info *page);
189 extern int get_page_type(struct page_info *page, u32 type);
191 static inline void put_page_and_type(struct page_info *page)
192 {
193 put_page_type(page);
194 put_page(page);
195 }
198 static inline int get_page_and_type(struct page_info *page,
199 struct domain *domain,
200 u32 type)
201 {
202 int rc = get_page(page, domain);
204 if ( likely(rc) && unlikely(!get_page_type(page, type)) )
205 {
206 put_page(page);
207 rc = 0;
208 }
210 return rc;
211 }
213 #define set_machinetophys(_mfn, _pfn) do { } while(0);
215 #ifdef MEMORY_GUARD
216 void *memguard_init(void *heap_start);
217 void memguard_guard_stack(void *p);
218 void memguard_guard_range(void *p, unsigned long l);
219 void memguard_unguard_range(void *p, unsigned long l);
220 #else
221 #define memguard_init(_s) (_s)
222 #define memguard_guard_stack(_p) ((void)0)
223 #define memguard_guard_range(_p,_l) ((void)0)
224 #define memguard_unguard_range(_p,_l) ((void)0)
225 #endif
227 // prototype of misc memory stuff
228 //unsigned long __get_free_pages(unsigned int mask, unsigned int order);
229 //void __free_pages(struct page_info *page, unsigned int order);
230 void *pgtable_quicklist_alloc(void);
231 void pgtable_quicklist_free(void *pgtable_entry);
233 // FOLLOWING FROM linux-2.6.7/include/mm.h
235 /*
236 * This struct defines a memory VMM memory area. There is one of these
237 * per VM-area/task. A VM area is any part of the process virtual memory
238 * space that has a special rule for the page-fault handlers (ie a shared
239 * library, the executable area etc).
240 */
241 struct vm_area_struct {
242 struct mm_struct * vm_mm; /* The address space we belong to. */
243 unsigned long vm_start; /* Our start address within vm_mm. */
244 unsigned long vm_end; /* The first byte after our end address
245 within vm_mm. */
247 /* linked list of VM areas per task, sorted by address */
248 struct vm_area_struct *vm_next;
250 pgprot_t vm_page_prot; /* Access permissions of this VMA. */
251 unsigned long vm_flags; /* Flags, listed below. */
253 #ifndef XEN
254 struct rb_node vm_rb;
256 // XEN doesn't need all the backing store stuff
257 /*
258 * For areas with an address space and backing store,
259 * linkage into the address_space->i_mmap prio tree, or
260 * linkage to the list of like vmas hanging off its node, or
261 * linkage of vma in the address_space->i_mmap_nonlinear list.
262 */
263 union {
264 struct {
265 struct list_head list;
266 void *parent; /* aligns with prio_tree_node parent */
267 struct vm_area_struct *head;
268 } vm_set;
270 struct prio_tree_node prio_tree_node;
271 } shared;
273 /*
274 * A file's MAP_PRIVATE vma can be in both i_mmap tree and anon_vma
275 * list, after a COW of one of the file pages. A MAP_SHARED vma
276 * can only be in the i_mmap tree. An anonymous MAP_PRIVATE, stack
277 * or brk vma (with NULL file) can only be in an anon_vma list.
278 */
279 struct list_head anon_vma_node; /* Serialized by anon_vma->lock */
280 struct anon_vma *anon_vma; /* Serialized by page_table_lock */
282 /* Function pointers to deal with this struct. */
283 struct vm_operations_struct * vm_ops;
285 /* Information about our backing store: */
286 unsigned long vm_pgoff; /* Offset (within vm_file) in PAGE_SIZE
287 units, *not* PAGE_CACHE_SIZE */
288 struct file * vm_file; /* File we map to (can be NULL). */
289 void * vm_private_data; /* was vm_pte (shared mem) */
291 #ifdef CONFIG_NUMA
292 struct mempolicy *vm_policy; /* NUMA policy for the VMA */
293 #endif
294 #endif
295 };
296 /*
297 * vm_flags..
298 */
299 #define VM_READ 0x00000001 /* currently active flags */
300 #define VM_WRITE 0x00000002
301 #define VM_EXEC 0x00000004
302 #define VM_SHARED 0x00000008
304 #define VM_MAYREAD 0x00000010 /* limits for mprotect() etc */
305 #define VM_MAYWRITE 0x00000020
306 #define VM_MAYEXEC 0x00000040
307 #define VM_MAYSHARE 0x00000080
309 #define VM_GROWSDOWN 0x00000100 /* general info on the segment */
310 #define VM_GROWSUP 0x00000200
311 #define VM_SHM 0x00000400 /* shared memory area, don't swap out */
312 #define VM_DENYWRITE 0x00000800 /* ETXTBSY on write attempts.. */
314 #define VM_EXECUTABLE 0x00001000
315 #define VM_LOCKED 0x00002000
316 #define VM_IO 0x00004000 /* Memory mapped I/O or similar */
318 /* Used by sys_madvise() */
319 #define VM_SEQ_READ 0x00008000 /* App will access data sequentially */
320 #define VM_RAND_READ 0x00010000 /* App will not benefit from clustered reads */
322 #define VM_DONTCOPY 0x00020000 /* Do not copy this vma on fork */
323 #define VM_DONTEXPAND 0x00040000 /* Cannot expand with mremap() */
324 #define VM_RESERVED 0x00080000 /* Don't unmap it from swap_out */
325 #define VM_ACCOUNT 0x00100000 /* Is a VM accounted object */
326 #define VM_HUGETLB 0x00400000 /* Huge TLB Page VM */
327 #define VM_NONLINEAR 0x00800000 /* Is non-linear (remap_file_pages) */
329 #ifndef VM_STACK_DEFAULT_FLAGS /* arch can override this */
330 #define VM_STACK_DEFAULT_FLAGS VM_DATA_DEFAULT_FLAGS
331 #endif
333 #ifdef CONFIG_STACK_GROWSUP
334 #define VM_STACK_FLAGS (VM_GROWSUP | VM_STACK_DEFAULT_FLAGS | VM_ACCOUNT)
335 #else
336 #define VM_STACK_FLAGS (VM_GROWSDOWN | VM_STACK_DEFAULT_FLAGS | VM_ACCOUNT)
337 #endif
339 #if 0 /* removed when rebasing to 2.6.13 */
340 /*
341 * The zone field is never updated after free_area_init_core()
342 * sets it, so none of the operations on it need to be atomic.
343 * We'll have up to (MAX_NUMNODES * MAX_NR_ZONES) zones total,
344 * so we use (MAX_NODES_SHIFT + MAX_ZONES_SHIFT) here to get enough bits.
345 */
346 #define NODEZONE_SHIFT (sizeof(page_flags_t)*8 - MAX_NODES_SHIFT - MAX_ZONES_SHIFT)
347 #define NODEZONE(node, zone) ((node << ZONES_SHIFT) | zone)
349 static inline unsigned long page_zonenum(struct page_info *page)
350 {
351 return (page->flags >> NODEZONE_SHIFT) & (~(~0UL << ZONES_SHIFT));
352 }
353 static inline unsigned long page_to_nid(struct page_info *page)
354 {
355 return (page->flags >> (NODEZONE_SHIFT + ZONES_SHIFT));
356 }
358 struct zone;
359 extern struct zone *zone_table[];
361 static inline struct zone *page_zone(struct page_info *page)
362 {
363 return zone_table[page->flags >> NODEZONE_SHIFT];
364 }
366 static inline void set_page_zone(struct page_info *page, unsigned long nodezone_num)
367 {
368 page->flags &= ~(~0UL << NODEZONE_SHIFT);
369 page->flags |= nodezone_num << NODEZONE_SHIFT;
370 }
371 #endif
373 #ifndef CONFIG_DISCONTIGMEM /* Don't use mapnrs, do it properly */
374 extern unsigned long max_mapnr;
375 #endif
377 static inline void *lowmem_page_address(struct page_info *page)
378 {
379 return __va(page_to_mfn(page) << PAGE_SHIFT);
380 }
382 #if defined(CONFIG_HIGHMEM) && !defined(WANT_PAGE_VIRTUAL)
383 #define HASHED_PAGE_VIRTUAL
384 #endif
386 #if defined(WANT_PAGE_VIRTUAL)
387 #define page_address(page) ((page)->virtual)
388 #define set_page_address(page, address) \
389 do { \
390 (page)->virtual = (address); \
391 } while(0)
392 #define page_address_init() do { } while(0)
393 #endif
395 #if defined(HASHED_PAGE_VIRTUAL)
396 void *page_address(struct page_info *page);
397 void set_page_address(struct page_info *page, void *virtual);
398 void page_address_init(void);
399 #endif
401 #if !defined(HASHED_PAGE_VIRTUAL) && !defined(WANT_PAGE_VIRTUAL)
402 #define page_address(page) lowmem_page_address(page)
403 #define set_page_address(page, address) do { } while(0)
404 #define page_address_init() do { } while(0)
405 #endif
408 #ifndef CONFIG_DEBUG_PAGEALLOC
409 static inline void
410 kernel_map_pages(struct page_info *page, int numpages, int enable)
411 {
412 }
413 #endif
415 extern unsigned long num_physpages;
416 extern unsigned long totalram_pages;
417 extern int nr_swap_pages;
419 extern void alloc_dom_xen_and_dom_io(void);
420 extern int mm_teardown(struct domain* d);
421 extern void mm_final_teardown(struct domain* d);
422 extern struct page_info * assign_new_domain_page(struct domain *d, unsigned long mpaddr);
423 extern void assign_new_domain0_page(struct domain *d, unsigned long mpaddr);
424 extern int __assign_domain_page(struct domain *d, unsigned long mpaddr, unsigned long physaddr, unsigned long flags);
425 extern void assign_domain_page(struct domain *d, unsigned long mpaddr, unsigned long physaddr);
426 extern void assign_domain_io_page(struct domain *d, unsigned long mpaddr, unsigned long flags);
427 struct p2m_entry;
428 extern unsigned long lookup_domain_mpa(struct domain *d, unsigned long mpaddr, struct p2m_entry* entry);
429 extern void *domain_mpa_to_imva(struct domain *d, unsigned long mpaddr);
430 extern volatile pte_t *lookup_noalloc_domain_pte(struct domain* d, unsigned long mpaddr);
431 extern unsigned long assign_domain_mmio_page(struct domain *d, unsigned long mpaddr, unsigned long phys_addr, unsigned long size, unsigned long flags);
432 extern unsigned long assign_domain_mach_page(struct domain *d, unsigned long mpaddr, unsigned long size, unsigned long flags);
433 int domain_page_mapped(struct domain *d, unsigned long mpaddr);
434 int efi_mmio(unsigned long physaddr, unsigned long size);
435 extern unsigned long ____lookup_domain_mpa(struct domain *d, unsigned long mpaddr);
436 extern unsigned long do_dom0vp_op(unsigned long cmd, unsigned long arg0, unsigned long arg1, unsigned long arg2, unsigned long arg3);
437 extern unsigned long dom0vp_zap_physmap(struct domain *d, unsigned long gpfn, unsigned int extent_order);
438 extern unsigned long dom0vp_add_physmap(struct domain* d, unsigned long gpfn, unsigned long mfn, unsigned long flags, domid_t domid);
439 extern unsigned long dom0vp_add_physmap_with_gmfn(struct domain* d, unsigned long gpfn, unsigned long gmfn, unsigned long flags, domid_t domid);
440 #ifdef CONFIG_XEN_IA64_EXPOSE_P2M
441 extern void expose_p2m_init(void);
442 extern unsigned long dom0vp_expose_p2m(struct domain* d, unsigned long conv_start_gpfn, unsigned long assign_start_gpfn, unsigned long expose_size, unsigned long granule_pfn);
443 #else
444 #define expose_p2m_init() do { } while (0)
445 #define dom0vp_expose_p2m(d, conv_start_gpfn, assign_start_gpfn, expose_size, granule_pfn) (-ENOSYS)
446 #endif
448 extern volatile unsigned long *mpt_table;
449 extern unsigned long gmfn_to_mfn_foreign(struct domain *d, unsigned long gpfn);
450 extern u64 translate_domain_pte(u64 pteval, u64 address, u64 itir__,
451 u64* itir, struct p2m_entry* entry);
452 #define machine_to_phys_mapping mpt_table
454 #define INVALID_M2P_ENTRY (~0UL)
455 #define VALID_M2P(_e) (!((_e) & (1UL<<63)))
457 #define set_gpfn_from_mfn(mfn, pfn) (machine_to_phys_mapping[(mfn)] = (pfn))
458 #define get_gpfn_from_mfn(mfn) (machine_to_phys_mapping[(mfn)])
460 /* If pmt table is provided by control pannel later, we need __get_user
461 * here. However if it's allocated by HV, we should access it directly
462 */
464 #define mfn_to_gmfn(_d, mfn) \
465 get_gpfn_from_mfn(mfn)
467 #define gmfn_to_mfn(_d, gpfn) \
468 gmfn_to_mfn_foreign((_d), (gpfn))
470 #define __gpfn_invalid(_d, gpfn) \
471 (lookup_domain_mpa((_d), ((gpfn)<<PAGE_SHIFT), NULL) & GPFN_INV_MASK)
473 #define __gmfn_valid(_d, gpfn) !__gpfn_invalid(_d, gpfn)
475 /* Return I/O type if trye */
476 #define __gpfn_is_io(_d, gpfn) \
477 ({ \
478 u64 pte, ret=0; \
479 pte = lookup_domain_mpa((_d), ((gpfn)<<PAGE_SHIFT), NULL); \
480 if(!(pte&GPFN_INV_MASK)) \
481 ret = pte & GPFN_IO_MASK; \
482 ret; \
483 })
485 #define __gpfn_is_mem(_d, gpfn) \
486 ({ \
487 u64 pte, ret=0; \
488 pte = lookup_domain_mpa((_d), ((gpfn)<<PAGE_SHIFT), NULL); \
489 if((!(pte&GPFN_INV_MASK))&&((pte & GPFN_IO_MASK)==GPFN_MEM)) \
490 ret = 1; \
491 ret; \
492 })
495 #define __gpa_to_mpa(_d, gpa) \
496 ((gmfn_to_mfn((_d),(gpa)>>PAGE_SHIFT)<<PAGE_SHIFT)|((gpa)&~PAGE_MASK))
498 #define __mpa_to_gpa(madr) \
499 ((get_gpfn_from_mfn((madr) >> PAGE_SHIFT) << PAGE_SHIFT) | \
500 ((madr) & ~PAGE_MASK))
502 /* Internal use only: returns 0 in case of bad address. */
503 extern unsigned long paddr_to_maddr(unsigned long paddr);
505 /* Arch-specific portion of memory_op hypercall. */
506 long arch_memory_op(int op, XEN_GUEST_HANDLE(void) arg);
508 int steal_page(
509 struct domain *d, struct page_info *page, unsigned int memflags);
511 #define domain_clamp_alloc_bitsize(d, b) (b)
513 unsigned long domain_get_maximum_gpfn(struct domain *d);
515 #endif /* __ASM_IA64_MM_H__ */