Commit | Line | Data |
---|---|---|
f88df14b DG |
1 | #ifndef _ASM_POWERPC_PGTABLE_PPC32_H |
2 | #define _ASM_POWERPC_PGTABLE_PPC32_H | |
3 | ||
4 | #include <asm-generic/4level-fixup.h> | |
5 | ||
6 | #ifndef __ASSEMBLY__ | |
7 | #include <linux/sched.h> | |
8 | #include <linux/threads.h> | |
9 | #include <asm/processor.h> /* For TASK_SIZE */ | |
10 | #include <asm/mmu.h> | |
11 | #include <asm/page.h> | |
12 | #include <asm/io.h> /* For sub-arch specific PPC_PIN_SIZE */ | |
13 | struct mm_struct; | |
14 | ||
15 | extern unsigned long va_to_phys(unsigned long address); | |
16 | extern pte_t *va_to_pte(unsigned long address); | |
17 | extern unsigned long ioremap_bot, ioremap_base; | |
18 | #endif /* __ASSEMBLY__ */ | |
19 | ||
20 | /* | |
21 | * The PowerPC MMU uses a hash table containing PTEs, together with | |
22 | * a set of 16 segment registers (on 32-bit implementations), to define | |
23 | * the virtual to physical address mapping. | |
24 | * | |
25 | * We use the hash table as an extended TLB, i.e. a cache of currently | |
26 | * active mappings. We maintain a two-level page table tree, much | |
27 | * like that used by the i386, for the sake of the Linux memory | |
28 | * management code. Low-level assembler code in hashtable.S | |
29 | * (procedure hash_page) is responsible for extracting ptes from the | |
30 | * tree and putting them into the hash table when necessary, and | |
31 | * updating the accessed and modified bits in the page table tree. | |
32 | */ | |
33 | ||
34 | /* | |
35 | * The PowerPC MPC8xx uses a TLB with hardware assisted, software tablewalk. | |
36 | * We also use the two level tables, but we can put the real bits in them | |
37 | * needed for the TLB and tablewalk. These definitions require Mx_CTR.PPM = 0, | |
38 | * Mx_CTR.PPCS = 0, and MD_CTR.TWAM = 1. The level 2 descriptor has | |
39 | * additional page protection (when Mx_CTR.PPCS = 1) that allows TLB hit | |
40 | * based upon user/super access. The TLB does not have accessed nor write | |
41 | * protect. We assume that if the TLB get loaded with an entry it is | |
42 | * accessed, and overload the changed bit for write protect. We use | |
43 | * two bits in the software pte that are supposed to be set to zero in | |
44 | * the TLB entry (24 and 25) for these indicators. Although the level 1 | |
45 | * descriptor contains the guarded and writethrough/copyback bits, we can | |
46 | * set these at the page level since they get copied from the Mx_TWC | |
47 | * register when the TLB entry is loaded. We will use bit 27 for guard, since | |
48 | * that is where it exists in the MD_TWC, and bit 26 for writethrough. | |
49 | * These will get masked from the level 2 descriptor at TLB load time, and | |
50 | * copied to the MD_TWC before it gets loaded. | |
51 | * Large page sizes added. We currently support two sizes, 4K and 8M. | |
52 | * This also allows a TLB hander optimization because we can directly | |
53 | * load the PMD into MD_TWC. The 8M pages are only used for kernel | |
54 | * mapping of well known areas. The PMD (PGD) entries contain control | |
55 | * flags in addition to the address, so care must be taken that the | |
56 | * software no longer assumes these are only pointers. | |
57 | */ | |
58 | ||
59 | /* | |
60 | * At present, all PowerPC 400-class processors share a similar TLB | |
61 | * architecture. The instruction and data sides share a unified, | |
62 | * 64-entry, fully-associative TLB which is maintained totally under | |
63 | * software control. In addition, the instruction side has a | |
64 | * hardware-managed, 4-entry, fully-associative TLB which serves as a | |
65 | * first level to the shared TLB. These two TLBs are known as the UTLB | |
66 | * and ITLB, respectively (see "mmu.h" for definitions). | |
67 | */ | |
68 | ||
69 | /* | |
70 | * The normal case is that PTEs are 32-bits and we have a 1-page | |
71 | * 1024-entry pgdir pointing to 1-page 1024-entry PTE pages. -- paulus | |
72 | * | |
73 | * For any >32-bit physical address platform, we can use the following | |
74 | * two level page table layout where the pgdir is 8KB and the MS 13 bits | |
75 | * are an index to the second level table. The combined pgdir/pmd first | |
76 | * level has 2048 entries and the second level has 512 64-bit PTE entries. | |
77 | * -Matt | |
78 | */ | |
79 | /* PMD_SHIFT determines the size of the area mapped by the PTE pages */ | |
80 | #define PMD_SHIFT (PAGE_SHIFT + PTE_SHIFT) | |
81 | #define PMD_SIZE (1UL << PMD_SHIFT) | |
82 | #define PMD_MASK (~(PMD_SIZE-1)) | |
83 | ||
84 | /* PGDIR_SHIFT determines what a top-level page table entry can map */ | |
85 | #define PGDIR_SHIFT PMD_SHIFT | |
86 | #define PGDIR_SIZE (1UL << PGDIR_SHIFT) | |
87 | #define PGDIR_MASK (~(PGDIR_SIZE-1)) | |
88 | ||
89 | /* | |
90 | * entries per page directory level: our page-table tree is two-level, so | |
91 | * we don't really have any PMD directory. | |
92 | */ | |
93 | #define PTRS_PER_PTE (1 << PTE_SHIFT) | |
94 | #define PTRS_PER_PMD 1 | |
95 | #define PTRS_PER_PGD (1 << (32 - PGDIR_SHIFT)) | |
96 | ||
97 | #define USER_PTRS_PER_PGD (TASK_SIZE / PGDIR_SIZE) | |
98 | #define FIRST_USER_ADDRESS 0 | |
99 | ||
100 | #define USER_PGD_PTRS (PAGE_OFFSET >> PGDIR_SHIFT) | |
101 | #define KERNEL_PGD_PTRS (PTRS_PER_PGD-USER_PGD_PTRS) | |
102 | ||
103 | #define pte_ERROR(e) \ | |
104 | printk("%s:%d: bad pte "PTE_FMT".\n", __FILE__, __LINE__, pte_val(e)) | |
105 | #define pmd_ERROR(e) \ | |
106 | printk("%s:%d: bad pmd %08lx.\n", __FILE__, __LINE__, pmd_val(e)) | |
107 | #define pgd_ERROR(e) \ | |
108 | printk("%s:%d: bad pgd %08lx.\n", __FILE__, __LINE__, pgd_val(e)) | |
109 | ||
110 | /* | |
111 | * Just any arbitrary offset to the start of the vmalloc VM area: the | |
112 | * current 64MB value just means that there will be a 64MB "hole" after the | |
113 | * physical memory until the kernel virtual memory starts. That means that | |
114 | * any out-of-bounds memory accesses will hopefully be caught. | |
115 | * The vmalloc() routines leaves a hole of 4kB between each vmalloced | |
116 | * area for the same reason. ;) | |
117 | * | |
118 | * We no longer map larger than phys RAM with the BATs so we don't have | |
119 | * to worry about the VMALLOC_OFFSET causing problems. We do have to worry | |
120 | * about clashes between our early calls to ioremap() that start growing down | |
121 | * from ioremap_base being run into the VM area allocations (growing upwards | |
122 | * from VMALLOC_START). For this reason we have ioremap_bot to check when | |
123 | * we actually run into our mappings setup in the early boot with the VM | |
124 | * system. This really does become a problem for machines with good amounts | |
125 | * of RAM. -- Cort | |
126 | */ | |
127 | #define VMALLOC_OFFSET (0x1000000) /* 16M */ | |
128 | #ifdef PPC_PIN_SIZE | |
129 | #define VMALLOC_START (((_ALIGN((long)high_memory, PPC_PIN_SIZE) + VMALLOC_OFFSET) & ~(VMALLOC_OFFSET-1))) | |
130 | #else | |
131 | #define VMALLOC_START ((((long)high_memory + VMALLOC_OFFSET) & ~(VMALLOC_OFFSET-1))) | |
132 | #endif | |
133 | #define VMALLOC_END ioremap_bot | |
134 | ||
135 | /* | |
136 | * Bits in a linux-style PTE. These match the bits in the | |
137 | * (hardware-defined) PowerPC PTE as closely as possible. | |
138 | */ | |
139 | ||
140 | #if defined(CONFIG_40x) | |
141 | ||
142 | /* There are several potential gotchas here. The 40x hardware TLBLO | |
143 | field looks like this: | |
144 | ||
145 | 0 1 2 3 4 ... 18 19 20 21 22 23 24 25 26 27 28 29 30 31 | |
146 | RPN..................... 0 0 EX WR ZSEL....... W I M G | |
147 | ||
148 | Where possible we make the Linux PTE bits match up with this | |
149 | ||
150 | - bits 20 and 21 must be cleared, because we use 4k pages (40x can | |
151 | support down to 1k pages), this is done in the TLBMiss exception | |
152 | handler. | |
153 | - We use only zones 0 (for kernel pages) and 1 (for user pages) | |
154 | of the 16 available. Bit 24-26 of the TLB are cleared in the TLB | |
155 | miss handler. Bit 27 is PAGE_USER, thus selecting the correct | |
156 | zone. | |
157 | - PRESENT *must* be in the bottom two bits because swap cache | |
158 | entries use the top 30 bits. Because 40x doesn't support SMP | |
159 | anyway, M is irrelevant so we borrow it for PAGE_PRESENT. Bit 30 | |
160 | is cleared in the TLB miss handler before the TLB entry is loaded. | |
161 | - All other bits of the PTE are loaded into TLBLO without | |
162 | modification, leaving us only the bits 20, 21, 24, 25, 26, 30 for | |
163 | software PTE bits. We actually use use bits 21, 24, 25, and | |
164 | 30 respectively for the software bits: ACCESSED, DIRTY, RW, and | |
165 | PRESENT. | |
166 | */ | |
167 | ||
168 | /* Definitions for 40x embedded chips. */ | |
169 | #define _PAGE_GUARDED 0x001 /* G: page is guarded from prefetch */ | |
170 | #define _PAGE_FILE 0x001 /* when !present: nonlinear file mapping */ | |
171 | #define _PAGE_PRESENT 0x002 /* software: PTE contains a translation */ | |
172 | #define _PAGE_NO_CACHE 0x004 /* I: caching is inhibited */ | |
173 | #define _PAGE_WRITETHRU 0x008 /* W: caching is write-through */ | |
174 | #define _PAGE_USER 0x010 /* matches one of the zone permission bits */ | |
175 | #define _PAGE_RW 0x040 /* software: Writes permitted */ | |
176 | #define _PAGE_DIRTY 0x080 /* software: dirty page */ | |
177 | #define _PAGE_HWWRITE 0x100 /* hardware: Dirty & RW, set in exception */ | |
178 | #define _PAGE_HWEXEC 0x200 /* hardware: EX permission */ | |
179 | #define _PAGE_ACCESSED 0x400 /* software: R: page referenced */ | |
180 | ||
181 | #define _PMD_PRESENT 0x400 /* PMD points to page of PTEs */ | |
182 | #define _PMD_BAD 0x802 | |
183 | #define _PMD_SIZE 0x0e0 /* size field, != 0 for large-page PMD entry */ | |
184 | #define _PMD_SIZE_4M 0x0c0 | |
185 | #define _PMD_SIZE_16M 0x0e0 | |
186 | #define PMD_PAGE_SIZE(pmdval) (1024 << (((pmdval) & _PMD_SIZE) >> 4)) | |
187 | ||
188 | #elif defined(CONFIG_44x) | |
189 | /* | |
190 | * Definitions for PPC440 | |
191 | * | |
192 | * Because of the 3 word TLB entries to support 36-bit addressing, | |
193 | * the attribute are difficult to map in such a fashion that they | |
194 | * are easily loaded during exception processing. I decided to | |
195 | * organize the entry so the ERPN is the only portion in the | |
196 | * upper word of the PTE and the attribute bits below are packed | |
197 | * in as sensibly as they can be in the area below a 4KB page size | |
198 | * oriented RPN. This at least makes it easy to load the RPN and | |
199 | * ERPN fields in the TLB. -Matt | |
200 | * | |
201 | * Note that these bits preclude future use of a page size | |
202 | * less than 4KB. | |
203 | * | |
204 | * | |
205 | * PPC 440 core has following TLB attribute fields; | |
206 | * | |
207 | * TLB1: | |
208 | * 0 1 2 3 4 ... 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 | |
209 | * RPN................................. - - - - - - ERPN....... | |
210 | * | |
211 | * TLB2: | |
212 | * 0 1 2 3 4 ... 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 | |
213 | * - - - - - - U0 U1 U2 U3 W I M G E - UX UW UR SX SW SR | |
214 | * | |
215 | * There are some constrains and options, to decide mapping software bits | |
216 | * into TLB entry. | |
217 | * | |
218 | * - PRESENT *must* be in the bottom three bits because swap cache | |
219 | * entries use the top 29 bits for TLB2. | |
220 | * | |
221 | * - FILE *must* be in the bottom three bits because swap cache | |
222 | * entries use the top 29 bits for TLB2. | |
223 | * | |
224 | * - CACHE COHERENT bit (M) has no effect on PPC440 core, because it | |
225 | * doesn't support SMP. So we can use this as software bit, like | |
226 | * DIRTY. | |
227 | * | |
228 | * With the PPC 44x Linux implementation, the 0-11th LSBs of the PTE are used | |
229 | * for memory protection related functions (see PTE structure in | |
230 | * include/asm-ppc/mmu.h). The _PAGE_XXX definitions in this file map to the | |
231 | * above bits. Note that the bit values are CPU specific, not architecture | |
232 | * specific. | |
233 | * | |
234 | * The kernel PTE entry holds an arch-dependent swp_entry structure under | |
235 | * certain situations. In other words, in such situations some portion of | |
236 | * the PTE bits are used as a swp_entry. In the PPC implementation, the | |
237 | * 3-24th LSB are shared with swp_entry, however the 0-2nd three LSB still | |
238 | * hold protection values. That means the three protection bits are | |
239 | * reserved for both PTE and SWAP entry at the most significant three | |
240 | * LSBs. | |
241 | * | |
242 | * There are three protection bits available for SWAP entry: | |
243 | * _PAGE_PRESENT | |
244 | * _PAGE_FILE | |
245 | * _PAGE_HASHPTE (if HW has) | |
246 | * | |
247 | * So those three bits have to be inside of 0-2nd LSB of PTE. | |
248 | * | |
249 | */ | |
250 | ||
251 | #define _PAGE_PRESENT 0x00000001 /* S: PTE valid */ | |
252 | #define _PAGE_RW 0x00000002 /* S: Write permission */ | |
253 | #define _PAGE_FILE 0x00000004 /* S: nonlinear file mapping */ | |
254 | #define _PAGE_ACCESSED 0x00000008 /* S: Page referenced */ | |
255 | #define _PAGE_HWWRITE 0x00000010 /* H: Dirty & RW */ | |
256 | #define _PAGE_HWEXEC 0x00000020 /* H: Execute permission */ | |
257 | #define _PAGE_USER 0x00000040 /* S: User page */ | |
258 | #define _PAGE_ENDIAN 0x00000080 /* H: E bit */ | |
259 | #define _PAGE_GUARDED 0x00000100 /* H: G bit */ | |
260 | #define _PAGE_DIRTY 0x00000200 /* S: Page dirty */ | |
261 | #define _PAGE_NO_CACHE 0x00000400 /* H: I bit */ | |
262 | #define _PAGE_WRITETHRU 0x00000800 /* H: W bit */ | |
263 | ||
264 | /* TODO: Add large page lowmem mapping support */ | |
265 | #define _PMD_PRESENT 0 | |
266 | #define _PMD_PRESENT_MASK (PAGE_MASK) | |
267 | #define _PMD_BAD (~PAGE_MASK) | |
268 | ||
269 | /* ERPN in a PTE never gets cleared, ignore it */ | |
270 | #define _PTE_NONE_MASK 0xffffffff00000000ULL | |
271 | ||
272 | #elif defined(CONFIG_FSL_BOOKE) | |
273 | /* | |
274 | MMU Assist Register 3: | |
275 | ||
276 | 32 33 34 35 36 ... 50 51 52 53 54 55 56 57 58 59 60 61 62 63 | |
277 | RPN...................... 0 0 U0 U1 U2 U3 UX SX UW SW UR SR | |
278 | ||
279 | - PRESENT *must* be in the bottom three bits because swap cache | |
280 | entries use the top 29 bits. | |
281 | ||
282 | - FILE *must* be in the bottom three bits because swap cache | |
283 | entries use the top 29 bits. | |
284 | */ | |
285 | ||
286 | /* Definitions for FSL Book-E Cores */ | |
287 | #define _PAGE_PRESENT 0x00001 /* S: PTE contains a translation */ | |
288 | #define _PAGE_USER 0x00002 /* S: User page (maps to UR) */ | |
289 | #define _PAGE_FILE 0x00002 /* S: when !present: nonlinear file mapping */ | |
290 | #define _PAGE_ACCESSED 0x00004 /* S: Page referenced */ | |
291 | #define _PAGE_HWWRITE 0x00008 /* H: Dirty & RW, set in exception */ | |
292 | #define _PAGE_RW 0x00010 /* S: Write permission */ | |
293 | #define _PAGE_HWEXEC 0x00020 /* H: UX permission */ | |
294 | ||
295 | #define _PAGE_ENDIAN 0x00040 /* H: E bit */ | |
296 | #define _PAGE_GUARDED 0x00080 /* H: G bit */ | |
297 | #define _PAGE_COHERENT 0x00100 /* H: M bit */ | |
298 | #define _PAGE_NO_CACHE 0x00200 /* H: I bit */ | |
299 | #define _PAGE_WRITETHRU 0x00400 /* H: W bit */ | |
300 | ||
301 | #ifdef CONFIG_PTE_64BIT | |
302 | #define _PAGE_DIRTY 0x08000 /* S: Page dirty */ | |
303 | ||
304 | /* ERPN in a PTE never gets cleared, ignore it */ | |
305 | #define _PTE_NONE_MASK 0xffffffffffff0000ULL | |
306 | #else | |
307 | #define _PAGE_DIRTY 0x00800 /* S: Page dirty */ | |
308 | #endif | |
309 | ||
310 | #define _PMD_PRESENT 0 | |
311 | #define _PMD_PRESENT_MASK (PAGE_MASK) | |
312 | #define _PMD_BAD (~PAGE_MASK) | |
313 | ||
314 | #elif defined(CONFIG_8xx) | |
315 | /* Definitions for 8xx embedded chips. */ | |
316 | #define _PAGE_PRESENT 0x0001 /* Page is valid */ | |
317 | #define _PAGE_FILE 0x0002 /* when !present: nonlinear file mapping */ | |
318 | #define _PAGE_NO_CACHE 0x0002 /* I: cache inhibit */ | |
319 | #define _PAGE_SHARED 0x0004 /* No ASID (context) compare */ | |
320 | ||
321 | /* These five software bits must be masked out when the entry is loaded | |
322 | * into the TLB. | |
323 | */ | |
324 | #define _PAGE_EXEC 0x0008 /* software: i-cache coherency required */ | |
325 | #define _PAGE_GUARDED 0x0010 /* software: guarded access */ | |
326 | #define _PAGE_DIRTY 0x0020 /* software: page changed */ | |
327 | #define _PAGE_RW 0x0040 /* software: user write access allowed */ | |
328 | #define _PAGE_ACCESSED 0x0080 /* software: page referenced */ | |
329 | ||
330 | /* Setting any bits in the nibble with the follow two controls will | |
331 | * require a TLB exception handler change. It is assumed unused bits | |
332 | * are always zero. | |
333 | */ | |
334 | #define _PAGE_HWWRITE 0x0100 /* h/w write enable: never set in Linux PTE */ | |
335 | #define _PAGE_USER 0x0800 /* One of the PP bits, the other is USER&~RW */ | |
336 | ||
337 | #define _PMD_PRESENT 0x0001 | |
338 | #define _PMD_BAD 0x0ff0 | |
339 | #define _PMD_PAGE_MASK 0x000c | |
340 | #define _PMD_PAGE_8M 0x000c | |
341 | ||
342 | /* | |
343 | * The 8xx TLB miss handler allegedly sets _PAGE_ACCESSED in the PTE | |
344 | * for an address even if _PAGE_PRESENT is not set, as a performance | |
345 | * optimization. This is a bug if you ever want to use swap unless | |
346 | * _PAGE_ACCESSED is 2, which it isn't, or unless you have 8xx-specific | |
347 | * definitions for __swp_entry etc. below, which would be gross. | |
348 | * -- paulus | |
349 | */ | |
350 | #define _PTE_NONE_MASK _PAGE_ACCESSED | |
351 | ||
352 | #else /* CONFIG_6xx */ | |
353 | /* Definitions for 60x, 740/750, etc. */ | |
354 | #define _PAGE_PRESENT 0x001 /* software: pte contains a translation */ | |
355 | #define _PAGE_HASHPTE 0x002 /* hash_page has made an HPTE for this pte */ | |
356 | #define _PAGE_FILE 0x004 /* when !present: nonlinear file mapping */ | |
357 | #define _PAGE_USER 0x004 /* usermode access allowed */ | |
358 | #define _PAGE_GUARDED 0x008 /* G: prohibit speculative access */ | |
359 | #define _PAGE_COHERENT 0x010 /* M: enforce memory coherence (SMP systems) */ | |
360 | #define _PAGE_NO_CACHE 0x020 /* I: cache inhibit */ | |
361 | #define _PAGE_WRITETHRU 0x040 /* W: cache write-through */ | |
362 | #define _PAGE_DIRTY 0x080 /* C: page changed */ | |
363 | #define _PAGE_ACCESSED 0x100 /* R: page referenced */ | |
364 | #define _PAGE_EXEC 0x200 /* software: i-cache coherency required */ | |
365 | #define _PAGE_RW 0x400 /* software: user write access allowed */ | |
366 | ||
367 | #define _PTE_NONE_MASK _PAGE_HASHPTE | |
368 | ||
369 | #define _PMD_PRESENT 0 | |
370 | #define _PMD_PRESENT_MASK (PAGE_MASK) | |
371 | #define _PMD_BAD (~PAGE_MASK) | |
372 | #endif | |
373 | ||
374 | /* | |
375 | * Some bits are only used on some cpu families... | |
376 | */ | |
377 | #ifndef _PAGE_HASHPTE | |
378 | #define _PAGE_HASHPTE 0 | |
379 | #endif | |
380 | #ifndef _PTE_NONE_MASK | |
381 | #define _PTE_NONE_MASK 0 | |
382 | #endif | |
383 | #ifndef _PAGE_SHARED | |
384 | #define _PAGE_SHARED 0 | |
385 | #endif | |
386 | #ifndef _PAGE_HWWRITE | |
387 | #define _PAGE_HWWRITE 0 | |
388 | #endif | |
389 | #ifndef _PAGE_HWEXEC | |
390 | #define _PAGE_HWEXEC 0 | |
391 | #endif | |
392 | #ifndef _PAGE_EXEC | |
393 | #define _PAGE_EXEC 0 | |
394 | #endif | |
395 | #ifndef _PMD_PRESENT_MASK | |
396 | #define _PMD_PRESENT_MASK _PMD_PRESENT | |
397 | #endif | |
398 | #ifndef _PMD_SIZE | |
399 | #define _PMD_SIZE 0 | |
400 | #define PMD_PAGE_SIZE(pmd) bad_call_to_PMD_PAGE_SIZE() | |
401 | #endif | |
402 | ||
403 | #define _PAGE_CHG_MASK (PAGE_MASK | _PAGE_ACCESSED | _PAGE_DIRTY) | |
404 | ||
405 | /* | |
406 | * Note: the _PAGE_COHERENT bit automatically gets set in the hardware | |
407 | * PTE if CONFIG_SMP is defined (hash_page does this); there is no need | |
408 | * to have it in the Linux PTE, and in fact the bit could be reused for | |
409 | * another purpose. -- paulus. | |
410 | */ | |
411 | ||
412 | #ifdef CONFIG_44x | |
413 | #define _PAGE_BASE (_PAGE_PRESENT | _PAGE_ACCESSED | _PAGE_GUARDED) | |
414 | #else | |
415 | #define _PAGE_BASE (_PAGE_PRESENT | _PAGE_ACCESSED) | |
416 | #endif | |
417 | #define _PAGE_WRENABLE (_PAGE_RW | _PAGE_DIRTY | _PAGE_HWWRITE) | |
418 | #define _PAGE_KERNEL (_PAGE_BASE | _PAGE_SHARED | _PAGE_WRENABLE) | |
419 | ||
420 | #ifdef CONFIG_PPC_STD_MMU | |
421 | /* On standard PPC MMU, no user access implies kernel read/write access, | |
422 | * so to write-protect kernel memory we must turn on user access */ | |
423 | #define _PAGE_KERNEL_RO (_PAGE_BASE | _PAGE_SHARED | _PAGE_USER) | |
424 | #else | |
425 | #define _PAGE_KERNEL_RO (_PAGE_BASE | _PAGE_SHARED) | |
426 | #endif | |
427 | ||
428 | #define _PAGE_IO (_PAGE_KERNEL | _PAGE_NO_CACHE | _PAGE_GUARDED) | |
429 | #define _PAGE_RAM (_PAGE_KERNEL | _PAGE_HWEXEC) | |
430 | ||
431 | #if defined(CONFIG_KGDB) || defined(CONFIG_XMON) || defined(CONFIG_BDI_SWITCH) | |
432 | /* We want the debuggers to be able to set breakpoints anywhere, so | |
433 | * don't write protect the kernel text */ | |
434 | #define _PAGE_RAM_TEXT _PAGE_RAM | |
435 | #else | |
436 | #define _PAGE_RAM_TEXT (_PAGE_KERNEL_RO | _PAGE_HWEXEC) | |
437 | #endif | |
438 | ||
439 | #define PAGE_NONE __pgprot(_PAGE_BASE) | |
440 | #define PAGE_READONLY __pgprot(_PAGE_BASE | _PAGE_USER) | |
441 | #define PAGE_READONLY_X __pgprot(_PAGE_BASE | _PAGE_USER | _PAGE_EXEC) | |
442 | #define PAGE_SHARED __pgprot(_PAGE_BASE | _PAGE_USER | _PAGE_RW) | |
443 | #define PAGE_SHARED_X __pgprot(_PAGE_BASE | _PAGE_USER | _PAGE_RW | _PAGE_EXEC) | |
444 | #define PAGE_COPY __pgprot(_PAGE_BASE | _PAGE_USER) | |
445 | #define PAGE_COPY_X __pgprot(_PAGE_BASE | _PAGE_USER | _PAGE_EXEC) | |
446 | ||
447 | #define PAGE_KERNEL __pgprot(_PAGE_RAM) | |
448 | #define PAGE_KERNEL_NOCACHE __pgprot(_PAGE_IO) | |
449 | ||
450 | /* | |
451 | * The PowerPC can only do execute protection on a segment (256MB) basis, | |
452 | * not on a page basis. So we consider execute permission the same as read. | |
453 | * Also, write permissions imply read permissions. | |
454 | * This is the closest we can get.. | |
455 | */ | |
456 | #define __P000 PAGE_NONE | |
457 | #define __P001 PAGE_READONLY_X | |
458 | #define __P010 PAGE_COPY | |
459 | #define __P011 PAGE_COPY_X | |
460 | #define __P100 PAGE_READONLY | |
461 | #define __P101 PAGE_READONLY_X | |
462 | #define __P110 PAGE_COPY | |
463 | #define __P111 PAGE_COPY_X | |
464 | ||
465 | #define __S000 PAGE_NONE | |
466 | #define __S001 PAGE_READONLY_X | |
467 | #define __S010 PAGE_SHARED | |
468 | #define __S011 PAGE_SHARED_X | |
469 | #define __S100 PAGE_READONLY | |
470 | #define __S101 PAGE_READONLY_X | |
471 | #define __S110 PAGE_SHARED | |
472 | #define __S111 PAGE_SHARED_X | |
473 | ||
474 | #ifndef __ASSEMBLY__ | |
475 | /* Make sure we get a link error if PMD_PAGE_SIZE is ever called on a | |
476 | * kernel without large page PMD support */ | |
477 | extern unsigned long bad_call_to_PMD_PAGE_SIZE(void); | |
478 | ||
479 | /* | |
480 | * Conversions between PTE values and page frame numbers. | |
481 | */ | |
482 | ||
483 | /* in some case we want to additionaly adjust where the pfn is in the pte to | |
484 | * allow room for more flags */ | |
485 | #if defined(CONFIG_FSL_BOOKE) && defined(CONFIG_PTE_64BIT) | |
486 | #define PFN_SHIFT_OFFSET (PAGE_SHIFT + 8) | |
487 | #else | |
488 | #define PFN_SHIFT_OFFSET (PAGE_SHIFT) | |
489 | #endif | |
490 | ||
491 | #define pte_pfn(x) (pte_val(x) >> PFN_SHIFT_OFFSET) | |
492 | #define pte_page(x) pfn_to_page(pte_pfn(x)) | |
493 | ||
494 | #define pfn_pte(pfn, prot) __pte(((pte_basic_t)(pfn) << PFN_SHIFT_OFFSET) |\ | |
495 | pgprot_val(prot)) | |
496 | #define mk_pte(page, prot) pfn_pte(page_to_pfn(page), prot) | |
497 | ||
498 | /* | |
499 | * ZERO_PAGE is a global shared page that is always zero: used | |
500 | * for zero-mapped memory areas etc.. | |
501 | */ | |
502 | extern unsigned long empty_zero_page[1024]; | |
503 | #define ZERO_PAGE(vaddr) (virt_to_page(empty_zero_page)) | |
504 | ||
505 | #endif /* __ASSEMBLY__ */ | |
506 | ||
507 | #define pte_none(pte) ((pte_val(pte) & ~_PTE_NONE_MASK) == 0) | |
508 | #define pte_present(pte) (pte_val(pte) & _PAGE_PRESENT) | |
509 | #define pte_clear(mm,addr,ptep) do { set_pte_at((mm), (addr), (ptep), __pte(0)); } while (0) | |
510 | ||
511 | #define pmd_none(pmd) (!pmd_val(pmd)) | |
512 | #define pmd_bad(pmd) (pmd_val(pmd) & _PMD_BAD) | |
513 | #define pmd_present(pmd) (pmd_val(pmd) & _PMD_PRESENT_MASK) | |
514 | #define pmd_clear(pmdp) do { pmd_val(*(pmdp)) = 0; } while (0) | |
515 | ||
516 | #ifndef __ASSEMBLY__ | |
517 | /* | |
518 | * The "pgd_xxx()" functions here are trivial for a folded two-level | |
519 | * setup: the pgd is never bad, and a pmd always exists (as it's folded | |
520 | * into the pgd entry) | |
521 | */ | |
522 | static inline int pgd_none(pgd_t pgd) { return 0; } | |
523 | static inline int pgd_bad(pgd_t pgd) { return 0; } | |
524 | static inline int pgd_present(pgd_t pgd) { return 1; } | |
525 | #define pgd_clear(xp) do { } while (0) | |
526 | ||
527 | #define pgd_page_vaddr(pgd) \ | |
528 | ((unsigned long) __va(pgd_val(pgd) & PAGE_MASK)) | |
529 | ||
530 | /* | |
531 | * The following only work if pte_present() is true. | |
532 | * Undefined behaviour if not.. | |
533 | */ | |
534 | static inline int pte_read(pte_t pte) { return pte_val(pte) & _PAGE_USER; } | |
535 | static inline int pte_write(pte_t pte) { return pte_val(pte) & _PAGE_RW; } | |
536 | static inline int pte_exec(pte_t pte) { return pte_val(pte) & _PAGE_EXEC; } | |
537 | static inline int pte_dirty(pte_t pte) { return pte_val(pte) & _PAGE_DIRTY; } | |
538 | static inline int pte_young(pte_t pte) { return pte_val(pte) & _PAGE_ACCESSED; } | |
539 | static inline int pte_file(pte_t pte) { return pte_val(pte) & _PAGE_FILE; } | |
540 | ||
541 | static inline void pte_uncache(pte_t pte) { pte_val(pte) |= _PAGE_NO_CACHE; } | |
542 | static inline void pte_cache(pte_t pte) { pte_val(pte) &= ~_PAGE_NO_CACHE; } | |
543 | ||
544 | static inline pte_t pte_rdprotect(pte_t pte) { | |
545 | pte_val(pte) &= ~_PAGE_USER; return pte; } | |
546 | static inline pte_t pte_wrprotect(pte_t pte) { | |
547 | pte_val(pte) &= ~(_PAGE_RW | _PAGE_HWWRITE); return pte; } | |
548 | static inline pte_t pte_exprotect(pte_t pte) { | |
549 | pte_val(pte) &= ~_PAGE_EXEC; return pte; } | |
550 | static inline pte_t pte_mkclean(pte_t pte) { | |
551 | pte_val(pte) &= ~(_PAGE_DIRTY | _PAGE_HWWRITE); return pte; } | |
552 | static inline pte_t pte_mkold(pte_t pte) { | |
553 | pte_val(pte) &= ~_PAGE_ACCESSED; return pte; } | |
554 | ||
555 | static inline pte_t pte_mkread(pte_t pte) { | |
556 | pte_val(pte) |= _PAGE_USER; return pte; } | |
557 | static inline pte_t pte_mkexec(pte_t pte) { | |
558 | pte_val(pte) |= _PAGE_USER | _PAGE_EXEC; return pte; } | |
559 | static inline pte_t pte_mkwrite(pte_t pte) { | |
560 | pte_val(pte) |= _PAGE_RW; return pte; } | |
561 | static inline pte_t pte_mkdirty(pte_t pte) { | |
562 | pte_val(pte) |= _PAGE_DIRTY; return pte; } | |
563 | static inline pte_t pte_mkyoung(pte_t pte) { | |
564 | pte_val(pte) |= _PAGE_ACCESSED; return pte; } | |
565 | ||
566 | static inline pte_t pte_modify(pte_t pte, pgprot_t newprot) | |
567 | { | |
568 | pte_val(pte) = (pte_val(pte) & _PAGE_CHG_MASK) | pgprot_val(newprot); | |
569 | return pte; | |
570 | } | |
571 | ||
572 | /* | |
573 | * When flushing the tlb entry for a page, we also need to flush the hash | |
574 | * table entry. flush_hash_pages is assembler (for speed) in hashtable.S. | |
575 | */ | |
576 | extern int flush_hash_pages(unsigned context, unsigned long va, | |
577 | unsigned long pmdval, int count); | |
578 | ||
579 | /* Add an HPTE to the hash table */ | |
580 | extern void add_hash_page(unsigned context, unsigned long va, | |
581 | unsigned long pmdval); | |
582 | ||
583 | /* | |
584 | * Atomic PTE updates. | |
585 | * | |
586 | * pte_update clears and sets bit atomically, and returns | |
587 | * the old pte value. In the 64-bit PTE case we lock around the | |
588 | * low PTE word since we expect ALL flag bits to be there | |
589 | */ | |
590 | #ifndef CONFIG_PTE_64BIT | |
591 | static inline unsigned long pte_update(pte_t *p, unsigned long clr, | |
592 | unsigned long set) | |
593 | { | |
594 | unsigned long old, tmp; | |
595 | ||
596 | __asm__ __volatile__("\ | |
597 | 1: lwarx %0,0,%3\n\ | |
598 | andc %1,%0,%4\n\ | |
599 | or %1,%1,%5\n" | |
600 | PPC405_ERR77(0,%3) | |
601 | " stwcx. %1,0,%3\n\ | |
602 | bne- 1b" | |
603 | : "=&r" (old), "=&r" (tmp), "=m" (*p) | |
604 | : "r" (p), "r" (clr), "r" (set), "m" (*p) | |
605 | : "cc" ); | |
606 | return old; | |
607 | } | |
608 | #else | |
609 | static inline unsigned long long pte_update(pte_t *p, unsigned long clr, | |
610 | unsigned long set) | |
611 | { | |
612 | unsigned long long old; | |
613 | unsigned long tmp; | |
614 | ||
615 | __asm__ __volatile__("\ | |
616 | 1: lwarx %L0,0,%4\n\ | |
617 | lwzx %0,0,%3\n\ | |
618 | andc %1,%L0,%5\n\ | |
619 | or %1,%1,%6\n" | |
620 | PPC405_ERR77(0,%3) | |
621 | " stwcx. %1,0,%4\n\ | |
622 | bne- 1b" | |
623 | : "=&r" (old), "=&r" (tmp), "=m" (*p) | |
624 | : "r" (p), "r" ((unsigned long)(p) + 4), "r" (clr), "r" (set), "m" (*p) | |
625 | : "cc" ); | |
626 | return old; | |
627 | } | |
628 | #endif | |
629 | ||
630 | /* | |
631 | * set_pte stores a linux PTE into the linux page table. | |
632 | * On machines which use an MMU hash table we avoid changing the | |
633 | * _PAGE_HASHPTE bit. | |
634 | */ | |
635 | static inline void set_pte_at(struct mm_struct *mm, unsigned long addr, | |
636 | pte_t *ptep, pte_t pte) | |
637 | { | |
638 | #if _PAGE_HASHPTE != 0 | |
639 | pte_update(ptep, ~_PAGE_HASHPTE, pte_val(pte) & ~_PAGE_HASHPTE); | |
640 | #else | |
641 | *ptep = pte; | |
642 | #endif | |
643 | } | |
644 | ||
645 | /* | |
646 | * 2.6 calles this without flushing the TLB entry, this is wrong | |
647 | * for our hash-based implementation, we fix that up here | |
648 | */ | |
649 | #define __HAVE_ARCH_PTEP_TEST_AND_CLEAR_YOUNG | |
650 | static inline int __ptep_test_and_clear_young(unsigned int context, unsigned long addr, pte_t *ptep) | |
651 | { | |
652 | unsigned long old; | |
653 | old = pte_update(ptep, _PAGE_ACCESSED, 0); | |
654 | #if _PAGE_HASHPTE != 0 | |
655 | if (old & _PAGE_HASHPTE) { | |
656 | unsigned long ptephys = __pa(ptep) & PAGE_MASK; | |
657 | flush_hash_pages(context, addr, ptephys, 1); | |
658 | } | |
659 | #endif | |
660 | return (old & _PAGE_ACCESSED) != 0; | |
661 | } | |
662 | #define ptep_test_and_clear_young(__vma, __addr, __ptep) \ | |
663 | __ptep_test_and_clear_young((__vma)->vm_mm->context.id, __addr, __ptep) | |
664 | ||
665 | #define __HAVE_ARCH_PTEP_TEST_AND_CLEAR_DIRTY | |
666 | static inline int ptep_test_and_clear_dirty(struct vm_area_struct *vma, | |
667 | unsigned long addr, pte_t *ptep) | |
668 | { | |
669 | return (pte_update(ptep, (_PAGE_DIRTY | _PAGE_HWWRITE), 0) & _PAGE_DIRTY) != 0; | |
670 | } | |
671 | ||
672 | #define __HAVE_ARCH_PTEP_GET_AND_CLEAR | |
673 | static inline pte_t ptep_get_and_clear(struct mm_struct *mm, unsigned long addr, | |
674 | pte_t *ptep) | |
675 | { | |
676 | return __pte(pte_update(ptep, ~_PAGE_HASHPTE, 0)); | |
677 | } | |
678 | ||
679 | #define __HAVE_ARCH_PTEP_SET_WRPROTECT | |
680 | static inline void ptep_set_wrprotect(struct mm_struct *mm, unsigned long addr, | |
681 | pte_t *ptep) | |
682 | { | |
683 | pte_update(ptep, (_PAGE_RW | _PAGE_HWWRITE), 0); | |
684 | } | |
685 | ||
686 | #define __HAVE_ARCH_PTEP_SET_ACCESS_FLAGS | |
687 | static inline void __ptep_set_access_flags(pte_t *ptep, pte_t entry, int dirty) | |
688 | { | |
689 | unsigned long bits = pte_val(entry) & | |
690 | (_PAGE_DIRTY | _PAGE_ACCESSED | _PAGE_RW); | |
691 | pte_update(ptep, 0, bits); | |
692 | } | |
693 | ||
694 | #define ptep_set_access_flags(__vma, __address, __ptep, __entry, __dirty) \ | |
695 | do { \ | |
696 | __ptep_set_access_flags(__ptep, __entry, __dirty); \ | |
697 | flush_tlb_page_nohash(__vma, __address); \ | |
698 | } while(0) | |
699 | ||
700 | /* | |
701 | * Macro to mark a page protection value as "uncacheable". | |
702 | */ | |
703 | #define pgprot_noncached(prot) (__pgprot(pgprot_val(prot) | _PAGE_NO_CACHE | _PAGE_GUARDED)) | |
704 | ||
705 | struct file; | |
706 | extern pgprot_t phys_mem_access_prot(struct file *file, unsigned long pfn, | |
707 | unsigned long size, pgprot_t vma_prot); | |
708 | #define __HAVE_PHYS_MEM_ACCESS_PROT | |
709 | ||
710 | #define __HAVE_ARCH_PTE_SAME | |
711 | #define pte_same(A,B) (((pte_val(A) ^ pte_val(B)) & ~_PAGE_HASHPTE) == 0) | |
712 | ||
713 | /* | |
714 | * Note that on Book E processors, the pmd contains the kernel virtual | |
715 | * (lowmem) address of the pte page. The physical address is less useful | |
716 | * because everything runs with translation enabled (even the TLB miss | |
717 | * handler). On everything else the pmd contains the physical address | |
718 | * of the pte page. -- paulus | |
719 | */ | |
720 | #ifndef CONFIG_BOOKE | |
721 | #define pmd_page_vaddr(pmd) \ | |
722 | ((unsigned long) __va(pmd_val(pmd) & PAGE_MASK)) | |
723 | #define pmd_page(pmd) \ | |
724 | (mem_map + (pmd_val(pmd) >> PAGE_SHIFT)) | |
725 | #else | |
726 | #define pmd_page_vaddr(pmd) \ | |
727 | ((unsigned long) (pmd_val(pmd) & PAGE_MASK)) | |
728 | #define pmd_page(pmd) \ | |
729 | (mem_map + (__pa(pmd_val(pmd)) >> PAGE_SHIFT)) | |
730 | #endif | |
731 | ||
732 | /* to find an entry in a kernel page-table-directory */ | |
733 | #define pgd_offset_k(address) pgd_offset(&init_mm, address) | |
734 | ||
735 | /* to find an entry in a page-table-directory */ | |
736 | #define pgd_index(address) ((address) >> PGDIR_SHIFT) | |
737 | #define pgd_offset(mm, address) ((mm)->pgd + pgd_index(address)) | |
738 | ||
739 | /* Find an entry in the second-level page table.. */ | |
740 | static inline pmd_t * pmd_offset(pgd_t * dir, unsigned long address) | |
741 | { | |
742 | return (pmd_t *) dir; | |
743 | } | |
744 | ||
745 | /* Find an entry in the third-level page table.. */ | |
746 | #define pte_index(address) \ | |
747 | (((address) >> PAGE_SHIFT) & (PTRS_PER_PTE - 1)) | |
748 | #define pte_offset_kernel(dir, addr) \ | |
749 | ((pte_t *) pmd_page_vaddr(*(dir)) + pte_index(addr)) | |
750 | #define pte_offset_map(dir, addr) \ | |
751 | ((pte_t *) kmap_atomic(pmd_page(*(dir)), KM_PTE0) + pte_index(addr)) | |
752 | #define pte_offset_map_nested(dir, addr) \ | |
753 | ((pte_t *) kmap_atomic(pmd_page(*(dir)), KM_PTE1) + pte_index(addr)) | |
754 | ||
755 | #define pte_unmap(pte) kunmap_atomic(pte, KM_PTE0) | |
756 | #define pte_unmap_nested(pte) kunmap_atomic(pte, KM_PTE1) | |
757 | ||
758 | extern pgd_t swapper_pg_dir[PTRS_PER_PGD]; | |
759 | ||
760 | extern void paging_init(void); | |
761 | ||
762 | /* | |
763 | * Encode and decode a swap entry. | |
764 | * Note that the bits we use in a PTE for representing a swap entry | |
765 | * must not include the _PAGE_PRESENT bit, the _PAGE_FILE bit, or the | |
766 | *_PAGE_HASHPTE bit (if used). -- paulus | |
767 | */ | |
768 | #define __swp_type(entry) ((entry).val & 0x1f) | |
769 | #define __swp_offset(entry) ((entry).val >> 5) | |
770 | #define __swp_entry(type, offset) ((swp_entry_t) { (type) | ((offset) << 5) }) | |
771 | #define __pte_to_swp_entry(pte) ((swp_entry_t) { pte_val(pte) >> 3 }) | |
772 | #define __swp_entry_to_pte(x) ((pte_t) { (x).val << 3 }) | |
773 | ||
774 | /* Encode and decode a nonlinear file mapping entry */ | |
775 | #define PTE_FILE_MAX_BITS 29 | |
776 | #define pte_to_pgoff(pte) (pte_val(pte) >> 3) | |
777 | #define pgoff_to_pte(off) ((pte_t) { ((off) << 3) | _PAGE_FILE }) | |
778 | ||
779 | /* CONFIG_APUS */ | |
780 | /* For virtual address to physical address conversion */ | |
781 | extern void cache_clear(__u32 addr, int length); | |
782 | extern void cache_push(__u32 addr, int length); | |
783 | extern int mm_end_of_chunk (unsigned long addr, int len); | |
784 | extern unsigned long iopa(unsigned long addr); | |
785 | extern unsigned long mm_ptov(unsigned long addr) __attribute_const__; | |
786 | ||
787 | /* Values for nocacheflag and cmode */ | |
788 | /* These are not used by the APUS kernel_map, but prevents | |
789 | compilation errors. */ | |
790 | #define KERNELMAP_FULL_CACHING 0 | |
791 | #define KERNELMAP_NOCACHE_SER 1 | |
792 | #define KERNELMAP_NOCACHE_NONSER 2 | |
793 | #define KERNELMAP_NO_COPYBACK 3 | |
794 | ||
795 | /* | |
796 | * Map some physical address range into the kernel address space. | |
797 | */ | |
798 | extern unsigned long kernel_map(unsigned long paddr, unsigned long size, | |
799 | int nocacheflag, unsigned long *memavailp ); | |
800 | ||
801 | /* | |
802 | * Set cache mode of (kernel space) address range. | |
803 | */ | |
804 | extern void kernel_set_cachemode (unsigned long address, unsigned long size, | |
805 | unsigned int cmode); | |
806 | ||
807 | /* Needs to be defined here and not in linux/mm.h, as it is arch dependent */ | |
808 | #define kern_addr_valid(addr) (1) | |
809 | ||
810 | #ifdef CONFIG_PHYS_64BIT | |
811 | extern int remap_pfn_range(struct vm_area_struct *vma, unsigned long from, | |
812 | unsigned long paddr, unsigned long size, pgprot_t prot); | |
813 | ||
814 | static inline int io_remap_pfn_range(struct vm_area_struct *vma, | |
815 | unsigned long vaddr, | |
816 | unsigned long pfn, | |
817 | unsigned long size, | |
818 | pgprot_t prot) | |
819 | { | |
820 | phys_addr_t paddr64 = fixup_bigphys_addr(pfn << PAGE_SHIFT, size); | |
821 | return remap_pfn_range(vma, vaddr, paddr64 >> PAGE_SHIFT, size, prot); | |
822 | } | |
823 | #else | |
824 | #define io_remap_pfn_range(vma, vaddr, pfn, size, prot) \ | |
825 | remap_pfn_range(vma, vaddr, pfn, size, prot) | |
826 | #endif | |
827 | ||
828 | /* | |
829 | * No page table caches to initialise | |
830 | */ | |
831 | #define pgtable_cache_init() do { } while (0) | |
832 | ||
833 | extern int get_pteptr(struct mm_struct *mm, unsigned long addr, pte_t **ptep, | |
834 | pmd_t **pmdp); | |
835 | ||
836 | #endif /* !__ASSEMBLY__ */ | |
837 | ||
838 | #endif /* _ASM_POWERPC_PGTABLE_PPC32_H */ |