]>
Commit | Line | Data |
---|---|---|
9b6b563c PM |
1 | /* |
2 | * Procedures for creating, accessing and interpreting the device tree. | |
3 | * | |
4 | * Paul Mackerras August 1996. | |
5 | * Copyright (C) 1996-2005 Paul Mackerras. | |
6 | * | |
7 | * Adapted for 64bit PowerPC by Dave Engebretsen and Peter Bergner. | |
8 | * {engebret|bergner}@us.ibm.com | |
9 | * | |
10 | * This program is free software; you can redistribute it and/or | |
11 | * modify it under the terms of the GNU General Public License | |
12 | * as published by the Free Software Foundation; either version | |
13 | * 2 of the License, or (at your option) any later version. | |
14 | */ | |
15 | ||
16 | #undef DEBUG | |
17 | ||
18 | #include <stdarg.h> | |
9b6b563c PM |
19 | #include <linux/kernel.h> |
20 | #include <linux/string.h> | |
21 | #include <linux/init.h> | |
22 | #include <linux/threads.h> | |
23 | #include <linux/spinlock.h> | |
24 | #include <linux/types.h> | |
25 | #include <linux/pci.h> | |
26 | #include <linux/stringify.h> | |
27 | #include <linux/delay.h> | |
28 | #include <linux/initrd.h> | |
29 | #include <linux/bitops.h> | |
4b16f8e2 | 30 | #include <linux/export.h> |
dcee3036 | 31 | #include <linux/kexec.h> |
7a4571ae | 32 | #include <linux/debugfs.h> |
0ebfff14 | 33 | #include <linux/irq.h> |
95f72d1e | 34 | #include <linux/memblock.h> |
9b6b563c PM |
35 | |
36 | #include <asm/prom.h> | |
37 | #include <asm/rtas.h> | |
9b6b563c PM |
38 | #include <asm/page.h> |
39 | #include <asm/processor.h> | |
40 | #include <asm/irq.h> | |
41 | #include <asm/io.h> | |
0cc4746c | 42 | #include <asm/kdump.h> |
9b6b563c PM |
43 | #include <asm/smp.h> |
44 | #include <asm/system.h> | |
45 | #include <asm/mmu.h> | |
1426d5a3 | 46 | #include <asm/paca.h> |
9b6b563c PM |
47 | #include <asm/pgtable.h> |
48 | #include <asm/pci.h> | |
49 | #include <asm/iommu.h> | |
50 | #include <asm/btext.h> | |
51 | #include <asm/sections.h> | |
52 | #include <asm/machdep.h> | |
53 | #include <asm/pSeries_reconfig.h> | |
40ef8cbc | 54 | #include <asm/pci-bridge.h> |
6ac26c8a | 55 | #include <asm/phyp_dump.h> |
2babf5c2 | 56 | #include <asm/kexec.h> |
14a43e69 BH |
57 | #include <asm/opal.h> |
58 | ||
37dd2bad | 59 | #include <mm/mmu_decl.h> |
9b6b563c PM |
60 | |
61 | #ifdef DEBUG | |
62 | #define DBG(fmt...) printk(KERN_ERR fmt) | |
63 | #else | |
64 | #define DBG(fmt...) | |
65 | #endif | |
66 | ||
9b6b563c | 67 | #ifdef CONFIG_PPC64 |
28897731 | 68 | int __initdata iommu_is_off; |
9b6b563c | 69 | int __initdata iommu_force_on; |
cf00a8d1 | 70 | unsigned long tce_alloc_start, tce_alloc_end; |
cd3db0c4 | 71 | u64 ppc64_rma_size; |
9b6b563c | 72 | #endif |
03bf469a | 73 | static phys_addr_t first_memblock_size; |
7ac87abb | 74 | static int __initdata boot_cpu_count; |
9b6b563c | 75 | |
2babf5c2 ME |
76 | static int __init early_parse_mem(char *p) |
77 | { | |
78 | if (!p) | |
79 | return 1; | |
80 | ||
81 | memory_limit = PAGE_ALIGN(memparse(p, &p)); | |
49a84965 | 82 | DBG("memory limit = 0x%llx\n", (unsigned long long)memory_limit); |
2babf5c2 ME |
83 | |
84 | return 0; | |
85 | } | |
86 | early_param("mem", early_parse_mem); | |
87 | ||
307cfe71 BH |
88 | /* |
89 | * overlaps_initrd - check for overlap with page aligned extension of | |
90 | * initrd. | |
91 | */ | |
92 | static inline int overlaps_initrd(unsigned long start, unsigned long size) | |
93 | { | |
94 | #ifdef CONFIG_BLK_DEV_INITRD | |
95 | if (!initrd_start) | |
96 | return 0; | |
97 | ||
98 | return (start + size) > _ALIGN_DOWN(initrd_start, PAGE_SIZE) && | |
99 | start <= _ALIGN_UP(initrd_end, PAGE_SIZE); | |
100 | #else | |
101 | return 0; | |
102 | #endif | |
103 | } | |
104 | ||
3c607ce2 LV |
105 | /** |
106 | * move_device_tree - move tree to an unused area, if needed. | |
107 | * | |
108 | * The device tree may be allocated beyond our memory limit, or inside the | |
307cfe71 BH |
109 | * crash kernel region for kdump, or within the page aligned range of initrd. |
110 | * If so, move it out of the way. | |
2babf5c2 | 111 | */ |
18f032cb | 112 | static void __init move_device_tree(void) |
2babf5c2 ME |
113 | { |
114 | unsigned long start, size; | |
115 | void *p; | |
116 | ||
117 | DBG("-> move_device_tree\n"); | |
118 | ||
119 | start = __pa(initial_boot_params); | |
087f79c4 | 120 | size = be32_to_cpu(initial_boot_params->totalsize); |
2babf5c2 | 121 | |
6dd22700 | 122 | if ((memory_limit && (start + size) > PHYSICAL_START + memory_limit) || |
307cfe71 BH |
123 | overlaps_crashkernel(start, size) || |
124 | overlaps_initrd(start, size)) { | |
e63075a3 | 125 | p = __va(memblock_alloc(size, PAGE_SIZE)); |
2babf5c2 ME |
126 | memcpy(p, initial_boot_params, size); |
127 | initial_boot_params = (struct boot_param_header *)p; | |
128 | DBG("Moved device tree to 0x%p\n", p); | |
129 | } | |
130 | ||
131 | DBG("<- move_device_tree\n"); | |
132 | } | |
9b6b563c | 133 | |
d205819e PM |
134 | /* |
135 | * ibm,pa-features is a per-cpu property that contains a string of | |
136 | * attribute descriptors, each of which has a 2 byte header plus up | |
137 | * to 254 bytes worth of processor attribute bits. First header | |
138 | * byte specifies the number of bytes following the header. | |
139 | * Second header byte is an "attribute-specifier" type, of which | |
140 | * zero is the only currently-defined value. | |
141 | * Implementation: Pass in the byte and bit offset for the feature | |
142 | * that we are interested in. The function will return -1 if the | |
143 | * pa-features property is missing, or a 1/0 to indicate if the feature | |
144 | * is supported/not supported. Note that the bit numbers are | |
145 | * big-endian to match the definition in PAPR. | |
146 | */ | |
147 | static struct ibm_pa_feature { | |
148 | unsigned long cpu_features; /* CPU_FTR_xxx bit */ | |
44ae3ab3 | 149 | unsigned long mmu_features; /* MMU_FTR_xxx bit */ |
d205819e PM |
150 | unsigned int cpu_user_ftrs; /* PPC_FEATURE_xxx bit */ |
151 | unsigned char pabyte; /* byte number in ibm,pa-features */ | |
152 | unsigned char pabit; /* bit number (big-endian) */ | |
153 | unsigned char invert; /* if 1, pa bit set => clear feature */ | |
154 | } ibm_pa_features[] __initdata = { | |
44ae3ab3 ME |
155 | {0, 0, PPC_FEATURE_HAS_MMU, 0, 0, 0}, |
156 | {0, 0, PPC_FEATURE_HAS_FPU, 0, 1, 0}, | |
157 | {0, MMU_FTR_SLB, 0, 0, 2, 0}, | |
158 | {CPU_FTR_CTRL, 0, 0, 0, 3, 0}, | |
159 | {CPU_FTR_NOEXECUTE, 0, 0, 0, 6, 0}, | |
160 | {CPU_FTR_NODSISRALIGN, 0, 0, 1, 1, 1}, | |
161 | {0, MMU_FTR_CI_LARGE_PAGE, 0, 1, 2, 0}, | |
339d76c5 | 162 | {CPU_FTR_REAL_LE, PPC_FEATURE_TRUE_LE, 5, 0, 0}, |
d205819e PM |
163 | }; |
164 | ||
974a76f5 PM |
165 | static void __init scan_features(unsigned long node, unsigned char *ftrs, |
166 | unsigned long tablelen, | |
167 | struct ibm_pa_feature *fp, | |
168 | unsigned long ft_size) | |
d205819e | 169 | { |
974a76f5 | 170 | unsigned long i, len, bit; |
d205819e PM |
171 | |
172 | /* find descriptor with type == 0 */ | |
173 | for (;;) { | |
174 | if (tablelen < 3) | |
175 | return; | |
974a76f5 | 176 | len = 2 + ftrs[0]; |
d205819e PM |
177 | if (tablelen < len) |
178 | return; /* descriptor 0 not found */ | |
974a76f5 | 179 | if (ftrs[1] == 0) |
d205819e PM |
180 | break; |
181 | tablelen -= len; | |
974a76f5 | 182 | ftrs += len; |
d205819e PM |
183 | } |
184 | ||
185 | /* loop over bits we know about */ | |
974a76f5 PM |
186 | for (i = 0; i < ft_size; ++i, ++fp) { |
187 | if (fp->pabyte >= ftrs[0]) | |
d205819e | 188 | continue; |
974a76f5 | 189 | bit = (ftrs[2 + fp->pabyte] >> (7 - fp->pabit)) & 1; |
d205819e PM |
190 | if (bit ^ fp->invert) { |
191 | cur_cpu_spec->cpu_features |= fp->cpu_features; | |
192 | cur_cpu_spec->cpu_user_features |= fp->cpu_user_ftrs; | |
44ae3ab3 | 193 | cur_cpu_spec->mmu_features |= fp->mmu_features; |
d205819e PM |
194 | } else { |
195 | cur_cpu_spec->cpu_features &= ~fp->cpu_features; | |
196 | cur_cpu_spec->cpu_user_features &= ~fp->cpu_user_ftrs; | |
44ae3ab3 | 197 | cur_cpu_spec->mmu_features &= ~fp->mmu_features; |
d205819e PM |
198 | } |
199 | } | |
200 | } | |
201 | ||
974a76f5 PM |
202 | static void __init check_cpu_pa_features(unsigned long node) |
203 | { | |
204 | unsigned char *pa_ftrs; | |
205 | unsigned long tablelen; | |
206 | ||
207 | pa_ftrs = of_get_flat_dt_prop(node, "ibm,pa-features", &tablelen); | |
208 | if (pa_ftrs == NULL) | |
209 | return; | |
210 | ||
211 | scan_features(node, pa_ftrs, tablelen, | |
212 | ibm_pa_features, ARRAY_SIZE(ibm_pa_features)); | |
213 | } | |
214 | ||
94491685 | 215 | #ifdef CONFIG_PPC_STD_MMU_64 |
584f8b71 MN |
216 | static void __init check_cpu_slb_size(unsigned long node) |
217 | { | |
218 | u32 *slb_size_ptr; | |
219 | ||
b60c31d8 MN |
220 | slb_size_ptr = of_get_flat_dt_prop(node, "slb-size", NULL); |
221 | if (slb_size_ptr != NULL) { | |
222 | mmu_slb_size = *slb_size_ptr; | |
223 | return; | |
224 | } | |
584f8b71 MN |
225 | slb_size_ptr = of_get_flat_dt_prop(node, "ibm,slb-size", NULL); |
226 | if (slb_size_ptr != NULL) { | |
227 | mmu_slb_size = *slb_size_ptr; | |
228 | } | |
229 | } | |
230 | #else | |
231 | #define check_cpu_slb_size(node) do { } while(0) | |
232 | #endif | |
233 | ||
974a76f5 PM |
234 | static struct feature_property { |
235 | const char *name; | |
236 | u32 min_value; | |
237 | unsigned long cpu_feature; | |
238 | unsigned long cpu_user_ftr; | |
239 | } feature_properties[] __initdata = { | |
240 | #ifdef CONFIG_ALTIVEC | |
241 | {"altivec", 0, CPU_FTR_ALTIVEC, PPC_FEATURE_HAS_ALTIVEC}, | |
242 | {"ibm,vmx", 1, CPU_FTR_ALTIVEC, PPC_FEATURE_HAS_ALTIVEC}, | |
243 | #endif /* CONFIG_ALTIVEC */ | |
b962ce9d MN |
244 | #ifdef CONFIG_VSX |
245 | /* Yes, this _really_ is ibm,vmx == 2 to enable VSX */ | |
246 | {"ibm,vmx", 2, CPU_FTR_VSX, PPC_FEATURE_HAS_VSX}, | |
247 | #endif /* CONFIG_VSX */ | |
974a76f5 PM |
248 | #ifdef CONFIG_PPC64 |
249 | {"ibm,dfp", 1, 0, PPC_FEATURE_HAS_DFP}, | |
250 | {"ibm,purr", 1, CPU_FTR_PURR, 0}, | |
251 | {"ibm,spurr", 1, CPU_FTR_SPURR, 0}, | |
252 | #endif /* CONFIG_PPC64 */ | |
253 | }; | |
254 | ||
14b3d926 VB |
255 | #if defined(CONFIG_44x) && defined(CONFIG_PPC_FPU) |
256 | static inline void identical_pvr_fixup(unsigned long node) | |
257 | { | |
258 | unsigned int pvr; | |
259 | char *model = of_get_flat_dt_prop(node, "model", NULL); | |
260 | ||
261 | /* | |
262 | * Since 440GR(x)/440EP(x) processors have the same pvr, | |
263 | * we check the node path and set bit 28 in the cur_cpu_spec | |
264 | * pvr for EP(x) processor version. This bit is always 0 in | |
265 | * the "real" pvr. Then we call identify_cpu again with | |
266 | * the new logical pvr to enable FPU support. | |
267 | */ | |
268 | if (model && strstr(model, "440EP")) { | |
269 | pvr = cur_cpu_spec->pvr_value | 0x8; | |
270 | identify_cpu(0, pvr); | |
271 | DBG("Using logical pvr %x for %s\n", pvr, model); | |
272 | } | |
273 | } | |
274 | #else | |
275 | #define identical_pvr_fixup(node) do { } while(0) | |
276 | #endif | |
277 | ||
974a76f5 PM |
278 | static void __init check_cpu_feature_properties(unsigned long node) |
279 | { | |
280 | unsigned long i; | |
281 | struct feature_property *fp = feature_properties; | |
282 | const u32 *prop; | |
283 | ||
284 | for (i = 0; i < ARRAY_SIZE(feature_properties); ++i, ++fp) { | |
285 | prop = of_get_flat_dt_prop(node, fp->name, NULL); | |
286 | if (prop && *prop >= fp->min_value) { | |
287 | cur_cpu_spec->cpu_features |= fp->cpu_feature; | |
288 | cur_cpu_spec->cpu_user_features |= fp->cpu_user_ftr; | |
289 | } | |
290 | } | |
291 | } | |
292 | ||
9b6b563c | 293 | static int __init early_init_dt_scan_cpus(unsigned long node, |
4df20460 AB |
294 | const char *uname, int depth, |
295 | void *data) | |
9b6b563c | 296 | { |
4df20460 | 297 | char *type = of_get_flat_dt_prop(node, "device_type", NULL); |
974a76f5 PM |
298 | const u32 *prop; |
299 | const u32 *intserv; | |
4df20460 AB |
300 | int i, nthreads; |
301 | unsigned long len; | |
9d07bc84 | 302 | int found = -1; |
208b3a4c | 303 | int found_thread = 0; |
9b6b563c PM |
304 | |
305 | /* We are scanning "cpu" nodes only */ | |
306 | if (type == NULL || strcmp(type, "cpu") != 0) | |
307 | return 0; | |
308 | ||
4df20460 AB |
309 | /* Get physical cpuid */ |
310 | intserv = of_get_flat_dt_prop(node, "ibm,ppc-interrupt-server#s", &len); | |
311 | if (intserv) { | |
312 | nthreads = len / sizeof(int); | |
9b6b563c | 313 | } else { |
4df20460 AB |
314 | intserv = of_get_flat_dt_prop(node, "reg", NULL); |
315 | nthreads = 1; | |
316 | } | |
317 | ||
318 | /* | |
319 | * Now see if any of these threads match our boot cpu. | |
320 | * NOTE: This must match the parsing done in smp_setup_cpu_maps. | |
321 | */ | |
322 | for (i = 0; i < nthreads; i++) { | |
323 | /* | |
324 | * version 2 of the kexec param format adds the phys cpuid of | |
325 | * booted proc. | |
326 | */ | |
208b3a4c BH |
327 | if (initial_boot_params->version >= 2) { |
328 | if (intserv[i] == initial_boot_params->boot_cpuid_phys) { | |
9d07bc84 | 329 | found = boot_cpu_count; |
208b3a4c BH |
330 | found_thread = i; |
331 | } | |
4df20460 AB |
332 | } else { |
333 | /* | |
334 | * Check if it's the boot-cpu, set it's hw index now, | |
335 | * unfortunately this format did not support booting | |
336 | * off secondary threads. | |
337 | */ | |
338 | if (of_get_flat_dt_prop(node, | |
9d07bc84 BH |
339 | "linux,boot-cpu", NULL) != NULL) |
340 | found = boot_cpu_count; | |
9b6b563c | 341 | } |
4df20460 AB |
342 | #ifdef CONFIG_SMP |
343 | /* logical cpu id is always 0 on UP kernels */ | |
9d07bc84 | 344 | boot_cpu_count++; |
4df20460 AB |
345 | #endif |
346 | } | |
347 | ||
9d07bc84 BH |
348 | if (found >= 0) { |
349 | DBG("boot cpu: logical %d physical %d\n", found, | |
208b3a4c | 350 | intserv[found_thread]); |
9d07bc84 | 351 | boot_cpuid = found; |
208b3a4c | 352 | set_hard_smp_processor_id(found, intserv[found_thread]); |
9b6b563c | 353 | |
974a76f5 PM |
354 | /* |
355 | * PAPR defines "logical" PVR values for cpus that | |
356 | * meet various levels of the architecture: | |
357 | * 0x0f000001 Architecture version 2.04 | |
358 | * 0x0f000002 Architecture version 2.05 | |
359 | * If the cpu-version property in the cpu node contains | |
360 | * such a value, we call identify_cpu again with the | |
361 | * logical PVR value in order to use the cpu feature | |
362 | * bits appropriate for the architecture level. | |
363 | * | |
364 | * A POWER6 partition in "POWER6 architected" mode | |
365 | * uses the 0x0f000002 PVR value; in POWER5+ mode | |
366 | * it uses 0x0f000001. | |
367 | */ | |
368 | prop = of_get_flat_dt_prop(node, "cpu-version", NULL); | |
369 | if (prop && (*prop & 0xff000000) == 0x0f000000) | |
370 | identify_cpu(0, *prop); | |
14b3d926 VB |
371 | |
372 | identical_pvr_fixup(node); | |
9b6b563c | 373 | } |
9b6b563c | 374 | |
974a76f5 | 375 | check_cpu_feature_properties(node); |
d205819e | 376 | check_cpu_pa_features(node); |
584f8b71 | 377 | check_cpu_slb_size(node); |
d205819e | 378 | |
9b6b563c | 379 | #ifdef CONFIG_PPC_PSERIES |
4df20460 | 380 | if (nthreads > 1) |
9b6b563c | 381 | cur_cpu_spec->cpu_features |= CPU_FTR_SMT; |
4df20460 AB |
382 | else |
383 | cur_cpu_spec->cpu_features &= ~CPU_FTR_SMT; | |
9b6b563c PM |
384 | #endif |
385 | ||
386 | return 0; | |
387 | } | |
388 | ||
32c97689 GL |
389 | int __init early_init_dt_scan_chosen_ppc(unsigned long node, const char *uname, |
390 | int depth, void *data) | |
9b6b563c | 391 | { |
9b6b563c | 392 | unsigned long *lprop; |
9b6b563c | 393 | |
32c97689 GL |
394 | /* Use common scan routine to determine if this is the chosen node */ |
395 | if (early_init_dt_scan_chosen(node, uname, depth, data) == 0) | |
396 | return 0; | |
397 | ||
9b6b563c PM |
398 | #ifdef CONFIG_PPC64 |
399 | /* check if iommu is forced on or off */ | |
3c726f8d | 400 | if (of_get_flat_dt_prop(node, "linux,iommu-off", NULL) != NULL) |
9b6b563c | 401 | iommu_is_off = 1; |
3c726f8d | 402 | if (of_get_flat_dt_prop(node, "linux,iommu-force-on", NULL) != NULL) |
9b6b563c PM |
403 | iommu_force_on = 1; |
404 | #endif | |
405 | ||
2babf5c2 | 406 | /* mem=x on the command line is the preferred mechanism */ |
86e03221 GL |
407 | lprop = of_get_flat_dt_prop(node, "linux,memory-limit", NULL); |
408 | if (lprop) | |
409 | memory_limit = *lprop; | |
9b6b563c PM |
410 | |
411 | #ifdef CONFIG_PPC64 | |
86e03221 GL |
412 | lprop = of_get_flat_dt_prop(node, "linux,tce-alloc-start", NULL); |
413 | if (lprop) | |
414 | tce_alloc_start = *lprop; | |
415 | lprop = of_get_flat_dt_prop(node, "linux,tce-alloc-end", NULL); | |
416 | if (lprop) | |
417 | tce_alloc_end = *lprop; | |
9b6b563c PM |
418 | #endif |
419 | ||
dcee3036 | 420 | #ifdef CONFIG_KEXEC |
63277161 | 421 | lprop = of_get_flat_dt_prop(node, "linux,crashkernel-base", NULL); |
70c6cc37 LV |
422 | if (lprop) |
423 | crashk_res.start = *lprop; | |
dcee3036 | 424 | |
63277161 | 425 | lprop = of_get_flat_dt_prop(node, "linux,crashkernel-size", NULL); |
70c6cc37 LV |
426 | if (lprop) |
427 | crashk_res.end = crashk_res.start + *lprop - 1; | |
dcee3036 | 428 | #endif |
32c97689 GL |
429 | |
430 | /* break now */ | |
431 | return 1; | |
9b6b563c PM |
432 | } |
433 | ||
0204568a PM |
434 | #ifdef CONFIG_PPC_PSERIES |
435 | /* | |
436 | * Interpret the ibm,dynamic-memory property in the | |
437 | * /ibm,dynamic-reconfiguration-memory node. | |
438 | * This contains a list of memory blocks along with NUMA affinity | |
439 | * information. | |
440 | */ | |
441 | static int __init early_init_dt_scan_drconf_memory(unsigned long node) | |
442 | { | |
0f0b56c3 | 443 | __be32 *dm, *ls, *usm; |
abe76885 | 444 | unsigned long l, n, flags; |
95f72d1e | 445 | u64 base, size, memblock_size; |
cf00085d | 446 | unsigned int is_kexec_kdump = 0, rngs; |
0204568a | 447 | |
3fdfd990 | 448 | ls = of_get_flat_dt_prop(node, "ibm,lmb-size", &l); |
0f0b56c3 | 449 | if (ls == NULL || l < dt_root_size_cells * sizeof(__be32)) |
0204568a | 450 | return 0; |
95f72d1e | 451 | memblock_size = dt_mem_next_cell(dt_root_size_cells, &ls); |
0204568a | 452 | |
63277161 | 453 | dm = of_get_flat_dt_prop(node, "ibm,dynamic-memory", &l); |
0f0b56c3 | 454 | if (dm == NULL || l < sizeof(__be32)) |
0204568a PM |
455 | return 0; |
456 | ||
457 | n = *dm++; /* number of entries */ | |
0f0b56c3 | 458 | if (l < (n * (dt_root_addr_cells + 4) + 1) * sizeof(__be32)) |
0204568a PM |
459 | return 0; |
460 | ||
cf00085d | 461 | /* check if this is a kexec/kdump kernel. */ |
63277161 | 462 | usm = of_get_flat_dt_prop(node, "linux,drconf-usable-memory", |
cf00085d C |
463 | &l); |
464 | if (usm != NULL) | |
465 | is_kexec_kdump = 1; | |
466 | ||
0204568a PM |
467 | for (; n != 0; --n) { |
468 | base = dt_mem_next_cell(dt_root_addr_cells, &dm); | |
469 | flags = dm[3]; | |
470 | /* skip DRC index, pad, assoc. list index, flags */ | |
471 | dm += 4; | |
472 | /* skip this block if the reserved bit is set in flags (0x80) | |
473 | or if the block is not assigned to this partition (0x8) */ | |
474 | if ((flags & 0x80) || !(flags & 0x8)) | |
475 | continue; | |
95f72d1e | 476 | size = memblock_size; |
cf00085d C |
477 | rngs = 1; |
478 | if (is_kexec_kdump) { | |
479 | /* | |
95f72d1e | 480 | * For each memblock in ibm,dynamic-memory, a corresponding |
cf00085d C |
481 | * entry in linux,drconf-usable-memory property contains |
482 | * a counter 'p' followed by 'p' (base, size) duple. | |
483 | * Now read the counter from | |
484 | * linux,drconf-usable-memory property | |
485 | */ | |
486 | rngs = dt_mem_next_cell(dt_root_size_cells, &usm); | |
487 | if (!rngs) /* there are no (base, size) duple */ | |
0204568a | 488 | continue; |
0204568a | 489 | } |
cf00085d C |
490 | do { |
491 | if (is_kexec_kdump) { | |
492 | base = dt_mem_next_cell(dt_root_addr_cells, | |
493 | &usm); | |
494 | size = dt_mem_next_cell(dt_root_size_cells, | |
495 | &usm); | |
496 | } | |
497 | if (iommu_is_off) { | |
498 | if (base >= 0x80000000ul) | |
499 | continue; | |
500 | if ((base + size) > 0x80000000ul) | |
501 | size = 0x80000000ul - base; | |
502 | } | |
95f72d1e | 503 | memblock_add(base, size); |
cf00085d | 504 | } while (--rngs); |
0204568a | 505 | } |
95f72d1e | 506 | memblock_dump_all(); |
0204568a PM |
507 | return 0; |
508 | } | |
509 | #else | |
510 | #define early_init_dt_scan_drconf_memory(node) 0 | |
511 | #endif /* CONFIG_PPC_PSERIES */ | |
9b6b563c | 512 | |
51975db0 GL |
513 | static int __init early_init_dt_scan_memory_ppc(unsigned long node, |
514 | const char *uname, | |
515 | int depth, void *data) | |
9b6b563c | 516 | { |
0204568a PM |
517 | if (depth == 1 && |
518 | strcmp(uname, "ibm,dynamic-reconfiguration-memory") == 0) | |
519 | return early_init_dt_scan_drconf_memory(node); | |
51975db0 GL |
520 | |
521 | return early_init_dt_scan_memory(node, uname, depth, data); | |
522 | } | |
0204568a | 523 | |
51975db0 GL |
524 | void __init early_init_dt_add_memory_arch(u64 base, u64 size) |
525 | { | |
cd3db0c4 | 526 | #ifdef CONFIG_PPC64 |
51975db0 GL |
527 | if (iommu_is_off) { |
528 | if (base >= 0x80000000ul) | |
529 | return; | |
530 | if ((base + size) > 0x80000000ul) | |
531 | size = 0x80000000ul - base; | |
532 | } | |
9b6b563c | 533 | #endif |
03bf469a BH |
534 | /* Keep track of the beginning of memory -and- the size of |
535 | * the very first block in the device-tree as it represents | |
536 | * the RMA on ppc64 server | |
537 | */ | |
538 | if (base < memstart_addr) { | |
539 | memstart_addr = base; | |
540 | first_memblock_size = size; | |
541 | } | |
cd3db0c4 BH |
542 | |
543 | /* Add the chunk to the MEMBLOCK list */ | |
544 | memblock_add(base, size); | |
9b6b563c PM |
545 | } |
546 | ||
672c5446 | 547 | void * __init early_init_dt_alloc_memory_arch(u64 size, u64 align) |
4ef7b373 | 548 | { |
672c5446 | 549 | return __va(memblock_alloc(size, align)); |
4ef7b373 JK |
550 | } |
551 | ||
1406bc2f JK |
552 | #ifdef CONFIG_BLK_DEV_INITRD |
553 | void __init early_init_dt_setup_initrd_arch(unsigned long start, | |
554 | unsigned long end) | |
555 | { | |
556 | initrd_start = (unsigned long)__va(start); | |
557 | initrd_end = (unsigned long)__va(end); | |
558 | initrd_below_start_ok = 1; | |
559 | } | |
560 | #endif | |
561 | ||
9b6b563c PM |
562 | static void __init early_reserve_mem(void) |
563 | { | |
cbbcf340 KG |
564 | u64 base, size; |
565 | u64 *reserve_map; | |
8a300887 JL |
566 | unsigned long self_base; |
567 | unsigned long self_size; | |
9b6b563c | 568 | |
cbbcf340 | 569 | reserve_map = (u64 *)(((unsigned long)initial_boot_params) + |
9b6b563c | 570 | initial_boot_params->off_mem_rsvmap); |
4d1f3f25 JX |
571 | |
572 | /* before we do anything, lets reserve the dt blob */ | |
8a300887 JL |
573 | self_base = __pa((unsigned long)initial_boot_params); |
574 | self_size = initial_boot_params->totalsize; | |
95f72d1e | 575 | memblock_reserve(self_base, self_size); |
4d1f3f25 | 576 | |
30437b3e DG |
577 | #ifdef CONFIG_BLK_DEV_INITRD |
578 | /* then reserve the initrd, if any */ | |
579 | if (initrd_start && (initrd_end > initrd_start)) | |
307cfe71 BH |
580 | memblock_reserve(_ALIGN_DOWN(__pa(initrd_start), PAGE_SIZE), |
581 | _ALIGN_UP(initrd_end, PAGE_SIZE) - | |
582 | _ALIGN_DOWN(initrd_start, PAGE_SIZE)); | |
30437b3e DG |
583 | #endif /* CONFIG_BLK_DEV_INITRD */ |
584 | ||
cbbcf340 KG |
585 | #ifdef CONFIG_PPC32 |
586 | /* | |
587 | * Handle the case where we might be booting from an old kexec | |
588 | * image that setup the mem_rsvmap as pairs of 32-bit values | |
589 | */ | |
590 | if (*reserve_map > 0xffffffffull) { | |
591 | u32 base_32, size_32; | |
592 | u32 *reserve_map_32 = (u32 *)reserve_map; | |
593 | ||
594 | while (1) { | |
595 | base_32 = *(reserve_map_32++); | |
596 | size_32 = *(reserve_map_32++); | |
597 | if (size_32 == 0) | |
598 | break; | |
8a300887 JL |
599 | /* skip if the reservation is for the blob */ |
600 | if (base_32 == self_base && size_32 == self_size) | |
601 | continue; | |
329dda08 | 602 | DBG("reserving: %x -> %x\n", base_32, size_32); |
95f72d1e | 603 | memblock_reserve(base_32, size_32); |
cbbcf340 KG |
604 | } |
605 | return; | |
606 | } | |
607 | #endif | |
9b6b563c PM |
608 | while (1) { |
609 | base = *(reserve_map++); | |
610 | size = *(reserve_map++); | |
611 | if (size == 0) | |
612 | break; | |
cbbcf340 | 613 | DBG("reserving: %llx -> %llx\n", base, size); |
95f72d1e | 614 | memblock_reserve(base, size); |
9b6b563c | 615 | } |
9b6b563c PM |
616 | } |
617 | ||
6ac26c8a | 618 | #ifdef CONFIG_PHYP_DUMP |
37ddd5d0 MA |
619 | /** |
620 | * phyp_dump_calculate_reserve_size() - reserve variable boot area 5% or arg | |
621 | * | |
622 | * Function to find the largest size we need to reserve | |
623 | * during early boot process. | |
624 | * | |
625 | * It either looks for boot param and returns that OR | |
626 | * returns larger of 256 or 5% rounded down to multiples of 256MB. | |
627 | * | |
628 | */ | |
629 | static inline unsigned long phyp_dump_calculate_reserve_size(void) | |
630 | { | |
631 | unsigned long tmp; | |
632 | ||
633 | if (phyp_dump_info->reserve_bootvar) | |
634 | return phyp_dump_info->reserve_bootvar; | |
635 | ||
636 | /* divide by 20 to get 5% of value */ | |
95f72d1e | 637 | tmp = memblock_end_of_DRAM(); |
37ddd5d0 MA |
638 | do_div(tmp, 20); |
639 | ||
640 | /* round it down in multiples of 256 */ | |
641 | tmp = tmp & ~0x0FFFFFFFUL; | |
642 | ||
643 | return (tmp > PHYP_DUMP_RMR_END ? tmp : PHYP_DUMP_RMR_END); | |
644 | } | |
645 | ||
6ac26c8a MA |
646 | /** |
647 | * phyp_dump_reserve_mem() - reserve all not-yet-dumped mmemory | |
648 | * | |
649 | * This routine may reserve memory regions in the kernel only | |
650 | * if the system is supported and a dump was taken in last | |
651 | * boot instance or if the hardware is supported and the | |
652 | * scratch area needs to be setup. In other instances it returns | |
653 | * without reserving anything. The memory in case of dump being | |
654 | * active is freed when the dump is collected (by userland tools). | |
655 | */ | |
656 | static void __init phyp_dump_reserve_mem(void) | |
657 | { | |
658 | unsigned long base, size; | |
37ddd5d0 MA |
659 | unsigned long variable_reserve_size; |
660 | ||
6ac26c8a MA |
661 | if (!phyp_dump_info->phyp_dump_configured) { |
662 | printk(KERN_ERR "Phyp-dump not supported on this hardware\n"); | |
663 | return; | |
664 | } | |
665 | ||
654f596d MA |
666 | if (!phyp_dump_info->phyp_dump_at_boot) { |
667 | printk(KERN_INFO "Phyp-dump disabled at boot time\n"); | |
668 | return; | |
669 | } | |
670 | ||
37ddd5d0 MA |
671 | variable_reserve_size = phyp_dump_calculate_reserve_size(); |
672 | ||
6ac26c8a MA |
673 | if (phyp_dump_info->phyp_dump_is_active) { |
674 | /* Reserve *everything* above RMR.Area freed by userland tools*/ | |
37ddd5d0 | 675 | base = variable_reserve_size; |
95f72d1e | 676 | size = memblock_end_of_DRAM() - base; |
6ac26c8a MA |
677 | |
678 | /* XXX crashed_ram_end is wrong, since it may be beyond | |
679 | * the memory_limit, it will need to be adjusted. */ | |
95f72d1e | 680 | memblock_reserve(base, size); |
6ac26c8a MA |
681 | |
682 | phyp_dump_info->init_reserve_start = base; | |
683 | phyp_dump_info->init_reserve_size = size; | |
684 | } else { | |
685 | size = phyp_dump_info->cpu_state_size + | |
686 | phyp_dump_info->hpte_region_size + | |
37ddd5d0 | 687 | variable_reserve_size; |
95f72d1e YL |
688 | base = memblock_end_of_DRAM() - size; |
689 | memblock_reserve(base, size); | |
6ac26c8a MA |
690 | phyp_dump_info->init_reserve_start = base; |
691 | phyp_dump_info->init_reserve_size = size; | |
692 | } | |
693 | } | |
694 | #else | |
695 | static inline void __init phyp_dump_reserve_mem(void) {} | |
696 | #endif /* CONFIG_PHYP_DUMP && CONFIG_PPC_RTAS */ | |
697 | ||
9b6b563c PM |
698 | void __init early_init_devtree(void *params) |
699 | { | |
49a84965 | 700 | phys_addr_t limit; |
6ca4f749 | 701 | |
44348105 | 702 | DBG(" -> early_init_devtree(%p)\n", params); |
9b6b563c PM |
703 | |
704 | /* Setup flat device-tree pointer */ | |
705 | initial_boot_params = params; | |
706 | ||
458148c0 ME |
707 | #ifdef CONFIG_PPC_RTAS |
708 | /* Some machines might need RTAS info for debugging, grab it now. */ | |
709 | of_scan_flat_dt(early_init_dt_scan_rtas, NULL); | |
710 | #endif | |
711 | ||
14a43e69 BH |
712 | #ifdef CONFIG_PPC_POWERNV |
713 | /* Some machines might need OPAL info for debugging, grab it now. */ | |
714 | of_scan_flat_dt(early_init_dt_scan_opal, NULL); | |
715 | #endif | |
716 | ||
6ac26c8a | 717 | #ifdef CONFIG_PHYP_DUMP |
25985edc | 718 | /* scan tree to see if dump occurred during last boot */ |
6ac26c8a MA |
719 | of_scan_flat_dt(early_init_dt_scan_phyp_dump, NULL); |
720 | #endif | |
721 | ||
817c21ad BH |
722 | /* Pre-initialize the cmd_line with the content of boot_commmand_line, |
723 | * which will be empty except when the content of the variable has | |
724 | * been overriden by a bootloading mechanism. This happens typically | |
725 | * with HAL takeover | |
726 | */ | |
727 | strlcpy(cmd_line, boot_command_line, COMMAND_LINE_SIZE); | |
728 | ||
9b6b563c PM |
729 | /* Retrieve various informations from the /chosen node of the |
730 | * device-tree, including the platform type, initrd location and | |
731 | * size, TCE reserve, and more ... | |
732 | */ | |
85f60ae4 | 733 | of_scan_flat_dt(early_init_dt_scan_chosen_ppc, cmd_line); |
9b6b563c | 734 | |
95f72d1e | 735 | /* Scan memory nodes and rebuild MEMBLOCKs */ |
3c726f8d | 736 | of_scan_flat_dt(early_init_dt_scan_root, NULL); |
51975db0 | 737 | of_scan_flat_dt(early_init_dt_scan_memory_ppc, NULL); |
846f77b0 ME |
738 | |
739 | /* Save command line for /proc/cmdline and then parse parameters */ | |
b8757b21 | 740 | strlcpy(boot_command_line, cmd_line, COMMAND_LINE_SIZE); |
846f77b0 ME |
741 | parse_early_param(); |
742 | ||
ba14f649 KG |
743 | /* make sure we've parsed cmdline for mem= before this */ |
744 | if (memory_limit) | |
745 | first_memblock_size = min(first_memblock_size, memory_limit); | |
746 | setup_initial_memory_limit(memstart_addr, first_memblock_size); | |
95f72d1e YL |
747 | /* Reserve MEMBLOCK regions used by kernel, initrd, dt, etc... */ |
748 | memblock_reserve(PHYSICAL_START, __pa(klimit) - PHYSICAL_START); | |
549e8152 PM |
749 | /* If relocatable, reserve first 32k for interrupt vectors etc. */ |
750 | if (PHYSICAL_START > MEMORY_START) | |
95f72d1e | 751 | memblock_reserve(MEMORY_START, 0x8000); |
47310413 | 752 | reserve_kdump_trampoline(); |
35dd5432 | 753 | reserve_crashkernel(); |
9b6b563c | 754 | early_reserve_mem(); |
6ac26c8a | 755 | phyp_dump_reserve_mem(); |
9b6b563c | 756 | |
6fbef13c TH |
757 | /* |
758 | * Ensure that total memory size is page-aligned, because otherwise | |
759 | * mark_bootmem() gets upset. | |
760 | */ | |
761 | memblock_analyze(); | |
762 | limit = ALIGN(memory_limit ?: memblock_phys_mem_size(), PAGE_SIZE); | |
95f72d1e | 763 | memblock_enforce_memory_limit(limit); |
6ca4f749 | 764 | |
95f72d1e YL |
765 | memblock_analyze(); |
766 | memblock_dump_all(); | |
2babf5c2 | 767 | |
95f72d1e | 768 | DBG("Phys. mem: %llx\n", memblock_phys_mem_size()); |
2babf5c2 ME |
769 | |
770 | /* We may need to relocate the flat tree, do it now. | |
771 | * FIXME .. and the initrd too? */ | |
772 | move_device_tree(); | |
773 | ||
1426d5a3 ME |
774 | allocate_pacas(); |
775 | ||
9b6b563c PM |
776 | DBG("Scanning CPUs ...\n"); |
777 | ||
25985edc | 778 | /* Retrieve CPU related informations from the flat tree |
3c726f8d | 779 | * (altivec support, boot CPU ID, ...) |
9b6b563c | 780 | */ |
3c726f8d | 781 | of_scan_flat_dt(early_init_dt_scan_cpus, NULL); |
9b6b563c | 782 | |
7ac87abb ME |
783 | #if defined(CONFIG_SMP) && defined(CONFIG_PPC64) |
784 | /* We'll later wait for secondaries to check in; there are | |
785 | * NCPUS-1 non-boot CPUs :-) | |
786 | */ | |
787 | spinning_secondaries = boot_cpu_count - 1; | |
788 | #endif | |
789 | ||
9b6b563c PM |
790 | DBG(" <- early_init_devtree()\n"); |
791 | } | |
792 | ||
9b6b563c PM |
793 | /******* |
794 | * | |
795 | * New implementation of the OF "find" APIs, return a refcounted | |
796 | * object, call of_node_put() when done. The device tree and list | |
797 | * are protected by a rw_lock. | |
798 | * | |
799 | * Note that property management will need some locking as well, | |
800 | * this isn't dealt with yet. | |
801 | * | |
802 | *******/ | |
803 | ||
e523f723 NL |
804 | /** |
805 | * of_find_next_cache_node - Find a node's subsidiary cache | |
806 | * @np: node of type "cpu" or "cache" | |
807 | * | |
808 | * Returns a node pointer with refcount incremented, use | |
809 | * of_node_put() on it when done. Caller should hold a reference | |
810 | * to np. | |
811 | */ | |
812 | struct device_node *of_find_next_cache_node(struct device_node *np) | |
813 | { | |
814 | struct device_node *child; | |
815 | const phandle *handle; | |
816 | ||
817 | handle = of_get_property(np, "l2-cache", NULL); | |
818 | if (!handle) | |
819 | handle = of_get_property(np, "next-level-cache", NULL); | |
820 | ||
821 | if (handle) | |
822 | return of_find_node_by_phandle(*handle); | |
823 | ||
824 | /* OF on pmac has nodes instead of properties named "l2-cache" | |
825 | * beneath CPU nodes. | |
826 | */ | |
827 | if (!strcmp(np->type, "cpu")) | |
828 | for_each_child_of_node(np, child) | |
829 | if (!strcmp(child->type, "cache")) | |
830 | return child; | |
831 | ||
832 | return NULL; | |
833 | } | |
834 | ||
9b6b563c PM |
835 | #ifdef CONFIG_PPC_PSERIES |
836 | /* | |
837 | * Fix up the uninitialized fields in a new device node: | |
0ebfff14 | 838 | * name, type and pci-specific fields |
9b6b563c PM |
839 | */ |
840 | ||
cc5d0189 | 841 | static int of_finish_dynamic_node(struct device_node *node) |
9b6b563c PM |
842 | { |
843 | struct device_node *parent = of_get_parent(node); | |
844 | int err = 0; | |
a7f67bdf | 845 | const phandle *ibm_phandle; |
9b6b563c | 846 | |
0e56efc7 SR |
847 | node->name = of_get_property(node, "name", NULL); |
848 | node->type = of_get_property(node, "device_type", NULL); | |
9b6b563c | 849 | |
847f5976 BH |
850 | if (!node->name) |
851 | node->name = "<NULL>"; | |
852 | if (!node->type) | |
853 | node->type = "<NULL>"; | |
854 | ||
9b6b563c PM |
855 | if (!parent) { |
856 | err = -ENODEV; | |
857 | goto out; | |
858 | } | |
859 | ||
860 | /* We don't support that function on PowerMac, at least | |
861 | * not yet | |
862 | */ | |
e8222502 | 863 | if (machine_is(powermac)) |
9b6b563c PM |
864 | return -ENODEV; |
865 | ||
6016a363 | 866 | /* fix up new node's phandle field */ |
0e56efc7 | 867 | if ((ibm_phandle = of_get_property(node, "ibm,phandle", NULL))) |
6016a363 | 868 | node->phandle = *ibm_phandle; |
9b6b563c PM |
869 | |
870 | out: | |
871 | of_node_put(parent); | |
872 | return err; | |
873 | } | |
874 | ||
875 | static int prom_reconfig_notifier(struct notifier_block *nb, | |
876 | unsigned long action, void *node) | |
877 | { | |
878 | int err; | |
879 | ||
880 | switch (action) { | |
881 | case PSERIES_RECONFIG_ADD: | |
cc5d0189 | 882 | err = of_finish_dynamic_node(node); |
de2780a3 | 883 | if (err < 0) |
9b6b563c | 884 | printk(KERN_ERR "finish_node returned %d\n", err); |
9b6b563c PM |
885 | break; |
886 | default: | |
de2780a3 | 887 | err = 0; |
9b6b563c PM |
888 | break; |
889 | } | |
de2780a3 | 890 | return notifier_from_errno(err); |
9b6b563c PM |
891 | } |
892 | ||
893 | static struct notifier_block prom_reconfig_nb = { | |
894 | .notifier_call = prom_reconfig_notifier, | |
895 | .priority = 10, /* This one needs to run first */ | |
896 | }; | |
897 | ||
898 | static int __init prom_reconfig_setup(void) | |
899 | { | |
900 | return pSeries_reconfig_notifier_register(&prom_reconfig_nb); | |
901 | } | |
902 | __initcall(prom_reconfig_setup); | |
903 | #endif | |
904 | ||
acf7d768 BH |
905 | /* Find the device node for a given logical cpu number, also returns the cpu |
906 | * local thread number (index in ibm,interrupt-server#s) if relevant and | |
907 | * asked for (non NULL) | |
908 | */ | |
909 | struct device_node *of_get_cpu_node(int cpu, unsigned int *thread) | |
910 | { | |
911 | int hardid; | |
912 | struct device_node *np; | |
913 | ||
914 | hardid = get_hard_smp_processor_id(cpu); | |
915 | ||
916 | for_each_node_by_type(np, "cpu") { | |
a7f67bdf | 917 | const u32 *intserv; |
acf7d768 BH |
918 | unsigned int plen, t; |
919 | ||
920 | /* Check for ibm,ppc-interrupt-server#s. If it doesn't exist | |
921 | * fallback to "reg" property and assume no threads | |
922 | */ | |
0e56efc7 | 923 | intserv = of_get_property(np, "ibm,ppc-interrupt-server#s", |
a7f67bdf | 924 | &plen); |
acf7d768 | 925 | if (intserv == NULL) { |
0e56efc7 | 926 | const u32 *reg = of_get_property(np, "reg", NULL); |
acf7d768 BH |
927 | if (reg == NULL) |
928 | continue; | |
929 | if (*reg == hardid) { | |
930 | if (thread) | |
931 | *thread = 0; | |
932 | return np; | |
933 | } | |
934 | } else { | |
935 | plen /= sizeof(u32); | |
936 | for (t = 0; t < plen; t++) { | |
937 | if (hardid == intserv[t]) { | |
938 | if (thread) | |
939 | *thread = t; | |
940 | return np; | |
941 | } | |
942 | } | |
943 | } | |
944 | } | |
945 | return NULL; | |
946 | } | |
36ca4ba4 | 947 | EXPORT_SYMBOL(of_get_cpu_node); |
7a4571ae | 948 | |
94a3807c | 949 | #if defined(CONFIG_DEBUG_FS) && defined(DEBUG) |
7a4571ae ME |
950 | static struct debugfs_blob_wrapper flat_dt_blob; |
951 | ||
952 | static int __init export_flat_device_tree(void) | |
953 | { | |
954 | struct dentry *d; | |
955 | ||
7a4571ae ME |
956 | flat_dt_blob.data = initial_boot_params; |
957 | flat_dt_blob.size = initial_boot_params->totalsize; | |
958 | ||
959 | d = debugfs_create_blob("flat-device-tree", S_IFREG | S_IRUSR, | |
94a3807c | 960 | powerpc_debugfs_root, &flat_dt_blob); |
7a4571ae ME |
961 | if (!d) |
962 | return 1; | |
963 | ||
964 | return 0; | |
965 | } | |
966 | __initcall(export_flat_device_tree); | |
967 | #endif |