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> | |
9b6b563c PM |
26 | #include <linux/delay.h> |
27 | #include <linux/initrd.h> | |
28 | #include <linux/bitops.h> | |
4b16f8e2 | 29 | #include <linux/export.h> |
dcee3036 | 30 | #include <linux/kexec.h> |
0ebfff14 | 31 | #include <linux/irq.h> |
95f72d1e | 32 | #include <linux/memblock.h> |
1cf3d8b3 | 33 | #include <linux/of.h> |
708b7eef | 34 | #include <linux/of_fdt.h> |
060f78c2 | 35 | #include <linux/libfdt.h> |
7f92bc56 | 36 | #include <linux/cpu.h> |
9b6b563c PM |
37 | |
38 | #include <asm/prom.h> | |
39 | #include <asm/rtas.h> | |
9b6b563c PM |
40 | #include <asm/page.h> |
41 | #include <asm/processor.h> | |
42 | #include <asm/irq.h> | |
43 | #include <asm/io.h> | |
0cc4746c | 44 | #include <asm/kdump.h> |
9b6b563c | 45 | #include <asm/smp.h> |
9b6b563c | 46 | #include <asm/mmu.h> |
1426d5a3 | 47 | #include <asm/paca.h> |
9b6b563c | 48 | #include <asm/pgtable.h> |
54820530 | 49 | #include <asm/powernv.h> |
9b6b563c PM |
50 | #include <asm/iommu.h> |
51 | #include <asm/btext.h> | |
52 | #include <asm/sections.h> | |
53 | #include <asm/machdep.h> | |
40ef8cbc | 54 | #include <asm/pci-bridge.h> |
2babf5c2 | 55 | #include <asm/kexec.h> |
14a43e69 | 56 | #include <asm/opal.h> |
eb39c880 | 57 | #include <asm/fadump.h> |
da6a97bf | 58 | #include <asm/epapr_hcalls.h> |
3808a889 | 59 | #include <asm/firmware.h> |
5a61ef74 | 60 | #include <asm/dt_cpu_ftrs.h> |
6c6ea537 | 61 | #include <asm/drmem.h> |
14a43e69 | 62 | |
37dd2bad | 63 | #include <mm/mmu_decl.h> |
9b6b563c PM |
64 | |
65 | #ifdef DEBUG | |
66 | #define DBG(fmt...) printk(KERN_ERR fmt) | |
67 | #else | |
68 | #define DBG(fmt...) | |
69 | #endif | |
70 | ||
9b6b563c | 71 | #ifdef CONFIG_PPC64 |
28897731 | 72 | int __initdata iommu_is_off; |
9b6b563c | 73 | int __initdata iommu_force_on; |
cf00a8d1 | 74 | unsigned long tce_alloc_start, tce_alloc_end; |
cd3db0c4 | 75 | u64 ppc64_rma_size; |
9b6b563c | 76 | #endif |
03bf469a | 77 | static phys_addr_t first_memblock_size; |
7ac87abb | 78 | static int __initdata boot_cpu_count; |
9b6b563c | 79 | |
2babf5c2 ME |
80 | static int __init early_parse_mem(char *p) |
81 | { | |
82 | if (!p) | |
83 | return 1; | |
84 | ||
85 | memory_limit = PAGE_ALIGN(memparse(p, &p)); | |
a84fcd46 | 86 | DBG("memory limit = 0x%llx\n", memory_limit); |
2babf5c2 ME |
87 | |
88 | return 0; | |
89 | } | |
90 | early_param("mem", early_parse_mem); | |
91 | ||
307cfe71 BH |
92 | /* |
93 | * overlaps_initrd - check for overlap with page aligned extension of | |
94 | * initrd. | |
95 | */ | |
96 | static inline int overlaps_initrd(unsigned long start, unsigned long size) | |
97 | { | |
98 | #ifdef CONFIG_BLK_DEV_INITRD | |
99 | if (!initrd_start) | |
100 | return 0; | |
101 | ||
102 | return (start + size) > _ALIGN_DOWN(initrd_start, PAGE_SIZE) && | |
103 | start <= _ALIGN_UP(initrd_end, PAGE_SIZE); | |
104 | #else | |
105 | return 0; | |
106 | #endif | |
107 | } | |
108 | ||
3c607ce2 LV |
109 | /** |
110 | * move_device_tree - move tree to an unused area, if needed. | |
111 | * | |
112 | * The device tree may be allocated beyond our memory limit, or inside the | |
307cfe71 BH |
113 | * crash kernel region for kdump, or within the page aligned range of initrd. |
114 | * If so, move it out of the way. | |
2babf5c2 | 115 | */ |
18f032cb | 116 | static void __init move_device_tree(void) |
2babf5c2 ME |
117 | { |
118 | unsigned long start, size; | |
119 | void *p; | |
120 | ||
121 | DBG("-> move_device_tree\n"); | |
122 | ||
123 | start = __pa(initial_boot_params); | |
060f78c2 | 124 | size = fdt_totalsize(initial_boot_params); |
2babf5c2 | 125 | |
6dd22700 | 126 | if ((memory_limit && (start + size) > PHYSICAL_START + memory_limit) || |
307cfe71 BH |
127 | overlaps_crashkernel(start, size) || |
128 | overlaps_initrd(start, size)) { | |
9a8dd708 | 129 | p = __va(memblock_phys_alloc(size, PAGE_SIZE)); |
2babf5c2 | 130 | memcpy(p, initial_boot_params, size); |
060f78c2 | 131 | initial_boot_params = p; |
2babf5c2 ME |
132 | DBG("Moved device tree to 0x%p\n", p); |
133 | } | |
134 | ||
135 | DBG("<- move_device_tree\n"); | |
136 | } | |
9b6b563c | 137 | |
d205819e PM |
138 | /* |
139 | * ibm,pa-features is a per-cpu property that contains a string of | |
140 | * attribute descriptors, each of which has a 2 byte header plus up | |
141 | * to 254 bytes worth of processor attribute bits. First header | |
142 | * byte specifies the number of bytes following the header. | |
143 | * Second header byte is an "attribute-specifier" type, of which | |
144 | * zero is the only currently-defined value. | |
145 | * Implementation: Pass in the byte and bit offset for the feature | |
146 | * that we are interested in. The function will return -1 if the | |
147 | * pa-features property is missing, or a 1/0 to indicate if the feature | |
148 | * is supported/not supported. Note that the bit numbers are | |
149 | * big-endian to match the definition in PAPR. | |
150 | */ | |
151 | static struct ibm_pa_feature { | |
152 | unsigned long cpu_features; /* CPU_FTR_xxx bit */ | |
44ae3ab3 | 153 | unsigned long mmu_features; /* MMU_FTR_xxx bit */ |
d205819e | 154 | unsigned int cpu_user_ftrs; /* PPC_FEATURE_xxx bit */ |
beff8237 | 155 | unsigned int cpu_user_ftrs2; /* PPC_FEATURE2_xxx bit */ |
d205819e PM |
156 | unsigned char pabyte; /* byte number in ibm,pa-features */ |
157 | unsigned char pabit; /* bit number (big-endian) */ | |
158 | unsigned char invert; /* if 1, pa bit set => clear feature */ | |
159 | } ibm_pa_features[] __initdata = { | |
e9eb0278 ME |
160 | { .pabyte = 0, .pabit = 0, .cpu_user_ftrs = PPC_FEATURE_HAS_MMU }, |
161 | { .pabyte = 0, .pabit = 1, .cpu_user_ftrs = PPC_FEATURE_HAS_FPU }, | |
162 | { .pabyte = 0, .pabit = 3, .cpu_features = CPU_FTR_CTRL }, | |
163 | { .pabyte = 0, .pabit = 6, .cpu_features = CPU_FTR_NOEXECUTE }, | |
164 | { .pabyte = 1, .pabit = 2, .mmu_features = MMU_FTR_CI_LARGE_PAGE }, | |
d957fb4d | 165 | #ifdef CONFIG_PPC_RADIX_MMU |
e9eb0278 | 166 | { .pabyte = 40, .pabit = 0, .mmu_features = MMU_FTR_TYPE_RADIX }, |
d957fb4d | 167 | #endif |
e9eb0278 ME |
168 | { .pabyte = 1, .pabit = 1, .invert = 1, .cpu_features = CPU_FTR_NODSISRALIGN }, |
169 | { .pabyte = 5, .pabit = 0, .cpu_features = CPU_FTR_REAL_LE, | |
170 | .cpu_user_ftrs = PPC_FEATURE_TRUE_LE }, | |
9e819963 | 171 | /* |
4705e024 AB |
172 | * If the kernel doesn't support TM (ie CONFIG_PPC_TRANSACTIONAL_MEM=n), |
173 | * we don't want to turn on TM here, so we use the *_COMP versions | |
174 | * which are 0 if the kernel doesn't support TM. | |
9e819963 | 175 | */ |
e9eb0278 ME |
176 | { .pabyte = 22, .pabit = 0, .cpu_features = CPU_FTR_TM_COMP, |
177 | .cpu_user_ftrs2 = PPC_FEATURE2_HTM_COMP | PPC_FEATURE2_HTM_NOSC_COMP }, | |
d205819e PM |
178 | }; |
179 | ||
9d0c4dfe | 180 | static void __init scan_features(unsigned long node, const unsigned char *ftrs, |
974a76f5 PM |
181 | unsigned long tablelen, |
182 | struct ibm_pa_feature *fp, | |
183 | unsigned long ft_size) | |
d205819e | 184 | { |
974a76f5 | 185 | unsigned long i, len, bit; |
d205819e PM |
186 | |
187 | /* find descriptor with type == 0 */ | |
188 | for (;;) { | |
189 | if (tablelen < 3) | |
190 | return; | |
974a76f5 | 191 | len = 2 + ftrs[0]; |
d205819e PM |
192 | if (tablelen < len) |
193 | return; /* descriptor 0 not found */ | |
974a76f5 | 194 | if (ftrs[1] == 0) |
d205819e PM |
195 | break; |
196 | tablelen -= len; | |
974a76f5 | 197 | ftrs += len; |
d205819e PM |
198 | } |
199 | ||
200 | /* loop over bits we know about */ | |
974a76f5 PM |
201 | for (i = 0; i < ft_size; ++i, ++fp) { |
202 | if (fp->pabyte >= ftrs[0]) | |
d205819e | 203 | continue; |
974a76f5 | 204 | bit = (ftrs[2 + fp->pabyte] >> (7 - fp->pabit)) & 1; |
d205819e PM |
205 | if (bit ^ fp->invert) { |
206 | cur_cpu_spec->cpu_features |= fp->cpu_features; | |
207 | cur_cpu_spec->cpu_user_features |= fp->cpu_user_ftrs; | |
beff8237 | 208 | cur_cpu_spec->cpu_user_features2 |= fp->cpu_user_ftrs2; |
44ae3ab3 | 209 | cur_cpu_spec->mmu_features |= fp->mmu_features; |
d205819e PM |
210 | } else { |
211 | cur_cpu_spec->cpu_features &= ~fp->cpu_features; | |
212 | cur_cpu_spec->cpu_user_features &= ~fp->cpu_user_ftrs; | |
beff8237 | 213 | cur_cpu_spec->cpu_user_features2 &= ~fp->cpu_user_ftrs2; |
44ae3ab3 | 214 | cur_cpu_spec->mmu_features &= ~fp->mmu_features; |
d205819e PM |
215 | } |
216 | } | |
217 | } | |
218 | ||
974a76f5 PM |
219 | static void __init check_cpu_pa_features(unsigned long node) |
220 | { | |
9d0c4dfe RH |
221 | const unsigned char *pa_ftrs; |
222 | int tablelen; | |
974a76f5 PM |
223 | |
224 | pa_ftrs = of_get_flat_dt_prop(node, "ibm,pa-features", &tablelen); | |
225 | if (pa_ftrs == NULL) | |
226 | return; | |
227 | ||
228 | scan_features(node, pa_ftrs, tablelen, | |
229 | ibm_pa_features, ARRAY_SIZE(ibm_pa_features)); | |
230 | } | |
231 | ||
4e003747 | 232 | #ifdef CONFIG_PPC_BOOK3S_64 |
9c61f7a0 | 233 | static void __init init_mmu_slb_size(unsigned long node) |
584f8b71 | 234 | { |
9d0c4dfe | 235 | const __be32 *slb_size_ptr; |
584f8b71 | 236 | |
9c61f7a0 AK |
237 | slb_size_ptr = of_get_flat_dt_prop(node, "slb-size", NULL) ? : |
238 | of_get_flat_dt_prop(node, "ibm,slb-size", NULL); | |
239 | ||
240 | if (slb_size_ptr) | |
dc0e643a | 241 | mmu_slb_size = be32_to_cpup(slb_size_ptr); |
584f8b71 MN |
242 | } |
243 | #else | |
9c61f7a0 | 244 | #define init_mmu_slb_size(node) do { } while(0) |
584f8b71 MN |
245 | #endif |
246 | ||
974a76f5 PM |
247 | static struct feature_property { |
248 | const char *name; | |
249 | u32 min_value; | |
250 | unsigned long cpu_feature; | |
251 | unsigned long cpu_user_ftr; | |
252 | } feature_properties[] __initdata = { | |
253 | #ifdef CONFIG_ALTIVEC | |
254 | {"altivec", 0, CPU_FTR_ALTIVEC, PPC_FEATURE_HAS_ALTIVEC}, | |
255 | {"ibm,vmx", 1, CPU_FTR_ALTIVEC, PPC_FEATURE_HAS_ALTIVEC}, | |
256 | #endif /* CONFIG_ALTIVEC */ | |
b962ce9d MN |
257 | #ifdef CONFIG_VSX |
258 | /* Yes, this _really_ is ibm,vmx == 2 to enable VSX */ | |
259 | {"ibm,vmx", 2, CPU_FTR_VSX, PPC_FEATURE_HAS_VSX}, | |
260 | #endif /* CONFIG_VSX */ | |
974a76f5 PM |
261 | #ifdef CONFIG_PPC64 |
262 | {"ibm,dfp", 1, 0, PPC_FEATURE_HAS_DFP}, | |
263 | {"ibm,purr", 1, CPU_FTR_PURR, 0}, | |
264 | {"ibm,spurr", 1, CPU_FTR_SPURR, 0}, | |
265 | #endif /* CONFIG_PPC64 */ | |
266 | }; | |
267 | ||
14b3d926 VB |
268 | #if defined(CONFIG_44x) && defined(CONFIG_PPC_FPU) |
269 | static inline void identical_pvr_fixup(unsigned long node) | |
270 | { | |
271 | unsigned int pvr; | |
9d0c4dfe | 272 | const char *model = of_get_flat_dt_prop(node, "model", NULL); |
14b3d926 VB |
273 | |
274 | /* | |
275 | * Since 440GR(x)/440EP(x) processors have the same pvr, | |
276 | * we check the node path and set bit 28 in the cur_cpu_spec | |
277 | * pvr for EP(x) processor version. This bit is always 0 in | |
278 | * the "real" pvr. Then we call identify_cpu again with | |
279 | * the new logical pvr to enable FPU support. | |
280 | */ | |
281 | if (model && strstr(model, "440EP")) { | |
282 | pvr = cur_cpu_spec->pvr_value | 0x8; | |
283 | identify_cpu(0, pvr); | |
284 | DBG("Using logical pvr %x for %s\n", pvr, model); | |
285 | } | |
286 | } | |
287 | #else | |
288 | #define identical_pvr_fixup(node) do { } while(0) | |
289 | #endif | |
290 | ||
974a76f5 PM |
291 | static void __init check_cpu_feature_properties(unsigned long node) |
292 | { | |
4f1f40f7 | 293 | int i; |
974a76f5 | 294 | struct feature_property *fp = feature_properties; |
dc0e643a | 295 | const __be32 *prop; |
974a76f5 | 296 | |
4f1f40f7 | 297 | for (i = 0; i < (int)ARRAY_SIZE(feature_properties); ++i, ++fp) { |
974a76f5 | 298 | prop = of_get_flat_dt_prop(node, fp->name, NULL); |
dc0e643a | 299 | if (prop && be32_to_cpup(prop) >= fp->min_value) { |
974a76f5 PM |
300 | cur_cpu_spec->cpu_features |= fp->cpu_feature; |
301 | cur_cpu_spec->cpu_user_features |= fp->cpu_user_ftr; | |
302 | } | |
303 | } | |
304 | } | |
305 | ||
9b6b563c | 306 | static int __init early_init_dt_scan_cpus(unsigned long node, |
4df20460 AB |
307 | const char *uname, int depth, |
308 | void *data) | |
9b6b563c | 309 | { |
9d0c4dfe | 310 | const char *type = of_get_flat_dt_prop(node, "device_type", NULL); |
dc0e643a IM |
311 | const __be32 *prop; |
312 | const __be32 *intserv; | |
4df20460 | 313 | int i, nthreads; |
9d0c4dfe | 314 | int len; |
9d07bc84 | 315 | int found = -1; |
208b3a4c | 316 | int found_thread = 0; |
9b6b563c PM |
317 | |
318 | /* We are scanning "cpu" nodes only */ | |
319 | if (type == NULL || strcmp(type, "cpu") != 0) | |
320 | return 0; | |
321 | ||
4df20460 AB |
322 | /* Get physical cpuid */ |
323 | intserv = of_get_flat_dt_prop(node, "ibm,ppc-interrupt-server#s", &len); | |
e16c8765 AF |
324 | if (!intserv) |
325 | intserv = of_get_flat_dt_prop(node, "reg", &len); | |
326 | ||
327 | nthreads = len / sizeof(int); | |
4df20460 AB |
328 | |
329 | /* | |
330 | * Now see if any of these threads match our boot cpu. | |
331 | * NOTE: This must match the parsing done in smp_setup_cpu_maps. | |
332 | */ | |
333 | for (i = 0; i < nthreads; i++) { | |
89c19062 ME |
334 | if (be32_to_cpu(intserv[i]) == |
335 | fdt_boot_cpuid_phys(initial_boot_params)) { | |
336 | found = boot_cpu_count; | |
337 | found_thread = i; | |
9b6b563c | 338 | } |
4df20460 AB |
339 | #ifdef CONFIG_SMP |
340 | /* logical cpu id is always 0 on UP kernels */ | |
9d07bc84 | 341 | boot_cpu_count++; |
4df20460 AB |
342 | #endif |
343 | } | |
344 | ||
7222f779 BH |
345 | /* Not the boot CPU */ |
346 | if (found < 0) | |
347 | return 0; | |
9b6b563c | 348 | |
7222f779 BH |
349 | DBG("boot cpu: logical %d physical %d\n", found, |
350 | be32_to_cpu(intserv[found_thread])); | |
351 | boot_cpuid = found; | |
14b3d926 | 352 | |
7222f779 BH |
353 | /* |
354 | * PAPR defines "logical" PVR values for cpus that | |
355 | * meet various levels of the architecture: | |
356 | * 0x0f000001 Architecture version 2.04 | |
357 | * 0x0f000002 Architecture version 2.05 | |
358 | * If the cpu-version property in the cpu node contains | |
359 | * such a value, we call identify_cpu again with the | |
360 | * logical PVR value in order to use the cpu feature | |
361 | * bits appropriate for the architecture level. | |
362 | * | |
363 | * A POWER6 partition in "POWER6 architected" mode | |
364 | * uses the 0x0f000002 PVR value; in POWER5+ mode | |
365 | * it uses 0x0f000001. | |
5a61ef74 NP |
366 | * |
367 | * If we're using device tree CPU feature discovery then we don't | |
368 | * support the cpu-version property, and it's the responsibility of the | |
369 | * firmware/hypervisor to provide the correct feature set for the | |
370 | * architecture level via the ibm,powerpc-cpu-features binding. | |
7222f779 | 371 | */ |
5a61ef74 NP |
372 | if (!dt_cpu_ftrs_in_use()) { |
373 | prop = of_get_flat_dt_prop(node, "cpu-version", NULL); | |
374 | if (prop && (be32_to_cpup(prop) & 0xff000000) == 0x0f000000) | |
375 | identify_cpu(0, be32_to_cpup(prop)); | |
7222f779 | 376 | |
5a61ef74 NP |
377 | check_cpu_feature_properties(node); |
378 | check_cpu_pa_features(node); | |
379 | } | |
9b6b563c | 380 | |
5a61ef74 | 381 | identical_pvr_fixup(node); |
9c61f7a0 | 382 | init_mmu_slb_size(node); |
d205819e | 383 | |
4a85b31d | 384 | #ifdef CONFIG_PPC64 |
5a61ef74 | 385 | if (nthreads == 1) |
4df20460 | 386 | cur_cpu_spec->cpu_features &= ~CPU_FTR_SMT; |
5a61ef74 NP |
387 | else if (!dt_cpu_ftrs_in_use()) |
388 | cur_cpu_spec->cpu_features |= CPU_FTR_SMT; | |
59f57774 | 389 | allocate_paca(boot_cpuid); |
9b6b563c | 390 | #endif |
59f57774 | 391 | set_hard_smp_processor_id(found, be32_to_cpu(intserv[found_thread])); |
5a61ef74 | 392 | |
9b6b563c PM |
393 | return 0; |
394 | } | |
395 | ||
e51df2c1 AB |
396 | static int __init early_init_dt_scan_chosen_ppc(unsigned long node, |
397 | const char *uname, | |
398 | int depth, void *data) | |
9b6b563c | 399 | { |
9d0c4dfe | 400 | const unsigned long *lprop; /* All these set by kernel, so no need to convert endian */ |
9b6b563c | 401 | |
32c97689 GL |
402 | /* Use common scan routine to determine if this is the chosen node */ |
403 | if (early_init_dt_scan_chosen(node, uname, depth, data) == 0) | |
404 | return 0; | |
405 | ||
9b6b563c PM |
406 | #ifdef CONFIG_PPC64 |
407 | /* check if iommu is forced on or off */ | |
3c726f8d | 408 | if (of_get_flat_dt_prop(node, "linux,iommu-off", NULL) != NULL) |
9b6b563c | 409 | iommu_is_off = 1; |
3c726f8d | 410 | if (of_get_flat_dt_prop(node, "linux,iommu-force-on", NULL) != NULL) |
9b6b563c PM |
411 | iommu_force_on = 1; |
412 | #endif | |
413 | ||
2babf5c2 | 414 | /* mem=x on the command line is the preferred mechanism */ |
86e03221 GL |
415 | lprop = of_get_flat_dt_prop(node, "linux,memory-limit", NULL); |
416 | if (lprop) | |
417 | memory_limit = *lprop; | |
9b6b563c PM |
418 | |
419 | #ifdef CONFIG_PPC64 | |
86e03221 GL |
420 | lprop = of_get_flat_dt_prop(node, "linux,tce-alloc-start", NULL); |
421 | if (lprop) | |
422 | tce_alloc_start = *lprop; | |
423 | lprop = of_get_flat_dt_prop(node, "linux,tce-alloc-end", NULL); | |
424 | if (lprop) | |
425 | tce_alloc_end = *lprop; | |
9b6b563c PM |
426 | #endif |
427 | ||
da665885 | 428 | #ifdef CONFIG_KEXEC_CORE |
63277161 | 429 | lprop = of_get_flat_dt_prop(node, "linux,crashkernel-base", NULL); |
70c6cc37 LV |
430 | if (lprop) |
431 | crashk_res.start = *lprop; | |
dcee3036 | 432 | |
63277161 | 433 | lprop = of_get_flat_dt_prop(node, "linux,crashkernel-size", NULL); |
70c6cc37 LV |
434 | if (lprop) |
435 | crashk_res.end = crashk_res.start + *lprop - 1; | |
dcee3036 | 436 | #endif |
32c97689 GL |
437 | |
438 | /* break now */ | |
439 | return 1; | |
9b6b563c PM |
440 | } |
441 | ||
6aba0c84 AK |
442 | /* |
443 | * Compare the range against max mem limit and update | |
444 | * size if it cross the limit. | |
445 | */ | |
446 | ||
447 | #ifdef CONFIG_SPARSEMEM | |
448 | static bool validate_mem_limit(u64 base, u64 *size) | |
449 | { | |
450 | u64 max_mem = 1UL << (MAX_PHYSMEM_BITS); | |
451 | ||
452 | if (base >= max_mem) | |
453 | return false; | |
454 | if ((base + *size) > max_mem) | |
455 | *size = max_mem - base; | |
456 | return true; | |
457 | } | |
458 | #else | |
459 | static bool validate_mem_limit(u64 base, u64 *size) | |
460 | { | |
461 | return true; | |
462 | } | |
463 | #endif | |
464 | ||
0204568a PM |
465 | #ifdef CONFIG_PPC_PSERIES |
466 | /* | |
6c6ea537 | 467 | * Interpret the ibm dynamic reconfiguration memory LMBs. |
0204568a PM |
468 | * This contains a list of memory blocks along with NUMA affinity |
469 | * information. | |
470 | */ | |
6c6ea537 NF |
471 | static void __init early_init_drmem_lmb(struct drmem_lmb *lmb, |
472 | const __be32 **usm) | |
0204568a | 473 | { |
6c6ea537 NF |
474 | u64 base, size; |
475 | int is_kexec_kdump = 0, rngs; | |
0204568a | 476 | |
6c6ea537 NF |
477 | base = lmb->base_addr; |
478 | size = drmem_lmb_size(); | |
479 | rngs = 1; | |
0204568a | 480 | |
6c6ea537 NF |
481 | /* |
482 | * Skip this block if the reserved bit is set in flags | |
483 | * or if the block is not assigned to this partition. | |
484 | */ | |
485 | if ((lmb->flags & DRCONF_MEM_RESERVED) || | |
486 | !(lmb->flags & DRCONF_MEM_ASSIGNED)) | |
487 | return; | |
0204568a | 488 | |
6c6ea537 | 489 | if (*usm) |
cf00085d C |
490 | is_kexec_kdump = 1; |
491 | ||
6c6ea537 NF |
492 | if (is_kexec_kdump) { |
493 | /* | |
494 | * For each memblock in ibm,dynamic-memory, a | |
495 | * corresponding entry in linux,drconf-usable-memory | |
496 | * property contains a counter 'p' followed by 'p' | |
497 | * (base, size) duple. Now read the counter from | |
498 | * linux,drconf-usable-memory property | |
499 | */ | |
500 | rngs = dt_mem_next_cell(dt_root_size_cells, usm); | |
501 | if (!rngs) /* there are no (base, size) duple */ | |
502 | return; | |
503 | } | |
504 | ||
505 | do { | |
cf00085d | 506 | if (is_kexec_kdump) { |
6c6ea537 NF |
507 | base = dt_mem_next_cell(dt_root_addr_cells, usm); |
508 | size = dt_mem_next_cell(dt_root_size_cells, usm); | |
509 | } | |
510 | ||
511 | if (iommu_is_off) { | |
512 | if (base >= 0x80000000ul) | |
0204568a | 513 | continue; |
6c6ea537 NF |
514 | if ((base + size) > 0x80000000ul) |
515 | size = 0x80000000ul - base; | |
0204568a | 516 | } |
6c6ea537 NF |
517 | |
518 | DBG("Adding: %llx -> %llx\n", base, size); | |
6aba0c84 AK |
519 | if (validate_mem_limit(base, &size)) |
520 | memblock_add(base, size); | |
6c6ea537 | 521 | } while (--rngs); |
0204568a | 522 | } |
0204568a | 523 | #endif /* CONFIG_PPC_PSERIES */ |
9b6b563c | 524 | |
51975db0 GL |
525 | static int __init early_init_dt_scan_memory_ppc(unsigned long node, |
526 | const char *uname, | |
527 | int depth, void *data) | |
9b6b563c | 528 | { |
6c6ea537 | 529 | #ifdef CONFIG_PPC_PSERIES |
0204568a | 530 | if (depth == 1 && |
6c6ea537 NF |
531 | strcmp(uname, "ibm,dynamic-reconfiguration-memory") == 0) { |
532 | walk_drmem_lmbs_early(node, early_init_drmem_lmb); | |
533 | return 0; | |
534 | } | |
535 | #endif | |
51975db0 GL |
536 | |
537 | return early_init_dt_scan_memory(node, uname, depth, data); | |
538 | } | |
0204568a | 539 | |
b27652dd KH |
540 | /* |
541 | * For a relocatable kernel, we need to get the memstart_addr first, | |
542 | * then use it to calculate the virtual kernel start address. This has | |
543 | * to happen at a very early stage (before machine_init). In this case, | |
544 | * we just want to get the memstart_address and would not like to mess the | |
545 | * memblock at this stage. So introduce a variable to skip the memblock_add() | |
546 | * for this reason. | |
547 | */ | |
548 | #ifdef CONFIG_RELOCATABLE | |
549 | static int add_mem_to_memblock = 1; | |
550 | #else | |
551 | #define add_mem_to_memblock 1 | |
552 | #endif | |
553 | ||
51975db0 GL |
554 | void __init early_init_dt_add_memory_arch(u64 base, u64 size) |
555 | { | |
cd3db0c4 | 556 | #ifdef CONFIG_PPC64 |
51975db0 GL |
557 | if (iommu_is_off) { |
558 | if (base >= 0x80000000ul) | |
559 | return; | |
560 | if ((base + size) > 0x80000000ul) | |
561 | size = 0x80000000ul - base; | |
562 | } | |
9b6b563c | 563 | #endif |
03bf469a BH |
564 | /* Keep track of the beginning of memory -and- the size of |
565 | * the very first block in the device-tree as it represents | |
566 | * the RMA on ppc64 server | |
567 | */ | |
568 | if (base < memstart_addr) { | |
569 | memstart_addr = base; | |
570 | first_memblock_size = size; | |
571 | } | |
cd3db0c4 BH |
572 | |
573 | /* Add the chunk to the MEMBLOCK list */ | |
6aba0c84 AK |
574 | if (add_mem_to_memblock) { |
575 | if (validate_mem_limit(base, &size)) | |
576 | memblock_add(base, size); | |
577 | } | |
9b6b563c PM |
578 | } |
579 | ||
c039e3a8 | 580 | static void __init early_reserve_mem_dt(void) |
0962e800 | 581 | { |
9d0c4dfe RH |
582 | unsigned long i, dt_root; |
583 | int len; | |
0962e800 JK |
584 | const __be32 *prop; |
585 | ||
24bbd929 | 586 | early_init_fdt_reserve_self(); |
edba274c RH |
587 | early_init_fdt_scan_reserved_mem(); |
588 | ||
0962e800 JK |
589 | dt_root = of_get_flat_dt_root(); |
590 | ||
591 | prop = of_get_flat_dt_prop(dt_root, "reserved-ranges", &len); | |
592 | ||
593 | if (!prop) | |
c039e3a8 BH |
594 | return; |
595 | ||
596 | DBG("Found new-style reserved-ranges\n"); | |
0962e800 JK |
597 | |
598 | /* Each reserved range is an (address,size) pair, 2 cells each, | |
599 | * totalling 4 cells per range. */ | |
600 | for (i = 0; i < len / (sizeof(*prop) * 4); i++) { | |
601 | u64 base, size; | |
602 | ||
603 | base = of_read_number(prop + (i * 4) + 0, 2); | |
604 | size = of_read_number(prop + (i * 4) + 2, 2); | |
605 | ||
c039e3a8 BH |
606 | if (size) { |
607 | DBG("reserving: %llx -> %llx\n", base, size); | |
0962e800 | 608 | memblock_reserve(base, size); |
c039e3a8 | 609 | } |
0962e800 | 610 | } |
0962e800 JK |
611 | } |
612 | ||
9b6b563c PM |
613 | static void __init early_reserve_mem(void) |
614 | { | |
dc0e643a | 615 | __be64 *reserve_map; |
9b6b563c | 616 | |
dc0e643a | 617 | reserve_map = (__be64 *)(((unsigned long)initial_boot_params) + |
060f78c2 | 618 | fdt_off_mem_rsvmap(initial_boot_params)); |
4d1f3f25 | 619 | |
c039e3a8 BH |
620 | /* Look for the new "reserved-regions" property in the DT */ |
621 | early_reserve_mem_dt(); | |
0962e800 | 622 | |
30437b3e | 623 | #ifdef CONFIG_BLK_DEV_INITRD |
c039e3a8 BH |
624 | /* Then reserve the initrd, if any */ |
625 | if (initrd_start && (initrd_end > initrd_start)) { | |
307cfe71 BH |
626 | memblock_reserve(_ALIGN_DOWN(__pa(initrd_start), PAGE_SIZE), |
627 | _ALIGN_UP(initrd_end, PAGE_SIZE) - | |
628 | _ALIGN_DOWN(initrd_start, PAGE_SIZE)); | |
c039e3a8 | 629 | } |
30437b3e DG |
630 | #endif /* CONFIG_BLK_DEV_INITRD */ |
631 | ||
cbbcf340 KG |
632 | #ifdef CONFIG_PPC32 |
633 | /* | |
634 | * Handle the case where we might be booting from an old kexec | |
635 | * image that setup the mem_rsvmap as pairs of 32-bit values | |
636 | */ | |
dc0e643a | 637 | if (be64_to_cpup(reserve_map) > 0xffffffffull) { |
cbbcf340 | 638 | u32 base_32, size_32; |
dc0e643a | 639 | __be32 *reserve_map_32 = (__be32 *)reserve_map; |
cbbcf340 | 640 | |
c039e3a8 BH |
641 | DBG("Found old 32-bit reserve map\n"); |
642 | ||
cbbcf340 | 643 | while (1) { |
dc0e643a IM |
644 | base_32 = be32_to_cpup(reserve_map_32++); |
645 | size_32 = be32_to_cpup(reserve_map_32++); | |
cbbcf340 KG |
646 | if (size_32 == 0) |
647 | break; | |
329dda08 | 648 | DBG("reserving: %x -> %x\n", base_32, size_32); |
95f72d1e | 649 | memblock_reserve(base_32, size_32); |
cbbcf340 KG |
650 | } |
651 | return; | |
652 | } | |
653 | #endif | |
9b6b563c PM |
654 | } |
655 | ||
07fd1761 CB |
656 | #ifdef CONFIG_PPC_TRANSACTIONAL_MEM |
657 | static bool tm_disabled __initdata; | |
658 | ||
659 | static int __init parse_ppc_tm(char *str) | |
660 | { | |
661 | bool res; | |
662 | ||
663 | if (kstrtobool(str, &res)) | |
664 | return -EINVAL; | |
665 | ||
666 | tm_disabled = !res; | |
667 | ||
668 | return 0; | |
669 | } | |
670 | early_param("ppc_tm", parse_ppc_tm); | |
671 | ||
672 | static void __init tm_init(void) | |
673 | { | |
674 | if (tm_disabled) { | |
675 | pr_info("Disabling hardware transactional memory (HTM)\n"); | |
676 | cur_cpu_spec->cpu_user_features2 &= | |
677 | ~(PPC_FEATURE2_HTM_NOSC | PPC_FEATURE2_HTM); | |
678 | cur_cpu_spec->cpu_features &= ~CPU_FTR_TM; | |
54820530 | 679 | return; |
07fd1761 | 680 | } |
54820530 ME |
681 | |
682 | pnv_tm_init(); | |
07fd1761 CB |
683 | } |
684 | #else | |
685 | static void tm_init(void) { } | |
686 | #endif /* CONFIG_PPC_TRANSACTIONAL_MEM */ | |
687 | ||
9b6b563c PM |
688 | void __init early_init_devtree(void *params) |
689 | { | |
49a84965 | 690 | phys_addr_t limit; |
6ca4f749 | 691 | |
44348105 | 692 | DBG(" -> early_init_devtree(%p)\n", params); |
9b6b563c | 693 | |
ad72a279 ME |
694 | /* Too early to BUG_ON(), do it by hand */ |
695 | if (!early_init_dt_verify(params)) | |
696 | panic("BUG: Failed verifying flat device tree, bad version?"); | |
697 | ||
458148c0 ME |
698 | #ifdef CONFIG_PPC_RTAS |
699 | /* Some machines might need RTAS info for debugging, grab it now. */ | |
700 | of_scan_flat_dt(early_init_dt_scan_rtas, NULL); | |
701 | #endif | |
702 | ||
14a43e69 BH |
703 | #ifdef CONFIG_PPC_POWERNV |
704 | /* Some machines might need OPAL info for debugging, grab it now. */ | |
705 | of_scan_flat_dt(early_init_dt_scan_opal, NULL); | |
706 | #endif | |
707 | ||
eb39c880 MS |
708 | #ifdef CONFIG_FA_DUMP |
709 | /* scan tree to see if dump is active during last boot */ | |
710 | of_scan_flat_dt(early_init_dt_scan_fw_dump, NULL); | |
711 | #endif | |
712 | ||
9b6b563c PM |
713 | /* Retrieve various informations from the /chosen node of the |
714 | * device-tree, including the platform type, initrd location and | |
715 | * size, TCE reserve, and more ... | |
716 | */ | |
3e47d147 | 717 | of_scan_flat_dt(early_init_dt_scan_chosen_ppc, boot_command_line); |
9b6b563c | 718 | |
95f72d1e | 719 | /* Scan memory nodes and rebuild MEMBLOCKs */ |
3c726f8d | 720 | of_scan_flat_dt(early_init_dt_scan_root, NULL); |
51975db0 | 721 | of_scan_flat_dt(early_init_dt_scan_memory_ppc, NULL); |
846f77b0 | 722 | |
846f77b0 ME |
723 | parse_early_param(); |
724 | ||
ba14f649 KG |
725 | /* make sure we've parsed cmdline for mem= before this */ |
726 | if (memory_limit) | |
a84fcd46 | 727 | first_memblock_size = min_t(u64, first_memblock_size, memory_limit); |
ba14f649 | 728 | setup_initial_memory_limit(memstart_addr, first_memblock_size); |
95f72d1e YL |
729 | /* Reserve MEMBLOCK regions used by kernel, initrd, dt, etc... */ |
730 | memblock_reserve(PHYSICAL_START, __pa(klimit) - PHYSICAL_START); | |
549e8152 PM |
731 | /* If relocatable, reserve first 32k for interrupt vectors etc. */ |
732 | if (PHYSICAL_START > MEMORY_START) | |
95f72d1e | 733 | memblock_reserve(MEMORY_START, 0x8000); |
47310413 | 734 | reserve_kdump_trampoline(); |
eb39c880 MS |
735 | #ifdef CONFIG_FA_DUMP |
736 | /* | |
737 | * If we fail to reserve memory for firmware-assisted dump then | |
738 | * fallback to kexec based kdump. | |
739 | */ | |
740 | if (fadump_reserve_mem() == 0) | |
741 | #endif | |
742 | reserve_crashkernel(); | |
9b6b563c PM |
743 | early_reserve_mem(); |
744 | ||
14ed7409 | 745 | /* Ensure that total memory size is page-aligned. */ |
6fbef13c | 746 | limit = ALIGN(memory_limit ?: memblock_phys_mem_size(), PAGE_SIZE); |
95f72d1e | 747 | memblock_enforce_memory_limit(limit); |
6ca4f749 | 748 | |
1aadc056 | 749 | memblock_allow_resize(); |
95f72d1e | 750 | memblock_dump_all(); |
2babf5c2 | 751 | |
95f72d1e | 752 | DBG("Phys. mem: %llx\n", memblock_phys_mem_size()); |
2babf5c2 ME |
753 | |
754 | /* We may need to relocate the flat tree, do it now. | |
755 | * FIXME .. and the initrd too? */ | |
756 | move_device_tree(); | |
757 | ||
59f57774 | 758 | allocate_paca_ptrs(); |
1426d5a3 | 759 | |
9b6b563c PM |
760 | DBG("Scanning CPUs ...\n"); |
761 | ||
5a61ef74 NP |
762 | dt_cpu_ftrs_scan(); |
763 | ||
25985edc | 764 | /* Retrieve CPU related informations from the flat tree |
3c726f8d | 765 | * (altivec support, boot CPU ID, ...) |
9b6b563c | 766 | */ |
3c726f8d | 767 | of_scan_flat_dt(early_init_dt_scan_cpus, NULL); |
36ae37e3 | 768 | if (boot_cpuid < 0) { |
d939be3a | 769 | printk("Failed to identify boot CPU !\n"); |
36ae37e3 BH |
770 | BUG(); |
771 | } | |
9b6b563c | 772 | |
7ac87abb ME |
773 | #if defined(CONFIG_SMP) && defined(CONFIG_PPC64) |
774 | /* We'll later wait for secondaries to check in; there are | |
775 | * NCPUS-1 non-boot CPUs :-) | |
776 | */ | |
777 | spinning_secondaries = boot_cpu_count - 1; | |
778 | #endif | |
779 | ||
1a01dc87 ME |
780 | mmu_early_init_devtree(); |
781 | ||
55672ecf MS |
782 | #ifdef CONFIG_PPC_POWERNV |
783 | /* Scan and build the list of machine check recoverable ranges */ | |
784 | of_scan_flat_dt(early_init_dt_scan_recoverable_ranges, NULL); | |
785 | #endif | |
da6a97bf | 786 | epapr_paravirt_early_init(); |
55672ecf | 787 | |
3808a889 BH |
788 | /* Now try to figure out if we are running on LPAR and so on */ |
789 | pseries_probe_fw_features(); | |
790 | ||
166dd7d3 BH |
791 | #ifdef CONFIG_PPC_PS3 |
792 | /* Identify PS3 firmware */ | |
793 | if (of_flat_dt_is_compatible(of_get_flat_dt_root(), "sony,ps3")) | |
794 | powerpc_firmware_features |= FW_FEATURE_PS3_POSSIBLE; | |
795 | #endif | |
796 | ||
07fd1761 CB |
797 | tm_init(); |
798 | ||
9b6b563c PM |
799 | DBG(" <- early_init_devtree()\n"); |
800 | } | |
801 | ||
b27652dd KH |
802 | #ifdef CONFIG_RELOCATABLE |
803 | /* | |
804 | * This function run before early_init_devtree, so we have to init | |
805 | * initial_boot_params. | |
806 | */ | |
807 | void __init early_get_first_memblock_info(void *params, phys_addr_t *size) | |
808 | { | |
809 | /* Setup flat device-tree pointer */ | |
810 | initial_boot_params = params; | |
811 | ||
812 | /* | |
813 | * Scan the memory nodes and set add_mem_to_memblock to 0 to avoid | |
814 | * mess the memblock. | |
815 | */ | |
816 | add_mem_to_memblock = 0; | |
817 | of_scan_flat_dt(early_init_dt_scan_root, NULL); | |
818 | of_scan_flat_dt(early_init_dt_scan_memory_ppc, NULL); | |
819 | add_mem_to_memblock = 1; | |
820 | ||
821 | if (size) | |
822 | *size = first_memblock_size; | |
823 | } | |
824 | #endif | |
825 | ||
9b6b563c PM |
826 | /******* |
827 | * | |
828 | * New implementation of the OF "find" APIs, return a refcounted | |
829 | * object, call of_node_put() when done. The device tree and list | |
830 | * are protected by a rw_lock. | |
831 | * | |
832 | * Note that property management will need some locking as well, | |
833 | * this isn't dealt with yet. | |
834 | * | |
835 | *******/ | |
836 | ||
b37193b7 BH |
837 | /** |
838 | * of_get_ibm_chip_id - Returns the IBM "chip-id" of a device | |
839 | * @np: device node of the device | |
840 | * | |
841 | * This looks for a property "ibm,chip-id" in the node or any | |
842 | * of its parents and returns its content, or -1 if it cannot | |
843 | * be found. | |
844 | */ | |
845 | int of_get_ibm_chip_id(struct device_node *np) | |
846 | { | |
847 | of_node_get(np); | |
1856f50c | 848 | while (np) { |
1856f50c | 849 | u32 chip_id; |
b37193b7 | 850 | |
1856f50c CJ |
851 | /* |
852 | * Skiboot may produce memory nodes that contain more than one | |
853 | * cell in chip-id, we only read the first one here. | |
854 | */ | |
855 | if (!of_property_read_u32(np, "ibm,chip-id", &chip_id)) { | |
b37193b7 | 856 | of_node_put(np); |
1856f50c | 857 | return chip_id; |
b37193b7 | 858 | } |
16c1d606 ME |
859 | |
860 | np = of_get_next_parent(np); | |
b37193b7 BH |
861 | } |
862 | return -1; | |
863 | } | |
b130e7c0 | 864 | EXPORT_SYMBOL(of_get_ibm_chip_id); |
b37193b7 | 865 | |
3eb906c6 ME |
866 | /** |
867 | * cpu_to_chip_id - Return the cpus chip-id | |
868 | * @cpu: The logical cpu number. | |
869 | * | |
870 | * Return the value of the ibm,chip-id property corresponding to the given | |
871 | * logical cpu number. If the chip-id can not be found, returns -1. | |
872 | */ | |
873 | int cpu_to_chip_id(int cpu) | |
874 | { | |
875 | struct device_node *np; | |
876 | ||
877 | np = of_get_cpu_node(cpu, NULL); | |
878 | if (!np) | |
879 | return -1; | |
880 | ||
881 | of_node_put(np); | |
882 | return of_get_ibm_chip_id(np); | |
883 | } | |
884 | EXPORT_SYMBOL(cpu_to_chip_id); | |
885 | ||
819d5965 SK |
886 | bool arch_match_cpu_phys_id(int cpu, u64 phys_id) |
887 | { | |
9f593f13 NP |
888 | #ifdef CONFIG_SMP |
889 | /* | |
890 | * Early firmware scanning must use this rather than | |
891 | * get_hard_smp_processor_id because we don't have pacas allocated | |
892 | * until memory topology is discovered. | |
893 | */ | |
894 | if (cpu_to_phys_id != NULL) | |
895 | return (int)phys_id == cpu_to_phys_id[cpu]; | |
896 | #endif | |
897 | ||
819d5965 SK |
898 | return (int)phys_id == get_hard_smp_processor_id(cpu); |
899 | } |