1 /*
2 * SCOM support for A2 platforms
3 *
4 * Copyright 2007-2011 Benjamin Herrenschmidt, David Gibson,
5 * Michael Ellerman, IBM Corp.
6 *
7 * This program is free software; you can redistribute it and/or
8 * modify it under the terms of the GNU General Public License
9 * as published by the Free Software Foundation; either version
10 * 2 of the License, or (at your option) any later version.
11 */
12
13 #include <linux/cpumask.h>
14 #include <linux/io.h>
15 #include <linux/of.h>
16 #include <linux/spinlock.h>
17 #include <linux/types.h>
18
19 #include <asm/cputhreads.h>
20 #include <asm/reg_a2.h>
21 #include <asm/scom.h>
22 #include <asm/udbg.h>
23
24 #include "wsp.h"
25
26 #define SCOM_RAMC 0x2a /* Ram Command */
27 #define SCOM_RAMC_TGT1_EXT 0x80000000
28 #define SCOM_RAMC_SRC1_EXT 0x40000000
29 #define SCOM_RAMC_SRC2_EXT 0x20000000
30 #define SCOM_RAMC_SRC3_EXT 0x10000000
31 #define SCOM_RAMC_ENABLE 0x00080000
32 #define SCOM_RAMC_THREADSEL 0x00060000
33 #define SCOM_RAMC_EXECUTE 0x00010000
34 #define SCOM_RAMC_MSR_OVERRIDE 0x00008000
35 #define SCOM_RAMC_MSR_PR 0x00004000
36 #define SCOM_RAMC_MSR_GS 0x00002000
37 #define SCOM_RAMC_FORCE 0x00001000
38 #define SCOM_RAMC_FLUSH 0x00000800
39 #define SCOM_RAMC_INTERRUPT 0x00000004
40 #define SCOM_RAMC_ERROR 0x00000002
41 #define SCOM_RAMC_DONE 0x00000001
42 #define SCOM_RAMI 0x29 /* Ram Instruction */
43 #define SCOM_RAMIC 0x28 /* Ram Instruction and Command */
44 #define SCOM_RAMIC_INSN 0xffffffff00000000
45 #define SCOM_RAMD 0x2d /* Ram Data */
46 #define SCOM_RAMDH 0x2e /* Ram Data High */
47 #define SCOM_RAMDL 0x2f /* Ram Data Low */
48 #define SCOM_PCCR0 0x33 /* PC Configuration Register 0 */
49 #define SCOM_PCCR0_ENABLE_DEBUG 0x80000000
50 #define SCOM_PCCR0_ENABLE_RAM 0x40000000
51 #define SCOM_THRCTL 0x30 /* Thread Control and Status */
52 #define SCOM_THRCTL_T0_STOP 0x80000000
53 #define SCOM_THRCTL_T1_STOP 0x40000000
54 #define SCOM_THRCTL_T2_STOP 0x20000000
55 #define SCOM_THRCTL_T3_STOP 0x10000000
56 #define SCOM_THRCTL_T0_STEP 0x08000000
57 #define SCOM_THRCTL_T1_STEP 0x04000000
58 #define SCOM_THRCTL_T2_STEP 0x02000000
59 #define SCOM_THRCTL_T3_STEP 0x01000000
60 #define SCOM_THRCTL_T0_RUN 0x00800000
61 #define SCOM_THRCTL_T1_RUN 0x00400000
62 #define SCOM_THRCTL_T2_RUN 0x00200000
63 #define SCOM_THRCTL_T3_RUN 0x00100000
64 #define SCOM_THRCTL_T0_PM 0x00080000
65 #define SCOM_THRCTL_T1_PM 0x00040000
66 #define SCOM_THRCTL_T2_PM 0x00020000
67 #define SCOM_THRCTL_T3_PM 0x00010000
68 #define SCOM_THRCTL_T0_UDE 0x00008000
69 #define SCOM_THRCTL_T1_UDE 0x00004000
70 #define SCOM_THRCTL_T2_UDE 0x00002000
71 #define SCOM_THRCTL_T3_UDE 0x00001000
72 #define SCOM_THRCTL_ASYNC_DIS 0x00000800
73 #define SCOM_THRCTL_TB_DIS 0x00000400
74 #define SCOM_THRCTL_DEC_DIS 0x00000200
75 #define SCOM_THRCTL_AND 0x31 /* Thread Control and Status */
76 #define SCOM_THRCTL_OR 0x32 /* Thread Control and Status */
77
78
79 static DEFINE_PER_CPU(scom_map_t, scom_ptrs);
80
get_scom(int cpu,struct device_node * np,int * first_thread)81 static scom_map_t get_scom(int cpu, struct device_node *np, int *first_thread)
82 {
83 scom_map_t scom = per_cpu(scom_ptrs, cpu);
84 int tcpu;
85
86 if (scom_map_ok(scom)) {
87 *first_thread = 0;
88 return scom;
89 }
90
91 *first_thread = 1;
92
93 scom = scom_map_device(np, 0);
94
95 for (tcpu = cpu_first_thread_sibling(cpu);
96 tcpu <= cpu_last_thread_sibling(cpu); tcpu++)
97 per_cpu(scom_ptrs, tcpu) = scom;
98
99 /* Hack: for the boot core, this will actually get called on
100 * the second thread up, not the first so our test above will
101 * set first_thread incorrectly. */
102 if (cpu_first_thread_sibling(cpu) == 0)
103 *first_thread = 0;
104
105 return scom;
106 }
107
a2_scom_ram(scom_map_t scom,int thread,u32 insn,int extmask)108 static int a2_scom_ram(scom_map_t scom, int thread, u32 insn, int extmask)
109 {
110 u64 cmd, mask, val;
111 int n = 0;
112
113 cmd = ((u64)insn << 32) | (((u64)extmask & 0xf) << 28)
114 | ((u64)thread << 17) | SCOM_RAMC_ENABLE | SCOM_RAMC_EXECUTE;
115 mask = SCOM_RAMC_DONE | SCOM_RAMC_INTERRUPT | SCOM_RAMC_ERROR;
116
117 scom_write(scom, SCOM_RAMIC, cmd);
118
119 while (!((val = scom_read(scom, SCOM_RAMC)) & mask)) {
120 pr_devel("Waiting on RAMC = 0x%llx\n", val);
121 if (++n == 3) {
122 pr_err("RAMC timeout on instruction 0x%08x, thread %d\n",
123 insn, thread);
124 return -1;
125 }
126 }
127
128 if (val & SCOM_RAMC_INTERRUPT) {
129 pr_err("RAMC interrupt on instruction 0x%08x, thread %d\n",
130 insn, thread);
131 return -SCOM_RAMC_INTERRUPT;
132 }
133
134 if (val & SCOM_RAMC_ERROR) {
135 pr_err("RAMC error on instruction 0x%08x, thread %d\n",
136 insn, thread);
137 return -SCOM_RAMC_ERROR;
138 }
139
140 return 0;
141 }
142
a2_scom_getgpr(scom_map_t scom,int thread,int gpr,int alt,u64 * out_gpr)143 static int a2_scom_getgpr(scom_map_t scom, int thread, int gpr, int alt,
144 u64 *out_gpr)
145 {
146 int rc;
147
148 /* or rN, rN, rN */
149 u32 insn = 0x7c000378 | (gpr << 21) | (gpr << 16) | (gpr << 11);
150 rc = a2_scom_ram(scom, thread, insn, alt ? 0xf : 0x0);
151 if (rc)
152 return rc;
153
154 *out_gpr = scom_read(scom, SCOM_RAMD);
155
156 return 0;
157 }
158
a2_scom_getspr(scom_map_t scom,int thread,int spr,u64 * out_spr)159 static int a2_scom_getspr(scom_map_t scom, int thread, int spr, u64 *out_spr)
160 {
161 int rc, sprhi, sprlo;
162 u32 insn;
163
164 sprhi = spr >> 5;
165 sprlo = spr & 0x1f;
166 insn = 0x7c2002a6 | (sprlo << 16) | (sprhi << 11); /* mfspr r1,spr */
167
168 if (spr == 0x0ff0)
169 insn = 0x7c2000a6; /* mfmsr r1 */
170
171 rc = a2_scom_ram(scom, thread, insn, 0xf);
172 if (rc)
173 return rc;
174 return a2_scom_getgpr(scom, thread, 1, 1, out_spr);
175 }
176
a2_scom_setgpr(scom_map_t scom,int thread,int gpr,int alt,u64 val)177 static int a2_scom_setgpr(scom_map_t scom, int thread, int gpr,
178 int alt, u64 val)
179 {
180 u32 lis = 0x3c000000 | (gpr << 21);
181 u32 li = 0x38000000 | (gpr << 21);
182 u32 oris = 0x64000000 | (gpr << 21) | (gpr << 16);
183 u32 ori = 0x60000000 | (gpr << 21) | (gpr << 16);
184 u32 rldicr32 = 0x780007c6 | (gpr << 21) | (gpr << 16);
185 u32 highest = val >> 48;
186 u32 higher = (val >> 32) & 0xffff;
187 u32 high = (val >> 16) & 0xffff;
188 u32 low = val & 0xffff;
189 int lext = alt ? 0x8 : 0x0;
190 int oext = alt ? 0xf : 0x0;
191 int rc = 0;
192
193 if (highest)
194 rc |= a2_scom_ram(scom, thread, lis | highest, lext);
195
196 if (higher) {
197 if (highest)
198 rc |= a2_scom_ram(scom, thread, oris | higher, oext);
199 else
200 rc |= a2_scom_ram(scom, thread, li | higher, lext);
201 }
202
203 if (highest || higher)
204 rc |= a2_scom_ram(scom, thread, rldicr32, oext);
205
206 if (high) {
207 if (highest || higher)
208 rc |= a2_scom_ram(scom, thread, oris | high, oext);
209 else
210 rc |= a2_scom_ram(scom, thread, lis | high, lext);
211 }
212
213 if (highest || higher || high)
214 rc |= a2_scom_ram(scom, thread, ori | low, oext);
215 else
216 rc |= a2_scom_ram(scom, thread, li | low, lext);
217
218 return rc;
219 }
220
a2_scom_setspr(scom_map_t scom,int thread,int spr,u64 val)221 static int a2_scom_setspr(scom_map_t scom, int thread, int spr, u64 val)
222 {
223 int sprhi = spr >> 5;
224 int sprlo = spr & 0x1f;
225 /* mtspr spr, r1 */
226 u32 insn = 0x7c2003a6 | (sprlo << 16) | (sprhi << 11);
227
228 if (spr == 0x0ff0)
229 insn = 0x7c200124; /* mtmsr r1 */
230
231 if (a2_scom_setgpr(scom, thread, 1, 1, val))
232 return -1;
233
234 return a2_scom_ram(scom, thread, insn, 0xf);
235 }
236
a2_scom_initial_tlb(scom_map_t scom,int thread)237 static int a2_scom_initial_tlb(scom_map_t scom, int thread)
238 {
239 extern u32 a2_tlbinit_code_start[], a2_tlbinit_code_end[];
240 extern u32 a2_tlbinit_after_iprot_flush[];
241 extern u32 a2_tlbinit_after_linear_map[];
242 u32 assoc, entries, i;
243 u64 epn, tlbcfg;
244 u32 *p;
245 int rc;
246
247 /* Invalidate all entries (including iprot) */
248
249 rc = a2_scom_getspr(scom, thread, SPRN_TLB0CFG, &tlbcfg);
250 if (rc)
251 goto scom_fail;
252 entries = tlbcfg & TLBnCFG_N_ENTRY;
253 assoc = (tlbcfg & TLBnCFG_ASSOC) >> 24;
254 epn = 0;
255
256 /* Set MMUCR2 to enable 4K, 64K, 1M, 16M and 1G pages */
257 a2_scom_setspr(scom, thread, SPRN_MMUCR2, 0x000a7531);
258 /* Set MMUCR3 to write all thids bit to the TLB */
259 a2_scom_setspr(scom, thread, SPRN_MMUCR3, 0x0000000f);
260
261 /* Set MAS1 for 1G page size, and MAS2 to our initial EPN */
262 a2_scom_setspr(scom, thread, SPRN_MAS1, MAS1_TSIZE(BOOK3E_PAGESZ_1GB));
263 a2_scom_setspr(scom, thread, SPRN_MAS2, epn);
264 for (i = 0; i < entries; i++) {
265
266 a2_scom_setspr(scom, thread, SPRN_MAS0, MAS0_ESEL(i % assoc));
267
268 /* tlbwe */
269 rc = a2_scom_ram(scom, thread, 0x7c0007a4, 0);
270 if (rc)
271 goto scom_fail;
272
273 /* Next entry is new address? */
274 if((i + 1) % assoc == 0) {
275 epn += (1 << 30);
276 a2_scom_setspr(scom, thread, SPRN_MAS2, epn);
277 }
278 }
279
280 /* Setup args for linear mapping */
281 rc = a2_scom_setgpr(scom, thread, 3, 0, MAS0_TLBSEL(0));
282 if (rc)
283 goto scom_fail;
284
285 /* Linear mapping */
286 for (p = a2_tlbinit_code_start; p < a2_tlbinit_after_linear_map; p++) {
287 rc = a2_scom_ram(scom, thread, *p, 0);
288 if (rc)
289 goto scom_fail;
290 }
291
292 /*
293 * For the boot thread, between the linear mapping and the debug
294 * mappings there is a loop to flush iprot mappings. Ramming doesn't do
295 * branches, but the secondary threads don't need to be nearly as smart
296 * (i.e. we don't need to worry about invalidating the mapping we're
297 * standing on).
298 */
299
300 /* Debug mappings. Expects r11 = MAS0 from linear map (set above) */
301 for (p = a2_tlbinit_after_iprot_flush; p < a2_tlbinit_code_end; p++) {
302 rc = a2_scom_ram(scom, thread, *p, 0);
303 if (rc)
304 goto scom_fail;
305 }
306
307 scom_fail:
308 if (rc)
309 pr_err("Setting up initial TLB failed, err %d\n", rc);
310
311 if (rc == -SCOM_RAMC_INTERRUPT) {
312 /* Interrupt, dump some status */
313 int rc[10];
314 u64 iar, srr0, srr1, esr, mas0, mas1, mas2, mas7_3, mas8, ccr2;
315 rc[0] = a2_scom_getspr(scom, thread, SPRN_IAR, &iar);
316 rc[1] = a2_scom_getspr(scom, thread, SPRN_SRR0, &srr0);
317 rc[2] = a2_scom_getspr(scom, thread, SPRN_SRR1, &srr1);
318 rc[3] = a2_scom_getspr(scom, thread, SPRN_ESR, &esr);
319 rc[4] = a2_scom_getspr(scom, thread, SPRN_MAS0, &mas0);
320 rc[5] = a2_scom_getspr(scom, thread, SPRN_MAS1, &mas1);
321 rc[6] = a2_scom_getspr(scom, thread, SPRN_MAS2, &mas2);
322 rc[7] = a2_scom_getspr(scom, thread, SPRN_MAS7_MAS3, &mas7_3);
323 rc[8] = a2_scom_getspr(scom, thread, SPRN_MAS8, &mas8);
324 rc[9] = a2_scom_getspr(scom, thread, SPRN_A2_CCR2, &ccr2);
325 pr_err(" -> retreived IAR =0x%llx (err %d)\n", iar, rc[0]);
326 pr_err(" retreived SRR0=0x%llx (err %d)\n", srr0, rc[1]);
327 pr_err(" retreived SRR1=0x%llx (err %d)\n", srr1, rc[2]);
328 pr_err(" retreived ESR =0x%llx (err %d)\n", esr, rc[3]);
329 pr_err(" retreived MAS0=0x%llx (err %d)\n", mas0, rc[4]);
330 pr_err(" retreived MAS1=0x%llx (err %d)\n", mas1, rc[5]);
331 pr_err(" retreived MAS2=0x%llx (err %d)\n", mas2, rc[6]);
332 pr_err(" retreived MS73=0x%llx (err %d)\n", mas7_3, rc[7]);
333 pr_err(" retreived MAS8=0x%llx (err %d)\n", mas8, rc[8]);
334 pr_err(" retreived CCR2=0x%llx (err %d)\n", ccr2, rc[9]);
335 }
336
337 return rc;
338 }
339
a2_scom_startup_cpu(unsigned int lcpu,int thr_idx,struct device_node * np)340 int __devinit a2_scom_startup_cpu(unsigned int lcpu, int thr_idx,
341 struct device_node *np)
342 {
343 u64 init_iar, init_msr, init_ccr2;
344 unsigned long start_here;
345 int rc, core_setup;
346 scom_map_t scom;
347 u64 pccr0;
348
349 scom = get_scom(lcpu, np, &core_setup);
350 if (!scom) {
351 printk(KERN_ERR "Couldn't map SCOM for CPU%d\n", lcpu);
352 return -1;
353 }
354
355 pr_devel("Bringing up CPU%d using SCOM...\n", lcpu);
356
357 pccr0 = scom_read(scom, SCOM_PCCR0);
358 scom_write(scom, SCOM_PCCR0, pccr0 | SCOM_PCCR0_ENABLE_DEBUG |
359 SCOM_PCCR0_ENABLE_RAM);
360
361 /* Stop the thead with THRCTL. If we are setting up the TLB we stop all
362 * threads. We also disable asynchronous interrupts while RAMing.
363 */
364 if (core_setup)
365 scom_write(scom, SCOM_THRCTL_OR,
366 SCOM_THRCTL_T0_STOP |
367 SCOM_THRCTL_T1_STOP |
368 SCOM_THRCTL_T2_STOP |
369 SCOM_THRCTL_T3_STOP |
370 SCOM_THRCTL_ASYNC_DIS);
371 else
372 scom_write(scom, SCOM_THRCTL_OR, SCOM_THRCTL_T0_STOP >> thr_idx);
373
374 /* Flush its pipeline just in case */
375 scom_write(scom, SCOM_RAMC, ((u64)thr_idx << 17) |
376 SCOM_RAMC_FLUSH | SCOM_RAMC_ENABLE);
377
378 a2_scom_getspr(scom, thr_idx, SPRN_IAR, &init_iar);
379 a2_scom_getspr(scom, thr_idx, 0x0ff0, &init_msr);
380 a2_scom_getspr(scom, thr_idx, SPRN_A2_CCR2, &init_ccr2);
381
382 /* Set MSR to MSR_CM (0x0ff0 is magic value for MSR_CM) */
383 rc = a2_scom_setspr(scom, thr_idx, 0x0ff0, MSR_CM);
384 if (rc) {
385 pr_err("Failed to set MSR ! err %d\n", rc);
386 return rc;
387 }
388
389 /* RAM in an sync/isync for the sake of it */
390 a2_scom_ram(scom, thr_idx, 0x7c0004ac, 0);
391 a2_scom_ram(scom, thr_idx, 0x4c00012c, 0);
392
393 if (core_setup) {
394 pr_devel("CPU%d is first thread in core, initializing TLB...\n",
395 lcpu);
396 rc = a2_scom_initial_tlb(scom, thr_idx);
397 if (rc)
398 goto fail;
399 }
400
401 start_here = *(unsigned long *)(core_setup ? generic_secondary_smp_init
402 : generic_secondary_thread_init);
403 pr_devel("CPU%d entry point at 0x%lx...\n", lcpu, start_here);
404
405 rc |= a2_scom_setspr(scom, thr_idx, SPRN_IAR, start_here);
406 rc |= a2_scom_setgpr(scom, thr_idx, 3, 0,
407 get_hard_smp_processor_id(lcpu));
408 /*
409 * Tell book3e_secondary_core_init not to set up the TLB, we've
410 * already done that.
411 */
412 rc |= a2_scom_setgpr(scom, thr_idx, 4, 0, 1);
413
414 rc |= a2_scom_setspr(scom, thr_idx, SPRN_TENS, 0x1 << thr_idx);
415
416 scom_write(scom, SCOM_RAMC, 0);
417 scom_write(scom, SCOM_THRCTL_AND, ~(SCOM_THRCTL_T0_STOP >> thr_idx));
418 scom_write(scom, SCOM_PCCR0, pccr0);
419 fail:
420 pr_devel(" SCOM initialization %s\n", rc ? "failed" : "succeeded");
421 if (rc) {
422 pr_err("Old IAR=0x%08llx MSR=0x%08llx CCR2=0x%08llx\n",
423 init_iar, init_msr, init_ccr2);
424 }
425
426 return rc;
427 }
428