xref: /illumos-gate/usr/src/uts/sun4u/seattle/os/seattle.c (revision ec5d0d67)
1 /*
2  * CDDL HEADER START
3  *
4  * The contents of this file are subject to the terms of the
5  * Common Development and Distribution License (the "License").
6  * You may not use this file except in compliance with the License.
7  *
8  * You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE
9  * or http://www.opensolaris.org/os/licensing.
10  * See the License for the specific language governing permissions
11  * and limitations under the License.
12  *
13  * When distributing Covered Code, include this CDDL HEADER in each
14  * file and include the License file at usr/src/OPENSOLARIS.LICENSE.
15  * If applicable, add the following below this CDDL HEADER, with the
16  * fields enclosed by brackets "[]" replaced with your own identifying
17  * information: Portions Copyright [yyyy] [name of copyright owner]
18  *
19  * CDDL HEADER END
20  */
21 
22 /*
23  * Copyright 2007 Sun Microsystems, Inc.  All rights reserved.
24  * Use is subject to license terms.
25  */
26 
27 #pragma ident	"%Z%%M%	%I%	%E% SMI"
28 
29 #include <sys/param.h>
30 #include <sys/systm.h>
31 #include <sys/sysmacros.h>
32 #include <sys/sunddi.h>
33 #include <sys/esunddi.h>
34 
35 #include <sys/platform_module.h>
36 #include <sys/errno.h>
37 #include <sys/cpu_sgnblk_defs.h>
38 #include <sys/rmc_comm_dp.h>
39 #include <sys/rmc_comm_drvintf.h>
40 #include <sys/modctl.h>
41 #include <sys/lgrp.h>
42 #include <sys/memnode.h>
43 #include <sys/promif.h>
44 
45 /* Anything related to shared i2c access applies to Seattle only */
46 #define	SHARED_MI2CV_PATH "/i2c@1f,530000"
47 static dev_info_t *shared_mi2cv_dip;
48 static kmutex_t mi2cv_mutex;
49 
50 int (*p2get_mem_unum)(int, uint64_t, char *, int, int *);
51 static void cpu_sgn_update(ushort_t, uchar_t, uchar_t, int);
52 int (*rmc_req_now)(rmc_comm_msg_t *, uint8_t) = NULL;
53 
54 void
55 startup_platform(void)
56 {
57 	mutex_init(&mi2cv_mutex, NULL, NULL, NULL);
58 }
59 
60 int
61 set_platform_tsb_spares()
62 {
63 	return (0);
64 }
65 
66 void
67 set_platform_defaults(void)
68 {
69 	extern char *tod_module_name;
70 	/* Set appropriate tod module */
71 	if (tod_module_name == NULL)
72 		tod_module_name = "todm5823";
73 
74 	cpu_sgn_func = cpu_sgn_update;
75 }
76 
77 /*
78  * these two dummy functions are loaded over the original
79  * todm5823 set and clear_power_alarm functions. On Seattle
80  * these functions are not supported, and thus we need to provide
81  * dummy functions that just returns.
82  * On Seattle, clock chip is not persistant across reboots,
83  * and moreover it has a bug sending memory access.
84  * This fix is done by writing over the original
85  * tod_ops function pointer with our dummy replacement functions.
86  */
87 /*ARGSUSED*/
88 static void
89 dummy_todm5823_set_power_alarm(timestruc_t ts)
90 {
91 }
92 
93 static void
94 dummy_todm5823_clear_power_alarm(void)
95 {
96 }
97 
98 /*
99  * Definitions for accessing the pci config space of the isa node
100  * of Southbridge.
101  */
102 static ddi_acc_handle_t isa_handle = NULL;	/* handle for isa pci space */
103 
104 /*
105  * Definition for accessing rmclomv
106  */
107 #define	RMCLOMV_PATHNAME	"/pseudo/rmclomv@0"
108 
109 void
110 load_platform_drivers(void)
111 {
112 	dev_info_t	*rmclomv_dip;
113 	/*
114 	 * It is OK to return error because 'us' driver is not available
115 	 * in all clusters (e.g. missing in Core cluster).
116 	 */
117 	(void) i_ddi_attach_hw_nodes("us");
118 
119 
120 	/*
121 	 * mc-us3i must stay loaded for plat_get_mem_unum()
122 	 */
123 	if (i_ddi_attach_hw_nodes("mc-us3i") != DDI_SUCCESS)
124 		cmn_err(CE_WARN, "mc-us3i driver failed to install");
125 	(void) ddi_hold_driver(ddi_name_to_major("mc-us3i"));
126 
127 	/*
128 	 * load the power button driver
129 	 */
130 	if (i_ddi_attach_hw_nodes("power") != DDI_SUCCESS)
131 		cmn_err(CE_WARN, "power button driver failed to install");
132 	else
133 		(void) ddi_hold_driver(ddi_name_to_major("power"));
134 
135 	/*
136 	 * load the GPIO driver for the ALOM reset and watchdog lines
137 	 */
138 	if (i_ddi_attach_hw_nodes("pmugpio") != DDI_SUCCESS)
139 		cmn_err(CE_WARN, "pmugpio failed to install");
140 	else {
141 		extern int watchdog_enable, watchdog_available;
142 		extern int disable_watchdog_on_exit;
143 
144 		/*
145 		 * Disable an active h/w watchdog timer upon exit to OBP.
146 		 */
147 		disable_watchdog_on_exit = 1;
148 
149 		watchdog_enable = 1;
150 		watchdog_available = 1;
151 	}
152 	(void) ddi_hold_driver(ddi_name_to_major("pmugpio"));
153 
154 	/*
155 	 * Figure out which mi2cv dip is shared with OBP for the nvram
156 	 * device, so the lock can be acquired.
157 	 */
158 	shared_mi2cv_dip = e_ddi_hold_devi_by_path(SHARED_MI2CV_PATH, 0);
159 	/*
160 	 * Load the environmentals driver (rmclomv)
161 	 *
162 	 * We need this driver to handle events from the RMC when state
163 	 * changes occur in the environmental data.
164 	 */
165 	if (i_ddi_attach_hw_nodes("rmc_comm") != DDI_SUCCESS) {
166 		cmn_err(CE_WARN, "rmc_comm failed to install");
167 	} else {
168 		(void) ddi_hold_driver(ddi_name_to_major("rmc_comm"));
169 
170 		rmclomv_dip = e_ddi_hold_devi_by_path(RMCLOMV_PATHNAME, 0);
171 		if (rmclomv_dip == NULL) {
172 			cmn_err(CE_WARN, "Could not install rmclomv driver\n");
173 		}
174 	}
175 
176 	/*
177 	 * These two dummy functions are loaded over the original
178 	 * todm5823 set and clear_power_alarm functions. On Seattle,
179 	 * these functionalities are not supported.
180 	 * The load_platform_drivers(void) is called from post_startup()
181 	 * which is after all the initialization of the tod module is
182 	 * finished, then we replace 2 of the tod_ops function pointers
183 	 * with our dummy version.
184 	 */
185 	tod_ops.tod_set_power_alarm = dummy_todm5823_set_power_alarm;
186 	tod_ops.tod_clear_power_alarm = dummy_todm5823_clear_power_alarm;
187 
188 	/*
189 	 * create a handle to the rmc_comm_request_nowait() function
190 	 * inside the rmc_comm module.
191 	 *
192 	 * The Seattle/Boston todm5823 driver will use this handle to
193 	 * use the rmc_comm_request_nowait() function to send time/date
194 	 * updates to ALOM.
195 	 */
196 	rmc_req_now = (int (*)(rmc_comm_msg_t *, uint8_t))
197 		modgetsymvalue("rmc_comm_request_nowait", 0);
198 }
199 
200 /*
201  * This routine is needed if a device error or timeout occurs before the
202  * driver is loaded.
203  */
204 /*ARGSUSED*/
205 int
206 plat_ide_chipreset(dev_info_t *dip, int chno)
207 {
208 	int	ret = DDI_SUCCESS;
209 
210 	if (isa_handle == NULL) {
211 		return (DDI_FAILURE);
212 	}
213 
214 	/*
215 	 * This will be filled in with the reset logic
216 	 * for the ULI1573 when that becomes available.
217 	 * currently this is just a stub.
218 	 */
219 	return (ret);
220 }
221 
222 
223 /*ARGSUSED*/
224 int
225 plat_cpu_poweron(struct cpu *cp)
226 {
227 	return (ENOTSUP);	/* not supported on this platform */
228 }
229 
230 /*ARGSUSED*/
231 int
232 plat_cpu_poweroff(struct cpu *cp)
233 {
234 	return (ENOTSUP);	/* not supported on this platform */
235 }
236 
237 /*ARGSUSED*/
238 void
239 plat_freelist_process(int mnode)
240 {
241 }
242 
243 char *platform_module_list[] = {
244 	"mi2cv",
245 	"pca9556",
246 	(char *)0
247 };
248 
249 /*ARGSUSED*/
250 void
251 plat_tod_fault(enum tod_fault_type tod_bad)
252 {
253 }
254 
255 /*ARGSUSED*/
256 int
257 plat_get_mem_unum(int synd_code, uint64_t flt_addr, int flt_bus_id,
258     int flt_in_memory, ushort_t flt_status, char *buf, int buflen, int *lenp)
259 {
260 	if (flt_in_memory && (p2get_mem_unum != NULL))
261 		return (p2get_mem_unum(synd_code, P2ALIGN(flt_addr, 8),
262 		    buf, buflen, lenp));
263 	else
264 		return (ENOTSUP);
265 }
266 
267 /*
268  * This platform hook gets called from mc_add_mem_unum_label() in the mc-us3i
269  * driver giving each platform the opportunity to add platform
270  * specific label information to the unum for ECC error logging purposes.
271  */
272 /*ARGSUSED*/
273 void
274 plat_add_mem_unum_label(char *unum, int mcid, int bank, int dimm)
275 {
276 	char old_unum[UNUM_NAMLEN];
277 	int printed;
278 	int buflen = UNUM_NAMLEN;
279 	strcpy(old_unum, unum);
280 	printed = snprintf(unum, buflen, "MB/P%d/B%d", mcid, bank);
281 	buflen -= printed;
282 	unum += printed;
283 
284 	if (dimm != -1) {
285 		printed = snprintf(unum, buflen, "/D%d", dimm);
286 		buflen -= printed;
287 		unum += printed;
288 	}
289 
290 	snprintf(unum, buflen, ": %s", old_unum);
291 }
292 
293 /*ARGSUSED*/
294 int
295 plat_get_cpu_unum(int cpuid, char *buf, int buflen, int *lenp)
296 {
297 	if (snprintf(buf, buflen, "MB") >= buflen) {
298 		return (ENOSPC);
299 	} else {
300 		*lenp = strlen(buf);
301 		return (0);
302 	}
303 }
304 
305 /*
306  * Our nodename has been set, pass it along to the RMC.
307  */
308 void
309 plat_nodename_set(void)
310 {
311 	rmc_comm_msg_t	req;	/* request */
312 	int (*rmc_req_res)(rmc_comm_msg_t *, rmc_comm_msg_t *, time_t) = NULL;
313 
314 	/*
315 	 * find the symbol for the mailbox routine
316 	 */
317 	rmc_req_res = (int (*)(rmc_comm_msg_t *, rmc_comm_msg_t *, time_t))
318 		modgetsymvalue("rmc_comm_request_response", 0);
319 
320 	if (rmc_req_res == NULL) {
321 		return;
322 	}
323 
324 	/*
325 	 * construct the message telling the RMC our nodename
326 	 */
327 	req.msg_type = DP_SET_CPU_NODENAME;
328 	req.msg_len = strlen(utsname.nodename) + 1;
329 	req.msg_bytes = 0;
330 	req.msg_buf = (caddr_t)utsname.nodename;
331 
332 	/*
333 	 * ship it
334 	 */
335 	(void) (rmc_req_res)(&req, NULL, 2000);
336 }
337 
338 sig_state_t current_sgn;
339 
340 /*
341  * cpu signatures - we're only interested in the overall system
342  * "signature" on this platform - not individual cpu signatures
343  */
344 /*ARGSUSED*/
345 static void
346 cpu_sgn_update(ushort_t sig, uchar_t state, uchar_t sub_state, int cpuid)
347 {
348 	dp_cpu_signature_t signature;
349 	rmc_comm_msg_t	req;	/* request */
350 	int (*rmc_req_now)(rmc_comm_msg_t *, uint8_t) = NULL;
351 
352 
353 	/*
354 	 * Differentiate a panic reboot from a non-panic reboot in the
355 	 * setting of the substate of the signature.
356 	 *
357 	 * If the new substate is REBOOT and we're rebooting due to a panic,
358 	 * then set the new substate to a special value indicating a panic
359 	 * reboot, SIGSUBST_PANIC_REBOOT.
360 	 *
361 	 * A panic reboot is detected by a current (previous) signature
362 	 * state of SIGST_EXIT, and a new signature substate of SIGSUBST_REBOOT.
363 	 * The domain signature state SIGST_EXIT is used as the panic flow
364 	 * progresses.
365 	 *
366 	 * At the end of the panic flow, the reboot occurs but we should know
367 	 * one that was involuntary, something that may be quite useful to know
368 	 * at OBP level.
369 	 */
370 	if (state == SIGST_EXIT && sub_state == SIGSUBST_REBOOT) {
371 		if (current_sgn.state_t.state == SIGST_EXIT &&
372 		    current_sgn.state_t.sub_state != SIGSUBST_REBOOT)
373 			sub_state = SIGSUBST_PANIC_REBOOT;
374 	}
375 
376 	/*
377 	 * offline and detached states only apply to a specific cpu
378 	 * so ignore them.
379 	 */
380 	if (state == SIGST_OFFLINE || state == SIGST_DETACHED) {
381 		return;
382 	}
383 
384 	current_sgn.signature = CPU_SIG_BLD(sig, state, sub_state);
385 
386 	/*
387 	 * find the symbol for the mailbox routine
388 	 */
389 	rmc_req_now = (int (*)(rmc_comm_msg_t *, uint8_t))
390 		modgetsymvalue("rmc_comm_request_nowait", 0);
391 	if (rmc_req_now == NULL) {
392 		return;
393 	}
394 
395 	signature.cpu_id = -1;
396 	signature.sig = sig;
397 	signature.states = state;
398 	signature.sub_state = sub_state;
399 	req.msg_type = DP_SET_CPU_SIGNATURE;
400 	req.msg_len = (int)(sizeof (signature));
401 	req.msg_bytes = 0;
402 	req.msg_buf = (caddr_t)&signature;
403 
404 	/*
405 	 * We need to tell the SP that the host is about to stop running.  The
406 	 * SP will then allow the date to be set at its console, it will change
407 	 * state of the activity indicator, it will display the correct host
408 	 * status, and it will stop sending console messages and alerts to the
409 	 * host communication channel.
410 	 *
411 	 * This requires the RMC_COMM_DREQ_URGENT as we want to
412 	 * be sure activity indicators will reflect the correct status.
413 	 *
414 	 * When sub_state SIGSUBST_DUMP is sent, the urgent flag
415 	 * (RMC_COMM_DREQ_URGENT) is not required as SIGSUBST_PANIC_REBOOT
416 	 * has already been sent and changed activity indicators.
417 	 */
418 	if (state == SIGST_EXIT && (sub_state == SIGSUBST_HALT ||
419 	    sub_state == SIGSUBST_REBOOT || sub_state == SIGSUBST_ENVIRON ||
420 	    sub_state == SIGSUBST_PANIC_REBOOT))
421 		(void) (rmc_req_now)(&req, RMC_COMM_DREQ_URGENT);
422 	else
423 		(void) (rmc_req_now)(&req, 0);
424 }
425 
426 /*
427  * Fiesta support for lgroups.
428  *
429  * On fiesta platform, an lgroup platform handle == CPU id
430  */
431 
432 /*
433  * Macro for extracting the CPU number from the CPU id
434  */
435 #define	CPUID_TO_LGRP(id)	((id) & 0x7)
436 #define	PLATFORM_MC_SHIFT	36
437 
438 /*
439  * Return the platform handle for the lgroup containing the given CPU
440  */
441 void *
442 plat_lgrp_cpu_to_hand(processorid_t id)
443 {
444 	return ((void *)(uintptr_t)CPUID_TO_LGRP(id));
445 }
446 
447 /*
448  * Platform specific lgroup initialization
449  */
450 void
451 plat_lgrp_init(void)
452 {
453 	pnode_t		curnode;
454 	char		tmp_name[sizeof (OBP_CPU) + 1];  /* extra padding */
455 	int		portid;
456 	int		cpucnt = 0;
457 	int		max_portid = -1;
458 	extern uint32_t lgrp_expand_proc_thresh;
459 	extern uint32_t lgrp_expand_proc_diff;
460 	extern pgcnt_t	lgrp_mem_free_thresh;
461 	extern uint32_t lgrp_loadavg_tolerance;
462 	extern uint32_t lgrp_loadavg_max_effect;
463 	extern uint32_t lgrp_load_thresh;
464 	extern lgrp_mem_policy_t  lgrp_mem_policy_root;
465 
466 	/*
467 	 * Count the number of CPUs installed to determine if
468 	 * NUMA optimization should be enabled or not.
469 	 *
470 	 * All CPU nodes reside in the root node and have a
471 	 * device type "cpu".
472 	 */
473 	curnode = prom_rootnode();
474 	for (curnode = prom_childnode(curnode); curnode;
475 	    curnode = prom_nextnode(curnode)) {
476 		bzero(tmp_name, sizeof (tmp_name));
477 		if (prom_bounded_getprop(curnode, OBP_DEVICETYPE, tmp_name,
478 		    sizeof (OBP_CPU)) == -1 || strcmp(tmp_name, OBP_CPU) != 0)
479 			continue;
480 
481 		cpucnt++;
482 
483 		if (prom_getprop(curnode, "portid", (caddr_t)&portid) !=
484 		    -1 && portid > max_portid)
485 			max_portid = portid;
486 	}
487 	if (cpucnt <= 1)
488 		max_mem_nodes = 1;
489 	else if (max_portid >= 0 && max_portid < MAX_MEM_NODES)
490 		max_mem_nodes = max_portid + 1;
491 
492 	/*
493 	 * Set tuneables for fiesta architecture
494 	 *
495 	 * lgrp_expand_proc_thresh is the minimum load on the lgroups
496 	 * this process is currently running on before considering
497 	 * expanding threads to another lgroup.
498 	 *
499 	 * lgrp_expand_proc_diff determines how much less the remote lgroup
500 	 * must be loaded before expanding to it.
501 	 *
502 	 * Optimize for memory bandwidth by spreading multi-threaded
503 	 * program to different lgroups.
504 	 */
505 	lgrp_expand_proc_thresh = lgrp_loadavg_max_effect - 1;
506 	lgrp_expand_proc_diff = lgrp_loadavg_max_effect / 2;
507 	lgrp_loadavg_tolerance = lgrp_loadavg_max_effect / 2;
508 	lgrp_mem_free_thresh = 1;	/* home lgrp must have some memory */
509 	lgrp_expand_proc_thresh = lgrp_loadavg_max_effect - 1;
510 	lgrp_mem_policy_root = LGRP_MEM_POLICY_NEXT;
511 	lgrp_load_thresh = 0;
512 
513 	mem_node_pfn_shift = PLATFORM_MC_SHIFT - MMU_PAGESHIFT;
514 }
515 
516 /*
517  * Return latency between "from" and "to" lgroups
518  *
519  * This latency number can only be used for relative comparison
520  * between lgroups on the running system, cannot be used across platforms,
521  * and may not reflect the actual latency.  It is platform and implementation
522  * specific, so platform gets to decide its value.  It would be nice if the
523  * number was at least proportional to make comparisons more meaningful though.
524  * NOTE: The numbers below are supposed to be load latencies for uncached
525  * memory divided by 10.
526  */
527 int
528 plat_lgrp_latency(void *from, void *to)
529 {
530 	/*
531 	 * Return remote latency when there are more than two lgroups
532 	 * (root and child) and getting latency between two different
533 	 * lgroups or root is involved
534 	 */
535 	if (lgrp_optimizations() && (from != to || from ==
536 	    (void *) LGRP_DEFAULT_HANDLE || to == (void *) LGRP_DEFAULT_HANDLE))
537 		return (17);
538 	else
539 		return (12);
540 }
541 
542 int
543 plat_pfn_to_mem_node(pfn_t pfn)
544 {
545 	ASSERT(max_mem_nodes > 1);
546 	return (pfn >> mem_node_pfn_shift);
547 }
548 
549 /*
550  * Assign memnode to lgroups
551  */
552 void
553 plat_fill_mc(pnode_t nodeid)
554 {
555 	int		portid;
556 
557 	/*
558 	 * Memory controller portid == global CPU id
559 	 */
560 	if ((prom_getprop(nodeid, "portid", (caddr_t)&portid) == -1) ||
561 	    (portid < 0))
562 		return;
563 
564 	if (portid < max_mem_nodes)
565 		plat_assign_lgrphand_to_mem_node((lgrp_handle_t)portid, portid);
566 }
567 
568 /* ARGSUSED */
569 void
570 plat_build_mem_nodes(u_longlong_t *list, size_t nelems)
571 {
572 	size_t	elem;
573 	pfn_t	basepfn;
574 	pgcnt_t	npgs;
575 
576 	/*
577 	 * Boot install lists are arranged <addr, len>, <addr, len>, ...
578 	 */
579 	for (elem = 0; elem < nelems; elem += 2) {
580 		basepfn = btop(list[elem]);
581 		npgs = btop(list[elem+1]);
582 		mem_node_add_slice(basepfn, basepfn + npgs - 1);
583 	}
584 }
585 
586 /*
587  * Common locking enter code
588  */
589 void
590 plat_setprop_enter(void)
591 {
592 	mutex_enter(&mi2cv_mutex);
593 }
594 
595 /*
596  * Common locking exit code
597  */
598 void
599 plat_setprop_exit(void)
600 {
601 	mutex_exit(&mi2cv_mutex);
602 }
603 
604 /*
605  * Called by mi2cv driver
606  */
607 void
608 plat_shared_i2c_enter(dev_info_t *i2cnexus_dip)
609 {
610 	if (i2cnexus_dip == shared_mi2cv_dip) {
611 		plat_setprop_enter();
612 	}
613 }
614 
615 /*
616  * Called by mi2cv driver
617  */
618 void
619 plat_shared_i2c_exit(dev_info_t *i2cnexus_dip)
620 {
621 	if (i2cnexus_dip == shared_mi2cv_dip) {
622 		plat_setprop_exit();
623 	}
624 }
625 /*
626  * Called by todm5823 driver
627  */
628 void
629 plat_rmc_comm_req(struct rmc_comm_msg *request)
630 {
631 	if (rmc_req_now)
632 		(void) rmc_req_now(request, 0);
633 }
634