powerpc/iseries: Fix oops reading from /proc/iSeries/mf/*/cmdline
[linux-2.6-block.git] / arch / powerpc / platforms / iseries / mf.c
1 /*
2  * Copyright (C) 2001 Troy D. Armstrong  IBM Corporation
3  * Copyright (C) 2004-2005 Stephen Rothwell  IBM Corporation
4  *
5  * This modules exists as an interface between a Linux secondary partition
6  * running on an iSeries and the primary partition's Virtual Service
7  * Processor (VSP) object.  The VSP has final authority over powering on/off
8  * all partitions in the iSeries.  It also provides miscellaneous low-level
9  * machine facility type operations.
10  *
11  *
12  * This program is free software; you can redistribute it and/or modify
13  * it under the terms of the GNU General Public License as published by
14  * the Free Software Foundation; either version 2 of the License, or
15  * (at your option) any later version.
16  *
17  * This program is distributed in the hope that it will be useful,
18  * but WITHOUT ANY WARRANTY; without even the implied warranty of
19  * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
20  * GNU General Public License for more details.
21  *
22  * You should have received a copy of the GNU General Public License
23  * along with this program; if not, write to the Free Software
24  * Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA  02111-1307 USA
25  */
26
27 #include <linux/types.h>
28 #include <linux/errno.h>
29 #include <linux/kernel.h>
30 #include <linux/init.h>
31 #include <linux/completion.h>
32 #include <linux/delay.h>
33 #include <linux/dma-mapping.h>
34 #include <linux/bcd.h>
35 #include <linux/rtc.h>
36
37 #include <asm/time.h>
38 #include <asm/uaccess.h>
39 #include <asm/paca.h>
40 #include <asm/abs_addr.h>
41 #include <asm/firmware.h>
42 #include <asm/iseries/mf.h>
43 #include <asm/iseries/hv_lp_config.h>
44 #include <asm/iseries/hv_lp_event.h>
45 #include <asm/iseries/it_lp_queue.h>
46
47 #include "setup.h"
48
49 static int mf_initialized;
50
51 /*
52  * This is the structure layout for the Machine Facilites LPAR event
53  * flows.
54  */
55 struct vsp_cmd_data {
56         u64 token;
57         u16 cmd;
58         HvLpIndex lp_index;
59         u8 result_code;
60         u32 reserved;
61         union {
62                 u64 state;      /* GetStateOut */
63                 u64 ipl_type;   /* GetIplTypeOut, Function02SelectIplTypeIn */
64                 u64 ipl_mode;   /* GetIplModeOut, Function02SelectIplModeIn */
65                 u64 page[4];    /* GetSrcHistoryIn */
66                 u64 flag;       /* GetAutoIplWhenPrimaryIplsOut,
67                                    SetAutoIplWhenPrimaryIplsIn,
68                                    WhiteButtonPowerOffIn,
69                                    Function08FastPowerOffIn,
70                                    IsSpcnRackPowerIncompleteOut */
71                 struct {
72                         u64 token;
73                         u64 address_type;
74                         u64 side;
75                         u32 length;
76                         u32 offset;
77                 } kern;         /* SetKernelImageIn, GetKernelImageIn,
78                                    SetKernelCmdLineIn, GetKernelCmdLineIn */
79                 u32 length_out; /* GetKernelImageOut, GetKernelCmdLineOut */
80                 u8 reserved[80];
81         } sub_data;
82 };
83
84 struct vsp_rsp_data {
85         struct completion com;
86         struct vsp_cmd_data *response;
87 };
88
89 struct alloc_data {
90         u16 size;
91         u16 type;
92         u32 count;
93         u16 reserved1;
94         u8 reserved2;
95         HvLpIndex target_lp;
96 };
97
98 struct ce_msg_data;
99
100 typedef void (*ce_msg_comp_hdlr)(void *token, struct ce_msg_data *vsp_cmd_rsp);
101
102 struct ce_msg_comp_data {
103         ce_msg_comp_hdlr handler;
104         void *token;
105 };
106
107 struct ce_msg_data {
108         u8 ce_msg[12];
109         char reserved[4];
110         struct ce_msg_comp_data *completion;
111 };
112
113 struct io_mf_lp_event {
114         struct HvLpEvent hp_lp_event;
115         u16 subtype_result_code;
116         u16 reserved1;
117         u32 reserved2;
118         union {
119                 struct alloc_data alloc;
120                 struct ce_msg_data ce_msg;
121                 struct vsp_cmd_data vsp_cmd;
122         } data;
123 };
124
125 #define subtype_data(a, b, c, d)        \
126                 (((a) << 24) + ((b) << 16) + ((c) << 8) + (d))
127
128 /*
129  * All outgoing event traffic is kept on a FIFO queue.  The first
130  * pointer points to the one that is outstanding, and all new
131  * requests get stuck on the end.  Also, we keep a certain number of
132  * preallocated pending events so that we can operate very early in
133  * the boot up sequence (before kmalloc is ready).
134  */
135 struct pending_event {
136         struct pending_event *next;
137         struct io_mf_lp_event event;
138         MFCompleteHandler hdlr;
139         char dma_data[72];
140         unsigned dma_data_length;
141         unsigned remote_address;
142 };
143 static spinlock_t pending_event_spinlock;
144 static struct pending_event *pending_event_head;
145 static struct pending_event *pending_event_tail;
146 static struct pending_event *pending_event_avail;
147 #define PENDING_EVENT_PREALLOC_LEN 16
148 static struct pending_event pending_event_prealloc[PENDING_EVENT_PREALLOC_LEN];
149
150 /*
151  * Put a pending event onto the available queue, so it can get reused.
152  * Attention! You must have the pending_event_spinlock before calling!
153  */
154 static void free_pending_event(struct pending_event *ev)
155 {
156         if (ev != NULL) {
157                 ev->next = pending_event_avail;
158                 pending_event_avail = ev;
159         }
160 }
161
162 /*
163  * Enqueue the outbound event onto the stack.  If the queue was
164  * empty to begin with, we must also issue it via the Hypervisor
165  * interface.  There is a section of code below that will touch
166  * the first stack pointer without the protection of the pending_event_spinlock.
167  * This is OK, because we know that nobody else will be modifying
168  * the first pointer when we do this.
169  */
170 static int signal_event(struct pending_event *ev)
171 {
172         int rc = 0;
173         unsigned long flags;
174         int go = 1;
175         struct pending_event *ev1;
176         HvLpEvent_Rc hv_rc;
177
178         /* enqueue the event */
179         if (ev != NULL) {
180                 ev->next = NULL;
181                 spin_lock_irqsave(&pending_event_spinlock, flags);
182                 if (pending_event_head == NULL)
183                         pending_event_head = ev;
184                 else {
185                         go = 0;
186                         pending_event_tail->next = ev;
187                 }
188                 pending_event_tail = ev;
189                 spin_unlock_irqrestore(&pending_event_spinlock, flags);
190         }
191
192         /* send the event */
193         while (go) {
194                 go = 0;
195
196                 /* any DMA data to send beforehand? */
197                 if (pending_event_head->dma_data_length > 0)
198                         HvCallEvent_dmaToSp(pending_event_head->dma_data,
199                                         pending_event_head->remote_address,
200                                         pending_event_head->dma_data_length,
201                                         HvLpDma_Direction_LocalToRemote);
202
203                 hv_rc = HvCallEvent_signalLpEvent(
204                                 &pending_event_head->event.hp_lp_event);
205                 if (hv_rc != HvLpEvent_Rc_Good) {
206                         printk(KERN_ERR "mf.c: HvCallEvent_signalLpEvent() "
207                                         "failed with %d\n", (int)hv_rc);
208
209                         spin_lock_irqsave(&pending_event_spinlock, flags);
210                         ev1 = pending_event_head;
211                         pending_event_head = pending_event_head->next;
212                         if (pending_event_head != NULL)
213                                 go = 1;
214                         spin_unlock_irqrestore(&pending_event_spinlock, flags);
215
216                         if (ev1 == ev)
217                                 rc = -EIO;
218                         else if (ev1->hdlr != NULL)
219                                 (*ev1->hdlr)((void *)ev1->event.hp_lp_event.xCorrelationToken, -EIO);
220
221                         spin_lock_irqsave(&pending_event_spinlock, flags);
222                         free_pending_event(ev1);
223                         spin_unlock_irqrestore(&pending_event_spinlock, flags);
224                 }
225         }
226
227         return rc;
228 }
229
230 /*
231  * Allocate a new pending_event structure, and initialize it.
232  */
233 static struct pending_event *new_pending_event(void)
234 {
235         struct pending_event *ev = NULL;
236         HvLpIndex primary_lp = HvLpConfig_getPrimaryLpIndex();
237         unsigned long flags;
238         struct HvLpEvent *hev;
239
240         spin_lock_irqsave(&pending_event_spinlock, flags);
241         if (pending_event_avail != NULL) {
242                 ev = pending_event_avail;
243                 pending_event_avail = pending_event_avail->next;
244         }
245         spin_unlock_irqrestore(&pending_event_spinlock, flags);
246         if (ev == NULL) {
247                 ev = kmalloc(sizeof(struct pending_event), GFP_ATOMIC);
248                 if (ev == NULL) {
249                         printk(KERN_ERR "mf.c: unable to kmalloc %ld bytes\n",
250                                         sizeof(struct pending_event));
251                         return NULL;
252                 }
253         }
254         memset(ev, 0, sizeof(struct pending_event));
255         hev = &ev->event.hp_lp_event;
256         hev->flags = HV_LP_EVENT_VALID | HV_LP_EVENT_DO_ACK | HV_LP_EVENT_INT;
257         hev->xType = HvLpEvent_Type_MachineFac;
258         hev->xSourceLp = HvLpConfig_getLpIndex();
259         hev->xTargetLp = primary_lp;
260         hev->xSizeMinus1 = sizeof(ev->event) - 1;
261         hev->xRc = HvLpEvent_Rc_Good;
262         hev->xSourceInstanceId = HvCallEvent_getSourceLpInstanceId(primary_lp,
263                         HvLpEvent_Type_MachineFac);
264         hev->xTargetInstanceId = HvCallEvent_getTargetLpInstanceId(primary_lp,
265                         HvLpEvent_Type_MachineFac);
266
267         return ev;
268 }
269
270 static int __maybe_unused
271 signal_vsp_instruction(struct vsp_cmd_data *vsp_cmd)
272 {
273         struct pending_event *ev = new_pending_event();
274         int rc;
275         struct vsp_rsp_data response;
276
277         if (ev == NULL)
278                 return -ENOMEM;
279
280         init_completion(&response.com);
281         response.response = vsp_cmd;
282         ev->event.hp_lp_event.xSubtype = 6;
283         ev->event.hp_lp_event.x.xSubtypeData =
284                 subtype_data('M', 'F',  'V',  'I');
285         ev->event.data.vsp_cmd.token = (u64)&response;
286         ev->event.data.vsp_cmd.cmd = vsp_cmd->cmd;
287         ev->event.data.vsp_cmd.lp_index = HvLpConfig_getLpIndex();
288         ev->event.data.vsp_cmd.result_code = 0xFF;
289         ev->event.data.vsp_cmd.reserved = 0;
290         memcpy(&(ev->event.data.vsp_cmd.sub_data),
291                         &(vsp_cmd->sub_data), sizeof(vsp_cmd->sub_data));
292         mb();
293
294         rc = signal_event(ev);
295         if (rc == 0)
296                 wait_for_completion(&response.com);
297         return rc;
298 }
299
300
301 /*
302  * Send a 12-byte CE message to the primary partition VSP object
303  */
304 static int signal_ce_msg(char *ce_msg, struct ce_msg_comp_data *completion)
305 {
306         struct pending_event *ev = new_pending_event();
307
308         if (ev == NULL)
309                 return -ENOMEM;
310
311         ev->event.hp_lp_event.xSubtype = 0;
312         ev->event.hp_lp_event.x.xSubtypeData =
313                 subtype_data('M',  'F',  'C',  'E');
314         memcpy(ev->event.data.ce_msg.ce_msg, ce_msg, 12);
315         ev->event.data.ce_msg.completion = completion;
316         return signal_event(ev);
317 }
318
319 /*
320  * Send a 12-byte CE message (with no data) to the primary partition VSP object
321  */
322 static int signal_ce_msg_simple(u8 ce_op, struct ce_msg_comp_data *completion)
323 {
324         u8 ce_msg[12];
325
326         memset(ce_msg, 0, sizeof(ce_msg));
327         ce_msg[3] = ce_op;
328         return signal_ce_msg(ce_msg, completion);
329 }
330
331 /*
332  * Send a 12-byte CE message and DMA data to the primary partition VSP object
333  */
334 static int dma_and_signal_ce_msg(char *ce_msg,
335                 struct ce_msg_comp_data *completion, void *dma_data,
336                 unsigned dma_data_length, unsigned remote_address)
337 {
338         struct pending_event *ev = new_pending_event();
339
340         if (ev == NULL)
341                 return -ENOMEM;
342
343         ev->event.hp_lp_event.xSubtype = 0;
344         ev->event.hp_lp_event.x.xSubtypeData =
345                 subtype_data('M', 'F', 'C', 'E');
346         memcpy(ev->event.data.ce_msg.ce_msg, ce_msg, 12);
347         ev->event.data.ce_msg.completion = completion;
348         memcpy(ev->dma_data, dma_data, dma_data_length);
349         ev->dma_data_length = dma_data_length;
350         ev->remote_address = remote_address;
351         return signal_event(ev);
352 }
353
354 /*
355  * Initiate a nice (hopefully) shutdown of Linux.  We simply are
356  * going to try and send the init process a SIGINT signal.  If
357  * this fails (why?), we'll simply force it off in a not-so-nice
358  * manner.
359  */
360 static int shutdown(void)
361 {
362         int rc = kill_cad_pid(SIGINT, 1);
363
364         if (rc) {
365                 printk(KERN_ALERT "mf.c: SIGINT to init failed (%d), "
366                                 "hard shutdown commencing\n", rc);
367                 mf_power_off();
368         } else
369                 printk(KERN_INFO "mf.c: init has been successfully notified "
370                                 "to proceed with shutdown\n");
371         return rc;
372 }
373
374 /*
375  * The primary partition VSP object is sending us a new
376  * event flow.  Handle it...
377  */
378 static void handle_int(struct io_mf_lp_event *event)
379 {
380         struct ce_msg_data *ce_msg_data;
381         struct ce_msg_data *pce_msg_data;
382         unsigned long flags;
383         struct pending_event *pev;
384
385         /* ack the interrupt */
386         event->hp_lp_event.xRc = HvLpEvent_Rc_Good;
387         HvCallEvent_ackLpEvent(&event->hp_lp_event);
388
389         /* process interrupt */
390         switch (event->hp_lp_event.xSubtype) {
391         case 0: /* CE message */
392                 ce_msg_data = &event->data.ce_msg;
393                 switch (ce_msg_data->ce_msg[3]) {
394                 case 0x5B:      /* power control notification */
395                         if ((ce_msg_data->ce_msg[5] & 0x20) != 0) {
396                                 printk(KERN_INFO "mf.c: Commencing partition shutdown\n");
397                                 if (shutdown() == 0)
398                                         signal_ce_msg_simple(0xDB, NULL);
399                         }
400                         break;
401                 case 0xC0:      /* get time */
402                         spin_lock_irqsave(&pending_event_spinlock, flags);
403                         pev = pending_event_head;
404                         if (pev != NULL)
405                                 pending_event_head = pending_event_head->next;
406                         spin_unlock_irqrestore(&pending_event_spinlock, flags);
407                         if (pev == NULL)
408                                 break;
409                         pce_msg_data = &pev->event.data.ce_msg;
410                         if (pce_msg_data->ce_msg[3] != 0x40)
411                                 break;
412                         if (pce_msg_data->completion != NULL) {
413                                 ce_msg_comp_hdlr handler =
414                                         pce_msg_data->completion->handler;
415                                 void *token = pce_msg_data->completion->token;
416
417                                 if (handler != NULL)
418                                         (*handler)(token, ce_msg_data);
419                         }
420                         spin_lock_irqsave(&pending_event_spinlock, flags);
421                         free_pending_event(pev);
422                         spin_unlock_irqrestore(&pending_event_spinlock, flags);
423                         /* send next waiting event */
424                         if (pending_event_head != NULL)
425                                 signal_event(NULL);
426                         break;
427                 }
428                 break;
429         case 1: /* IT sys shutdown */
430                 printk(KERN_INFO "mf.c: Commencing system shutdown\n");
431                 shutdown();
432                 break;
433         }
434 }
435
436 /*
437  * The primary partition VSP object is acknowledging the receipt
438  * of a flow we sent to them.  If there are other flows queued
439  * up, we must send another one now...
440  */
441 static void handle_ack(struct io_mf_lp_event *event)
442 {
443         unsigned long flags;
444         struct pending_event *two = NULL;
445         unsigned long free_it = 0;
446         struct ce_msg_data *ce_msg_data;
447         struct ce_msg_data *pce_msg_data;
448         struct vsp_rsp_data *rsp;
449
450         /* handle current event */
451         if (pending_event_head == NULL) {
452                 printk(KERN_ERR "mf.c: stack empty for receiving ack\n");
453                 return;
454         }
455
456         switch (event->hp_lp_event.xSubtype) {
457         case 0:     /* CE msg */
458                 ce_msg_data = &event->data.ce_msg;
459                 if (ce_msg_data->ce_msg[3] != 0x40) {
460                         free_it = 1;
461                         break;
462                 }
463                 if (ce_msg_data->ce_msg[2] == 0)
464                         break;
465                 free_it = 1;
466                 pce_msg_data = &pending_event_head->event.data.ce_msg;
467                 if (pce_msg_data->completion != NULL) {
468                         ce_msg_comp_hdlr handler =
469                                 pce_msg_data->completion->handler;
470                         void *token = pce_msg_data->completion->token;
471
472                         if (handler != NULL)
473                                 (*handler)(token, ce_msg_data);
474                 }
475                 break;
476         case 4: /* allocate */
477         case 5: /* deallocate */
478                 if (pending_event_head->hdlr != NULL)
479                         (*pending_event_head->hdlr)((void *)event->hp_lp_event.xCorrelationToken, event->data.alloc.count);
480                 free_it = 1;
481                 break;
482         case 6:
483                 free_it = 1;
484                 rsp = (struct vsp_rsp_data *)event->data.vsp_cmd.token;
485                 if (rsp == NULL) {
486                         printk(KERN_ERR "mf.c: no rsp\n");
487                         break;
488                 }
489                 if (rsp->response != NULL)
490                         memcpy(rsp->response, &event->data.vsp_cmd,
491                                         sizeof(event->data.vsp_cmd));
492                 complete(&rsp->com);
493                 break;
494         }
495
496         /* remove from queue */
497         spin_lock_irqsave(&pending_event_spinlock, flags);
498         if ((pending_event_head != NULL) && (free_it == 1)) {
499                 struct pending_event *oldHead = pending_event_head;
500
501                 pending_event_head = pending_event_head->next;
502                 two = pending_event_head;
503                 free_pending_event(oldHead);
504         }
505         spin_unlock_irqrestore(&pending_event_spinlock, flags);
506
507         /* send next waiting event */
508         if (two != NULL)
509                 signal_event(NULL);
510 }
511
512 /*
513  * This is the generic event handler we are registering with
514  * the Hypervisor.  Ensure the flows are for us, and then
515  * parse it enough to know if it is an interrupt or an
516  * acknowledge.
517  */
518 static void hv_handler(struct HvLpEvent *event)
519 {
520         if ((event != NULL) && (event->xType == HvLpEvent_Type_MachineFac)) {
521                 if (hvlpevent_is_ack(event))
522                         handle_ack((struct io_mf_lp_event *)event);
523                 else
524                         handle_int((struct io_mf_lp_event *)event);
525         } else
526                 printk(KERN_ERR "mf.c: alien event received\n");
527 }
528
529 /*
530  * Global kernel interface to allocate and seed events into the
531  * Hypervisor.
532  */
533 void mf_allocate_lp_events(HvLpIndex target_lp, HvLpEvent_Type type,
534                 unsigned size, unsigned count, MFCompleteHandler hdlr,
535                 void *user_token)
536 {
537         struct pending_event *ev = new_pending_event();
538         int rc;
539
540         if (ev == NULL) {
541                 rc = -ENOMEM;
542         } else {
543                 ev->event.hp_lp_event.xSubtype = 4;
544                 ev->event.hp_lp_event.xCorrelationToken = (u64)user_token;
545                 ev->event.hp_lp_event.x.xSubtypeData =
546                         subtype_data('M', 'F', 'M', 'A');
547                 ev->event.data.alloc.target_lp = target_lp;
548                 ev->event.data.alloc.type = type;
549                 ev->event.data.alloc.size = size;
550                 ev->event.data.alloc.count = count;
551                 ev->hdlr = hdlr;
552                 rc = signal_event(ev);
553         }
554         if ((rc != 0) && (hdlr != NULL))
555                 (*hdlr)(user_token, rc);
556 }
557 EXPORT_SYMBOL(mf_allocate_lp_events);
558
559 /*
560  * Global kernel interface to unseed and deallocate events already in
561  * Hypervisor.
562  */
563 void mf_deallocate_lp_events(HvLpIndex target_lp, HvLpEvent_Type type,
564                 unsigned count, MFCompleteHandler hdlr, void *user_token)
565 {
566         struct pending_event *ev = new_pending_event();
567         int rc;
568
569         if (ev == NULL)
570                 rc = -ENOMEM;
571         else {
572                 ev->event.hp_lp_event.xSubtype = 5;
573                 ev->event.hp_lp_event.xCorrelationToken = (u64)user_token;
574                 ev->event.hp_lp_event.x.xSubtypeData =
575                         subtype_data('M', 'F', 'M', 'D');
576                 ev->event.data.alloc.target_lp = target_lp;
577                 ev->event.data.alloc.type = type;
578                 ev->event.data.alloc.count = count;
579                 ev->hdlr = hdlr;
580                 rc = signal_event(ev);
581         }
582         if ((rc != 0) && (hdlr != NULL))
583                 (*hdlr)(user_token, rc);
584 }
585 EXPORT_SYMBOL(mf_deallocate_lp_events);
586
587 /*
588  * Global kernel interface to tell the VSP object in the primary
589  * partition to power this partition off.
590  */
591 void mf_power_off(void)
592 {
593         printk(KERN_INFO "mf.c: Down it goes...\n");
594         signal_ce_msg_simple(0x4d, NULL);
595         for (;;)
596                 ;
597 }
598
599 /*
600  * Global kernel interface to tell the VSP object in the primary
601  * partition to reboot this partition.
602  */
603 void mf_reboot(char *cmd)
604 {
605         printk(KERN_INFO "mf.c: Preparing to bounce...\n");
606         signal_ce_msg_simple(0x4e, NULL);
607         for (;;)
608                 ;
609 }
610
611 /*
612  * Display a single word SRC onto the VSP control panel.
613  */
614 void mf_display_src(u32 word)
615 {
616         u8 ce[12];
617
618         memset(ce, 0, sizeof(ce));
619         ce[3] = 0x4a;
620         ce[7] = 0x01;
621         ce[8] = word >> 24;
622         ce[9] = word >> 16;
623         ce[10] = word >> 8;
624         ce[11] = word;
625         signal_ce_msg(ce, NULL);
626 }
627
628 /*
629  * Display a single word SRC of the form "PROGXXXX" on the VSP control panel.
630  */
631 static __init void mf_display_progress_src(u16 value)
632 {
633         u8 ce[12];
634         u8 src[72];
635
636         memcpy(ce, "\x00\x00\x04\x4A\x00\x00\x00\x48\x00\x00\x00\x00", 12);
637         memcpy(src, "\x01\x00\x00\x01\x00\x00\x00\x00\x00\x00\x00\x00"
638                 "\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00"
639                 "\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00"
640                 "\x00\x00\x00\x00PROGxxxx                        ",
641                 72);
642         src[6] = value >> 8;
643         src[7] = value & 255;
644         src[44] = "0123456789ABCDEF"[(value >> 12) & 15];
645         src[45] = "0123456789ABCDEF"[(value >> 8) & 15];
646         src[46] = "0123456789ABCDEF"[(value >> 4) & 15];
647         src[47] = "0123456789ABCDEF"[value & 15];
648         dma_and_signal_ce_msg(ce, NULL, src, sizeof(src), 9 * 64 * 1024);
649 }
650
651 /*
652  * Clear the VSP control panel.  Used to "erase" an SRC that was
653  * previously displayed.
654  */
655 static void mf_clear_src(void)
656 {
657         signal_ce_msg_simple(0x4b, NULL);
658 }
659
660 void __init mf_display_progress(u16 value)
661 {
662         if (!mf_initialized)
663                 return;
664
665         if (0xFFFF == value)
666                 mf_clear_src();
667         else
668                 mf_display_progress_src(value);
669 }
670
671 /*
672  * Initialization code here.
673  */
674 void __init mf_init(void)
675 {
676         int i;
677
678         spin_lock_init(&pending_event_spinlock);
679
680         for (i = 0; i < PENDING_EVENT_PREALLOC_LEN; i++)
681                 free_pending_event(&pending_event_prealloc[i]);
682
683         HvLpEvent_registerHandler(HvLpEvent_Type_MachineFac, &hv_handler);
684
685         /* virtual continue ack */
686         signal_ce_msg_simple(0x57, NULL);
687
688         mf_initialized = 1;
689         mb();
690
691         printk(KERN_NOTICE "mf.c: iSeries Linux LPAR Machine Facilities "
692                         "initialized\n");
693 }
694
695 struct rtc_time_data {
696         struct completion com;
697         struct ce_msg_data ce_msg;
698         int rc;
699 };
700
701 static void get_rtc_time_complete(void *token, struct ce_msg_data *ce_msg)
702 {
703         struct rtc_time_data *rtc = token;
704
705         memcpy(&rtc->ce_msg, ce_msg, sizeof(rtc->ce_msg));
706         rtc->rc = 0;
707         complete(&rtc->com);
708 }
709
710 static int mf_set_rtc(struct rtc_time *tm)
711 {
712         char ce_time[12];
713         u8 day, mon, hour, min, sec, y1, y2;
714         unsigned year;
715
716         year = 1900 + tm->tm_year;
717         y1 = year / 100;
718         y2 = year % 100;
719
720         sec = tm->tm_sec;
721         min = tm->tm_min;
722         hour = tm->tm_hour;
723         day = tm->tm_mday;
724         mon = tm->tm_mon + 1;
725
726         sec = bin2bcd(sec);
727         min = bin2bcd(min);
728         hour = bin2bcd(hour);
729         mon = bin2bcd(mon);
730         day = bin2bcd(day);
731         y1 = bin2bcd(y1);
732         y2 = bin2bcd(y2);
733
734         memset(ce_time, 0, sizeof(ce_time));
735         ce_time[3] = 0x41;
736         ce_time[4] = y1;
737         ce_time[5] = y2;
738         ce_time[6] = sec;
739         ce_time[7] = min;
740         ce_time[8] = hour;
741         ce_time[10] = day;
742         ce_time[11] = mon;
743
744         return signal_ce_msg(ce_time, NULL);
745 }
746
747 static int rtc_set_tm(int rc, u8 *ce_msg, struct rtc_time *tm)
748 {
749         tm->tm_wday = 0;
750         tm->tm_yday = 0;
751         tm->tm_isdst = 0;
752         if (rc) {
753                 tm->tm_sec = 0;
754                 tm->tm_min = 0;
755                 tm->tm_hour = 0;
756                 tm->tm_mday = 15;
757                 tm->tm_mon = 5;
758                 tm->tm_year = 52;
759                 return rc;
760         }
761
762         if ((ce_msg[2] == 0xa9) ||
763             (ce_msg[2] == 0xaf)) {
764                 /* TOD clock is not set */
765                 tm->tm_sec = 1;
766                 tm->tm_min = 1;
767                 tm->tm_hour = 1;
768                 tm->tm_mday = 10;
769                 tm->tm_mon = 8;
770                 tm->tm_year = 71;
771                 mf_set_rtc(tm);
772         }
773         {
774                 u8 year = ce_msg[5];
775                 u8 sec = ce_msg[6];
776                 u8 min = ce_msg[7];
777                 u8 hour = ce_msg[8];
778                 u8 day = ce_msg[10];
779                 u8 mon = ce_msg[11];
780
781                 sec = bcd2bin(sec);
782                 min = bcd2bin(min);
783                 hour = bcd2bin(hour);
784                 day = bcd2bin(day);
785                 mon = bcd2bin(mon);
786                 year = bcd2bin(year);
787
788                 if (year <= 69)
789                         year += 100;
790
791                 tm->tm_sec = sec;
792                 tm->tm_min = min;
793                 tm->tm_hour = hour;
794                 tm->tm_mday = day;
795                 tm->tm_mon = mon;
796                 tm->tm_year = year;
797         }
798
799         return 0;
800 }
801
802 static int mf_get_rtc(struct rtc_time *tm)
803 {
804         struct ce_msg_comp_data ce_complete;
805         struct rtc_time_data rtc_data;
806         int rc;
807
808         memset(&ce_complete, 0, sizeof(ce_complete));
809         memset(&rtc_data, 0, sizeof(rtc_data));
810         init_completion(&rtc_data.com);
811         ce_complete.handler = &get_rtc_time_complete;
812         ce_complete.token = &rtc_data;
813         rc = signal_ce_msg_simple(0x40, &ce_complete);
814         if (rc)
815                 return rc;
816         wait_for_completion(&rtc_data.com);
817         return rtc_set_tm(rtc_data.rc, rtc_data.ce_msg.ce_msg, tm);
818 }
819
820 struct boot_rtc_time_data {
821         int busy;
822         struct ce_msg_data ce_msg;
823         int rc;
824 };
825
826 static void get_boot_rtc_time_complete(void *token, struct ce_msg_data *ce_msg)
827 {
828         struct boot_rtc_time_data *rtc = token;
829
830         memcpy(&rtc->ce_msg, ce_msg, sizeof(rtc->ce_msg));
831         rtc->rc = 0;
832         rtc->busy = 0;
833 }
834
835 static int mf_get_boot_rtc(struct rtc_time *tm)
836 {
837         struct ce_msg_comp_data ce_complete;
838         struct boot_rtc_time_data rtc_data;
839         int rc;
840
841         memset(&ce_complete, 0, sizeof(ce_complete));
842         memset(&rtc_data, 0, sizeof(rtc_data));
843         rtc_data.busy = 1;
844         ce_complete.handler = &get_boot_rtc_time_complete;
845         ce_complete.token = &rtc_data;
846         rc = signal_ce_msg_simple(0x40, &ce_complete);
847         if (rc)
848                 return rc;
849         /* We need to poll here as we are not yet taking interrupts */
850         while (rtc_data.busy) {
851                 if (hvlpevent_is_pending())
852                         process_hvlpevents();
853         }
854         return rtc_set_tm(rtc_data.rc, rtc_data.ce_msg.ce_msg, tm);
855 }
856
857 #ifdef CONFIG_PROC_FS
858
859 static int proc_mf_dump_cmdline(char *page, char **start, off_t off,
860                 int count, int *eof, void *data)
861 {
862         int len;
863         char *p;
864         struct vsp_cmd_data vsp_cmd;
865         int rc;
866         dma_addr_t dma_addr;
867
868         /* The HV appears to return no more than 256 bytes of command line */
869         if (off >= 256)
870                 return 0;
871         if ((off + count) > 256)
872                 count = 256 - off;
873
874         dma_addr = iseries_hv_map(page, off + count, DMA_FROM_DEVICE);
875         if (dma_addr == DMA_ERROR_CODE)
876                 return -ENOMEM;
877         memset(page, 0, off + count);
878         memset(&vsp_cmd, 0, sizeof(vsp_cmd));
879         vsp_cmd.cmd = 33;
880         vsp_cmd.sub_data.kern.token = dma_addr;
881         vsp_cmd.sub_data.kern.address_type = HvLpDma_AddressType_TceIndex;
882         vsp_cmd.sub_data.kern.side = (u64)data;
883         vsp_cmd.sub_data.kern.length = off + count;
884         mb();
885         rc = signal_vsp_instruction(&vsp_cmd);
886         iseries_hv_unmap(dma_addr, off + count, DMA_FROM_DEVICE);
887         if (rc)
888                 return rc;
889         if (vsp_cmd.result_code != 0)
890                 return -ENOMEM;
891         p = page;
892         len = 0;
893         while (len < (off + count)) {
894                 if ((*p == '\0') || (*p == '\n')) {
895                         if (*p == '\0')
896                                 *p = '\n';
897                         p++;
898                         len++;
899                         *eof = 1;
900                         break;
901                 }
902                 p++;
903                 len++;
904         }
905
906         if (len < off) {
907                 *eof = 1;
908                 len = 0;
909         }
910         return len;
911 }
912
913 #if 0
914 static int mf_getVmlinuxChunk(char *buffer, int *size, int offset, u64 side)
915 {
916         struct vsp_cmd_data vsp_cmd;
917         int rc;
918         int len = *size;
919         dma_addr_t dma_addr;
920
921         dma_addr = iseries_hv_map(buffer, len, DMA_FROM_DEVICE);
922         memset(buffer, 0, len);
923         memset(&vsp_cmd, 0, sizeof(vsp_cmd));
924         vsp_cmd.cmd = 32;
925         vsp_cmd.sub_data.kern.token = dma_addr;
926         vsp_cmd.sub_data.kern.address_type = HvLpDma_AddressType_TceIndex;
927         vsp_cmd.sub_data.kern.side = side;
928         vsp_cmd.sub_data.kern.offset = offset;
929         vsp_cmd.sub_data.kern.length = len;
930         mb();
931         rc = signal_vsp_instruction(&vsp_cmd);
932         if (rc == 0) {
933                 if (vsp_cmd.result_code == 0)
934                         *size = vsp_cmd.sub_data.length_out;
935                 else
936                         rc = -ENOMEM;
937         }
938
939         iseries_hv_unmap(dma_addr, len, DMA_FROM_DEVICE);
940
941         return rc;
942 }
943
944 static int proc_mf_dump_vmlinux(char *page, char **start, off_t off,
945                 int count, int *eof, void *data)
946 {
947         int sizeToGet = count;
948
949         if (!capable(CAP_SYS_ADMIN))
950                 return -EACCES;
951
952         if (mf_getVmlinuxChunk(page, &sizeToGet, off, (u64)data) == 0) {
953                 if (sizeToGet != 0) {
954                         *start = page + off;
955                         return sizeToGet;
956                 }
957                 *eof = 1;
958                 return 0;
959         }
960         *eof = 1;
961         return 0;
962 }
963 #endif
964
965 static int proc_mf_dump_side(char *page, char **start, off_t off,
966                 int count, int *eof, void *data)
967 {
968         int len;
969         char mf_current_side = ' ';
970         struct vsp_cmd_data vsp_cmd;
971
972         memset(&vsp_cmd, 0, sizeof(vsp_cmd));
973         vsp_cmd.cmd = 2;
974         vsp_cmd.sub_data.ipl_type = 0;
975         mb();
976
977         if (signal_vsp_instruction(&vsp_cmd) == 0) {
978                 if (vsp_cmd.result_code == 0) {
979                         switch (vsp_cmd.sub_data.ipl_type) {
980                         case 0: mf_current_side = 'A';
981                                 break;
982                         case 1: mf_current_side = 'B';
983                                 break;
984                         case 2: mf_current_side = 'C';
985                                 break;
986                         default:        mf_current_side = 'D';
987                                 break;
988                         }
989                 }
990         }
991
992         len = sprintf(page, "%c\n", mf_current_side);
993
994         if (len <= (off + count))
995                 *eof = 1;
996         *start = page + off;
997         len -= off;
998         if (len > count)
999                 len = count;
1000         if (len < 0)
1001                 len = 0;
1002         return len;
1003 }
1004
1005 static int proc_mf_change_side(struct file *file, const char __user *buffer,
1006                 unsigned long count, void *data)
1007 {
1008         char side;
1009         u64 newSide;
1010         struct vsp_cmd_data vsp_cmd;
1011
1012         if (!capable(CAP_SYS_ADMIN))
1013                 return -EACCES;
1014
1015         if (count == 0)
1016                 return 0;
1017
1018         if (get_user(side, buffer))
1019                 return -EFAULT;
1020
1021         switch (side) {
1022         case 'A':       newSide = 0;
1023                         break;
1024         case 'B':       newSide = 1;
1025                         break;
1026         case 'C':       newSide = 2;
1027                         break;
1028         case 'D':       newSide = 3;
1029                         break;
1030         default:
1031                 printk(KERN_ERR "mf_proc.c: proc_mf_change_side: invalid side\n");
1032                 return -EINVAL;
1033         }
1034
1035         memset(&vsp_cmd, 0, sizeof(vsp_cmd));
1036         vsp_cmd.sub_data.ipl_type = newSide;
1037         vsp_cmd.cmd = 10;
1038
1039         (void)signal_vsp_instruction(&vsp_cmd);
1040
1041         return count;
1042 }
1043
1044 #if 0
1045 static void mf_getSrcHistory(char *buffer, int size)
1046 {
1047         struct IplTypeReturnStuff return_stuff;
1048         struct pending_event *ev = new_pending_event();
1049         int rc = 0;
1050         char *pages[4];
1051
1052         pages[0] = kmalloc(4096, GFP_ATOMIC);
1053         pages[1] = kmalloc(4096, GFP_ATOMIC);
1054         pages[2] = kmalloc(4096, GFP_ATOMIC);
1055         pages[3] = kmalloc(4096, GFP_ATOMIC);
1056         if ((ev == NULL) || (pages[0] == NULL) || (pages[1] == NULL)
1057                          || (pages[2] == NULL) || (pages[3] == NULL))
1058                 return -ENOMEM;
1059
1060         return_stuff.xType = 0;
1061         return_stuff.xRc = 0;
1062         return_stuff.xDone = 0;
1063         ev->event.hp_lp_event.xSubtype = 6;
1064         ev->event.hp_lp_event.x.xSubtypeData =
1065                 subtype_data('M', 'F', 'V', 'I');
1066         ev->event.data.vsp_cmd.xEvent = &return_stuff;
1067         ev->event.data.vsp_cmd.cmd = 4;
1068         ev->event.data.vsp_cmd.lp_index = HvLpConfig_getLpIndex();
1069         ev->event.data.vsp_cmd.result_code = 0xFF;
1070         ev->event.data.vsp_cmd.reserved = 0;
1071         ev->event.data.vsp_cmd.sub_data.page[0] = iseries_hv_addr(pages[0]);
1072         ev->event.data.vsp_cmd.sub_data.page[1] = iseries_hv_addr(pages[1]);
1073         ev->event.data.vsp_cmd.sub_data.page[2] = iseries_hv_addr(pages[2]);
1074         ev->event.data.vsp_cmd.sub_data.page[3] = iseries_hv_addr(pages[3]);
1075         mb();
1076         if (signal_event(ev) != 0)
1077                 return;
1078
1079         while (return_stuff.xDone != 1)
1080                 udelay(10);
1081         if (return_stuff.xRc == 0)
1082                 memcpy(buffer, pages[0], size);
1083         kfree(pages[0]);
1084         kfree(pages[1]);
1085         kfree(pages[2]);
1086         kfree(pages[3]);
1087 }
1088 #endif
1089
1090 static int proc_mf_dump_src(char *page, char **start, off_t off,
1091                 int count, int *eof, void *data)
1092 {
1093 #if 0
1094         int len;
1095
1096         mf_getSrcHistory(page, count);
1097         len = count;
1098         len -= off;
1099         if (len < count) {
1100                 *eof = 1;
1101                 if (len <= 0)
1102                         return 0;
1103         } else
1104                 len = count;
1105         *start = page + off;
1106         return len;
1107 #else
1108         return 0;
1109 #endif
1110 }
1111
1112 static int proc_mf_change_src(struct file *file, const char __user *buffer,
1113                 unsigned long count, void *data)
1114 {
1115         char stkbuf[10];
1116
1117         if (!capable(CAP_SYS_ADMIN))
1118                 return -EACCES;
1119
1120         if ((count < 4) && (count != 1)) {
1121                 printk(KERN_ERR "mf_proc: invalid src\n");
1122                 return -EINVAL;
1123         }
1124
1125         if (count > (sizeof(stkbuf) - 1))
1126                 count = sizeof(stkbuf) - 1;
1127         if (copy_from_user(stkbuf, buffer, count))
1128                 return -EFAULT;
1129
1130         if ((count == 1) && (*stkbuf == '\0'))
1131                 mf_clear_src();
1132         else
1133                 mf_display_src(*(u32 *)stkbuf);
1134
1135         return count;
1136 }
1137
1138 static int proc_mf_change_cmdline(struct file *file, const char __user *buffer,
1139                 unsigned long count, void *data)
1140 {
1141         struct vsp_cmd_data vsp_cmd;
1142         dma_addr_t dma_addr;
1143         char *page;
1144         int ret = -EACCES;
1145
1146         if (!capable(CAP_SYS_ADMIN))
1147                 goto out;
1148
1149         dma_addr = 0;
1150         page = iseries_hv_alloc(count, &dma_addr, GFP_ATOMIC);
1151         ret = -ENOMEM;
1152         if (page == NULL)
1153                 goto out;
1154
1155         ret = -EFAULT;
1156         if (copy_from_user(page, buffer, count))
1157                 goto out_free;
1158
1159         memset(&vsp_cmd, 0, sizeof(vsp_cmd));
1160         vsp_cmd.cmd = 31;
1161         vsp_cmd.sub_data.kern.token = dma_addr;
1162         vsp_cmd.sub_data.kern.address_type = HvLpDma_AddressType_TceIndex;
1163         vsp_cmd.sub_data.kern.side = (u64)data;
1164         vsp_cmd.sub_data.kern.length = count;
1165         mb();
1166         (void)signal_vsp_instruction(&vsp_cmd);
1167         ret = count;
1168
1169 out_free:
1170         iseries_hv_free(count, page, dma_addr);
1171 out:
1172         return ret;
1173 }
1174
1175 static ssize_t proc_mf_change_vmlinux(struct file *file,
1176                                       const char __user *buf,
1177                                       size_t count, loff_t *ppos)
1178 {
1179         struct proc_dir_entry *dp = PDE(file->f_path.dentry->d_inode);
1180         ssize_t rc;
1181         dma_addr_t dma_addr;
1182         char *page;
1183         struct vsp_cmd_data vsp_cmd;
1184
1185         rc = -EACCES;
1186         if (!capable(CAP_SYS_ADMIN))
1187                 goto out;
1188
1189         dma_addr = 0;
1190         page = iseries_hv_alloc(count, &dma_addr, GFP_ATOMIC);
1191         rc = -ENOMEM;
1192         if (page == NULL) {
1193                 printk(KERN_ERR "mf.c: couldn't allocate memory to set vmlinux chunk\n");
1194                 goto out;
1195         }
1196         rc = -EFAULT;
1197         if (copy_from_user(page, buf, count))
1198                 goto out_free;
1199
1200         memset(&vsp_cmd, 0, sizeof(vsp_cmd));
1201         vsp_cmd.cmd = 30;
1202         vsp_cmd.sub_data.kern.token = dma_addr;
1203         vsp_cmd.sub_data.kern.address_type = HvLpDma_AddressType_TceIndex;
1204         vsp_cmd.sub_data.kern.side = (u64)dp->data;
1205         vsp_cmd.sub_data.kern.offset = *ppos;
1206         vsp_cmd.sub_data.kern.length = count;
1207         mb();
1208         rc = signal_vsp_instruction(&vsp_cmd);
1209         if (rc)
1210                 goto out_free;
1211         rc = -ENOMEM;
1212         if (vsp_cmd.result_code != 0)
1213                 goto out_free;
1214
1215         *ppos += count;
1216         rc = count;
1217 out_free:
1218         iseries_hv_free(count, page, dma_addr);
1219 out:
1220         return rc;
1221 }
1222
1223 static const struct file_operations proc_vmlinux_operations = {
1224         .write          = proc_mf_change_vmlinux,
1225 };
1226
1227 static int __init mf_proc_init(void)
1228 {
1229         struct proc_dir_entry *mf_proc_root;
1230         struct proc_dir_entry *ent;
1231         struct proc_dir_entry *mf;
1232         char name[2];
1233         int i;
1234
1235         if (!firmware_has_feature(FW_FEATURE_ISERIES))
1236                 return 0;
1237
1238         mf_proc_root = proc_mkdir("iSeries/mf", NULL);
1239         if (!mf_proc_root)
1240                 return 1;
1241
1242         name[1] = '\0';
1243         for (i = 0; i < 4; i++) {
1244                 name[0] = 'A' + i;
1245                 mf = proc_mkdir(name, mf_proc_root);
1246                 if (!mf)
1247                         return 1;
1248
1249                 ent = create_proc_entry("cmdline", S_IFREG|S_IRUSR|S_IWUSR, mf);
1250                 if (!ent)
1251                         return 1;
1252                 ent->data = (void *)(long)i;
1253                 ent->read_proc = proc_mf_dump_cmdline;
1254                 ent->write_proc = proc_mf_change_cmdline;
1255
1256                 if (i == 3)     /* no vmlinux entry for 'D' */
1257                         continue;
1258
1259                 ent = proc_create_data("vmlinux", S_IFREG|S_IWUSR, mf,
1260                                        &proc_vmlinux_operations,
1261                                        (void *)(long)i);
1262                 if (!ent)
1263                         return 1;
1264         }
1265
1266         ent = create_proc_entry("side", S_IFREG|S_IRUSR|S_IWUSR, mf_proc_root);
1267         if (!ent)
1268                 return 1;
1269         ent->data = (void *)0;
1270         ent->read_proc = proc_mf_dump_side;
1271         ent->write_proc = proc_mf_change_side;
1272
1273         ent = create_proc_entry("src", S_IFREG|S_IRUSR|S_IWUSR, mf_proc_root);
1274         if (!ent)
1275                 return 1;
1276         ent->data = (void *)0;
1277         ent->read_proc = proc_mf_dump_src;
1278         ent->write_proc = proc_mf_change_src;
1279
1280         return 0;
1281 }
1282
1283 __initcall(mf_proc_init);
1284
1285 #endif /* CONFIG_PROC_FS */
1286
1287 /*
1288  * Get the RTC from the virtual service processor
1289  * This requires flowing LpEvents to the primary partition
1290  */
1291 void iSeries_get_rtc_time(struct rtc_time *rtc_tm)
1292 {
1293         mf_get_rtc(rtc_tm);
1294         rtc_tm->tm_mon--;
1295 }
1296
1297 /*
1298  * Set the RTC in the virtual service processor
1299  * This requires flowing LpEvents to the primary partition
1300  */
1301 int iSeries_set_rtc_time(struct rtc_time *tm)
1302 {
1303         mf_set_rtc(tm);
1304         return 0;
1305 }
1306
1307 unsigned long iSeries_get_boot_time(void)
1308 {
1309         struct rtc_time tm;
1310
1311         mf_get_boot_rtc(&tm);
1312         return mktime(tm.tm_year + 1900, tm.tm_mon, tm.tm_mday,
1313                       tm.tm_hour, tm.tm_min, tm.tm_sec);
1314 }