sgi-xp: cleanup naming of partition defines
[linux-2.6/linux-acpi-2.6/ibm-acpi-2.6.git] / drivers / misc / sgi-xp / xpc_partition.c
blobb5fb21641130d95c25b81e5f377e80d00ac238ee
1 /*
2 * This file is subject to the terms and conditions of the GNU General Public
3 * License. See the file "COPYING" in the main directory of this archive
4 * for more details.
6 * Copyright (c) 2004-2008 Silicon Graphics, Inc. All Rights Reserved.
7 */
9 /*
10 * Cross Partition Communication (XPC) partition support.
12 * This is the part of XPC that detects the presence/absence of
13 * other partitions. It provides a heartbeat and monitors the
14 * heartbeats of other partitions.
18 #include <linux/device.h>
19 #include <linux/hardirq.h>
20 #include "xpc.h"
22 /* XPC is exiting flag */
23 int xpc_exiting;
25 /* this partition's reserved page pointers */
26 struct xpc_rsvd_page *xpc_rsvd_page;
27 static unsigned long *xpc_part_nasids;
28 unsigned long *xpc_mach_nasids;
30 static int xpc_nasid_mask_nbytes; /* #of bytes in nasid mask */
31 int xpc_nasid_mask_nlongs; /* #of longs in nasid mask */
33 struct xpc_partition *xpc_partitions;
36 * Guarantee that the kmalloc'd memory is cacheline aligned.
38 void *
39 xpc_kmalloc_cacheline_aligned(size_t size, gfp_t flags, void **base)
41 /* see if kmalloc will give us cachline aligned memory by default */
42 *base = kmalloc(size, flags);
43 if (*base == NULL)
44 return NULL;
46 if ((u64)*base == L1_CACHE_ALIGN((u64)*base))
47 return *base;
49 kfree(*base);
51 /* nope, we'll have to do it ourselves */
52 *base = kmalloc(size + L1_CACHE_BYTES, flags);
53 if (*base == NULL)
54 return NULL;
56 return (void *)L1_CACHE_ALIGN((u64)*base);
60 * Given a nasid, get the physical address of the partition's reserved page
61 * for that nasid. This function returns 0 on any error.
63 static unsigned long
64 xpc_get_rsvd_page_pa(int nasid)
66 enum xp_retval ret;
67 u64 cookie = 0;
68 unsigned long rp_pa = nasid; /* seed with nasid */
69 size_t len = 0;
70 size_t buf_len = 0;
71 void *buf = buf;
72 void *buf_base = NULL;
74 while (1) {
76 ret = xpc_get_partition_rsvd_page_pa(buf, &cookie, &rp_pa,
77 &len);
79 dev_dbg(xpc_part, "SAL returned with ret=%d, cookie=0x%016lx, "
80 "address=0x%016lx, len=0x%016lx\n", ret,
81 (unsigned long)cookie, rp_pa, len);
83 if (ret != xpNeedMoreInfo)
84 break;
86 /* !!! L1_CACHE_ALIGN() is only a sn2-bte_copy requirement */
87 if (L1_CACHE_ALIGN(len) > buf_len) {
88 kfree(buf_base);
89 buf_len = L1_CACHE_ALIGN(len);
90 buf = xpc_kmalloc_cacheline_aligned(buf_len, GFP_KERNEL,
91 &buf_base);
92 if (buf_base == NULL) {
93 dev_err(xpc_part, "unable to kmalloc "
94 "len=0x%016lx\n", buf_len);
95 ret = xpNoMemory;
96 break;
100 ret = xp_remote_memcpy(xp_pa(buf), rp_pa, buf_len);
101 if (ret != xpSuccess) {
102 dev_dbg(xpc_part, "xp_remote_memcpy failed %d\n", ret);
103 break;
107 kfree(buf_base);
109 if (ret != xpSuccess)
110 rp_pa = 0;
112 dev_dbg(xpc_part, "reserved page at phys address 0x%016lx\n", rp_pa);
113 return rp_pa;
117 * Fill the partition reserved page with the information needed by
118 * other partitions to discover we are alive and establish initial
119 * communications.
121 struct xpc_rsvd_page *
122 xpc_setup_rsvd_page(void)
124 struct xpc_rsvd_page *rp;
125 unsigned long rp_pa;
126 unsigned long new_ts_jiffies;
128 /* get the local reserved page's address */
130 preempt_disable();
131 rp_pa = xpc_get_rsvd_page_pa(xp_cpu_to_nasid(smp_processor_id()));
132 preempt_enable();
133 if (rp_pa == 0) {
134 dev_err(xpc_part, "SAL failed to locate the reserved page\n");
135 return NULL;
137 rp = (struct xpc_rsvd_page *)__va(rp_pa);
139 if (rp->SAL_version < 3) {
140 /* SAL_versions < 3 had a SAL_partid defined as a u8 */
141 rp->SAL_partid &= 0xff;
143 BUG_ON(rp->SAL_partid != xp_partition_id);
145 if (rp->SAL_partid < 0 || rp->SAL_partid >= xp_max_npartitions) {
146 dev_err(xpc_part, "the reserved page's partid of %d is outside "
147 "supported range (< 0 || >= %d)\n", rp->SAL_partid,
148 xp_max_npartitions);
149 return NULL;
152 rp->version = XPC_RP_VERSION;
153 rp->max_npartitions = xp_max_npartitions;
155 /* establish the actual sizes of the nasid masks */
156 if (rp->SAL_version == 1) {
157 /* SAL_version 1 didn't set the nasids_size field */
158 rp->SAL_nasids_size = 128;
160 xpc_nasid_mask_nbytes = rp->SAL_nasids_size;
161 xpc_nasid_mask_nlongs = BITS_TO_LONGS(rp->SAL_nasids_size *
162 BITS_PER_BYTE);
164 /* setup the pointers to the various items in the reserved page */
165 xpc_part_nasids = XPC_RP_PART_NASIDS(rp);
166 xpc_mach_nasids = XPC_RP_MACH_NASIDS(rp);
168 if (xpc_rsvd_page_init(rp) != xpSuccess)
169 return NULL;
172 * Set timestamp of when reserved page was setup by XPC.
173 * This signifies to the remote partition that our reserved
174 * page is initialized.
176 new_ts_jiffies = jiffies;
177 if (new_ts_jiffies == 0 || new_ts_jiffies == rp->ts_jiffies)
178 new_ts_jiffies++;
179 rp->ts_jiffies = new_ts_jiffies;
181 return rp;
185 * Get a copy of a portion of the remote partition's rsvd page.
187 * remote_rp points to a buffer that is cacheline aligned for BTE copies and
188 * is large enough to contain a copy of their reserved page header and
189 * part_nasids mask.
191 enum xp_retval
192 xpc_get_remote_rp(int nasid, unsigned long *discovered_nasids,
193 struct xpc_rsvd_page *remote_rp, unsigned long *remote_rp_pa)
195 int l;
196 enum xp_retval ret;
198 /* get the reserved page's physical address */
200 *remote_rp_pa = xpc_get_rsvd_page_pa(nasid);
201 if (*remote_rp_pa == 0)
202 return xpNoRsvdPageAddr;
204 /* pull over the reserved page header and part_nasids mask */
205 ret = xp_remote_memcpy(xp_pa(remote_rp), *remote_rp_pa,
206 XPC_RP_HEADER_SIZE + xpc_nasid_mask_nbytes);
207 if (ret != xpSuccess)
208 return ret;
210 if (discovered_nasids != NULL) {
211 unsigned long *remote_part_nasids =
212 XPC_RP_PART_NASIDS(remote_rp);
214 for (l = 0; l < xpc_nasid_mask_nlongs; l++)
215 discovered_nasids[l] |= remote_part_nasids[l];
218 /* zero timestamp indicates the reserved page has not been setup */
219 if (remote_rp->ts_jiffies == 0)
220 return xpRsvdPageNotSet;
222 if (XPC_VERSION_MAJOR(remote_rp->version) !=
223 XPC_VERSION_MAJOR(XPC_RP_VERSION)) {
224 return xpBadVersion;
227 /* check that both remote and local partids are valid for each side */
228 if (remote_rp->SAL_partid < 0 ||
229 remote_rp->SAL_partid >= xp_max_npartitions ||
230 remote_rp->max_npartitions <= xp_partition_id) {
231 return xpInvalidPartid;
234 if (remote_rp->SAL_partid == xp_partition_id)
235 return xpLocalPartid;
237 return xpSuccess;
241 * See if the other side has responded to a partition deactivate request
242 * from us. Though we requested the remote partition to deactivate with regard
243 * to us, we really only need to wait for the other side to disengage from us.
246 xpc_partition_disengaged(struct xpc_partition *part)
248 short partid = XPC_PARTID(part);
249 int disengaged;
251 disengaged = !xpc_partition_engaged(partid);
252 if (part->disengage_timeout) {
253 if (!disengaged) {
254 if (time_is_after_jiffies(part->disengage_timeout)) {
255 /* timelimit hasn't been reached yet */
256 return 0;
260 * Other side hasn't responded to our deactivate
261 * request in a timely fashion, so assume it's dead.
264 dev_info(xpc_part, "deactivate request to remote "
265 "partition %d timed out\n", partid);
266 xpc_disengage_timedout = 1;
267 xpc_assume_partition_disengaged(partid);
268 disengaged = 1;
270 part->disengage_timeout = 0;
272 /* cancel the timer function, provided it's not us */
273 if (!in_interrupt())
274 del_singleshot_timer_sync(&part->disengage_timer);
276 DBUG_ON(part->act_state != XPC_P_AS_DEACTIVATING &&
277 part->act_state != XPC_P_AS_INACTIVE);
278 if (part->act_state != XPC_P_AS_INACTIVE)
279 xpc_wakeup_channel_mgr(part);
281 xpc_cancel_partition_deactivation_request(part);
283 return disengaged;
287 * Mark specified partition as active.
289 enum xp_retval
290 xpc_mark_partition_active(struct xpc_partition *part)
292 unsigned long irq_flags;
293 enum xp_retval ret;
295 dev_dbg(xpc_part, "setting partition %d to ACTIVE\n", XPC_PARTID(part));
297 spin_lock_irqsave(&part->act_lock, irq_flags);
298 if (part->act_state == XPC_P_AS_ACTIVATING) {
299 part->act_state = XPC_P_AS_ACTIVE;
300 ret = xpSuccess;
301 } else {
302 DBUG_ON(part->reason == xpSuccess);
303 ret = part->reason;
305 spin_unlock_irqrestore(&part->act_lock, irq_flags);
307 return ret;
311 * Start the process of deactivating the specified partition.
313 void
314 xpc_deactivate_partition(const int line, struct xpc_partition *part,
315 enum xp_retval reason)
317 unsigned long irq_flags;
319 spin_lock_irqsave(&part->act_lock, irq_flags);
321 if (part->act_state == XPC_P_AS_INACTIVE) {
322 XPC_SET_REASON(part, reason, line);
323 spin_unlock_irqrestore(&part->act_lock, irq_flags);
324 if (reason == xpReactivating) {
325 /* we interrupt ourselves to reactivate partition */
326 xpc_request_partition_reactivation(part);
328 return;
330 if (part->act_state == XPC_P_AS_DEACTIVATING) {
331 if ((part->reason == xpUnloading && reason != xpUnloading) ||
332 reason == xpReactivating) {
333 XPC_SET_REASON(part, reason, line);
335 spin_unlock_irqrestore(&part->act_lock, irq_flags);
336 return;
339 part->act_state = XPC_P_AS_DEACTIVATING;
340 XPC_SET_REASON(part, reason, line);
342 spin_unlock_irqrestore(&part->act_lock, irq_flags);
344 /* ask remote partition to deactivate with regard to us */
345 xpc_request_partition_deactivation(part);
347 /* set a timelimit on the disengage phase of the deactivation request */
348 part->disengage_timeout = jiffies + (xpc_disengage_timelimit * HZ);
349 part->disengage_timer.expires = part->disengage_timeout;
350 add_timer(&part->disengage_timer);
352 dev_dbg(xpc_part, "bringing partition %d down, reason = %d\n",
353 XPC_PARTID(part), reason);
355 xpc_partition_going_down(part, reason);
359 * Mark specified partition as inactive.
361 void
362 xpc_mark_partition_inactive(struct xpc_partition *part)
364 unsigned long irq_flags;
366 dev_dbg(xpc_part, "setting partition %d to INACTIVE\n",
367 XPC_PARTID(part));
369 spin_lock_irqsave(&part->act_lock, irq_flags);
370 part->act_state = XPC_P_AS_INACTIVE;
371 spin_unlock_irqrestore(&part->act_lock, irq_flags);
372 part->remote_rp_pa = 0;
376 * SAL has provided a partition and machine mask. The partition mask
377 * contains a bit for each even nasid in our partition. The machine
378 * mask contains a bit for each even nasid in the entire machine.
380 * Using those two bit arrays, we can determine which nasids are
381 * known in the machine. Each should also have a reserved page
382 * initialized if they are available for partitioning.
384 void
385 xpc_discovery(void)
387 void *remote_rp_base;
388 struct xpc_rsvd_page *remote_rp;
389 unsigned long remote_rp_pa;
390 int region;
391 int region_size;
392 int max_regions;
393 int nasid;
394 struct xpc_rsvd_page *rp;
395 unsigned long *discovered_nasids;
396 enum xp_retval ret;
398 remote_rp = xpc_kmalloc_cacheline_aligned(XPC_RP_HEADER_SIZE +
399 xpc_nasid_mask_nbytes,
400 GFP_KERNEL, &remote_rp_base);
401 if (remote_rp == NULL)
402 return;
404 discovered_nasids = kzalloc(sizeof(long) * xpc_nasid_mask_nlongs,
405 GFP_KERNEL);
406 if (discovered_nasids == NULL) {
407 kfree(remote_rp_base);
408 return;
411 rp = (struct xpc_rsvd_page *)xpc_rsvd_page;
414 * The term 'region' in this context refers to the minimum number of
415 * nodes that can comprise an access protection grouping. The access
416 * protection is in regards to memory, IOI and IPI.
418 max_regions = 64;
419 region_size = xp_region_size;
421 switch (region_size) {
422 case 128:
423 max_regions *= 2;
424 case 64:
425 max_regions *= 2;
426 case 32:
427 max_regions *= 2;
428 region_size = 16;
429 DBUG_ON(!is_shub2());
432 for (region = 0; region < max_regions; region++) {
434 if (xpc_exiting)
435 break;
437 dev_dbg(xpc_part, "searching region %d\n", region);
439 for (nasid = (region * region_size * 2);
440 nasid < ((region + 1) * region_size * 2); nasid += 2) {
442 if (xpc_exiting)
443 break;
445 dev_dbg(xpc_part, "checking nasid %d\n", nasid);
447 if (test_bit(nasid / 2, xpc_part_nasids)) {
448 dev_dbg(xpc_part, "PROM indicates Nasid %d is "
449 "part of the local partition; skipping "
450 "region\n", nasid);
451 break;
454 if (!(test_bit(nasid / 2, xpc_mach_nasids))) {
455 dev_dbg(xpc_part, "PROM indicates Nasid %d was "
456 "not on Numa-Link network at reset\n",
457 nasid);
458 continue;
461 if (test_bit(nasid / 2, discovered_nasids)) {
462 dev_dbg(xpc_part, "Nasid %d is part of a "
463 "partition which was previously "
464 "discovered\n", nasid);
465 continue;
468 /* pull over the rsvd page header & part_nasids mask */
470 ret = xpc_get_remote_rp(nasid, discovered_nasids,
471 remote_rp, &remote_rp_pa);
472 if (ret != xpSuccess) {
473 dev_dbg(xpc_part, "unable to get reserved page "
474 "from nasid %d, reason=%d\n", nasid,
475 ret);
477 if (ret == xpLocalPartid)
478 break;
480 continue;
483 xpc_request_partition_activation(remote_rp,
484 remote_rp_pa, nasid);
488 kfree(discovered_nasids);
489 kfree(remote_rp_base);
493 * Given a partid, get the nasids owned by that partition from the
494 * remote partition's reserved page.
496 enum xp_retval
497 xpc_initiate_partid_to_nasids(short partid, void *nasid_mask)
499 struct xpc_partition *part;
500 unsigned long part_nasid_pa;
502 part = &xpc_partitions[partid];
503 if (part->remote_rp_pa == 0)
504 return xpPartitionDown;
506 memset(nasid_mask, 0, xpc_nasid_mask_nbytes);
508 part_nasid_pa = (unsigned long)XPC_RP_PART_NASIDS(part->remote_rp_pa);
510 return xp_remote_memcpy(xp_pa(nasid_mask), part_nasid_pa,
511 xpc_nasid_mask_nbytes);