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
6 * Copyright (c) 2004-2006 Silicon Graphics, Inc. All Rights Reserved.
11 * Cross Partition Communication (XPC) partition support.
13 * This is the part of XPC that detects the presence/absence of
14 * other partitions. It provides a heartbeat and monitors the
15 * heartbeats of other partitions.
20 #include <linux/kernel.h>
21 #include <linux/sysctl.h>
22 #include <linux/cache.h>
23 #include <linux/mmzone.h>
24 #include <linux/nodemask.h>
25 #include <asm/uncached.h>
26 #include <asm/sn/bte.h>
27 #include <asm/sn/intr.h>
28 #include <asm/sn/sn_sal.h>
29 #include <asm/sn/nodepda.h>
30 #include <asm/sn/addrs.h>
31 #include <asm/sn/xpc.h>
34 /* XPC is exiting flag */
38 /* SH_IPI_ACCESS shub register value on startup */
39 static u64 xpc_sh1_IPI_access;
40 static u64 xpc_sh2_IPI_access0;
41 static u64 xpc_sh2_IPI_access1;
42 static u64 xpc_sh2_IPI_access2;
43 static u64 xpc_sh2_IPI_access3;
46 /* original protection values for each node */
47 u64 xpc_prot_vec[MAX_NUMNODES];
50 /* this partition's reserved page pointers */
51 struct xpc_rsvd_page *xpc_rsvd_page;
52 static u64 *xpc_part_nasids;
53 static u64 *xpc_mach_nasids;
54 struct xpc_vars *xpc_vars;
55 struct xpc_vars_part *xpc_vars_part;
57 static int xp_nasid_mask_bytes; /* actual size in bytes of nasid mask */
58 static int xp_nasid_mask_words; /* actual size in words of nasid mask */
62 * For performance reasons, each entry of xpc_partitions[] is cacheline
63 * aligned. And xpc_partitions[] is padded with an additional entry at the
64 * end so that the last legitimate entry doesn't share its cacheline with
67 struct xpc_partition xpc_partitions[XP_MAX_PARTITIONS + 1];
71 * Generic buffer used to store a local copy of portions of a remote
72 * partition's reserved page (either its header and part_nasids mask,
75 char *xpc_remote_copy_buffer;
76 void *xpc_remote_copy_buffer_base;
80 * Given a nasid, get the physical address of the partition's reserved page
81 * for that nasid. This function returns 0 on any error.
84 xpc_get_rsvd_page_pa(int nasid)
89 u64 rp_pa = nasid; /* seed with nasid */
93 void *buf_base = NULL;
98 status = sn_partition_reserved_page_pa(buf, &cookie, &rp_pa,
101 dev_dbg(xpc_part, "SAL returned with status=%li, cookie="
102 "0x%016lx, address=0x%016lx, len=0x%016lx\n",
103 status, cookie, rp_pa, len);
105 if (status != SALRET_MORE_PASSES) {
109 if (L1_CACHE_ALIGN(len) > buf_len) {
110 if (buf_base != NULL) {
113 buf_len = L1_CACHE_ALIGN(len);
114 buf = (u64) xpc_kmalloc_cacheline_aligned(buf_len,
115 GFP_KERNEL, &buf_base);
116 if (buf_base == NULL) {
117 dev_err(xpc_part, "unable to kmalloc "
118 "len=0x%016lx\n", buf_len);
119 status = SALRET_ERROR;
124 bte_res = xp_bte_copy(rp_pa, buf, buf_len,
125 (BTE_NOTIFY | BTE_WACQUIRE), NULL);
126 if (bte_res != BTE_SUCCESS) {
127 dev_dbg(xpc_part, "xp_bte_copy failed %i\n", bte_res);
128 status = SALRET_ERROR;
133 if (buf_base != NULL) {
137 if (status != SALRET_OK) {
140 dev_dbg(xpc_part, "reserved page at phys address 0x%016lx\n", rp_pa);
146 * Fill the partition reserved page with the information needed by
147 * other partitions to discover we are alive and establish initial
150 struct xpc_rsvd_page *
151 xpc_rsvd_page_init(void)
153 struct xpc_rsvd_page *rp;
155 u64 rp_pa, nasid_array = 0;
159 /* get the local reserved page's address */
162 rp_pa = xpc_get_rsvd_page_pa(cpuid_to_nasid(smp_processor_id()));
165 dev_err(xpc_part, "SAL failed to locate the reserved page\n");
168 rp = (struct xpc_rsvd_page *) __va(rp_pa);
170 if (rp->partid != sn_partition_id) {
171 dev_err(xpc_part, "the reserved page's partid of %d should be "
172 "%d\n", rp->partid, sn_partition_id);
176 rp->version = XPC_RP_VERSION;
178 /* establish the actual sizes of the nasid masks */
179 if (rp->SAL_version == 1) {
180 /* SAL_version 1 didn't set the nasids_size field */
181 rp->nasids_size = 128;
183 xp_nasid_mask_bytes = rp->nasids_size;
184 xp_nasid_mask_words = xp_nasid_mask_bytes / 8;
186 /* setup the pointers to the various items in the reserved page */
187 xpc_part_nasids = XPC_RP_PART_NASIDS(rp);
188 xpc_mach_nasids = XPC_RP_MACH_NASIDS(rp);
189 xpc_vars = XPC_RP_VARS(rp);
190 xpc_vars_part = XPC_RP_VARS_PART(rp);
193 * Before clearing xpc_vars, see if a page of AMOs had been previously
194 * allocated. If not we'll need to allocate one and set permissions
195 * so that cross-partition AMOs are allowed.
197 * The allocated AMO page needs MCA reporting to remain disabled after
198 * XPC has unloaded. To make this work, we keep a copy of the pointer
199 * to this page (i.e., amos_page) in the struct xpc_vars structure,
200 * which is pointed to by the reserved page, and re-use that saved copy
201 * on subsequent loads of XPC. This AMO page is never freed, and its
202 * memory protections are never restricted.
204 if ((amos_page = xpc_vars->amos_page) == NULL) {
205 amos_page = (AMO_t *) TO_AMO(uncached_alloc_page(0));
206 if (amos_page == NULL) {
207 dev_err(xpc_part, "can't allocate page of AMOs\n");
212 * Open up AMO-R/W to cpu. This is done for Shub 1.1 systems
213 * when xpc_allow_IPI_ops() is called via xpc_hb_init().
215 if (!enable_shub_wars_1_1()) {
216 ret = sn_change_memprotect(ia64_tpa((u64) amos_page),
217 PAGE_SIZE, SN_MEMPROT_ACCESS_CLASS_1,
220 dev_err(xpc_part, "can't change memory "
222 uncached_free_page(__IA64_UNCACHED_OFFSET |
223 TO_PHYS((u64) amos_page));
227 } else if (!IS_AMO_ADDRESS((u64) amos_page)) {
229 * EFI's XPBOOT can also set amos_page in the reserved page,
230 * but it happens to leave it as an uncached physical address
231 * and we need it to be an uncached virtual, so we'll have to
234 if (!IS_AMO_PHYS_ADDRESS((u64) amos_page)) {
235 dev_err(xpc_part, "previously used amos_page address "
236 "is bad = 0x%p\n", (void *) amos_page);
239 amos_page = (AMO_t *) TO_AMO((u64) amos_page);
243 memset(xpc_vars, 0, sizeof(struct xpc_vars));
245 xpc_vars->version = XPC_V_VERSION;
246 xpc_vars->act_nasid = cpuid_to_nasid(0);
247 xpc_vars->act_phys_cpuid = cpu_physical_id(0);
248 xpc_vars->vars_part_pa = __pa(xpc_vars_part);
249 xpc_vars->amos_page_pa = ia64_tpa((u64) amos_page);
250 xpc_vars->amos_page = amos_page; /* save for next load of XPC */
253 /* clear xpc_vars_part */
254 memset((u64 *) xpc_vars_part, 0, sizeof(struct xpc_vars_part) *
257 /* initialize the activate IRQ related AMO variables */
258 for (i = 0; i < xp_nasid_mask_words; i++) {
259 (void) xpc_IPI_init(XPC_ACTIVATE_IRQ_AMOS + i);
262 /* initialize the engaged remote partitions related AMO variables */
263 (void) xpc_IPI_init(XPC_ENGAGED_PARTITIONS_AMO);
264 (void) xpc_IPI_init(XPC_DISENGAGE_REQUEST_AMO);
266 /* timestamp of when reserved page was setup by XPC */
267 rp->stamp = CURRENT_TIME;
270 * This signifies to the remote partition that our reserved
271 * page is initialized.
273 rp->vars_pa = __pa(xpc_vars);
280 * Change protections to allow IPI operations (and AMO operations on
284 xpc_allow_IPI_ops(void)
290 // >>> Change SH_IPI_ACCESS code to use SAL call once it is available.
293 xpc_sh2_IPI_access0 =
294 (u64) HUB_L((u64 *) LOCAL_MMR_ADDR(SH2_IPI_ACCESS0));
295 xpc_sh2_IPI_access1 =
296 (u64) HUB_L((u64 *) LOCAL_MMR_ADDR(SH2_IPI_ACCESS1));
297 xpc_sh2_IPI_access2 =
298 (u64) HUB_L((u64 *) LOCAL_MMR_ADDR(SH2_IPI_ACCESS2));
299 xpc_sh2_IPI_access3 =
300 (u64) HUB_L((u64 *) LOCAL_MMR_ADDR(SH2_IPI_ACCESS3));
302 for_each_online_node(node) {
303 nasid = cnodeid_to_nasid(node);
304 HUB_S((u64 *) GLOBAL_MMR_ADDR(nasid, SH2_IPI_ACCESS0),
306 HUB_S((u64 *) GLOBAL_MMR_ADDR(nasid, SH2_IPI_ACCESS1),
308 HUB_S((u64 *) GLOBAL_MMR_ADDR(nasid, SH2_IPI_ACCESS2),
310 HUB_S((u64 *) GLOBAL_MMR_ADDR(nasid, SH2_IPI_ACCESS3),
316 (u64) HUB_L((u64 *) LOCAL_MMR_ADDR(SH1_IPI_ACCESS));
318 for_each_online_node(node) {
319 nasid = cnodeid_to_nasid(node);
320 HUB_S((u64 *) GLOBAL_MMR_ADDR(nasid, SH1_IPI_ACCESS),
324 * Since the BIST collides with memory operations on
325 * SHUB 1.1 sn_change_memprotect() cannot be used.
327 if (enable_shub_wars_1_1()) {
328 /* open up everything */
329 xpc_prot_vec[node] = (u64) HUB_L((u64 *)
330 GLOBAL_MMR_ADDR(nasid,
331 SH1_MD_DQLP_MMR_DIR_PRIVEC0));
332 HUB_S((u64 *) GLOBAL_MMR_ADDR(nasid,
333 SH1_MD_DQLP_MMR_DIR_PRIVEC0),
335 HUB_S((u64 *) GLOBAL_MMR_ADDR(nasid,
336 SH1_MD_DQRP_MMR_DIR_PRIVEC0),
345 * Restrict protections to disallow IPI operations (and AMO operations on
349 xpc_restrict_IPI_ops(void)
355 // >>> Change SH_IPI_ACCESS code to use SAL call once it is available.
359 for_each_online_node(node) {
360 nasid = cnodeid_to_nasid(node);
361 HUB_S((u64 *) GLOBAL_MMR_ADDR(nasid, SH2_IPI_ACCESS0),
362 xpc_sh2_IPI_access0);
363 HUB_S((u64 *) GLOBAL_MMR_ADDR(nasid, SH2_IPI_ACCESS1),
364 xpc_sh2_IPI_access1);
365 HUB_S((u64 *) GLOBAL_MMR_ADDR(nasid, SH2_IPI_ACCESS2),
366 xpc_sh2_IPI_access2);
367 HUB_S((u64 *) GLOBAL_MMR_ADDR(nasid, SH2_IPI_ACCESS3),
368 xpc_sh2_IPI_access3);
373 for_each_online_node(node) {
374 nasid = cnodeid_to_nasid(node);
375 HUB_S((u64 *) GLOBAL_MMR_ADDR(nasid, SH1_IPI_ACCESS),
378 if (enable_shub_wars_1_1()) {
379 HUB_S((u64 *) GLOBAL_MMR_ADDR(nasid,
380 SH1_MD_DQLP_MMR_DIR_PRIVEC0),
382 HUB_S((u64 *) GLOBAL_MMR_ADDR(nasid,
383 SH1_MD_DQRP_MMR_DIR_PRIVEC0),
392 * At periodic intervals, scan through all active partitions and ensure
393 * their heartbeat is still active. If not, the partition is deactivated.
396 xpc_check_remote_hb(void)
398 struct xpc_vars *remote_vars;
399 struct xpc_partition *part;
404 remote_vars = (struct xpc_vars *) xpc_remote_copy_buffer;
406 for (partid = 1; partid < XP_MAX_PARTITIONS; partid++) {
412 if (partid == sn_partition_id) {
416 part = &xpc_partitions[partid];
418 if (part->act_state == XPC_P_INACTIVE ||
419 part->act_state == XPC_P_DEACTIVATING) {
423 /* pull the remote_hb cache line */
424 bres = xp_bte_copy(part->remote_vars_pa,
427 (BTE_NOTIFY | BTE_WACQUIRE), NULL);
428 if (bres != BTE_SUCCESS) {
429 XPC_DEACTIVATE_PARTITION(part,
430 xpc_map_bte_errors(bres));
434 dev_dbg(xpc_part, "partid = %d, heartbeat = %ld, last_heartbeat"
435 " = %ld, heartbeat_offline = %ld, HB_mask = 0x%lx\n",
436 partid, remote_vars->heartbeat, part->last_heartbeat,
437 remote_vars->heartbeat_offline,
438 remote_vars->heartbeating_to_mask);
440 if (((remote_vars->heartbeat == part->last_heartbeat) &&
441 (remote_vars->heartbeat_offline == 0)) ||
442 !xpc_hb_allowed(sn_partition_id, remote_vars)) {
444 XPC_DEACTIVATE_PARTITION(part, xpcNoHeartbeat);
448 part->last_heartbeat = remote_vars->heartbeat;
454 * Get a copy of a portion of the remote partition's rsvd page.
456 * remote_rp points to a buffer that is cacheline aligned for BTE copies and
457 * is large enough to contain a copy of their reserved page header and
460 static enum xpc_retval
461 xpc_get_remote_rp(int nasid, u64 *discovered_nasids,
462 struct xpc_rsvd_page *remote_rp, u64 *remote_rp_pa)
467 /* get the reserved page's physical address */
469 *remote_rp_pa = xpc_get_rsvd_page_pa(nasid);
470 if (*remote_rp_pa == 0) {
471 return xpcNoRsvdPageAddr;
475 /* pull over the reserved page header and part_nasids mask */
476 bres = xp_bte_copy(*remote_rp_pa, (u64) remote_rp,
477 XPC_RP_HEADER_SIZE + xp_nasid_mask_bytes,
478 (BTE_NOTIFY | BTE_WACQUIRE), NULL);
479 if (bres != BTE_SUCCESS) {
480 return xpc_map_bte_errors(bres);
484 if (discovered_nasids != NULL) {
485 u64 *remote_part_nasids = XPC_RP_PART_NASIDS(remote_rp);
488 for (i = 0; i < xp_nasid_mask_words; i++) {
489 discovered_nasids[i] |= remote_part_nasids[i];
494 /* check that the partid is for another partition */
496 if (remote_rp->partid < 1 ||
497 remote_rp->partid > (XP_MAX_PARTITIONS - 1)) {
498 return xpcInvalidPartid;
501 if (remote_rp->partid == sn_partition_id) {
502 return xpcLocalPartid;
506 if (XPC_VERSION_MAJOR(remote_rp->version) !=
507 XPC_VERSION_MAJOR(XPC_RP_VERSION)) {
508 return xpcBadVersion;
516 * Get a copy of the remote partition's XPC variables from the reserved page.
518 * remote_vars points to a buffer that is cacheline aligned for BTE copies and
519 * assumed to be of size XPC_RP_VARS_SIZE.
521 static enum xpc_retval
522 xpc_get_remote_vars(u64 remote_vars_pa, struct xpc_vars *remote_vars)
527 if (remote_vars_pa == 0) {
528 return xpcVarsNotSet;
531 /* pull over the cross partition variables */
532 bres = xp_bte_copy(remote_vars_pa, (u64) remote_vars, XPC_RP_VARS_SIZE,
533 (BTE_NOTIFY | BTE_WACQUIRE), NULL);
534 if (bres != BTE_SUCCESS) {
535 return xpc_map_bte_errors(bres);
538 if (XPC_VERSION_MAJOR(remote_vars->version) !=
539 XPC_VERSION_MAJOR(XPC_V_VERSION)) {
540 return xpcBadVersion;
548 * Update the remote partition's info.
551 xpc_update_partition_info(struct xpc_partition *part, u8 remote_rp_version,
552 struct timespec *remote_rp_stamp, u64 remote_rp_pa,
553 u64 remote_vars_pa, struct xpc_vars *remote_vars)
555 part->remote_rp_version = remote_rp_version;
556 dev_dbg(xpc_part, " remote_rp_version = 0x%016lx\n",
557 part->remote_rp_version);
559 part->remote_rp_stamp = *remote_rp_stamp;
560 dev_dbg(xpc_part, " remote_rp_stamp (tv_sec = 0x%lx tv_nsec = 0x%lx\n",
561 part->remote_rp_stamp.tv_sec, part->remote_rp_stamp.tv_nsec);
563 part->remote_rp_pa = remote_rp_pa;
564 dev_dbg(xpc_part, " remote_rp_pa = 0x%016lx\n", part->remote_rp_pa);
566 part->remote_vars_pa = remote_vars_pa;
567 dev_dbg(xpc_part, " remote_vars_pa = 0x%016lx\n",
568 part->remote_vars_pa);
570 part->last_heartbeat = remote_vars->heartbeat;
571 dev_dbg(xpc_part, " last_heartbeat = 0x%016lx\n",
572 part->last_heartbeat);
574 part->remote_vars_part_pa = remote_vars->vars_part_pa;
575 dev_dbg(xpc_part, " remote_vars_part_pa = 0x%016lx\n",
576 part->remote_vars_part_pa);
578 part->remote_act_nasid = remote_vars->act_nasid;
579 dev_dbg(xpc_part, " remote_act_nasid = 0x%x\n",
580 part->remote_act_nasid);
582 part->remote_act_phys_cpuid = remote_vars->act_phys_cpuid;
583 dev_dbg(xpc_part, " remote_act_phys_cpuid = 0x%x\n",
584 part->remote_act_phys_cpuid);
586 part->remote_amos_page_pa = remote_vars->amos_page_pa;
587 dev_dbg(xpc_part, " remote_amos_page_pa = 0x%lx\n",
588 part->remote_amos_page_pa);
590 part->remote_vars_version = remote_vars->version;
591 dev_dbg(xpc_part, " remote_vars_version = 0x%x\n",
592 part->remote_vars_version);
597 * Prior code has determined the nasid which generated an IPI. Inspect
598 * that nasid to determine if its partition needs to be activated or
601 * A partition is consider "awaiting activation" if our partition
602 * flags indicate it is not active and it has a heartbeat. A
603 * partition is considered "awaiting deactivation" if our partition
604 * flags indicate it is active but it has no heartbeat or it is not
605 * sending its heartbeat to us.
607 * To determine the heartbeat, the remote nasid must have a properly
608 * initialized reserved page.
611 xpc_identify_act_IRQ_req(int nasid)
613 struct xpc_rsvd_page *remote_rp;
614 struct xpc_vars *remote_vars;
617 int remote_rp_version;
620 struct timespec remote_rp_stamp = { 0, 0 };
622 struct xpc_partition *part;
626 /* pull over the reserved page structure */
628 remote_rp = (struct xpc_rsvd_page *) xpc_remote_copy_buffer;
630 ret = xpc_get_remote_rp(nasid, NULL, remote_rp, &remote_rp_pa);
631 if (ret != xpcSuccess) {
632 dev_warn(xpc_part, "unable to get reserved page from nasid %d, "
633 "which sent interrupt, reason=%d\n", nasid, ret);
637 remote_vars_pa = remote_rp->vars_pa;
638 remote_rp_version = remote_rp->version;
639 if (XPC_SUPPORTS_RP_STAMP(remote_rp_version)) {
640 remote_rp_stamp = remote_rp->stamp;
642 partid = remote_rp->partid;
643 part = &xpc_partitions[partid];
646 /* pull over the cross partition variables */
648 remote_vars = (struct xpc_vars *) xpc_remote_copy_buffer;
650 ret = xpc_get_remote_vars(remote_vars_pa, remote_vars);
651 if (ret != xpcSuccess) {
653 dev_warn(xpc_part, "unable to get XPC variables from nasid %d, "
654 "which sent interrupt, reason=%d\n", nasid, ret);
656 XPC_DEACTIVATE_PARTITION(part, ret);
661 part->act_IRQ_rcvd++;
663 dev_dbg(xpc_part, "partid for nasid %d is %d; IRQs = %d; HB = "
664 "%ld:0x%lx\n", (int) nasid, (int) partid, part->act_IRQ_rcvd,
665 remote_vars->heartbeat, remote_vars->heartbeating_to_mask);
667 if (xpc_partition_disengaged(part) &&
668 part->act_state == XPC_P_INACTIVE) {
670 xpc_update_partition_info(part, remote_rp_version,
671 &remote_rp_stamp, remote_rp_pa,
672 remote_vars_pa, remote_vars);
674 if (XPC_SUPPORTS_DISENGAGE_REQUEST(part->remote_vars_version)) {
675 if (xpc_partition_disengage_requested(1UL << partid)) {
677 * Other side is waiting on us to disengage,
678 * even though we already have.
683 /* other side doesn't support disengage requests */
684 xpc_clear_partition_disengage_request(1UL << partid);
687 xpc_activate_partition(part);
691 DBUG_ON(part->remote_rp_version == 0);
692 DBUG_ON(part->remote_vars_version == 0);
694 if (!XPC_SUPPORTS_RP_STAMP(part->remote_rp_version)) {
695 DBUG_ON(XPC_SUPPORTS_DISENGAGE_REQUEST(part->
696 remote_vars_version));
698 if (!XPC_SUPPORTS_RP_STAMP(remote_rp_version)) {
699 DBUG_ON(XPC_SUPPORTS_DISENGAGE_REQUEST(remote_vars->
701 /* see if the other side rebooted */
702 if (part->remote_amos_page_pa ==
703 remote_vars->amos_page_pa &&
704 xpc_hb_allowed(sn_partition_id,
706 /* doesn't look that way, so ignore the IPI */
712 * Other side rebooted and previous XPC didn't support the
713 * disengage request, so we don't need to do anything special.
716 xpc_update_partition_info(part, remote_rp_version,
717 &remote_rp_stamp, remote_rp_pa,
718 remote_vars_pa, remote_vars);
719 part->reactivate_nasid = nasid;
720 XPC_DEACTIVATE_PARTITION(part, xpcReactivating);
724 DBUG_ON(!XPC_SUPPORTS_DISENGAGE_REQUEST(part->remote_vars_version));
726 if (!XPC_SUPPORTS_RP_STAMP(remote_rp_version)) {
727 DBUG_ON(!XPC_SUPPORTS_DISENGAGE_REQUEST(remote_vars->version));
730 * Other side rebooted and previous XPC did support the
731 * disengage request, but the new one doesn't.
734 xpc_clear_partition_engaged(1UL << partid);
735 xpc_clear_partition_disengage_request(1UL << partid);
737 xpc_update_partition_info(part, remote_rp_version,
738 &remote_rp_stamp, remote_rp_pa,
739 remote_vars_pa, remote_vars);
743 DBUG_ON(!XPC_SUPPORTS_DISENGAGE_REQUEST(remote_vars->version));
745 stamp_diff = xpc_compare_stamps(&part->remote_rp_stamp,
747 if (stamp_diff != 0) {
748 DBUG_ON(stamp_diff >= 0);
751 * Other side rebooted and the previous XPC did support
752 * the disengage request, as does the new one.
755 DBUG_ON(xpc_partition_engaged(1UL << partid));
756 DBUG_ON(xpc_partition_disengage_requested(1UL <<
759 xpc_update_partition_info(part, remote_rp_version,
760 &remote_rp_stamp, remote_rp_pa,
761 remote_vars_pa, remote_vars);
766 if (part->disengage_request_timeout > 0 &&
767 !xpc_partition_disengaged(part)) {
768 /* still waiting on other side to disengage from us */
773 part->reactivate_nasid = nasid;
774 XPC_DEACTIVATE_PARTITION(part, xpcReactivating);
776 } else if (XPC_SUPPORTS_DISENGAGE_REQUEST(part->remote_vars_version) &&
777 xpc_partition_disengage_requested(1UL << partid)) {
778 XPC_DEACTIVATE_PARTITION(part, xpcOtherGoingDown);
784 * Loop through the activation AMO variables and process any bits
785 * which are set. Each bit indicates a nasid sending a partition
786 * activation or deactivation request.
788 * Return #of IRQs detected.
791 xpc_identify_act_IRQ_sender(void)
795 u64 nasid; /* remote nasid */
796 int n_IRQs_detected = 0;
800 act_amos = xpc_vars->amos_page + XPC_ACTIVATE_IRQ_AMOS;
803 /* scan through act AMO variable looking for non-zero entries */
804 for (word = 0; word < xp_nasid_mask_words; word++) {
810 nasid_mask = xpc_IPI_receive(&act_amos[word]);
811 if (nasid_mask == 0) {
812 /* no IRQs from nasids in this variable */
816 dev_dbg(xpc_part, "AMO[%d] gave back 0x%lx\n", word,
821 * If this nasid has been added to the machine since
822 * our partition was reset, this will retain the
823 * remote nasid in our reserved pages machine mask.
824 * This is used in the event of module reload.
826 xpc_mach_nasids[word] |= nasid_mask;
829 /* locate the nasid(s) which sent interrupts */
831 for (bit = 0; bit < (8 * sizeof(u64)); bit++) {
832 if (nasid_mask & (1UL << bit)) {
834 nasid = XPC_NASID_FROM_W_B(word, bit);
835 dev_dbg(xpc_part, "interrupt from nasid %ld\n",
837 xpc_identify_act_IRQ_req(nasid);
841 return n_IRQs_detected;
846 * See if the other side has responded to a partition disengage request
850 xpc_partition_disengaged(struct xpc_partition *part)
852 partid_t partid = XPC_PARTID(part);
856 disengaged = (xpc_partition_engaged(1UL << partid) == 0);
857 if (part->disengage_request_timeout) {
859 if (jiffies < part->disengage_request_timeout) {
860 /* timelimit hasn't been reached yet */
865 * Other side hasn't responded to our disengage
866 * request in a timely fashion, so assume it's dead.
869 dev_info(xpc_part, "disengage from remote partition %d "
870 "timed out\n", partid);
871 xpc_disengage_request_timedout = 1;
872 xpc_clear_partition_engaged(1UL << partid);
875 part->disengage_request_timeout = 0;
877 /* cancel the timer function, provided it's not us */
878 if (!in_interrupt()) {
879 del_singleshot_timer_sync(&part->
880 disengage_request_timer);
883 DBUG_ON(part->act_state != XPC_P_DEACTIVATING &&
884 part->act_state != XPC_P_INACTIVE);
885 if (part->act_state != XPC_P_INACTIVE) {
886 xpc_wakeup_channel_mgr(part);
889 if (XPC_SUPPORTS_DISENGAGE_REQUEST(part->remote_vars_version)) {
890 xpc_cancel_partition_disengage_request(part);
898 * Mark specified partition as active.
901 xpc_mark_partition_active(struct xpc_partition *part)
903 unsigned long irq_flags;
907 dev_dbg(xpc_part, "setting partition %d to ACTIVE\n", XPC_PARTID(part));
909 spin_lock_irqsave(&part->act_lock, irq_flags);
910 if (part->act_state == XPC_P_ACTIVATING) {
911 part->act_state = XPC_P_ACTIVE;
914 DBUG_ON(part->reason == xpcSuccess);
917 spin_unlock_irqrestore(&part->act_lock, irq_flags);
924 * Notify XPC that the partition is down.
927 xpc_deactivate_partition(const int line, struct xpc_partition *part,
928 enum xpc_retval reason)
930 unsigned long irq_flags;
933 spin_lock_irqsave(&part->act_lock, irq_flags);
935 if (part->act_state == XPC_P_INACTIVE) {
936 XPC_SET_REASON(part, reason, line);
937 spin_unlock_irqrestore(&part->act_lock, irq_flags);
938 if (reason == xpcReactivating) {
939 /* we interrupt ourselves to reactivate partition */
940 xpc_IPI_send_reactivate(part);
944 if (part->act_state == XPC_P_DEACTIVATING) {
945 if ((part->reason == xpcUnloading && reason != xpcUnloading) ||
946 reason == xpcReactivating) {
947 XPC_SET_REASON(part, reason, line);
949 spin_unlock_irqrestore(&part->act_lock, irq_flags);
953 part->act_state = XPC_P_DEACTIVATING;
954 XPC_SET_REASON(part, reason, line);
956 spin_unlock_irqrestore(&part->act_lock, irq_flags);
958 if (XPC_SUPPORTS_DISENGAGE_REQUEST(part->remote_vars_version)) {
959 xpc_request_partition_disengage(part);
960 xpc_IPI_send_disengage(part);
962 /* set a timelimit on the disengage request */
963 part->disengage_request_timeout = jiffies +
964 (xpc_disengage_request_timelimit * HZ);
965 part->disengage_request_timer.expires =
966 part->disengage_request_timeout;
967 add_timer(&part->disengage_request_timer);
970 dev_dbg(xpc_part, "bringing partition %d down, reason = %d\n",
971 XPC_PARTID(part), reason);
973 xpc_partition_going_down(part, reason);
978 * Mark specified partition as inactive.
981 xpc_mark_partition_inactive(struct xpc_partition *part)
983 unsigned long irq_flags;
986 dev_dbg(xpc_part, "setting partition %d to INACTIVE\n",
989 spin_lock_irqsave(&part->act_lock, irq_flags);
990 part->act_state = XPC_P_INACTIVE;
991 spin_unlock_irqrestore(&part->act_lock, irq_flags);
992 part->remote_rp_pa = 0;
997 * SAL has provided a partition and machine mask. The partition mask
998 * contains a bit for each even nasid in our partition. The machine
999 * mask contains a bit for each even nasid in the entire machine.
1001 * Using those two bit arrays, we can determine which nasids are
1002 * known in the machine. Each should also have a reserved page
1003 * initialized if they are available for partitioning.
1008 void *remote_rp_base;
1009 struct xpc_rsvd_page *remote_rp;
1010 struct xpc_vars *remote_vars;
1017 struct xpc_rsvd_page *rp;
1019 struct xpc_partition *part;
1020 u64 *discovered_nasids;
1021 enum xpc_retval ret;
1024 remote_rp = xpc_kmalloc_cacheline_aligned(XPC_RP_HEADER_SIZE +
1025 xp_nasid_mask_bytes,
1026 GFP_KERNEL, &remote_rp_base);
1027 if (remote_rp == NULL) {
1030 remote_vars = (struct xpc_vars *) remote_rp;
1033 discovered_nasids = kmalloc(sizeof(u64) * xp_nasid_mask_words,
1035 if (discovered_nasids == NULL) {
1036 kfree(remote_rp_base);
1039 memset(discovered_nasids, 0, sizeof(u64) * xp_nasid_mask_words);
1041 rp = (struct xpc_rsvd_page *) xpc_rsvd_page;
1044 * The term 'region' in this context refers to the minimum number of
1045 * nodes that can comprise an access protection grouping. The access
1046 * protection is in regards to memory, IOI and IPI.
1049 region_size = sn_region_size;
1051 switch (region_size) {
1059 DBUG_ON(!is_shub2());
1062 for (region = 0; region < max_regions; region++) {
1064 if ((volatile int) xpc_exiting) {
1068 dev_dbg(xpc_part, "searching region %d\n", region);
1070 for (nasid = (region * region_size * 2);
1071 nasid < ((region + 1) * region_size * 2);
1074 if ((volatile int) xpc_exiting) {
1078 dev_dbg(xpc_part, "checking nasid %d\n", nasid);
1081 if (XPC_NASID_IN_ARRAY(nasid, xpc_part_nasids)) {
1082 dev_dbg(xpc_part, "PROM indicates Nasid %d is "
1083 "part of the local partition; skipping "
1088 if (!(XPC_NASID_IN_ARRAY(nasid, xpc_mach_nasids))) {
1089 dev_dbg(xpc_part, "PROM indicates Nasid %d was "
1090 "not on Numa-Link network at reset\n",
1095 if (XPC_NASID_IN_ARRAY(nasid, discovered_nasids)) {
1096 dev_dbg(xpc_part, "Nasid %d is part of a "
1097 "partition which was previously "
1098 "discovered\n", nasid);
1103 /* pull over the reserved page structure */
1105 ret = xpc_get_remote_rp(nasid, discovered_nasids,
1106 remote_rp, &remote_rp_pa);
1107 if (ret != xpcSuccess) {
1108 dev_dbg(xpc_part, "unable to get reserved page "
1109 "from nasid %d, reason=%d\n", nasid,
1112 if (ret == xpcLocalPartid) {
1118 remote_vars_pa = remote_rp->vars_pa;
1120 partid = remote_rp->partid;
1121 part = &xpc_partitions[partid];
1124 /* pull over the cross partition variables */
1126 ret = xpc_get_remote_vars(remote_vars_pa, remote_vars);
1127 if (ret != xpcSuccess) {
1128 dev_dbg(xpc_part, "unable to get XPC variables "
1129 "from nasid %d, reason=%d\n", nasid,
1132 XPC_DEACTIVATE_PARTITION(part, ret);
1136 if (part->act_state != XPC_P_INACTIVE) {
1137 dev_dbg(xpc_part, "partition %d on nasid %d is "
1138 "already activating\n", partid, nasid);
1143 * Register the remote partition's AMOs with SAL so it
1144 * can handle and cleanup errors within that address
1145 * range should the remote partition go down. We don't
1146 * unregister this range because it is difficult to
1147 * tell when outstanding writes to the remote partition
1148 * are finished and thus when it is thus safe to
1149 * unregister. This should not result in wasted space
1150 * in the SAL xp_addr_region table because we should
1151 * get the same page for remote_act_amos_pa after
1152 * module reloads and system reboots.
1154 if (sn_register_xp_addr_region(
1155 remote_vars->amos_page_pa,
1156 PAGE_SIZE, 1) < 0) {
1157 dev_dbg(xpc_part, "partition %d failed to "
1158 "register xp_addr region 0x%016lx\n",
1159 partid, remote_vars->amos_page_pa);
1161 XPC_SET_REASON(part, xpcPhysAddrRegFailed,
1167 * The remote nasid is valid and available.
1168 * Send an interrupt to that nasid to notify
1169 * it that we are ready to begin activation.
1171 dev_dbg(xpc_part, "sending an interrupt to AMO 0x%lx, "
1172 "nasid %d, phys_cpuid 0x%x\n",
1173 remote_vars->amos_page_pa,
1174 remote_vars->act_nasid,
1175 remote_vars->act_phys_cpuid);
1177 if (XPC_SUPPORTS_DISENGAGE_REQUEST(remote_vars->
1179 part->remote_amos_page_pa =
1180 remote_vars->amos_page_pa;
1181 xpc_mark_partition_disengaged(part);
1182 xpc_cancel_partition_disengage_request(part);
1184 xpc_IPI_send_activate(remote_vars);
1188 kfree(discovered_nasids);
1189 kfree(remote_rp_base);
1194 * Given a partid, get the nasids owned by that partition from the
1195 * remote partition's reserved page.
1198 xpc_initiate_partid_to_nasids(partid_t partid, void *nasid_mask)
1200 struct xpc_partition *part;
1205 part = &xpc_partitions[partid];
1206 if (part->remote_rp_pa == 0) {
1207 return xpcPartitionDown;
1210 memset(nasid_mask, 0, XP_NASID_MASK_BYTES);
1212 part_nasid_pa = (u64) XPC_RP_PART_NASIDS(part->remote_rp_pa);
1214 bte_res = xp_bte_copy(part_nasid_pa, (u64) nasid_mask,
1215 xp_nasid_mask_bytes, (BTE_NOTIFY | BTE_WACQUIRE), NULL);
1217 return xpc_map_bte_errors(bte_res);