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) 2004-2005 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>
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_COMPACT_NODES];
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 * xpc_discovery runs only once and is a seperate thread that is
76 * very likely going to be processing in parallel with receiving
79 char ____cacheline_aligned xpc_remote_copy_buffer[XPC_RP_HEADER_SIZE +
84 * Given a nasid, get the physical address of the partition's reserved page
85 * for that nasid. This function returns 0 on any error.
88 xpc_get_rsvd_page_pa(int nasid, u64 buf, u64 buf_size)
93 u64 rp_pa = nasid; /* seed with nasid */
99 status = sn_partition_reserved_page_pa(buf, &cookie, &rp_pa,
102 dev_dbg(xpc_part, "SAL returned with status=%li, cookie="
103 "0x%016lx, address=0x%016lx, len=0x%016lx\n",
104 status, cookie, rp_pa, len);
106 if (status != SALRET_MORE_PASSES) {
110 if (len > buf_size) {
111 dev_err(xpc_part, "len (=0x%016lx) > buf_size\n", len);
112 status = SALRET_ERROR;
116 bte_res = xp_bte_copy(rp_pa, ia64_tpa(buf), buf_size,
117 (BTE_NOTIFY | BTE_WACQUIRE), NULL);
118 if (bte_res != BTE_SUCCESS) {
119 dev_dbg(xpc_part, "xp_bte_copy failed %i\n", bte_res);
120 status = SALRET_ERROR;
125 if (status != SALRET_OK) {
128 dev_dbg(xpc_part, "reserved page at phys address 0x%016lx\n", rp_pa);
134 * Fill the partition reserved page with the information needed by
135 * other partitions to discover we are alive and establish initial
138 struct xpc_rsvd_page *
139 xpc_rsvd_page_init(void)
141 struct xpc_rsvd_page *rp;
143 u64 rp_pa, nasid_array = 0;
147 /* get the local reserved page's address */
149 rp_pa = xpc_get_rsvd_page_pa(cnodeid_to_nasid(0),
150 (u64) xpc_remote_copy_buffer,
154 dev_err(xpc_part, "SAL failed to locate the reserved page\n");
157 rp = (struct xpc_rsvd_page *) __va(rp_pa);
159 if (rp->partid != sn_partition_id) {
160 dev_err(xpc_part, "the reserved page's partid of %d should be "
161 "%d\n", rp->partid, sn_partition_id);
165 rp->version = XPC_RP_VERSION;
167 /* establish the actual sizes of the nasid masks */
168 if (rp->SAL_version == 1) {
169 /* SAL_version 1 didn't set the nasids_size field */
170 rp->nasids_size = 128;
172 xp_nasid_mask_bytes = rp->nasids_size;
173 xp_nasid_mask_words = xp_nasid_mask_bytes / 8;
175 /* setup the pointers to the various items in the reserved page */
176 xpc_part_nasids = XPC_RP_PART_NASIDS(rp);
177 xpc_mach_nasids = XPC_RP_MACH_NASIDS(rp);
178 xpc_vars = XPC_RP_VARS(rp);
179 xpc_vars_part = XPC_RP_VARS_PART(rp);
182 * Before clearing xpc_vars, see if a page of AMOs had been previously
183 * allocated. If not we'll need to allocate one and set permissions
184 * so that cross-partition AMOs are allowed.
186 * The allocated AMO page needs MCA reporting to remain disabled after
187 * XPC has unloaded. To make this work, we keep a copy of the pointer
188 * to this page (i.e., amos_page) in the struct xpc_vars structure,
189 * which is pointed to by the reserved page, and re-use that saved copy
190 * on subsequent loads of XPC. This AMO page is never freed, and its
191 * memory protections are never restricted.
193 if ((amos_page = xpc_vars->amos_page) == NULL) {
194 amos_page = (AMO_t *) TO_AMO(uncached_alloc_page(0));
195 if (amos_page == NULL) {
196 dev_err(xpc_part, "can't allocate page of AMOs\n");
201 * Open up AMO-R/W to cpu. This is done for Shub 1.1 systems
202 * when xpc_allow_IPI_ops() is called via xpc_hb_init().
204 if (!enable_shub_wars_1_1()) {
205 ret = sn_change_memprotect(ia64_tpa((u64) amos_page),
206 PAGE_SIZE, SN_MEMPROT_ACCESS_CLASS_1,
209 dev_err(xpc_part, "can't change memory "
211 uncached_free_page(__IA64_UNCACHED_OFFSET |
212 TO_PHYS((u64) amos_page));
216 } else if (!IS_AMO_ADDRESS((u64) amos_page)) {
218 * EFI's XPBOOT can also set amos_page in the reserved page,
219 * but it happens to leave it as an uncached physical address
220 * and we need it to be an uncached virtual, so we'll have to
223 if (!IS_AMO_PHYS_ADDRESS((u64) amos_page)) {
224 dev_err(xpc_part, "previously used amos_page address "
225 "is bad = 0x%p\n", (void *) amos_page);
228 amos_page = (AMO_t *) TO_AMO((u64) amos_page);
232 memset(xpc_vars, 0, sizeof(struct xpc_vars));
234 xpc_vars->version = XPC_V_VERSION;
235 xpc_vars->act_nasid = cpuid_to_nasid(0);
236 xpc_vars->act_phys_cpuid = cpu_physical_id(0);
237 xpc_vars->vars_part_pa = __pa(xpc_vars_part);
238 xpc_vars->amos_page_pa = ia64_tpa((u64) amos_page);
239 xpc_vars->amos_page = amos_page; /* save for next load of XPC */
242 /* clear xpc_vars_part */
243 memset((u64 *) xpc_vars_part, 0, sizeof(struct xpc_vars_part) *
246 /* initialize the activate IRQ related AMO variables */
247 for (i = 0; i < xp_nasid_mask_words; i++) {
248 (void) xpc_IPI_init(XPC_ACTIVATE_IRQ_AMOS + i);
251 /* initialize the engaged remote partitions related AMO variables */
252 (void) xpc_IPI_init(XPC_ENGAGED_PARTITIONS_AMO);
253 (void) xpc_IPI_init(XPC_DISENGAGE_REQUEST_AMO);
255 /* timestamp of when reserved page was setup by XPC */
256 rp->stamp = CURRENT_TIME;
259 * This signifies to the remote partition that our reserved
260 * page is initialized.
262 rp->vars_pa = __pa(xpc_vars);
269 * Change protections to allow IPI operations (and AMO operations on
273 xpc_allow_IPI_ops(void)
279 // >>> Change SH_IPI_ACCESS code to use SAL call once it is available.
282 xpc_sh2_IPI_access0 =
283 (u64) HUB_L((u64 *) LOCAL_MMR_ADDR(SH2_IPI_ACCESS0));
284 xpc_sh2_IPI_access1 =
285 (u64) HUB_L((u64 *) LOCAL_MMR_ADDR(SH2_IPI_ACCESS1));
286 xpc_sh2_IPI_access2 =
287 (u64) HUB_L((u64 *) LOCAL_MMR_ADDR(SH2_IPI_ACCESS2));
288 xpc_sh2_IPI_access3 =
289 (u64) HUB_L((u64 *) LOCAL_MMR_ADDR(SH2_IPI_ACCESS3));
291 for_each_online_node(node) {
292 nasid = cnodeid_to_nasid(node);
293 HUB_S((u64 *) GLOBAL_MMR_ADDR(nasid, SH2_IPI_ACCESS0),
295 HUB_S((u64 *) GLOBAL_MMR_ADDR(nasid, SH2_IPI_ACCESS1),
297 HUB_S((u64 *) GLOBAL_MMR_ADDR(nasid, SH2_IPI_ACCESS2),
299 HUB_S((u64 *) GLOBAL_MMR_ADDR(nasid, SH2_IPI_ACCESS3),
305 (u64) HUB_L((u64 *) LOCAL_MMR_ADDR(SH1_IPI_ACCESS));
307 for_each_online_node(node) {
308 nasid = cnodeid_to_nasid(node);
309 HUB_S((u64 *) GLOBAL_MMR_ADDR(nasid, SH1_IPI_ACCESS),
313 * Since the BIST collides with memory operations on
314 * SHUB 1.1 sn_change_memprotect() cannot be used.
316 if (enable_shub_wars_1_1()) {
317 /* open up everything */
318 xpc_prot_vec[node] = (u64) HUB_L((u64 *)
319 GLOBAL_MMR_ADDR(nasid,
320 SH1_MD_DQLP_MMR_DIR_PRIVEC0));
321 HUB_S((u64 *) GLOBAL_MMR_ADDR(nasid,
322 SH1_MD_DQLP_MMR_DIR_PRIVEC0),
324 HUB_S((u64 *) GLOBAL_MMR_ADDR(nasid,
325 SH1_MD_DQRP_MMR_DIR_PRIVEC0),
334 * Restrict protections to disallow IPI operations (and AMO operations on
338 xpc_restrict_IPI_ops(void)
344 // >>> Change SH_IPI_ACCESS code to use SAL call once it is available.
348 for_each_online_node(node) {
349 nasid = cnodeid_to_nasid(node);
350 HUB_S((u64 *) GLOBAL_MMR_ADDR(nasid, SH2_IPI_ACCESS0),
351 xpc_sh2_IPI_access0);
352 HUB_S((u64 *) GLOBAL_MMR_ADDR(nasid, SH2_IPI_ACCESS1),
353 xpc_sh2_IPI_access1);
354 HUB_S((u64 *) GLOBAL_MMR_ADDR(nasid, SH2_IPI_ACCESS2),
355 xpc_sh2_IPI_access2);
356 HUB_S((u64 *) GLOBAL_MMR_ADDR(nasid, SH2_IPI_ACCESS3),
357 xpc_sh2_IPI_access3);
362 for_each_online_node(node) {
363 nasid = cnodeid_to_nasid(node);
364 HUB_S((u64 *) GLOBAL_MMR_ADDR(nasid, SH1_IPI_ACCESS),
367 if (enable_shub_wars_1_1()) {
368 HUB_S((u64 *) GLOBAL_MMR_ADDR(nasid,
369 SH1_MD_DQLP_MMR_DIR_PRIVEC0),
371 HUB_S((u64 *) GLOBAL_MMR_ADDR(nasid,
372 SH1_MD_DQRP_MMR_DIR_PRIVEC0),
381 * At periodic intervals, scan through all active partitions and ensure
382 * their heartbeat is still active. If not, the partition is deactivated.
385 xpc_check_remote_hb(void)
387 struct xpc_vars *remote_vars;
388 struct xpc_partition *part;
393 remote_vars = (struct xpc_vars *) xpc_remote_copy_buffer;
395 for (partid = 1; partid < XP_MAX_PARTITIONS; partid++) {
401 if (partid == sn_partition_id) {
405 part = &xpc_partitions[partid];
407 if (part->act_state == XPC_P_INACTIVE ||
408 part->act_state == XPC_P_DEACTIVATING) {
412 /* pull the remote_hb cache line */
413 bres = xp_bte_copy(part->remote_vars_pa,
414 ia64_tpa((u64) remote_vars),
416 (BTE_NOTIFY | BTE_WACQUIRE), NULL);
417 if (bres != BTE_SUCCESS) {
418 XPC_DEACTIVATE_PARTITION(part,
419 xpc_map_bte_errors(bres));
423 dev_dbg(xpc_part, "partid = %d, heartbeat = %ld, last_heartbeat"
424 " = %ld, kdb_status = %ld, HB_mask = 0x%lx\n", partid,
425 remote_vars->heartbeat, part->last_heartbeat,
426 remote_vars->kdb_status,
427 remote_vars->heartbeating_to_mask);
429 if (((remote_vars->heartbeat == part->last_heartbeat) &&
430 (remote_vars->kdb_status == 0)) ||
431 !xpc_hb_allowed(sn_partition_id, remote_vars)) {
433 XPC_DEACTIVATE_PARTITION(part, xpcNoHeartbeat);
437 part->last_heartbeat = remote_vars->heartbeat;
443 * Get a copy of a portion of the remote partition's rsvd page.
445 * remote_rp points to a buffer that is cacheline aligned for BTE copies and
446 * is large enough to contain a copy of their reserved page header and
449 static enum xpc_retval
450 xpc_get_remote_rp(int nasid, u64 *discovered_nasids,
451 struct xpc_rsvd_page *remote_rp, u64 *remote_rp_pa)
456 /* get the reserved page's physical address */
458 *remote_rp_pa = xpc_get_rsvd_page_pa(nasid, (u64) remote_rp,
460 xp_nasid_mask_bytes);
461 if (*remote_rp_pa == 0) {
462 return xpcNoRsvdPageAddr;
466 /* pull over the reserved page header and part_nasids mask */
468 bres = xp_bte_copy(*remote_rp_pa, ia64_tpa((u64) remote_rp),
469 XPC_RP_HEADER_SIZE + xp_nasid_mask_bytes,
470 (BTE_NOTIFY | BTE_WACQUIRE), NULL);
471 if (bres != BTE_SUCCESS) {
472 return xpc_map_bte_errors(bres);
476 if (discovered_nasids != NULL) {
477 u64 *remote_part_nasids = XPC_RP_PART_NASIDS(remote_rp);
480 for (i = 0; i < xp_nasid_mask_words; i++) {
481 discovered_nasids[i] |= remote_part_nasids[i];
486 /* check that the partid is for another partition */
488 if (remote_rp->partid < 1 ||
489 remote_rp->partid > (XP_MAX_PARTITIONS - 1)) {
490 return xpcInvalidPartid;
493 if (remote_rp->partid == sn_partition_id) {
494 return xpcLocalPartid;
498 if (XPC_VERSION_MAJOR(remote_rp->version) !=
499 XPC_VERSION_MAJOR(XPC_RP_VERSION)) {
500 return xpcBadVersion;
508 * Get a copy of the remote partition's XPC variables from the reserved page.
510 * remote_vars points to a buffer that is cacheline aligned for BTE copies and
511 * assumed to be of size XPC_RP_VARS_SIZE.
513 static enum xpc_retval
514 xpc_get_remote_vars(u64 remote_vars_pa, struct xpc_vars *remote_vars)
519 if (remote_vars_pa == 0) {
520 return xpcVarsNotSet;
524 /* pull over the cross partition variables */
526 bres = xp_bte_copy(remote_vars_pa, ia64_tpa((u64) remote_vars),
528 (BTE_NOTIFY | BTE_WACQUIRE), NULL);
529 if (bres != BTE_SUCCESS) {
530 return xpc_map_bte_errors(bres);
533 if (XPC_VERSION_MAJOR(remote_vars->version) !=
534 XPC_VERSION_MAJOR(XPC_V_VERSION)) {
535 return xpcBadVersion;
543 * Update the remote partition's info.
546 xpc_update_partition_info(struct xpc_partition *part, u8 remote_rp_version,
547 struct timespec *remote_rp_stamp, u64 remote_rp_pa,
548 u64 remote_vars_pa, struct xpc_vars *remote_vars)
550 part->remote_rp_version = remote_rp_version;
551 dev_dbg(xpc_part, " remote_rp_version = 0x%016lx\n",
552 part->remote_rp_version);
554 part->remote_rp_stamp = *remote_rp_stamp;
555 dev_dbg(xpc_part, " remote_rp_stamp (tv_sec = 0x%lx tv_nsec = 0x%lx\n",
556 part->remote_rp_stamp.tv_sec, part->remote_rp_stamp.tv_nsec);
558 part->remote_rp_pa = remote_rp_pa;
559 dev_dbg(xpc_part, " remote_rp_pa = 0x%016lx\n", part->remote_rp_pa);
561 part->remote_vars_pa = remote_vars_pa;
562 dev_dbg(xpc_part, " remote_vars_pa = 0x%016lx\n",
563 part->remote_vars_pa);
565 part->last_heartbeat = remote_vars->heartbeat;
566 dev_dbg(xpc_part, " last_heartbeat = 0x%016lx\n",
567 part->last_heartbeat);
569 part->remote_vars_part_pa = remote_vars->vars_part_pa;
570 dev_dbg(xpc_part, " remote_vars_part_pa = 0x%016lx\n",
571 part->remote_vars_part_pa);
573 part->remote_act_nasid = remote_vars->act_nasid;
574 dev_dbg(xpc_part, " remote_act_nasid = 0x%x\n",
575 part->remote_act_nasid);
577 part->remote_act_phys_cpuid = remote_vars->act_phys_cpuid;
578 dev_dbg(xpc_part, " remote_act_phys_cpuid = 0x%x\n",
579 part->remote_act_phys_cpuid);
581 part->remote_amos_page_pa = remote_vars->amos_page_pa;
582 dev_dbg(xpc_part, " remote_amos_page_pa = 0x%lx\n",
583 part->remote_amos_page_pa);
585 part->remote_vars_version = remote_vars->version;
586 dev_dbg(xpc_part, " remote_vars_version = 0x%x\n",
587 part->remote_vars_version);
592 * Prior code has determined the nasid which generated an IPI. Inspect
593 * that nasid to determine if its partition needs to be activated or
596 * A partition is consider "awaiting activation" if our partition
597 * flags indicate it is not active and it has a heartbeat. A
598 * partition is considered "awaiting deactivation" if our partition
599 * flags indicate it is active but it has no heartbeat or it is not
600 * sending its heartbeat to us.
602 * To determine the heartbeat, the remote nasid must have a properly
603 * initialized reserved page.
606 xpc_identify_act_IRQ_req(int nasid)
608 struct xpc_rsvd_page *remote_rp;
609 struct xpc_vars *remote_vars;
612 int remote_rp_version;
615 struct timespec remote_rp_stamp = { 0, 0 };
617 struct xpc_partition *part;
621 /* pull over the reserved page structure */
623 remote_rp = (struct xpc_rsvd_page *) xpc_remote_copy_buffer;
625 ret = xpc_get_remote_rp(nasid, NULL, remote_rp, &remote_rp_pa);
626 if (ret != xpcSuccess) {
627 dev_warn(xpc_part, "unable to get reserved page from nasid %d, "
628 "which sent interrupt, reason=%d\n", nasid, ret);
632 remote_vars_pa = remote_rp->vars_pa;
633 remote_rp_version = remote_rp->version;
634 if (XPC_SUPPORTS_RP_STAMP(remote_rp_version)) {
635 remote_rp_stamp = remote_rp->stamp;
637 partid = remote_rp->partid;
638 part = &xpc_partitions[partid];
641 /* pull over the cross partition variables */
643 remote_vars = (struct xpc_vars *) xpc_remote_copy_buffer;
645 ret = xpc_get_remote_vars(remote_vars_pa, remote_vars);
646 if (ret != xpcSuccess) {
648 dev_warn(xpc_part, "unable to get XPC variables from nasid %d, "
649 "which sent interrupt, reason=%d\n", nasid, ret);
651 XPC_DEACTIVATE_PARTITION(part, ret);
656 part->act_IRQ_rcvd++;
658 dev_dbg(xpc_part, "partid for nasid %d is %d; IRQs = %d; HB = "
659 "%ld:0x%lx\n", (int) nasid, (int) partid, part->act_IRQ_rcvd,
660 remote_vars->heartbeat, remote_vars->heartbeating_to_mask);
662 if (xpc_partition_disengaged(part) &&
663 part->act_state == XPC_P_INACTIVE) {
665 xpc_update_partition_info(part, remote_rp_version,
666 &remote_rp_stamp, remote_rp_pa,
667 remote_vars_pa, remote_vars);
669 if (XPC_SUPPORTS_DISENGAGE_REQUEST(part->remote_vars_version)) {
670 if (xpc_partition_disengage_requested(1UL << partid)) {
672 * Other side is waiting on us to disengage,
673 * even though we already have.
678 /* other side doesn't support disengage requests */
679 xpc_clear_partition_disengage_request(1UL << partid);
682 xpc_activate_partition(part);
686 DBUG_ON(part->remote_rp_version == 0);
687 DBUG_ON(part->remote_vars_version == 0);
689 if (!XPC_SUPPORTS_RP_STAMP(part->remote_rp_version)) {
690 DBUG_ON(XPC_SUPPORTS_DISENGAGE_REQUEST(part->
691 remote_vars_version));
693 if (!XPC_SUPPORTS_RP_STAMP(remote_rp_version)) {
694 DBUG_ON(XPC_SUPPORTS_DISENGAGE_REQUEST(remote_vars->
696 /* see if the other side rebooted */
697 if (part->remote_amos_page_pa ==
698 remote_vars->amos_page_pa &&
699 xpc_hb_allowed(sn_partition_id,
701 /* doesn't look that way, so ignore the IPI */
707 * Other side rebooted and previous XPC didn't support the
708 * disengage request, so we don't need to do anything special.
711 xpc_update_partition_info(part, remote_rp_version,
712 &remote_rp_stamp, remote_rp_pa,
713 remote_vars_pa, remote_vars);
714 part->reactivate_nasid = nasid;
715 XPC_DEACTIVATE_PARTITION(part, xpcReactivating);
719 DBUG_ON(!XPC_SUPPORTS_DISENGAGE_REQUEST(part->remote_vars_version));
721 if (!XPC_SUPPORTS_RP_STAMP(remote_rp_version)) {
722 DBUG_ON(!XPC_SUPPORTS_DISENGAGE_REQUEST(remote_vars->version));
725 * Other side rebooted and previous XPC did support the
726 * disengage request, but the new one doesn't.
729 xpc_clear_partition_engaged(1UL << partid);
730 xpc_clear_partition_disengage_request(1UL << partid);
732 xpc_update_partition_info(part, remote_rp_version,
733 &remote_rp_stamp, remote_rp_pa,
734 remote_vars_pa, remote_vars);
738 DBUG_ON(!XPC_SUPPORTS_DISENGAGE_REQUEST(remote_vars->version));
740 stamp_diff = xpc_compare_stamps(&part->remote_rp_stamp,
742 if (stamp_diff != 0) {
743 DBUG_ON(stamp_diff >= 0);
746 * Other side rebooted and the previous XPC did support
747 * the disengage request, as does the new one.
750 DBUG_ON(xpc_partition_engaged(1UL << partid));
751 DBUG_ON(xpc_partition_disengage_requested(1UL <<
754 xpc_update_partition_info(part, remote_rp_version,
755 &remote_rp_stamp, remote_rp_pa,
756 remote_vars_pa, remote_vars);
761 if (!xpc_partition_disengaged(part)) {
762 /* still waiting on other side to disengage from us */
767 part->reactivate_nasid = nasid;
768 XPC_DEACTIVATE_PARTITION(part, xpcReactivating);
770 } else if (XPC_SUPPORTS_DISENGAGE_REQUEST(part->remote_vars_version) &&
771 xpc_partition_disengage_requested(1UL << partid)) {
772 XPC_DEACTIVATE_PARTITION(part, xpcOtherGoingDown);
778 * Loop through the activation AMO variables and process any bits
779 * which are set. Each bit indicates a nasid sending a partition
780 * activation or deactivation request.
782 * Return #of IRQs detected.
785 xpc_identify_act_IRQ_sender(void)
789 u64 nasid; /* remote nasid */
790 int n_IRQs_detected = 0;
794 act_amos = xpc_vars->amos_page + XPC_ACTIVATE_IRQ_AMOS;
797 /* scan through act AMO variable looking for non-zero entries */
798 for (word = 0; word < xp_nasid_mask_words; word++) {
804 nasid_mask = xpc_IPI_receive(&act_amos[word]);
805 if (nasid_mask == 0) {
806 /* no IRQs from nasids in this variable */
810 dev_dbg(xpc_part, "AMO[%d] gave back 0x%lx\n", word,
815 * If this nasid has been added to the machine since
816 * our partition was reset, this will retain the
817 * remote nasid in our reserved pages machine mask.
818 * This is used in the event of module reload.
820 xpc_mach_nasids[word] |= nasid_mask;
823 /* locate the nasid(s) which sent interrupts */
825 for (bit = 0; bit < (8 * sizeof(u64)); bit++) {
826 if (nasid_mask & (1UL << bit)) {
828 nasid = XPC_NASID_FROM_W_B(word, bit);
829 dev_dbg(xpc_part, "interrupt from nasid %ld\n",
831 xpc_identify_act_IRQ_req(nasid);
835 return n_IRQs_detected;
840 * See if the other side has responded to a partition disengage request
844 xpc_partition_disengaged(struct xpc_partition *part)
846 partid_t partid = XPC_PARTID(part);
850 disengaged = (xpc_partition_engaged(1UL << partid) == 0);
851 if (part->disengage_request_timeout) {
853 if (jiffies < part->disengage_request_timeout) {
854 /* timelimit hasn't been reached yet */
859 * Other side hasn't responded to our disengage
860 * request in a timely fashion, so assume it's dead.
863 xpc_clear_partition_engaged(1UL << partid);
866 part->disengage_request_timeout = 0;
868 /* cancel the timer function, provided it's not us */
869 if (!in_interrupt()) {
870 del_singleshot_timer_sync(&part->
871 disengage_request_timer);
874 DBUG_ON(part->act_state != XPC_P_DEACTIVATING &&
875 part->act_state != XPC_P_INACTIVE);
876 if (part->act_state != XPC_P_INACTIVE) {
877 xpc_wakeup_channel_mgr(part);
880 if (XPC_SUPPORTS_DISENGAGE_REQUEST(part->remote_vars_version)) {
881 xpc_cancel_partition_disengage_request(part);
889 * Mark specified partition as active.
892 xpc_mark_partition_active(struct xpc_partition *part)
894 unsigned long irq_flags;
898 dev_dbg(xpc_part, "setting partition %d to ACTIVE\n", XPC_PARTID(part));
900 spin_lock_irqsave(&part->act_lock, irq_flags);
901 if (part->act_state == XPC_P_ACTIVATING) {
902 part->act_state = XPC_P_ACTIVE;
905 DBUG_ON(part->reason == xpcSuccess);
908 spin_unlock_irqrestore(&part->act_lock, irq_flags);
915 * Notify XPC that the partition is down.
918 xpc_deactivate_partition(const int line, struct xpc_partition *part,
919 enum xpc_retval reason)
921 unsigned long irq_flags;
924 spin_lock_irqsave(&part->act_lock, irq_flags);
926 if (part->act_state == XPC_P_INACTIVE) {
927 XPC_SET_REASON(part, reason, line);
928 spin_unlock_irqrestore(&part->act_lock, irq_flags);
929 if (reason == xpcReactivating) {
930 /* we interrupt ourselves to reactivate partition */
931 xpc_IPI_send_reactivate(part);
935 if (part->act_state == XPC_P_DEACTIVATING) {
936 if ((part->reason == xpcUnloading && reason != xpcUnloading) ||
937 reason == xpcReactivating) {
938 XPC_SET_REASON(part, reason, line);
940 spin_unlock_irqrestore(&part->act_lock, irq_flags);
944 part->act_state = XPC_P_DEACTIVATING;
945 XPC_SET_REASON(part, reason, line);
947 spin_unlock_irqrestore(&part->act_lock, irq_flags);
949 if (XPC_SUPPORTS_DISENGAGE_REQUEST(part->remote_vars_version)) {
950 xpc_request_partition_disengage(part);
951 xpc_IPI_send_disengage(part);
953 /* set a timelimit on the disengage request */
954 part->disengage_request_timeout = jiffies +
955 (xpc_disengage_request_timelimit * HZ);
956 part->disengage_request_timer.expires =
957 part->disengage_request_timeout;
958 add_timer(&part->disengage_request_timer);
961 dev_dbg(xpc_part, "bringing partition %d down, reason = %d\n",
962 XPC_PARTID(part), reason);
964 xpc_partition_going_down(part, reason);
969 * Mark specified partition as inactive.
972 xpc_mark_partition_inactive(struct xpc_partition *part)
974 unsigned long irq_flags;
977 dev_dbg(xpc_part, "setting partition %d to INACTIVE\n",
980 spin_lock_irqsave(&part->act_lock, irq_flags);
981 part->act_state = XPC_P_INACTIVE;
982 spin_unlock_irqrestore(&part->act_lock, irq_flags);
983 part->remote_rp_pa = 0;
988 * SAL has provided a partition and machine mask. The partition mask
989 * contains a bit for each even nasid in our partition. The machine
990 * mask contains a bit for each even nasid in the entire machine.
992 * Using those two bit arrays, we can determine which nasids are
993 * known in the machine. Each should also have a reserved page
994 * initialized if they are available for partitioning.
999 void *remote_rp_base;
1000 struct xpc_rsvd_page *remote_rp;
1001 struct xpc_vars *remote_vars;
1008 struct xpc_rsvd_page *rp;
1010 struct xpc_partition *part;
1011 u64 *discovered_nasids;
1012 enum xpc_retval ret;
1015 remote_rp = xpc_kmalloc_cacheline_aligned(XPC_RP_HEADER_SIZE +
1016 xp_nasid_mask_bytes,
1017 GFP_KERNEL, &remote_rp_base);
1018 if (remote_rp == NULL) {
1021 remote_vars = (struct xpc_vars *) remote_rp;
1024 discovered_nasids = kmalloc(sizeof(u64) * xp_nasid_mask_words,
1026 if (discovered_nasids == NULL) {
1027 kfree(remote_rp_base);
1030 memset(discovered_nasids, 0, sizeof(u64) * xp_nasid_mask_words);
1032 rp = (struct xpc_rsvd_page *) xpc_rsvd_page;
1035 * The term 'region' in this context refers to the minimum number of
1036 * nodes that can comprise an access protection grouping. The access
1037 * protection is in regards to memory, IOI and IPI.
1040 region_size = sn_region_size;
1042 switch (region_size) {
1050 DBUG_ON(!is_shub2());
1053 for (region = 0; region < max_regions; region++) {
1055 if ((volatile int) xpc_exiting) {
1059 dev_dbg(xpc_part, "searching region %d\n", region);
1061 for (nasid = (region * region_size * 2);
1062 nasid < ((region + 1) * region_size * 2);
1065 if ((volatile int) xpc_exiting) {
1069 dev_dbg(xpc_part, "checking nasid %d\n", nasid);
1072 if (XPC_NASID_IN_ARRAY(nasid, xpc_part_nasids)) {
1073 dev_dbg(xpc_part, "PROM indicates Nasid %d is "
1074 "part of the local partition; skipping "
1079 if (!(XPC_NASID_IN_ARRAY(nasid, xpc_mach_nasids))) {
1080 dev_dbg(xpc_part, "PROM indicates Nasid %d was "
1081 "not on Numa-Link network at reset\n",
1086 if (XPC_NASID_IN_ARRAY(nasid, discovered_nasids)) {
1087 dev_dbg(xpc_part, "Nasid %d is part of a "
1088 "partition which was previously "
1089 "discovered\n", nasid);
1094 /* pull over the reserved page structure */
1096 ret = xpc_get_remote_rp(nasid, discovered_nasids,
1097 remote_rp, &remote_rp_pa);
1098 if (ret != xpcSuccess) {
1099 dev_dbg(xpc_part, "unable to get reserved page "
1100 "from nasid %d, reason=%d\n", nasid,
1103 if (ret == xpcLocalPartid) {
1109 remote_vars_pa = remote_rp->vars_pa;
1111 partid = remote_rp->partid;
1112 part = &xpc_partitions[partid];
1115 /* pull over the cross partition variables */
1117 ret = xpc_get_remote_vars(remote_vars_pa, remote_vars);
1118 if (ret != xpcSuccess) {
1119 dev_dbg(xpc_part, "unable to get XPC variables "
1120 "from nasid %d, reason=%d\n", nasid,
1123 XPC_DEACTIVATE_PARTITION(part, ret);
1127 if (part->act_state != XPC_P_INACTIVE) {
1128 dev_dbg(xpc_part, "partition %d on nasid %d is "
1129 "already activating\n", partid, nasid);
1134 * Register the remote partition's AMOs with SAL so it
1135 * can handle and cleanup errors within that address
1136 * range should the remote partition go down. We don't
1137 * unregister this range because it is difficult to
1138 * tell when outstanding writes to the remote partition
1139 * are finished and thus when it is thus safe to
1140 * unregister. This should not result in wasted space
1141 * in the SAL xp_addr_region table because we should
1142 * get the same page for remote_act_amos_pa after
1143 * module reloads and system reboots.
1145 if (sn_register_xp_addr_region(
1146 remote_vars->amos_page_pa,
1147 PAGE_SIZE, 1) < 0) {
1148 dev_dbg(xpc_part, "partition %d failed to "
1149 "register xp_addr region 0x%016lx\n",
1150 partid, remote_vars->amos_page_pa);
1152 XPC_SET_REASON(part, xpcPhysAddrRegFailed,
1158 * The remote nasid is valid and available.
1159 * Send an interrupt to that nasid to notify
1160 * it that we are ready to begin activation.
1162 dev_dbg(xpc_part, "sending an interrupt to AMO 0x%lx, "
1163 "nasid %d, phys_cpuid 0x%x\n",
1164 remote_vars->amos_page_pa,
1165 remote_vars->act_nasid,
1166 remote_vars->act_phys_cpuid);
1168 if (XPC_SUPPORTS_DISENGAGE_REQUEST(remote_vars->
1170 part->remote_amos_page_pa =
1171 remote_vars->amos_page_pa;
1172 xpc_mark_partition_disengaged(part);
1173 xpc_cancel_partition_disengage_request(part);
1175 xpc_IPI_send_activate(remote_vars);
1179 kfree(discovered_nasids);
1180 kfree(remote_rp_base);
1185 * Given a partid, get the nasids owned by that partition from the
1186 * remote partition's reserved page.
1189 xpc_initiate_partid_to_nasids(partid_t partid, void *nasid_mask)
1191 struct xpc_partition *part;
1196 part = &xpc_partitions[partid];
1197 if (part->remote_rp_pa == 0) {
1198 return xpcPartitionDown;
1201 memset(nasid_mask, 0, XP_NASID_MASK_BYTES);
1203 part_nasid_pa = (u64) XPC_RP_PART_NASIDS(part->remote_rp_pa);
1205 bte_res = xp_bte_copy(part_nasid_pa, ia64_tpa((u64) nasid_mask),
1206 xp_nasid_mask_bytes, (BTE_NOTIFY | BTE_WACQUIRE), NULL);
1208 return xpc_map_bte_errors(bte_res);