perf tools: Fix find tids routine by excluding "." and ".."
[cris-mirror.git] / arch / powerpc / platforms / pseries / lpar.c
blobcf79b46d8f885539d28bc4ca9bcb4f766a2e107c
1 /*
2 * pSeries_lpar.c
3 * Copyright (C) 2001 Todd Inglett, IBM Corporation
5 * pSeries LPAR support.
6 *
7 * This program is free software; you can redistribute it and/or modify
8 * it under the terms of the GNU General Public License as published by
9 * the Free Software Foundation; either version 2 of the License, or
10 * (at your option) any later version.
12 * This program is distributed in the hope that it will be useful,
13 * but WITHOUT ANY WARRANTY; without even the implied warranty of
14 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
15 * GNU General Public License for more details.
17 * You should have received a copy of the GNU General Public License
18 * along with this program; if not, write to the Free Software
19 * Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA
22 /* Enables debugging of low-level hash table routines - careful! */
23 #undef DEBUG
25 #include <linux/kernel.h>
26 #include <linux/dma-mapping.h>
27 #include <linux/console.h>
28 #include <asm/processor.h>
29 #include <asm/mmu.h>
30 #include <asm/page.h>
31 #include <asm/pgtable.h>
32 #include <asm/machdep.h>
33 #include <asm/abs_addr.h>
34 #include <asm/mmu_context.h>
35 #include <asm/iommu.h>
36 #include <asm/tlbflush.h>
37 #include <asm/tlb.h>
38 #include <asm/prom.h>
39 #include <asm/cputable.h>
40 #include <asm/udbg.h>
41 #include <asm/smp.h>
42 #include <asm/trace.h>
44 #include "plpar_wrappers.h"
45 #include "pseries.h"
48 /* in hvCall.S */
49 EXPORT_SYMBOL(plpar_hcall);
50 EXPORT_SYMBOL(plpar_hcall9);
51 EXPORT_SYMBOL(plpar_hcall_norets);
53 extern void pSeries_find_serial_port(void);
56 static int vtermno; /* virtual terminal# for udbg */
58 #define __ALIGNED__ __attribute__((__aligned__(sizeof(long))))
59 static void udbg_hvsi_putc(char c)
61 /* packet's seqno isn't used anyways */
62 uint8_t packet[] __ALIGNED__ = { 0xff, 5, 0, 0, c };
63 int rc;
65 if (c == '\n')
66 udbg_hvsi_putc('\r');
68 do {
69 rc = plpar_put_term_char(vtermno, sizeof(packet), packet);
70 } while (rc == H_BUSY);
73 static long hvsi_udbg_buf_len;
74 static uint8_t hvsi_udbg_buf[256];
76 static int udbg_hvsi_getc_poll(void)
78 unsigned char ch;
79 int rc, i;
81 if (hvsi_udbg_buf_len == 0) {
82 rc = plpar_get_term_char(vtermno, &hvsi_udbg_buf_len, hvsi_udbg_buf);
83 if (rc != H_SUCCESS || hvsi_udbg_buf[0] != 0xff) {
84 /* bad read or non-data packet */
85 hvsi_udbg_buf_len = 0;
86 } else {
87 /* remove the packet header */
88 for (i = 4; i < hvsi_udbg_buf_len; i++)
89 hvsi_udbg_buf[i-4] = hvsi_udbg_buf[i];
90 hvsi_udbg_buf_len -= 4;
94 if (hvsi_udbg_buf_len <= 0 || hvsi_udbg_buf_len > 256) {
95 /* no data ready */
96 hvsi_udbg_buf_len = 0;
97 return -1;
100 ch = hvsi_udbg_buf[0];
101 /* shift remaining data down */
102 for (i = 1; i < hvsi_udbg_buf_len; i++) {
103 hvsi_udbg_buf[i-1] = hvsi_udbg_buf[i];
105 hvsi_udbg_buf_len--;
107 return ch;
110 static int udbg_hvsi_getc(void)
112 int ch;
113 for (;;) {
114 ch = udbg_hvsi_getc_poll();
115 if (ch == -1) {
116 /* This shouldn't be needed...but... */
117 volatile unsigned long delay;
118 for (delay=0; delay < 2000000; delay++)
120 } else {
121 return ch;
126 static void udbg_putcLP(char c)
128 char buf[16];
129 unsigned long rc;
131 if (c == '\n')
132 udbg_putcLP('\r');
134 buf[0] = c;
135 do {
136 rc = plpar_put_term_char(vtermno, 1, buf);
137 } while(rc == H_BUSY);
140 /* Buffered chars getc */
141 static long inbuflen;
142 static long inbuf[2]; /* must be 2 longs */
144 static int udbg_getc_pollLP(void)
146 /* The interface is tricky because it may return up to 16 chars.
147 * We save them statically for future calls to udbg_getc().
149 char ch, *buf = (char *)inbuf;
150 int i;
151 long rc;
152 if (inbuflen == 0) {
153 /* get some more chars. */
154 inbuflen = 0;
155 rc = plpar_get_term_char(vtermno, &inbuflen, buf);
156 if (rc != H_SUCCESS)
157 inbuflen = 0; /* otherwise inbuflen is garbage */
159 if (inbuflen <= 0 || inbuflen > 16) {
160 /* Catch error case as well as other oddities (corruption) */
161 inbuflen = 0;
162 return -1;
164 ch = buf[0];
165 for (i = 1; i < inbuflen; i++) /* shuffle them down. */
166 buf[i-1] = buf[i];
167 inbuflen--;
168 return ch;
171 static int udbg_getcLP(void)
173 int ch;
174 for (;;) {
175 ch = udbg_getc_pollLP();
176 if (ch == -1) {
177 /* This shouldn't be needed...but... */
178 volatile unsigned long delay;
179 for (delay=0; delay < 2000000; delay++)
181 } else {
182 return ch;
187 /* call this from early_init() for a working debug console on
188 * vterm capable LPAR machines
190 void __init udbg_init_debug_lpar(void)
192 vtermno = 0;
193 udbg_putc = udbg_putcLP;
194 udbg_getc = udbg_getcLP;
195 udbg_getc_poll = udbg_getc_pollLP;
197 register_early_udbg_console();
200 /* returns 0 if couldn't find or use /chosen/stdout as console */
201 void __init find_udbg_vterm(void)
203 struct device_node *stdout_node;
204 const u32 *termno;
205 const char *name;
207 /* find the boot console from /chosen/stdout */
208 if (!of_chosen)
209 return;
210 name = of_get_property(of_chosen, "linux,stdout-path", NULL);
211 if (name == NULL)
212 return;
213 stdout_node = of_find_node_by_path(name);
214 if (!stdout_node)
215 return;
216 name = of_get_property(stdout_node, "name", NULL);
217 if (!name) {
218 printk(KERN_WARNING "stdout node missing 'name' property!\n");
219 goto out;
222 /* Check if it's a virtual terminal */
223 if (strncmp(name, "vty", 3) != 0)
224 goto out;
225 termno = of_get_property(stdout_node, "reg", NULL);
226 if (termno == NULL)
227 goto out;
228 vtermno = termno[0];
230 if (of_device_is_compatible(stdout_node, "hvterm1")) {
231 udbg_putc = udbg_putcLP;
232 udbg_getc = udbg_getcLP;
233 udbg_getc_poll = udbg_getc_pollLP;
234 add_preferred_console("hvc", termno[0] & 0xff, NULL);
235 } else if (of_device_is_compatible(stdout_node, "hvterm-protocol")) {
236 vtermno = termno[0];
237 udbg_putc = udbg_hvsi_putc;
238 udbg_getc = udbg_hvsi_getc;
239 udbg_getc_poll = udbg_hvsi_getc_poll;
240 add_preferred_console("hvsi", termno[0] & 0xff, NULL);
242 out:
243 of_node_put(stdout_node);
246 void vpa_init(int cpu)
248 int hwcpu = get_hard_smp_processor_id(cpu);
249 unsigned long addr;
250 long ret;
252 if (cpu_has_feature(CPU_FTR_ALTIVEC))
253 lppaca[cpu].vmxregs_in_use = 1;
255 addr = __pa(&lppaca[cpu]);
256 ret = register_vpa(hwcpu, addr);
258 if (ret) {
259 printk(KERN_ERR "WARNING: vpa_init: VPA registration for "
260 "cpu %d (hw %d) of area %lx returns %ld\n",
261 cpu, hwcpu, addr, ret);
262 return;
265 * PAPR says this feature is SLB-Buffer but firmware never
266 * reports that. All SPLPAR support SLB shadow buffer.
268 addr = __pa(&slb_shadow[cpu]);
269 if (firmware_has_feature(FW_FEATURE_SPLPAR)) {
270 ret = register_slb_shadow(hwcpu, addr);
271 if (ret)
272 printk(KERN_ERR
273 "WARNING: vpa_init: SLB shadow buffer "
274 "registration for cpu %d (hw %d) of area %lx "
275 "returns %ld\n", cpu, hwcpu, addr, ret);
279 static long pSeries_lpar_hpte_insert(unsigned long hpte_group,
280 unsigned long va, unsigned long pa,
281 unsigned long rflags, unsigned long vflags,
282 int psize, int ssize)
284 unsigned long lpar_rc;
285 unsigned long flags;
286 unsigned long slot;
287 unsigned long hpte_v, hpte_r;
289 if (!(vflags & HPTE_V_BOLTED))
290 pr_devel("hpte_insert(group=%lx, va=%016lx, pa=%016lx, "
291 "rflags=%lx, vflags=%lx, psize=%d)\n",
292 hpte_group, va, pa, rflags, vflags, psize);
294 hpte_v = hpte_encode_v(va, psize, ssize) | vflags | HPTE_V_VALID;
295 hpte_r = hpte_encode_r(pa, psize) | rflags;
297 if (!(vflags & HPTE_V_BOLTED))
298 pr_devel(" hpte_v=%016lx, hpte_r=%016lx\n", hpte_v, hpte_r);
300 /* Now fill in the actual HPTE */
301 /* Set CEC cookie to 0 */
302 /* Zero page = 0 */
303 /* I-cache Invalidate = 0 */
304 /* I-cache synchronize = 0 */
305 /* Exact = 0 */
306 flags = 0;
308 /* Make pHyp happy */
309 if ((rflags & _PAGE_NO_CACHE) & !(rflags & _PAGE_WRITETHRU))
310 hpte_r &= ~_PAGE_COHERENT;
312 lpar_rc = plpar_pte_enter(flags, hpte_group, hpte_v, hpte_r, &slot);
313 if (unlikely(lpar_rc == H_PTEG_FULL)) {
314 if (!(vflags & HPTE_V_BOLTED))
315 pr_devel(" full\n");
316 return -1;
320 * Since we try and ioremap PHBs we don't own, the pte insert
321 * will fail. However we must catch the failure in hash_page
322 * or we will loop forever, so return -2 in this case.
324 if (unlikely(lpar_rc != H_SUCCESS)) {
325 if (!(vflags & HPTE_V_BOLTED))
326 pr_devel(" lpar err %lu\n", lpar_rc);
327 return -2;
329 if (!(vflags & HPTE_V_BOLTED))
330 pr_devel(" -> slot: %lu\n", slot & 7);
332 /* Because of iSeries, we have to pass down the secondary
333 * bucket bit here as well
335 return (slot & 7) | (!!(vflags & HPTE_V_SECONDARY) << 3);
338 static DEFINE_SPINLOCK(pSeries_lpar_tlbie_lock);
340 static long pSeries_lpar_hpte_remove(unsigned long hpte_group)
342 unsigned long slot_offset;
343 unsigned long lpar_rc;
344 int i;
345 unsigned long dummy1, dummy2;
347 /* pick a random slot to start at */
348 slot_offset = mftb() & 0x7;
350 for (i = 0; i < HPTES_PER_GROUP; i++) {
352 /* don't remove a bolted entry */
353 lpar_rc = plpar_pte_remove(H_ANDCOND, hpte_group + slot_offset,
354 (0x1UL << 4), &dummy1, &dummy2);
355 if (lpar_rc == H_SUCCESS)
356 return i;
357 BUG_ON(lpar_rc != H_NOT_FOUND);
359 slot_offset++;
360 slot_offset &= 0x7;
363 return -1;
366 static void pSeries_lpar_hptab_clear(void)
368 unsigned long size_bytes = 1UL << ppc64_pft_size;
369 unsigned long hpte_count = size_bytes >> 4;
370 struct {
371 unsigned long pteh;
372 unsigned long ptel;
373 } ptes[4];
374 long lpar_rc;
375 int i, j;
377 /* Read in batches of 4,
378 * invalidate only valid entries not in the VRMA
379 * hpte_count will be a multiple of 4
381 for (i = 0; i < hpte_count; i += 4) {
382 lpar_rc = plpar_pte_read_4_raw(0, i, (void *)ptes);
383 if (lpar_rc != H_SUCCESS)
384 continue;
385 for (j = 0; j < 4; j++){
386 if ((ptes[j].pteh & HPTE_V_VRMA_MASK) ==
387 HPTE_V_VRMA_MASK)
388 continue;
389 if (ptes[j].pteh & HPTE_V_VALID)
390 plpar_pte_remove_raw(0, i + j, 0,
391 &(ptes[j].pteh), &(ptes[j].ptel));
397 * This computes the AVPN and B fields of the first dword of a HPTE,
398 * for use when we want to match an existing PTE. The bottom 7 bits
399 * of the returned value are zero.
401 static inline unsigned long hpte_encode_avpn(unsigned long va, int psize,
402 int ssize)
404 unsigned long v;
406 v = (va >> 23) & ~(mmu_psize_defs[psize].avpnm);
407 v <<= HPTE_V_AVPN_SHIFT;
408 v |= ((unsigned long) ssize) << HPTE_V_SSIZE_SHIFT;
409 return v;
413 * NOTE: for updatepp ops we are fortunate that the linux "newpp" bits and
414 * the low 3 bits of flags happen to line up. So no transform is needed.
415 * We can probably optimize here and assume the high bits of newpp are
416 * already zero. For now I am paranoid.
418 static long pSeries_lpar_hpte_updatepp(unsigned long slot,
419 unsigned long newpp,
420 unsigned long va,
421 int psize, int ssize, int local)
423 unsigned long lpar_rc;
424 unsigned long flags = (newpp & 7) | H_AVPN;
425 unsigned long want_v;
427 want_v = hpte_encode_avpn(va, psize, ssize);
429 pr_devel(" update: avpnv=%016lx, hash=%016lx, f=%lx, psize: %d ...",
430 want_v, slot, flags, psize);
432 lpar_rc = plpar_pte_protect(flags, slot, want_v);
434 if (lpar_rc == H_NOT_FOUND) {
435 pr_devel("not found !\n");
436 return -1;
439 pr_devel("ok\n");
441 BUG_ON(lpar_rc != H_SUCCESS);
443 return 0;
446 static unsigned long pSeries_lpar_hpte_getword0(unsigned long slot)
448 unsigned long dword0;
449 unsigned long lpar_rc;
450 unsigned long dummy_word1;
451 unsigned long flags;
453 /* Read 1 pte at a time */
454 /* Do not need RPN to logical page translation */
455 /* No cross CEC PFT access */
456 flags = 0;
458 lpar_rc = plpar_pte_read(flags, slot, &dword0, &dummy_word1);
460 BUG_ON(lpar_rc != H_SUCCESS);
462 return dword0;
465 static long pSeries_lpar_hpte_find(unsigned long va, int psize, int ssize)
467 unsigned long hash;
468 unsigned long i;
469 long slot;
470 unsigned long want_v, hpte_v;
472 hash = hpt_hash(va, mmu_psize_defs[psize].shift, ssize);
473 want_v = hpte_encode_avpn(va, psize, ssize);
475 /* Bolted entries are always in the primary group */
476 slot = (hash & htab_hash_mask) * HPTES_PER_GROUP;
477 for (i = 0; i < HPTES_PER_GROUP; i++) {
478 hpte_v = pSeries_lpar_hpte_getword0(slot);
480 if (HPTE_V_COMPARE(hpte_v, want_v) && (hpte_v & HPTE_V_VALID))
481 /* HPTE matches */
482 return slot;
483 ++slot;
486 return -1;
489 static void pSeries_lpar_hpte_updateboltedpp(unsigned long newpp,
490 unsigned long ea,
491 int psize, int ssize)
493 unsigned long lpar_rc, slot, vsid, va, flags;
495 vsid = get_kernel_vsid(ea, ssize);
496 va = hpt_va(ea, vsid, ssize);
498 slot = pSeries_lpar_hpte_find(va, psize, ssize);
499 BUG_ON(slot == -1);
501 flags = newpp & 7;
502 lpar_rc = plpar_pte_protect(flags, slot, 0);
504 BUG_ON(lpar_rc != H_SUCCESS);
507 static void pSeries_lpar_hpte_invalidate(unsigned long slot, unsigned long va,
508 int psize, int ssize, int local)
510 unsigned long want_v;
511 unsigned long lpar_rc;
512 unsigned long dummy1, dummy2;
514 pr_devel(" inval : slot=%lx, va=%016lx, psize: %d, local: %d\n",
515 slot, va, psize, local);
517 want_v = hpte_encode_avpn(va, psize, ssize);
518 lpar_rc = plpar_pte_remove(H_AVPN, slot, want_v, &dummy1, &dummy2);
519 if (lpar_rc == H_NOT_FOUND)
520 return;
522 BUG_ON(lpar_rc != H_SUCCESS);
525 static void pSeries_lpar_hpte_removebolted(unsigned long ea,
526 int psize, int ssize)
528 unsigned long slot, vsid, va;
530 vsid = get_kernel_vsid(ea, ssize);
531 va = hpt_va(ea, vsid, ssize);
533 slot = pSeries_lpar_hpte_find(va, psize, ssize);
534 BUG_ON(slot == -1);
536 pSeries_lpar_hpte_invalidate(slot, va, psize, ssize, 0);
539 /* Flag bits for H_BULK_REMOVE */
540 #define HBR_REQUEST 0x4000000000000000UL
541 #define HBR_RESPONSE 0x8000000000000000UL
542 #define HBR_END 0xc000000000000000UL
543 #define HBR_AVPN 0x0200000000000000UL
544 #define HBR_ANDCOND 0x0100000000000000UL
547 * Take a spinlock around flushes to avoid bouncing the hypervisor tlbie
548 * lock.
550 static void pSeries_lpar_flush_hash_range(unsigned long number, int local)
552 unsigned long i, pix, rc;
553 unsigned long flags = 0;
554 struct ppc64_tlb_batch *batch = &__get_cpu_var(ppc64_tlb_batch);
555 int lock_tlbie = !cpu_has_feature(CPU_FTR_LOCKLESS_TLBIE);
556 unsigned long param[9];
557 unsigned long va;
558 unsigned long hash, index, shift, hidx, slot;
559 real_pte_t pte;
560 int psize, ssize;
562 if (lock_tlbie)
563 spin_lock_irqsave(&pSeries_lpar_tlbie_lock, flags);
565 psize = batch->psize;
566 ssize = batch->ssize;
567 pix = 0;
568 for (i = 0; i < number; i++) {
569 va = batch->vaddr[i];
570 pte = batch->pte[i];
571 pte_iterate_hashed_subpages(pte, psize, va, index, shift) {
572 hash = hpt_hash(va, shift, ssize);
573 hidx = __rpte_to_hidx(pte, index);
574 if (hidx & _PTEIDX_SECONDARY)
575 hash = ~hash;
576 slot = (hash & htab_hash_mask) * HPTES_PER_GROUP;
577 slot += hidx & _PTEIDX_GROUP_IX;
578 if (!firmware_has_feature(FW_FEATURE_BULK_REMOVE)) {
579 pSeries_lpar_hpte_invalidate(slot, va, psize,
580 ssize, local);
581 } else {
582 param[pix] = HBR_REQUEST | HBR_AVPN | slot;
583 param[pix+1] = hpte_encode_avpn(va, psize,
584 ssize);
585 pix += 2;
586 if (pix == 8) {
587 rc = plpar_hcall9(H_BULK_REMOVE, param,
588 param[0], param[1], param[2],
589 param[3], param[4], param[5],
590 param[6], param[7]);
591 BUG_ON(rc != H_SUCCESS);
592 pix = 0;
595 } pte_iterate_hashed_end();
597 if (pix) {
598 param[pix] = HBR_END;
599 rc = plpar_hcall9(H_BULK_REMOVE, param, param[0], param[1],
600 param[2], param[3], param[4], param[5],
601 param[6], param[7]);
602 BUG_ON(rc != H_SUCCESS);
605 if (lock_tlbie)
606 spin_unlock_irqrestore(&pSeries_lpar_tlbie_lock, flags);
609 void __init hpte_init_lpar(void)
611 ppc_md.hpte_invalidate = pSeries_lpar_hpte_invalidate;
612 ppc_md.hpte_updatepp = pSeries_lpar_hpte_updatepp;
613 ppc_md.hpte_updateboltedpp = pSeries_lpar_hpte_updateboltedpp;
614 ppc_md.hpte_insert = pSeries_lpar_hpte_insert;
615 ppc_md.hpte_remove = pSeries_lpar_hpte_remove;
616 ppc_md.hpte_removebolted = pSeries_lpar_hpte_removebolted;
617 ppc_md.flush_hash_range = pSeries_lpar_flush_hash_range;
618 ppc_md.hpte_clear_all = pSeries_lpar_hptab_clear;
621 #ifdef CONFIG_PPC_SMLPAR
622 #define CMO_FREE_HINT_DEFAULT 1
623 static int cmo_free_hint_flag = CMO_FREE_HINT_DEFAULT;
625 static int __init cmo_free_hint(char *str)
627 char *parm;
628 parm = strstrip(str);
630 if (strcasecmp(parm, "no") == 0 || strcasecmp(parm, "off") == 0) {
631 printk(KERN_INFO "cmo_free_hint: CMO free page hinting is not active.\n");
632 cmo_free_hint_flag = 0;
633 return 1;
636 cmo_free_hint_flag = 1;
637 printk(KERN_INFO "cmo_free_hint: CMO free page hinting is active.\n");
639 if (strcasecmp(parm, "yes") == 0 || strcasecmp(parm, "on") == 0)
640 return 1;
642 return 0;
645 __setup("cmo_free_hint=", cmo_free_hint);
647 static void pSeries_set_page_state(struct page *page, int order,
648 unsigned long state)
650 int i, j;
651 unsigned long cmo_page_sz, addr;
653 cmo_page_sz = cmo_get_page_size();
654 addr = __pa((unsigned long)page_address(page));
656 for (i = 0; i < (1 << order); i++, addr += PAGE_SIZE) {
657 for (j = 0; j < PAGE_SIZE; j += cmo_page_sz)
658 plpar_hcall_norets(H_PAGE_INIT, state, addr + j, 0);
662 void arch_free_page(struct page *page, int order)
664 if (!cmo_free_hint_flag || !firmware_has_feature(FW_FEATURE_CMO))
665 return;
667 pSeries_set_page_state(page, order, H_PAGE_SET_UNUSED);
669 EXPORT_SYMBOL(arch_free_page);
671 #endif
673 #ifdef CONFIG_TRACEPOINTS
675 * We optimise our hcall path by placing hcall_tracepoint_refcount
676 * directly in the TOC so we can check if the hcall tracepoints are
677 * enabled via a single load.
680 /* NB: reg/unreg are called while guarded with the tracepoints_mutex */
681 extern long hcall_tracepoint_refcount;
683 void hcall_tracepoint_regfunc(void)
685 hcall_tracepoint_refcount++;
688 void hcall_tracepoint_unregfunc(void)
690 hcall_tracepoint_refcount--;
693 void __trace_hcall_entry(unsigned long opcode, unsigned long *args)
695 trace_hcall_entry(opcode, args);
698 void __trace_hcall_exit(long opcode, unsigned long retval,
699 unsigned long *retbuf)
701 trace_hcall_exit(opcode, retval, retbuf);
703 #endif