aboutsummaryrefslogtreecommitdiffstats
path: root/arch/ia64
diff options
context:
space:
mode:
authorLinus Torvalds <torvalds@linux-foundation.org>2008-04-18 12:44:55 -0400
committerLinus Torvalds <torvalds@linux-foundation.org>2008-04-18 12:44:55 -0400
commit4786b4ee22de6304e841b12ee22b849230d7fba3 (patch)
tree08793b8fbcd63204d5d3355ac755745adcfef170 /arch/ia64
parent253ba4e79edc695b2925bd2ef34de06ff4d4070c (diff)
parent71b264f85ff50c14fe945ffff06ae0d5e9a9124e (diff)
Merge branch 'release' of git://git.kernel.org/pub/scm/linux/kernel/git/aegl/linux-2.6
* 'release' of git://git.kernel.org/pub/scm/linux/kernel/git/aegl/linux-2.6: (27 commits) [IA64] kdump: Add crash_save_vmcoreinfo for INIT [IA64] Fix NUMA configuration issue [IA64] Itanium Spec updates [IA64] Untangle sync_icache_dcache() page size determination [IA64] arch/ia64/kernel/: use time_* macros [IA64] remove redundant display of free swap space in show_mem() [IA64] make IOMMU respect the segment boundary limits [IA64] kprobes: kprobe-booster for ia64 [IA64] fix getpid and set_tid_address fast system calls for pid namespaces [IA64] Replace explicit jiffies tests with time_* macros. [IA64] use goto to jump out do/while_each_thread [IA64] Fix unlock ordering in smp_callin [IA64] pgd_offset() constfication. [IA64] kdump: crash.c coding style fix [IA64] kdump: add kdump_on_fatal_mca [IA64] Minimize per_cpu reservations. [IA64] Correct pernodesize calculation. [IA64] Kernel parameter for max number of concurrent global TLB purges [IA64] Multiple outstanding ptc.g instruction support [IA64] Implement smp_call_function_mask for ia64 ...
Diffstat (limited to 'arch/ia64')
-rw-r--r--arch/ia64/Kconfig14
-rw-r--r--arch/ia64/hp/common/sba_iommu.c56
-rw-r--r--arch/ia64/ia32/elfcore32.h14
-rw-r--r--arch/ia64/ia32/sys_ia32.c649
-rw-r--r--arch/ia64/kernel/acpi.c4
-rw-r--r--arch/ia64/kernel/asm-offsets.c13
-rw-r--r--arch/ia64/kernel/crash.c56
-rw-r--r--arch/ia64/kernel/efi.c46
-rw-r--r--arch/ia64/kernel/entry.S65
-rw-r--r--arch/ia64/kernel/fsys.S88
-rw-r--r--arch/ia64/kernel/head.S20
-rw-r--r--arch/ia64/kernel/irq_ia64.c2
-rw-r--r--arch/ia64/kernel/ivt.S69
-rw-r--r--arch/ia64/kernel/kprobes.c133
-rw-r--r--arch/ia64/kernel/mca.c60
-rw-r--r--arch/ia64/kernel/mca_asm.S5
-rw-r--r--arch/ia64/kernel/minstate.h14
-rw-r--r--arch/ia64/kernel/numa.c2
-rw-r--r--arch/ia64/kernel/patch.c8
-rw-r--r--arch/ia64/kernel/perfmon.c4
-rw-r--r--arch/ia64/kernel/process.c30
-rw-r--r--arch/ia64/kernel/ptrace.c1217
-rw-r--r--arch/ia64/kernel/setup.c31
-rw-r--r--arch/ia64/kernel/smp.c82
-rw-r--r--arch/ia64/kernel/smpboot.c2
-rw-r--r--arch/ia64/kernel/time.c78
-rw-r--r--arch/ia64/kernel/unaligned.c3
-rw-r--r--arch/ia64/mm/contig.c4
-rw-r--r--arch/ia64/mm/discontig.c17
-rw-r--r--arch/ia64/mm/init.c14
-rw-r--r--arch/ia64/mm/numa.c4
-rw-r--r--arch/ia64/mm/tlb.c357
-rw-r--r--arch/ia64/sn/kernel/xpc_main.c8
-rw-r--r--arch/ia64/sn/kernel/xpc_partition.c2
34 files changed, 2645 insertions, 526 deletions
diff --git a/arch/ia64/Kconfig b/arch/ia64/Kconfig
index 8fa3faf5ef1b..ed21737a00c5 100644
--- a/arch/ia64/Kconfig
+++ b/arch/ia64/Kconfig
@@ -283,6 +283,17 @@ config FORCE_MAX_ZONEORDER
283 default "17" if HUGETLB_PAGE 283 default "17" if HUGETLB_PAGE
284 default "11" 284 default "11"
285 285
286config VIRT_CPU_ACCOUNTING
287 bool "Deterministic task and CPU time accounting"
288 default n
289 help
290 Select this option to enable more accurate task and CPU time
291 accounting. This is done by reading a CPU counter on each
292 kernel entry and exit and on transitions within the kernel
293 between system, softirq and hardirq state, so there is a
294 small performance impact.
295 If in doubt, say N here.
296
286config SMP 297config SMP
287 bool "Symmetric multi-processing support" 298 bool "Symmetric multi-processing support"
288 help 299 help
@@ -611,6 +622,9 @@ config IRQ_PER_CPU
611 bool 622 bool
612 default y 623 default y
613 624
625config IOMMU_HELPER
626 def_bool (IA64_HP_ZX1 || IA64_HP_ZX1_SWIOTLB || IA64_GENERIC)
627
614source "arch/ia64/hp/sim/Kconfig" 628source "arch/ia64/hp/sim/Kconfig"
615 629
616source "arch/ia64/Kconfig.debug" 630source "arch/ia64/Kconfig.debug"
diff --git a/arch/ia64/hp/common/sba_iommu.c b/arch/ia64/hp/common/sba_iommu.c
index 523eae6d3e49..9409de5c9441 100644
--- a/arch/ia64/hp/common/sba_iommu.c
+++ b/arch/ia64/hp/common/sba_iommu.c
@@ -35,6 +35,7 @@
35#include <linux/nodemask.h> 35#include <linux/nodemask.h>
36#include <linux/bitops.h> /* hweight64() */ 36#include <linux/bitops.h> /* hweight64() */
37#include <linux/crash_dump.h> 37#include <linux/crash_dump.h>
38#include <linux/iommu-helper.h>
38 39
39#include <asm/delay.h> /* ia64_get_itc() */ 40#include <asm/delay.h> /* ia64_get_itc() */
40#include <asm/io.h> 41#include <asm/io.h>
@@ -460,6 +461,13 @@ get_iovp_order (unsigned long size)
460 return order; 461 return order;
461} 462}
462 463
464static unsigned long ptr_to_pide(struct ioc *ioc, unsigned long *res_ptr,
465 unsigned int bitshiftcnt)
466{
467 return (((unsigned long)res_ptr - (unsigned long)ioc->res_map) << 3)
468 + bitshiftcnt;
469}
470
463/** 471/**
464 * sba_search_bitmap - find free space in IO PDIR resource bitmap 472 * sba_search_bitmap - find free space in IO PDIR resource bitmap
465 * @ioc: IO MMU structure which owns the pdir we are interested in. 473 * @ioc: IO MMU structure which owns the pdir we are interested in.
@@ -471,15 +479,25 @@ get_iovp_order (unsigned long size)
471 * Cool perf optimization: search for log2(size) bits at a time. 479 * Cool perf optimization: search for log2(size) bits at a time.
472 */ 480 */
473static SBA_INLINE unsigned long 481static SBA_INLINE unsigned long
474sba_search_bitmap(struct ioc *ioc, unsigned long bits_wanted, int use_hint) 482sba_search_bitmap(struct ioc *ioc, struct device *dev,
483 unsigned long bits_wanted, int use_hint)
475{ 484{
476 unsigned long *res_ptr; 485 unsigned long *res_ptr;
477 unsigned long *res_end = (unsigned long *) &(ioc->res_map[ioc->res_size]); 486 unsigned long *res_end = (unsigned long *) &(ioc->res_map[ioc->res_size]);
478 unsigned long flags, pide = ~0UL; 487 unsigned long flags, pide = ~0UL, tpide;
488 unsigned long boundary_size;
489 unsigned long shift;
490 int ret;
479 491
480 ASSERT(((unsigned long) ioc->res_hint & (sizeof(unsigned long) - 1UL)) == 0); 492 ASSERT(((unsigned long) ioc->res_hint & (sizeof(unsigned long) - 1UL)) == 0);
481 ASSERT(res_ptr < res_end); 493 ASSERT(res_ptr < res_end);
482 494
495 boundary_size = (unsigned long long)dma_get_seg_boundary(dev) + 1;
496 boundary_size = ALIGN(boundary_size, 1ULL << iovp_shift) >> iovp_shift;
497
498 BUG_ON(ioc->ibase & ~iovp_mask);
499 shift = ioc->ibase >> iovp_shift;
500
483 spin_lock_irqsave(&ioc->res_lock, flags); 501 spin_lock_irqsave(&ioc->res_lock, flags);
484 502
485 /* Allow caller to force a search through the entire resource space */ 503 /* Allow caller to force a search through the entire resource space */
@@ -504,9 +522,7 @@ sba_search_bitmap(struct ioc *ioc, unsigned long bits_wanted, int use_hint)
504 if (likely(*res_ptr != ~0UL)) { 522 if (likely(*res_ptr != ~0UL)) {
505 bitshiftcnt = ffz(*res_ptr); 523 bitshiftcnt = ffz(*res_ptr);
506 *res_ptr |= (1UL << bitshiftcnt); 524 *res_ptr |= (1UL << bitshiftcnt);
507 pide = ((unsigned long)res_ptr - (unsigned long)ioc->res_map); 525 pide = ptr_to_pide(ioc, res_ptr, bitshiftcnt);
508 pide <<= 3; /* convert to bit address */
509 pide += bitshiftcnt;
510 ioc->res_bitshift = bitshiftcnt + bits_wanted; 526 ioc->res_bitshift = bitshiftcnt + bits_wanted;
511 goto found_it; 527 goto found_it;
512 } 528 }
@@ -535,11 +551,13 @@ sba_search_bitmap(struct ioc *ioc, unsigned long bits_wanted, int use_hint)
535 DBG_RES(" %p %lx %lx\n", res_ptr, mask, *res_ptr); 551 DBG_RES(" %p %lx %lx\n", res_ptr, mask, *res_ptr);
536 ASSERT(0 != mask); 552 ASSERT(0 != mask);
537 for (; mask ; mask <<= o, bitshiftcnt += o) { 553 for (; mask ; mask <<= o, bitshiftcnt += o) {
538 if(0 == ((*res_ptr) & mask)) { 554 tpide = ptr_to_pide(ioc, res_ptr, bitshiftcnt);
555 ret = iommu_is_span_boundary(tpide, bits_wanted,
556 shift,
557 boundary_size);
558 if ((0 == ((*res_ptr) & mask)) && !ret) {
539 *res_ptr |= mask; /* mark resources busy! */ 559 *res_ptr |= mask; /* mark resources busy! */
540 pide = ((unsigned long)res_ptr - (unsigned long)ioc->res_map); 560 pide = tpide;
541 pide <<= 3; /* convert to bit address */
542 pide += bitshiftcnt;
543 ioc->res_bitshift = bitshiftcnt + bits_wanted; 561 ioc->res_bitshift = bitshiftcnt + bits_wanted;
544 goto found_it; 562 goto found_it;
545 } 563 }
@@ -560,6 +578,11 @@ sba_search_bitmap(struct ioc *ioc, unsigned long bits_wanted, int use_hint)
560 end = res_end - qwords; 578 end = res_end - qwords;
561 579
562 for (; res_ptr < end; res_ptr++) { 580 for (; res_ptr < end; res_ptr++) {
581 tpide = ptr_to_pide(ioc, res_ptr, 0);
582 ret = iommu_is_span_boundary(tpide, bits_wanted,
583 shift, boundary_size);
584 if (ret)
585 goto next_ptr;
563 for (i = 0 ; i < qwords ; i++) { 586 for (i = 0 ; i < qwords ; i++) {
564 if (res_ptr[i] != 0) 587 if (res_ptr[i] != 0)
565 goto next_ptr; 588 goto next_ptr;
@@ -572,8 +595,7 @@ sba_search_bitmap(struct ioc *ioc, unsigned long bits_wanted, int use_hint)
572 res_ptr[i] = ~0UL; 595 res_ptr[i] = ~0UL;
573 res_ptr[i] |= RESMAP_MASK(bits); 596 res_ptr[i] |= RESMAP_MASK(bits);
574 597
575 pide = ((unsigned long)res_ptr - (unsigned long)ioc->res_map); 598 pide = tpide;
576 pide <<= 3; /* convert to bit address */
577 res_ptr += qwords; 599 res_ptr += qwords;
578 ioc->res_bitshift = bits; 600 ioc->res_bitshift = bits;
579 goto found_it; 601 goto found_it;
@@ -605,7 +627,7 @@ found_it:
605 * resource bit map. 627 * resource bit map.
606 */ 628 */
607static int 629static int
608sba_alloc_range(struct ioc *ioc, size_t size) 630sba_alloc_range(struct ioc *ioc, struct device *dev, size_t size)
609{ 631{
610 unsigned int pages_needed = size >> iovp_shift; 632 unsigned int pages_needed = size >> iovp_shift;
611#ifdef PDIR_SEARCH_TIMING 633#ifdef PDIR_SEARCH_TIMING
@@ -622,9 +644,9 @@ sba_alloc_range(struct ioc *ioc, size_t size)
622 /* 644 /*
623 ** "seek and ye shall find"...praying never hurts either... 645 ** "seek and ye shall find"...praying never hurts either...
624 */ 646 */
625 pide = sba_search_bitmap(ioc, pages_needed, 1); 647 pide = sba_search_bitmap(ioc, dev, pages_needed, 1);
626 if (unlikely(pide >= (ioc->res_size << 3))) { 648 if (unlikely(pide >= (ioc->res_size << 3))) {
627 pide = sba_search_bitmap(ioc, pages_needed, 0); 649 pide = sba_search_bitmap(ioc, dev, pages_needed, 0);
628 if (unlikely(pide >= (ioc->res_size << 3))) { 650 if (unlikely(pide >= (ioc->res_size << 3))) {
629#if DELAYED_RESOURCE_CNT > 0 651#if DELAYED_RESOURCE_CNT > 0
630 unsigned long flags; 652 unsigned long flags;
@@ -653,7 +675,7 @@ sba_alloc_range(struct ioc *ioc, size_t size)
653 } 675 }
654 spin_unlock_irqrestore(&ioc->saved_lock, flags); 676 spin_unlock_irqrestore(&ioc->saved_lock, flags);
655 677
656 pide = sba_search_bitmap(ioc, pages_needed, 0); 678 pide = sba_search_bitmap(ioc, dev, pages_needed, 0);
657 if (unlikely(pide >= (ioc->res_size << 3))) 679 if (unlikely(pide >= (ioc->res_size << 3)))
658 panic(__FILE__ ": I/O MMU @ %p is out of mapping resources\n", 680 panic(__FILE__ ": I/O MMU @ %p is out of mapping resources\n",
659 ioc->ioc_hpa); 681 ioc->ioc_hpa);
@@ -936,7 +958,7 @@ sba_map_single(struct device *dev, void *addr, size_t size, int dir)
936 spin_unlock_irqrestore(&ioc->res_lock, flags); 958 spin_unlock_irqrestore(&ioc->res_lock, flags);
937#endif 959#endif
938 960
939 pide = sba_alloc_range(ioc, size); 961 pide = sba_alloc_range(ioc, dev, size);
940 962
941 iovp = (dma_addr_t) pide << iovp_shift; 963 iovp = (dma_addr_t) pide << iovp_shift;
942 964
@@ -1373,7 +1395,7 @@ sba_coalesce_chunks(struct ioc *ioc, struct device *dev,
1373 dma_len = (dma_len + dma_offset + ~iovp_mask) & iovp_mask; 1395 dma_len = (dma_len + dma_offset + ~iovp_mask) & iovp_mask;
1374 ASSERT(dma_len <= DMA_CHUNK_SIZE); 1396 ASSERT(dma_len <= DMA_CHUNK_SIZE);
1375 dma_sg->dma_address = (dma_addr_t) (PIDE_FLAG 1397 dma_sg->dma_address = (dma_addr_t) (PIDE_FLAG
1376 | (sba_alloc_range(ioc, dma_len) << iovp_shift) 1398 | (sba_alloc_range(ioc, dev, dma_len) << iovp_shift)
1377 | dma_offset); 1399 | dma_offset);
1378 n_mappings++; 1400 n_mappings++;
1379 } 1401 }
diff --git a/arch/ia64/ia32/elfcore32.h b/arch/ia64/ia32/elfcore32.h
index 446c9aac924d..9a3abf58cea3 100644
--- a/arch/ia64/ia32/elfcore32.h
+++ b/arch/ia64/ia32/elfcore32.h
@@ -30,7 +30,19 @@ struct elf_siginfo
30 int si_errno; /* errno */ 30 int si_errno; /* errno */
31}; 31};
32 32
33#define jiffies_to_timeval(a,b) do { (b)->tv_usec = 0; (b)->tv_sec = (a)/HZ; }while(0) 33#ifdef CONFIG_VIRT_CPU_ACCOUNTING
34/*
35 * Hacks are here since types between compat_timeval (= pair of s32) and
36 * ia64-native timeval (= pair of s64) are not compatible, at least a file
37 * arch/ia64/ia32/../../../fs/binfmt_elf.c will get warnings from compiler on
38 * use of cputime_to_timeval(), which usually an alias of jiffies_to_timeval().
39 */
40#define cputime_to_timeval(a,b) \
41 do { (b)->tv_usec = 0; (b)->tv_sec = (a)/NSEC_PER_SEC; } while(0)
42#else
43#define jiffies_to_timeval(a,b) \
44 do { (b)->tv_usec = 0; (b)->tv_sec = (a)/HZ; } while(0)
45#endif
34 46
35struct elf_prstatus 47struct elf_prstatus
36{ 48{
diff --git a/arch/ia64/ia32/sys_ia32.c b/arch/ia64/ia32/sys_ia32.c
index b1bf51fe97b4..7e028ceb93ba 100644
--- a/arch/ia64/ia32/sys_ia32.c
+++ b/arch/ia64/ia32/sys_ia32.c
@@ -38,6 +38,7 @@
38#include <linux/eventpoll.h> 38#include <linux/eventpoll.h>
39#include <linux/personality.h> 39#include <linux/personality.h>
40#include <linux/ptrace.h> 40#include <linux/ptrace.h>
41#include <linux/regset.h>
41#include <linux/stat.h> 42#include <linux/stat.h>
42#include <linux/ipc.h> 43#include <linux/ipc.h>
43#include <linux/capability.h> 44#include <linux/capability.h>
@@ -2387,16 +2388,45 @@ get_free_idx (void)
2387 return -ESRCH; 2388 return -ESRCH;
2388} 2389}
2389 2390
2391static void set_tls_desc(struct task_struct *p, int idx,
2392 const struct ia32_user_desc *info, int n)
2393{
2394 struct thread_struct *t = &p->thread;
2395 struct desc_struct *desc = &t->tls_array[idx - GDT_ENTRY_TLS_MIN];
2396 int cpu;
2397
2398 /*
2399 * We must not get preempted while modifying the TLS.
2400 */
2401 cpu = get_cpu();
2402
2403 while (n-- > 0) {
2404 if (LDT_empty(info)) {
2405 desc->a = 0;
2406 desc->b = 0;
2407 } else {
2408 desc->a = LDT_entry_a(info);
2409 desc->b = LDT_entry_b(info);
2410 }
2411
2412 ++info;
2413 ++desc;
2414 }
2415
2416 if (t == &current->thread)
2417 load_TLS(t, cpu);
2418
2419 put_cpu();
2420}
2421
2390/* 2422/*
2391 * Set a given TLS descriptor: 2423 * Set a given TLS descriptor:
2392 */ 2424 */
2393asmlinkage int 2425asmlinkage int
2394sys32_set_thread_area (struct ia32_user_desc __user *u_info) 2426sys32_set_thread_area (struct ia32_user_desc __user *u_info)
2395{ 2427{
2396 struct thread_struct *t = &current->thread;
2397 struct ia32_user_desc info; 2428 struct ia32_user_desc info;
2398 struct desc_struct *desc; 2429 int idx;
2399 int cpu, idx;
2400 2430
2401 if (copy_from_user(&info, u_info, sizeof(info))) 2431 if (copy_from_user(&info, u_info, sizeof(info)))
2402 return -EFAULT; 2432 return -EFAULT;
@@ -2416,18 +2446,7 @@ sys32_set_thread_area (struct ia32_user_desc __user *u_info)
2416 if (idx < GDT_ENTRY_TLS_MIN || idx > GDT_ENTRY_TLS_MAX) 2446 if (idx < GDT_ENTRY_TLS_MIN || idx > GDT_ENTRY_TLS_MAX)
2417 return -EINVAL; 2447 return -EINVAL;
2418 2448
2419 desc = t->tls_array + idx - GDT_ENTRY_TLS_MIN; 2449 set_tls_desc(current, idx, &info, 1);
2420
2421 cpu = smp_processor_id();
2422
2423 if (LDT_empty(&info)) {
2424 desc->a = 0;
2425 desc->b = 0;
2426 } else {
2427 desc->a = LDT_entry_a(&info);
2428 desc->b = LDT_entry_b(&info);
2429 }
2430 load_TLS(t, cpu);
2431 return 0; 2450 return 0;
2432} 2451}
2433 2452
@@ -2451,6 +2470,20 @@ sys32_set_thread_area (struct ia32_user_desc __user *u_info)
2451#define GET_PRESENT(desc) (((desc)->b >> 15) & 1) 2470#define GET_PRESENT(desc) (((desc)->b >> 15) & 1)
2452#define GET_USEABLE(desc) (((desc)->b >> 20) & 1) 2471#define GET_USEABLE(desc) (((desc)->b >> 20) & 1)
2453 2472
2473static void fill_user_desc(struct ia32_user_desc *info, int idx,
2474 const struct desc_struct *desc)
2475{
2476 info->entry_number = idx;
2477 info->base_addr = GET_BASE(desc);
2478 info->limit = GET_LIMIT(desc);
2479 info->seg_32bit = GET_32BIT(desc);
2480 info->contents = GET_CONTENTS(desc);
2481 info->read_exec_only = !GET_WRITABLE(desc);
2482 info->limit_in_pages = GET_LIMIT_PAGES(desc);
2483 info->seg_not_present = !GET_PRESENT(desc);
2484 info->useable = GET_USEABLE(desc);
2485}
2486
2454asmlinkage int 2487asmlinkage int
2455sys32_get_thread_area (struct ia32_user_desc __user *u_info) 2488sys32_get_thread_area (struct ia32_user_desc __user *u_info)
2456{ 2489{
@@ -2464,22 +2497,588 @@ sys32_get_thread_area (struct ia32_user_desc __user *u_info)
2464 return -EINVAL; 2497 return -EINVAL;
2465 2498
2466 desc = current->thread.tls_array + idx - GDT_ENTRY_TLS_MIN; 2499 desc = current->thread.tls_array + idx - GDT_ENTRY_TLS_MIN;
2467 2500 fill_user_desc(&info, idx, desc);
2468 info.entry_number = idx;
2469 info.base_addr = GET_BASE(desc);
2470 info.limit = GET_LIMIT(desc);
2471 info.seg_32bit = GET_32BIT(desc);
2472 info.contents = GET_CONTENTS(desc);
2473 info.read_exec_only = !GET_WRITABLE(desc);
2474 info.limit_in_pages = GET_LIMIT_PAGES(desc);
2475 info.seg_not_present = !GET_PRESENT(desc);
2476 info.useable = GET_USEABLE(desc);
2477 2501
2478 if (copy_to_user(u_info, &info, sizeof(info))) 2502 if (copy_to_user(u_info, &info, sizeof(info)))
2479 return -EFAULT; 2503 return -EFAULT;
2480 return 0; 2504 return 0;
2481} 2505}
2482 2506
2507struct regset_get {
2508 void *kbuf;
2509 void __user *ubuf;
2510};
2511
2512struct regset_set {
2513 const void *kbuf;
2514 const void __user *ubuf;
2515};
2516
2517struct regset_getset {
2518 struct task_struct *target;
2519 const struct user_regset *regset;
2520 union {
2521 struct regset_get get;
2522 struct regset_set set;
2523 } u;
2524 unsigned int pos;
2525 unsigned int count;
2526 int ret;
2527};
2528
2529static void getfpreg(struct task_struct *task, int regno, int *val)
2530{
2531 switch (regno / sizeof(int)) {
2532 case 0:
2533 *val = task->thread.fcr & 0xffff;
2534 break;
2535 case 1:
2536 *val = task->thread.fsr & 0xffff;
2537 break;
2538 case 2:
2539 *val = (task->thread.fsr>>16) & 0xffff;
2540 break;
2541 case 3:
2542 *val = task->thread.fir;
2543 break;
2544 case 4:
2545 *val = (task->thread.fir>>32) & 0xffff;
2546 break;
2547 case 5:
2548 *val = task->thread.fdr;
2549 break;
2550 case 6:
2551 *val = (task->thread.fdr >> 32) & 0xffff;
2552 break;
2553 }
2554}
2555
2556static void setfpreg(struct task_struct *task, int regno, int val)
2557{
2558 switch (regno / sizeof(int)) {
2559 case 0:
2560 task->thread.fcr = (task->thread.fcr & (~0x1f3f))
2561 | (val & 0x1f3f);
2562 break;
2563 case 1:
2564 task->thread.fsr = (task->thread.fsr & (~0xffff)) | val;
2565 break;
2566 case 2:
2567 task->thread.fsr = (task->thread.fsr & (~0xffff0000))
2568 | (val << 16);
2569 break;
2570 case 3:
2571 task->thread.fir = (task->thread.fir & (~0xffffffff)) | val;
2572 break;
2573 case 5:
2574 task->thread.fdr = (task->thread.fdr & (~0xffffffff)) | val;
2575 break;
2576 }
2577}
2578
2579static void access_fpreg_ia32(int regno, void *reg,
2580 struct pt_regs *pt, struct switch_stack *sw,
2581 int tos, int write)
2582{
2583 void *f;
2584
2585 if ((regno += tos) >= 8)
2586 regno -= 8;
2587 if (regno < 4)
2588 f = &pt->f8 + regno;
2589 else if (regno <= 7)
2590 f = &sw->f12 + (regno - 4);
2591 else {
2592 printk(KERN_ERR "regno must be less than 7 \n");
2593 return;
2594 }
2595
2596 if (write)
2597 memcpy(f, reg, sizeof(struct _fpreg_ia32));
2598 else
2599 memcpy(reg, f, sizeof(struct _fpreg_ia32));
2600}
2601
2602static void do_fpregs_get(struct unw_frame_info *info, void *arg)
2603{
2604 struct regset_getset *dst = arg;
2605 struct task_struct *task = dst->target;
2606 struct pt_regs *pt;
2607 int start, end, tos;
2608 char buf[80];
2609
2610 if (dst->count == 0 || unw_unwind_to_user(info) < 0)
2611 return;
2612 if (dst->pos < 7 * sizeof(int)) {
2613 end = min((dst->pos + dst->count),
2614 (unsigned int)(7 * sizeof(int)));
2615 for (start = dst->pos; start < end; start += sizeof(int))
2616 getfpreg(task, start, (int *)(buf + start));
2617 dst->ret = user_regset_copyout(&dst->pos, &dst->count,
2618 &dst->u.get.kbuf, &dst->u.get.ubuf, buf,
2619 0, 7 * sizeof(int));
2620 if (dst->ret || dst->count == 0)
2621 return;
2622 }
2623 if (dst->pos < sizeof(struct ia32_user_i387_struct)) {
2624 pt = task_pt_regs(task);
2625 tos = (task->thread.fsr >> 11) & 7;
2626 end = min(dst->pos + dst->count,
2627 (unsigned int)(sizeof(struct ia32_user_i387_struct)));
2628 start = (dst->pos - 7 * sizeof(int)) /
2629 sizeof(struct _fpreg_ia32);
2630 end = (end - 7 * sizeof(int)) / sizeof(struct _fpreg_ia32);
2631 for (; start < end; start++)
2632 access_fpreg_ia32(start,
2633 (struct _fpreg_ia32 *)buf + start,
2634 pt, info->sw, tos, 0);
2635 dst->ret = user_regset_copyout(&dst->pos, &dst->count,
2636 &dst->u.get.kbuf, &dst->u.get.ubuf,
2637 buf, 7 * sizeof(int),
2638 sizeof(struct ia32_user_i387_struct));
2639 if (dst->ret || dst->count == 0)
2640 return;
2641 }
2642}
2643
2644static void do_fpregs_set(struct unw_frame_info *info, void *arg)
2645{
2646 struct regset_getset *dst = arg;
2647 struct task_struct *task = dst->target;
2648 struct pt_regs *pt;
2649 char buf[80];
2650 int end, start, tos;
2651
2652 if (dst->count == 0 || unw_unwind_to_user(info) < 0)
2653 return;
2654
2655 if (dst->pos < 7 * sizeof(int)) {
2656 start = dst->pos;
2657 dst->ret = user_regset_copyin(&dst->pos, &dst->count,
2658 &dst->u.set.kbuf, &dst->u.set.ubuf, buf,
2659 0, 7 * sizeof(int));
2660 if (dst->ret)
2661 return;
2662 for (; start < dst->pos; start += sizeof(int))
2663 setfpreg(task, start, *((int *)(buf + start)));
2664 if (dst->count == 0)
2665 return;
2666 }
2667 if (dst->pos < sizeof(struct ia32_user_i387_struct)) {
2668 start = (dst->pos - 7 * sizeof(int)) /
2669 sizeof(struct _fpreg_ia32);
2670 dst->ret = user_regset_copyin(&dst->pos, &dst->count,
2671 &dst->u.set.kbuf, &dst->u.set.ubuf,
2672 buf, 7 * sizeof(int),
2673 sizeof(struct ia32_user_i387_struct));
2674 if (dst->ret)
2675 return;
2676 pt = task_pt_regs(task);
2677 tos = (task->thread.fsr >> 11) & 7;
2678 end = (dst->pos - 7 * sizeof(int)) / sizeof(struct _fpreg_ia32);
2679 for (; start < end; start++)
2680 access_fpreg_ia32(start,
2681 (struct _fpreg_ia32 *)buf + start,
2682 pt, info->sw, tos, 1);
2683 if (dst->count == 0)
2684 return;
2685 }
2686}
2687
2688#define OFFSET(member) ((int)(offsetof(struct ia32_user_fxsr_struct, member)))
2689static void getfpxreg(struct task_struct *task, int start, int end, char *buf)
2690{
2691 int min_val;
2692
2693 min_val = min(end, OFFSET(fop));
2694 while (start < min_val) {
2695 if (start == OFFSET(cwd))
2696 *((short *)buf) = task->thread.fcr & 0xffff;
2697 else if (start == OFFSET(swd))
2698 *((short *)buf) = task->thread.fsr & 0xffff;
2699 else if (start == OFFSET(twd))
2700 *((short *)buf) = (task->thread.fsr>>16) & 0xffff;
2701 buf += 2;
2702 start += 2;
2703 }
2704 /* skip fop element */
2705 if (start == OFFSET(fop)) {
2706 start += 2;
2707 buf += 2;
2708 }
2709 while (start < end) {
2710 if (start == OFFSET(fip))
2711 *((int *)buf) = task->thread.fir;
2712 else if (start == OFFSET(fcs))
2713 *((int *)buf) = (task->thread.fir>>32) & 0xffff;
2714 else if (start == OFFSET(foo))
2715 *((int *)buf) = task->thread.fdr;
2716 else if (start == OFFSET(fos))
2717 *((int *)buf) = (task->thread.fdr>>32) & 0xffff;
2718 else if (start == OFFSET(mxcsr))
2719 *((int *)buf) = ((task->thread.fcr>>32) & 0xff80)
2720 | ((task->thread.fsr>>32) & 0x3f);
2721 buf += 4;
2722 start += 4;
2723 }
2724}
2725
2726static void setfpxreg(struct task_struct *task, int start, int end, char *buf)
2727{
2728 int min_val, num32;
2729 short num;
2730 unsigned long num64;
2731
2732 min_val = min(end, OFFSET(fop));
2733 while (start < min_val) {
2734 num = *((short *)buf);
2735 if (start == OFFSET(cwd)) {
2736 task->thread.fcr = (task->thread.fcr & (~0x1f3f))
2737 | (num & 0x1f3f);
2738 } else if (start == OFFSET(swd)) {
2739 task->thread.fsr = (task->thread.fsr & (~0xffff)) | num;
2740 } else if (start == OFFSET(twd)) {
2741 task->thread.fsr = (task->thread.fsr & (~0xffff0000))
2742 | (((int)num) << 16);
2743 }
2744 buf += 2;
2745 start += 2;
2746 }
2747 /* skip fop element */
2748 if (start == OFFSET(fop)) {
2749 start += 2;
2750 buf += 2;
2751 }
2752 while (start < end) {
2753 num32 = *((int *)buf);
2754 if (start == OFFSET(fip))
2755 task->thread.fir = (task->thread.fir & (~0xffffffff))
2756 | num32;
2757 else if (start == OFFSET(foo))
2758 task->thread.fdr = (task->thread.fdr & (~0xffffffff))
2759 | num32;
2760 else if (start == OFFSET(mxcsr)) {
2761 num64 = num32 & 0xff10;
2762 task->thread.fcr = (task->thread.fcr &
2763 (~0xff1000000000UL)) | (num64<<32);
2764 num64 = num32 & 0x3f;
2765 task->thread.fsr = (task->thread.fsr &
2766 (~0x3f00000000UL)) | (num64<<32);
2767 }
2768 buf += 4;
2769 start += 4;
2770 }
2771}
2772
2773static void do_fpxregs_get(struct unw_frame_info *info, void *arg)
2774{
2775 struct regset_getset *dst = arg;
2776 struct task_struct *task = dst->target;
2777 struct pt_regs *pt;
2778 char buf[128];
2779 int start, end, tos;
2780
2781 if (dst->count == 0 || unw_unwind_to_user(info) < 0)
2782 return;
2783 if (dst->pos < OFFSET(st_space[0])) {
2784 end = min(dst->pos + dst->count, (unsigned int)32);
2785 getfpxreg(task, dst->pos, end, buf);
2786 dst->ret = user_regset_copyout(&dst->pos, &dst->count,
2787 &dst->u.get.kbuf, &dst->u.get.ubuf, buf,
2788 0, OFFSET(st_space[0]));
2789 if (dst->ret || dst->count == 0)
2790 return;
2791 }
2792 if (dst->pos < OFFSET(xmm_space[0])) {
2793 pt = task_pt_regs(task);
2794 tos = (task->thread.fsr >> 11) & 7;
2795 end = min(dst->pos + dst->count,
2796 (unsigned int)OFFSET(xmm_space[0]));
2797 start = (dst->pos - OFFSET(st_space[0])) / 16;
2798 end = (end - OFFSET(st_space[0])) / 16;
2799 for (; start < end; start++)
2800 access_fpreg_ia32(start, buf + 16 * start, pt,
2801 info->sw, tos, 0);
2802 dst->ret = user_regset_copyout(&dst->pos, &dst->count,
2803 &dst->u.get.kbuf, &dst->u.get.ubuf,
2804 buf, OFFSET(st_space[0]), OFFSET(xmm_space[0]));
2805 if (dst->ret || dst->count == 0)
2806 return;
2807 }
2808 if (dst->pos < OFFSET(padding[0]))
2809 dst->ret = user_regset_copyout(&dst->pos, &dst->count,
2810 &dst->u.get.kbuf, &dst->u.get.ubuf,
2811 &info->sw->f16, OFFSET(xmm_space[0]),
2812 OFFSET(padding[0]));
2813}
2814
2815static void do_fpxregs_set(struct unw_frame_info *info, void *arg)
2816{
2817 struct regset_getset *dst = arg;
2818 struct task_struct *task = dst->target;
2819 char buf[128];
2820 int start, end;
2821
2822 if (dst->count == 0 || unw_unwind_to_user(info) < 0)
2823 return;
2824
2825 if (dst->pos < OFFSET(st_space[0])) {
2826 start = dst->pos;
2827 dst->ret = user_regset_copyin(&dst->pos, &dst->count,
2828 &dst->u.set.kbuf, &dst->u.set.ubuf,
2829 buf, 0, OFFSET(st_space[0]));
2830 if (dst->ret)
2831 return;
2832 setfpxreg(task, start, dst->pos, buf);
2833 if (dst->count == 0)
2834 return;
2835 }
2836 if (dst->pos < OFFSET(xmm_space[0])) {
2837 struct pt_regs *pt;
2838 int tos;
2839 pt = task_pt_regs(task);
2840 tos = (task->thread.fsr >> 11) & 7;
2841 start = (dst->pos - OFFSET(st_space[0])) / 16;
2842 dst->ret = user_regset_copyin(&dst->pos, &dst->count,
2843 &dst->u.set.kbuf, &dst->u.set.ubuf,
2844 buf, OFFSET(st_space[0]), OFFSET(xmm_space[0]));
2845 if (dst->ret)
2846 return;
2847 end = (dst->pos - OFFSET(st_space[0])) / 16;
2848 for (; start < end; start++)
2849 access_fpreg_ia32(start, buf + 16 * start, pt, info->sw,
2850 tos, 1);
2851 if (dst->count == 0)
2852 return;
2853 }
2854 if (dst->pos < OFFSET(padding[0]))
2855 dst->ret = user_regset_copyin(&dst->pos, &dst->count,
2856 &dst->u.set.kbuf, &dst->u.set.ubuf,
2857 &info->sw->f16, OFFSET(xmm_space[0]),
2858 OFFSET(padding[0]));
2859}
2860#undef OFFSET
2861
2862static int do_regset_call(void (*call)(struct unw_frame_info *, void *),
2863 struct task_struct *target,
2864 const struct user_regset *regset,
2865 unsigned int pos, unsigned int count,
2866 const void *kbuf, const void __user *ubuf)
2867{
2868 struct regset_getset info = { .target = target, .regset = regset,
2869 .pos = pos, .count = count,
2870 .u.set = { .kbuf = kbuf, .ubuf = ubuf },
2871 .ret = 0 };
2872
2873 if (target == current)
2874 unw_init_running(call, &info);
2875 else {
2876 struct unw_frame_info ufi;
2877 memset(&ufi, 0, sizeof(ufi));
2878 unw_init_from_blocked_task(&ufi, target);
2879 (*call)(&ufi, &info);
2880 }
2881
2882 return info.ret;
2883}
2884
2885static int ia32_fpregs_get(struct task_struct *target,
2886 const struct user_regset *regset,
2887 unsigned int pos, unsigned int count,
2888 void *kbuf, void __user *ubuf)
2889{
2890 return do_regset_call(do_fpregs_get, target, regset, pos, count,
2891 kbuf, ubuf);
2892}
2893
2894static int ia32_fpregs_set(struct task_struct *target,
2895 const struct user_regset *regset,
2896 unsigned int pos, unsigned int count,
2897 const void *kbuf, const void __user *ubuf)
2898{
2899 return do_regset_call(do_fpregs_set, target, regset, pos, count,
2900 kbuf, ubuf);
2901}
2902
2903static int ia32_fpxregs_get(struct task_struct *target,
2904 const struct user_regset *regset,
2905 unsigned int pos, unsigned int count,
2906 void *kbuf, void __user *ubuf)
2907{
2908 return do_regset_call(do_fpxregs_get, target, regset, pos, count,
2909 kbuf, ubuf);
2910}
2911
2912static int ia32_fpxregs_set(struct task_struct *target,
2913 const struct user_regset *regset,
2914 unsigned int pos, unsigned int count,
2915 const void *kbuf, const void __user *ubuf)
2916{
2917 return do_regset_call(do_fpxregs_set, target, regset, pos, count,
2918 kbuf, ubuf);
2919}
2920
2921static int ia32_genregs_get(struct task_struct *target,
2922 const struct user_regset *regset,
2923 unsigned int pos, unsigned int count,
2924 void *kbuf, void __user *ubuf)
2925{
2926 if (kbuf) {
2927 u32 *kp = kbuf;
2928 while (count > 0) {
2929 *kp++ = getreg(target, pos);
2930 pos += 4;
2931 count -= 4;
2932 }
2933 } else {
2934 u32 __user *up = ubuf;
2935 while (count > 0) {
2936 if (__put_user(getreg(target, pos), up++))
2937 return -EFAULT;
2938 pos += 4;
2939 count -= 4;
2940 }
2941 }
2942 return 0;
2943}
2944
2945static int ia32_genregs_set(struct task_struct *target,
2946 const struct user_regset *regset,
2947 unsigned int pos, unsigned int count,
2948 const void *kbuf, const void __user *ubuf)
2949{
2950 int ret = 0;
2951
2952 if (kbuf) {
2953 const u32 *kp = kbuf;
2954 while (!ret && count > 0) {
2955 putreg(target, pos, *kp++);
2956 pos += 4;
2957 count -= 4;
2958 }
2959 } else {
2960 const u32 __user *up = ubuf;
2961 u32 val;
2962 while (!ret && count > 0) {
2963 ret = __get_user(val, up++);
2964 if (!ret)
2965 putreg(target, pos, val);
2966 pos += 4;
2967 count -= 4;
2968 }
2969 }
2970 return ret;
2971}
2972
2973static int ia32_tls_active(struct task_struct *target,
2974 const struct user_regset *regset)
2975{
2976 struct thread_struct *t = &target->thread;
2977 int n = GDT_ENTRY_TLS_ENTRIES;
2978 while (n > 0 && desc_empty(&t->tls_array[n -1]))
2979 --n;
2980 return n;
2981}
2982
2983static int ia32_tls_get(struct task_struct *target,
2984 const struct user_regset *regset, unsigned int pos,
2985 unsigned int count, void *kbuf, void __user *ubuf)
2986{
2987 const struct desc_struct *tls;
2988
2989 if (pos > GDT_ENTRY_TLS_ENTRIES * sizeof(struct ia32_user_desc) ||
2990 (pos % sizeof(struct ia32_user_desc)) != 0 ||
2991 (count % sizeof(struct ia32_user_desc)) != 0)
2992 return -EINVAL;
2993
2994 pos /= sizeof(struct ia32_user_desc);
2995 count /= sizeof(struct ia32_user_desc);
2996
2997 tls = &target->thread.tls_array[pos];
2998
2999 if (kbuf) {
3000 struct ia32_user_desc *info = kbuf;
3001 while (count-- > 0)
3002 fill_user_desc(info++, GDT_ENTRY_TLS_MIN + pos++,
3003 tls++);
3004 } else {
3005 struct ia32_user_desc __user *u_info = ubuf;
3006 while (count-- > 0) {
3007 struct ia32_user_desc info;
3008 fill_user_desc(&info, GDT_ENTRY_TLS_MIN + pos++, tls++);
3009 if (__copy_to_user(u_info++, &info, sizeof(info)))
3010 return -EFAULT;
3011 }
3012 }
3013
3014 return 0;
3015}
3016
3017static int ia32_tls_set(struct task_struct *target,
3018 const struct user_regset *regset, unsigned int pos,
3019 unsigned int count, const void *kbuf, const void __user *ubuf)
3020{
3021 struct ia32_user_desc infobuf[GDT_ENTRY_TLS_ENTRIES];
3022 const struct ia32_user_desc *info;
3023
3024 if (pos > GDT_ENTRY_TLS_ENTRIES * sizeof(struct ia32_user_desc) ||
3025 (pos % sizeof(struct ia32_user_desc)) != 0 ||
3026 (count % sizeof(struct ia32_user_desc)) != 0)
3027 return -EINVAL;
3028
3029 if (kbuf)
3030 info = kbuf;
3031 else if (__copy_from_user(infobuf, ubuf, count))
3032 return -EFAULT;
3033 else
3034 info = infobuf;
3035
3036 set_tls_desc(target,
3037 GDT_ENTRY_TLS_MIN + (pos / sizeof(struct ia32_user_desc)),
3038 info, count / sizeof(struct ia32_user_desc));
3039
3040 return 0;
3041}
3042
3043/*
3044 * This should match arch/i386/kernel/ptrace.c:native_regsets.
3045 * XXX ioperm? vm86?
3046 */
3047static const struct user_regset ia32_regsets[] = {
3048 {
3049 .core_note_type = NT_PRSTATUS,
3050 .n = sizeof(struct user_regs_struct32)/4,
3051 .size = 4, .align = 4,
3052 .get = ia32_genregs_get, .set = ia32_genregs_set
3053 },
3054 {
3055 .core_note_type = NT_PRFPREG,
3056 .n = sizeof(struct ia32_user_i387_struct) / 4,
3057 .size = 4, .align = 4,
3058 .get = ia32_fpregs_get, .set = ia32_fpregs_set
3059 },
3060 {
3061 .core_note_type = NT_PRXFPREG,
3062 .n = sizeof(struct ia32_user_fxsr_struct) / 4,
3063 .size = 4, .align = 4,
3064 .get = ia32_fpxregs_get, .set = ia32_fpxregs_set
3065 },
3066 {
3067 .core_note_type = NT_386_TLS,
3068 .n = GDT_ENTRY_TLS_ENTRIES,
3069 .bias = GDT_ENTRY_TLS_MIN,
3070 .size = sizeof(struct ia32_user_desc),
3071 .align = sizeof(struct ia32_user_desc),
3072 .active = ia32_tls_active,
3073 .get = ia32_tls_get, .set = ia32_tls_set,
3074 },
3075};
3076
3077const struct user_regset_view user_ia32_view = {
3078 .name = "i386", .e_machine = EM_386,
3079 .regsets = ia32_regsets, .n = ARRAY_SIZE(ia32_regsets)
3080};
3081
2483long sys32_fadvise64_64(int fd, __u32 offset_low, __u32 offset_high, 3082long sys32_fadvise64_64(int fd, __u32 offset_low, __u32 offset_high,
2484 __u32 len_low, __u32 len_high, int advice) 3083 __u32 len_low, __u32 len_high, int advice)
2485{ 3084{
diff --git a/arch/ia64/kernel/acpi.c b/arch/ia64/kernel/acpi.c
index 78f28d825f30..c7467f863c7a 100644
--- a/arch/ia64/kernel/acpi.c
+++ b/arch/ia64/kernel/acpi.c
@@ -423,6 +423,7 @@ static u32 __devinitdata pxm_flag[PXM_FLAG_LEN];
423#define pxm_bit_set(bit) (set_bit(bit,(void *)pxm_flag)) 423#define pxm_bit_set(bit) (set_bit(bit,(void *)pxm_flag))
424#define pxm_bit_test(bit) (test_bit(bit,(void *)pxm_flag)) 424#define pxm_bit_test(bit) (test_bit(bit,(void *)pxm_flag))
425static struct acpi_table_slit __initdata *slit_table; 425static struct acpi_table_slit __initdata *slit_table;
426cpumask_t early_cpu_possible_map = CPU_MASK_NONE;
426 427
427static int get_processor_proximity_domain(struct acpi_srat_cpu_affinity *pa) 428static int get_processor_proximity_domain(struct acpi_srat_cpu_affinity *pa)
428{ 429{
@@ -482,6 +483,7 @@ acpi_numa_processor_affinity_init(struct acpi_srat_cpu_affinity *pa)
482 (pa->apic_id << 8) | (pa->local_sapic_eid); 483 (pa->apic_id << 8) | (pa->local_sapic_eid);
483 /* nid should be overridden as logical node id later */ 484 /* nid should be overridden as logical node id later */
484 node_cpuid[srat_num_cpus].nid = pxm; 485 node_cpuid[srat_num_cpus].nid = pxm;
486 cpu_set(srat_num_cpus, early_cpu_possible_map);
485 srat_num_cpus++; 487 srat_num_cpus++;
486} 488}
487 489
@@ -559,7 +561,7 @@ void __init acpi_numa_arch_fixup(void)
559 } 561 }
560 562
561 /* set logical node id in cpu structure */ 563 /* set logical node id in cpu structure */
562 for (i = 0; i < srat_num_cpus; i++) 564 for_each_possible_early_cpu(i)
563 node_cpuid[i].nid = pxm_to_node(node_cpuid[i].nid); 565 node_cpuid[i].nid = pxm_to_node(node_cpuid[i].nid);
564 566
565 printk(KERN_INFO "Number of logical nodes in system = %d\n", 567 printk(KERN_INFO "Number of logical nodes in system = %d\n",
diff --git a/arch/ia64/kernel/asm-offsets.c b/arch/ia64/kernel/asm-offsets.c
index 0aebc6f79e95..230a6f92367f 100644
--- a/arch/ia64/kernel/asm-offsets.c
+++ b/arch/ia64/kernel/asm-offsets.c
@@ -7,6 +7,7 @@
7#define ASM_OFFSETS_C 1 7#define ASM_OFFSETS_C 1
8 8
9#include <linux/sched.h> 9#include <linux/sched.h>
10#include <linux/pid.h>
10#include <linux/clocksource.h> 11#include <linux/clocksource.h>
11 12
12#include <asm-ia64/processor.h> 13#include <asm-ia64/processor.h>
@@ -34,17 +35,29 @@ void foo(void)
34 DEFINE(SIGFRAME_SIZE, sizeof (struct sigframe)); 35 DEFINE(SIGFRAME_SIZE, sizeof (struct sigframe));
35 DEFINE(UNW_FRAME_INFO_SIZE, sizeof (struct unw_frame_info)); 36 DEFINE(UNW_FRAME_INFO_SIZE, sizeof (struct unw_frame_info));
36 37
38 BUILD_BUG_ON(sizeof(struct upid) != 32);
39 DEFINE(IA64_UPID_SHIFT, 5);
40
37 BLANK(); 41 BLANK();
38 42
39 DEFINE(TI_FLAGS, offsetof(struct thread_info, flags)); 43 DEFINE(TI_FLAGS, offsetof(struct thread_info, flags));
40 DEFINE(TI_CPU, offsetof(struct thread_info, cpu)); 44 DEFINE(TI_CPU, offsetof(struct thread_info, cpu));
41 DEFINE(TI_PRE_COUNT, offsetof(struct thread_info, preempt_count)); 45 DEFINE(TI_PRE_COUNT, offsetof(struct thread_info, preempt_count));
46#ifdef CONFIG_VIRT_CPU_ACCOUNTING
47 DEFINE(TI_AC_STAMP, offsetof(struct thread_info, ac_stamp));
48 DEFINE(TI_AC_LEAVE, offsetof(struct thread_info, ac_leave));
49 DEFINE(TI_AC_STIME, offsetof(struct thread_info, ac_stime));
50 DEFINE(TI_AC_UTIME, offsetof(struct thread_info, ac_utime));
51#endif
42 52
43 BLANK(); 53 BLANK();
44 54
45 DEFINE(IA64_TASK_BLOCKED_OFFSET,offsetof (struct task_struct, blocked)); 55 DEFINE(IA64_TASK_BLOCKED_OFFSET,offsetof (struct task_struct, blocked));
46 DEFINE(IA64_TASK_CLEAR_CHILD_TID_OFFSET,offsetof (struct task_struct, clear_child_tid)); 56 DEFINE(IA64_TASK_CLEAR_CHILD_TID_OFFSET,offsetof (struct task_struct, clear_child_tid));
47 DEFINE(IA64_TASK_GROUP_LEADER_OFFSET, offsetof (struct task_struct, group_leader)); 57 DEFINE(IA64_TASK_GROUP_LEADER_OFFSET, offsetof (struct task_struct, group_leader));
58 DEFINE(IA64_TASK_TGIDLINK_OFFSET, offsetof (struct task_struct, pids[PIDTYPE_PID].pid));
59 DEFINE(IA64_PID_LEVEL_OFFSET, offsetof (struct pid, level));
60 DEFINE(IA64_PID_UPID_OFFSET, offsetof (struct pid, numbers[0]));
48 DEFINE(IA64_TASK_PENDING_OFFSET,offsetof (struct task_struct, pending)); 61 DEFINE(IA64_TASK_PENDING_OFFSET,offsetof (struct task_struct, pending));
49 DEFINE(IA64_TASK_PID_OFFSET, offsetof (struct task_struct, pid)); 62 DEFINE(IA64_TASK_PID_OFFSET, offsetof (struct task_struct, pid));
50 DEFINE(IA64_TASK_REAL_PARENT_OFFSET, offsetof (struct task_struct, real_parent)); 63 DEFINE(IA64_TASK_REAL_PARENT_OFFSET, offsetof (struct task_struct, real_parent));
diff --git a/arch/ia64/kernel/crash.c b/arch/ia64/kernel/crash.c
index fbe742ad2fde..90ef338cf46f 100644
--- a/arch/ia64/kernel/crash.c
+++ b/arch/ia64/kernel/crash.c
@@ -24,6 +24,7 @@ int kdump_status[NR_CPUS];
24static atomic_t kdump_cpu_frozen; 24static atomic_t kdump_cpu_frozen;
25atomic_t kdump_in_progress; 25atomic_t kdump_in_progress;
26static int kdump_on_init = 1; 26static int kdump_on_init = 1;
27static int kdump_on_fatal_mca = 1;
27 28
28static inline Elf64_Word 29static inline Elf64_Word
29*append_elf_note(Elf64_Word *buf, char *name, unsigned type, void *data, 30*append_elf_note(Elf64_Word *buf, char *name, unsigned type, void *data,
@@ -118,6 +119,7 @@ machine_crash_shutdown(struct pt_regs *pt)
118static void 119static void
119machine_kdump_on_init(void) 120machine_kdump_on_init(void)
120{ 121{
122 crash_save_vmcoreinfo();
121 local_irq_disable(); 123 local_irq_disable();
122 kexec_disable_iosapic(); 124 kexec_disable_iosapic();
123 machine_kexec(ia64_kimage); 125 machine_kexec(ia64_kimage);
@@ -148,7 +150,7 @@ kdump_init_notifier(struct notifier_block *self, unsigned long val, void *data)
148 struct ia64_mca_notify_die *nd; 150 struct ia64_mca_notify_die *nd;
149 struct die_args *args = data; 151 struct die_args *args = data;
150 152
151 if (!kdump_on_init) 153 if (!kdump_on_init && !kdump_on_fatal_mca)
152 return NOTIFY_DONE; 154 return NOTIFY_DONE;
153 155
154 if (!ia64_kimage) { 156 if (!ia64_kimage) {
@@ -173,32 +175,38 @@ kdump_init_notifier(struct notifier_block *self, unsigned long val, void *data)
173 return NOTIFY_DONE; 175 return NOTIFY_DONE;
174 176
175 switch (val) { 177 switch (val) {
176 case DIE_INIT_MONARCH_PROCESS: 178 case DIE_INIT_MONARCH_PROCESS:
179 if (kdump_on_init) {
177 atomic_set(&kdump_in_progress, 1); 180 atomic_set(&kdump_in_progress, 1);
178 *(nd->monarch_cpu) = -1; 181 *(nd->monarch_cpu) = -1;
179 break; 182 }
180 case DIE_INIT_MONARCH_LEAVE: 183 break;
184 case DIE_INIT_MONARCH_LEAVE:
185 if (kdump_on_init)
181 machine_kdump_on_init(); 186 machine_kdump_on_init();
182 break; 187 break;
183 case DIE_INIT_SLAVE_LEAVE: 188 case DIE_INIT_SLAVE_LEAVE:
184 if (atomic_read(&kdump_in_progress)) 189 if (atomic_read(&kdump_in_progress))
185 unw_init_running(kdump_cpu_freeze, NULL); 190 unw_init_running(kdump_cpu_freeze, NULL);
186 break; 191 break;
187 case DIE_MCA_RENDZVOUS_LEAVE: 192 case DIE_MCA_RENDZVOUS_LEAVE:
188 if (atomic_read(&kdump_in_progress)) 193 if (atomic_read(&kdump_in_progress))
189 unw_init_running(kdump_cpu_freeze, NULL); 194 unw_init_running(kdump_cpu_freeze, NULL);
190 break; 195 break;
191 case DIE_MCA_MONARCH_LEAVE: 196 case DIE_MCA_MONARCH_LEAVE:
192 /* die_register->signr indicate if MCA is recoverable */ 197 /* die_register->signr indicate if MCA is recoverable */
193 if (!args->signr) 198 if (kdump_on_fatal_mca && !args->signr) {
194 machine_kdump_on_init(); 199 atomic_set(&kdump_in_progress, 1);
195 break; 200 *(nd->monarch_cpu) = -1;
201 machine_kdump_on_init();
202 }
203 break;
196 } 204 }
197 return NOTIFY_DONE; 205 return NOTIFY_DONE;
198} 206}
199 207
200#ifdef CONFIG_SYSCTL 208#ifdef CONFIG_SYSCTL
201static ctl_table kdump_on_init_table[] = { 209static ctl_table kdump_ctl_table[] = {
202 { 210 {
203 .ctl_name = CTL_UNNUMBERED, 211 .ctl_name = CTL_UNNUMBERED,
204 .procname = "kdump_on_init", 212 .procname = "kdump_on_init",
@@ -207,6 +215,14 @@ static ctl_table kdump_on_init_table[] = {
207 .mode = 0644, 215 .mode = 0644,
208 .proc_handler = &proc_dointvec, 216 .proc_handler = &proc_dointvec,
209 }, 217 },
218 {
219 .ctl_name = CTL_UNNUMBERED,
220 .procname = "kdump_on_fatal_mca",
221 .data = &kdump_on_fatal_mca,
222 .maxlen = sizeof(int),
223 .mode = 0644,
224 .proc_handler = &proc_dointvec,
225 },
210 { .ctl_name = 0 } 226 { .ctl_name = 0 }
211}; 227};
212 228
@@ -215,7 +231,7 @@ static ctl_table sys_table[] = {
215 .ctl_name = CTL_KERN, 231 .ctl_name = CTL_KERN,
216 .procname = "kernel", 232 .procname = "kernel",
217 .mode = 0555, 233 .mode = 0555,
218 .child = kdump_on_init_table, 234 .child = kdump_ctl_table,
219 }, 235 },
220 { .ctl_name = 0 } 236 { .ctl_name = 0 }
221}; 237};
diff --git a/arch/ia64/kernel/efi.c b/arch/ia64/kernel/efi.c
index 728d7247a1a6..d45f215bc8fc 100644
--- a/arch/ia64/kernel/efi.c
+++ b/arch/ia64/kernel/efi.c
@@ -37,6 +37,7 @@
37#include <asm/pgtable.h> 37#include <asm/pgtable.h>
38#include <asm/processor.h> 38#include <asm/processor.h>
39#include <asm/mca.h> 39#include <asm/mca.h>
40#include <asm/tlbflush.h>
40 41
41#define EFI_DEBUG 0 42#define EFI_DEBUG 0
42 43
@@ -403,6 +404,41 @@ efi_get_pal_addr (void)
403 return NULL; 404 return NULL;
404} 405}
405 406
407
408static u8 __init palo_checksum(u8 *buffer, u32 length)
409{
410 u8 sum = 0;
411 u8 *end = buffer + length;
412
413 while (buffer < end)
414 sum = (u8) (sum + *(buffer++));
415
416 return sum;
417}
418
419/*
420 * Parse and handle PALO table which is published at:
421 * http://www.dig64.org/home/DIG64_PALO_R1_0.pdf
422 */
423static void __init handle_palo(unsigned long palo_phys)
424{
425 struct palo_table *palo = __va(palo_phys);
426 u8 checksum;
427
428 if (strncmp(palo->signature, PALO_SIG, sizeof(PALO_SIG) - 1)) {
429 printk(KERN_INFO "PALO signature incorrect.\n");
430 return;
431 }
432
433 checksum = palo_checksum((u8 *)palo, palo->length);
434 if (checksum) {
435 printk(KERN_INFO "PALO checksum incorrect.\n");
436 return;
437 }
438
439 setup_ptcg_sem(palo->max_tlb_purges, NPTCG_FROM_PALO);
440}
441
406void 442void
407efi_map_pal_code (void) 443efi_map_pal_code (void)
408{ 444{
@@ -432,6 +468,7 @@ efi_init (void)
432 u64 efi_desc_size; 468 u64 efi_desc_size;
433 char *cp, vendor[100] = "unknown"; 469 char *cp, vendor[100] = "unknown";
434 int i; 470 int i;
471 unsigned long palo_phys;
435 472
436 /* 473 /*
437 * It's too early to be able to use the standard kernel command line 474 * It's too early to be able to use the standard kernel command line
@@ -496,6 +533,8 @@ efi_init (void)
496 efi.hcdp = EFI_INVALID_TABLE_ADDR; 533 efi.hcdp = EFI_INVALID_TABLE_ADDR;
497 efi.uga = EFI_INVALID_TABLE_ADDR; 534 efi.uga = EFI_INVALID_TABLE_ADDR;
498 535
536 palo_phys = EFI_INVALID_TABLE_ADDR;
537
499 for (i = 0; i < (int) efi.systab->nr_tables; i++) { 538 for (i = 0; i < (int) efi.systab->nr_tables; i++) {
500 if (efi_guidcmp(config_tables[i].guid, MPS_TABLE_GUID) == 0) { 539 if (efi_guidcmp(config_tables[i].guid, MPS_TABLE_GUID) == 0) {
501 efi.mps = config_tables[i].table; 540 efi.mps = config_tables[i].table;
@@ -515,10 +554,17 @@ efi_init (void)
515 } else if (efi_guidcmp(config_tables[i].guid, HCDP_TABLE_GUID) == 0) { 554 } else if (efi_guidcmp(config_tables[i].guid, HCDP_TABLE_GUID) == 0) {
516 efi.hcdp = config_tables[i].table; 555 efi.hcdp = config_tables[i].table;
517 printk(" HCDP=0x%lx", config_tables[i].table); 556 printk(" HCDP=0x%lx", config_tables[i].table);
557 } else if (efi_guidcmp(config_tables[i].guid,
558 PROCESSOR_ABSTRACTION_LAYER_OVERWRITE_GUID) == 0) {
559 palo_phys = config_tables[i].table;
560 printk(" PALO=0x%lx", config_tables[i].table);
518 } 561 }
519 } 562 }
520 printk("\n"); 563 printk("\n");
521 564
565 if (palo_phys != EFI_INVALID_TABLE_ADDR)
566 handle_palo(palo_phys);
567
522 runtime = __va(efi.systab->runtime); 568 runtime = __va(efi.systab->runtime);
523 efi.get_time = phys_get_time; 569 efi.get_time = phys_get_time;
524 efi.set_time = phys_set_time; 570 efi.set_time = phys_set_time;
diff --git a/arch/ia64/kernel/entry.S b/arch/ia64/kernel/entry.S
index 3c331c464b40..b0be4a280174 100644
--- a/arch/ia64/kernel/entry.S
+++ b/arch/ia64/kernel/entry.S
@@ -710,6 +710,16 @@ ENTRY(ia64_leave_syscall)
710(pUStk) cmp.eq.unc p6,p0=r0,r0 // p6 <- pUStk 710(pUStk) cmp.eq.unc p6,p0=r0,r0 // p6 <- pUStk
711#endif 711#endif
712.work_processed_syscall: 712.work_processed_syscall:
713#ifdef CONFIG_VIRT_CPU_ACCOUNTING
714 adds r2=PT(LOADRS)+16,r12
715(pUStk) mov.m r22=ar.itc // fetch time at leave
716 adds r18=TI_FLAGS+IA64_TASK_SIZE,r13
717 ;;
718(p6) ld4 r31=[r18] // load current_thread_info()->flags
719 ld8 r19=[r2],PT(B6)-PT(LOADRS) // load ar.rsc value for "loadrs"
720 adds r3=PT(AR_BSPSTORE)+16,r12 // deferred
721 ;;
722#else
713 adds r2=PT(LOADRS)+16,r12 723 adds r2=PT(LOADRS)+16,r12
714 adds r3=PT(AR_BSPSTORE)+16,r12 724 adds r3=PT(AR_BSPSTORE)+16,r12
715 adds r18=TI_FLAGS+IA64_TASK_SIZE,r13 725 adds r18=TI_FLAGS+IA64_TASK_SIZE,r13
@@ -718,6 +728,7 @@ ENTRY(ia64_leave_syscall)
718 ld8 r19=[r2],PT(B6)-PT(LOADRS) // load ar.rsc value for "loadrs" 728 ld8 r19=[r2],PT(B6)-PT(LOADRS) // load ar.rsc value for "loadrs"
719 nop.i 0 729 nop.i 0
720 ;; 730 ;;
731#endif
721 mov r16=ar.bsp // M2 get existing backing store pointer 732 mov r16=ar.bsp // M2 get existing backing store pointer
722 ld8 r18=[r2],PT(R9)-PT(B6) // load b6 733 ld8 r18=[r2],PT(R9)-PT(B6) // load b6
723(p6) and r15=TIF_WORK_MASK,r31 // any work other than TIF_SYSCALL_TRACE? 734(p6) and r15=TIF_WORK_MASK,r31 // any work other than TIF_SYSCALL_TRACE?
@@ -737,12 +748,21 @@ ENTRY(ia64_leave_syscall)
737 748
738 ld8 r29=[r2],16 // M0|1 load cr.ipsr 749 ld8 r29=[r2],16 // M0|1 load cr.ipsr
739 ld8 r28=[r3],16 // M0|1 load cr.iip 750 ld8 r28=[r3],16 // M0|1 load cr.iip
751#ifdef CONFIG_VIRT_CPU_ACCOUNTING
752(pUStk) add r14=TI_AC_LEAVE+IA64_TASK_SIZE,r13
753 ;;
754 ld8 r30=[r2],16 // M0|1 load cr.ifs
755 ld8 r25=[r3],16 // M0|1 load ar.unat
756(pUStk) add r15=IA64_TASK_THREAD_ON_USTACK_OFFSET,r13
757 ;;
758#else
740 mov r22=r0 // A clear r22 759 mov r22=r0 // A clear r22
741 ;; 760 ;;
742 ld8 r30=[r2],16 // M0|1 load cr.ifs 761 ld8 r30=[r2],16 // M0|1 load cr.ifs
743 ld8 r25=[r3],16 // M0|1 load ar.unat 762 ld8 r25=[r3],16 // M0|1 load ar.unat
744(pUStk) add r14=IA64_TASK_THREAD_ON_USTACK_OFFSET,r13 763(pUStk) add r14=IA64_TASK_THREAD_ON_USTACK_OFFSET,r13
745 ;; 764 ;;
765#endif
746 ld8 r26=[r2],PT(B0)-PT(AR_PFS) // M0|1 load ar.pfs 766 ld8 r26=[r2],PT(B0)-PT(AR_PFS) // M0|1 load ar.pfs
747(pKStk) mov r22=psr // M2 read PSR now that interrupts are disabled 767(pKStk) mov r22=psr // M2 read PSR now that interrupts are disabled
748 nop 0 768 nop 0
@@ -759,7 +779,11 @@ ENTRY(ia64_leave_syscall)
759 ld8.fill r1=[r3],16 // M0|1 load r1 779 ld8.fill r1=[r3],16 // M0|1 load r1
760(pUStk) mov r17=1 // A 780(pUStk) mov r17=1 // A
761 ;; 781 ;;
782#ifdef CONFIG_VIRT_CPU_ACCOUNTING
783(pUStk) st1 [r15]=r17 // M2|3
784#else
762(pUStk) st1 [r14]=r17 // M2|3 785(pUStk) st1 [r14]=r17 // M2|3
786#endif
763 ld8.fill r13=[r3],16 // M0|1 787 ld8.fill r13=[r3],16 // M0|1
764 mov f8=f0 // F clear f8 788 mov f8=f0 // F clear f8
765 ;; 789 ;;
@@ -775,12 +799,22 @@ ENTRY(ia64_leave_syscall)
775 shr.u r18=r19,16 // I0|1 get byte size of existing "dirty" partition 799 shr.u r18=r19,16 // I0|1 get byte size of existing "dirty" partition
776 cover // B add current frame into dirty partition & set cr.ifs 800 cover // B add current frame into dirty partition & set cr.ifs
777 ;; 801 ;;
802#ifdef CONFIG_VIRT_CPU_ACCOUNTING
803 mov r19=ar.bsp // M2 get new backing store pointer
804 st8 [r14]=r22 // M save time at leave
805 mov f10=f0 // F clear f10
806
807 mov r22=r0 // A clear r22
808 movl r14=__kernel_syscall_via_epc // X
809 ;;
810#else
778 mov r19=ar.bsp // M2 get new backing store pointer 811 mov r19=ar.bsp // M2 get new backing store pointer
779 mov f10=f0 // F clear f10 812 mov f10=f0 // F clear f10
780 813
781 nop.m 0 814 nop.m 0
782 movl r14=__kernel_syscall_via_epc // X 815 movl r14=__kernel_syscall_via_epc // X
783 ;; 816 ;;
817#endif
784 mov.m ar.csd=r0 // M2 clear ar.csd 818 mov.m ar.csd=r0 // M2 clear ar.csd
785 mov.m ar.ccv=r0 // M2 clear ar.ccv 819 mov.m ar.ccv=r0 // M2 clear ar.ccv
786 mov b7=r14 // I0 clear b7 (hint with __kernel_syscall_via_epc) 820 mov b7=r14 // I0 clear b7 (hint with __kernel_syscall_via_epc)
@@ -913,10 +947,18 @@ GLOBAL_ENTRY(ia64_leave_kernel)
913 adds r16=PT(CR_IPSR)+16,r12 947 adds r16=PT(CR_IPSR)+16,r12
914 adds r17=PT(CR_IIP)+16,r12 948 adds r17=PT(CR_IIP)+16,r12
915 949
950#ifdef CONFIG_VIRT_CPU_ACCOUNTING
951 .pred.rel.mutex pUStk,pKStk
952(pKStk) mov r22=psr // M2 read PSR now that interrupts are disabled
953(pUStk) mov.m r22=ar.itc // M fetch time at leave
954 nop.i 0
955 ;;
956#else
916(pKStk) mov r22=psr // M2 read PSR now that interrupts are disabled 957(pKStk) mov r22=psr // M2 read PSR now that interrupts are disabled
917 nop.i 0 958 nop.i 0
918 nop.i 0 959 nop.i 0
919 ;; 960 ;;
961#endif
920 ld8 r29=[r16],16 // load cr.ipsr 962 ld8 r29=[r16],16 // load cr.ipsr
921 ld8 r28=[r17],16 // load cr.iip 963 ld8 r28=[r17],16 // load cr.iip
922 ;; 964 ;;
@@ -938,15 +980,37 @@ GLOBAL_ENTRY(ia64_leave_kernel)
938 ;; 980 ;;
939 ld8.fill r12=[r16],16 981 ld8.fill r12=[r16],16
940 ld8.fill r13=[r17],16 982 ld8.fill r13=[r17],16
983#ifdef CONFIG_VIRT_CPU_ACCOUNTING
984(pUStk) adds r3=TI_AC_LEAVE+IA64_TASK_SIZE,r18
985#else
941(pUStk) adds r18=IA64_TASK_THREAD_ON_USTACK_OFFSET,r18 986(pUStk) adds r18=IA64_TASK_THREAD_ON_USTACK_OFFSET,r18
987#endif
942 ;; 988 ;;
943 ld8 r20=[r16],16 // ar.fpsr 989 ld8 r20=[r16],16 // ar.fpsr
944 ld8.fill r15=[r17],16 990 ld8.fill r15=[r17],16
991#ifdef CONFIG_VIRT_CPU_ACCOUNTING
992(pUStk) adds r18=IA64_TASK_THREAD_ON_USTACK_OFFSET,r18 // deferred
993#endif
945 ;; 994 ;;
946 ld8.fill r14=[r16],16 995 ld8.fill r14=[r16],16
947 ld8.fill r2=[r17] 996 ld8.fill r2=[r17]
948(pUStk) mov r17=1 997(pUStk) mov r17=1
949 ;; 998 ;;
999#ifdef CONFIG_VIRT_CPU_ACCOUNTING
1000 // mmi_ : ld8 st1 shr;; mmi_ : st8 st1 shr;;
1001 // mib : mov add br -> mib : ld8 add br
1002 // bbb_ : br nop cover;; mbb_ : mov br cover;;
1003 //
1004 // no one require bsp in r16 if (pKStk) branch is selected.
1005(pUStk) st8 [r3]=r22 // save time at leave
1006(pUStk) st1 [r18]=r17 // restore current->thread.on_ustack
1007 shr.u r18=r19,16 // get byte size of existing "dirty" partition
1008 ;;
1009 ld8.fill r3=[r16] // deferred
1010 LOAD_PHYS_STACK_REG_SIZE(r17)
1011(pKStk) br.cond.dpnt skip_rbs_switch
1012 mov r16=ar.bsp // get existing backing store pointer
1013#else
950 ld8.fill r3=[r16] 1014 ld8.fill r3=[r16]
951(pUStk) st1 [r18]=r17 // restore current->thread.on_ustack 1015(pUStk) st1 [r18]=r17 // restore current->thread.on_ustack
952 shr.u r18=r19,16 // get byte size of existing "dirty" partition 1016 shr.u r18=r19,16 // get byte size of existing "dirty" partition
@@ -954,6 +1018,7 @@ GLOBAL_ENTRY(ia64_leave_kernel)
954 mov r16=ar.bsp // get existing backing store pointer 1018 mov r16=ar.bsp // get existing backing store pointer
955 LOAD_PHYS_STACK_REG_SIZE(r17) 1019 LOAD_PHYS_STACK_REG_SIZE(r17)
956(pKStk) br.cond.dpnt skip_rbs_switch 1020(pKStk) br.cond.dpnt skip_rbs_switch
1021#endif
957 1022
958 /* 1023 /*
959 * Restore user backing store. 1024 * Restore user backing store.
diff --git a/arch/ia64/kernel/fsys.S b/arch/ia64/kernel/fsys.S
index 44841971f077..c1625c7e1779 100644
--- a/arch/ia64/kernel/fsys.S
+++ b/arch/ia64/kernel/fsys.S
@@ -61,13 +61,29 @@ ENTRY(fsys_getpid)
61 .prologue 61 .prologue
62 .altrp b6 62 .altrp b6
63 .body 63 .body
64 add r17=IA64_TASK_GROUP_LEADER_OFFSET,r16
65 ;;
66 ld8 r17=[r17] // r17 = current->group_leader
64 add r9=TI_FLAGS+IA64_TASK_SIZE,r16 67 add r9=TI_FLAGS+IA64_TASK_SIZE,r16
65 ;; 68 ;;
66 ld4 r9=[r9] 69 ld4 r9=[r9]
67 add r8=IA64_TASK_TGID_OFFSET,r16 70 add r17=IA64_TASK_TGIDLINK_OFFSET,r17
68 ;; 71 ;;
69 and r9=TIF_ALLWORK_MASK,r9 72 and r9=TIF_ALLWORK_MASK,r9
70 ld4 r8=[r8] // r8 = current->tgid 73 ld8 r17=[r17] // r17 = current->group_leader->pids[PIDTYPE_PID].pid
74 ;;
75 add r8=IA64_PID_LEVEL_OFFSET,r17
76 ;;
77 ld4 r8=[r8] // r8 = pid->level
78 add r17=IA64_PID_UPID_OFFSET,r17 // r17 = &pid->numbers[0]
79 ;;
80 shl r8=r8,IA64_UPID_SHIFT
81 ;;
82 add r17=r17,r8 // r17 = &pid->numbers[pid->level]
83 ;;
84 ld4 r8=[r17] // r8 = pid->numbers[pid->level].nr
85 ;;
86 mov r17=0
71 ;; 87 ;;
72 cmp.ne p8,p0=0,r9 88 cmp.ne p8,p0=0,r9
73(p8) br.spnt.many fsys_fallback_syscall 89(p8) br.spnt.many fsys_fallback_syscall
@@ -126,15 +142,25 @@ ENTRY(fsys_set_tid_address)
126 .altrp b6 142 .altrp b6
127 .body 143 .body
128 add r9=TI_FLAGS+IA64_TASK_SIZE,r16 144 add r9=TI_FLAGS+IA64_TASK_SIZE,r16
145 add r17=IA64_TASK_TGIDLINK_OFFSET,r16
129 ;; 146 ;;
130 ld4 r9=[r9] 147 ld4 r9=[r9]
131 tnat.z p6,p7=r32 // check argument register for being NaT 148 tnat.z p6,p7=r32 // check argument register for being NaT
149 ld8 r17=[r17] // r17 = current->pids[PIDTYPE_PID].pid
132 ;; 150 ;;
133 and r9=TIF_ALLWORK_MASK,r9 151 and r9=TIF_ALLWORK_MASK,r9
134 add r8=IA64_TASK_PID_OFFSET,r16 152 add r8=IA64_PID_LEVEL_OFFSET,r17
135 add r18=IA64_TASK_CLEAR_CHILD_TID_OFFSET,r16 153 add r18=IA64_TASK_CLEAR_CHILD_TID_OFFSET,r16
136 ;; 154 ;;
137 ld4 r8=[r8] 155 ld4 r8=[r8] // r8 = pid->level
156 add r17=IA64_PID_UPID_OFFSET,r17 // r17 = &pid->numbers[0]
157 ;;
158 shl r8=r8,IA64_UPID_SHIFT
159 ;;
160 add r17=r17,r8 // r17 = &pid->numbers[pid->level]
161 ;;
162 ld4 r8=[r17] // r8 = pid->numbers[pid->level].nr
163 ;;
138 cmp.ne p8,p0=0,r9 164 cmp.ne p8,p0=0,r9
139 mov r17=-1 165 mov r17=-1
140 ;; 166 ;;
@@ -210,27 +236,25 @@ ENTRY(fsys_gettimeofday)
210 // Note that instructions are optimized for McKinley. McKinley can 236 // Note that instructions are optimized for McKinley. McKinley can
211 // process two bundles simultaneously and therefore we continuously 237 // process two bundles simultaneously and therefore we continuously
212 // try to feed the CPU two bundles and then a stop. 238 // try to feed the CPU two bundles and then a stop.
213 // 239
214 // Additional note that code has changed a lot. Optimization is TBD.
215 // Comments begin with "?" are maybe outdated.
216 tnat.nz p6,p0 = r31 // ? branch deferred to fit later bundle
217 mov pr = r30,0xc000 // Set predicates according to function
218 add r2 = TI_FLAGS+IA64_TASK_SIZE,r16 240 add r2 = TI_FLAGS+IA64_TASK_SIZE,r16
241 tnat.nz p6,p0 = r31 // guard against Nat argument
242(p6) br.cond.spnt.few .fail_einval
219 movl r20 = fsyscall_gtod_data // load fsyscall gettimeofday data address 243 movl r20 = fsyscall_gtod_data // load fsyscall gettimeofday data address
220 ;; 244 ;;
245 ld4 r2 = [r2] // process work pending flags
221 movl r29 = itc_jitter_data // itc_jitter 246 movl r29 = itc_jitter_data // itc_jitter
222 add r22 = IA64_GTOD_WALL_TIME_OFFSET,r20 // wall_time 247 add r22 = IA64_GTOD_WALL_TIME_OFFSET,r20 // wall_time
223 ld4 r2 = [r2] // process work pending flags
224 ;;
225(p15) add r22 = IA64_GTOD_MONO_TIME_OFFSET,r20 // monotonic_time
226 add r21 = IA64_CLKSRC_MMIO_OFFSET,r20 248 add r21 = IA64_CLKSRC_MMIO_OFFSET,r20
227 add r19 = IA64_ITC_LASTCYCLE_OFFSET,r29 249 mov pr = r30,0xc000 // Set predicates according to function
250 ;;
228 and r2 = TIF_ALLWORK_MASK,r2 251 and r2 = TIF_ALLWORK_MASK,r2
229(p6) br.cond.spnt.few .fail_einval // ? deferred branch 252 add r19 = IA64_ITC_LASTCYCLE_OFFSET,r29
253(p15) add r22 = IA64_GTOD_MONO_TIME_OFFSET,r20 // monotonic_time
230 ;; 254 ;;
231 add r26 = IA64_CLKSRC_CYCLE_LAST_OFFSET,r20 // clksrc_cycle_last 255 add r26 = IA64_CLKSRC_CYCLE_LAST_OFFSET,r20 // clksrc_cycle_last
232 cmp.ne p6, p0 = 0, r2 // Fallback if work is scheduled 256 cmp.ne p6, p0 = 0, r2 // Fallback if work is scheduled
233(p6) br.cond.spnt.many fsys_fallback_syscall 257(p6) br.cond.spnt.many fsys_fallback_syscall
234 ;; 258 ;;
235 // Begin critical section 259 // Begin critical section
236.time_redo: 260.time_redo:
@@ -258,7 +282,6 @@ ENTRY(fsys_gettimeofday)
258(p8) mov r2 = ar.itc // CPU_TIMER. 36 clocks latency!!! 282(p8) mov r2 = ar.itc // CPU_TIMER. 36 clocks latency!!!
259(p9) ld8 r2 = [r30] // MMIO_TIMER. Could also have latency issues.. 283(p9) ld8 r2 = [r30] // MMIO_TIMER. Could also have latency issues..
260(p13) ld8 r25 = [r19] // get itc_lastcycle value 284(p13) ld8 r25 = [r19] // get itc_lastcycle value
261 ;; // ? could be removed by moving the last add upward
262 ld8 r9 = [r22],IA64_TIMESPEC_TV_NSEC_OFFSET // tv_sec 285 ld8 r9 = [r22],IA64_TIMESPEC_TV_NSEC_OFFSET // tv_sec
263 ;; 286 ;;
264 ld8 r8 = [r22],-IA64_TIMESPEC_TV_NSEC_OFFSET // tv_nsec 287 ld8 r8 = [r22],-IA64_TIMESPEC_TV_NSEC_OFFSET // tv_nsec
@@ -285,13 +308,12 @@ ENTRY(fsys_gettimeofday)
285EX(.fail_efault, probe.w.fault r31, 3) 308EX(.fail_efault, probe.w.fault r31, 3)
286 xmpy.l f8 = f8,f7 // nsec_per_cyc*(counter-last_counter) 309 xmpy.l f8 = f8,f7 // nsec_per_cyc*(counter-last_counter)
287 ;; 310 ;;
288 // ? simulate tbit.nz.or p7,p0 = r28,0
289 getf.sig r2 = f8 311 getf.sig r2 = f8
290 mf 312 mf
291 ;; 313 ;;
292 ld4 r10 = [r20] // gtod_lock.sequence 314 ld4 r10 = [r20] // gtod_lock.sequence
293 shr.u r2 = r2,r23 // shift by factor 315 shr.u r2 = r2,r23 // shift by factor
294 ;; // ? overloaded 3 bundles! 316 ;;
295 add r8 = r8,r2 // Add xtime.nsecs 317 add r8 = r8,r2 // Add xtime.nsecs
296 cmp4.ne p7,p0 = r28,r10 318 cmp4.ne p7,p0 = r28,r10
297(p7) br.cond.dpnt.few .time_redo // sequence number changed, redo 319(p7) br.cond.dpnt.few .time_redo // sequence number changed, redo
@@ -319,9 +341,9 @@ EX(.fail_efault, probe.w.fault r31, 3)
319EX(.fail_efault, probe.w.fault r23, 3) // This also costs 5 cycles 341EX(.fail_efault, probe.w.fault r23, 3) // This also costs 5 cycles
320(p14) xmpy.hu f8 = f8, f7 // xmpy has 5 cycles latency so use it 342(p14) xmpy.hu f8 = f8, f7 // xmpy has 5 cycles latency so use it
321 ;; 343 ;;
322 mov r8 = r0
323(p14) getf.sig r2 = f8 344(p14) getf.sig r2 = f8
324 ;; 345 ;;
346 mov r8 = r0
325(p14) shr.u r21 = r2, 4 347(p14) shr.u r21 = r2, 4
326 ;; 348 ;;
327EX(.fail_efault, st8 [r31] = r9) 349EX(.fail_efault, st8 [r31] = r9)
@@ -660,7 +682,11 @@ GLOBAL_ENTRY(fsys_bubble_down)
660 nop.i 0 682 nop.i 0
661 ;; 683 ;;
662 mov ar.rsc=0 // M2 set enforced lazy mode, pl 0, LE, loadrs=0 684 mov ar.rsc=0 // M2 set enforced lazy mode, pl 0, LE, loadrs=0
685#ifdef CONFIG_VIRT_CPU_ACCOUNTING
686 mov.m r30=ar.itc // M get cycle for accounting
687#else
663 nop.m 0 688 nop.m 0
689#endif
664 nop.i 0 690 nop.i 0
665 ;; 691 ;;
666 mov r23=ar.bspstore // M2 (12 cyc) save ar.bspstore 692 mov r23=ar.bspstore // M2 (12 cyc) save ar.bspstore
@@ -682,6 +708,28 @@ GLOBAL_ENTRY(fsys_bubble_down)
682 cmp.ne pKStk,pUStk=r0,r0 // A set pKStk <- 0, pUStk <- 1 708 cmp.ne pKStk,pUStk=r0,r0 // A set pKStk <- 0, pUStk <- 1
683 br.call.sptk.many b7=ia64_syscall_setup // B 709 br.call.sptk.many b7=ia64_syscall_setup // B
684 ;; 710 ;;
711#ifdef CONFIG_VIRT_CPU_ACCOUNTING
712 // mov.m r30=ar.itc is called in advance
713 add r16=TI_AC_STAMP+IA64_TASK_SIZE,r2
714 add r17=TI_AC_LEAVE+IA64_TASK_SIZE,r2
715 ;;
716 ld8 r18=[r16],TI_AC_STIME-TI_AC_STAMP // time at last check in kernel
717 ld8 r19=[r17],TI_AC_UTIME-TI_AC_LEAVE // time at leave kernel
718 ;;
719 ld8 r20=[r16],TI_AC_STAMP-TI_AC_STIME // cumulated stime
720 ld8 r21=[r17] // cumulated utime
721 sub r22=r19,r18 // stime before leave kernel
722 ;;
723 st8 [r16]=r30,TI_AC_STIME-TI_AC_STAMP // update stamp
724 sub r18=r30,r19 // elapsed time in user mode
725 ;;
726 add r20=r20,r22 // sum stime
727 add r21=r21,r18 // sum utime
728 ;;
729 st8 [r16]=r20 // update stime
730 st8 [r17]=r21 // update utime
731 ;;
732#endif
685 mov ar.rsc=0x3 // M2 set eager mode, pl 0, LE, loadrs=0 733 mov ar.rsc=0x3 // M2 set eager mode, pl 0, LE, loadrs=0
686 mov rp=r14 // I0 set the real return addr 734 mov rp=r14 // I0 set the real return addr
687 and r3=_TIF_SYSCALL_TRACEAUDIT,r3 // A 735 and r3=_TIF_SYSCALL_TRACEAUDIT,r3 // A
diff --git a/arch/ia64/kernel/head.S b/arch/ia64/kernel/head.S
index d3a41d5f8d12..ddeab4e36fd5 100644
--- a/arch/ia64/kernel/head.S
+++ b/arch/ia64/kernel/head.S
@@ -1002,6 +1002,26 @@ GLOBAL_ENTRY(sched_clock)
1002 br.ret.sptk.many rp 1002 br.ret.sptk.many rp
1003END(sched_clock) 1003END(sched_clock)
1004 1004
1005#ifdef CONFIG_VIRT_CPU_ACCOUNTING
1006GLOBAL_ENTRY(cycle_to_cputime)
1007 alloc r16=ar.pfs,1,0,0,0
1008 addl r8=THIS_CPU(cpu_info) + IA64_CPUINFO_NSEC_PER_CYC_OFFSET,r0
1009 ;;
1010 ldf8 f8=[r8]
1011 ;;
1012 setf.sig f9=r32
1013 ;;
1014 xmpy.lu f10=f9,f8 // calculate low 64 bits of 128-bit product (4 cyc)
1015 xmpy.hu f11=f9,f8 // calculate high 64 bits of 128-bit product
1016 ;;
1017 getf.sig r8=f10 // (5 cyc)
1018 getf.sig r9=f11
1019 ;;
1020 shrp r8=r9,r8,IA64_NSEC_PER_CYC_SHIFT
1021 br.ret.sptk.many rp
1022END(cycle_to_cputime)
1023#endif /* CONFIG_VIRT_CPU_ACCOUNTING */
1024
1005GLOBAL_ENTRY(start_kernel_thread) 1025GLOBAL_ENTRY(start_kernel_thread)
1006 .prologue 1026 .prologue
1007 .save rp, r0 // this is the end of the call-chain 1027 .save rp, r0 // this is the end of the call-chain
diff --git a/arch/ia64/kernel/irq_ia64.c b/arch/ia64/kernel/irq_ia64.c
index d8be23fbe6bc..5538471e8d68 100644
--- a/arch/ia64/kernel/irq_ia64.c
+++ b/arch/ia64/kernel/irq_ia64.c
@@ -472,7 +472,7 @@ ia64_handle_irq (ia64_vector vector, struct pt_regs *regs)
472 static unsigned char count; 472 static unsigned char count;
473 static long last_time; 473 static long last_time;
474 474
475 if (jiffies - last_time > 5*HZ) 475 if (time_after(jiffies, last_time + 5 * HZ))
476 count = 0; 476 count = 0;
477 if (++count < 5) { 477 if (++count < 5) {
478 last_time = jiffies; 478 last_time = jiffies;
diff --git a/arch/ia64/kernel/ivt.S b/arch/ia64/kernel/ivt.S
index 34f44d8be00d..6678c49daba3 100644
--- a/arch/ia64/kernel/ivt.S
+++ b/arch/ia64/kernel/ivt.S
@@ -805,8 +805,13 @@ ENTRY(break_fault)
805 805
806(p8) adds r28=16,r28 // A switch cr.iip to next bundle 806(p8) adds r28=16,r28 // A switch cr.iip to next bundle
807(p9) adds r8=1,r8 // A increment ei to next slot 807(p9) adds r8=1,r8 // A increment ei to next slot
808#ifdef CONFIG_VIRT_CPU_ACCOUNTING
809 ;;
810 mov b6=r30 // I0 setup syscall handler branch reg early
811#else
808 nop.i 0 812 nop.i 0
809 ;; 813 ;;
814#endif
810 815
811 mov.m r25=ar.unat // M2 (5 cyc) 816 mov.m r25=ar.unat // M2 (5 cyc)
812 dep r29=r8,r29,41,2 // I0 insert new ei into cr.ipsr 817 dep r29=r8,r29,41,2 // I0 insert new ei into cr.ipsr
@@ -817,7 +822,11 @@ ENTRY(break_fault)
817 // 822 //
818/////////////////////////////////////////////////////////////////////// 823///////////////////////////////////////////////////////////////////////
819 st1 [r16]=r0 // M2|3 clear current->thread.on_ustack flag 824 st1 [r16]=r0 // M2|3 clear current->thread.on_ustack flag
825#ifdef CONFIG_VIRT_CPU_ACCOUNTING
826 mov.m r30=ar.itc // M get cycle for accounting
827#else
820 mov b6=r30 // I0 setup syscall handler branch reg early 828 mov b6=r30 // I0 setup syscall handler branch reg early
829#endif
821 cmp.eq pKStk,pUStk=r0,r17 // A were we on kernel stacks already? 830 cmp.eq pKStk,pUStk=r0,r17 // A were we on kernel stacks already?
822 831
823 and r9=_TIF_SYSCALL_TRACEAUDIT,r9 // A mask trace or audit 832 and r9=_TIF_SYSCALL_TRACEAUDIT,r9 // A mask trace or audit
@@ -829,6 +838,30 @@ ENTRY(break_fault)
829 cmp.eq p14,p0=r9,r0 // A are syscalls being traced/audited? 838 cmp.eq p14,p0=r9,r0 // A are syscalls being traced/audited?
830 br.call.sptk.many b7=ia64_syscall_setup // B 839 br.call.sptk.many b7=ia64_syscall_setup // B
8311: 8401:
841#ifdef CONFIG_VIRT_CPU_ACCOUNTING
842 // mov.m r30=ar.itc is called in advance, and r13 is current
843 add r16=TI_AC_STAMP+IA64_TASK_SIZE,r13 // A
844 add r17=TI_AC_LEAVE+IA64_TASK_SIZE,r13 // A
845(pKStk) br.cond.spnt .skip_accounting // B unlikely skip
846 ;;
847 ld8 r18=[r16],TI_AC_STIME-TI_AC_STAMP // M get last stamp
848 ld8 r19=[r17],TI_AC_UTIME-TI_AC_LEAVE // M time at leave
849 ;;
850 ld8 r20=[r16],TI_AC_STAMP-TI_AC_STIME // M cumulated stime
851 ld8 r21=[r17] // M cumulated utime
852 sub r22=r19,r18 // A stime before leave
853 ;;
854 st8 [r16]=r30,TI_AC_STIME-TI_AC_STAMP // M update stamp
855 sub r18=r30,r19 // A elapsed time in user
856 ;;
857 add r20=r20,r22 // A sum stime
858 add r21=r21,r18 // A sum utime
859 ;;
860 st8 [r16]=r20 // M update stime
861 st8 [r17]=r21 // M update utime
862 ;;
863.skip_accounting:
864#endif
832 mov ar.rsc=0x3 // M2 set eager mode, pl 0, LE, loadrs=0 865 mov ar.rsc=0x3 // M2 set eager mode, pl 0, LE, loadrs=0
833 nop 0 866 nop 0
834 bsw.1 // B (6 cyc) regs are saved, switch to bank 1 867 bsw.1 // B (6 cyc) regs are saved, switch to bank 1
@@ -928,6 +961,7 @@ END(interrupt)
928 * - r27: saved ar.rsc 961 * - r27: saved ar.rsc
929 * - r28: saved cr.iip 962 * - r28: saved cr.iip
930 * - r29: saved cr.ipsr 963 * - r29: saved cr.ipsr
964 * - r30: ar.itc for accounting (don't touch)
931 * - r31: saved pr 965 * - r31: saved pr
932 * - b0: original contents (to be saved) 966 * - b0: original contents (to be saved)
933 * On exit: 967 * On exit:
@@ -1090,6 +1124,41 @@ END(dispatch_illegal_op_fault)
1090 DBG_FAULT(16) 1124 DBG_FAULT(16)
1091 FAULT(16) 1125 FAULT(16)
1092 1126
1127#ifdef CONFIG_VIRT_CPU_ACCOUNTING
1128 /*
1129 * There is no particular reason for this code to be here, other than
1130 * that there happens to be space here that would go unused otherwise.
1131 * If this fault ever gets "unreserved", simply moved the following
1132 * code to a more suitable spot...
1133 *
1134 * account_sys_enter is called from SAVE_MIN* macros if accounting is
1135 * enabled and if the macro is entered from user mode.
1136 */
1137ENTRY(account_sys_enter)
1138 // mov.m r20=ar.itc is called in advance, and r13 is current
1139 add r16=TI_AC_STAMP+IA64_TASK_SIZE,r13
1140 add r17=TI_AC_LEAVE+IA64_TASK_SIZE,r13
1141 ;;
1142 ld8 r18=[r16],TI_AC_STIME-TI_AC_STAMP // time at last check in kernel
1143 ld8 r19=[r17],TI_AC_UTIME-TI_AC_LEAVE // time at left from kernel
1144 ;;
1145 ld8 r23=[r16],TI_AC_STAMP-TI_AC_STIME // cumulated stime
1146 ld8 r21=[r17] // cumulated utime
1147 sub r22=r19,r18 // stime before leave kernel
1148 ;;
1149 st8 [r16]=r20,TI_AC_STIME-TI_AC_STAMP // update stamp
1150 sub r18=r20,r19 // elapsed time in user mode
1151 ;;
1152 add r23=r23,r22 // sum stime
1153 add r21=r21,r18 // sum utime
1154 ;;
1155 st8 [r16]=r23 // update stime
1156 st8 [r17]=r21 // update utime
1157 ;;
1158 br.ret.sptk.many rp
1159END(account_sys_enter)
1160#endif
1161
1093 .org ia64_ivt+0x4400 1162 .org ia64_ivt+0x4400
1094///////////////////////////////////////////////////////////////////////////////////////// 1163/////////////////////////////////////////////////////////////////////////////////////////
1095// 0x4400 Entry 17 (size 64 bundles) Reserved 1164// 0x4400 Entry 17 (size 64 bundles) Reserved
diff --git a/arch/ia64/kernel/kprobes.c b/arch/ia64/kernel/kprobes.c
index 8d9a446a0d17..233434f4f88f 100644
--- a/arch/ia64/kernel/kprobes.c
+++ b/arch/ia64/kernel/kprobes.c
@@ -78,6 +78,20 @@ static enum instruction_type bundle_encoding[32][3] = {
78 { u, u, u }, /* 1F */ 78 { u, u, u }, /* 1F */
79}; 79};
80 80
81/* Insert a long branch code */
82static void __kprobes set_brl_inst(void *from, void *to)
83{
84 s64 rel = ((s64) to - (s64) from) >> 4;
85 bundle_t *brl;
86 brl = (bundle_t *) ((u64) from & ~0xf);
87 brl->quad0.template = 0x05; /* [MLX](stop) */
88 brl->quad0.slot0 = NOP_M_INST; /* nop.m 0x0 */
89 brl->quad0.slot1_p0 = ((rel >> 20) & 0x7fffffffff) << 2;
90 brl->quad1.slot1_p1 = (((rel >> 20) & 0x7fffffffff) << 2) >> (64 - 46);
91 /* brl.cond.sptk.many.clr rel<<4 (qp=0) */
92 brl->quad1.slot2 = BRL_INST(rel >> 59, rel & 0xfffff);
93}
94
81/* 95/*
82 * In this function we check to see if the instruction 96 * In this function we check to see if the instruction
83 * is IP relative instruction and update the kprobe 97 * is IP relative instruction and update the kprobe
@@ -496,6 +510,77 @@ void __kprobes arch_prepare_kretprobe(struct kretprobe_instance *ri,
496 regs->b0 = ((struct fnptr *)kretprobe_trampoline)->ip; 510 regs->b0 = ((struct fnptr *)kretprobe_trampoline)->ip;
497} 511}
498 512
513/* Check the instruction in the slot is break */
514static int __kprobes __is_ia64_break_inst(bundle_t *bundle, uint slot)
515{
516 unsigned int major_opcode;
517 unsigned int template = bundle->quad0.template;
518 unsigned long kprobe_inst;
519
520 /* Move to slot 2, if bundle is MLX type and kprobe slot is 1 */
521 if (slot == 1 && bundle_encoding[template][1] == L)
522 slot++;
523
524 /* Get Kprobe probe instruction at given slot*/
525 get_kprobe_inst(bundle, slot, &kprobe_inst, &major_opcode);
526
527 /* For break instruction,
528 * Bits 37:40 Major opcode to be zero
529 * Bits 27:32 X6 to be zero
530 * Bits 32:35 X3 to be zero
531 */
532 if (major_opcode || ((kprobe_inst >> 27) & 0x1FF)) {
533 /* Not a break instruction */
534 return 0;
535 }
536
537 /* Is a break instruction */
538 return 1;
539}
540
541/*
542 * In this function, we check whether the target bundle modifies IP or
543 * it triggers an exception. If so, it cannot be boostable.
544 */
545static int __kprobes can_boost(bundle_t *bundle, uint slot,
546 unsigned long bundle_addr)
547{
548 unsigned int template = bundle->quad0.template;
549
550 do {
551 if (search_exception_tables(bundle_addr + slot) ||
552 __is_ia64_break_inst(bundle, slot))
553 return 0; /* exception may occur in this bundle*/
554 } while ((++slot) < 3);
555 template &= 0x1e;
556 if (template >= 0x10 /* including B unit */ ||
557 template == 0x04 /* including X unit */ ||
558 template == 0x06) /* undefined */
559 return 0;
560
561 return 1;
562}
563
564/* Prepare long jump bundle and disables other boosters if need */
565static void __kprobes prepare_booster(struct kprobe *p)
566{
567 unsigned long addr = (unsigned long)p->addr & ~0xFULL;
568 unsigned int slot = (unsigned long)p->addr & 0xf;
569 struct kprobe *other_kp;
570
571 if (can_boost(&p->ainsn.insn[0].bundle, slot, addr)) {
572 set_brl_inst(&p->ainsn.insn[1].bundle, (bundle_t *)addr + 1);
573 p->ainsn.inst_flag |= INST_FLAG_BOOSTABLE;
574 }
575
576 /* disables boosters in previous slots */
577 for (; addr < (unsigned long)p->addr; addr++) {
578 other_kp = get_kprobe((void *)addr);
579 if (other_kp)
580 other_kp->ainsn.inst_flag &= ~INST_FLAG_BOOSTABLE;
581 }
582}
583
499int __kprobes arch_prepare_kprobe(struct kprobe *p) 584int __kprobes arch_prepare_kprobe(struct kprobe *p)
500{ 585{
501 unsigned long addr = (unsigned long) p->addr; 586 unsigned long addr = (unsigned long) p->addr;
@@ -530,6 +615,8 @@ int __kprobes arch_prepare_kprobe(struct kprobe *p)
530 615
531 prepare_break_inst(template, slot, major_opcode, kprobe_inst, p, qp); 616 prepare_break_inst(template, slot, major_opcode, kprobe_inst, p, qp);
532 617
618 prepare_booster(p);
619
533 return 0; 620 return 0;
534} 621}
535 622
@@ -543,7 +630,9 @@ void __kprobes arch_arm_kprobe(struct kprobe *p)
543 src = &p->opcode.bundle; 630 src = &p->opcode.bundle;
544 631
545 flush_icache_range((unsigned long)p->ainsn.insn, 632 flush_icache_range((unsigned long)p->ainsn.insn,
546 (unsigned long)p->ainsn.insn + sizeof(kprobe_opcode_t)); 633 (unsigned long)p->ainsn.insn +
634 sizeof(kprobe_opcode_t) * MAX_INSN_SIZE);
635
547 switch (p->ainsn.slot) { 636 switch (p->ainsn.slot) {
548 case 0: 637 case 0:
549 dest->quad0.slot0 = src->quad0.slot0; 638 dest->quad0.slot0 = src->quad0.slot0;
@@ -584,13 +673,13 @@ void __kprobes arch_disarm_kprobe(struct kprobe *p)
584void __kprobes arch_remove_kprobe(struct kprobe *p) 673void __kprobes arch_remove_kprobe(struct kprobe *p)
585{ 674{
586 mutex_lock(&kprobe_mutex); 675 mutex_lock(&kprobe_mutex);
587 free_insn_slot(p->ainsn.insn, 0); 676 free_insn_slot(p->ainsn.insn, p->ainsn.inst_flag & INST_FLAG_BOOSTABLE);
588 mutex_unlock(&kprobe_mutex); 677 mutex_unlock(&kprobe_mutex);
589} 678}
590/* 679/*
591 * We are resuming execution after a single step fault, so the pt_regs 680 * We are resuming execution after a single step fault, so the pt_regs
592 * structure reflects the register state after we executed the instruction 681 * structure reflects the register state after we executed the instruction
593 * located in the kprobe (p->ainsn.insn.bundle). We still need to adjust 682 * located in the kprobe (p->ainsn.insn->bundle). We still need to adjust
594 * the ip to point back to the original stack address. To set the IP address 683 * the ip to point back to the original stack address. To set the IP address
595 * to original stack address, handle the case where we need to fixup the 684 * to original stack address, handle the case where we need to fixup the
596 * relative IP address and/or fixup branch register. 685 * relative IP address and/or fixup branch register.
@@ -607,7 +696,7 @@ static void __kprobes resume_execution(struct kprobe *p, struct pt_regs *regs)
607 if (slot == 1 && bundle_encoding[template][1] == L) 696 if (slot == 1 && bundle_encoding[template][1] == L)
608 slot = 2; 697 slot = 2;
609 698
610 if (p->ainsn.inst_flag) { 699 if (p->ainsn.inst_flag & ~INST_FLAG_BOOSTABLE) {
611 700
612 if (p->ainsn.inst_flag & INST_FLAG_FIX_RELATIVE_IP_ADDR) { 701 if (p->ainsn.inst_flag & INST_FLAG_FIX_RELATIVE_IP_ADDR) {
613 /* Fix relative IP address */ 702 /* Fix relative IP address */
@@ -686,33 +775,12 @@ static void __kprobes prepare_ss(struct kprobe *p, struct pt_regs *regs)
686static int __kprobes is_ia64_break_inst(struct pt_regs *regs) 775static int __kprobes is_ia64_break_inst(struct pt_regs *regs)
687{ 776{
688 unsigned int slot = ia64_psr(regs)->ri; 777 unsigned int slot = ia64_psr(regs)->ri;
689 unsigned int template, major_opcode;
690 unsigned long kprobe_inst;
691 unsigned long *kprobe_addr = (unsigned long *)regs->cr_iip; 778 unsigned long *kprobe_addr = (unsigned long *)regs->cr_iip;
692 bundle_t bundle; 779 bundle_t bundle;
693 780
694 memcpy(&bundle, kprobe_addr, sizeof(bundle_t)); 781 memcpy(&bundle, kprobe_addr, sizeof(bundle_t));
695 template = bundle.quad0.template;
696
697 /* Move to slot 2, if bundle is MLX type and kprobe slot is 1 */
698 if (slot == 1 && bundle_encoding[template][1] == L)
699 slot++;
700 782
701 /* Get Kprobe probe instruction at given slot*/ 783 return __is_ia64_break_inst(&bundle, slot);
702 get_kprobe_inst(&bundle, slot, &kprobe_inst, &major_opcode);
703
704 /* For break instruction,
705 * Bits 37:40 Major opcode to be zero
706 * Bits 27:32 X6 to be zero
707 * Bits 32:35 X3 to be zero
708 */
709 if (major_opcode || ((kprobe_inst >> 27) & 0x1FF) ) {
710 /* Not a break instruction */
711 return 0;
712 }
713
714 /* Is a break instruction */
715 return 1;
716} 784}
717 785
718static int __kprobes pre_kprobes_handler(struct die_args *args) 786static int __kprobes pre_kprobes_handler(struct die_args *args)
@@ -802,6 +870,19 @@ static int __kprobes pre_kprobes_handler(struct die_args *args)
802 return 1; 870 return 1;
803 871
804ss_probe: 872ss_probe:
873#if !defined(CONFIG_PREEMPT) || defined(CONFIG_PM)
874 if (p->ainsn.inst_flag == INST_FLAG_BOOSTABLE && !p->post_handler) {
875 /* Boost up -- we can execute copied instructions directly */
876 ia64_psr(regs)->ri = p->ainsn.slot;
877 regs->cr_iip = (unsigned long)&p->ainsn.insn->bundle & ~0xFULL;
878 /* turn single stepping off */
879 ia64_psr(regs)->ss = 0;
880
881 reset_current_kprobe();
882 preempt_enable_no_resched();
883 return 1;
884 }
885#endif
805 prepare_ss(p, regs); 886 prepare_ss(p, regs);
806 kcb->kprobe_status = KPROBE_HIT_SS; 887 kcb->kprobe_status = KPROBE_HIT_SS;
807 return 1; 888 return 1;
diff --git a/arch/ia64/kernel/mca.c b/arch/ia64/kernel/mca.c
index 6c18221dba36..e51bced3b0fa 100644
--- a/arch/ia64/kernel/mca.c
+++ b/arch/ia64/kernel/mca.c
@@ -69,6 +69,7 @@
69 * 2007-04-27 Russ Anderson <rja@sgi.com> 69 * 2007-04-27 Russ Anderson <rja@sgi.com>
70 * Support multiple cpus going through OS_MCA in the same event. 70 * Support multiple cpus going through OS_MCA in the same event.
71 */ 71 */
72#include <linux/jiffies.h>
72#include <linux/types.h> 73#include <linux/types.h>
73#include <linux/init.h> 74#include <linux/init.h>
74#include <linux/sched.h> 75#include <linux/sched.h>
@@ -97,6 +98,7 @@
97 98
98#include <asm/irq.h> 99#include <asm/irq.h>
99#include <asm/hw_irq.h> 100#include <asm/hw_irq.h>
101#include <asm/tlb.h>
100 102
101#include "mca_drv.h" 103#include "mca_drv.h"
102#include "entry.h" 104#include "entry.h"
@@ -112,6 +114,7 @@ DEFINE_PER_CPU(u64, ia64_mca_data); /* == __per_cpu_mca[smp_processor_id()] */
112DEFINE_PER_CPU(u64, ia64_mca_per_cpu_pte); /* PTE to map per-CPU area */ 114DEFINE_PER_CPU(u64, ia64_mca_per_cpu_pte); /* PTE to map per-CPU area */
113DEFINE_PER_CPU(u64, ia64_mca_pal_pte); /* PTE to map PAL code */ 115DEFINE_PER_CPU(u64, ia64_mca_pal_pte); /* PTE to map PAL code */
114DEFINE_PER_CPU(u64, ia64_mca_pal_base); /* vaddr PAL code granule */ 116DEFINE_PER_CPU(u64, ia64_mca_pal_base); /* vaddr PAL code granule */
117DEFINE_PER_CPU(u64, ia64_mca_tr_reload); /* Flag for TR reload */
115 118
116unsigned long __per_cpu_mca[NR_CPUS]; 119unsigned long __per_cpu_mca[NR_CPUS];
117 120
@@ -293,7 +296,8 @@ static void ia64_mlogbuf_dump_from_init(void)
293 if (mlogbuf_finished) 296 if (mlogbuf_finished)
294 return; 297 return;
295 298
296 if (mlogbuf_timestamp && (mlogbuf_timestamp + 30*HZ > jiffies)) { 299 if (mlogbuf_timestamp &&
300 time_before(jiffies, mlogbuf_timestamp + 30 * HZ)) {
297 printk(KERN_ERR "INIT: mlogbuf_dump is interrupted by INIT " 301 printk(KERN_ERR "INIT: mlogbuf_dump is interrupted by INIT "
298 " and the system seems to be messed up.\n"); 302 " and the system seems to be messed up.\n");
299 ia64_mlogbuf_finish(0); 303 ia64_mlogbuf_finish(0);
@@ -1182,6 +1186,49 @@ all_in:
1182 return; 1186 return;
1183} 1187}
1184 1188
1189/* mca_insert_tr
1190 *
1191 * Switch rid when TR reload and needed!
1192 * iord: 1: itr, 2: itr;
1193 *
1194*/
1195static void mca_insert_tr(u64 iord)
1196{
1197
1198 int i;
1199 u64 old_rr;
1200 struct ia64_tr_entry *p;
1201 unsigned long psr;
1202 int cpu = smp_processor_id();
1203
1204 psr = ia64_clear_ic();
1205 for (i = IA64_TR_ALLOC_BASE; i < IA64_TR_ALLOC_MAX; i++) {
1206 p = &__per_cpu_idtrs[cpu][iord-1][i];
1207 if (p->pte & 0x1) {
1208 old_rr = ia64_get_rr(p->ifa);
1209 if (old_rr != p->rr) {
1210 ia64_set_rr(p->ifa, p->rr);
1211 ia64_srlz_d();
1212 }
1213 ia64_ptr(iord, p->ifa, p->itir >> 2);
1214 ia64_srlz_i();
1215 if (iord & 0x1) {
1216 ia64_itr(0x1, i, p->ifa, p->pte, p->itir >> 2);
1217 ia64_srlz_i();
1218 }
1219 if (iord & 0x2) {
1220 ia64_itr(0x2, i, p->ifa, p->pte, p->itir >> 2);
1221 ia64_srlz_i();
1222 }
1223 if (old_rr != p->rr) {
1224 ia64_set_rr(p->ifa, old_rr);
1225 ia64_srlz_d();
1226 }
1227 }
1228 }
1229 ia64_set_psr(psr);
1230}
1231
1185/* 1232/*
1186 * ia64_mca_handler 1233 * ia64_mca_handler
1187 * 1234 *
@@ -1266,16 +1313,17 @@ ia64_mca_handler(struct pt_regs *regs, struct switch_stack *sw,
1266 } else { 1313 } else {
1267 /* Dump buffered message to console */ 1314 /* Dump buffered message to console */
1268 ia64_mlogbuf_finish(1); 1315 ia64_mlogbuf_finish(1);
1269#ifdef CONFIG_KEXEC
1270 atomic_set(&kdump_in_progress, 1);
1271 monarch_cpu = -1;
1272#endif
1273 } 1316 }
1317
1318 if (__get_cpu_var(ia64_mca_tr_reload)) {
1319 mca_insert_tr(0x1); /*Reload dynamic itrs*/
1320 mca_insert_tr(0x2); /*Reload dynamic itrs*/
1321 }
1322
1274 if (notify_die(DIE_MCA_MONARCH_LEAVE, "MCA", regs, (long)&nd, 0, recover) 1323 if (notify_die(DIE_MCA_MONARCH_LEAVE, "MCA", regs, (long)&nd, 0, recover)
1275 == NOTIFY_STOP) 1324 == NOTIFY_STOP)
1276 ia64_mca_spin(__func__); 1325 ia64_mca_spin(__func__);
1277 1326
1278
1279 if (atomic_dec_return(&mca_count) > 0) { 1327 if (atomic_dec_return(&mca_count) > 0) {
1280 int i; 1328 int i;
1281 1329
diff --git a/arch/ia64/kernel/mca_asm.S b/arch/ia64/kernel/mca_asm.S
index 8bc7d259e0c6..a06d46548ff9 100644
--- a/arch/ia64/kernel/mca_asm.S
+++ b/arch/ia64/kernel/mca_asm.S
@@ -219,8 +219,13 @@ ia64_reload_tr:
219 mov r20=IA64_TR_CURRENT_STACK 219 mov r20=IA64_TR_CURRENT_STACK
220 ;; 220 ;;
221 itr.d dtr[r20]=r16 221 itr.d dtr[r20]=r16
222 GET_THIS_PADDR(r2, ia64_mca_tr_reload)
223 mov r18 = 1
222 ;; 224 ;;
223 srlz.d 225 srlz.d
226 ;;
227 st8 [r2] =r18
228 ;;
224 229
225done_tlb_purge_and_reload: 230done_tlb_purge_and_reload:
226 231
diff --git a/arch/ia64/kernel/minstate.h b/arch/ia64/kernel/minstate.h
index c9ac8bada786..7c548ac52bbc 100644
--- a/arch/ia64/kernel/minstate.h
+++ b/arch/ia64/kernel/minstate.h
@@ -3,6 +3,18 @@
3 3
4#include "entry.h" 4#include "entry.h"
5 5
6#ifdef CONFIG_VIRT_CPU_ACCOUNTING
7/* read ar.itc in advance, and use it before leaving bank 0 */
8#define ACCOUNT_GET_STAMP \
9(pUStk) mov.m r20=ar.itc;
10#define ACCOUNT_SYS_ENTER \
11(pUStk) br.call.spnt rp=account_sys_enter \
12 ;;
13#else
14#define ACCOUNT_GET_STAMP
15#define ACCOUNT_SYS_ENTER
16#endif
17
6/* 18/*
7 * DO_SAVE_MIN switches to the kernel stacks (if necessary) and saves 19 * DO_SAVE_MIN switches to the kernel stacks (if necessary) and saves
8 * the minimum state necessary that allows us to turn psr.ic back 20 * the minimum state necessary that allows us to turn psr.ic back
@@ -122,11 +134,13 @@
122 ;; \ 134 ;; \
123.mem.offset 0,0; st8.spill [r16]=r2,16; \ 135.mem.offset 0,0; st8.spill [r16]=r2,16; \
124.mem.offset 8,0; st8.spill [r17]=r3,16; \ 136.mem.offset 8,0; st8.spill [r17]=r3,16; \
137 ACCOUNT_GET_STAMP \
125 adds r2=IA64_PT_REGS_R16_OFFSET,r1; \ 138 adds r2=IA64_PT_REGS_R16_OFFSET,r1; \
126 ;; \ 139 ;; \
127 EXTRA; \ 140 EXTRA; \
128 movl r1=__gp; /* establish kernel global pointer */ \ 141 movl r1=__gp; /* establish kernel global pointer */ \
129 ;; \ 142 ;; \
143 ACCOUNT_SYS_ENTER \
130 bsw.1; /* switch back to bank 1 (must be last in insn group) */ \ 144 bsw.1; /* switch back to bank 1 (must be last in insn group) */ \
131 ;; 145 ;;
132 146
diff --git a/arch/ia64/kernel/numa.c b/arch/ia64/kernel/numa.c
index a78b45f5fe2f..c93420c97409 100644
--- a/arch/ia64/kernel/numa.c
+++ b/arch/ia64/kernel/numa.c
@@ -73,7 +73,7 @@ void __init build_cpu_to_node_map(void)
73 for(node=0; node < MAX_NUMNODES; node++) 73 for(node=0; node < MAX_NUMNODES; node++)
74 cpus_clear(node_to_cpu_mask[node]); 74 cpus_clear(node_to_cpu_mask[node]);
75 75
76 for(cpu = 0; cpu < NR_CPUS; ++cpu) { 76 for_each_possible_early_cpu(cpu) {
77 node = -1; 77 node = -1;
78 for (i = 0; i < NR_CPUS; ++i) 78 for (i = 0; i < NR_CPUS; ++i)
79 if (cpu_physical_id(cpu) == node_cpuid[i].phys_id) { 79 if (cpu_physical_id(cpu) == node_cpuid[i].phys_id) {
diff --git a/arch/ia64/kernel/patch.c b/arch/ia64/kernel/patch.c
index 2cb9425e0421..e0dca8743dbb 100644
--- a/arch/ia64/kernel/patch.c
+++ b/arch/ia64/kernel/patch.c
@@ -135,10 +135,10 @@ ia64_patch_mckinley_e9 (unsigned long start, unsigned long end)
135 135
136 while (offp < (s32 *) end) { 136 while (offp < (s32 *) end) {
137 wp = (u64 *) ia64_imva((char *) offp + *offp); 137 wp = (u64 *) ia64_imva((char *) offp + *offp);
138 wp[0] = 0x0000000100000000UL; /* nop.m 0; nop.i 0; nop.i 0 */ 138 wp[0] = 0x0000000100000011UL; /* nop.m 0; nop.i 0; br.ret.sptk.many b6 */
139 wp[1] = 0x0004000000000200UL; 139 wp[1] = 0x0084006880000200UL;
140 wp[2] = 0x0000000100000011UL; /* nop.m 0; nop.i 0; br.ret.sptk.many b6 */ 140 wp[2] = 0x0000000100000000UL; /* nop.m 0; nop.i 0; nop.i 0 */
141 wp[3] = 0x0084006880000200UL; 141 wp[3] = 0x0004000000000200UL;
142 ia64_fc(wp); ia64_fc(wp + 2); 142 ia64_fc(wp); ia64_fc(wp + 2);
143 ++offp; 143 ++offp;
144 } 144 }
diff --git a/arch/ia64/kernel/perfmon.c b/arch/ia64/kernel/perfmon.c
index a2aabfdc80d9..d1d24f4598da 100644
--- a/arch/ia64/kernel/perfmon.c
+++ b/arch/ia64/kernel/perfmon.c
@@ -4204,10 +4204,10 @@ pfm_check_task_exist(pfm_context_t *ctx)
4204 do_each_thread (g, t) { 4204 do_each_thread (g, t) {
4205 if (t->thread.pfm_context == ctx) { 4205 if (t->thread.pfm_context == ctx) {
4206 ret = 0; 4206 ret = 0;
4207 break; 4207 goto out;
4208 } 4208 }
4209 } while_each_thread (g, t); 4209 } while_each_thread (g, t);
4210 4210out:
4211 read_unlock(&tasklist_lock); 4211 read_unlock(&tasklist_lock);
4212 4212
4213 DPRINT(("pfm_check_task_exist: ret=%d ctx=%p\n", ret, ctx)); 4213 DPRINT(("pfm_check_task_exist: ret=%d ctx=%p\n", ret, ctx));
diff --git a/arch/ia64/kernel/process.c b/arch/ia64/kernel/process.c
index 49937a383b23..a5ea817cbcbf 100644
--- a/arch/ia64/kernel/process.c
+++ b/arch/ia64/kernel/process.c
@@ -625,21 +625,6 @@ do_dump_fpu (struct unw_frame_info *info, void *arg)
625 do_dump_task_fpu(current, info, arg); 625 do_dump_task_fpu(current, info, arg);
626} 626}
627 627
628int
629dump_task_regs(struct task_struct *task, elf_gregset_t *regs)
630{
631 struct unw_frame_info tcore_info;
632
633 if (current == task) {
634 unw_init_running(do_copy_regs, regs);
635 } else {
636 memset(&tcore_info, 0, sizeof(tcore_info));
637 unw_init_from_blocked_task(&tcore_info, task);
638 do_copy_task_regs(task, &tcore_info, regs);
639 }
640 return 1;
641}
642
643void 628void
644ia64_elf_core_copy_regs (struct pt_regs *pt, elf_gregset_t dst) 629ia64_elf_core_copy_regs (struct pt_regs *pt, elf_gregset_t dst)
645{ 630{
@@ -647,21 +632,6 @@ ia64_elf_core_copy_regs (struct pt_regs *pt, elf_gregset_t dst)
647} 632}
648 633
649int 634int
650dump_task_fpu (struct task_struct *task, elf_fpregset_t *dst)
651{
652 struct unw_frame_info tcore_info;
653
654 if (current == task) {
655 unw_init_running(do_dump_fpu, dst);
656 } else {
657 memset(&tcore_info, 0, sizeof(tcore_info));
658 unw_init_from_blocked_task(&tcore_info, task);
659 do_dump_task_fpu(task, &tcore_info, dst);
660 }
661 return 1;
662}
663
664int
665dump_fpu (struct pt_regs *pt, elf_fpregset_t dst) 635dump_fpu (struct pt_regs *pt, elf_fpregset_t dst)
666{ 636{
667 unw_init_running(do_dump_fpu, dst); 637 unw_init_running(do_dump_fpu, dst);
diff --git a/arch/ia64/kernel/ptrace.c b/arch/ia64/kernel/ptrace.c
index ab784ec4319d..2a9943b5947f 100644
--- a/arch/ia64/kernel/ptrace.c
+++ b/arch/ia64/kernel/ptrace.c
@@ -3,6 +3,9 @@
3 * 3 *
4 * Copyright (C) 1999-2005 Hewlett-Packard Co 4 * Copyright (C) 1999-2005 Hewlett-Packard Co
5 * David Mosberger-Tang <davidm@hpl.hp.com> 5 * David Mosberger-Tang <davidm@hpl.hp.com>
6 * Copyright (C) 2006 Intel Co
7 * 2006-08-12 - IA64 Native Utrace implementation support added by
8 * Anil S Keshavamurthy <anil.s.keshavamurthy@intel.com>
6 * 9 *
7 * Derived from the x86 and Alpha versions. 10 * Derived from the x86 and Alpha versions.
8 */ 11 */
@@ -17,6 +20,8 @@
17#include <linux/security.h> 20#include <linux/security.h>
18#include <linux/audit.h> 21#include <linux/audit.h>
19#include <linux/signal.h> 22#include <linux/signal.h>
23#include <linux/regset.h>
24#include <linux/elf.h>
20 25
21#include <asm/pgtable.h> 26#include <asm/pgtable.h>
22#include <asm/processor.h> 27#include <asm/processor.h>
@@ -740,25 +745,6 @@ ia64_sync_fph (struct task_struct *task)
740 psr->dfh = 1; 745 psr->dfh = 1;
741} 746}
742 747
743static int
744access_fr (struct unw_frame_info *info, int regnum, int hi,
745 unsigned long *data, int write_access)
746{
747 struct ia64_fpreg fpval;
748 int ret;
749
750 ret = unw_get_fr(info, regnum, &fpval);
751 if (ret < 0)
752 return ret;
753
754 if (write_access) {
755 fpval.u.bits[hi] = *data;
756 ret = unw_set_fr(info, regnum, fpval);
757 } else
758 *data = fpval.u.bits[hi];
759 return ret;
760}
761
762/* 748/*
763 * Change the machine-state of CHILD such that it will return via the normal 749 * Change the machine-state of CHILD such that it will return via the normal
764 * kernel exit-path, rather than the syscall-exit path. 750 * kernel exit-path, rather than the syscall-exit path.
@@ -860,309 +846,7 @@ access_nat_bits (struct task_struct *child, struct pt_regs *pt,
860 846
861static int 847static int
862access_uarea (struct task_struct *child, unsigned long addr, 848access_uarea (struct task_struct *child, unsigned long addr,
863 unsigned long *data, int write_access) 849 unsigned long *data, int write_access);
864{
865 unsigned long *ptr, regnum, urbs_end, cfm;
866 struct switch_stack *sw;
867 struct pt_regs *pt;
868# define pt_reg_addr(pt, reg) ((void *) \
869 ((unsigned long) (pt) \
870 + offsetof(struct pt_regs, reg)))
871
872
873 pt = task_pt_regs(child);
874 sw = (struct switch_stack *) (child->thread.ksp + 16);
875
876 if ((addr & 0x7) != 0) {
877 dprintk("ptrace: unaligned register address 0x%lx\n", addr);
878 return -1;
879 }
880
881 if (addr < PT_F127 + 16) {
882 /* accessing fph */
883 if (write_access)
884 ia64_sync_fph(child);
885 else
886 ia64_flush_fph(child);
887 ptr = (unsigned long *)
888 ((unsigned long) &child->thread.fph + addr);
889 } else if ((addr >= PT_F10) && (addr < PT_F11 + 16)) {
890 /* scratch registers untouched by kernel (saved in pt_regs) */
891 ptr = pt_reg_addr(pt, f10) + (addr - PT_F10);
892 } else if (addr >= PT_F12 && addr < PT_F15 + 16) {
893 /*
894 * Scratch registers untouched by kernel (saved in
895 * switch_stack).
896 */
897 ptr = (unsigned long *) ((long) sw
898 + (addr - PT_NAT_BITS - 32));
899 } else if (addr < PT_AR_LC + 8) {
900 /* preserved state: */
901 struct unw_frame_info info;
902 char nat = 0;
903 int ret;
904
905 unw_init_from_blocked_task(&info, child);
906 if (unw_unwind_to_user(&info) < 0)
907 return -1;
908
909 switch (addr) {
910 case PT_NAT_BITS:
911 return access_nat_bits(child, pt, &info,
912 data, write_access);
913
914 case PT_R4: case PT_R5: case PT_R6: case PT_R7:
915 if (write_access) {
916 /* read NaT bit first: */
917 unsigned long dummy;
918
919 ret = unw_get_gr(&info, (addr - PT_R4)/8 + 4,
920 &dummy, &nat);
921 if (ret < 0)
922 return ret;
923 }
924 return unw_access_gr(&info, (addr - PT_R4)/8 + 4, data,
925 &nat, write_access);
926
927 case PT_B1: case PT_B2: case PT_B3:
928 case PT_B4: case PT_B5:
929 return unw_access_br(&info, (addr - PT_B1)/8 + 1, data,
930 write_access);
931
932 case PT_AR_EC:
933 return unw_access_ar(&info, UNW_AR_EC, data,
934 write_access);
935
936 case PT_AR_LC:
937 return unw_access_ar(&info, UNW_AR_LC, data,
938 write_access);
939
940 default:
941 if (addr >= PT_F2 && addr < PT_F5 + 16)
942 return access_fr(&info, (addr - PT_F2)/16 + 2,
943 (addr & 8) != 0, data,
944 write_access);
945 else if (addr >= PT_F16 && addr < PT_F31 + 16)
946 return access_fr(&info,
947 (addr - PT_F16)/16 + 16,
948 (addr & 8) != 0,
949 data, write_access);
950 else {
951 dprintk("ptrace: rejecting access to register "
952 "address 0x%lx\n", addr);
953 return -1;
954 }
955 }
956 } else if (addr < PT_F9+16) {
957 /* scratch state */
958 switch (addr) {
959 case PT_AR_BSP:
960 /*
961 * By convention, we use PT_AR_BSP to refer to
962 * the end of the user-level backing store.
963 * Use ia64_rse_skip_regs(PT_AR_BSP, -CFM.sof)
964 * to get the real value of ar.bsp at the time
965 * the kernel was entered.
966 *
967 * Furthermore, when changing the contents of
968 * PT_AR_BSP (or PT_CFM) while the task is
969 * blocked in a system call, convert the state
970 * so that the non-system-call exit
971 * path is used. This ensures that the proper
972 * state will be picked up when resuming
973 * execution. However, it *also* means that
974 * once we write PT_AR_BSP/PT_CFM, it won't be
975 * possible to modify the syscall arguments of
976 * the pending system call any longer. This
977 * shouldn't be an issue because modifying
978 * PT_AR_BSP/PT_CFM generally implies that
979 * we're either abandoning the pending system
980 * call or that we defer it's re-execution
981 * (e.g., due to GDB doing an inferior
982 * function call).
983 */
984 urbs_end = ia64_get_user_rbs_end(child, pt, &cfm);
985 if (write_access) {
986 if (*data != urbs_end) {
987 if (in_syscall(pt))
988 convert_to_non_syscall(child,
989 pt,
990 cfm);
991 /*
992 * Simulate user-level write
993 * of ar.bsp:
994 */
995 pt->loadrs = 0;
996 pt->ar_bspstore = *data;
997 }
998 } else
999 *data = urbs_end;
1000 return 0;
1001
1002 case PT_CFM:
1003 urbs_end = ia64_get_user_rbs_end(child, pt, &cfm);
1004 if (write_access) {
1005 if (((cfm ^ *data) & PFM_MASK) != 0) {
1006 if (in_syscall(pt))
1007 convert_to_non_syscall(child,
1008 pt,
1009 cfm);
1010 pt->cr_ifs = ((pt->cr_ifs & ~PFM_MASK)
1011 | (*data & PFM_MASK));
1012 }
1013 } else
1014 *data = cfm;
1015 return 0;
1016
1017 case PT_CR_IPSR:
1018 if (write_access) {
1019 unsigned long tmp = *data;
1020 /* psr.ri==3 is a reserved value: SDM 2:25 */
1021 if ((tmp & IA64_PSR_RI) == IA64_PSR_RI)
1022 tmp &= ~IA64_PSR_RI;
1023 pt->cr_ipsr = ((tmp & IPSR_MASK)
1024 | (pt->cr_ipsr & ~IPSR_MASK));
1025 } else
1026 *data = (pt->cr_ipsr & IPSR_MASK);
1027 return 0;
1028
1029 case PT_AR_RSC:
1030 if (write_access)
1031 pt->ar_rsc = *data | (3 << 2); /* force PL3 */
1032 else
1033 *data = pt->ar_rsc;
1034 return 0;
1035
1036 case PT_AR_RNAT:
1037 ptr = pt_reg_addr(pt, ar_rnat);
1038 break;
1039 case PT_R1:
1040 ptr = pt_reg_addr(pt, r1);
1041 break;
1042 case PT_R2: case PT_R3:
1043 ptr = pt_reg_addr(pt, r2) + (addr - PT_R2);
1044 break;
1045 case PT_R8: case PT_R9: case PT_R10: case PT_R11:
1046 ptr = pt_reg_addr(pt, r8) + (addr - PT_R8);
1047 break;
1048 case PT_R12: case PT_R13:
1049 ptr = pt_reg_addr(pt, r12) + (addr - PT_R12);
1050 break;
1051 case PT_R14:
1052 ptr = pt_reg_addr(pt, r14);
1053 break;
1054 case PT_R15:
1055 ptr = pt_reg_addr(pt, r15);
1056 break;
1057 case PT_R16: case PT_R17: case PT_R18: case PT_R19:
1058 case PT_R20: case PT_R21: case PT_R22: case PT_R23:
1059 case PT_R24: case PT_R25: case PT_R26: case PT_R27:
1060 case PT_R28: case PT_R29: case PT_R30: case PT_R31:
1061 ptr = pt_reg_addr(pt, r16) + (addr - PT_R16);
1062 break;
1063 case PT_B0:
1064 ptr = pt_reg_addr(pt, b0);
1065 break;
1066 case PT_B6:
1067 ptr = pt_reg_addr(pt, b6);
1068 break;
1069 case PT_B7:
1070 ptr = pt_reg_addr(pt, b7);
1071 break;
1072 case PT_F6: case PT_F6+8: case PT_F7: case PT_F7+8:
1073 case PT_F8: case PT_F8+8: case PT_F9: case PT_F9+8:
1074 ptr = pt_reg_addr(pt, f6) + (addr - PT_F6);
1075 break;
1076 case PT_AR_BSPSTORE:
1077 ptr = pt_reg_addr(pt, ar_bspstore);
1078 break;
1079 case PT_AR_UNAT:
1080 ptr = pt_reg_addr(pt, ar_unat);
1081 break;
1082 case PT_AR_PFS:
1083 ptr = pt_reg_addr(pt, ar_pfs);
1084 break;
1085 case PT_AR_CCV:
1086 ptr = pt_reg_addr(pt, ar_ccv);
1087 break;
1088 case PT_AR_FPSR:
1089 ptr = pt_reg_addr(pt, ar_fpsr);
1090 break;
1091 case PT_CR_IIP:
1092 ptr = pt_reg_addr(pt, cr_iip);
1093 break;
1094 case PT_PR:
1095 ptr = pt_reg_addr(pt, pr);
1096 break;
1097 /* scratch register */
1098
1099 default:
1100 /* disallow accessing anything else... */
1101 dprintk("ptrace: rejecting access to register "
1102 "address 0x%lx\n", addr);
1103 return -1;
1104 }
1105 } else if (addr <= PT_AR_SSD) {
1106 ptr = pt_reg_addr(pt, ar_csd) + (addr - PT_AR_CSD);
1107 } else {
1108 /* access debug registers */
1109
1110 if (addr >= PT_IBR) {
1111 regnum = (addr - PT_IBR) >> 3;
1112 ptr = &child->thread.ibr[0];
1113 } else {
1114 regnum = (addr - PT_DBR) >> 3;
1115 ptr = &child->thread.dbr[0];
1116 }
1117
1118 if (regnum >= 8) {
1119 dprintk("ptrace: rejecting access to register "
1120 "address 0x%lx\n", addr);
1121 return -1;
1122 }
1123#ifdef CONFIG_PERFMON
1124 /*
1125 * Check if debug registers are used by perfmon. This
1126 * test must be done once we know that we can do the
1127 * operation, i.e. the arguments are all valid, but
1128 * before we start modifying the state.
1129 *
1130 * Perfmon needs to keep a count of how many processes
1131 * are trying to modify the debug registers for system
1132 * wide monitoring sessions.
1133 *
1134 * We also include read access here, because they may
1135 * cause the PMU-installed debug register state
1136 * (dbr[], ibr[]) to be reset. The two arrays are also
1137 * used by perfmon, but we do not use
1138 * IA64_THREAD_DBG_VALID. The registers are restored
1139 * by the PMU context switch code.
1140 */
1141 if (pfm_use_debug_registers(child)) return -1;
1142#endif
1143
1144 if (!(child->thread.flags & IA64_THREAD_DBG_VALID)) {
1145 child->thread.flags |= IA64_THREAD_DBG_VALID;
1146 memset(child->thread.dbr, 0,
1147 sizeof(child->thread.dbr));
1148 memset(child->thread.ibr, 0,
1149 sizeof(child->thread.ibr));
1150 }
1151
1152 ptr += regnum;
1153
1154 if ((regnum & 1) && write_access) {
1155 /* don't let the user set kernel-level breakpoints: */
1156 *ptr = *data & ~(7UL << 56);
1157 return 0;
1158 }
1159 }
1160 if (write_access)
1161 *ptr = *data;
1162 else
1163 *data = *ptr;
1164 return 0;
1165}
1166 850
1167static long 851static long
1168ptrace_getregs (struct task_struct *child, struct pt_all_user_regs __user *ppr) 852ptrace_getregs (struct task_struct *child, struct pt_all_user_regs __user *ppr)
@@ -1626,3 +1310,892 @@ syscall_trace_leave (long arg0, long arg1, long arg2, long arg3,
1626 if (test_thread_flag(TIF_RESTORE_RSE)) 1310 if (test_thread_flag(TIF_RESTORE_RSE))
1627 ia64_sync_krbs(); 1311 ia64_sync_krbs();
1628} 1312}
1313
1314/* Utrace implementation starts here */
1315struct regset_get {
1316 void *kbuf;
1317 void __user *ubuf;
1318};
1319
1320struct regset_set {
1321 const void *kbuf;
1322 const void __user *ubuf;
1323};
1324
1325struct regset_getset {
1326 struct task_struct *target;
1327 const struct user_regset *regset;
1328 union {
1329 struct regset_get get;
1330 struct regset_set set;
1331 } u;
1332 unsigned int pos;
1333 unsigned int count;
1334 int ret;
1335};
1336
1337static int
1338access_elf_gpreg(struct task_struct *target, struct unw_frame_info *info,
1339 unsigned long addr, unsigned long *data, int write_access)
1340{
1341 struct pt_regs *pt;
1342 unsigned long *ptr = NULL;
1343 int ret;
1344 char nat = 0;
1345
1346 pt = task_pt_regs(target);
1347 switch (addr) {
1348 case ELF_GR_OFFSET(1):
1349 ptr = &pt->r1;
1350 break;
1351 case ELF_GR_OFFSET(2):
1352 case ELF_GR_OFFSET(3):
1353 ptr = (void *)&pt->r2 + (addr - ELF_GR_OFFSET(2));
1354 break;
1355 case ELF_GR_OFFSET(4) ... ELF_GR_OFFSET(7):
1356 if (write_access) {
1357 /* read NaT bit first: */
1358 unsigned long dummy;
1359
1360 ret = unw_get_gr(info, addr/8, &dummy, &nat);
1361 if (ret < 0)
1362 return ret;
1363 }
1364 return unw_access_gr(info, addr/8, data, &nat, write_access);
1365 case ELF_GR_OFFSET(8) ... ELF_GR_OFFSET(11):
1366 ptr = (void *)&pt->r8 + addr - ELF_GR_OFFSET(8);
1367 break;
1368 case ELF_GR_OFFSET(12):
1369 case ELF_GR_OFFSET(13):
1370 ptr = (void *)&pt->r12 + addr - ELF_GR_OFFSET(12);
1371 break;
1372 case ELF_GR_OFFSET(14):
1373 ptr = &pt->r14;
1374 break;
1375 case ELF_GR_OFFSET(15):
1376 ptr = &pt->r15;
1377 }
1378 if (write_access)
1379 *ptr = *data;
1380 else
1381 *data = *ptr;
1382 return 0;
1383}
1384
1385static int
1386access_elf_breg(struct task_struct *target, struct unw_frame_info *info,
1387 unsigned long addr, unsigned long *data, int write_access)
1388{
1389 struct pt_regs *pt;
1390 unsigned long *ptr = NULL;
1391
1392 pt = task_pt_regs(target);
1393 switch (addr) {
1394 case ELF_BR_OFFSET(0):
1395 ptr = &pt->b0;
1396 break;
1397 case ELF_BR_OFFSET(1) ... ELF_BR_OFFSET(5):
1398 return unw_access_br(info, (addr - ELF_BR_OFFSET(0))/8,
1399 data, write_access);
1400 case ELF_BR_OFFSET(6):
1401 ptr = &pt->b6;
1402 break;
1403 case ELF_BR_OFFSET(7):
1404 ptr = &pt->b7;
1405 }
1406 if (write_access)
1407 *ptr = *data;
1408 else
1409 *data = *ptr;
1410 return 0;
1411}
1412
1413static int
1414access_elf_areg(struct task_struct *target, struct unw_frame_info *info,
1415 unsigned long addr, unsigned long *data, int write_access)
1416{
1417 struct pt_regs *pt;
1418 unsigned long cfm, urbs_end;
1419 unsigned long *ptr = NULL;
1420
1421 pt = task_pt_regs(target);
1422 if (addr >= ELF_AR_RSC_OFFSET && addr <= ELF_AR_SSD_OFFSET) {
1423 switch (addr) {
1424 case ELF_AR_RSC_OFFSET:
1425 /* force PL3 */
1426 if (write_access)
1427 pt->ar_rsc = *data | (3 << 2);
1428 else
1429 *data = pt->ar_rsc;
1430 return 0;
1431 case ELF_AR_BSP_OFFSET:
1432 /*
1433 * By convention, we use PT_AR_BSP to refer to
1434 * the end of the user-level backing store.
1435 * Use ia64_rse_skip_regs(PT_AR_BSP, -CFM.sof)
1436 * to get the real value of ar.bsp at the time
1437 * the kernel was entered.
1438 *
1439 * Furthermore, when changing the contents of
1440 * PT_AR_BSP (or PT_CFM) while the task is
1441 * blocked in a system call, convert the state
1442 * so that the non-system-call exit
1443 * path is used. This ensures that the proper
1444 * state will be picked up when resuming
1445 * execution. However, it *also* means that
1446 * once we write PT_AR_BSP/PT_CFM, it won't be
1447 * possible to modify the syscall arguments of
1448 * the pending system call any longer. This
1449 * shouldn't be an issue because modifying
1450 * PT_AR_BSP/PT_CFM generally implies that
1451 * we're either abandoning the pending system
1452 * call or that we defer it's re-execution
1453 * (e.g., due to GDB doing an inferior
1454 * function call).
1455 */
1456 urbs_end = ia64_get_user_rbs_end(target, pt, &cfm);
1457 if (write_access) {
1458 if (*data != urbs_end) {
1459 if (in_syscall(pt))
1460 convert_to_non_syscall(target,
1461 pt,
1462 cfm);
1463 /*
1464 * Simulate user-level write
1465 * of ar.bsp:
1466 */
1467 pt->loadrs = 0;
1468 pt->ar_bspstore = *data;
1469 }
1470 } else
1471 *data = urbs_end;
1472 return 0;
1473 case ELF_AR_BSPSTORE_OFFSET:
1474 ptr = &pt->ar_bspstore;
1475 break;
1476 case ELF_AR_RNAT_OFFSET:
1477 ptr = &pt->ar_rnat;
1478 break;
1479 case ELF_AR_CCV_OFFSET:
1480 ptr = &pt->ar_ccv;
1481 break;
1482 case ELF_AR_UNAT_OFFSET:
1483 ptr = &pt->ar_unat;
1484 break;
1485 case ELF_AR_FPSR_OFFSET:
1486 ptr = &pt->ar_fpsr;
1487 break;
1488 case ELF_AR_PFS_OFFSET:
1489 ptr = &pt->ar_pfs;
1490 break;
1491 case ELF_AR_LC_OFFSET:
1492 return unw_access_ar(info, UNW_AR_LC, data,
1493 write_access);
1494 case ELF_AR_EC_OFFSET:
1495 return unw_access_ar(info, UNW_AR_EC, data,
1496 write_access);
1497 case ELF_AR_CSD_OFFSET:
1498 ptr = &pt->ar_csd;
1499 break;
1500 case ELF_AR_SSD_OFFSET:
1501 ptr = &pt->ar_ssd;
1502 }
1503 } else if (addr >= ELF_CR_IIP_OFFSET && addr <= ELF_CR_IPSR_OFFSET) {
1504 switch (addr) {
1505 case ELF_CR_IIP_OFFSET:
1506 ptr = &pt->cr_iip;
1507 break;
1508 case ELF_CFM_OFFSET:
1509 urbs_end = ia64_get_user_rbs_end(target, pt, &cfm);
1510 if (write_access) {
1511 if (((cfm ^ *data) & PFM_MASK) != 0) {
1512 if (in_syscall(pt))
1513 convert_to_non_syscall(target,
1514 pt,
1515 cfm);
1516 pt->cr_ifs = ((pt->cr_ifs & ~PFM_MASK)
1517 | (*data & PFM_MASK));
1518 }
1519 } else
1520 *data = cfm;
1521 return 0;
1522 case ELF_CR_IPSR_OFFSET:
1523 if (write_access) {
1524 unsigned long tmp = *data;
1525 /* psr.ri==3 is a reserved value: SDM 2:25 */
1526 if ((tmp & IA64_PSR_RI) == IA64_PSR_RI)
1527 tmp &= ~IA64_PSR_RI;
1528 pt->cr_ipsr = ((tmp & IPSR_MASK)
1529 | (pt->cr_ipsr & ~IPSR_MASK));
1530 } else
1531 *data = (pt->cr_ipsr & IPSR_MASK);
1532 return 0;
1533 }
1534 } else if (addr == ELF_NAT_OFFSET)
1535 return access_nat_bits(target, pt, info,
1536 data, write_access);
1537 else if (addr == ELF_PR_OFFSET)
1538 ptr = &pt->pr;
1539 else
1540 return -1;
1541
1542 if (write_access)
1543 *ptr = *data;
1544 else
1545 *data = *ptr;
1546
1547 return 0;
1548}
1549
1550static int
1551access_elf_reg(struct task_struct *target, struct unw_frame_info *info,
1552 unsigned long addr, unsigned long *data, int write_access)
1553{
1554 if (addr >= ELF_GR_OFFSET(1) && addr <= ELF_GR_OFFSET(15))
1555 return access_elf_gpreg(target, info, addr, data, write_access);
1556 else if (addr >= ELF_BR_OFFSET(0) && addr <= ELF_BR_OFFSET(7))
1557 return access_elf_breg(target, info, addr, data, write_access);
1558 else
1559 return access_elf_areg(target, info, addr, data, write_access);
1560}
1561
1562void do_gpregs_get(struct unw_frame_info *info, void *arg)
1563{
1564 struct pt_regs *pt;
1565 struct regset_getset *dst = arg;
1566 elf_greg_t tmp[16];
1567 unsigned int i, index, min_copy;
1568
1569 if (unw_unwind_to_user(info) < 0)
1570 return;
1571
1572 /*
1573 * coredump format:
1574 * r0-r31
1575 * NaT bits (for r0-r31; bit N == 1 iff rN is a NaT)
1576 * predicate registers (p0-p63)
1577 * b0-b7
1578 * ip cfm user-mask
1579 * ar.rsc ar.bsp ar.bspstore ar.rnat
1580 * ar.ccv ar.unat ar.fpsr ar.pfs ar.lc ar.ec
1581 */
1582
1583
1584 /* Skip r0 */
1585 if (dst->count > 0 && dst->pos < ELF_GR_OFFSET(1)) {
1586 dst->ret = user_regset_copyout_zero(&dst->pos, &dst->count,
1587 &dst->u.get.kbuf,
1588 &dst->u.get.ubuf,
1589 0, ELF_GR_OFFSET(1));
1590 if (dst->ret || dst->count == 0)
1591 return;
1592 }
1593
1594 /* gr1 - gr15 */
1595 if (dst->count > 0 && dst->pos < ELF_GR_OFFSET(16)) {
1596 index = (dst->pos - ELF_GR_OFFSET(1)) / sizeof(elf_greg_t);
1597 min_copy = ELF_GR_OFFSET(16) > (dst->pos + dst->count) ?
1598 (dst->pos + dst->count) : ELF_GR_OFFSET(16);
1599 for (i = dst->pos; i < min_copy; i += sizeof(elf_greg_t),
1600 index++)
1601 if (access_elf_reg(dst->target, info, i,
1602 &tmp[index], 0) < 0) {
1603 dst->ret = -EIO;
1604 return;
1605 }
1606 dst->ret = user_regset_copyout(&dst->pos, &dst->count,
1607 &dst->u.get.kbuf, &dst->u.get.ubuf, tmp,
1608 ELF_GR_OFFSET(1), ELF_GR_OFFSET(16));
1609 if (dst->ret || dst->count == 0)
1610 return;
1611 }
1612
1613 /* r16-r31 */
1614 if (dst->count > 0 && dst->pos < ELF_NAT_OFFSET) {
1615 pt = task_pt_regs(dst->target);
1616 dst->ret = user_regset_copyout(&dst->pos, &dst->count,
1617 &dst->u.get.kbuf, &dst->u.get.ubuf, &pt->r16,
1618 ELF_GR_OFFSET(16), ELF_NAT_OFFSET);
1619 if (dst->ret || dst->count == 0)
1620 return;
1621 }
1622
1623 /* nat, pr, b0 - b7 */
1624 if (dst->count > 0 && dst->pos < ELF_CR_IIP_OFFSET) {
1625 index = (dst->pos - ELF_NAT_OFFSET) / sizeof(elf_greg_t);
1626 min_copy = ELF_CR_IIP_OFFSET > (dst->pos + dst->count) ?
1627 (dst->pos + dst->count) : ELF_CR_IIP_OFFSET;
1628 for (i = dst->pos; i < min_copy; i += sizeof(elf_greg_t),
1629 index++)
1630 if (access_elf_reg(dst->target, info, i,
1631 &tmp[index], 0) < 0) {
1632 dst->ret = -EIO;
1633 return;
1634 }
1635 dst->ret = user_regset_copyout(&dst->pos, &dst->count,
1636 &dst->u.get.kbuf, &dst->u.get.ubuf, tmp,
1637 ELF_NAT_OFFSET, ELF_CR_IIP_OFFSET);
1638 if (dst->ret || dst->count == 0)
1639 return;
1640 }
1641
1642 /* ip cfm psr ar.rsc ar.bsp ar.bspstore ar.rnat
1643 * ar.ccv ar.unat ar.fpsr ar.pfs ar.lc ar.ec ar.csd ar.ssd
1644 */
1645 if (dst->count > 0 && dst->pos < (ELF_AR_END_OFFSET)) {
1646 index = (dst->pos - ELF_CR_IIP_OFFSET) / sizeof(elf_greg_t);
1647 min_copy = ELF_AR_END_OFFSET > (dst->pos + dst->count) ?
1648 (dst->pos + dst->count) : ELF_AR_END_OFFSET;
1649 for (i = dst->pos; i < min_copy; i += sizeof(elf_greg_t),
1650 index++)
1651 if (access_elf_reg(dst->target, info, i,
1652 &tmp[index], 0) < 0) {
1653 dst->ret = -EIO;
1654 return;
1655 }
1656 dst->ret = user_regset_copyout(&dst->pos, &dst->count,
1657 &dst->u.get.kbuf, &dst->u.get.ubuf, tmp,
1658 ELF_CR_IIP_OFFSET, ELF_AR_END_OFFSET);
1659 }
1660}
1661
1662void do_gpregs_set(struct unw_frame_info *info, void *arg)
1663{
1664 struct pt_regs *pt;
1665 struct regset_getset *dst = arg;
1666 elf_greg_t tmp[16];
1667 unsigned int i, index;
1668
1669 if (unw_unwind_to_user(info) < 0)
1670 return;
1671
1672 /* Skip r0 */
1673 if (dst->count > 0 && dst->pos < ELF_GR_OFFSET(1)) {
1674 dst->ret = user_regset_copyin_ignore(&dst->pos, &dst->count,
1675 &dst->u.set.kbuf,
1676 &dst->u.set.ubuf,
1677 0, ELF_GR_OFFSET(1));
1678 if (dst->ret || dst->count == 0)
1679 return;
1680 }
1681
1682 /* gr1-gr15 */
1683 if (dst->count > 0 && dst->pos < ELF_GR_OFFSET(16)) {
1684 i = dst->pos;
1685 index = (dst->pos - ELF_GR_OFFSET(1)) / sizeof(elf_greg_t);
1686 dst->ret = user_regset_copyin(&dst->pos, &dst->count,
1687 &dst->u.set.kbuf, &dst->u.set.ubuf, tmp,
1688 ELF_GR_OFFSET(1), ELF_GR_OFFSET(16));
1689 if (dst->ret)
1690 return;
1691 for ( ; i < dst->pos; i += sizeof(elf_greg_t), index++)
1692 if (access_elf_reg(dst->target, info, i,
1693 &tmp[index], 1) < 0) {
1694 dst->ret = -EIO;
1695 return;
1696 }
1697 if (dst->count == 0)
1698 return;
1699 }
1700
1701 /* gr16-gr31 */
1702 if (dst->count > 0 && dst->pos < ELF_NAT_OFFSET) {
1703 pt = task_pt_regs(dst->target);
1704 dst->ret = user_regset_copyin(&dst->pos, &dst->count,
1705 &dst->u.set.kbuf, &dst->u.set.ubuf, &pt->r16,
1706 ELF_GR_OFFSET(16), ELF_NAT_OFFSET);
1707 if (dst->ret || dst->count == 0)
1708 return;
1709 }
1710
1711 /* nat, pr, b0 - b7 */
1712 if (dst->count > 0 && dst->pos < ELF_CR_IIP_OFFSET) {
1713 i = dst->pos;
1714 index = (dst->pos - ELF_NAT_OFFSET) / sizeof(elf_greg_t);
1715 dst->ret = user_regset_copyin(&dst->pos, &dst->count,
1716 &dst->u.set.kbuf, &dst->u.set.ubuf, tmp,
1717 ELF_NAT_OFFSET, ELF_CR_IIP_OFFSET);
1718 if (dst->ret)
1719 return;
1720 for (; i < dst->pos; i += sizeof(elf_greg_t), index++)
1721 if (access_elf_reg(dst->target, info, i,
1722 &tmp[index], 1) < 0) {
1723 dst->ret = -EIO;
1724 return;
1725 }
1726 if (dst->count == 0)
1727 return;
1728 }
1729
1730 /* ip cfm psr ar.rsc ar.bsp ar.bspstore ar.rnat
1731 * ar.ccv ar.unat ar.fpsr ar.pfs ar.lc ar.ec ar.csd ar.ssd
1732 */
1733 if (dst->count > 0 && dst->pos < (ELF_AR_END_OFFSET)) {
1734 i = dst->pos;
1735 index = (dst->pos - ELF_CR_IIP_OFFSET) / sizeof(elf_greg_t);
1736 dst->ret = user_regset_copyin(&dst->pos, &dst->count,
1737 &dst->u.set.kbuf, &dst->u.set.ubuf, tmp,
1738 ELF_CR_IIP_OFFSET, ELF_AR_END_OFFSET);
1739 if (dst->ret)
1740 return;
1741 for ( ; i < dst->pos; i += sizeof(elf_greg_t), index++)
1742 if (access_elf_reg(dst->target, info, i,
1743 &tmp[index], 1) < 0) {
1744 dst->ret = -EIO;
1745 return;
1746 }
1747 }
1748}
1749
1750#define ELF_FP_OFFSET(i) (i * sizeof(elf_fpreg_t))
1751
1752void do_fpregs_get(struct unw_frame_info *info, void *arg)
1753{
1754 struct regset_getset *dst = arg;
1755 struct task_struct *task = dst->target;
1756 elf_fpreg_t tmp[30];
1757 int index, min_copy, i;
1758
1759 if (unw_unwind_to_user(info) < 0)
1760 return;
1761
1762 /* Skip pos 0 and 1 */
1763 if (dst->count > 0 && dst->pos < ELF_FP_OFFSET(2)) {
1764 dst->ret = user_regset_copyout_zero(&dst->pos, &dst->count,
1765 &dst->u.get.kbuf,
1766 &dst->u.get.ubuf,
1767 0, ELF_FP_OFFSET(2));
1768 if (dst->count == 0 || dst->ret)
1769 return;
1770 }
1771
1772 /* fr2-fr31 */
1773 if (dst->count > 0 && dst->pos < ELF_FP_OFFSET(32)) {
1774 index = (dst->pos - ELF_FP_OFFSET(2)) / sizeof(elf_fpreg_t);
1775
1776 min_copy = min(((unsigned int)ELF_FP_OFFSET(32)),
1777 dst->pos + dst->count);
1778 for (i = dst->pos; i < min_copy; i += sizeof(elf_fpreg_t),
1779 index++)
1780 if (unw_get_fr(info, i / sizeof(elf_fpreg_t),
1781 &tmp[index])) {
1782 dst->ret = -EIO;
1783 return;
1784 }
1785 dst->ret = user_regset_copyout(&dst->pos, &dst->count,
1786 &dst->u.get.kbuf, &dst->u.get.ubuf, tmp,
1787 ELF_FP_OFFSET(2), ELF_FP_OFFSET(32));
1788 if (dst->count == 0 || dst->ret)
1789 return;
1790 }
1791
1792 /* fph */
1793 if (dst->count > 0) {
1794 ia64_flush_fph(dst->target);
1795 if (task->thread.flags & IA64_THREAD_FPH_VALID)
1796 dst->ret = user_regset_copyout(
1797 &dst->pos, &dst->count,
1798 &dst->u.get.kbuf, &dst->u.get.ubuf,
1799 &dst->target->thread.fph,
1800 ELF_FP_OFFSET(32), -1);
1801 else
1802 /* Zero fill instead. */
1803 dst->ret = user_regset_copyout_zero(
1804 &dst->pos, &dst->count,
1805 &dst->u.get.kbuf, &dst->u.get.ubuf,
1806 ELF_FP_OFFSET(32), -1);
1807 }
1808}
1809
1810void do_fpregs_set(struct unw_frame_info *info, void *arg)
1811{
1812 struct regset_getset *dst = arg;
1813 elf_fpreg_t fpreg, tmp[30];
1814 int index, start, end;
1815
1816 if (unw_unwind_to_user(info) < 0)
1817 return;
1818
1819 /* Skip pos 0 and 1 */
1820 if (dst->count > 0 && dst->pos < ELF_FP_OFFSET(2)) {
1821 dst->ret = user_regset_copyin_ignore(&dst->pos, &dst->count,
1822 &dst->u.set.kbuf,
1823 &dst->u.set.ubuf,
1824 0, ELF_FP_OFFSET(2));
1825 if (dst->count == 0 || dst->ret)
1826 return;
1827 }
1828
1829 /* fr2-fr31 */
1830 if (dst->count > 0 && dst->pos < ELF_FP_OFFSET(32)) {
1831 start = dst->pos;
1832 end = min(((unsigned int)ELF_FP_OFFSET(32)),
1833 dst->pos + dst->count);
1834 dst->ret = user_regset_copyin(&dst->pos, &dst->count,
1835 &dst->u.set.kbuf, &dst->u.set.ubuf, tmp,
1836 ELF_FP_OFFSET(2), ELF_FP_OFFSET(32));
1837 if (dst->ret)
1838 return;
1839
1840 if (start & 0xF) { /* only write high part */
1841 if (unw_get_fr(info, start / sizeof(elf_fpreg_t),
1842 &fpreg)) {
1843 dst->ret = -EIO;
1844 return;
1845 }
1846 tmp[start / sizeof(elf_fpreg_t) - 2].u.bits[0]
1847 = fpreg.u.bits[0];
1848 start &= ~0xFUL;
1849 }
1850 if (end & 0xF) { /* only write low part */
1851 if (unw_get_fr(info, end / sizeof(elf_fpreg_t),
1852 &fpreg)) {
1853 dst->ret = -EIO;
1854 return;
1855 }
1856 tmp[end / sizeof(elf_fpreg_t) - 2].u.bits[1]
1857 = fpreg.u.bits[1];
1858 end = (end + 0xF) & ~0xFUL;
1859 }
1860
1861 for ( ; start < end ; start += sizeof(elf_fpreg_t)) {
1862 index = start / sizeof(elf_fpreg_t);
1863 if (unw_set_fr(info, index, tmp[index - 2])) {
1864 dst->ret = -EIO;
1865 return;
1866 }
1867 }
1868 if (dst->ret || dst->count == 0)
1869 return;
1870 }
1871
1872 /* fph */
1873 if (dst->count > 0 && dst->pos < ELF_FP_OFFSET(128)) {
1874 ia64_sync_fph(dst->target);
1875 dst->ret = user_regset_copyin(&dst->pos, &dst->count,
1876 &dst->u.set.kbuf,
1877 &dst->u.set.ubuf,
1878 &dst->target->thread.fph,
1879 ELF_FP_OFFSET(32), -1);
1880 }
1881}
1882
1883static int
1884do_regset_call(void (*call)(struct unw_frame_info *, void *),
1885 struct task_struct *target,
1886 const struct user_regset *regset,
1887 unsigned int pos, unsigned int count,
1888 const void *kbuf, const void __user *ubuf)
1889{
1890 struct regset_getset info = { .target = target, .regset = regset,
1891 .pos = pos, .count = count,
1892 .u.set = { .kbuf = kbuf, .ubuf = ubuf },
1893 .ret = 0 };
1894
1895 if (target == current)
1896 unw_init_running(call, &info);
1897 else {
1898 struct unw_frame_info ufi;
1899 memset(&ufi, 0, sizeof(ufi));
1900 unw_init_from_blocked_task(&ufi, target);
1901 (*call)(&ufi, &info);
1902 }
1903
1904 return info.ret;
1905}
1906
1907static int
1908gpregs_get(struct task_struct *target,
1909 const struct user_regset *regset,
1910 unsigned int pos, unsigned int count,
1911 void *kbuf, void __user *ubuf)
1912{
1913 return do_regset_call(do_gpregs_get, target, regset, pos, count,
1914 kbuf, ubuf);
1915}
1916
1917static int gpregs_set(struct task_struct *target,
1918 const struct user_regset *regset,
1919 unsigned int pos, unsigned int count,
1920 const void *kbuf, const void __user *ubuf)
1921{
1922 return do_regset_call(do_gpregs_set, target, regset, pos, count,
1923 kbuf, ubuf);
1924}
1925
1926static void do_gpregs_writeback(struct unw_frame_info *info, void *arg)
1927{
1928 do_sync_rbs(info, ia64_sync_user_rbs);
1929}
1930
1931/*
1932 * This is called to write back the register backing store.
1933 * ptrace does this before it stops, so that a tracer reading the user
1934 * memory after the thread stops will get the current register data.
1935 */
1936static int
1937gpregs_writeback(struct task_struct *target,
1938 const struct user_regset *regset,
1939 int now)
1940{
1941 if (test_and_set_tsk_thread_flag(target, TIF_RESTORE_RSE))
1942 return 0;
1943 tsk_set_notify_resume(target);
1944 return do_regset_call(do_gpregs_writeback, target, regset, 0, 0,
1945 NULL, NULL);
1946}
1947
1948static int
1949fpregs_active(struct task_struct *target, const struct user_regset *regset)
1950{
1951 return (target->thread.flags & IA64_THREAD_FPH_VALID) ? 128 : 32;
1952}
1953
1954static int fpregs_get(struct task_struct *target,
1955 const struct user_regset *regset,
1956 unsigned int pos, unsigned int count,
1957 void *kbuf, void __user *ubuf)
1958{
1959 return do_regset_call(do_fpregs_get, target, regset, pos, count,
1960 kbuf, ubuf);
1961}
1962
1963static int fpregs_set(struct task_struct *target,
1964 const struct user_regset *regset,
1965 unsigned int pos, unsigned int count,
1966 const void *kbuf, const void __user *ubuf)
1967{
1968 return do_regset_call(do_fpregs_set, target, regset, pos, count,
1969 kbuf, ubuf);
1970}
1971
1972static int
1973access_uarea(struct task_struct *child, unsigned long addr,
1974 unsigned long *data, int write_access)
1975{
1976 unsigned int pos = -1; /* an invalid value */
1977 int ret;
1978 unsigned long *ptr, regnum;
1979
1980 if ((addr & 0x7) != 0) {
1981 dprintk("ptrace: unaligned register address 0x%lx\n", addr);
1982 return -1;
1983 }
1984 if ((addr >= PT_NAT_BITS + 8 && addr < PT_F2) ||
1985 (addr >= PT_R7 + 8 && addr < PT_B1) ||
1986 (addr >= PT_AR_LC + 8 && addr < PT_CR_IPSR) ||
1987 (addr >= PT_AR_SSD + 8 && addr < PT_DBR)) {
1988 dprintk("ptrace: rejecting access to register "
1989 "address 0x%lx\n", addr);
1990 return -1;
1991 }
1992
1993 switch (addr) {
1994 case PT_F32 ... (PT_F127 + 15):
1995 pos = addr - PT_F32 + ELF_FP_OFFSET(32);
1996 break;
1997 case PT_F2 ... (PT_F5 + 15):
1998 pos = addr - PT_F2 + ELF_FP_OFFSET(2);
1999 break;
2000 case PT_F10 ... (PT_F31 + 15):
2001 pos = addr - PT_F10 + ELF_FP_OFFSET(10);
2002 break;
2003 case PT_F6 ... (PT_F9 + 15):
2004 pos = addr - PT_F6 + ELF_FP_OFFSET(6);
2005 break;
2006 }
2007
2008 if (pos != -1) {
2009 if (write_access)
2010 ret = fpregs_set(child, NULL, pos,
2011 sizeof(unsigned long), data, NULL);
2012 else
2013 ret = fpregs_get(child, NULL, pos,
2014 sizeof(unsigned long), data, NULL);
2015 if (ret != 0)
2016 return -1;
2017 return 0;
2018 }
2019
2020 switch (addr) {
2021 case PT_NAT_BITS:
2022 pos = ELF_NAT_OFFSET;
2023 break;
2024 case PT_R4 ... PT_R7:
2025 pos = addr - PT_R4 + ELF_GR_OFFSET(4);
2026 break;
2027 case PT_B1 ... PT_B5:
2028 pos = addr - PT_B1 + ELF_BR_OFFSET(1);
2029 break;
2030 case PT_AR_EC:
2031 pos = ELF_AR_EC_OFFSET;
2032 break;
2033 case PT_AR_LC:
2034 pos = ELF_AR_LC_OFFSET;
2035 break;
2036 case PT_CR_IPSR:
2037 pos = ELF_CR_IPSR_OFFSET;
2038 break;
2039 case PT_CR_IIP:
2040 pos = ELF_CR_IIP_OFFSET;
2041 break;
2042 case PT_CFM:
2043 pos = ELF_CFM_OFFSET;
2044 break;
2045 case PT_AR_UNAT:
2046 pos = ELF_AR_UNAT_OFFSET;
2047 break;
2048 case PT_AR_PFS:
2049 pos = ELF_AR_PFS_OFFSET;
2050 break;
2051 case PT_AR_RSC:
2052 pos = ELF_AR_RSC_OFFSET;
2053 break;
2054 case PT_AR_RNAT:
2055 pos = ELF_AR_RNAT_OFFSET;
2056 break;
2057 case PT_AR_BSPSTORE:
2058 pos = ELF_AR_BSPSTORE_OFFSET;
2059 break;
2060 case PT_PR:
2061 pos = ELF_PR_OFFSET;
2062 break;
2063 case PT_B6:
2064 pos = ELF_BR_OFFSET(6);
2065 break;
2066 case PT_AR_BSP:
2067 pos = ELF_AR_BSP_OFFSET;
2068 break;
2069 case PT_R1 ... PT_R3:
2070 pos = addr - PT_R1 + ELF_GR_OFFSET(1);
2071 break;
2072 case PT_R12 ... PT_R15:
2073 pos = addr - PT_R12 + ELF_GR_OFFSET(12);
2074 break;
2075 case PT_R8 ... PT_R11:
2076 pos = addr - PT_R8 + ELF_GR_OFFSET(8);
2077 break;
2078 case PT_R16 ... PT_R31:
2079 pos = addr - PT_R16 + ELF_GR_OFFSET(16);
2080 break;
2081 case PT_AR_CCV:
2082 pos = ELF_AR_CCV_OFFSET;
2083 break;
2084 case PT_AR_FPSR:
2085 pos = ELF_AR_FPSR_OFFSET;
2086 break;
2087 case PT_B0:
2088 pos = ELF_BR_OFFSET(0);
2089 break;
2090 case PT_B7:
2091 pos = ELF_BR_OFFSET(7);
2092 break;
2093 case PT_AR_CSD:
2094 pos = ELF_AR_CSD_OFFSET;
2095 break;
2096 case PT_AR_SSD:
2097 pos = ELF_AR_SSD_OFFSET;
2098 break;
2099 }
2100
2101 if (pos != -1) {
2102 if (write_access)
2103 ret = gpregs_set(child, NULL, pos,
2104 sizeof(unsigned long), data, NULL);
2105 else
2106 ret = gpregs_get(child, NULL, pos,
2107 sizeof(unsigned long), data, NULL);
2108 if (ret != 0)
2109 return -1;
2110 return 0;
2111 }
2112
2113 /* access debug registers */
2114 if (addr >= PT_IBR) {
2115 regnum = (addr - PT_IBR) >> 3;
2116 ptr = &child->thread.ibr[0];
2117 } else {
2118 regnum = (addr - PT_DBR) >> 3;
2119 ptr = &child->thread.dbr[0];
2120 }
2121
2122 if (regnum >= 8) {
2123 dprintk("ptrace: rejecting access to register "
2124 "address 0x%lx\n", addr);
2125 return -1;
2126 }
2127#ifdef CONFIG_PERFMON
2128 /*
2129 * Check if debug registers are used by perfmon. This
2130 * test must be done once we know that we can do the
2131 * operation, i.e. the arguments are all valid, but
2132 * before we start modifying the state.
2133 *
2134 * Perfmon needs to keep a count of how many processes
2135 * are trying to modify the debug registers for system
2136 * wide monitoring sessions.
2137 *
2138 * We also include read access here, because they may
2139 * cause the PMU-installed debug register state
2140 * (dbr[], ibr[]) to be reset. The two arrays are also
2141 * used by perfmon, but we do not use
2142 * IA64_THREAD_DBG_VALID. The registers are restored
2143 * by the PMU context switch code.
2144 */
2145 if (pfm_use_debug_registers(child))
2146 return -1;
2147#endif
2148
2149 if (!(child->thread.flags & IA64_THREAD_DBG_VALID)) {
2150 child->thread.flags |= IA64_THREAD_DBG_VALID;
2151 memset(child->thread.dbr, 0,
2152 sizeof(child->thread.dbr));
2153 memset(child->thread.ibr, 0,
2154 sizeof(child->thread.ibr));
2155 }
2156
2157 ptr += regnum;
2158
2159 if ((regnum & 1) && write_access) {
2160 /* don't let the user set kernel-level breakpoints: */
2161 *ptr = *data & ~(7UL << 56);
2162 return 0;
2163 }
2164 if (write_access)
2165 *ptr = *data;
2166 else
2167 *data = *ptr;
2168 return 0;
2169}
2170
2171static const struct user_regset native_regsets[] = {
2172 {
2173 .core_note_type = NT_PRSTATUS,
2174 .n = ELF_NGREG,
2175 .size = sizeof(elf_greg_t), .align = sizeof(elf_greg_t),
2176 .get = gpregs_get, .set = gpregs_set,
2177 .writeback = gpregs_writeback
2178 },
2179 {
2180 .core_note_type = NT_PRFPREG,
2181 .n = ELF_NFPREG,
2182 .size = sizeof(elf_fpreg_t), .align = sizeof(elf_fpreg_t),
2183 .get = fpregs_get, .set = fpregs_set, .active = fpregs_active
2184 },
2185};
2186
2187static const struct user_regset_view user_ia64_view = {
2188 .name = "ia64",
2189 .e_machine = EM_IA_64,
2190 .regsets = native_regsets, .n = ARRAY_SIZE(native_regsets)
2191};
2192
2193const struct user_regset_view *task_user_regset_view(struct task_struct *tsk)
2194{
2195#ifdef CONFIG_IA32_SUPPORT
2196 extern const struct user_regset_view user_ia32_view;
2197 if (IS_IA32_PROCESS(task_pt_regs(tsk)))
2198 return &user_ia32_view;
2199#endif
2200 return &user_ia64_view;
2201}
diff --git a/arch/ia64/kernel/setup.c b/arch/ia64/kernel/setup.c
index 4aa9eaea76c3..5015ca1275ca 100644
--- a/arch/ia64/kernel/setup.c
+++ b/arch/ia64/kernel/setup.c
@@ -59,6 +59,7 @@
59#include <asm/setup.h> 59#include <asm/setup.h>
60#include <asm/smp.h> 60#include <asm/smp.h>
61#include <asm/system.h> 61#include <asm/system.h>
62#include <asm/tlbflush.h>
62#include <asm/unistd.h> 63#include <asm/unistd.h>
63#include <asm/hpsim.h> 64#include <asm/hpsim.h>
64 65
@@ -176,6 +177,29 @@ filter_rsvd_memory (unsigned long start, unsigned long end, void *arg)
176 return 0; 177 return 0;
177} 178}
178 179
180/*
181 * Similar to "filter_rsvd_memory()", but the reserved memory ranges
182 * are not filtered out.
183 */
184int __init
185filter_memory(unsigned long start, unsigned long end, void *arg)
186{
187 void (*func)(unsigned long, unsigned long, int);
188
189#if IGNORE_PFN0
190 if (start == PAGE_OFFSET) {
191 printk(KERN_WARNING "warning: skipping physical page 0\n");
192 start += PAGE_SIZE;
193 if (start >= end)
194 return 0;
195 }
196#endif
197 func = arg;
198 if (start < end)
199 call_pernode_memory(__pa(start), end - start, func);
200 return 0;
201}
202
179static void __init 203static void __init
180sort_regions (struct rsvd_region *rsvd_region, int max) 204sort_regions (struct rsvd_region *rsvd_region, int max)
181{ 205{
@@ -493,6 +517,8 @@ setup_arch (char **cmdline_p)
493 acpi_table_init(); 517 acpi_table_init();
494# ifdef CONFIG_ACPI_NUMA 518# ifdef CONFIG_ACPI_NUMA
495 acpi_numa_init(); 519 acpi_numa_init();
520 per_cpu_scan_finalize((cpus_weight(early_cpu_possible_map) == 0 ?
521 32 : cpus_weight(early_cpu_possible_map)), additional_cpus);
496# endif 522# endif
497#else 523#else
498# ifdef CONFIG_SMP 524# ifdef CONFIG_SMP
@@ -946,9 +972,10 @@ cpu_init (void)
946#endif 972#endif
947 973
948 /* set ia64_ctx.max_rid to the maximum RID that is supported by all CPUs: */ 974 /* set ia64_ctx.max_rid to the maximum RID that is supported by all CPUs: */
949 if (ia64_pal_vm_summary(NULL, &vmi) == 0) 975 if (ia64_pal_vm_summary(NULL, &vmi) == 0) {
950 max_ctx = (1U << (vmi.pal_vm_info_2_s.rid_size - 3)) - 1; 976 max_ctx = (1U << (vmi.pal_vm_info_2_s.rid_size - 3)) - 1;
951 else { 977 setup_ptcg_sem(vmi.pal_vm_info_2_s.max_purges, NPTCG_FROM_PAL);
978 } else {
952 printk(KERN_WARNING "cpu_init: PAL VM summary failed, assuming 18 RID bits\n"); 979 printk(KERN_WARNING "cpu_init: PAL VM summary failed, assuming 18 RID bits\n");
953 max_ctx = (1U << 15) - 1; /* use architected minimum */ 980 max_ctx = (1U << 15) - 1; /* use architected minimum */
954 } 981 }
diff --git a/arch/ia64/kernel/smp.c b/arch/ia64/kernel/smp.c
index 4e446aa5f4ac..9a9d4c489330 100644
--- a/arch/ia64/kernel/smp.c
+++ b/arch/ia64/kernel/smp.c
@@ -213,6 +213,19 @@ send_IPI_allbutself (int op)
213 * Called with preemption disabled. 213 * Called with preemption disabled.
214 */ 214 */
215static inline void 215static inline void
216send_IPI_mask(cpumask_t mask, int op)
217{
218 unsigned int cpu;
219
220 for_each_cpu_mask(cpu, mask) {
221 send_IPI_single(cpu, op);
222 }
223}
224
225/*
226 * Called with preemption disabled.
227 */
228static inline void
216send_IPI_all (int op) 229send_IPI_all (int op)
217{ 230{
218 int i; 231 int i;
@@ -401,6 +414,75 @@ smp_call_function_single (int cpuid, void (*func) (void *info), void *info, int
401} 414}
402EXPORT_SYMBOL(smp_call_function_single); 415EXPORT_SYMBOL(smp_call_function_single);
403 416
417/**
418 * smp_call_function_mask(): Run a function on a set of other CPUs.
419 * <mask> The set of cpus to run on. Must not include the current cpu.
420 * <func> The function to run. This must be fast and non-blocking.
421 * <info> An arbitrary pointer to pass to the function.
422 * <wait> If true, wait (atomically) until function
423 * has completed on other CPUs.
424 *
425 * Returns 0 on success, else a negative status code.
426 *
427 * If @wait is true, then returns once @func has returned; otherwise
428 * it returns just before the target cpu calls @func.
429 *
430 * You must not call this function with disabled interrupts or from a
431 * hardware interrupt handler or from a bottom half handler.
432 */
433int smp_call_function_mask(cpumask_t mask,
434 void (*func)(void *), void *info,
435 int wait)
436{
437 struct call_data_struct data;
438 cpumask_t allbutself;
439 int cpus;
440
441 spin_lock(&call_lock);
442 allbutself = cpu_online_map;
443 cpu_clear(smp_processor_id(), allbutself);
444
445 cpus_and(mask, mask, allbutself);
446 cpus = cpus_weight(mask);
447 if (!cpus) {
448 spin_unlock(&call_lock);
449 return 0;
450 }
451
452 /* Can deadlock when called with interrupts disabled */
453 WARN_ON(irqs_disabled());
454
455 data.func = func;
456 data.info = info;
457 atomic_set(&data.started, 0);
458 data.wait = wait;
459 if (wait)
460 atomic_set(&data.finished, 0);
461
462 call_data = &data;
463 mb(); /* ensure store to call_data precedes setting of IPI_CALL_FUNC*/
464
465 /* Send a message to other CPUs */
466 if (cpus_equal(mask, allbutself))
467 send_IPI_allbutself(IPI_CALL_FUNC);
468 else
469 send_IPI_mask(mask, IPI_CALL_FUNC);
470
471 /* Wait for response */
472 while (atomic_read(&data.started) != cpus)
473 cpu_relax();
474
475 if (wait)
476 while (atomic_read(&data.finished) != cpus)
477 cpu_relax();
478 call_data = NULL;
479
480 spin_unlock(&call_lock);
481 return 0;
482
483}
484EXPORT_SYMBOL(smp_call_function_mask);
485
404/* 486/*
405 * this function sends a 'generic call function' IPI to all other CPUs 487 * this function sends a 'generic call function' IPI to all other CPUs
406 * in the system. 488 * in the system.
diff --git a/arch/ia64/kernel/smpboot.c b/arch/ia64/kernel/smpboot.c
index 32ee5979a042..16483be18c0b 100644
--- a/arch/ia64/kernel/smpboot.c
+++ b/arch/ia64/kernel/smpboot.c
@@ -400,9 +400,9 @@ smp_callin (void)
400 /* Setup the per cpu irq handling data structures */ 400 /* Setup the per cpu irq handling data structures */
401 __setup_vector_irq(cpuid); 401 __setup_vector_irq(cpuid);
402 cpu_set(cpuid, cpu_online_map); 402 cpu_set(cpuid, cpu_online_map);
403 unlock_ipi_calllock();
404 per_cpu(cpu_state, cpuid) = CPU_ONLINE; 403 per_cpu(cpu_state, cpuid) = CPU_ONLINE;
405 spin_unlock(&vector_lock); 404 spin_unlock(&vector_lock);
405 unlock_ipi_calllock();
406 406
407 smp_setup_percpu_timer(); 407 smp_setup_percpu_timer();
408 408
diff --git a/arch/ia64/kernel/time.c b/arch/ia64/kernel/time.c
index 17fda5293c67..48e15a51782f 100644
--- a/arch/ia64/kernel/time.c
+++ b/arch/ia64/kernel/time.c
@@ -59,6 +59,84 @@ static struct clocksource clocksource_itc = {
59}; 59};
60static struct clocksource *itc_clocksource; 60static struct clocksource *itc_clocksource;
61 61
62#ifdef CONFIG_VIRT_CPU_ACCOUNTING
63
64#include <linux/kernel_stat.h>
65
66extern cputime_t cycle_to_cputime(u64 cyc);
67
68/*
69 * Called from the context switch with interrupts disabled, to charge all
70 * accumulated times to the current process, and to prepare accounting on
71 * the next process.
72 */
73void ia64_account_on_switch(struct task_struct *prev, struct task_struct *next)
74{
75 struct thread_info *pi = task_thread_info(prev);
76 struct thread_info *ni = task_thread_info(next);
77 cputime_t delta_stime, delta_utime;
78 __u64 now;
79
80 now = ia64_get_itc();
81
82 delta_stime = cycle_to_cputime(pi->ac_stime + (now - pi->ac_stamp));
83 account_system_time(prev, 0, delta_stime);
84 account_system_time_scaled(prev, delta_stime);
85
86 if (pi->ac_utime) {
87 delta_utime = cycle_to_cputime(pi->ac_utime);
88 account_user_time(prev, delta_utime);
89 account_user_time_scaled(prev, delta_utime);
90 }
91
92 pi->ac_stamp = ni->ac_stamp = now;
93 ni->ac_stime = ni->ac_utime = 0;
94}
95
96/*
97 * Account time for a transition between system, hard irq or soft irq state.
98 * Note that this function is called with interrupts enabled.
99 */
100void account_system_vtime(struct task_struct *tsk)
101{
102 struct thread_info *ti = task_thread_info(tsk);
103 unsigned long flags;
104 cputime_t delta_stime;
105 __u64 now;
106
107 local_irq_save(flags);
108
109 now = ia64_get_itc();
110
111 delta_stime = cycle_to_cputime(ti->ac_stime + (now - ti->ac_stamp));
112 account_system_time(tsk, 0, delta_stime);
113 account_system_time_scaled(tsk, delta_stime);
114 ti->ac_stime = 0;
115
116 ti->ac_stamp = now;
117
118 local_irq_restore(flags);
119}
120
121/*
122 * Called from the timer interrupt handler to charge accumulated user time
123 * to the current process. Must be called with interrupts disabled.
124 */
125void account_process_tick(struct task_struct *p, int user_tick)
126{
127 struct thread_info *ti = task_thread_info(p);
128 cputime_t delta_utime;
129
130 if (ti->ac_utime) {
131 delta_utime = cycle_to_cputime(ti->ac_utime);
132 account_user_time(p, delta_utime);
133 account_user_time_scaled(p, delta_utime);
134 ti->ac_utime = 0;
135 }
136}
137
138#endif /* CONFIG_VIRT_CPU_ACCOUNTING */
139
62static irqreturn_t 140static irqreturn_t
63timer_interrupt (int irq, void *dev_id) 141timer_interrupt (int irq, void *dev_id)
64{ 142{
diff --git a/arch/ia64/kernel/unaligned.c b/arch/ia64/kernel/unaligned.c
index 6903361d11a5..ff0e7c10faa7 100644
--- a/arch/ia64/kernel/unaligned.c
+++ b/arch/ia64/kernel/unaligned.c
@@ -13,6 +13,7 @@
13 * 2001/08/13 Correct size of extended floats (float_fsz) from 16 to 10 bytes. 13 * 2001/08/13 Correct size of extended floats (float_fsz) from 16 to 10 bytes.
14 * 2001/01/17 Add support emulation of unaligned kernel accesses. 14 * 2001/01/17 Add support emulation of unaligned kernel accesses.
15 */ 15 */
16#include <linux/jiffies.h>
16#include <linux/kernel.h> 17#include <linux/kernel.h>
17#include <linux/sched.h> 18#include <linux/sched.h>
18#include <linux/tty.h> 19#include <linux/tty.h>
@@ -1290,7 +1291,7 @@ within_logging_rate_limit (void)
1290{ 1291{
1291 static unsigned long count, last_time; 1292 static unsigned long count, last_time;
1292 1293
1293 if (jiffies - last_time > 5*HZ) 1294 if (time_after(jiffies, last_time + 5 * HZ))
1294 count = 0; 1295 count = 0;
1295 if (count < 5) { 1296 if (count < 5) {
1296 last_time = jiffies; 1297 last_time = jiffies;
diff --git a/arch/ia64/mm/contig.c b/arch/ia64/mm/contig.c
index 344f64eca7a9..798bf9835a51 100644
--- a/arch/ia64/mm/contig.c
+++ b/arch/ia64/mm/contig.c
@@ -45,8 +45,6 @@ void show_mem(void)
45 45
46 printk(KERN_INFO "Mem-info:\n"); 46 printk(KERN_INFO "Mem-info:\n");
47 show_free_areas(); 47 show_free_areas();
48 printk(KERN_INFO "Free swap: %6ldkB\n",
49 nr_swap_pages<<(PAGE_SHIFT-10));
50 printk(KERN_INFO "Node memory in pages:\n"); 48 printk(KERN_INFO "Node memory in pages:\n");
51 for_each_online_pgdat(pgdat) { 49 for_each_online_pgdat(pgdat) {
52 unsigned long present; 50 unsigned long present;
@@ -255,7 +253,7 @@ paging_init (void)
255 max_zone_pfns[ZONE_NORMAL] = max_low_pfn; 253 max_zone_pfns[ZONE_NORMAL] = max_low_pfn;
256 254
257#ifdef CONFIG_VIRTUAL_MEM_MAP 255#ifdef CONFIG_VIRTUAL_MEM_MAP
258 efi_memmap_walk(register_active_ranges, NULL); 256 efi_memmap_walk(filter_memory, register_active_ranges);
259 efi_memmap_walk(find_largest_hole, (u64 *)&max_gap); 257 efi_memmap_walk(find_largest_hole, (u64 *)&max_gap);
260 if (max_gap < LARGE_GAP) { 258 if (max_gap < LARGE_GAP) {
261 vmem_map = (struct page *) 0; 259 vmem_map = (struct page *) 0;
diff --git a/arch/ia64/mm/discontig.c b/arch/ia64/mm/discontig.c
index ee5e68b2af94..544dc420c65e 100644
--- a/arch/ia64/mm/discontig.c
+++ b/arch/ia64/mm/discontig.c
@@ -104,7 +104,7 @@ static int __meminit early_nr_cpus_node(int node)
104{ 104{
105 int cpu, n = 0; 105 int cpu, n = 0;
106 106
107 for (cpu = 0; cpu < NR_CPUS; cpu++) 107 for_each_possible_early_cpu(cpu)
108 if (node == node_cpuid[cpu].nid) 108 if (node == node_cpuid[cpu].nid)
109 n++; 109 n++;
110 110
@@ -124,6 +124,7 @@ static unsigned long __meminit compute_pernodesize(int node)
124 pernodesize += node * L1_CACHE_BYTES; 124 pernodesize += node * L1_CACHE_BYTES;
125 pernodesize += L1_CACHE_ALIGN(sizeof(pg_data_t)); 125 pernodesize += L1_CACHE_ALIGN(sizeof(pg_data_t));
126 pernodesize += L1_CACHE_ALIGN(sizeof(struct ia64_node_data)); 126 pernodesize += L1_CACHE_ALIGN(sizeof(struct ia64_node_data));
127 pernodesize += L1_CACHE_ALIGN(sizeof(pg_data_t));
127 pernodesize = PAGE_ALIGN(pernodesize); 128 pernodesize = PAGE_ALIGN(pernodesize);
128 return pernodesize; 129 return pernodesize;
129} 130}
@@ -142,7 +143,7 @@ static void *per_cpu_node_setup(void *cpu_data, int node)
142#ifdef CONFIG_SMP 143#ifdef CONFIG_SMP
143 int cpu; 144 int cpu;
144 145
145 for (cpu = 0; cpu < NR_CPUS; cpu++) { 146 for_each_possible_early_cpu(cpu) {
146 if (node == node_cpuid[cpu].nid) { 147 if (node == node_cpuid[cpu].nid) {
147 memcpy(__va(cpu_data), __phys_per_cpu_start, 148 memcpy(__va(cpu_data), __phys_per_cpu_start,
148 __per_cpu_end - __per_cpu_start); 149 __per_cpu_end - __per_cpu_start);
@@ -345,7 +346,7 @@ static void __init initialize_pernode_data(void)
345 346
346#ifdef CONFIG_SMP 347#ifdef CONFIG_SMP
347 /* Set the node_data pointer for each per-cpu struct */ 348 /* Set the node_data pointer for each per-cpu struct */
348 for (cpu = 0; cpu < NR_CPUS; cpu++) { 349 for_each_possible_early_cpu(cpu) {
349 node = node_cpuid[cpu].nid; 350 node = node_cpuid[cpu].nid;
350 per_cpu(cpu_info, cpu).node_data = mem_data[node].node_data; 351 per_cpu(cpu_info, cpu).node_data = mem_data[node].node_data;
351 } 352 }
@@ -444,7 +445,7 @@ void __init find_memory(void)
444 mem_data[node].min_pfn = ~0UL; 445 mem_data[node].min_pfn = ~0UL;
445 } 446 }
446 447
447 efi_memmap_walk(register_active_ranges, NULL); 448 efi_memmap_walk(filter_memory, register_active_ranges);
448 449
449 /* 450 /*
450 * Initialize the boot memory maps in reverse order since that's 451 * Initialize the boot memory maps in reverse order since that's
@@ -493,13 +494,9 @@ void __cpuinit *per_cpu_init(void)
493 int cpu; 494 int cpu;
494 static int first_time = 1; 495 static int first_time = 1;
495 496
496
497 if (smp_processor_id() != 0)
498 return __per_cpu_start + __per_cpu_offset[smp_processor_id()];
499
500 if (first_time) { 497 if (first_time) {
501 first_time = 0; 498 first_time = 0;
502 for (cpu = 0; cpu < NR_CPUS; cpu++) 499 for_each_possible_early_cpu(cpu)
503 per_cpu(local_per_cpu_offset, cpu) = __per_cpu_offset[cpu]; 500 per_cpu(local_per_cpu_offset, cpu) = __per_cpu_offset[cpu];
504 } 501 }
505 502
@@ -522,8 +519,6 @@ void show_mem(void)
522 519
523 printk(KERN_INFO "Mem-info:\n"); 520 printk(KERN_INFO "Mem-info:\n");
524 show_free_areas(); 521 show_free_areas();
525 printk(KERN_INFO "Free swap: %6ldkB\n",
526 nr_swap_pages<<(PAGE_SHIFT-10));
527 printk(KERN_INFO "Node memory in pages:\n"); 522 printk(KERN_INFO "Node memory in pages:\n");
528 for_each_online_pgdat(pgdat) { 523 for_each_online_pgdat(pgdat) {
529 unsigned long present; 524 unsigned long present;
diff --git a/arch/ia64/mm/init.c b/arch/ia64/mm/init.c
index a4ca657c72c6..5c1de53c8c1c 100644
--- a/arch/ia64/mm/init.c
+++ b/arch/ia64/mm/init.c
@@ -58,7 +58,6 @@ __ia64_sync_icache_dcache (pte_t pte)
58{ 58{
59 unsigned long addr; 59 unsigned long addr;
60 struct page *page; 60 struct page *page;
61 unsigned long order;
62 61
63 page = pte_page(pte); 62 page = pte_page(pte);
64 addr = (unsigned long) page_address(page); 63 addr = (unsigned long) page_address(page);
@@ -66,12 +65,7 @@ __ia64_sync_icache_dcache (pte_t pte)
66 if (test_bit(PG_arch_1, &page->flags)) 65 if (test_bit(PG_arch_1, &page->flags))
67 return; /* i-cache is already coherent with d-cache */ 66 return; /* i-cache is already coherent with d-cache */
68 67
69 if (PageCompound(page)) { 68 flush_icache_range(addr, addr + (PAGE_SIZE << compound_order(page)));
70 order = compound_order(page);
71 flush_icache_range(addr, addr + (1UL << order << PAGE_SHIFT));
72 }
73 else
74 flush_icache_range(addr, addr + PAGE_SIZE);
75 set_bit(PG_arch_1, &page->flags); /* mark page as clean */ 69 set_bit(PG_arch_1, &page->flags); /* mark page as clean */
76} 70}
77 71
@@ -553,12 +547,10 @@ find_largest_hole (u64 start, u64 end, void *arg)
553#endif /* CONFIG_VIRTUAL_MEM_MAP */ 547#endif /* CONFIG_VIRTUAL_MEM_MAP */
554 548
555int __init 549int __init
556register_active_ranges(u64 start, u64 end, void *arg) 550register_active_ranges(u64 start, u64 len, int nid)
557{ 551{
558 int nid = paddr_to_nid(__pa(start)); 552 u64 end = start + len;
559 553
560 if (nid < 0)
561 nid = 0;
562#ifdef CONFIG_KEXEC 554#ifdef CONFIG_KEXEC
563 if (start > crashk_res.start && start < crashk_res.end) 555 if (start > crashk_res.start && start < crashk_res.end)
564 start = crashk_res.end; 556 start = crashk_res.end;
diff --git a/arch/ia64/mm/numa.c b/arch/ia64/mm/numa.c
index 7807fc5c0422..b73bf1838e57 100644
--- a/arch/ia64/mm/numa.c
+++ b/arch/ia64/mm/numa.c
@@ -27,7 +27,9 @@
27 */ 27 */
28int num_node_memblks; 28int num_node_memblks;
29struct node_memblk_s node_memblk[NR_NODE_MEMBLKS]; 29struct node_memblk_s node_memblk[NR_NODE_MEMBLKS];
30struct node_cpuid_s node_cpuid[NR_CPUS]; 30struct node_cpuid_s node_cpuid[NR_CPUS] =
31 { [0 ... NR_CPUS-1] = { .phys_id = 0, .nid = NUMA_NO_NODE } };
32
31/* 33/*
32 * This is a matrix with "distances" between nodes, they should be 34 * This is a matrix with "distances" between nodes, they should be
33 * proportional to the memory access latency ratios. 35 * proportional to the memory access latency ratios.
diff --git a/arch/ia64/mm/tlb.c b/arch/ia64/mm/tlb.c
index 655da240d13c..d52ec4e83409 100644
--- a/arch/ia64/mm/tlb.c
+++ b/arch/ia64/mm/tlb.c
@@ -11,6 +11,9 @@
11 * Rohit Seth <rohit.seth@intel.com> 11 * Rohit Seth <rohit.seth@intel.com>
12 * Ken Chen <kenneth.w.chen@intel.com> 12 * Ken Chen <kenneth.w.chen@intel.com>
13 * Christophe de Dinechin <ddd@hp.com>: Avoid ptc.e on memory allocation 13 * Christophe de Dinechin <ddd@hp.com>: Avoid ptc.e on memory allocation
14 * Copyright (C) 2007 Intel Corp
15 * Fenghua Yu <fenghua.yu@intel.com>
16 * Add multiple ptc.g/ptc.ga instruction support in global tlb purge.
14 */ 17 */
15#include <linux/module.h> 18#include <linux/module.h>
16#include <linux/init.h> 19#include <linux/init.h>
@@ -26,6 +29,9 @@
26#include <asm/pal.h> 29#include <asm/pal.h>
27#include <asm/tlbflush.h> 30#include <asm/tlbflush.h>
28#include <asm/dma.h> 31#include <asm/dma.h>
32#include <asm/processor.h>
33#include <asm/sal.h>
34#include <asm/tlb.h>
29 35
30static struct { 36static struct {
31 unsigned long mask; /* mask of supported purge page-sizes */ 37 unsigned long mask; /* mask of supported purge page-sizes */
@@ -39,6 +45,10 @@ struct ia64_ctx ia64_ctx = {
39}; 45};
40 46
41DEFINE_PER_CPU(u8, ia64_need_tlb_flush); 47DEFINE_PER_CPU(u8, ia64_need_tlb_flush);
48DEFINE_PER_CPU(u8, ia64_tr_num); /*Number of TR slots in current processor*/
49DEFINE_PER_CPU(u8, ia64_tr_used); /*Max Slot number used by kernel*/
50
51struct ia64_tr_entry __per_cpu_idtrs[NR_CPUS][2][IA64_TR_ALLOC_MAX];
42 52
43/* 53/*
44 * Initializes the ia64_ctx.bitmap array based on max_ctx+1. 54 * Initializes the ia64_ctx.bitmap array based on max_ctx+1.
@@ -84,14 +94,140 @@ wrap_mmu_context (struct mm_struct *mm)
84 local_flush_tlb_all(); 94 local_flush_tlb_all();
85} 95}
86 96
97/*
98 * Implement "spinaphores" ... like counting semaphores, but they
99 * spin instead of sleeping. If there are ever any other users for
100 * this primitive it can be moved up to a spinaphore.h header.
101 */
102struct spinaphore {
103 atomic_t cur;
104};
105
106static inline void spinaphore_init(struct spinaphore *ss, int val)
107{
108 atomic_set(&ss->cur, val);
109}
110
111static inline void down_spin(struct spinaphore *ss)
112{
113 while (unlikely(!atomic_add_unless(&ss->cur, -1, 0)))
114 while (atomic_read(&ss->cur) == 0)
115 cpu_relax();
116}
117
118static inline void up_spin(struct spinaphore *ss)
119{
120 atomic_add(1, &ss->cur);
121}
122
123static struct spinaphore ptcg_sem;
124static u16 nptcg = 1;
125static int need_ptcg_sem = 1;
126static int toolatetochangeptcgsem = 0;
127
128/*
129 * Kernel parameter "nptcg=" overrides max number of concurrent global TLB
130 * purges which is reported from either PAL or SAL PALO.
131 *
132 * We don't have sanity checking for nptcg value. It's the user's responsibility
133 * for valid nptcg value on the platform. Otherwise, kernel may hang in some
134 * cases.
135 */
136static int __init
137set_nptcg(char *str)
138{
139 int value = 0;
140
141 get_option(&str, &value);
142 setup_ptcg_sem(value, NPTCG_FROM_KERNEL_PARAMETER);
143
144 return 1;
145}
146
147__setup("nptcg=", set_nptcg);
148
149/*
150 * Maximum number of simultaneous ptc.g purges in the system can
151 * be defined by PAL_VM_SUMMARY (in which case we should take
152 * the smallest value for any cpu in the system) or by the PAL
153 * override table (in which case we should ignore the value from
154 * PAL_VM_SUMMARY).
155 *
156 * Kernel parameter "nptcg=" overrides maximum number of simultanesous ptc.g
157 * purges defined in either PAL_VM_SUMMARY or PAL override table. In this case,
158 * we should ignore the value from either PAL_VM_SUMMARY or PAL override table.
159 *
160 * Complicating the logic here is the fact that num_possible_cpus()
161 * isn't fully setup until we start bringing cpus online.
162 */
163void
164setup_ptcg_sem(int max_purges, int nptcg_from)
165{
166 static int kp_override;
167 static int palo_override;
168 static int firstcpu = 1;
169
170 if (toolatetochangeptcgsem) {
171 BUG_ON(max_purges < nptcg);
172 return;
173 }
174
175 if (nptcg_from == NPTCG_FROM_KERNEL_PARAMETER) {
176 kp_override = 1;
177 nptcg = max_purges;
178 goto resetsema;
179 }
180 if (kp_override) {
181 need_ptcg_sem = num_possible_cpus() > nptcg;
182 return;
183 }
184
185 if (nptcg_from == NPTCG_FROM_PALO) {
186 palo_override = 1;
187
188 /* In PALO max_purges == 0 really means it! */
189 if (max_purges == 0)
190 panic("Whoa! Platform does not support global TLB purges.\n");
191 nptcg = max_purges;
192 if (nptcg == PALO_MAX_TLB_PURGES) {
193 need_ptcg_sem = 0;
194 return;
195 }
196 goto resetsema;
197 }
198 if (palo_override) {
199 if (nptcg != PALO_MAX_TLB_PURGES)
200 need_ptcg_sem = (num_possible_cpus() > nptcg);
201 return;
202 }
203
204 /* In PAL_VM_SUMMARY max_purges == 0 actually means 1 */
205 if (max_purges == 0) max_purges = 1;
206
207 if (firstcpu) {
208 nptcg = max_purges;
209 firstcpu = 0;
210 }
211 if (max_purges < nptcg)
212 nptcg = max_purges;
213 if (nptcg == PAL_MAX_PURGES) {
214 need_ptcg_sem = 0;
215 return;
216 } else
217 need_ptcg_sem = (num_possible_cpus() > nptcg);
218
219resetsema:
220 spinaphore_init(&ptcg_sem, max_purges);
221}
222
87void 223void
88ia64_global_tlb_purge (struct mm_struct *mm, unsigned long start, 224ia64_global_tlb_purge (struct mm_struct *mm, unsigned long start,
89 unsigned long end, unsigned long nbits) 225 unsigned long end, unsigned long nbits)
90{ 226{
91 static DEFINE_SPINLOCK(ptcg_lock);
92
93 struct mm_struct *active_mm = current->active_mm; 227 struct mm_struct *active_mm = current->active_mm;
94 228
229 toolatetochangeptcgsem = 1;
230
95 if (mm != active_mm) { 231 if (mm != active_mm) {
96 /* Restore region IDs for mm */ 232 /* Restore region IDs for mm */
97 if (mm && active_mm) { 233 if (mm && active_mm) {
@@ -102,19 +238,20 @@ ia64_global_tlb_purge (struct mm_struct *mm, unsigned long start,
102 } 238 }
103 } 239 }
104 240
105 /* HW requires global serialization of ptc.ga. */ 241 if (need_ptcg_sem)
106 spin_lock(&ptcg_lock); 242 down_spin(&ptcg_sem);
107 { 243
108 do { 244 do {
109 /* 245 /*
110 * Flush ALAT entries also. 246 * Flush ALAT entries also.
111 */ 247 */
112 ia64_ptcga(start, (nbits<<2)); 248 ia64_ptcga(start, (nbits << 2));
113 ia64_srlz_i(); 249 ia64_srlz_i();
114 start += (1UL << nbits); 250 start += (1UL << nbits);
115 } while (start < end); 251 } while (start < end);
116 } 252
117 spin_unlock(&ptcg_lock); 253 if (need_ptcg_sem)
254 up_spin(&ptcg_sem);
118 255
119 if (mm != active_mm) { 256 if (mm != active_mm) {
120 activate_context(active_mm); 257 activate_context(active_mm);
@@ -190,6 +327,9 @@ ia64_tlb_init (void)
190 ia64_ptce_info_t uninitialized_var(ptce_info); /* GCC be quiet */ 327 ia64_ptce_info_t uninitialized_var(ptce_info); /* GCC be quiet */
191 unsigned long tr_pgbits; 328 unsigned long tr_pgbits;
192 long status; 329 long status;
330 pal_vm_info_1_u_t vm_info_1;
331 pal_vm_info_2_u_t vm_info_2;
332 int cpu = smp_processor_id();
193 333
194 if ((status = ia64_pal_vm_page_size(&tr_pgbits, &purge.mask)) != 0) { 334 if ((status = ia64_pal_vm_page_size(&tr_pgbits, &purge.mask)) != 0) {
195 printk(KERN_ERR "PAL_VM_PAGE_SIZE failed with status=%ld; " 335 printk(KERN_ERR "PAL_VM_PAGE_SIZE failed with status=%ld; "
@@ -206,4 +346,191 @@ ia64_tlb_init (void)
206 local_cpu_data->ptce_stride[1] = ptce_info.stride[1]; 346 local_cpu_data->ptce_stride[1] = ptce_info.stride[1];
207 347
208 local_flush_tlb_all(); /* nuke left overs from bootstrapping... */ 348 local_flush_tlb_all(); /* nuke left overs from bootstrapping... */
349 status = ia64_pal_vm_summary(&vm_info_1, &vm_info_2);
350
351 if (status) {
352 printk(KERN_ERR "ia64_pal_vm_summary=%ld\n", status);
353 per_cpu(ia64_tr_num, cpu) = 8;
354 return;
355 }
356 per_cpu(ia64_tr_num, cpu) = vm_info_1.pal_vm_info_1_s.max_itr_entry+1;
357 if (per_cpu(ia64_tr_num, cpu) >
358 (vm_info_1.pal_vm_info_1_s.max_dtr_entry+1))
359 per_cpu(ia64_tr_num, cpu) =
360 vm_info_1.pal_vm_info_1_s.max_dtr_entry+1;
361 if (per_cpu(ia64_tr_num, cpu) > IA64_TR_ALLOC_MAX) {
362 per_cpu(ia64_tr_num, cpu) = IA64_TR_ALLOC_MAX;
363 printk(KERN_DEBUG "TR register number exceeds IA64_TR_ALLOC_MAX!"
364 "IA64_TR_ALLOC_MAX should be extended\n");
365 }
366}
367
368/*
369 * is_tr_overlap
370 *
371 * Check overlap with inserted TRs.
372 */
373static int is_tr_overlap(struct ia64_tr_entry *p, u64 va, u64 log_size)
374{
375 u64 tr_log_size;
376 u64 tr_end;
377 u64 va_rr = ia64_get_rr(va);
378 u64 va_rid = RR_TO_RID(va_rr);
379 u64 va_end = va + (1<<log_size) - 1;
380
381 if (va_rid != RR_TO_RID(p->rr))
382 return 0;
383 tr_log_size = (p->itir & 0xff) >> 2;
384 tr_end = p->ifa + (1<<tr_log_size) - 1;
385
386 if (va > tr_end || p->ifa > va_end)
387 return 0;
388 return 1;
389
390}
391
392/*
393 * ia64_insert_tr in virtual mode. Allocate a TR slot
394 *
395 * target_mask : 0x1 : itr, 0x2 : dtr, 0x3 : idtr
396 *
397 * va : virtual address.
398 * pte : pte entries inserted.
399 * log_size: range to be covered.
400 *
401 * Return value: <0 : error No.
402 *
403 * >=0 : slot number allocated for TR.
404 * Must be called with preemption disabled.
405 */
406int ia64_itr_entry(u64 target_mask, u64 va, u64 pte, u64 log_size)
407{
408 int i, r;
409 unsigned long psr;
410 struct ia64_tr_entry *p;
411 int cpu = smp_processor_id();
412
413 r = -EINVAL;
414 /*Check overlap with existing TR entries*/
415 if (target_mask & 0x1) {
416 p = &__per_cpu_idtrs[cpu][0][0];
417 for (i = IA64_TR_ALLOC_BASE; i <= per_cpu(ia64_tr_used, cpu);
418 i++, p++) {
419 if (p->pte & 0x1)
420 if (is_tr_overlap(p, va, log_size)) {
421 printk(KERN_DEBUG "Overlapped Entry"
422 "Inserted for TR Reigster!!\n");
423 goto out;
424 }
425 }
426 }
427 if (target_mask & 0x2) {
428 p = &__per_cpu_idtrs[cpu][1][0];
429 for (i = IA64_TR_ALLOC_BASE; i <= per_cpu(ia64_tr_used, cpu);
430 i++, p++) {
431 if (p->pte & 0x1)
432 if (is_tr_overlap(p, va, log_size)) {
433 printk(KERN_DEBUG "Overlapped Entry"
434 "Inserted for TR Reigster!!\n");
435 goto out;
436 }
437 }
438 }
439
440 for (i = IA64_TR_ALLOC_BASE; i < per_cpu(ia64_tr_num, cpu); i++) {
441 switch (target_mask & 0x3) {
442 case 1:
443 if (!(__per_cpu_idtrs[cpu][0][i].pte & 0x1))
444 goto found;
445 continue;
446 case 2:
447 if (!(__per_cpu_idtrs[cpu][1][i].pte & 0x1))
448 goto found;
449 continue;
450 case 3:
451 if (!(__per_cpu_idtrs[cpu][0][i].pte & 0x1) &&
452 !(__per_cpu_idtrs[cpu][1][i].pte & 0x1))
453 goto found;
454 continue;
455 default:
456 r = -EINVAL;
457 goto out;
458 }
459 }
460found:
461 if (i >= per_cpu(ia64_tr_num, cpu))
462 return -EBUSY;
463
464 /*Record tr info for mca hander use!*/
465 if (i > per_cpu(ia64_tr_used, cpu))
466 per_cpu(ia64_tr_used, cpu) = i;
467
468 psr = ia64_clear_ic();
469 if (target_mask & 0x1) {
470 ia64_itr(0x1, i, va, pte, log_size);
471 ia64_srlz_i();
472 p = &__per_cpu_idtrs[cpu][0][i];
473 p->ifa = va;
474 p->pte = pte;
475 p->itir = log_size << 2;
476 p->rr = ia64_get_rr(va);
477 }
478 if (target_mask & 0x2) {
479 ia64_itr(0x2, i, va, pte, log_size);
480 ia64_srlz_i();
481 p = &__per_cpu_idtrs[cpu][1][i];
482 p->ifa = va;
483 p->pte = pte;
484 p->itir = log_size << 2;
485 p->rr = ia64_get_rr(va);
486 }
487 ia64_set_psr(psr);
488 r = i;
489out:
490 return r;
491}
492EXPORT_SYMBOL_GPL(ia64_itr_entry);
493
494/*
495 * ia64_purge_tr
496 *
497 * target_mask: 0x1: purge itr, 0x2 : purge dtr, 0x3 purge idtr.
498 * slot: slot number to be freed.
499 *
500 * Must be called with preemption disabled.
501 */
502void ia64_ptr_entry(u64 target_mask, int slot)
503{
504 int cpu = smp_processor_id();
505 int i;
506 struct ia64_tr_entry *p;
507
508 if (slot < IA64_TR_ALLOC_BASE || slot >= per_cpu(ia64_tr_num, cpu))
509 return;
510
511 if (target_mask & 0x1) {
512 p = &__per_cpu_idtrs[cpu][0][slot];
513 if ((p->pte&0x1) && is_tr_overlap(p, p->ifa, p->itir>>2)) {
514 p->pte = 0;
515 ia64_ptr(0x1, p->ifa, p->itir>>2);
516 ia64_srlz_i();
517 }
518 }
519
520 if (target_mask & 0x2) {
521 p = &__per_cpu_idtrs[cpu][1][slot];
522 if ((p->pte & 0x1) && is_tr_overlap(p, p->ifa, p->itir>>2)) {
523 p->pte = 0;
524 ia64_ptr(0x2, p->ifa, p->itir>>2);
525 ia64_srlz_i();
526 }
527 }
528
529 for (i = per_cpu(ia64_tr_used, cpu); i >= IA64_TR_ALLOC_BASE; i--) {
530 if ((__per_cpu_idtrs[cpu][0][i].pte & 0x1) ||
531 (__per_cpu_idtrs[cpu][1][i].pte & 0x1))
532 break;
533 }
534 per_cpu(ia64_tr_used, cpu) = i;
209} 535}
536EXPORT_SYMBOL_GPL(ia64_ptr_entry);
diff --git a/arch/ia64/sn/kernel/xpc_main.c b/arch/ia64/sn/kernel/xpc_main.c
index 81785b78bc1e..9e0b164da9c2 100644
--- a/arch/ia64/sn/kernel/xpc_main.c
+++ b/arch/ia64/sn/kernel/xpc_main.c
@@ -199,7 +199,7 @@ xpc_timeout_partition_disengage_request(unsigned long data)
199 struct xpc_partition *part = (struct xpc_partition *) data; 199 struct xpc_partition *part = (struct xpc_partition *) data;
200 200
201 201
202 DBUG_ON(jiffies < part->disengage_request_timeout); 202 DBUG_ON(time_before(jiffies, part->disengage_request_timeout));
203 203
204 (void) xpc_partition_disengaged(part); 204 (void) xpc_partition_disengaged(part);
205 205
@@ -230,7 +230,7 @@ xpc_hb_beater(unsigned long dummy)
230{ 230{
231 xpc_vars->heartbeat++; 231 xpc_vars->heartbeat++;
232 232
233 if (jiffies >= xpc_hb_check_timeout) { 233 if (time_after_eq(jiffies, xpc_hb_check_timeout)) {
234 wake_up_interruptible(&xpc_act_IRQ_wq); 234 wake_up_interruptible(&xpc_act_IRQ_wq);
235 } 235 }
236 236
@@ -270,7 +270,7 @@ xpc_hb_checker(void *ignore)
270 270
271 271
272 /* checking of remote heartbeats is skewed by IRQ handling */ 272 /* checking of remote heartbeats is skewed by IRQ handling */
273 if (jiffies >= xpc_hb_check_timeout) { 273 if (time_after_eq(jiffies, xpc_hb_check_timeout)) {
274 dev_dbg(xpc_part, "checking remote heartbeats\n"); 274 dev_dbg(xpc_part, "checking remote heartbeats\n");
275 xpc_check_remote_hb(); 275 xpc_check_remote_hb();
276 276
@@ -305,7 +305,7 @@ xpc_hb_checker(void *ignore)
305 /* wait for IRQ or timeout */ 305 /* wait for IRQ or timeout */
306 (void) wait_event_interruptible(xpc_act_IRQ_wq, 306 (void) wait_event_interruptible(xpc_act_IRQ_wq,
307 (last_IRQ_count < atomic_read(&xpc_act_IRQ_rcvd) || 307 (last_IRQ_count < atomic_read(&xpc_act_IRQ_rcvd) ||
308 jiffies >= xpc_hb_check_timeout || 308 time_after_eq(jiffies, xpc_hb_check_timeout) ||
309 (volatile int) xpc_exiting)); 309 (volatile int) xpc_exiting));
310 } 310 }
311 311
diff --git a/arch/ia64/sn/kernel/xpc_partition.c b/arch/ia64/sn/kernel/xpc_partition.c
index 7ba403232cb8..9e97c2684832 100644
--- a/arch/ia64/sn/kernel/xpc_partition.c
+++ b/arch/ia64/sn/kernel/xpc_partition.c
@@ -877,7 +877,7 @@ xpc_partition_disengaged(struct xpc_partition *part)
877 disengaged = (xpc_partition_engaged(1UL << partid) == 0); 877 disengaged = (xpc_partition_engaged(1UL << partid) == 0);
878 if (part->disengage_request_timeout) { 878 if (part->disengage_request_timeout) {
879 if (!disengaged) { 879 if (!disengaged) {
880 if (jiffies < part->disengage_request_timeout) { 880 if (time_before(jiffies, part->disengage_request_timeout)) {
881 /* timelimit hasn't been reached yet */ 881 /* timelimit hasn't been reached yet */
882 return 0; 882 return 0;
883 } 883 }
generated by cgit v1.2.2 (git 2.25.0) at 2024-11-11 04:05:02 -0500