/* libunwind - a platform-independent unwind library Copyright (C) 2001-2005 Hewlett-Packard Co Contributed by David Mosberger-Tang <davidm@hpl.hp.com> This file is part of libunwind. Permission is hereby granted, free of charge, to any person obtaining a copy of this software and associated documentation files (the "Software"), to deal in the Software without restriction, including without limitation the rights to use, copy, modify, merge, publish, distribute, sublicense, and/or sell copies of the Software, and to permit persons to whom the Software is furnished to do so, subject to the following conditions: The above copyright notice and this permission notice shall be included in all copies or substantial portions of the Software. THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. */ #include "offsets.h" #include "regs.h" #include "unwind_i.h" enum ia64_script_insn_opcode { IA64_INSN_INC_PSP, /* psp += val */ IA64_INSN_LOAD_PSP, /* psp = *psp_loc */ IA64_INSN_ADD_PSP, /* s[dst] = (s.psp + val) */ IA64_INSN_ADD_PSP_NAT, /* like above, but with NaT info */ IA64_INSN_ADD_SP, /* s[dst] = (s.sp + val) */ IA64_INSN_ADD_SP_NAT, /* like above, but with NaT info */ IA64_INSN_MOVE, /* s[dst] = s[val] */ IA64_INSN_MOVE_NAT, /* like above, but with NaT info */ IA64_INSN_MOVE_NO_NAT, /* like above, but clear NaT info */ IA64_INSN_MOVE_STACKED, /* s[dst] = rse_skip(*s.bsp_loc, val) */ IA64_INSN_MOVE_STACKED_NAT, /* like above, but with NaT info */ IA64_INSN_MOVE_SCRATCH, /* s[dst] = scratch reg "val" */ IA64_INSN_MOVE_SCRATCH_NAT, /* like above, but with NaT info */ IA64_INSN_MOVE_SCRATCH_NO_NAT /* like above, but clear NaT info */ }; #ifdef HAVE___THREAD static __thread struct ia64_script_cache ia64_per_thread_cache = { #ifdef HAVE_ATOMIC_OPS_H .busy = AO_TS_INITIALIZER #else .lock = PTHREAD_MUTEX_INITIALIZER #endif }; #endif static inline unw_hash_index_t CONST_ATTR hash (unw_word_t ip) { /* based on (sqrt(5)/2-1)*2^64 */ # define magic ((unw_word_t) 0x9e3779b97f4a7c16ULL) return (ip >> 4) * magic >> (64 - IA64_LOG_UNW_HASH_SIZE); } static inline long cache_match (struct ia64_script *script, unw_word_t ip, unw_word_t pr) { if (ip == script->ip && ((pr ^ script->pr_val) & script->pr_mask) == 0) return 1; return 0; } static inline void flush_script_cache (struct ia64_script_cache *cache) { int i; cache->lru_head = IA64_UNW_CACHE_SIZE - 1; cache->lru_tail = 0; for (i = 0; i < IA64_UNW_CACHE_SIZE; ++i) { if (i > 0) cache->buckets[i].lru_chain = (i - 1); cache->buckets[i].coll_chain = -1; cache->buckets[i].ip = 0; } for (i = 0; i<IA64_UNW_HASH_SIZE; ++i) cache->hash[i] = -1; } static inline struct ia64_script_cache * get_script_cache (unw_addr_space_t as, intrmask_t *saved_maskp) { struct ia64_script_cache *cache = &as->global_cache; unw_caching_policy_t caching = as->caching_policy; if (caching == UNW_CACHE_NONE) return NULL; #ifdef HAVE_ATOMIC_H if (!spin_trylock_irqsave (&cache->busy, *saved_maskp)) return NULL; #else # ifdef HAVE___THREAD if (as->caching_policy == UNW_CACHE_PER_THREAD) cache = &ia64_per_thread_cache; # endif # ifdef HAVE_ATOMIC_OPS_H if (AO_test_and_set (&cache->busy) == AO_TS_SET) return NULL; # else if (likely (caching == UNW_CACHE_GLOBAL)) { Debug (16, "acquiring lock\n"); lock_acquire (&cache->lock, *saved_maskp); } # endif #endif if (atomic_read (&as->cache_generation) != atomic_read (&cache->generation)) { flush_script_cache (cache); cache->generation = as->cache_generation; } return cache; } static inline void put_script_cache (unw_addr_space_t as, struct ia64_script_cache *cache, intrmask_t *saved_maskp) { assert (as->caching_policy != UNW_CACHE_NONE); Debug (16, "unmasking signals/interrupts and releasing lock\n"); #ifdef HAVE_ATOMIC_H spin_unlock_irqrestore (&cache->busy, *saved_maskp); #else # ifdef HAVE_ATOMIC_OPS_H AO_CLEAR (&cache->busy); # else if (likely (as->caching_policy == UNW_CACHE_GLOBAL)) lock_release (&cache->lock, *saved_maskp); # endif #endif } static struct ia64_script * script_lookup (struct ia64_script_cache *cache, struct cursor *c) { struct ia64_script *script = cache->buckets + c->hint; unsigned short index; unw_word_t ip, pr; ip = c->ip; pr = c->pr; if (cache_match (script, ip, pr)) return script; index = cache->hash[hash (ip)]; if (index >= IA64_UNW_CACHE_SIZE) return 0; script = cache->buckets + index; while (1) { if (cache_match (script, ip, pr)) { /* update hint; no locking needed: single-word writes are atomic */ c->hint = cache->buckets[c->prev_script].hint = (script - cache->buckets); return script; } if (script->coll_chain >= IA64_UNW_HASH_SIZE) return 0; script = cache->buckets + script->coll_chain; } } static inline void script_init (struct ia64_script *script, unw_word_t ip) { script->ip = ip; script->hint = 0; script->count = 0; script->abi_marker = 0; } static inline struct ia64_script * script_new (struct ia64_script_cache *cache, unw_word_t ip) { struct ia64_script *script, *prev, *tmp; unw_hash_index_t index; unsigned short head; head = cache->lru_head; script = cache->buckets + head; cache->lru_head = script->lru_chain; /* re-insert script at the tail of the LRU chain: */ cache->buckets[cache->lru_tail].lru_chain = head; cache->lru_tail = head; /* remove the old script from the hash table (if it's there): */ if (script->ip) { index = hash (script->ip); tmp = cache->buckets + cache->hash[index]; prev = 0; while (1) { if (tmp == script) { if (prev) prev->coll_chain = tmp->coll_chain; else cache->hash[index] = tmp->coll_chain; break; } else prev = tmp; if (tmp->coll_chain >= IA64_UNW_CACHE_SIZE) /* old script wasn't in the hash-table */ break; tmp = cache->buckets + tmp->coll_chain; } } /* enter new script in the hash table */ index = hash (ip); script->coll_chain = cache->hash[index]; cache->hash[index] = script - cache->buckets; script_init (script, ip); return script; } static inline void script_finalize (struct ia64_script *script, struct cursor *c, struct ia64_state_record *sr) { script->pr_mask = sr->pr_mask; script->pr_val = sr->pr_val; script->pi = c->pi; } static inline void script_emit (struct ia64_script *script, struct ia64_script_insn insn) { if (script->count >= IA64_MAX_SCRIPT_LEN) { Dprintf ("%s: script exceeds maximum size of %u instructions!\n", __FUNCTION__, IA64_MAX_SCRIPT_LEN); return; } script->insn[script->count++] = insn; } static void compile_reg (struct ia64_state_record *sr, int i, struct ia64_reg_info *r, struct ia64_script *script) { enum ia64_script_insn_opcode opc; unsigned long val, rval; struct ia64_script_insn insn; long is_preserved_gr; if (r->where == IA64_WHERE_NONE || r->when >= sr->when_target) return; opc = IA64_INSN_MOVE; val = rval = r->val; is_preserved_gr = (i >= IA64_REG_R4 && i <= IA64_REG_R7); if (r->where == IA64_WHERE_GR) { /* Handle most common case first... */ if (rval >= 32) { /* register got spilled to a stacked register */ if (is_preserved_gr) opc = IA64_INSN_MOVE_STACKED_NAT; else opc = IA64_INSN_MOVE_STACKED; val = rval; } else if (rval >= 4 && rval <= 7) { /* register got spilled to a preserved register */ val = IA64_REG_R4 + (rval - 4); if (is_preserved_gr) opc = IA64_INSN_MOVE_NAT; } else { /* register got spilled to a scratch register */ if (is_preserved_gr) opc = IA64_INSN_MOVE_SCRATCH_NAT; else opc = IA64_INSN_MOVE_SCRATCH; val = UNW_IA64_GR + rval; } } else { switch (r->where) { case IA64_WHERE_FR: /* Note: There is no need to handle NaT-bit info here (indepent of is_preserved_gr), because for floating-point NaTs are represented as NaTVal, so the NaT-info never needs to be consulated. */ if (rval >= 2 && rval <= 5) val = IA64_REG_F2 + (rval - 2); else if (rval >= 16 && rval <= 31) val = IA64_REG_F16 + (rval - 16); else { opc = IA64_INSN_MOVE_SCRATCH; val = UNW_IA64_FR + rval; } break; case IA64_WHERE_BR: if (rval >= 1 && rval <= 5) { val = IA64_REG_B1 + (rval - 1); if (is_preserved_gr) opc = IA64_INSN_MOVE_NO_NAT; } else { opc = IA64_INSN_MOVE_SCRATCH; if (is_preserved_gr) opc = IA64_INSN_MOVE_SCRATCH_NO_NAT; val = UNW_IA64_BR + rval; } break; case IA64_WHERE_SPREL: if (is_preserved_gr) opc = IA64_INSN_ADD_SP_NAT; else { opc = IA64_INSN_ADD_SP; if (i >= IA64_REG_F2 && i <= IA64_REG_F31) val |= IA64_LOC_TYPE_FP; } break; case IA64_WHERE_PSPREL: if (is_preserved_gr) opc = IA64_INSN_ADD_PSP_NAT; else { opc = IA64_INSN_ADD_PSP; if (i >= IA64_REG_F2 && i <= IA64_REG_F31) val |= IA64_LOC_TYPE_FP; } break; default: Dprintf ("%s: register %u has unexpected `where' value of %u\n", __FUNCTION__, i, r->where); break; } } insn.opc = opc; insn.dst = i; insn.val = val; script_emit (script, insn); if (i == IA64_REG_PSP) { /* c->psp must contain the _value_ of the previous sp, not it's save-location. We get this by dereferencing the value we just stored in loc[IA64_REG_PSP]: */ insn.opc = IA64_INSN_LOAD_PSP; script_emit (script, insn); } } /* Sort the registers which got saved in decreasing order of WHEN value. This is needed to ensure that the save-locations are updated in the proper order. For example, suppose r4 gets spilled to memory and then r5 gets saved in r4. In this case, we need to update the save location of r5 before the one of r4. */ static inline int sort_regs (struct ia64_state_record *sr, int regorder[]) { int r, i, j, max, max_reg, max_when, num_regs = 0; assert (IA64_REG_BSP == 3); for (r = IA64_REG_BSP; r < IA64_NUM_PREGS; ++r) { if (sr->curr.reg[r].where == IA64_WHERE_NONE || sr->curr.reg[r].when >= sr->when_target) continue; regorder[num_regs++] = r; } /* Simple insertion-sort. Involves about N^2/2 comparisons and N exchanges. N is often small (say, 2-5) so a fancier sorting algorithm may not be worthwhile. */ for (i = max = 0; i < num_regs - 1; ++i) { max_reg = regorder[max]; max_when = sr->curr.reg[max_reg].when; for (j = i + 1; j < num_regs; ++j) if (sr->curr.reg[regorder[j]].when > max_when) { max = j; max_reg = regorder[j]; max_when = sr->curr.reg[max_reg].when; } if (i != max) { regorder[max] = regorder[i]; regorder[i] = max_reg; } } return num_regs; } /* Build an unwind script that unwinds from state OLD_STATE to the entrypoint of the function that called OLD_STATE. */ static inline int build_script (struct cursor *c, struct ia64_script *script) { int num_regs, i, ret, regorder[IA64_NUM_PREGS - 3]; struct ia64_reg_info *pri_unat; struct ia64_state_record sr; struct ia64_script_insn insn; ret = ia64_create_state_record (c, &sr); if (ret < 0) return ret; /* First, compile the update for IA64_REG_PSP. This is important because later save-locations may depend on it's correct (updated) value. Fixed-size frames are handled specially and variable-size frames get handled via the normal compile_reg(). */ if (sr.when_target > sr.curr.reg[IA64_REG_PSP].when && (sr.curr.reg[IA64_REG_PSP].where == IA64_WHERE_NONE) && sr.curr.reg[IA64_REG_PSP].val != 0) { /* new psp is psp plus frame size */ insn.opc = IA64_INSN_INC_PSP; insn.val = sr.curr.reg[IA64_REG_PSP].val; /* frame size */ script_emit (script, insn); } else compile_reg (&sr, IA64_REG_PSP, sr.curr.reg + IA64_REG_PSP, script); /* Second, compile the update for the primary UNaT, if any: */ if (sr.when_target >= sr.curr.reg[IA64_REG_PRI_UNAT_GR].when || sr.when_target >= sr.curr.reg[IA64_REG_PRI_UNAT_MEM].when) { if (sr.when_target < sr.curr.reg[IA64_REG_PRI_UNAT_GR].when) /* (primary) NaT bits were saved to memory only */ pri_unat = sr.curr.reg + IA64_REG_PRI_UNAT_MEM; else if (sr.when_target < sr.curr.reg[IA64_REG_PRI_UNAT_MEM].when) /* (primary) NaT bits were saved to a register only */ pri_unat = sr.curr.reg + IA64_REG_PRI_UNAT_GR; else if (sr.curr.reg[IA64_REG_PRI_UNAT_MEM].when > sr.curr.reg[IA64_REG_PRI_UNAT_GR].when) /* (primary) NaT bits were last saved to memory */ pri_unat = sr.curr.reg + IA64_REG_PRI_UNAT_MEM; else /* (primary) NaT bits were last saved to a register */ pri_unat = sr.curr.reg + IA64_REG_PRI_UNAT_GR; /* Note: we always store the final primary-UNaT location in UNAT_MEM. */ compile_reg (&sr, IA64_REG_PRI_UNAT_MEM, pri_unat, script); } /* Third, compile the other register in decreasing order of WHEN values. */ num_regs = sort_regs (&sr, regorder); for (i = 0; i < num_regs; ++i) compile_reg (&sr, regorder[i], sr.curr.reg + regorder[i], script); script->abi_marker = sr.abi_marker; script_finalize (script, c, &sr); ia64_free_state_record (&sr); return 0; } static inline void set_nat_info (struct cursor *c, unsigned long dst, ia64_loc_t nat_loc, uint8_t bitnr) { assert (dst >= IA64_REG_R4 && dst <= IA64_REG_R7); c->loc[dst - IA64_REG_R4 + IA64_REG_NAT4] = nat_loc; c->nat_bitnr[dst - IA64_REG_R4] = bitnr; } /* Apply the unwinding actions represented by OPS and update SR to reflect the state that existed upon entry to the function that this unwinder represents. */ static inline int run_script (struct ia64_script *script, struct cursor *c) { struct ia64_script_insn *ip, *limit, next_insn; ia64_loc_t loc, nat_loc; unsigned long opc, dst; uint8_t nat_bitnr; unw_word_t val; int ret; c->pi = script->pi; ip = script->insn; limit = script->insn + script->count; next_insn = *ip; c->abi_marker = script->abi_marker; while (ip++ < limit) { opc = next_insn.opc; dst = next_insn.dst; val = next_insn.val; next_insn = *ip; /* This is by far the most common operation: */ if (likely (opc == IA64_INSN_MOVE_STACKED)) { if ((ret = ia64_get_stacked (c, val, &loc, NULL)) < 0) return ret; } else switch (opc) { case IA64_INSN_INC_PSP: c->psp += val; continue; case IA64_INSN_LOAD_PSP: if ((ret = ia64_get (c, c->loc[IA64_REG_PSP], &c->psp)) < 0) return ret; continue; case IA64_INSN_ADD_PSP: loc = IA64_LOC_ADDR (c->psp + val, (val & IA64_LOC_TYPE_FP)); break; case IA64_INSN_ADD_SP: loc = IA64_LOC_ADDR (c->sp + val, (val & IA64_LOC_TYPE_FP)); break; case IA64_INSN_MOVE_NO_NAT: set_nat_info (c, dst, IA64_NULL_LOC, 0); case IA64_INSN_MOVE: loc = c->loc[val]; break; case IA64_INSN_MOVE_SCRATCH_NO_NAT: set_nat_info (c, dst, IA64_NULL_LOC, 0); case IA64_INSN_MOVE_SCRATCH: loc = ia64_scratch_loc (c, val, NULL); break; case IA64_INSN_ADD_PSP_NAT: loc = IA64_LOC_ADDR (c->psp + val, 0); assert (!IA64_IS_REG_LOC (loc)); set_nat_info (c, dst, c->loc[IA64_REG_PRI_UNAT_MEM], ia64_unat_slot_num (IA64_GET_ADDR (loc))); break; case IA64_INSN_ADD_SP_NAT: loc = IA64_LOC_ADDR (c->sp + val, 0); assert (!IA64_IS_REG_LOC (loc)); set_nat_info (c, dst, c->loc[IA64_REG_PRI_UNAT_MEM], ia64_unat_slot_num (IA64_GET_ADDR (loc))); break; case IA64_INSN_MOVE_NAT: loc = c->loc[val]; set_nat_info (c, dst, c->loc[val - IA64_REG_R4 + IA64_REG_NAT4], c->nat_bitnr[val - IA64_REG_R4]); break; case IA64_INSN_MOVE_STACKED_NAT: if ((ret = ia64_get_stacked (c, val, &loc, &nat_loc)) < 0) return ret; assert (!IA64_IS_REG_LOC (loc)); set_nat_info (c, dst, nat_loc, rse_slot_num (IA64_GET_ADDR (loc))); break; case IA64_INSN_MOVE_SCRATCH_NAT: loc = ia64_scratch_loc (c, val, NULL); nat_loc = ia64_scratch_loc (c, val + (UNW_IA64_NAT - UNW_IA64_GR), &nat_bitnr); set_nat_info (c, dst, nat_loc, nat_bitnr); break; } c->loc[dst] = loc; } return 0; } static int uncached_find_save_locs (struct cursor *c) { struct ia64_script script; int ret = 0; if ((ret = ia64_fetch_proc_info (c, c->ip, 1)) < 0) return ret; script_init (&script, c->ip); if ((ret = build_script (c, &script)) < 0) { if (ret != -UNW_ESTOPUNWIND) Dprintf ("%s: failed to build unwind script for ip %lx\n", __FUNCTION__, (long) c->ip); return ret; } return run_script (&script, c); } HIDDEN int ia64_find_save_locs (struct cursor *c) { struct ia64_script_cache *cache = NULL; struct ia64_script *script = NULL; intrmask_t saved_mask; int ret = 0; if (c->as->caching_policy == UNW_CACHE_NONE) return uncached_find_save_locs (c); cache = get_script_cache (c->as, &saved_mask); if (!cache) { Debug (1, "contention on script-cache; doing uncached lookup\n"); return uncached_find_save_locs (c); } { script = script_lookup (cache, c); Debug (8, "ip %lx %s in script cache\n", (long) c->ip, script ? "hit" : "missed"); if (!script || (script->count == 0 && !script->pi.unwind_info)) { if ((ret = ia64_fetch_proc_info (c, c->ip, 1)) < 0) goto out; } if (!script) { script = script_new (cache, c->ip); if (!script) { Dprintf ("%s: failed to create unwind script\n", __FUNCTION__); ret = -UNW_EUNSPEC; goto out; } } cache->buckets[c->prev_script].hint = script - cache->buckets; if (script->count == 0) ret = build_script (c, script); assert (script->count > 0); c->hint = script->hint; c->prev_script = script - cache->buckets; if (ret < 0) { if (ret != -UNW_ESTOPUNWIND) Dprintf ("%s: failed to locate/build unwind script for ip %lx\n", __FUNCTION__, (long) c->ip); goto out; } ret = run_script (script, c); } out: put_script_cache (c->as, cache, &saved_mask); return ret; } HIDDEN void ia64_validate_cache (unw_addr_space_t as, void *arg) { #ifndef UNW_REMOTE_ONLY if (as == unw_local_addr_space && ia64_local_validate_cache (as, arg) == 1) return; #endif #ifndef UNW_LOCAL_ONLY /* local info is up-to-date, check dynamic info. */ unwi_dyn_validate_cache (as, arg); #endif } HIDDEN int ia64_cache_proc_info (struct cursor *c) { struct ia64_script_cache *cache; struct ia64_script *script; intrmask_t saved_mask; int ret = 0; cache = get_script_cache (c->as, &saved_mask); if (!cache) return ret; /* cache is busy */ /* Re-check to see if a cache entry has been added in the meantime: */ script = script_lookup (cache, c); if (script) goto out; script = script_new (cache, c->ip); if (!script) { Dprintf ("%s: failed to create unwind script\n", __FUNCTION__); ret = -UNW_EUNSPEC; goto out; } script->pi = c->pi; out: put_script_cache (c->as, cache, &saved_mask); return ret; } HIDDEN int ia64_get_cached_proc_info (struct cursor *c) { struct ia64_script_cache *cache; struct ia64_script *script; intrmask_t saved_mask; cache = get_script_cache (c->as, &saved_mask); if (!cache) return -UNW_ENOINFO; /* cache is busy */ { script = script_lookup (cache, c); if (script) c->pi = script->pi; } put_script_cache (c->as, cache, &saved_mask); return script ? 0 : -UNW_ENOINFO; }