From: Zecheng Li Date: Mon, 13 Oct 2025 18:16:00 +0000 (+0000) Subject: perf annotate: Track arithmetic instructions on pointers X-Git-Tag: v6.19-rc1~61^2~166 X-Git-Url: http://git.ipfire.org/cgi-bin/gitweb.cgi?a=commitdiff_plain;h=1f4cc4ae3f8ae661bae3722e973a0f5be650fcbf;p=thirdparty%2Fkernel%2Flinux.git perf annotate: Track arithmetic instructions on pointers Track the arithmetic operations on registers with pointer types. We handle only add, sub and lea instructions. The original pointer information needs to be preserved for getting outermost struct types. For example, reg0 points to a struct cfs_rq, when we add 0x10 to reg0, it should preserve the information of struct cfs_rq + 0x10 in the register instead of a pointer type to the child field at 0x10. Details: 1. struct type_state_reg now includes an offset, indicating if the register points to the start or an internal part of its associated type. This offset is used in mem to reg and reg to stack mem transfers, and also applied to the final type offset. 2. lea offset(%sp/%fp), reg is now treated as taking the address of a stack variable. It worked fine in most cases, but an issue with this approach is the pointer type may not exist. 3. lea offset(%base), reg is handled by moving the type from %base and adding an offset, similar to an add operation followed by a mov reg to reg. 4. Non-stack variables from DWARF with non-zero offsets in their location expressions are now accepted with register offset tracking. Multi-register addressing modes in LEA are not supported. Signed-off-by: Zecheng Li Signed-off-by: Namhyung Kim --- diff --git a/tools/perf/arch/x86/annotate/instructions.c b/tools/perf/arch/x86/annotate/instructions.c index 746443e040896..def7729a394c9 100644 --- a/tools/perf/arch/x86/annotate/instructions.c +++ b/tools/perf/arch/x86/annotate/instructions.c @@ -248,6 +248,7 @@ static void update_insn_state_x86(struct type_state *state, tsr = &state->regs[state->ret_reg]; tsr->type = type_die; tsr->kind = TSR_KIND_TYPE; + tsr->offset = 0; tsr->ok = true; pr_debug_dtp("call [%x] return -> reg%d", @@ -284,6 +285,7 @@ static void update_insn_state_x86(struct type_state *state, !strcmp(var_name, "this_cpu_off") && tsr->kind == TSR_KIND_CONST) { tsr->kind = TSR_KIND_PERCPU_BASE; + tsr->offset = 0; tsr->ok = true; imm_value = tsr->imm_value; } @@ -291,6 +293,19 @@ static void update_insn_state_x86(struct type_state *state, else return; + /* Ignore add to non-pointer or non-const types */ + if (tsr->kind == TSR_KIND_POINTER || + (dwarf_tag(&tsr->type) == DW_TAG_pointer_type && + src->reg1 != DWARF_REG_PC && tsr->kind == TSR_KIND_TYPE && !dst->mem_ref)) { + tsr->offset += imm_value; + pr_debug_dtp("add [%x] offset %#"PRIx64" to reg%d", + insn_offset, imm_value, dst->reg1); + pr_debug_type_name(&tsr->type, tsr->kind); + } + + if (tsr->kind == TSR_KIND_CONST) + tsr->imm_value += imm_value; + if (tsr->kind != TSR_KIND_PERCPU_BASE) return; @@ -302,6 +317,7 @@ static void update_insn_state_x86(struct type_state *state, */ tsr->type = type_die; tsr->kind = TSR_KIND_PERCPU_POINTER; + tsr->offset = 0; tsr->ok = true; pr_debug_dtp("add [%x] percpu %#"PRIx64" -> reg%d", @@ -311,6 +327,106 @@ static void update_insn_state_x86(struct type_state *state, return; } + if (!strncmp(dl->ins.name, "sub", 3)) { + u64 imm_value = -1ULL; + + if (!has_reg_type(state, dst->reg1)) + return; + + tsr = &state->regs[dst->reg1]; + tsr->copied_from = -1; + + if (src->imm) + imm_value = src->offset; + else if (has_reg_type(state, src->reg1) && + state->regs[src->reg1].kind == TSR_KIND_CONST) + imm_value = state->regs[src->reg1].imm_value; + + if (tsr->kind == TSR_KIND_POINTER || + (dwarf_tag(&tsr->type) == DW_TAG_pointer_type && + src->reg1 != DWARF_REG_PC && tsr->kind == TSR_KIND_TYPE && !dst->mem_ref)) { + tsr->offset -= imm_value; + pr_debug_dtp("sub [%x] offset %#"PRIx64" to reg%d", + insn_offset, imm_value, dst->reg1); + pr_debug_type_name(&tsr->type, tsr->kind); + } + + if (tsr->kind == TSR_KIND_CONST) + tsr->imm_value -= imm_value; + + return; + } + + if (!strncmp(dl->ins.name, "lea", 3)) { + int sreg = src->reg1; + struct type_state_reg src_tsr; + + if (!has_reg_type(state, sreg) || + !has_reg_type(state, dst->reg1) || + !src->mem_ref) + return; + + src_tsr = state->regs[sreg]; + tsr = &state->regs[dst->reg1]; + + tsr->copied_from = -1; + tsr->ok = false; + + /* Case 1: Based on stack pointer or frame pointer */ + if (sreg == fbreg || sreg == state->stack_reg) { + struct type_state_stack *stack; + int offset = src->offset - fboff; + + stack = find_stack_state(state, offset); + if (!stack) + return; + + tsr->type = stack->type; + tsr->kind = TSR_KIND_POINTER; + tsr->offset = offset - stack->offset; + tsr->ok = true; + + if (sreg == fbreg) { + pr_debug_dtp("lea [%x] address of -%#x(stack) -> reg%d", + insn_offset, -src->offset, dst->reg1); + } else { + pr_debug_dtp("lea [%x] address of %#x(reg%d) -> reg%d", + insn_offset, src->offset, sreg, dst->reg1); + } + + pr_debug_type_name(&tsr->type, tsr->kind); + } + /* Case 2: Based on a register holding a typed pointer */ + else if (src_tsr.ok && (src_tsr.kind == TSR_KIND_POINTER || + (dwarf_tag(&src_tsr.type) == DW_TAG_pointer_type && + src_tsr.kind == TSR_KIND_TYPE))) { + + if (src_tsr.kind == TSR_KIND_TYPE && + __die_get_real_type(&state->regs[sreg].type, &type_die) == NULL) + return; + + if (src_tsr.kind == TSR_KIND_POINTER) + type_die = state->regs[sreg].type; + + /* Check if the target type has a member at the new offset */ + if (die_get_member_type(&type_die, + src->offset + src_tsr.offset, &type_die) == NULL) + return; + + tsr->type = src_tsr.type; + tsr->kind = src_tsr.kind; + tsr->offset = src->offset + src_tsr.offset; + tsr->ok = true; + + pr_debug_dtp("lea [%x] address of %s%#x(reg%d) -> reg%d", + insn_offset, src->offset < 0 ? "-" : "", + abs(src->offset), sreg, dst->reg1); + + pr_debug_type_name(&tsr->type, tsr->kind); + } + return; + } + if (strncmp(dl->ins.name, "mov", 3)) return; @@ -345,6 +461,7 @@ static void update_insn_state_x86(struct type_state *state, if (var_addr == 40) { tsr->kind = TSR_KIND_CANARY; + tsr->offset = 0; tsr->ok = true; pr_debug_dtp("mov [%x] stack canary -> reg%d\n", @@ -361,6 +478,7 @@ static void update_insn_state_x86(struct type_state *state, tsr->type = type_die; tsr->kind = TSR_KIND_TYPE; + tsr->offset = 0; tsr->ok = true; pr_debug_dtp("mov [%x] this-cpu addr=%#"PRIx64" -> reg%d", @@ -372,6 +490,7 @@ static void update_insn_state_x86(struct type_state *state, if (src->imm) { tsr->kind = TSR_KIND_CONST; tsr->imm_value = src->offset; + tsr->offset = 0; tsr->ok = true; pr_debug_dtp("mov [%x] imm=%#x -> reg%d\n", @@ -388,6 +507,7 @@ static void update_insn_state_x86(struct type_state *state, tsr->type = state->regs[src->reg1].type; tsr->kind = state->regs[src->reg1].kind; tsr->imm_value = state->regs[src->reg1].imm_value; + tsr->offset = state->regs[src->reg1].offset; tsr->ok = true; /* To copy back the variable type later (hopefully) */ @@ -421,12 +541,14 @@ retry: } else if (!stack->compound) { tsr->type = stack->type; tsr->kind = stack->kind; + tsr->offset = 0; tsr->ok = true; } else if (die_get_member_type(&stack->type, offset - stack->offset, &type_die)) { tsr->type = type_die; tsr->kind = TSR_KIND_TYPE; + tsr->offset = 0; tsr->ok = true; } else { tsr->ok = false; @@ -446,9 +568,10 @@ retry: else if (has_reg_type(state, sreg) && state->regs[sreg].ok && state->regs[sreg].kind == TSR_KIND_TYPE && die_deref_ptr_type(&state->regs[sreg].type, - src->offset, &type_die)) { + src->offset + state->regs[sreg].offset, &type_die)) { tsr->type = type_die; tsr->kind = TSR_KIND_TYPE; + tsr->offset = 0; tsr->ok = true; pr_debug_dtp("mov [%x] %#x(reg%d) -> reg%d", @@ -459,9 +582,10 @@ retry: else if (has_reg_type(state, sreg) && state->regs[sreg].ok && state->regs[sreg].kind == TSR_KIND_POINTER && die_get_member_type(&state->regs[sreg].type, - src->offset, &type_die)) { + src->offset + state->regs[sreg].offset, &type_die)) { tsr->type = state->regs[sreg].type; tsr->kind = TSR_KIND_TYPE; + tsr->offset = src->offset + state->regs[sreg].offset; tsr->ok = true; pr_debug_dtp("mov [%x] addr %#x(reg%d) -> reg%d", @@ -486,6 +610,7 @@ retry: tsr->type = type_die; tsr->kind = TSR_KIND_TYPE; + tsr->offset = 0; tsr->ok = true; pr_debug_dtp("mov [%x] global addr=%"PRIx64" -> reg%d", @@ -517,6 +642,7 @@ retry: die_get_member_type(&type_die, offset, &type_die)) { tsr->type = type_die; tsr->kind = TSR_KIND_TYPE; + tsr->offset = 0; tsr->ok = true; if (src->multi_regs) { @@ -539,6 +665,7 @@ retry: src->offset, &type_die)) { tsr->type = type_die; tsr->kind = TSR_KIND_TYPE; + tsr->offset = 0; tsr->ok = true; pr_debug_dtp("mov [%x] pointer %#x(reg%d) -> reg%d", @@ -561,6 +688,7 @@ retry: &var_name, &offset) && !strcmp(var_name, "__per_cpu_offset")) { tsr->kind = TSR_KIND_PERCPU_BASE; + tsr->offset = 0; tsr->ok = true; pr_debug_dtp("mov [%x] percpu base reg%d\n", @@ -609,6 +737,11 @@ retry: pr_debug_dtp("mov [%x] reg%d -> %#x(reg%d)", insn_offset, src->reg1, offset, dst->reg1); } + if (tsr->offset != 0) { + pr_debug_dtp(" reg%d offset %#x ->", + src->reg1, tsr->offset); + } + pr_debug_type_name(&tsr->type, tsr->kind); } /* diff --git a/tools/perf/util/annotate-data.c b/tools/perf/util/annotate-data.c index 48f3bf20070f0..4204a7956ee5a 100644 --- a/tools/perf/util/annotate-data.c +++ b/tools/perf/util/annotate-data.c @@ -905,7 +905,7 @@ static void update_var_state(struct type_state *state, struct data_loc_info *dlo insn_offset, -offset); } pr_debug_type_name(&mem_die, TSR_KIND_TYPE); - } else if (has_reg_type(state, var->reg) && var->offset == 0) { + } else if (has_reg_type(state, var->reg)) { struct type_state_reg *reg; Dwarf_Die orig_type; @@ -921,6 +921,7 @@ static void update_var_state(struct type_state *state, struct data_loc_info *dlo !is_better_type(®->type, &mem_die)) continue; + reg->offset = -var->offset; reg->type = mem_die; reg->kind = TSR_KIND_POINTER; reg->ok = true; @@ -932,13 +933,17 @@ static void update_var_state(struct type_state *state, struct data_loc_info *dlo } orig_type = reg->type; - + /* + * var->offset + reg value is the beginning of the struct + * reg->offset is the offset the reg points + */ + reg->offset = -var->offset; reg->type = mem_die; reg->kind = TSR_KIND_TYPE; reg->ok = true; - pr_debug_dtp("var [%"PRIx64"] reg%d", - insn_offset, var->reg); + pr_debug_dtp("var [%"PRIx64"] reg%d offset %x", + insn_offset, var->reg, var->offset); pr_debug_type_name(&mem_die, TSR_KIND_TYPE); /* @@ -1126,7 +1131,7 @@ again: if (__die_get_real_type(&state->regs[reg].type, type_die) == NULL) return PERF_TMR_NO_POINTER; - dloc->type_offset = dloc->op->offset; + dloc->type_offset = dloc->op->offset + state->regs[reg].offset; if (dwarf_tag(type_die) == DW_TAG_typedef) die_get_real_type(type_die, &sized_type); @@ -1155,7 +1160,7 @@ again: */ *type_die = state->regs[reg].type; - dloc->type_offset = dloc->op->offset; + dloc->type_offset = dloc->op->offset + state->regs[reg].offset; /* Get the size of the actual type */ if (dwarf_aggregate_size(type_die, &size) < 0 || diff --git a/tools/perf/util/annotate-data.h b/tools/perf/util/annotate-data.h index 0267834420569..14655b76db65a 100644 --- a/tools/perf/util/annotate-data.h +++ b/tools/perf/util/annotate-data.h @@ -174,6 +174,12 @@ extern struct annotated_data_stat ann_data_stat; struct type_state_reg { Dwarf_Die type; u32 imm_value; + /* + * The offset within the struct that the register points to. + * A value of 0 means the register points to the beginning. + * type_offset = op->offset + reg->offset + */ + s32 offset; bool ok; bool caller_saved; u8 kind;