From cc8b0b92a1699bc32f7fec71daa2bfc90de43a4d Mon Sep 17 00:00:00 2001 From: Alexei Starovoitov Date: Thu, 14 Dec 2017 17:55:05 -0800 Subject: bpf: introduce function calls (function boundaries) Allow arbitrary function calls from bpf function to another bpf function. Since the beginning of bpf all bpf programs were represented as a single function and program authors were forced to use always_inline for all functions in their C code. That was causing llvm to unnecessary inflate the code size and forcing developers to move code to header files with little code reuse. With a bit of additional complexity teach verifier to recognize arbitrary function calls from one bpf function to another as long as all of functions are presented to the verifier as a single bpf program. New program layout: r6 = r1 // some code .. r1 = .. // arg1 r2 = .. // arg2 call pc+1 // function call pc-relative exit .. = r1 // access arg1 .. = r2 // access arg2 .. call pc+20 // second level of function call ... It allows for better optimized code and finally allows to introduce the core bpf libraries that can be reused in different projects, since programs are no longer limited by single elf file. With function calls bpf can be compiled into multiple .o files. This patch is the first step. It detects programs that contain multiple functions and checks that calls between them are valid. It splits the sequence of bpf instructions (one program) into a set of bpf functions that call each other. Calls to only known functions are allowed. In the future the verifier may allow calls to unresolved functions and will do dynamic linking. This logic supports statically linked bpf functions only. Such function boundary detection could have been done as part of control flow graph building in check_cfg(), but it's cleaner to separate function boundary detection vs control flow checks within a subprogram (function) into logically indepedent steps. Follow up patches may split check_cfg() further, but not check_subprogs(). Only allow bpf-to-bpf calls for root only and for non-hw-offloaded programs. These restrictions can be relaxed in the future. Signed-off-by: Alexei Starovoitov Acked-by: Daniel Borkmann Signed-off-by: Daniel Borkmann --- kernel/bpf/disasm.c | 8 ++- kernel/bpf/verifier.c | 141 +++++++++++++++++++++++++++++++++++++++++++++++++- 2 files changed, 145 insertions(+), 4 deletions(-) (limited to 'kernel') diff --git a/kernel/bpf/disasm.c b/kernel/bpf/disasm.c index e682850c9715..883f88fa5bfc 100644 --- a/kernel/bpf/disasm.c +++ b/kernel/bpf/disasm.c @@ -189,8 +189,12 @@ void print_bpf_insn(bpf_insn_print_cb verbose, struct bpf_verifier_env *env, u8 opcode = BPF_OP(insn->code); if (opcode == BPF_CALL) { - verbose(env, "(%02x) call %s#%d\n", insn->code, - func_id_name(insn->imm), insn->imm); + if (insn->src_reg == BPF_PSEUDO_CALL) + verbose(env, "(%02x) call pc%+d\n", insn->code, + insn->imm); + else + verbose(env, "(%02x) call %s#%d\n", insn->code, + func_id_name(insn->imm), insn->imm); } else if (insn->code == (BPF_JMP | BPF_JA)) { verbose(env, "(%02x) goto pc%+d\n", insn->code, insn->off); diff --git a/kernel/bpf/verifier.c b/kernel/bpf/verifier.c index e807bda7fe29..1d0f7ff0b9a9 100644 --- a/kernel/bpf/verifier.c +++ b/kernel/bpf/verifier.c @@ -20,6 +20,8 @@ #include #include #include +#include +#include #include "disasm.h" @@ -636,6 +638,113 @@ enum reg_arg_type { DST_OP_NO_MARK /* same as above, check only, don't mark */ }; +static int cmp_subprogs(const void *a, const void *b) +{ + return *(int *)a - *(int *)b; +} + +static int find_subprog(struct bpf_verifier_env *env, int off) +{ + u32 *p; + + p = bsearch(&off, env->subprog_starts, env->subprog_cnt, + sizeof(env->subprog_starts[0]), cmp_subprogs); + if (!p) + return -ENOENT; + return p - env->subprog_starts; + +} + +static int add_subprog(struct bpf_verifier_env *env, int off) +{ + int insn_cnt = env->prog->len; + int ret; + + if (off >= insn_cnt || off < 0) { + verbose(env, "call to invalid destination\n"); + return -EINVAL; + } + ret = find_subprog(env, off); + if (ret >= 0) + return 0; + if (env->subprog_cnt >= BPF_MAX_SUBPROGS) { + verbose(env, "too many subprograms\n"); + return -E2BIG; + } + env->subprog_starts[env->subprog_cnt++] = off; + sort(env->subprog_starts, env->subprog_cnt, + sizeof(env->subprog_starts[0]), cmp_subprogs, NULL); + return 0; +} + +static int check_subprogs(struct bpf_verifier_env *env) +{ + int i, ret, subprog_start, subprog_end, off, cur_subprog = 0; + struct bpf_insn *insn = env->prog->insnsi; + int insn_cnt = env->prog->len; + + /* determine subprog starts. The end is one before the next starts */ + for (i = 0; i < insn_cnt; i++) { + if (insn[i].code != (BPF_JMP | BPF_CALL)) + continue; + if (insn[i].src_reg != BPF_PSEUDO_CALL) + continue; + if (!env->allow_ptr_leaks) { + verbose(env, "function calls to other bpf functions are allowed for root only\n"); + return -EPERM; + } + if (bpf_prog_is_dev_bound(env->prog->aux)) { + verbose(env, "funcation calls in offloaded programs are not supported yet\n"); + return -EINVAL; + } + ret = add_subprog(env, i + insn[i].imm + 1); + if (ret < 0) + return ret; + } + + if (env->log.level > 1) + for (i = 0; i < env->subprog_cnt; i++) + verbose(env, "func#%d @%d\n", i, env->subprog_starts[i]); + + /* now check that all jumps are within the same subprog */ + subprog_start = 0; + if (env->subprog_cnt == cur_subprog) + subprog_end = insn_cnt; + else + subprog_end = env->subprog_starts[cur_subprog++]; + for (i = 0; i < insn_cnt; i++) { + u8 code = insn[i].code; + + if (BPF_CLASS(code) != BPF_JMP) + goto next; + if (BPF_OP(code) == BPF_EXIT || BPF_OP(code) == BPF_CALL) + goto next; + off = i + insn[i].off + 1; + if (off < subprog_start || off >= subprog_end) { + verbose(env, "jump out of range from insn %d to %d\n", i, off); + return -EINVAL; + } +next: + if (i == subprog_end - 1) { + /* to avoid fall-through from one subprog into another + * the last insn of the subprog should be either exit + * or unconditional jump back + */ + if (code != (BPF_JMP | BPF_EXIT) && + code != (BPF_JMP | BPF_JA)) { + verbose(env, "last insn is not an exit or jmp\n"); + return -EINVAL; + } + subprog_start = subprog_end; + if (env->subprog_cnt == cur_subprog) + subprog_end = insn_cnt; + else + subprog_end = env->subprog_starts[cur_subprog++]; + } + } + return 0; +} + static void mark_reg_read(const struct bpf_verifier_state *state, u32 regno) { struct bpf_verifier_state *parent = state->parent; @@ -3284,6 +3393,10 @@ static int check_cfg(struct bpf_verifier_env *env) int ret = 0; int i, t; + ret = check_subprogs(env); + if (ret < 0) + return ret; + insn_state = kcalloc(insn_cnt, sizeof(int), GFP_KERNEL); if (!insn_state) return -ENOMEM; @@ -3316,6 +3429,14 @@ peek_stack: goto err_free; if (t + 1 < insn_cnt) env->explored_states[t + 1] = STATE_LIST_MARK; + if (insns[t].src_reg == BPF_PSEUDO_CALL) { + env->explored_states[t] = STATE_LIST_MARK; + ret = push_insn(t, t + insns[t].imm + 1, BRANCH, env); + if (ret == 1) + goto peek_stack; + else if (ret < 0) + goto err_free; + } } else if (opcode == BPF_JA) { if (BPF_SRC(insns[t].code) != BPF_K) { ret = -EINVAL; @@ -4245,6 +4366,19 @@ static int adjust_insn_aux_data(struct bpf_verifier_env *env, u32 prog_len, return 0; } +static void adjust_subprog_starts(struct bpf_verifier_env *env, u32 off, u32 len) +{ + int i; + + if (len == 1) + return; + for (i = 0; i < env->subprog_cnt; i++) { + if (env->subprog_starts[i] < off) + continue; + env->subprog_starts[i] += len - 1; + } +} + static struct bpf_prog *bpf_patch_insn_data(struct bpf_verifier_env *env, u32 off, const struct bpf_insn *patch, u32 len) { @@ -4255,6 +4389,7 @@ static struct bpf_prog *bpf_patch_insn_data(struct bpf_verifier_env *env, u32 of return NULL; if (adjust_insn_aux_data(env, new_prog->len, off, len)) return NULL; + adjust_subprog_starts(env, off, len); return new_prog; } @@ -4408,6 +4543,8 @@ static int fixup_bpf_calls(struct bpf_verifier_env *env) for (i = 0; i < insn_cnt; i++, insn++) { if (insn->code != (BPF_JMP | BPF_CALL)) continue; + if (insn->src_reg == BPF_PSEUDO_CALL) + continue; if (insn->imm == BPF_FUNC_get_route_realm) prog->dst_needed = 1; @@ -4589,12 +4726,12 @@ int bpf_check(struct bpf_prog **prog, union bpf_attr *attr) if (!env->explored_states) goto skip_full_check; + env->allow_ptr_leaks = capable(CAP_SYS_ADMIN); + ret = check_cfg(env); if (ret < 0) goto skip_full_check; - env->allow_ptr_leaks = capable(CAP_SYS_ADMIN); - ret = do_check(env); if (env->cur_state) { free_verifier_state(env->cur_state, true); -- cgit v1.2.3-71-gd317 From f4d7e40a5b7157e1329c3c5b10f60d8289fc2941 Mon Sep 17 00:00:00 2001 From: Alexei Starovoitov Date: Thu, 14 Dec 2017 17:55:06 -0800 Subject: bpf: introduce function calls (verification) Allow arbitrary function calls from bpf function to another bpf function. To recognize such set of bpf functions the verifier does: 1. runs control flow analysis to detect function boundaries 2. proceeds with verification of all functions starting from main(root) function It recognizes that the stack of the caller can be accessed by the callee (if the caller passed a pointer to its stack to the callee) and the callee can store map_value and other pointers into the stack of the caller. 3. keeps track of the stack_depth of each function to make sure that total stack depth is still less than 512 bytes 4. disallows pointers to the callee stack to be stored into the caller stack, since they will be invalid as soon as the callee returns 5. to reuse all of the existing state_pruning logic each function call is considered to be independent call from the verifier point of view. The verifier pretends to inline all function calls it sees are being called. It stores the callsite instruction index as part of the state to make sure that two calls to the same callee from two different places in the caller will be different from state pruning point of view 6. more safety checks are added to liveness analysis Implementation details: . struct bpf_verifier_state is now consists of all stack frames that led to this function . struct bpf_func_state represent one stack frame. It consists of registers in the given frame and its stack . propagate_liveness() logic had a premature optimization where mark_reg_read() and mark_stack_slot_read() were manually inlined with loop iterating over parents for each register or stack slot. Undo this optimization to reuse more complex mark_*_read() logic . skip_callee() logic is not necessary from safety point of view, but without it mark_*_read() markings become too conservative, since after returning from the funciton call a read of r6-r9 will incorrectly propagate the read marks into callee causing inefficient pruning later . mark_*_read() logic is now aware of control flow which makes it more complex. In the future the plan is to rewrite liveness to be hierarchical. So that liveness can be done within basic block only and control flow will be responsible for propagation of liveness information along cfg and between calls. . tail_calls and ld_abs insns are not allowed in the programs with bpf-to-bpf calls . returning stack pointers to the caller or storing them into stack frame of the caller is not allowed Testing: . no difference in cilium processed_insn numbers . large number of tests follows in next patches Signed-off-by: Alexei Starovoitov Acked-by: John Fastabend Acked-by: Daniel Borkmann Signed-off-by: Daniel Borkmann --- include/linux/bpf_verifier.h | 36 ++- kernel/bpf/verifier.c | 702 +++++++++++++++++++++++++++++++++---------- 2 files changed, 585 insertions(+), 153 deletions(-) (limited to 'kernel') diff --git a/include/linux/bpf_verifier.h b/include/linux/bpf_verifier.h index 91a583bb3fa7..1f23408024ee 100644 --- a/include/linux/bpf_verifier.h +++ b/include/linux/bpf_verifier.h @@ -76,6 +76,14 @@ struct bpf_reg_state { s64 smax_value; /* maximum possible (s64)value */ u64 umin_value; /* minimum possible (u64)value */ u64 umax_value; /* maximum possible (u64)value */ + /* Inside the callee two registers can be both PTR_TO_STACK like + * R1=fp-8 and R2=fp-8, but one of them points to this function stack + * while another to the caller's stack. To differentiate them 'frameno' + * is used which is an index in bpf_verifier_state->frame[] array + * pointing to bpf_func_state. + * This field must be second to last, for states_equal() reasons. + */ + u32 frameno; /* This field must be last, for states_equal() reasons. */ enum bpf_reg_liveness live; }; @@ -96,13 +104,34 @@ struct bpf_stack_state { /* state of the program: * type of all registers and stack info */ -struct bpf_verifier_state { +struct bpf_func_state { struct bpf_reg_state regs[MAX_BPF_REG]; struct bpf_verifier_state *parent; + /* index of call instruction that called into this func */ + int callsite; + /* stack frame number of this function state from pov of + * enclosing bpf_verifier_state. + * 0 = main function, 1 = first callee. + */ + u32 frameno; + /* subprog number == index within subprog_stack_depth + * zero == main subprog + */ + u32 subprogno; + + /* should be second to last. See copy_func_state() */ int allocated_stack; struct bpf_stack_state *stack; }; +#define MAX_CALL_FRAMES 8 +struct bpf_verifier_state { + /* call stack tracking */ + struct bpf_func_state *frame[MAX_CALL_FRAMES]; + struct bpf_verifier_state *parent; + u32 curframe; +}; + /* linked list of verifier states used to prune search */ struct bpf_verifier_state_list { struct bpf_verifier_state state; @@ -163,12 +192,15 @@ struct bpf_verifier_env { struct bpf_insn_aux_data *insn_aux_data; /* array of per-insn state */ struct bpf_verifer_log log; u32 subprog_starts[BPF_MAX_SUBPROGS]; + u16 subprog_stack_depth[BPF_MAX_SUBPROGS + 1]; u32 subprog_cnt; }; static inline struct bpf_reg_state *cur_regs(struct bpf_verifier_env *env) { - return env->cur_state->regs; + struct bpf_verifier_state *cur = env->cur_state; + + return cur->frame[cur->curframe]->regs; } #if defined(CONFIG_NET) && defined(CONFIG_BPF_SYSCALL) diff --git a/kernel/bpf/verifier.c b/kernel/bpf/verifier.c index 1d0f7ff0b9a9..f6e09d84a96f 100644 --- a/kernel/bpf/verifier.c +++ b/kernel/bpf/verifier.c @@ -229,13 +229,23 @@ static void print_liveness(struct bpf_verifier_env *env, verbose(env, "w"); } +static struct bpf_func_state *func(struct bpf_verifier_env *env, + const struct bpf_reg_state *reg) +{ + struct bpf_verifier_state *cur = env->cur_state; + + return cur->frame[reg->frameno]; +} + static void print_verifier_state(struct bpf_verifier_env *env, - struct bpf_verifier_state *state) + const struct bpf_func_state *state) { - struct bpf_reg_state *reg; + const struct bpf_reg_state *reg; enum bpf_reg_type t; int i; + if (state->frameno) + verbose(env, " frame%d:", state->frameno); for (i = 0; i < MAX_BPF_REG; i++) { reg = &state->regs[i]; t = reg->type; @@ -248,6 +258,8 @@ static void print_verifier_state(struct bpf_verifier_env *env, tnum_is_const(reg->var_off)) { /* reg->off should be 0 for SCALAR_VALUE */ verbose(env, "%lld", reg->var_off.value + reg->off); + if (t == PTR_TO_STACK) + verbose(env, ",call_%d", func(env, reg)->callsite); } else { verbose(env, "(id=%d", reg->id); if (t != SCALAR_VALUE) @@ -303,8 +315,8 @@ static void print_verifier_state(struct bpf_verifier_env *env, verbose(env, "\n"); } -static int copy_stack_state(struct bpf_verifier_state *dst, - const struct bpf_verifier_state *src) +static int copy_stack_state(struct bpf_func_state *dst, + const struct bpf_func_state *src) { if (!src->stack) return 0; @@ -320,13 +332,13 @@ static int copy_stack_state(struct bpf_verifier_state *dst, /* do_check() starts with zero-sized stack in struct bpf_verifier_state to * make it consume minimal amount of memory. check_stack_write() access from - * the program calls into realloc_verifier_state() to grow the stack size. + * the program calls into realloc_func_state() to grow the stack size. * Note there is a non-zero 'parent' pointer inside bpf_verifier_state * which this function copies over. It points to previous bpf_verifier_state * which is never reallocated */ -static int realloc_verifier_state(struct bpf_verifier_state *state, int size, - bool copy_old) +static int realloc_func_state(struct bpf_func_state *state, int size, + bool copy_old) { u32 old_size = state->allocated_stack; struct bpf_stack_state *new_stack; @@ -359,10 +371,21 @@ static int realloc_verifier_state(struct bpf_verifier_state *state, int size, return 0; } +static void free_func_state(struct bpf_func_state *state) +{ + kfree(state->stack); + kfree(state); +} + static void free_verifier_state(struct bpf_verifier_state *state, bool free_self) { - kfree(state->stack); + int i; + + for (i = 0; i <= state->curframe; i++) { + free_func_state(state->frame[i]); + state->frame[i] = NULL; + } if (free_self) kfree(state); } @@ -370,18 +393,46 @@ static void free_verifier_state(struct bpf_verifier_state *state, /* copy verifier state from src to dst growing dst stack space * when necessary to accommodate larger src stack */ -static int copy_verifier_state(struct bpf_verifier_state *dst, - const struct bpf_verifier_state *src) +static int copy_func_state(struct bpf_func_state *dst, + const struct bpf_func_state *src) { int err; - err = realloc_verifier_state(dst, src->allocated_stack, false); + err = realloc_func_state(dst, src->allocated_stack, false); if (err) return err; - memcpy(dst, src, offsetof(struct bpf_verifier_state, allocated_stack)); + memcpy(dst, src, offsetof(struct bpf_func_state, allocated_stack)); return copy_stack_state(dst, src); } +static int copy_verifier_state(struct bpf_verifier_state *dst_state, + const struct bpf_verifier_state *src) +{ + struct bpf_func_state *dst; + int i, err; + + /* if dst has more stack frames then src frame, free them */ + for (i = src->curframe + 1; i <= dst_state->curframe; i++) { + free_func_state(dst_state->frame[i]); + dst_state->frame[i] = NULL; + } + dst_state->curframe = src->curframe; + dst_state->parent = src->parent; + for (i = 0; i <= src->curframe; i++) { + dst = dst_state->frame[i]; + if (!dst) { + dst = kzalloc(sizeof(*dst), GFP_KERNEL); + if (!dst) + return -ENOMEM; + dst_state->frame[i] = dst; + } + err = copy_func_state(dst, src->frame[i]); + if (err) + return err; + } + return 0; +} + static int pop_stack(struct bpf_verifier_env *env, int *prev_insn_idx, int *insn_idx) { @@ -443,6 +494,10 @@ err: static const int caller_saved[CALLER_SAVED_REGS] = { BPF_REG_0, BPF_REG_1, BPF_REG_2, BPF_REG_3, BPF_REG_4, BPF_REG_5 }; +#define CALLEE_SAVED_REGS 5 +static const int callee_saved[CALLEE_SAVED_REGS] = { + BPF_REG_6, BPF_REG_7, BPF_REG_8, BPF_REG_9 +}; static void __mark_reg_not_init(struct bpf_reg_state *reg); @@ -578,6 +633,7 @@ static void __mark_reg_unknown(struct bpf_reg_state *reg) reg->id = 0; reg->off = 0; reg->var_off = tnum_unknown; + reg->frameno = 0; __mark_reg_unbounded(reg); } @@ -614,8 +670,9 @@ static void mark_reg_not_init(struct bpf_verifier_env *env, } static void init_reg_state(struct bpf_verifier_env *env, - struct bpf_reg_state *regs) + struct bpf_func_state *state) { + struct bpf_reg_state *regs = state->regs; int i; for (i = 0; i < MAX_BPF_REG; i++) { @@ -626,12 +683,24 @@ static void init_reg_state(struct bpf_verifier_env *env, /* frame pointer */ regs[BPF_REG_FP].type = PTR_TO_STACK; mark_reg_known_zero(env, regs, BPF_REG_FP); + regs[BPF_REG_FP].frameno = state->frameno; /* 1st arg to a function */ regs[BPF_REG_1].type = PTR_TO_CTX; mark_reg_known_zero(env, regs, BPF_REG_1); } +#define BPF_MAIN_FUNC (-1) +static void init_func_state(struct bpf_verifier_env *env, + struct bpf_func_state *state, + int callsite, int frameno, int subprogno) +{ + state->callsite = callsite; + state->frameno = frameno; + state->subprogno = subprogno; + init_reg_state(env, state); +} + enum reg_arg_type { SRC_OP, /* register is used as source operand */ DST_OP, /* register is used as destination operand */ @@ -745,29 +814,86 @@ next: return 0; } -static void mark_reg_read(const struct bpf_verifier_state *state, u32 regno) +struct bpf_verifier_state *skip_callee(struct bpf_verifier_env *env, + const struct bpf_verifier_state *state, + struct bpf_verifier_state *parent, + u32 regno) { - struct bpf_verifier_state *parent = state->parent; + struct bpf_verifier_state *tmp = NULL; + + /* 'parent' could be a state of caller and + * 'state' could be a state of callee. In such case + * parent->curframe < state->curframe + * and it's ok for r1 - r5 registers + * + * 'parent' could be a callee's state after it bpf_exit-ed. + * In such case parent->curframe > state->curframe + * and it's ok for r0 only + */ + if (parent->curframe == state->curframe || + (parent->curframe < state->curframe && + regno >= BPF_REG_1 && regno <= BPF_REG_5) || + (parent->curframe > state->curframe && + regno == BPF_REG_0)) + return parent; + + if (parent->curframe > state->curframe && + regno >= BPF_REG_6) { + /* for callee saved regs we have to skip the whole chain + * of states that belong to callee and mark as LIVE_READ + * the registers before the call + */ + tmp = parent; + while (tmp && tmp->curframe != state->curframe) { + tmp = tmp->parent; + } + if (!tmp) + goto bug; + parent = tmp; + } else { + goto bug; + } + return parent; +bug: + verbose(env, "verifier bug regno %d tmp %p\n", regno, tmp); + verbose(env, "regno %d parent frame %d current frame %d\n", + regno, parent->curframe, state->curframe); + return 0; +} + +static int mark_reg_read(struct bpf_verifier_env *env, + const struct bpf_verifier_state *state, + struct bpf_verifier_state *parent, + u32 regno) +{ + bool writes = parent == state->parent; /* Observe write marks */ if (regno == BPF_REG_FP) /* We don't need to worry about FP liveness because it's read-only */ - return; + return 0; while (parent) { /* if read wasn't screened by an earlier write ... */ - if (state->regs[regno].live & REG_LIVE_WRITTEN) + if (writes && state->frame[state->curframe]->regs[regno].live & REG_LIVE_WRITTEN) break; + parent = skip_callee(env, state, parent, regno); + if (!parent) + return -EFAULT; /* ... then we depend on parent's value */ - parent->regs[regno].live |= REG_LIVE_READ; + parent->frame[parent->curframe]->regs[regno].live |= REG_LIVE_READ; state = parent; parent = state->parent; + writes = true; } + return 0; } static int check_reg_arg(struct bpf_verifier_env *env, u32 regno, enum reg_arg_type t) { - struct bpf_reg_state *regs = env->cur_state->regs; + struct bpf_verifier_state *vstate = env->cur_state; + struct bpf_func_state *state = vstate->frame[vstate->curframe]; + struct bpf_reg_state *regs = state->regs; if (regno >= MAX_BPF_REG) { verbose(env, "R%d is invalid\n", regno); @@ -780,7 +906,7 @@ static int check_reg_arg(struct bpf_verifier_env *env, u32 regno, verbose(env, "R%d !read_ok\n", regno); return -EACCES; } - mark_reg_read(env->cur_state, regno); + return mark_reg_read(env, vstate, vstate->parent, regno); } else { /* check whether register used as dest operand can be written to */ if (regno == BPF_REG_FP) { @@ -815,13 +941,15 @@ static bool is_spillable_regtype(enum bpf_reg_type type) * stack boundary and alignment are checked in check_mem_access() */ static int check_stack_write(struct bpf_verifier_env *env, - struct bpf_verifier_state *state, int off, - int size, int value_regno) + struct bpf_func_state *state, /* func where register points to */ + int off, int size, int value_regno) { + struct bpf_func_state *cur; /* state of the current function */ int i, slot = -off - 1, spi = slot / BPF_REG_SIZE, err; + enum bpf_reg_type type; - err = realloc_verifier_state(state, round_up(slot + 1, BPF_REG_SIZE), - true); + err = realloc_func_state(state, round_up(slot + 1, BPF_REG_SIZE), + true); if (err) return err; /* caller checked that off % size == 0 and -MAX_BPF_STACK <= off < 0, @@ -834,8 +962,9 @@ static int check_stack_write(struct bpf_verifier_env *env, return -EACCES; } + cur = env->cur_state->frame[env->cur_state->curframe]; if (value_regno >= 0 && - is_spillable_regtype(state->regs[value_regno].type)) { + is_spillable_regtype((type = cur->regs[value_regno].type))) { /* register containing pointer is being spilled into stack */ if (size != BPF_REG_SIZE) { @@ -843,8 +972,13 @@ static int check_stack_write(struct bpf_verifier_env *env, return -EACCES; } + if (state != cur && type == PTR_TO_STACK) { + verbose(env, "cannot spill pointers to stack into stack frame of the caller\n"); + return -EINVAL; + } + /* save register state */ - state->stack[spi].spilled_ptr = state->regs[value_regno]; + state->stack[spi].spilled_ptr = cur->regs[value_regno]; state->stack[spi].spilled_ptr.live |= REG_LIVE_WRITTEN; for (i = 0; i < BPF_REG_SIZE; i++) @@ -860,34 +994,68 @@ static int check_stack_write(struct bpf_verifier_env *env, return 0; } -static void mark_stack_slot_read(const struct bpf_verifier_state *state, int slot) +/* registers of every function are unique and mark_reg_read() propagates + * the liveness in the following cases: + * - from callee into caller for R1 - R5 that were used as arguments + * - from caller into callee for R0 that used as result of the call + * - from caller to the same caller skipping states of the callee for R6 - R9, + * since R6 - R9 are callee saved by implicit function prologue and + * caller's R6 != callee's R6, so when we propagate liveness up to + * parent states we need to skip callee states for R6 - R9. + * + * stack slot marking is different, since stacks of caller and callee are + * accessible in both (since caller can pass a pointer to caller's stack to + * callee which can pass it to another function), hence mark_stack_slot_read() + * has to propagate the stack liveness to all parent states at given frame number. + * Consider code: + * f1() { + * ptr = fp - 8; + * *ptr = ctx; + * call f2 { + * .. = *ptr; + * } + * .. = *ptr; + * } + * First *ptr is reading from f1's stack and mark_stack_slot_read() has + * to mark liveness at the f1's frame and not f2's frame. + * Second *ptr is also reading from f1's stack and mark_stack_slot_read() has + * to propagate liveness to f2 states at f1's frame level and further into + * f1 states at f1's frame level until write into that stack slot + */ +static void mark_stack_slot_read(struct bpf_verifier_env *env, + const struct bpf_verifier_state *state, + struct bpf_verifier_state *parent, + int slot, int frameno) { - struct bpf_verifier_state *parent = state->parent; + bool writes = parent == state->parent; /* Observe write marks */ while (parent) { /* if read wasn't screened by an earlier write ... */ - if (state->stack[slot].spilled_ptr.live & REG_LIVE_WRITTEN) + if (writes && state->frame[frameno]->stack[slot].spilled_ptr.live & REG_LIVE_WRITTEN) break; /* ... then we depend on parent's value */ - parent->stack[slot].spilled_ptr.live |= REG_LIVE_READ; + parent->frame[frameno]->stack[slot].spilled_ptr.live |= REG_LIVE_READ; state = parent; parent = state->parent; + writes = true; } } static int check_stack_read(struct bpf_verifier_env *env, - struct bpf_verifier_state *state, int off, int size, - int value_regno) + struct bpf_func_state *reg_state /* func where register points to */, + int off, int size, int value_regno) { + struct bpf_verifier_state *vstate = env->cur_state; + struct bpf_func_state *state = vstate->frame[vstate->curframe]; int i, slot = -off - 1, spi = slot / BPF_REG_SIZE; u8 *stype; - if (state->allocated_stack <= slot) { + if (reg_state->allocated_stack <= slot) { verbose(env, "invalid read from stack off %d+0 size %d\n", off, size); return -EACCES; } - stype = state->stack[spi].slot_type; + stype = reg_state->stack[spi].slot_type; if (stype[0] == STACK_SPILL) { if (size != BPF_REG_SIZE) { @@ -903,13 +1071,14 @@ static int check_stack_read(struct bpf_verifier_env *env, if (value_regno >= 0) { /* restore register state from stack */ - state->regs[value_regno] = state->stack[spi].spilled_ptr; + state->regs[value_regno] = reg_state->stack[spi].spilled_ptr; /* mark reg as written since spilled pointer state likely * has its liveness marks cleared by is_state_visited() * which resets stack/reg liveness for state transitions */ state->regs[value_regno].live |= REG_LIVE_WRITTEN; - mark_stack_slot_read(state, spi); + mark_stack_slot_read(env, vstate, vstate->parent, spi, + reg_state->frameno); } return 0; } else { @@ -947,7 +1116,8 @@ static int __check_map_access(struct bpf_verifier_env *env, u32 regno, int off, static int check_map_access(struct bpf_verifier_env *env, u32 regno, int off, int size, bool zero_size_allowed) { - struct bpf_verifier_state *state = env->cur_state; + struct bpf_verifier_state *vstate = env->cur_state; + struct bpf_func_state *state = vstate->frame[vstate->curframe]; struct bpf_reg_state *reg = &state->regs[regno]; int err; @@ -1197,6 +1367,39 @@ static int check_ptr_alignment(struct bpf_verifier_env *env, strict); } +static int update_stack_depth(struct bpf_verifier_env *env, + const struct bpf_func_state *func, + int off) +{ + u16 stack = env->subprog_stack_depth[func->subprogno], total = 0; + struct bpf_verifier_state *cur = env->cur_state; + int i; + + if (stack >= -off) + return 0; + + /* update known max for given subprogram */ + env->subprog_stack_depth[func->subprogno] = -off; + + /* compute the total for current call chain */ + for (i = 0; i <= cur->curframe; i++) { + u32 depth = env->subprog_stack_depth[cur->frame[i]->subprogno]; + + /* round up to 32-bytes, since this is granularity + * of interpreter stack sizes + */ + depth = round_up(depth, 32); + total += depth; + } + + if (total > MAX_BPF_STACK) { + verbose(env, "combined stack size of %d calls is %d. Too large\n", + cur->curframe, total); + return -EACCES; + } + return 0; +} + /* check whether memory at (regno + off) is accessible for t = (read | write) * if t==write, value_regno is a register which value is stored into memory * if t==read, value_regno is a register which will receive the value from memory @@ -1207,9 +1410,9 @@ static int check_mem_access(struct bpf_verifier_env *env, int insn_idx, u32 regn int bpf_size, enum bpf_access_type t, int value_regno) { - struct bpf_verifier_state *state = env->cur_state; struct bpf_reg_state *regs = cur_regs(env); struct bpf_reg_state *reg = regs + regno; + struct bpf_func_state *state; int size, err = 0; size = bpf_size_to_bytes(bpf_size); @@ -1298,8 +1501,10 @@ static int check_mem_access(struct bpf_verifier_env *env, int insn_idx, u32 regn return -EACCES; } - if (env->prog->aux->stack_depth < -off) - env->prog->aux->stack_depth = -off; + state = func(env, reg); + err = update_stack_depth(env, state, off); + if (err) + return err; if (t == BPF_WRITE) err = check_stack_write(env, state, off, size, @@ -1390,7 +1595,7 @@ static int check_stack_boundary(struct bpf_verifier_env *env, int regno, struct bpf_call_arg_meta *meta) { struct bpf_reg_state *reg = cur_regs(env) + regno; - struct bpf_verifier_state *state = env->cur_state; + struct bpf_func_state *state = func(env, reg); int off, i, slot, spi; if (reg->type != PTR_TO_STACK) { @@ -1421,9 +1626,6 @@ static int check_stack_boundary(struct bpf_verifier_env *env, int regno, return -EACCES; } - if (env->prog->aux->stack_depth < -off) - env->prog->aux->stack_depth = -off; - if (meta && meta->raw_mode) { meta->access_size = access_size; meta->regno = regno; @@ -1441,7 +1643,7 @@ static int check_stack_boundary(struct bpf_verifier_env *env, int regno, return -EACCES; } } - return 0; + return update_stack_depth(env, state, off); } static int check_helper_mem_access(struct bpf_verifier_env *env, int regno, @@ -1694,6 +1896,10 @@ static int check_map_func_compatibility(struct bpf_verifier_env *env, case BPF_FUNC_tail_call: if (map->map_type != BPF_MAP_TYPE_PROG_ARRAY) goto error; + if (env->subprog_cnt) { + verbose(env, "tail_calls are not allowed in programs with bpf-to-bpf calls\n"); + return -EINVAL; + } break; case BPF_FUNC_perf_event_read: case BPF_FUNC_perf_event_output: @@ -1755,9 +1961,9 @@ static int check_raw_mode(const struct bpf_func_proto *fn) /* Packet data might have moved, any old PTR_TO_PACKET[_META,_END] * are now invalid, so turn them into unknown SCALAR_VALUE. */ -static void clear_all_pkt_pointers(struct bpf_verifier_env *env) +static void __clear_all_pkt_pointers(struct bpf_verifier_env *env, + struct bpf_func_state *state) { - struct bpf_verifier_state *state = env->cur_state; struct bpf_reg_state *regs = state->regs, *reg; int i; @@ -1774,7 +1980,121 @@ static void clear_all_pkt_pointers(struct bpf_verifier_env *env) } } -static int check_call(struct bpf_verifier_env *env, int func_id, int insn_idx) +static void clear_all_pkt_pointers(struct bpf_verifier_env *env) +{ + struct bpf_verifier_state *vstate = env->cur_state; + int i; + + for (i = 0; i <= vstate->curframe; i++) + __clear_all_pkt_pointers(env, vstate->frame[i]); +} + +static int check_func_call(struct bpf_verifier_env *env, struct bpf_insn *insn, + int *insn_idx) +{ + struct bpf_verifier_state *state = env->cur_state; + struct bpf_func_state *caller, *callee; + int i, subprog, target_insn; + + if (state->curframe >= MAX_CALL_FRAMES) { + verbose(env, "the call stack of %d frames is too deep\n", + state->curframe); + return -E2BIG; + } + + target_insn = *insn_idx + insn->imm; + subprog = find_subprog(env, target_insn + 1); + if (subprog < 0) { + verbose(env, "verifier bug. No program starts at insn %d\n", + target_insn + 1); + return -EFAULT; + } + + caller = state->frame[state->curframe]; + if (state->frame[state->curframe + 1]) { + verbose(env, "verifier bug. Frame %d already allocated\n", + state->curframe + 1); + return -EFAULT; + } + + callee = kzalloc(sizeof(*callee), GFP_KERNEL); + if (!callee) + return -ENOMEM; + state->frame[state->curframe + 1] = callee; + + /* callee cannot access r0, r6 - r9 for reading and has to write + * into its own stack before reading from it. + * callee can read/write into caller's stack + */ + init_func_state(env, callee, + /* remember the callsite, it will be used by bpf_exit */ + *insn_idx /* callsite */, + state->curframe + 1 /* frameno within this callchain */, + subprog + 1 /* subprog number within this prog */); + + /* copy r1 - r5 args that callee can access */ + for (i = BPF_REG_1; i <= BPF_REG_5; i++) + callee->regs[i] = caller->regs[i]; + + /* after the call regsiters r0 - r5 were scratched */ + for (i = 0; i < CALLER_SAVED_REGS; i++) { + mark_reg_not_init(env, caller->regs, caller_saved[i]); + check_reg_arg(env, caller_saved[i], DST_OP_NO_MARK); + } + + /* only increment it after check_reg_arg() finished */ + state->curframe++; + + /* and go analyze first insn of the callee */ + *insn_idx = target_insn; + + if (env->log.level) { + verbose(env, "caller:\n"); + print_verifier_state(env, caller); + verbose(env, "callee:\n"); + print_verifier_state(env, callee); + } + return 0; +} + +static int prepare_func_exit(struct bpf_verifier_env *env, int *insn_idx) +{ + struct bpf_verifier_state *state = env->cur_state; + struct bpf_func_state *caller, *callee; + struct bpf_reg_state *r0; + + callee = state->frame[state->curframe]; + r0 = &callee->regs[BPF_REG_0]; + if (r0->type == PTR_TO_STACK) { + /* technically it's ok to return caller's stack pointer + * (or caller's caller's pointer) back to the caller, + * since these pointers are valid. Only current stack + * pointer will be invalid as soon as function exits, + * but let's be conservative + */ + verbose(env, "cannot return stack pointer to the caller\n"); + return -EINVAL; + } + + state->curframe--; + caller = state->frame[state->curframe]; + /* return to the caller whatever r0 had in the callee */ + caller->regs[BPF_REG_0] = *r0; + + *insn_idx = callee->callsite + 1; + if (env->log.level) { + verbose(env, "returning from callee:\n"); + print_verifier_state(env, callee); + verbose(env, "to caller at %d:\n", *insn_idx); + print_verifier_state(env, caller); + } + /* clear everything in the callee */ + free_func_state(callee); + state->frame[state->curframe + 1] = NULL; + return 0; +} + +static int check_helper_call(struct bpf_verifier_env *env, int func_id, int insn_idx) { const struct bpf_func_proto *fn = NULL; struct bpf_reg_state *regs; @@ -1934,7 +2254,9 @@ static int adjust_ptr_min_max_vals(struct bpf_verifier_env *env, const struct bpf_reg_state *ptr_reg, const struct bpf_reg_state *off_reg) { - struct bpf_reg_state *regs = cur_regs(env), *dst_reg; + struct bpf_verifier_state *vstate = env->cur_state; + struct bpf_func_state *state = vstate->frame[vstate->curframe]; + struct bpf_reg_state *regs = state->regs, *dst_reg; bool known = tnum_is_const(off_reg->var_off); s64 smin_val = off_reg->smin_value, smax_val = off_reg->smax_value, smin_ptr = ptr_reg->smin_value, smax_ptr = ptr_reg->smax_value; @@ -1946,13 +2268,13 @@ static int adjust_ptr_min_max_vals(struct bpf_verifier_env *env, dst_reg = ®s[dst]; if (WARN_ON_ONCE(known && (smin_val != smax_val))) { - print_verifier_state(env, env->cur_state); + print_verifier_state(env, state); verbose(env, "verifier internal error: known but bad sbounds\n"); return -EINVAL; } if (WARN_ON_ONCE(known && (umin_val != umax_val))) { - print_verifier_state(env, env->cur_state); + print_verifier_state(env, state); verbose(env, "verifier internal error: known but bad ubounds\n"); return -EINVAL; @@ -2354,7 +2676,9 @@ static int adjust_scalar_min_max_vals(struct bpf_verifier_env *env, static int adjust_reg_min_max_vals(struct bpf_verifier_env *env, struct bpf_insn *insn) { - struct bpf_reg_state *regs = cur_regs(env), *dst_reg, *src_reg; + struct bpf_verifier_state *vstate = env->cur_state; + struct bpf_func_state *state = vstate->frame[vstate->curframe]; + struct bpf_reg_state *regs = state->regs, *dst_reg, *src_reg; struct bpf_reg_state *ptr_reg = NULL, off_reg = {0}; u8 opcode = BPF_OP(insn->code); int rc; @@ -2428,12 +2752,12 @@ static int adjust_reg_min_max_vals(struct bpf_verifier_env *env, /* Got here implies adding two SCALAR_VALUEs */ if (WARN_ON_ONCE(ptr_reg)) { - print_verifier_state(env, env->cur_state); + print_verifier_state(env, state); verbose(env, "verifier internal error: unexpected ptr_reg\n"); return -EINVAL; } if (WARN_ON(!src_reg)) { - print_verifier_state(env, env->cur_state); + print_verifier_state(env, state); verbose(env, "verifier internal error: no src_reg\n"); return -EINVAL; } @@ -2587,14 +2911,15 @@ static int check_alu_op(struct bpf_verifier_env *env, struct bpf_insn *insn) return 0; } -static void find_good_pkt_pointers(struct bpf_verifier_state *state, +static void find_good_pkt_pointers(struct bpf_verifier_state *vstate, struct bpf_reg_state *dst_reg, enum bpf_reg_type type, bool range_right_open) { + struct bpf_func_state *state = vstate->frame[vstate->curframe]; struct bpf_reg_state *regs = state->regs, *reg; u16 new_range; - int i; + int i, j; if (dst_reg->off < 0 || (dst_reg->off == 0 && range_right_open)) @@ -2664,12 +2989,15 @@ static void find_good_pkt_pointers(struct bpf_verifier_state *state, /* keep the maximum range already checked */ regs[i].range = max(regs[i].range, new_range); - for (i = 0; i < state->allocated_stack / BPF_REG_SIZE; i++) { - if (state->stack[i].slot_type[0] != STACK_SPILL) - continue; - reg = &state->stack[i].spilled_ptr; - if (reg->type == type && reg->id == dst_reg->id) - reg->range = max(reg->range, new_range); + for (j = 0; j <= vstate->curframe; j++) { + state = vstate->frame[j]; + for (i = 0; i < state->allocated_stack / BPF_REG_SIZE; i++) { + if (state->stack[i].slot_type[0] != STACK_SPILL) + continue; + reg = &state->stack[i].spilled_ptr; + if (reg->type == type && reg->id == dst_reg->id) + reg->range = max(reg->range, new_range); + } } } @@ -2907,20 +3235,24 @@ static void mark_map_reg(struct bpf_reg_state *regs, u32 regno, u32 id, /* The logic is similar to find_good_pkt_pointers(), both could eventually * be folded together at some point. */ -static void mark_map_regs(struct bpf_verifier_state *state, u32 regno, +static void mark_map_regs(struct bpf_verifier_state *vstate, u32 regno, bool is_null) { + struct bpf_func_state *state = vstate->frame[vstate->curframe]; struct bpf_reg_state *regs = state->regs; u32 id = regs[regno].id; - int i; + int i, j; for (i = 0; i < MAX_BPF_REG; i++) mark_map_reg(regs, i, id, is_null); - for (i = 0; i < state->allocated_stack / BPF_REG_SIZE; i++) { - if (state->stack[i].slot_type[0] != STACK_SPILL) - continue; - mark_map_reg(&state->stack[i].spilled_ptr, 0, id, is_null); + for (j = 0; j <= vstate->curframe; j++) { + state = vstate->frame[j]; + for (i = 0; i < state->allocated_stack / BPF_REG_SIZE; i++) { + if (state->stack[i].slot_type[0] != STACK_SPILL) + continue; + mark_map_reg(&state->stack[i].spilled_ptr, 0, id, is_null); + } } } @@ -3020,8 +3352,10 @@ static bool try_match_pkt_pointers(const struct bpf_insn *insn, static int check_cond_jmp_op(struct bpf_verifier_env *env, struct bpf_insn *insn, int *insn_idx) { - struct bpf_verifier_state *other_branch, *this_branch = env->cur_state; - struct bpf_reg_state *regs = this_branch->regs, *dst_reg; + struct bpf_verifier_state *this_branch = env->cur_state; + struct bpf_verifier_state *other_branch; + struct bpf_reg_state *regs = this_branch->frame[this_branch->curframe]->regs; + struct bpf_reg_state *dst_reg, *other_branch_regs; u8 opcode = BPF_OP(insn->code); int err; @@ -3084,6 +3418,7 @@ static int check_cond_jmp_op(struct bpf_verifier_env *env, other_branch = push_stack(env, *insn_idx + insn->off + 1, *insn_idx); if (!other_branch) return -EFAULT; + other_branch_regs = other_branch->frame[other_branch->curframe]->regs; /* detect if we are comparing against a constant value so we can adjust * our min/max values for our dst register. @@ -3096,22 +3431,22 @@ static int check_cond_jmp_op(struct bpf_verifier_env *env, if (dst_reg->type == SCALAR_VALUE && regs[insn->src_reg].type == SCALAR_VALUE) { if (tnum_is_const(regs[insn->src_reg].var_off)) - reg_set_min_max(&other_branch->regs[insn->dst_reg], + reg_set_min_max(&other_branch_regs[insn->dst_reg], dst_reg, regs[insn->src_reg].var_off.value, opcode); else if (tnum_is_const(dst_reg->var_off)) - reg_set_min_max_inv(&other_branch->regs[insn->src_reg], + reg_set_min_max_inv(&other_branch_regs[insn->src_reg], ®s[insn->src_reg], dst_reg->var_off.value, opcode); else if (opcode == BPF_JEQ || opcode == BPF_JNE) /* Comparing for equality, we can combine knowledge */ - reg_combine_min_max(&other_branch->regs[insn->src_reg], - &other_branch->regs[insn->dst_reg], + reg_combine_min_max(&other_branch_regs[insn->src_reg], + &other_branch_regs[insn->dst_reg], ®s[insn->src_reg], ®s[insn->dst_reg], opcode); } } else if (dst_reg->type == SCALAR_VALUE) { - reg_set_min_max(&other_branch->regs[insn->dst_reg], + reg_set_min_max(&other_branch_regs[insn->dst_reg], dst_reg, insn->imm, opcode); } @@ -3132,7 +3467,7 @@ static int check_cond_jmp_op(struct bpf_verifier_env *env, return -EACCES; } if (env->log.level) - print_verifier_state(env, this_branch); + print_verifier_state(env, this_branch->frame[this_branch->curframe]); return 0; } @@ -3217,6 +3552,18 @@ static int check_ld_abs(struct bpf_verifier_env *env, struct bpf_insn *insn) return -EINVAL; } + if (env->subprog_cnt) { + /* when program has LD_ABS insn JITs and interpreter assume + * that r1 == ctx == skb which is not the case for callees + * that can have arbitrary arguments. It's problematic + * for main prog as well since JITs would need to analyze + * all functions in order to make proper register save/restore + * decisions in the main prog. Hence disallow LD_ABS with calls + */ + verbose(env, "BPF_LD_[ABS|IND] instructions cannot be mixed with bpf-to-bpf calls\n"); + return -EINVAL; + } + if (insn->dst_reg != BPF_REG_0 || insn->off != 0 || BPF_SIZE(insn->code) == BPF_DW || (mode == BPF_ABS && insn->src_reg != BPF_REG_0)) { @@ -3555,11 +3902,21 @@ static bool check_ids(u32 old_id, u32 cur_id, struct idpair *idmap) static bool regsafe(struct bpf_reg_state *rold, struct bpf_reg_state *rcur, struct idpair *idmap) { + bool equal; + if (!(rold->live & REG_LIVE_READ)) /* explored state didn't use this */ return true; - if (memcmp(rold, rcur, offsetof(struct bpf_reg_state, live)) == 0) + equal = memcmp(rold, rcur, offsetof(struct bpf_reg_state, frameno)) == 0; + + if (rold->type == PTR_TO_STACK) + /* two stack pointers are equal only if they're pointing to + * the same stack frame, since fp-8 in foo != fp-8 in bar + */ + return equal && rold->frameno == rcur->frameno; + + if (equal) return true; if (rold->type == NOT_INIT) @@ -3632,7 +3989,6 @@ static bool regsafe(struct bpf_reg_state *rold, struct bpf_reg_state *rcur, tnum_in(rold->var_off, rcur->var_off); case PTR_TO_CTX: case CONST_PTR_TO_MAP: - case PTR_TO_STACK: case PTR_TO_PACKET_END: /* Only valid matches are exact, which memcmp() above * would have accepted @@ -3647,8 +4003,8 @@ static bool regsafe(struct bpf_reg_state *rold, struct bpf_reg_state *rcur, return false; } -static bool stacksafe(struct bpf_verifier_state *old, - struct bpf_verifier_state *cur, +static bool stacksafe(struct bpf_func_state *old, + struct bpf_func_state *cur, struct idpair *idmap) { int i, spi; @@ -3724,9 +4080,8 @@ static bool stacksafe(struct bpf_verifier_state *old, * whereas register type in current state is meaningful, it means that * the current state will reach 'bpf_exit' instruction safely */ -static bool states_equal(struct bpf_verifier_env *env, - struct bpf_verifier_state *old, - struct bpf_verifier_state *cur) +static bool func_states_equal(struct bpf_func_state *old, + struct bpf_func_state *cur) { struct idpair *idmap; bool ret = false; @@ -3750,71 +4105,76 @@ out_free: return ret; } +static bool states_equal(struct bpf_verifier_env *env, + struct bpf_verifier_state *old, + struct bpf_verifier_state *cur) +{ + int i; + + if (old->curframe != cur->curframe) + return false; + + /* for states to be equal callsites have to be the same + * and all frame states need to be equivalent + */ + for (i = 0; i <= old->curframe; i++) { + if (old->frame[i]->callsite != cur->frame[i]->callsite) + return false; + if (!func_states_equal(old->frame[i], cur->frame[i])) + return false; + } + return true; +} + /* A write screens off any subsequent reads; but write marks come from the - * straight-line code between a state and its parent. When we arrive at a - * jump target (in the first iteration of the propagate_liveness() loop), - * we didn't arrive by the straight-line code, so read marks in state must - * propagate to parent regardless of state's write marks. + * straight-line code between a state and its parent. When we arrive at an + * equivalent state (jump target or such) we didn't arrive by the straight-line + * code, so read marks in the state must propagate to the parent regardless + * of the state's write marks. That's what 'parent == state->parent' comparison + * in mark_reg_read() and mark_stack_slot_read() is for. */ -static bool do_propagate_liveness(const struct bpf_verifier_state *state, - struct bpf_verifier_state *parent) +static int propagate_liveness(struct bpf_verifier_env *env, + const struct bpf_verifier_state *vstate, + struct bpf_verifier_state *vparent) { - bool writes = parent == state->parent; /* Observe write marks */ - bool touched = false; /* any changes made? */ - int i; + int i, frame, err = 0; + struct bpf_func_state *state, *parent; - if (!parent) - return touched; + if (vparent->curframe != vstate->curframe) { + WARN(1, "propagate_live: parent frame %d current frame %d\n", + vparent->curframe, vstate->curframe); + return -EFAULT; + } /* Propagate read liveness of registers... */ BUILD_BUG_ON(BPF_REG_FP + 1 != MAX_BPF_REG); /* We don't need to worry about FP liveness because it's read-only */ for (i = 0; i < BPF_REG_FP; i++) { - if (parent->regs[i].live & REG_LIVE_READ) + if (vparent->frame[vparent->curframe]->regs[i].live & REG_LIVE_READ) continue; - if (writes && (state->regs[i].live & REG_LIVE_WRITTEN)) - continue; - if (state->regs[i].live & REG_LIVE_READ) { - parent->regs[i].live |= REG_LIVE_READ; - touched = true; + if (vstate->frame[vstate->curframe]->regs[i].live & REG_LIVE_READ) { + err = mark_reg_read(env, vstate, vparent, i); + if (err) + return err; } } + /* ... and stack slots */ - for (i = 0; i < state->allocated_stack / BPF_REG_SIZE && - i < parent->allocated_stack / BPF_REG_SIZE; i++) { - if (parent->stack[i].slot_type[0] != STACK_SPILL) - continue; - if (state->stack[i].slot_type[0] != STACK_SPILL) - continue; - if (parent->stack[i].spilled_ptr.live & REG_LIVE_READ) - continue; - if (writes && - (state->stack[i].spilled_ptr.live & REG_LIVE_WRITTEN)) - continue; - if (state->stack[i].spilled_ptr.live & REG_LIVE_READ) { - parent->stack[i].spilled_ptr.live |= REG_LIVE_READ; - touched = true; + for (frame = 0; frame <= vstate->curframe; frame++) { + state = vstate->frame[frame]; + parent = vparent->frame[frame]; + for (i = 0; i < state->allocated_stack / BPF_REG_SIZE && + i < parent->allocated_stack / BPF_REG_SIZE; i++) { + if (parent->stack[i].slot_type[0] != STACK_SPILL) + continue; + if (state->stack[i].slot_type[0] != STACK_SPILL) + continue; + if (parent->stack[i].spilled_ptr.live & REG_LIVE_READ) + continue; + if (state->stack[i].spilled_ptr.live & REG_LIVE_READ) + mark_stack_slot_read(env, vstate, vparent, i, frame); } } - return touched; -} - -/* "parent" is "a state from which we reach the current state", but initially - * it is not the state->parent (i.e. "the state whose straight-line code leads - * to the current state"), instead it is the state that happened to arrive at - * a (prunable) equivalent of the current state. See comment above - * do_propagate_liveness() for consequences of this. - * This function is just a more efficient way of calling mark_reg_read() or - * mark_stack_slot_read() on each reg in "parent" that is read in "state", - * though it requires that parent != state->parent in the call arguments. - */ -static void propagate_liveness(const struct bpf_verifier_state *state, - struct bpf_verifier_state *parent) -{ - while (do_propagate_liveness(state, parent)) { - /* Something changed, so we need to feed those changes onward */ - state = parent; - parent = state->parent; - } + return err; } static int is_state_visited(struct bpf_verifier_env *env, int insn_idx) @@ -3822,7 +4182,7 @@ static int is_state_visited(struct bpf_verifier_env *env, int insn_idx) struct bpf_verifier_state_list *new_sl; struct bpf_verifier_state_list *sl; struct bpf_verifier_state *cur = env->cur_state; - int i, err; + int i, j, err; sl = env->explored_states[insn_idx]; if (!sl) @@ -3843,7 +4203,9 @@ static int is_state_visited(struct bpf_verifier_env *env, int insn_idx) * they'll be immediately forgotten as we're pruning * this state and will pop a new one. */ - propagate_liveness(&sl->state, cur); + err = propagate_liveness(env, &sl->state, cur); + if (err) + return err; return 1; } sl = sl->next; @@ -3851,9 +4213,10 @@ static int is_state_visited(struct bpf_verifier_env *env, int insn_idx) /* there were no equivalent states, remember current one. * technically the current state is not proven to be safe yet, - * but it will either reach bpf_exit (which means it's safe) or - * it will be rejected. Since there are no loops, we won't be - * seeing this 'insn_idx' instruction again on the way to bpf_exit + * but it will either reach outer most bpf_exit (which means it's safe) + * or it will be rejected. Since there are no loops, we won't be + * seeing this tuple (frame[0].callsite, frame[1].callsite, .. insn_idx) + * again on the way to bpf_exit */ new_sl = kzalloc(sizeof(struct bpf_verifier_state_list), GFP_KERNEL); if (!new_sl) @@ -3877,10 +4240,16 @@ static int is_state_visited(struct bpf_verifier_env *env, int insn_idx) * explored_states can get read marks.) */ for (i = 0; i < BPF_REG_FP; i++) - cur->regs[i].live = REG_LIVE_NONE; - for (i = 0; i < cur->allocated_stack / BPF_REG_SIZE; i++) - if (cur->stack[i].slot_type[0] == STACK_SPILL) - cur->stack[i].spilled_ptr.live = REG_LIVE_NONE; + cur->frame[cur->curframe]->regs[i].live = REG_LIVE_NONE; + + /* all stack frames are accessible from callee, clear them all */ + for (j = 0; j <= cur->curframe; j++) { + struct bpf_func_state *frame = cur->frame[j]; + + for (i = 0; i < frame->allocated_stack / BPF_REG_SIZE; i++) + if (frame->stack[i].slot_type[0] == STACK_SPILL) + frame->stack[i].spilled_ptr.live = REG_LIVE_NONE; + } return 0; } @@ -3898,7 +4267,7 @@ static int do_check(struct bpf_verifier_env *env) struct bpf_verifier_state *state; struct bpf_insn *insns = env->prog->insnsi; struct bpf_reg_state *regs; - int insn_cnt = env->prog->len; + int insn_cnt = env->prog->len, i; int insn_idx, prev_insn_idx = 0; int insn_processed = 0; bool do_print_state = false; @@ -3906,9 +4275,18 @@ static int do_check(struct bpf_verifier_env *env) state = kzalloc(sizeof(struct bpf_verifier_state), GFP_KERNEL); if (!state) return -ENOMEM; - env->cur_state = state; - init_reg_state(env, state->regs); + state->curframe = 0; state->parent = NULL; + state->frame[0] = kzalloc(sizeof(struct bpf_func_state), GFP_KERNEL); + if (!state->frame[0]) { + kfree(state); + return -ENOMEM; + } + env->cur_state = state; + init_func_state(env, state->frame[0], + BPF_MAIN_FUNC /* callsite */, + 0 /* frameno */, + 0 /* subprogno, zero == main subprog */); insn_idx = 0; for (;;) { struct bpf_insn *insn; @@ -3955,7 +4333,7 @@ static int do_check(struct bpf_verifier_env *env) else verbose(env, "\nfrom %d to %d:", prev_insn_idx, insn_idx); - print_verifier_state(env, state); + print_verifier_state(env, state->frame[state->curframe]); do_print_state = false; } @@ -4088,13 +4466,17 @@ static int do_check(struct bpf_verifier_env *env) if (opcode == BPF_CALL) { if (BPF_SRC(insn->code) != BPF_K || insn->off != 0 || - insn->src_reg != BPF_REG_0 || + (insn->src_reg != BPF_REG_0 && + insn->src_reg != BPF_PSEUDO_CALL) || insn->dst_reg != BPF_REG_0) { verbose(env, "BPF_CALL uses reserved fields\n"); return -EINVAL; } - err = check_call(env, insn->imm, insn_idx); + if (insn->src_reg == BPF_PSEUDO_CALL) + err = check_func_call(env, insn, &insn_idx); + else + err = check_helper_call(env, insn->imm, insn_idx); if (err) return err; @@ -4119,6 +4501,16 @@ static int do_check(struct bpf_verifier_env *env) return -EINVAL; } + if (state->curframe) { + /* exit from nested function */ + prev_insn_idx = insn_idx; + err = prepare_func_exit(env, &insn_idx); + if (err) + return err; + do_print_state = true; + continue; + } + /* eBPF calling convetion is such that R0 is used * to return the value from eBPF program. * Make sure that it's readable at this time @@ -4179,8 +4571,16 @@ process_bpf_exit: insn_idx++; } - verbose(env, "processed %d insns, stack depth %d\n", insn_processed, - env->prog->aux->stack_depth); + verbose(env, "processed %d insns, stack depth ", insn_processed); + for (i = 0; i < env->subprog_cnt + 1; i++) { + u32 depth = env->subprog_stack_depth[i]; + + verbose(env, "%d", depth); + if (i + 1 < env->subprog_cnt + 1) + verbose(env, "+"); + } + verbose(env, "\n"); + env->prog->aux->stack_depth = env->subprog_stack_depth[0]; return 0; } -- cgit v1.2.3-71-gd317 From cc2b14d51053eb055c06f45e1a5cdbfcf2b79e94 Mon Sep 17 00:00:00 2001 From: Alexei Starovoitov Date: Thu, 14 Dec 2017 17:55:08 -0800 Subject: bpf: teach verifier to recognize zero initialized stack programs with function calls are often passing various pointers via stack. When all calls are inlined llvm flattens stack accesses and optimizes away extra branches. When functions are not inlined it becomes the job of the verifier to recognize zero initialized stack to avoid exploring paths that program will not take. The following program would fail otherwise: ptr = &buffer_on_stack; *ptr = 0; ... func_call(.., ptr, ...) { if (..) *ptr = bpf_map_lookup(); } ... if (*ptr != 0) { // Access (*ptr)->field is valid. // Without stack_zero tracking such (*ptr)->field access // will be rejected } since stack slots are no longer uniform invalid | spill | misc add liveness marking to all slots, but do it in 8 byte chunks. So if nothing was read or written in [fp-16, fp-9] range it will be marked as LIVE_NONE. If any byte in that range was read, it will be marked LIVE_READ and stacksafe() check will perform byte-by-byte verification. If all bytes in the range were written the slot will be marked as LIVE_WRITTEN. This significantly speeds up state equality comparison and reduces total number of states processed. before after bpf_lb-DLB_L3.o 2051 2003 bpf_lb-DLB_L4.o 3287 3164 bpf_lb-DUNKNOWN.o 1080 1080 bpf_lxc-DDROP_ALL.o 24980 12361 bpf_lxc-DUNKNOWN.o 34308 16605 bpf_netdev.o 15404 10962 bpf_overlay.o 7191 6679 Signed-off-by: Alexei Starovoitov Acked-by: Daniel Borkmann Signed-off-by: Daniel Borkmann --- include/linux/bpf_verifier.h | 3 +- kernel/bpf/verifier.c | 129 +++++++++++++++++++++++++++++++++---------- 2 files changed, 103 insertions(+), 29 deletions(-) (limited to 'kernel') diff --git a/include/linux/bpf_verifier.h b/include/linux/bpf_verifier.h index 1f23408024ee..585d4e17ea88 100644 --- a/include/linux/bpf_verifier.h +++ b/include/linux/bpf_verifier.h @@ -91,7 +91,8 @@ struct bpf_reg_state { enum bpf_stack_slot_type { STACK_INVALID, /* nothing was stored in this stack slot */ STACK_SPILL, /* register spilled into stack */ - STACK_MISC /* BPF program wrote some data into this slot */ + STACK_MISC, /* BPF program wrote some data into this slot */ + STACK_ZERO, /* BPF program wrote constant zero */ }; #define BPF_REG_SIZE 8 /* size of eBPF register in bytes */ diff --git a/kernel/bpf/verifier.c b/kernel/bpf/verifier.c index f6e09d84a96f..df1ded3faf1d 100644 --- a/kernel/bpf/verifier.c +++ b/kernel/bpf/verifier.c @@ -311,6 +311,8 @@ static void print_verifier_state(struct bpf_verifier_env *env, verbose(env, "=%s", reg_type_str[state->stack[i].spilled_ptr.type]); } + if (state->stack[i].slot_type[0] == STACK_ZERO) + verbose(env, " fp%d=0", (-i - 1) * BPF_REG_SIZE); } verbose(env, "\n"); } @@ -522,6 +524,13 @@ static void __mark_reg_known_zero(struct bpf_reg_state *reg) __mark_reg_known(reg, 0); } +static void __mark_reg_const_zero(struct bpf_reg_state *reg) +{ + __mark_reg_known(reg, 0); + reg->off = 0; + reg->type = SCALAR_VALUE; +} + static void mark_reg_known_zero(struct bpf_verifier_env *env, struct bpf_reg_state *regs, u32 regno) { @@ -937,6 +946,12 @@ static bool is_spillable_regtype(enum bpf_reg_type type) } } +/* Does this register contain a constant zero? */ +static bool register_is_null(struct bpf_reg_state *reg) +{ + return reg->type == SCALAR_VALUE && tnum_equals_const(reg->var_off, 0); +} + /* check_stack_read/write functions track spill/fill of registers, * stack boundary and alignment are checked in check_mem_access() */ @@ -984,12 +999,30 @@ static int check_stack_write(struct bpf_verifier_env *env, for (i = 0; i < BPF_REG_SIZE; i++) state->stack[spi].slot_type[i] = STACK_SPILL; } else { + u8 type = STACK_MISC; + /* regular write of data into stack */ state->stack[spi].spilled_ptr = (struct bpf_reg_state) {}; + /* only mark the slot as written if all 8 bytes were written + * otherwise read propagation may incorrectly stop too soon + * when stack slots are partially written. + * This heuristic means that read propagation will be + * conservative, since it will add reg_live_read marks + * to stack slots all the way to first state when programs + * writes+reads less than 8 bytes + */ + if (size == BPF_REG_SIZE) + state->stack[spi].spilled_ptr.live |= REG_LIVE_WRITTEN; + + /* when we zero initialize stack slots mark them as such */ + if (value_regno >= 0 && + register_is_null(&cur->regs[value_regno])) + type = STACK_ZERO; + for (i = 0; i < size; i++) state->stack[spi].slot_type[(slot - i) % BPF_REG_SIZE] = - STACK_MISC; + type; } return 0; } @@ -1030,6 +1063,14 @@ static void mark_stack_slot_read(struct bpf_verifier_env *env, bool writes = parent == state->parent; /* Observe write marks */ while (parent) { + if (parent->frame[frameno]->allocated_stack <= slot * BPF_REG_SIZE) + /* since LIVE_WRITTEN mark is only done for full 8-byte + * write the read marks are conservative and parent + * state may not even have the stack allocated. In such case + * end the propagation, since the loop reached beginning + * of the function + */ + break; /* if read wasn't screened by an earlier write ... */ if (writes && state->frame[frameno]->stack[slot].spilled_ptr.live & REG_LIVE_WRITTEN) break; @@ -1077,21 +1118,38 @@ static int check_stack_read(struct bpf_verifier_env *env, * which resets stack/reg liveness for state transitions */ state->regs[value_regno].live |= REG_LIVE_WRITTEN; - mark_stack_slot_read(env, vstate, vstate->parent, spi, - reg_state->frameno); } + mark_stack_slot_read(env, vstate, vstate->parent, spi, + reg_state->frameno); return 0; } else { + int zeros = 0; + for (i = 0; i < size; i++) { - if (stype[(slot - i) % BPF_REG_SIZE] != STACK_MISC) { - verbose(env, "invalid read from stack off %d+%d size %d\n", - off, i, size); - return -EACCES; + if (stype[(slot - i) % BPF_REG_SIZE] == STACK_MISC) + continue; + if (stype[(slot - i) % BPF_REG_SIZE] == STACK_ZERO) { + zeros++; + continue; } + verbose(env, "invalid read from stack off %d+%d size %d\n", + off, i, size); + return -EACCES; + } + mark_stack_slot_read(env, vstate, vstate->parent, spi, + reg_state->frameno); + if (value_regno >= 0) { + if (zeros == size) { + /* any size read into register is zero extended, + * so the whole register == const_zero + */ + __mark_reg_const_zero(&state->regs[value_regno]); + } else { + /* have read misc data from the stack */ + mark_reg_unknown(env, state->regs, value_regno); + } + state->regs[value_regno].live |= REG_LIVE_WRITTEN; } - if (value_regno >= 0) - /* have read misc data from the stack */ - mark_reg_unknown(env, state->regs, value_regno); return 0; } } @@ -1578,12 +1636,6 @@ static int check_xadd(struct bpf_verifier_env *env, int insn_idx, struct bpf_ins BPF_SIZE(insn->code), BPF_WRITE, -1); } -/* Does this register contain a constant zero? */ -static bool register_is_null(struct bpf_reg_state *reg) -{ - return reg->type == SCALAR_VALUE && tnum_equals_const(reg->var_off, 0); -} - /* when register 'regno' is passed into function that will read 'access_size' * bytes from that pointer, make sure that it's within stack boundary * and all elements of stack are initialized. @@ -1633,15 +1685,30 @@ static int check_stack_boundary(struct bpf_verifier_env *env, int regno, } for (i = 0; i < access_size; i++) { + u8 *stype; + slot = -(off + i) - 1; spi = slot / BPF_REG_SIZE; - if (state->allocated_stack <= slot || - state->stack[spi].slot_type[slot % BPF_REG_SIZE] != - STACK_MISC) { - verbose(env, "invalid indirect read from stack off %d+%d size %d\n", - off, i, access_size); - return -EACCES; + if (state->allocated_stack <= slot) + goto err; + stype = &state->stack[spi].slot_type[slot % BPF_REG_SIZE]; + if (*stype == STACK_MISC) + goto mark; + if (*stype == STACK_ZERO) { + /* helper can write anything into the stack */ + *stype = STACK_MISC; + goto mark; } +err: + verbose(env, "invalid indirect read from stack off %d+%d size %d\n", + off, i, access_size); + return -EACCES; +mark: + /* reading any byte out of 8-byte 'spill_slot' will cause + * the whole slot to be marked as 'read' + */ + mark_stack_slot_read(env, env->cur_state, env->cur_state->parent, + spi, state->frameno); } return update_stack_depth(env, state, off); } @@ -4022,8 +4089,19 @@ static bool stacksafe(struct bpf_func_state *old, for (i = 0; i < old->allocated_stack; i++) { spi = i / BPF_REG_SIZE; + if (!(old->stack[spi].spilled_ptr.live & REG_LIVE_READ)) + /* explored state didn't use this */ + return true; + if (old->stack[spi].slot_type[i % BPF_REG_SIZE] == STACK_INVALID) continue; + /* if old state was safe with misc data in the stack + * it will be safe with zero-initialized stack. + * The opposite is not true + */ + if (old->stack[spi].slot_type[i % BPF_REG_SIZE] == STACK_MISC && + cur->stack[spi].slot_type[i % BPF_REG_SIZE] == STACK_ZERO) + continue; if (old->stack[spi].slot_type[i % BPF_REG_SIZE] != cur->stack[spi].slot_type[i % BPF_REG_SIZE]) /* Ex: old explored (safe) state has STACK_SPILL in @@ -4164,10 +4242,6 @@ static int propagate_liveness(struct bpf_verifier_env *env, parent = vparent->frame[frame]; for (i = 0; i < state->allocated_stack / BPF_REG_SIZE && i < parent->allocated_stack / BPF_REG_SIZE; i++) { - if (parent->stack[i].slot_type[0] != STACK_SPILL) - continue; - if (state->stack[i].slot_type[0] != STACK_SPILL) - continue; if (parent->stack[i].spilled_ptr.live & REG_LIVE_READ) continue; if (state->stack[i].spilled_ptr.live & REG_LIVE_READ) @@ -4247,8 +4321,7 @@ static int is_state_visited(struct bpf_verifier_env *env, int insn_idx) struct bpf_func_state *frame = cur->frame[j]; for (i = 0; i < frame->allocated_stack / BPF_REG_SIZE; i++) - if (frame->stack[i].slot_type[0] == STACK_SPILL) - frame->stack[i].spilled_ptr.live = REG_LIVE_NONE; + frame->stack[i].spilled_ptr.live = REG_LIVE_NONE; } return 0; } -- cgit v1.2.3-71-gd317 From 1ea47e01ad6ea0fe99697c54c2413d81dd21fe32 Mon Sep 17 00:00:00 2001 From: Alexei Starovoitov Date: Thu, 14 Dec 2017 17:55:13 -0800 Subject: bpf: add support for bpf_call to interpreter though bpf_call is still the same call instruction and calling convention 'bpf to bpf' and 'bpf to helper' is the same the interpreter has to oparate on 'struct bpf_insn *'. To distinguish these two cases add a kernel internal opcode and mark call insns with it. This opcode is seen by interpreter only. JITs will never see it. Also add tiny bit of debug code to aid interpreter debugging. Signed-off-by: Alexei Starovoitov Acked-by: Daniel Borkmann Signed-off-by: Daniel Borkmann --- include/linux/bpf.h | 1 + include/linux/filter.h | 6 ++++ kernel/bpf/core.c | 90 ++++++++++++++++++++++++++++++++++++++++---------- kernel/bpf/verifier.c | 36 ++++++++++++++++++++ 4 files changed, 116 insertions(+), 17 deletions(-) (limited to 'kernel') diff --git a/include/linux/bpf.h b/include/linux/bpf.h index 54dc7cae2949..8935f6f63d5f 100644 --- a/include/linux/bpf.h +++ b/include/linux/bpf.h @@ -402,6 +402,7 @@ static inline void bpf_long_memcpy(void *dst, const void *src, u32 size) /* verify correctness of eBPF program */ int bpf_check(struct bpf_prog **fp, union bpf_attr *attr); +void bpf_patch_call_args(struct bpf_insn *insn, u32 stack_depth); /* Map specifics */ struct net_device *__dev_map_lookup_elem(struct bpf_map *map, u32 key); diff --git a/include/linux/filter.h b/include/linux/filter.h index 5feb441d3dd9..f26e6da1007b 100644 --- a/include/linux/filter.h +++ b/include/linux/filter.h @@ -58,6 +58,9 @@ struct bpf_prog_aux; /* unused opcode to mark special call to bpf_tail_call() helper */ #define BPF_TAIL_CALL 0xf0 +/* unused opcode to mark call to interpreter with arguments */ +#define BPF_CALL_ARGS 0xe0 + /* As per nm, we expose JITed images as text (code) section for * kallsyms. That way, tools like perf can find it to match * addresses. @@ -710,6 +713,9 @@ bool sk_filter_charge(struct sock *sk, struct sk_filter *fp); void sk_filter_uncharge(struct sock *sk, struct sk_filter *fp); u64 __bpf_call_base(u64 r1, u64 r2, u64 r3, u64 r4, u64 r5); +#define __bpf_call_base_args \ + ((u64 (*)(u64, u64, u64, u64, u64, const struct bpf_insn *)) \ + __bpf_call_base) struct bpf_prog *bpf_int_jit_compile(struct bpf_prog *prog); void bpf_jit_compile(struct bpf_prog *prog); diff --git a/kernel/bpf/core.c b/kernel/bpf/core.c index d32bebf4f2de..dc12c4fd006e 100644 --- a/kernel/bpf/core.c +++ b/kernel/bpf/core.c @@ -217,30 +217,40 @@ int bpf_prog_calc_tag(struct bpf_prog *fp) return 0; } -static bool bpf_is_jmp_and_has_target(const struct bpf_insn *insn) -{ - return BPF_CLASS(insn->code) == BPF_JMP && - /* Call and Exit are both special jumps with no - * target inside the BPF instruction image. - */ - BPF_OP(insn->code) != BPF_CALL && - BPF_OP(insn->code) != BPF_EXIT; -} - static void bpf_adj_branches(struct bpf_prog *prog, u32 pos, u32 delta) { struct bpf_insn *insn = prog->insnsi; u32 i, insn_cnt = prog->len; + bool pseudo_call; + u8 code; + int off; for (i = 0; i < insn_cnt; i++, insn++) { - if (!bpf_is_jmp_and_has_target(insn)) + code = insn->code; + if (BPF_CLASS(code) != BPF_JMP) continue; + if (BPF_OP(code) == BPF_EXIT) + continue; + if (BPF_OP(code) == BPF_CALL) { + if (insn->src_reg == BPF_PSEUDO_CALL) + pseudo_call = true; + else + continue; + } else { + pseudo_call = false; + } + off = pseudo_call ? insn->imm : insn->off; /* Adjust offset of jmps if we cross boundaries. */ - if (i < pos && i + insn->off + 1 > pos) - insn->off += delta; - else if (i > pos + delta && i + insn->off + 1 <= pos + delta) - insn->off -= delta; + if (i < pos && i + off + 1 > pos) + off += delta; + else if (i > pos + delta && i + off + 1 <= pos + delta) + off -= delta; + + if (pseudo_call) + insn->imm = off; + else + insn->off = off; } } @@ -774,8 +784,7 @@ EXPORT_SYMBOL_GPL(__bpf_call_base); * * Decode and execute eBPF instructions. */ -static unsigned int ___bpf_prog_run(u64 *regs, const struct bpf_insn *insn, - u64 *stack) +static u64 ___bpf_prog_run(u64 *regs, const struct bpf_insn *insn, u64 *stack) { u64 tmp; static const void *jumptable[256] = { @@ -835,6 +844,7 @@ static unsigned int ___bpf_prog_run(u64 *regs, const struct bpf_insn *insn, [BPF_ALU64 | BPF_NEG] = &&ALU64_NEG, /* Call instruction */ [BPF_JMP | BPF_CALL] = &&JMP_CALL, + [BPF_JMP | BPF_CALL_ARGS] = &&JMP_CALL_ARGS, [BPF_JMP | BPF_TAIL_CALL] = &&JMP_TAIL_CALL, /* Jumps */ [BPF_JMP | BPF_JA] = &&JMP_JA, @@ -1025,6 +1035,13 @@ select_insn: BPF_R4, BPF_R5); CONT; + JMP_CALL_ARGS: + BPF_R0 = (__bpf_call_base_args + insn->imm)(BPF_R1, BPF_R2, + BPF_R3, BPF_R4, + BPF_R5, + insn + insn->off + 1); + CONT; + JMP_TAIL_CALL: { struct bpf_map *map = (struct bpf_map *) (unsigned long) BPF_R2; struct bpf_array *array = container_of(map, struct bpf_array, map); @@ -1297,6 +1314,23 @@ static unsigned int PROG_NAME(stack_size)(const void *ctx, const struct bpf_insn return ___bpf_prog_run(regs, insn, stack); \ } +#define PROG_NAME_ARGS(stack_size) __bpf_prog_run_args##stack_size +#define DEFINE_BPF_PROG_RUN_ARGS(stack_size) \ +static u64 PROG_NAME_ARGS(stack_size)(u64 r1, u64 r2, u64 r3, u64 r4, u64 r5, \ + const struct bpf_insn *insn) \ +{ \ + u64 stack[stack_size / sizeof(u64)]; \ + u64 regs[MAX_BPF_REG]; \ +\ + FP = (u64) (unsigned long) &stack[ARRAY_SIZE(stack)]; \ + BPF_R1 = r1; \ + BPF_R2 = r2; \ + BPF_R3 = r3; \ + BPF_R4 = r4; \ + BPF_R5 = r5; \ + return ___bpf_prog_run(regs, insn, stack); \ +} + #define EVAL1(FN, X) FN(X) #define EVAL2(FN, X, Y...) FN(X) EVAL1(FN, Y) #define EVAL3(FN, X, Y...) FN(X) EVAL2(FN, Y) @@ -1308,6 +1342,10 @@ EVAL6(DEFINE_BPF_PROG_RUN, 32, 64, 96, 128, 160, 192); EVAL6(DEFINE_BPF_PROG_RUN, 224, 256, 288, 320, 352, 384); EVAL4(DEFINE_BPF_PROG_RUN, 416, 448, 480, 512); +EVAL6(DEFINE_BPF_PROG_RUN_ARGS, 32, 64, 96, 128, 160, 192); +EVAL6(DEFINE_BPF_PROG_RUN_ARGS, 224, 256, 288, 320, 352, 384); +EVAL4(DEFINE_BPF_PROG_RUN_ARGS, 416, 448, 480, 512); + #define PROG_NAME_LIST(stack_size) PROG_NAME(stack_size), static unsigned int (*interpreters[])(const void *ctx, @@ -1316,6 +1354,24 @@ EVAL6(PROG_NAME_LIST, 32, 64, 96, 128, 160, 192) EVAL6(PROG_NAME_LIST, 224, 256, 288, 320, 352, 384) EVAL4(PROG_NAME_LIST, 416, 448, 480, 512) }; +#undef PROG_NAME_LIST +#define PROG_NAME_LIST(stack_size) PROG_NAME_ARGS(stack_size), +static u64 (*interpreters_args[])(u64 r1, u64 r2, u64 r3, u64 r4, u64 r5, + const struct bpf_insn *insn) = { +EVAL6(PROG_NAME_LIST, 32, 64, 96, 128, 160, 192) +EVAL6(PROG_NAME_LIST, 224, 256, 288, 320, 352, 384) +EVAL4(PROG_NAME_LIST, 416, 448, 480, 512) +}; +#undef PROG_NAME_LIST + +void bpf_patch_call_args(struct bpf_insn *insn, u32 stack_depth) +{ + stack_depth = max_t(u32, stack_depth, 1); + insn->off = (s16) insn->imm; + insn->imm = interpreters_args[(round_up(stack_depth, 32) / 32) - 1] - + __bpf_call_base_args; + insn->code = BPF_JMP | BPF_CALL_ARGS; +} bool bpf_prog_array_compatible(struct bpf_array *array, const struct bpf_prog *fp) diff --git a/kernel/bpf/verifier.c b/kernel/bpf/verifier.c index df1ded3faf1d..cdc1f043c69b 100644 --- a/kernel/bpf/verifier.c +++ b/kernel/bpf/verifier.c @@ -1458,6 +1458,21 @@ static int update_stack_depth(struct bpf_verifier_env *env, return 0; } +static int get_callee_stack_depth(struct bpf_verifier_env *env, + const struct bpf_insn *insn, int idx) +{ + int start = idx + insn->imm + 1, subprog; + + subprog = find_subprog(env, start); + if (subprog < 0) { + WARN_ONCE(1, "verifier bug. No program starts at insn %d\n", + start); + return -EFAULT; + } + subprog++; + return env->subprog_stack_depth[subprog]; +} + /* check whether memory at (regno + off) is accessible for t = (read | write) * if t==write, value_regno is a register which value is stored into memory * if t==read, value_regno is a register which will receive the value from memory @@ -4997,6 +5012,24 @@ static int convert_ctx_accesses(struct bpf_verifier_env *env) return 0; } +static int fixup_call_args(struct bpf_verifier_env *env) +{ + struct bpf_prog *prog = env->prog; + struct bpf_insn *insn = prog->insnsi; + int i, depth; + + for (i = 0; i < prog->len; i++, insn++) { + if (insn->code != (BPF_JMP | BPF_CALL) || + insn->src_reg != BPF_PSEUDO_CALL) + continue; + depth = get_callee_stack_depth(env, insn, i); + if (depth < 0) + return depth; + bpf_patch_call_args(insn, depth); + } + return 0; +} + /* fixup insn->imm field of bpf_call instructions * and inline eligible helpers as explicit sequence of BPF instructions * @@ -5225,6 +5258,9 @@ skip_full_check: if (ret == 0) ret = fixup_bpf_calls(env); + if (ret == 0) + ret = fixup_call_args(env); + if (log->level && bpf_verifier_log_full(log)) ret = -ENOSPC; if (log->level && !log->ubuf) { -- cgit v1.2.3-71-gd317 From 60b58afc96c9df71871df2dbad42037757ceef26 Mon Sep 17 00:00:00 2001 From: Alexei Starovoitov Date: Thu, 14 Dec 2017 17:55:14 -0800 Subject: bpf: fix net.core.bpf_jit_enable race global bpf_jit_enable variable is tested multiple times in JITs, blinding and verifier core. The malicious root can try to toggle it while loading the programs. This race condition was accounted for and there should be no issues, but it's safer to avoid this race condition. Signed-off-by: Alexei Starovoitov Acked-by: Daniel Borkmann Signed-off-by: Daniel Borkmann --- arch/arm/net/bpf_jit_32.c | 2 +- arch/arm64/net/bpf_jit_comp.c | 2 +- arch/mips/net/ebpf_jit.c | 2 +- arch/powerpc/net/bpf_jit_comp64.c | 2 +- arch/s390/net/bpf_jit_comp.c | 2 +- arch/sparc/net/bpf_jit_comp_64.c | 2 +- arch/x86/net/bpf_jit_comp.c | 2 +- include/linux/filter.h | 5 +++-- kernel/bpf/core.c | 3 ++- kernel/bpf/verifier.c | 2 +- 10 files changed, 13 insertions(+), 11 deletions(-) (limited to 'kernel') diff --git a/arch/arm/net/bpf_jit_32.c b/arch/arm/net/bpf_jit_32.c index c199990e12b6..4425189bb24c 100644 --- a/arch/arm/net/bpf_jit_32.c +++ b/arch/arm/net/bpf_jit_32.c @@ -1824,7 +1824,7 @@ struct bpf_prog *bpf_int_jit_compile(struct bpf_prog *prog) /* If BPF JIT was not enabled then we must fall back to * the interpreter. */ - if (!bpf_jit_enable) + if (!prog->jit_requested) return orig_prog; /* If constant blinding was enabled and we failed during blinding diff --git a/arch/arm64/net/bpf_jit_comp.c b/arch/arm64/net/bpf_jit_comp.c index ba38d403abb2..288137cb0871 100644 --- a/arch/arm64/net/bpf_jit_comp.c +++ b/arch/arm64/net/bpf_jit_comp.c @@ -844,7 +844,7 @@ struct bpf_prog *bpf_int_jit_compile(struct bpf_prog *prog) int image_size; u8 *image_ptr; - if (!bpf_jit_enable) + if (!prog->jit_requested) return orig_prog; tmp = bpf_jit_blind_constants(prog); diff --git a/arch/mips/net/ebpf_jit.c b/arch/mips/net/ebpf_jit.c index 962b0259b4b6..97069a1b6f43 100644 --- a/arch/mips/net/ebpf_jit.c +++ b/arch/mips/net/ebpf_jit.c @@ -1869,7 +1869,7 @@ struct bpf_prog *bpf_int_jit_compile(struct bpf_prog *prog) unsigned int image_size; u8 *image_ptr; - if (!bpf_jit_enable || !cpu_has_mips64r2) + if (!prog->jit_requested || !cpu_has_mips64r2) return prog; tmp = bpf_jit_blind_constants(prog); diff --git a/arch/powerpc/net/bpf_jit_comp64.c b/arch/powerpc/net/bpf_jit_comp64.c index 46d74e81aff1..d5a5bc43cf8f 100644 --- a/arch/powerpc/net/bpf_jit_comp64.c +++ b/arch/powerpc/net/bpf_jit_comp64.c @@ -993,7 +993,7 @@ struct bpf_prog *bpf_int_jit_compile(struct bpf_prog *fp) struct bpf_prog *tmp_fp; bool bpf_blinded = false; - if (!bpf_jit_enable) + if (!fp->jit_requested) return org_fp; tmp_fp = bpf_jit_blind_constants(org_fp); diff --git a/arch/s390/net/bpf_jit_comp.c b/arch/s390/net/bpf_jit_comp.c index e81c16838b90..f4baa8c514d3 100644 --- a/arch/s390/net/bpf_jit_comp.c +++ b/arch/s390/net/bpf_jit_comp.c @@ -1300,7 +1300,7 @@ struct bpf_prog *bpf_int_jit_compile(struct bpf_prog *fp) struct bpf_jit jit; int pass; - if (!bpf_jit_enable) + if (!fp->jit_requested) return orig_fp; tmp = bpf_jit_blind_constants(fp); diff --git a/arch/sparc/net/bpf_jit_comp_64.c b/arch/sparc/net/bpf_jit_comp_64.c index 5765e7e711f7..a2f1b5e774a7 100644 --- a/arch/sparc/net/bpf_jit_comp_64.c +++ b/arch/sparc/net/bpf_jit_comp_64.c @@ -1517,7 +1517,7 @@ struct bpf_prog *bpf_int_jit_compile(struct bpf_prog *prog) u8 *image_ptr; int pass; - if (!bpf_jit_enable) + if (!prog->jit_requested) return orig_prog; tmp = bpf_jit_blind_constants(prog); diff --git a/arch/x86/net/bpf_jit_comp.c b/arch/x86/net/bpf_jit_comp.c index 0554e8aef4d5..68859b58ab84 100644 --- a/arch/x86/net/bpf_jit_comp.c +++ b/arch/x86/net/bpf_jit_comp.c @@ -1121,7 +1121,7 @@ struct bpf_prog *bpf_int_jit_compile(struct bpf_prog *prog) int pass; int i; - if (!bpf_jit_enable) + if (!prog->jit_requested) return orig_prog; tmp = bpf_jit_blind_constants(prog); diff --git a/include/linux/filter.h b/include/linux/filter.h index f26e6da1007b..3d6edc34932c 100644 --- a/include/linux/filter.h +++ b/include/linux/filter.h @@ -458,6 +458,7 @@ struct bpf_binary_header { struct bpf_prog { u16 pages; /* Number of allocated pages */ u16 jited:1, /* Is our filter JIT'ed? */ + jit_requested:1,/* archs need to JIT the prog */ locked:1, /* Program image locked? */ gpl_compatible:1, /* Is filter GPL compatible? */ cb_access:1, /* Is control block accessed? */ @@ -804,7 +805,7 @@ static inline bool bpf_prog_ebpf_jited(const struct bpf_prog *fp) return fp->jited && bpf_jit_is_ebpf(); } -static inline bool bpf_jit_blinding_enabled(void) +static inline bool bpf_jit_blinding_enabled(struct bpf_prog *prog) { /* These are the prerequisites, should someone ever have the * idea to call blinding outside of them, we make sure to @@ -812,7 +813,7 @@ static inline bool bpf_jit_blinding_enabled(void) */ if (!bpf_jit_is_ebpf()) return false; - if (!bpf_jit_enable) + if (!prog->jit_requested) return false; if (!bpf_jit_harden) return false; diff --git a/kernel/bpf/core.c b/kernel/bpf/core.c index dc12c4fd006e..bda911644b1c 100644 --- a/kernel/bpf/core.c +++ b/kernel/bpf/core.c @@ -94,6 +94,7 @@ struct bpf_prog *bpf_prog_alloc(unsigned int size, gfp_t gfp_extra_flags) fp->pages = size / PAGE_SIZE; fp->aux = aux; fp->aux->prog = fp; + fp->jit_requested = ebpf_jit_enabled(); INIT_LIST_HEAD_RCU(&fp->aux->ksym_lnode); @@ -721,7 +722,7 @@ struct bpf_prog *bpf_jit_blind_constants(struct bpf_prog *prog) struct bpf_insn *insn; int i, rewritten; - if (!bpf_jit_blinding_enabled()) + if (!bpf_jit_blinding_enabled(prog)) return prog; clone = bpf_prog_clone_create(prog, GFP_USER); diff --git a/kernel/bpf/verifier.c b/kernel/bpf/verifier.c index cdc1f043c69b..8e0e4cd0d5e4 100644 --- a/kernel/bpf/verifier.c +++ b/kernel/bpf/verifier.c @@ -5080,7 +5080,7 @@ static int fixup_bpf_calls(struct bpf_verifier_env *env) /* BPF_EMIT_CALL() assumptions in some of the map_gen_lookup * handlers are currently limited to 64 bit only. */ - if (ebpf_jit_enabled() && BITS_PER_LONG == 64 && + if (prog->jit_requested && BITS_PER_LONG == 64 && insn->imm == BPF_FUNC_map_lookup_elem) { map_ptr = env->insn_aux_data[i + delta].map_ptr; if (map_ptr == BPF_MAP_PTR_POISON || -- cgit v1.2.3-71-gd317 From 1c2a088a6626d4f51d2f2c97b0cbedbfbf3637f6 Mon Sep 17 00:00:00 2001 From: Alexei Starovoitov Date: Thu, 14 Dec 2017 17:55:15 -0800 Subject: bpf: x64: add JIT support for multi-function programs Typical JIT does several passes over bpf instructions to compute total size and relative offsets of jumps and calls. With multitple bpf functions calling each other all relative calls will have invalid offsets intially therefore we need to additional last pass over the program to emit calls with correct offsets. For example in case of three bpf functions: main: call foo call bpf_map_lookup exit foo: call bar exit bar: exit We will call bpf_int_jit_compile() indepedently for main(), foo() and bar() x64 JIT typically does 4-5 passes to converge. After these initial passes the image for these 3 functions will be good except call targets, since start addresses of foo() and bar() are unknown when we were JITing main() (note that call bpf_map_lookup will be resolved properly during initial passes). Once start addresses of 3 functions are known we patch call_insn->imm to point to right functions and call bpf_int_jit_compile() again which needs only one pass. Additional safety checks are done to make sure this last pass doesn't produce image that is larger or smaller than previous pass. When constant blinding is on it's applied to all functions at the first pass, since doing it once again at the last pass can change size of the JITed code. Tested on x64 and arm64 hw with JIT on/off, blinding on/off. x64 jits bpf-to-bpf calls correctly while arm64 falls back to interpreter. All other JITs that support normal BPF_CALL will behave the same way since bpf-to-bpf call is equivalent to bpf-to-kernel call from JITs point of view. Signed-off-by: Alexei Starovoitov Acked-by: Daniel Borkmann Signed-off-by: Daniel Borkmann --- arch/x86/net/bpf_jit_comp.c | 47 ++++++++++++++-- include/linux/bpf.h | 3 ++ include/linux/bpf_verifier.h | 1 + include/linux/filter.h | 2 + kernel/bpf/core.c | 13 ++++- kernel/bpf/syscall.c | 3 +- kernel/bpf/verifier.c | 126 +++++++++++++++++++++++++++++++++++++++++++ 7 files changed, 189 insertions(+), 6 deletions(-) (limited to 'kernel') diff --git a/arch/x86/net/bpf_jit_comp.c b/arch/x86/net/bpf_jit_comp.c index 68859b58ab84..87f214fbe66e 100644 --- a/arch/x86/net/bpf_jit_comp.c +++ b/arch/x86/net/bpf_jit_comp.c @@ -1109,13 +1109,23 @@ common_load: return proglen; } +struct x64_jit_data { + struct bpf_binary_header *header; + int *addrs; + u8 *image; + int proglen; + struct jit_context ctx; +}; + struct bpf_prog *bpf_int_jit_compile(struct bpf_prog *prog) { struct bpf_binary_header *header = NULL; struct bpf_prog *tmp, *orig_prog = prog; + struct x64_jit_data *jit_data; int proglen, oldproglen = 0; struct jit_context ctx = {}; bool tmp_blinded = false; + bool extra_pass = false; u8 *image = NULL; int *addrs; int pass; @@ -1135,10 +1145,28 @@ struct bpf_prog *bpf_int_jit_compile(struct bpf_prog *prog) prog = tmp; } + jit_data = prog->aux->jit_data; + if (!jit_data) { + jit_data = kzalloc(sizeof(*jit_data), GFP_KERNEL); + if (!jit_data) { + prog = orig_prog; + goto out; + } + prog->aux->jit_data = jit_data; + } + addrs = jit_data->addrs; + if (addrs) { + ctx = jit_data->ctx; + oldproglen = jit_data->proglen; + image = jit_data->image; + header = jit_data->header; + extra_pass = true; + goto skip_init_addrs; + } addrs = kmalloc(prog->len * sizeof(*addrs), GFP_KERNEL); if (!addrs) { prog = orig_prog; - goto out; + goto out_addrs; } /* Before first pass, make a rough estimation of addrs[] @@ -1149,6 +1177,7 @@ struct bpf_prog *bpf_int_jit_compile(struct bpf_prog *prog) addrs[i] = proglen; } ctx.cleanup_addr = proglen; +skip_init_addrs: /* JITed image shrinks with every pass and the loop iterates * until the image stops shrinking. Very large bpf programs @@ -1189,7 +1218,15 @@ struct bpf_prog *bpf_int_jit_compile(struct bpf_prog *prog) if (image) { bpf_flush_icache(header, image + proglen); - bpf_jit_binary_lock_ro(header); + if (!prog->is_func || extra_pass) { + bpf_jit_binary_lock_ro(header); + } else { + jit_data->addrs = addrs; + jit_data->ctx = ctx; + jit_data->proglen = proglen; + jit_data->image = image; + jit_data->header = header; + } prog->bpf_func = (void *)image; prog->jited = 1; prog->jited_len = proglen; @@ -1197,8 +1234,12 @@ struct bpf_prog *bpf_int_jit_compile(struct bpf_prog *prog) prog = orig_prog; } + if (!prog->is_func || extra_pass) { out_addrs: - kfree(addrs); + kfree(addrs); + kfree(jit_data); + prog->aux->jit_data = NULL; + } out: if (tmp_blinded) bpf_jit_prog_release_other(prog, prog == orig_prog ? diff --git a/include/linux/bpf.h b/include/linux/bpf.h index 8935f6f63d5f..da54ef644fcd 100644 --- a/include/linux/bpf.h +++ b/include/linux/bpf.h @@ -200,6 +200,9 @@ struct bpf_prog_aux { u32 max_ctx_offset; u32 stack_depth; u32 id; + u32 func_cnt; + struct bpf_prog **func; + void *jit_data; /* JIT specific data. arch dependent */ struct latch_tree_node ksym_tnode; struct list_head ksym_lnode; const struct bpf_prog_ops *ops; diff --git a/include/linux/bpf_verifier.h b/include/linux/bpf_verifier.h index 585d4e17ea88..aaac589e490c 100644 --- a/include/linux/bpf_verifier.h +++ b/include/linux/bpf_verifier.h @@ -143,6 +143,7 @@ struct bpf_insn_aux_data { union { enum bpf_reg_type ptr_type; /* pointer type for load/store insns */ struct bpf_map *map_ptr; /* pointer for call insn into lookup_elem */ + s32 call_imm; /* saved imm field of call insn */ }; int ctx_field_size; /* the ctx field size for load insn, maybe 0 */ bool seen; /* this insn was processed by the verifier */ diff --git a/include/linux/filter.h b/include/linux/filter.h index 3d6edc34932c..e872b4ebaa57 100644 --- a/include/linux/filter.h +++ b/include/linux/filter.h @@ -463,6 +463,8 @@ struct bpf_prog { gpl_compatible:1, /* Is filter GPL compatible? */ cb_access:1, /* Is control block accessed? */ dst_needed:1, /* Do we need dst entry? */ + blinded:1, /* Was blinded */ + is_func:1, /* program is a bpf function */ kprobe_override:1; /* Do we override a kprobe? */ enum bpf_prog_type type; /* Type of BPF program */ u32 len; /* Number of filter blocks */ diff --git a/kernel/bpf/core.c b/kernel/bpf/core.c index bda911644b1c..768e0a02d8c8 100644 --- a/kernel/bpf/core.c +++ b/kernel/bpf/core.c @@ -722,7 +722,7 @@ struct bpf_prog *bpf_jit_blind_constants(struct bpf_prog *prog) struct bpf_insn *insn; int i, rewritten; - if (!bpf_jit_blinding_enabled(prog)) + if (!bpf_jit_blinding_enabled(prog) || prog->blinded) return prog; clone = bpf_prog_clone_create(prog, GFP_USER); @@ -764,6 +764,7 @@ struct bpf_prog *bpf_jit_blind_constants(struct bpf_prog *prog) i += insn_delta; } + clone->blinded = 1; return clone; } #endif /* CONFIG_BPF_JIT */ @@ -1629,11 +1630,19 @@ int bpf_prog_array_copy_info(struct bpf_prog_array __rcu *array, static void bpf_prog_free_deferred(struct work_struct *work) { struct bpf_prog_aux *aux; + int i; aux = container_of(work, struct bpf_prog_aux, work); if (bpf_prog_is_dev_bound(aux)) bpf_prog_offload_destroy(aux->prog); - bpf_jit_free(aux->prog); + for (i = 0; i < aux->func_cnt; i++) + bpf_jit_free(aux->func[i]); + if (aux->func_cnt) { + kfree(aux->func); + bpf_prog_unlock_free(aux->prog); + } else { + bpf_jit_free(aux->prog); + } } /* Free internal BPF program */ diff --git a/kernel/bpf/syscall.c b/kernel/bpf/syscall.c index 2c4cfeaa8d5e..e2e1c78ce1dc 100644 --- a/kernel/bpf/syscall.c +++ b/kernel/bpf/syscall.c @@ -1194,7 +1194,8 @@ static int bpf_prog_load(union bpf_attr *attr) goto free_used_maps; /* eBPF program is ready to be JITed */ - prog = bpf_prog_select_runtime(prog, &err); + if (!prog->bpf_func) + prog = bpf_prog_select_runtime(prog, &err); if (err < 0) goto free_used_maps; diff --git a/kernel/bpf/verifier.c b/kernel/bpf/verifier.c index 8e0e4cd0d5e4..48b2901cf483 100644 --- a/kernel/bpf/verifier.c +++ b/kernel/bpf/verifier.c @@ -5012,12 +5012,138 @@ static int convert_ctx_accesses(struct bpf_verifier_env *env) return 0; } +static int jit_subprogs(struct bpf_verifier_env *env) +{ + struct bpf_prog *prog = env->prog, **func, *tmp; + int i, j, subprog_start, subprog_end = 0, len, subprog; + struct bpf_insn *insn = prog->insnsi; + void *old_bpf_func; + int err = -ENOMEM; + + if (env->subprog_cnt == 0) + return 0; + + for (i = 0; i < prog->len; i++, insn++) { + if (insn->code != (BPF_JMP | BPF_CALL) || + insn->src_reg != BPF_PSEUDO_CALL) + continue; + subprog = find_subprog(env, i + insn->imm + 1); + if (subprog < 0) { + WARN_ONCE(1, "verifier bug. No program starts at insn %d\n", + i + insn->imm + 1); + return -EFAULT; + } + /* temporarily remember subprog id inside insn instead of + * aux_data, since next loop will split up all insns into funcs + */ + insn->off = subprog + 1; + /* remember original imm in case JIT fails and fallback + * to interpreter will be needed + */ + env->insn_aux_data[i].call_imm = insn->imm; + /* point imm to __bpf_call_base+1 from JITs point of view */ + insn->imm = 1; + } + + func = kzalloc(sizeof(prog) * (env->subprog_cnt + 1), GFP_KERNEL); + if (!func) + return -ENOMEM; + + for (i = 0; i <= env->subprog_cnt; i++) { + subprog_start = subprog_end; + if (env->subprog_cnt == i) + subprog_end = prog->len; + else + subprog_end = env->subprog_starts[i]; + + len = subprog_end - subprog_start; + func[i] = bpf_prog_alloc(bpf_prog_size(len), GFP_USER); + if (!func[i]) + goto out_free; + memcpy(func[i]->insnsi, &prog->insnsi[subprog_start], + len * sizeof(struct bpf_insn)); + func[i]->len = len; + func[i]->is_func = 1; + /* Use bpf_prog_F_tag to indicate functions in stack traces. + * Long term would need debug info to populate names + */ + func[i]->aux->name[0] = 'F'; + func[i]->aux->stack_depth = env->subprog_stack_depth[i]; + func[i]->jit_requested = 1; + func[i] = bpf_int_jit_compile(func[i]); + if (!func[i]->jited) { + err = -ENOTSUPP; + goto out_free; + } + cond_resched(); + } + /* at this point all bpf functions were successfully JITed + * now populate all bpf_calls with correct addresses and + * run last pass of JIT + */ + for (i = 0; i <= env->subprog_cnt; i++) { + insn = func[i]->insnsi; + for (j = 0; j < func[i]->len; j++, insn++) { + if (insn->code != (BPF_JMP | BPF_CALL) || + insn->src_reg != BPF_PSEUDO_CALL) + continue; + subprog = insn->off; + insn->off = 0; + insn->imm = (u64 (*)(u64, u64, u64, u64, u64)) + func[subprog]->bpf_func - + __bpf_call_base; + } + } + for (i = 0; i <= env->subprog_cnt; i++) { + old_bpf_func = func[i]->bpf_func; + tmp = bpf_int_jit_compile(func[i]); + if (tmp != func[i] || func[i]->bpf_func != old_bpf_func) { + verbose(env, "JIT doesn't support bpf-to-bpf calls\n"); + err = -EFAULT; + goto out_free; + } + cond_resched(); + } + + /* finally lock prog and jit images for all functions and + * populate kallsysm + */ + for (i = 0; i <= env->subprog_cnt; i++) { + bpf_prog_lock_ro(func[i]); + bpf_prog_kallsyms_add(func[i]); + } + prog->jited = 1; + prog->bpf_func = func[0]->bpf_func; + prog->aux->func = func; + prog->aux->func_cnt = env->subprog_cnt + 1; + return 0; +out_free: + for (i = 0; i <= env->subprog_cnt; i++) + if (func[i]) + bpf_jit_free(func[i]); + kfree(func); + /* cleanup main prog to be interpreted */ + prog->jit_requested = 0; + for (i = 0, insn = prog->insnsi; i < prog->len; i++, insn++) { + if (insn->code != (BPF_JMP | BPF_CALL) || + insn->src_reg != BPF_PSEUDO_CALL) + continue; + insn->off = 0; + insn->imm = env->insn_aux_data[i].call_imm; + } + return err; +} + static int fixup_call_args(struct bpf_verifier_env *env) { struct bpf_prog *prog = env->prog; struct bpf_insn *insn = prog->insnsi; int i, depth; + if (env->prog->jit_requested) + if (jit_subprogs(env) == 0) + return 0; + for (i = 0; i < prog->len; i++, insn++) { if (insn->code != (BPF_JMP | BPF_CALL) || insn->src_reg != BPF_PSEUDO_CALL) -- cgit v1.2.3-71-gd317