bpf, x64: Fix tailcall infinite loop
Subsystems (6)
v2 cover bpf, x64: Fix tailcall infinite loop
This patch series fixes a tailcall infinite loop.
From commit ebf7d1f508a73871 ("bpf, x64: rework pro/epilogue and tailcall handling in JIT"), the tailcall on x64 works better than before.
From commit e411901c0b775a3a ("bpf: allow for tailcalls in BPF subprograms for x64 JIT"), tailcall is able to run in BPF subprograms on x64.
From commit 5b92a28aae4dd0f8 ("bpf: Support attaching tracing BPF program to other BPF programs"), BPF program is able to trace other BPF programs.
How about combining them all together?
1. FENTRY/FEXIT on a BPF subprogram.
2. A tailcall runs in the BPF subprogram.
3. The tailcall calls itself.
As a result, a tailcall infinite loop comes up. And the loop would halt the machine.
As we know, in tail call context, the tail_call_cnt propagates by stack and RAX register between BPF subprograms. So do it in trampolines.
How did I discover the bug?
From commit 7f6e4312e15a5c37 ("bpf: Limit caller's stack depth 256 for subprogs with tailcalls"), the total stack size limits to around 8KiB. Then, I write some bpf progs to validate the stack consuming, that are tailcalls running in bpf2bpf and FENTRY/FEXIT tracing on bpf2bpf[1].
At that time, accidently, I made a tailcall loop. And then the loop halted my VM. Without the loop, the bpf progs would consume over 8KiB stack size. But the _stack-overflow_ did not halt my VM.
With bpf_printk(), I confirmed that the tailcall count limit did not work expectedly. Next, read the code and fix it.
Finally, unfortunately, I only fix it on x64 but other arches. As a result, CI tests failed because this bug hasn't been fixed on s390x.
Some helps are requested.
[1]: https://github.com/Asphaltt/learn-by-example/tree/main/ebpf/tailcall-stackoverflow
Leon Hwang (2):
bpf, x64: Fix tailcall infinite loop
selftests/bpf: Add testcases for tailcall infinite loop fixing
arch/x86/net/bpf_jit_comp.c | 40 +++- include/linux/bpf.h | 5 + kernel/bpf/trampoline.c | 4 +- kernel/bpf/verifier.c | 31 ++- .../selftests/bpf/prog_tests/tailcalls.c | 194 +++++++++++++++++- .../bpf/progs/tailcall_bpf2bpf_fentry.c | 18 ++ .../bpf/progs/tailcall_bpf2bpf_fexit.c | 18 ++ 7 files changed, 292 insertions(+), 18 deletions(-) create mode 100644 tools/testing/selftests/bpf/progs/tailcall_bpf2bpf_fentry.c create mode 100644 tools/testing/selftests/bpf/progs/tailcall_bpf2bpf_fexit.c
base-commit: 9930e4af4b509bcf6f060b09b16884f26102d110
Signature · 1 line
2.41.0
RFC PATCH v2 1/2 bpf, x64: Fix tailcall infinite loop
From commit ebf7d1f508a73871 ("bpf, x64: rework pro/epilogue and tailcall handling in JIT"), the tailcall on x64 works better than before.
From commit e411901c0b775a3a ("bpf: allow for tailcalls in BPF subprograms for x64 JIT"), tailcall is able to run in BPF subprograms on x64.
From commit 5b92a28aae4dd0f8 ("bpf: Support attaching tracing BPF program to other BPF programs"), BPF program is able to trace other BPF programs.
How about combining them all together?
1. FENTRY/FEXIT on a BPF subprogram.
2. A tailcall runs in the BPF subprogram.
3. The tailcall calls itself.
As a result, a tailcall infinite loop comes up. And the loop would halt the machine.
As we know, in tail call context, the tail_call_cnt propagates by stack and RAX register between BPF subprograms. So do it in trampolines.
--- arch/x86/net/bpf_jit_comp.c | 40 +++++++++++++++++++++++++++++-------- include/linux/bpf.h | 5 +++++ kernel/bpf/trampoline.c | 4 ++-- kernel/bpf/verifier.c | 31 +++++++++++++++++++++------- 4 files changed, 63 insertions(+), 17 deletions(-)
arch/x86/net/bpf_jit_comp.c +32 -8
@@ -303,8 +303,12 @@ static void emit_prologue(u8 **pprog, u32 stack_depth, bool ebpf_from_cbpf, prog +=X86_PATCH_SIZE ;if (!ebpf_from_cbpf) {if (tail_call_reachable && !is_subprog) +/* When it's the entry of the whole tailcall context, +* zeroing rax means initialising tail_call_cnt. +*/ EMIT2 (0x31 ,0xC0 );/* xor eax, eax */ else +// Keep the same instruction layout. EMIT2 (0x66 ,0x90 );/* nop2 */ }EMIT1 (0x55 );/* push rbp */
@@ -1018,6 +1022,10 @@ static void emit_shiftx(u8 **pprog, u32 dst_reg, u8 src_reg, bool is64, u8 op)#define INSN_SZ_DIFF (((addrs[i] - addrs[i - 1]) - (prog - temp))) +/* mov rax, qword ptr [rbp - rounded_stack_depth - 8] */ +#define RESTORE_TAIL_CALL_CNT (stack) \ + EMIT3_off32(0x48, 0x8B, 0x85, -round_up(stack, 8) - 8) +static int do_jit (struct bpf_prog *bpf_prog,int *addrs,u8 *image,u8 *rw_image,int oldproglen,struct jit_context *ctx,bool jmp_padding) {
@@ -1623,9 +1631,7 @@ st: if (is_imm8(insn->off)) func = (u8 *) __bpf_call_base + imm32;if (tail_call_reachable) { -/* mov rax, qword ptr [rbp - rounded_stack_depth - 8] */ -EMIT3_off32 (0x48 ,0x8B ,0x85 , - -round_up (bpf_prog->aux->stack_depth,8 ) -8 ); +RESTORE_TAIL_CALL_CNT (bpf_prog->aux->stack_depth);if (!imm32)return -EINVAL ; offs =7 +x86_call_depth_emit_accounting (&prog, func);
@@ -2298,7 +2304,9 @@ static int invoke_bpf_mod_ret(const struct btf_func_model *m, u8 **pprog,* push rbp * mov rbp, rsp * sub rsp, 16 // space for skb and dev -* push rbx // temp regs to pass start time +* mov qword ptr [rbp - 40], rbx // temp regs to pass start time +* mov rax, 2 // cache number of argument to rax +* mov qword ptr [rbp - 32], rax // save number of argument to stack * mov qword ptr [rbp - 16], rdi // save skb pointer to stack * mov qword ptr [rbp - 8], rsi // save dev pointer to stack * call __bpf_prog_enter // rcu_read_lock and preempt_disable
@@ -2323,7 +2331,9 @@ static int invoke_bpf_mod_ret(const struct btf_func_model *m, u8 **pprog,* push rbp * mov rbp, rsp * sub rsp, 24 // space for skb, dev, return value -* push rbx // temp regs to pass start time +* mov qword ptr [rbp - 40], rbx // temp regs to pass start time +* mov rax, 2 // cache number of argument to rax +* mov qword ptr [rbp - 32], rax // save number of argument to stack * mov qword ptr [rbp - 24], rdi // save skb pointer to stack * mov qword ptr [rbp - 16], rsi // save dev pointer to stack * call __bpf_prog_enter // rcu_read_lock and preempt_disable
@@ -2400,6 +2410,7 @@ int arch_prepare_bpf_trampoline(struct bpf_tramp_image *im, void *image, void *i* [ ... ] * [ stack_arg2 ] * RBP - arg_stack_off [ stack_arg1 ] +* RSP [ tail_call_cnt ] BPF_TRAMP_F_TAIL_CALL_CTX */ /* room for return value of orig_call or fentry prog */
@@ -2464,6 +2475,8 @@ int arch_prepare_bpf_trampoline(struct bpf_tramp_image *im, void *image, void *ielse /* sub rsp, stack_size */ EMIT4 (0x48 ,0x83 ,0xEC , stack_size); +if (flags &BPF_TRAMP_F_TAIL_CALL_CTX ) +EMIT1 (0x50 );/* push rax */ /* mov QWORD PTR [rbp - rbx_off], rbx */ emit_stx (&prog,BPF_DW ,BPF_REG_FP ,BPF_REG_6 , -rbx_off);
@@ -2516,9 +2529,15 @@ int arch_prepare_bpf_trampoline(struct bpf_tramp_image *im, void *image, void *irestore_regs (m, &prog, regs_off);save_args (m, &prog, arg_stack_off, true); +if (flags &BPF_TRAMP_F_TAIL_CALL_CTX ) +/* Before calling the original function, restore the +* tail_call_cnt from stack to rax. +*/ +RESTORE_TAIL_CALL_CNT (stack_size); +if (flags &BPF_TRAMP_F_ORIG_STACK ) { -emit_ldx (&prog,BPF_DW ,BPF_REG_0 ,BPF_REG_FP ,8 ); -EMIT2 (0xff ,0xd0 );/* call *rax */ +emit_ldx (&prog,BPF_DW ,BPF_REG_6 ,BPF_REG_FP ,8 ); +EMIT2 (0xff ,0xd3 );/* call *rbx */ // FIXME: Confirm 0xd3? }else {/* call original function */ if (emit_rsb_call (&prog, orig_call, prog)) {
@@ -2569,7 +2588,12 @@ int arch_prepare_bpf_trampoline(struct bpf_tramp_image *im, void *image, void *i ret = -EINVAL ; goto cleanup; } - } + }else if (flags &BPF_TRAMP_F_TAIL_CALL_CTX ) +/* Before running the original function, restore the +* tail_call_cnt from stack to rax. +*/ +RESTORE_TAIL_CALL_CNT (stack_size); +/* restore return value of orig_call or fentry prog back into RAX */ if (save_ret)emit_ldx (&prog,BPF_DW ,BPF_REG_0 ,BPF_REG_FP ,-8 );
include/linux/bpf.h +5 -0
@@ -1028,6 +1028,11 @@ struct btf_func_model {*/ #define BPF_TRAMP_F_SHARE_IPMODIFY BIT(6) +/* Indicate that current trampoline is in a tail call context. Then, it has to +* cache and restore tail_call_cnt to avoid infinite tail call loop. +*/ +#define BPF_TRAMP_F_TAIL_CALL_CTX BIT(7) +/* Each call __bpf_prog_enter + call bpf_func + call __bpf_prog_exit is ~50 * bytes on x86. */
kernel/bpf/trampoline.c +2 -2
@@ -415,8 +415,8 @@ static int bpf_trampoline_update(struct bpf_trampoline *tr, bool lock_direct_mut goto out; } -/* clear all bits except SHARE_IPMODIFY */ - tr->flags &=BPF_TRAMP_F_SHARE_IPMODIFY ; +/* clear all bits except SHARE_IPMODIFY and TAIL_CALL_CTX */ + tr->flags &= (BPF_TRAMP_F_SHARE_IPMODIFY |BPF_TRAMP_F_TAIL_CALL_CTX );if (tlinks[BPF_TRAMP_FEXIT ].nr_links || tlinks[BPF_TRAMP_MODIFY_RETURN ].nr_links) {
kernel/bpf/verifier.c +24 -7
@@ -19246,6 +19246,21 @@ static int check_non_sleepable_error_inject(u32 btf_id)return btf_id_set_contains (&btf_non_sleepable_error_inject, btf_id); } +static inline int find_subprog_index (const struct bpf_prog *prog, +u32 btf_id) +{ +struct bpf_prog_aux *aux = prog->aux; +int i, subprog =-1 ; + +for (i =0 ; i < aux->func_info_cnt; i++) +if (aux->func_info[i].type_id == btf_id) { + subprog = i; +break ; + } + +return subprog; +} + int bpf_check_attach_target(struct bpf_verifier_log *log, const struct bpf_prog *prog, const struct bpf_prog *tgt_prog,
@@ -19254,9 +19269,9 @@ int bpf_check_attach_target(struct bpf_verifier_log *log, {bool prog_extension = prog->type ==BPF_PROG_TYPE_EXT ;const char prefix[] ="btf_trace_" ; -int ret =0 , subprog =-1 , i;const struct btf_type *t;bool conservative = true; +int ret =0 , subprog;const char *tname;struct btf *btf;long addr =0 ;
@@ -19291,11 +19306,7 @@ int bpf_check_attach_target(struct bpf_verifier_log *log,return -EINVAL ; } -for (i =0 ; i < aux->func_info_cnt; i++) -if (aux->func_info[i].type_id == btf_id) { - subprog = i; -break ; - } + subprog =find_subprog_index (tgt_prog, btf_id);if (subprog ==-1 ) {bpf_log (log,"Subprog %s doesn't exist\n" , tname);return -EINVAL ;
@@ -19559,7 +19570,7 @@ static int check_attach_btf_id(struct bpf_verifier_env *env)struct bpf_attach_target_info tgt_info = {};u32 btf_id = prog->aux->attach_btf_id;struct bpf_trampoline *tr; -int ret; +int ret, subprog;u64 key;if (prog->type ==BPF_PROG_TYPE_SYSCALL ) {
@@ -19629,6 +19640,12 @@ static int check_attach_btf_id(struct bpf_verifier_env *env)if (!tr)return -ENOMEM ; +if (tgt_prog && tgt_prog->aux->tail_call_reachable) { + subprog =find_subprog_index (tgt_prog, btf_id); + tr->flags = subprog >0 && tgt_prog->aux->func[subprog]->is_func ? +BPF_TRAMP_F_TAIL_CALL_CTX :0 ; + } + prog->aux->dst_trampoline = tr;return 0 ; }
Signature · 1 line
2.41.0
Quoted 112 lines · Leon Hwang
From commit
ebf7d1f508a73871("bpf, x64: rework pro/epilogue and tailcall handling in JIT"), the tailcall on x64 works better than before.From commit
e411901c0b775a3a("bpf: allow for tailcalls in BPF subprograms for x64 JIT"), tailcall is able to run in BPF subprograms on x64.From commit
5b92a28aae4dd0f8("bpf: Support attaching tracing BPF program to other BPF programs"), BPF program is able to trace other BPF programs.How about combining them all together?
1. FENTRY/FEXIT on a BPF subprogram.
2. A tailcall runs in the BPF subprogram.
3. The tailcall calls itself.As a result, a tailcall infinite loop comes up. And the loop would halt the machine.
As we know, in tail call context, the
tail_call_cntpropagates by stack and RAX register between BPF subprograms. So do it in trampolines.Signed-off-by Leon Hwang <[email protected]>--- arch/x86/net/bpf_jit_comp.c | 40 +++++++++++++++++++++++++++++-------- include/linux/bpf.h | 5 +++++ kernel/bpf/trampoline.c | 4 ++-- kernel/bpf/verifier.c | 31 +++++++++++++++++++++------- 4 files changed, 63 insertions(+), 17 deletions(-)arch/x86/net/bpf_jit_comp.c +19
-5@@ -303,8 +303,12 @@ static void emit_prologue(u8 **pprog, u32 stack_depth, bool ebpf_from_cbpf, prog +=X86_PATCH_SIZE ;if (!ebpf_from_cbpf) {if (tail_call_reachable && !is_subprog) +/* When it's the entry of the whole tailcall context, +* zeroing rax means initialising tail_call_cnt. +*/ EMIT2 (0x31 ,0xC0 );/* xor eax, eax */ else +// Keep the same instruction layout. EMIT2 (0x66 ,0x90 );/* nop2 */ }EMIT1 (0x55 );/* push rbp */ @@ -1018,6 +1022,10 @@ static void emit_shiftx(u8 **pprog, u32 dst_reg, u8 src_reg, bool is64, u8 op)#define INSN_SZ_DIFF (((addrs[i] - addrs[i - 1]) - (prog - temp))) +/* mov rax, qword ptr [rbp - rounded_stack_depth - 8] */ +#define RESTORE_TAIL_CALL_CNT (stack) \ + EMIT3_off32(0x48, 0x8B, 0x85, -round_up(stack, 8) - 8) +static int do_jit (struct bpf_prog *bpf_prog,int *addrs,u8 *image,u8 *rw_image,int oldproglen,struct jit_context *ctx,bool jmp_padding) {@@ -1623,9 +1631,7 @@ st: if (is_imm8(insn->off)) func = (u8 *) __bpf_call_base + imm32;if (tail_call_reachable) { -/* mov rax, qword ptr [rbp - rounded_stack_depth - 8] */ -EMIT3_off32 (0x48 ,0x8B ,0x85 , - -round_up (bpf_prog->aux->stack_depth,8 ) -8 ); +RESTORE_TAIL_CALL_CNT (bpf_prog->aux->stack_depth);if (!imm32)return -EINVAL ; offs =7 +x86_call_depth_emit_accounting (&prog, func);@@ -2298,7 +2304,9 @@ static int invoke_bpf_mod_ret(const struct btf_func_model *m, u8 **pprog,* push rbp * mov rbp, rsp * sub rsp, 16 // space for skb and dev -* push rbx // temp regs to pass start time +* mov qword ptr [rbp - 40], rbx // temp regs to pass start time +* mov rax, 2 // cache number of argument to rax +* mov qword ptr [rbp - 32], rax // save number of argument to stack * mov qword ptr [rbp - 16], rdi // save skb pointer to stack * mov qword ptr [rbp - 8], rsi // save dev pointer to stack * call __bpf_prog_enter // rcu_read_lock and preempt_disable @@ -2323,7 +2331,9 @@ static int invoke_bpf_mod_ret(const struct btf_func_model *m, u8 **pprog,* push rbp * mov rbp, rsp * sub rsp, 24 // space for skb, dev, return value -* push rbx // temp regs to pass start time +* mov qword ptr [rbp - 40], rbx // temp regs to pass start time +* mov rax, 2 // cache number of argument to rax +* mov qword ptr [rbp - 32], rax // save number of argument to stack * mov qword ptr [rbp - 24], rdi // save skb pointer to stack * mov qword ptr [rbp - 16], rsi // save dev pointer to stack * call __bpf_prog_enter // rcu_read_lock and preempt_disable @@ -2400,6 +2410,7 @@ int arch_prepare_bpf_trampoline(struct bpf_tramp_image *im, void *image, void *i* [ ... ] * [ stack_arg2 ] * RBP - arg_stack_off [ stack_arg1 ] +* RSP [ tail_call_cnt ] BPF_TRAMP_F_TAIL_CALL_CTX */ /* room for return value of orig_call or fentry prog */ @@ -2464,6 +2475,8 @@ int arch_prepare_bpf_trampoline(struct bpf_tramp_image *im, void *image, void *ielse /* sub rsp, stack_size */ EMIT4 (0x48 ,0x83 ,0xEC , stack_size); +if (flags &BPF_TRAMP_F_TAIL_CALL_CTX ) +EMIT1 (0x50 );/* push rax */ /* mov QWORD PTR [rbp - rbx_off], rbx */ emit_stx (&prog,BPF_DW ,BPF_REG_FP ,BPF_REG_6 , -rbx_off);@@ -2516,9 +2529,15 @@ int arch_prepare_bpf_trampoline(struct bpf_tramp_image *im, void *image, void *irestore_regs (m, &prog, regs_off);save_args (m, &prog, arg_stack_off, true); +if (flags &BPF_TRAMP_F_TAIL_CALL_CTX )
arch/x86/net/bpf_jit_comp.c int arch_prepare_bpf_trampoline(struct bpf_tramp_image *im, void *image, void *i · line 2540
Quoted 10 lines
arch/x86/net/bpf_jit_comp.c +7
-2+/* Before calling the original function, restore the +* tail_call_cnt from stack to rax. +*/ +RESTORE_TAIL_CALL_CNT (stack_size); +if (flags &BPF_TRAMP_F_ORIG_STACK ) { -emit_ldx (&prog,BPF_DW ,BPF_REG_0 ,BPF_REG_FP ,8 ); -EMIT2 (0xff ,0xd0 );/* call *rax */ +emit_ldx (&prog,BPF_DW ,BPF_REG_6 ,BPF_REG_FP ,8 ); +EMIT2 (0xff ,0xd3 );/* call *rbx */ // FIXME: Confirm 0xd3?
To avoiding rax conflict with tail call, change the calling register from rax to rbx
But, I'm unable to confirm the opcode.
Then, I asked chatGPT to list call and corresponding opcode:
Certainly! Here's a table that provides call instructions along with their corresponding opcodes in x86-64 assembly:
| `call` Register | Opcode (Hex) | Opcode (Binary) | |-----------------|--------------|-----------------| | `rax` | `FF D0` | `11111111 11010000` | | `rcx` | `FF D1` | `11111111 11010001` | | `rdx` | `FF D2` | `11111111 11010010` | | `rbx` | `FF D3` | `11111111 11010011` | | `rsp` | `FF D4` | `11111111 11010100` | | `rbp` | `FF D5` | `11111111 11010101` | | `rsi` | `FF D6` | `11111111 11010110` | | `rdi` | `FF D7` | `11111111 11010111` | | `r8` | `41 FF D0` | `01000001 11111111 11010000` | | `r9` | `41 FF D1` | `01000001 11111111 11010001` | | `r10` | `41 FF D2` | `01000001 11111111 11010010` | | `r11` | `41 FF D3` | `01000001 11111111 11010011` | | `r12` | `41 FF D4` | `01000001 11111111 11010100` | | `r13` | `41 FF D5` | `01000001 11111111 11010101` | | `r14` | `41 FF D6` | `01000001 11111111 11010110` | | `r15` | `41 FF D7` | `01000001 11111111 11010111` |
EMIT2(0xff, 0xd3); /* call *rbx */, is it right?
Thanks,
Leon
[...]
Quoted 35 lines · Leon Hwang
From commit
ebf7d1f508a73871("bpf, x64: rework pro/epilogue and tailcall handling in JIT"), the tailcall on x64 works better than before.From commit
e411901c0b775a3a("bpf: allow for tailcalls in BPF subprograms for x64 JIT"), tailcall is able to run in BPF subprograms on x64.From commit
5b92a28aae4dd0f8("bpf: Support attaching tracing BPF program to other BPF programs"), BPF program is able to trace other BPF programs.How about combining them all together?
1. FENTRY/FEXIT on a BPF subprogram.
2. A tailcall runs in the BPF subprogram.
3. The tailcall calls itself.As a result, a tailcall infinite loop comes up. And the loop would halt the machine.
As we know, in tail call context, the
tail_call_cntpropagates by stack and RAX register between BPF subprograms. So do it in trampolines.Signed-off-by Leon Hwang <[email protected]>--- arch/x86/net/bpf_jit_comp.c | 40 +++++++++++++++++++++++++++++-------- include/linux/bpf.h | 5 +++++ kernel/bpf/trampoline.c | 4 ++-- kernel/bpf/verifier.c | 31 +++++++++++++++++++++------- 4 files changed, 63 insertions(+), 17 deletions(-)arch/x86/net/bpf_jit_comp.c +0
-0
arch/x86/net/bpf_jit_comp.c static void emit_prologue(u8 **pprog, u32 stack_depth, bool ebpf_from_cbpf, · line 311
Quoted 9 lines
arch/x86/net/bpf_jit_comp.c +4
-0@@ -303,8 +303,12 @@ static void emit_prologue(u8 **pprog, u32 stack_depth, bool ebpf_from_cbpf, prog +=X86_PATCH_SIZE ;if (!ebpf_from_cbpf) {if (tail_call_reachable && !is_subprog) +/* When it's the entry of the whole tailcall context, +* zeroing rax means initialising tail_call_cnt. +*/ EMIT2 (0x31 ,0xC0 );/* xor eax, eax */ else +// Keep the same instruction layout.
No c++ style comments please.
Quoted 26 lines
arch/x86/net/bpf_jit_comp.c +5
-3EMIT2 (0x66 ,0x90 );/* nop2 */ }EMIT1 (0x55 );/* push rbp */ @@ -1018,6 +1022,10 @@ static void emit_shiftx(u8 **pprog, u32 dst_reg, u8 src_reg, bool is64, u8 op)#define INSN_SZ_DIFF (((addrs[i] - addrs[i - 1]) - (prog - temp))) +/* mov rax, qword ptr [rbp - rounded_stack_depth - 8] */ +#define RESTORE_TAIL_CALL_CNT (stack) \ + EMIT3_off32(0x48, 0x8B, 0x85, -round_up(stack, 8) - 8) +static int do_jit (struct bpf_prog *bpf_prog,int *addrs,u8 *image,u8 *rw_image,int oldproglen,struct jit_context *ctx,bool jmp_padding) {@@ -1623,9 +1631,7 @@ st: if (is_imm8(insn->off)) func = (u8 *) __bpf_call_base + imm32;if (tail_call_reachable) { -/* mov rax, qword ptr [rbp - rounded_stack_depth - 8] */ -EMIT3_off32 (0x48 ,0x8B ,0x85 , - -round_up (bpf_prog->aux->stack_depth,8 ) -8 ); +RESTORE_TAIL_CALL_CNT (bpf_prog->aux->stack_depth);if (!imm32)return -EINVAL ; offs =7 +x86_call_depth_emit_accounting (&prog, func);
arch/x86/net/bpf_jit_comp.c static int invoke_bpf_mod_ret(const struct btf_func_model *m, u8 **pprog, · line 2308
Quoted 6 lines
arch/x86/net/bpf_jit_comp.c +2
-1@@ -2298,7 +2304,9 @@ static int invoke_bpf_mod_ret(const struct btf_func_model *m, u8 **pprog,* push rbp * mov rbp, rsp * sub rsp, 16 // space for skb and dev -* push rbx // temp regs to pass start time +* mov qword ptr [rbp - 40], rbx // temp regs to pass start time +* mov rax, 2 // cache number of argument to rax
What does it mean?
arch/x86/net/bpf_jit_comp.c static int invoke_bpf_mod_ret(const struct btf_func_model *m, u8 **pprog, · line 2309
Quoted 1 line
arch/x86/net/bpf_jit_comp.c +1
-0+* mov qword ptr [rbp - 32], rax // save number of argument to stack
Here // is ok since it's inside /* */
Quoted 36 lines
arch/x86/net/bpf_jit_comp.c +7
-1* mov qword ptr [rbp - 16], rdi // save skb pointer to stack * mov qword ptr [rbp - 8], rsi // save dev pointer to stack * call __bpf_prog_enter // rcu_read_lock and preempt_disable @@ -2323,7 +2331,9 @@ static int invoke_bpf_mod_ret(const struct btf_func_model *m, u8 **pprog,* push rbp * mov rbp, rsp * sub rsp, 24 // space for skb, dev, return value -* push rbx // temp regs to pass start time +* mov qword ptr [rbp - 40], rbx // temp regs to pass start time +* mov rax, 2 // cache number of argument to rax +* mov qword ptr [rbp - 32], rax // save number of argument to stack * mov qword ptr [rbp - 24], rdi // save skb pointer to stack * mov qword ptr [rbp - 16], rsi // save dev pointer to stack * call __bpf_prog_enter // rcu_read_lock and preempt_disable @@ -2400,6 +2410,7 @@ int arch_prepare_bpf_trampoline(struct bpf_tramp_image *im, void *image, void *i* [ ... ] * [ stack_arg2 ] * RBP - arg_stack_off [ stack_arg1 ] +* RSP [ tail_call_cnt ] BPF_TRAMP_F_TAIL_CALL_CTX */ /* room for return value of orig_call or fentry prog */ @@ -2464,6 +2475,8 @@ int arch_prepare_bpf_trampoline(struct bpf_tramp_image *im, void *image, void *ielse /* sub rsp, stack_size */ EMIT4 (0x48 ,0x83 ,0xEC , stack_size); +if (flags &BPF_TRAMP_F_TAIL_CALL_CTX ) +EMIT1 (0x50 );/* push rax */ /* mov QWORD PTR [rbp - rbx_off], rbx */ emit_stx (&prog,BPF_DW ,BPF_REG_FP ,BPF_REG_6 , -rbx_off);@@ -2516,9 +2529,15 @@ int arch_prepare_bpf_trampoline(struct bpf_tramp_image *im, void *image, void *irestore_regs (m, &prog, regs_off);save_args (m, &prog, arg_stack_off, true); +if (flags &BPF_TRAMP_F_TAIL_CALL_CTX )
arch/x86/net/bpf_jit_comp.c int arch_prepare_bpf_trampoline(struct bpf_tramp_image *im, void *image, void *i · line 2540
Quoted 10 lines
arch/x86/net/bpf_jit_comp.c +7
-2+/* Before calling the original function, restore the +* tail_call_cnt from stack to rax. +*/ +RESTORE_TAIL_CALL_CNT (stack_size); +if (flags &BPF_TRAMP_F_ORIG_STACK ) { -emit_ldx (&prog,BPF_DW ,BPF_REG_0 ,BPF_REG_FP ,8 ); -EMIT2 (0xff ,0xd0 );/* call *rax */ +emit_ldx (&prog,BPF_DW ,BPF_REG_6 ,BPF_REG_FP ,8 ); +EMIT2 (0xff ,0xd3 );/* call *rbx */ // FIXME: Confirm 0xd3?
please no FIXME like comments.
You have to be confident in the code you're submitting. llvm-mc -triple=x86_64 -show-encoding -x86-asm-syntax=intel -output-asm-variant=1 <<< 'call rbx'
Quoted 108 lines
arch/x86/net/bpf_jit_comp.c +6
-1}else {/* call original function */ if (emit_rsb_call (&prog, orig_call, prog)) {@@ -2569,7 +2588,12 @@ int arch_prepare_bpf_trampoline(struct bpf_tramp_image *im, void *image, void *i ret = -EINVAL ; goto cleanup; } - } + }else if (flags &BPF_TRAMP_F_TAIL_CALL_CTX ) +/* Before running the original function, restore the +* tail_call_cnt from stack to rax. +*/ +RESTORE_TAIL_CALL_CNT (stack_size); +/* restore return value of orig_call or fentry prog back into RAX */ if (save_ret)emit_ldx (&prog,BPF_DW ,BPF_REG_0 ,BPF_REG_FP ,-8 );include/linux/bpf.h +5
-0@@ -1028,6 +1028,11 @@ struct btf_func_model {*/ #define BPF_TRAMP_F_SHARE_IPMODIFY BIT(6) +/* Indicate that current trampoline is in a tail call context. Then, it has to +* cache and restore tail_call_cnt to avoid infinite tail call loop. +*/ +#define BPF_TRAMP_F_TAIL_CALL_CTX BIT(7) +/* Each call __bpf_prog_enter + call bpf_func + call __bpf_prog_exit is ~50 * bytes on x86. */ kernel/bpf/trampoline.c +2
-2@@ -415,8 +415,8 @@ static int bpf_trampoline_update(struct bpf_trampoline *tr, bool lock_direct_mut goto out; } -/* clear all bits except SHARE_IPMODIFY */ - tr->flags &=BPF_TRAMP_F_SHARE_IPMODIFY ; +/* clear all bits except SHARE_IPMODIFY and TAIL_CALL_CTX */ + tr->flags &= (BPF_TRAMP_F_SHARE_IPMODIFY |BPF_TRAMP_F_TAIL_CALL_CTX );if (tlinks[BPF_TRAMP_FEXIT ].nr_links || tlinks[BPF_TRAMP_MODIFY_RETURN ].nr_links) {kernel/bpf/verifier.c +18
-7@@ -19246,6 +19246,21 @@ static int check_non_sleepable_error_inject(u32 btf_id)return btf_id_set_contains (&btf_non_sleepable_error_inject, btf_id); } +static inline int find_subprog_index (const struct bpf_prog *prog, +u32 btf_id) +{ +struct bpf_prog_aux *aux = prog->aux; +int i, subprog =-1 ; + +for (i =0 ; i < aux->func_info_cnt; i++) +if (aux->func_info[i].type_id == btf_id) { + subprog = i; +break ; + } + +return subprog; +} + int bpf_check_attach_target(struct bpf_verifier_log *log, const struct bpf_prog *prog, const struct bpf_prog *tgt_prog,@@ -19254,9 +19269,9 @@ int bpf_check_attach_target(struct bpf_verifier_log *log, {bool prog_extension = prog->type ==BPF_PROG_TYPE_EXT ;const char prefix[] ="btf_trace_" ; -int ret =0 , subprog =-1 , i;const struct btf_type *t;bool conservative = true; +int ret =0 , subprog;const char *tname;struct btf *btf;long addr =0 ;@@ -19291,11 +19306,7 @@ int bpf_check_attach_target(struct bpf_verifier_log *log,return -EINVAL ; } -for (i =0 ; i < aux->func_info_cnt; i++) -if (aux->func_info[i].type_id == btf_id) { - subprog = i; -break ; - } + subprog =find_subprog_index (tgt_prog, btf_id);if (subprog ==-1 ) {bpf_log (log,"Subprog %s doesn't exist\n" , tname);return -EINVAL ;@@ -19559,7 +19570,7 @@ static int check_attach_btf_id(struct bpf_verifier_env *env)struct bpf_attach_target_info tgt_info = {};u32 btf_id = prog->aux->attach_btf_id;struct bpf_trampoline *tr; -int ret; +int ret, subprog;u64 key;if (prog->type ==BPF_PROG_TYPE_SYSCALL ) {
kernel/bpf/verifier.c static int check_attach_btf_id(struct bpf_verifier_env *env) · line 19646
Quoted 7 lines
kernel/bpf/verifier.c +4
-0@@ -19629,6 +19640,12 @@ static int check_attach_btf_id(struct bpf_verifier_env *env)if (!tr)return -ENOMEM ; +if (tgt_prog && tgt_prog->aux->tail_call_reachable) { + subprog =find_subprog_index (tgt_prog, btf_id); + tr->flags = subprog >0 && tgt_prog->aux->func[subprog]->is_func ? +BPF_TRAMP_F_TAIL_CALL_CTX :0 ;
If prog has subprogs all of them will 'is_func', no?
What's the point of the search ?
Just tgt_prog->aux->tail_call_reachable and func_cnt > 0 would be enough?
Quoted 31 lines · Alexei Starovoitov
Quoted 30 lines · Leon Hwang
From commit
ebf7d1f508a73871("bpf, x64: rework pro/epilogue and tailcall handling in JIT"), the tailcall on x64 works better than before.From commit
e411901c0b775a3a("bpf: allow for tailcalls in BPF subprograms for x64 JIT"), tailcall is able to run in BPF subprograms on x64.From commit
5b92a28aae4dd0f8("bpf: Support attaching tracing BPF program to other BPF programs"), BPF program is able to trace other BPF programs.How about combining them all together?
1. FENTRY/FEXIT on a BPF subprogram.
2. A tailcall runs in the BPF subprogram.
3. The tailcall calls itself.As a result, a tailcall infinite loop comes up. And the loop would halt the machine.
As we know, in tail call context, the
tail_call_cntpropagates by stack and RAX register between BPF subprograms. So do it in trampolines.Signed-off-by Leon Hwang <[email protected]>--- arch/x86/net/bpf_jit_comp.c | 40 +++++++++++++++++++++++++++++-------- include/linux/bpf.h | 5 +++++ kernel/bpf/trampoline.c | 4 ++-- kernel/bpf/verifier.c | 31 +++++++++++++++++++++------- 4 files changed, 63 insertions(+), 17 deletions(-)
Quoted 16 lines
Quoted 5 lines
arch/x86/net/bpf_jit_comp.c +0
-0arch/x86/net/bpf_jit_comp.c static void emit_prologue(u8 **pprog, u32 stack_depth, bool ebpf_from_cbpf, · line 311
Quoted 9 lines
arch/x86/net/bpf_jit_comp.c +4
-0@@ -303,8 +303,12 @@ static void emit_prologue(u8 **pprog, u32 stack_depth, bool ebpf_from_cbpf, prog +=X86_PATCH_SIZE ;if (!ebpf_from_cbpf) {if (tail_call_reachable && !is_subprog) +/* When it's the entry of the whole tailcall context, +* zeroing rax means initialising tail_call_cnt. +*/ EMIT2 (0x31 ,0xC0 );/* xor eax, eax */ else +// Keep the same instruction layout. No c++ style comments please.
Got it.
Quoted 27 lines
Quoted 27 lines
arch/x86/net/bpf_jit_comp.c +5
-3EMIT2 (0x66 ,0x90 );/* nop2 */ }EMIT1 (0x55 );/* push rbp */ @@ -1018,6 +1022,10 @@ static void emit_shiftx(u8 **pprog, u32 dst_reg, u8 src_reg, bool is64, u8 op)#define INSN_SZ_DIFF (((addrs[i] - addrs[i - 1]) - (prog - temp))) +/* mov rax, qword ptr [rbp - rounded_stack_depth - 8] */ +#define RESTORE_TAIL_CALL_CNT (stack) \ + EMIT3_off32(0x48, 0x8B, 0x85, -round_up(stack, 8) - 8) +static int do_jit (struct bpf_prog *bpf_prog,int *addrs,u8 *image,u8 *rw_image,int oldproglen,struct jit_context *ctx,bool jmp_padding) {@@ -1623,9 +1631,7 @@ st: if (is_imm8(insn->off)) func = (u8 *) __bpf_call_base + imm32;if (tail_call_reachable) { -/* mov rax, qword ptr [rbp - rounded_stack_depth - 8] */ -EMIT3_off32 (0x48 ,0x8B ,0x85 , - -round_up (bpf_prog->aux->stack_depth,8 ) -8 ); +RESTORE_TAIL_CALL_CNT (bpf_prog->aux->stack_depth);if (!imm32)return -EINVAL ; offs =7 +x86_call_depth_emit_accounting (&prog, func);@@ -2298,7 +2304,9 @@ static int invoke_bpf_mod_ret(const struct btf_func_model *m, u8 **pprog,
Quoted 8 lines
arch/x86/net/bpf_jit_comp.c static int invoke_bpf_mod_ret(const struct btf_func_model *m, u8 **pprog, · line 2308
Quoted 6 lines
arch/x86/net/bpf_jit_comp.c +2
-1* push rbp * mov rbp, rsp * sub rsp, 16 // space for skb and dev -* push rbx // temp regs to pass start time +* mov qword ptr [rbp - 40], rbx // temp regs to pass start time +* mov rax, 2 // cache number of argument to rax What does it mean?
I think it's the corresponding instruction to the following code snippet in arch_prepare_bpf_trampoline().
/* Store number of argument registers of the traced function: * mov rax, nr_regs * mov QWORD PTR [rbp - nregs_off], rax */ emit_mov_imm64(&prog, BPF_REG_0, 0, (u32) nr_regs); emit_stx(&prog, BPF_DW, BPF_REG_FP, BPF_REG_0, -nregs_off);
Quoted 4 lines
arch/x86/net/bpf_jit_comp.c static int invoke_bpf_mod_ret(const struct btf_func_model *m, u8 **pprog, · line 2309
Quoted 2 lines
arch/x86/net/bpf_jit_comp.c +1
-0+* mov qword ptr [rbp - 32], rax // save number of argument to stack Here // is ok since it's inside /* */
Got it.
Quoted 36 lines
Quoted 36 lines
arch/x86/net/bpf_jit_comp.c +6
-1* mov qword ptr [rbp - 16], rdi // save skb pointer to stack * mov qword ptr [rbp - 8], rsi // save dev pointer to stack * call __bpf_prog_enter // rcu_read_lock and preempt_disable @@ -2323,7 +2331,9 @@ static int invoke_bpf_mod_ret(const struct btf_func_model *m, u8 **pprog,* push rbp * mov rbp, rsp * sub rsp, 24 // space for skb, dev, return value -* push rbx // temp regs to pass start time +* mov qword ptr [rbp - 40], rbx // temp regs to pass start time +* mov rax, 2 // cache number of argument to rax +* mov qword ptr [rbp - 32], rax // save number of argument to stack * mov qword ptr [rbp - 24], rdi // save skb pointer to stack * mov qword ptr [rbp - 16], rsi // save dev pointer to stack * call __bpf_prog_enter // rcu_read_lock and preempt_disable @@ -2400,6 +2410,7 @@ int arch_prepare_bpf_trampoline(struct bpf_tramp_image *im, void *image, void *i* [ ... ] * [ stack_arg2 ] * RBP - arg_stack_off [ stack_arg1 ] +* RSP [ tail_call_cnt ] BPF_TRAMP_F_TAIL_CALL_CTX */ /* room for return value of orig_call or fentry prog */ @@ -2464,6 +2475,8 @@ int arch_prepare_bpf_trampoline(struct bpf_tramp_image *im, void *image, void *ielse /* sub rsp, stack_size */ EMIT4 (0x48 ,0x83 ,0xEC , stack_size); +if (flags &BPF_TRAMP_F_TAIL_CALL_CTX ) +EMIT1 (0x50 );/* push rax */ /* mov QWORD PTR [rbp - rbx_off], rbx */ emit_stx (&prog,BPF_DW ,BPF_REG_FP ,BPF_REG_6 , -rbx_off);@@ -2516,9 +2529,15 @@ int arch_prepare_bpf_trampoline(struct bpf_tramp_image *im, void *image, void *irestore_regs (m, &prog, regs_off);save_args (m, &prog, arg_stack_off, true);
Quoted 16 lines
arch/x86/net/bpf_jit_comp.c int arch_prepare_bpf_trampoline(struct bpf_tramp_image *im, void *image, void *i · line 2540
Quoted 11 lines
arch/x86/net/bpf_jit_comp.c +8
-2+if (flags &BPF_TRAMP_F_TAIL_CALL_CTX ) +/* Before calling the original function, restore the +* tail_call_cnt from stack to rax. +*/ +RESTORE_TAIL_CALL_CNT (stack_size); +if (flags &BPF_TRAMP_F_ORIG_STACK ) { -emit_ldx (&prog,BPF_DW ,BPF_REG_0 ,BPF_REG_FP ,8 ); -EMIT2 (0xff ,0xd0 );/* call *rax */ +emit_ldx (&prog,BPF_DW ,BPF_REG_6 ,BPF_REG_FP ,8 ); +EMIT2 (0xff ,0xd3 );/* call *rbx */ // FIXME: Confirm 0xd3? please no FIXME like comments.
You have to be confident in the code you're submitting. llvm-mc -triple=x86_64 -show-encoding -x86-asm-syntax=intel -output-asm-variant=1 <<< 'call rbx'
Got it. Thanks for the guide.
Quoted 112 lines
Quoted 112 lines
arch/x86/net/bpf_jit_comp.c +6
-1}else {/* call original function */ if (emit_rsb_call (&prog, orig_call, prog)) {@@ -2569,7 +2588,12 @@ int arch_prepare_bpf_trampoline(struct bpf_tramp_image *im, void *image, void *i ret = -EINVAL ; goto cleanup; } - } + }else if (flags &BPF_TRAMP_F_TAIL_CALL_CTX ) +/* Before running the original function, restore the +* tail_call_cnt from stack to rax. +*/ +RESTORE_TAIL_CALL_CNT (stack_size); +/* restore return value of orig_call or fentry prog back into RAX */ if (save_ret)emit_ldx (&prog,BPF_DW ,BPF_REG_0 ,BPF_REG_FP ,-8 );include/linux/bpf.h +5
-0@@ -1028,6 +1028,11 @@ struct btf_func_model {*/ #define BPF_TRAMP_F_SHARE_IPMODIFY BIT(6) +/* Indicate that current trampoline is in a tail call context. Then, it has to +* cache and restore tail_call_cnt to avoid infinite tail call loop. +*/ +#define BPF_TRAMP_F_TAIL_CALL_CTX BIT(7) +/* Each call __bpf_prog_enter + call bpf_func + call __bpf_prog_exit is ~50 * bytes on x86. */ kernel/bpf/trampoline.c +2
-2@@ -415,8 +415,8 @@ static int bpf_trampoline_update(struct bpf_trampoline *tr, bool lock_direct_mut goto out; } -/* clear all bits except SHARE_IPMODIFY */ - tr->flags &=BPF_TRAMP_F_SHARE_IPMODIFY ; +/* clear all bits except SHARE_IPMODIFY and TAIL_CALL_CTX */ + tr->flags &= (BPF_TRAMP_F_SHARE_IPMODIFY |BPF_TRAMP_F_TAIL_CALL_CTX );if (tlinks[BPF_TRAMP_FEXIT ].nr_links || tlinks[BPF_TRAMP_MODIFY_RETURN ].nr_links) {kernel/bpf/verifier.c +18
-7@@ -19246,6 +19246,21 @@ static int check_non_sleepable_error_inject(u32 btf_id)return btf_id_set_contains (&btf_non_sleepable_error_inject, btf_id); } +static inline int find_subprog_index (const struct bpf_prog *prog, +u32 btf_id) +{ +struct bpf_prog_aux *aux = prog->aux; +int i, subprog =-1 ; + +for (i =0 ; i < aux->func_info_cnt; i++) +if (aux->func_info[i].type_id == btf_id) { + subprog = i; +break ; + } + +return subprog; +} + int bpf_check_attach_target(struct bpf_verifier_log *log, const struct bpf_prog *prog, const struct bpf_prog *tgt_prog,@@ -19254,9 +19269,9 @@ int bpf_check_attach_target(struct bpf_verifier_log *log, {bool prog_extension = prog->type ==BPF_PROG_TYPE_EXT ;const char prefix[] ="btf_trace_" ; -int ret =0 , subprog =-1 , i;const struct btf_type *t;bool conservative = true; +int ret =0 , subprog;const char *tname;struct btf *btf;long addr =0 ;@@ -19291,11 +19306,7 @@ int bpf_check_attach_target(struct bpf_verifier_log *log,return -EINVAL ; } -for (i =0 ; i < aux->func_info_cnt; i++) -if (aux->func_info[i].type_id == btf_id) { - subprog = i; -break ; - } + subprog =find_subprog_index (tgt_prog, btf_id);if (subprog ==-1 ) {bpf_log (log,"Subprog %s doesn't exist\n" , tname);return -EINVAL ;@@ -19559,7 +19570,7 @@ static int check_attach_btf_id(struct bpf_verifier_env *env)struct bpf_attach_target_info tgt_info = {};u32 btf_id = prog->aux->attach_btf_id;struct bpf_trampoline *tr; -int ret; +int ret, subprog;u64 key;if (prog->type ==BPF_PROG_TYPE_SYSCALL ) {@@ -19629,6 +19640,12 @@ static int check_attach_btf_id(struct bpf_verifier_env *env)if (!tr)return -ENOMEM ;
Quoted 8 lines
kernel/bpf/verifier.c static int check_attach_btf_id(struct bpf_verifier_env *env) · line 19646
Quoted 4 lines
kernel/bpf/verifier.c +4
-0+if (tgt_prog && tgt_prog->aux->tail_call_reachable) { + subprog =find_subprog_index (tgt_prog, btf_id); + tr->flags = subprog >0 && tgt_prog->aux->func[subprog]->is_func ? +BPF_TRAMP_F_TAIL_CALL_CTX :0 ;If prog has subprogs all of them will '
is_func', no?
What's the point of the search ?
Justtgt_prog->aux->tail_call_reachableandfunc_cnt> 0 would be enough?
tgt_prog->aux->tail_call_reachable and subprog > 0 would be enough? It has to confirm that the attaching target is a subprog of tgt_prog instead of tgt_prog itself.
In tail call context, when 'call' a func, tail_call_cnt will be restored to rax.
static int do_jit() {
/* call */
case BPF_JMP | BPF_CALL: {
int offs;
func = (u8 *) __bpf_call_base + imm32;
if (tail_call_reachable) {
/* mov rax, qword ptr [rbp - rounded_stack_depth - 8] */
EMIT3_off32(0x48, 0x8B, 0x85,
-round_up(bpf_prog->aux->stack_depth, 8) - 8);
/* ... */
}
}As a result, when 'call' a subprog, tail_call_cnt will be transferred by rax. Do all of subprogs run by 'call', including not-'is_func' subprogs?
The point of the search is to confirm that the attaching subprog runs by 'call'.
Currently, I'm sure that tgt_prog->aux->tail_call_reachable, subprog > 0 and tgt_prog->aux->func[subprog]->is_func is the case to be fixed.
Thanks,
Leon
Quoted 93 lines · Leon Hwang
Quoted 31 lines · Alexei Starovoitov
Quoted 30 lines · Leon Hwang
From commit
ebf7d1f508a73871("bpf, x64: rework pro/epilogue and tailcall handling in JIT"), the tailcall on x64 works better than before.From commit
e411901c0b775a3a("bpf: allow for tailcalls in BPF subprograms for x64 JIT"), tailcall is able to run in BPF subprograms on x64.From commit
5b92a28aae4dd0f8("bpf: Support attaching tracing BPF program to other BPF programs"), BPF program is able to trace other BPF programs.How about combining them all together?
1. FENTRY/FEXIT on a BPF subprogram.
2. A tailcall runs in the BPF subprogram.
3. The tailcall calls itself.As a result, a tailcall infinite loop comes up. And the loop would halt the machine.
As we know, in tail call context, the
tail_call_cntpropagates by stack and RAX register between BPF subprograms. So do it in trampolines.Signed-off-by Leon Hwang <[email protected]>--- arch/x86/net/bpf_jit_comp.c | 40 +++++++++++++++++++++++++++++-------- include/linux/bpf.h | 5 +++++ kernel/bpf/trampoline.c | 4 ++-- kernel/bpf/verifier.c | 31 +++++++++++++++++++++------- 4 files changed, 63 insertions(+), 17 deletions(-)Quoted 16 lines
Quoted 5 lines
arch/x86/net/bpf_jit_comp.c +0
-0arch/x86/net/bpf_jit_comp.c static void emit_prologue(u8 **pprog, u32 stack_depth, bool ebpf_from_cbpf, · line 311
Quoted 9 lines
arch/x86/net/bpf_jit_comp.c +4
-0@@ -303,8 +303,12 @@ static void emit_prologue(u8 **pprog, u32 stack_depth, bool ebpf_from_cbpf, prog +=X86_PATCH_SIZE ;if (!ebpf_from_cbpf) {if (tail_call_reachable && !is_subprog) +/* When it's the entry of the whole tailcall context, +* zeroing rax means initialising tail_call_cnt. +*/ EMIT2 (0x31 ,0xC0 );/* xor eax, eax */ else +// Keep the same instruction layout. No c++ style comments please.
Got it.
Quoted 27 lines
Quoted 27 lines
arch/x86/net/bpf_jit_comp.c +5
-3EMIT2 (0x66 ,0x90 );/* nop2 */ }EMIT1 (0x55 );/* push rbp */ @@ -1018,6 +1022,10 @@ static void emit_shiftx(u8 **pprog, u32 dst_reg, u8 src_reg, bool is64, u8 op)#define INSN_SZ_DIFF (((addrs[i] - addrs[i - 1]) - (prog - temp))) +/* mov rax, qword ptr [rbp - rounded_stack_depth - 8] */ +#define RESTORE_TAIL_CALL_CNT (stack) \ + EMIT3_off32(0x48, 0x8B, 0x85, -round_up(stack, 8) - 8) +static int do_jit (struct bpf_prog *bpf_prog,int *addrs,u8 *image,u8 *rw_image,int oldproglen,struct jit_context *ctx,bool jmp_padding) {@@ -1623,9 +1631,7 @@ st: if (is_imm8(insn->off)) func = (u8 *) __bpf_call_base + imm32;if (tail_call_reachable) { -/* mov rax, qword ptr [rbp - rounded_stack_depth - 8] */ -EMIT3_off32 (0x48 ,0x8B ,0x85 , - -round_up (bpf_prog->aux->stack_depth,8 ) -8 ); +RESTORE_TAIL_CALL_CNT (bpf_prog->aux->stack_depth);if (!imm32)return -EINVAL ; offs =7 +x86_call_depth_emit_accounting (&prog, func);@@ -2298,7 +2304,9 @@ static int invoke_bpf_mod_ret(const struct btf_func_model *m, u8 **pprog,Quoted 8 lines
arch/x86/net/bpf_jit_comp.c static int invoke_bpf_mod_ret(const struct btf_func_model *m, u8 **pprog, · line 2308
Quoted 6 lines
arch/x86/net/bpf_jit_comp.c +2
-1* push rbp * mov rbp, rsp * sub rsp, 16 // space for skb and dev -* push rbx // temp regs to pass start time +* mov qword ptr [rbp - 40], rbx // temp regs to pass start time +* mov rax, 2 // cache number of argument to rax What does it mean?
I think it's the corresponding instruction to the following code snippet in
arch_prepare_bpf_trampoline().
Quoted 6 lines
/* Store number of argument registers of the traced function:
* mov rax, nr_regs
* mov QWORD PTR [rbp - nregs_off], rax
*/
emit_mov_imm64(&prog, BPF_REG_0, 0, (u32) nr_regs);
emit_stx(&prog, BPF_DW, BPF_REG_FP, BPF_REG_0, -nregs_off);Ahh. I see.
The comment on top of arch_prepare_bpf_trampoline() is hopelessly obsolete. Don't touch it in this patch set. We probably should delete it at some point or take an effort to update it thoroughly.
Earlier recommendation to you was to update this comment:
/* Generated trampoline stack layout:
Quoted 203 lines
Quoted 4 lines
arch/x86/net/bpf_jit_comp.c static int invoke_bpf_mod_ret(const struct btf_func_model *m, u8 **pprog, · line 2309
Quoted 2 lines
arch/x86/net/bpf_jit_comp.c +1
-0+* mov qword ptr [rbp - 32], rax // save number of argument to stack Here // is ok since it's inside /* */
Got it.
Quoted 36 lines
Quoted 36 lines
arch/x86/net/bpf_jit_comp.c +6
-1* mov qword ptr [rbp - 16], rdi // save skb pointer to stack * mov qword ptr [rbp - 8], rsi // save dev pointer to stack * call __bpf_prog_enter // rcu_read_lock and preempt_disable @@ -2323,7 +2331,9 @@ static int invoke_bpf_mod_ret(const struct btf_func_model *m, u8 **pprog,* push rbp * mov rbp, rsp * sub rsp, 24 // space for skb, dev, return value -* push rbx // temp regs to pass start time +* mov qword ptr [rbp - 40], rbx // temp regs to pass start time +* mov rax, 2 // cache number of argument to rax +* mov qword ptr [rbp - 32], rax // save number of argument to stack * mov qword ptr [rbp - 24], rdi // save skb pointer to stack * mov qword ptr [rbp - 16], rsi // save dev pointer to stack * call __bpf_prog_enter // rcu_read_lock and preempt_disable @@ -2400,6 +2410,7 @@ int arch_prepare_bpf_trampoline(struct bpf_tramp_image *im, void *image, void *i* [ ... ] * [ stack_arg2 ] * RBP - arg_stack_off [ stack_arg1 ] +* RSP [ tail_call_cnt ] BPF_TRAMP_F_TAIL_CALL_CTX */ /* room for return value of orig_call or fentry prog */ @@ -2464,6 +2475,8 @@ int arch_prepare_bpf_trampoline(struct bpf_tramp_image *im, void *image, void *ielse /* sub rsp, stack_size */ EMIT4 (0x48 ,0x83 ,0xEC , stack_size); +if (flags &BPF_TRAMP_F_TAIL_CALL_CTX ) +EMIT1 (0x50 );/* push rax */ /* mov QWORD PTR [rbp - rbx_off], rbx */ emit_stx (&prog,BPF_DW ,BPF_REG_FP ,BPF_REG_6 , -rbx_off);@@ -2516,9 +2529,15 @@ int arch_prepare_bpf_trampoline(struct bpf_tramp_image *im, void *image, void *irestore_regs (m, &prog, regs_off);save_args (m, &prog, arg_stack_off, true);Quoted 16 lines
arch/x86/net/bpf_jit_comp.c int arch_prepare_bpf_trampoline(struct bpf_tramp_image *im, void *image, void *i · line 2540
Quoted 11 lines
arch/x86/net/bpf_jit_comp.c +8
-2+if (flags &BPF_TRAMP_F_TAIL_CALL_CTX ) +/* Before calling the original function, restore the +* tail_call_cnt from stack to rax. +*/ +RESTORE_TAIL_CALL_CNT (stack_size); +if (flags &BPF_TRAMP_F_ORIG_STACK ) { -emit_ldx (&prog,BPF_DW ,BPF_REG_0 ,BPF_REG_FP ,8 ); -EMIT2 (0xff ,0xd0 );/* call *rax */ +emit_ldx (&prog,BPF_DW ,BPF_REG_6 ,BPF_REG_FP ,8 ); +EMIT2 (0xff ,0xd3 );/* call *rbx */ // FIXME: Confirm 0xd3? please no FIXME like comments.
You have to be confident in the code you're submitting. llvm-mc -triple=x86_64 -show-encoding -x86-asm-syntax=intel -output-asm-variant=1 <<< 'call rbx'Got it. Thanks for the guide.
Quoted 112 lines
Quoted 112 lines
arch/x86/net/bpf_jit_comp.c +6
-1}else {/* call original function */ if (emit_rsb_call (&prog, orig_call, prog)) {@@ -2569,7 +2588,12 @@ int arch_prepare_bpf_trampoline(struct bpf_tramp_image *im, void *image, void *i ret = -EINVAL ; goto cleanup; } - } + }else if (flags &BPF_TRAMP_F_TAIL_CALL_CTX ) +/* Before running the original function, restore the +* tail_call_cnt from stack to rax. +*/ +RESTORE_TAIL_CALL_CNT (stack_size); +/* restore return value of orig_call or fentry prog back into RAX */ if (save_ret)emit_ldx (&prog,BPF_DW ,BPF_REG_0 ,BPF_REG_FP ,-8 );include/linux/bpf.h +5
-0@@ -1028,6 +1028,11 @@ struct btf_func_model {*/ #define BPF_TRAMP_F_SHARE_IPMODIFY BIT(6) +/* Indicate that current trampoline is in a tail call context. Then, it has to +* cache and restore tail_call_cnt to avoid infinite tail call loop. +*/ +#define BPF_TRAMP_F_TAIL_CALL_CTX BIT(7) +/* Each call __bpf_prog_enter + call bpf_func + call __bpf_prog_exit is ~50 * bytes on x86. */ kernel/bpf/trampoline.c +2
-2@@ -415,8 +415,8 @@ static int bpf_trampoline_update(struct bpf_trampoline *tr, bool lock_direct_mut goto out; } -/* clear all bits except SHARE_IPMODIFY */ - tr->flags &=BPF_TRAMP_F_SHARE_IPMODIFY ; +/* clear all bits except SHARE_IPMODIFY and TAIL_CALL_CTX */ + tr->flags &= (BPF_TRAMP_F_SHARE_IPMODIFY |BPF_TRAMP_F_TAIL_CALL_CTX );if (tlinks[BPF_TRAMP_FEXIT ].nr_links || tlinks[BPF_TRAMP_MODIFY_RETURN ].nr_links) {kernel/bpf/verifier.c +18
-7@@ -19246,6 +19246,21 @@ static int check_non_sleepable_error_inject(u32 btf_id)return btf_id_set_contains (&btf_non_sleepable_error_inject, btf_id); } +static inline int find_subprog_index (const struct bpf_prog *prog, +u32 btf_id) +{ +struct bpf_prog_aux *aux = prog->aux; +int i, subprog =-1 ; + +for (i =0 ; i < aux->func_info_cnt; i++) +if (aux->func_info[i].type_id == btf_id) { + subprog = i; +break ; + } + +return subprog; +} + int bpf_check_attach_target(struct bpf_verifier_log *log, const struct bpf_prog *prog, const struct bpf_prog *tgt_prog,@@ -19254,9 +19269,9 @@ int bpf_check_attach_target(struct bpf_verifier_log *log, {bool prog_extension = prog->type ==BPF_PROG_TYPE_EXT ;const char prefix[] ="btf_trace_" ; -int ret =0 , subprog =-1 , i;const struct btf_type *t;bool conservative = true; +int ret =0 , subprog;const char *tname;struct btf *btf;long addr =0 ;@@ -19291,11 +19306,7 @@ int bpf_check_attach_target(struct bpf_verifier_log *log,return -EINVAL ; } -for (i =0 ; i < aux->func_info_cnt; i++) -if (aux->func_info[i].type_id == btf_id) { - subprog = i; -break ; - } + subprog =find_subprog_index (tgt_prog, btf_id);if (subprog ==-1 ) {bpf_log (log,"Subprog %s doesn't exist\n" , tname);return -EINVAL ;@@ -19559,7 +19570,7 @@ static int check_attach_btf_id(struct bpf_verifier_env *env)struct bpf_attach_target_info tgt_info = {};u32 btf_id = prog->aux->attach_btf_id;struct bpf_trampoline *tr; -int ret; +int ret, subprog;u64 key;if (prog->type ==BPF_PROG_TYPE_SYSCALL ) {@@ -19629,6 +19640,12 @@ static int check_attach_btf_id(struct bpf_verifier_env *env)if (!tr)return -ENOMEM ;Quoted 8 lines
kernel/bpf/verifier.c static int check_attach_btf_id(struct bpf_verifier_env *env) · line 19646
Quoted 4 lines
kernel/bpf/verifier.c +4
-0+if (tgt_prog && tgt_prog->aux->tail_call_reachable) { + subprog =find_subprog_index (tgt_prog, btf_id); + tr->flags = subprog >0 && tgt_prog->aux->func[subprog]->is_func ? +BPF_TRAMP_F_TAIL_CALL_CTX :0 ;If prog has subprogs all of them will '
is_func', no?
What's the point of the search ?
Justtgt_prog->aux->tail_call_reachableandfunc_cnt> 0 would be enough?
tgt_prog->aux->tail_call_reachableand subprog > 0 would be enough? It has to confirm that the attaching target is a subprog oftgt_proginstead oftgt_progitself.In tail call context, when 'call' a func,
tail_call_cntwill be restored to rax.static int do_jit() { /* call */ case BPF_JMP | BPF_CALL: { int offs; func = (u8 *) __bpf_call_base + imm32; if (tail_call_reachable) { /* mov rax, qword ptr [rbp - rounded_stack_depth - 8] */ EMIT3_off32(0x48, 0x8B, 0x85, -round_up(bpf_prog->aux->stack_depth, 8) - 8); /* ... */ } }
Quoted 2 lines
As a result, when 'call' a subprog,
tail_call_cntwill be transferred by rax. Do all of subprogs run by 'call', including not-'is_func' subprogs?
Let me ask again. Do you see a subprog that has is_func==0 ?
Quoted 7 lines · Alexei Starovoitov
Quoted 6 lines · Leon Hwang
Quoted 2 lines · Alexei Starovoitov
Quoted 1 line · Leon Hwang
[SNIP]
Quoted 17 lines
Quoted 10 lines
Quoted 6 lines
Quoted 4 lines
* sub rsp, 16 // space for skb and dev - * push rbx // temp regs to pass start time + * mov qword ptr [rbp - 40], rbx // temp regs to pass start time + * mov rax, 2 // cache number of argument to raxWhat does it mean?
I think it's the corresponding instruction to the following code snippet in
arch_prepare_bpf_trampoline().Quoted 6 lines
/* Store number of argument registers of the traced function: * mov rax, nr_regs * mov QWORD PTR [rbp - nregs_off], rax */ emit_mov_imm64(&prog, BPF_REG_0, 0, (u32) nr_regs); emit_stx(&prog, BPF_DW, BPF_REG_FP, BPF_REG_0, -nregs_off);
Quoted 4 lines
Ahh. I see.
The comment on top ofarch_prepare_bpf_trampoline()is hopelessly obsolete. Don't touch it in this patch set. We probably should delete it at some point or take an effort to update it thoroughly.
Got it.
Quoted 69 lines
Earlier recommendation to you was to update this comment:
/* Generated trampoline stack layout:Quoted 66 lines
Quoted 4 lines
Quoted 2 lines
+ * mov qword ptr [rbp - 32], rax // save number of argument to stack
Here // is ok since it's inside /* */
Got it.
Quoted 36 lines
Quoted 36 lines
* mov qword ptr [rbp - 16], rdi // save skb pointer to stack * mov qword ptr [rbp - 8], rsi // save dev pointer to stack * call __bpf_prog_enter // rcu_read_lock and preempt_disablearch/x86/net/bpf_jit_comp.c +6
-1@@ -2323,7 +2331,9 @@ static int invoke_bpf_mod_ret(const struct btf_func_model *m, u8 **pprog,* push rbp * mov rbp, rsp * sub rsp, 24 // space for skb, dev, return value -* push rbx // temp regs to pass start time +* mov qword ptr [rbp - 40], rbx // temp regs to pass start time +* mov rax, 2 // cache number of argument to rax +* mov qword ptr [rbp - 32], rax // save number of argument to stack * mov qword ptr [rbp - 24], rdi // save skb pointer to stack * mov qword ptr [rbp - 16], rsi // save dev pointer to stack * call __bpf_prog_enter // rcu_read_lock and preempt_disable @@ -2400,6 +2410,7 @@ int arch_prepare_bpf_trampoline(struct bpf_tramp_image *im, void *image, void *i* [ ... ] * [ stack_arg2 ] * RBP - arg_stack_off [ stack_arg1 ] +* RSP [ tail_call_cnt ] BPF_TRAMP_F_TAIL_CALL_CTX */ /* room for return value of orig_call or fentry prog */ @@ -2464,6 +2475,8 @@ int arch_prepare_bpf_trampoline(struct bpf_tramp_image *im, void *image, void *ielse /* sub rsp, stack_size */ EMIT4 (0x48 ,0x83 ,0xEC , stack_size); +if (flags &BPF_TRAMP_F_TAIL_CALL_CTX ) +EMIT1 (0x50 );/* push rax */ /* mov QWORD PTR [rbp - rbx_off], rbx */ emit_stx (&prog,BPF_DW ,BPF_REG_FP ,BPF_REG_6 , -rbx_off);@@ -2516,9 +2529,15 @@ int arch_prepare_bpf_trampoline(struct bpf_tramp_image *im, void *image, void *irestore_regs (m, &prog, regs_off);save_args (m, &prog, arg_stack_off, true);Quoted 16 lines
arch/x86/net/bpf_jit_comp.c int arch_prepare_bpf_trampoline(struct bpf_tramp_image *im, void *image, void *i · line 2540
Quoted 11 lines
arch/x86/net/bpf_jit_comp.c +8
-2+if (flags &BPF_TRAMP_F_TAIL_CALL_CTX ) +/* Before calling the original function, restore the +* tail_call_cnt from stack to rax. +*/ +RESTORE_TAIL_CALL_CNT (stack_size); +if (flags &BPF_TRAMP_F_ORIG_STACK ) { -emit_ldx (&prog,BPF_DW ,BPF_REG_0 ,BPF_REG_FP ,8 ); -EMIT2 (0xff ,0xd0 );/* call *rax */ +emit_ldx (&prog,BPF_DW ,BPF_REG_6 ,BPF_REG_FP ,8 ); +EMIT2 (0xff ,0xd3 );/* call *rbx */ // FIXME: Confirm 0xd3? please no FIXME like comments.
You have to be confident in the code you're submitting. llvm-mc -triple=x86_64 -show-encoding -x86-asm-syntax=intel -output-asm-variant=1 <<< 'call rbx'Got it. Thanks for the guide.
Quoted 4 lines
Quoted 4 lines
arch/x86/net/bpf_jit_comp.c +0
-0}else {/* call original function */ if (emit_rsb_call (&prog, orig_call, prog)) {
[SNIP]
Quoted 35 lines
Quoted 35 lines
Quoted 6 lines
Quoted 6 lines
if (prog->type == BPF_PROG_TYPE_SYSCALL) {kernel/bpf/verifier.c +0
-0@@ -19629,6 +19640,12 @@ static int check_attach_btf_id(struct bpf_verifier_env *env)if (!tr)return -ENOMEM ;Quoted 8 lines
kernel/bpf/verifier.c static int check_attach_btf_id(struct bpf_verifier_env *env) · line 19646
Quoted 4 lines
kernel/bpf/verifier.c +4
-0+if (tgt_prog && tgt_prog->aux->tail_call_reachable) { + subprog =find_subprog_index (tgt_prog, btf_id); + tr->flags = subprog >0 && tgt_prog->aux->func[subprog]->is_func ? +BPF_TRAMP_F_TAIL_CALL_CTX :0 ;If prog has subprogs all of them will '
is_func', no?
What's the point of the search ?
Justtgt_prog->aux->tail_call_reachableandfunc_cnt> 0 would be enough?
tgt_prog->aux->tail_call_reachableand subprog > 0 would be enough? It has to confirm that the attaching target is a subprog oftgt_proginstead oftgt_progitself.In tail call context, when 'call' a func,
tail_call_cntwill be restored to rax.static int do_jit() { /* call */ case BPF_JMP | BPF_CALL: { int offs; func = (u8 *) __bpf_call_base + imm32; if (tail_call_reachable) { /* mov rax, qword ptr [rbp - rounded_stack_depth - 8] */ EMIT3_off32(0x48, 0x8B, 0x85, -round_up(bpf_prog->aux->stack_depth, 8) - 8); /* ... */ } }
Quoted 4 lines
Quoted 2 lines
As a result, when 'call' a subprog,
tail_call_cntwill be transferred by rax. Do all of subprogs run by 'call', including not-'is_func' subprogs?Let me ask again. Do you see a subprog that has is_func==0 ?
Oh, I get it.
In jit_subprogs(), all of subprogs are 'is_func'.
So, it's unnecessary to check tgt_prog->aux->func[subprog]->is_func.
I'll submit a new RFC PATCH later.
Thanks,
Leon
RFC PATCH v2 2/2 selftests/bpf: Add testcases for tailcall infinite loop fixing
In reply to Leon HwangAdd 3 test cases to confirm the tailcall infinite loop bug has been fixed.
Like tailcall_bpf2bpf cases, do fentry/fexit on the bpf2bpf, and then check the final count result.
tools/testing/selftests/bpf/test_progs -t tailcalls
226/13 tailcalls/tailcall_bpf2bpf_fentry:OK 226/14 tailcalls/tailcall_bpf2bpf_fexit:OK 226/15 tailcalls/tailcall_bpf2bpf_fentry_fexit:OK 226 tailcalls:OK
Summary: 1/15 PASSED, 0 SKIPPED, 0 FAILED
--- .../selftests/bpf/prog_tests/tailcalls.c | 194 +++++++++++++++++- .../bpf/progs/tailcall_bpf2bpf_fentry.c | 18 ++ .../bpf/progs/tailcall_bpf2bpf_fexit.c | 18 ++ 3 files changed, 229 insertions(+), 1 deletion(-) create mode 100644 tools/testing/selftests/bpf/progs/tailcall_bpf2bpf_fentry.c create mode 100644 tools/testing/selftests/bpf/progs/tailcall_bpf2bpf_fexit.c
tools/testing/selftests/bpf/prog_tests/tailcalls.c +193 -1
@@ -634,7 +634,7 @@ static void test_tailcall_bpf2bpf_2(void)return ; data_fd =bpf_map__fd (data_map); -if (CHECK_FAIL (map_fd <0 )) +if (CHECK_FAIL (data_fd <0 ))return ; i =0 ;
@@ -884,6 +884,191 @@ static void test_tailcall_bpf2bpf_6(void)tailcall_bpf2bpf6__destroy (obj); } +static void __tailcall_bpf2bpf_fentry_fexit (bool test_fentry,bool test_fexit) +{ +struct bpf_object *tgt_obj =NULL , *fentry_obj =NULL , *fexit_obj =NULL ; +struct bpf_link *fentry_link =NULL , *fexit_link =NULL ; +int err, map_fd, prog_fd, main_fd, data_fd, i, val; +struct bpf_map *prog_array, *data_map; +struct bpf_program *prog; +char buff[128 ] = {}; + +LIBBPF_OPTS (bpf_test_run_opts, topts, + .data_in = buff, + .data_size_in = sizeof(buff), + .repeat = 1, + ); + + err =bpf_prog_test_load ("tailcall_bpf2bpf2.bpf.o" , +BPF_PROG_TYPE_SCHED_CLS , + &tgt_obj, &prog_fd); +if (!ASSERT_OK (err,"load tgt_obj" )) +return ; + + prog =bpf_object__find_program_by_name (tgt_obj,"entry" ); +if (!ASSERT_OK_PTR (prog,"find entry prog" )) + goto out; + + main_fd =bpf_program__fd (prog); +if (!ASSERT_FALSE (main_fd <0 ,"find entry prog fd" )) + goto out; + + prog_array =bpf_object__find_map_by_name (tgt_obj,"jmp_table" ); +if (!ASSERT_OK_PTR (prog_array,"find jmp_table map" )) + goto out; + + map_fd =bpf_map__fd (prog_array); +if (!ASSERT_FALSE (map_fd <0 ,"find jmp_table map fd" )) + goto out; + + prog =bpf_object__find_program_by_name (tgt_obj,"classifier_0" ); +if (!ASSERT_OK_PTR (prog,"find classifier_0 prog" )) + goto out; + + prog_fd =bpf_program__fd (prog); +if (!ASSERT_FALSE (prog_fd <0 ,"find classifier_0 prog fd" )) + goto out; + + i =0 ; + err =bpf_map_update_elem (map_fd, &i, &prog_fd,BPF_ANY ); +if (!ASSERT_OK (err,"update jmp_table" )) + goto out; + +if (test_fentry) { + fentry_obj =bpf_object__open_file ("tailcall_bpf2bpf_fentry.bpf.o" , +NULL ); +if (!ASSERT_OK_PTR (fentry_obj,"open fentry_obj file" )) + goto out; + + prog =bpf_object__find_program_by_name (fentry_obj,"fentry" ); +if (!ASSERT_OK_PTR (prog,"find fentry prog" )) + goto out; + + err =bpf_program__set_attach_target (prog, prog_fd, +"subprog_tail" ); +if (!ASSERT_OK (err,"set_attach_target subprog_tail" )) + goto out; + + err =bpf_object__load (fentry_obj); +if (!ASSERT_OK (err,"load fentry_obj" )) + goto out; + + fentry_link =bpf_program__attach_trace (prog); +if (!ASSERT_OK_PTR (fentry_link,"attach_trace" )) + goto out; + } + +if (test_fexit) { + fexit_obj =bpf_object__open_file ("tailcall_bpf2bpf_fexit.bpf.o" , +NULL ); +if (!ASSERT_OK_PTR (fexit_obj,"open fexit_obj file" )) + goto out; + + prog =bpf_object__find_program_by_name (fexit_obj,"fexit" ); +if (!ASSERT_OK_PTR (prog,"find fexit prog" )) + goto out; + + err =bpf_program__set_attach_target (prog, prog_fd, +"subprog_tail" ); +if (!ASSERT_OK (err,"set_attach_target subprog_tail" )) + goto out; + + err =bpf_object__load (fexit_obj); +if (!ASSERT_OK (err,"load fexit_obj" )) + goto out; + + fexit_link =bpf_program__attach_trace (prog); +if (!ASSERT_OK_PTR (fexit_link,"attach_trace" )) + goto out; + } + + err =bpf_prog_test_run_opts (main_fd, &topts); +ASSERT_OK (err,"tailcall" ); +ASSERT_EQ (topts.retval,1 ,"tailcall retval" ); + + data_map =bpf_object__find_map_by_name (tgt_obj,"tailcall.bss" ); +if (!ASSERT_FALSE (!data_map || !bpf_map__is_internal (data_map), +"find tailcall.bss map" )) + goto out; + + data_fd =bpf_map__fd (data_map); +if (!ASSERT_FALSE (data_fd <0 ,"find tailcall.bss map fd" )) + goto out; + + i =0 ; + err =bpf_map_lookup_elem (data_fd, &i, &val); +ASSERT_OK (err,"tailcall count" ); +ASSERT_EQ (val,33 ,"tailcall count" ); + +if (test_fentry) { + data_map =bpf_object__find_map_by_name (fentry_obj,".bss" ); +if (!ASSERT_FALSE (!data_map || !bpf_map__is_internal (data_map), +"find tailcall_bpf2bpf_fentry.bss.bss map" )) + goto out; + + data_fd =bpf_map__fd (data_map); +if (!ASSERT_FALSE (data_fd <0 , +"find tailcall_bpf2bpf_fentry.bss.bss map fd" )) + goto out; + + i =0 ; + err =bpf_map_lookup_elem (data_fd, &i, &val); +ASSERT_OK (err,"fentry count" ); +ASSERT_EQ (val,33 ,"fentry count" ); + } + +if (test_fexit) { + data_map =bpf_object__find_map_by_name (fexit_obj,".bss" ); +if (!ASSERT_FALSE (!data_map || !bpf_map__is_internal (data_map), +"find tailcall_bpf2bpf_fexit.bss map" )) + goto out; + + data_fd =bpf_map__fd (data_map); +if (!ASSERT_FALSE (data_fd <0 , +"find tailcall_bpf2bpf_fexit.bss map fd" )) + goto out; + + i =0 ; + err =bpf_map_lookup_elem (data_fd, &i, &val); +ASSERT_OK (err,"fexit count" ); +ASSERT_EQ (val,33 ,"fexit count" ); + } + +out: +bpf_link__destroy (fentry_link); +bpf_link__destroy (fexit_link); +bpf_object__close (fentry_obj); +bpf_object__close (fexit_obj); +bpf_object__close (tgt_obj); +} + +/* test_tailcall_bpf2bpf_fentry checks that the count value of the tail call +* limit enforcement matches with expectations when tailcall is preceded with +* bpf2bpf call, and the bpf2bpf call is traced by fentry. +*/ +static void test_tailcall_bpf2bpf_fentry (void ) +{ +__tailcall_bpf2bpf_fentry_fexit (true, false); +} + +/* test_tailcall_bpf2bpf_fexit checks that the count value of the tail call +* limit enforcement matches with expectations when tailcall is preceded with +* bpf2bpf call, and the bpf2bpf call is traced by fexit. +*/ +static void test_tailcall_bpf2bpf_fexit (void ) +{ +__tailcall_bpf2bpf_fentry_fexit (false, true); +} + +/* test_tailcall_bpf2bpf_fentry_fexit checks that the count value of the tail call +* limit enforcement matches with expectations when tailcall is preceded with +* bpf2bpf call, and the bpf2bpf call is traced by both fentry and fexit. +*/ +static void test_tailcall_bpf2bpf_fentry_fexit (void ) +{ +__tailcall_bpf2bpf_fentry_fexit (true, true); +} +void test_tailcalls (void ) {if (test__start_subtest ("tailcall_1" ))
@@ -910,4 +1095,11 @@ void test_tailcalls(void)test_tailcall_bpf2bpf_4 (true);if (test__start_subtest ("tailcall_bpf2bpf_6" ))test_tailcall_bpf2bpf_6 (); +if (test__start_subtest ("tailcall_bpf2bpf_fentry" )) +test_tailcall_bpf2bpf_fentry (); +if (test__start_subtest ("tailcall_bpf2bpf_fexit" )) +test_tailcall_bpf2bpf_fexit (); +if (test__start_subtest ("tailcall_bpf2bpf_fentry_fexit" )) +test_tailcall_bpf2bpf_fentry_fexit (); } +
tools/testing/selftests/bpf/progs/tailcall_bpf2bpf_fentry.c new file +18 -0
@@ -0,0 +1,18 @@ +// SPDX-License-Identifier: GPL-2.0 +/* Copyright Leon Hwang */ + +#include "vmlinux.h" +#include <bpf/bpf_helpers.h> +#include <bpf/bpf_tracing.h> + +int count =0 ; + +SEC("fentry/subprog_tail") +int BPF_PROG (fentry ,struct sk_buff *skb) +{ + count++; + +return 0 ; +} + +char _license[] SEC("license") = "GPL";
tools/testing/selftests/bpf/progs/tailcall_bpf2bpf_fexit.c new file +18 -0
@@ -0,0 +1,18 @@ +// SPDX-License-Identifier: GPL-2.0 +/* Copyright Leon Hwang */ + +#include "vmlinux.h" +#include <bpf/bpf_helpers.h> +#include <bpf/bpf_tracing.h> + +int count =0 ; + +SEC("fexit/subprog_tail") +int BPF_PROG (fexit ,struct sk_buff *skb) +{ + count++; + +return 0 ; +} + +char _license[] SEC("license") = "GPL";
Signature · 1 line
2.41.0