Skip to contentSkip to conversation

bpf: hide unused bpf_patch_call_args

PATCH Superseded Arnd Bergmann 3 participants 8 messages May 23, 2023 – May 24, 2023

Subsystems (6)
Stable branch BPF Tracing BPF core +2 more subsystems

v2 of 3 Superseded Latest is v3 →

lore
Share & tools
lore
#0 lore raw

PATCH v2 1/2 bpf: hide unused bpf_patch_call_args

From: Arnd Bergmann <[email protected]>

This function has no callers and no declaration when CONFIG_BPF_JIT_ALWAYS_ON is enabled:

kernel/bpf/core.c:2075:6: error: no previous prototype for 'bpf_patch_call_args' [-Werror=missing-prototypes]

Hide the definition as well.

---
v2: change indentation to align arguments better. Still not great as the line is just too long

---
 kernel/bpf/core.c | 7 +++++--
 1 file changed, 5 insertions(+), 2 deletions(-)
kernel/bpf/core.c +5 -2
index 7421487422d4..0926714641eb 100644
@@ -2064,14 +2064,16 @@ 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) = {
+static __maybe_unused
+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
 
+#ifdef CONFIG_BPF_SYSCALL
 void bpf_patch_call_args(struct bpf_insn *insn, u32 stack_depth)
 {
 	stack_depth = max_t(u32, stack_depth, 1);
@@ -2080,6 +2082,7 @@ void bpf_patch_call_args(struct bpf_insn *insn, u32 stack_depth)
 		__bpf_call_base_args;
 	insn->code = BPF_JMP | BPF_CALL_ARGS;
 }
+#endif
 
 #else
 static unsigned int __bpf_prog_ret0_warn(const void *ctx,
Signature · 1 line
2.39.2
#1 lore raw

PATCH v2 2/2 bpf: fix bpf_probe_read_kernel prototype mismatch

From: Arnd Bergmann <[email protected]>

bpf_probe_read_kernel() has a __weak definition in core.c and another definition with an incompatible prototype in kernel/trace/bpf_trace.c, when CONFIG_BPF_EVENTS is enabled.

Since the two are incompatible, there cannot be a shared declaration in a header file, but the lack of a prototype causes a W=1 warning:

kernel/bpf/core.c:1638:12: error: no previous prototype for 'bpf_probe_read_kernel' [-Werror=missing-prototypes]

Change the core.c file to only reference the inner bpf_probe_read_kernel_common() helper and provide a prototype for that.

Aside from the warning, this addresses a bug on 32-bit architectures from incorrect argument passing with the mismatched prototype.

--
v2: rewrite completely to fix the mismatch.

---
 include/linux/bpf.h      | 2 ++
 kernel/bpf/core.c        | 9 ++++++---
 kernel/trace/bpf_trace.c | 3 +--
 3 files changed, 9 insertions(+), 5 deletions(-)
include/linux/bpf.h +2 -0
index 36e4b2d8cca2..6a231ec61a87 100644
@@ -2619,6 +2619,8 @@ static inline void bpf_dynptr_set_rdonly(struct bpf_dynptr_kern *ptr)
 }
 #endif /* CONFIG_BPF_SYSCALL */
 
+int bpf_probe_read_kernel_common(void *dst, u32 size, const void *unsafe_ptr);
+
 void __bpf_free_used_btfs(struct bpf_prog_aux *aux,
 			  struct btf_mod_pair *used_btfs, u32 len);
 
kernel/bpf/core.c +6 -3
index 0926714641eb..e7f0d5f146fa 100644
@@ -35,6 +35,7 @@
 #include <linux/bpf_verifier.h>
 #include <linux/nodemask.h>
 #include <linux/nospec.h>
+#include <linux/bpf.h>
 #include <linux/bpf_mem_alloc.h>
 #include <linux/memcontrol.h>
 
@@ -1635,11 +1636,13 @@ bool bpf_opcode_in_insntable(u8 code)
 }
 
 #ifndef CONFIG_BPF_JIT_ALWAYS_ON
-u64 __weak bpf_probe_read_kernel(void *dst, u32 size, const void *unsafe_ptr)
+#ifndef CONFIG_BPF_EVENTS
+int bpf_probe_read_kernel_common(void * dst, u32 size, const void *unsafe_ptr)
 {
 	memset(dst, 0, size);
 	return -EFAULT;
 }
+#endif
 
 /**
  *	___bpf_prog_run - run eBPF program on a given context
@@ -1931,8 +1934,8 @@ static u64 ___bpf_prog_run(u64 *regs, const struct bpf_insn *insn)
 		DST = *(SIZE *)(unsigned long) (SRC + insn->off);	\
 		CONT;							\
 	LDX_PROBE_MEM_##SIZEOP:						\
-		bpf_probe_read_kernel(&DST, sizeof(SIZE),		\
-				      (const void *)(long) (SRC + insn->off));	\
+		bpf_probe_read_kernel_common(&DST, sizeof(SIZE),	\
+			      (const void *)(long) (SRC + insn->off));	\
 		DST = *((SIZE *)&DST);					\
 		CONT;
 
kernel/trace/bpf_trace.c +1 -2
index 2bc41e6ac9fe..290fdce2ce53 100644
@@ -223,8 +223,7 @@ const struct bpf_func_proto bpf_probe_read_user_str_proto = {
 	.arg3_type	= ARG_ANYTHING,
 };
 
-static __always_inline int
-bpf_probe_read_kernel_common(void *dst, u32 size, const void *unsafe_ptr)
+int bpf_probe_read_kernel_common(void *dst, u32 size, const void *unsafe_ptr)
 {
 	int ret;
 
Signature · 1 line
2.39.2
#2 lore raw

Re: [PATCH 2/2] [v2] bpf: fix bpf_probe_read_kernel prototype mismatch

Quoted 9 lines · Arnd Bergmann

From: Arnd Bergmann <[email protected]>

bpf_probe_read_kernel() has a __weak definition in core.c and another definition with an incompatible prototype in kernel/trace/bpf_trace.c, when CONFIG_BPF_EVENTS is enabled.

Since the two are incompatible, there cannot be a shared declaration in a header file, but the lack of a prototype causes a W=1 warning:

Quoted 7 lines

kernel/bpf/core.c:1638:12: error: no previous prototype for 'bpf_probe_read_kernel' [-Werror=missing-prototypes]

Change the core.c file to only reference the inner bpf_probe_read_kernel_common() helper and provide a prototype for that.

Aside from the warning, this addresses a bug on 32-bit architectures from incorrect argument passing with the mismatched prototype.

Could you explain what is this '32-bit architectures ... incorrect argument passing' thing?

Quoted 37 lines

--
v2: rewrite completely to fix the mismatch.

---
  include/linux/bpf.h      | 2 ++
  kernel/bpf/core.c        | 9 ++++++---
  kernel/trace/bpf_trace.c | 3 +--
  3 files changed, 9 insertions(+), 5 deletions(-)
include/linux/bpf.h +2 -0
index 36e4b2d8cca2..6a231ec61a87 100644
@@ -2619,6 +2619,8 @@ static inline void bpf_dynptr_set_rdonly(struct bpf_dynptr_kern *ptr)
  }
  #endif /* CONFIG_BPF_SYSCALL */
 
+int bpf_probe_read_kernel_common(void *dst, u32 size, const void *unsafe_ptr);
+
  void __bpf_free_used_btfs(struct bpf_prog_aux *aux,
  			  struct btf_mod_pair *used_btfs, u32 len);
 
kernel/bpf/core.c +1 -0
index 0926714641eb..e7f0d5f146fa 100644
@@ -35,6 +35,7 @@
  #include <linux/bpf_verifier.h>
  #include <linux/nodemask.h>
  #include <linux/nospec.h>
+#include <linux/bpf.h>
  #include <linux/bpf_mem_alloc.h>
  #include <linux/memcontrol.h>
 

kernel/bpf/core.c bool bpf_opcode_in_insntable(u8 code) · line 1640

Quoted 6 lines
kernel/bpf/core.c +2 -1
@@ -1635,11 +1636,13 @@ bool bpf_opcode_in_insntable(u8 code)
  }
 
  #ifndef CONFIG_BPF_JIT_ALWAYS_ON
-u64 __weak bpf_probe_read_kernel(void *dst, u32 size, const void *unsafe_ptr)
+#ifndef CONFIG_BPF_EVENTS
+int bpf_probe_read_kernel_common(void * dst, u32 size, const void *unsafe_ptr)

void * dst => void *dst

Quoted 33 lines
kernel/bpf/core.c +3 -2
  {
  	memset(dst, 0, size);
  	return -EFAULT;
  }
+#endif
 
  /**
   *	___bpf_prog_run - run eBPF program on a given context
@@ -1931,8 +1934,8 @@ static u64 ___bpf_prog_run(u64 *regs, const struct bpf_insn *insn)
  		DST = *(SIZE *)(unsigned long) (SRC + insn->off);	\
  		CONT;							\
  	LDX_PROBE_MEM_##SIZEOP:						\
-		bpf_probe_read_kernel(&DST, sizeof(SIZE),		\
-				      (const void *)(long) (SRC + insn->off));	\
+		bpf_probe_read_kernel_common(&DST, sizeof(SIZE),	\
+			      (const void *)(long) (SRC + insn->off));	\
  		DST = *((SIZE *)&DST);					\
  		CONT;
 
kernel/trace/bpf_trace.c +1 -2
index 2bc41e6ac9fe..290fdce2ce53 100644
@@ -223,8 +223,7 @@ const struct bpf_func_proto bpf_probe_read_user_str_proto = {
  	.arg3_type	= ARG_ANYTHING,
  };
 
-static __always_inline int
-bpf_probe_read_kernel_common(void *dst, u32 size, const void *unsafe_ptr)
+int bpf_probe_read_kernel_common(void *dst, u32 size, const void *unsafe_ptr)
  {
  	int ret;
 
#3 lore raw

Re: [PATCH 2/2] [v2] bpf: fix bpf_probe_read_kernel prototype mismatch

Quoted 7 lines · Yonghong Song
Quoted 2 lines · Arnd Bergmann

Aside from the warning, this addresses a bug on 32-bit architectures from incorrect argument passing with the mismatched prototype.

Could you explain what is this '32-bit architectures ... incorrect argument passing' thing?

I've expanded that paragraph now:

| Aside from the warning, this addresses a bug on 32-bit architectures
| from incorrect argument passing with the mismatched prototype:
| BPF_CALL_x() functions use 64-bit arguments that are passed in
| pairs of register or on the stack on 32-bit architectures, while the
| normal function uses one register per argument.

Let me know if you think I should put more details in there.

Quoted 10 lines

kernel/bpf/core.c bool bpf_opcode_in_insntable(u8 code) · line 1640

Quoted 7 lines
kernel/bpf/core.c +2 -1
@@ -1635,11 +1636,13 @@ bool bpf_opcode_in_insntable(u8 code)
  }
 
  #ifndef CONFIG_BPF_JIT_ALWAYS_ON
-u64 __weak bpf_probe_read_kernel(void *dst, u32 size, const void *unsafe_ptr)
+#ifndef CONFIG_BPF_EVENTS
+int bpf_probe_read_kernel_common(void * dst, u32 size, const void *unsafe_ptr)

void * dst => void *dst

Fixed now.

Thanks,

Arnd

#4 lore raw

Re: [PATCH 2/2] [v2] bpf: fix bpf_probe_read_kernel prototype mismatch

Quoted 11 lines · Arnd Bergmann
Quoted 7 lines · Yonghong Song
Quoted 2 lines · Arnd Bergmann

Aside from the warning, this addresses a bug on 32-bit architectures from incorrect argument passing with the mismatched prototype.

Could you explain what is this '32-bit architectures ... incorrect argument passing' thing?

I've expanded that paragraph now:

Quoted 7 lines
| Aside from the warning, this addresses a bug on 32-bit architectures
| from incorrect argument passing with the mismatched prototype:
| BPF_CALL_x() functions use 64-bit arguments that are passed in
| pairs of register or on the stack on 32-bit architectures, while the
| normal function uses one register per argument.

Let me know if you think I should put more details in there.

Please mention the function you try to address for the bug on 32-bit architecture is:

u64 __weak bpf_probe_read_kernel(void *dst, u32 size, const void *unsafe_ptr)

which will be incompatible with BPF_CALL_3(bpf_probe_read_kernel, void *, dst, u32, size,
const void *, unsafe_ptr)
in bpf_trace.c.

So you fixed this bug by using internal function bpf_probe_read_kernel_common() instead.

Thanks.

Quoted 17 lines
Quoted 11 lines

kernel/bpf/core.c bool bpf_opcode_in_insntable(u8 code) · line 1640

Quoted 8 lines
kernel/bpf/core.c +2 -1
@@ -1635,11 +1636,13 @@ bool bpf_opcode_in_insntable(u8 code)
   }
 
   #ifndef CONFIG_BPF_JIT_ALWAYS_ON
-u64 __weak bpf_probe_read_kernel(void *dst, u32 size, const void *unsafe_ptr)
+#ifndef CONFIG_BPF_EVENTS
+int bpf_probe_read_kernel_common(void * dst, u32 size, const void *unsafe_ptr)

void * dst => void *dst

Fixed now.

Thanks,

Arnd

#5 lore raw
In reply to Arnd Bergmann
Quoted 6 lines · Arnd Bergmann

From: Arnd Bergmann <[email protected]>

This function has no callers and no declaration when CONFIG_BPF_JIT_ALWAYS_ON is enabled:

kernel/bpf/core.c:2075:6: error: no previous prototype for 'bpf_patch_call_args' [-Werror=missing-prototypes]

If CONFIG_BPF_JIT_ALWAYS_ON is enabled, the definition of bpf_patch_call_args should be invisible. Maybe I missed something. Could you list *ALL& bpf related config options in your setup so people can reproduce you above error messages?

Quoted 42 lines

Hide the definition as well.

---
v2: change indentation to align arguments better. Still not great as the line is just too long

---
  kernel/bpf/core.c | 7 +++++--
  1 file changed, 5 insertions(+), 2 deletions(-)
kernel/bpf/core.c +5 -2
index 7421487422d4..0926714641eb 100644
@@ -2064,14 +2064,16 @@ 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) = {
+static __maybe_unused
+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
 
+#ifdef CONFIG_BPF_SYSCALL
  void bpf_patch_call_args(struct bpf_insn *insn, u32 stack_depth)
  {
  	stack_depth = max_t(u32, stack_depth, 1);
@@ -2080,6 +2082,7 @@ void bpf_patch_call_args(struct bpf_insn *insn, u32 stack_depth)
  		__bpf_call_base_args;
  	insn->code = BPF_JMP | BPF_CALL_ARGS;
  }
+#endif
 
  #else
  static unsigned int __bpf_prog_ret0_warn(const void *ctx,
#6 lore raw
Quoted 6 lines · Yonghong Song
Quoted 5 lines · Arnd Bergmann

From: Arnd Bergmann <[email protected]>

This function has no callers and no declaration when CONFIG_BPF_JIT_ALWAYS_ON is enabled:

Quoted 6 lines
Quoted 1 line

kernel/bpf/core.c:2075:6: error: no previous prototype for 'bpf_patch_call_args' [-Werror=missing-prototypes]

If CONFIG_BPF_JIT_ALWAYS_ON is enabled, the definition of bpf_patch_call_args should be invisible. Maybe I missed something. Could you list *ALL& bpf related config options in your setup so people can reproduce you above error messages?

Sorry, my mistake. I've reworded the changelog now to fix this:

| This function is only used when CONFIG_BPF_JIT_ALWAYS_ON is disabled,
| but CONFIG_BPF_SYSCALL is enabled. When both are turned off, the
| prototype is missing but the unused function is still compiled,
| as seen from this W=1 warning:
| 
| kernel/bpf/core.c:2075:6: error: no previous prototype for 'bpf_patch_call_args' [-Werror=missing-prototypes]
| 
| Add a matching #ifdef for the definition to leave it out.

If this makes sense now, I'll send out a v3.

Arnd

#7 lore raw
Quoted 25 lines · Arnd Bergmann
Quoted 6 lines · Yonghong Song
Quoted 5 lines · Arnd Bergmann

From: Arnd Bergmann <[email protected]>

This function has no callers and no declaration when CONFIG_BPF_JIT_ALWAYS_ON is enabled:

Quoted 6 lines
Quoted 1 line

kernel/bpf/core.c:2075:6: error: no previous prototype for 'bpf_patch_call_args' [-Werror=missing-prototypes]

If CONFIG_BPF_JIT_ALWAYS_ON is enabled, the definition of bpf_patch_call_args should be invisible. Maybe I missed something. Could you list *ALL& bpf related config options in your setup so people can reproduce you above error messages?

Sorry, my mistake. I've reworded the changelog now to fix this:

| This function is only used when CONFIG_BPF_JIT_ALWAYS_ON is disabled,
| but CONFIG_BPF_SYSCALL is enabled. When both are turned off, the
| prototype is missing but the unused function is still compiled,
| as seen from this W=1 warning:
|
| kernel/bpf/core.c:2075:6: error: no previous prototype for 'bpf_patch_call_args' [-Werror=missing-prototypes]
|
| Add a matching #ifdef for the definition to leave it out.
Quoted 1 line

If this makes sense now, I'll send out a v3.

Yes, the above new commit message sounds good to me. Thanks.

Quoted 2 lines

Arnd