[PATCHv2] btf_encoder: Match ftrace addresses within elf functions

Currently when processing DWARF function, we check its entrypoint
against ftrace addresses, assuming that the ftrace address matches
with function's entrypoint.

This is not the case on some architectures as reported by Nathan
when building kernel on arm [1].

Fixing the check to take into account the whole function not
just the entrypoint.

Most of the is_ftrace_func code was contributed by Andrii.

[1] https://lore.kernel.org/bpf/20210209034416.GA1669105@ubuntu-m3-large-x86/
Acked-by: Andrii Nakryiko <andrii@kernel.org>
Signed-off-by: Jiri Olsa <jolsa@kernel.org>
---
v2 changes:
  - update functions addr directly [Andrii]

 btf_encoder.c | 40 ++++++++++++++++++++++++++++++++++++++--
 1 file changed, 38 insertions(+), 2 deletions(-)

diff --git a/btf_encoder.c b/btf_encoder.c
index b124ec20a689..80e896961d4e 100644
--- a/btf_encoder.c
+++ b/btf_encoder.c

@@ -36,6 +36,7 @@ struct funcs_layout {
 struct elf_function {
 	const char	*name;
 	unsigned long	 addr;
+	unsigned long	 size;
 	unsigned long	 sh_addr;
 	bool		 generated;
 };

@@ -98,6 +99,7 @@ static int collect_function(struct btf_elf *btfe, GElf_Sym *sym,
 
 	functions[functions_cnt].name = name;
 	functions[functions_cnt].addr = elf_sym__value(sym);
+	functions[functions_cnt].size = elf_sym__size(sym);
 	functions[functions_cnt].sh_addr = sh.sh_addr;
 	functions[functions_cnt].generated = false;
 	functions_cnt++;

@@ -236,6 +238,39 @@ get_kmod_addrs(struct btf_elf *btfe, __u64 **paddrs, __u64 *pcount)
 	return 0;
 }
 
+static int is_ftrace_func(struct elf_function *func, __u64 *addrs, __u64 count)
+{
+	__u64 start = func->addr;
+	__u64 addr, end = func->addr + func->size;
+
+	/*
+	 * The invariant here is addr[r] that is the smallest address
+	 * that is >= than function start addr. Except the corner case
+	 * where there is no such r, but for that we have a final check
+	 * in the return.
+	 */
+	size_t l = 0, r = count - 1, m;
+
+	/* make sure we don't use invalid r */
+	if (count == 0)
+		return false;
+
+	while (l < r) {
+		m = l + (r - l) / 2;
+		addr = addrs[m];
+
+		if (addr >= start) {
+			/* we satisfy invariant, so tighten r */
+			r = m;
+		} else {
+			/* m is not good enough as l, maybe m + 1 will be */
+			l = m + 1;
+		}
+	}
+
+	return start <= addrs[r] && addrs[r] < end;
+}
+
 static int setup_functions(struct btf_elf *btfe, struct funcs_layout *fl)
 {
 	__u64 *addrs, count, i;

@@ -283,10 +318,11 @@ static int setup_functions(struct btf_elf *btfe, struct funcs_layout *fl)
 		 * functions[x]::addr is relative address within section
 		 * and needs to be relocated by adding sh_addr.
 		 */
-		__u64 addr = kmod ? func->addr + func->sh_addr : func->addr;
+		if (kmod)
+			func->addr += func->sh_addr;
 
 		/* Make sure function is within ftrace addresses. */
-		if (bsearch(&addr, addrs, count, sizeof(addrs[0]), addrs_cmp)) {
+		if (is_ftrace_func(func, addrs, count)) {
 			/*
 			 * We iterate over sorted array, so we can easily skip
 			 * not valid item and move following valid field into

-- 
2.29.2

On Sat, Feb 13, 2021 at 5:46 PM Jiri Olsa [off-list ref] wrote:

Currently when processing DWARF function, we check its entrypoint
against ftrace addresses, assuming that the ftrace address matches
with function's entrypoint.

This is not the case on some architectures as reported by Nathan
when building kernel on arm [1].

Fixing the check to take into account the whole function not
just the entrypoint.

Most of the is_ftrace_func code was contributed by Andrii.

[1] https://lore.kernel.org/bpf/20210209034416.GA1669105@ubuntu-m3-large-x86/
Acked-by: Andrii Nakryiko <andrii@kernel.org>
Signed-off-by: Jiri Olsa <jolsa@kernel.org>

Tested this v2 together with "btf_encoder: sanitize non-regular int
base type" v2 on top of pahole v1.20

Tested-by: Sedat Dilek <redacted> # Linux v5.11-rc7+ and
LLVM/Clang v12.0.0-rc1 on x86 (64bit)

- Sedat -

On Sat, Feb 13, 2021 at 05:46:48PM +0100, Jiri Olsa wrote:

Currently when processing DWARF function, we check its entrypoint
against ftrace addresses, assuming that the ftrace address matches
with function's entrypoint.

This is not the case on some architectures as reported by Nathan
when building kernel on arm [1].

Fixing the check to take into account the whole function not
just the entrypoint.

Most of the is_ftrace_func code was contributed by Andrii.

[1] https://lore.kernel.org/bpf/20210209034416.GA1669105@ubuntu-m3-large-x86/
Acked-by: Andrii Nakryiko <andrii@kernel.org>
Signed-off-by: Jiri Olsa <jolsa@kernel.org>

I did several builds with CONFIG_DEBUG_INFO_BTF enabled (arm64, ppc64le,
and x86_64) and saw no build errors. I did not do any runtime testing.

Tested-by: Nathan Chancellor <nathan@kernel.org>

Em Sat, Feb 13, 2021 at 05:46:48PM +0100, Jiri Olsa escreveu:

Currently when processing DWARF function, we check its entrypoint
against ftrace addresses, assuming that the ftrace address matches
with function's entrypoint.

This is not the case on some architectures as reported by Nathan
when building kernel on arm [1].

Fixing the check to take into account the whole function not
just the entrypoint.

Most of the is_ftrace_func code was contributed by Andrii.

Applied locally, will go out after tests,

- Arnaldo
 

-- 

- Arnaldo

Em Sat, Feb 13, 2021 at 08:11:40PM +0100, Sedat Dilek escreveu:

On Sat, Feb 13, 2021 at 5:46 PM Jiri Olsa [off-list ref] wrote:

quoted

Currently when processing DWARF function, we check its entrypoint
against ftrace addresses, assuming that the ftrace address matches
with function's entrypoint.

This is not the case on some architectures as reported by Nathan
when building kernel on arm [1].

Fixing the check to take into account the whole function not
just the entrypoint.

Most of the is_ftrace_func code was contributed by Andrii.

[1] https://lore.kernel.org/bpf/20210209034416.GA1669105@ubuntu-m3-large-x86/
Acked-by: Andrii Nakryiko <andrii@kernel.org>
Signed-off-by: Jiri Olsa <jolsa@kernel.org>

Tested this v2 together with "btf_encoder: sanitize non-regular int
base type" v2 on top of pahole v1.20

Tested-by: Sedat Dilek <redacted> # Linux v5.11-rc7+ and
LLVM/Clang v12.0.0-rc1 on x86 (64bit)

Thanks, added to the cset,

- Arnaldo
 

- Sedat -

quoted

---
v2 changes:
  - update functions addr directly [Andrii]

 btf_encoder.c | 40 ++++++++++++++++++++++++++++++++++++++--
 1 file changed, 38 insertions(+), 2 deletions(-)

diff --git a/btf_encoder.c b/btf_encoder.c
index b124ec20a689..80e896961d4e 100644
--- a/btf_encoder.c
+++ b/btf_encoder.c

@@ -36,6 +36,7 @@ struct funcs_layout {
 struct elf_function {
        const char      *name;
        unsigned long    addr;
+       unsigned long    size;
        unsigned long    sh_addr;
        bool             generated;
 };

@@ -98,6 +99,7 @@ static int collect_function(struct btf_elf *btfe, GElf_Sym *sym,

        functions[functions_cnt].name = name;
        functions[functions_cnt].addr = elf_sym__value(sym);
+       functions[functions_cnt].size = elf_sym__size(sym);
        functions[functions_cnt].sh_addr = sh.sh_addr;
        functions[functions_cnt].generated = false;
        functions_cnt++;

@@ -236,6 +238,39 @@ get_kmod_addrs(struct btf_elf *btfe, __u64 **paddrs, __u64 *pcount)
        return 0;
 }

+static int is_ftrace_func(struct elf_function *func, __u64 *addrs, __u64 count)
+{
+       __u64 start = func->addr;
+       __u64 addr, end = func->addr + func->size;
+
+       /*
+        * The invariant here is addr[r] that is the smallest address
+        * that is >= than function start addr. Except the corner case
+        * where there is no such r, but for that we have a final check
+        * in the return.
+        */
+       size_t l = 0, r = count - 1, m;
+
+       /* make sure we don't use invalid r */
+       if (count == 0)
+               return false;
+
+       while (l < r) {
+               m = l + (r - l) / 2;
+               addr = addrs[m];
+
+               if (addr >= start) {
+                       /* we satisfy invariant, so tighten r */
+                       r = m;
+               } else {
+                       /* m is not good enough as l, maybe m + 1 will be */
+                       l = m + 1;
+               }
+       }
+
+       return start <= addrs[r] && addrs[r] < end;
+}
+
 static int setup_functions(struct btf_elf *btfe, struct funcs_layout *fl)
 {
        __u64 *addrs, count, i;

@@ -283,10 +318,11 @@ static int setup_functions(struct btf_elf *btfe, struct funcs_layout *fl)
                 * functions[x]::addr is relative address within section
                 * and needs to be relocated by adding sh_addr.
                 */
-               __u64 addr = kmod ? func->addr + func->sh_addr : func->addr;
+               if (kmod)
+                       func->addr += func->sh_addr;

                /* Make sure function is within ftrace addresses. */
-               if (bsearch(&addr, addrs, count, sizeof(addrs[0]), addrs_cmp)) {
+               if (is_ftrace_func(func, addrs, count)) {
                        /*
                         * We iterate over sorted array, so we can easily skip
                         * not valid item and move following valid field into
--

2.29.2

-- 

- Arnaldo

Em Sat, Feb 13, 2021 at 07:30:48PM -0700, Nathan Chancellor escreveu:

On Sat, Feb 13, 2021 at 05:46:48PM +0100, Jiri Olsa wrote:

quoted

Currently when processing DWARF function, we check its entrypoint
against ftrace addresses, assuming that the ftrace address matches
with function's entrypoint.

This is not the case on some architectures as reported by Nathan
when building kernel on arm [1].

Fixing the check to take into account the whole function not
just the entrypoint.

Most of the is_ftrace_func code was contributed by Andrii.

[1] https://lore.kernel.org/bpf/20210209034416.GA1669105@ubuntu-m3-large-x86/
Acked-by: Andrii Nakryiko <andrii@kernel.org>
Signed-off-by: Jiri Olsa <jolsa@kernel.org>

I did several builds with CONFIG_DEBUG_INFO_BTF enabled (arm64, ppc64le,
and x86_64) and saw no build errors. I did not do any runtime testing.

Tested-by: Nathan Chancellor <nathan@kernel.org>

Thanks, added to the cset,

- Arnaldo
 

quoted

---
v2 changes:
  - update functions addr directly [Andrii]

 btf_encoder.c | 40 ++++++++++++++++++++++++++++++++++++++--
 1 file changed, 38 insertions(+), 2 deletions(-)

diff --git a/btf_encoder.c b/btf_encoder.c
index b124ec20a689..80e896961d4e 100644
--- a/btf_encoder.c
+++ b/btf_encoder.c

@@ -36,6 +36,7 @@ struct funcs_layout {
 struct elf_function {
 	const char	*name;
 	unsigned long	 addr;
+	unsigned long	 size;
 	unsigned long	 sh_addr;
 	bool		 generated;
 };

@@ -98,6 +99,7 @@ static int collect_function(struct btf_elf *btfe, GElf_Sym *sym,
 
 	functions[functions_cnt].name = name;
 	functions[functions_cnt].addr = elf_sym__value(sym);
+	functions[functions_cnt].size = elf_sym__size(sym);
 	functions[functions_cnt].sh_addr = sh.sh_addr;
 	functions[functions_cnt].generated = false;
 	functions_cnt++;

@@ -236,6 +238,39 @@ get_kmod_addrs(struct btf_elf *btfe, __u64 **paddrs, __u64 *pcount)
 	return 0;
 }
 
+static int is_ftrace_func(struct elf_function *func, __u64 *addrs, __u64 count)
+{
+	__u64 start = func->addr;
+	__u64 addr, end = func->addr + func->size;
+
+	/*
+	 * The invariant here is addr[r] that is the smallest address
+	 * that is >= than function start addr. Except the corner case
+	 * where there is no such r, but for that we have a final check
+	 * in the return.
+	 */
+	size_t l = 0, r = count - 1, m;
+
+	/* make sure we don't use invalid r */
+	if (count == 0)
+		return false;
+
+	while (l < r) {
+		m = l + (r - l) / 2;
+		addr = addrs[m];
+
+		if (addr >= start) {
+			/* we satisfy invariant, so tighten r */
+			r = m;
+		} else {
+			/* m is not good enough as l, maybe m + 1 will be */
+			l = m + 1;
+		}
+	}
+
+	return start <= addrs[r] && addrs[r] < end;
+}
+
 static int setup_functions(struct btf_elf *btfe, struct funcs_layout *fl)
 {
 	__u64 *addrs, count, i;

@@ -283,10 +318,11 @@ static int setup_functions(struct btf_elf *btfe, struct funcs_layout *fl)
 		 * functions[x]::addr is relative address within section
 		 * and needs to be relocated by adding sh_addr.
 		 */
-		__u64 addr = kmod ? func->addr + func->sh_addr : func->addr;
+		if (kmod)
+			func->addr += func->sh_addr;
 
 		/* Make sure function is within ftrace addresses. */
-		if (bsearch(&addr, addrs, count, sizeof(addrs[0]), addrs_cmp)) {
+		if (is_ftrace_func(func, addrs, count)) {
 			/*
 			 * We iterate over sorted array, so we can easily skip
 			 * not valid item and move following valid field into

-- 
2.29.2

-- 

- Arnaldo

On Wed, Feb 17, 2021 at 1:44 PM Arnaldo Carvalho de Melo
[off-list ref] wrote:

Em Sat, Feb 13, 2021 at 05:46:48PM +0100, Jiri Olsa escreveu:

quoted

Currently when processing DWARF function, we check its entrypoint
against ftrace addresses, assuming that the ftrace address matches
with function's entrypoint.

This is not the case on some architectures as reported by Nathan
when building kernel on arm [1].

Fixing the check to take into account the whole function not
just the entrypoint.

Most of the is_ftrace_func code was contributed by Andrii.

Applied locally, will go out after tests,

Hi Arnaldo,

Is it possible to have a pahole version 1.21 with this patch and the
one from Yonghong Son?

From my local pahole Git:

$ git log --oneline --no-merges v1.20..
2f83aefdbddf (for-1.20/btf_encoder-ftrace_elf-clang-jolsa-v2)
btf_encoder: Match ftrace addresses within elf functions
f21eafdfc877 (for-1.20/btf_encoder-sanitized_int-clang-yhs-v2)
btf_encoder: sanitize non-regular int base type

Both patches fixes all issues seen so far with LLVM/Clang >=
12.0.0-rc1 and DWARF-v5 and BTF (debug-info) and pahole on
Linux/x86_64 and according to Nathan on Linux/arm64.
Yesterday, I tried with LLVM/Clang 13-git from <apt.llvm.org>.

BTW, Nick's DWARF-v5 patches are pending in <kbuild.git#kbuild> (see [1]).

Personally, I can wait until [1] is in Linus Git.

Please, let me/us know what you are planning.
( I know it is Linux v5.12 merge-window. )

Thanks.

Regards,
- Sedat -

[1] https://git.kernel.org/pub/scm/linux/kernel/git/masahiroy/linux-kbuild.git/log/?h=kbuild

- Arnaldo

quoted

[1] https://lore.kernel.org/bpf/20210209034416.GA1669105@ubuntu-m3-large-x86/
Acked-by: Andrii Nakryiko <andrii@kernel.org>
Signed-off-by: Jiri Olsa <jolsa@kernel.org>
---
v2 changes:
  - update functions addr directly [Andrii]

 btf_encoder.c | 40 ++++++++++++++++++++++++++++++++++++++--
 1 file changed, 38 insertions(+), 2 deletions(-)

diff --git a/btf_encoder.c b/btf_encoder.c
index b124ec20a689..80e896961d4e 100644
--- a/btf_encoder.c
+++ b/btf_encoder.c

@@ -36,6 +36,7 @@ struct funcs_layout {
 struct elf_function {
      const char      *name;
      unsigned long    addr;
+     unsigned long    size;
      unsigned long    sh_addr;
      bool             generated;
 };

@@ -98,6 +99,7 @@ static int collect_function(struct btf_elf *btfe, GElf_Sym *sym,

      functions[functions_cnt].name = name;
      functions[functions_cnt].addr = elf_sym__value(sym);
+     functions[functions_cnt].size = elf_sym__size(sym);
      functions[functions_cnt].sh_addr = sh.sh_addr;
      functions[functions_cnt].generated = false;
      functions_cnt++;

@@ -236,6 +238,39 @@ get_kmod_addrs(struct btf_elf *btfe, __u64 **paddrs, __u64 *pcount)
      return 0;
 }

+static int is_ftrace_func(struct elf_function *func, __u64 *addrs, __u64 count)
+{
+     __u64 start = func->addr;
+     __u64 addr, end = func->addr + func->size;
+
+     /*
+      * The invariant here is addr[r] that is the smallest address
+      * that is >= than function start addr. Except the corner case
+      * where there is no such r, but for that we have a final check
+      * in the return.
+      */
+     size_t l = 0, r = count - 1, m;
+
+     /* make sure we don't use invalid r */
+     if (count == 0)
+             return false;
+
+     while (l < r) {
+             m = l + (r - l) / 2;
+             addr = addrs[m];
+
+             if (addr >= start) {
+                     /* we satisfy invariant, so tighten r */
+                     r = m;
+             } else {
+                     /* m is not good enough as l, maybe m + 1 will be */
+                     l = m + 1;
+             }
+     }
+
+     return start <= addrs[r] && addrs[r] < end;
+}
+
 static int setup_functions(struct btf_elf *btfe, struct funcs_layout *fl)
 {
      __u64 *addrs, count, i;

@@ -283,10 +318,11 @@ static int setup_functions(struct btf_elf *btfe, struct funcs_layout *fl)
               * functions[x]::addr is relative address within section
               * and needs to be relocated by adding sh_addr.
               */
-             __u64 addr = kmod ? func->addr + func->sh_addr : func->addr;
+             if (kmod)
+                     func->addr += func->sh_addr;

              /* Make sure function is within ftrace addresses. */
-             if (bsearch(&addr, addrs, count, sizeof(addrs[0]), addrs_cmp)) {
+             if (is_ftrace_func(func, addrs, count)) {
                      /*
                       * We iterate over sorted array, so we can easily skip
                       * not valid item and move following valid field into
--

2.29.2

--

- Arnaldo

On February 17, 2021 10:40:43 AM GMT-03:00, Sedat Dilek [off-list ref] wrote:

On Wed, Feb 17, 2021 at 1:44 PM Arnaldo Carvalho de Melo
[off-list ref] wrote:

quoted

Em Sat, Feb 13, 2021 at 05:46:48PM +0100, Jiri Olsa escreveu:

quoted

Currently when processing DWARF function, we check its entrypoint
against ftrace addresses, assuming that the ftrace address matches
with function's entrypoint.

This is not the case on some architectures as reported by Nathan
when building kernel on arm [1].

Fixing the check to take into account the whole function not
just the entrypoint.

Most of the is_ftrace_func code was contributed by Andrii.

Applied locally, will go out after tests,

Hi Arnaldo,

Is it possible to have a pahole version 1.21 with this patch and the
one from Yonghong Son?

From my local pahole Git:

$ git log --oneline --no-merges v1.20..
2f83aefdbddf (for-1.20/btf_encoder-ftrace_elf-clang-jolsa-v2)
btf_encoder: Match ftrace addresses within elf functions
f21eafdfc877 (for-1.20/btf_encoder-sanitized_int-clang-yhs-v2)
btf_encoder: sanitize non-regular int base type

Both patches fixes all issues seen so far with LLVM/Clang >=
12.0.0-rc1 and DWARF-v5 and BTF (debug-info) and pahole on
Linux/x86_64 and according to Nathan on Linux/arm64.
Yesterday, I tried with LLVM/Clang 13-git from <apt.llvm.org>.

BTW, Nick's DWARF-v5 patches are pending in <kbuild.git#kbuild> (see
[1]).

Personally, I can wait until [1] is in Linus Git.

Please, let me/us know what you are planning.
( I know it is Linux v5.12 merge-window. )

Sure, next week.

- Arnaldo

Regards,
- Sedat -

[1]
https://git.kernel.org/pub/scm/linux/kernel/git/masahiroy/linux-kbuild.git/log/?h=kbuild

-- 
Sent from my Android device with K-9 Mail. Please excuse my brevity.

On Wed, Feb 17, 2021 at 2:56 PM Arnaldo Carvalho de Melo
[off-list ref] wrote:

On February 17, 2021 10:40:43 AM GMT-03:00, Sedat Dilek [off-list ref] wrote:

quoted

On Wed, Feb 17, 2021 at 1:44 PM Arnaldo Carvalho de Melo
[off-list ref] wrote:

quoted

Em Sat, Feb 13, 2021 at 05:46:48PM +0100, Jiri Olsa escreveu:

quoted

Currently when processing DWARF function, we check its entrypoint
against ftrace addresses, assuming that the ftrace address matches
with function's entrypoint.

This is not the case on some architectures as reported by Nathan
when building kernel on arm [1].

Fixing the check to take into account the whole function not
just the entrypoint.

Most of the is_ftrace_func code was contributed by Andrii.

Applied locally, will go out after tests,

Hi Arnaldo,

Is it possible to have a pahole version 1.21 with this patch and the
one from Yonghong Son?

From my local pahole Git:

$ git log --oneline --no-merges v1.20..
2f83aefdbddf (for-1.20/btf_encoder-ftrace_elf-clang-jolsa-v2)
btf_encoder: Match ftrace addresses within elf functions
f21eafdfc877 (for-1.20/btf_encoder-sanitized_int-clang-yhs-v2)
btf_encoder: sanitize non-regular int base type

Both patches fixes all issues seen so far with LLVM/Clang >=
12.0.0-rc1 and DWARF-v5 and BTF (debug-info) and pahole on
Linux/x86_64 and according to Nathan on Linux/arm64.
Yesterday, I tried with LLVM/Clang 13-git from <apt.llvm.org>.

BTW, Nick's DWARF-v5 patches are pending in <kbuild.git#kbuild> (see
[1]).

Personally, I can wait until [1] is in Linus Git.

Please, let me/us know what you are planning.
( I know it is Linux v5.12 merge-window. )

Sure, next week.

That's OK with me.

- Sedat -

- Arnaldo

quoted

Regards,
- Sedat -

[1]
https://git.kernel.org/pub/scm/linux/kernel/git/masahiroy/linux-kbuild.git/log/?h=kbuild

--
Sent from my Android device with K-9 Mail. Please excuse my brevity.