[PATCH bpf-next v5] bpf: Support llvm-objcopy and llvm-objdump for vmlinux BTF

[PATCH bpf-next v5] bpf: Support llvm-objcopy and llvm-objdump for vmlinux BTF

[Fangrui Song]

Simplify gen_btf logic to make it work with llvm-objcopy and llvm-objdump.
The existing 'file format' and 'architecture' parsing logic is brittle
and does not work with llvm-objcopy/llvm-objdump.

.BTF in .tmp_vmlinux.btf is non-SHF_ALLOC. Add the SHF_ALLOC flag and
rename .BTF to BTF so that C code can reference the section via linker
synthesized __start_BTF and __stop_BTF. This fixes a small problem that
previous .BTF had the SHF_WRITE flag. Additionally, `objcopy -I binary`
synthesized symbols _binary__btf_vmlinux_bin_start and
_binary__btf_vmlinux_bin_start (not used elsewhere) are replaced with
more common __start_BTF and __stop_BTF.

Add 2>/dev/null because GNU objcopy (but not llvm-objcopy) warns
"empty loadable segment detected at vaddr=0xffffffff81000000, is this intentional?"

We use a dd command to change the e_type field in the ELF header from
ET_EXEC to ET_REL so that lld will accept .btf.vmlinux.bin.o.  Accepting
ET_EXEC as an input file is an extremely rare GNU ld feature that lld
does not intend to support, because this is error-prone.

Fixes: df786c9b9476 ("bpf: Force .BTF section start to zero when dumping from vmlinux")
Cc: Stanislav Fomichev <sdf@google.com>
Cc: Nick Desaulniers <ndesaulniers@google.com>
Reported-by: Nathan Chancellor <natechancellor@gmail.com>
Link: https://github.com/ClangBuiltLinux/linux/issues/871
Signed-off-by: Fangrui Song <maskray@google.com>
---
  kernel/bpf/btf.c        |  9 ++++-----
  kernel/bpf/sysfs_btf.c  | 11 +++++------
  scripts/link-vmlinux.sh | 17 ++++++-----------
  3 files changed, 15 insertions(+), 22 deletions(-)

diff --git a/kernel/bpf/btf.c b/kernel/bpf/btf.c
index 50080add2ab9..6f397c4da05e 100644
--- a/kernel/bpf/btf.c
+++ b/kernel/bpf/btf.c
@@ -3477,8 +3477,8 @@ static struct btf *btf_parse(void __user *btf_data, u32 btf_data_size,
  	return ERR_PTR(err);
  }
  
-extern char __weak _binary__btf_vmlinux_bin_start[];
-extern char __weak _binary__btf_vmlinux_bin_end[];
+extern char __weak __start_BTF[];
+extern char __weak __stop_BTF[];
  extern struct btf *btf_vmlinux;
  
  #define BPF_MAP_TYPE(_id, _ops)
@@ -3605,9 +3605,8 @@ struct btf *btf_parse_vmlinux(void)
  	}
  	env->btf = btf;
  
-	btf->data = _binary__btf_vmlinux_bin_start;
-	btf->data_size = _binary__btf_vmlinux_bin_end -
-		_binary__btf_vmlinux_bin_start;
+	btf->data = __start_BTF;
+	btf->data_size = __stop_BTF - __start_BTF;
  
  	err = btf_parse_hdr(env);
  	if (err)
diff --git a/kernel/bpf/sysfs_btf.c b/kernel/bpf/sysfs_btf.c
index 7ae5dddd1fe6..3b495773de5a 100644
--- a/kernel/bpf/sysfs_btf.c
+++ b/kernel/bpf/sysfs_btf.c
@@ -9,15 +9,15 @@
  #include <linux/sysfs.h>
  
  /* See scripts/link-vmlinux.sh, gen_btf() func for details */
-extern char __weak _binary__btf_vmlinux_bin_start[];
-extern char __weak _binary__btf_vmlinux_bin_end[];
+extern char __weak __start_BTF[];
+extern char __weak __stop_BTF[];
  
  static ssize_t
  btf_vmlinux_read(struct file *file, struct kobject *kobj,
  		 struct bin_attribute *bin_attr,
  		 char *buf, loff_t off, size_t len)
  {
-	memcpy(buf, _binary__btf_vmlinux_bin_start + off, len);
+	memcpy(buf, __start_BTF + off, len);
  	return len;
  }
  
@@ -30,15 +30,14 @@ static struct kobject *btf_kobj;
  
  static int __init btf_vmlinux_init(void)
  {
-	if (!_binary__btf_vmlinux_bin_start)
+	if (!__start_BTF)
  		return 0;
  
  	btf_kobj = kobject_create_and_add("btf", kernel_kobj);
  	if (!btf_kobj)
  		return -ENOMEM;
  
-	bin_attr_btf_vmlinux.size = _binary__btf_vmlinux_bin_end -
-				    _binary__btf_vmlinux_bin_start;
+	bin_attr_btf_vmlinux.size = __stop_BTF - __start_BTF;
  
  	return sysfs_create_bin_file(btf_kobj, &bin_attr_btf_vmlinux);
  }
diff --git a/scripts/link-vmlinux.sh b/scripts/link-vmlinux.sh
index ac569e197bfa..ae2048625f1e 100755
--- a/scripts/link-vmlinux.sh
+++ b/scripts/link-vmlinux.sh
@@ -133,17 +133,12 @@ gen_btf()
  	info "BTF" ${2}
  	LLVM_OBJCOPY=${OBJCOPY} ${PAHOLE} -J ${1}
  
-	# dump .BTF section into raw binary file to link with final vmlinux
-	bin_arch=$(LANG=C ${OBJDUMP} -f ${1} | grep architecture | \
-		cut -d, -f1 | cut -d' ' -f2)
-	bin_format=$(LANG=C ${OBJDUMP} -f ${1} | grep 'file format' | \
-		awk '{print $4}')
-	bin_file=.btf.vmlinux.bin
-	${OBJCOPY} --change-section-address .BTF=0 \
-		--set-section-flags .BTF=alloc -O binary \
-		--only-section=.BTF ${1} $bin_file
-	${OBJCOPY} -I binary -O ${bin_format} -B ${bin_arch} \
-		--rename-section .data=.BTF $bin_file ${2}
+	# Extract .BTF, add SHF_ALLOC, rename to BTF so that we can reference
+	# it via linker synthesized __start_BTF and __stop_BTF. Change e_type
+	# to ET_REL so that it can be used to link final vmlinux.
+	${OBJCOPY} --only-section=.BTF --set-section-flags .BTF=alloc,readonly \
+		--rename-section .BTF=BTF ${1} ${2} 2>/dev/null && \
+		printf '\1' | dd of=${2} conv=notrunc bs=1 seek=16 status=none
  }
  
  # Create ${2} .o file with all symbols from the ${1} object file
-- 
2.25.1.481.gfbce0eb801-goog

Re: [PATCH bpf-next v5] bpf: Support llvm-objcopy and llvm-objdump for vmlinux BTF

[Stanislav Fomichev]

On 03/17, Fangrui Song wrote:
> Simplify gen_btf logic to make it work with llvm-objcopy and llvm-objdump.
> The existing 'file format' and 'architecture' parsing logic is brittle
> and does not work with llvm-objcopy/llvm-objdump.
Thanks, it all makes sense and looks much better/portable (too much
dependence on binutils :-).
I've left a bunch of questions/nits below.

> 
> .BTF in .tmp_vmlinux.btf is non-SHF_ALLOC. Add the SHF_ALLOC flag and
> rename .BTF to BTF so that C code can reference the section via linker
> synthesized __start_BTF and __stop_BTF. This fixes a small problem that
> previous .BTF had the SHF_WRITE flag. Additionally, `objcopy -I binary`
> synthesized symbols _binary__btf_vmlinux_bin_start and
> _binary__btf_vmlinux_bin_start (not used elsewhere) are replaced with
> more common __start_BTF and __stop_BTF.
> 
> Add 2>/dev/null because GNU objcopy (but not llvm-objcopy) warns
> "empty loadable segment detected at vaddr=0xffffffff81000000, is this intentional?"
> 
> We use a dd command to change the e_type field in the ELF header from
> ET_EXEC to ET_REL so that lld will accept .btf.vmlinux.bin.o.  Accepting
> ET_EXEC as an input file is an extremely rare GNU ld feature that lld
> does not intend to support, because this is error-prone.
Please keep small changelog here, for example:

v5:
* rebased on top of bpfnext

Btw, I tried to pull and test it and failed:
$ curl -LO https://lore.kernel.org/bpf/20200317211649.o4fzaxrzy6qxvz4f@google.com/raw
  % Total    % Received % Xferd  Average Speed   Time    Time     Time  Current
                                 Dload  Upload   Total   Spent    Left  Speed
100  9627  100  9627    0     0  47191      0 --:--:-- --:--:-- --:--:-- 47191
$ git am raw
warning: Patch sent with format=flowed; space at the end of lines might
be lost.
Applying: bpf: Support llvm-objcopy and llvm-objdump for vmlinux BTF
error: corrupt patch at line 20
Patch failed at 0001 bpf: Support llvm-objcopy and llvm-objdump for
vmlinux BTF
hint: Use 'git am --show-current-patch' to see the failed patch
When you have resolved this problem, run "git am --continue".
If you prefer to skip this patch, run "git am --skip" instead.
To restore the original branch and stop patching, run "git am --abort".
$ git describe
v5.6-rc5-1621-g230021539e8c

Are you sure it's on top of the bpf-next? Or am I doing something wrong?

> Fixes: df786c9b9476 ("bpf: Force .BTF section start to zero when dumping from vmlinux")
> Cc: Stanislav Fomichev <sdf@google.com>
> Cc: Nick Desaulniers <ndesaulniers@google.com>
> Reported-by: Nathan Chancellor <natechancellor@gmail.com>
> Link: https://github.com/ClangBuiltLinux/linux/issues/871
> Signed-off-by: Fangrui Song <maskray@google.com>
> ---
>  kernel/bpf/btf.c        |  9 ++++-----
>  kernel/bpf/sysfs_btf.c  | 11 +++++------
>  scripts/link-vmlinux.sh | 17 ++++++-----------
>  3 files changed, 15 insertions(+), 22 deletions(-)
> 
> diff --git a/kernel/bpf/btf.c b/kernel/bpf/btf.c
> index 50080add2ab9..6f397c4da05e 100644
> --- a/kernel/bpf/btf.c
> +++ b/kernel/bpf/btf.c
> @@ -3477,8 +3477,8 @@ static struct btf *btf_parse(void __user *btf_data, u32 btf_data_size,
>  	return ERR_PTR(err);
>  }
> -extern char __weak _binary__btf_vmlinux_bin_start[];
> -extern char __weak _binary__btf_vmlinux_bin_end[];
> +extern char __weak __start_BTF[];
> +extern char __weak __stop_BTF[];
>  extern struct btf *btf_vmlinux;
>  #define BPF_MAP_TYPE(_id, _ops)
> @@ -3605,9 +3605,8 @@ struct btf *btf_parse_vmlinux(void)
>  	}
>  	env->btf = btf;
> -	btf->data = _binary__btf_vmlinux_bin_start;
> -	btf->data_size = _binary__btf_vmlinux_bin_end -
> -		_binary__btf_vmlinux_bin_start;
> +	btf->data = __start_BTF;
> +	btf->data_size = __stop_BTF - __start_BTF;
>  	err = btf_parse_hdr(env);
>  	if (err)
> diff --git a/kernel/bpf/sysfs_btf.c b/kernel/bpf/sysfs_btf.c
> index 7ae5dddd1fe6..3b495773de5a 100644
> --- a/kernel/bpf/sysfs_btf.c
> +++ b/kernel/bpf/sysfs_btf.c
> @@ -9,15 +9,15 @@
>  #include <linux/sysfs.h>
>  /* See scripts/link-vmlinux.sh, gen_btf() func for details */
> -extern char __weak _binary__btf_vmlinux_bin_start[];
> -extern char __weak _binary__btf_vmlinux_bin_end[];
> +extern char __weak __start_BTF[];
> +extern char __weak __stop_BTF[];
>  static ssize_t
>  btf_vmlinux_read(struct file *file, struct kobject *kobj,
>  		 struct bin_attribute *bin_attr,
>  		 char *buf, loff_t off, size_t len)
>  {
> -	memcpy(buf, _binary__btf_vmlinux_bin_start + off, len);
> +	memcpy(buf, __start_BTF + off, len);
>  	return len;
>  }
> @@ -30,15 +30,14 @@ static struct kobject *btf_kobj;
>  static int __init btf_vmlinux_init(void)
>  {
> -	if (!_binary__btf_vmlinux_bin_start)
> +	if (!__start_BTF)
>  		return 0;
>  	btf_kobj = kobject_create_and_add("btf", kernel_kobj);
>  	if (!btf_kobj)
>  		return -ENOMEM;
> -	bin_attr_btf_vmlinux.size = _binary__btf_vmlinux_bin_end -
> -				    _binary__btf_vmlinux_bin_start;
> +	bin_attr_btf_vmlinux.size = __stop_BTF - __start_BTF;
>  	return sysfs_create_bin_file(btf_kobj, &bin_attr_btf_vmlinux);
>  }
> diff --git a/scripts/link-vmlinux.sh b/scripts/link-vmlinux.sh
> index ac569e197bfa..ae2048625f1e 100755
> --- a/scripts/link-vmlinux.sh
> +++ b/scripts/link-vmlinux.sh
> @@ -133,17 +133,12 @@ gen_btf()
>  	info "BTF" ${2}
>  	LLVM_OBJCOPY=${OBJCOPY} ${PAHOLE} -J ${1}
> -	# dump .BTF section into raw binary file to link with final vmlinux
> -	bin_arch=$(LANG=C ${OBJDUMP} -f ${1} | grep architecture | \
> -		cut -d, -f1 | cut -d' ' -f2)
> -	bin_format=$(LANG=C ${OBJDUMP} -f ${1} | grep 'file format' | \
> -		awk '{print $4}')
> -	bin_file=.btf.vmlinux.bin
We still have the following in gen_btf that you need to remove:
	local bin_arch
	local bin_format
	local bin_file

> -	${OBJCOPY} --change-section-address .BTF=0 \
> -		--set-section-flags .BTF=alloc -O binary \
> -		--only-section=.BTF ${1} $bin_file
> -	${OBJCOPY} -I binary -O ${bin_format} -B ${bin_arch} \
> -		--rename-section .data=.BTF $bin_file ${2}
> +	# Extract .BTF, add SHF_ALLOC, rename to BTF so that we can reference
> +	# it via linker synthesized __start_BTF and __stop_BTF. Change e_type
> +	# to ET_REL so that it can be used to link final vmlinux.
> +	${OBJCOPY} --only-section=.BTF --set-section-flags .BTF=alloc,readonly \
> +		--rename-section .BTF=BTF ${1} ${2} 2>/dev/null && \
> +		printf '\1' | dd of=${2} conv=notrunc bs=1 seek=16 status=none
	We have 'set -e' so && can be omitted. Maybe it will be a bit
	more clear:

	# Extract .BTF, add SHF_ALLOC, rename to BTF so that we can reference
	# it via linker synthesized __start_BTF and __stop_BTF.
	${OBJCOPY} --only-section=.BTF --set-section-flags .BTF=alloc,readonly \
		--rename-section .BTF=BTF ${1} ${2} 2>/dev/null

	# Change e_type to ET_REL (0x01) so that it can be used to link final
	# vmlinux.
	printf '\1' | dd of=${2} conv=notrunc bs=1 seek=16 status=none

	^^^ maybe also point out that this is required for llvm's ldd.
	I know you point it out in the commit message, but still nice
	to have as a comment here.

>  }
>  # Create ${2} .o file with all symbols from the ${1} object file
> -- 
> 2.25.1.481.gfbce0eb801-goog

Re: [PATCH bpf-next v5] bpf: Support llvm-objcopy and llvm-objdump for vmlinux BTF

[Fangrui Song]

On 2020-03-17, Stanislav Fomichev wrote:
>On 03/17, Fangrui Song wrote:
>> Simplify gen_btf logic to make it work with llvm-objcopy and llvm-objdump.
>> The existing 'file format' and 'architecture' parsing logic is brittle
>> and does not work with llvm-objcopy/llvm-objdump.
>Thanks, it all makes sense and looks much better/portable (too much
>dependence on binutils :-).
>I've left a bunch of questions/nits below.
>
>>
>> .BTF in .tmp_vmlinux.btf is non-SHF_ALLOC. Add the SHF_ALLOC flag and
>> rename .BTF to BTF so that C code can reference the section via linker
>> synthesized __start_BTF and __stop_BTF. This fixes a small problem that
>> previous .BTF had the SHF_WRITE flag. Additionally, `objcopy -I binary`
>> synthesized symbols _binary__btf_vmlinux_bin_start and
>> _binary__btf_vmlinux_bin_start (not used elsewhere) are replaced with
>> more common __start_BTF and __stop_BTF.
>>
>> Add 2>/dev/null because GNU objcopy (but not llvm-objcopy) warns
>> "empty loadable segment detected at vaddr=0xffffffff81000000, is this intentional?"
>>
>> We use a dd command to change the e_type field in the ELF header from
>> ET_EXEC to ET_REL so that lld will accept .btf.vmlinux.bin.o.  Accepting
>> ET_EXEC as an input file is an extremely rare GNU ld feature that lld
>> does not intend to support, because this is error-prone.
>Please keep small changelog here, for example:
>
>v5:
>* rebased on top of bpfnext

Thanks for the tip. Add them at the bottom?

>Btw, I tried to pull and test it and failed:
>$ curl -LO https://lore.kernel.org/bpf/20200317211649.o4fzaxrzy6qxvz4f@google.com/raw
>  % Total    % Received % Xferd  Average Speed   Time    Time     Time  Current
>                                 Dload  Upload   Total   Spent    Left  Speed
>100  9627  100  9627    0     0  47191      0 --:--:-- --:--:-- --:--:-- 47191
>$ git am raw
>warning: Patch sent with format=flowed; space at the end of lines might
>be lost.
>Applying: bpf: Support llvm-objcopy and llvm-objdump for vmlinux BTF
>error: corrupt patch at line 20
>Patch failed at 0001 bpf: Support llvm-objcopy and llvm-objdump for
>vmlinux BTF
>hint: Use 'git am --show-current-patch' to see the failed patch
>When you have resolved this problem, run "git am --continue".
>If you prefer to skip this patch, run "git am --skip" instead.
>To restore the original branch and stop patching, run "git am --abort".
>$ git describe
>v5.6-rc5-1621-g230021539e8c
>
>Are you sure it's on top of the bpf-next? Or am I doing something wrong?

It can be my fault when messing with various patches..
Will be more cautious.

>> Fixes: df786c9b9476 ("bpf: Force .BTF section start to zero when dumping from vmlinux")
>> Cc: Stanislav Fomichev <sdf@google.com>
>> Cc: Nick Desaulniers <ndesaulniers@google.com>
>> Reported-by: Nathan Chancellor <natechancellor@gmail.com>
>> Link: https://github.com/ClangBuiltLinux/linux/issues/871
>> Signed-off-by: Fangrui Song <maskray@google.com>
>> ---
>>  kernel/bpf/btf.c        |  9 ++++-----
>>  kernel/bpf/sysfs_btf.c  | 11 +++++------
>>  scripts/link-vmlinux.sh | 17 ++++++-----------
>>  3 files changed, 15 insertions(+), 22 deletions(-)
>>
>> diff --git a/kernel/bpf/btf.c b/kernel/bpf/btf.c
>> index 50080add2ab9..6f397c4da05e 100644
>> --- a/kernel/bpf/btf.c
>> +++ b/kernel/bpf/btf.c
>> @@ -3477,8 +3477,8 @@ static struct btf *btf_parse(void __user *btf_data, u32 btf_data_size,
>>  	return ERR_PTR(err);
>>  }
>> -extern char __weak _binary__btf_vmlinux_bin_start[];
>> -extern char __weak _binary__btf_vmlinux_bin_end[];
>> +extern char __weak __start_BTF[];
>> +extern char __weak __stop_BTF[];
>>  extern struct btf *btf_vmlinux;
>>  #define BPF_MAP_TYPE(_id, _ops)
>> @@ -3605,9 +3605,8 @@ struct btf *btf_parse_vmlinux(void)
>>  	}
>>  	env->btf = btf;
>> -	btf->data = _binary__btf_vmlinux_bin_start;
>> -	btf->data_size = _binary__btf_vmlinux_bin_end -
>> -		_binary__btf_vmlinux_bin_start;
>> +	btf->data = __start_BTF;
>> +	btf->data_size = __stop_BTF - __start_BTF;
>>  	err = btf_parse_hdr(env);
>>  	if (err)
>> diff --git a/kernel/bpf/sysfs_btf.c b/kernel/bpf/sysfs_btf.c
>> index 7ae5dddd1fe6..3b495773de5a 100644
>> --- a/kernel/bpf/sysfs_btf.c
>> +++ b/kernel/bpf/sysfs_btf.c
>> @@ -9,15 +9,15 @@
>>  #include <linux/sysfs.h>
>>  /* See scripts/link-vmlinux.sh, gen_btf() func for details */
>> -extern char __weak _binary__btf_vmlinux_bin_start[];
>> -extern char __weak _binary__btf_vmlinux_bin_end[];
>> +extern char __weak __start_BTF[];
>> +extern char __weak __stop_BTF[];
>>  static ssize_t
>>  btf_vmlinux_read(struct file *file, struct kobject *kobj,
>>  		 struct bin_attribute *bin_attr,
>>  		 char *buf, loff_t off, size_t len)
>>  {
>> -	memcpy(buf, _binary__btf_vmlinux_bin_start + off, len);
>> +	memcpy(buf, __start_BTF + off, len);
>>  	return len;
>>  }
>> @@ -30,15 +30,14 @@ static struct kobject *btf_kobj;
>>  static int __init btf_vmlinux_init(void)
>>  {
>> -	if (!_binary__btf_vmlinux_bin_start)
>> +	if (!__start_BTF)
>>  		return 0;
>>  	btf_kobj = kobject_create_and_add("btf", kernel_kobj);
>>  	if (!btf_kobj)
>>  		return -ENOMEM;
>> -	bin_attr_btf_vmlinux.size = _binary__btf_vmlinux_bin_end -
>> -				    _binary__btf_vmlinux_bin_start;
>> +	bin_attr_btf_vmlinux.size = __stop_BTF - __start_BTF;
>>  	return sysfs_create_bin_file(btf_kobj, &bin_attr_btf_vmlinux);
>>  }
>> diff --git a/scripts/link-vmlinux.sh b/scripts/link-vmlinux.sh
>> index ac569e197bfa..ae2048625f1e 100755
>> --- a/scripts/link-vmlinux.sh
>> +++ b/scripts/link-vmlinux.sh
>> @@ -133,17 +133,12 @@ gen_btf()
>>  	info "BTF" ${2}
>>  	LLVM_OBJCOPY=${OBJCOPY} ${PAHOLE} -J ${1}
>> -	# dump .BTF section into raw binary file to link with final vmlinux
>> -	bin_arch=$(LANG=C ${OBJDUMP} -f ${1} | grep architecture | \
>> -		cut -d, -f1 | cut -d' ' -f2)
>> -	bin_format=$(LANG=C ${OBJDUMP} -f ${1} | grep 'file format' | \
>> -		awk '{print $4}')
>> -	bin_file=.btf.vmlinux.bin
>We still have the following in gen_btf that you need to remove:
>	local bin_arch
>	local bin_format
>	local bin_file

Thanks. Will delete them.

>> -	${OBJCOPY} --change-section-address .BTF=0 \
>> -		--set-section-flags .BTF=alloc -O binary \
>> -		--only-section=.BTF ${1} $bin_file
>> -	${OBJCOPY} -I binary -O ${bin_format} -B ${bin_arch} \
>> -		--rename-section .data=.BTF $bin_file ${2}
>> +	# Extract .BTF, add SHF_ALLOC, rename to BTF so that we can reference
>> +	# it via linker synthesized __start_BTF and __stop_BTF. Change e_type
>> +	# to ET_REL so that it can be used to link final vmlinux.
>> +	${OBJCOPY} --only-section=.BTF --set-section-flags .BTF=alloc,readonly \
>> +		--rename-section .BTF=BTF ${1} ${2} 2>/dev/null && \
>> +		printf '\1' | dd of=${2} conv=notrunc bs=1 seek=16 status=none
>	We have 'set -e' so && can be omitted. Maybe it will be a bit
>	more clear:

Agree.

>	# Extract .BTF, add SHF_ALLOC, rename to BTF so that we can reference
>	# it via linker synthesized __start_BTF and __stop_BTF.
>	${OBJCOPY} --only-section=.BTF --set-section-flags .BTF=alloc,readonly \
>		--rename-section .BTF=BTF ${1} ${2} 2>/dev/null
>
>	# Change e_type to ET_REL (0x01) so that it can be used to link final
>	# vmlinux.
>	printf '\1' | dd of=${2} conv=notrunc bs=1 seek=16 status=none
>
>	^^^ maybe also point out that this is required for llvm's ldd.
>	I know you point it out in the commit message, but still nice
>	to have as a comment here.

Will mention lld in v6.

>>  }
>>  # Create ${2} .o file with all symbols from the ${1} object file
>> --
>> 2.25.1.481.gfbce0eb801-goog

Re: [PATCH bpf-next v5] bpf: Support llvm-objcopy and llvm-objdump for vmlinux BTF

[Nick Desaulniers]

On Tue, Mar 17, 2020 at 3:01 PM Fangrui Song <maskray@google.com> wrote:
>
> On 2020-03-17, Stanislav Fomichev wrote:
> >Please keep small changelog here, for example:
> >
> >v5:
> >* rebased on top of bpfnext
>
> Thanks for the tip. Add them at the bottom?

"Below the fold" see this patch I just sent out:
https://lore.kernel.org/lkml/20200317215515.226917-1-ndesaulniers@google.com/T/#u
grep "Changes"

$ git format-patch -v2 HEAD~
$ vim 0001-...patch
<manually add changelog "below the fold">
-- 
Thanks,
~Nick Desaulniers

Re: [PATCH bpf-next v5] bpf: Support llvm-objcopy and llvm-objdump for vmlinux BTF

[Stanislav Fomichev]

On Tue, Mar 17, 2020 at 3:08 PM Nick Desaulniers
<ndesaulniers@google.com> wrote:
>
> On Tue, Mar 17, 2020 at 3:01 PM Fangrui Song <maskray@google.com> wrote:
> >
> > On 2020-03-17, Stanislav Fomichev wrote:
> > >Please keep small changelog here, for example:
> > >
> > >v5:
> > >* rebased on top of bpfnext
> >
> > Thanks for the tip. Add them at the bottom?
>
> "Below the fold" see this patch I just sent out:
> https://lore.kernel.org/lkml/20200317215515.226917-1-ndesaulniers@google.com/T/#u
> grep "Changes"
>
> $ git format-patch -v2 HEAD~
> $ vim 0001-...patch
> <manually add changelog "below the fold">
BPF subtree prefers the changelog in the commit body, not the comments
(iow, before ---).
Add them at the end of you message, see, for example:
https://lore.kernel.org/bpf/a428fb88-9b53-27dd-a195-497755944921@iogearbox.net/T/

> --
> Thanks,
> ~Nick Desaulniers

Re: [PATCH bpf-next v5] bpf: Support llvm-objcopy and llvm-objdump for vmlinux BTF

[Nick Desaulniers]

On Tue, Mar 17, 2020 at 3:13 PM Stanislav Fomichev <sdf@google.com> wrote:
>
> On Tue, Mar 17, 2020 at 3:08 PM Nick Desaulniers
> <ndesaulniers@google.com> wrote:
> >
> > On Tue, Mar 17, 2020 at 3:01 PM Fangrui Song <maskray@google.com> wrote:
> > >
> > > On 2020-03-17, Stanislav Fomichev wrote:
> > > >Please keep small changelog here, for example:
> > > >
> > > >v5:
> > > >* rebased on top of bpfnext
> > >
> > > Thanks for the tip. Add them at the bottom?
> >
> > "Below the fold" see this patch I just sent out:
> > https://lore.kernel.org/lkml/20200317215515.226917-1-ndesaulniers@google.com/T/#u
> > grep "Changes"
> >
> > $ git format-patch -v2 HEAD~
> > $ vim 0001-...patch
> > <manually add changelog "below the fold">
> BPF subtree prefers the changelog in the commit body, not the comments
> (iow, before ---).
> Add them at the end of you message, see, for example:
> https://lore.kernel.org/bpf/a428fb88-9b53-27dd-a195-497755944921@iogearbox.net/T/

Sigh, every maintainer is a special snowflake.  In our tree, you're
only allowed to commit on Thursdays under a blood moon. /s

But thanks for the note.
-- 
Thanks,
~Nick Desaulniers

Re: [PATCH bpf-next v5] bpf: Support llvm-objcopy and llvm-objdump for vmlinux BTF

[Daniel Borkmann]

On 3/18/20 12:14 AM, Nick Desaulniers wrote:
> On Tue, Mar 17, 2020 at 3:13 PM Stanislav Fomichev <sdf@google.com> wrote:
>> On Tue, Mar 17, 2020 at 3:08 PM Nick Desaulniers
>> <ndesaulniers@google.com> wrote:
>>> On Tue, Mar 17, 2020 at 3:01 PM Fangrui Song <maskray@google.com> wrote:
>>>> On 2020-03-17, Stanislav Fomichev wrote:
>>>>> Please keep small changelog here, for example:
>>>>>
>>>>> v5:
>>>>> * rebased on top of bpfnext
>>>>
>>>> Thanks for the tip. Add them at the bottom?
>>>
>>> "Below the fold" see this patch I just sent out:
>>> https://lore.kernel.org/lkml/20200317215515.226917-1-ndesaulniers@google.com/T/#u
>>> grep "Changes"
>>>
>>> $ git format-patch -v2 HEAD~
>>> $ vim 0001-...patch
>>> <manually add changelog "below the fold">
>> BPF subtree prefers the changelog in the commit body, not the comments
>> (iow, before ---).
>> Add them at the end of you message, see, for example:
>> https://lore.kernel.org/bpf/a428fb88-9b53-27dd-a195-497755944921@iogearbox.net/T/
> 
> Sigh, every maintainer is a special snowflake.  In our tree, you're
> only allowed to commit on Thursdays under a blood moon. /s
> 
> But thanks for the note.

For every commit into bpf/bpf-next we automatically put a 'Link:' tag pointing
to the lore.kernel.org/bpf long-term archive, so there is always an easy way to
retrieve additional information or discussions ([0] as one example). So whether
you put it above or below the '---' line I personally don't mind either way. All
crucial information that helps a reader later on to understand /why/ we ended up
exactly with the current change should be part of the main commit log though.

  [0] https://git.kernel.org/pub/scm/linux/kernel/git/bpf/bpf-next.git/commit/?id=483d7a30f538e2f8addd32aa9a3d2e94ae55fa65

Re: [PATCH bpf-next v5] bpf: Support llvm-objcopy and llvm-objdump for vmlinux BTF

[Kees Cook]

On Tue, Mar 17, 2020 at 02:16:49PM -0700, Fangrui Song wrote:
> Simplify gen_btf logic to make it work with llvm-objcopy and llvm-objdump.
> The existing 'file format' and 'architecture' parsing logic is brittle
> and does not work with llvm-objcopy/llvm-objdump.
> 
> .BTF in .tmp_vmlinux.btf is non-SHF_ALLOC. Add the SHF_ALLOC flag and
> rename .BTF to BTF so that C code can reference the section via linker
> synthesized __start_BTF and __stop_BTF. This fixes a small problem that
> previous .BTF had the SHF_WRITE flag. Additionally, `objcopy -I binary`
> synthesized symbols _binary__btf_vmlinux_bin_start and
> _binary__btf_vmlinux_bin_start (not used elsewhere) are replaced with
> more common __start_BTF and __stop_BTF.

I'm glad to see the name change benefit here. Just reducing the number
of execs in this path is only worth this change, IMO. Going from 2
objdump and 2 objcopy calls to a single objcopy is very nice. :)

> 
> Add 2>/dev/null because GNU objcopy (but not llvm-objcopy) warns
> "empty loadable segment detected at vaddr=0xffffffff81000000, is this intentional?"
> 
> We use a dd command to change the e_type field in the ELF header from
> ET_EXEC to ET_REL so that lld will accept .btf.vmlinux.bin.o.  Accepting
> ET_EXEC as an input file is an extremely rare GNU ld feature that lld
> does not intend to support, because this is error-prone.

Nice ELF trick. :)

> 
> Fixes: df786c9b9476 ("bpf: Force .BTF section start to zero when dumping from vmlinux")
> Cc: Stanislav Fomichev <sdf@google.com>
> Cc: Nick Desaulniers <ndesaulniers@google.com>
> Reported-by: Nathan Chancellor <natechancellor@gmail.com>
> Link: https://github.com/ClangBuiltLinux/linux/issues/871
> Signed-off-by: Fangrui Song <maskray@google.com>

Reviewed-by: Kees Cook <keescook@chromium.org>

-Kees

> ---
>  kernel/bpf/btf.c        |  9 ++++-----
>  kernel/bpf/sysfs_btf.c  | 11 +++++------
>  scripts/link-vmlinux.sh | 17 ++++++-----------
>  3 files changed, 15 insertions(+), 22 deletions(-)
> 
> diff --git a/kernel/bpf/btf.c b/kernel/bpf/btf.c
> index 50080add2ab9..6f397c4da05e 100644
> --- a/kernel/bpf/btf.c
> +++ b/kernel/bpf/btf.c
> @@ -3477,8 +3477,8 @@ static struct btf *btf_parse(void __user *btf_data, u32 btf_data_size,
>  	return ERR_PTR(err);
>  }
> -extern char __weak _binary__btf_vmlinux_bin_start[];
> -extern char __weak _binary__btf_vmlinux_bin_end[];
> +extern char __weak __start_BTF[];
> +extern char __weak __stop_BTF[];
>  extern struct btf *btf_vmlinux;
>  #define BPF_MAP_TYPE(_id, _ops)
> @@ -3605,9 +3605,8 @@ struct btf *btf_parse_vmlinux(void)
>  	}
>  	env->btf = btf;
> -	btf->data = _binary__btf_vmlinux_bin_start;
> -	btf->data_size = _binary__btf_vmlinux_bin_end -
> -		_binary__btf_vmlinux_bin_start;
> +	btf->data = __start_BTF;
> +	btf->data_size = __stop_BTF - __start_BTF;
>  	err = btf_parse_hdr(env);
>  	if (err)
> diff --git a/kernel/bpf/sysfs_btf.c b/kernel/bpf/sysfs_btf.c
> index 7ae5dddd1fe6..3b495773de5a 100644
> --- a/kernel/bpf/sysfs_btf.c
> +++ b/kernel/bpf/sysfs_btf.c
> @@ -9,15 +9,15 @@
>  #include <linux/sysfs.h>
>  /* See scripts/link-vmlinux.sh, gen_btf() func for details */
> -extern char __weak _binary__btf_vmlinux_bin_start[];
> -extern char __weak _binary__btf_vmlinux_bin_end[];
> +extern char __weak __start_BTF[];
> +extern char __weak __stop_BTF[];
>  static ssize_t
>  btf_vmlinux_read(struct file *file, struct kobject *kobj,
>  		 struct bin_attribute *bin_attr,
>  		 char *buf, loff_t off, size_t len)
>  {
> -	memcpy(buf, _binary__btf_vmlinux_bin_start + off, len);
> +	memcpy(buf, __start_BTF + off, len);
>  	return len;
>  }
> @@ -30,15 +30,14 @@ static struct kobject *btf_kobj;
>  static int __init btf_vmlinux_init(void)
>  {
> -	if (!_binary__btf_vmlinux_bin_start)
> +	if (!__start_BTF)
>  		return 0;
>  	btf_kobj = kobject_create_and_add("btf", kernel_kobj);
>  	if (!btf_kobj)
>  		return -ENOMEM;
> -	bin_attr_btf_vmlinux.size = _binary__btf_vmlinux_bin_end -
> -				    _binary__btf_vmlinux_bin_start;
> +	bin_attr_btf_vmlinux.size = __stop_BTF - __start_BTF;
>  	return sysfs_create_bin_file(btf_kobj, &bin_attr_btf_vmlinux);
>  }
> diff --git a/scripts/link-vmlinux.sh b/scripts/link-vmlinux.sh
> index ac569e197bfa..ae2048625f1e 100755
> --- a/scripts/link-vmlinux.sh
> +++ b/scripts/link-vmlinux.sh
> @@ -133,17 +133,12 @@ gen_btf()
>  	info "BTF" ${2}
>  	LLVM_OBJCOPY=${OBJCOPY} ${PAHOLE} -J ${1}
> -	# dump .BTF section into raw binary file to link with final vmlinux
> -	bin_arch=$(LANG=C ${OBJDUMP} -f ${1} | grep architecture | \
> -		cut -d, -f1 | cut -d' ' -f2)
> -	bin_format=$(LANG=C ${OBJDUMP} -f ${1} | grep 'file format' | \
> -		awk '{print $4}')
> -	bin_file=.btf.vmlinux.bin
> -	${OBJCOPY} --change-section-address .BTF=0 \
> -		--set-section-flags .BTF=alloc -O binary \
> -		--only-section=.BTF ${1} $bin_file
> -	${OBJCOPY} -I binary -O ${bin_format} -B ${bin_arch} \
> -		--rename-section .data=.BTF $bin_file ${2}
> +	# Extract .BTF, add SHF_ALLOC, rename to BTF so that we can reference
> +	# it via linker synthesized __start_BTF and __stop_BTF. Change e_type
> +	# to ET_REL so that it can be used to link final vmlinux.
> +	${OBJCOPY} --only-section=.BTF --set-section-flags .BTF=alloc,readonly \
> +		--rename-section .BTF=BTF ${1} ${2} 2>/dev/null && \
> +		printf '\1' | dd of=${2} conv=notrunc bs=1 seek=16 status=none
>  }
>  # Create ${2} .o file with all symbols from the ${1} object file
> -- 
> 2.25.1.481.gfbce0eb801-goog

-- 
Kees Cook