Re: [PATCH v2] pahole: Add --kabi_prefix flag

[Shuyi Cheng <chengshuyi@linux.alibaba.com>]

On 5/20/21 7:49 PM, Jiri Olsa wrote:
> On Thu, May 20, 2021 at 06:27:48PM +0800, Shuyi Cheng wrote:
>>
>>
>> On 5/20/21 4:07 AM, Jiri Olsa wrote:
>>> On Wed, May 19, 2021 at 10:44:44AM +0800, Shuyi Cheng wrote:
>>>> To solve problems similar to _RH_KABI_REPLACE. The _RH_KABI_REPLACE(_orig,
>>>> _new) macros perserve size alignment and kabi agreement between _orig and
>>>> _new.Below is the definition of this macro:
>>>>
>>>> # define _RH_KABI_REPLACE(_orig, _new)            \
>>>>       union {                        \
>>>>           _new;                    \
>>>>           struct {                \
>>>>               _orig;                \
>>>>           } __UNIQUE_ID(rh_kabi_hide);        \
>>>>           __RH_KABI_CHECK_SIZE_ALIGN(_orig, _new);    \
>>>>       }
>>>
>>> hi,
>>> that macro sounds familiar ;-) I think this should be already
>>> solved directly in the header file by this one:
>>>
>>>     https://gitlab.com/cki-project/kernel-ark/-/commit/331be9c5a436057ee852075c102d9d90a9046a30
>>>
>>> jirka
>>>
>>
>> Well, this patch solves this problem very well from the kernel level. But
>> there are many mirrors [here](http://debuginfo.centos.org/7/x86_64/) that
>> still have this problem. And these mirrors support ebpf, so it is very
>> important to effectively extract the btf segment from these mirrors. :-)
> 
> hum, the link shows 3.10.* centos kernel, right?
> AFAIK there's no BTF support in those kernels..
> 
> but maybe I'm missing some backporting channel,
> could you please point me to related sources?
> 
> jirka
> 

Yes, it is the centos kernel of version 3.10. This version of the centos 
kernel supports ebpf. As far as I think, BTF features don't depend on 
the kernel much. Only when using the STRUCT OPS feature does the kernel 
support BTF, see 
[here](https://github.com/libbpf/libbpf/blob/57375504c6c9766d23f178c40f71bf5e8df9363d/src/libbpf.c#L2549) 
and [here](https://www.spinics.net/lists/netdev/msg637060.html). 
Therefore, in the 3.10 version of the Centos kernel, with the help of 
the vmlinux btf segment, libbpf CO-RE can be easily supported.


>>
>>>>
>>>>
>>>> __UNIQUE_ID uses the __COUNTER__ macro, and the __COUNTER__ macro is
>>>> automatically incremented by 1 every time it is precompiled. Therefore, in
>>>> different compilation units, the same structure has different names.Here is
>>>> a concrete example:
>>>>
>>>> struct acpi_dev_node {
>>>>       union {
>>>>           struct acpi_device *companion;
>>>>           struct {
>>>>               void *handle;
>>>>           } __UNIQUE_ID_rh_kabi_hide29;
>>>>           union {        };
>>>>       };
>>>> };
>>>> struct acpi_dev_node {
>>>>       union {
>>>>           struct acpi_device *companion;
>>>>           struct {
>>>>               void *handle;
>>>>           } __UNIQUE_ID_rh_kabi_hide31;
>>>>           union {        };
>>>>       };
>>>> };
>>>>
>>>> Finally, it will cause the btf algorithm to de-duplication efficiency is not
>>>> high, and time-consuming. For example, running ./pahole -J
>>>> vmlinux-3.10.0-1160.el7.x86_64 without --kabi_prefix flag, the running time
>>>> is:
>>>>                   real 8m28.912s
>>>>                   user 8m27.271s
>>>>                   sys 0m1.471s
>>>> And the size of the generated btf segment is 30678240 bytes.
>>>>
>>>> After adding the patch, running ./pahole
>>>> --kabi_prefix=__UNIQUE_ID_rh_kabi_hide -J vmlinux-3.10.0-1160.el7.x86_64.
>>>> The running time of the command is:
>>>>                   real 0m19.634s
>>>>                   user 0m18.457s
>>>>                   sys 0m1.169s
>>>> And the size of the generated btf segment is 3117719 bytes.
>>>>
>>>> ---
>>>>
>>>> v1:https://lore.kernel.org/dwarves/CAEf4Bzazh4RNCY3rGRRXfO2wJ7DSiMx8w+61B_hjhu9FrOffpQ@mail.gmail.com
>>>> v1->v2:
>>>> --Change btf_prefix to --kabi_prefix.
>>>> --Add man page.
>>>> --Add space after if and comma.
>>>>
>>>>    man-pages/pahole.1 |  4 ++++
>>>>    pahole.c           | 10 ++++++++++
>>>>    pahole_strings.h   |  2 ++
>>>>    strings.c          |  9 ++++++++-
>>>>    4 files changed, 24 insertions(+), 1 deletion(-)
>>>>
>>>> diff --git a/man-pages/pahole.1 b/man-pages/pahole.1
>>>> index a10738f..2659fe6 100644
>>>> --- a/man-pages/pahole.1
>>>> +++ b/man-pages/pahole.1
>>>> @@ -340,6 +340,10 @@ Show a traditional string version, i.e.: "v1.18".
>>>>    Show a numeric only version, suitable for use in Makefiles and scripts
>>>> where
>>>>    one wants to know what if the installed version has some feature, i.e.: 118
>>>> instead of "v1.18".
>>>>
>>>> +.TP
>>>> +.B \-\-kabi_prefix=STRING
>>>> +When the prefix of the string is STRING, treat the string as STRING.
>>>> +
>>>>    .SH NOTES
>>>>
>>>>    To enable the generation of debugging information in the Linux kernel build
>>>> diff --git a/pahole.c b/pahole.c
>>>> index dc40ccf..6a700d9 100644
>>>> --- a/pahole.c
>>>> +++ b/pahole.c
>>>> @@ -24,6 +24,7 @@
>>>>    #include "btf_encoder.h"
>>>>    #include "libbtf.h"
>>>>    #include "lib/bpf/src/libbpf.h"
>>>> +#include "pahole_strings.h"
>>>>
>>>>    static bool btf_encode;
>>>>    static bool ctf_encode;
>>>> @@ -855,6 +856,7 @@ ARGP_PROGRAM_VERSION_HOOK_DEF = dwarves_print_version;
>>>>    #define ARGP_btf_gen_floats	   322
>>>>    #define ARGP_btf_gen_all	   323
>>>>    #define ARGP_with_flexible_array   324
>>>> +#define ARGP_kabi_prefix   325
>>>>
>>>>    static const struct argp_option pahole__options[] = {
>>>>    	{
>>>> @@ -1140,6 +1142,12 @@ static const struct argp_option pahole__options[] = {
>>>>    		.doc  = "Path to the base BTF file",
>>>>    	},
>>>>    	{
>>>> +		.name = "kabi_prefix",
>>>> +		.key = ARGP_kabi_prefix,
>>>> +		.arg = "STRING",
>>>> +		.doc = "When the prefix of the string is STRING, treat the string as
>>>> STRING.",
>>>> +	},
>>>> +	{
>>>>    		.name = "btf_encode",
>>>>    		.key  = 'J',
>>>>    		.doc  = "Encode as BTF",
>>>> @@ -1297,6 +1305,8 @@ static error_t pahole__options_parser(int key, char
>>>> *arg,
>>>>    		btf_encode_force = true;		break;
>>>>    	case ARGP_btf_base:
>>>>    		base_btf_file = arg;			break;
>>>> +	case ARGP_kabi_prefix:
>>>> +		kabi_prefix = arg;		break;
>>>>    	case ARGP_numeric_version:
>>>>    		print_numeric_version = true;		break;
>>>>    	case ARGP_btf_gen_floats:
>>>> diff --git a/pahole_strings.h b/pahole_strings.h
>>>> index 522fbf2..a836ba8 100644
>>>> --- a/pahole_strings.h
>>>> +++ b/pahole_strings.h
>>>> @@ -14,6 +14,8 @@ struct strings {
>>>>    	struct btf *btf;
>>>>    };
>>>>
>>>> +extern const char *kabi_prefix;
>>>> +
>>>>    struct strings *strings__new(void);
>>>>
>>>>    void strings__delete(struct strings *strings);
>>>> diff --git a/strings.c b/strings.c
>>>> index d37f49d..d1a54ec 100644
>>>> --- a/strings.c
>>>> +++ b/strings.c
>>>> @@ -17,6 +17,8 @@
>>>>    #include "dutil.h"
>>>>    #include "lib/bpf/src/libbpf.h"
>>>>
>>>> +const char *kabi_prefix;
>>>> +
>>>>    struct strings *strings__new(void)
>>>>    {
>>>>    	struct strings *strs = malloc(sizeof(*strs));
>>>> @@ -48,7 +50,12 @@ strings_t strings__add(struct strings *strs, const char
>>>> *str)
>>>>    	if (str == NULL)
>>>>    		return 0;
>>>>
>>>> -	index = btf__add_str(strs->btf, str);
>>>> +	if (kabi_prefix &&
>>>> +		strncmp(str, kabi_prefix, strlen(kabi_prefix)) == 0)
>>>> +		index = btf__add_str(strs->btf, kabi_prefix);
>>>> +	else
>>>> +		index = btf__add_str(strs->btf, str);
>>>> +		
>>>>    	if (index < 0)
>>>>    		return 0;
>>>>
>>>> -- 
>>>> 1.8.3.1
>>>>
>>