BPF List
 help / color / mirror / Atom feed
From: Alan Maguire <alan.maguire@oracle.com>
To: Eduard Zingerman <eddyz87@gmail.com>,
	acme@kernel.org, andrii.nakryiko@gmail.com
Cc: jolsa@kernel.org, ast@kernel.org, daniel@iogearbox.net,
	martin.lau@linux.dev, song@kernel.org, yhs@fb.com,
	john.fastabend@gmail.com, kpsingh@kernel.org, sdf@google.com,
	haoluo@google.com, mykolal@fb.com, bpf@vger.kernel.org,
	Andrii Nakryiko <andrii@kernel.org>
Subject: Re: [RFC dwarves 3/4] pahole: add --btf_features=feature1[,feature2...] support
Date: Wed, 11 Oct 2023 17:41:10 +0100	[thread overview]
Message-ID: <f822334f-335e-bd38-09c7-95c69086ba6f@oracle.com> (raw)
In-Reply-To: <b7b61031f41ab4082205ed061bb66cb859bd1f0d.camel@gmail.com>

On 11/10/2023 17:28, Eduard Zingerman wrote:
> On Wed, 2023-10-11 at 10:17 +0100, Alan Maguire wrote:
>> This allows consumers to specify an opt-in set of features
>> they want to use in BTF encoding.
>>
>> Supported features are
>>
>> 	encode_force  Ignore invalid symbols when encoding BTF.
>> 	var           Encode variables using BTF_KIND_VAR in BTF.
>> 	float         Encode floating-point types in BTF.
>> 	decl_tag      Encode declaration tags using BTF_KIND_DECL_TAG.
>> 	type_tag      Encode type tags using BTF_KIND_TYPE_TAG.
>> 	enum64        Encode enum64 values with BTF_KIND_ENUM64.
>> 	optimized     Encode representations of optimized functions
>> 	              with suffixes like ".isra.0" etc
>> 	consistent    Avoid encoding inconsistent static functions.
>> 	              These occur when a parameter is optimized out
>> 	              in some CUs and not others, or when the same
>> 	              function name has inconsistent BTF descriptions
>> 	              in different CUs.
>>
>> Specifying "--btf_features=all" is the equivalent to setting
>> all of the above.  If pahole does not know about a feature
>> it silently ignores it.  These properties allow us to use
>> the --btf_features option in the kernel pahole_flags.sh
>> script to specify the desired set of features.  If a new
>> feature is not present in pahole but requested, pahole
>> BTF encoding will not complain (but will not encode the
>> feature).
>>
>> Suggested-by: Andrii Nakryiko <andrii@kernel.org>
>> Signed-off-by: Alan Maguire <alan.maguire@oracle.com>
>> ---
>>  man-pages/pahole.1 | 20 +++++++++++
>>  pahole.c           | 87 +++++++++++++++++++++++++++++++++++++++++++++-
>>  2 files changed, 106 insertions(+), 1 deletion(-)
>>
>> diff --git a/man-pages/pahole.1 b/man-pages/pahole.1
>> index c1b48de..7c072dc 100644
>> --- a/man-pages/pahole.1
>> +++ b/man-pages/pahole.1
>> @@ -273,6 +273,26 @@ Generate BTF for functions with optimization-related suffixes (.isra, .constprop
>>  .B \-\-btf_gen_all
>>  Allow using all the BTF features supported by pahole.
>>  
>> +.TP
>> +.B \-\-btf_features=FEATURE_LIST
>> +Encode BTF using the specified feature list, or specify 'all' for all features supported.  This single parameter value can be used as an alternative to unsing multiple BTF-related options. Supported features are
>> +
>> +.nf
>> +	encode_force  Ignore invalid symbols when encoding BTF.
>> +	var           Encode variables using BTF_KIND_VAR in BTF.
>> +	float         Encode floating-point types in BTF.
>> +	decl_tag      Encode declaration tags using BTF_KIND_DECL_TAG.
>> +	type_tag      Encode type tags using BTF_KIND_TYPE_TAG.
>> +	enum64        Encode enum64 values with BTF_KIND_ENUM64.
>> +	optimized     Encode representations of optimized functions
>> +	              with suffixes like ".isra.0" etc
>> +	consistent    Avoid encoding inconsistent static functions.
>> +	              These occur when a parameter is optimized out
>> +	              in some CUs and not others, or when the same
>> +	              function name has inconsistent BTF descriptions
>> +	              in different CUs.
>> +.fi
>> +
>>  .TP
>>  .B \-l, \-\-show_first_biggest_size_base_type_member
>>  Show first biggest size base_type member.
>> diff --git a/pahole.c b/pahole.c
>> index 7a41dc3..4f00b08 100644
>> --- a/pahole.c
>> +++ b/pahole.c
>> @@ -1229,6 +1229,83 @@ ARGP_PROGRAM_VERSION_HOOK_DEF = dwarves_print_version;
>>  #define ARGP_skip_emitting_atomic_typedefs 338
>>  #define ARGP_btf_gen_optimized  339
>>  #define ARGP_skip_encoding_btf_inconsistent_proto 340
>> +#define ARGP_btf_features	341
>> +
>> +/* --btf_features=feature1[,feature2,..] option allows us to specify
>> + * opt-in features (or "all"); these are translated into conf_load
>> + * values by specifying the associated bool offset and whether it
>> + * is a skip option or not; btf_features is for opting _into_ features
>> + * so for skip options we have to reverse the logic.  For example
>> + * "--skip_encoding_btf_type_tag --btf_gen_floats" translate to
>> + * "--btf_features=type_tag,float"
>> + */
>> +#define BTF_FEATURE(name, alias, skip)				\
>> +	{ #name, #alias, offsetof(struct conf_load, alias), skip }
>> +
>> +struct btf_feature {
>> +	const char      *name;
>> +	const char      *option_alias;
>> +	size_t          conf_load_offset;
>> +	bool		skip;
>> +} btf_features[] = {
>> +	BTF_FEATURE(encode_force, btf_encode_force, false),
>> +	BTF_FEATURE(var, skip_encoding_btf_vars, true),
>> +	BTF_FEATURE(float, btf_gen_floats, false),
>> +	BTF_FEATURE(decl_tag, skip_encoding_btf_decl_tag, true),
>> +	BTF_FEATURE(type_tag, skip_encoding_btf_type_tag, true),
>> +	BTF_FEATURE(enum64, skip_encoding_btf_enum64, true),
>> +	BTF_FEATURE(optimized, btf_gen_optimized, false),
>> +	/* the "skip" in skip_encoding_btf_inconsistent_proto is misleading
>> +	 * here; this is a positive feature to ensure consistency of
>> +	 * representation rather than a negative option which we want
>> +	 * to invert.  So as a result, "skip" is false here.
>> +	 */
>> +	BTF_FEATURE(consistent, skip_encoding_btf_inconsistent_proto, false),
>> +};
>> +
>> +#define BTF_MAX_FEATURES	32
>> +#define BTF_MAX_FEATURE_STR	256
>> +
>> +/* Translate --btf_features=feature1[,feature2] into conf_load values.
>> + * Explicitly ignores unrecognized features to allow future specification
>> + * of new opt-in features.
>> + */
>> +static void parse_btf_features(const char *features, struct conf_load *conf_load)
>> +{
>> +	char *feature_list[BTF_MAX_FEATURES] = {};
>> +	char f[BTF_MAX_FEATURE_STR];
>> +	bool encode_all = false;
>> +	int i, j, n = 0;
>> +
>> +	strncpy(f, features, sizeof(f));
>> +
>> +	if (strcmp(features, "all") == 0) {
>> +		encode_all = true;
>> +	} else {
>> +		char *saveptr = NULL, *s = f, *t;
>> +
>> +		while ((t = strtok_r(s, ",", &saveptr)) != NULL) {
>> +			s = NULL;
>> +			feature_list[n++] = t;
> 
> Maybe guard against `n` >= BTF_MAX_FEATURES here?
>

good point - will fix.

>> +		}
>> +	}
>> +
>> +	for (i = 0; i < ARRAY_SIZE(btf_features); i++) {
>> +		bool *bval = (bool *)(((void *)conf_load) + btf_features[i].conf_load_offset);
>> +		bool match = encode_all;
>> +
>> +		if (!match) {
>> +			for (j = 0; j < n; j++) {
>> +				if (strcmp(feature_list[j], btf_features[i].name) == 0) {
>> +					match = true;
>> +					break;
>> +				}
>> +			}
>> +		}
>> +		if (match)
>> +			*bval = btf_features[i].skip ? false : true;
> 
> I'm not sure I understand the logic behind "skip" features.
> Take `decl_tag` for example:
> - by default conf_load->skip_encoding_btf_decl_tag is 0;
> - if `--btf_features=decl_tag` is passed it is still 0 because of the
>   `skip ? false : true` logic.
> 
> If there is no way to change "skip" features why listing these at all?
> 
You're right; in the case of a skip feature, I think we need the
following behaviour

1. we skip the encoding by default (so the equivalent of
--skip_encoding_btf_decl_tag, setting skip_encoding_btf_decl_tag
to true
2. if the user however specifies the logical inversion of the skip
feature in --btf_features (in this case "decl_tag" - or "all")
skip_encoding_btf_decl_tag is set to false.

So in my code we had 2 above but not 1. If both were in place I think
we'd have the right set of behaviours. Does that sound right?

Maybe a better way to express all this would be to rename the "skip"
field in "struct btf_feature" to "default" - so in the case of a "skip"
feature, the default is true, but for opt-in features, the default is false.

> Other than that I tested the patch-set with current kernel master and
> a change to pahole-flags.sh and bpf tests pass.
> 

Thanks so much for testing this!

Alan

>> +	}
>> +}
>>  
>>  static const struct argp_option pahole__options[] = {
>>  	{
>> @@ -1651,6 +1728,12 @@ static const struct argp_option pahole__options[] = {
>>  		.key = ARGP_skip_encoding_btf_inconsistent_proto,
>>  		.doc = "Skip functions that have multiple inconsistent function prototypes sharing the same name, or that use unexpected registers for parameter values."
>>  	},
>> +	{
>> +		.name = "btf_features",
>> +		.key = ARGP_btf_features,
>> +		.arg = "FEATURE_LIST",
>> +		.doc = "Specify supported BTF features in FEATURE_LIST or 'all' for all supported features. See the pahole manual page for the list of supported features."
>> +	},
>>  	{
>>  		.name = NULL,
>>  	}
>> @@ -1796,7 +1879,7 @@ static error_t pahole__options_parser(int key, char *arg,
>>  	case ARGP_btf_gen_floats:
>>  		conf_load.btf_gen_floats = true;	break;
>>  	case ARGP_btf_gen_all:
>> -		conf_load.btf_gen_floats = true;	break;
>> +		parse_btf_features("all", &conf_load);	break;
>>  	case ARGP_with_flexible_array:
>>  		show_with_flexible_array = true;	break;
>>  	case ARGP_prettify_input_filename:
>> @@ -1826,6 +1909,8 @@ static error_t pahole__options_parser(int key, char *arg,
>>  		conf_load.btf_gen_optimized = true;		break;
>>  	case ARGP_skip_encoding_btf_inconsistent_proto:
>>  		conf_load.skip_encoding_btf_inconsistent_proto = true; break;
>> +	case ARGP_btf_features:
>> +		parse_btf_features(arg, &conf_load);	break;
>>  	default:
>>  		return ARGP_ERR_UNKNOWN;
>>  	}
> 

  reply	other threads:[~2023-10-11 16:41 UTC|newest]

Thread overview: 21+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2023-10-11  9:17 [RFC dwarves 0/4] pahole, btf_encoder: support --btf_features Alan Maguire
2023-10-11  9:17 ` [RFC dwarves 1/4] btf_encoder, pahole: move btf encoding options into conf_load Alan Maguire
2023-10-12 12:54   ` Jiri Olsa
2023-10-11  9:17 ` [RFC dwarves 2/4] dwarves: move ARRAY_SIZE() to dwarves.h Alan Maguire
2023-10-11  9:17 ` [RFC dwarves 3/4] pahole: add --btf_features=feature1[,feature2...] support Alan Maguire
2023-10-11 16:28   ` Eduard Zingerman
2023-10-11 16:41     ` Alan Maguire [this message]
2023-10-11 19:08       ` Eduard Zingerman
2023-10-11 22:05         ` Alan Maguire
2023-10-11 22:14           ` Eduard Zingerman
2023-10-12 12:35             ` Alan Maguire
2023-10-13 14:17               ` Arnaldo Carvalho de Melo
2023-10-13 14:43                 ` Alan Maguire
2023-10-12 12:53   ` Jiri Olsa
2023-10-12 13:48     ` Alan Maguire
2023-10-12 21:19       ` Jiri Olsa
2023-10-13  0:21   ` Andrii Nakryiko
2023-10-13 11:54     ` Alan Maguire
2023-10-13 18:39       ` Andrii Nakryiko
2023-10-11  9:17 ` [RFC dwarves 4/4] pahole: add --supported_btf_features to display feature support Alan Maguire
2023-10-13  0:14 ` [RFC dwarves 0/4] pahole, btf_encoder: support --btf_features Andrii Nakryiko

Reply instructions:

You may reply publicly to this message via plain-text email
using any one of the following methods:

* Save the following mbox file, import it into your mail client,
  and reply-to-all from there: mbox

  Avoid top-posting and favor interleaved quoting:
  https://en.wikipedia.org/wiki/Posting_style#Interleaved_style

* Reply using the --to, --cc, and --in-reply-to
  switches of git-send-email(1):

  git send-email \
    --in-reply-to=f822334f-335e-bd38-09c7-95c69086ba6f@oracle.com \
    --to=alan.maguire@oracle.com \
    --cc=acme@kernel.org \
    --cc=andrii.nakryiko@gmail.com \
    --cc=andrii@kernel.org \
    --cc=ast@kernel.org \
    --cc=bpf@vger.kernel.org \
    --cc=daniel@iogearbox.net \
    --cc=eddyz87@gmail.com \
    --cc=haoluo@google.com \
    --cc=john.fastabend@gmail.com \
    --cc=jolsa@kernel.org \
    --cc=kpsingh@kernel.org \
    --cc=martin.lau@linux.dev \
    --cc=mykolal@fb.com \
    --cc=sdf@google.com \
    --cc=song@kernel.org \
    --cc=yhs@fb.com \
    /path/to/YOUR_REPLY

  https://kernel.org/pub/software/scm/git/docs/git-send-email.html

* If your mail client supports setting the In-Reply-To header
  via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox