Message-ID: <530C5C69.9080800@linux.vnet.ibm.com>
Date:	Tue, 25 Feb 2014 14:33:37 +0530
From:	Hemant Kumar <hkshaw@...ux.vnet.ibm.com>
To:	Namhyung Kim <namhyung@...nel.org>
CC:	linux-kernel@...r.kernel.org, srikar@...ux.vnet.ibm.com,
	peterz@...radead.org, oleg@...hat.com,
	hegdevasant@...ux.vnet.ibm.com, mingo@...hat.com, anton@...hat.com,
	systemtap@...rceware.org, masami.hiramatsu.pt@...achi.com,
	aravinda@...ux.vnet.ibm.com, penberg@....fi
Subject: Re: [RFC PATCH v1 1/2] perf/sdt : Listing of SDT markers by perf

On 02/25/2014 12:26 PM, Namhyung Kim wrote:
> Hi Hemant,
>
> On Mon, 24 Feb 2014 14:45:43 +0530, Hemant Kumar wrote:
>> This patch enables perf to list the SDT markers present in a system. It looks
>> in dsos given by ldconfig --print-cache and for other binaries, it looks into
>> the PATH environment variable. After preparing a list of the binaries, then
>> it starts searching for SDT markers in them.
>> To find the SDT markers, first an elf section named .note.stapsdt is searched
>> for. And then the SDT notes are retreived one by one from that section.
>> To counter the effect of prelinking, the section ".stapsdt.base" is searched.
>> If its found, then the location of the SDT marker is adjusted.
>>
>> All these markers' info is written into a cache file
>> "/var/cache/perf/perf-sdt.cache".
>> Since, the presence of SDT markers is quite common these days, hence, its better
>> to make them visible to a user easily. Also, creating a cache file will help a user
>> to probe (to be implemented) these markers without much hussle. This cache file will
>> hold most of the SDT markers.
>> The format of each entry is -
>>
>> %provider : marker : file_path : build_id : location : semaphore_loc
>>
>> % - marks the beginning of each entry.
>> provider - The provider name of the SDT marker.
>> marker - The marker name of the SDT marker.
>> file_path - Full/absolute path of the file in which this marker is present.
>> location : Adjusted location of the SDT marker inside the program.
>> semaphore_loc : The semaphore address if present otherwise 0x0.
>>
>> This format should help when probing will be implemented. The adjusted address
>> from the required entry can be directly used in probing if the build_id matches.
>>
>> To use this feature, invoke :
>> # perf list sdt --scan
>>
>> "--scan" should be used for the first time and whenever there is any change in
>> the files containing the SDT markers.
>>
>> And then use :
>> # perf list sdt
>>
>> This displays a list of SDT markers (read from the cache file) present in most of
>> the dsos and other binaries.
> The default action of perf list (with no argument) prints all available
> events on the system - so it should also print SDT markers IMHO.

Right, but won't that list be huge?

>> However, not all the SDT markers are present in the cache file. Only the ELFs
>> present in the directories in PATH variable are looked into.
>> An individual file argument can be given to "perf list" to find out the SDT markers
>> present in that file. Usage is as below :
>>
>> # perf list sdt /home/hemant/tmp
>>
>> /home/hemant/tmp:
>> %user : foo
>> %user : bar
>>
>> Signed-off-by : hkshaw@...ux.vnet.ibm.com
>> ---
> [SNIP]
>> +/*
>> + * Finds out the libraries present in a system as shown by the command
>> + * "ldconfig --print-cache". Uses "=>" and '/' to find out the start of a
>> + * dso path.
>> + */
>> +static void parse_lib_name(char *str, char *path)
>> +{
>> +	char *ptr, *q, *r;
>> +	char c = '=';
>> +
>> +	while (str != NULL) {
>> +		/* look for "=>" and then '/' */
>> +		ptr = strchr(str, c);
>> +		if (ptr && (*(ptr + 1) == '>')) {
> Hmm.. why not searching "=>" by strstr() then?

Right! will change it to strstr.

>
>> +			ptr++;
>> +			q = strchr(ptr, '/');
>> +			if (!q)
>> +				return;
>> +			r = strchr(ptr, '\n');
>> +			*r = '\0';
>> +			strcpy(path, q);
>> +			return;
>> +		} else if (ptr == NULL) {
>> +			return;
>> +		} else {
>> +			str = ptr + 1;
>> +			continue;
>> +		}
>> +	}
>> +}
> [SNIP]
>> +	/*
>> +	 * Take a line from the o/p of ldconfig and
>> +	 * parse it to find a library path
>> +	 */
>> +	while (getline(&line, &len, fp) != -1) {
>> +		memset(path, '\0', PATH_MAX);
>> +		parse_lib_name(line, path);
>> +		if (strcmp(path, ""))
> I think it'd better changing the parse_lib_name to return a value
> instead of checking whether the path is an empty string.

Ok. Seems better!

>
>
>> +			append_path(path, &lib_list.list);
>> +	}
>> +	/* line and fp no more required */
>> +	free(line);
>> +	pclose(fp);
>> +
>> +	/* Find and write the SDT markers in the perf-sdt cache*/
>> +	flush_sdt_list(&lib_list.list, cache);
>> +
>> +	cleanup_path_list(&lib_list.list);
>> +	return;
>> +}
> [SNIP]
>> +static int populate_sdt_note(Elf **elf, const char *data, size_t len, int type,
>> +			     struct sdt_note **note)
>> +{
>> +	const char *provider, *name;
>> +	struct sdt_note *tmp = NULL;
>> +	GElf_Ehdr ehdr;
>> +	GElf_Addr base_off = 0;
>> +	GElf_Shdr shdr;
>> +	int ret = -1;
>> +	int i;
>> +
>> +	union {
>> +		Elf64_Addr a64[3];
>> +		Elf32_Addr a32[3];
>> +	} buf;
>> +
>> +	Elf_Data dst = {
>> +		.d_buf = &buf, .d_type = ELF_T_ADDR, .d_version = EV_CURRENT,
>> +		.d_size = gelf_fsize((*elf), ELF_T_ADDR, 3, EV_CURRENT),
>> +		.d_off = 0, .d_align = 0
>> +	};
>> +	Elf_Data src = {
>> +		.d_buf = (void *) data, .d_type = ELF_T_ADDR,
>> +		.d_version = EV_CURRENT, .d_size = dst.d_size, .d_off = 0,
>> +		.d_align = 0
>> +	};
>> +
>> +	/* Check the type of each of the notes */
>> +	if (type != SDT_NOTE_TYPE)
>> +		goto out_err;
>> +
>> +	tmp = (struct sdt_note *)calloc(1, sizeof(struct sdt_note));
>> +	if (!tmp) {
>> +		ret = -ENOMEM;
>> +		goto out_err;
>> +	}
>> +
>> +	INIT_LIST_HEAD(&tmp->note_list);
>> +
>> +	if (len < dst.d_size + 3)
>> +		goto out_free_note;
>> +
>> +	/* Translation from file representation to memory representation */
>> +	if (gelf_xlatetom(*elf, &dst, &src,
>> +			  elf_getident(*elf, NULL)[EI_DATA]) == NULL)
> Do we really need this xlate function?  It seems elf_getdata() already
> did necessary conversions so only thing we need to do is checking its
> class and read out the addresses in a proper length, no?
>

Hmm, alright. I thought the conversion was necessary for cross developed 
binaries.
But I guess elf_getdata() should do all these conversions. Will remove that.

>> +		printf("gelf_xlatetom : %s\n", elf_errmsg(-1));
>> +
>> +	/* Populate the fields of sdt_note */
>> +	provider = data + dst.d_size;
>> +
>> +	name = (const char *)memchr(provider, '\0', data + len - provider);
>> +	if (name++ == NULL)
>> +		goto out_free_note;
>> +
>> +	tmp->provider = strdup(provider);
>> +	if (!tmp->provider) {
>> +		ret = -ENOMEM;
>> +		goto out_free_note;
>> +	}
>> +	tmp->name = strdup(name);
>> +	if (!tmp->name) {
>> +		ret = -ENOMEM;
>> +		goto out_free_prov;
>> +	}
>> +
>> +	/* Obtain the addresses */
>> +	if (gelf_getclass(*elf) == ELFCLASS32) {
>> +		for (i = 0; i < 3; i++)
>> +			tmp->addr.a32[i] = buf.a32[i];
>> +		tmp->bit32 = true;
>> +	} else {
>> +		for (i = 0; i < 3; i++)
>> +			tmp->addr.a64[i] = buf.a64[i];
>> +		tmp->bit32 = false;
>> +	}
>> +
>> +	/* Now Adjust the prelink effect */
>> +	if (!gelf_getehdr(*elf, &ehdr)) {
>> +		pr_debug("%s : cannot get elf header.\n", __func__);
>> +		ret = -EBADF;
>> +		goto out_free_name;
>> +	}
>> +
>> +	/*
>> +	 * Find out the .stapsdt.base section.
>> +	 * This scn will help us to handle prelinking (if present).
>> +	 * Compare the retrieved file offset of the base section with the
>> +	 * base address in the description of the SDT note. If its different,
>> +	 * then accordingly, adjust the note location.
>> +	 */
>> +	if (elf_section_by_name(*elf, &ehdr, &shdr, SDT_BASE_SCN, NULL))
>> +		base_off = shdr.sh_offset;
>> +	if (base_off) {
> Why not moving this block under the above if block?
>

Yeah, better! Will do it.

>> +		if (tmp->bit32)
>> +			tmp->addr.a32[0] = tmp->addr.a32[0] + base_off -
>> +				tmp->addr.a32[1];
>> +		else
>> +			tmp->addr.a64[0] = tmp->addr.a64[0] + base_off -
>> +				tmp->addr.a64[1];
>> +	}
>> +
>> +	*note = tmp;
>> +	return 0;
>> +
>> +out_free_name:
>> +	free(tmp->name);
>> +out_free_prov:
>> +	free(tmp->provider);
>> +out_free_note:
>> +	free(tmp);
>> +out_err:
>> +	return ret;
>> +}
>> +
>> +static int construct_sdt_notes_list(Elf *elf, struct list_head *sdt_notes)
>> +{
>> +	GElf_Ehdr ehdr;
>> +	Elf_Scn *scn = NULL;
>> +	Elf_Data *data;
>> +	GElf_Shdr shdr;
>> +	size_t shstrndx, next;
>> +	GElf_Nhdr nhdr;
>> +	size_t name_off, desc_off, offset;
>> +	struct sdt_note *tmp = NULL;
>> +	int ret = 0, val = 0;
>> +
>> +	if (gelf_getehdr(elf, &ehdr) == NULL) {
>> +		ret = -EBADF;
>> +		goto out_ret;
>> +	}
>> +	if (elf_getshdrstrndx(elf, &shstrndx) != 0) {
>> +		ret = -EBADF;
>> +		goto out_ret;
>> +	}
>> +
>> +	/* Look for the required section */
>> +	scn = elf_section_by_name(elf, &ehdr, &shdr, SDT_NOTE_SCN, NULL);
>> +	if (!scn) {
>> +		ret = -ENOENT;
>> +		goto out_ret;
>> +	}
>> +
>> +	if (!(shdr.sh_type == SHT_NOTE) || (shdr.sh_flags & SHF_ALLOC)) {
>> +		ret = -ENOENT;
>> +		goto out_ret;
>> +	}
>> +
>> +	data = elf_getdata(scn, NULL);
>> +
>> +	/* Get the SDT notes */
>> +	for (offset = 0; (next = gelf_getnote(data, offset, &nhdr, &name_off,
>> +					      &desc_off)) > 0; offset = next) {
>> +		if (nhdr.n_namesz == sizeof(SDT_NOTE_NAME) &&
>> +		    !memcmp(data->d_buf + name_off, SDT_NOTE_NAME,
>> +			    sizeof(SDT_NOTE_NAME))) {
>> +			val = populate_sdt_note(&elf, ((data->d_buf) + desc_off),
>> +						nhdr.n_descsz, nhdr.n_type,
>> +						&tmp);
>> +			if (!val)
>> +				list_add_tail(&tmp->note_list, sdt_notes);
>> +			if (val == -ENOMEM) {
>> +				ret = val;
>> +				goto out_ret;
>> +			}
>> +		}
>> +	}
>> +	if (!sdt_notes)
> Did you mean list_empty(sdt_notes) ? :)

Oops! Yes, will fix it.

[SNIP]

Thanks a lot for reviewing the patch.

-- 
Thanks
Hemant Kumar

--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@...r.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/

Powered by blists - more mailing lists