[lttng-tools.git] / src / bin / lttng-relayd / backward-compatibility-group-by.c

/*
 * Copyright (C) 2019 Jonathan Rajotte <jonathan.rajotte-julien@efficios.com>
 *
 * SPDX-License-Identifier: GPL-2.0-only
 *
 */

#include "common/time.h"
#include <assert.h>
#include <regex.h>
#include <stdio.h>
#include <stdlib.h>
#include <string.h>

#include <common/common.h>
#include <common/defaults.h>
#include <common/utils.h>

#include "backward-compatibility-group-by.h"

#define DATETIME_REGEX \
	".*-[1-2][0-9][0-9][0-9][0-1][0-9][0-3][0-9]-[0-2][0-9][0-5][0-9][0-5][0-9]$"

/*
 * Provide support for --group-output-by-session for producer >= 2.4 and < 2.11.
 * Take the stream path, extract all available information, craft a new path to
 * the best of our ability enforcing the group by session.
 *
 * Return the allocated string containing the new stream path or else NULL.
 */
char *backward_compat_group_by_session(const char *path,
		const char *local_session_name,
		time_t relay_session_creation_time)
{
	int ret;
	size_t len;
	char *leftover_ptr;
	char *local_copy = NULL;
	char *datetime = NULL;
	char *partial_base_path = NULL;
	char *filepath_per_session = NULL;
	const char *second_token_ptr;
	const char *leftover_second_token_ptr;
	const char *hostname_ptr;
	regex_t regex;

	assert(path);
	assert(local_session_name);
	assert(local_session_name[0] != '\0');

	DBG("Parsing path \"%s\" of session \"%s\" to create a new path that is grouped by session",
			path, local_session_name);

	/* Get a local copy for strtok */
	local_copy = strdup(path);
	if (!local_copy) {
		PERROR("Failed to parse session path: couldn't copy input path");
		goto error;
	}

	/*
	 * The use of strtok with '/' as delimiter is valid since we refuse '/'
	 * in session name and '/' is not a valid hostname character based on
	 * RFC-952 [1], RFC-921 [2] and refined in RFC-1123 [3].
	 * [1] https://tools.ietf.org/html/rfc952
	 * [2] https://tools.ietf.org/html/rfc921
	 * [3] https://tools.ietf.org/html/rfc1123#page-13
	 */

	/*
	 * Get the hostname and possible session_name.
	 * Note that we can get the hostname and session name from the
	 * relay_session object we already have. Still, it is easier to
	 * tokenized the passed path to obtain the start of the path leftover.
	 */
	hostname_ptr = strtok_r(local_copy, "/", &leftover_ptr);
	if (!hostname_ptr) {
		ERR("Failed to parse session path \"%s\": couldn't identify hostname",
				path);
		goto error;
	}

	second_token_ptr = strtok_r(NULL, "/", &leftover_ptr);
	if (!second_token_ptr) {
		ERR("Failed to parse session path \"%s\": couldn't identify session name",
				path);
		goto error;
	}

	/*
	 * Check if the second token is a base path set at url level. This is
	 * legal in streaming, live and snapshot [1]. Otherwise it is the
	 * session name with possibly a datetime attached [2]. Note that when
	 * "adding" snapshot output (lttng snapshot add-output), no session name
	 * is present in the path by default. The handling for "base path" take
	 * care of this case as well.
	 * [1] e.g --set-url net://localhost/my_marvellous_path
	 * [2] Can be:
	 *            <session_name>
	 *                When using --snapshot on session create.
	 *            <session_name>-<date>-<time>
	 *            <auto>-<date>-<time>
	 */
	if (strncmp(second_token_ptr, local_session_name,
			    strlen(local_session_name)) != 0) {
		/*
		 * Token does not start with session name.
		 * This mean this is an extra path scenario.
		 * Duplicate the current token since it is part of an
		 * base_path.
		 * Set secDuplicate the current token since it is part of an
		 * base_path. The rest is the leftover.
		 * Set second_token_ptr to the local_session_name for further
		 * processing.
		 */
		partial_base_path = strdup(second_token_ptr);
		if (!partial_base_path) {
			PERROR("Failed to parse session path: couldn't copy partial base path");
			goto error;
		}

		second_token_ptr = local_session_name;
	}

	/*
	 * Based on the previous test, we can move inside the token ptr to
	 * remove the "local_session_name" and inspect the rest of the token.
	 * We are looking into extracting the creation datetime from either the
	 * session_name or the token. We need to to all this gymnastic because
	 * an extra path could decide to append a datetime to its first
	 * subdirectory.
	 * Possible scenario:
	 *     <session_name>
	 *     <session_name>-<date>-<time>
	 *     <auto>-<date>-<time>
	 *     <session_name>_base_path_foo_bar
	 *     <session_name>-<false date>-<false-time> (via a base path)
	 *
	 * We have no way to discern from the basic scenario of:
	 *     <session_name>-<date>-<time>
	 * and one done using a base path with the exact format we normally
	 * expect.
	 *
	 * e.g:
	 *     lttng create my_session -U
	 *         net://localhost/my_session-19910319-120000/
	 */
	ret = regcomp(&regex, DATETIME_REGEX, 0);
	if (ret) {
		ERR("Failed to parse session path: regex compilation failed with code %d", ret);
		goto error;
	}

	leftover_second_token_ptr =
			second_token_ptr + strlen(local_session_name);
	len = strlen(leftover_second_token_ptr);
	if (len == 0) {
		/*
		 * We are either dealing with an auto session name or only the
		 * session_name. If this is a auto session name, we need to
		 * fetch the creation datetime.
		 */
		ret = regexec(&regex, local_session_name, 0, NULL, 0);
		if (ret == 0) {
			const ssize_t local_session_name_offset =
					strlen(local_session_name) - DATETIME_STR_LEN + 1;

			assert(local_session_name_offset >= 0);
			datetime = strdup(local_session_name +
					local_session_name_offset);
			if (!datetime) {
				PERROR("Failed to parse session path: couldn't copy datetime on regex match");
				goto error_regex;
			}
		} else {
			datetime = zmalloc(DATETIME_STR_LEN);
			if (!datetime) {
				PERROR("Failed to allocate DATETIME string");
				goto error;
			}

			ret = time_to_datetime_str(relay_session_creation_time,
					datetime, DATETIME_STR_LEN);
			if (ret) {
				/* time_to_datetime_str already logs errors. */
				goto error;
			}
		}
	} else if (len == DATETIME_STR_LEN &&
			!regexec(&regex, leftover_second_token_ptr, 0, NULL,
					0)) {
		/*
		 * The leftover from the second token is of format
		 * "-<datetime>", use it as the creation time.
		 * Ignore leading "-".
		 */
		datetime = strdup(&leftover_second_token_ptr[1]);
		if (!datetime) {
			PERROR("Failed to parse session path: couldn't copy datetime on regex match");
			goto error_regex;
		}
	} else {
		/*
		 * Base path scenario.
		 * We cannot try to extract the datetime from the session name
		 * since nothing prevent a user to name a session in the
		 * "name-<datetime>" format. Using the datetime from such a
		 * session would be invalid.
		 * */
		assert(partial_base_path == NULL);
		assert(datetime == NULL);

		partial_base_path = strdup(second_token_ptr);
		if (!partial_base_path) {
			PERROR("Failed to parse session path: couldn't copy partial base path");
			goto error_regex;
		}
	}

	ret = asprintf(&filepath_per_session, "%s/%s%s%s/%s%s%s",
			local_session_name, hostname_ptr, datetime ? "-" : "",
			datetime ? datetime : "",
			partial_base_path ? partial_base_path : "",
			partial_base_path ? "/" : "", leftover_ptr);
	if (ret < 0) {
		filepath_per_session = NULL;
		goto error;
	}
error_regex:
	regfree(&regex);
error:
	free(local_copy);
	free(partial_base_path);
	free(datetime);
	return filepath_per_session;
}
Commit	Line	Data
2a635488	1	/*
ab5be9fa	2	* Copyright (C) 2019 Jonathan Rajotte <jonathan.rajotte-julien@efficios.com>
2a635488	3	*
ab5be9fa	4	* SPDX-License-Identifier: GPL-2.0-only
2a635488	5	*
2a635488 JR	6	*/
2a635488 JR	7
d2cb4a90	8	#include "common/time.h"
2a635488 JR	9	#include <assert.h>
	10	#include <regex.h>
	11	#include <stdio.h>
	12	#include <stdlib.h>
	13	#include <string.h>
	14
	15	#include <common/common.h>
	16	#include <common/defaults.h>
	17	#include <common/utils.h>
	18
	19	#include "backward-compatibility-group-by.h"
	20
2a635488	21	#define DATETIME_REGEX \
eb60c7af	22	".*-[1-2][0-9][0-9][0-9][0-1][0-9][0-3][0-9]-[0-2][0-9][0-5][0-9][0-5][0-9]$"
2a635488 JR	23
	24	/*
	25	* Provide support for --group-output-by-session for producer >= 2.4 and < 2.11.
	26	* Take the stream path, extract all available information, craft a new path to
	27	* the best of our ability enforcing the group by session.
	28	*
	29	* Return the allocated string containing the new stream path or else NULL.
	30	*/
d2cb4a90 JG	31	char backward_compat_group_by_session(const char path,
	32	const char *local_session_name,
	33	time_t relay_session_creation_time)
2a635488 JR	34	{
	35	int ret;
	36	size_t len;
	37	char *leftover_ptr;
	38	char *local_copy = NULL;
	39	char *datetime = NULL;
	40	char *partial_base_path = NULL;
	41	char *filepath_per_session = NULL;
	42	const char *second_token_ptr;
	43	const char *leftover_second_token_ptr;
	44	const char *hostname_ptr;
	45	regex_t regex;
	46
	47	assert(path);
	48	assert(local_session_name);
	49	assert(local_session_name[0] != '\0');
	50
	51	DBG("Parsing path \"%s\" of session \"%s\" to create a new path that is grouped by session",
	52	path, local_session_name);
	53
	54	/* Get a local copy for strtok */
	55	local_copy = strdup(path);
	56	if (!local_copy) {
	57	PERROR("Failed to parse session path: couldn't copy input path");
	58	goto error;
	59	}
	60
	61	/*
	62	* The use of strtok with '/' as delimiter is valid since we refuse '/'
	63	* in session name and '/' is not a valid hostname character based on
4f173935	64	* RFC-952 [1], RFC-921 [2] and refined in RFC-1123 [3].
2a635488 JR	65	* [1] https://tools.ietf.org/html/rfc952
	66	* [2] https://tools.ietf.org/html/rfc921
	67	* [3] https://tools.ietf.org/html/rfc1123#page-13
	68	*/
	69
	70	/*
	71	* Get the hostname and possible session_name.
	72	* Note that we can get the hostname and session name from the
	73	* relay_session object we already have. Still, it is easier to
	74	* tokenized the passed path to obtain the start of the path leftover.
	75	*/
	76	hostname_ptr = strtok_r(local_copy, "/", &leftover_ptr);
	77	if (!hostname_ptr) {
	78	ERR("Failed to parse session path \"%s\": couldn't identify hostname",
	79	path);
	80	goto error;
	81	}
	82
	83	second_token_ptr = strtok_r(NULL, "/", &leftover_ptr);
	84	if (!second_token_ptr) {
	85	ERR("Failed to parse session path \"%s\": couldn't identify session name",
	86	path);
	87	goto error;
	88	}
	89
	90	/*
	91	* Check if the second token is a base path set at url level. This is
	92	* legal in streaming, live and snapshot [1]. Otherwise it is the
	93	* session name with possibly a datetime attached [2]. Note that when
	94	* "adding" snapshot output (lttng snapshot add-output), no session name
	95	* is present in the path by default. The handling for "base path" take
	96	* care of this case as well.
	97	* [1] e.g --set-url net://localhost/my_marvellous_path
	98	* [2] Can be:
	99	* <session_name>
	100	* When using --snapshot on session create.
	101	* <session_name>-<date>-<time>
	102	* <auto>-<date>-<time>
	103	*/
	104	if (strncmp(second_token_ptr, local_session_name,
	105	strlen(local_session_name)) != 0) {
	106	/*
	107	* Token does not start with session name.
	108	* This mean this is an extra path scenario.
	109	* Duplicate the current token since it is part of an
	110	* base_path.
	111	* Set secDuplicate the current token since it is part of an
	112	* base_path. The rest is the leftover.
	113	* Set second_token_ptr to the local_session_name for further
	114	* processing.
	115	*/
	116	partial_base_path = strdup(second_token_ptr);
	117	if (!partial_base_path) {
	118	PERROR("Failed to parse session path: couldn't copy partial base path");
	119	goto error;
	120	}
	121
	122	second_token_ptr = local_session_name;
	123	}
	124
	125	/*
	126	* Based on the previous test, we can move inside the token ptr to
	127	* remove the "local_session_name" and inspect the rest of the token.
	128	* We are looking into extracting the creation datetime from either the
129	* session_name or the token. We need to to all this gymnastic because
130	* an extra path could decide to append a datetime to its first
131	* subdirectory.
132	* Possible scenario:
133	* <session_name>
134	* <session_name>-<date>-<time>
135	* <auto>-<date>-<time>
136	* <session_name>_base_path_foo_bar
137	* <session_name>-<false date>-<false-time> (via a base path)
138	*
139	* We have no way to discern from the basic scenario of:
140	* <session_name>-<date>-<time>
141	* and one done using a base path with the exact format we normally
142	* expect.
143	*
144	* e.g:
145	* lttng create my_session -U
146	* net://localhost/my_session-19910319-120000/
147	*/
148	ret = regcomp(&regex, DATETIME_REGEX, 0);
149	if (ret) {
150	ERR("Failed to parse session path: regex compilation failed with code %d", ret);
151	goto error;
152	}
153
154	leftover_second_token_ptr =
155	second_token_ptr + strlen(local_session_name);
156	len = strlen(leftover_second_token_ptr);
157	if (len == 0) {
158	/*
159	* We are either dealing with an auto session name or only the
160	* session_name. If this is a auto session name, we need to
161	* fetch the creation datetime.
162	*/
163	ret = regexec(&regex, local_session_name, 0, NULL, 0);
164	if (ret == 0) {
165	const ssize_t local_session_name_offset =
d2cb4a90	166	strlen(local_session_name) - DATETIME_STR_LEN + 1;
2a635488 JR	167
	168	assert(local_session_name_offset >= 0);
	169	datetime = strdup(local_session_name +
	170	local_session_name_offset);
	171	if (!datetime) {
	172	PERROR("Failed to parse session path: couldn't copy datetime on regex match");
	173	goto error_regex;
	174	}
d2cb4a90 JG	175	} else {
	176	datetime = zmalloc(DATETIME_STR_LEN);
	177	if (!datetime) {
	178	PERROR("Failed to allocate DATETIME string");
	179	goto error;
	180	}
	181
	182	ret = time_to_datetime_str(relay_session_creation_time,
	183	datetime, DATETIME_STR_LEN);
	184	if (ret) {
	185	/* time_to_datetime_str already logs errors. */
	186	goto error;
	187	}
2a635488	188	}
d2cb4a90	189	} else if (len == DATETIME_STR_LEN &&
2a635488 JR	190	!regexec(&regex, leftover_second_token_ptr, 0, NULL,
	191	0)) {
	192	/*
	193	* The leftover from the second token is of format
	194	* "-<datetime>", use it as the creation time.
	195	* Ignore leading "-".
	196	*/
	197	datetime = strdup(&leftover_second_token_ptr[1]);
	198	if (!datetime) {
	199	PERROR("Failed to parse session path: couldn't copy datetime on regex match");
	200	goto error_regex;
	201	}
	202	} else {
	203	/*
	204	* Base path scenario.
	205	* We cannot try to extract the datetime from the session name
	206	* since nothing prevent a user to name a session in the
	207	* "name-<datetime>" format. Using the datetime from such a
	208	* session would be invalid.
	209	* */
	210	assert(partial_base_path == NULL);
	211	assert(datetime == NULL);
	212
	213	partial_base_path = strdup(second_token_ptr);
	214	if (!partial_base_path) {
	215	PERROR("Failed to parse session path: couldn't copy partial base path");
	216	goto error_regex;
	217	}
	218	}
	219
	220	ret = asprintf(&filepath_per_session, "%s/%s%s%s/%s%s%s",
	221	local_session_name, hostname_ptr, datetime ? "-" : "",
	222	datetime ? datetime : "",
	223	partial_base_path ? partial_base_path : "",
	224	partial_base_path ? "/" : "", leftover_ptr);
	225	if (ret < 0) {
	226	filepath_per_session = NULL;
	227	goto error;
	228	}
	229	error_regex:
	230	regfree(&regex);
	231	error:
	232	free(local_copy);
	233	free(partial_base_path);
	234	free(datetime);
	235	return filepath_per_session;
	236	}