From a53faa6b9d92be769fad0d47cd7e6b7bc023fc3f Mon Sep 17 00:00:00 2001 From: Raghavendra Bhat Date: Tue, 21 Feb 2012 11:29:06 +0530 Subject: arequal: take arguments from user to ignore directories for checksum calculation With changes such as gfid based backend in glusterfs, we have to ignore some hidden directories from the backend (which creates the hidden directories for better management of the volume). So this patch takes care of taking the list of ignored directories from user as an input argument. Those directories which are present in the path given by the user (of which arequal is calculating the checksum), are not considered while crawling the path. Change-Id: If6b2209088300c5c59a8b0fb74e854f3ce030571 Signed-off-by: Raghavendra Bhat --- arequal/arequal-checksum.c | 322 +++++++++++++++++++++++++++++++++++---------- 1 file changed, 252 insertions(+), 70 deletions(-) diff --git a/arequal/arequal-checksum.c b/arequal/arequal-checksum.c index 8df7b02..107c0df 100644 --- a/arequal/arequal-checksum.c +++ b/arequal/arequal-checksum.c @@ -1,5 +1,5 @@ /* - Copyright (c) 2006-2011 Gluster, Inc. + Copyright (c) 2006-2011 Gluster, Inc. This file is part of GlusterFS. GlusterFS is free software; you can redistribute it and/or modify @@ -17,14 +17,17 @@ . */ +#ifndef _GNU_SOURCE +#define _GNU_SOURCE +#endif #define _XOPEN_SOURCE 600 +#include #include #include #include #include -#include #include #include #include @@ -34,10 +37,29 @@ #include #include #include +#include int debug = 0; +typedef struct { + char test_directory[4096]; + char **ignored_directory; + unsigned int directories_ignored; +} arequal_config_t; + +static arequal_config_t arequal_config; + +static error_t +arequal_parse_opts (int key, char *arg, struct argp_state *_state); + +static struct argp_option arequal_options[] = { + { "ignore", 'i', "IGNORED", 0, + "entry in the given path to be ignored"}, + { "path", 'p', "PATH", 0, "path where arequal has to be run"}, + {0, 0, 0, 0, 0} +}; + #define DBG(fmt ...) do { \ if (debug) { \ fprintf (stderr, "D "); \ @@ -45,12 +67,96 @@ int debug = 0; } \ } while (0) +void +add_to_list (char *arg); +void +get_absolute_path (char directory[], char *arg); static inline int roof(int a, int b) { return ((((a)+(b)-1)/((b)?(b):1))*(b)); } +void +add_to_list (char *arg) +{ + char *string = NULL; + int index = 0; + + index = arequal_config.directories_ignored - 1; + string = strdup (arg); + + if (!arequal_config.ignored_directory) { + arequal_config.ignored_directory = calloc (1, sizeof (char *)); + } else + arequal_config.ignored_directory = + realloc (arequal_config.ignored_directory, + sizeof (char *) * (index+1)); + + arequal_config.ignored_directory[index] = string; +} + +static error_t +arequal_parse_opts (int key, char *arg, struct argp_state *_state) +{ + switch (key) { + case 'i': + { + arequal_config.directories_ignored++; + add_to_list (arg); + } + break; + case 'p': + { + if (arg[0] == '/') + strcpy (arequal_config.test_directory, arg); + else + get_absolute_path (arequal_config.test_directory, arg); + + if (arequal_config.test_directory + [strlen(arequal_config.test_directory) - 1] == '/') + arequal_config.test_directory + [strlen(arequal_config.test_directory) - 1] = '\0'; + } + break; + + case ARGP_KEY_NO_ARGS: + break; + case ARGP_KEY_ARG: + break; + case ARGP_KEY_END: + if (_state->argc == 1) { + argp_usage (_state); + } + + } + + return 0; +} + +void +get_absolute_path (char directory[], char *arg) +{ + char cwd[4096] = {0,}; + + if (getcwd (cwd, sizeof (cwd)) == NULL) + printf ("some error in getting cwd\n"); + + if (strcmp (arg, ".") != 0) { + if (cwd[strlen(cwd)] != '/') + cwd[strlen (cwd)] = '/'; + strcat (cwd, arg); + } + strcpy (directory, cwd); +} + +static struct argp argp = { + arequal_options, + arequal_parse_opts, + "", + "arequal - Tool which calculates the checksum of all the entries" + "present in a given directory" +}; /* All this runs in single thread, hence using 'global' variables */ @@ -109,103 +215,103 @@ checksum_path (const char *path) return csum; } - int checksum_md5 (const char *path, const struct stat *sb) { uint64_t this_data_checksum = 0; - FILE *filep = NULL; - char *cmd = NULL; - char strvalue[17] = {0,}; - int ret = -1; - int len = 0; - const char *pos = NULL; - char *cpos = NULL; - - /* Have to escape single-quotes in filename. - * First, calculate the size of the buffer I'll need. - */ - for (pos = path; *pos; pos++) { - if ( *pos == '\'' ) - len += 4; - else - len += 1; - } - - cmd = malloc(sizeof(char) * (len + 20)); - cmd[0] = '\0'; - - /* Now, build the command with single quotes escaped. */ - - cpos = cmd; - strcpy(cpos, "md5sum '"); - cpos += 8; - - /* Add the file path, with every single quotes replaced with this sequence: - * '\'' - */ - - for (pos = path; *pos; pos++) { - if ( *pos == '\'' ) { - strcpy(cpos, "'\\''"); - cpos += 4; - } else { - *cpos = *pos; - cpos++; - } - } - - /* Add on the trailing single-quote and null-terminate. */ - strcpy(cpos, "'"); - + FILE *filep = NULL; + char *cmd = NULL; + char strvalue[17] = {0,}; + int ret = -1; + int len = 0; + const char *pos = NULL; + char *cpos = NULL; + + /* Have to escape single-quotes in filename. + * First, calculate the size of the buffer I'll need. + */ + for (pos = path; *pos; pos++) { + if ( *pos == '\'' ) + len += 4; + else + len += 1; + } + + cmd = malloc(sizeof(char) * (len + 20)); + cmd[0] = '\0'; + + /* Now, build the command with single quotes escaped. */ + + cpos = cmd; + strcpy(cpos, "md5sum '"); + cpos += 8; + + /* Add the file path, with every single quotes replaced with this sequence: + * '\'' + */ + + for (pos = path; *pos; pos++) { + if ( *pos == '\'' ) { + strcpy(cpos, "'\\''"); + cpos += 4; + } else { + *cpos = *pos; + cpos++; + } + } + + /* Add on the trailing single-quote and null-terminate. */ + strcpy(cpos, "'"); + filep = popen (cmd, "r"); if (!filep) { - perror (path); - goto out; + perror (path); + goto out; } if (fread (strvalue, sizeof (char), 16, filep) != 16) { - fprintf (stderr, "%s: short read\n", path); - goto out; - } + fprintf (stderr, "%s: short read\n", path); + goto out; + } this_data_checksum = strtoull (strvalue, NULL, 16); if (-1 == this_data_checksum) { fprintf (stderr, "%s: %s\n", strvalue, strerror (errno)); - goto out; + goto out; } checksum_file1 ^= this_data_checksum; if (fread (strvalue, sizeof (char), 16, filep) != 16) { - fprintf (stderr, "%s: short read\n", path); - goto out; - } + fprintf (stderr, "%s: short read\n", path); + goto out; + } this_data_checksum = strtoull (strvalue, NULL, 16); if (-1 == this_data_checksum) { fprintf (stderr, "%s: %s\n", strvalue, strerror (errno)); - goto out; + goto out; } checksum_file2 ^= this_data_checksum; - ret = 0; + ret = 0; out: - if (filep) - pclose (filep); - - if (cmd) - free(cmd); + if (filep) + pclose (filep); + + if (cmd) + free(cmd); return ret; } - int checksum_filenames (const char *path, const struct stat *sb) { DIR *dirp = NULL; struct dirent *entry = NULL; unsigned long long csum = 0; + int i = 0; + int found = 0; dirp = opendir (path); if (!dirp) { @@ -215,6 +321,24 @@ checksum_filenames (const char *path, const struct stat *sb) errno = 0; while ((entry = readdir (dirp))) { + /* do not calculate the checksum of the entries which user has + told to ignore and proceed to other siblings.*/ + if (arequal_config.ignored_directory) { + for (i = 0;i < arequal_config.directories_ignored;i++) { + if ((strcmp (entry->d_name, + arequal_config.ignored_directory[i]) + == 0)) { + found = 1; + DBG ("ignoring the entry %s\n", + entry->d_name); + break; + } + } + if (found == 1) { + found = 0; + continue; + } + } csum = checksum_path (entry->d_name); checksum_dir ^= csum; } @@ -321,6 +445,49 @@ process_entry (const char *path, const struct stat *sb, int typeflag, struct FTW *ftwbuf) { int ret = 0; + char *name = NULL; + char *bname = NULL; + char *dname = NULL; + int i = 0; + + /* The if condition below helps in ignoring some directories in + the given path. If the name of the entry is one of the directory + names that the user told to ignore, then that directory will not + be processed and will return FTW_SKIP_SUBTREE to nftw which will + not crawl this directory and move on to other siblings. + Note that for nftw to recognize FTW_SKIP_TREE, FTW_ACTIONRETVAL + should be passed as an argument to nftw. + + This mainly helps in calculating the checksum of network filesystems + (client-server), where the server might have some hidden directories + for managing the filesystem. So to calculate the sanity of filesytem + one has to get the checksum of the client and then the export directory + of server by telling arequal to ignore some of the directories which + are not part of the namespace. + */ + + if (arequal_config.ignored_directory) { + name = strdup (path); + + name[strlen(name)] == '\0'; + + bname = strrchr (name, '/'); + if (bname) + bname++; + + dname = dirname (name); + for ( i = 0; i < arequal_config.directories_ignored; i++) { + if ((strcmp (bname, arequal_config.ignored_directory[i]) + == 0) && (strcmp (arequal_config.test_directory, + dname) == 0)) { + DBG ("ignoring %s\n", bname); + ret = FTW_SKIP_SUBTREE; + if (name) + free (name); + return ret; + } + } + } DBG ("processing entry %s\n", path); @@ -339,6 +506,8 @@ process_entry (const char *path, const struct stat *sb, break; } + if (name) + free (name); return ret; } @@ -409,14 +578,19 @@ int main(int argc, char *argv[]) { int ret = 0; + int i = 0; - if (argc != 2) { - fprintf (stderr, "Usage: %s \n", - argv[0]); - return -1; - } + ret = argp_parse (&argp, argc, argv, 0, 0, NULL); + if (ret != 0) { + fprintf (stderr, "parsing arguments failed\n"); + return -2; + } - ret = nftw (argv[1], process_entry, 30, FTW_PHYS|FTW_MOUNT); + /* Use FTW_ACTIONRETVAL to take decision on what to do depending upon */ + /* the return value of the callback function */ + /* (process_entry in this case) */ + ret = nftw (arequal_config.test_directory, process_entry, 30, + FTW_ACTIONRETVAL|FTW_PHYS|FTW_MOUNT); if (ret != 0) { fprintf (stderr, "ftw (%s) returned %d (%s), terminating\n", argv[1], ret, strerror (errno)); @@ -425,5 +599,13 @@ main(int argc, char *argv[]) display_stats (stdout); + if (arequal_config.ignored_directory) { + for (i = 0; i < arequal_config.directories_ignored; i++) { + if (arequal_config.ignored_directory[i]) + free (arequal_config.ignored_directory[i]); + } + free (arequal_config.ignored_directory); + } + return 0; } -- cgit