[Toybox] [New Toy] - modprobe
ibid.ag at gmail.com
ibid.ag at gmail.com
Tue Dec 17 15:33:32 PST 2013
On Tue, Dec 17, 2013 at 04:39:39PM +0530, Ashwini Sharma wrote:
> Hi List,
>
> An implementation for modprobe is attached.
> go thru the same and let me know your inputs.
>
> regards,
> Ashwini
For comments see below.
No big issues with style that I can see, except that it should be
default n until Rob's looked over it.
At 575 lines, it's largeish, but it seems reasonable to me...
except perhaps for use of modules.dep:
http://lists.landley.net/pipermail/toybox-landley.net/2013-August/001294.html
http://lists.landley.net/pipermail/toybox-landley.net/2013-August/001297.html
> /* modprobe.c - modprobe utility.
> *
> * Copyright 2012 Madhur Verma <mad.flexi at gmail.com>
> * Copyright 2013 Kyungwan Han <asura321 at gmail.com>
> *
> * No Standard.
>
> USE_MODPROBE(NEWTOY(modprobe, "alrqvsDb", TOYFLAG_SBIN))
>
> config MODPROBE
> bool "modprobe"
> default y
> help
> usage: modprobe [-alrqvsDb] MODULE [symbol=value][...]
>
> modprobe utility - inserts modules and dependencies.
>
> -a Load multiple MODULEs
> -l List (MODULE is a pattern)
> -r Remove MODULE (stacks) or do autoclean
> -q Quiet
> -v Verbose
> -s Log to syslog
> -D Show dependencies
> -b Apply blacklist to module names too
> */
> #define FOR_modprobe
> #include "toys.h"
> #include <sys/syscall.h>
> #include <fnmatch.h>
>
> GLOBALS(
> struct arg_list *probes;
> struct arg_list *dbase[256];
> char *cmdopts;
> int nudeps;
> uint8_t symreq;
> )
>
> /* Note: if "#define DBASE_SIZE" modified,
> * Please update GLOBALS dbase[256] accordingly.
> */
> #define DBASE_SIZE 256
> #define MODNAME_LEN 256
Any reason not to use sizeof(TT.dbase)?
>
> // Modules flag definations
> #define MOD_ALOADED 0x0001
> #define MOD_BLACKLIST 0x0002
> #define MOD_FNDDEPMOD 0x0004
> #define MOD_NDDEPS 0x0008
I can see using the macros, but I seem to remember Rob mentioning somewhere
before that defining things just before you need them can clarify the
code a little.
That would mean moving them just above config_action().
> static void (*dbg)(char *format, ...);
> // dummy interface for debugging.
> static void dummy(char *format, ...)
> {
> }
>
> // Current probing modules info
> struct module_s {
> uint32_t flags;
> char *cmdname, *name, *depent, *opts;
> struct arg_list *rnames, *dep;
> };
>
> // Converts path name FILE to module name.
> static char *path2mod(char *file, char *mod)
> {
> int i;
> char *from, *lslash;
>
> if (!file) return NULL;
> if (!mod) mod = xmalloc(MODNAME_LEN);
>
> lslash = strrchr(file, '/');
> if (!lslash || (lslash == file && !lslash[1])) from = file;
> else from = lslash + 1;
A home-made basename()?
> for (i = 0; i < (MODNAME_LEN-1) && from[i] && from[i] != '.'; i++)
> mod[i] = (from[i] == '-') ? '_' : from[i];
> mod[i] = '\0';
> return mod;
> }
I presume you are only using this part for display.
modprobe dm_crypt or dm-crypt will load dm-crypt.ko
modprobe micheal-mic or micheal_mic will load micheal_mic.ko
Whatever you do, either gets displayed with an '_'
> // locate character in string.
> static char *strchr_nul(char *s, int c)
> {
> while(*s != '\0' && *s != c) s++;
> return (char*)s;
> }
I'm guessing this is because strchrnul is a non-standard glibc
extension?
I don't know about uclibc or newlib (if the latter is relevant),
but musl does include it.
$ nm /opt/musl/lib/libc.a |grep strchrnul
U __strchrnul
U __strchrnul
strchrnul.o:
00000000 T __strchrnul
00000000 W strchrnul
U __strchrnul
> // Add options in opts from toadd.
> static char *add_opts(char *opts, char *toadd)
> {
> if (toadd) {
> int optlen = 0;
>
> if (opts) optlen = strlen(opts);
> opts = xrealloc(opts, optlen + strlen(toadd) + 2);
> sprintf(opts + optlen, " %s", toadd);
> }
> return opts;
> }
>
> // Remove first element from the list and return it.
> static void *llist_popme(struct arg_list **head)
> {
> char *data = NULL;
> struct arg_list *temp = *head;
>
> if (temp) {
> data = temp->arg;
> *head = temp->next;
> free(temp);
> }
> return data;
> }
>
> // Add new node at the beginning of the list.
> static void llist_add(struct arg_list **old, void *data)
> {
> struct arg_list *new = xmalloc(sizeof(struct arg_list));
>
> new->arg = (char*)data;
> new->next = *old;
> *old = new;
> }
>
> // Add new node at tail of list.
> static void llist_add_tail(struct arg_list **head, void *data)
> {
> while (*head) head = &(*head)->next;
> *head = xzalloc(sizeof(struct arg_list));
> (*head)->arg = (char*)data;
> }
>
> // Reverse list order.
> static struct arg_list *llist_rev(struct arg_list *list)
> {
> struct arg_list *rev = NULL;
>
> while (list) {
> struct arg_list *next = list->next;
>
> list->next = rev;
> rev = list;
> list = next;
> }
> return rev;
> }
>
> /*
> * Returns struct module_s from the data base if found, NULL otherwise.
> * if ps - create module entry, add it to data base and return the same mod.
> */
> static struct module_s *get_mod(char *mod, uint8_t ps)
What on earth does "ps" indicate?
"add" would make the function much clearer.
> {
> char name[MODNAME_LEN];
> struct module_s *modentry;
> struct arg_list *temp;
> unsigned i, hash = 0;
>
> path2mod(mod, name);
> for (i = 0; name[i]; i++) hash = ((hash*31) + hash) + name[i];
> hash %= DBASE_SIZE;
> for (temp = TT.dbase[hash]; temp; temp = temp->next) {
> modentry = (struct module_s *) temp->arg;
> if (!strcmp(modentry->name, name)) return modentry;
> }
> if (!ps) return NULL;
> modentry = xzalloc(sizeof(*modentry));
> modentry->name = xstrdup(name);
> llist_add(&TT.dbase[hash], modentry);
> return modentry;
> }
>
> /*
> * Read a line from file with \ continuation and escape commented line.
> * Return the line in allocated string (*li)
> */
> static int read_line(FILE *fl, char **li)
> {
> char *nxtline = NULL, *line;
> int len, nxtlen, linelen, nxtlinelen;
>
> while (1) {
> line = NULL;
> linelen = nxtlinelen = 0;
> len = getline(&line, (size_t*)&linelen, fl);
> if (len <= 0) return len;
> // checking for commented lines.
> if (line[0] != '#') break;
> free(line);
> }
> for (;;) {
> if (line[len - 1] == '\n') len--;
> // checking line continuation.
> if (!len || line[len - 1] != '\\') break;
> len--;
> nxtlen = getline(&nxtline, (size_t*)&nxtlinelen, fl);
EINVAL is the only documented error for getline(), but I'd expect it to
also fail if ENOMEM (since getline uses malloc()/realloc()).
Which suggests that (not currently existent) xgetline() should be used.
> if (nxtlen <= 0) break;
> if (linelen < len + nxtlen + 1) {
> linelen = len + nxtlen + 1;
> line = xrealloc(line, linelen);
> }
> memcpy(&line[len], nxtline, nxtlen);
> len += nxtlen;
> }
> line[len] = '\0';
> *li = xstrdup(line);
> if (line) free(line);
> if (nxtline) free(nxtline);
> return len;
> }
>
> /*
> * Action to be taken on all config files in default directories
> * checks for aliases, options, install, remove and blacklist
> */
> static int config_action(struct dirtree *node)
> {
> FILE *fc;
> char *filename, *tokens[3], *line, *linecp;
> struct module_s *modent;
> int tcount = 0;
>
> if (!dirtree_notdotdot(node)) return 0;
> if (S_ISDIR(node->st.st_mode)) return DIRTREE_RECURSE;
>
> if (!S_ISREG(node->st.st_mode)) return 0; // process only regular file
> filename = dirtree_path(node, NULL);
> if (!(fc = fopen(filename, "r"))) {
> free(filename);
> return 0;
> }
> for (line = linecp = NULL; read_line(fc, &line) > 0;
> free(line), free(linecp), line = linecp = NULL) {
> char *tk = NULL;
>
> if (!strlen(line)) continue;
> linecp = xstrdup(line);
> for (tk = strtok(linecp, "# \t"), tcount = 0; tk;
> tk = strtok(NULL, "# \t"), tcount++) {
> tokens[tcount] = tk;
> if (tcount == 2) {
> tokens[2] = line + strlen(tokens[0]) + strlen(tokens[1]) + 2;
> break;
> }
> }
> if (!tk) continue;
> // process the tokens[0] contains first word of config line.
> if (!strcmp(tokens[0], "alias")) {
> struct arg_list *temp;
> char aliase[MODNAME_LEN], *realname;
>
> if (!tokens[2]) continue;
> path2mod(tokens[1], aliase);
> for (temp = TT.probes; temp; temp = temp->next) {
> modent = (struct module_s *) temp->arg;
> if (fnmatch(aliase, modent->name, 0)) continue;
> realname = path2mod(tokens[2], NULL);
> llist_add(&modent->rnames, realname);
> if (modent->flags & MOD_NDDEPS) {
> modent->flags &= ~MOD_NDDEPS;
> TT.nudeps--;
> }
> modent = get_mod(realname, 1);
> if (!(modent->flags & MOD_NDDEPS)) {
> modent->flags |= MOD_NDDEPS;
> TT.nudeps++;
> }
> }
> } else if (!strcmp(tokens[0], "options")) {
> if (!tokens[2]) continue;
> modent = get_mod(tokens[1], 1);
> modent->opts = add_opts(modent->opts, tokens[2]);
> } else if (!strcmp(tokens[0], "include"))
> dirtree_read(tokens[1], config_action);
> else if (!strcmp(tokens[0], "blacklist"))
> get_mod(tokens[1], 1)->flags |= MOD_BLACKLIST;
> else if (!strcmp(tokens[0], "install")) continue;
> else if (!strcmp(tokens[0], "remove")) continue;
> else error_msg("Invalid option %s found in file %s", tokens[0], filename);
> }
> fclose(fc);
> free(filename);
> return 0;
> }
>
> // Show matched modules else return -1 on failure.
> static int depmode_read_entry(char *cmdname)
> {
> char *line;
> int ret = -1;
> FILE *fe = xfopen("modules.dep", "r");
>
> while (read_line(fe, &line) > 0) {
> char *tmp = strchr(line, ':');
>
> if (tmp) {
> *tmp = '\0';
> char *name = basename(line);
>
> tmp = strchr(name, '.');
> if (tmp) *tmp = '\0';
> if (!cmdname) {
> if (tmp) *tmp = '.';
> xprintf("%s\n", line);
> ret = 0;
> } else if (!fnmatch(cmdname, name, 0)) {
> if (tmp) *tmp = '.';
> xprintf("%s\n", line);
> ret = 0;
> }
> }
> free(line);
> }
> return ret;
> }
>
> // Finds dependencies for modules from the modules.dep file.
See comment at the top about using modules.dep.
I suppose if the goal were to keep from duplicating the guts of nm and
running that for every module, that might be a sensible reason...
> static void find_dep(void)
> {
> char *line = NULL;
> struct module_s *mod;
> FILE *fe = xfopen("modules.dep", "r");
>
> for (; read_line(fe, &line) > 0; free(line)) {
> char *tmp = strchr(line, ':');
>
> if (tmp) {
> *tmp = '\0';
> mod = get_mod(line, 0);
> if (!mod) continue;
> if ((mod->flags & MOD_ALOADED) &&
> !(toys.optflags & (FLAG_r | FLAG_D))) continue;
>
> mod->flags |= MOD_FNDDEPMOD;
> if ((mod->flags & MOD_NDDEPS) && (!mod->dep)) {
> TT.nudeps--;
> llist_add(&mod->dep, xstrdup(line));
> tmp++;
> if (*tmp) {
> char *tok;
>
> while ((tok = strsep(&tmp, " \t"))) {
> if (!*tok) continue;
> llist_add_tail(&mod->dep, xstrdup(tok));
> }
> }
> }
> }
> }
> fclose(fe);
> }
>
> // Remove a module from the Linux Kernel. if !modules does auto remove.
> static int rm_mod(char *modules, uint32_t flags)
> {
> errno = 0;
> if (modules) {
> int len = strlen(modules);
>
> if (len > 3 && !strcmp(&modules[len-3], ".ko" )) modules[len-3] = 0;
> }
> if (!flags) flags = O_NONBLOCK|O_EXCL;
> syscall(__NR_delete_module, modules, flags);
> return errno;
> }
>
> // Insert module same as insmod implementation.
If it's the same as insmod, it should be shared with insmod.
That said, I can see an initial version duplicating it.
> static int ins_mod(char *modules, char *flags)
> {
> char *buf = NULL;
> int len, res;
> int fd = xopen(modules, O_RDONLY);
>
> len = fdlength(fd);
> buf = xmalloc(len);
> xreadall(fd, buf, len);
> xclose(fd);
>
> while (flags && strlen(toybuf) + strlen(flags) + 2 < sizeof(toybuf)) {
> strcat(toybuf, flags);
> strcat(toybuf, " ");
> }
> res = syscall(__NR_init_module, buf, len, toybuf);
> if (CFG_TOYBOX_FREE && buf != toybuf) free(buf);
> if (res) perror_exit("failed to load %s ", toys.optargs[0]);
> return res;
> }
>
> // Add module in probes list, if not loaded.
> static void add_mod(char *name)
> {
> struct module_s *mod = get_mod(name, 1);
>
> if (!(toys.optflags & (FLAG_r | FLAG_D)) && (mod->flags & MOD_ALOADED)) {
> dbg("skipping %s, it is already loaded\n", name);
> return;
> }
> dbg("queuing %s\n", name);
> mod->cmdname = name;
> mod->flags |= MOD_NDDEPS;
> llist_add_tail(&TT.probes, mod);
> TT.nudeps++;
> if (!strncmp(mod->name, "symbol:", 7)) TT.symreq = 1;
> }
>
> // Parse cmdline options suplied for module.
> static char *add_cmdopt(char **argv)
> {
> char *opt = xzalloc(1);
> int lopt = 0;
>
> while (*++argv) {
> char *fmt, *var, *val;
>
> var = *argv;
> opt = xrealloc(opt, lopt + 2 + strlen(var) + 2);
> // check for key=val or key = val.
> fmt = "%.*s%s ";
> val = strchr_nul(var, '=');
> if (*val) {
> val++;
> if (strchr(val, ' ')) fmt = "%.*s\"%s\" ";
> }
> lopt += sprintf(opt + lopt, fmt, (int) (val - var), var, val);
> }
> return opt;
> }
>
> // Probes a single module and loads all its dependencies.
> static int go_probe(struct module_s *m)
> {
> int rc = 0, first = 1;
>
> if (!(m->flags & MOD_FNDDEPMOD)) {
> if (!(toys.optflags & FLAG_s))
> error_msg("module %s not found in modules.dep", m->name);
> return -ENOENT;
> }
> dbg("go_prob'ing %s\n", m->name);
> if (!(toys.optflags & FLAG_r)) m->dep = llist_rev(m->dep);
>
> while (m->dep) {
> struct module_s *m2;
> char *fn, *options;
>
> rc = 0;
> fn = llist_popme(&m->dep);
> m2 = get_mod(fn, 1);
> // are we removing ?
> if (toys.optflags & FLAG_r) {
> if (m2->flags & MOD_ALOADED) {
> if ((rc = rm_mod(m2->name, O_EXCL))) {
> if (first) {
> perror_msg("can't unload module %s", m2->name);
> break;
> }
> } else m2->flags &= ~MOD_ALOADED;
> }
> first = 0;
> continue;
> }
> options = m2->opts;
> m2->opts = NULL;
> if (m == m2) options = add_opts(options, TT.cmdopts);
>
> // are we only checking dependencies ?
> if (toys.optflags & FLAG_D) {
> dbg(options ? "insmod %s %s\n" : "insmod %s\n", fn, options);
> if (options) free(options);
> continue;
> }
> if (m2->flags & MOD_ALOADED) {
> dbg("%s is already loaded, skipping\n", fn);
> if (options) free(options);
> continue;
> }
> // none of above is true insert the module.
> rc = ins_mod(fn, options);
> dbg("loaded %s '%s', rc:%d\n", fn, options, rc);
> if (rc == EEXIST) rc = 0;
> if (options) free(options);
> if (rc) {
> perror_msg("can't load module %s (%s)", m2->name, fn);
> break;
> }
> m2->flags |= MOD_ALOADED;
> }
> return rc;
> }
>
> void modprobe_main(void)
> {
> struct utsname uts;
> char **argv = toys.optargs, *procline = NULL;
> FILE *fs;
> struct module_s *module;
> unsigned flags = toys.optflags;
>
> dbg = dummy;
> if (flags & FLAG_v) dbg = xprintf;
>
> if ((toys.optc < 1) && (((flags & FLAG_r) && (flags & FLAG_l))
> ||(!((flags & FLAG_r)||(flags & FLAG_l))))) {
> toys.exithelp++;
> error_exit(" Syntex Error.");
> }
> // Check for -r flag without arg if yes then do auto remove.
> if ((flags & FLAG_r) && (!toys.optc)) {
> if (rm_mod(NULL, O_NONBLOCK | O_EXCL) != 0) perror_exit("rmmod");
> return;
> }
>
> // change directory to /lib/modules/<release>/
> xchdir("/lib/modules");
> uname(&uts);
> xchdir(uts.release);
>
> // modules.dep processing for dependency check.
> if (flags & FLAG_l) {
> if (depmode_read_entry(toys.optargs[0])) error_exit("no module found.");
> return;
> }
> // Read /proc/modules to get loadded modules.
> fs = xfopen("/proc/modules", "r");
>
> while (read_line(fs, &procline) > 0) {
> *(strchr(procline, ' ')) = '\0';
> get_mod(procline, 1)->flags = MOD_ALOADED;
> free(procline);
> procline = NULL;
> }
> fclose(fs);
> if ((flags & FLAG_a) || (flags & FLAG_r)) {
> do {
> add_mod(*argv++);
> } while (*argv);
> } else {
> add_mod(argv[0]);
> TT.cmdopts = add_cmdopt(argv);
> }
> if (!TT.probes) {
> fprintf(stderr, "All modules loaded successfully. \n");
> return;
> }
> dirtree_read("/etc/modprobe.conf", config_action);
> dirtree_read("/etc/modprobe.d", config_action);
> if (TT.symreq) dirtree_read("modules.symbols", config_action);
> if (TT.nudeps) dirtree_read("modules.alias", config_action);
> find_dep();
> while ((module = llist_popme(&TT.probes))) {
> if (!module->rnames) {
> dbg("probing by module name\n");
> /* This is not an alias. Literal names are blacklisted
> * only if '-b' is given.
> */
> if (!(flags & FLAG_b) || !(module->flags & MOD_BLACKLIST))
> go_probe(module);
> continue;
> }
> do { // Probe all real names for the alias.
> char *real = llist_pop(&module->rnames);
> struct module_s *m2 = get_mod(real, 0);
>
> dbg("probing alias %s by realname %s\n", module->name, real);
> if (!m2) continue;
> if (!(m2->flags & MOD_BLACKLIST)
> && (!(m2->flags & MOD_ALOADED) || (flags & (FLAG_r | FLAG_D))))
> go_probe(m2);
> free(real);
> } while (module->rnames);
> }
> }
> _______________________________________________
> Toybox mailing list
> Toybox at lists.landley.net
> http://lists.landley.net/listinfo.cgi/toybox-landley.net
More information about the Toybox
mailing list