Re: [PATCH] [WIP]tools: add devname2tmpfile

[Dave Reisner <d@falconindy.com>]

On Thu, Apr 11, 2013 at 05:47:55PM +0200, Tom Gundersen wrote:
> This tool reads modules.devname from the current kernel directory and writes the information
> out in the tmpfiles format so that systemd-tmpfiles can use this to create the required device
> nodes before systemd-udevd is started on boot.

I'm a little confused as to why this should live in kmod if the output
is only useful to systemd. Rather, I would have suspected that this
would be part of src/core/kmod-setup.c in systemd. modules.devname is
easily parseable and requires no knowledge of what kmod does internally.
In fact, your code initializes a kmod context but then never uses it for
anything.

I've left some comments for you regardless of where this ends up
living...

> This makes sure nothing but kmod reads the private files under
> /usr/lib/modules/.

The code says otherwise -- it reads from /lib/modules/...

> Cc: <linux-hotplug@vger.kernel.org>
> Cc: <systemd-devel@lists.freedesktop.org>
> ---
>  Makefile.am             |   3 +-
>  tools/devname2tmpfile.c | 126 ++++++++++++++++++++++++++++++++++++++++++++++++
>  tools/kmod.c            |   1 +
>  tools/kmod.h            |   1 +
>  4 files changed, 130 insertions(+), 1 deletion(-)
>  create mode 100644 tools/devname2tmpfile.c
>
> diff --git a/Makefile.am b/Makefile.am
> index 9feaf96..50cd380 100644
> --- a/Makefile.am
> +++ b/Makefile.am
> @@ -109,7 +109,8 @@ noinst_SCRIPTS = tools/insmod tools/rmmod tools/lsmod \
>  tools_kmod_SOURCES = tools/kmod.c tools/kmod.h tools/lsmod.c \
>  		     tools/rmmod.c tools/insmod.c \
>  		     tools/modinfo.c tools/modprobe.c \
> -		     tools/depmod.c tools/log.h tools/log.c
> +		     tools/depmod.c tools/log.h tools/log.c \
> +		     tools/devname2tmpfile.c
>  tools_kmod_LDADD = libkmod/libkmod-util.la \
>  		   libkmod/libkmod.la
>
> diff --git a/tools/devname2tmpfile.c b/tools/devname2tmpfile.c
> new file mode 100644
> index 0000000..05a2b4e
> --- /dev/null
> +++ b/tools/devname2tmpfile.c
> @@ -0,0 +1,126 @@
> +/*
> + * kmod-devname2tmpfile - write devnames of current kernel in tmpfiles.d format
> + *
> + * Copyright (C) 2004-2012 Kay Sievers <kay@vrfy.org>
> + * Copyright (C) 2011-2013  ProFUSION embedded systems
> + * Copyright (C) 2013 Tom Gundersen <teg@jklm.no>
> + *
> + * This program is free software: you can redistribute it and/or modify
> + * it under the terms of the GNU General Public License as published by
> + * the Free Software Foundation, either version 2 of the License, or
> + * (at your option) any later version.
> + *
> + * This program is distributed in the hope that it will be useful,
> + * but WITHOUT ANY WARRANTY; without even the implied warranty of
> + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
> + * GNU General Public License for more details.
> + *
> + * You should have received a copy of the GNU General Public License
> + * along with this program.  If not, see <http://www.gnu.org/licenses/>.
> + */
> +
> +#include <stdio.h>
> +#include <stdlib.h>
> +#include <stddef.h>
> +#include <errno.h>
> +#include <unistd.h>
> +#include <string.h>
> +#include <limits.h>
> +#include <sys/utsname.h>
> +#include <sys/stat.h>
> +#include <sys/types.h>
> +#include "libkmod.h"
> +
> +#include "kmod.h"
> +
> +static int do_devname2tmpfile(int argc, char *argv[])
> +{
> +        struct kmod_ctx *ctx;
> +        struct utsname kernel;
> +        const char *null_config = NULL;
> +        char modules[PATH_MAX];
> +        char buf[4096];
> +        FILE *f, *out;
> +
> +	if (argc != 1) {
> +		fprintf(stderr, "Usage: %s\n", argv[0]);
> +		return EXIT_FAILURE;
> +	}
> +
> +	ctx = kmod_new(NULL, &null_config);

As mentioned above, this is never actually used, but it seems the only
thing it could be potentially used for is logging, rather than writing
directly to a stream.

> +	if (ctx = NULL) {
> +		fputs("Error: kmod_new() failed!\n", stderr);
> +		return EXIT_FAILURE;
> +	}
> +
> +        if (uname(&kernel) < 0) {
> +                fputs("Error: uname failed!\n", stderr);
> +                return EXIT_FAILURE;
> +        }
> +        snprintf(modules, sizeof(modules), "/lib/modules/%s/modules.devname", kernel.release);
> +        f = fopen(modules, "re");
> +        if (f = NULL) {
> +		fprintf(stderr, "Error: could not open %s/modules.devname!\n", kernel.release);

Full path here instead of just a trailing fragment?

> +                return EXIT_FAILURE;

I disagree that this should be an error or a failure.

> +        }
> +
> +        if (mkdir("/run/tmpfiles.d/", 755) != 0 && errno != EEXIST) {
> +		fputs("Error: /run/tmpfiles.d does not exist and could not be created!\n", stderr);
> +                return EXIT_FAILURE;
> +        }
> +
> +        out = fopen("/run/tmpfiles.d/kmod.conf", "we");
> +        if (out = NULL) {
> +		fputs("Error: could not create /run/tmpfiles.d/kmod.conf!\n", stderr);
> +                return EXIT_FAILURE;
> +        }

You appear to have some inconsistent whitespace above here in the
fprintf and fputs calls. There's more below which I haven't pointed out.

> +
> +        while (fgets(buf, sizeof(buf), f) != NULL) {
> +                char *s;
> +                const char *modname;
> +                const char *devname;
> +                const char *devno;
> +                int maj, min;

You declare these as int (signed), but then treat them as unsigned (%u)
in sscanf and fprintf.

> +                char type;
> +
> +                if (buf[0] = '#')
> +                        continue;
> +
> +                modname = buf;
> +                s = strchr(modname, ' ');
> +                if (s = NULL)
> +                        continue;
> +                s[0] = '\0';
> +
> +                devname = &s[1];
> +                s = strchr(devname, ' ');
> +                if (s = NULL)
> +                        continue;
> +                s[0] = '\0';
> +
> +                devno = &s[1];
> +                s = strchr(devno, ' ');
> +                if (s = NULL)
> +                        s = strchr(devno, '\n');
> +                if (s != NULL)
> +                        s[0] = '\0';
> +                if (sscanf(devno, "%c%u:%u", &type, &maj, &min) != 3)
> +                        continue;

This whole section is unnecessarily verbose. This should be a well
formed file where non-comment data matches the format string "%ms %ms
%c%u:%u".  You can simply do your validation on that (i.e. sscanf
returns exactly 5). Note that %ms will allocate memory for the data, so
you need to free it after printing it below. Alternatively, just use
adequately sized stack allocated char[] with %s.

> +
> +                if (type != 'c' && type != 'b')
> +                    continue;

I think this is worth logging about given that depmod currently never
writes anything but 'b' or 'c' here. I suspect this won't be changing
any time soon...

> +                fprintf(out, "%c /dev/%s 0600 - - - %u:%u\n", type, devname, maj, min);
> +        }
> +
> +        fclose(f);
> +	kmod_unref(ctx);
> +
> +        return EXIT_SUCCESS;
> +}
> +
> +const struct kmod_cmd kmod_cmd_devname2tmpfile = {
> +	.name = "devname2tmpfile",
> +	.cmd = do_devname2tmpfile,
> +	.help = "write devnames of current kernel in tmpfiles.d format",
> +};
> diff --git a/tools/kmod.c b/tools/kmod.c
> index ebb8875..ff6518c 100644
> --- a/tools/kmod.c
> +++ b/tools/kmod.c
> @@ -37,6 +37,7 @@ static const struct kmod_cmd kmod_cmd_help;
>  static const struct kmod_cmd *kmod_cmds[] = {
>  	&kmod_cmd_help,
>  	&kmod_cmd_list,
> +	&kmod_cmd_devname2tmpfile,
>  };
>
>  static const struct kmod_cmd *kmod_compat_cmds[] = {
> diff --git a/tools/kmod.h b/tools/kmod.h
> index 80fa4c2..6410ed5 100644
> --- a/tools/kmod.h
> +++ b/tools/kmod.h
> @@ -35,5 +35,6 @@ extern const struct kmod_cmd kmod_cmd_compat_modprobe;
>  extern const struct kmod_cmd kmod_cmd_compat_depmod;
>
>  extern const struct kmod_cmd kmod_cmd_list;
> +extern const struct kmod_cmd kmod_cmd_devname2tmpfile;
>
>  #include "log.h"
> --
> 1.8.2.1
>
> --
> To unsubscribe from this list: send the line "unsubscribe linux-modules" in
> the body of a message to majordomo@vger.kernel.org
> More majordomo info at  http://vger.kernel.org/majordomo-info.html