lists.openwall.net   lists  /  announce  owl-users  owl-dev  john-users  john-dev  passwdqc-users  yescrypt  popa3d-users  /  oss-security  kernel-hardening  musl  sabotage  tlsify  passwords  /  crypt-dev  xvendor  /  Bugtraq  Full-Disclosure  linux-kernel  linux-netdev  linux-ext4  linux-hardening  linux-cve-announce  PHC 
Open Source and information security mailing list archives
 
Hash Suite: Windows password security audit tool. GUI, reports in PDF.
[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]
Message-ID: <1385129826.1747.22.camel@leonhard>
Date:	Fri, 22 Nov 2013 23:17:06 +0900
From:	Namhyung Kim <namhyung@...nel.org>
To:	Jiri Olsa <jolsa@...hat.com>
Cc:	linux-kernel@...r.kernel.org,
	Corey Ashford <cjashfor@...ux.vnet.ibm.com>,
	Frederic Weisbecker <fweisbec@...il.com>,
	Ingo Molnar <mingo@...e.hu>, Paul Mackerras <paulus@...ba.org>,
	Peter Zijlstra <a.p.zijlstra@...llo.nl>,
	Arnaldo Carvalho de Melo <acme@...hat.com>,
	Steven Rostedt <rostedt@...dmis.org>,
	David Ahern <dsahern@...il.com>
Subject: Re: [PATCH 01/22] tools lib traceevent: Add plugin support

Hi Jiri,

2013-11-21 (목), 12:00 +0100, Jiri Olsa:
> Backporting plugin support for traceevent lib.
> 
> Backported from Steven Rostedt's trace-cmd repo (HEAD 0f2c2fb):
> git://git.kernel.org/pub/scm/linux/kernel/git/rostedt/trace-cmd.git
> 
> It's now possible to use following interface to load plugins
> (shared objects) to enahnce pevent object functionality.
> 
> The plugin interface/hooks are as follows:
> (taken from event-parse.h comments)
> 
> - 'pevent_plugin_loader' (required)
>     The function name to initialized the plugin.
> 
>     int pevent_plugin_loader(struct pevent *pevent)
> 
> - 'pevent_plugin_unloader' (optional)
>     The function called just before unloading
> 
>     int pevent_plugin_unloader(void)
> 
> - 'pevent_plugin_options'  (optional)
>     Plugin options that can be set before loading
> 
>     struct plugin_option pevent_plugin_options[] = {
>        {
>                .name = "option-name",
>                .plugin_alias = "overide-file-name", (optional)
>                .description = "description of option to show users",
>        },
>        {
>                .name = NULL,
>        },
>     };
> 
>     Array must end with .name = NULL;
> 
>     The plugin_alias (below) can be used to give a shorter
>     name to access the vairable. Useful if a plugin handles
>     more than one event.
> 
>     NOTE options support is not backported yet.
> 
> - 'pevent_plugin_alias' (optional)
>     The name to use for finding options (uses filename if not defined)
> 
> New traceevent functions are added to search and load
> available plugins:
> 
>   struct plugin_list*
>   traceevent_load_plugins(struct pevent *pevent)
>     - loads plusing for 'struct pevent' object and returns
>       loaded plugins list
> 
>   void traceevent_unload_plugins(struct plugin_list *plugin_list);
>     - unload plugin list

First of all, thank you for your nice work! :)

> 
[SNIP[
> +static void
> +load_plugin(struct pevent *pevent, const char *path,
> +	    const char *file, void *data)
> +{
> +	struct plugin_list **plugin_list = data;
> +	pevent_plugin_load_func func;
> +	struct plugin_list *list;
> +	const char *alias;
> +	char *plugin;
> +	void *handle;
> +
> +	plugin = malloc_or_die(strlen(path) + strlen(file) + 2);

I'd like not to see this malloc_or_die() anymore in a new code.  Just
returning after showing a warning looks enough here.

> +
> +	strcpy(plugin, path);
> +	strcat(plugin, "/");
> +	strcat(plugin, file);
> +
> +	handle = dlopen(plugin, RTLD_NOW | RTLD_GLOBAL);

Why RTLD_NOW and RTLD_GLOBAL?  Hmm.. maybe using _NOW is needed to
prevent a runtime error, but not sure why _GLOBAL is needed.

IIUC _GLOBAL is for exporting symbols to *other libraries*.  Is it
intended for this plugin support?


> +	if (!handle) {
> +		warning("could not load plugin '%s'\n%s\n",
> +			plugin, dlerror());
> +		goto out_free;
> +	}
> +
> +	alias = dlsym(handle, PEVENT_PLUGIN_ALIAS_NAME);
> +	if (!alias)
> +		alias = file;

So this 'alias' is not used anywhere in the current code, right?
Do you plan to add the option processing soon?

> +
> +	func = dlsym(handle, PEVENT_PLUGIN_LOADER_NAME);
> +	if (!func) {
> +		warning("could not find func '%s' in plugin '%s'\n%s\n",
> +			PEVENT_PLUGIN_LOADER_NAME, plugin, dlerror());
> +		goto out_free;
> +	}
> +
> +	list = malloc_or_die(sizeof(*list));

Ditto.  Please check return value and handle error properly.


> +	list->next = *plugin_list;
> +	list->handle = handle;
> +	list->name = plugin;
> +	*plugin_list = list;
> +
> +	pr_stat("registering plugin: %s", plugin);
> +	func(pevent);
> +	return;
> +
> + out_free:
> +	free(plugin);
> +}
> +
> +static void
> +load_plugins_dir(struct pevent *pevent, const char *suffix,
> +		 const char *path,
> +		 void (*load_plugin)(struct pevent *pevent,
> +				     const char *path,
> +				     const char *name,
> +				     void *data),

Doesn't gcc complain about the name shadows other function?  Anyway I
think we should avoid using same name - probably using typedef for the
function might be helpful.


> +		 void *data)
[SNIP]
> +	/*
> +	 * Now let the home directory override the environment
> +	 * or system defaults.
> +	 */
> +	home = getenv("HOME");
> +	if (!home)
> +		return;
> +
> +	path = malloc_or_die(strlen(home) + strlen(LOCAL_PLUGIN_DIR) + 2);

Ditto.

Thanks,
Namhyung

> +
> +	strcpy(path, home);
> +	strcat(path, "/");
> +	strcat(path, LOCAL_PLUGIN_DIR);
> +
> +	load_plugins_dir(pevent, suffix, path, load_plugin, data);
> +
> +	free(path);
> +}
> +
> +struct plugin_list*
> +traceevent_load_plugins(struct pevent *pevent)
> +{
> +	struct plugin_list *list = NULL;
> +
> +	load_plugins(pevent, ".so", load_plugin, &list);
> +	return list;
> +}
> +
> +void
> +traceevent_unload_plugins(struct plugin_list *plugin_list)
> +{
> +	pevent_plugin_unload_func func;
> +	struct plugin_list *list;
> +
> +	while (plugin_list) {
> +		list = plugin_list;
> +		plugin_list = list->next;
> +		func = dlsym(list->handle, PEVENT_PLUGIN_UNLOADER_NAME);
> +		if (func)
> +			func();
> +		dlclose(list->handle);
> +		free(list->name);
> +		free(list);
> +	}
> +}



--
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

Powered by Openwall GNU/*/Linux Powered by OpenVZ