dm-devel.redhat.com archive mirror
 help / color / mirror / Atom feed
From: mwilck@suse.com
To: Christophe Varoqui <christophe.varoqui@opensvc.com>,
	Benjamin Marzinski <bmarzins@redhat.com>
Cc: dm-devel@redhat.com, Martin Wilck <mwilck@suse.com>
Subject: [PATCH v2 17/21] libmultipath: add udev and logsink symbols
Date: Thu, 24 Sep 2020 15:37:12 +0200	[thread overview]
Message-ID: <20200924133716.14120-18-mwilck@suse.com> (raw)
In-Reply-To: <20200924133716.14120-1-mwilck@suse.com>

From: Martin Wilck <mwilck@suse.com>

With these symbols added, applications using libmultipath don't
need to define global variables "udev" and "logsink" any more.
This comes at the cost of having to call an init function.
Currently, libmultipath_init() does nothing but initialize
"udev".

The linker's symbol lookup order still allows applications to use
their own "logsink" and "udev" variables, which will take precendence
over libmultipath's internal ones. In this case, calling
libmultipath_init() can be skipped, but like before,
udev should be initialized (using udev_new()) before making any
libmultipath calls.

Signed-off-by: Martin Wilck <mwilck@suse.com>
---
 libmultipath/config.c             | 46 +++++++++++++++++++++++++++++++
 libmultipath/config.h             | 46 ++++++++++++++++++++++++++++++-
 libmultipath/debug.c              |  2 ++
 libmultipath/libmultipath.version |  8 ++++++
 4 files changed, 101 insertions(+), 1 deletion(-)

diff --git a/libmultipath/config.c b/libmultipath/config.c
index 01b77df..fbb66b3 100644
--- a/libmultipath/config.c
+++ b/libmultipath/config.c
@@ -27,6 +27,52 @@
 #include "mpath_cmd.h"
 #include "propsel.h"
 
+/*
+ * We don't support re-initialization after
+ * libmultipath_exit().
+ */
+static bool libmultipath_exit_called;
+static pthread_once_t _init_once = PTHREAD_ONCE_INIT;
+static pthread_once_t _exit_once = PTHREAD_ONCE_INIT;
+struct udev *udev;
+
+static void _udev_init(void)
+{
+	if (udev)
+		udev_ref(udev);
+	else
+		udev = udev_new();
+	if (!udev)
+		condlog(0, "%s: failed to initialize udev", __func__);
+}
+
+static void _libmultipath_init(void)
+{
+	_udev_init();
+}
+
+static bool _is_libmultipath_initialized(void)
+{
+	return !libmultipath_exit_called && !!udev;
+}
+
+int libmultipath_init(void)
+{
+	pthread_once(&_init_once, _libmultipath_init);
+	return !_is_libmultipath_initialized();
+}
+
+static void _libmultipath_exit(void)
+{
+	libmultipath_exit_called = true;
+	udev_unref(udev);
+}
+
+void libmultipath_exit(void)
+{
+	pthread_once(&_exit_once, _libmultipath_exit);
+}
+
 static struct config __internal_config;
 struct config *libmp_get_multipath_config(void)
 {
diff --git a/libmultipath/config.h b/libmultipath/config.h
index 5997b71..dac4e8f 100644
--- a/libmultipath/config.h
+++ b/libmultipath/config.h
@@ -232,7 +232,51 @@ struct config {
 	char *enable_foreign;
 };
 
-extern struct udev * udev;
+/**
+ * extern variable: udev
+ *
+ * A &struct udev instance used by libmultipath. libmultipath expects
+ * a valid, initialized &struct udev in this variable.
+ * An application can define this variable itself, in which case
+ * the applications's instance will take precedence.
+ * The application can initialize and destroy this variable by
+ * calling libmultipath_init() and libmultipath_exit(), respectively,
+ * whether or not it defines the variable itself.
+ * An application can initialize udev with udev_new() before calling
+ * libmultipath_init(), e.g. if it has to make libudev calls before
+ * libmultipath calls. If an application wants to keep using the
+ * udev variable after calling libmultipath_exit(), it should have taken
+ * an additional reference on it beforehand. This is the case e.g.
+ * after initiazing udev with udev_new().
+ */
+extern struct udev *udev;
+
+/**
+ * libmultipath_init() - library initialization
+ *
+ * This function initializes libmultipath data structures.
+ * It is light-weight; some other initializations, like device-mapper
+ * initialization, are done lazily when the respective functionality
+ * is required.
+ *
+ * Clean up by libmultipath_exit() when the program terminates.
+ * It is an error to call libmultipath_init() after libmultipath_exit().
+ * Return: 0 on success, 1 on failure.
+ */
+int libmultipath_init(void);
+
+/**
+ * libmultipath_exit() - library un-initialization
+ *
+ * This function un-initializes libmultipath data structures.
+ * It is recommended to call this function at program exit.
+ *
+ * Calls to libmultipath_init() after libmultipath_exit() will fail
+ * (in other words, libmultipath can't be re-initialized).
+ * Any other libmultipath calls after libmultipath_exit() may cause
+ * undefined behavior.
+ */
+void libmultipath_exit(void);
 
 int find_hwe (const struct _vector *hwtable,
 	      const char * vendor, const char * product, const char *revision,
diff --git a/libmultipath/debug.c b/libmultipath/debug.c
index 4128cb9..b3a1de9 100644
--- a/libmultipath/debug.c
+++ b/libmultipath/debug.c
@@ -15,6 +15,8 @@
 #include "defaults.h"
 #include "debug.h"
 
+int logsink;
+
 void dlog (int sink, int prio, const char * fmt, ...)
 {
 	va_list ap;
diff --git a/libmultipath/libmultipath.version b/libmultipath/libmultipath.version
index 81bcc9d..2e531ef 100644
--- a/libmultipath/libmultipath.version
+++ b/libmultipath/libmultipath.version
@@ -228,3 +228,11 @@ global:
 	init_config;
 	uninit_config;
 } LIBMULTIPATH_0.8.4.2;
+
+LIBMULTIPATH_0.8.4.4 {
+global:
+	udev;
+	logsink;
+	libmultipath_init;
+	libmultipath_exit;
+} LIBMULTIPATH_0.8.4.3;
-- 
2.28.0

  parent reply	other threads:[~2020-09-24 13:37 UTC|newest]

Thread overview: 35+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2020-09-24 13:36 [PATCH v2 00/21] multipath-tools: shutdown, libdevmapper races, globals mwilck
2020-09-24 13:36 ` [PATCH v2 01/21] multipathd: allow shutdown during configure() mwilck
2020-09-24 13:36 ` [PATCH v2 02/21] multipathd: avoid sending "READY=1" to systemd on early shutdown mwilck
2020-09-24 13:36 ` [PATCH v2 03/21] multipathd: send "STOPPING=1" to systemd on shutdown mwilck
2020-09-24 13:36 ` [PATCH v2 04/21] multipathd: send "RELOADING=1" to systemd on DAEMON_CONFIGURE state mwilck
2020-09-24 13:37 ` [PATCH v2 05/21] multipathd: use volatile qualifier for running_state mwilck
2020-09-24 13:37 ` [PATCH v2 06/21] multipathd: generalize and fix wait_for_state_change_if() mwilck
2020-09-24 13:37 ` [PATCH v2 07/21] multipathd: set_config_state(): avoid code duplication mwilck
2020-09-24 13:37 ` [PATCH v2 08/21] multipathd: cancel threads early during shutdown mwilck
2020-09-24 13:37 ` [PATCH v2 09/21] multipath-tools: don't call dm_lib_release() any more mwilck
2020-09-24 13:37 ` [PATCH v2 10/21] libmultipath: devmapper: refactor libdm version determination mwilck
2020-09-25 23:48   ` Benjamin Marzinski
2020-09-24 13:37 ` [PATCH v2 11/21] libmultipath: protect racy libdevmapper calls with a mutex mwilck
2020-09-29  5:57   ` Benjamin Marzinski
2020-09-29  9:09     ` Martin Wilck
2020-09-24 13:37 ` [PATCH v2 12/21] libmultipath: constify file argument in config parser mwilck
2020-09-24 13:37 ` [PATCH v2 13/21] libmultipath: provide defaults for {get, put}_multipath_config mwilck
2020-09-25  4:34   ` Benjamin Marzinski
2020-09-25 20:00     ` Martin Wilck
2020-09-25 21:57       ` Benjamin Marzinski
2020-09-26  9:43         ` Martin Wilck
2020-09-27  1:03           ` Benjamin Marzinski
2020-09-24 13:37 ` [PATCH v2 14/21] libmpathpersist: allow using libmultipath " mwilck
2020-09-24 13:37 ` [PATCH v2 15/21] multipath: use {get_put}_multipath_config from libmultipath mwilck
2020-09-24 13:37 ` [PATCH v2 16/21] mpathpersist: use {get, put}_multipath_config() " mwilck
2020-09-24 13:37 ` mwilck [this message]
2020-09-25 23:03   ` [PATCH v2 17/21] libmultipath: add udev and logsink symbols Benjamin Marzinski
2020-09-25 23:05     ` Martin Wilck
2020-09-24 13:37 ` [PATCH v2 18/21] multipath: remove logsink and udev mwilck
2020-09-24 13:37 ` [PATCH v2 19/21] libmpathpersist: call libmultipath_{init, exit}() mwilck
2020-09-25 23:55   ` Benjamin Marzinski
2020-09-24 13:37 ` [PATCH v2 20/21] mpathpersist: remove logsink and udev mwilck
2020-09-25 23:56   ` Benjamin Marzinski
2020-09-24 13:37 ` [PATCH v2 21/21] multipathd: " mwilck
2020-09-26  0:19   ` Benjamin Marzinski

Reply instructions:

You may reply publicly to this message via plain-text email
using any one of the following methods:

* Save the following mbox file, import it into your mail client,
  and reply-to-all from there: mbox

  Avoid top-posting and favor interleaved quoting:
  https://en.wikipedia.org/wiki/Posting_style#Interleaved_style

* Reply using the --to, --cc, and --in-reply-to
  switches of git-send-email(1):

  git send-email \
    --in-reply-to=20200924133716.14120-18-mwilck@suse.com \
    --to=mwilck@suse.com \
    --cc=bmarzins@redhat.com \
    --cc=christophe.varoqui@opensvc.com \
    --cc=dm-devel@redhat.com \
    /path/to/YOUR_REPLY

  https://kernel.org/pub/software/scm/git/docs/git-send-email.html

* If your mail client supports setting the In-Reply-To header
  via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for NNTP newsgroup(s).