public inbox for netdev@vger.kernel.org
 help / color / mirror / Atom feed
From: Jiri Pirko <jiri@resnulli.us>
To: Jakub Kicinski <kuba@kernel.org>
Cc: jacob.e.keller@intel.com, leon@kernel.org, netdev@vger.kernel.org
Subject: Re: [RFC net-next 04/10] devlink: always check if the devlink instance is registered
Date: Mon, 2 Jan 2023 15:57:24 +0100	[thread overview]
Message-ID: <Y7Lw1GSGml1E8SXw@nanopsycho> (raw)
In-Reply-To: <20221217011953.152487-5-kuba@kernel.org>

Sat, Dec 17, 2022 at 02:19:47AM CET, kuba@kernel.org wrote:
>Always check under the instance lock whether the devlink instance
>is still / already registered.
>
>This is a no-op for the most part, as the unregistration path currently
>waits for all references. On the init path, however, we may temporarily
>open up a race with netdev code, if netdevs are registered before the
>devlink instance. This is temporary, the next change fixes it, and this
>commit has been split out for the ease of review.
>
>Note that in case of iterating over sub-objects which have their
>own lock (regions and line cards) we assume an implicit dependency
>between those objects existing and devlink unregistration.
>
>Signed-off-by: Jakub Kicinski <kuba@kernel.org>
>---
> include/net/devlink.h |  1 +
> net/devlink/basic.c   | 35 +++++++++++++++++++++++++++++------
> net/devlink/core.c    | 25 +++++++++++++++++++++----
> net/devlink/netlink.c | 10 ++++++++--
> 4 files changed, 59 insertions(+), 12 deletions(-)
>
>diff --git a/include/net/devlink.h b/include/net/devlink.h
>index 6a2e4f21779f..36e013d3aa52 100644
>--- a/include/net/devlink.h
>+++ b/include/net/devlink.h
>@@ -1626,6 +1626,7 @@ struct device *devlink_to_dev(const struct devlink *devlink);
> void devl_lock(struct devlink *devlink);
> int devl_trylock(struct devlink *devlink);
> void devl_unlock(struct devlink *devlink);
>+bool devl_is_alive(struct devlink *devlink);
> void devl_assert_locked(struct devlink *devlink);
> bool devl_lock_is_held(struct devlink *devlink);
> 
>diff --git a/net/devlink/basic.c b/net/devlink/basic.c
>index 5f33d74eef83..6b18e70a39fd 100644
>--- a/net/devlink/basic.c
>+++ b/net/devlink/basic.c
>@@ -2130,6 +2130,9 @@ static int devlink_nl_cmd_linecard_get_dumpit(struct sk_buff *msg,
> 		int idx = 0;
> 
> 		mutex_lock(&devlink->linecards_lock);
>+		if (!devl_is_alive(devlink))
>+			goto next_devlink;


Thinking about this a bit more, things would be cleaner if reporters and
linecards are converted to rely on instance lock as well. I don't see a
good reason for a separate lock in both cases, really.

Also, we could introduce devlinks_xa_for_each_registered_get_lock()
iterator that would lock the instance as well right away to avoid
this devl_is_alive() dance on multiple places when you iterate devlinks.


>+
> 		list_for_each_entry(linecard, &devlink->linecard_list, list) {
> 			if (idx < dump->idx) {
> 				idx++;
>@@ -2151,6 +2154,7 @@ static int devlink_nl_cmd_linecard_get_dumpit(struct sk_buff *msg,
> 			}
> 			idx++;
> 		}
>+next_devlink:
> 		mutex_unlock(&devlink->linecards_lock);
> 		devlink_put(devlink);
> 	}
>@@ -7809,6 +7813,12 @@ devlink_nl_cmd_health_reporter_get_dumpit(struct sk_buff *msg,
> 		int idx = 0;
> 
> 		mutex_lock(&devlink->reporters_lock);
>+		if (!devl_is_alive(devlink)) {
>+			mutex_unlock(&devlink->reporters_lock);
>+			devlink_put(devlink);
>+			continue;
>+		}
>+
> 		list_for_each_entry(reporter, &devlink->reporter_list,
> 				    list) {
> 			if (idx < dump->idx) {
>@@ -7830,6 +7840,9 @@ devlink_nl_cmd_health_reporter_get_dumpit(struct sk_buff *msg,
> 		mutex_unlock(&devlink->reporters_lock);
> 
> 		devl_lock(devlink);
>+		if (!devl_is_alive(devlink))
>+			goto next_devlink;
>+
> 		xa_for_each(&devlink->ports, port_index, port) {
> 			mutex_lock(&port->reporters_lock);
> 			list_for_each_entry(reporter, &port->reporter_list, list) {
>@@ -7853,6 +7866,7 @@ devlink_nl_cmd_health_reporter_get_dumpit(struct sk_buff *msg,
> 			}
> 			mutex_unlock(&port->reporters_lock);
> 		}
>+next_devlink:
> 		devl_unlock(devlink);
> 		devlink_put(devlink);
> 	}
>@@ -12218,7 +12232,8 @@ void devlink_compat_running_version(struct devlink *devlink,
> 		return;
> 
> 	devl_lock(devlink);

How about to have a helper, something like devl_lock_alive() (or
devl_lock_registered() with the naming scheme I suggest in the other
thread)? Then you can do:

	if (!devl_lock_alive(devlink))
		return;
	__devlink_compat_running_version(devlink, buf, len);
	devl_unlock(devlink);



>-	__devlink_compat_running_version(devlink, buf, len);
>+	if (devl_is_alive(devlink))
>+		__devlink_compat_running_version(devlink, buf, len);
> 	devl_unlock(devlink);
> }
> 
>@@ -12227,20 +12242,28 @@ int devlink_compat_flash_update(struct devlink *devlink, const char *file_name)
> 	struct devlink_flash_update_params params = {};
> 	int ret;
> 
>-	if (!devlink->ops->flash_update)
>-		return -EOPNOTSUPP;
>+	devl_lock(devlink);
>+	if (!devl_is_alive(devlink)) {
>+		ret = -ENODEV;
>+		goto out_unlock;
>+	}
>+
>+	if (!devlink->ops->flash_update) {
>+		ret = -EOPNOTSUPP;
>+		goto out_unlock;
>+	}
> 
> 	ret = request_firmware(&params.fw, file_name, devlink->dev);
> 	if (ret)
>-		return ret;
>+		goto out_unlock;
> 
>-	devl_lock(devlink);
> 	devlink_flash_update_begin_notify(devlink);
> 	ret = devlink->ops->flash_update(devlink, &params, NULL);
> 	devlink_flash_update_end_notify(devlink);
>-	devl_unlock(devlink);
> 
> 	release_firmware(params.fw);
>+out_unlock:
>+	devl_unlock(devlink);
> 
> 	return ret;
> }
>diff --git a/net/devlink/core.c b/net/devlink/core.c
>index d3b8336946fd..2abad8247597 100644
>--- a/net/devlink/core.c
>+++ b/net/devlink/core.c
>@@ -67,6 +67,21 @@ void devl_unlock(struct devlink *devlink)
> }
> EXPORT_SYMBOL_GPL(devl_unlock);
> 
>+bool devl_is_alive(struct devlink *devlink)
>+{
>+	return xa_get_mark(&devlinks, devlink->index, DEVLINK_REGISTERED);
>+}
>+EXPORT_SYMBOL_GPL(devl_is_alive);
>+
>+/**
>+ * devlink_try_get() - try to obtain a reference on a devlink instance
>+ * @devlink: instance to reference
>+ *
>+ * Obtain a reference on a devlink instance. A reference on a devlink instance
>+ * only implies that it's safe to take the instance lock. It does not imply
>+ * that the instance is registered, use devl_is_alive() after taking
>+ * the instance lock to check registration status.
>+ */
> struct devlink *__must_check devlink_try_get(struct devlink *devlink)
> {
> 	if (refcount_inc_not_zero(&devlink->refcount))
>@@ -300,10 +315,12 @@ static void __net_exit devlink_pernet_pre_exit(struct net *net)
> 	devlinks_xa_for_each_registered_get(net, index, devlink) {
> 		WARN_ON(!(devlink->features & DEVLINK_F_RELOAD));
> 		devl_lock(devlink);
>-		err = devlink_reload(devlink, &init_net,
>-				     DEVLINK_RELOAD_ACTION_DRIVER_REINIT,
>-				     DEVLINK_RELOAD_LIMIT_UNSPEC,
>-				     &actions_performed, NULL);
>+		err = 0;
>+		if (devl_is_alive(devlink))
>+			err = devlink_reload(devlink, &init_net,
>+					     DEVLINK_RELOAD_ACTION_DRIVER_REINIT,
>+					     DEVLINK_RELOAD_LIMIT_UNSPEC,
>+					     &actions_performed, NULL);
> 		devl_unlock(devlink);
> 		devlink_put(devlink);
> 
>diff --git a/net/devlink/netlink.c b/net/devlink/netlink.c
>index b38df704be1c..773efaabb6ad 100644
>--- a/net/devlink/netlink.c
>+++ b/net/devlink/netlink.c
>@@ -98,7 +98,8 @@ devlink_get_from_attrs_lock(struct net *net, struct nlattr **attrs)
> 
> 	devlinks_xa_for_each_registered_get(net, index, devlink) {
> 		devl_lock(devlink);
>-		if (strcmp(devlink->dev->bus->name, busname) == 0 &&
>+		if (devl_is_alive(devlink) &&
>+		    strcmp(devlink->dev->bus->name, busname) == 0 &&
> 		    strcmp(dev_name(devlink->dev), devname) == 0)
> 			return devlink;
> 		devl_unlock(devlink);
>@@ -210,7 +211,12 @@ int devlink_instance_iter_dump(struct sk_buff *msg, struct netlink_callback *cb)
> 
> 	devlink_dump_for_each_instance_get(msg, dump, devlink) {
> 		devl_lock(devlink);
>-		err = cmd->dump_one(msg, devlink, cb);
>+
>+		if (devl_is_alive(devlink))
>+			err = cmd->dump_one(msg, devlink, cb);
>+		else
>+			err = 0;
>+
> 		devl_unlock(devlink);
> 		devlink_put(devlink);
> 
>-- 
>2.38.1
>

  parent reply	other threads:[~2023-01-02 14:57 UTC|newest]

Thread overview: 50+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2022-12-17  1:19 [RFC net-next 00/10] devlink: remove the wait-for-references on unregister Jakub Kicinski
2022-12-17  1:19 ` [RFC net-next 01/10] devlink: bump the instance index directly when iterating Jakub Kicinski
2023-01-02 13:24   ` Jiri Pirko
2023-01-02 22:48     ` Jakub Kicinski
2023-01-03  7:35       ` Jiri Pirko
2023-01-04  2:31         ` Jakub Kicinski
2023-01-02 22:56     ` Jakub Kicinski
2022-12-17  1:19 ` [RFC net-next 02/10] devlink: update the code in netns move to latest helpers Jakub Kicinski
2023-01-02 13:45   ` Jiri Pirko
2022-12-17  1:19 ` [RFC net-next 03/10] devlink: protect devlink->dev by the instance lock Jakub Kicinski
2022-12-17  1:19 ` [RFC net-next 04/10] devlink: always check if the devlink instance is registered Jakub Kicinski
2022-12-19 17:48   ` Jacob Keller
2022-12-19 21:55     ` Jakub Kicinski
2022-12-19 22:08       ` Jacob Keller
2023-01-02 13:58   ` Jiri Pirko
2023-01-02 23:05     ` Jakub Kicinski
2023-01-03  9:26       ` Jiri Pirko
2023-01-04  2:49         ` Jakub Kicinski
2023-01-04 16:14           ` Jiri Pirko
2023-01-02 14:57   ` Jiri Pirko [this message]
2023-01-02 15:12     ` Jiri Pirko
2023-01-02 23:16     ` Jakub Kicinski
2023-01-03  9:30       ` Jiri Pirko
2023-01-03 12:26   ` Jiri Pirko
2023-01-04  2:50     ` Jakub Kicinski
2022-12-17  1:19 ` [RFC net-next 05/10] devlink: remove the registration guarantee of references Jakub Kicinski
2022-12-19 17:56   ` Jacob Keller
2022-12-19 22:02     ` Jakub Kicinski
2022-12-19 22:14       ` Jacob Keller
2022-12-19 22:31         ` Jakub Kicinski
2023-01-02 14:18       ` Jiri Pirko
2023-01-02 14:32   ` Jiri Pirko
2023-01-02 23:18     ` Jakub Kicinski
2022-12-17  1:19 ` [RFC net-next 06/10] devlink: don't require setting features before registration Jakub Kicinski
2023-01-02 15:25   ` Jiri Pirko
2023-01-02 23:24     ` Jakub Kicinski
2023-01-02 23:32       ` Jakub Kicinski
2023-01-03  9:46         ` Jiri Pirko
2022-12-17  1:19 ` [RFC net-next 07/10] netdevsim: rename a label Jakub Kicinski
2022-12-19 18:01   ` Jacob Keller
2022-12-17  1:19 ` [RFC net-next 08/10] netdevsim: move devlink registration under the instance lock Jakub Kicinski
2022-12-17  1:19 ` [RFC net-next 09/10] devlink: allow registering parameters after the instance Jakub Kicinski
2022-12-17  1:19 ` [RFC net-next 10/10] netdevsim: register devlink instance before sub-objects Jakub Kicinski
2023-01-02 13:34   ` Jiri Pirko
2023-01-02 23:25     ` Jakub Kicinski
2023-01-03  9:51       ` Jiri Pirko
2023-01-04  2:52         ` Jakub Kicinski
2022-12-19 17:38 ` [RFC net-next 00/10] devlink: remove the wait-for-references on unregister Jacob Keller
2022-12-19 22:10   ` Jakub Kicinski
2022-12-19 22:16     ` Jacob Keller

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=Y7Lw1GSGml1E8SXw@nanopsycho \
    --to=jiri@resnulli.us \
    --cc=jacob.e.keller@intel.com \
    --cc=kuba@kernel.org \
    --cc=leon@kernel.org \
    --cc=netdev@vger.kernel.org \
    /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