diff mbox series

[v3,1/3] driver core: fw_devlink: Improve handling of cyclic dependencies

Message ID 20210915170940.617415-2-saravanak@google.com (mailing list archive)
State Not Applicable, archived
Headers show
Series fw_devlink bug fixes | expand

Commit Message

Saravana Kannan Sept. 15, 2021, 5:09 p.m. UTC
When we have a dependency of the form:

Device-A -> Device-C
	Device-B

Device-C -> Device-B

Where,
* Indentation denotes "child of" parent in previous line.
* X -> Y denotes X is consumer of Y based on firmware (Eg: DT).

We have cyclic dependency: device-A -> device-C -> device-B -> device-A

fw_devlink current treats device-C -> device-B dependency as an invalid
dependency and doesn't enforce it but leaves the rest of the
dependencies as is.

While the current behavior is necessary, it is not sufficient if the
false dependency in this example is actually device-A -> device-C. When
this is the case, device-C will correctly probe defer waiting for
device-B to be added, but device-A will be incorrectly probe deferred by
fw_devlink waiting on device-C to probe successfully. Due to this, none
of the devices in the cycle will end up probing.

To fix this, we need to go relax all the dependencies in the cycle like
we already do in the other instances where fw_devlink detects cycles.
A real world example of this was reported[1] and analyzed[2].

[1] - https://lore.kernel.org/lkml/0a2c4106-7f48-2bb5-048e-8c001a7c3fda@samsung.com/
[2] - https://lore.kernel.org/lkml/CAGETcx8peaew90SWiux=TyvuGgvTQOmO4BFALz7aj0Za5QdNFQ@mail.gmail.com/
Fixes: f9aa460672c9 ("driver core: Refactor fw_devlink feature")
Reported-by: Marek Szyprowski <m.szyprowski@samsung.com>
Signed-off-by: Saravana Kannan <saravanak@google.com>
Tested-by: Marek Szyprowski <m.szyprowski@samsung.com>
---
 drivers/base/core.c | 17 ++++++++++++-----
 1 file changed, 12 insertions(+), 5 deletions(-)

Comments

Rafael J. Wysocki Sept. 18, 2021, 2:38 p.m. UTC | #1
On Wed, Sep 15, 2021 at 7:09 PM Saravana Kannan <saravanak@google.com> wrote:
>
> When we have a dependency of the form:
>
> Device-A -> Device-C
>         Device-B
>
> Device-C -> Device-B
>
> Where,
> * Indentation denotes "child of" parent in previous line.
> * X -> Y denotes X is consumer of Y based on firmware (Eg: DT).
>
> We have cyclic dependency: device-A -> device-C -> device-B -> device-A
>
> fw_devlink current treats device-C -> device-B dependency as an invalid
> dependency and doesn't enforce it but leaves the rest of the
> dependencies as is.
>
> While the current behavior is necessary, it is not sufficient if the
> false dependency in this example is actually device-A -> device-C. When
> this is the case, device-C will correctly probe defer waiting for
> device-B to be added, but device-A will be incorrectly probe deferred by
> fw_devlink waiting on device-C to probe successfully. Due to this, none
> of the devices in the cycle will end up probing.
>
> To fix this, we need to go relax all the dependencies in the cycle like
> we already do in the other instances where fw_devlink detects cycles.
> A real world example of this was reported[1] and analyzed[2].
>
> [1] - https://lore.kernel.org/lkml/0a2c4106-7f48-2bb5-048e-8c001a7c3fda@samsung.com/
> [2] - https://lore.kernel.org/lkml/CAGETcx8peaew90SWiux=TyvuGgvTQOmO4BFALz7aj0Za5QdNFQ@mail.gmail.com/
> Fixes: f9aa460672c9 ("driver core: Refactor fw_devlink feature")
> Reported-by: Marek Szyprowski <m.szyprowski@samsung.com>
> Signed-off-by: Saravana Kannan <saravanak@google.com>
> Tested-by: Marek Szyprowski <m.szyprowski@samsung.com>
> ---
>  drivers/base/core.c | 17 ++++++++++++-----
>  1 file changed, 12 insertions(+), 5 deletions(-)
>
> diff --git a/drivers/base/core.c b/drivers/base/core.c
> index e65dd803a453..316df6027093 100644
> --- a/drivers/base/core.c
> +++ b/drivers/base/core.c
> @@ -1772,14 +1772,21 @@ static int fw_devlink_create_devlink(struct device *con,
>          * be broken by applying logic. Check for these types of cycles and
>          * break them so that devices in the cycle probe properly.
>          *
> -        * If the supplier's parent is dependent on the consumer, then
> -        * the consumer-supplier dependency is a false dependency. So,
> -        * treat it as an invalid link.
> +        * If the supplier's parent is dependent on the consumer, then the
> +        * consumer and supplier have a cyclic dependency. Since fw_devlink
> +        * can't tell which of the inferred dependencies are incorrect, don't
> +        * enforce probe ordering between any of the devices in this cyclic
> +        * dependency. Do this by relaxing all the fw_devlink device links in
> +        * this cycle and by treating the fwnode link between the consumer and
> +        * the supplier as an invalid dependency.
>          */
>         sup_dev = fwnode_get_next_parent_dev(sup_handle);
>         if (sup_dev && device_is_dependent(con, sup_dev)) {
> -               dev_dbg(con, "Not linking to %pfwP - False link\n",
> -                       sup_handle);
> +               dev_info(con, "Fixing up cyclic dependency with %pfwP (%s)\n",
> +                        sup_handle, dev_name(sup_dev));

Why not dev_dbg()?

Other than this, the change makes sense to me.

> +               device_links_write_lock();
> +               fw_devlink_relax_cycle(con, sup_dev);
> +               device_links_write_unlock();
>                 ret = -EINVAL;
>         } else {
>                 /*
> --
> 2.33.0.309.g3052b89438-goog
>
diff mbox series

Patch

diff --git a/drivers/base/core.c b/drivers/base/core.c
index e65dd803a453..316df6027093 100644
--- a/drivers/base/core.c
+++ b/drivers/base/core.c
@@ -1772,14 +1772,21 @@  static int fw_devlink_create_devlink(struct device *con,
 	 * be broken by applying logic. Check for these types of cycles and
 	 * break them so that devices in the cycle probe properly.
 	 *
-	 * If the supplier's parent is dependent on the consumer, then
-	 * the consumer-supplier dependency is a false dependency. So,
-	 * treat it as an invalid link.
+	 * If the supplier's parent is dependent on the consumer, then the
+	 * consumer and supplier have a cyclic dependency. Since fw_devlink
+	 * can't tell which of the inferred dependencies are incorrect, don't
+	 * enforce probe ordering between any of the devices in this cyclic
+	 * dependency. Do this by relaxing all the fw_devlink device links in
+	 * this cycle and by treating the fwnode link between the consumer and
+	 * the supplier as an invalid dependency.
 	 */
 	sup_dev = fwnode_get_next_parent_dev(sup_handle);
 	if (sup_dev && device_is_dependent(con, sup_dev)) {
-		dev_dbg(con, "Not linking to %pfwP - False link\n",
-			sup_handle);
+		dev_info(con, "Fixing up cyclic dependency with %pfwP (%s)\n",
+			 sup_handle, dev_name(sup_dev));
+		device_links_write_lock();
+		fw_devlink_relax_cycle(con, sup_dev);
+		device_links_write_unlock();
 		ret = -EINVAL;
 	} else {
 		/*