Thread (57 messages) 57 messages, 6 authors, 2021-09-30
STALE1738d

[PATCH v1 1/2] driver core: fw_devlink: Add support for FWNODE_FLAG_BROKEN_PARENT

From: Saravana Kannan <hidden>
Date: 2021-08-26 07:45:42
Also in: linux-acpi, lkml
Subsystem: acpi, driver core, kobjects, debugfs and sysfs, software nodes and device properties, the rest · Maintainers: "Rafael J. Wysocki", Greg Kroah-Hartman, Danilo Krummrich, Linus Torvalds

If a parent device is also a supplier to a child device, fw_devlink=on
(correctly) delays the probe() of the child device until the probe() of
the parent finishes successfully.

However, some drivers of such parent devices (where parent is also a
supplier) incorrectly expect the child device to finish probing
successfully as soon as they are added using device_add() and before the
probe() of the parent device has completed successfully. While this
might have worked before, this is not guaranteed by driver core.
fw_devlink=on catches/breaks such drivers. One example of such a case is
discussed in the link mentioned below.

Add a flag to make fw_devlink=on not enforce these supplier-consumer
relationships, so these drivers can continue working. The flag is
intentionally called BROKEN_PARENT so it's clear that this flag
shouldn't be used in the normal case and that there's a problem with the
driver.

Link: https://lore.kernel.org/netdev/CAGETcx_uj0V4DChME-gy5HGKTYnxLBX=TH2rag29f_p=UcG+Tg@mail.gmail.com/ (local)
Signed-off-by: Saravana Kannan <redacted>
---
 drivers/base/core.c    | 22 ++++++++++++++++++++++
 include/linux/fwnode.h |  3 +++
 2 files changed, 25 insertions(+)
diff --git a/drivers/base/core.c b/drivers/base/core.c
index f6360490a4a3..2cc34f8ff051 100644
--- a/drivers/base/core.c
+++ b/drivers/base/core.c
@@ -1719,6 +1719,28 @@ static int fw_devlink_create_devlink(struct device *con,
 	struct device *sup_dev;
 	int ret = 0;
 
+	/*
+	 * In some cases, a device P might also be a supplier to its child node
+	 * C. However, this would defer the probe of C until the probe of P
+	 * completes successfully. This is perfectly fine in the device driver
+	 * model. device_add() doesn't guarantee probe completion of the device
+	 * by the time it returns.
+	 *
+	 * However, there are a few drivers that assume C will finish probing
+	 * as soon as it's added and before P finishes probing. While this is a
+	 * broken assumption that needs the driver to be fixed, we don't want
+	 * to block fw_devlink improvements because of these drivers.
+	 *
+	 * So, we provide a flag to let fw_devlink know not to delay the probe
+	 * of C until the probe of P completes successfully.
+	 *
+	 * When such a flag is set, we can't create device links with P as the
+	 * supplier of C as that would delay the probe of C.
+	 */
+	if (sup_handle->flags & FWNODE_FLAG_BROKEN_PARENT &&
+	    fwnode_is_ancestor_of(sup_handle, con->fwnode))
+		return -EINVAL;
+
 	sup_dev = get_dev_from_fwnode(sup_handle);
 	if (sup_dev) {
 		/*
diff --git a/include/linux/fwnode.h b/include/linux/fwnode.h
index 59828516ebaf..9382065e6ff8 100644
--- a/include/linux/fwnode.h
+++ b/include/linux/fwnode.h
@@ -22,10 +22,13 @@ struct device;
  * LINKS_ADDED:	The fwnode has already be parsed to add fwnode links.
  * NOT_DEVICE:	The fwnode will never be populated as a struct device.
  * INITIALIZED: The hardware corresponding to fwnode has been initialized.
+ * BROKEN_PARENT: The driver of this fwnode/device expects the child devices to
+ *		  probe as soon as they are added.
  */
 #define FWNODE_FLAG_LINKS_ADDED		BIT(0)
 #define FWNODE_FLAG_NOT_DEVICE		BIT(1)
 #define FWNODE_FLAG_INITIALIZED		BIT(2)
+#define FWNODE_FLAG_BROKEN_PARENT	BIT(3)
 
 struct fwnode_handle {
 	struct fwnode_handle *secondary;
-- 
2.33.0.rc2.250.ged5fa647cd-goog
Keyboard shortcuts
hback out one level
jnext message in thread
kprevious message in thread
ldrill in
Escclose help / fold thread tree
?toggle this help