Skip to contentSkip to conversation

fw_devlink: Improve cycle detection in DT

PATCH In review Saravana Kannan 3 participants 6 messages Jun 10, 2020 – Jun 18, 2020

Subsystems (2)
lore
Share & tools
lore
#0 lore raw

v2 cover fw_devlink: Improve cycle detection in DT

Patch 2/2 explain the series. Just using a cover letter to thread the series and add CC's.

-Saravana

v1 -> v2:
Patch 2/2:
- Added more comments
- Fixed missing put_device()
- Fixed stupid fall through in the error case

Saravana Kannan (2):
driver core: Add device_is_dependent() to linux/device.h
of: property: Improve cycle detection when one of the devices is never
added

 drivers/base/core.c    |  2 +-
 drivers/of/property.c  | 62 ++++++++++++++++++++++++++++++++++++++----
 include/linux/device.h |  1 +
 3 files changed, 58 insertions(+), 7 deletions(-)
Signature · 1 line
2.27.0.278.ge193c7cf3a9-goog
#1 lore raw

PATCH v2 1/2 driver core: Add device_is_dependent() to linux/device.h

DT implementation of fw_devlink needs this function to detect cycles. So make it available.

---
 drivers/base/core.c    | 2 +-
 include/linux/device.h | 1 +
 2 files changed, 2 insertions(+), 1 deletion(-)
drivers/base/core.c +1 -1
index 139cdf7e7327..509b13610b56 100644
@@ -118,7 +118,7 @@ int device_links_read_lock_held(void)
  * Check if @target depends on @dev or any device dependent on it (its child or
  * its consumer etc).  Return 1 if that is the case or 0 otherwise.
  */
-static int device_is_dependent(struct device *dev, void *target)
+int device_is_dependent(struct device *dev, void *target)
 {
 	struct device_link *link;
 	int ret;
include/linux/device.h +1 -0
index ac8e37cd716a..5c618effc51e 100644
@@ -829,6 +829,7 @@ extern int device_change_owner(struct device *dev, kuid_t kuid, kgid_t kgid);
 extern const char *device_get_devnode(struct device *dev,
 				      umode_t *mode, kuid_t *uid, kgid_t *gid,
 				      const char **tmp);
+extern int device_is_dependent(struct device *dev, void *target);
 
 static inline bool device_supports_offline(struct device *dev)
 {
Signature · 1 line
2.27.0.278.ge193c7cf3a9-goog
#2 lore raw

PATCH v2 2/2 of: property: Improve cycle detection when one of the devices is never added

In reply to Saravana Kannan

Consider this example where -> means LHS device is a consumer of RHS device and indentation represents "child of" of the previous device.

Device A -> Device C

Device B -> Device A
Device C

Without this commit:
1. Device A is added.
2. Device A is added to waiting for supplier list (Device C)
3. Device B is added
4. Device B is linked as a consumer to Device A
5. Device A doesn't probe because it's waiting for Device C to be added.
6. Device B doesn't probe because Device A hasn't probed.
7. Device C will never be added because it's parent hasn't probed.

So, Device A, B and C will be in a probe/add deadlock.

This commit detects this scenario and stops trying to create a device link between Device A and Device C since doing so would create the following cycle:
Device A -> Devic C -(parent)-> Device B -> Device A.

With this commit:
1. Device A is added.
3. Device B is added
4. Device B is linked as a consumer to Device A
5. Device A probes.
6. Device B probes because Device A has probed.
7. Device C is added and probed.

---
 drivers/of/property.c | 62 ++++++++++++++++++++++++++++++++++++++-----
 1 file changed, 56 insertions(+), 6 deletions(-)
drivers/of/property.c +56 -6
index 1f2086f4e7ce..ef09e4372ce8 100644
@@ -1014,6 +1014,30 @@ static bool of_is_ancestor_of(struct device_node *test_ancestor,
 	return false;
 }
 
+/**
+ * of_get_next_parent_dev - Add device link to supplier from supplier phandle
+ * @np: device tree node
+ *
+ * Given a device tree node (@np), this function finds its closest ancestor
+ * device tree node that has a corresponding struct device.
+ *
+ * The caller of this function is expected to call put_device() on the returned
+ * device when they are done.
+ */
+static struct device *of_get_next_parent_dev(struct device_node *np)
+{
+	struct device *dev = NULL;
+
+	of_node_get(np);
+	do {
+		np = of_get_next_parent(np);
+		if (np)
+			dev = get_dev_from_fwnode(&np->fwnode);
+	} while (np && !dev);
+	of_node_put(np);
+	return dev;
+}
+
 /**
  * of_link_to_phandle - Add device link to supplier from supplier phandle
  * @dev: consumer device
@@ -1035,10 +1059,9 @@ static bool of_is_ancestor_of(struct device_node *test_ancestor,
 static int of_link_to_phandle(struct device *dev, struct device_node *sup_np,
 			      u32 dl_flags)
 {
-	struct device *sup_dev;
+	struct device *sup_dev, *sup_par_dev;
 	int ret = 0;
 	struct device_node *tmp_np = sup_np;
-	int is_populated;
 
 	of_node_get(sup_np);
 	/*
@@ -1075,16 +1098,43 @@ static int of_link_to_phandle(struct device *dev, struct device_node *sup_np,
 		return -EINVAL;
 	}
 	sup_dev = get_dev_from_fwnode(&sup_np->fwnode);
-	is_populated = of_node_check_flag(sup_np, OF_POPULATED);
-	of_node_put(sup_np);
-	if (!sup_dev && is_populated) {
+	if (!sup_dev && of_node_check_flag(sup_np, OF_POPULATED)) {
 		/* Early device without struct device. */
 		dev_dbg(dev, "Not linking to %pOFP - No struct device\n",
 			sup_np);
+		of_node_put(sup_np);
 		return -ENODEV;
 	} else if (!sup_dev) {
-		return -EAGAIN;
+		/*
+		 * DL_FLAG_SYNC_STATE_ONLY doesn't block probing and supports
+		 * cycles. So cycle detection isn't necessary and shouldn't be
+		 * done.
+		 */
+		if (dl_flags & DL_FLAG_SYNC_STATE_ONLY) {
+			of_node_put(sup_np);
+			return -EAGAIN;
+		}
+
+		sup_par_dev = of_get_next_parent_dev(sup_np);
+
+		if (sup_par_dev && device_is_dependent(dev, sup_par_dev)) {
+			/* Cyclic dependency detected, don't try to link */
+			dev_dbg(dev, "Not linking to %pOFP - cycle detected\n",
+				sup_np);
+			ret = -EINVAL;
+		} else {
+			/*
+			 * Can't check for cycles or no cycles. So let's try
+			 * again later.
+			 */
+			ret = -EAGAIN;
+		}
+
+		of_node_put(sup_np);
+		put_device(sup_par_dev);
+		return ret;
 	}
+	of_node_put(sup_np);
 	if (!device_link_add(dev, sup_dev, dl_flags))
 		ret = -EINVAL;
 	put_device(sup_dev);
Signature · 1 line
2.27.0.278.ge193c7cf3a9-goog
maintainer
#3 lore raw

Re: [PATCH v2 2/2] of: property: Improve cycle detection when one of the devices is never added

Quoted 32 lines · Saravana Kannan

Consider this example where -> means LHS device is a consumer of RHS device and indentation represents "child of" of the previous device.

Device A -> Device C

Device B -> Device A
Device C

Without this commit:
1. Device A is added.
2. Device A is added to waiting for supplier list (Device C)
3. Device B is added
4. Device B is linked as a consumer to Device A
5. Device A doesn't probe because it's waiting for Device C to be added.
6. Device B doesn't probe because Device A hasn't probed.
7. Device C will never be added because it's parent hasn't probed.

So, Device A, B and C will be in a probe/add deadlock.

This commit detects this scenario and stops trying to create a device link between Device A and Device C since doing so would create the following cycle:
Device A -> Devic C -(parent)-> Device B -> Device A.

With this commit:
1. Device A is added.
3. Device B is added
4. Device B is linked as a consumer to Device A
5. Device A probes.
6. Device B probes because Device A has probed.
7. Device C is added and probed.

Quoted 5 lines
---
 drivers/of/property.c | 62 ++++++++++++++++++++++++++++++++++++++-----
 1 file changed, 56 insertions(+), 6 deletions(-)

Both patches applied.

Rob

#4 lore raw

Re: [PATCH v2 2/2] of: property: Improve cycle detection when one of the devices is never added

Quoted 34 lines · Rob Herring
Quoted 32 lines · Saravana Kannan

Consider this example where -> means LHS device is a consumer of RHS device and indentation represents "child of" of the previous device.

Device A -> Device C

Device B -> Device A
Device C

Without this commit:
1. Device A is added.
2. Device A is added to waiting for supplier list (Device C)
3. Device B is added
4. Device B is linked as a consumer to Device A
5. Device A doesn't probe because it's waiting for Device C to be added.
6. Device B doesn't probe because Device A hasn't probed.
7. Device C will never be added because it's parent hasn't probed.

So, Device A, B and C will be in a probe/add deadlock.

This commit detects this scenario and stops trying to create a device link between Device A and Device C since doing so would create the following cycle:
Device A -> Devic C -(parent)-> Device B -> Device A.

With this commit:
1. Device A is added.
3. Device B is added
4. Device B is linked as a consumer to Device A
5. Device A probes.
6. Device B probes because Device A has probed.
7. Device C is added and probed.

Quoted 7 lines
Quoted 5 lines
---
 drivers/of/property.c | 62 ++++++++++++++++++++++++++++++++++++++-----
 1 file changed, 56 insertions(+), 6 deletions(-)

Both patches applied.

Thanks!

-Saravana

Tested-by on the v2 series
#5 lore raw
In reply to Saravana Kannan
Quoted 17 lines · Saravana Kannan

Patch 2/2 explain the series. Just using a cover letter to thread the series and add CC's.

-Saravana

v1 -> v2:
Patch 2/2:
- Added more comments
- Fixed missing put_device()
- Fixed stupid fall through in the error case

Saravana Kannan (2):
driver core: Add device_is_dependent() to linux/device.h
of: property: Improve cycle detection when one of the devices is never
added

Quoted 4 lines
 drivers/base/core.c    |  2 +-
 drivers/of/property.c  | 62 ++++++++++++++++++++++++++++++++++++++----
 include/linux/device.h |  1 +
 3 files changed, 58 insertions(+), 7 deletions(-)

With both patches, booting with fw_devlink=on (instead of deferred_probe_timeout=30) this allows all modules to properly load on the db845c, and without these patches, we fail to get display.

thanks
-john