Skip to contentSkip to conversation

pci cleanup/prep patches

PATCH Applied Keith Busch 4 participants 14 messages Oct 22, 2024 – Nov 11, 2024 PCI

lore
Share & tools
lore
#0 lore raw

v3 cover pci cleanup/prep patches

From: Keith Busch <[email protected]>

This is a subset of a previous RFC bus lock patches that are simply good cleanups that should help make it easier to introduce different locking later.

Changes from v2:

Rebased to pci/next

Added memory barriers around bit ops for patch 1.

Added reviews.

Keith Busch (5):
pci: make pci_stop_dev concurrent safe
pci: make pci_destroy_dev concurrent safe
pci: move the walk bus lock to where its needed
pci: walk bus recursively
pci: unexport pci_walk_bus_locked

 drivers/pci/bus.c    | 49 +++++++++++++++-----------------------------
 drivers/pci/pci.h    | 17 +++++++++++++--
 drivers/pci/remove.c | 22 +++++++++-----------
 3 files changed, 41 insertions(+), 47 deletions(-)
Signature · 1 line
2.43.5
#1 lore raw

PATCH v3 1/5 pci: make pci_stop_dev concurrent safe

From: Keith Busch <[email protected]>

Use the atomic ADDED flag to safely ensure concurrent callers can't attempt to stop the device multiple times. Callers should currently all be holding the pci_rescan_remove_lock, so there shouldn't be an existing race. But that global lock can cause lock dependency issues, so this is preparing to reduce reliance on that lock by using the existing existing atomic bit ops.

---
 drivers/pci/bus.c    |  2 +-
 drivers/pci/pci.h    | 11 +++++++++--
 drivers/pci/remove.c | 20 +++++++++-----------
 3 files changed, 19 insertions(+), 14 deletions(-)
drivers/pci/bus.c +1 -1
index e0a2441be6d32..aec08e81abff7 100644
@@ -358,7 +358,7 @@ void pci_bus_add_device(struct pci_dev *dev)
 	if (retval < 0 && retval != -EPROBE_DEFER)
 		pci_warn(dev, "device attach failed (%d)\n", retval);
 
-	pci_dev_assign_added(dev, true);
+	pci_dev_assign_added(dev);
 
 	if (dev_of_node(&dev->dev) && pci_is_bridge(dev)) {
 		retval = of_platform_populate(dev_of_node(&dev->dev), NULL, NULL,
drivers/pci/pci.h +9 -2
index d89fdbf04f363..c4cceec006d0d 100644
@@ -470,9 +470,16 @@ static inline int pci_dev_set_disconnected(struct pci_dev *dev, void *unused)
 #define PCI_DPC_RECOVERED 1
 #define PCI_DPC_RECOVERING 2
 
-static inline void pci_dev_assign_added(struct pci_dev *dev, bool added)
+static inline void pci_dev_assign_added(struct pci_dev *dev)
 {
-	assign_bit(PCI_DEV_ADDED, &dev->priv_flags, added);
+	smp_mb__before_atomic();
+	set_bit(PCI_DEV_ADDED, &dev->priv_flags);
+	smp_mb__after_atomic();
+}
+
+static inline bool pci_dev_test_and_clear_added(struct pci_dev *dev)
+{
+	return test_and_clear_bit(PCI_DEV_ADDED, &dev->priv_flags);
 }
 
 static inline bool pci_dev_is_added(const struct pci_dev *dev)
drivers/pci/remove.c +9 -11
index e4ce1145aa3ee..e0d4402ec3391 100644
@@ -31,18 +31,16 @@ static int pci_pwrctl_unregister(struct device *dev, void *data)
 
 static void pci_stop_dev(struct pci_dev *dev)
 {
-	pci_pme_active(dev, false);
-
-	if (pci_dev_is_added(dev)) {
-		device_for_each_child(dev->dev.parent, dev_of_node(&dev->dev),
-				      pci_pwrctl_unregister);
-		device_release_driver(&dev->dev);
-		pci_proc_detach_device(dev);
-		pci_remove_sysfs_dev_files(dev);
-		of_pci_remove_node(dev);
+	if (!pci_dev_test_and_clear_added(dev))
+		return;
 
-		pci_dev_assign_added(dev, false);
-	}
+	pci_pme_active(dev, false);
+	device_for_each_child(dev->dev.parent, dev_of_node(&dev->dev),
+			      pci_pwrctl_unregister);
+	device_release_driver(&dev->dev);
+	pci_proc_detach_device(dev);
+	pci_remove_sysfs_dev_files(dev);
+	of_pci_remove_node(dev);
 }
 
 static void pci_destroy_dev(struct pci_dev *dev)
Signature · 1 line
2.43.5
#2 lore raw

Re: [PATCHv3 1/5] pci: make pci_stop_dev concurrent safe

Quoted 18 lines · Keith Busch
drivers/pci/remove.c +2 -9
@@ -31,18 +31,16 @@ static int pci_pwrctl_unregister(struct device *dev, void *data)
 
 static void pci_stop_dev(struct pci_dev *dev)
 {
-	pci_pme_active(dev, false);
-
-	if (pci_dev_is_added(dev)) {
-		device_for_each_child(dev->dev.parent, dev_of_node(&dev->dev),
-				      pci_pwrctl_unregister);
-		device_release_driver(&dev->dev);
-		pci_proc_detach_device(dev);
-		pci_remove_sysfs_dev_files(dev);
-		of_pci_remove_node(dev);
+	if (!pci_dev_test_and_clear_added(dev))
+		return;
 

drivers/pci/remove.c static int pci_pwrctl_unregister(struct device *dev, void *data) · line 44

Quoted 10 lines
drivers/pci/remove.c +7 -2
-		pci_dev_assign_added(dev, false);
-	}
+	pci_pme_active(dev, false);
+	device_for_each_child(dev->dev.parent, dev_of_node(&dev->dev),
+			      pci_pwrctl_unregister);
+	device_release_driver(&dev->dev);
+	pci_proc_detach_device(dev);
+	pci_remove_sysfs_dev_files(dev);
+	of_pci_remove_node(dev);
 }

The above is now queued for v6.13 as commit 6d6d962a8dc2 on pci/locking.

I note there's a behavioral change here:

Previously "pci_pme_active(dev, false)" was called unconditionally, now only if the "added" flag has been set. The commit message doesn't explain why this change is fine, so perhaps it's inadvertent?

Thanks,

Lukas

#3 lore raw

Re: [PATCHv3 1/5] pci: make pci_stop_dev concurrent safe

Quoted 32 lines · Lukas Wunner
Quoted 18 lines · Keith Busch
drivers/pci/remove.c +2 -9
@@ -31,18 +31,16 @@ static int pci_pwrctl_unregister(struct device *dev, void *data)
 
 static void pci_stop_dev(struct pci_dev *dev)
 {
-	pci_pme_active(dev, false);
-
-	if (pci_dev_is_added(dev)) {
-		device_for_each_child(dev->dev.parent, dev_of_node(&dev->dev),
-				      pci_pwrctl_unregister);
-		device_release_driver(&dev->dev);
-		pci_proc_detach_device(dev);
-		pci_remove_sysfs_dev_files(dev);
-		of_pci_remove_node(dev);
+	if (!pci_dev_test_and_clear_added(dev))
+		return;
 

drivers/pci/remove.c static int pci_pwrctl_unregister(struct device *dev, void *data) · line 44

Quoted 10 lines
drivers/pci/remove.c +7 -2
-		pci_dev_assign_added(dev, false);
-	}
+	pci_pme_active(dev, false);
+	device_for_each_child(dev->dev.parent, dev_of_node(&dev->dev),
+			      pci_pwrctl_unregister);
+	device_release_driver(&dev->dev);
+	pci_proc_detach_device(dev);
+	pci_remove_sysfs_dev_files(dev);
+	of_pci_remove_node(dev);
 }

The above is now queued for v6.13 as commit 6d6d962a8dc2 on pci/locking.

Quoted 5 lines

I note there's a behavioral change here:

Previously "pci_pme_active(dev, false)" was called unconditionally, now only if the "added" flag has been set. The commit message doesn't explain why this change is fine, so perhaps it's inadvertent?

Hm, not exactly intentional. It doesn't appear to accomplish anything to call it multiple times, but it also looks hamrless to do so. Looking at the history of this, it looks like it was purposefully done unconditionally with the understanding it's "safe" to do that. With that in mind, I'm happy to move it back where it was.

#4 lore raw

Re: [PATCHv3 1/5] pci: make pci_stop_dev concurrent safe

Quoted 39 lines · Keith Busch
Quoted 32 lines · Lukas Wunner
Quoted 18 lines · Keith Busch
drivers/pci/remove.c +2 -9
@@ -31,18 +31,16 @@ static int pci_pwrctl_unregister(struct device *dev, void *data)
 
 static void pci_stop_dev(struct pci_dev *dev)
 {
-	pci_pme_active(dev, false);
-
-	if (pci_dev_is_added(dev)) {
-		device_for_each_child(dev->dev.parent, dev_of_node(&dev->dev),
-				      pci_pwrctl_unregister);
-		device_release_driver(&dev->dev);
-		pci_proc_detach_device(dev);
-		pci_remove_sysfs_dev_files(dev);
-		of_pci_remove_node(dev);
+	if (!pci_dev_test_and_clear_added(dev))
+		return;
 

drivers/pci/remove.c static int pci_pwrctl_unregister(struct device *dev, void *data) · line 44

Quoted 10 lines
drivers/pci/remove.c +7 -2
-		pci_dev_assign_added(dev, false);
-	}
+	pci_pme_active(dev, false);
+	device_for_each_child(dev->dev.parent, dev_of_node(&dev->dev),
+			      pci_pwrctl_unregister);
+	device_release_driver(&dev->dev);
+	pci_proc_detach_device(dev);
+	pci_remove_sysfs_dev_files(dev);
+	of_pci_remove_node(dev);
 }

The above is now queued for v6.13 as commit 6d6d962a8dc2 on pci/locking.

Quoted 5 lines

I note there's a behavioral change here:

Previously "pci_pme_active(dev, false)" was called unconditionally, now only if the "added" flag has been set. The commit message doesn't explain why this change is fine, so perhaps it's inadvertent?

Quoted 5 lines

Hm, not exactly intentional. It doesn't appear to accomplish anything to call it multiple times, but it also looks hamrless to do so. Looking at the history of this, it looks like it was purposefully done unconditionally with the understanding it's "safe" to do that. With that in mind, I'm happy to move it back where it was.

Yes I think it would be good if you could submit a fixup that Bjorn could fold into 6d6d962a8dc2, just to minimize regression potential.

Thanks,

Lukas

#5 lore raw

PATCH v3 2/5 pci: make pci_destroy_dev concurrent safe

In reply to Keith Busch

From: Keith Busch <[email protected]>

Use an atomic flag instead of the racey check against the device's kobj parent. We shouldn't be poking into device implementation details at this level anyway.

---
 drivers/pci/pci.h    | 6 ++++++
 drivers/pci/remove.c | 2 +-
 2 files changed, 7 insertions(+), 1 deletion(-)
drivers/pci/pci.h +6 -0
index c4cceec006d0d..cae4f55d5a4be 100644
@@ -469,6 +469,7 @@ static inline int pci_dev_set_disconnected(struct pci_dev *dev, void *unused)
 #define PCI_DEV_ADDED 0
 #define PCI_DPC_RECOVERED 1
 #define PCI_DPC_RECOVERING 2
+#define PCI_DEV_REMOVED 3
 
 static inline void pci_dev_assign_added(struct pci_dev *dev)
 {
@@ -487,6 +488,11 @@ static inline bool pci_dev_is_added(const struct pci_dev *dev)
 	return test_bit(PCI_DEV_ADDED, &dev->priv_flags);
 }
 
+static inline bool pci_dev_test_and_set_removed(struct pci_dev *dev)
+{
+	return test_and_set_bit(PCI_DEV_REMOVED, &dev->priv_flags);
+}
+
 #ifdef CONFIG_PCIEAER
 #include <linux/aer.h>
 
drivers/pci/remove.c +1 -1
index e0d4402ec3391..2e940101ce1bf 100644
@@ -45,7 +45,7 @@ static void pci_stop_dev(struct pci_dev *dev)
 
 static void pci_destroy_dev(struct pci_dev *dev)
 {
-	if (!dev->dev.kobj.parent)
+	if (pci_dev_test_and_set_removed(dev))
 		return;
 
 	pci_npem_remove(dev);
Signature · 1 line
2.43.5
#6 lore raw

PATCH v3 3/5 pci: move the walk bus lock to where its needed

In reply to Keith Busch

From: Keith Busch <[email protected]>

Simplify the common function by removing an unnecessary parameter that can be more easily handled in the only caller that wants it.

---
 drivers/pci/bus.c | 12 +++++-------
 1 file changed, 5 insertions(+), 7 deletions(-)
drivers/pci/bus.c +5 -7
index aec08e81abff7..b863b8bdd6ee6 100644
@@ -400,7 +400,7 @@ void pci_bus_add_devices(const struct pci_bus *bus)
 EXPORT_SYMBOL(pci_bus_add_devices);
 
 static void __pci_walk_bus(struct pci_bus *top, int (*cb)(struct pci_dev *, void *),
-			   void *userdata, bool locked)
+			   void *userdata)
 {
 	struct pci_dev *dev;
 	struct pci_bus *bus;
@@ -408,8 +408,6 @@ static void __pci_walk_bus(struct pci_bus *top, int (*cb)(struct pci_dev *, void
 	int retval;
 
 	bus = top;
-	if (!locked)
-		down_read(&pci_bus_sem);
 	next = top->devices.next;
 	for (;;) {
 		if (next == &bus->devices) {
@@ -432,8 +430,6 @@ static void __pci_walk_bus(struct pci_bus *top, int (*cb)(struct pci_dev *, void
 		if (retval)
 			break;
 	}
-	if (!locked)
-		up_read(&pci_bus_sem);
 }
 
 /**
@@ -451,7 +447,9 @@ static void __pci_walk_bus(struct pci_bus *top, int (*cb)(struct pci_dev *, void
  */
 void pci_walk_bus(struct pci_bus *top, int (*cb)(struct pci_dev *, void *), void *userdata)
 {
-	__pci_walk_bus(top, cb, userdata, false);
+	down_read(&pci_bus_sem);
+	__pci_walk_bus(top, cb, userdata);
+	up_read(&pci_bus_sem);
 }
 EXPORT_SYMBOL_GPL(pci_walk_bus);
 
@@ -459,7 +457,7 @@ void pci_walk_bus_locked(struct pci_bus *top, int (*cb)(struct pci_dev *, void *
 {
 	lockdep_assert_held(&pci_bus_sem);
 
-	__pci_walk_bus(top, cb, userdata, true);
+	__pci_walk_bus(top, cb, userdata);
 }
 EXPORT_SYMBOL_GPL(pci_walk_bus_locked);
 
Signature · 1 line
2.43.5
#7 lore raw

PATCH v3 4/5 pci: walk bus recursively

In reply to Keith Busch

From: Keith Busch <[email protected]>

The original implementation purposefully chose a non-recursive walk, presumably as a precaution on stack use. We do recursive bus walking in other places though. For example:

pci_bus_resettable
pci_stop_bus_device
pci_remove_bus_device
pci_bus_allocate_dev_resources

So, recursive pci bus walking is well tested and safe, and the implementation is easier to follow. The motivation for changing it now is to make it easier to introduce finer grain locking in the future.

---
 drivers/pci/bus.c | 36 +++++++++++-------------------------
 1 file changed, 11 insertions(+), 25 deletions(-)
drivers/pci/bus.c +11 -25
index b863b8bdd6ee6..9c552396241e0 100644
@@ -399,37 +399,23 @@ void pci_bus_add_devices(const struct pci_bus *bus)
 }
 EXPORT_SYMBOL(pci_bus_add_devices);
 
-static void __pci_walk_bus(struct pci_bus *top, int (*cb)(struct pci_dev *, void *),
-			   void *userdata)
+static int __pci_walk_bus(struct pci_bus *top, int (*cb)(struct pci_dev *, void *),
+			  void *userdata)
 {
 	struct pci_dev *dev;
-	struct pci_bus *bus;
-	struct list_head *next;
-	int retval;
+	int ret = 0;
 
-	bus = top;
-	next = top->devices.next;
-	for (;;) {
-		if (next == &bus->devices) {
-			/* end of this bus, go up or finish */
-			if (bus == top)
+	list_for_each_entry(dev, &top->devices, bus_list) {
+		ret = cb(dev, userdata);
+		if (ret)
+			break;
+		if (dev->subordinate) {
+			ret = __pci_walk_bus(dev->subordinate, cb, userdata);
+			if (ret)
 				break;
-			next = bus->self->bus_list.next;
-			bus = bus->self->bus;
-			continue;
 		}
-		dev = list_entry(next, struct pci_dev, bus_list);
-		if (dev->subordinate) {
-			/* this is a pci-pci bridge, do its devices next */
-			next = dev->subordinate->devices.next;
-			bus = dev->subordinate;
-		} else
-			next = dev->bus_list.next;
-
-		retval = cb(dev, userdata);
-		if (retval)
-			break;
 	}
+	return ret;
 }
 
 /**
Signature · 1 line
2.43.5
#8 lore raw

Re: [PATCHv3 4/5] pci: walk bus recursively

Quoted 4 lines · Keith Busch

The original implementation purposefully chose a non-recursive walk, presumably as a precaution on stack use. We do recursive bus walking in other places though. For example:

Quoted 8 lines

pci_bus_resettable
pci_stop_bus_device
pci_remove_bus_device
pci_bus_allocate_dev_resources

So, recursive pci bus walking is well tested and safe, and the implementation is easier to follow. The motivation for changing it now is to make it easier to introduce finer grain locking in the future.

Hm, I find the loss of non-recursive bus walking regrettable.

If the solution proposed earlier today...

https://lore.kernel.org/all/[email protected]/

... is workable and we find it doesn't need recursive bus walking, perhaps we should consider reverting this change (which is now e24eafdda271 on pci/locking).

Thanks,

Lukas

#9 lore raw

PATCH v3 5/5 pci: unexport pci_walk_bus_locked

In reply to Keith Busch

From: Keith Busch <[email protected]>

There's only one user of this, and it's internal, so no need to export it.

---
 drivers/pci/bus.c | 1 -
 1 file changed, 1 deletion(-)
drivers/pci/bus.c +0 -1
index 9c552396241e0..d015d5821cefc 100644
@@ -445,7 +445,6 @@ void pci_walk_bus_locked(struct pci_bus *top, int (*cb)(struct pci_dev *, void *
 
 	__pci_walk_bus(top, cb, userdata);
 }
-EXPORT_SYMBOL_GPL(pci_walk_bus_locked);
 
 struct pci_bus *pci_bus_get(struct pci_bus *bus)
 {
Signature · 1 line
2.43.5
#10 lore raw

Re: [PATCHv3 5/5] pci: unexport pci_walk_bus_locked

Quoted 18 lines · Keith Busch

From: Keith Busch <[email protected]>

There's only one user of this, and it's internal, so no need to export it.

---
 drivers/pci/bus.c | 1 -
 1 file changed, 1 deletion(-)
drivers/pci/bus.c +0 -0
index 9c552396241e0..d015d5821cefc 100644

drivers/pci/bus.c void pci_walk_bus_locked(struct pci_bus *top, int (*cb)(struct pci_dev *, void * · line 448

Quoted 4 lines
drivers/pci/bus.c +0 -1
@@ -445,7 +445,6 @@ void pci_walk_bus_locked(struct pci_bus *top, int (*cb)(struct pci_dev *, void *
 
 	__pci_walk_bus(top, cb, userdata);
 }
-EXPORT_SYMBOL_GPL(pci_walk_bus_locked);

I think we could also move the declaration from include/linux/pci.h to drivers/pci/pci.h, right?

I guess there's some argument for keeping it in include/linux/pci.h next to the pci_walk_bus() declaration, but I certainly don't want to encourage more use of either one, especially outside the PCI core.

#11 lore raw

Re: [PATCHv3 5/5] pci: unexport pci_walk_bus_locked

Quoted 10 lines · Bjorn Helgaas

drivers/pci/bus.c void pci_walk_bus_locked(struct pci_bus *top, int (*cb)(struct pci_dev *, void * · line 448

Quoted 5 lines · Keith Busch
drivers/pci/bus.c +0 -1
@@ -445,7 +445,6 @@ void pci_walk_bus_locked(struct pci_bus *top, int (*cb)(struct pci_dev *, void *
 
 	__pci_walk_bus(top, cb, userdata);
 }
-EXPORT_SYMBOL_GPL(pci_walk_bus_locked);

I think we could also move the declaration from include/linux/pci.h to drivers/pci/pci.h, right?

Quoted 3 lines

I guess there's some argument for keeping it in include/linux/pci.h next to the pci_walk_bus() declaration, but I certainly don't want to encourage more use of either one, especially outside the PCI core.

Thanks, that's a good point! Not only are modules not using this, neither does any kernel code outside drivers/pci/, so no need to declare this in include/linux/.

#12 lore raw

Re: [PATCHv3 5/5] pci: unexport pci_walk_bus_locked

Quoted 11 lines · Keith Busch
Quoted 10 lines · Bjorn Helgaas

drivers/pci/bus.c void pci_walk_bus_locked(struct pci_bus *top, int (*cb)(struct pci_dev *, void * · line 448

Quoted 5 lines · Keith Busch
drivers/pci/bus.c +0 -1
@@ -445,7 +445,6 @@ void pci_walk_bus_locked(struct pci_bus *top, int (*cb)(struct pci_dev *, void *
 
 	__pci_walk_bus(top, cb, userdata);
 }
-EXPORT_SYMBOL_GPL(pci_walk_bus_locked);

I think we could also move the declaration from include/linux/pci.h to drivers/pci/pci.h, right?

Quoted 7 lines
Quoted 3 lines

I guess there's some argument for keeping it in include/linux/pci.h next to the pci_walk_bus() declaration, but I certainly don't want to encourage more use of either one, especially outside the PCI core.

Thanks, that's a good point! Not only are modules not using this, neither does any kernel code outside drivers/pci/, so no need to declare this in include/linux/.

I'll make this change locally so you don't need to repost for it.

#13 lore raw
In reply to Keith Busch
Quoted 21 lines · Keith Busch

From: Keith Busch <[email protected]>

This is a subset of a previous RFC bus lock patches that are simply good cleanups that should help make it easier to introduce different locking later.

Changes from v2:

Rebased to pci/next

Added memory barriers around bit ops for patch 1.

Added reviews.

Keith Busch (5):
pci: make pci_stop_dev concurrent safe
pci: make pci_destroy_dev concurrent safe
pci: move the walk bus lock to where its needed
pci: walk bus recursively
pci: unexport pci_walk_bus_locked

Quoted 4 lines
 drivers/pci/bus.c    | 49 +++++++++++++++-----------------------------
 drivers/pci/pci.h    | 17 +++++++++++++--
 drivers/pci/remove.c | 22 +++++++++-----------
 3 files changed, 41 insertions(+), 47 deletions(-)

Applied to pci/locking for v6.13, thanks Keith!