Skip to contentSkip to conversation

PCI/hotplug: Add 'uevent' sysfs attribute to trigger slot events

PATCH Superseded Ramesh Errabolu 3 participants 8 messages Apr 16 – Apr 22 1 message not in the archive

Subsystems (2)

v2 of 6 Superseded Latest is v6 →

lore
Share & tools
lore
#0 lore raw

v2 cover PCI/hotplug: Add 'uevent' sysfs attribute to trigger slot events

Hi all,

This is v2 of the patch adding write-only uevent support for PCI slots. The original v1 was posted in February 2026 and received valuable feedback from Leon Romanovsky and others. This v2 addresses the feedback and is rebased onto the current kernel tree.

Link to v1:
https://lore.kernel.org/all/[email protected]/

Testing: Tested on s390x with multiple PCI slots
- Write-only uevent attribute is created for each slot
- Writing "add" triggers proper uevent generation
- udevadm-trigger successfully generates slot uevents
- Events are properly received by udev monitoring tools

Changes since v1 (February 2026):
- Improved commit message clarity, per reviewer feedback
- Rebased onto current kernel tree (v7.x)

Ramesh Errabolu (1):
PCI/hotplug: Add 'uevent' sysfs attribute to trigger slot events

 drivers/pci/hotplug/pci_hotplug_core.c | 25 +++++++++++++++++++++++++
 1 file changed, 25 insertions(+)
Signature · 1 line
2.43.0
#1 lore raw

PATCH v2 1/1 PCI/hotplug: Add 'uevent' sysfs attribute to trigger slot events

Add a write-only 'uevent' sysfs attribute for synthesizing uevents for a PCI slot. This extends the existing uevent support which emits a KOBJ_ADD uevent in pci_hp_add() with the ability to replay such uevents for cold plugged devices. As such events are only emitted by hotplug capable PCI slots so is the support for synthesizing them.

The change was validated by manually triggering 'add' uevent for a specific hotplug PCI slot:

    $ echo "add $(uuidgen)" | sudo tee   \
                /sys/bus/pci/slots/<slot-id>/uevent
---
 drivers/pci/hotplug/pci_hotplug_core.c | 25 +++++++++++++++++++++++++
 1 file changed, 25 insertions(+)
drivers/pci/hotplug/pci_hotplug_core.c +25 -0
index fadcf98a8a66..c3634b1cc7a8 100644
@@ -176,6 +176,21 @@ static struct pci_slot_attribute hotplug_slot_attr_presence = {
 	.show = presence_read_file,
 };
 
+static ssize_t uevent_write_file(struct pci_slot *slot,
+				 const char *buf, size_t len)
+{
+	int rc;
+
+	rc = kobject_synth_uevent(&slot->kobj, buf, len);
+	return rc ? rc : len;
+}
+
+static struct pci_slot_attribute hotplug_slot_attr_uevent = {
+	.attr = {.name = "uevent", .mode = S_IFREG | 0200},
+	.show = NULL,
+	.store = uevent_write_file
+};
+
 static ssize_t test_write_file(struct pci_slot *pci_slot, const char *buf,
 			       size_t count)
 {
@@ -254,6 +269,11 @@ static int fs_add_slot(struct hotplug_slot *slot, struct pci_slot *pci_slot)
 		kobject_put(kobj);
 	}
 
+	retval = sysfs_create_file(&pci_slot->kobj,
+				   &hotplug_slot_attr_uevent.attr);
+	if (retval)
+		goto exit_uevent;
+
 	if (has_power_file(slot)) {
 		retval = sysfs_create_file(&pci_slot->kobj,
 					   &hotplug_slot_attr_power.attr);
@@ -306,6 +326,9 @@ static int fs_add_slot(struct hotplug_slot *slot, struct pci_slot *pci_slot)
 	if (has_power_file(slot))
 		sysfs_remove_file(&pci_slot->kobj, &hotplug_slot_attr_power.attr);
 exit_power:
+	sysfs_remove_file(&pci_slot->kobj,
+			  &hotplug_slot_attr_uevent.attr);
+exit_uevent:
 	sysfs_remove_link(&pci_slot->kobj, "module");
 exit:
 	return retval;
@@ -313,6 +336,8 @@ static int fs_add_slot(struct hotplug_slot *slot, struct pci_slot *pci_slot)
 
 static void fs_remove_slot(struct hotplug_slot *slot, struct pci_slot *pci_slot)
 {
+	sysfs_remove_file(&pci_slot->kobj, &hotplug_slot_attr_uevent.attr);
+
 	if (has_power_file(slot))
 		sysfs_remove_file(&pci_slot->kobj, &hotplug_slot_attr_power.attr);
 
Signature · 1 line
2.43.0
Reviewed-by Tested-by on v2 1/1
#2 lore raw
Quoted 10 lines · Ramesh Errabolu

Add a write-only 'uevent' sysfs attribute for synthesizing uevents for a PCI slot. This extends the existing uevent support which emits a KOBJ_ADD uevent in pci_hp_add() with the ability to replay such uevents for cold plugged devices. As such events are only emitted by hotplug capable PCI slots so is the support for synthesizing them.

The change was validated by manually triggering 'add' uevent for a specific hotplug PCI slot:

Quoted 8 lines
    $ echo "add $(uuidgen)" | sudo tee   \
                /sys/bus/pci/slots/<slot-id>/uevent
---
 drivers/pci/hotplug/pci_hotplug_core.c | 25 +++++++++++++++++++++++++
 1 file changed, 25 insertions(+)

Sorry for the review delay. This looks good to me and will really help with udev rule execution for devices which are discovered in standby state during boot and thus currently don't create any uevents. Of course to work by default this still requires a systemd/udev addition that I know you're also working on.

Consequently, I tested the uevent synthesization with both slots with a configured (slot powered on) and standby (slot powered off) PCI function.

Feel free to add:

Thanks,
Niklas

#3 lore raw

Thanks for testing and also giving me the review-by

Regards,
Ramesh

Quoted 36 lines · Niklas Schnelle
Quoted 10 lines · Ramesh Errabolu

Add a write-only 'uevent' sysfs attribute for synthesizing uevents for a PCI slot. This extends the existing uevent support which emits a KOBJ_ADD uevent in pci_hp_add() with the ability to replay such uevents for cold plugged devices. As such events are only emitted by hotplug capable PCI slots so is the support for synthesizing them.

The change was validated by manually triggering 'add' uevent for a specific hotplug PCI slot:

Quoted 8 lines
     $ echo "add $(uuidgen)" | sudo tee   \
                 /sys/bus/pci/slots/<slot-id>/uevent
---
  drivers/pci/hotplug/pci_hotplug_core.c | 25 +++++++++++++++++++++++++
  1 file changed, 25 insertions(+)

Sorry for the review delay. This looks good to me and will really help with udev rule execution for devices which are discovered in standby state during boot and thus currently don't create any uevents. Of course to work by default this still requires a systemd/udev addition that I know you're also working on.

Consequently, I tested the uevent synthesization with both slots with a configured (slot powered on) and standby (slot powered off) PCI function.

Feel free to add:

Thanks,
Niklas

#4 lore raw
In reply to Ramesh Errabolu

Hello,

drivers/pci/hotplug/pci_hotplug_core.c static struct pci_slot_attribute hotplug_slot_attr_presence = { · line 192

Quoted 5 lines
drivers/pci/hotplug/pci_hotplug_core.c +5 -0
+static struct pci_slot_attribute hotplug_slot_attr_uevent = {
+	.attr = {.name = "uevent", .mode = S_IFREG | 0200},
+	.show = NULL,
+	.store = uevent_write_file
+};

I think, you could use the __ATTR_WO() macro here.

Thank you!

Krzysztof

#5 lore raw
Quoted 8 lines · Krzysztof Wilczyński

Hello,

drivers/pci/hotplug/pci_hotplug_core.c static struct pci_slot_attribute hotplug_slot_attr_presence = { · line 192

Quoted 5 lines
drivers/pci/hotplug/pci_hotplug_core.c +5 -0
+static struct pci_slot_attribute hotplug_slot_attr_uevent = {
+	.attr = {.name = "uevent", .mode = S_IFREG | 0200},
+	.show = NULL,
+	.store = uevent_write_file
+};

I think, you could use the __ATTR_WO() macro here.

Use of the suggested macro is not appropriate

Quoted 4 lines

Thank you!

Krzysztof

#6 lore raw

Hello,

Quoted 7 lines
Quoted 6 lines

drivers/pci/hotplug/pci_hotplug_core.c static struct pci_slot_attribute hotplug_slot_attr_presence = { · line 192

Quoted 5 lines
drivers/pci/hotplug/pci_hotplug_core.c +5 -0
+static struct pci_slot_attribute hotplug_slot_attr_uevent = {
+	.attr = {.name = "uevent", .mode = S_IFREG | 0200},
+	.show = NULL,
+	.store = uevent_write_file
+};

I think, you could use the __ATTR_WO() macro here.

Use of the suggested macro is not appropriate

Just to expand on the "not appropriate" bit here.

Not wanting to change the name to have the _store suffix would be fine.

The __ATTR() would work here. But, this file is old and wanting to keep the style aligned with rest of it would be fine, too.

So, both of these reasons would be fine, to avoid using any new macro.

Thank you!

Krzysztof

A message that is not in the archive (sent off-list or not yet received): [email protected]

#8 lore raw

Hello,

Quoted 15 lines
Quoted 15 lines
Quoted 7 lines
Quoted 6 lines

drivers/pci/hotplug/pci_hotplug_core.c static struct pci_slot_attribute hotplug_slot_attr_presence = { · line 192

Quoted 5 lines
drivers/pci/hotplug/pci_hotplug_core.c +5 -0
+static struct pci_slot_attribute hotplug_slot_attr_uevent = {
+	.attr = {.name = "uevent", .mode = S_IFREG | 0200},
+	.show = NULL,
+	.store = uevent_write_file
+};

I think, you could use the __ATTR_WO() macro here.

Use of the suggested macro is not appropriate

Just to expand on the "not appropriate" bit here.

Not wanting to change the name to have the _store suffix would be fine.

The __ATTR() would work here. But, this file is old and wanting to keep the style aligned with rest of it would be fine, too.

So, both of these reasons would be fine, to avoid using any new macro.

[...]

Quoted 4 lines

Hello Krzysztof, let me first apologize to you in the way I said it. I should have elaborated the reasoning the way you did. I am open to making the changing if anyone else feels it is a better way.

Oh, no worries. No hard feelings. :)

I just wanted to add a little bit more details for posterity.

Thank you!

Krzysztof