Skip to contentSkip to conversation

PCI: Fix warning without CONFIG_VIDEO

PATCH In review Mario Limonciello 2 participants 9 messages Jul 18, 2025 – Jul 21, 2025 PCI

lore
Share & tools
lore
#0 lore raw

PATCH 1/1 PCI: Fix warning without CONFIG_VIDEO

From: Mario Limonciello <[email protected]>

When compiled without CONFIG_VIDEO pci_create_boot_display_file() will never create a sysfs file for boot_display. Guard the sysfs file declaration against CONFIG_VIDEO.

---
 drivers/pci/pci-sysfs.c | 2 ++
 1 file changed, 2 insertions(+)
drivers/pci/pci-sysfs.c +2 -0
index 6b1a0ae254d3a..f6540a72204d3 100644
@@ -680,12 +680,14 @@ const struct attribute_group *pcibus_groups[] = {
 	NULL,
 };
 
+#ifdef CONFIG_VIDEO
 static ssize_t boot_display_show(struct device *dev, struct device_attribute *attr,
 				 char *buf)
 {
 	return sysfs_emit(buf, "1\n");
 }
 static DEVICE_ATTR_RO(boot_display);
+#endif /* CONFIG_VIDEO */
 
 static ssize_t boot_vga_show(struct device *dev, struct device_attribute *attr,
 			     char *buf)
Signature · 1 line
2.43.0
#1 lore raw
Quoted 18 lines · Mario Limonciello

From: Mario Limonciello <[email protected]>

When compiled without CONFIG_VIDEO pci_create_boot_display_file() will never create a sysfs file for boot_display. Guard the sysfs file declaration against CONFIG_VIDEO.

---
 drivers/pci/pci-sysfs.c | 2 ++
 1 file changed, 2 insertions(+)
drivers/pci/pci-sysfs.c +0 -0
index 6b1a0ae254d3a..f6540a72204d3 100644

drivers/pci/pci-sysfs.c const struct attribute_group *pcibus_groups[] = { · line 689

Quoted 10 lines
drivers/pci/pci-sysfs.c +1 -0
@@ -680,12 +680,14 @@ const struct attribute_group *pcibus_groups[] = {
 	NULL,
 };
 
+#ifdef CONFIG_VIDEO
 static ssize_t boot_display_show(struct device *dev, struct device_attribute *attr,
 				 char *buf)
 {
 	return sysfs_emit(buf, "1\n");
 }
 static DEVICE_ATTR_RO(boot_display);

I failed to give my comment during the offending series itself, but it is never late than never. Why are we adding non-PCI attributes under bus/pci in the first place? Though the underlying device uses PCI as a transport, only the PCI bus specific attrbutes should be placed under bus/pci and the driver/peripheral specific attrbutes should belong to the respective bus/class/device hierarchy.

Now, if other peripherals (like netdev) start adding these device specific attributes under bus/pci, it will turn out to be a mess.

- Mani

Signature · 1 line
மணிவண்ணன் சதாசிவம்
#2 lore raw
Quoted 36 lines · Manivannan Sadhasivam
Quoted 18 lines · Mario Limonciello

From: Mario Limonciello <[email protected]>

When compiled without CONFIG_VIDEO pci_create_boot_display_file() will never create a sysfs file for boot_display. Guard the sysfs file declaration against CONFIG_VIDEO.

---
  drivers/pci/pci-sysfs.c | 2 ++
  1 file changed, 2 insertions(+)
drivers/pci/pci-sysfs.c +0 -0
index 6b1a0ae254d3a..f6540a72204d3 100644

drivers/pci/pci-sysfs.c const struct attribute_group *pcibus_groups[] = { · line 689

Quoted 10 lines
drivers/pci/pci-sysfs.c +1 -0
@@ -680,12 +680,14 @@ const struct attribute_group *pcibus_groups[] = {
  	NULL,
  };
 
+#ifdef CONFIG_VIDEO
  static ssize_t boot_display_show(struct device *dev, struct device_attribute *attr,
  				 char *buf)
  {
  	return sysfs_emit(buf, "1\n");
  }
  static DEVICE_ATTR_RO(boot_display);

I failed to give my comment during the offending series itself, but it is never late than never. Why are we adding non-PCI attributes under bus/pci in the first place? Though the underlying device uses PCI as a transport, only the PCI bus specific attrbutes should be placed under bus/pci and the driver/peripheral specific attrbutes should belong to the respective bus/class/device hierarchy.

Quoted 5 lines

Now, if other peripherals (like netdev) start adding these device specific attributes under bus/pci, it will turn out to be a mess.

- Mani

It was mostly to mirror the location of where boot_vga is, which arguably has the same issue you raise.

I would be incredibly surprised if there was a proposal to add a 'boot_display' attribute from netdev..

#3 lore raw
Quoted 36 lines · Mario Limonciello
Quoted 35 lines · Manivannan Sadhasivam
Quoted 18 lines · Mario Limonciello

From: Mario Limonciello <[email protected]>

When compiled without CONFIG_VIDEO pci_create_boot_display_file() will never create a sysfs file for boot_display. Guard the sysfs file declaration against CONFIG_VIDEO.

---
  drivers/pci/pci-sysfs.c | 2 ++
  1 file changed, 2 insertions(+)
drivers/pci/pci-sysfs.c +0 -0
index 6b1a0ae254d3a..f6540a72204d3 100644

drivers/pci/pci-sysfs.c const struct attribute_group *pcibus_groups[] = { · line 688

Quoted 9 lines
drivers/pci/pci-sysfs.c +1 -0
@@ -680,12 +680,14 @@ const struct attribute_group *pcibus_groups[] = {
  	NULL,
  };
+#ifdef CONFIG_VIDEO
  static ssize_t boot_display_show(struct device *dev, struct device_attribute *attr,
  				 char *buf)
  {
  	return sysfs_emit(buf, "1\n");
  }
  static DEVICE_ATTR_RO(boot_display);

I failed to give my comment during the offending series itself, but it is never late than never. Why are we adding non-PCI attributes under bus/pci in the first place? Though the underlying device uses PCI as a transport, only the PCI bus specific attrbutes should be placed under bus/pci and the driver/peripheral specific attrbutes should belong to the respective bus/class/device hierarchy.

Quoted 9 lines
Quoted 5 lines

Now, if other peripherals (like netdev) start adding these device specific attributes under bus/pci, it will turn out to be a mess.

- Mani

It was mostly to mirror the location of where boot_vga is, which arguably has the same issue you raise.

Yes, I agree. But 'boot_vga' has set a bad precedence IMO.

Quoted 2 lines

I would be incredibly surprised if there was a proposal to add a 'boot_display' attribute from netdev..

Not 'boot_display' but why not 'boot_network' or something else. I was just merely pointing out the fact that the other subsystems can start dumping device/usecase specific attributes under bus/pci.

- Mani

Signature · 1 line
மணிவண்ணன் சதாசிவம்
#4 lore raw
Quoted 49 lines · Manivannan Sadhasivam
Quoted 36 lines · Mario Limonciello
Quoted 35 lines · Manivannan Sadhasivam
Quoted 18 lines · Mario Limonciello

From: Mario Limonciello <[email protected]>

When compiled without CONFIG_VIDEO pci_create_boot_display_file() will never create a sysfs file for boot_display. Guard the sysfs file declaration against CONFIG_VIDEO.

---
   drivers/pci/pci-sysfs.c | 2 ++
   1 file changed, 2 insertions(+)
drivers/pci/pci-sysfs.c +0 -0
index 6b1a0ae254d3a..f6540a72204d3 100644

drivers/pci/pci-sysfs.c const struct attribute_group *pcibus_groups[] = { · line 688

Quoted 9 lines
drivers/pci/pci-sysfs.c +1 -0
@@ -680,12 +680,14 @@ const struct attribute_group *pcibus_groups[] = {
   	NULL,
   };
+#ifdef CONFIG_VIDEO
   static ssize_t boot_display_show(struct device *dev, struct device_attribute *attr,
   				 char *buf)
   {
   	return sysfs_emit(buf, "1\n");
   }
   static DEVICE_ATTR_RO(boot_display);

I failed to give my comment during the offending series itself, but it is never late than never. Why are we adding non-PCI attributes under bus/pci in the first place? Though the underlying device uses PCI as a transport, only the PCI bus specific attrbutes should be placed under bus/pci and the driver/peripheral specific attrbutes should belong to the respective bus/class/device hierarchy.

Quoted 9 lines
Quoted 5 lines

Now, if other peripherals (like netdev) start adding these device specific attributes under bus/pci, it will turn out to be a mess.

- Mani

It was mostly to mirror the location of where boot_vga is, which arguably has the same issue you raise.

Yes, I agree. But 'boot_vga' has set a bad precedence IMO.

Quoted 9 lines
Quoted 2 lines

I would be incredibly surprised if there was a proposal to add a 'boot_display' attribute from netdev..

Not 'boot_display' but why not 'boot_network' or something else. I was just merely pointing out the fact that the other subsystems can start dumping device/usecase specific attributes under bus/pci.

- Mani

This is a pretty general problem that exists that attributes are first come first served. For example amdgpu adds mem_busy_percent and it has certain semantics. Now PCI core can't add that.

And if nouveau.ko wants to add the same thing they need to follow the same semantics because userspace will look for those.

#5 lore raw
Quoted 60 lines · Mario Limonciello
Quoted 49 lines · Manivannan Sadhasivam
Quoted 36 lines · Mario Limonciello
Quoted 35 lines · Manivannan Sadhasivam
Quoted 18 lines · Mario Limonciello

From: Mario Limonciello <[email protected]>

When compiled without CONFIG_VIDEO pci_create_boot_display_file() will never create a sysfs file for boot_display. Guard the sysfs file declaration against CONFIG_VIDEO.

---
   drivers/pci/pci-sysfs.c | 2 ++
   1 file changed, 2 insertions(+)
drivers/pci/pci-sysfs.c +0 -0
index 6b1a0ae254d3a..f6540a72204d3 100644

drivers/pci/pci-sysfs.c const struct attribute_group *pcibus_groups[] = { · line 688

Quoted 9 lines
drivers/pci/pci-sysfs.c +1 -0
@@ -680,12 +680,14 @@ const struct attribute_group *pcibus_groups[] = {
   	NULL,
   };
+#ifdef CONFIG_VIDEO
   static ssize_t boot_display_show(struct device *dev, struct device_attribute *attr,
   				 char *buf)
   {
   	return sysfs_emit(buf, "1\n");
   }
   static DEVICE_ATTR_RO(boot_display);

I failed to give my comment during the offending series itself, but it is never late than never. Why are we adding non-PCI attributes under bus/pci in the first place? Though the underlying device uses PCI as a transport, only the PCI bus specific attrbutes should be placed under bus/pci and the driver/peripheral specific attrbutes should belong to the respective bus/class/device hierarchy.

Quoted 9 lines
Quoted 5 lines

Now, if other peripherals (like netdev) start adding these device specific attributes under bus/pci, it will turn out to be a mess.

- Mani

It was mostly to mirror the location of where boot_vga is, which arguably has the same issue you raise.

Yes, I agree. But 'boot_vga' has set a bad precedence IMO.

Quoted 9 lines
Quoted 2 lines

I would be incredibly surprised if there was a proposal to add a 'boot_display' attribute from netdev..

Not 'boot_display' but why not 'boot_network' or something else. I was just merely pointing out the fact that the other subsystems can start dumping device/usecase specific attributes under bus/pci.

- Mani

Quoted 6 lines

This is a pretty general problem that exists that attributes are first come first served. For example amdgpu adds mem_busy_percent and it has certain semantics. Now PCI core can't add that.

And if nouveau.ko wants to add the same thing they need to follow the same semantics because userspace will look for those.

But here, userspace is not yet looking for 'boot_display' isn't it? Why do you need to mirror 'boot_vga' attribute?

- Mani

Signature · 1 line
மணிவண்ணன் சதாசிவம்
#6 lore raw
Quoted 65 lines · Manivannan Sadhasivam
Quoted 64 lines · Mario Limonciello
Quoted 49 lines · Manivannan Sadhasivam
Quoted 36 lines · Mario Limonciello
Quoted 35 lines · Manivannan Sadhasivam
Quoted 18 lines · Mario Limonciello

From: Mario Limonciello <[email protected]>

When compiled without CONFIG_VIDEO pci_create_boot_display_file() will never create a sysfs file for boot_display. Guard the sysfs file declaration against CONFIG_VIDEO.

---
    drivers/pci/pci-sysfs.c | 2 ++
    1 file changed, 2 insertions(+)
drivers/pci/pci-sysfs.c +0 -0
index 6b1a0ae254d3a..f6540a72204d3 100644

drivers/pci/pci-sysfs.c const struct attribute_group *pcibus_groups[] = { · line 688

Quoted 9 lines
drivers/pci/pci-sysfs.c +1 -0
@@ -680,12 +680,14 @@ const struct attribute_group *pcibus_groups[] = {
    	NULL,
    };
+#ifdef CONFIG_VIDEO
    static ssize_t boot_display_show(struct device *dev, struct device_attribute *attr,
    				 char *buf)
    {
    	return sysfs_emit(buf, "1\n");
    }
    static DEVICE_ATTR_RO(boot_display);

I failed to give my comment during the offending series itself, but it is never late than never. Why are we adding non-PCI attributes under bus/pci in the first place? Though the underlying device uses PCI as a transport, only the PCI bus specific attrbutes should be placed under bus/pci and the driver/peripheral specific attrbutes should belong to the respective bus/class/device hierarchy.

Quoted 9 lines
Quoted 5 lines

Now, if other peripherals (like netdev) start adding these device specific attributes under bus/pci, it will turn out to be a mess.

- Mani

It was mostly to mirror the location of where boot_vga is, which arguably has the same issue you raise.

Yes, I agree. But 'boot_vga' has set a bad precedence IMO.

Quoted 9 lines
Quoted 2 lines

I would be incredibly surprised if there was a proposal to add a 'boot_display' attribute from netdev..

Not 'boot_display' but why not 'boot_network' or something else. I was just merely pointing out the fact that the other subsystems can start dumping device/usecase specific attributes under bus/pci.

- Mani

This is a pretty general problem that exists that attributes are first come first served. For example amdgpu adds mem_busy_percent and it has certain semantics. Now PCI core can't add that.

Quoted 6 lines
Quoted 2 lines

And if nouveau.ko wants to add the same thing they need to follow the same semantics because userspace will look for those.

But here, userspace is not yet looking for 'boot_display' isn't it? Why do you need to mirror 'boot_vga' attribute?

I have a pull request opened in libpciaccess; but correct it has not been merged.

/If/ we're to change this, what would you propose the path to be for this sysfs file?

#7 lore raw
Quoted 75 lines · Mario Limonciello
Quoted 65 lines · Manivannan Sadhasivam
Quoted 64 lines · Mario Limonciello
Quoted 49 lines · Manivannan Sadhasivam
Quoted 36 lines · Mario Limonciello
Quoted 35 lines · Manivannan Sadhasivam
Quoted 18 lines · Mario Limonciello

From: Mario Limonciello <[email protected]>

When compiled without CONFIG_VIDEO pci_create_boot_display_file() will never create a sysfs file for boot_display. Guard the sysfs file declaration against CONFIG_VIDEO.

---
    drivers/pci/pci-sysfs.c | 2 ++
    1 file changed, 2 insertions(+)
drivers/pci/pci-sysfs.c +0 -0
index 6b1a0ae254d3a..f6540a72204d3 100644

drivers/pci/pci-sysfs.c const struct attribute_group *pcibus_groups[] = { · line 688

Quoted 9 lines
drivers/pci/pci-sysfs.c +1 -0
@@ -680,12 +680,14 @@ const struct attribute_group *pcibus_groups[] = {
    	NULL,
    };
+#ifdef CONFIG_VIDEO
    static ssize_t boot_display_show(struct device *dev, struct device_attribute *attr,
    				 char *buf)
    {
    	return sysfs_emit(buf, "1\n");
    }
    static DEVICE_ATTR_RO(boot_display);

I failed to give my comment during the offending series itself, but it is never late than never. Why are we adding non-PCI attributes under bus/pci in the first place? Though the underlying device uses PCI as a transport, only the PCI bus specific attrbutes should be placed under bus/pci and the driver/peripheral specific attrbutes should belong to the respective bus/class/device hierarchy.

Quoted 9 lines
Quoted 5 lines

Now, if other peripherals (like netdev) start adding these device specific attributes under bus/pci, it will turn out to be a mess.

- Mani

It was mostly to mirror the location of where boot_vga is, which arguably has the same issue you raise.

Yes, I agree. But 'boot_vga' has set a bad precedence IMO.

Quoted 9 lines
Quoted 2 lines

I would be incredibly surprised if there was a proposal to add a 'boot_display' attribute from netdev..

Not 'boot_display' but why not 'boot_network' or something else. I was just merely pointing out the fact that the other subsystems can start dumping device/usecase specific attributes under bus/pci.

- Mani

This is a pretty general problem that exists that attributes are first come first served. For example amdgpu adds mem_busy_percent and it has certain semantics. Now PCI core can't add that.

Quoted 6 lines
Quoted 2 lines

And if nouveau.ko wants to add the same thing they need to follow the same semantics because userspace will look for those.

But here, userspace is not yet looking for 'boot_display' isn't it? Why do you need to mirror 'boot_vga' attribute?

Quoted 5 lines

I have a pull request opened in libpciaccess; but correct it has not been merged.

/If/ we're to change this, what would you propose the path to be for this sysfs file?

Obviously, DRM.

- Mani

Signature · 1 line
மணிவண்ணன் சதாசிவம்
#8 lore raw
Quoted 79 lines · Manivannan Sadhasivam
Quoted 78 lines · Mario Limonciello
Quoted 65 lines · Manivannan Sadhasivam
Quoted 64 lines · Mario Limonciello
Quoted 49 lines · Manivannan Sadhasivam
Quoted 36 lines · Mario Limonciello
Quoted 35 lines · Manivannan Sadhasivam
Quoted 18 lines · Mario Limonciello

From: Mario Limonciello <[email protected]>

When compiled without CONFIG_VIDEO pci_create_boot_display_file() will never create a sysfs file for boot_display. Guard the sysfs file declaration against CONFIG_VIDEO.

---
     drivers/pci/pci-sysfs.c | 2 ++
     1 file changed, 2 insertions(+)
drivers/pci/pci-sysfs.c +0 -0
index 6b1a0ae254d3a..f6540a72204d3 100644

drivers/pci/pci-sysfs.c const struct attribute_group *pcibus_groups[] = { · line 688

Quoted 9 lines
drivers/pci/pci-sysfs.c +1 -0
@@ -680,12 +680,14 @@ const struct attribute_group *pcibus_groups[] = {
     	NULL,
     };
+#ifdef CONFIG_VIDEO
     static ssize_t boot_display_show(struct device *dev, struct device_attribute *attr,
     				 char *buf)
     {
     	return sysfs_emit(buf, "1\n");
     }
     static DEVICE_ATTR_RO(boot_display);

I failed to give my comment during the offending series itself, but it is never late than never. Why are we adding non-PCI attributes under bus/pci in the first place? Though the underlying device uses PCI as a transport, only the PCI bus specific attrbutes should be placed under bus/pci and the driver/peripheral specific attrbutes should belong to the respective bus/class/device hierarchy.

Quoted 9 lines
Quoted 5 lines

Now, if other peripherals (like netdev) start adding these device specific attributes under bus/pci, it will turn out to be a mess.

- Mani

It was mostly to mirror the location of where boot_vga is, which arguably has the same issue you raise.

Yes, I agree. But 'boot_vga' has set a bad precedence IMO.

Quoted 9 lines
Quoted 2 lines

I would be incredibly surprised if there was a proposal to add a 'boot_display' attribute from netdev..

Not 'boot_display' but why not 'boot_network' or something else. I was just merely pointing out the fact that the other subsystems can start dumping device/usecase specific attributes under bus/pci.

- Mani

This is a pretty general problem that exists that attributes are first come first served. For example amdgpu adds mem_busy_percent and it has certain semantics. Now PCI core can't add that.

Quoted 6 lines
Quoted 2 lines

And if nouveau.ko wants to add the same thing they need to follow the same semantics because userspace will look for those.

But here, userspace is not yet looking for 'boot_display' isn't it? Why do you need to mirror 'boot_vga' attribute?

I have a pull request opened in libpciaccess; but correct it has not been merged.

Quoted 7 lines
Quoted 2 lines

/If/ we're to change this, what would you propose the path to be for this sysfs file?

Obviously, DRM.

- Mani

The problem is sysfs for drm is oriented around connectors. This is definitely not a connector, it's a property of the device itself.

❯ cd /sys/class/drm/
❯ ls

lrwxrwxrwx    - root 21 Jul 10:41  card0 -> ../../devices/pci0000:00/0000:00:08.1/0000:c2:00.0/drm/card0
lrwxrwxrwx    - root 21 Jul 10:41  card0-DP-1 -> ../../devices/pci0000:00/0000:00:08.1/0000:c2:00.0/drm/card0/card0-DP-1
lrwxrwxrwx    - root 21 Jul 10:41  card0-DP-2 -> ../../devices/pci0000:00/0000:00:08.1/0000:c2:00.0/drm/card0/card0-DP-2
lrwxrwxrwx    - root 21 Jul 10:41  card0-DP-3 -> ../../devices/pci0000:00/0000:00:08.1/0000:c2:00.0/drm/card0/card0-DP-3
lrwxrwxrwx    - root 21 Jul 10:41  card0-DP-4 -> ../../devices/pci0000:00/0000:00:08.1/0000:c2:00.0/drm/card0/card0-DP-4
lrwxrwxrwx    - root 21 Jul 10:41  card0-DP-5 -> ../../devices/pci0000:00/0000:00:08.1/0000:c2:00.0/drm/card0/card0-DP-5
lrwxrwxrwx    - root 21 Jul 10:41  card0-DP-6 -> ../../devices/pci0000:00/0000:00:08.1/0000:c2:00.0/drm/card0/card0-DP-6
lrwxrwxrwx    - root 21 Jul 10:41  card0-DP-7 -> ../../devices/pci0000:00/0000:00:08.1/0000:c2:00.0/drm/card0/card0-DP-7
lrwxrwxrwx    - root 21 Jul 10:41  card0-DP-8 -> ../../devices/pci0000:00/0000:00:08.1/0000:c2:00.0/drm/card0/card0-DP-8
lrwxrwxrwx    - root 21 Jul 10:41  card0-DP-9 -> ../../devices/pci0000:00/0000:00:08.1/0000:c2:00.0/drm/card0/card0-DP-9
lrwxrwxrwx    - root 21 Jul 10:41  card0-DP-10 -> ../../devices/pci0000:00/0000:00:08.1/0000:c2:00.0/drm/card0/card0-DP-10
lrwxrwxrwx    - root 21 Jul 10:41  card0-DP-11 -> ../../devices/pci0000:00/0000:00:08.1/0000:c2:00.0/drm/card0/card0-DP-11
lrwxrwxrwx    - root 21 Jul 10:41  card0-HDMI-A-1 -> ../../devices/pci0000:00/0000:00:08.1/0000:c2:00.0/drm/card0/card0-HDMI-A-1
lrwxrwxrwx    - root 21 Jul 10:41  card0-Writeback-1 -> ../../devices/pci0000:00/0000:00:08.1/0000:c2:00.0/drm/card0/card0-Writeback-1
lrwxrwxrwx    - root 21 Jul 10:41  renderD128 -> ../../devices/pci0000:00/0000:00:08.1/0000:c2:00.0/drm/renderD128

.r--r--r-- 4.1k root 21 Jul 10:41 󰡯 version
❯ cat version
drm 1.1.0 20060810
❯ cd card0
❯ ls

drwxr-xr-x    - root 21 Jul 08:14  card0-DP-1
drwxr-xr-x    - root 21 Jul 08:14  card0-DP-2
drwxr-xr-x    - root 21 Jul 08:14  card0-DP-3
drwxr-xr-x    - root 21 Jul 08:14  card0-DP-4
drwxr-xr-x    - root 21 Jul 08:14  card0-DP-5
drwxr-xr-x    - root 21 Jul 08:14  card0-DP-6
drwxr-xr-x    - root 21 Jul 08:14  card0-DP-7
drwxr-xr-x    - root 21 Jul 08:14  card0-DP-8
drwxr-xr-x    - root 21 Jul 08:14  card0-DP-9
drwxr-xr-x    - root 21 Jul 08:14  card0-DP-10
drwxr-xr-x    - root 21 Jul 08:14  card0-DP-11
drwxr-xr-x    - root 21 Jul 08:14  card0-HDMI-A-1
drwxr-xr-x    - root 21 Jul 08:14  card0-Writeback-1
lrwxrwxrwx    - root 21 Jul 08:14  device -> ../../../0000:c2:00.0
drwxr-xr-x    - root 21 Jul 08:14  power
lrwxrwxrwx    - root 21 Jul 08:14  subsystem -> ../../../../../../class/drm

.r--r--r-- 4.1k root 21 Jul 08:14 󰡯 dev
.rw-r--r-- 4.1k root 21 Jul 08:14 󰡯 uevent

If you look at the device you can see there are dozens of properties for the device at the device folder that come from all different parts of the stack that populate them.

❯ cd device
❯ ls

lrwxrwxrwx    - root 16 Jul 11:34  consumer:pci:0000:c2:00.1 -> ../../../virtual/devlink/pci:0000:c2:00.0--pci:0000:c2:00.1
lrwxrwxrwx    - root 16 Jul 11:34  driver -> ../../../../bus/pci/drivers/amdgpu
drwxr-xr-x    - root 16 Jul 11:34  drm
lrwxrwxrwx    - root 16 Jul 11:34  firmware_node -> ../../../LNXSYSTM:00/LNXSYBUS:00/PNP0A08:00/device:18/LNXVIDEO:00
drwxr-xr-x    - root 16 Jul 11:34  fw_version
drwxr-xr-x    - root 16 Jul 11:34  graphics
drwxr-xr-x    - root 16 Jul 11:34  hwmon
drwxr-xr-x    - root 16 Jul 11:34  i2c-0
drwxr-xr-x    - root 16 Jul 11:34  i2c-1
drwxr-xr-x    - root 16 Jul 11:34  i2c-2
drwxr-xr-x    - root 16 Jul 11:34  i2c-3
drwxr-xr-x    - root 16 Jul 11:34  i2c-4
drwxr-xr-x    - root 16 Jul 11:34  i2c-5
drwxr-xr-x    - root 16 Jul 11:34  i2c-6
drwxr-xr-x    - root 16 Jul 11:34  i2c-7
drwxr-xr-x    - root 16 Jul 11:34  i2c-8
drwxr-xr-x    - root 16 Jul 11:34  i2c-17
drwxr-xr-x    - root 16 Jul 11:34  i2c-18
lrwxrwxrwx    - root 16 Jul 11:34  iommu -> ../../0000:00:00.2/iommu/ivhd0
lrwxrwxrwx    - root 16 Jul 11:34  iommu_group -> ../../../../kernel/iommu_groups/19
drwxr-xr-x    - root 16 Jul 11:34  ip_discovery
drwxr-xr-x    - root 16 Jul 11:34  link
drwxr-xr-x    - root 16 Jul 11:34  msi_irqs
drwxr-xr-x    - root 16 Jul 11:34  power
lrwxrwxrwx    - root 16 Jul 11:34  subsystem -> ../../../../bus/pci

.r--r--r-- 4.1k root 16 Jul 11:34 󰡯 ari_enabled
.rw-r--r-- 4.1k root 16 Jul 11:34 󰡯 broken_parity_status .r--r--r-- 4.1k root 16 Jul 11:34 󰡯 class
.r--r--r-- 4.1k root 16 Jul 11:34 󰡯 compute_reset_mask .rw-r--r-- 4.1k root 16 Jul 11:34 󱁻 config
.r--r--r-- 4.1k root 16 Jul 11:34 󰡯 consistent_dma_mask_bits .r--r--r-- 4.1k root 16 Jul 11:34 󰡯 current_link_speed .r--r--r-- 4.1k root 16 Jul 11:34 󰡯 current_link_width .rw-r--r-- 4.1k root 16 Jul 11:34 󰡯 d3cold_allowed .r--r--r-- 4.1k root 16 Jul 11:34 󰡯 device
.r--r--r-- 4.1k root 16 Jul 11:34 󰡯 dma_mask_bits .rw-r--r-- 4.1k root 16 Jul 11:34 󰡯 driver_override .rw-r--r-- 4.1k root 16 Jul 11:34 󰡯 enable
.rw-r--r-- 4.1k root 16 Jul 11:34 󰡯 enforce_isolation .r--r--r-- 4.1k root 16 Jul 11:34 󰡯 gfx_reset_mask .r--r--r-- 4.1k root 16 Jul 11:34 󰡯 gpu_busy_percent .r--r--r-- 4.1k root 16 Jul 11:34 󰡯 gpu_metrics
.rw-rw-r-- 5.1k root 16 Jul 11:34 󰡯 hdcp_srm
.r--r--r-- 4.1k root 16 Jul 11:34 󰡯 irq
.r--r--r-- 4.1k root 16 Jul 11:34 󰡯 jpeg_reset_mask .r--r--r-- 4.1k root 16 Jul 11:34 󰡯 local_cpulist .r--r--r-- 4.1k root 16 Jul 11:34 󰡯 local_cpus
.r--r--r-- 4.1k root 16 Jul 11:34 󰡯 max_link_speed .r--r--r-- 4.1k root 16 Jul 11:34 󰡯 max_link_width .r--r--r-- 4.1k root 16 Jul 11:34 󰡯 mem_info_gtt_total .r--r--r-- 4.1k root 16 Jul 11:34 󰡯 mem_info_gtt_used .r--r--r-- 4.1k root 16 Jul 11:34 󰡯 mem_info_preempt_used .r--r--r-- 4.1k root 16 Jul 11:34 󰡯 mem_info_vis_vram_total .r--r--r-- 4.1k root 16 Jul 11:34 󰡯 mem_info_vis_vram_used .r--r--r-- 4.1k root 16 Jul 11:34 󰡯 mem_info_vram_total .r--r--r-- 4.1k root 16 Jul 11:34 󰡯 mem_info_vram_used .r--r--r-- 4.1k root 16 Jul 11:34 󰡯 modalias
.rw-r--r-- 4.1k root 16 Jul 11:34 󰡯 msi_bus
.rw-r--r-- 4.1k root 16 Jul 11:34 󰡯 numa_node
.r--r--r-- 4.1k root 16 Jul 11:34 󰡯 pcie_replay_count .rw-r--r-- 4.1k root 16 Jul 11:34 󰡯 power_dpm_force_performance_level .rw-r--r-- 4.1k root 16 Jul 11:34 󰡯 power_dpm_state .r--r--r-- 4.1k root 16 Jul 11:34 󰡯 power_state
.r--r--r-- 4.1k root 16 Jul 11:34 󰡯 pp_cur_state .r--r--r-- 4.1k root 16 Jul 11:34 󰡯 pp_dpm_dcefclk .rw-r--r-- 4.1k root 16 Jul 11:34 󰡯 pp_dpm_fclk
.rw-r--r-- 4.1k root 16 Jul 11:34 󰡯 pp_dpm_mclk
.rw-r--r-- 4.1k root 16 Jul 11:34 󰡯 pp_dpm_pcie
.rw-r--r-- 4.1k root 16 Jul 11:34 󰡯 pp_dpm_sclk
.rw-r--r-- 4.1k root 16 Jul 11:34 󰡯 pp_dpm_socclk .rw-r--r-- 4.1k root 16 Jul 11:34 󰡯 pp_force_state .r--r--r-- 4.1k root 16 Jul 11:34 󰡯 pp_num_states .rw-r--r-- 4.1k root 16 Jul 11:34 󰡯 pp_od_clk_voltage .rw-r--r-- 4.1k root 16 Jul 11:34 󰡯 pp_table
.-w--w---- 4.1k root 16 Jul 11:34 󰡯 remove
.-w------- 4.1k root 16 Jul 11:34 󰡯 rescan
.-w------- 4.1k root 16 Jul 11:34 󰡯 reset
.rw-r--r-- 4.1k root 16 Jul 11:34 󰡯 reset_method .r--r--r-- 4.1k root 16 Jul 11:34 󰡯 resource
.rw------- 8.6G root 16 Jul 11:34 󰡯 resource0
.rw-r--r-- 4.1k root 16 Jul 11:34 󰡯 resource0_resize .rw------- 8.6G root 16 Jul 11:34 󰡯 resource0_wc .rw------- 268M root 16 Jul 11:34 󰡯 resource2
.rw-r--r-- 4.1k root 16 Jul 11:34 󰡯 resource2_resize .rw------- 268M root 16 Jul 11:34 󰡯 resource2_wc .rw------- 256 root 16 Jul 11:34 󰡯 resource4
.rw------- 1.0M root 16 Jul 11:34 󰡯 resource5
.r--r--r-- 4.1k root 16 Jul 11:34 󰡯 revision
.-w------- 4.1k root 16 Jul 11:34 󰡯 run_cleaner_shader .r--r--r-- 4.1k root 16 Jul 11:34 󰡯 sdma_reset_mask .r--r--r-- 4.1k root 16 Jul 11:34 󰡯 subsystem_device .r--r--r-- 4.1k root 16 Jul 11:34 󰡯 subsystem_vendor .rw-r--r-- 4.1k root 16 Jul 11:34 󰡯 thermal_throttling_logging .rw-r--r-- 4.1k root 16 Jul 11:34 󰡯 uevent
.r--r--r-- 4.1k root 16 Jul 11:34 󰡯 vbios_version .r--r--r-- 4.1k root 16 Jul 11:34 󰡯 vcn_busy_percent .r--r--r-- 4.1k root 16 Jul 11:34 󰡯 vendor
.r--r--r-- 4.1k root 16 Jul 11:34 󰡯 vpe_reset_mask