usb: xhci: Fix debugfs bandwidth reporting

PATCH Superseded Michal Pecio 2 participants 6 messages Mar 4 – Mar 25

Subsystems (2)

USB USB xHCI driver

lore

Share & tools

lore

Michal Pecio

author

2026-03-04 10:49 UTC #0 lore raw

PATCH 1/1 usb: xhci: Fix debugfs bandwidth reporting

Replace kernel USB speed numbers with xHCI protocol IDs expected by HW. They are numerically equal up to high speed, but instead of SuperSpeed we were querying SuperSpeed+.

Gen1 hardware rejects such commands with TRB Error, which resulted in zero available bandwidth being shown.

While at that, report command failure as IO error, not zero bandwidth.

Signed-off-by Michal Pecio <[email protected]>

---
 drivers/usb/host/xhci-debugfs.c | 6 +++---
 drivers/usb/host/xhci.c         | 2 ++
 2 files changed, 5 insertions(+), 3 deletions(-)

drivers/usb/host/xhci-debugfs.c +3 -3

index c2080ee4df50..d45fa29a9691 100644
@@ -709,7 +709,7 @@ static int xhci_ss_bw_show(struct seq_file *s, void *unused)
 	int ret;
 	struct xhci_hcd		*xhci = (struct xhci_hcd *)s->private;
 
-	ret = xhci_port_bw_show(xhci, USB_SPEED_SUPER, s);
+	ret = xhci_port_bw_show(xhci, DEV_PORT_SPEED(XDEV_SS), s);
 	return ret;
 }

@@ -718,7 +718,7 @@ static int xhci_hs_bw_show(struct seq_file *s, void *unused)
 	int ret;
 	struct xhci_hcd		*xhci = (struct xhci_hcd *)s->private;
 
-	ret = xhci_port_bw_show(xhci, USB_SPEED_HIGH, s);
+	ret = xhci_port_bw_show(xhci, DEV_PORT_SPEED(XDEV_HS), s);
 	return ret;
 }

@@ -727,7 +727,7 @@ static int xhci_fs_bw_show(struct seq_file *s, void *unused)
 	int ret;
 	struct xhci_hcd		*xhci = (struct xhci_hcd *)s->private;
 
-	ret = xhci_port_bw_show(xhci, USB_SPEED_FULL, s);
+	ret = xhci_port_bw_show(xhci, DEV_PORT_SPEED(XDEV_FS), s);
 	return ret;
 }

drivers/usb/host/xhci.c +2 -0

index 6b312966c103..2f23b796ca12 100644
@@ -3277,6 +3277,8 @@ int xhci_get_port_bandwidth(struct xhci_hcd *xhci, struct xhci_container_ctx *ct
 	spin_unlock_irqrestore(&xhci->lock, flags);
 
 	wait_for_completion(cmd->completion);
+	if (cmd->status != COMP_SUCCESS)
+		ret = -EIO;
 err_out:
 	kfree(cmd->completion);
 	kfree(cmd);

Signature · 1 line

2.48.1

Mathias Nyman

2026-03-19 21:04 UTC #1 lore raw

Quoted 10 lines · Michal Pecio

Replace kernel USB speed numbers with xHCI protocol IDs expected by HW. They are numerically equal up to high speed, but instead of SuperSpeed we were querying SuperSpeed+.
Gen1 hardware rejects such commands with TRB Error, which resulted in zero available bandwidth being shown.
While at that, report command failure as IO error, not zero bandwidth.
Signed-off-by Michal Pecio <[email protected]>

Added to queue

Thanks
Mathias

Michal Pecio

author

2026-03-25 10:32 UTC #2 lore raw

Quoted 5 lines · Mathias Nyman

Quoted 4 lines · Michal Pecio
Replace kernel USB speed numbers with xHCI protocol IDs expected by HW. They are numerically equal up to high speed, but instead of SuperSpeed we were querying SuperSpeed+.

Quoted 8 lines

Quoted 6 lines
Gen1 hardware rejects such commands with TRB Error, which resulted in zero available bandwidth being shown.
While at that, report command failure as IO error, not zero bandwidth.
Signed-off-by Michal Pecio <[email protected]>
Added to queue

Hi,

Thanks for taking the patch, but can we have a last minute swap for a v2 or optionally v3?

Problem is that returning -EIO is common (the command is optional) and it upsets userspace: "grep -r" spams the console with errors, "zip -R" terminates and doesn't include remaining files, etc. So I would prefer to print an error string in this case. "Real" errors will still be returned ordinarily.

The optional v3 also renames the new directory for consistency. Technically it's a breaking change, but I believe it's permissible in debugfs, particularly for an interface only added recently.

Regards,
Michal

Michal Pecio

author

2026-03-25 10:33 UTC #3 lore raw

PATCH v2 1/1 usb: xhci: Fix debugfs bandwidth reporting

Replace kernel USB speed numbers with xHCI protocol IDs expected by HW. They are numerically equal up to high speed, but instead of SuperSpeed we were querying SuperSpeed+.

Gen1 hardware rejects such commands with TRB Error, which resulted in zero available bandwidth being shown.

While at that, report failures properly. No attempt made at "tunneling" all possible comp codes through errno, debugfs users may inspect the result through event-ring/trbs.

Signed-off-by Michal Pecio <[email protected]>

---

New in v2:
- print an error string instead of returning -EIO
- document known and sometimes unobvious error cases

 drivers/usb/host/xhci-debugfs.c | 10 +++++++---
 drivers/usb/host/xhci.c         |  9 ++++++++-
 2 files changed, 15 insertions(+), 4 deletions(-)

drivers/usb/host/xhci-debugfs.c +7 -3

index 6cc0f60da771..da528d07b815 100644
@@ -703,6 +703,10 @@ static int xhci_port_bw_show(struct xhci_hcd *xhci, u8 dev_speed,
 		seq_printf(s, "port[%d] available bw: %d%%.\n", i,
 				ctx->bytes[i]);
 err_out:
+	if (ret == -EIO) {
+		seq_puts(s, "Get Port Bandwidth failed\n");
+		ret = 0;
+	}
 	pm_runtime_put_sync(dev);
 	xhci_free_port_bw_ctx(xhci, ctx);
 	return ret;

@@ -713,7 +717,7 @@ static int xhci_ss_bw_show(struct seq_file *s, void *unused)
 	int ret;
 	struct xhci_hcd		*xhci = (struct xhci_hcd *)s->private;
 
-	ret = xhci_port_bw_show(xhci, USB_SPEED_SUPER, s);
+	ret = xhci_port_bw_show(xhci, DEV_PORT_SPEED(XDEV_SS), s);
 	return ret;
 }

@@ -722,7 +726,7 @@ static int xhci_hs_bw_show(struct seq_file *s, void *unused)
 	int ret;
 	struct xhci_hcd		*xhci = (struct xhci_hcd *)s->private;
 
-	ret = xhci_port_bw_show(xhci, USB_SPEED_HIGH, s);
+	ret = xhci_port_bw_show(xhci, DEV_PORT_SPEED(XDEV_HS), s);
 	return ret;
 }

@@ -731,7 +735,7 @@ static int xhci_fs_bw_show(struct seq_file *s, void *unused)
 	int ret;
 	struct xhci_hcd		*xhci = (struct xhci_hcd *)s->private;
 
-	ret = xhci_port_bw_show(xhci, USB_SPEED_FULL, s);
+	ret = xhci_port_bw_show(xhci, DEV_PORT_SPEED(XDEV_FS), s);
 	return ret;
 }

drivers/usb/host/xhci.c +8 -1

index e6edafdfcdb4..c86819afdede 100644
@@ -3262,7 +3262,12 @@ void xhci_reset_bandwidth(struct usb_hcd *hcd, struct usb_device *udev)
 }
 EXPORT_SYMBOL_GPL(xhci_reset_bandwidth);
 
-/* Get the available bandwidth of the ports under the xhci roothub */
+/*
+ * Get the available bandwidth of the ports under the xhci roothub.
+ * EIO means the command failed: command not implemented or unsupported
+ * speed (TRB Error), some ASMedia complete with Parameter Error when
+ * querying the root hub (slot_id = 0), or other error or timeout.
+ */
 int xhci_get_port_bandwidth(struct xhci_hcd *xhci, struct xhci_container_ctx *ctx,
 			    u8 dev_speed)
 {

@@ -3291,6 +3296,8 @@ int xhci_get_port_bandwidth(struct xhci_hcd *xhci, struct xhci_container_ctx *ct
 	spin_unlock_irqrestore(&xhci->lock, flags);
 
 	wait_for_completion(cmd->completion);
+	if (cmd->status != COMP_SUCCESS)
+		ret = -EIO;
 err_out:
 	kfree(cmd->completion);
 	kfree(cmd);

Signature · 1 line

2.48.1

Michal Pecio

author

2026-03-25 10:34 UTC #4 lore raw

PATCH v3 1/1 usb: xhci: Fix debugfs bandwidth reporting

In reply to Michal Pecio

Replace kernel USB speed numbers with xHCI protocol IDs expected by HW. They are numerically equal up to high speed, but instead of SuperSpeed we were querying SuperSpeed+.

Gen1 hardware rejects such commands with TRB Error, which resulted in zero available bandwidth being shown.

While at that, report failures properly. No attempt made at "tunneling" all possible comp codes through errno, debugfs users may inspect the result through event-ring/trbs.

Rename port_bandwidth to port-bandwidth for consistency with others.

Signed-off-by Michal Pecio <[email protected]>

---

New in v3:
- rename port_bandwidth to port-bandwidth

New in v2:
- print an error string instead of returning -EIO
- document known and sometimes unobvious error cases

 drivers/usb/host/xhci-debugfs.c | 12 ++++++++----
 drivers/usb/host/xhci.c         |  9 ++++++++-
 2 files changed, 16 insertions(+), 5 deletions(-)

drivers/usb/host/xhci-debugfs.c +8 -4

index 6cc0f60da771..55778c8cab78 100644
@@ -703,6 +703,10 @@ static int xhci_port_bw_show(struct xhci_hcd *xhci, u8 dev_speed,
 		seq_printf(s, "port[%d] available bw: %d%%.\n", i,
 				ctx->bytes[i]);
 err_out:
+	if (ret == -EIO) {
+		seq_puts(s, "Get Port Bandwidth failed\n");
+		ret = 0;
+	}
 	pm_runtime_put_sync(dev);
 	xhci_free_port_bw_ctx(xhci, ctx);
 	return ret;

@@ -713,7 +717,7 @@ static int xhci_ss_bw_show(struct seq_file *s, void *unused)
 	int ret;
 	struct xhci_hcd		*xhci = (struct xhci_hcd *)s->private;
 
-	ret = xhci_port_bw_show(xhci, USB_SPEED_SUPER, s);
+	ret = xhci_port_bw_show(xhci, DEV_PORT_SPEED(XDEV_SS), s);
 	return ret;
 }

@@ -722,7 +726,7 @@ static int xhci_hs_bw_show(struct seq_file *s, void *unused)
 	int ret;
 	struct xhci_hcd		*xhci = (struct xhci_hcd *)s->private;
 
-	ret = xhci_port_bw_show(xhci, USB_SPEED_HIGH, s);
+	ret = xhci_port_bw_show(xhci, DEV_PORT_SPEED(XDEV_HS), s);
 	return ret;
 }

@@ -731,7 +735,7 @@ static int xhci_fs_bw_show(struct seq_file *s, void *unused)
 	int ret;
 	struct xhci_hcd		*xhci = (struct xhci_hcd *)s->private;
 
-	ret = xhci_port_bw_show(xhci, USB_SPEED_FULL, s);
+	ret = xhci_port_bw_show(xhci, DEV_PORT_SPEED(XDEV_FS), s);
 	return ret;
 }

@@ -767,7 +771,7 @@ static const struct file_operations bw_fops = {
 static void xhci_debugfs_create_bandwidth(struct xhci_hcd *xhci,
 					struct dentry *parent)
 {
-	parent = debugfs_create_dir("port_bandwidth", parent);
+	parent = debugfs_create_dir("port-bandwidth", parent);
 
 	xhci_debugfs_create_files(xhci, bw_context_files,
 			  ARRAY_SIZE(bw_context_files),

drivers/usb/host/xhci.c +8 -1

index e6edafdfcdb4..c86819afdede 100644
@@ -3262,7 +3262,12 @@ void xhci_reset_bandwidth(struct usb_hcd *hcd, struct usb_device *udev)
 }
 EXPORT_SYMBOL_GPL(xhci_reset_bandwidth);
 
-/* Get the available bandwidth of the ports under the xhci roothub */
+/*
+ * Get the available bandwidth of the ports under the xhci roothub.
+ * EIO means the command failed: command not implemented or unsupported
+ * speed (TRB Error), some ASMedia complete with Parameter Error when
+ * querying the root hub (slot_id = 0), or other error or timeout.
+ */
 int xhci_get_port_bandwidth(struct xhci_hcd *xhci, struct xhci_container_ctx *ctx,
 			    u8 dev_speed)
 {

@@ -3291,6 +3296,8 @@ int xhci_get_port_bandwidth(struct xhci_hcd *xhci, struct xhci_container_ctx *ct
 	spin_unlock_irqrestore(&xhci->lock, flags);
 
 	wait_for_completion(cmd->completion);
+	if (cmd->status != COMP_SUCCESS)
+		ret = -EIO;
 err_out:
 	kfree(cmd->completion);
 	kfree(cmd);

Signature · 1 line

2.48.1

Mathias Nyman

2026-03-25 12:24 UTC #5 lore raw

In reply to Michal Pecio

Quoted 11 lines · Michal Pecio

Quoted 10 lines · Mathias Nyman
Quoted 9 lines · Michal Pecio
Replace kernel USB speed numbers with xHCI protocol IDs expected by HW. They are numerically equal up to high speed, but instead of SuperSpeed we were querying SuperSpeed+.
Gen1 hardware rejects such commands with TRB Error, which resulted in zero available bandwidth being shown.
While at that, report command failure as IO error, not zero bandwidth.

Quoted 9 lines

Quoted 3 lines
Quoted 1 line
Signed-off-by Michal Pecio <[email protected]>
Added to queue
Hi,
Thanks for taking the patch, but can we have a last minute swap for a v2 or optionally v3?

Sure, no problem.

Quoted 10 lines

Problem is that returning -EIO is common (the command is optional) and it upsets userspace: "grep -r" spams the console with errors, "zip -R" terminates and doesn't include remaining files, etc. So I would prefer to print an error string in this case. "Real" errors will still be returned ordinarily.
The optional v3 also renames the new directory for consistency. Technically it's a breaking change, but I believe it's permissible in debugfs, particularly for an interface only added recently.

lets stick with v2, avoiding possible breakage due to debugfs file rename.

Thanks
Mathias