Kirill A. Korinsky <kirill@korins.ky> wrote:

> my update libfido accidently reverted a hardcoded USB device metadata which
> leads broken pledge in, for example, chromium.

I'm more willing to allow pledge to reach ioctl USB_GET_DEVICEINFO, if
there are some incremental safety improvements made to the kernel code
that is reached.

I think usbd_fill_deviceinfo() should return int, look carefully at
what it is going to provide to userland, and return an error if the
usb configuration is not sane.

But first, I think we should purge old data in the structure.

Index: usb_subr.c
===================================================================
RCS file: /cvs/src/sys/dev/usb/usb_subr.c,v
diff -u -p -u -r1.167 usb_subr.c
--- usb_subr.c	15 Aug 2026 22:06:40 -0000	1.167
+++ usb_subr.c	7 Sep 2026 18:36:54 -0000
@@ -1298,6 +1298,7 @@ usbd_fill_deviceinfo(struct usbd_device 
 	struct usbd_port *p;
 	int i;
 
+	memset(di, 0, sizeof(*di));
 	di->udi_bus = dev->bus->usbctl->dv_unit;
 	di->udi_addr = dev->address;
 	strlcpy(di->udi_vendor, dev->vendor, sizeof(di->udi_vendor));