CVE-2026-64339

In the Linux kernel, the following vulnerability has been resolved:

usb: misc: usbio: bound bulk IN response length to the received transfer

usbio_bulk_msg() copies bpkt_len = le16_to_cpu(bpkt->len) bytes out of the bulk IN buffer (usbio->rxbuf, allocated with size usbio->rxbuf_len) into the callers buffer. bpkt_len is fully controlled by the device and is only checked against ibuf_len; ibuf_len in turn is checked against usbio->txbuf_len, not against rxbuf_len:

if ((obuf_len > (usbio->txbuf_len - sizeof(*bpkt))) ||
    (ibuf_len > (usbio->txbuf_len - sizeof(*bpkt))))
	return -EMSGSIZE;

txbuf_len and rxbuf_len are taken independently from the bulk OUT and bulk IN endpoint wMaxPacketSize in usbio_probe(). A malicious or malfunctioning device that advertises a large bulk OUT endpoint and a small bulk IN endpoint (e.g. by claiming one of the quirk-free IDs such as the Lattice NX33U, 0x2ac1:0x20cb) therefore makes ibuf_len, and hence the device-supplied bpkt_len, exceed rxbuf_len. memcpy() then reads up to txbuf_len - rxbuf_len bytes past the end of the rxbuf slab object. The over-read bytes are handed back to the i2c layer and on to user space through i2c-dev, disclosing adjacent slab memory; with KASAN this is reported as a slab-out-of-bounds read.

The number of bytes actually received is already known: act equals the URB actual_length and is bounded by rxbuf_len. Reject any response that claims more payload than was received, mirroring the existing act < sizeof(*bpkt) check just above.

The control path (usbio_ctrl_msg()) is not affected: it uses a single buffer (ctrlbuf) for both directions, so its analogous copy can never leave the allocation.

Found by code review. The out-of-bounds read was confirmed under AddressSanitizer with a faithful userspace model of usbio_bulk_msg()s receive path (an rxbuf_len-sized buffer, the same act/ibuf_len/bpkt_len checks and the memcpy). A USB raw-gadget + dummy_hcd reproducer is also available.

Weakness

The product reads data past the end, or before the beginning, of the intended buffer.

Affected Software

Potential Mitigations

• Assume all input is malicious. Use an “accept known good” input validation strategy, i.e., use a list of acceptable inputs that strictly conform to specifications. Reject any input that does not strictly conform to specifications, or transform it into something that does.
• When performing input validation, consider all potentially relevant properties, including length, type of input, the full range of acceptable values, missing or extra inputs, syntax, consistency across related fields, and conformance to business rules. As an example of business rule logic, “boat” may be syntactically valid because it only contains alphanumeric characters, but it is not valid if the input is only expected to contain colors such as “red” or “blue.”
• Do not rely exclusively on looking for malicious or malformed inputs. This is likely to miss at least one undesirable input, especially if the code’s environment changes. This can give attackers enough room to bypass the intended validation. However, denylists can be useful for detecting potential attacks or determining which inputs are so malformed that they should be rejected outright.
• To reduce the likelihood of introducing an out-of-bounds read, ensure that you validate and ensure correct calculations for any length argument, buffer size calculation, or offset. Be especially careful of relying on a sentinel (i.e. special character such as NUL) in untrusted inputs.

Related Attack Patterns

• https://cwe.mitre.org/data/definitions/540.html

References

• https://git.kernel.org/stable/c/48394f94211cf8fe0ea8604fc441633abf90fc94
• https://git.kernel.org/stable/c/8c6314489550fa81d41723a0ff33f655b5b6c7b6
• https://git.kernel.org/stable/c/fc1b546973c1442d5b947fcdd03581f20ecc5bd2