linux-kernel - Highly critical bug in XHCI Controller

lists.openwall.net		lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC
Open Source and information security mailing list archives

Hash Suite: Windows password security audit tool. GUI, reports in PDF.

[<prev] [next>] [thread-next>] [day] [month] [year] [list]

Message-ID: <3905c1c88695e0ffcfabf700c06dd7223decef8d.camel@sundtek.de>
Date: Sun, 17 Nov 2024 15:33:28 +0800
From: Markus Rechberger <linuxusb.ml@...dtek.de>
To: linux-usb@...r.kernel.org, linux-kernel@...r.kernel.org
Subject: Highly critical bug in XHCI Controller

Hi,

the issue was first reported at vdr-portal.de
https://www-vdr--portal-de.translate.goog/forum/index.php?thread/136541-empfehlung-dvb-s2-tuner-oder-satip/&postID=1376196&_x_tr_sl=de&_x_tr_tl=en&_x_tr_hl=de&_x_tr_pto=wapp#post1376196

we've got around a highly critical bug in the xhci driver.

https://sundtek.de/support/uxvd32.txt

In xhci.c

The bug is still active in Mainline:
https://github.com/torvalds/linux/blob/master/drivers/usb/host/xhci.c#L2382

static int xhci_check_bw_table(struct xhci_hcd *xhci,
        struct xhci_virt_device *virt_dev,
        int old_active_eps)

bw_table can end up with a NULL pointer.

This problem will lead to a complete kernel crash, rendering the entire
system unusable without any access to the actual linux system.

How to trigger the problem?
Short D+/D- or pull them to ground on a USB device while connecting the
device.

The problem can happen due to following cases:
* a device is getting suddenly disconnected during enumeration
* a faulty cable
* a faulty device 
* a malicious device triggers this issue on purpose
* if there are electrical issues during connecting a device.

A quick hotfix would be to check if bw_table is NULL in
xhci_check_bw_table, however the check should be performed earlier - in
the area where bw_table is supposed to be assigned.

Best Regards,
Markus