CVE-2023-52886

In the Linux kernel, the following vulnerability has been resolved:<br /> <br /> USB: core: Fix race by not overwriting udev-&gt;descriptor in hub_port_init()<br /> <br /> Syzbot reported an out-of-bounds read in sysfs.c:read_descriptors():<br /> <br /> BUG: KASAN: slab-out-of-bounds in read_descriptors+0x263/0x280 drivers/usb/core/sysfs.c:883<br /> Read of size 8 at addr ffff88801e78b8c8 by task udevd/5011<br /> <br /> CPU: 0 PID: 5011 Comm: udevd Not tainted 6.4.0-rc6-syzkaller-00195-g40f71e7cd3c6 #0<br /> Hardware name: Google Google Compute Engine/Google Compute Engine, BIOS Google 05/27/2023<br /> Call Trace:<br /> <br /> __dump_stack lib/dump_stack.c:88 [inline]<br /> dump_stack_lvl+0xd9/0x150 lib/dump_stack.c:106<br /> print_address_description.constprop.0+0x2c/0x3c0 mm/kasan/report.c:351<br /> print_report mm/kasan/report.c:462 [inline]<br /> kasan_report+0x11c/0x130 mm/kasan/report.c:572<br /> read_descriptors+0x263/0x280 drivers/usb/core/sysfs.c:883<br /> ...<br /> Allocated by task 758:<br /> ...<br /> __do_kmalloc_node mm/slab_common.c:966 [inline]<br /> __kmalloc+0x5e/0x190 mm/slab_common.c:979<br /> kmalloc include/linux/slab.h:563 [inline]<br /> kzalloc include/linux/slab.h:680 [inline]<br /> usb_get_configuration+0x1f7/0x5170 drivers/usb/core/config.c:887<br /> usb_enumerate_device drivers/usb/core/hub.c:2407 [inline]<br /> usb_new_device+0x12b0/0x19d0 drivers/usb/core/hub.c:2545<br /> <br /> As analyzed by Khazhy Kumykov, the cause of this bug is a race between<br /> read_descriptors() and hub_port_init(): The first routine uses a field<br /> in udev-&gt;descriptor, not expecting it to change, while the second<br /> overwrites it.<br /> <br /> Prior to commit 45bf39f8df7f ("USB: core: Don&amp;#39;t hold device lock while<br /> reading the "descriptors" sysfs file") this race couldn&amp;#39;t occur,<br /> because the routines were mutually exclusive thanks to the device<br /> locking. Removing that locking from read_descriptors() exposed it to<br /> the race.<br /> <br /> The best way to fix the bug is to keep hub_port_init() from changing<br /> udev-&gt;descriptor once udev has been initialized and registered.<br /> Drivers expect the descriptors stored in the kernel to be immutable;<br /> we should not undermine this expectation. In fact, this change should<br /> have been made long ago.<br /> <br /> So now hub_port_init() will take an additional argument, specifying a<br /> buffer in which to store the device descriptor it reads. (If udev has<br /> not yet been initialized, the buffer pointer will be NULL and then<br /> hub_port_init() will store the device descriptor in udev as before.)<br /> This eliminates the data race responsible for the out-of-bounds read.<br /> <br /> The changes to hub_port_init() appear more extensive than they really<br /> are, because of indentation changes resulting from an attempt to avoid<br /> writing to other parts of the usb_device structure after it has been<br /> initialized. Similar changes should be made to the code that reads<br /> the BOS descriptor, but that can be handled in a separate patch later<br /> on. This patch is sufficient to fix the bug found by syzbot.