commit 2ef47001b3ee3ded579b7532ebdcf8680e4d8c54 upstream. The USB kerneldoc says that the actual_length field "is read in non-iso completion functions", but the usbfs driver uses it for all URB types in processcompl(). Since not all of the host controller drivers set actual_length for isochronous URBs, programs using usbfs with some host controllers don't work properly. For example, Minas reports that a USB camera controlled by libusb doesn't work properly with a dwc2 controller. It doesn't seem worthwhile to change the HCDs and the documentation, since the in-kernel USB class drivers evidently don't rely on actual_length for isochronous transfers. The easiest solution is for usbfs to calculate the actual_length value for itself, by adding up the lengths of the individual packets in an isochronous transfer. Signed-off-by: Alan Stern <stern@rowland.harvard.edu> CC: Minas Harutyunyan <Minas.Harutyunyan@synopsys.com> Reported-and-tested-by: wlf <wulf@rock-chips.com> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Taking the uurb->buffer_length userspace passes in as a maximum for the
actual urbs transfer_buffer_length causes 2 serious issues:
1) It breaks isochronous support for all userspace apps using libusb,
as existing libusb versions pass in 0 for uurb->buffer_length,
relying on the kernel using the lenghts of the usbdevfs_iso_packet_desc
descriptors passed in added together as buffer length.
This for example causes redirection of USB audio and Webcam's into
virtual machines using qemu-kvm to no longer work. This is a userspace
ABI break and as such must be reverted.
Note that the original commit does not protect other users / the
kernels memory, it only stops the userspace process making the call
from shooting itself in the foot.
2) It may cause the kernel to program host controllers to DMA over random
memory. Just as the devio code used to only look at the iso_packet_desc
lenghts, the host drivers do the same, relying on the submitter of the
urbs to make sure the entire buffer is large enough and not checking
transfer_buffer_length.
But the "USB: devio: Don't corrupt user memory" commit now takes the
userspace provided uurb->buffer_length for the buffer-size while copying
over the user-provided iso_packet_desc lengths 1:1, allowing the user
to specify a small buffer size while programming the host controller to
dma a lot more data.
(Atleast the ohci, uhci, xhci and fhci drivers do not check
transfer_buffer_length for isoc transfers.)
This reverts commit fa1ed74eb1c2 ("USB: devio: Don't corrupt user memory")
fixing both these issues.
Cc: Dan Carpenter <dan.carpenter@oracle.com>
Cc: stable@vger.kernel.org
Signed-off-by: Hans de Goede <hdegoede@redhat.com>
Acked-by: Alan Stern <stern@rowland.harvard.edu>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
The user buffer has "uurb->buffer_length" bytes. If the kernel has more information than that, we should truncate it instead of writing past the end of the user's buffer. I added a WARN_ONCE() to help the user debug the issue. Reported-by: Alan Stern <stern@rowland.harvard.edu> Cc: stable <stable@vger.kernel.org> Signed-off-by: Dan Carpenter <dan.carpenter@oracle.com> Acked-by: Alan Stern <stern@rowland.harvard.edu> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
There used to be an integer overflow check in proc_do_submiturb() but
we removed it. It turns out that it's still required. The
uurb->buffer_length variable is a signed integer and it's controlled by
the user. It can lead to an integer overflow when we do:
num_sgs = DIV_ROUND_UP(uurb->buffer_length, USB_SG_SIZE);
If we strip away the macro then that line looks like this:
num_sgs = (uurb->buffer_length + USB_SG_SIZE - 1) / USB_SG_SIZE;
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
It's the first addition which can overflow.
Fixes: 1129d270cbfb ("USB: Increase usbfs transfer limit")
Cc: stable <stable@vger.kernel.org>
Signed-off-by: Dan Carpenter <dan.carpenter@oracle.com>
Acked-by: Alan Stern <stern@rowland.harvard.edu>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
vm_operations_struct are not supposed to change at runtime. All functions working with const vm_operations_struct. So mark the non-const structs as const. Signed-off-by: Arvind Yadav <arvind.yadav.cs@gmail.com> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
While running reboot tests w/ a specific set of USB devices (and slub_debug enabled), I found that once every few hours my device would be crashed with a stack that looked like this: [ 14.012445] BUG: spinlock bad magic on CPU#0, modprobe/2091 [ 14.012460] lock: 0xffffffc0cb055978, .magic: ffffffc0, .owner: cryption contexts: %lu/%lu [ 14.012460] /1025536097, .owner_cpu: 0 [ 14.012466] CPU: 0 PID: 2091 Comm: modprobe Not tainted 4.4.79 #352 [ 14.012468] Hardware name: Google Kevin (DT) [ 14.012471] Call trace: [ 14.012483] [<....>] dump_backtrace+0x0/0x160 [ 14.012487] [<....>] show_stack+0x20/0x28 [ 14.012494] [<....>] dump_stack+0xb4/0xf0 [ 14.012500] [<....>] spin_dump+0x8c/0x98 [ 14.012504] [<....>] spin_bug+0x30/0x3c [ 14.012508] [<....>] do_raw_spin_lock+0x40/0x164 [ 14.012515] [<....>] _raw_spin_lock_irqsave+0x64/0x74 [ 14.012521] [<....>] __wake_up+0x2c/0x60 [ 14.012528] [<....>] async_completed+0x2d0/0x300 [ 14.012534] [<....>] __usb_hcd_giveback_urb+0xc4/0x138 [ 14.012538] [<....>] usb_hcd_giveback_urb+0x54/0xf0 [ 14.012544] [<....>] xhci_irq+0x1314/0x1348 [ 14.012548] [<....>] usb_hcd_irq+0x40/0x50 [ 14.012553] [<....>] handle_irq_event_percpu+0x1b4/0x3f0 [ 14.012556] [<....>] handle_irq_event+0x4c/0x7c [ 14.012561] [<....>] handle_fasteoi_irq+0x158/0x1c8 [ 14.012564] [<....>] generic_handle_irq+0x30/0x44 [ 14.012568] [<....>] __handle_domain_irq+0x90/0xbc [ 14.012572] [<....>] gic_handle_irq+0xcc/0x18c Investigation using kgdb() found that the wait queue that was passed into wake_up() had been freed (it was filled with slub_debug poison). I analyzed and instrumented the code and reproduced. My current belief is that this is happening: 1. async_completed() is called (from IRQ). Moves "as" onto the completed list. 2. On another CPU, proc_reapurbnonblock_compat() calls async_getcompleted(). Blocks on spinlock. 3. async_completed() releases the lock; keeps running; gets blocked midway through wake_up(). 4. proc_reapurbnonblock_compat() => async_getcompleted() gets the lock; removes "as" from completed list and frees it. 5. usbdev_release() is called. Frees "ps". 6. async_completed() finally continues running wake_up(). ...but wake_up() has a pointer to the freed "ps". The instrumentation that led me to believe this was based on adding some trace_printk() calls in a select few functions and then using kdb's "ftdump" at crash time. The trace follows (NOTE: in the trace below I cheated a little bit and added a udelay(1000) in async_completed() after releasing the spinlock because I wanted it to trigger quicker): <...>-2104 0d.h2 13759034us!: async_completed at start: as=ffffffc0cc638200 mtpd-2055 3.... 13759356us : async_getcompleted before spin_lock_irqsave mtpd-2055 3d..1 13759362us : async_getcompleted after list_del_init: as=ffffffc0cc638200 mtpd-2055 3.... 13759371us+: proc_reapurbnonblock_compat: free_async(ffffffc0cc638200) mtpd-2055 3.... 13759422us+: async_getcompleted before spin_lock_irqsave mtpd-2055 3.... 13759479us : usbdev_release at start: ps=ffffffc0cc042080 mtpd-2055 3.... 13759487us : async_getcompleted before spin_lock_irqsave mtpd-2055 3.... 13759497us!: usbdev_release after kfree(ps): ps=ffffffc0cc042080 <...>-2104 0d.h2 13760294us : async_completed before wake_up(): as=ffffffc0cc638200 To fix this problem we can just move the wake_up() under the ps->lock. There should be no issues there that I'm aware of. Signed-off-by: Douglas Anderson <dianders@chromium.org> Acked-by: Alan Stern <stern@rowland.harvard.edu> Cc: stable <stable@vger.kernel.org> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Pull misc compat stuff updates from Al Viro:
"This part is basically untangling various compat stuff. Compat
syscalls moved to their native counterparts, getting rid of quite a
bit of double-copying and/or set_fs() uses. A lot of field-by-field
copyin/copyout killed off.
- kernel/compat.c is much closer to containing just the
copyin/copyout of compat structs. Not all compat syscalls are gone
from it yet, but it's getting there.
- ipc/compat_mq.c killed off completely.
- block/compat_ioctl.c cleaned up; floppy compat ioctls moved to
drivers/block/floppy.c where they belong. Yes, there are several
drivers that implement some of the same ioctls. Some are m68k and
one is 32bit-only pmac. drivers/block/floppy.c is the only one in
that bunch that can be built on biarch"
* 'misc.compat' of git://git.kernel.org/pub/scm/linux/kernel/git/viro/vfs:
mqueue: move compat syscalls to native ones
usbdevfs: get rid of field-by-field copyin
compat_hdio_ioctl: get rid of set_fs()
take floppy compat ioctls to sodding floppy.c
ipmi: get rid of field-by-field __get_user()
ipmi: get COMPAT_IPMICTL_RECEIVE_MSG in sync with the native one
rt_sigtimedwait(): move compat to native
select: switch compat_{get,put}_fd_set() to compat_{get,put}_bitmap()
put_compat_rusage(): switch to copy_to_user()
sigpending(): move compat to native
getrlimit()/setrlimit(): move compat to native
times(2): move compat to native
compat_{get,put}_bitmap(): use unsafe_{get,put}_user()
fb_get_fscreeninfo(): don't bother with do_fb_ioctl()
do_sigaltstack(): lift copying to/from userland into callers
take compat_sys_old_getrlimit() to native syscall
trim __ARCH_WANT_SYS_OLD_GETRLIMIT
Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
The usbfs interface does not provide any way for the user to learn the speed at which a device is connected. The current API includes a USBDEVFS_CONNECTINFO ioctl, but all it provides is the device's address and a one-bit value indicating whether the connection is low speed. That may have sufficed in the era of USB-1.1, but it isn't good enough today. This patch introduces a new ioctl, USBDEVFS_GET_SPEED, which returns a numeric value indicating the speed of the connection: unknown, low, full, high, wireless, super, or super-plus. Similar information (not exactly the same) is available through sysfs, but it seems reasonable to provide the actual value in usbfs. Signed-off-by: Alan Stern <stern@rowland.harvard.edu> Reported-by: Reinhard Huck <reinhard.huck@thesycon.de> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Format specifier %p can leak kernel addresses while not valuing the kptr_restrict system settings. When kptr_restrict is set to (1), kernel pointers printed using the %pK format specifier will be replaced with Zeros. Debugging Note : &pK prints only Zeros as address. If you need actual address information, write 0 to kptr_restrict. echo 0 > /proc/sys/kernel/kptr_restrict [Found by poking around in a random vendor kernel tree, it would be nice if someone would actually send these types of patches upstream - gkh] Signed-off-by: Vamsi Krishna Samavedam <vskrishn@codeaurora.org> Cc: stable <stable@vger.kernel.org> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>