[ Upstream commit 508595515f ]
In some cases the "Allocate & copy" block in ffs_epfile_io() is not
executed. Consequently, in such a case ffs_alloc_buffer() is never called
and struct ffs_io_data is not initialized properly. This in turn leads to
problems when ffs_free_buffer() is called at the end of ffs_epfile_io().
This patch uses kzalloc() instead of kmalloc() in the aio case and memset()
in non-aio case to properly initialize struct ffs_io_data.
Signed-off-by: Andrzej Pietrasiewicz <andrzej.p@collabora.com>
Signed-off-by: Felipe Balbi <felipe.balbi@linux.intel.com>
Signed-off-by: Sasha Levin <sashal@kernel.org>
commit d29fcf7078 upstream.
On spin lock release in rx_submit, gether_disconnect get a chance to
run, it makes port_usb NULL, rx_submit access NULL port USB, hence null
pointer crash.
Fixed by releasing the lock in rx_submit after port_usb is used.
Fixes: 2b3d942c48 ("usb ethernet gadget: split out network core")
Cc: <stable@vger.kernel.org>
Signed-off-by: Kiruthika Varadarajan <Kiruthika.Varadarajan@harman.com>
Signed-off-by: Felipe Balbi <felipe.balbi@linux.intel.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
[ Upstream commit 54f64d5c98 ]
Since the 5.0 merge window opened, I've been seeing frequent
crashes on suspend and reboot with the trace:
[ 36.911170] Unable to handle kernel paging request at virtual address ffffff801153d660
[ 36.912769] Unable to handle kernel paging request at virtual address ffffff800004b564
...
[ 36.950666] Call trace:
[ 36.950670] queued_spin_lock_slowpath+0x1cc/0x2c8
[ 36.950681] _raw_spin_lock_irqsave+0x64/0x78
[ 36.950692] complete+0x28/0x70
[ 36.950703] ffs_epfile_io_complete+0x3c/0x50
[ 36.950713] usb_gadget_giveback_request+0x34/0x108
[ 36.950721] dwc3_gadget_giveback+0x50/0x68
[ 36.950723] dwc3_thread_interrupt+0x358/0x1488
[ 36.950731] irq_thread_fn+0x30/0x88
[ 36.950734] irq_thread+0x114/0x1b0
[ 36.950739] kthread+0x104/0x130
[ 36.950747] ret_from_fork+0x10/0x1c
I isolated this down to in ffs_epfile_io():
https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git/tree/drivers/usb/gadget/function/f_fs.c#n1065
Where the completion done is setup on the stack:
DECLARE_COMPLETION_ONSTACK(done);
Then later we setup a request and queue it, and wait for it:
if (unlikely(wait_for_completion_interruptible(&done))) {
/*
* To avoid race condition with ffs_epfile_io_complete,
* dequeue the request first then check
* status. usb_ep_dequeue API should guarantee no race
* condition with req->complete callback.
*/
usb_ep_dequeue(ep->ep, req);
interrupted = ep->status < 0;
}
The problem is, that we end up being interrupted, dequeue the
request, and exit.
But then the irq triggers and we try calling complete() on the
context pointer which points to now random stack space, which
results in the panic.
Alan Stern pointed out there is a bug here, in that the snippet
above "assumes that usb_ep_dequeue() waits until the request has
been completed." And that:
wait_for_completion(&done);
Is needed right after the usb_ep_dequeue().
Thus this patch implements that change. With it I no longer see
the crashes on suspend or reboot.
This issue seems to have been uncovered by behavioral changes in
the dwc3 driver in commit fec9095bde ("usb: dwc3: gadget:
remove wait_end_transfer").
Cc: Alan Stern <stern@rowland.harvard.edu>
Cc: Felipe Balbi <balbi@kernel.org>
Cc: Zeng Tao <prime.zeng@hisilicon.com>
Cc: Jack Pham <jackp@codeaurora.org>
Cc: Thinh Nguyen <thinh.nguyen@synopsys.com>
Cc: Chen Yu <chenyu56@huawei.com>
Cc: Jerry Zhang <zhangjerry@google.com>
Cc: Lars-Peter Clausen <lars@metafoo.de>
Cc: Vincent Pelletier <plr.vincent@gmail.com>
Cc: Andrzej Pietrasiewicz <andrzej.p@samsung.com>
Cc: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Cc: Linux USB List <linux-usb@vger.kernel.org>
Suggested-by: Alan Stern <stern@rowland.harvard.edu>
Signed-off-by: John Stultz <john.stultz@linaro.org>
Signed-off-by: Felipe Balbi <felipe.balbi@linux.intel.com>
Signed-off-by: Sasha Levin <sashal@kernel.org>
commit 072684e8c5 upstream.
In f_hidg_write() the write_spinlock is acquired before calling
usb_ep_queue() which causes a deadlock when dummy_hcd is being used.
This is because dummy_queue() callbacks into f_hidg_req_complete() which
tries to acquire the same spinlock. This is (part of) the backtrace when
the deadlock occurs:
0xffffffffc06b1410 in f_hidg_req_complete
0xffffffffc06a590a in usb_gadget_giveback_request
0xffffffffc06cfff2 in dummy_queue
0xffffffffc06a4b96 in usb_ep_queue
0xffffffffc06b1eb6 in f_hidg_write
0xffffffff8127730b in __vfs_write
0xffffffff812774d1 in vfs_write
0xffffffff81277725 in SYSC_write
Fix this by releasing the write_spinlock before calling usb_ep_queue()
Reviewed-by: James Bottomley <James.Bottomley@HansenPartnership.com>
Tested-by: James Bottomley <James.Bottomley@HansenPartnership.com>
Cc: stable@vger.kernel.org # 4.11+
Fixes: 749494b6bd ("usb: gadget: f_hid: fix: Move IN request allocation to set_alt()")
Signed-off-by: Radoslav Gerganov <rgerganov@vmware.com>
Signed-off-by: Felipe Balbi <felipe.balbi@linux.intel.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
[ Upstream commit df28169e15 ]
The source_sink_alloc_func() function is supposed to return error
pointers on error. The function is called from usb_get_function() which
doesn't check for NULL returns so it would result in an Oops.
Of course, in the current kernel, small allocations always succeed so
this doesn't affect runtime.
Signed-off-by: Dan Carpenter <dan.carpenter@oracle.com>
Signed-off-by: Felipe Balbi <felipe.balbi@linux.intel.com>
Signed-off-by: Sasha Levin <sashal@kernel.org>
commit 9ae24af366 upstream.
num can be indirectly controlled by user-space, hence leading to
a potential exploitation of the Spectre variant 1 vulnerability.
This issue was detected with the help of Smatch:
drivers/usb/gadget/function/f_mass_storage.c:3177 fsg_lun_make() warn:
potential spectre issue 'fsg_opts->common->luns' [r] (local cap)
Fix this by sanitizing num before using it to index
fsg_opts->common->luns
Notice that given that speculation windows are large, the policy is
to kill the speculation on the first load and not worry if it can be
completed with a dependent load/store [1].
[1] https://marc.info/?l=linux-kernel&m=152449131114778&w=2
Cc: stable@vger.kernel.org
Signed-off-by: Gustavo A. R. Silva <gustavo@embeddedor.com>
Acked-by: Felipe Balbi <felipe.balbi@intel.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
commit daa35bd956 upstream.
When the gadget serial device has no associated TTY, do not pass any
received data into the TTY layer for processing; simply drop it instead.
This prevents the TTY layer from calling back into the gadget serial
driver, which will then crash in e.g. gs_write_room() due to lack of
gadget serial device to TTY association (i.e. a NULL pointer dereference).
Signed-off-by: Stephen Warren <swarren@nvidia.com>
Signed-off-by: Felipe Balbi <felipe.balbi@linux.intel.com>
Signed-off-by: Krzysztof Kozlowski <krzk@kernel.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
[ Upstream commit eec24f2a0d ]
The list [1] of commits doing endianness fixes in USB subsystem is long
due to below quote from USB spec Revision 2.0 from April 27, 2000:
------------
8.1 Byte/Bit Ordering
Multiple byte fields in standard descriptors, requests, and responses
are interpreted as and moved over the bus in little-endian order, i.e.
LSB to MSB.
------------
This commit belongs to the same family.
[1] Example of endianness fixes in USB subsystem:
commit 14e1d56cbe ("usb: gadget: f_uac2: endianness fixes.")
commit 42370b8211 ("usb: gadget: f_uac1: endianness fixes.")
commit 63afd5cc78 ("USB: chaoskey: fix Alea quirk on big-endian hosts")
commit 74098c4ac7 ("usb: gadget: acm: fix endianness in notifications")
commit cdd7928df0 ("ACM gadget: fix endianness in notifications")
commit 323ece54e0 ("cdc-wdm: fix endianness bug in debug statements")
commit e102609f10 ("usb: gadget: uvc: Fix endianness mismatches")
list goes on
Fixes: 132fcb4608 ("usb: gadget: Add Audio Class 2.0 Driver")
Signed-off-by: Eugeniu Rosca <erosca@de.adit-jv.com>
Reviewed-by: Ruslan Bilovol <ruslan.bilovol@gmail.com>
Signed-off-by: Felipe Balbi <felipe.balbi@linux.intel.com>
Signed-off-by: Sasha Levin <alexander.levin@microsoft.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
commit 4d644abf25 upstream.
Commit 1b9ba000 ("Allow function drivers to pause control
transfers") states that USB_GADGET_DELAYED_STATUS is only
supported if data phase is 0 bytes.
It seems that when the length is not 0 bytes, there is no
need to explicitly delay the data stage since the transfer
is not completed until the user responds. However, when the
length is 0, there is no data stage and the transfer is
finished once setup() returns, hence there is a need to
explicitly delay completion.
This manifests as the following bugs:
Prior to 946ef68ad4 ('Let setup() return
USB_GADGET_DELAYED_STATUS'), when setup is 0 bytes, ffs
would require user to queue a 0 byte request in order to
clear setup state. However, that 0 byte request was actually
not needed and would hang and cause errors in other setup
requests.
After the above commit, 0 byte setups work since the gadget
now accepts empty queues to ep0 to clear the delay, but all
other setups hang.
Fixes: 946ef68ad4 ("Let setup() return USB_GADGET_DELAYED_STATUS")
Signed-off-by: Jerry Zhang <zhangjerry@google.com>
Cc: stable <stable@vger.kernel.org>
Acked-by: Felipe Balbi <felipe.balbi@linux.intel.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
[ Upstream commit 4058ebf33c ]
When using a AIO read() operation on the function FS gadget driver a URB is
submitted asynchronously and on URB completion the received data is copied
to the userspace buffer associated with the read operation.
This is done from a kernel worker thread invoking copy_to_user() (through
copy_to_iter()). And while the user space process memory is made available
to the kernel thread using use_mm(), some architecture require in addition
to this that the operation runs with USER_DS set. Otherwise the userspace
memory access will fail.
For example on ARM64 with Privileged Access Never (PAN) and User Access
Override (UAO) enabled the following crash occurs.
Internal error: Accessing user space memory with fs=KERNEL_DS: 9600004f [#1] SMP
Modules linked in:
CPU: 2 PID: 1636 Comm: kworker/2:1 Not tainted 4.9.0-04081-g8ab2dfb-dirty #487
Hardware name: ZynqMP ZCU102 Rev1.0 (DT)
Workqueue: events ffs_user_copy_worker
task: ffffffc87afc8080 task.stack: ffffffc87a00c000
PC is at __arch_copy_to_user+0x190/0x220
LR is at copy_to_iter+0x78/0x3c8
[...]
[<ffffff800847b790>] __arch_copy_to_user+0x190/0x220
[<ffffff80086f25d8>] ffs_user_copy_worker+0x70/0x130
[<ffffff80080b8c64>] process_one_work+0x1dc/0x460
[<ffffff80080b8f38>] worker_thread+0x50/0x4b0
[<ffffff80080bf5a0>] kthread+0xd8/0xf0
[<ffffff8008083680>] ret_from_fork+0x10/0x50
Address this by placing a set_fs(USER_DS) before of the copy operation
and revert it again once the copy operation has finished.
This patch is analogous to commit d7ffde35e3 ("vhost: use USER_DS in
vhost_worker thread") which addresses the same underlying issue.
Signed-off-by: Lars-Peter Clausen <lars@metafoo.de>
Signed-off-by: Felipe Balbi <felipe.balbi@linux.intel.com>
Signed-off-by: Sasha Levin <alexander.levin@microsoft.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
[ Upstream commit 946ef68ad4 ]
Some UDC drivers (like the DWC3) expect that the response to a setup()
request is queued from within the setup function itself so that it is
available as soon as setup() has completed.
Upon receiving a setup request the function fs driver creates an event that
is made available to userspace. And only once userspace has acknowledged
that event the response to the setup request is queued.
So it violates the requirement of those UDC drivers and random failures can
be observed. This is basically a race condition and if userspace is able to
read the event and queue the response fast enough all is good. But if it is
not, for example because other processes are currently scheduled to run,
the USB host that sent the setup request will observe an error.
To avoid this the gadget framework provides the USB_GADGET_DELAYED_STATUS
return code. If a setup() callback returns this value the UDC driver is
aware that response is not yet available and can uses the appropriate
methods to handle this case.
Since in the case of function fs the response will never be available when
the setup() function returns make sure that this status code is used.
This fixed random occasional failures that were previously observed on a
DWC3 based system under high system load.
Signed-off-by: Lars-Peter Clausen <lars@metafoo.de>
Signed-off-by: Felipe Balbi <felipe.balbi@linux.intel.com>
Signed-off-by: Sasha Levin <alexander.levin@microsoft.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
[ Upstream commit 8813a59ed8 ]
If there are multiple functions associated with a configuration, then
the UAC2 interfaces may not start at zero. Set the correct first
interface number in the association descriptor so that the audio
interfaces are enumerated correctly in this case.
Reviewed-by: Krzysztof Opasiak <k.opasiak@samsung.com>
Signed-off-by: John Keeping <john@metanate.com>
Signed-off-by: Felipe Balbi <felipe.balbi@linux.intel.com>
Signed-off-by: Sasha Levin <alexander.levin@microsoft.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
commit 7fafcfdf63 upstream.
It looks like there is a possibility of a double-free vulnerability on an
error path of the f_midi_set_alt function in the f_midi driver. If the
path is feasible then free_ep_req gets called twice:
req->complete = f_midi_complete;
err = usb_ep_queue(midi->out_ep, req, GFP_ATOMIC);
=> ...
usb_gadget_giveback_request
=>
f_midi_complete (CALLBACK)
(inside f_midi_complete, for various cases of status)
free_ep_req(ep, req); // first kfree
if (err) {
ERROR(midi, "%s: couldn't enqueue request: %d\n",
midi->out_ep->name, err);
free_ep_req(midi->out_ep, req); // second kfree
return err;
}
The double-free possibility was introduced with commit ad0d1a058e
("usb: gadget: f_midi: fix leak on failed to enqueue out requests").
Found by MOXCAFE tool.
Signed-off-by: Tuba Yavuz <tuba@ece.ufl.edu>
Fixes: ad0d1a058e ("usb: gadget: f_midi: fix leak on failed to enqueue out requests")
Acked-by: Felipe Balbi <felipe.balbi@linux.intel.com>
Cc: stable <stable@vger.kernel.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
commit 749494b6bd upstream.
Since commit: ba1582f222 ("usb: gadget: f_hid: use alloc_ep_req()")
we cannot allocate any requests in bind() as we check if we should
align request buffer based on endpoint descriptor which is assigned
in set_alt().
Allocating request in bind() function causes a NULL pointer
dereference.
This commit moves allocation of IN request from bind() to set_alt()
to prevent this issue.
Fixes: ba1582f222 ("usb: gadget: f_hid: use alloc_ep_req()")
Cc: stable@vger.kernel.org
Tested-by: David Lechner <david@lechnology.com>
Signed-off-by: Krzysztof Opasiak <k.opasiak@samsung.com>
Signed-off-by: Felipe Balbi <felipe.balbi@linux.intel.com>
Cc: Bin Liu <b-liu@ti.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
commit 1a087f0321 upstream.
When I debug a kernel crash issue in funcitonfs, found ffs_data.ref
overflowed, While functionfs is unmounting, ffs_data is put twice.
Commit 43938613c6 ("drivers, usb: convert ffs_data.ref from atomic_t to
refcount_t") can avoid refcount overflow, but that is risk some situations.
So no need put ffs data in ffs_fs_kill_sb, already put in ffs_data_closed.
The issue can be reproduced in Mediatek mt6763 SoC, ffs for ADB device.
KASAN enabled configuration reports use-after-free errro.
BUG: KASAN: use-after-free in refcount_dec_and_test+0x14/0xe0 at addr ffffffc0579386a0
Read of size 4 by task umount/4650
====================================================
BUG kmalloc-512 (Tainted: P W O ): kasan: bad access detected
-----------------------------------------------------------------------------
INFO: Allocated in ffs_fs_mount+0x194/0x844 age=22856 cpu=2 pid=566
alloc_debug_processing+0x1ac/0x1e8
___slab_alloc.constprop.63+0x640/0x648
__slab_alloc.isra.57.constprop.62+0x24/0x34
kmem_cache_alloc_trace+0x1a8/0x2bc
ffs_fs_mount+0x194/0x844
mount_fs+0x6c/0x1d0
vfs_kern_mount+0x50/0x1b4
do_mount+0x258/0x1034
INFO: Freed in ffs_data_put+0x25c/0x320 age=0 cpu=3 pid=4650
free_debug_processing+0x22c/0x434
__slab_free+0x2d8/0x3a0
kfree+0x254/0x264
ffs_data_put+0x25c/0x320
ffs_data_closed+0x124/0x15c
ffs_fs_kill_sb+0xb8/0x110
deactivate_locked_super+0x6c/0x98
deactivate_super+0xb0/0xbc
INFO: Object 0xffffffc057938600 @offset=1536 fp=0x (null)
......
Call trace:
[<ffffff900808cf5c>] dump_backtrace+0x0/0x250
[<ffffff900808d3a0>] show_stack+0x14/0x1c
[<ffffff90084a8c04>] dump_stack+0xa0/0xc8
[<ffffff900826c2b4>] print_trailer+0x158/0x260
[<ffffff900826d9d8>] object_err+0x3c/0x40
[<ffffff90082745f0>] kasan_report_error+0x2a8/0x754
[<ffffff9008274f84>] kasan_report+0x5c/0x60
[<ffffff9008273208>] __asan_load4+0x70/0x88
[<ffffff90084cd81c>] refcount_dec_and_test+0x14/0xe0
[<ffffff9008d98f9c>] ffs_data_put+0x80/0x320
[<ffffff9008d9d904>] ffs_fs_kill_sb+0xc8/0x110
[<ffffff90082852a0>] deactivate_locked_super+0x6c/0x98
[<ffffff900828537c>] deactivate_super+0xb0/0xbc
[<ffffff90082af0c0>] cleanup_mnt+0x64/0xec
[<ffffff90082af1b0>] __cleanup_mnt+0x10/0x18
[<ffffff90080d9e68>] task_work_run+0xcc/0x124
[<ffffff900808c8c0>] do_notify_resume+0x60/0x70
[<ffffff90080866e4>] work_pending+0x10/0x14
Cc: stable@vger.kernel.org
Signed-off-by: Xinyong <xinyong.fang@linux.alibaba.com>
Signed-off-by: Felipe Balbi <felipe.balbi@linux.intel.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
commit 6cf439e0d3 upstream.
During _ffs_func_bind(), the received descriptors are evaluated
to prepare for binding with the gadget in order to allocate
endpoints and optionally set up OS descriptors. However, the
high- and super-speed descriptors are only parsed based on
whether the gadget_is_dualspeed() and gadget_is_superspeed()
calls are true, respectively.
This is a problem in case a userspace program always provides
all of the {full,high,super,OS} descriptors when configuring a
function. Then, for example if a gadget device is not capable
of SuperSpeed, the call to ffs_do_descs() for the SS descriptors
is skipped, resulting in an incorrect offset calculation for
the vla_ptr when moving on to the OS descriptors that follow.
This causes ffs_do_os_descs() to fail as it is now looking at
the SS descriptors' offset within the raw_descs buffer instead.
_ffs_func_bind() should evaluate the descriptors unconditionally,
so remove the checks for gadget speed.
Fixes: f0175ab519 ("usb: gadget: f_fs: OS descriptors support")
Cc: stable@vger.kernel.org
Co-Developed-by: Mayank Rana <mrana@codeaurora.org>
Signed-off-by: Mayank Rana <mrana@codeaurora.org>
Signed-off-by: Jack Pham <jackp@codeaurora.org>
Signed-off-by: Felipe Balbi <felipe.balbi@linux.intel.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
commit c8cd751060 upstream.
Commit 76e0da34c7 ("usb-gadget/uvc: use per-attribute show and store
methods") caused a stringification of an undefined macro argument "aname",
so three UVC parameters (streaming_interval, streaming_maxpacket and
streaming_maxburst) were named "aname".
Add the definition of "aname" to the main macro and name the filenames as
originaly intended.
Signed-off-by: Petr Cvek <petr.cvek@tul.cz>
Signed-off-by: Felipe Balbi <felipe.balbi@linux.intel.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
commit ce5bf9a50d upstream.
Upon usb composition switch there is possibility of ep0 file
release happening after gadget driver bind. In case of composition
switch from adb to a non-adb composition gadget will never gets
bound again resulting into failure of usb device enumeration. Fix
this issue by checking FFS_FL_BOUND flag and avoid extra
gadget driver unbind if it is already done as part of composition
switch.
This fixes adb reconnection error reported on Android running
v4.4 and above kernel versions. Verified on Hikey running vanilla
v4.15-rc7 + few out of tree Mali patches.
Reviewed-at: https://android-review.googlesource.com/#/c/582632/
Cc: Felipe Balbi <balbi@kernel.org>
Cc: Greg KH <gregkh@linux-foundation.org>
Cc: Michal Nazarewicz <mina86@mina86.com>
Cc: John Stultz <john.stultz@linaro.org>
Cc: Dmitry Shmidt <dimitrysh@google.com>
Cc: Badhri <badhri@google.com>
Cc: Android Kernel Team <kernel-team@android.com>
Signed-off-by: Hemant Kumar <hemantk@codeaurora.org>
[AmitP: Cherry-picked it from android-4.14 and updated the commit log]
Signed-off-by: Amit Pundir <amit.pundir@linaro.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
commit a3acc69608 upstream.
The specification says that the Reserved1 field in OS_DESC_EXT_COMPAT
must have the value "1", but when this feature was first implemented we
rejected any non-zero values.
This was adjusted to accept all non-zero values (while now rejecting
zero) in commit 53642399aa ("usb: gadget: f_fs: Fix wrong check on
reserved1 of OS_DESC_EXT_COMPAT"), but that breaks any userspace
programs that worked previously by returning EINVAL when Reserved1 == 0
which was previously the only value that succeeded!
If we just set the field to "1" ourselves, both old and new userspace
programs continue to work correctly and, as a bonus, old programs are
now compliant with the specification without having to fix anything
themselves.
Fixes: 53642399aa ("usb: gadget: f_fs: Fix wrong check on reserved1 of OS_DESC_EXT_COMPAT")
Signed-off-by: John Keeping <john@metanate.com>
Signed-off-by: Felipe Balbi <felipe.balbi@linux.intel.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
commit cdafb6d8b8 upstream.
KASAN enabled configuration reports an error
BUG: KASAN: use-after-free in ffs_free_inst+... [usb_f_fs] at addr ...
Write of size 8 by task ...
This is observed after "ffs-test" is run and interrupted. If after that
functionfs is unmounted and g_ffs module is unloaded, that use-after-free
occurs during g_ffs module removal.
Although the report indicates ffs_free_inst() function, the actual
use-after-free condition occurs in _ffs_free_dev() function, which
is probably inlined into ffs_free_inst().
This happens due to keeping the ffs_data reference in device structure
during functionfs unmounting, while ffs_data itself is freed as no longer
needed. The fix is to clear that reference in ffs_closed() function,
which is a counterpart of ffs_ready(), where the reference is stored.
Fixes: 3262ad8243 ("usb: gadget: f_fs: Stop ffs_closed NULL pointer dereference")
Signed-off-by: Andrew Gabbasov <andrew_gabbasov@mentor.com>
Acked-by: Felipe Balbi <felipe.balbi@linux.intel.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
commit ff74745e6d upstream.
Kmemleak checking configuration reports a memory leak in
usb_os_desc_prepare_interf_dir function when rndis function
instance is freed and then allocated again. For example, this
happens with FunctionFS driver with RNDIS function enabled
when "ffs-test" test application is run several times in a row.
The data for intermediate "os_desc" group for interface directories
is allocated as a single VLA chunk and (after a change of default
groups handling) is not ever freed and actually not stored anywhere
besides inside a list of default groups of a parent group.
The fix is to make usb_os_desc_prepare_interf_dir function return
a pointer to allocated data (as a pointer to the first VLA item)
instead of (an unused) integer and to make the caller component
(currently the only one is RNDIS function) responsible for storing
the pointer and freeing the memory when appropriate.
Fixes: 1ae1602de0 ("configfs: switch ->default groups to a linked list")
Signed-off-by: Andrew Gabbasov <andrew_gabbasov@mentor.com>
Signed-off-by: Felipe Balbi <felipe.balbi@linux.intel.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
commit 1fbbb78f25 upstream.
As a holdover from the old g_file_storage gadget, the g_mass_storage
legacy gadget driver attempts to unregister itself when its main
operating thread terminates (if it hasn't been unregistered already).
This is not strictly necessary; it was never more than an attempt to
have the gadget fail cleanly if something went wrong and the main
thread was killed.
However, now that the UDC core manages gadget drivers independently of
UDC drivers, this scheme doesn't work any more. A simple test:
modprobe dummy-hcd
modprobe g-mass-storage file=...
rmmod dummy-hcd
ends up in a deadlock with the following backtrace:
sysrq: SysRq : Show Blocked State
task PC stack pid father
file-storage D 0 1130 2 0x00000000
Call Trace:
__schedule+0x53e/0x58c
schedule+0x6e/0x77
schedule_preempt_disabled+0xd/0xf
__mutex_lock.isra.1+0x129/0x224
? _raw_spin_unlock_irqrestore+0x12/0x14
__mutex_lock_slowpath+0x12/0x14
mutex_lock+0x28/0x2b
usb_gadget_unregister_driver+0x29/0x9b [udc_core]
usb_composite_unregister+0x10/0x12 [libcomposite]
msg_cleanup+0x1d/0x20 [g_mass_storage]
msg_thread_exits+0xd/0xdd7 [g_mass_storage]
fsg_main_thread+0x1395/0x13d6 [usb_f_mass_storage]
? __schedule+0x573/0x58c
kthread+0xd9/0xdb
? do_set_interface+0x25c/0x25c [usb_f_mass_storage]
? init_completion+0x1e/0x1e
ret_from_fork+0x19/0x24
rmmod D 0 1155 683 0x00000000
Call Trace:
__schedule+0x53e/0x58c
schedule+0x6e/0x77
schedule_timeout+0x26/0xbc
? __schedule+0x573/0x58c
do_wait_for_common+0xb3/0x128
? usleep_range+0x81/0x81
? wake_up_q+0x3f/0x3f
wait_for_common+0x2e/0x45
wait_for_completion+0x17/0x19
fsg_common_put+0x34/0x81 [usb_f_mass_storage]
fsg_free_inst+0x13/0x1e [usb_f_mass_storage]
usb_put_function_instance+0x1a/0x25 [libcomposite]
msg_unbind+0x2a/0x42 [g_mass_storage]
__composite_unbind+0x4a/0x6f [libcomposite]
composite_unbind+0x12/0x14 [libcomposite]
usb_gadget_remove_driver+0x4f/0x77 [udc_core]
usb_del_gadget_udc+0x52/0xcc [udc_core]
dummy_udc_remove+0x27/0x2c [dummy_hcd]
platform_drv_remove+0x1d/0x31
device_release_driver_internal+0xe9/0x16d
device_release_driver+0x11/0x13
bus_remove_device+0xd2/0xe2
device_del+0x19f/0x221
? selinux_capable+0x22/0x27
platform_device_del+0x21/0x63
platform_device_unregister+0x10/0x1a
cleanup+0x20/0x817 [dummy_hcd]
SyS_delete_module+0x10c/0x197
? ____fput+0xd/0xf
? task_work_run+0x55/0x62
? prepare_exit_to_usermode+0x65/0x75
do_fast_syscall_32+0x86/0xc3
entry_SYSENTER_32+0x4e/0x7c
What happens is that removing the dummy-hcd driver causes the UDC core
to unbind the gadget driver, which it does while holding the udc_lock
mutex. The unbind routine in g_mass_storage tells the main thread to
exit and waits for it to terminate.
But as mentioned above, when the main thread exits it tries to
unregister the mass-storage function driver. Via the composite
framework this ends up calling usb_gadget_unregister_driver(), which
tries to acquire the udc_lock mutex. The result is deadlock.
The simplest way to fix the problem is not to be so clever: The main
thread doesn't have to unregister the function driver. The side
effects won't be so terrible; if the gadget is still attached to a USB
host when the main thread is killed, it will appear to the host as
though the gadget's firmware has crashed -- a reasonably accurate
interpretation, and an all-too-common occurrence for USB mass-storage
devices.
In fact, the code to unregister the driver when the main thread exits
is specific to g-mass-storage; it is not used when f-mass-storage is
included as a function in a larger composite device. Therefore the
entire mechanism responsible for this (the fsg_operations structure
with its ->thread_exits method, the fsg_common_set_ops() routine, and
the msg_thread_exits() callback routine) can all be eliminated. Even
the msg_registered bitflag can be removed, because now the driver is
unregistered in only one place rather than in two places.
Signed-off-by: Alan Stern <stern@rowland.harvard.edu>
Acked-by: Felipe Balbi <felipe.balbi@linux.intel.com>
Acked-by: Michal Nazarewicz <mina86@mina86.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
[ Upstream commit 43aef5c2ca ]
This fixes an error message that was probably copied and pasted. The same
message is used for both the in and out endpoints, so it makes it impossible
to know which one actually failed because both cases say "IN".
Make the out endpoint error message say "OUT".
Signed-off-by: David Lechner <david@lechnology.com>
Signed-off-by: Felipe Balbi <felipe.balbi@linux.intel.com>
Signed-off-by: Sasha Levin <alexander.levin@verizon.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
commit b7f73850bb upstream.
Companion descriptor is only used for SuperSpeed endpoints,
if the endpoints are HighSpeed or FullSpeed, the Companion
descriptor will not allocated, so we can only access it if
gadget is SuperSpeed.
I can reproduce this issue on Rockchip platform rk3368 SoC
which supports USB 2.0, and use functionfs for ADB. Kernel
build with CONFIG_KASAN=y and CONFIG_SLUB_DEBUG=y report
the following BUG:
==================================================================
BUG: KASAN: slab-out-of-bounds in ffs_func_set_alt+0x224/0x3a0 at addr ffffffc0601f6509
Read of size 1 by task swapper/0/0
============================================================================
BUG kmalloc-256 (Not tainted): kasan: bad access detected
----------------------------------------------------------------------------
Disabling lock debugging due to kernel taint
INFO: Allocated in ffs_func_bind+0x52c/0x99c age=1275 cpu=0 pid=1
alloc_debug_processing+0x128/0x17c
___slab_alloc.constprop.58+0x50c/0x610
__slab_alloc.isra.55.constprop.57+0x24/0x34
__kmalloc+0xe0/0x250
ffs_func_bind+0x52c/0x99c
usb_add_function+0xd8/0x1d4
configfs_composite_bind+0x48c/0x570
udc_bind_to_driver+0x6c/0x170
usb_udc_attach_driver+0xa4/0xd0
gadget_dev_desc_UDC_store+0xcc/0x118
configfs_write_file+0x1a0/0x1f8
__vfs_write+0x64/0x174
vfs_write+0xe4/0x200
SyS_write+0x68/0xc8
el0_svc_naked+0x24/0x28
INFO: Freed in inode_doinit_with_dentry+0x3f0/0x7c4 age=1275 cpu=7 pid=247
...
Call trace:
[<ffffff900808aab4>] dump_backtrace+0x0/0x230
[<ffffff900808acf8>] show_stack+0x14/0x1c
[<ffffff90084ad420>] dump_stack+0xa0/0xc8
[<ffffff90082157cc>] print_trailer+0x188/0x198
[<ffffff9008215948>] object_err+0x3c/0x4c
[<ffffff900821b5ac>] kasan_report+0x324/0x4dc
[<ffffff900821aa38>] __asan_load1+0x24/0x50
[<ffffff90089eb750>] ffs_func_set_alt+0x224/0x3a0
[<ffffff90089d3760>] composite_setup+0xdcc/0x1ac8
[<ffffff90089d7394>] android_setup+0x124/0x1a0
[<ffffff90089acd18>] _setup+0x54/0x74
[<ffffff90089b6b98>] handle_ep0+0x3288/0x4390
[<ffffff90089b9b44>] dwc_otg_pcd_handle_out_ep_intr+0x14dc/0x2ae4
[<ffffff90089be85c>] dwc_otg_pcd_handle_intr+0x1ec/0x298
[<ffffff90089ad680>] dwc_otg_pcd_irq+0x10/0x20
[<ffffff9008116328>] handle_irq_event_percpu+0x124/0x3ac
[<ffffff9008116610>] handle_irq_event+0x60/0xa0
[<ffffff900811af30>] handle_fasteoi_irq+0x10c/0x1d4
[<ffffff9008115568>] generic_handle_irq+0x30/0x40
[<ffffff90081159b4>] __handle_domain_irq+0xac/0xdc
[<ffffff9008080e9c>] gic_handle_irq+0x64/0xa4
...
Memory state around the buggy address:
ffffffc0601f6400: 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00
ffffffc0601f6480: 00 00 00 00 00 00 00 00 00 00 06 fc fc fc fc fc
>ffffffc0601f6500: fc fc fc fc fc fc fc fc fc fc fc fc fc fc fc fc
^
ffffffc0601f6580: fc fc fc fc fc fc fc fc fc fc fc fc fc fc fc fc
ffffffc0601f6600: fc fc fc fc fc fc fc fc 00 00 00 00 00 00 00 00
==================================================================
Signed-off-by: William Wu <william.wu@rock-chips.com>
Signed-off-by: Felipe Balbi <felipe.balbi@linux.intel.com>
Cc: Jerry Zhang <zhangjerry@google.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
commit dc9217b69d upstream.
f_mass_storage has a memorry barrier issue with the sleep and wake
functions that can cause a deadlock. This results in intermittent hangs
during MSC file transfer. The host will reset the device after receiving
no response to resume the transfer. This issue is seen when dwc3 is
processing 2 transfer-in-progress events at the same time, invoking
completion handlers for CSW and CBW. Also this issue occurs depending on
the system timing and latency.
To increase the chance to hit this issue, you can force dwc3 driver to
wait and process those 2 events at once by adding a small delay (~100us)
in dwc3_check_event_buf() whenever the request is for CSW and read the
event count again. Avoid debugging with printk and ftrace as extra
delays and memory barrier will mask this issue.
Scenario which can lead to failure:
-----------------------------------
1) The main thread sleeps and waits for the next command in
get_next_command().
2) bulk_in_complete() wakes up main thread for CSW.
3) bulk_out_complete() tries to wake up the running main thread for CBW.
4) thread_wakeup_needed is not loaded with correct value in
sleep_thread().
5) Main thread goes to sleep again.
The pattern is shown below. Note the 2 critical variables.
* common->thread_wakeup_needed
* bh->state
CPU 0 (sleep_thread) CPU 1 (wakeup_thread)
============================== ===============================
bh->state = BH_STATE_FULL;
smp_wmb();
thread_wakeup_needed = 0; thread_wakeup_needed = 1;
smp_rmb();
if (bh->state != BH_STATE_FULL)
sleep again ...
As pointed out by Alan Stern, this is an R-pattern issue. The issue can
be seen when there are two wakeups in quick succession. The
thread_wakeup_needed can be overwritten in sleep_thread, and the read of
the bh->state maybe reordered before the write to thread_wakeup_needed.
This patch applies full memory barrier smp_mb() in both sleep_thread()
and wakeup_thread() to ensure the order which the thread_wakeup_needed
and bh->state are written and loaded.
However, a better solution in the future would be to use wait_queue
method that takes care of managing memory barrier between waker and
waiter.
Acked-by: Alan Stern <stern@rowland.harvard.edu>
Signed-off-by: Thinh Nguyen <thinhn@synopsys.com>
Signed-off-by: Felipe Balbi <felipe.balbi@linux.intel.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
commit 09424c50b7 upstream.
The streaming_maxburst module parameter is 0 offset (0..15)
so we must add 1 while using it for wBytesPerInterval
calculation for the SuperSpeed companion descriptor.
Without this host uvcvideo driver will always see the wrong
wBytesPerInterval for SuperSpeed uvc gadget and may not find
a suitable video interface endpoint.
e.g. for streaming_maxburst = 0 case it will always
fail as wBytePerInterval was evaluating to 0.
Reviewed-by: Laurent Pinchart <laurent.pinchart@ideasonboard.com>
Signed-off-by: Roger Quadros <rogerq@ti.com>
Signed-off-by: Felipe Balbi <felipe.balbi@linux.intel.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
commit cdd7928df0 upstream.
The gadget code exports the bitfield for serial status changes
over the wire in its internal endianness. The fix is to convert
to little endian before sending it over the wire.
Signed-off-by: Oliver Neukum <oneukum@suse.com>
Tested-by: 家瑋 <momo1208@gmail.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
commit eb38d913c2 upstream.
This reverts commit 4fbac5206a.
This commit breaks g_webcam when used with uvc-gadget [1].
The user space application (e.g. uvc-gadget) is responsible for
sending response to UVC class specific requests on control endpoint
in uvc_send_response() in uvc_v4l2.c.
The bad commit was causing a duplicate response to be sent with
incorrect response data thus causing UVC probe to fail at the host
and broken control transfer endpoint at the gadget.
[1] - git://git.ideasonboard.org/uvc-gadget.git
Acked-by: Laurent Pinchart <laurent.pinchart@ideasonboard.com>
Signed-off-by: Roger Quadros <rogerq@ti.com>
Signed-off-by: Felipe Balbi <felipe.balbi@linux.intel.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
commit 2bfa0719ac upstream.
If we're dealing with SuperSpeed endpoints, we need
to make sure to pass along the companion descriptor
and initialize fields needed by the Gadget
API. Eventually, f_fs.c should be converted to use
config_ep_by_speed() like all other functions,
though.
Signed-off-by: Felipe Balbi <felipe.balbi@linux.intel.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
commit 33e4c1a998 upstream.
As IN request has to be allocated in set_alt() and released in
disable() we cannot use mutex to protect it as we cannot sleep
in those funcitons. Let's replace this mutex with a spinlock.
Tested-by: David Lechner <david@lechnology.com>
Signed-off-by: Krzysztof Opasiak <k.opasiak@samsung.com>
Signed-off-by: Felipe Balbi <felipe.balbi@linux.intel.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
commit aa65d11aa0 upstream.
When we unlock our spinlock to copy data to user we may get
disabled by USB host and free the whole list of completed out
requests including the one from which we are copying the data
to user memory.
To prevent from this let's remove our working element from
the list and place it back only if there is sth left when we
finish with it.
Fixes: 99c5150058 ("usb: gadget: hidg: register OUT INT endpoint for SET_REPORT")
Tested-by: David Lechner <david@lechnology.com>
Signed-off-by: Krzysztof Opasiak <k.opasiak@samsung.com>
Signed-off-by: Felipe Balbi <felipe.balbi@linux.intel.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
commit 20d2ca955b upstream.
Requests for out endpoint are allocated in bind() function
but never released.
This commit ensures that all pending requests are released
when we disable out endpoint.
Fixes: 99c5150058 ("usb: gadget: hidg: register OUT INT endpoint for SET_REPORT")
Tested-by: David Lechner <david@lechnology.com>
Signed-off-by: Krzysztof Opasiak <k.opasiak@samsung.com>
Signed-off-by: Felipe Balbi <felipe.balbi@linux.intel.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
commit 83e526f2a2 upstream.
OS descriptor head, when flagged as provided, is accessed without
checking if it fits in provided buffer. Verify length before access.
Also, there are other places where buffer length it checked
after accessing offsets which are potentially past the end. Check
buffer length before as well to fail cleanly.
Signed-off-by: Vincent Pelletier <plr.vincent@gmail.com>
Acked-by: Felipe Balbi <felipe.balbi@linux.intel.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
commit eaa496ffaa upstream.
ep->mult is supposed to be set to Isochronous and
Interrupt Endapoint's multiplier value. This value
is computed from different places depending on the
link speed.
If we're dealing with HighSpeed, then it's part of
bits [12:11] of wMaxPacketSize. This case wasn't
taken into consideration before.
While at that, also make sure the ep->mult defaults
to one so drivers can use it unconditionally and
assume they'll never multiply ep->maxpacket to zero.
Signed-off-by: Felipe Balbi <felipe.balbi@linux.intel.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
commit eaa496ffaa upstream.
ep->mult is supposed to be set to Isochronous and
Interrupt Endapoint's multiplier value. This value
is computed from different places depending on the
link speed.
If we're dealing with HighSpeed, then it's part of
bits [12:11] of wMaxPacketSize. This case wasn't
taken into consideration before.
While at that, also make sure the ep->mult defaults
to one so drivers can use it unconditionally and
assume they'll never multiply ep->maxpacket to zero.
Cc: <stable@vger.kernel.org>
Signed-off-by: Felipe Balbi <felipe.balbi@linux.intel.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Properly check the return code of ffs_func_revmap_intf() and
ffs_func_revmap_ep() for a non-negative value.
Instead of checking the return code, the comparison was performed for the last
parameter of the function calls, because of wrong parenthesis.
This also fixes the following static checker warning:
drivers/usb/gadget/function/f_fs.c:3152 ffs_func_req_match()
warn: always true condition '(((creq->wIndex)) >= 0) => (0-u16max >= 0)'
Reported-by: Dan Carpenter <dan.carpenter@oracle.com>
Signed-off-by: Felix Hädicke <felixhaedicke@web.de>
Signed-off-by: Felipe Balbi <felipe.balbi@linux.intel.com>
According to Dave Miller "the networking stack has a
hard requirement that all SKBs which are transmitted
must have their completion signalled in a fininte
amount of time. This is because, until the SKB is
freed by the driver, it holds onto socket,
netfilter, and other subsystem resources."
In summary, this means that using TX IRQ throttling
for the networking gadgets is, at least, complex and
we should avoid it for the time being.
Cc: <stable@vger.kernel.org>
Reported-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Tested-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Suggested-by: David Miller <davem@davemloft.net>
Signed-off-by: Felipe Balbi <felipe.balbi@linux.intel.com>
ffs_func_eps_disable is called from atomic context so it cannot sleep
thus cannot grab a mutex. Change the handling of epfile->read_buffer
to use non-sleeping synchronisation method.
Reported-by: Chen Yu <chenyu56@huawei.com>
Signed-off-by: Michał Nazarewicz <mina86@mina86.com>
Fixes: 9353afbbfa ("buffer data from ‘oversized’ OUT requests")
Tested-by: John Stultz <john.stultz@linaro.org>
Tested-by: Chen Yu <chenyu56@huawei.com>
Signed-off-by: Felipe Balbi <felipe.balbi@linux.intel.com>
If we don't guarantee that we will always get an
interrupt at least when we're queueing our very last
request, we could fall into situation where we queue
every request with 'no_interrupt' set. This will
cause the link to get stuck.
The behavior above has been triggered with g_ether
and dwc3.
Cc: <stable@vger.kernel.org>
Reported-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Signed-off-by: Felipe Balbi <felipe.balbi@linux.intel.com>
Pull more vfs updates from Al Viro:
">rename2() work from Miklos + current_time() from Deepa"
* 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/viro/vfs:
fs: Replace current_fs_time() with current_time()
fs: Replace CURRENT_TIME_SEC with current_time() for inode timestamps
fs: Replace CURRENT_TIME with current_time() for inode timestamps
fs: proc: Delete inode time initializations in proc_alloc_inode()
vfs: Add current_time() api
vfs: add note about i_op->rename changes to porting
fs: rename "rename2" i_op to "rename"
vfs: remove unused i_op->rename
fs: make remaining filesystems use .rename2
libfs: support RENAME_NOREPLACE in simple_rename()
fs: support RENAME_NOREPLACE for local filesystems
ncpfs: fix unused variable warning
CURRENT_TIME macro is not appropriate for filesystems as it
doesn't use the right granularity for filesystem timestamps.
Use current_time() instead.
CURRENT_TIME is also not y2038 safe.
This is also in preparation for the patch that transitions
vfs timestamps to use 64 bit time and hence make them
y2038 safe. As part of the effort current_time() will be
extended to do range checks. Hence, it is necessary for all
file system timestamps to use current_time(). Also,
current_time() will be transitioned along with vfs to be
y2038 safe.
Note that whenever a single call to current_time() is used
to change timestamps in different inodes, it is because they
share the same time granularity.
Signed-off-by: Deepa Dinamani <deepa.kernel@gmail.com>
Reviewed-by: Arnd Bergmann <arnd@arndb.de>
Acked-by: Felipe Balbi <balbi@kernel.org>
Acked-by: Steven Whitehouse <swhiteho@redhat.com>
Acked-by: Ryusuke Konishi <konishi.ryusuke@lab.ntt.co.jp>
Acked-by: David Sterba <dsterba@suse.com>
Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
