The use of variable length arrays in structs (VLAIS) in the Linux Kernel code
precludes the use of compilers which don't implement VLAIS (for instance the
Clang compiler). This alternate patch calculates offsets into the kmalloc-ed
memory buffer using macros. The previous patch required multiple kmalloc and
kfree calls. This version uses "group" vs "struct" since it really is not a
struct and is essentially a group of VLA in a common allocated block. This
version also fixes the issues pointed out by Andrzej Pietrasiewicz and
Michal Nazarewicz.
Signed-off-by: NMark Charlebois <charlebm@gmail.com>
Signed-off-by: NBehan Webster <behanw@converseincode.com>

[elimination of miexed declaration and code, checkpatch cleanup]
[fixes after Michal's review]
Signed-off-by: NAndrzej Pietrasiewicz <andrzej.p@samsung.com>
Signed-off-by: NKyungmin Park <kyungmin.park@samsung.com>
Acked-by: NMichal Nazarewicz <mina86@mina86.com>
Signed-off-by: NFelipe Balbi <balbi@ti.com>

The rndis function's users use only the new interface, so the two modules
can be merged.
Signed-off-by: NAndrzej Pietrasiewicz <andrzej.p@samsung.com>
Signed-off-by: NKyungmin Park <kyungmin.park@samsung.com>
Acked-by: NMichal Nazarewicz <mina86@mina86.com>
Signed-off-by: NFelipe Balbi <balbi@ti.com>

There are no old function interface users left, so the old interface
can be removed.
Signed-off-by: NAndrzej Pietrasiewicz <andrzej.p@samsung.com>
Signed-off-by: NKyungmin Park <kyungmin.park@samsung.com>
Acked-by: NMichal Nazarewicz <mina86@mina86.com>
Signed-off-by: NFelipe Balbi <balbi@ti.com>

There is a new interface of f_rndis and g_ffs is the last to use the old one.
Signed-off-by: NAndrzej Pietrasiewicz <andrzej.p@samsung.com>
Signed-off-by: NKyungmin Park <kyungmin.park@samsung.com>
Acked-by: NMichal Nazarewicz <mina86@mina86.com>
Signed-off-by: NFelipe Balbi <balbi@ti.com>

There are no old function interface users left, so the old interface
can be removed.
Signed-off-by: NAndrzej Pietrasiewicz <andrzej.p@samsung.com>
Signed-off-by: NKyungmin Park <kyungmin.park@samsung.com>
Acked-by: NMichal Nazarewicz <mina86@mina86.com>
Signed-off-by: NFelipe Balbi <balbi@ti.com>

There is a new function interface of f_subset and g_ffs is the last to use
the old one.
Signed-off-by: NAndrzej Pietrasiewicz <andrzej.p@samsung.com>
Signed-off-by: NKyungmin Park <kyungmin.park@samsung.com>
Acked-by: NMichal Nazarewicz <mina86@mina86.com>
Signed-off-by: NFelipe Balbi <balbi@ti.com>

There are no old function interface users left, so the old interface
can be removed.
Signed-off-by: NAndrzej Pietrasiewicz <andrzej.p@samsung.com>
Signed-off-by: NKyungmim Park <kyungmin.park@samsung.com>
Acked-by: NMichal Nazarewicz <mina86@mina86.com>
Signed-off-by: NFelipe Balbi <balbi@ti.com>

There is a new funtion interface and g_ffs is the last gadget to use the old.
Signed-off-by: NAndrzej Pietrasiewicz <andrzej.p@samsung.com>
Signed-off-by: NKyungmin Park <kyungmin.park@samsung.com>
Acked-by: NMichal Nazarewicz <mina86@mina86.com>
Signed-off-by: NFelipe Balbi <balbi@ti.com>

Since d6a01439
usb: gadget: move the global the_dev variable to their users
"the_dev" variable can be used as a "setup done" flag; non-NULL
meaning "setup done", NULL meaning "setup not done". Moreover,
gether_cleanup() can be safely called with a NULL argument.

Corrected a comment to be consistent with the code.
Signed-off-by: NAndrzej Pietrasiewicz <andrzej.p@samsung.com>
Signed-off-by: NKyungmin Park <kyungmin.park@samsung.com>
Acked-by: NMichal Nazarewicz <mina86@mina86.com>
Signed-off-by: NFelipe Balbi <balbi@ti.com>

USB function's configfs config group is created in a generic way in
usb/gadget/configfs.c:function_make(), which in turn delegates actual
allocation and setup of the USB function instance to a particular
implementation, e.g. in f_acm.c. The said implementation does its job
in a parameter-less function e.g. acm_alloc_instance(), which results
in creating an unnamed config group, whose name is set later in
function_make(). function_make() creates the name by parsing a string
of the form:

<function name>.<instance name>

which comes from userspace as a parameter to mkdir invocation.

Up to now only <function name> has been used, while <instance name>
has been ignored. This patch adds a set_inst_name() operation to
struct usb_function_instance which allows passing the <instance name>
from function_make() so that it is not ignored. It is entirely up to the
implementor of set_inst_name() what to do with the <instance name>.

In a typical case, the struct usb_function_instance is embedded in a
larger struct which is retrieved in set_inst_name() with container_of(),
and the larger struct contains a field to store the <instance name>.
Signed-off-by: NAndrzej Pietrasiewicz <andrzej.p@samsung.com>
Signed-off-by: NKyungmin Park <kyungmin.park@samsung.com>
Acked-by: NMichal Nazarewicz <mina86@mina86.com>
Signed-off-by: NFelipe Balbi <balbi@ti.com>

This patch adds flushing TX FIFO in kill_all_requests() function in
dedicated-fifo mode. It's because when requests are killed (when endpoint is
disabled or in case of device reset/disconnection) in FIFO can stay some
unsent data. In the worst case FIFO can stay full, and then if endpoint will
be back enabled, sending new data will be impossible.
Signed-off-by: NRobert Baldyga <r.baldyga@samsung.com>
Signed-off-by: NKyungmin Park <kyungmin.park@samsung.com>
Signed-off-by: NFelipe Balbi <balbi@ti.com>

This patch set maximum possible maxpacket value for each speed. Previous
values didn't allow to use maxpacket sizes greater than 64 in full speed
and 512 in high speed, although hardware is able to handle up to 1023 in fs
and 1024 in hs.
Tested-by: NMatt Porter <mporter@linaro.org>
Signed-off-by: NRobert Baldyga <r.baldyga@samsung.com>
Signed-off-by: NKyungmin Park <kyungmin.park@samsung.com>
Signed-off-by: NFelipe Balbi <balbi@ti.com>

We need to initialize the notifer before use it.

So lets initialize it when add a new phy device to
reduce the code redundancy.
Signed-off-by: NNeil Zhang <zhangwm@marvell.com>
Signed-off-by: NFelipe Balbi <balbi@ti.com>

DMAEngine uses IRQ if dma_request_channel() was called,
and it is using devm_request_irq() today,
OTOH, dma_request_channel() will be called when each
USB connection happened on this driver.
This means same IRQ will be requested many times
whenever each USB connected,
and this IRQ isn't freed when USB disconnected.
Request DMAEngine once.
Signed-off-by: NKuninori Morimoto <kuninori.morimoto.gx@renesas.com>
Signed-off-by: NFelipe Balbi <balbi@ti.com>

Add Keystone platform USB PHY driver support. Current main purpose
of this driver is to enable the PHY reference clock gate on the
Keystone SoC. Otherwise it is a nop PHY.

[ balbi@ti.com : add COMPILE_TEST as a possible dependency
	make sure drvdata is initialized before adding PHY ]

Cc: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Acked-by: NSantosh Shilimkar <santosh.shilimkar@ti.com>
Signed-off-by: NWingMan Kwok <w-kwok2@ti.com>
Signed-off-by: NFelipe Balbi <balbi@ti.com>

Add Keystone platform specific glue layer to support
USB3 Host mode.

[ balbi@ti.com : fix order of clk_disable() and
	platform_device_unregister() ]

Cc: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Acked-by: NSantosh Shilimkar <santosh.shilimkar@ti.com>
Signed-off-by: NWingMan Kwok <w-kwok2@ti.com>
Signed-off-by: NFelipe Balbi <balbi@ti.com>

DWC3 requires epout to have buffer size aligned to MaxPacketSize value.
This patch sets necessary quirk for it.
Signed-off-by: NDavid Cohen <david.a.cohen@linux.intel.com>
Signed-off-by: NFelipe Balbi <balbi@ti.com>

Check gadget.quirk_ep_out_aligned_size to decide if buffer size requires
to be aligned to maxpacketsize of an out endpoint.  ffs_epfile_io() needs
to pad epout buffer to match above condition if quirk is found.
Signed-off-by: NMichal Nazarewicz <mina86@mina86.com>
Signed-off-by: NDavid Cohen <david.a.cohen@linux.intel.com>
Signed-off-by: NFelipe Balbi <balbi@ti.com>

When endpoint changes (due to it being disabled or alt setting changed),
mimic the action as if the change happened after the request has been
queued, instead of retrying with the new endpoint.
Signed-off-by: NMichal Nazarewicz <mina86@mina86.com>
Cc: David Cohen <david.a.cohen@linux.intel.com>
Signed-off-by: NFelipe Balbi <balbi@ti.com>

Use usb_add_phy_dev instead of usb_add_phy, so that devices can
be bound to the phy. This is needed to set up USB phy for
some internal PCI USB host controllers on R-Car Gen2.
Signed-off-by: NValentine Barshak <valentine.barshak@cogentembedded.com>
Signed-off-by: NFelipe Balbi <balbi@ti.com>

the lock was only taken inside the hardirq
handler, which runs with IRQs disabled. There's
no chance of any race condition happening, even
on SMP machines. It's safe to remove that
spinlock.
Acked-by: NSantosh Shilimkar <santosh.shilimkar@ti.com>
Signed-off-by: NFelipe Balbi <balbi@ti.com>

Add Tahvo USB transceiver driver.

Based on old code from linux-omap tree. The original driver was written
by Juha Yrjölä, Tony Lindgren, and Timo Teräs.
Signed-off-by: NAaro Koskinen <aaro.koskinen@iki.fi>
Signed-off-by: NFelipe Balbi <balbi@ti.com>

Transceivers need to manage OTG controller state on OMAP1 to enable
switching between peripheral and host modes. Provide a driver for that.
Signed-off-by: NAaro Koskinen <aaro.koskinen@iki.fi>
Signed-off-by: NFelipe Balbi <balbi@ti.com>

clk_prepare_enable() may fail, so let's check its return value and propagate it
in the case of error.
Acked-by: NPeter Chen <peter.chen@freescale.com>
Signed-off-by: NFabio Estevam <fabio.estevam@freescale.com>
Signed-off-by: NFelipe Balbi <balbi@ti.com>

Don't use DEFINE_PCI_DEVICE_TABLE macro, because this macro
is not preferred.
Signed-off-by: NJingoo Han <jg1.han@samsung.com>
Signed-off-by: NFelipe Balbi <balbi@ti.com>

Don't use DEFINE_PCI_DEVICE_TABLE macro, because this macro
is not preferred.
Signed-off-by: NJingoo Han <jg1.han@samsung.com>
Signed-off-by: NFelipe Balbi <balbi@ti.com>

Don't use DEFINE_PCI_DEVICE_TABLE macro, because this macro
is not preferred.
Signed-off-by: NJingoo Han <jg1.han@samsung.com>
Signed-off-by: NFelipe Balbi <balbi@ti.com>

Add support for using the sourcesink function in gadgets composed with
configfs.
Signed-off-by: NAndrzej Pietrasiewicz <andrzej.p@samsung.com>
Signed-off-by: NKyungmin Park <kyungmin.park@samsung.com>
Signed-off-by: NFelipe Balbi <balbi@ti.com>

Add support for using the loopback USB function in gadgets composed with
configfs.
Signed-off-by: NAndrzej Pietrasiewicz <andrzej.p@samsung.com>
Signed-off-by: NKyungmin Park <kyungmin.park@samsung.com>
Signed-off-by: NFelipe Balbi <balbi@ti.com>

alloc_ep_req() is a function repeated in several modules.
Make a common implementation and use it.
Signed-off-by: NAndrzej Pietrasiewicz <andrzej.p@samsung.com>
Signed-off-by: NKyungmin Park <kyungmin.park@samsung.com>
Acked-by: NMichal Nazarewicz <mina86@mina86.com>
Signed-off-by: NFelipe Balbi <balbi@ti.com>

If there are setup requests not directed to an endpont or an interface,
current config's setup() has been attempted so far.
This patch, in case the above fails, adds code which tries the setup() of
configuration's function if there is only one function in the configuration.

This behavior is required to provide equivalent of gadget zero with configfs.

The gadget zero has a "config driver" for sourcesink, but all it does is
delegating the request to the function proper. So when the equivalent gadget
is set up with configfs it needs to handle requests directed to
"config driver", but with configfs it is not possible to specify
"config drivers".
Signed-off-by: NAndrzej Pietrasiewicz <andrzej.p@samsung.com>
Signed-off-by: NKyungmin Park <kyungmin.park@samsung.com>
Acked-by: NMichal Nazarewicz <mina86@mina86.com>
Signed-off-by: NFelipe Balbi <balbi@ti.com>

Only start the otg_timer in dual role mode; otherwise in peripheral mode
when musb is disconnected from the host port, otg_timer starts and
continuously toggles the session, which causes VBUS pulse.
Signed-off-by: NBin Liu <b-liu@ti.com>
Acked-by: NSebastian Andrzej Siewior <bigeasy@linutronix.de>
Signed-off-by: NFelipe Balbi <balbi@ti.com>

This patch removes dependency on Freescale USB UTG Transceiver
driver and makes OTG FSM implementation selectable.
Signed-off-by: NAnton Tikhomirov <av.tikhomirov@samsung.com>
Signed-off-by: NFelipe Balbi <balbi@ti.com>

Other USB drivers may want to use OTG final state machine
implementation, so make this header available for them.
Signed-off-by: NAnton Tikhomirov <av.tikhomirov@samsung.com>
Signed-off-by: NFelipe Balbi <balbi@ti.com>

OTG Final State Machine calls functions which may sleep.

For example, start_gadget callback implementation can use
usb_gadget_vbus_connect(), whose context: can sleep.

If so, mutex should be used instead of spinlock.
Signed-off-by: NAnton Tikhomirov <av.tikhomirov@samsung.com>
Signed-off-by: NFelipe Balbi <balbi@ti.com>

if this header is included twice, we would have
redefinition build errors. Fix this.
Signed-off-by: NAnton Tikhomirov <av.tikhomirov@samsung.com>
Signed-off-by: NFelipe Balbi <balbi@ti.com>

this header uses spinlocks and errno values, so
we must include <linux/spinlock.h> and <linux/errno.h>
to avoid build errors.
Signed-off-by: NFelipe Balbi <balbi@ti.com>

These functions used to return negative errror codes but now they return
ERR_PTRs.
Signed-off-by: NDan Carpenter <dan.carpenter@oracle.com>
Acked-by: NSebastian Andrzej Siewior <bigeasy@linutronix.de>
Signed-off-by: NFelipe Balbi <balbi@ti.com>

device_driver.name is "const char *"
Signed-off-by: NGeert Uytterhoeven <geert@linux-m68k.org>
Cc: linux-usb@vger.kernel.org
Signed-off-by: NFelipe Balbi <balbi@ti.com>

The stop_activity function never uses driver argument (even
though it modifies it) and thus it is safe to remove it.
Signed-off-by: NMichal Nazarewicz <mina86@mina86.com>
Signed-off-by: NFelipe Balbi <balbi@ti.com>