openwrt/staging/blogic.git
8 years agousb: gadget: uvc: add V4L2 dependency
Arnd Bergmann [Mon, 12 Sep 2016 15:34:57 +0000 (17:34 +0200)]
usb: gadget: uvc: add V4L2 dependency

Building the UVC gadget into the kernel fails to build when
CONFIG_VIDEO_V4L2 is a loadable module:

drivers/usb/gadget/function/usb_f_uvc.o: In function `uvc_function_ep0_complete':
uvc_configfs.c:(.text.uvc_function_ep0_complete+0x84): undefined reference to `v4l2_event_queue'
drivers/usb/gadget/function/usb_f_uvc.o: In function `uvc_function_disable':
uvc_configfs.c:(.text.uvc_function_disable+0x34): undefined reference to `v4l2_event_queue'

Adding a dependency in USB_CONFIGFS_F_UVC (which is a bool symbol)
make the 'select USB_F_UVC' statement turn the USB_F_UVC into 'm'
whenever CONFIG_VIDEO_V4L2=m too, avoiding the link failure.

Signed-off-by: Arnd Bergmann <arnd@arndb.de>
Signed-off-by: Felipe Balbi <felipe.balbi@linux.intel.com>
8 years agousb: dwc3: of-simple: add compatible for Cavium
Felipe Balbi [Mon, 12 Sep 2016 18:24:58 +0000 (21:24 +0300)]
usb: dwc3: of-simple: add compatible for Cavium

Add necessary compatible flag for Cavium's DWC3 so
dwc3-of-simple will probe.

Tested-by: Steven J. Hill <Steven.Hill@cavium.com>
Signed-off-by: Felipe Balbi <felipe.balbi@linux.intel.com>
8 years agousb: dwc3: of-simple: allow glues without clocks
Felipe Balbi [Mon, 12 Sep 2016 18:20:22 +0000 (21:20 +0300)]
usb: dwc3: of-simple: allow glues without clocks

Instead of erroring out when we don't have clocks,
let's just avoid any calls to the clk API.

Tested-by: Steven J. Hill <Steven.Hill@cavium.com>
Signed-off-by: Felipe Balbi <felipe.balbi@linux.intel.com>
8 years agousb: dwc3: of-simple: Fix warning during unbind
Anurag Kumar Vulisha [Fri, 9 Sep 2016 13:28:37 +0000 (18:58 +0530)]
usb: dwc3: of-simple: Fix warning during unbind

In dwc3_of_simple_remove() we are using clk_unprepare() before doing
any clk_disable(). If we enable Common CLK framework (CCF) and try to
unbind dwc3-of-simple driver, we see kernel WARN messages.

This patch fixes this kernel warning by using clk_disable_unprepare()
instead of clk_unprepare().

Signed-off-by: Anurag Kumar Vulisha <anuragku@xilinx.com>
Signed-off-by: Felipe Balbi <felipe.balbi@linux.intel.com>
8 years agousb: dwc3: fix Clear Stall EP command failure
Lu Baolu [Fri, 9 Sep 2016 04:51:27 +0000 (12:51 +0800)]
usb: dwc3: fix Clear Stall EP command failure

Commit 50c763f8c1bac ("usb: dwc3: Set the ClearPendIN bit on Clear
Stall EP command") sets ClearPendIN bit for all IN endpoints of
v2.60a+ cores. This causes ClearStall command fails on 2.60+ cores
operating in HighSpeed mode.

In page 539 of 2.60a specification:

"When issuing Clear Stall command for IN endpoints in SuperSpeed
mode, the software must set the "ClearPendIN" bit to '1' to
clear any pending IN transcations, so that the device does not
expect any ACK TP from the host for the data sent earlier."

It's obvious that we only need to apply this rule to those IN
endpoints that currently operating in SuperSpeed mode.

Fixes: 50c763f8c1bac ("usb: dwc3: Set the ClearPendIN bit on Clear Stall EP command")
Cc: <stable@vger.kernel.org> # v4.7+
Signed-off-by: Lu Baolu <baolu.lu@linux.intel.com>
Signed-off-by: Felipe Balbi <felipe.balbi@linux.intel.com>
8 years agousb: dwc3: avoid -Wmaybe-uninitialized warning
Arnd Bergmann [Fri, 9 Sep 2016 10:01:51 +0000 (12:01 +0200)]
usb: dwc3: avoid -Wmaybe-uninitialized warning

Cleaning up the loop in dwc3_cleanup_done_reqs() introduced a
gcc warning if built with "-Wmaybe-uninitialized":

drivers/usb/dwc3/gadget.c: In function 'dwc3_endpoint_transfer_complete':
drivers/usb/dwc3/gadget.c:2015:9: 'trb' may be used uninitialized in this function [-Wmaybe-uninitialized]

I believe it is a false positive and we always have a valid 'trb'
pointer at the end of the function, but neither I nor the compiler
are able to prove that.

This works around the warning by computing a flag earlier in the function
when it's guaranteed to be valid, which tells the compiler that it's
safe and makes it easier to understand to me.

Signed-off-by: Arnd Bergmann <arnd@arndb.de>
Fixes: 31162af447d7 ("usb: dwc3: gadget: avoid while (1) loop on completion")
Signed-off-by: Felipe Balbi <felipe.balbi@linux.intel.com>
8 years agousb: gadget: NCM: Protect dev->port_usb using dev->lock
Harish Jenny K N [Fri, 9 Sep 2016 09:30:42 +0000 (11:30 +0200)]
usb: gadget: NCM: Protect dev->port_usb using dev->lock

This commit incorporates findings from
https://lkml.org/lkml/2016/4/25/594

The function has been modified to make sure we hold
the dev lock when accessing the net device pointer.

Acked-by: Jim Baxter <jim_baxter@mentor.com>
Signed-off-by: Harish Jenny K N <harish_kandiga@mentor.com>
Signed-off-by: Felipe Balbi <felipe.balbi@linux.intel.com>
8 years agousb: gadget: u_ether: fix another dereference after null check
Harish Jenny K N [Fri, 9 Sep 2016 09:30:41 +0000 (11:30 +0200)]
usb: gadget: u_ether: fix another dereference after null check

dev->port_usb is checked for null pointer previously, so dev->port_usb
might be null during no zlp check, fix it by adding null pointer check.

Acked-by: Jim Baxter <jim_baxter@mentor.com>
Signed-off-by: Harish Jenny K N <harish_kandiga@mentor.com>
Signed-off-by: Felipe Balbi <felipe.balbi@linux.intel.com>
8 years agousb: gadget: remove variable ret and remove unnecessary if statement
Colin Ian King [Thu, 8 Sep 2016 20:09:30 +0000 (21:09 +0100)]
usb: gadget: remove variable ret and remove unnecessary if statement

the if statement in lb_modinit is unnecessary so we can totally
remove the variable ret and just return the return value from
the call to usb_function_register.

Signed-off-by: Colin Ian King <colin.king@canonical.com>
Signed-off-by: Felipe Balbi <felipe.balbi@linux.intel.com>
8 years agousb: dwc2: Properly account for the force mode delays
John Youn [Thu, 8 Sep 2016 02:39:43 +0000 (19:39 -0700)]
usb: dwc2: Properly account for the force mode delays

When a force mode bit is set and the IDDIG debounce filter is enabled,
there is a delay for the forced mode to take effect. This delay is due
to the IDDIG debounce filter and is variable depending on the platform's
PHY clock speed. To account for this delay we can poll for the expected
mode.

On a clear force mode, since we don't know what mode to poll for, delay
for a fixed 100 ms. This is the maximum delay based on the slowest PHY
clock speed.

Tested-by: Stefan Wahren <stefan.wahren@i2se.com>
Signed-off-by: John Youn <johnyoun@synopsys.com>
Signed-off-by: Felipe Balbi <felipe.balbi@linux.intel.com>
8 years agousb: dwc2: Add delay to core soft reset
John Youn [Thu, 8 Sep 2016 02:39:40 +0000 (19:39 -0700)]
usb: dwc2: Add delay to core soft reset

Add a delay to the core soft reset function to account for the IDDIG
debounce filter.

If the current mode is host, either due to the force mode bit being
set (which persists after core reset) or the connector id pin, a core
soft reset will temporarily reset the mode to device and a delay from
the IDDIG debounce filter will occur before going back to host mode.

Tested-by: Stefan Wahren <stefan.wahren@i2se.com>
Signed-off-by: John Youn <johnyoun@synopsys.com>
Signed-off-by: Felipe Balbi <felipe.balbi@linux.intel.com>
8 years agousb: dwc2: gadget: Only initialize device if in device mode
John Youn [Thu, 8 Sep 2016 02:39:37 +0000 (19:39 -0700)]
usb: dwc2: gadget: Only initialize device if in device mode

In dwc2_hsotg_udc_start(), don't initialize the controller for device
mode unless we are actually in device mode.

Tested-by: Heiko Stuebner <heiko@sntech.de>
Tested-by: Stefan Wahren <stefan.wahren@i2se.com>
Signed-off-by: John Youn <johnyoun@synopsys.com>
Signed-off-by: Felipe Balbi <felipe.balbi@linux.intel.com>
8 years agousb: gadget: f_fs: Stop ffs_closed NULL pointer dereference
Jim Baxter [Thu, 8 Sep 2016 09:18:16 +0000 (11:18 +0200)]
usb: gadget: f_fs: Stop ffs_closed NULL pointer dereference

The struct ffs_data::private_data has a pointer to
ffs_dev stored in it during the ffs_fs_mount() function
however it is not cleared when the ffs_dev is freed
later which causes the ffs_closed function to crash
with "Unable to handle kernel NULL pointer dereference"
error when using the data in ffs_data::private_data.

This clears this pointer during the ffs_free_dev clean
up function.

Signed-off-by: Jim Baxter <jim_baxter@mentor.com>
Signed-off-by: Jiada Wang <jiada_wang@mentor.com>
Signed-off-by: Harish Jenny K N <harish_kandiga@mentor.com>
Signed-off-by: Felipe Balbi <felipe.balbi@linux.intel.com>
8 years agousb: dwc3: Fix dr_mode validation
Thinh Nguyen [Wed, 7 Sep 2016 02:22:03 +0000 (19:22 -0700)]
usb: dwc3: Fix dr_mode validation

This patch follows the similar fix in dwc2. See
commit 5268ed9d2e3b ("usb: dwc2: Fix dr_mode validation")

Currently, the dr_mode is only checked against the module configuration.
It also needs to be checked against the hardware capablities.

The driver now checks if both the module configuration and hardware are
capable of the dr_mode value. If not, then it will issue a warning and
fall back to a supported value. If it is unable to fall back to a
suitable value, then the probe will fail.

Behavior summary:

      module          :  actual
 HW   config  dr_mode :  dr_mode
---------------------------------
 host  host   any     :  host
 host  dev    any     :  INVALID
 host  otg    any     :  host

 dev   host   any     :  INVALID
 dev   dev    any     :  dev
 dev   otg    any     :  dev

 otg   host   any     :  host
 otg   dev    any     :  dev
 otg   otg    any     :  dr_mode

Signed-off-by: Thinh Nguyen <thinhn@synopsys.com>
Signed-off-by: John Youn <johnyoun@synopsys.com>
Signed-off-by: Felipe Balbi <felipe.balbi@linux.intel.com>
8 years agousb: dwc3: ulpi: make dwc3_ulpi_ops constant
Tal Shorer [Tue, 16 Aug 2016 16:04:53 +0000 (19:04 +0300)]
usb: dwc3: ulpi: make dwc3_ulpi_ops constant

ulpi_register_interface() accepts a const struct ulpi_ops and dwc3
doesn't perform any changes to this struct at runtime, so there's no
reason it shouldn't be constant.

Reviewed-by: Heikki Krogerus <heikki.krogerus@linux.intel.com>
Signed-off-by: Tal Shorer <tal.shorer@gmail.com>
Signed-off-by: Felipe Balbi <felipe.balbi@linux.intel.com>
8 years agousb: ulpi: make ops struct constant
Tal Shorer [Tue, 16 Aug 2016 16:04:52 +0000 (19:04 +0300)]
usb: ulpi: make ops struct constant

None of the core ulpi functions perform any changes to the operations
struct, and logically as a struct that contains function pointers
there's no reason it shouldn't be constant.

Acked-by: Heikki Krogerus <heikki.krogerus@linux.intel.com>
Signed-off-by: Tal Shorer <tal.shorer@gmail.com>
Signed-off-by: Felipe Balbi <felipe.balbi@linux.intel.com>
8 years agousb: ulpi: remove "dev" field from struct ulpi_ops
Tal Shorer [Tue, 16 Aug 2016 16:04:51 +0000 (19:04 +0300)]
usb: ulpi: remove "dev" field from struct ulpi_ops

Operations now use ulpi->dev.parent directly instead of via the
ulpi_ops struct, making this field unused. Remove it.

Acked-by: Heikki Krogerus <heikki.krogerus@linux.intel.com>
Signed-off-by: Tal Shorer <tal.shorer@gmail.com>
Signed-off-by: Felipe Balbi <felipe.balbi@linux.intel.com>
8 years agousb: ulpi: rename operations {read|write}_dev to simply {read|write}
Tal Shorer [Tue, 16 Aug 2016 16:04:50 +0000 (19:04 +0300)]
usb: ulpi: rename operations {read|write}_dev to simply {read|write}

With the removal of the old {read|write} operations, we can now safely
rename the new api operations {read|write}_dev to use the shorter and
clearer names {read|write}, respectively.

Acked-by: Heikki Krogerus <heikki.krogerus@linux.intel.com>
Signed-off-by: Tal Shorer <tal.shorer@gmail.com>
Signed-off-by: Felipe Balbi <felipe.balbi@linux.intel.com>
8 years agousb: ulpi: remove calls to old api callbacks
Tal Shorer [Tue, 16 Aug 2016 16:04:49 +0000 (19:04 +0300)]
usb: ulpi: remove calls to old api callbacks

Now that all users use the new api callbacks, remove the old api
callbacks and force new interface drivers to use the new api.

Acked-by: Heikki Krogerus <heikki.krogerus@linux.intel.com>
Signed-off-by: Tal Shorer <tal.shorer@gmail.com>
Signed-off-by: Felipe Balbi <felipe.balbi@linux.intel.com>
8 years agousb: dwc3: ulpi: use new api
Tal Shorer [Tue, 16 Aug 2016 16:04:48 +0000 (19:04 +0300)]
usb: dwc3: ulpi: use new api

The old read, write callbacks in struct ulpi_ops have been deprecated
in favor of new callbacks that pass the parent device directly.
Replace the used callbacks in dwc3's ulpi component with the new api.

Reviewed-by: Heikki Krogerus <heikki.krogerus@linux.intel.com>
Signed-off-by: Tal Shorer <tal.shorer@gmail.com>
Signed-off-by: Felipe Balbi <felipe.balbi@linux.intel.com>
8 years agousb: ulpi: add new api functions, {read|write}_dev()
Tal Shorer [Tue, 16 Aug 2016 16:04:47 +0000 (19:04 +0300)]
usb: ulpi: add new api functions, {read|write}_dev()

Add these two new api callbacks to struct ulpi_ops. These are different
than read, write in that they pass the parent device directly instead
of via the ops argument.
They are intended to replace the old api functions.

If the new api callbacks are missing, revert to calling the old ones
as before.

Acked-by: Heikki Krogerus <heikki.krogerus@linux.intel.com>
Signed-off-by: Tal Shorer <tal.shorer@gmail.com>
Signed-off-by: Felipe Balbi <felipe.balbi@linux.intel.com>
8 years agousb: ulpi: move setting of ulpi->dev parent up in ulpi_register()
Tal Shorer [Tue, 16 Aug 2016 16:04:46 +0000 (19:04 +0300)]
usb: ulpi: move setting of ulpi->dev parent up in ulpi_register()

Once ulpi operations use the parent device directly, this will be
needed during the operations used in ulpi_register() itself, so set
the parent field before calling any ulpi operations.

Acked-by: Heikki Krogerus <heikki.krogerus@linux.intel.com>
Signed-off-by: Tal Shorer <tal.shorer@gmail.com>
Signed-off-by: Felipe Balbi <felipe.balbi@linux.intel.com>
8 years agoDocumentation: devicetree: dwc2: Deprecate g-tx-fifo-size
John Youn [Mon, 29 Aug 2016 20:39:03 +0000 (13:39 -0700)]
Documentation: devicetree: dwc2: Deprecate g-tx-fifo-size

This property is not needed because the periodic fifos are not
configurable. So it was incorrect to add this property in the first
place.

Acked-by: Rob Herring <robh@kernel.org>
Signed-off-by: John Youn <johnyoun@synopsys.com>
Signed-off-by: Felipe Balbi <felipe.balbi@linux.intel.com>
8 years agousb: dwc2: gadget: free TX FIFO after killing requests
Robert Baldyga [Mon, 29 Aug 2016 20:39:00 +0000 (13:39 -0700)]
usb: dwc2: gadget: free TX FIFO after killing requests

As kill_all_requests() potentially flushes TX FIFO, we should should
free FIFO after calling it. Otherwise FIFO could stay unflushed properly.

Signed-off-by: Robert Baldyga <r.baldyga@samsung.com>
Signed-off-by: John Youn <johnyoun@synopsys.com>
Signed-off-by: Felipe Balbi <felipe.balbi@linux.intel.com>
8 years agousb: dwc2: gadget: remove dead code from dwc2_hsotg_ep_enable()
Robert Baldyga [Mon, 29 Aug 2016 20:38:57 +0000 (13:38 -0700)]
usb: dwc2: gadget: remove dead code from dwc2_hsotg_ep_enable()

Since FIFO is always freed in dwc2_hsotg_ep_disable(), ep->fifo_index
is always 0 in dwc2_hsotg_ep_enable(), hence code inside if() block is
never executed.

Signed-off-by: Robert Baldyga <r.baldyga@samsung.com>
Signed-off-by: John Youn <johnyoun@synopsys.com>
Signed-off-by: Felipe Balbi <felipe.balbi@linux.intel.com>
8 years agousb: dwc2: gadget: change variable name to more meaningful
Robert Baldyga [Mon, 29 Aug 2016 20:38:55 +0000 (13:38 -0700)]
usb: dwc2: gadget: change variable name to more meaningful

Since we handle FIFOs and endpoint separately, using variable named 'ep'
in context of FIFO is misleading, hence we rename it to 'fifo'.

Signed-off-by: Robert Baldyga <r.baldyga@samsung.com>
Signed-off-by: John Youn <johnyoun@synopsys.com>
Signed-off-by: Felipe Balbi <felipe.balbi@linux.intel.com>
8 years agousb: dwc2: gadget: fix TX FIFO size and address initialization
Robert Baldyga [Mon, 29 Aug 2016 20:38:52 +0000 (13:38 -0700)]
usb: dwc2: gadget: fix TX FIFO size and address initialization

According to DWC2 documentation, DPTxFSize field of DPTXFSIZn register
is read only, which means that software cannot change FIFO size.

Register description says:
"The value of this register is the Largest Device Mode Periodic Tx Data
FIFO Depth (parameter OTG_TX_DPERIO_DFIFO_DEPTH_n), as specified during
coreConsultant configuration."

That means, that we have to setup only FIFO start addresses (DPTxFStAddr),
taking into account reset values of DPTxFSize.

Initialize FIFO start addresses properly and remove unneeded core related
to incorrect FIFO size initialization.

Signed-off-by: Robert Baldyga <r.baldyga@samsung.com>
Signed-off-by: John Youn <johnyoun@synopsys.com>
Signed-off-by: Felipe Balbi <felipe.balbi@linux.intel.com>
8 years agousb: dwc2: gadget: use ep->fifo_index in context of FIFO registers
Robert Baldyga [Mon, 29 Aug 2016 20:38:50 +0000 (13:38 -0700)]
usb: dwc2: gadget: use ep->fifo_index in context of FIFO registers

In context of FIFO registers we use ep->fifo_index instead of ep->index.

Signed-off-by: Robert Baldyga <r.baldyga@samsung.com>
Signed-off-by: John Youn <johnyoun@synopsys.com>
Signed-off-by: Felipe Balbi <felipe.balbi@linux.intel.com>
8 years agousb: gadget: configfs: log function unbinding as debug
Romain Izard [Mon, 29 Aug 2016 09:22:29 +0000 (12:22 +0300)]
usb: gadget: configfs: log function unbinding as debug

Disabling USB gadget functions configured through configfs is something
that can happen in normal use cases. Keep the existing log for this type
of event, but only as debug, not as an error.

Signed-off-by: Romain Izard <romain.izard.pro@gmail.com>
Signed-off-by: Felipe Balbi <felipe.balbi@linux.intel.com>
8 years agousb: gadget: uvc: Add missing call for additional setup data
Petr Cvek [Wed, 17 Aug 2016 10:36:57 +0000 (12:36 +0200)]
usb: gadget: uvc: Add missing call for additional setup data

Some UVC commands require additional data (non zero uvc->event_length).
Add usb_ep_queue() call, so uvc_function_ep0_complete() can be called
and send received data to the userspace.

Signed-off-by: Petr Cvek <petr.cvek@tul.cz>
Signed-off-by: Felipe Balbi <felipe.balbi@linux.intel.com>
8 years agousb: renesas_usbhs: set quirk_avoids_skb_reserve if USB-DMAC is used
Yoshihiro Shimoda [Mon, 22 Aug 2016 08:48:28 +0000 (17:48 +0900)]
usb: renesas_usbhs: set quirk_avoids_skb_reserve if USB-DMAC is used

This patch sets the quirk_avoids_skb_reserve flag to improve performance
of a network gadget driver (e.g. f_ncm.c) if USB-DMAC is used.

For example (on r8a7795 board + f_ncm.c + iperf udp mode / receiving):
  - without this patch:  90.3 Mbits/sec
  - with    this patch: 273   Mbits/sec

Signed-off-by: Yoshihiro Shimoda <yoshihiro.shimoda.uh@renesas.com>
Signed-off-by: Felipe Balbi <felipe.balbi@linux.intel.com>
8 years agousb: gadget: f_ncm: add support for no_skb_reserve
Yoshihiro Shimoda [Mon, 22 Aug 2016 08:48:27 +0000 (17:48 +0900)]
usb: gadget: f_ncm: add support for no_skb_reserve

This patch adds to support no_skb_reserve function to improve
performance for some platforms. About the detail, please refer to
the commit log of "quirk_avoids_skb_reserve" in
include/linux/usb/gadget.h.

Signed-off-by: Yoshihiro Shimoda <yoshihiro.shimoda.uh@renesas.com>
Signed-off-by: Felipe Balbi <felipe.balbi@linux.intel.com>
8 years agousb: gadget: u_ether: add a flag to avoid skb_reserve() calling
Yoshihiro Shimoda [Mon, 22 Aug 2016 08:48:26 +0000 (17:48 +0900)]
usb: gadget: u_ether: add a flag to avoid skb_reserve() calling

This patch adds a flag "no_skb_reserve" in struct eth_dev.
So, if a peripheral driver sets the quirk_avoids_skb_reserve flag,
upper network gadget drivers (e.g. f_ncm.c) can avoid skb_reserve()
calling using the flag as well.

Signed-off-by: Yoshihiro Shimoda <yoshihiro.shimoda.uh@renesas.com>
Signed-off-by: Felipe Balbi <felipe.balbi@linux.intel.com>
8 years agousb: gadget: add a new quirk to avoid skb_reserve in u_ether.c
Yoshihiro Shimoda [Mon, 22 Aug 2016 08:48:25 +0000 (17:48 +0900)]
usb: gadget: add a new quirk to avoid skb_reserve in u_ether.c

Some platforms (e.g. USB-DMAC on R-Car SoCs) has memory alignment
restriction. If memory alignment is not match, the usb peripheral
driver decides not to use the DMA controller. Then, the performance
is not good.

In the case of u_ether.c, since it calls skb_reserve() in rx_submit(),
it is possible to cause memory alignment mismatch.

So, this patch adds a new quirk "quirk_avoids_skb_reserve" to avoid
skb_reserve() calling in u_ether.c to improve performance.

A peripheral driver will set this flag and network gadget drivers
(e.g. f_ncm.c) will reference the flag via gadget_avoids_skb_reserve().

Signed-off-by: Yoshihiro Shimoda <yoshihiro.shimoda.uh@renesas.com>
Signed-off-by: Felipe Balbi <felipe.balbi@linux.intel.com>
8 years agousb: gadget: net2280: fix typo: "Inavlid" -> "Invalid"
Colin Ian King [Wed, 24 Aug 2016 06:44:19 +0000 (07:44 +0100)]
usb: gadget: net2280: fix typo: "Inavlid" -> "Invalid"

trivial typo fix in dev_err message

Signed-off-by: Colin Ian King <colin.king@canonical.com>
Signed-off-by: Felipe Balbi <felipe.balbi@linux.intel.com>
8 years agousb: renesas_usbhs: add a compatible string for r8a7796
Yoshihiro Shimoda [Wed, 24 Aug 2016 07:39:53 +0000 (16:39 +0900)]
usb: renesas_usbhs: add a compatible string for r8a7796

This patch adds support for r8a7796 (R-Car M3-W).

Acked-by: Geert Uytterhoeven <geert+renesas@glider.be>
Acked-by: Rob Herring <robh@kernel.org>
Signed-off-by: Yoshihiro Shimoda <yoshihiro.shimoda.uh@renesas.com>
Signed-off-by: Felipe Balbi <felipe.balbi@linux.intel.com>
8 years agousb: gadget: goku_udc: fix memory leak in goku_probe()
Alexey Khoroshilov [Fri, 26 Aug 2016 00:06:02 +0000 (03:06 +0300)]
usb: gadget: goku_udc: fix memory leak in goku_probe()

Memory allocated for goku_udc device is not deallocated at error
paths in goku_probe(), because gadget_release() destructor
is not registered yet.

Found by Linux Driver Verification project (linuxtesting.org).

Signed-off-by: Alexey Khoroshilov <khoroshilov@ispras.ru>
Signed-off-by: Felipe Balbi <felipe.balbi@linux.intel.com>
8 years agousb: renesas_usbhs: pipe: don't print on ENOMEM
Wolfram Sang [Thu, 25 Aug 2016 17:39:30 +0000 (19:39 +0200)]
usb: renesas_usbhs: pipe: don't print on ENOMEM

All kmalloc-based functions print enough information on failures.

Signed-off-by: Wolfram Sang <wsa-dev@sang-engineering.com>
Signed-off-by: Felipe Balbi <felipe.balbi@linux.intel.com>
8 years agousb: renesas_usbhs: mod_host: don't print on ENOMEM
Wolfram Sang [Thu, 25 Aug 2016 17:39:29 +0000 (19:39 +0200)]
usb: renesas_usbhs: mod_host: don't print on ENOMEM

All kmalloc-based functions print enough information on failures.

Signed-off-by: Wolfram Sang <wsa-dev@sang-engineering.com>
Signed-off-by: Felipe Balbi <felipe.balbi@linux.intel.com>
8 years agousb: renesas_usbhs: mod_gadget: don't print on ENOMEM
Wolfram Sang [Thu, 25 Aug 2016 17:39:28 +0000 (19:39 +0200)]
usb: renesas_usbhs: mod_gadget: don't print on ENOMEM

All kmalloc-based functions print enough information on failures.

Signed-off-by: Wolfram Sang <wsa-dev@sang-engineering.com>
Signed-off-by: Felipe Balbi <felipe.balbi@linux.intel.com>
8 years agousb: gadget: udc: udc-xilinx: don't print on ENOMEM
Wolfram Sang [Thu, 25 Aug 2016 17:39:05 +0000 (19:39 +0200)]
usb: gadget: udc: udc-xilinx: don't print on ENOMEM

All kmalloc-based functions print enough information on failures.

Signed-off-by: Wolfram Sang <wsa-dev@sang-engineering.com>
Signed-off-by: Felipe Balbi <felipe.balbi@linux.intel.com>
8 years agousb: gadget: udc: goku_udc: don't print on ENOMEM
Wolfram Sang [Thu, 25 Aug 2016 17:39:04 +0000 (19:39 +0200)]
usb: gadget: udc: goku_udc: don't print on ENOMEM

All kmalloc-based functions print enough information on failures.

Signed-off-by: Wolfram Sang <wsa-dev@sang-engineering.com>
Signed-off-by: Felipe Balbi <felipe.balbi@linux.intel.com>
8 years agousb: gadget: udc: fsl_qe_udc: don't print on ENOMEM
Wolfram Sang [Thu, 25 Aug 2016 17:39:03 +0000 (19:39 +0200)]
usb: gadget: udc: fsl_qe_udc: don't print on ENOMEM

All kmalloc-based functions print enough information on failures.

Signed-off-by: Wolfram Sang <wsa-dev@sang-engineering.com>
Signed-off-by: Felipe Balbi <felipe.balbi@linux.intel.com>
8 years agousb: dwc2: gadget: don't print on ENOMEM
Wolfram Sang [Thu, 25 Aug 2016 17:39:02 +0000 (19:39 +0200)]
usb: dwc2: gadget: don't print on ENOMEM

All kmalloc-based functions print enough information on failures.

Signed-off-by: Wolfram Sang <wsa-dev@sang-engineering.com>
Signed-off-by: Felipe Balbi <felipe.balbi@linux.intel.com>
8 years agousb: phy: ab8500-usb: fix spelling mistake "regester" -> "register"
Colin Ian King [Sun, 28 Aug 2016 11:16:19 +0000 (12:16 +0100)]
usb: phy: ab8500-usb: fix spelling mistake "regester" -> "register"

Trivial fix to spelling mistakes in dev_err messages.

Signed-off-by: Colin Ian King <colin.king@canonical.com>
Signed-off-by: Felipe Balbi <felipe.balbi@linux.intel.com>
8 years agousb: gadget: don't couple configfs to legacy gadgets
Felipe Balbi [Fri, 26 Aug 2016 09:21:34 +0000 (12:21 +0300)]
usb: gadget: don't couple configfs to legacy gadgets

It's perfectly fine to have all configfs functions
built-in while having modular legacy gadgets. Let's
allow for that.

Signed-off-by: Felipe Balbi <felipe.balbi@linux.intel.com>
8 years agousb: dwc3: gadget: stop kicking if we run out of space
Felipe Balbi [Wed, 24 Aug 2016 11:38:10 +0000 (14:38 +0300)]
usb: dwc3: gadget: stop kicking if we run out of space

In case our TRB ring is full, we can avoid trying to
kick transfers which won't start and just add requests
to the queue.

Signed-off-by: Felipe Balbi <felipe.balbi@linux.intel.com>
8 years agousb: dwc3: gadget: abolish trbs_left
Felipe Balbi [Wed, 24 Aug 2016 11:37:22 +0000 (14:37 +0300)]
usb: dwc3: gadget: abolish trbs_left

Instead, we can always rely on dwc3_calc_trbs_left()
directly.

Signed-off-by: Felipe Balbi <felipe.balbi@linux.intel.com>
8 years agousb: gadget: f_hid: use alloc_ep_req()
Felipe F. Tonello [Tue, 23 Aug 2016 17:24:51 +0000 (18:24 +0100)]
usb: gadget: f_hid: use alloc_ep_req()

Use gadget's framework allocation function instead of directly calling
usb_ep_alloc_request().

Signed-off-by: Felipe F. Tonello <eu@felipetonello.com>
Signed-off-by: Felipe Balbi <felipe.balbi@linux.intel.com>
8 years agousb: gadget: f_hid: use free_ep_req()
Felipe F. Tonello [Tue, 23 Aug 2016 17:24:50 +0000 (18:24 +0100)]
usb: gadget: f_hid: use free_ep_req()

We should always use free_ep_req() when allocating requests with
alloc_ep_req().

Signed-off-by: Felipe F. Tonello <eu@felipetonello.com>
Signed-off-by: Felipe Balbi <felipe.balbi@linux.intel.com>
8 years agousb: gadget: remove useless parameter in alloc_ep_req()
Felipe F. Tonello [Tue, 23 Aug 2016 17:24:49 +0000 (18:24 +0100)]
usb: gadget: remove useless parameter in alloc_ep_req()

The default_length parameter of alloc_ep_req was not really necessary
and gadget drivers would almost always create an inline function to pass
the same value to len and default_len.

This patch removes that parameter and updates all calls to alloc_ep_req() to
use the new API.

Signed-off-by: Felipe F. Tonello <eu@felipetonello.com>
Signed-off-by: Felipe Balbi <felipe.balbi@linux.intel.com>
8 years agousb: dwc3: Add ENDXFER command polling
John Youn [Mon, 22 Aug 2016 22:39:13 +0000 (15:39 -0700)]
usb: dwc3: Add ENDXFER command polling

ENDXFER polling is available on version 3.10a and later of the
DWC_usb3 (USB 3.0) controller. With this feature, the software can poll
the CMDACT bit in the DEPCMD register after issuing an ENDXFER command.
This feature is enabled by writing GUCTL2[14].

This feature is NOT available on the DWC_usb31 (USB 3.1) IP.

Signed-off-by: John Youn <johnyoun@synopsys.com>
Signed-off-by: Felipe Balbi <felipe.balbi@linux.intel.com>
8 years agousb: phy: mxs: Add DT bindings to configure TX settings
Jaret Cantu [Tue, 16 Aug 2016 22:31:48 +0000 (18:31 -0400)]
usb: phy: mxs: Add DT bindings to configure TX settings

The TX settings can be calibrated for particular hardware.  The
phy is reset by Linux, so this cannot be handled by the bootloader.

The TRM mentions that the maximum resistance should be used for the
DN/DP calibration in order to pass USB certification.

The values for the TX registers are poorly described in the TRM.
The meanings of the register values were taken from another
NXP-provided document:
https://community.nxp.com/message/566147#comment-566912

Acked-by: Peter Chen <peter.chen@nxp.com>
Acked-by: Rob Herring <robh@kernel.org>
Signed-off-by: Jaret Cantu <jaret.cantu@timesys.com>
Signed-off-by: Felipe Balbi <felipe.balbi@linux.intel.com>
8 years agoDocumentation: dt: dwc3: note the supported phy-names
Brian Norris [Thu, 18 Aug 2016 21:37:16 +0000 (14:37 -0700)]
Documentation: dt: dwc3: note the supported phy-names

The dwc3 driver expicitly looks for "usb2-phy" or "usb3-phy", but we
never noted these names in the documentation.

Acked-by: Rob Herring <robh@kernel.org>
Signed-off-by: Brian Norris <briannorris@chromium.org>
Signed-off-by: Felipe Balbi <felipe.balbi@linux.intel.com>
8 years agousb: dwc3: Add revision numbers for the USB 3.0 IP
John Youn [Fri, 19 Aug 2016 18:57:52 +0000 (11:57 -0700)]
usb: dwc3: Add revision numbers for the USB 3.0 IP

Add revision number constants for the 3.00a and 3.10a releases.

Signed-off-by: John Youn <johnyoun@synopsys.com>
Signed-off-by: Felipe Balbi <felipe.balbi@linux.intel.com>
8 years agousb: gadget: f_fs: handle control requests in config 0
Felix Hädicke [Tue, 21 Jun 2016 23:12:09 +0000 (01:12 +0200)]
usb: gadget: f_fs: handle control requests in config 0

Introduces a new FunctionFS descriptor flag named
FUNCTIONFS_CONFIG0_SETUP.

When this flag is enabled, FunctionFS userspace drivers can process
non-standard control requests in configuration 0.

Signed-off-by: Felix Hädicke <felixhaedicke@web.de>
Signed-off-by: Felipe Balbi <felipe.balbi@linux.intel.com>
8 years agousb: gadget: composite: let USB functions process ctrl reqs in cfg0
Felix Hädicke [Tue, 21 Jun 2016 23:12:08 +0000 (01:12 +0200)]
usb: gadget: composite: let USB functions process ctrl reqs in cfg0

It can sometimes be necessary for gadget drivers to process non-standard
control requests, which host devices can send without having sent
USB_REQ_SET_CONFIGURATION.

Therefore, the req_match() usb_function method is enhanced with the new
parameter "config0". When a USB configuration is active, this parameter
is false. When a non-core control request is processed in
composite_setup(), without an active configuration, req_match() of the
USB functions of all available configurations which implement this
function, is called with config0=true. Then the control request gets
processed by the first usb_function instance whose req_match() returns
true.

Signed-off-by: Felix Hädicke <felixhaedicke@web.de>
Signed-off-by: Felipe Balbi <felipe.balbi@linux.intel.com>
8 years agousb: gadget: f_fs: handle control requests not directed to interface or endpoint
Felix Hädicke [Tue, 21 Jun 2016 23:12:07 +0000 (01:12 +0200)]
usb: gadget: f_fs: handle control requests not directed to interface or endpoint

Introduces a new FunctionFS descriptor flag named
FUNCTIONFS_ALL_CTRL_RECIP. When this flag is enabled, control requests,
which are not explicitly directed to an interface or endpoint, can be
handled.

This allows FunctionFS userspace drivers to process non-standard
control requests.

Signed-off-by: Felix Hädicke <felixhaedicke@web.de>
Signed-off-by: Felipe Balbi <felipe.balbi@linux.intel.com>
8 years agousb: gadget: f_hid: add dev to configfs
Johannes Berg [Thu, 23 Jun 2016 20:28:54 +0000 (22:28 +0200)]
usb: gadget: f_hid: add dev to configfs

Even if the /dev/hidg* chardev is automatically created, one
has to guess which one belongs to which function. In the case
of multiple HID functions, or maybe even multiple peripherals,
this becomes difficult.

Add the dev (with major and minor number) to configfs to allow
looking up (or even creating) the right device node for each
function. This file is read-only.

Signed-off-by: Johannes Berg <johannes@sipsolutions.net>
Signed-off-by: Felipe Balbi <felipe.balbi@linux.intel.com>
8 years agousb: dwc3: core: Move the mode setting to the right place
Baolin Wang [Fri, 15 Jul 2016 09:13:27 +0000 (17:13 +0800)]
usb: dwc3: core: Move the mode setting to the right place

When dwc3 core enters into suspend mode, the system (especially for mobile
device) may power off the dwc3 controller for power saving, that will cause
dwc3 controller lost the mode operation when resuming dwc3 core.

Thus we can move the mode setting into dwc3_core_init() function to avoid this
issue.

Signed-off-by: Baolin Wang <baolin.wang@linaro.org>
Signed-off-by: Felipe Balbi <felipe.balbi@linux.intel.com>
8 years agousb: gadget: f_midi: drop substreams when disabling endpoint
Felipe F. Tonello [Mon, 8 Aug 2016 20:30:10 +0000 (21:30 +0100)]
usb: gadget: f_midi: drop substreams when disabling endpoint

This change makes sure that the ALSA buffers are cleaned if an endpoint
becomes disabled.

Before this change, if the internal ALSA buffer did overflow, the MIDI
function would stop sending MIDI to the host.

Signed-off-by: Felipe F. Tonello <eu@felipetonello.com>
Signed-off-by: Felipe Balbi <felipe.balbi@linux.intel.com>
8 years agousb: gadget: f_midi: refactor state machine
Felipe F. Tonello [Mon, 8 Aug 2016 20:30:09 +0000 (21:30 +0100)]
usb: gadget: f_midi: refactor state machine

This refactor results in a cleaner state machine code and promotes
consistency, readability, and maintanability of this driver.

This refactor state machine was well tested and it is currently running in
production code and devices.

Signed-off-by: Felipe F. Tonello <eu@felipetonello.com>
Signed-off-by: Felipe Balbi <felipe.balbi@linux.intel.com>
8 years agousb: gadget: f_midi: defaults buflen sizes to 512
Felipe F. Tonello [Mon, 8 Aug 2016 20:30:08 +0000 (21:30 +0100)]
usb: gadget: f_midi: defaults buflen sizes to 512

512 is the value used by wMaxPacketSize, as specified by the USB Spec. This
makes sure this driver uses, by default, the most optimal value for IN and OUT
endpoint requests.

Signed-off-by: Felipe F. Tonello <eu@felipetonello.com>
Signed-off-by: Felipe Balbi <felipe.balbi@linux.intel.com>
8 years agousb: gadget: f_midi: remove alignment code for OUT endpoint
Felipe F. Tonello [Mon, 8 Aug 2016 20:30:07 +0000 (21:30 +0100)]
usb: gadget: f_midi: remove alignment code for OUT endpoint

The new version of alloc_ep_req() already aligns the buffer size to
wMaxPacketSize on OUT endpoints.

Signed-off-by: Felipe F. Tonello <eu@felipetonello.com>
Signed-off-by: Felipe Balbi <felipe.balbi@linux.intel.com>
8 years agousb: gadget: align buffer size when allocating for OUT endpoint
Felipe F. Tonello [Mon, 8 Aug 2016 20:30:06 +0000 (21:30 +0100)]
usb: gadget: align buffer size when allocating for OUT endpoint

Using usb_ep_align() makes sure that the buffer size for OUT endpoints is
always aligned with wMaxPacketSize (512 usually). This makes sure
that no buffer has the wrong size, which can cause nasty bugs.

Signed-off-by: Felipe F. Tonello <eu@felipetonello.com>
Signed-off-by: Felipe Balbi <felipe.balbi@linux.intel.com>
8 years agousb: gadget: change len to size_t on alloc_ep_req()
Felipe F. Tonello [Mon, 8 Aug 2016 20:30:05 +0000 (21:30 +0100)]
usb: gadget: change len to size_t on alloc_ep_req()

Length of buffers should be of type size_t whenever possible. Altough
recommended, this change has no real practical change, unless a driver has a
uses a huge or negative buffer size - it might help find these bugs.

Signed-off-by: Felipe F. Tonello <eu@felipetonello.com>
Signed-off-by: Felipe Balbi <felipe.balbi@linux.intel.com>
8 years agousb: gadget: fix usb_ep_align_maybe endianness and new usb_ep_align
Felipe F. Tonello [Mon, 8 Aug 2016 20:30:04 +0000 (21:30 +0100)]
usb: gadget: fix usb_ep_align_maybe endianness and new usb_ep_align

USB spec specifies wMaxPacketSize to be little endian (as other properties),
so when using this variable in the driver we should convert to the current
CPU endianness if necessary.

This patch also introduces usb_ep_align() which does always returns the
aligned buffer size for an endpoint. This is useful to be used by USB requests
allocator functions.

Signed-off-by: Felipe F. Tonello <eu@felipetonello.com>
Signed-off-by: Felipe Balbi <felipe.balbi@linux.intel.com>
8 years agousb: gadget: Add per-lun inquiry string
Philipp Gesang [Mon, 1 Aug 2016 15:04:19 +0000 (17:04 +0200)]
usb: gadget: Add per-lun inquiry string

Introduce an attribute "inquiry_string" to the lun.

In some environments, e. g. BIOS boot menus, the inquiry string
is the only information about devices presented to the user. The
default string depends on the "cdrom" bit of the first lun as
well as the kernel version and allows no further customization.
So without access to the client it is not obvious which gadget is
active at a given point and what any of the available luns might
contain.

If "inquiry_string" is ignored or set to the empty string, the
old behavior is preserved.

Signed-off-by: Philipp Gesang <philipp.gesang@intra2net.com>
Signed-off-by: Felipe Balbi <felipe.balbi@linux.intel.com>
8 years agousb: dwc3: of-simple: add compatible for rockchip rk3399
William Wu [Tue, 16 Aug 2016 14:44:36 +0000 (22:44 +0800)]
usb: dwc3: of-simple: add compatible for rockchip rk3399

Rockchip platform merely enable usb3 clocks and
populate its children. So we can use this generic
glue layer to support Rockchip dwc3.

Signed-off-by: William Wu <william.wu@rock-chips.com>
Signed-off-by: Felipe Balbi <felipe.balbi@linux.intel.com>
8 years agousb: dwc3: rockchip: add devicetree bindings documentation
William Wu [Tue, 16 Aug 2016 14:46:41 +0000 (22:46 +0800)]
usb: dwc3: rockchip: add devicetree bindings documentation

This patch adds the devicetree documentation required for Rockchip
USB3.0 core wrapper consisting of USB3.0 IP from Synopsys.

It supports DRD mode, and could operate in device mode (SS, HS, FS)
and host mode (SS, HS, FS, LS).

Signed-off-by: William Wu <william.wu@rock-chips.com>
Acked-by: Rob Herring <robh@kernel.org>
Signed-off-by: Felipe Balbi <felipe.balbi@linux.intel.com>
8 years agousb: dwc3: add dis_del_phy_power_chg_quirk
William Wu [Tue, 16 Aug 2016 14:44:39 +0000 (22:44 +0800)]
usb: dwc3: add dis_del_phy_power_chg_quirk

Add a quirk to clear the GUSB3PIPECTL.DELAYP1TRANS bit,
which specifies whether disable delay PHY power change
from P0 to P1/P2/P3 when link state changing from U0
to U1/U2/U3 respectively.

Signed-off-by: William Wu <william.wu@rock-chips.com>
Acked-by: Rob Herring <robh@kernel.org>
Signed-off-by: Felipe Balbi <felipe.balbi@linux.intel.com>
8 years agousb: dwc3: make usb2 phy utmi interface configurable
William Wu [Tue, 16 Aug 2016 14:44:38 +0000 (22:44 +0800)]
usb: dwc3: make usb2 phy utmi interface configurable

Support to configure the UTMI+ PHY with an 8- or 16-bit
interface via DT. The UTMI+ PHY interface is a hardware
capability, and it's platform dependent. Normally, the
PHYIF can be configured during coreconsultant.

But for some specific USB cores(e.g. rk3399 SoC DWC3),
the default PHYIF configuration value is false, so we
need to reconfigure it by software.

Signed-off-by: William Wu <william.wu@rock-chips.com>
Acked-by: Rob Herring <robh@kernel.org>
Signed-off-by: Felipe Balbi <felipe.balbi@linux.intel.com>
8 years agousb: dwc3: add dis_u2_freeclk_exists_quirk
William Wu [Tue, 16 Aug 2016 14:44:37 +0000 (22:44 +0800)]
usb: dwc3: add dis_u2_freeclk_exists_quirk

Add a quirk to clear the GUSB2PHYCFG.U2_FREECLK_EXISTS bit,
which specifies whether the USB2.0 PHY provides a free-running
PHY clock, which is active when the clock control input is active.

Signed-off-by: William Wu <william.wu@rock-chips.com>
Acked-by: Rob Herring <robh@kernel.org>
Signed-off-by: Felipe Balbi <felipe.balbi@linux.intel.com>
8 years agousb: gadget: net2280: match interrupt endpoints to PIO endpoints and DMA to bulk
Jussi Kivilinna [Fri, 12 Aug 2016 14:29:35 +0000 (17:29 +0300)]
usb: gadget: net2280: match interrupt endpoints to PIO endpoints and DMA to bulk

With composite gadget (ACM + NCM), USB3380 to host TCP transfer
speed dropped to 150 Mbit/s compared to 900 Mbit/s with NCM
gadget. Problem seems to be that net2280/USB3380 has only four
DMA channels and those DMA channels are allocated to first HW
endpoints. Endpoint match function was mapping endpoint names
directly, so NCM did not get DMA for bulk endpoints.

This patch changed match_ep to prefer DMA enabled hw endpoints
for bulk usb endpoints and PIO for interrupt usb endpoints.

Signed-off-by: Jussi Kivilinna <jussi.kivilinna@haltian.com>
Signed-off-by: Felipe Balbi <felipe.balbi@linux.intel.com>
8 years agousb: gadget: net2280: fix infinite loop in irq handler
Jussi Kivilinna [Fri, 12 Aug 2016 14:29:34 +0000 (17:29 +0300)]
usb: gadget: net2280: fix infinite loop in irq handler

With SuperSpeed CDC NCM gadget, net2280 would get stuck in
'handle_ep_small' function. Triggering issue requires large
TCP transfer from host to USB3380.

Patch adds check for stuck condition and prevents hard lockup.

Signed-off-by: Jussi Kivilinna <jussi.kivilinna@haltian.com>
Signed-off-by: Felipe Balbi <felipe.balbi@linux.intel.com>
8 years agousb: gadget: f_ncm: add SuperSpeed descriptors for CDC NCM
Jussi Kivilinna [Fri, 12 Aug 2016 14:28:05 +0000 (17:28 +0300)]
usb: gadget: f_ncm: add SuperSpeed descriptors for CDC NCM

Patch enables SuperSpeed for NCM gadget.

Tested with USB3380 and measured TCP throughput with two Intel PCs:
  udc to host: 920 Mbit/s
  host to udc: 550 Mbit/s

Signed-off-by: Jussi Kivilinna <jussi.kivilinna@haltian.com>
Signed-off-by: Felipe Balbi <felipe.balbi@linux.intel.com>
8 years agousb: dwc2: Remove deprecated create_singlethread_workqueue
Bhaktipriya Shridhar [Thu, 28 Jul 2016 08:27:29 +0000 (13:57 +0530)]
usb: dwc2: Remove deprecated create_singlethread_workqueue

alloc_ordered_workqueue replaces the deprecated
create_singlethread_workqueue.

There are multiple work items on the work queue, which require
ordering. Hence, an ordered workqueue has been used.

The workqueue "wq_otg" is not being used on a memory reclaim path.
Hence, WQ_MEM_RECLAIM has not been set.

Acked-by: John Youn <johnyoun@synopsys.com>
Signed-off-by: Bhaktipriya Shridhar <bhaktipriya96@gmail.com>
Signed-off-by: Felipe Balbi <felipe.balbi@linux.intel.com>
8 years agousb: gadget: remove redundant self assignment
Colin Ian King [Mon, 25 Jul 2016 21:57:36 +0000 (22:57 +0100)]
usb: gadget: remove redundant self assignment

The assignment ret = ret is redundant and can be removed.

Reviewed-by: Peter Chen <peter.chen@nxp.com>
Signed-off-by: Colin Ian King <colin.king@canonical.com>
Signed-off-by: Felipe Balbi <felipe.balbi@linux.intel.com>
8 years agousb: gadget: configfs: log function unbinding as information
Romain Izard [Tue, 26 Jul 2016 16:21:46 +0000 (18:21 +0200)]
usb: gadget: configfs: log function unbinding as information

Disabling USB gadget functions configured through configfs is something
that can happen in normal use cases. Keep the existing log for this type
of event, but only as information, not as an error.

Signed-off-by: Romain Izard <romain.izard.pro@gmail.com>
Signed-off-by: Felipe Balbi <felipe.balbi@linux.intel.com>
8 years agousb: phy: generic: remove the vbus dependency
Robert Jarzmik [Fri, 8 Jul 2016 20:09:07 +0000 (22:09 +0200)]
usb: phy: generic: remove the vbus dependency

As the last known user, ie. pxa27x_udc relying on calls to
usb_gadget_xxx() was amended to use the phy notifier, remove a bit the
USB stack adherence.

Actually the driver still uses the gadget API for structures definition,
but the implementation of USB gadget specific function usb_gadget_*() is
not necessary anymore.

Signed-off-by: Robert Jarzmik <robert.jarzmik@free.fr>
Signed-off-by: Felipe Balbi <felipe.balbi@linux.intel.com>
8 years agousb: phy: generic: cope with initial state
Robert Jarzmik [Fri, 8 Jul 2016 20:09:06 +0000 (22:09 +0200)]
usb: phy: generic: cope with initial state

In the gpio based case, the status of the phy is known at start by
reading the VBus gpio.

Actually, this is a fix, as this initial state, when not set up,
prevents a gadget to answer to the enumeration phase, as there is no
notification in this case (the VBus is already high when kernel boots)
so no interrupt is triggered, and the flow is :
 - gadget initializes
 - gadget gets its phy-generic with a xxx_get_phy_xxx() call type
 - gadget does a "set_peripheral()" call type
   => here if the otg->state is correctly filled, the proper vbus
   handling will be called, and the gadget will be aware it should
   answer enumeration and go forth

Without this fix, the USB cable must be removed and replugged for any
gadget relying on phy-generic and its gpio vbus handling to work.

The problem was seen on a pxa27x architecture based board on a
devicetree build.

Signed-off-by: Robert Jarzmik <robert.jarzmik@free.fr>
Signed-off-by: Felipe Balbi <felipe.balbi@linux.intel.com>
8 years agousb: gadget: pxa27x: add phy notifier event handler
Robert Jarzmik [Fri, 8 Jul 2016 20:09:05 +0000 (22:09 +0200)]
usb: gadget: pxa27x: add phy notifier event handler

In the legacy behavior, and USB phy, upon detection a VBus signal, was
calling usb_gadget_vbus_(dis)connect().

This model doesn't work if the phy is generic and doesn't have an
adherence to the gadget API.

Instead of relying on the phy to call the gadget API, hook up the phy
notifier to report the VBus event, and upon it call the usb gadget API
ourselves.

This brings a new ordering problem, as before even if the usb_get_phy()
was failing because the UDC was probed before the phy, the phy would
call the gadget anyway, making the VBus connection event forwarded to
the gadget. Now we rely on the notifier, we have to ensure the
xxx_get_phy() does indeed work.

In order to cope with this, it is assumed that :
 - for legacy platform_data machine, as the ordering cannot be ensured,
   the phy must call usb_gadget_vbus_(dis)connect, such as
   phy-gpio-vbus-usb.c
 - for new devicetree platforms, we'll rely on the probe deferral, and
   the phy can be gadget API agnostic.

Signed-off-by: Robert Jarzmik <robert.jarzmik@free.fr>
Signed-off-by: Felipe Balbi <felipe.balbi@linux.intel.com>
8 years agousb: dwc3: gadget: improve increment request->actual
Felipe Balbi [Fri, 12 Aug 2016 10:20:32 +0000 (13:20 +0300)]
usb: dwc3: gadget: improve increment request->actual

No functional changes, just a slight cosmetic
change.

Signed-off-by: Felipe Balbi <felipe.balbi@linux.intel.com>
8 years agousb: dwc3: gadget: remove condition that never happens
Felipe Balbi [Fri, 12 Aug 2016 10:19:20 +0000 (13:19 +0300)]
usb: dwc3: gadget: remove condition that never happens

We don't use LST bit anymore, so this condition will
never trigger.

Signed-off-by: Felipe Balbi <felipe.balbi@linux.intel.com>
8 years agousb: dwc3: gadget: add remaining sg entries to ring
Felipe Balbi [Fri, 12 Aug 2016 10:17:27 +0000 (13:17 +0300)]
usb: dwc3: gadget: add remaining sg entries to ring

Upon transfer completion after a full ring, let's
add more TRBs to our ring in order to complete our
request successfully.

Signed-off-by: Felipe Balbi <felipe.balbi@linux.intel.com>
8 years agousb: dwc3: gadget: interrupt on ring full too
Felipe Balbi [Fri, 12 Aug 2016 10:13:10 +0000 (13:13 +0300)]
usb: dwc3: gadget: interrupt on ring full too

If the ring is full and we are processing a big
sglist, then let's interrupt so we can, later, add
more TRBs to the ring.

Signed-off-by: Felipe Balbi <felipe.balbi@linux.intel.com>
8 years agousb: dwc3: gadget: add sg and num_pending_sgs to dwc3_request
Felipe Balbi [Fri, 12 Aug 2016 10:10:10 +0000 (13:10 +0300)]
usb: dwc3: gadget: add sg and num_pending_sgs to dwc3_request

These two fields will be used in a follow-up patch
to track how many entries of request's sglist we
have already processed. The reason is that if a
gadget driver sends an sglist with more entries then
we can fit in the ring, we will have to continue
processing remaining afterwards.

Signed-off-by: Felipe Balbi <felipe.balbi@linux.intel.com>
8 years agousb: dwc3: gadget: avoid while (1) loop on completion
Felipe Balbi [Thu, 11 Aug 2016 11:38:37 +0000 (14:38 +0300)]
usb: dwc3: gadget: avoid while (1) loop on completion

We know that we have to iterate over the list of
started requests. Instead of looping forever, we can
rely on list_for_each_entry(). Likewise, instead of
a do {} while loop over all, maybe available,
scatterlist entries, we can detect if $this request
uses scatterlist and rely on for_each_sg().

This makes the code easier to follow while making
sure that we will *always* break out of the loop.

Signed-off-by: Felipe Balbi <felipe.balbi@linux.intel.com>
8 years agousb: dwc3: gadget: simplify __dwc3_gadget_ep_queue()
Felipe Balbi [Thu, 11 Aug 2016 11:27:52 +0000 (14:27 +0300)]
usb: dwc3: gadget: simplify __dwc3_gadget_ep_queue()

Many of the comments in that function are really
outdated and don't match what the driver is
doing. Moreover, recent patches combined programming
model for all non-control endpoints, this gives us
an opportunity to get rid of our special cases in
__dwc3_gadget_ep_queue().

Signed-off-by: Felipe Balbi <felipe.balbi@linux.intel.com>
8 years agousb: dwc3: gadget: simplify dwc3_ep_prev_trb()
Felipe Balbi [Thu, 11 Aug 2016 09:26:59 +0000 (12:26 +0300)]
usb: dwc3: gadget: simplify dwc3_ep_prev_trb()

We always need to decrement our index by at least
one. Simplify the implementation by using a
temporary local variable and making sure that we
will always decrement one extra if tmp == 0.

Signed-off-by: Felipe Balbi <felipe.balbi@linux.intel.com>
8 years agousb: dwc3: gadget: increment dequeue pointer on completion
Felipe Balbi [Thu, 11 Aug 2016 09:24:27 +0000 (12:24 +0300)]
usb: dwc3: gadget: increment dequeue pointer on completion

Instead of waiting until giveback before
incrementing the dequeue pointer, we can increment
it from dwc3_cleanup_done_reqs(), that way we avoid
an extra loop over all TRBs during giveback.

While at that, also avoid using req->first_trb_index
as that's completely unnecessary. A follow-up patch
will clean up further uses of that and remove the
field altogether.

Signed-off-by: Felipe Balbi <felipe.balbi@linux.intel.com>
8 years agousb: dwc3: gadget: retire LST bit completely
Felipe Balbi [Wed, 10 Aug 2016 13:04:33 +0000 (16:04 +0300)]
usb: dwc3: gadget: retire LST bit completely

The only endpoint which actually requires LST bit
and XferComplete is ep0/1. Let's save some time by
completely removing LST bit support and
XferComplete.

This simplifies and consolidates endpoint handling
for all other 3 transfer types while also avoiding
extra interrupts.

Signed-off-by: Felipe Balbi <felipe.balbi@linux.intel.com>
8 years agoLinux 4.8-rc3
Linus Torvalds [Sun, 21 Aug 2016 23:14:10 +0000 (16:14 -0700)]
Linux 4.8-rc3

8 years agoMerge branch 'parisc-4.8-2' of git://git.kernel.org/pub/scm/linux/kernel/git/deller...
Linus Torvalds [Sun, 21 Aug 2016 21:28:24 +0000 (14:28 -0700)]
Merge branch 'parisc-4.8-2' of git://git./linux/kernel/git/deller/parisc-linux

Pull two parisc fixes from Helge Deller:
 "The first patch ensures that the high-res cr16 clocksource (which was
  added in kernel 4.7) gets choosen as default clocksource for parisc.

  The second patch moves the #define of EREFUSED down inside errno.h and
  thus unbreaks building the gccgo compiler"

* 'parisc-4.8-2' of git://git.kernel.org/pub/scm/linux/kernel/git/deller/parisc-linux:
  parisc: Fix order of EREFUSED define in errno.h
  parisc: Fix automatic selection of cr16 clocksource

8 years agoEDAC, skx_edac: Add EDAC driver for Skylake
Tony Luck [Sat, 20 Aug 2016 23:27:58 +0000 (16:27 -0700)]
EDAC, skx_edac: Add EDAC driver for Skylake

This is an entirely new driver instead of yet another set of patches
to sb_edac.c because:

1) Mapping from PCI devices to socket/memory controller is significantly
   different. Skylake scatters devices on a socket across a number of
   PCI buses.
2) There is an extra level of interleaving via the "mcroute" register
   that would be a little messy to squeeze into the old driver.
3) Validation is getting too expensive. Changes to sb_edac need to
   be checked against Sandy Bridge, Ivy Bridge, Haswell, Broadwell and
   Knights Landing.

Acked-by: Aristeu Rozanski <aris@redhat.com>
Acked-by: Borislav Petkov <bp@suse.de>
Signed-off-by: Tony Luck <tony.luck@intel.com>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
8 years agoparisc: Fix order of EREFUSED define in errno.h
Helge Deller [Sat, 20 Aug 2016 09:51:38 +0000 (11:51 +0200)]
parisc: Fix order of EREFUSED define in errno.h

When building gccgo in userspace, errno.h gets parsed and the go include file
sysinfo.go is generated.

Since EREFUSED is defined to the same value as ECONNREFUSED, and ECONNREFUSED
is defined later on in errno.h, this leads to go complaining that EREFUSED
isn't defined yet.

Fix this trivial problem by moving the define of EREFUSED down after
ECONNREFUSED in errno.h (and clean up the indenting while touching this line).

Signed-off-by: Helge Deller <deller@gmx.de>
Cc: stable@vger.kernel.org
8 years agoparisc: Fix automatic selection of cr16 clocksource
Helge Deller [Fri, 19 Aug 2016 20:39:02 +0000 (22:39 +0200)]
parisc: Fix automatic selection of cr16 clocksource

Commit 54b66800907 (parisc: Add native high-resolution sched_clock()
implementation) added support to use the CPU-internal cr16 counters as reliable
clocksource with the help of HAVE_UNSTABLE_SCHED_CLOCK.

Sadly the commit missed to remove the hack which prevented cr16 to become the
default clocksource even on SMP systems.

Signed-off-by: Helge Deller <deller@gmx.de>
Cc: stable@vger.kernel.org # 4.7+
8 years agoMake the hardened user-copy code depend on having a hardened allocator
Linus Torvalds [Fri, 19 Aug 2016 19:47:01 +0000 (12:47 -0700)]
Make the hardened user-copy code depend on having a hardened allocator

The kernel test robot reported a usercopy failure in the new hardened
sanity checks, due to a page-crossing copy of the FPU state into the
task structure.

This happened because the kernel test robot was testing with SLOB, which
doesn't actually do the required book-keeping for slab allocations, and
as a result the hardening code didn't realize that the task struct
allocation was one single allocation - and the sanity checks fail.

Since SLOB doesn't even claim to support hardening (and you really
shouldn't use it), the straightforward solution is to just make the
usercopy hardening code depend on the allocator supporting it.

Reported-by: kernel test robot <xiaolong.ye@intel.com>
Cc: Kees Cook <keescook@chromium.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
8 years agoMerge branch 'i2c/for-current' of git://git.kernel.org/pub/scm/linux/kernel/git/wsa...
Linus Torvalds [Fri, 19 Aug 2016 19:10:06 +0000 (12:10 -0700)]
Merge branch 'i2c/for-current' of git://git./linux/kernel/git/wsa/linux

Pull i2c fixes from Wolfram Sang:
 "I2C has some pretty standard driver bugfixes and one minor cleanup"

* 'i2c/for-current' of git://git.kernel.org/pub/scm/linux/kernel/git/wsa/linux:
  i2c: meson: Use complete() instead of complete_all()
  i2c: brcmstb: Use complete() instead of complete_all()
  i2c: bcm-kona: Use complete() instead of complete_all()
  i2c: bcm-iproc: Use complete() instead of complete_all()
  i2c: at91: fix support of the "alternative command" feature
  i2c: ocores: add missed clk_disable_unprepare() on failure paths
  i2c: cros-ec-tunnel: Fix usage of cros_ec_cmd_xfer()
  i2c: mux: demux-pinctrl: properly roll back when adding adapter fails

8 years agoMerge tag 'dm-4.8-fixes-2' of git://git.kernel.org/pub/scm/linux/kernel/git/device...
Linus Torvalds [Fri, 19 Aug 2016 16:32:48 +0000 (09:32 -0700)]
Merge tag 'dm-4.8-fixes-2' of git://git./linux/kernel/git/device-mapper/linux-dm

Pull device mapper fixes from Mike Snitzer:

 - a stable fix for DM round robin multipath path selector to disable
   preemption before using this_cpu_ptr()

 - a slight increase in DM crypt's mempool reserves to make swap ontop
   of DM crypt more performant

 - a few DM raid fixes to issues found while testing changes that were
   merged in v4.8-rc1

* tag 'dm-4.8-fixes-2' of git://git.kernel.org/pub/scm/linux/kernel/git/device-mapper/linux-dm:
  dm raid: support raid0 with missing metadata devices
  dm raid: enhance attempt_restore_of_faulty_devices() to support more devices
  dm raid: fix restoring of failed devices regression
  dm raid: fix frozen recovery regression
  dm crypt: increase mempool reserve to better support swapping
  dm round robin: do not use this_cpu_ptr() without having preemption disabled