We are trying to kick transfers on Isochronous endpoints in a more
controlled manner now. And this ended up rendering this piece of code
unnecessary.
Signed-off-by: Felipe Balbi <felipe.balbi@linux.intel.com>
instead of having one big loop, let's split it down into two smaller
handlers: one for linear buffers and one for scatterlist.
Signed-off-by: Felipe Balbi <felipe.balbi@linux.intel.com>
After all the previous changes, it's now a lot clearer how isoc
transfers should be managed. We don't need to try to End Transfers
from ep_queue since that's already done by cleanup_requests.
Signed-off-by: Felipe Balbi <felipe.balbi@linux.intel.com>
Now, this part of the code is duplicated and brings no extra value to
the driver. Let's remove it.
Signed-off-by: Felipe Balbi <felipe.balbi@linux.intel.com>
In case we get an event with status set to Missed Isoc, this means we
have missed an isochronous interval and should issue End Transfer
command and wait for the following XferNotReady.
Let's do that early, rather than late.
Signed-off-by: Felipe Balbi <felipe.balbi@linux.intel.com>
By now, it has the same semantics as DWC3_EP_TRANSFER_STARTED, but
that has a much more descriptive name.
Signed-off-by: Felipe Balbi <felipe.balbi@linux.intel.com>
We will need an up-to-date frame_number from XferInProgress too when
future patches improve our handling of Isoc endpoints.
Signed-off-by: Felipe Balbi <felipe.balbi@linux.intel.com>
The core requires the extra two bits of information for properly
scheduling Isochronous transfers. This means that we can't rely on
__dwc3_gadget_get_frame(). Let's always cache uFrame number from
XferNotReady instead.
Signed-off-by: Felipe Balbi <felipe.balbi@linux.intel.com>
Endpoint handlers need to know about endpoints, not dwc. If they
really need access to dwc (e.g. for printing error messages) we have a
reference to it tucked inside the endpoint.
This patch has no functional changes, it's simply moving things around.
Signed-off-by: Felipe Balbi <felipe.balbi@linux.intel.com>
We *KNOW* which events we enable for which endpoint types and
we *KNOW* when they'll trigger. The endpoint type checks are
pointless.
Signed-off-by: Felipe Balbi <felipe.balbi@linux.intel.com>
This patch simply renames two functions to more descriptive names so
that it's easier to understand what they're doing.
Cleanup only, no functional changes.
Signed-off-by: Felipe Balbi <felipe.balbi@linux.intel.com>
Instead of constantly calling kick transfer everything some event
shows up, let's just rely on the fact that we send Update Transfer
every time a new request is queued.
Signed-off-by: Felipe Balbi <felipe.balbi@linux.intel.com>
Now that we're making sure we don't have XferComplete events, we can
rename this function to what it actually handles:
dwc3_gadget_endpoint_transfer_in_progress()
Signed-off-by: Felipe Balbi <felipe.balbi@linux.intel.com>
XferComplete is enabled only for the default control pipe, let's make
that clear in the code.
Signed-off-by: Felipe Balbi <felipe.balbi@linux.intel.com>
We don't use XferNotReady for anything other than Default Control
Pipe, which is handled in ep0.c, and Isochronous endpoints. Let's make
that clear in the code.
Signed-off-by: Felipe Balbi <felipe.balbi@linux.intel.com>
Interrupt endpoints behave much like Bulk endpoints with the exception
that they are periodic. We can pre-issue Start Transfer exactly as we
do for Bulk endpoints.
While at that, remove one trailing blank line which is unnecessary.
Signed-off-by: Felipe Balbi <felipe.balbi@linux.intel.com>
The present code correctly fetches the req which were previously not
queued from the started_list but fails to continue queuing from the sg
where it previously stopped queuing (because of the unavailable TRB's).
This patch correct's the code to continue queuing from the correct sg
present in the sglist.
For example, consider 5 sgs in req. Because of limited TRB's among the
5 sgs only 3 got queued. This patch corrects the code to start queuing
from correct sg i.e 4th sg when the TRBs are available.
Signed-off-by: Anurag Kumar Vulisha <anuragku@xilinx.com>
Signed-off-by: Felipe Balbi <felipe.balbi@linux.intel.com>
The code logic in dwc3_prepare_one_trb() incorrectly uses the address
and length fields present in req packet for mapping TRB's instead of
using the address and length fields of scattergather lists. This patch
correct's the code to use sg->address and sg->length when scattergather
lists are present.
Signed-off-by: Anurag Kumar Vulisha <anuragku@xilinx.com>
Signed-off-by: Felipe Balbi <felipe.balbi@linux.intel.com>
extcon device is used to detect host/device connection. Since extcon
OF property is deprecated, alternative method should be added.
This method uses OF graph bindings to locate extcon.
Signed-off-by: Andrzej Hajda <a.hajda@samsung.com>
Signed-off-by: Felipe Balbi <felipe.balbi@linux.intel.com>
Historically, the clocks and resets are handled on the glue layer
side instead of the DWC3 core. For simple cases, dwc3-of-simple.c
takes care of arbitrary number of clocks and resets. The DT node
structure typically looks like as follows:
dwc3-glue {
compatible = "foo,dwc3";
clocks = ...;
resets = ...;
...
dwc3 {
compatible = "snps,dwc3";
...
};
}
By supporting the clocks and the reset in the dwc3/core.c, it will
be turned into a single node:
dwc3 {
compatible = "foo,dwc3", "snps,dwc3";
clocks = ...;
resets = ...;
...
}
This commit adds the binding of clocks and resets specific to this IP.
The number of clocks should generally be the same across SoCs, it is
just some SoCs either tie clocks together or do not provide software
control of some of the clocks.
I took the clock names from the Synopsys datasheet: "ref" (ref_clk),
"bus_early" (bus_clk_early), and "suspend" (suspend_clk).
I found only one reset line in the datasheet, hence the reset-names
property is omitted.
Those clocks are required for new platforms. Enforcing the new
binding breaks existing platforms since they specify clocks (and
resets) in their glue layer node, but nothing in the core node.
I listed such exceptional cases in the DT binding. The driver
code has been relaxed to accept no clock. This change is based
on the discussion [1].
I inserted reset_control_deassert() and clk_bulk_enable() before the
first register access, i.e. dwc3_cache_hwparams().
[1] https://patchwork.kernel.org/patch/10284265/
Signed-off-by: Masahiro Yamada <yamada.masahiro@socionext.com>
Reviewed-by: Rob Herring <robh@kernel.org>
Signed-off-by: Felipe Balbi <felipe.balbi@linux.intel.com>
Fix the fallout of the conversion to GPIO descriptors in 3df0340810.
Acked-by: Ludovic Desroches <ludovic.desroches@microchip.com>
Acked-by: Nicolas Ferre <nicolas.ferre@microchip.com>
Signed-off-by: Romain Izard <romain.izard.pro@gmail.com>
Signed-off-by: Felipe Balbi <felipe.balbi@linux.intel.com>
The include defines the private platform_data structure used with AVR
platforms. It has no user since 7c55984e19. Remove it.
Acked-by: Ludovic Desroches <ludovic.desroches@microchip.com>
Acked-by: Nicolas Ferre <nicolas.ferre@microchip.com>
Signed-off-by: Romain Izard <romain.izard.pro@gmail.com>
Signed-off-by: Felipe Balbi <felipe.balbi@linux.intel.com>
When converting to GPIO descriptors, gpiod_get_value automatically
handles the line inversion flags from the device tree.
Do not invert the line twice.
Fixes: 3df0340810 ("usb: gadget: udc: atmel: convert to use GPIO descriptors")
Acked-by: Ludovic Desroches <ludovic.desroches@microchip.com>
Acked-by: Nicolas Ferre <nicolas.ferre@microchip.com>
Signed-off-by: Romain Izard <romain.izard.pro@gmail.com>
Signed-off-by: Felipe Balbi <felipe.balbi@linux.intel.com>
A message can be forged to crash the stack; here we make sure we don't
completely break the system if this occurs
Signed-off-by: Michel Pollet <michel.pollet@bp.renesas.com>
Signed-off-by: Felipe Balbi <felipe.balbi@linux.intel.com>
It is not a good idea to directly modify the resource of a platform
device. Modify its local copy, and pass it to devm_ioremap_resource()
so that we do not need to restore it in the failure path and the remove
hook.
Signed-off-by: Masahiro Yamada <yamada.masahiro@socionext.com>
Reviewed-by: Masami Hiramatsu <mhiramat@kernel.org>
Signed-off-by: Felipe Balbi <felipe.balbi@linux.intel.com>
The USB3CV version 2.1.80 (March 26, 2018) requires all devices
( gen1, gen2, single lane, dual lane) to return the value of 0x0320
in the bcdUSB field
Signed-off-by: Chunfeng Yun <chunfeng.yun@mediatek.com>
Signed-off-by: Felipe Balbi <felipe.balbi@linux.intel.com>
Some PHY drivers (e.g. for Qualcomm QUSB2 and QMP PHYs) support
runtime PM to reduce PHY power consumption during bus_suspend.
Add changes to let core auto-suspend PHYs on host bus-suspend
using GUSB2PHYCFG register if needed for a platform. Also perform
PHYs runtime suspend/resume and let platform glue drivers e.g.
dwc3-qcom handle remote wakeup during bus suspend by waking up
devices on receiving wakeup event from PHY.
Signed-off-by: Manu Gautam <mgautam@codeaurora.org>
Signed-off-by: Felipe Balbi <felipe.balbi@linux.intel.com>
DWC3 controller on Qualcomm SOCs has a Qscratch wrapper.
Some of its uses are described below resulting in need to
have a separate glue driver instead of using dwc3-of-simple:
- It exposes register interface to override vbus-override
and lane0-pwr-present signals going to hardware. These
must be updated in peripheral mode for DWC3 if vbus lines
are not connected to hardware block. Otherwise RX termination
in SS mode or DP pull-up is not applied by device controller.
- pwr_events_irq_stat support to check if USB2 PHY is in L2 state
before glue driver proceeds with suspend.
- Support for wakeup interrupts lines that are asserted whenever
there is any wakeup event on USB3 or USB2 bus.
- Support to replace pip3 clock going to DWC3 with utmi clock
for hardware configuration where SSPHY is not used with DWC3.
Signed-off-by: Manu Gautam <mgautam@codeaurora.org>
Signed-off-by: Felipe Balbi <felipe.balbi@linux.intel.com>
Add compatible string to use this generic glue layer to support
Allwinner H6 platform's dwc3 controller.
Signed-off-by: Icenowy Zheng <icenowy@aosc.io>
Signed-off-by: Felipe Balbi <felipe.balbi@linux.intel.com>
In case there are multiple ecm instances, either for multiple
otg controllers, or multiple virtual links using libcomposite,
each instance needs to have its own host mac address string
value for iMACAddress.
Update the source array (ecm_string_defs), every time before
usb_gstrings_attach(). Without that, all links wrongly were
getting the same, last allocated, host mac address, rather
than the correct one, as requested via configfs.
Signed-off-by: Lukasz Nowak <lukasz.nowak@exablue.com>
Signed-off-by: Felipe Balbi <felipe.balbi@linux.intel.com>
Fixes: <b3e3893e1253> ("net: use core MTU range checking")
which patched only one of two functions used to setup the
USB Gadget Ethernet driver, causing a serious performance
regression in the ability to increase mtu size above 1500.
Signed-off-by: John Greb <h3x4m3r0n@gmail.com>
Signed-off-by: Felipe Balbi <felipe.balbi@linux.intel.com>
The current frame_number is read from core for both
device and host modes. Reading of the current frame
number needs to be performed ASAP due to IRQ latency's.
This is why, it is moved to common interrupt handler.
Accordingly updated dwc2_gadget_target_frame_elapsed()
function which uses stored frame_number instead of
reading frame number.
In cases when target frame value is incremented
the frame_number is required to read again.
Signed-off-by: Artur Petrosyan <arturp@synopsys.com>
Signed-off-by: Felipe Balbi <felipe.balbi@linux.intel.com>
Dumping the registers via debugfs makes USB on Raspberry Pi completely
unusable. The read of register GRXSTSP ("Receive Status Read and Pop
Register") is responsible for this behaviour, because it pops the RX FIFO.
So avoid this by omitting the relevant register.
CC: Mian Yousaf Kaukab <yousaf.kaukab@intel.com>
Fixes: 563cf017c4 ("usb: dwc2: debugfs: add support for complete register dump")
Acked-by: Minas Harutyunyan <hminas@synopsys.com>
Signed-off-by: Stefan Wahren <stefan.wahren@i2se.com>
Signed-off-by: Felipe Balbi <felipe.balbi@linux.intel.com>
Crash caused by going out of "eps_out" array range.
Iteration on "eps_out" changed to less than "num_of_eps".
Signed-off-by: Artur Petrosyan <arturp@synopsys.com>
Signed-off-by: Felipe Balbi <felipe.balbi@linux.intel.com>
Added GHWCFG4_IPG_ISOC_SUPPORTED and DCFG_IPG_ISOC_SUPPORDED
bits definitions to enable/disable IPG feature.
Added ipg_isoc_en core parameter which will indicate IPG support
enable/disable and initialize it.
Signed-off-by: Grigor Tovmasyan <tovmasya@synopsys.com>
Signed-off-by: Felipe Balbi <felipe.balbi@linux.intel.com>
In DDMA mode required to enable BNA interrupt for
both directions.
Signed-off-by: Minas Harutyunyan <hminas@synopsys.com>
Signed-off-by: Felipe Balbi <felipe.balbi@linux.intel.com>
Changed existing two descriptor-chain flow to one chain.
In two-chain implementation BNA interrupt used for switching between
two chains. BNA interrupt asserted because of returning to
beginning of the chain based on L-bit of last descriptor.
Because of that we lose packets. This issue resolved by using one
desc-chain.
Removed all staff related to two desc-chain flow from
DDMA ISOC related functions.
Removed request length checking from dwc2_gadget_fill_isoc_desc()
function. Request length checking added to dwc2_hsotg_ep_queue()
function. If request length greater than descriptor limits then
request not added to queue. Additional checking done for High
Bandwidth ISOC OUT's which not supported by driver. In
dwc2_gadget_fill_isoc_desc() function also checked desc-chain
status (full or not) to avoid of reusing not yet processed
descriptors.
In dwc2_gadget_start_isoc_ddma() function creation of desc-chain
always started from descriptor 0. Before filling descriptors, they
were initialized by HOST BUSY status.
In dwc2_gadget_complete_isoc_request_ddma() added checking for
desc-chain rollover. Also added checking completion status.
Request completed successfully if DEV_DMA_STS is DEV_DMA_STS_SUCC,
otherwise complete with actual=0. For systems with high IRQ latency
added pointer compl_desc to next descriptor to be completed by
XferCompl interrupt. This pointer replace descriptor index calculation
based on DxEPDMA register. On descriptor completion interrupt
processing all descriptors starting from compl_desc till descriptor
which Buffer Status field not equal DMA_DONE status.
Actually removed dwc2_gadget_start_next_isoc_ddma() function because
now driver use only one desc-chain and instead that function added
dwc2_gadget_handle_isoc_bna() function for handling BNA interrupts.
Handling BNA interrupt done by flushing TxFIFOs for OUT EPs,
completing request with actual=0 and resetting desc-chain number and
target frame to initial values for restarting transfers.
On handling NAK request completed with actual=0. Incremented target
frame to allow fill desc chain and start transfers.
In DDMA mode avoided of frame number incrementing, because tracking
of frame number performed in dwc2_gadget_fill_isoc_desc() function.
When core assert XferCompl along with BNA, we should ignore XferCompl
in dwc2_hsotg_epint() function.
On BNA interrupt replaced dwc2_gadget_start_next_isoc_ddma() by above
mentioned BNA handler.
In dwc2_hsotg_ep_enable() function added sanity check of bInterval
for ISOC IN in DDMA mode, because HW doesn't supported EP's with
bInterval more than 10 and check for mc for ISOC OUT transfers,
because core doesn't support high bandwidth transfers.
Signed-off-by: Minas Harutyunyan <hminas@synopsys.com>
Signed-off-by: Felipe Balbi <felipe.balbi@linux.intel.com>
In fact the driver depends on EXTCON only when it's configed as
USB_MTU3_DUAL_ROLE, so make USB_MTU3_DUAL_ROLE depend on EXTCON but
not USB_MTU3.
Signed-off-by: Chunfeng Yun <chunfeng.yun@mediatek.com>
Signed-off-by: Felipe Balbi <felipe.balbi@linux.intel.com>
There is an error dialog popped up in PC when test TEST_J/K
by EHSETT tool, due to not waiting for the completion of
control transfer. Here fix it by entering test mode after
Status Stage finish.
Signed-off-by: Chunfeng Yun <chunfeng.yun@mediatek.com>
Signed-off-by: Felipe Balbi <felipe.balbi@linux.intel.com>
When boot on the platform with the USB cable connected to Win7,
the Win7 will pop up an error dialog: "USB Device not recognized",
but finally the Win7 can enumerate it successfully.
The root cause is as the following:
When the xHCI driver set PORT_POWER of the OTG port, and if both
IDPIN and VBUS_VALID are high at the same time, the MTU3 controller
will set SESSION and pull up DP, so the Win7 can detect existence
of USB device, but if the mtu3 driver can't switch to device mode
during the debounce time, the Win7 can not enumerate it.
Here to fix it by removing the 1s delayed EXTCON register to speed up
mode switch.
Signed-off-by: Chunfeng Yun <chunfeng.yun@mediatek.com>
Signed-off-by: Felipe Balbi <felipe.balbi@linux.intel.com>
The usb_add_gadget_udc() will set the gadget state as
USB_STATE_NOTATTACHED, so we needn't set it again.
Signed-off-by: Chunfeng Yun <chunfeng.yun@mediatek.com>
Signed-off-by: Felipe Balbi <felipe.balbi@linux.intel.com>
The variable of 'count' is declared as u8, this will cause an issue
due to value truncated when works in SS or SSP mode and data length
is greater than 255, so change it as u32.
Signed-off-by: Chunfeng Yun <chunfeng.yun@mediatek.com>
Signed-off-by: Felipe Balbi <felipe.balbi@linux.intel.com>
We should get drvdata from struct device directly. Going via
platform_device is an unneeded step back and forth.
Signed-off-by: Wolfram Sang <wsa+renesas@sang-engineering.com>
Signed-off-by: Felipe Balbi <felipe.balbi@linux.intel.com>
Felipe Balbi