In the DWC3 databook, for a device initiated disconnect or bus reset, the
driver is required to send dependxfer commands for any pending transfers.
In addition, before the controller can move to the halted state, the SW
needs to acknowledge any pending events. If the controller is not halted
properly, there is a chance the controller will continue accessing stale or
freed TRBs and buffers.
Signed-off-by: Wesley Cheng <wcheng@codeaurora.org>
Reviewed-by: Thinh Nguyen <thinhn@synopsys.com>
Signed-off-by: Felipe Balbi <balbi@kernel.org>
At Android ADB and MTP use case, it uses f_fs which supports scatter list,
it means one request may need several TRBs for it. Besides, TRB consumes
very fast compared to TRB has prepared for above use case, there are at
most 120 pending requests, the date size is 16KB for each request, so four
TRBs (4KB per TRB) per sg entry at worst case. so we need to enlarge the
TRB ring length to avoid "no free TRB error". Since each TRB only consumes
12 bytes (3 * 32 bits), we enlarge the TRB length to 600, it leaves some
buffers for potential "no free TRB error", and only increases a little
memory cost.
Signed-off-by: Peter Chen <peter.chen@nxp.com>
Signed-off-by: Felipe Balbi <balbi@kernel.org>
The scatter buffer list support earlier than DEV_VER_V2 is not
good enough, software can't know well about short transfer for it.
Cc: Pawel Laszczak <pawell@cadence.com>
Signed-off-by: Peter Chen <peter.chen@nxp.com>
Signed-off-by: Felipe Balbi <balbi@kernel.org>
Add sg case for workaround 2, the workaround 2 is described at the
beginning of this file.
Signed-off-by: Peter Chen <peter.chen@nxp.com>
Signed-off-by: Felipe Balbi <balbi@kernel.org>
- Judge each TRB has been handled at cdns3_trb_handled, since
the DMA pointer may be at the middle of the TD, we can't consider
this TD has finished at that time.
- Calculate req->actual according to finished TRBs.
- Handle short transfer for sg list use case correctly. When the
short transfer occurs, we check OUT_SMM at TRB to see if it is
the last TRB.
Signed-off-by: Peter Chen <peter.chen@nxp.com>
Signed-off-by: Felipe Balbi <balbi@kernel.org>
For sg buffer list use case, we need to add ISP for each TRB, and
add CHAIN bit for each TRB except for the last TRB.
Signed-off-by: Peter Chen <peter.chen@nxp.com>
Signed-off-by: Felipe Balbi <balbi@kernel.org>
It only dumps the first TRB per request, it is not useful if only dump
the first TRB when there are several TRBs per request. We improve it by
dumpping all TRBs per request in this commit.
Signed-off-by: Peter Chen <peter.chen@nxp.com>
Signed-off-by: Felipe Balbi <balbi@kernel.org>
It needs to use request->num_mapped_sgs to indicate mapped sg number,
the request->num_sgs is the sg number before the mapping. These two
entries have different values for the platforms which iommu or
swiotlb is used. Besides, it needs to use correct sg APIs for
mapped sg list for TRB assignment.
Signed-off-by: Peter Chen <peter.chen@nxp.com>
Signed-off-by: Felipe Balbi <balbi@kernel.org>
SPLIT_BOUNDARY_DISABLE should be set for DesignWare USB3 DRD Core
of Hisilicon Kirin Soc when dwc3 core act as host.
[mchehab: dropped a dev_dbg() as only traces are now allowwed on this driver]
Signed-off-by: Yu Chen <chenyu56@huawei.com>
Signed-off-by: Mauro Carvalho Chehab <mchehab+huawei@kernel.org>
Signed-off-by: Felipe Balbi <balbi@kernel.org>
Delete unused initialized value of 'ret', because it will
be assigned by the function clk_prepare_enable().
Signed-off-by: Zhang Shengju <zhangshengju@cmss.chinamobile.com>
Signed-off-by: Tang Bin <tangbin@cmss.chinamobile.com>
Signed-off-by: Felipe Balbi <balbi@kernel.org>
If an error occurred before calling the 'v4l2_device_register' func,
and then goto error, but no need to call 'v4l2_device_unregister'
func.
Signed-off-by: Zqiang <qiang.zhang@windriver.com>
Signed-off-by: Felipe Balbi <balbi@kernel.org>
The functions dwc3_prepare_one_trb_sg and dwc3_prepare_one_trb_linear
are not necessarily preparing "one" TRB, it can prepare multiple TRBs.
Rename these functions as follow:
dwc3_prepare_one_trb_sg -> dwc3_prepare_trbs_sg
dwc3_prepare_one_trb_linear -> dwc3_prepare_trbs_linear
Signed-off-by: Thinh Nguyen <Thinh.Nguyen@synopsys.com>
Signed-off-by: Felipe Balbi <balbi@kernel.org>
There are a lot of common codes for preparing SG and linear TRBs.
Refactor them for easier read.
Signed-off-by: Thinh Nguyen <Thinh.Nguyen@synopsys.com>
Signed-off-by: Felipe Balbi <balbi@kernel.org>
If we run out of TRBs because we need extra TRBs, make sure to set the
IOC bit for the previously prepared TRB to get completion notification
to free up TRBs to resume later.
Signed-off-by: Thinh Nguyen <Thinh.Nguyen@synopsys.com>
Signed-off-by: Felipe Balbi <balbi@kernel.org>
By returning the number of TRBs prepared, we know whether to execute
__dwc3_gadget_kick_transfer(). This allows us to check if we ran out of
TRBs when extra TRBs are needed for OUT transfers. It also allows us to
properly handle usb_gadget_map_request_by_dev() error.
Fixes: c6267a5163 ("usb: dwc3: gadget: align transfers to wMaxPacketSize")
Signed-off-by: Thinh Nguyen <Thinh.Nguyen@synopsys.com>
Signed-off-by: Felipe Balbi <balbi@kernel.org>
In preparation for fixing the check for number of remaining TRBs,
revise dwc3_prepare_one_trb_linear() and dwc3_prepare_one_trb_sg() to
return the number of prepared TRBs.
Signed-off-by: Thinh Nguyen <Thinh.Nguyen@synopsys.com>
Signed-off-by: Felipe Balbi <balbi@kernel.org>
The current ZLP handling for ep0 requests is only for control IN
requests. For OUT direction, DWC3 needs to check and setup for MPS
alignment.
Usually, control OUT requests can indicate its transfer size via the
wLength field of the control message. So usb_request->zero is usually
not needed for OUT direction. To handle ZLP OUT for control endpoint,
make sure the TRB is MPS size.
Cc: stable@vger.kernel.org
Fixes: c7fcdeb262 ("usb: dwc3: ep0: simplify EP0 state machine")
Fixes: d6e5a549cc ("usb: dwc3: simplify ZLP handling")
Signed-off-by: Thinh Nguyen <Thinh.Nguyen@synopsys.com>
Signed-off-by: Felipe Balbi <balbi@kernel.org>
For OUT requests that requires extra TRBs for ZLP. We don't need to
prepare the 0-length TRB and simply prepare the MPS size TRB. This
reduces 1 TRB needed to prepare for ZLP.
Signed-off-by: Thinh Nguyen <Thinh.Nguyen@synopsys.com>
Signed-off-by: Felipe Balbi <balbi@kernel.org>
When the driver prepares the extra TRB, it uses bounce buffer. If we
just add a new parameter to dwc3_prepare_one_trb() to indicate this,
then we can refactor and simplify the driver quite a bit.
dwc3_prepare_one_trb() also checks if a request had been moved to the
started list. This is a prerequisite to subsequence patches improving
the handling of extra TRBs.
Signed-off-by: Thinh Nguyen <Thinh.Nguyen@synopsys.com>
Signed-off-by: Felipe Balbi <balbi@kernel.org>
An SG request may be partially completed (due to no available TRBs).
Don't reclaim extra TRBs and clear the needs_extra_trb flag until the
request is fully completed. Otherwise, the driver will reclaim the wrong
TRB.
Cc: stable@vger.kernel.org
Fixes: 1f512119a0 ("usb: dwc3: gadget: add remaining sg entries to ring")
Signed-off-by: Thinh Nguyen <Thinh.Nguyen@synopsys.com>
Signed-off-by: Felipe Balbi <balbi@kernel.org>
When preparing for SG, not all the entries are prepared at once. When
resume, don't use the remaining request length to calculate for MPS
alignment. Use the entire request->length to do that.
Cc: stable@vger.kernel.org
Fixes: 5d187c0454 ("usb: dwc3: gadget: Don't setup more than requested")
Signed-off-by: Thinh Nguyen <Thinh.Nguyen@synopsys.com>
Signed-off-by: Felipe Balbi <balbi@kernel.org>
Probe deferral is an expected condition and can happen multiple times
during boot. Make sure not to output an error message in that case
because they are not useful.
Signed-off-by: Thierry Reding <treding@nvidia.com>
Signed-off-by: Felipe Balbi <balbi@kernel.org>
Align parameters on subsequent lines with the parameters on the first
line for consistency.
Signed-off-by: Thierry Reding <treding@nvidia.com>
Signed-off-by: Felipe Balbi <balbi@kernel.org>
There is a spelling mistake in a literal string. Fix it.
Signed-off-by: Colin Ian King <colin.king@canonical.com>
Signed-off-by: Felipe Balbi <balbi@kernel.org>
Make sure to use consistent spelling and formatting in error messages.
Signed-off-by: Thierry Reding <treding@nvidia.com>
Signed-off-by: Felipe Balbi <balbi@kernel.org>
There is no need to use GFP_ATOMIC here. It is a probe function, no
spinlock is taken.
Reviewed-by: JC Kuo <jckuo@nvidia.com>
Acked-by: Thierry Reding <treding@nvidia.com>
Signed-off-by: Christophe JAILLET <christophe.jaillet@wanadoo.fr>
Signed-off-by: Felipe Balbi <balbi@kernel.org>
Current UDC core connects gadget during the loading gadget flow
(udc_bind_to_driver->usb_udc_connect_control), but for
platforms which do not connect gadget if the VBUS is not there,
they call usb_gadget_disconnect, but the gadget is not connected
at this time, notify disconnecton for the gadget driver is meaningless
at this situation.
Signed-off-by: Peter Chen <peter.chen@nxp.com>
Signed-off-by: Felipe Balbi <balbi@kernel.org>
Refactor END_TRANSFER command completion handling and move it outside of
the switch statement to its own function. This makes it cleaner and
consistent with other event handler functions. No functional change
here.
Signed-off-by: Thinh Nguyen <thinhn@synopsys.com>
Signed-off-by: Felipe Balbi <balbi@kernel.org>
Remove unused 'udc' variable to fix compile warnings:
drivers/usb/gadget/udc/s3c2410_udc.c: In function 's3c2410_udc_dequeue':
drivers/usb/gadget/udc/s3c2410_udc.c:1268:22: warning: variable 'udc' set but not used [-Wunused-but-set-variable]
Reviewed-by: Nathan Chancellor <natechancellor@gmail.com>
Signed-off-by: Krzysztof Kozlowski <krzk@kernel.org>
Signed-off-by: Felipe Balbi <balbi@kernel.org>
Add the phy cleanup if dwc3 mode init fail, which is the missing part of
de-init for dwc3 core init.
Fixes: c499ff71ff ("usb: dwc3: core: re-factor init and exit paths")
Cc: <stable@vger.kernel.org>
Signed-off-by: Li Jun <jun.li@nxp.com>
Signed-off-by: Felipe Balbi <balbi@kernel.org>
We have already allocated gadget structure dynamically at UDC (dwc3)
driver, so commit fac323471d ("usb: udc: allow adding and removing
the same gadget device")could be reverted.
Cc: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Cc: Alan Stern <stern@rowland.harvard.edu>
Reviewed-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Reviewed-by: Alan Stern <stern@rowland.harvard.edu>
Signed-off-by: Peter Chen <peter.chen@nxp.com>
Signed-off-by: Felipe Balbi <balbi@kernel.org>
The current code uses commit fac323471d ("usb: udc: allow adding
and removing the same gadget device") as the workaround to let
the gadget device is re-used, but it is not allowed from driver
core point. In this commit, we allocate gadget structure dynamically,
and free it at its release function. Since the gadget device's
driver_data has already occupied by usb_composite_dev structure, we have
to use gadget device's platform data to store dwc3 structure.
Cc: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Cc: Alan Stern <stern@rowland.harvard.edu>
Reviewed-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Signed-off-by: Peter Chen <peter.chen@nxp.com>
Signed-off-by: Felipe Balbi <balbi@kernel.org>
If cdns3_gadget_start is failed, it never frees cdns3_device structure.
Meanwhile, there is no release function for gadget device, it causes
there is no sync with driver core.
To fix this, we add release function for gadget device, and free
cdns3_device structure at there. Meanwhile, With the new UDC core
APIs, we could work with driver core better to handle memory leak
issue.
Reviewed-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Signed-off-by: Peter Chen <peter.chen@nxp.com>
Signed-off-by: Felipe Balbi <balbi@kernel.org>
Like net2280 (on which it was based), the net2272 UDC driver has a
problem with leaking memory along some of its failure pathways. It
also has another problem, not previously noted, in that some of the
failure pathways will call usb_del_gadget_udc() without first calling
usb_add_gadget_udc_release(). And it leaks memory by calling kfree()
when it should call put_device().
Until now it has been impossible to handle the memory leaks, because of
lack of support in the UDC core for separately initializing and adding
gadgets, or for separately deleting and freeing gadgets. An earlier
patch in this series adds the necessary support, making it possible to
fix the outstanding problems properly.
This patch adds an "added" flag to the net2272 structure to indicate
whether or not the gadget has been registered (and thus whether or not
to call usb_del_gadget()), and it fixes the deallocation issues by
calling usb_put_gadget() at the appropriate places.
A similar memory leak issue, apparently never before recognized, stems
from the fact that the driver never initializes the drvdata field in
the gadget's embedded struct device! Evidently this wasn't noticed
because the pointer is only ever used as an argument to kfree(), which
doesn't mind getting called with a NULL pointer. In fact, the drvdata
for gadget device will be written by usb_composite_dev structure if
any gadget class is loaded, so it needs to use usb_gadget structure
to get net2280 private data.
CC: Anton Vasilyev <vasilyev@ispras.ru>
CC: Evgeny Novikov <novikov@ispras.ru>
CC: Benjamin Herrenschmidt <benh@kernel.crashing.org>
Reviewed-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Signed-off-by: Alan Stern <stern@rowland.harvard.edu>
Signed-off-by: Peter Chen <peter.chen@nxp.com>
Signed-off-by: Felipe Balbi <balbi@kernel.org>
As Anton and Evgeny have noted, the net2280 UDC driver has a problem
with leaking memory along some of its failure pathways. It also has
another problem, not previously noted, in that some of the failure
pathways will call usb_del_gadget_udc() without first calling
usb_add_gadget_udc_release(). And it leaks memory by calling kfree()
when it should call put_device().
Previous attempts to fix the problems have failed because of lack of
support in the UDC core for separately initializing and adding
gadgets, or for separately deleting and freeing gadgets. The previous
patch in this series adds the necessary support, making it possible to
fix the outstanding problems properly.
This patch adds an "added" flag to the net2280 structure to indicate
whether or not the gadget has been registered (and thus whether or not
to call usb_del_gadget()), and it fixes the deallocation issues by
calling usb_put_gadget() at the appropriate point.
A similar memory leak issue, apparently never before recognized, stems
from the fact that the driver never initializes the drvdata field in
the gadget's embedded struct device! Evidently this wasn't noticed
because the pointer is only ever used as an argument to kfree(), which
doesn't mind getting called with a NULL pointer. In fact, the drvdata
for gadget device will be written by usb_composite_dev structure if
any gadget class is loaded, so it needs to use usb_gadget structure
to get net2280 private data.
CC: Benjamin Herrenschmidt <benh@kernel.crashing.org>
Reported-by: Anton Vasilyev <vasilyev@ispras.ru>
Reported-by: Evgeny Novikov <novikov@ispras.ru>
Reviewed-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Signed-off-by: Alan Stern <stern@rowland.harvard.edu>
Signed-off-by: Peter Chen <peter.chen@nxp.com>
Signed-off-by: Felipe Balbi <balbi@kernel.org>
The routines used by the UDC core to interface with the kernel's
device model, namely usb_add_gadget_udc(),
usb_add_gadget_udc_release(), and usb_del_gadget_udc(), provide access
to only a subset of the device model's full API. They include
functionality equivalent to device_register() and device_unregister()
for gadgets, but they omit device_initialize(), device_add(),
device_del(), get_device(), and put_device().
This patch expands the UDC API by adding usb_initialize_gadget(),
usb_add_gadget(), usb_del_gadget(), usb_get_gadget(), and
usb_put_gadget() to fill in the gap. It rewrites the existing
routines to call the new ones.
CC: Anton Vasilyev <vasilyev@ispras.ru>
CC: Evgeny Novikov <novikov@ispras.ru>
CC: Benjamin Herrenschmidt <benh@kernel.crashing.org>
Reviewed-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Signed-off-by: Alan Stern <stern@rowland.harvard.edu>
Signed-off-by: Peter Chen <peter.chen@nxp.com>
Signed-off-by: Felipe Balbi <balbi@kernel.org>
It is found by sparse.
Reported-by: kbuild test robot <lkp@intel.com>
Signed-off-by: Peter Chen <peter.chen@nxp.com>
Signed-off-by: Felipe Balbi <balbi@kernel.org>
udc_controller->irq is "unsigned int" always >= 0, but platform_get_irq may
return little than zero. So "dc_controller->irq < 0" condition is never
accessible.
Acked-by: Li Yang <leoyang.li@nxp.com>
Signed-off-by: Ye Bin <yebin10@huawei.com>
Signed-off-by: Felipe Balbi <balbi@kernel.org>
The u_ether driver has a qmult setting that multiplies the
transmit queue length (which by default is 2).
The intent is that it should be enabled at high/super speed, but
because the code does not explicitly check for USB_SUPER_PLUS,
it is disabled at that speed.
Fix this by ensuring that the queue multiplier is enabled for any
wired link at high speed or above. Using >= for USB_SPEED_*
constants seems correct because it is what the gadget_is_xxxspeed
functions do.
The queue multiplier substantially helps performance at higher
speeds. On a direct SuperSpeed Plus link to a Linux laptop,
iperf3 single TCP stream:
Before (qmult=1): 1.3 Gbps
After (qmult=5): 3.2 Gbps
Fixes: 04617db7aa ("usb: gadget: add SS descriptors to Ethernet gadget")
Reviewed-by: Maciej Żenczykowski <maze@google.com>
Signed-off-by: Lorenzo Colitti <lorenzo@google.com>
Signed-off-by: Felipe Balbi <balbi@kernel.org>
The commit aba3a8d01d ("usb: gadget: u_serial: add suspend resume
callbacks") set/cleared the suspended flag in USB bus suspend/resume
only. But, when a USB cable is disconnected in the suspend, since some
controllers will not detect USB bus resume, the suspended flag is not
cleared. After that, user cannot send any data. To fix the issue,
clears the suspended flag in the gserial_disconnect().
Fixes: aba3a8d01d ("usb: gadget: u_serial: add suspend resume callbacks")
Signed-off-by: Yoshihiro Shimoda <yoshihiro.shimoda.uh@renesas.com>
Tested-by: Linh Phung <linh.phung.jy@renesas.com>
Tested-by: Tam Nguyen <tam.nguyen.xa@renesas.com>
Signed-off-by: Felipe Balbi <balbi@kernel.org>
Currently, enabling f_ncm at SuperSpeed Plus speeds results in an
oops in config_ep_by_speed because ncm_set_alt passes in NULL
ssp_descriptors. Fix this by re-using the SuperSpeed descriptors.
This is safe because usb_assign_descriptors calls
usb_copy_descriptors.
Tested: enabled f_ncm on a dwc3 gadget and 10Gbps link, ran iperf
Reviewed-by: Maciej Żenczykowski <maze@google.com>
Signed-off-by: Lorenzo Colitti <lorenzo@google.com>
Signed-off-by: Felipe Balbi <balbi@kernel.org>
This improves performance on fast connections. When directly
connecting to a Linux laptop running 5.6, single-stream iperf3
goes from ~1.7Gbps to ~2.3Gbps out, and from ~620Mbps to ~720Mbps
in.
Reviewed-by: Maciej Żenczykowski <maze@google.com>
Signed-off-by: Lorenzo Colitti <lorenzo@google.com>
Signed-off-by: Felipe Balbi <balbi@kernel.org>
Currently, SuperSpeed NCM gadgets report a speed of 851 Mbps
in USB_CDC_NOTIFY_SPEED_CHANGE. But the calculation appears to
assume 16 packets per microframe, and USB 3 and above no longer
use microframes.
Maximum speed is actually much higher. On a direct connection,
theoretical throughput is at most 3.86 Gbps for gen1x1 and
9.36 Gbps for gen2x1, and I have seen gadget->host iperf
throughput of >2 Gbps for gen1x1 and >4 Gbps for gen2x1.
Unfortunately the ConnectionSpeedChange defined in the CDC spec
only uses 32-bit values, so we can't report accurate numbers for
10Gbps and above. So, report 3.75Gbps for SuperSpeed (which is
roughly maximum theoretical performance) and 4.25Gbps for
SuperSpeed Plus (which is close to the maximum that we can report
in a 32-bit unsigned integer).
This results in:
[50879.191272] cdc_ncm 2-2:1.0 enx228b127e050c: renamed from usb0
[50879.234778] cdc_ncm 2-2:1.0 enx228b127e050c: 3750 mbit/s downlink 3750 mbit/s uplink
on SuperSpeed and:
[50798.434527] cdc_ncm 8-2:1.0 enx228b127e050c: renamed from usb0
[50798.524278] cdc_ncm 8-2:1.0 enx228b127e050c: 4250 mbit/s downlink 4250 mbit/s uplink
on SuperSpeed Plus.
Fixes: 1650113888 ("usb: gadget: f_ncm: add SuperSpeed descriptors for CDC NCM")
Reviewed-by: Maciej Żenczykowski <maze@google.com>
Signed-off-by: Lorenzo Colitti <lorenzo@google.com>
Signed-off-by: Felipe Balbi <balbi@kernel.org>
After commit f4cfe5ce60 ("usb: cdns3: gadget: improve the
set_configuration handling"), the software will inform the
hardware the request has finished at cdns3_ep0_complete_setup.
The configuration set bit is only set after request has finished,
so it needs to move waiting operation after that. Meanwhile,
if it is timeout, it will show warning message and return error.
Signed-off-by: Peter Chen <peter.chen@nxp.com>
Signed-off-by: Felipe Balbi <balbi@kernel.org>
Some PHYs may need to enter related mode, and do some settings.
Signed-off-by: Peter Chen <peter.chen@nxp.com>
Signed-off-by: Felipe Balbi <balbi@kernel.org>
It is meaningless to handle any interrupts after disconnecting
with host
Signed-off-by: Peter Chen <peter.chen@nxp.com>
Signed-off-by: Felipe Balbi <balbi@kernel.org>
