When the TRB ring is full, the dwc3 driver must make sure that there's
at least 1 TRB with Interrupt On Completion (IOC) set to notify of
available TRBs.
The current logic just sets the TRB's IOC whenever we run out of TRBs,
but it doesn't consider that there may be other TRBs with IOC/LST set
already. This creates more events and unnecessary delay from interrupt
handling. Only forcefully set IOC when we run out of TRBs and none of
the TRBs in the TRB ring has had IOC set.
Signed-off-by: Thinh Nguyen <Thinh.Nguyen@synopsys.com>
Link: https://lore.kernel.org/r/72a1fa448eb1201b152e65be7902a5d1c75b9f3a.1667867687.git.Thinh.Nguyen@synopsys.com
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
The gadget driver may have a certain expectation of how the request
completion flow should be from to its configuration. Make sure the
controller driver respect that. That is, don't set IMI (Interrupt on
Missed Isoc) when usb_request->no_interrupt is set. Also, the driver
should only set IMI to the last TRB of a chain.
Fixes: 72246da40f ("usb: Introduce DesignWare USB3 DRD Driver")
Cc: stable@vger.kernel.org
Signed-off-by: Thinh Nguyen <Thinh.Nguyen@synopsys.com>
Reviewed-by: Jeff Vanhoof <jdv1029@gmail.com>
Tested-by: Jeff Vanhoof <jdv1029@gmail.com>
Link: https://lore.kernel.org/r/ced336c84434571340c07994e3667a0ee284fefe.1666735451.git.Thinh.Nguyen@synopsys.com
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
When servicing a transfer completion event, the dwc3 driver will reclaim
TRBs of started requests up to the request associated with the interrupt
event. Currently we don't check for interrupt due to missed isoc, and
the driver may attempt to reclaim TRBs beyond the associated event. This
causes invalid memory access when the hardware still owns the TRB. If
there's a missed isoc TRB with IMI (interrupt on missed isoc), make sure
to stop servicing further.
Note that only the last TRB of chained TRBs has its status updated with
missed isoc.
Fixes: 72246da40f ("usb: Introduce DesignWare USB3 DRD Driver")
Cc: stable@vger.kernel.org
Reported-by: Jeff Vanhoof <jdv1029@gmail.com>
Reported-by: Dan Vacura <w36195@motorola.com>
Signed-off-by: Thinh Nguyen <Thinh.Nguyen@synopsys.com>
Reviewed-by: Jeff Vanhoof <jdv1029@gmail.com>
Tested-by: Jeff Vanhoof <jdv1029@gmail.com>
Link: https://lore.kernel.org/r/b29acbeab531b666095dfdafd8cb5c7654fbb3e1.1666735451.git.Thinh.Nguyen@synopsys.com
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
The gadget driver may wait on the request completion when it sets the
USB_GADGET_DELAYED_STATUS. Make sure that the End Transfer command can
go through if the dwc->delayed_status is set so that the request can
complete. When the delayed_status is set, the Setup packet is already
processed, and the next phase should be either Data or Status. It's
unlikely that the host would cancel the control transfer and send a new
Setup packet during End Transfer command. But if that's the case, we can
try again when ep0state returns to EP0_SETUP_PHASE.
Fixes: e1ee843488 ("usb: dwc3: gadget: Force sending delayed status during soft disconnect")
Cc: stable@vger.kernel.org
Signed-off-by: Thinh Nguyen <Thinh.Nguyen@synopsys.com>
Link: https://lore.kernel.org/r/3f9f59e5d74efcbaee444cf4b30ef639cc7b124e.1666146954.git.Thinh.Nguyen@synopsys.com
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
DWC3_EP_DELAYED_STOP is utilized to defer issuing the end transfer command
until the subsequent SETUP stage, in order to avoid end transfer timeouts.
During cable disconnect scenarios, __dwc3_gadget_ep_disable() is
responsible for ensuring endpoints have no active transfers pending. Since
dwc3_remove_request() can now exit early if the EP delayed stop is set,
avoid clearing all DEP flags, otherwise the transition back into the SETUP
stage won't issue an endxfer command.
Fixes: 2b2da6574e ("usb: dwc3: Avoid unmapping USB requests if endxfer is not complete")
Reviewed-by: Thinh Nguyen <Thinh.Nguyen@synopsys.com>
Signed-off-by: Wesley Cheng <quic_wcheng@quicinc.com>
Link: https://lore.kernel.org/r/20220919231213.21364-1-quic_wcheng@quicinc.com
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
During a cable disconnect sequence, if ep0state is not in the SETUP phase,
then nothing will trigger any pending end transfer commands. Force
stopping of any pending SETUP transaction, and move back to the SETUP
phase.
Reviewed-by: Thinh Nguyen <Thinh.Nguyen@synopsys.com>
Signed-off-by: Wesley Cheng <quic_wcheng@quicinc.com>
Link: https://lore.kernel.org/r/20220901193625.8727-6-quic_wcheng@quicinc.com
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
For endxfer commands that do not require an endpoint complete interrupt,
avoid having to wait for the command active bit to clear. This allows for
EP0 events to continue to be handled, which allows for the controller to
complete it. Otherwise, it is known that the endxfer command will fail if
there is a pending SETUP token that needs to be read.
Suggested-by: Thinh Nguyen <Thinh.Nguyen@synopsys.com>
Reviewed-by: Thinh Nguyen <Thinh.Nguyen@synopsys.com>
Signed-off-by: Wesley Cheng <quic_wcheng@quicinc.com>
Link: https://lore.kernel.org/r/20220901193625.8727-5-quic_wcheng@quicinc.com
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Since EP0 transactions need to be completed before the controller halt
sequence is finished, this may take some time depending on the host and the
enabled functions. Increase the controller halt timeout, so that we give
the controller sufficient time to handle EP0 transfers.
Signed-off-by: Wesley Cheng <quic_wcheng@quicinc.com>
Link: https://lore.kernel.org/r/20220901193625.8727-4-quic_wcheng@quicinc.com
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Remove the need for making dwc3_gadget_suspend() and dwc3_gadget_resume()
to be called in a spinlock, as dwc3_gadget_run_stop() could potentially
take some time to complete.
Signed-off-by: Wesley Cheng <quic_wcheng@quicinc.com>
Link: https://lore.kernel.org/r/20220901193625.8727-3-quic_wcheng@quicinc.com
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
If DWC3_EP_DELAYED_STOP is set during stop active transfers, then do not
continue attempting to unmap request buffers during dwc3_remove_requests().
This can lead to SMMU faults, as the controller has not stopped the
processing of the TRB. Defer this sequence to the EP0 out start, which
ensures that there are no pending SETUP transactions before issuing the
endxfer.
Reviewed-by: Thinh Nguyen <Thinh.Nguyen@synopsys.com>
Signed-off-by: Wesley Cheng <quic_wcheng@quicinc.com>
Link: https://lore.kernel.org/r/20220901193625.8727-2-quic_wcheng@quicinc.com
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
We need the USB fixes in here and this resolves the merge issue in:
drivers/usb/dwc3/gadget.c
Reported-by: Stephen Rothwell <sfr@canb.auug.org.au>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
During soft disconnect, EP0 events are expected to be handled in order to
allow the controller to successfully move into the halted state. Since
__dwc3_gadget_stop() is executed before polling, EP0 has been disabled, and
events are being blocked. Allow xfercomplete events to be handled, so that
cached SETUP packets can be read out from the internal controller memory.
Without doing so, it will lead to endxfer timeouts, which results to
controller halt failures.
Reviewed-by: Thinh Nguyen <Thinh.Nguyen@synopsys.com>
Signed-off-by: Wesley Cheng <quic_wcheng@quicinc.com>
Link: https://lore.kernel.org/r/20220817182359.13550-5-quic_wcheng@quicinc.com
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Ensure that there are no pending events being handled in between soft
connect/disconnect transitions. As we are keeping interrupts enabled,
and EP0 events are still being serviced, this avoids any stale events from
being serviced.
Reviewed-by: Thinh Nguyen <Thinh.Nguyen@synopsys.com>
Signed-off-by: Wesley Cheng <quic_wcheng@quicinc.com>
Link: https://lore.kernel.org/r/20220817182359.13550-4-quic_wcheng@quicinc.com
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
If any function drivers request for a delayed status phase, this leads to a
SETUP transfer timeout error, since the function may take longer to process
the DATA stage. This eventually results in end transfer timeouts, as there
is a pending SETUP transaction.
In addition, allow the DWC3_EP_DELAY_STOP to be set for if there is a
delayed status requested. Ocasionally, a host may abort the current SETUP
transaction, by issuing a subsequent SETUP token. In those situations, it
would result in an endxfer timeout as well.
Reviewed-by: Thinh Nguyen <Thinh.Nguyen@synopsys.com>
Signed-off-by: Wesley Cheng <quic_wcheng@quicinc.com>
Link: https://lore.kernel.org/r/20220817182359.13550-3-quic_wcheng@quicinc.com
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
There are some operations that need to be ignored if there is a soft
disconnect in progress. This is to avoid having a pending EP0 transfer in
progress while attempting to stop active transfers and halting the
controller.
There were several instances seen where a soft disconnect was able to occur
during early link negotiation, i.e. bus reset/conndone, which leads to the
conndone handler re-configuring EPs while attempting to halt the
controller, as DEP flags are cleared as part of the soft disconnect path.
ep0out: cmd 'Start New Configuration'
ep0out: cmd 'Set Endpoint Transfer Resource'
ep0in: cmd 'Set Endpoint Transfer Resource'
ep1out: cmd 'Set Endpoint Transfer Resource'
...
event (00030601): Suspend [U3]
event (00000101): Reset [U0]
ep0out: req ffffff87e5c9e100 length 0/0 zsI ==> 0
event (00000201): Connection Done [U0]
ep0out: cmd 'Start New Configuration'
ep0out: cmd 'Set Endpoint Transfer Resource'
In addition, if a soft disconnect occurs, EP0 events are still allowed to
process, however, it will stall/restart during the SETUP phase. The
host is still able to query for the DATA phase, leading to a
xfernotready(DATA) event. Since none of the SETUP transfer parameters are
populated, the xfernotready is treated as a "wrong direction" error,
leading to a duplicate stall/restart routine.
Add the proper softconnect/connected checks in sequences that are
potentially involved during soft disconnect processing.
Reviewed-by: Thinh Nguyen <Thinh.Nguyen@synopsys.com>
Signed-off-by: Wesley Cheng <quic_wcheng@quicinc.com>
Link: https://lore.kernel.org/r/20220817182359.13550-2-quic_wcheng@quicinc.com
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
The functions stop_active_transfers and ep_disable are both calling
remove_requests. This functions in both cases will giveback the requests
with status ESHUTDOWN, which also represents an physical disconnection.
For ep_disable this is not true. This patch adds the status parameter to
remove_requests and sets the status to ECONNRESET on ep_disable.
Signed-off-by: Michael Grzeschik <m.grzeschik@pengutronix.de>
Link: https://lore.kernel.org/r/20220720213523.1055897-1-m.grzeschik@pengutronix.de
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Relocate the pullups_connected check until after it is ensured that there
are no runtime PM transitions. If another context triggered the DWC3
core's runtime resume, it may have already enabled the Run/Stop. Do not
re-run the entire pullup sequence again, as it may issue a core soft
reset while Run/Stop is already set.
This patch depends on
commit 69e131d1ac ("usb: dwc3: gadget: Prevent repeat pullup()")
Fixes: 77adb8bdf4 ("usb: dwc3: gadget: Allow runtime suspend if UDC unbinded")
Cc: stable <stable@kernel.org>
Signed-off-by: Wesley Cheng <quic_wcheng@quicinc.com>
Link: https://lore.kernel.org/r/20220728020647.9377-1-quic_wcheng@quicinc.com
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
For High-Speed Transfers the prepare_one_trb function is calculating the
multiplier setting for the trb based on the length parameter of the trb
currently prepared. This assumption is wrong. For trbs with a sg list,
the length of the actual request has to be taken instead.
Fixes: 40d829fb2e ("usb: dwc3: gadget: Correct ISOC DATA PIDs for short packets")
Cc: stable <stable@kernel.org>
Signed-off-by: Michael Grzeschik <m.grzeschik@pengutronix.de>
Link: https://lore.kernel.org/r/20220704141812.1532306-3-m.grzeschik@pengutronix.de
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
The function __dwc3_prepare_one_trb has many parameters. Since it is
only used in dwc3_prepare_one_trb there is no point in keeping the
function. We merge both functions and get rid of the big list of
parameters.
Fixes: 40d829fb2e ("usb: dwc3: gadget: Correct ISOC DATA PIDs for short packets")
Cc: stable <stable@kernel.org>
Signed-off-by: Michael Grzeschik <m.grzeschik@pengutronix.de>
Link: https://lore.kernel.org/r/20220704141812.1532306-2-m.grzeschik@pengutronix.de
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
The DWC3_EVENT_PENDING flag is used to protect against invalid call to
top-half interrupt handler, which can occur when there's a delay in
software detection of the interrupt line deassertion.
However, the clearing of this flag was done prior to unmasking the
interrupt line, creating opportunity where the top-half handler can
come. This breaks the serialization and creates a race between the
top-half and bottom-half handler, resulting in losing synchronization
between the controller and the driver when processing events.
To fix this, make sure the clearing of the DWC3_EVENT_PENDING is done at
the end of the bottom-half handler.
Fixes: d325a1de49 ("usb: dwc3: gadget: Prevent losing events in event cache")
Cc: stable@vger.kernel.org
Signed-off-by: Thinh Nguyen <Thinh.Nguyen@synopsys.com>
Link: https://lore.kernel.org/r/8670aaf1cf52e7d1e6df2a827af2d77263b93b75.1656380429.git.Thinh.Nguyen@synopsys.com
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
The multiplier parameter of dwc3_gadget_calc_tx_fifo_size() was
not documented:
drivers/usb/dwc3/gadget.c:675: warning: Function parameter or member 'mult' not described in 'dwc3_gadget_calc_tx_fifo_size'
Signed-off-by: Mauro Carvalho Chehab <mchehab@kernel.org>
Link: https://lore.kernel.org/r/bd599a18cea45c57d91c69d3e30d8b1e8ea69dd1.1656409369.git.mchehab@kernel.org
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
The current logic to assign the max packet limit for IN endpoints attempts
to take the default HW value and apply the optimal endpoint settings based
on it. However, if the default value reports a TxFIFO size large enough
for only one max packet, it will divide the value and assign a smaller ep
max packet limit.
For example, if the default TxFIFO size fits 1024B, current logic will
assign 1024/3 = 341B to ep max packet size. If function drivers attempt to
request for an endpoint with a wMaxPacketSize of 1024B (SS BULK max packet
size) then it will fail, as the gadget is unable to find an endpoint which
can fit the requested size.
Functionally, if the TxFIFO has enough space to fit one max packet, it will
be sufficient, at least when initializing the endpoints.
Fixes: d94ea53198 ("usb: dwc3: gadget: Properly set maxpacket limit")
Cc: stable <stable@kernel.org>
Signed-off-by: Wesley Cheng <quic_wcheng@quicinc.com>
Link: https://lore.kernel.org/r/20220523213948.22142-1-quic_wcheng@quicinc.com
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
When dwc3_gadget_ep_cleanup_completed_requests() called to
dwc3_gadget_giveback() where the dwc3 lock is released, other thread is
able to execute. In this situation, usb_ep_disable() gets the chance to
clear endpoint descriptor pointer which leds to the null pointer
dereference problem. So needs to move the null pointer check to a proper
place.
Example call stack:
Thread#1:
dwc3_thread_interrupt()
spin_lock
-> dwc3_process_event_buf()
-> dwc3_process_event_entry()
-> dwc3_endpoint_interrupt()
-> dwc3_gadget_endpoint_trbs_complete()
-> dwc3_gadget_ep_cleanup_completed_requests()
...
-> dwc3_giveback()
spin_unlock
Thread#2 executes
Thread#2:
configfs_composite_disconnect()
-> __composite_disconnect()
-> ffs_func_disable()
-> ffs_func_set_alt()
-> ffs_func_eps_disable()
-> usb_ep_disable()
wait for dwc3 spin_lock
Thread#1 released lock
clear endpoint.desc
Fixes: 2628844812 ("usb: dwc3: gadget: Fix null pointer exception")
Cc: stable <stable@kernel.org>
Signed-off-by: Albert Wang <albertccwang@google.com>
Link: https://lore.kernel.org/r/20220518061315.3359198-1-albertccwang@google.com
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
According to the databook ep0 should be in setup phase during reset.
If host issues reset between control transfers, ep0 will be in an
invalid state. Fix this by issuing stall and restart on ep0 if it
is not in setup phase.
Also SW needs to complete pending control transfer and setup core for
next setup stage as per data book. Hence check ep0 state during reset
interrupt handling and make sure active transfers on ep0 out/in
endpoint are stopped by queuing ENDXFER command for that endpoint and
restart ep0 out again to receive next setup packet.
Signed-off-by: Mayank Rana <quic_mrana@quicinc.com>
Link: https://lore.kernel.org/r/1651693001-29891-1-git-send-email-quic_mrana@quicinc.com
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
If the controller hasn't DMA'ed the Setup data from its fifo, it won't
process the End Transfer command. Polling for the command completion may
block the driver from servicing the Setup phase and cause a timeout.
Previously we only check and delay issuing End Transfer in the case of
endpoint dequeue. Let's do that for all End Transfer scenarios.
Signed-off-by: Thinh Nguyen <Thinh.Nguyen@synopsys.com>
Link: https://lore.kernel.org/r/2fcf3b5d90068d549589a57a27a79f76c6769b04.1650593829.git.Thinh.Nguyen@synopsys.com
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
The driver shouldn't be able to issue End Transfer to the control
endpoint at anytime. Typically we should only do so in error cases such
as invalid/unexpected direction of Data Phase as described in the
control transfer flow of the programming guide. It _may_ end started
data phase during controller deinitialization from soft disconnect or
driver removal. However, that should not happen because the driver
should be maintained in EP0_SETUP_PHASE during driver tear-down. On
soft-connect, the controller should be reset from a soft-reset and there
should be no issue starting the control endpoint.
Signed-off-by: Thinh Nguyen <Thinh.Nguyen@synopsys.com>
Link: https://lore.kernel.org/r/3c6643678863a26702e4115e9e19d7d94a30d49c.1650593829.git.Thinh.Nguyen@synopsys.com
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Since we can't guarantee that the host won't send new Setup packet
before going through the device-initiated disconnect, don't prepare
beyond the Setup stage and keep the device in EP0_SETUP_PHASE. This
ensures that the device-initated disconnect sequence can go through
gracefully. Note that the controller won't service the End Transfer
command if it can't DMA out the Setup packet.
Signed-off-by: Thinh Nguyen <Thinh.Nguyen@synopsys.com>
Link: https://lore.kernel.org/r/6bacec56ecabb2c6e49a09cedfcac281fdc97de0.1650593829.git.Thinh.Nguyen@synopsys.com
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
If the GEVNTCOUNT indicates events in the event buffer, the driver needs
to acknowledge them before the controller can halt. Simply let the
interrupt handler acknowledges the remaining event generated by the
controller while polling for DSTS.DEVCTLHLT. This avoids disabling irq
and taking care of race condition between the interrupt handlers and
pullup().
Signed-off-by: Thinh Nguyen <Thinh.Nguyen@synopsys.com>
Link: https://lore.kernel.org/r/ea306ec93c41ccafbdb5d16404ff3b6eca299613.1650593829.git.Thinh.Nguyen@synopsys.com
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Don't do soft-disconnect if it's previously done. Likewise, don't do
soft-connect if the device is currently connected and running. It would
break normal operation.
Currently the caller of pullup() (udc's sysfs soft_connect) only checks
if it had initiated disconnect to prevent repeating soft-disconnect. It
doesn't check for soft-connect. To be safe, let's keep the check here
regardless whether the udc core is fixed.
Signed-off-by: Thinh Nguyen <Thinh.Nguyen@synopsys.com>
Link: https://lore.kernel.org/r/1c1345bd66c97a9d32f77d63aaadd04b7b037143.1650593829.git.Thinh.Nguyen@synopsys.com
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
-----BEGIN PGP SIGNATURE-----
iQFSBAABCAA8FiEEq68RxlopcLEwq+PEeb4+QwBBGIYFAmJu9FYeHHRvcnZhbGRz
QGxpbnV4LWZvdW5kYXRpb24ub3JnAAoJEHm+PkMAQRiGAyEH/16xtJSpLmLwrQzG
o+4ToQxSQ+/9UHyu0RTEvHg2THm9/8emtIuYyc/5FgdoWctcSa3AaDcveWmuWmkS
KYcdhfJsaEqjNHS3OPYXN84fmo9Hel7263shu5+IYmP/sN0DfQp6UWTryX1q4B3Q
4Pdutkuq63Uwd8nBZ5LXQBumaBrmkkuMgWEdT4+6FOo1mPzwdIGBxCuz1UsNNl5k
chLWxkQfe2eqgWbYJrgCQfrVdORXVtoU2fGilZUNrHRVGkkldXkkz5clJfapyZD3
odmZCEbrE4GPKgZwCmDERMfD1hzhZDtYKiHfOQ506szH5ykJjPBcOjHed7dA60eB
J3+wdek=
=39Ca
-----END PGP SIGNATURE-----
Merge 5.18-rc5 into usb-next
We need the USB fixes in here, and this resolves a merge issue in
drivers/usb/dwc3/drd.c
Reported-by: Stephen Rothwell <sfr@canb.auug.org.au>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
If the user sets the usb_request's no_interrupt, then there will be no
completion event for the request. Currently the driver incorrectly uses
the event status of a different request to report the status for a
request with no_interrupt. The dwc3 driver needs to check the TRB status
associated with the request when reporting its status.
Note: this is only applicable to missed_isoc TRB completion status, but
the other status are also listed for completeness/documentation.
Fixes: 6d8a019614 ("usb: dwc3: gadget: check for Missed Isoc from event status")
Cc: <stable@vger.kernel.org>
Signed-off-by: Thinh Nguyen <Thinh.Nguyen@synopsys.com>
Link: https://lore.kernel.org/r/db2c80108286cfd108adb05bad52138b78d7c3a7.1650673655.git.Thinh.Nguyen@synopsys.com
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
The list_for_each_entry_safe() macro saves the current item (n) and
the item after (n+1), so that n can be safely removed without
corrupting the list. However, when traversing the list and removing
items using gadget giveback, the DWC3 lock is briefly released,
allowing other routines to execute. There is a situation where, while
items are being removed from the cancelled_list using
dwc3_gadget_ep_cleanup_cancelled_requests(), the pullup disable
routine is running in parallel (due to UDC unbind). As the cleanup
routine removes n, and the pullup disable removes n+1, once the
cleanup retakes the DWC3 lock, it references a request who was already
removed/handled. With list debug enabled, this leads to a panic.
Ensure all instances of the macro are replaced where gadget giveback
is used.
Example call stack:
Thread#1:
__dwc3_gadget_ep_set_halt() - CLEAR HALT
-> dwc3_gadget_ep_cleanup_cancelled_requests()
->list_for_each_entry_safe()
->dwc3_gadget_giveback(n)
->dwc3_gadget_del_and_unmap_request()- n deleted[cancelled_list]
->spin_unlock
->Thread#2 executes
...
->dwc3_gadget_giveback(n+1)
->Already removed!
Thread#2:
dwc3_gadget_pullup()
->waiting for dwc3 spin_lock
...
->Thread#1 released lock
->dwc3_stop_active_transfers()
->dwc3_remove_requests()
->fetches n+1 item from cancelled_list (n removed by Thread#1)
->dwc3_gadget_giveback()
->dwc3_gadget_del_and_unmap_request()- n+1 deleted[cancelled_list]
->spin_unlock
Fixes: d4f1afe5e8 ("usb: dwc3: gadget: move requests to cancelled_list")
Signed-off-by: Wesley Cheng <quic_wcheng@quicinc.com>
Link: https://lore.kernel.org/r/20220414183521.23451-1-quic_wcheng@quicinc.com
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
The usb_ep_clear_halt() API can be called from the function driver, and
translates to dwc3_gadget_ep_set_halt(). This routine is shared with when
the host issues a clear feature ENDPOINT_HALT, and is differentiated by the
protocol argument. If the following sequence occurs, there can be a
situation where the delayed_status flag is improperly cleared for the wrong
SETUP transaction:
1. Vendor specific control transfer returns USB_GADGET_DELAYED_STATUS.
2. DWC3 gadget sets dwc->delayed_status to '1'.
3. Another function driver issues a usb_ep_clear_halt() call.
4. DWC3 gadget issues dwc3_stop_active_transfer() and sets
DWC3_EP_PENDING_CLEAR_STALL.
5. EP command complete interrupt triggers for the end transfer, and
dwc3_ep0_send_delayed_status() is allowed to run, as delayed_status
is '1' due to step#1.
6. STATUS phase is sent, and delayed_status is cleared.
7. Vendor specific control transfer is finished being handled, and issues
usb_composite_setup_continue(). This results in queuing of a data
phase.
Cache the protocol flag so that DWC3 gadget is aware of when the clear halt
is due to a SETUP request from the host versus when it is sourced from a
function driver. This allows for the EP command complete interrupt to know
if it needs to issue a delayed status phase.
Signed-off-by: Wesley Cheng <quic_wcheng@quicinc.com>
Link: https://lore.kernel.org/r/20220414073902.21960-1-quic_wcheng@quicinc.com
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Improve performance of isoc transfers in superspeed by increasing the
number of possible endpoints available, matching that of what was
configured for bulk transfers.
Signed-off-by: Dan Vacura <w36195@motorola.com>
Link: https://lore.kernel.org/r/20220420183338.445622-1-w36195@motorola.com
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Two kerneldoc comments in gadget.c have excess function parameter description or wrong
prototype name and one kerneldoc comment in core.c has an excess function parameter
description, resulting in these three doc-build warnings:
1. ./drivers/usb/dwc3/gadget.c:675: warning: Excess function parameter
'nfifos' description in 'dwc3_gadget_calc_tx_fifo_size'
2. ./drivers/usb/dwc3/gadget.c:700: warning: expecting prototype for
dwc3_gadget_clear_tx_fifo_size(). Prototype was for dwc3_gadget_clear_tx_fifos()
instead
3. ./drivers/usb/dwc3/core.c:347: warning: Excess function parameter 'ref_clk_per'
description in 'dwc3_ref_clk_period'
Fix the warnings by correcting the prototype name and removing excess parameter descriptions.
Signed-off-by: Kushagra Verma <kushagra765@outlook.com>
Link: https://lore.kernel.org/r/SI2PR01MB392995043CACD80884A13764F81C9@SI2PR01MB3929.apcprd01.prod.exchangelabs.com
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
It is recommended by the Synopsis databook to issue a DCTL.CSftReset
when reconnecting from a device-initiated disconnect routine. This
resolves issues with enumeration during fast composition switching
cases, which result in an unknown device on the host.
Reviewed-by: Thinh Nguyen <Thinh.Nguyen@synopsys.com>
Signed-off-by: Wesley Cheng <quic_wcheng@quicinc.com>
Link: https://lore.kernel.org/r/20220316011358.3057-1-quic_wcheng@quicinc.com
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
If a Setup packet is received but yet to DMA out, the controller will
not process the End Transfer command of any endpoint. Polling of its
DEPCMD.CmdAct may block setting up TRB for Setup packet, causing a
command timeout.
This may occur if the driver doesn’t service the completion interrupt of
the control status stage yet due to system latency, then it won’t
prepare TRB and start the transfer for the next Setup Stage. To the host
side, the control transfer had completed, and the host can send a new
Setup packet at this point.
In the meanwhile, if the driver receives an async call to dequeue a
request (triggering End Transfer) to any endpoint, then the driver will
service that End transfer first, blocking the control status stage
completion handler. Since no TRB is available for the Setup stage, the
Setup packet can’t be DMA’ed out and the End Transfer gets hung.
The driver must not block setting up of the Setup stage. So track and
only issue the End Transfer command only when there’s Setup TRB prepared
so that the controller can DMA out the Setup packet. Delay the End
transfer command if there's no Setup TRB available. This is applicable to
all DWC_usb3x IPs.
Co-developed-by: Wesley Cheng <quic_wcheng@quicinc.com>
Signed-off-by: Thinh Nguyen <Thinh.Nguyen@synopsys.com>
Signed-off-by: Wesley Cheng <quic_wcheng@quicinc.com>
Link: https://lore.kernel.org/r/20220309205402.4467-1-quic_wcheng@quicinc.com
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
To improve reading the code this patch moves the cases to start_isoc or
return the function under one common condition check.
Reviewed-by: Thinh Nguyen <Thinh.Nguyen@synopsys.com>
Signed-off-by: Michael Grzeschik <m.grzeschik@pengutronix.de>
Link: https://lore.kernel.org/r/20220306211251.2281335-2-m.grzeschik@pengutronix.de
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Currently the driver will schedule isoc transfers immediately on the
next interval, which is quite aggressive when the interval is 125us.
There's report that some platforms may need more time to process the
transfer, otherwise the controller may miss the first interval. Let's
keep it simple and give the controller at least 500us to schedule the
isoc transfer.
Link: https://lore.kernel.org/linux-usb/20220302143539.GI11577@pengutronix.de/
Tested-by: Michael Grzeschik <m.grzeschik@pengutronix.de>
Signed-off-by: Thinh Nguyen <Thinh.Nguyen@synopsys.com>
Link: https://lore.kernel.org/r/deb8146b8e1f7f8495ef2d5647017270934cb2d8.1646708142.git.Thinh.Nguyen@synopsys.com
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
The interrupt service routine registered for the gadget is a primary
handler which mask the interrupt source and a threaded handler which
handles the source of the interrupt. Since the threaded handler is
voluntary threaded, the IRQ-core does not disable bottom halves before
invoke the handler like it does for the forced-threaded handler.
Due to changes in networking it became visible that a network gadget's
completions handler may schedule a softirq which remains unprocessed.
The gadget's completion handler is usually invoked either in hard-IRQ or
soft-IRQ context. In this context it is enough to just raise the softirq
because the softirq itself will be handled once that context is left.
In the case of the voluntary threaded handler, there is nothing that
will process pending softirqs. Which means it remain queued until
another random interrupt (on this CPU) fires and handles it on its exit
path or another thread locks and unlocks a lock with the bh suffix.
Worst case is that the CPU goes idle and the NOHZ complains about
unhandled softirqs.
Disable bottom halves before acquiring the lock (and disabling
interrupts) and enable them after dropping the lock. This ensures that
any pending softirqs will handled right away.
Link: https://lkml.kernel.org/r/c2a64979-73d1-2c22-e048-c275c9f81558@samsung.com
Fixes: e5f68b4a3e ("Revert "usb: dwc3: gadget: remove unnecessary _irqsave()"")
Cc: stable <stable@kernel.org>
Reported-by: Marek Szyprowski <m.szyprowski@samsung.com>
Tested-by: Marek Szyprowski <m.szyprowski@samsung.com>
Signed-off-by: Sebastian Andrzej Siewior <bigeasy@linutronix.de>
Link: https://lore.kernel.org/r/Yg/YPejVQH3KkRVd@linutronix.de
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
With CPU re-ordering on write instructions, there might
be a chance that the HWO is set before the TRB is updated
with the new mapped buffer address.
And in the case where core is processing a list of TRBs
it is possible that it fetched the TRBs when the HWO is set
but before the buffer address is updated.
Prevent this by adding a memory barrier before the HWO
is updated to ensure that the core always process the
updated TRBs.
Fixes: f6bafc6a1c ("usb: dwc3: convert TRBs into bitshifts")
Cc: stable <stable@vger.kernel.org>
Reviewed-by: Pavankumar Kondeti <quic_pkondeti@quicinc.com>
Signed-off-by: Udipto Goswami <quic_ugoswami@quicinc.com>
Link: https://lore.kernel.org/r/1644207958-18287-1-git-send-email-quic_ugoswami@quicinc.com
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Synopsys introduced a new enhancement to DWC_usb32 called Multi-Stream
Transfer (MST) to improve bulk streams performance for SuperSpeed and
SuperSpeed Plus. This enhancement allows the controller to look ahead
and process multiple bulk streams.
Previously, to initiate a bulk stream transfer, the driver has to issue
Start Transfer command and wait for the stream to complete before
initiating a new stream. As a result, the controller does not process
TRBs beyond a single stream. With the enhancement, as long as there are
new requests, the dwc3 driver can keep preparing new TRBs and the
controller can keep caching and processing them without waiting for the
transfer completion.
The programming flow is similar to regular bulk endpoint with a few
additional rules:
1) Chained TRBs of the same stream must have a matching stream ID
2) The last TRB of a stream must have CHN=0
3) All the TRBs with LST=0 must have CSP=1
Depends on the application and usage, internal tests show significant
performance improvement in UASP transfers with this enhancement.
Signed-off-by: Thinh Nguyen <Thinh.Nguyen@synopsys.com>
Link: https://lore.kernel.org/r/cd9c7a8bf11f790983ac546222dd114893f16b3a.1638242424.git.Thinh.Nguyen@synopsys.com
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
The driver knows what it needs to set for GEVNTSIZn, and the controller
doesn't modify this register unless there's a hard reset. To save a few
microseconds of register read in read-modify-write operation, simply do
register write with the expected values. This can improve performance
when there are many interrupts generated, which the driver needs to
check and handle.
Signed-off-by: Thinh Nguyen <Thinh.Nguyen@synopsys.com>
Link: https://lore.kernel.org/r/efddf4ee5821c4bc5ae7ad90d629ec7a0ebcbf9a.1638240306.git.Thinh.Nguyen@synopsys.com
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
The controller doesn't check for Update Transfer command parameters
DEPCMDPAR{0,1,2}. Writing to these registers is unnecessary. Ignoring
this improves performance slightly by removing the register write
delay.
Signed-off-by: Thinh Nguyen <Thinh.Nguyen@synopsys.com>
Link: https://lore.kernel.org/r/997d9ebf38c6bba920d4ee77bd8c77bf81978a55.1638240306.git.Thinh.Nguyen@synopsys.com
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Greg Kroah-Hartman