Rather than keeping the asynchronous schedule running always, keep it
running only across USB mass storage transfers for now, as it seems
that keeping it running all the time interferes with certain control
transfers during device enumeration.
Note that running the async schedule all the time should not be an
issue, especially on EHCI HCD, as that one implements most of the
transfers using async schedule.
Note that we have usb_disable_asynch(), which however is utterly broken.
The usb_disable_asynch() blocks the USB core from doing async transfers
by setting a global flag. The async schedule should however be disabled
per USB controller. Moreover, setting a global flag does not prevent the
controller from using the async schedule, which e.g. the EHCI HCD does.
This patch implements additional callback to the controller, which
permits it to lock the async schedule and keep it running across
multiple transfers. Once the schedule is unlocked, it must also be
disabled. This thus prevents the async schedule from running outside
of the USB mass storage transfers.
Signed-off-by: Marek Vasut <marek.vasut+renesas@gmail.com>
Cc: Lukasz Majewski <lukma@denx.de>
Cc: Tom Rini <trini@konsulko.com>
Tested-by: Tom Rini <trini@konsulko.com> [omap3_beagle, previously failing]
At present dm/device.h includes the linux-compatible features. This
requires including linux/compat.h which in turn includes a lot of headers.
One of these is malloc.h which we thus end up including in every file in
U-Boot. Apart from the inefficiency of this, it is problematic for sandbox
which needs to use the system malloc() in some files.
Move the compatibility features into a separate header file.
Signed-off-by: Simon Glass <sjg@chromium.org>
These functions are CPU-related and do not use driver model. Move them to
cpu_func.h
Signed-off-by: Simon Glass <sjg@chromium.org>
Reviewed-by: Daniel Schwierzeck <daniel.schwierzeck@gmail.com>
Reviewed-by: Tom Rini <trini@konsulko.com>
Profiling the EHCI driver shows a significant performance problem in
ehci_submit_async(). Specifically, this function keeps enabling and
disabling async schedule back and forth for every single transaction.
However, enabling/disabling the async schedule does not take effect
immediatelly, but instead may take up to 1 mS (8 uFrames) to complete.
This impacts USB storage significantly, esp. since the recent reduction
of maximum transfer size to support more USB storage devices. This in
turn results in sharp increase in the number of ehci_submit_async()
calls. Since one USB storage BBB transfer does three such calls and
the maximum transfer size is 120 kiB, the overhead is 6 mS per 120 kiB,
which is unacceptable.
However, this overhead can be removed simply by keeping the async
schedule running. Specifically, the first transfer starts the async
schedule and then each and every subsequent transfer only adds a new
QH into that schedule, waits until the QH is completed and does NOT
disable the async schedule. The async schedule is stopped only by
shutting down the controller, which must happen before moving out
of U-Boot, otherwise the controller will corrupt memory.
Signed-off-by: Marek Vasut <marek.vasut+renesas@gmail.com>
Cc: Bin Meng <bmeng.cn@gmail.com>
Cc: Simon Glass <sjg@chromium.org>
This allows to disable the USB driver model in SPL because it checks
the CONFIG_SPL_DM_USB variable for SPL builds. Nothing changes for
regular non-SPL builds.
Signed-off-by: Sven Schwermer <sven@svenschwermer.de>
Workaround makes FS as default mode on all affected socs.
Add support to check erratum-A005275 validity for an soc. This info is
required to determine whether a given soc is affected by this erratum.
Add quirk for this erratum "has_fsl_erratum_a005275" . This quirk is used
to enable workaround for the errata
Force FS mode as default by:
- making EPS as FS
- setting PFSC bit to disable HS chirping
This workaround can be disabled by mentioning "no_erratum_a005275" in
hwconfig string
Signed-off-by: Chris Packham <judge.packham@gmail.com>
Reviewed-by: York Sun <york.sun@nxp.com>
When U-Boot started using SPDX tags we were among the early adopters and
there weren't a lot of other examples to borrow from. So we picked the
area of the file that usually had a full license text and replaced it
with an appropriate SPDX-License-Identifier: entry. Since then, the
Linux Kernel has adopted SPDX tags and they place it as the very first
line in a file (except where shebangs are used, then it's second line)
and with slightly different comment styles than us.
In part due to community overlap, in part due to better tag visibility
and in part for other minor reasons, switch over to that style.
This commit changes all instances where we have a single declared
license in the tag as both the before and after are identical in tag
contents. There's also a few places where I found we did not have a tag
and have introduced one.
Signed-off-by: Tom Rini <trini@konsulko.com>
Its a valid use case to call ehci_submit_async() with a NULL buffer
with length 0. E.g. from usb_set_configuration().
As invalidate_dcache_range() isn't able to judge if the address
NULL is valid or not (depending on the SoC hardware configuration it
might be valid) do the check in ehci_submit_async() as here we know
that we don't have to invalidate such a buffer.
Signed-off-by: Dirk Behme <dirk.behme@de.bosch.com>
Currently we check in ehci_shutdown() if ctrl is NULL after
dereferencing it.
Before this we have already dereferenced ctrl, ctrl->hccr,
and ctrl->hcor in ehci_get_portsc_register(), ehci_submit_root(),
and hci_common_init().
A better approach is to already check ctrl, ctrl->hccr, and ctrl->hcor
during the initialization in ehci_register() and usb_lowlevel_init()
and signal an error here via the return code.
Signed-off-by: Heinrich Schuchardt <xypron.glpk@gmx.de>
EHCD can handle any transfer length as long as there is enough free
heap space left, hence set the theoretical max number SIZE_MAX.
Signed-off-by: Bin Meng <bmeng.cn@gmail.com>
EHC reports supported maximum number of ports in the HCSPARAMS
register, so it's unnecessary to use a hardcoded config option
CONFIG_SYS_USB_EHCI_MAX_ROOT_PORTS.
Signed-off-by: Bin Meng <bmeng.cn@gmail.com>
Reviewed-by: Simon Glass <sjg@chromium.org>
Reviewed-by: Stefan Roese <sr@denx.de>
Tested-by: Stefan Roese <sr@denx.de>
USB 3.0 hubs have a slightly different hub descriptor than USB 2.0
hubs, with a fixed (rather than variable length) size. Change the
host controller drivers that access those last two fields
(DeviceRemovable and PortPowerCtrlMask) to use the union.
Signed-off-by: Bin Meng <bmeng.cn@gmail.com>
Reviewed-by: Simon Glass <sjg@chromium.org>
Reviewed-by: Stefan Roese <sr@denx.de>
Tested-by: Stefan Roese <sr@denx.de>
If the usb controller is not running, no need to shutdown it,
otherwise `usb stop` complains about:
"EHCI failed to shut down host controller".
To i.MX7D SDB, there are two usb ports, one Host, one OTG.
If we only plug one udisk to the Host port and then `usb start`,
the OTG controller for OTG port does not run actually. Then,
if `usb stop`, the OTG controller for OTG port will also be
shutdown, but it is not running.
This patch adds a check to only shutdown the running controller.
Signed-off-by: Peng Fan <peng.fan@nxp.com>
Cc: Marek Vasut <marex@denx.de>
Cc: Simon Glass <sjg@chromium.org>
Cc: Mateusz Kulikowski <mateusz.kulikowski@gmail.com>
Cc: Hans de Goede <hdegoede@redhat.com>
Cc: "Stefan Brüns" <stefan.bruens@rwth-aachen.de>
Cc: Stephen Warren <swarren@nvidia.com>
This commit fixes crash on BananaPi (and possibly others)
casued by 3f9f8a5b83.
Crash reason:
When no ops were passed to ehci_register(), USB host driver caused
NULL pointer dereference.
Signed-off-by: Mateusz Kulikowski <mateusz.kulikowski@gmail.com>
Some host controllers need addidional initialization after ehci_reset()
In non-dm implementation it is possible to use CONFIG_EHCI_HCD_INIT_AFTER_RESET.
This patch adds similar option to ehci drivers using dm.
Signed-off-by: Mateusz Kulikowski <mateusz.kulikowski@gmail.com>
Acked-by: Marek Vasut <marex@denx.de>
Reviewed-by: Tom Rini <trini@konsulko.com>
Reviewed-by: Simon Glass <sjg@chromium.org>
Tested-by: Simon Glass <sjg@chromium.org>
Fix the following warning on aarch64 introduced by using p2v/v2p
functions in the code:
In file included from ./arch/arm/include/asm/byteorder.h:29:0,
from include/compiler.h:125,
from include/image.h:19,
from include/common.h:88,
from drivers/usb/host/ehci-hcd.c:10:
drivers/usb/host/ehci-hcd.c: In function ‘ehci_td_buffer’:
drivers/usb/host/ehci-hcd.c:250:49: warning: cast to pointer from integer of different size [-Wint-to-pointer-cast]
td->qt_buffer[idx] = cpu_to_hc32(virt_to_phys((void *)addr));
^
include/linux/byteorder/little_endian.h:34:51: note: in definition of macro ‘__cpu_to_le32’
#define __cpu_to_le32(x) ((__force __le32)(__u32)(x))
^
drivers/usb/host/ehci-hcd.c:250:24: note: in expansion of macro ‘cpu_to_hc32’
td->qt_buffer[idx] = cpu_to_hc32(virt_to_phys((void *)addr));
Signed-off-by: Marek Vasut <marex@denx.de>
Cc: Stephen Warren <swarren@wwwdotorg.org>
Cc: Tom Rini <trini@konsulko.com>
If the USB EHCI is configured for little endian MMIO, make sure to
clear the USBMODE_BE flag from the USBMODE register.
Signed-off-by: Marek Vasut <marex@denx.de>
Cc: Daniel Schwierzeck <daniel.schwierzeck@gmail.com>
Cc: Hans de Goede <hdegoede@redhat.com>
Certain processor architectures, like MIPS, require that the USB
structures and transfer buffers are passed with their PA to the
USB controller. If VA is passed, the USB will not work. Add the
necessary virt_to_phys() calls into the USB EHCI code to make it
work.
Signed-off-by: Marek Vasut <marex@denx.de>
Cc: Daniel Schwierzeck <daniel.schwierzeck@gmail.com>
Cc: Hans de Goede <hdegoede@redhat.com>
Start split and complete split tokens need the hub address and the
downstream port of the first HS hub (device view).
The core of the function was duplicated in both host/ehci_hcd and
musb-new/usb-compat.h.
Signed-off-by: Stefan Brüns <stefan.bruens@rwth-aachen.de>
Reviewed-by: Marek Vasut <marex@denx.de>
Reviewed-by: Hans de Goede <hdegoede@redhat.com>
Tested-by: Hans de Goede <hdegoede@redhat.com>
The current name is inconsistent with other driver model data access
functions. Rename it and fix up all users.
Signed-off-by: Simon Glass <sjg@chromium.org>
Reviewed-by: Joe Hershberger <joe.hershberger@ni.com>
Now that we have a new header file for cache-aligned allocation, we should
move the stack-based allocation macro there also.
Signed-off-by: Simon Glass <sjg@chromium.org>
When an EHCI device is registered in device mode, the HW isn't actually
initialized at all, and hence isn't left in a running state. Consequently,
when the device is deregistered, ehci_shutdown() will fail, since the HW
bits it expects to see set in response to its shutdown requests will not
be sent, and the message "EHCI failed to shut down host controller." will
be printed.
Fix ehci-hcd.c to remember whether the device was registered in host or
device mode, and only call ehci_shutdown() for host mode registrations.
Signed-off-by: Stephen Warren <swarren@nvidia.com>
Without this we loose every other interrupt packet. We never noticed this
because with keyboards the packets which we were loosing would normally
be key release packets.
But now that we do keyrepeat in software instead of relying on the hid
idle functionality, missing a release will result in key repeat triggering.
This commit fixes this.
Signed-off-by: Hans de Goede <hdegoede@redhat.com>
Interrupt endpoints typically are polled for a long time by the usb
controller before they return anything, so calls to submit_int_msg() can
take a long time to complete this.
To avoid this the u-boot code has the an interrupt queue mechanism / API,
add support for this to the driver-model usb code and implement it for the
dm ehci code.
See the added doc comments for more details.
Signed-off-by: Hans de Goede <hdegoede@redhat.com>
Acked-by: Simon Glass <sjg@chromium.org>
This is a preparation patch for adding interrupt-queue support to the
ehci dm code.
Signed-off-by: Hans de Goede <hdegoede@redhat.com>
Acked-by: Simon Glass <sjg@chromium.org>
Short circuit the retry loop in legacy_hub_port_reset() by returning an
error from usb_control_msg() when a device was handed over to a companion
by the ehci code. This avoids trying to reset low / fullspeed devices 5
times needlessly. Also do not print an error when a device has been handed
over.
Signed-off-by: Hans de Goede <hdegoede@redhat.com>
When after a reset the port status connection bit is still set and the enable
bit is not then we're dealing with a full-speed device and should hand it over
to the companion controller.
Signed-off-by: Hans de Goede <hdegoede@redhat.com>
Without this usb-1 device descriptors do not get read properly.
Signed-off-by: Hans de Goede <hdegoede@redhat.com>
Acked-by: Simon Glass <sjg@chromium.org>
The ehci driver model code for finding the first upstream usb-2 hub before
this commit has a number of issues:
1) "if (!ttdev->speed != USB_SPEED_HIGH)" does not work because the '!'
takes presedence over the '!=' this should simply be
"if (ttdev->speed == USB_SPEED_HIGH)"
2) It makes ttdev point to the first upstream usb-2 hub, but ttdev should
point to the last usb-1 device before the first usb-2 hub (when going
upstream from the device), as ttdev is used to find the port of the
first usb-2 hub to which the the last usb-1 device is connected.
3) parent_devnum however should be set to the devnum of the first usb-2
hub, so we need to keep pointers around to both usb_device structs.
To complicate things further during enumeration usb_device.dev will point
to the parent udevice, where as during normal use it will point to
the actual udevice, we must handle both cases correctly.
This commit fixes all this making usb-1 devices attached to usb-2 hubs,
including usb-1 devices attached to usb-1 hubs attached to usb-2 hubs, work.
Signed-off-by: Hans de Goede <hdegoede@redhat.com>
Acked-by: Simon Glass <sjg@chromium.org>
Use usb_get_bus in dm ehci code rather then re-implementing it.
Signed-off-by: Hans de Goede <hdegoede@redhat.com>
Acked-by: Simon Glass <sjg@chromium.org>
Add a way for EHCI controller drivers to support driver model. Drivers can
call ehci_register() to register themselves in their probe() methods.
Signed-off-by: Simon Glass <sjg@chromium.org>
Reviewed-by: Marek Vasut <marex@denx.de>
The index cannot be used with driver model, and isn't needed anyway. Change
the parameter to a pointer.
Signed-off-by: Simon Glass <sjg@chromium.org>
Reviewed-by: Marek Vasut <marex@denx.de>
These are a pain with driver model because we might have different EHCI
drivers which want to implement them differently. Now that they use
consistent function signatures, we can in good conscience move them to
a struct.
Signed-off-by: Simon Glass <sjg@chromium.org>
Reviewed-by: Marek Vasut <marex@denx.de>
Fix non-driver-model EHCI to set up the EHCI operations correctly:
Signed-off-by: Tom Rini <trini@konsulko.com>
Move the bulk of the code in usb_lowlevel_init() into a separate function
which will also be used by driver model. Keep the CONFIG options out of
this function by providing a tweak flag for Faraday. We need to avoid using
CONFIG options in driver model code where possible, since it makes it
impossible to use multiple controllers in that code where they have
different options.
The CONFIG_EHCI_HCD_INIT_AFTER_RESET option is also kept out of the
common init function. With driver model the controller will be able to
perform this extra init itself after registering with the EHCI layer.
Signed-off-by: Simon Glass <sjg@chromium.org>
Reviewed-by: Marek Vasut <marex@denx.de>
With driver model we want to remove the controller pointer in struct udevice
and use driver model data structures instead. To prepare for this, move
access to this field to a function which can provide a different
implementation for driver model.
Signed-off-by: Simon Glass <sjg@chromium.org>
Reviewed-by: Marek Vasut <marex@denx.de>
Adjust this function so that it is passed an EHCI controller pointer so that
implementations can look up their controller. This makes the weak functions
use a consistent API.
Signed-off-by: Simon Glass <sjg@chromium.org>
Reviewed-by: Marek Vasut <marex@denx.de>
Adjust this function so that it is passed an EHCI controller pointer so that
implementations can look up their controller. This makes the weak functions
use a consistent API.
Signed-off-by: Simon Glass <sjg@chromium.org>
Reviewed-by: Marek Vasut <marex@denx.de>
Adjust this function so that it is passed an EHCI controller pointer so that
implementations can look up their controller.
Signed-off-by: Simon Glass <sjg@chromium.org>
Reviewed-by: Marek Vasut <marex@denx.de>
Add a private data pointer that clients of EHCI can use to access their
private information. This establishes a link between struct ehci_ctrl and
its associated controller data structure.
Signed-off-by: Simon Glass <sjg@chromium.org>
Reviewed-by: Marek Vasut <marex@denx.de>
Adjust this function so that it is passed an EHCI controller pointer so that
implementations can look up their controller.
Signed-off-by: Simon Glass <sjg@chromium.org>
Reviewed-by: Marek Vasut <marex@denx.de>
Put these at the top of the file so they are in one place. Also add function
prototypes to the header file to avoid call site mismatches.
Signed-off-by: Simon Glass <sjg@chromium.org>
Reviewed-by: Marek Vasut <marex@denx.de>
Change addresses to unsigned long to be compatible with 64-bit builds.
Regardless of fixing warnings, the device is still only 32-bit capable.
Signed-off-by: Rob Herring <robh@kernel.org>
Cc: Marek Vasut <marex@denx.de>
Currently create_int_queue is only implemented by the ehci code, and that
does not honor interrupt intervals, but other drivers which might also want
to implement create_int_queue may honor intervals, so add an interval param.
Signed-off-by: Hans de Goede <hdegoede@redhat.com>
