usb: ehci: Do not use memset on dma coherent memory ...

memset is an optimized operation that at least on ARM64 may only be
called on cached memory, see 32e8842c40 ("ARM: lib64: Make string
functions aware of MMU configuration"). To avoid crashes in the ehci
driver we no longer call memset on memory allocated with
dma_alloc_coherent(), but use a simple memzero32 function instead.

Reported-by: Elmar Albert <EAlbert@data-modul.com>
Signed-off-by: Sascha Hauer <s.hauer@pengutronix.de>

usb: host: ehci: Allocate only NUM_QH queue heads for qh_list ...

Use NUM_QH, not NUM_TD to specify allocation size.

Signed-off-by: Andrey Smirnov <andrew.smirnov@gmail.com>
Signed-off-by: Sascha Hauer <s.hauer@pengutronix.de>

ush: host: ehci: Simplify QH initialization in ehci_submit_async() ...

We use a fixed circular list for asynchronous schedule that never
changes, so if we drop an explicit memset() that zeros out an entire
struct and replace it with code initializine all of the fields
explicitly, we can set QH list once in ehci_init() and never touch it
again.

While at it move qt_altnext initialization to ehci_init() as well
since we never change that field either.

Signed-off-by: Andrey Smirnov <andrew.smirnov@gmail.com>
Signed-off-by: Sascha Hauer <s.hauer@pengutronix.de>

usb: host: ehci: Treat ehci->qh_list as an array ...

Since ehci->qh_list is an array, convert the code to use
ehci->qh_list[0] to access its first element to make things a bit more
clear. No functional change intended.

Signed-off-by: Andrey Smirnov <andrew.smirnov@gmail.com>
Signed-off-by: Sascha Hauer <s.hauer@pengutronix.de>

usb: host: ehci: Drop explicit memset() in ehci_init() ...

Memory allocated by dma_alloc_coherent() is already zeroed out,
ehci->qh_list[0] never changes during the operation of the driver and
ehci->qh_list[1] will be explicitly initialized by ehci_submit_async()
so this additional memset() shouldn't be necessary.

Signed-off-by: Andrey Smirnov <andrew.smirnov@gmail.com>
Signed-off-by: Sascha Hauer <s.hauer@pengutronix.de>

usb: host: ehci: Replace explicit printf() with dev_err() ...

Signed-off-by: Andrey Smirnov <andrew.smirnov@gmail.com>
Signed-off-by: Sascha Hauer <s.hauer@pengutronix.de>

usb: host: ehci: Add a comment explaing IOC setting for data ...

Add a comment explaining the rationale behinde QT_TOKEN_IOC(req ==
NULL).

Signed-off-by: Andrey Smirnov <andrew.smirnov@gmail.com>
Signed-off-by: Sascha Hauer <s.hauer@pengutronix.de>

usb: host: ehci: Use bool to simplify ehci_submit_async() ...

Change the type of 'c' to bool and drop the trigrah.

Signed-off-by: Andrey Smirnov <andrew.smirnov@gmail.com>
Signed-off-by: Sascha Hauer <s.hauer@pengutronix.de>

usb: host: ehci: Drop needless assignments in ehci_submit_async() ...

Neither of those is necessary, since both will be overwritten further
down in the code.

Signed-off-by: Andrey Smirnov <andrew.smirnov@gmail.com>
Signed-off-by: Sascha Hauer <s.hauer@pengutronix.de>

usb: host: ehci: Simplify ehci_submit_async()'s epilogue ...

Re-organize the epilogue of ehci_submit_async() to bail out early if
token is still marked as "active" and drop no longer necessary check
for "dev->status != USB_ST_NOT_PROC". While at it return -EIO instead
of -1 in the case of error.

Signed-off-by: Andrey Smirnov <andrew.smirnov@gmail.com>
Signed-off-by: Sascha Hauer <s.hauer@pengutronix.de>

usb: host: ehci: Initialize qTDs explicitly ...

Convert the code to initialize only used qTDs as well as initializing
all of the fields explicitly without doing a wholesale memset() first.

Signed-off-by: Andrey Smirnov <andrew.smirnov@gmail.com>
Signed-off-by: Sascha Hauer <s.hauer@pengutronix.de>

usb: host: ehci: Simplify ehci_td_buffer() ...

Rework the code of ehci_td_buffer() with following trivial changes:

    * Switch to using dma_addr_t for 'delta' and 'next'
    * Convert while to for loop
    * Replace explicit magic number with dedicated contants derived
      via ARRAY_SIZE
    * Use ALIGN_DOWN to calculate 'next'
    * Return -ENOMEM instead of -1 when we ran out of buffers

Signed-off-by: Andrey Smirnov <andrew.smirnov@gmail.com>
Signed-off-by: Sascha Hauer <s.hauer@pengutronix.de>

usb: host: ehci: Simplify qTD buffer synchronization ...

Only two buffers ever need to be sychronized and DMA direction of it
is always known based on the type of request we are
processing. Simplify the code by moving synchronization to be a part
of ehci_prepare_qtd(). While at it convert the code to use
dma_map_single() API.

Signed-off-by: Andrey Smirnov <andrew.smirnov@gmail.com>
Signed-off-by: Sascha Hauer <s.hauer@pengutronix.de>

usb: host: ehci: Introduce ehci_prepare_qtd() ...

Move shared struct qTD initialization code into a subroutine and
convert the rest of the code to use it. No functional change intended.

Signed-off-by: Andrey Smirnov <andrew.smirnov@gmail.com>
Signed-off-by: Sascha Hauer <s.hauer@pengutronix.de>

usb: host: ehci: Drop unnecessary cleanup code ...

This looks like a leftover from U-Boot. Analogous code there would
walk all qTDs and call ehci_free() on each one of them, but since we
don't have anything of the sort it seems the whole loop can be
dropped. Note that U-Boot dropped that code in
de98e8b

Signed-off-by: Andrey Smirnov <andrew.smirnov@gmail.com>
Signed-off-by: Sascha Hauer <s.hauer@pengutronix.de>

usb: host: ehci: Replace magic number with macros ...

Import a number of missing macros from U-Boot and convert ehci-hcd to
use them instead of explicitly specifying magic shifts. No functional
change intended.

Signed-off-by: Andrey Smirnov <andrew.smirnov@gmail.com>
Signed-off-by: Sascha Hauer <s.hauer@pengutronix.de>

usb: host: ehci: Use to USBSTS to wait for transfer completion ...

Since we already set up our last qTD to have IOC flag set we may as
well take advange of that and poll USBSTS to wait for transfer
completion. Doesn't change much, but allows us to drop a custom
polling loop and re-use handshake() instead.

Signed-off-by: Andrey Smirnov <andrew.smirnov@gmail.com>
Signed-off-by: Sascha Hauer <s.hauer@pengutronix.de>

usb: host: ehci: Share code to enable/disable async schedule ...

Signed-off-by: Andrey Smirnov <andrew.smirnov@gmail.com>
Signed-off-by: Sascha Hauer <s.hauer@pengutronix.de>

usb: host: ehci: Do not zero out DMA coherent memory ...

Memory returned by dma_alloc_coherent() should already be zeroed
out, so there's no need to do this explicitly.

Signed-off-by: Andrey Smirnov <andrew.smirnov@gmail.com>
Signed-off-by: Sascha Hauer <s.hauer@pengutronix.de>

usb: host: ehci: do not use dev->priv ...

An ehci can be registered with ehci_register which is passed a struct
device_d *. In that case the priv pointer may already be used by the
caller, so we must not use it in the ehci code. At least for the Atmel
ehci driver this fixes a bug, here dev->priv is set two times to
different values.
Since we need dev->priv in the ehci code to get the controller in
ehci_detect() we can no longer implement that without the help of the
caller, hence we eport ehci_detect() and epect it to be called by
the code which registers a ehci host.

Signed-off-by: Sascha Hauer <s.hauer@pengutronix.de>

usb: host: ehci: add ehci_unregister() ...

ehci_register() allocates data and registers a ehci host. Add
ehci_unregister() to properly halt the controller and to free the memory
again.. To do so, change ehci_register() to return the ehci host rather
than an error code.

Signed-off-by: Sascha Hauer <s.hauer@pengutronix.de>

usb: host: ehci: rename ehci_priv to ehci_host ...

As we are going to export the structure as a cookie to others
rename it to ehci_host.

Signed-off-by: Sascha Hauer <s.hauer@pengutronix.de>

usb: ehci: pass full speed devices to companion controller ...

According to the "Enhanced Host Controller Interface Specification for
Universal Serial Bus" after a USB port reset the EHCI Driver checks the
PortEnable bit in the PORTSC register. If set to a one, the connected device is
a high-speed device [...]. At the time the EHCI Driver receives the port reset
and enable request the LineStatus bits might indicate a low-speed device.
Additionally, when the port reset process is complete, the PortEnable field may
indicate that a full-speed device is attached. In either case the EHCI driver
sets the PortOwner bit in the PORTSC register to a one to release port
ownership to a companion host controller.

Signed-off-by: Peter Mamonov <pmamonov@gmail.com>
Signed-off-by: Sascha Hauer <s.hauer@pengutronix.de>

usb: ehci: forward phy given in registration data to host ...

Allow to pass a phy in the registration data and forward it to
the usb_host structure.

Signed-off-by: Sascha Hauer <s.hauer@pengutronix.de>

usb: ehci: honour timeout value ...

usb_control_msg() and usb_bulk_msg() have a timeout parameter. Honour
this in the ehci driver instead of using a default timeout.

Signed-off-by: Sascha Hauer <s.hauer@pengutronix.de>

Merge branch 'for-next/misc'

ehci-hcd: preserve DTC in QH for interrupt transactions ...

In Linux ehci-q Data Toggle Control bit for interrupt transactions
is preserved in Queue Head (EHCI 3.6.2). So don't set 14's bit in
qh_endpt1 and don't manage toggle bit anywhere in int transactions.

This fixes an idle issue for USB keyboards, where keys
were physically changed but barebox haven't seen them in polling
until the next USB keyboard polling occurs.

Signed-off-by: Aleksey Kuleshov <rndfax@yandex.ru>
Signed-off-by: Sascha Hauer <s.hauer@pengutronix.de>

ehci-hcd: remove useless timeout ...

This is just a regular iteration loop so no timeout needed

Signed-off-by: Aleksey Kuleshov <rndfax@yandex.ru>
Signed-off-by: Sascha Hauer <s.hauer@pengutronix.de>

driver: replace dev_request_mem_region with dev_request_mem_resource ...

dev_request_mem_region doesn't work properly one some SoCs on which
PTR_ERR() values clash with valid return values from dev_request_mem_region.
Replace them with dev_request_mem_resource where possible.

This patch has been generated with the following semantic patch:

// <smpl>
@@
expression d;
expression n;
expression io;
identifier func;
@@
func(...) {
+struct resource *iores;
<+...
-io = dev_request_mem_region(d, n);
-if (IS_ERR(io)) {
+iores = dev_request_mem_resource(d, n);
+if (IS_ERR(iores)) {
...
-	return PTR_ERR(io);
-}
+	return PTR_ERR(iores);
+}
+io = IOMEM(iores->start);
...+>
}

@@
expression d;
expression n;
expression io;
identifier func;
@@
func(...) {
+struct resource *iores;
<+...
-io = dev_request_mem_region(d, n);
-if (IS_ERR(io)) {
+iores = dev_request_mem_resource(d, n);
+if (IS_ERR(iores))
-	return PTR_ERR(io);
-}
+	return PTR_ERR(iores);
+io = IOMEM(iores->start);
...+>
}

@@
expression d;
expression n;
expression io;
identifier func;
@@
func(...) {
+struct resource *iores;
<+...
-io = dev_request_mem_region(d, n);
-if (IS_ERR(io)) {
-	ret = PTR_ERR(io);
+iores = dev_request_mem_resource(d, n);
+if (IS_ERR(iores)) {
+	ret = PTR_ERR(iores);
...
}
+io = IOMEM(iores->start);
...+>
}

@@
expression d;
expression n;
expression io;
identifier func;
@@
func(...) {
+struct resource *iores;
<+...
-io = dev_request_mem_region(d, n);
+iores = dev_request_mem_resource(d, n);
+if (IS_ERR(iores))
+	return PTR_ERR(iores);
+io = IOMEM(iores->start);
...+>
}

@@
identifier func;
@@
func(...) {
<+...
struct resource *iores;
-struct resource *iores;
...+>
}
// </smpl>

Signed-off-by: Sascha Hauer <s.hauer@pengutronix.de>

usb: ehci-hcd: Use regular mdelay() ...

The USB core will make sure we do not re-enter the ehci driver, so
we can use regular mdelay instead of mdelay_non_interruptible().

Signed-off-by: Sascha Hauer <s.hauer@pengutronix.de>