Paul Burton [Sun, 30 Apr 2017 19:57:08 +0000 (21:57 +0200)]
net: pch_gbe: Add cache maintenance
On MIPS systems DMA isn't coherent with the CPU caches unless an IOCU is
present. When there is no IOCU we need to writeback or invalidate the
data caches at appropriate points. Perform this cache maintenance in
the pch_gbe driver which is used on the MIPS Boston development board.
Signed-off-by: Paul Burton <paul.burton@imgtec.com> Reviewed-by: Bin Meng <bmeng.cn@gmail.com> Tested-by: Bin Meng <bmeng.cn@gmail.com> Signed-off-by: Daniel Schwierzeck <daniel.schwierzeck@gmail.com> Reviewed-by: Simon Glass <sjg@chromium.org> Acked-by: Joe Hershberger <joe.hershberger@ni.com>
Paul Burton [Sun, 30 Apr 2017 19:57:07 +0000 (21:57 +0200)]
net: pch_gbe: CPU accessible addresses are virtual
Use the virt_to_bus & bus_to_virt functions rather than phys_to_bus &
bus_to_phys, since the addresses accessed by the CPU will be virtual
rather than physical. On MIPS physical & virtual addresses differ as we
use virtual addresses in kseg0, and attempting to use physical addresses
directly caused problems as they're in the user segment which would be
mapped via the uninitialised TLB.
Signed-off-by: Paul Burton <paul.burton@imgtec.com> Signed-off-by: Daniel Schwierzeck <daniel.schwierzeck@gmail.com> Reviewed-by: Bin Meng <bmeng.cn@gmail.com> Tested-by: Bin Meng <bmeng.cn@gmail.com> Reviewed-by: Simon Glass <sjg@chromium.org> Acked-by: Joe Hershberger <joe.hershberger@ni.com>
Paul Burton [Sun, 30 Apr 2017 19:57:06 +0000 (21:57 +0200)]
net: pch_gbe: Fix rx descriptor buffer addresses
The loop to set up buffer addresses in rx descriptors always operated on
descriptor 0, rather than on each descriptor sequentially. Fix this in
order to setup correct buffer addresses for each descriptor.
Signed-off-by: Paul Burton <paul.burton@imgtec.com> Signed-off-by: Daniel Schwierzeck <daniel.schwierzeck@gmail.com> Reviewed-by: Bin Meng <bmeng.cn@gmail.com> Tested-by: Bin Meng <bmeng.cn@gmail.com> Reviewed-by: Simon Glass <sjg@chromium.org> Acked-by: Joe Hershberger <joe.hershberger@ni.com>
Paul Burton [Sun, 30 Apr 2017 19:57:05 +0000 (21:57 +0200)]
net: pch_gbe: Reset during probe
Using the EG20T gigabit ethernet controller on the MIPS Boston board, we
find that we have to reset the controller in order for the RGMII link to
the PHY to become functional. Without doing so we constantly time out in
pch_gbe_mdio_ready.
Signed-off-by: Paul Burton <paul.burton@imgtec.com> Reviewed-by: Bin Meng <bmeng.cn@gmail.com> Tested-by: Bin Meng <bmeng.cn@gmail.com> Signed-off-by: Daniel Schwierzeck <daniel.schwierzeck@gmail.com> Reviewed-by: Simon Glass <sjg@chromium.org> Acked-by: Joe Hershberger <joe.hershberger@ni.com>
net: zynq_gem: Use wait_for_bit with non breakable
Use wait_for_bit to be non breakable as using it with
breakable causes issue of un interruptible auto negotiation.
This is due to the ctrlc pressed will taken for wait_for_bit()
abort during phy_read() and hence not coming out of
auto negotiation.
Signed-off-by: Siva Durga Prasad Paladugu <sivadur@xilinx.com> Signed-off-by: Michal Simek <michal.simek@xilinx.com> Acked-by: Joe Hershberger <joe.hershberger@ni.com>
Sekhar Nori [Mon, 8 May 2017 15:19:56 +0000 (20:49 +0530)]
drivers: net: cpsw: abort init() on aneg timeout
Abort CPSW driver init when auto-negotiation of link
times out. Currently, the code ignores return status
of phy_startup(), and goes ahead with network operation
(like DHCP) even though the link may be down.
Instead, abort init process if link is down or if there
is another error, so phy_startup() can easily be retried
again. This also helps quick fallback to next network interface
(like USB RNDIS) without inordinate delay.
Tested on AM571x IDK and AM335x BeagleBone black.
Signed-off-by: Sekhar Nori <nsekhar@ti.com> Reviewed-by: Tom Rini <trini@konsulko.com> Acked-by: Joe Hershberger <joe.hershberger@ni.com>
The .read_rom_hwaddr net_ops hook does not check the return value, which
is why it was never caught that we are currently returning 0 if the
read_rom_hwaddr function return -ENOSYS and -ENOSYS otherwise.
In this case we can simplify this by just returning the result of the
function.
Signed-off-by: Olliver Schinagl <oliver@schinagl.nl> Acked-by: Joe Hershberger <joe.hershberger@ni.com> Acked-by: Michal Simek <michal.simek@xilinx.com>
Jacob Chen [Mon, 27 Mar 2017 08:54:17 +0000 (16:54 +0800)]
net: designware: Add phy supply support
Some board need a regulator for gmac phy, so add this code to handle it. Signed-off-by: Jacob Chen <jacob2.chen@rock-chips.com> Acked-by: Joe Hershberger <joe.hershberger@ni.com>
Philipp Tomsich [Sun, 26 Mar 2017 16:50:23 +0000 (18:50 +0200)]
net: Kconfig:make PHY_GIGE and individual Micrel PHYs selectable
This change migrate the following configuration options for Kconfig:
* PHY_GIGE, indicates that a controller (with an appropriate PHY) is
Gigabit capable and enables extra support in the miiutil for
parsing the status of Gigabit PHYs
* adds configuration options for Micrel KSZ9021 and KSZ9031 GbE PHYs,
which previously had to enabled through a board-specific config file
Signed-off-by: Philipp Tomsich <philipp.tomsich@theobroma-systems.com> Acked-by: Maxime Ripard <maxime.ripard@free-electrons.com> Acked-by: Joe Hershberger <joe.hershberger@ni.com>
Simon Glass [Fri, 19 May 2017 02:09:43 +0000 (20:09 -0600)]
dm: test: Separate out the bus DT offset test
We cannot access the device tree via an offset when running in livetree
mode. Separate out that part of the bus' children tests and mark it as
for the flat tree only.
Simon Glass [Fri, 19 May 2017 02:09:18 +0000 (20:09 -0600)]
dm: gpio: Refactor to prepare for live tree support
Move the main part of the GPIO request function into a separate function
so that it can be used by the live tree function when added. Update the
xlate method to use a node reference.
Update all GPIO drivers to handle the modified xlate() method.
Simon Glass [Fri, 19 May 2017 02:09:17 +0000 (20:09 -0600)]
dm: core: Run tests with both livetree and flat tree
Some tests require either livetree or flat tree. Add flags to allow the
tests to specify this. Adjust the test runner to run with livetree (if
supported) and then flat tree.
Some video tests are quite slow and running on flat tree adds little extra
test value, so run these on livetree only.
Simon Glass [Fri, 19 May 2017 02:09:16 +0000 (20:09 -0600)]
dm: test: Add support for running tests with livetree
It is useful to run the driver model tests with both livetree and flat
tree in case something is different between the two. Add this feature to
the test runner.
Simon Glass [Fri, 19 May 2017 02:09:13 +0000 (20:09 -0600)]
sandbox: Add a way to reset sandbox state for tests
Running a new test should reset the sandbox state to avoid tests
interferring with each other. Move the existing state-reset code into a
function so it can be used from tests.
Also update the code to reset the SPI devices and adjust the test code to
call it.
Simon Glass [Fri, 19 May 2017 02:09:03 +0000 (20:09 -0600)]
dm: core: Add device-based 'read' functions to access DT
It is common to read a device-tree property from the node associated with
a device. Add convenience functions to do this so that drivers do not need
to deal with accessing the ofnode from the device.
These functions all start with 'dev_read_' to provide consistent naming
for all functions which read information from a device's device tree node.
These are inlined when using the flat DT to save code size. The live tree
implementation is added in a later commit.
Simon Glass [Fri, 19 May 2017 02:09:02 +0000 (20:09 -0600)]
dm: core: Add a place to put extra device-tree reading functions
Some functions deal with structured data rather than simple data types.
It makes sense to have these in their own file. For now this just has a
function to read a flashmap entry. Move the data types also.
Simon Glass [Fri, 19 May 2017 02:08:58 +0000 (20:08 -0600)]
dm: core: Add operations on device tree references
Since U-Boot supports both a live tree and a flat tree, we need an easy
way to access the tree without worrying about which is currently active.
To support this, U-Boot has the concept of an ofnode, which can refer
either to a live tree node or a flat tree node.
For the live tree, the reference contains a pointer to the node (struct
device_node *) or NULL if the node is invalid. For the flat tree, the
reference contains the node offset or -1 if the node is invalid.
Add a basic set of operations using ofnodes. These are implemented by
using either libfdt functions (in the case of a flat DT reference) or
the live-tree of_...() functions.
Note that it is not possible to have both live and flat references active
at the same time. As soon as the live tree is available, everything in
U-Boot should switch to using that. This avoids confusion and allows us to
assume that the type of a reference is simply based on whether we have a
live tree yet, or not.
Simon Glass [Fri, 19 May 2017 02:08:54 +0000 (20:08 -0600)]
dm: core: Add livetree access functions
Add a basic assortment of functions to access the live device tree. These
come from Linux v4.9 and are modified for U-Boot to the minimum extent
possible. While these functions are now very stable in Linux, it will be
possible to merge in fixes if needed.
Simon Glass [Mon, 24 Apr 2017 02:02:10 +0000 (20:02 -0600)]
dm: mmc: Rewrite mmc_blk_probe()
This function is called when the MMC block device is being probed. There
is a recursive call in this function since find_mmc_device() itself can
cause the MMC device to be probed.
Admittedly the MMC device should already be probed, since we would not be
probing its child otherwise, but the current code is unnecessarily
convoluted.
Rewrite this to access the MMC structure directly.
Simon Glass [Mon, 24 Apr 2017 02:02:09 +0000 (20:02 -0600)]
dm: mmc: Check that drivers have operations
When binding a new MMC device, make sure that it has the required
operations. Since for now we still support *not* having the operations
(with CONFIG_DM_MMC_OPS not enabled) it makes sense to add this check.
we expect that the block devices for mmc0 and mmc1 would have device
numbers of 0 and 1 respectively.
Unfortunately this does not currently always happen. If there is another
MMC device earlier in the driver model data structures its block device
will be created first. It will therefore get device number 0 and mmc0
will therefore miss out. In this case the MMC device will have sequence
number 0 but its block device will not.
To avoid this, allow a device to request a device number and bump any
existing device number that is using it. This all happens during the
binding phase so it is safe to change these numbers around. This allows
device numbers to match the aliases in all circumstances.
Simon Glass [Sun, 23 Apr 2017 01:10:56 +0000 (19:10 -0600)]
dm: mmc: Don't call board_mmc_power_init() with driver model
We should not call out to board code from drivers. With driver model,
mmc_power_init() already has code to use a named regulator, but the
legacy code path remains. Update the code to make this clear.
Simon Glass [Wed, 17 May 2017 23:18:10 +0000 (17:18 -0600)]
dm: core: Add ofnode to represent device tree nodes
With live tree we need a struct device_node * to reference a node. With
the existing flat tree, we need an int offset. We need to unify these into
a single value which can represent both.
Add an ofnode union for this and adjust existing code to move to this.
Simon Glass [Wed, 17 May 2017 23:18:09 +0000 (17:18 -0600)]
dm: core: Replace of_offset with accessor (part 2)
At present devices use a simple integer offset to record the device tree
node associated with the device. In preparation for supporting a live
device tree, which uses a node pointer instead, refactor existing code to
access this field through an inline function.
Simon Glass [Wed, 17 May 2017 23:18:08 +0000 (17:18 -0600)]
dm: core: Dont export dm_scan_fdt_node()
This function is only used in one place. It is better to just declare it
internally since there is a simpler replacement for use outside the
driver-model core code.
projects / u-boot / log