Device sync APIs are actually wrappers for semaphores.
Let's replace them with semaphores.
Jira: ZEP-1411
Change-Id: I02c7cba21d21ff9288e452121e3b7ebb7d251bb4
Signed-off-by: Baohong Liu <baohong.liu@intel.com>
Make pointers to struct config_info const in prepration for a const
config_info.
Change-Id: I28789a7f1f26e4a0d499f5a89a567ae8c61eae51
Signed-off-by: Marcus Shawcroft <marcus.shawcroft@gmail.com>
The TX fifo threshold is pretty arbitrary.
Set this too big and too many interrupts will occur
for no good reason. Set it too small, and latency
in the SPI transactions is introduced. User's will probably
have to tune this per their application and SPI frequency, etc.
I think setting this to 50% is a good guess for now.
Change-Id: Ib325d40bc7ee10473d99443b3b3cd00fd6e4b95f
Signed-off-by: Chuck Jordan <cjordan@synopsys.com>
I've found many problems with the SPI driver and this repairs many of them.
The baud rate divisor was being derived from the CPU clock. But, some
targets may have a seperate clock attached to SPI. If the soc.h file
defines the symbol SPI_DW_SPI_CLOCK, it will use this instead
of CONFIG_SYS_CLOCK_HW_CYCLES_PER_SEC for the baud rate calculation.
completed() had a mistake where it would terminate the SPI transaction
too early, well before the tx data has cleared the FIFO. I found I couldn't
drive an OLED display correctly because completed() was wrong.
The repair is to now consider a new flag called spi->last_tx,
which will be set after the TX interrupt occurs with nothing to send any
longer. There is also a while loop added to SPIN until BUSY drops.
Another improvement is that push_data will NOT consider RX fifo size
if there is no RX going on. The calculation here when RX is going on
could go negative. I've added a check for that and prevent TX handling
if RX buffer is full. I think that is the intention -- to deal with RX first
if its fifos are more full.
In spi_dw_transceive, if we are only doing spi_write w/o reading,
don't enable RX interrupts at all. The OLED I'm working with failed
to have a pull-up on MISO SPI signal. As a result, a huge number of
garbage RX events arrive, and the interrupt handler finds there is
no rx buffer, so it tosses the data. But this is a waist of realtime.
It seems WRONG to enable RX interrupts if its something your not using,
so software can GATE these spurious events in this way.
With these changes, SPI can be used much more reliably, with FIFOs
that are deeper, and SPI devices that only require TX.
Change-Id: I0fe0745f2381c61c8a19ce086496b422a32a30a5
Signed-off-by: Chuck Jordan <cjordan@synopsys.com>
When using the Synopsys DesignWare Synchronous Serial Interface (SSI),
the FIFO depth can vary from 2-256, depending upon how this module is built.
For quark_se_ss, it was using a depth of 8. For EM Starterkit, it will be
32. Adding this now as a configurable option. A larger FIFO really helps
reduce SPI interrupts.
Change-Id: Id2bc8470bfc08ab447d38b89c7904cff010c63bd
Signed-off-by: Chuck Jordan <cjordan@synopsys.com>
There's no reason to require callers to cast their data to uint8_t *
when the data might e.g. originate in a packed struct or some other
data type. Instead, be nice to callers and let them use any pointer
they want. Additionally, declare the TX buffer as a const pointer so
unnecessary typecasts aren't needed for that either (if the data
originates in a const location).
Change-Id: I1482ca4e350b5a7fbda6871ed9f54f255af3aa9e
Signed-off-by: Johan Hedberg <johan.hedberg@intel.com>
Though it's an ARC core, Quark SE SS does not follow the same registers
mapping as the official DesignWare document. Some parts are common, some
not.
Instead of bloating spi_dw.c with a lot of #ifdef or rewriting a whole
new driver though the logic is 99% the same, it's then better to:
- centralize common macros and definitions into spi_dw.h
- have a specific spi_dw_quark_se_ss_reg.h for register map, clock
gating and register helpers dedicated to Quark SE SS.
- have a spi_dw_regs.h for the common case, i.e. not Quark SE SS.
GPIO CS emulation and interrupt masking ends up then in spi_dw.h.
Clock gating is specific thus found in respective *_regs.h header.
Adding proper interrupt masks to quark_se_ss soc.h file as well.
One of the main difference is also the interrupt management: through one
line or multiple lines (one for each interrupt: rx, tx and error). On
Quark SE Sensor Sub-System it has been set to use multiple lines, thus
introducing relevant Kconfig options and managing those when configuring
the IRQs.
Quark SE SS SPI controller is also working on a lower level, i.e. it
requires a tiny bit more logic from the driver. Main example is the data
register which needs to be told what is happening from the driver.
Taking the opportunity to fix minor logic issues:
- ICR register should be cleared by reading, only on error in the ISR
handler, but it does not harm doing it anyway and because Quark SE SS
requires to clear up interrupt as soon as they have been handled,
introducing a clear_interrupts() function called at the and of the ISR
handler.
- TXFTLR should be set after each spi_transceive() since last pull_data
might set it to 0.
- Enable the clock (i.e. open the clock gate) at initialization.
- No need to mask interrupts at spi_configure() since these are already
masked at initialization and at the end of a transaction.
- Let's use BIT() macro when relevant.
Change-Id: I24344aaf8bff3390383a84436f516951c1a2d2a4
Signed-off-by: Tomasz Bursztyka <tomasz.bursztyka@linux.intel.com>
Adds extern "C" { } blocks to header files so that they can be
safely used by C++ source files.
Change-Id: Ia4db0c36a5dac5d3de351184a297d2af0df64532
Signed-off-by: Peter Mitsis <peter.mitsis@windriver.com>
It might be necessary to emulate CS through a GPIO pin depending on
these 2 conditions:
- the controller's CS pin is not wired, and thus a GPIO pin is the only
option
- The controller is unstable at a certain frequency and cannot set/unset
CS reliably. This is actually a possible issue on DesignWare's SPI
controller in Quark SE or Quarks D2000 where it has been found
unstable at 1Mhz and above.
Change-Id: Ib6a06577906c005ddd347070d476a367a9c3da8a
Signed-off-by: Tomasz Bursztyka <tomasz.bursztyka@linux.intel.com>
- Refine how DFS is calulated now that it is strictely used to
manipulate buffer lengths.
- Fix threshold limit
- Tune RX threshold relevantly (reduce it if rx_len is lower than actual)
- Don't push more than available left space in FIFO
- Tune the private structure to lower memory space occupation
Change-Id: I65b1b48b996b2104cebcb24cc366fb4dcbf7d53b
Signed-off-by: Tomasz Bursztyka <tomasz.bursztyka@linux.intel.com>
IMR and ISR bits are same, but it stil better to differentiate them
properly. Also fixing naming where all ISR ends with an 'S'.
Change-Id: I2fc1e1d8d2743c3d98f5da40a5f4720a85c4f9a7
Signed-off-by: Tomasz Bursztyka <tomasz.bursztyka@linux.intel.com>
Registers offsets are hopefully all the same, but size differs.
On x86, thus 32bits support, CTRL0 or DR for instance are 32 bits r/w.
And DFS on 32 bits support is placed differently as well.
Change-Id: I5115d5c3c9bba71ece4a6f4a1d3d2fdc203c8da1
Signed-off-by: Tomasz Bursztyka <tomasz.bursztyka@linux.intel.com>
Only 8 bit frames were supported. Added support for bigger data frames
which can go up to 32 bits (on 32bits version of the controller, 16 bits
otherwise). Store the frame size in bytes during configure, and use it
during pull/push to read/write correct frame size.
Change-Id: Iae8c55442e0a205403aa3febd1811b36aaf4c5b6
Signed-off-by: Ramesh Thomas <ramesh.thomas@intel.com>
Signed-off-by: Tomasz Bursztyka <tomasz.bursztyka@linux.intel.com>
Baohong Liu