As explained in previous patch, this SPI controller seems to be
tested on SPI flash only before mass production and some bits are
swizzled under other SPI modes probably due to incorrect wiring
inside the silicon. Drop implementation of SPI mode 1/2/3 since
they are broken.
Also drop RT2880_SPI_MODE_BITS macro because we now have only
SPI_LSB_FIRST implemented and the mode_bits is so short that we
don't need a macro there.
Signed-off-by: Chuanhong Guo <gch981213@gmail.com>
Reviewed-by: NeilBrown <neil@brown.name>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
According to John Crispin (aka blogic) on IRC on Nov 26 2018:
so basically i made cs1 work for MTK/labs when i built
the linkit smart for them. the req-sheet said that cs1 should be proper
duplex spi. however ....
1) the core will always send 1 byte before any transfer, this is the
m25p80 command.
2) mode 3 is broken and bit reversed (?)
3) some bit are incorrectly wired in hw for mode2/3
we wrote a test script and test for [0-0xffff] on all modes and certain
bits are swizzled under certain conditions and it was not possible to
fix this even using a hack.
we then decided to use spi-gpio and i never removed the errornous code
basically the spi is fecked for anything but half duplex spi mode0
running a sflash on it
The controller will always send some data from OPCODE register under half
duplex mode before starting a full-duplex transfer, so the full-duplex
mode is broken.
This piece of code also make CS1 unavailable since it forces the
broken full-duplex mode to be used on CS1.
Signed-off-by: Chuanhong Guo <gch981213@gmail.com>
Reviewed-by: NeilBrown <neil@brown.name>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
The patch fixes following checkpatch.pl issue:
WARNING : braces {} are not necessary for single statement blocks
Signed-off-by: Sankalp Negi <sankalpnegi2310@gmail.com>
Reviewed-by: NeilBrown <neil@brown.name>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
The patch fixes following checkpatch.pl issue:
ERROR : space required before the open parenthesis
Signed-off-by: Sankalp Negi <sankalpnegi2310@gmail.com>
Reviewed-by: NeilBrown <neil@brown.name>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
The patch fixes following checkpatch.pl issue:
ERROR : code indent should use tabs where possible
Signed-off-by: Sankalp Negi <sankalpnegi2310@gmail.com>
Reviewed-by: NeilBrown <neil@brown.name>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
The patch fixes following checkpatch.pl issue:
ERROR : switch and case should be at the same indent
Signed-off-by: Sankalp Negi <sankalpnegi2310@gmail.com>
Reviewed-by: NeilBrown <neil@brown.name>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
The mt7621 SPI engine has a 32 byte buffer and the driver
currently only allows 32-byte read requests and 36 bytes writes
(there is a 4byte op/addr buffer).
This is an unnecessary limitation. As the SPI clock is controlled
by the host it is quite acceptable to send a larger message in
multiple smaller transactions. As long as Chip Select is kept asserted
the whole time, the SPI engine can be run multiple times for
a single SPI message.
This patch factors out the transaction logic and calls for each
transfer in the message. A write transfer might leave bytes in the
buffer to be combined with a following read transfer, as this is
a common pattern.
With this in place, we can remove the current max_transfer_size limit.
In testing, this increases the read throughput for a NOR flash chip
from 1.4MB/s to 2.3MB/s, a 50% improvement.
Signed-off-by: NeilBrown <neil@brown.name>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
All callers have a 'struct mt7621_spi' and that is all
mt7621_spi_wait_till_ready() needs. So just pass it
instead of the spi_device.
Signed-off-by: NeilBrown <neil@brown.name>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
This lock is never initialized, locked once, and never unlocked.
Clearly it is pointless - so remove it.
Signed-off-by: NeilBrown <neil@brown.name>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Remove the owner assignment form the platform driver as
platform_driver_register() already initializes the owner.
Found using coccinelle.
Signed-off-by: Christian Lütke-Stetzkamp <christian@lkamp.de>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
NeilBrown:
The code will fail with a warning if asked to transfer
more than 32 bytes at a time. So used max_transfer_size
interface to tell users about this.
Signed-off-by: John Crispin <blogic@openwrt.org>
Signed-off-by: NeilBrown <neil@brown.name>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Chuanhong Guo
Sankalp Negi
NeilBrown
Christian Lütke-Stetzkamp
John Crispin