On 31/01/2019 23:47, NeilBrown wrote:
On Thu, Jan 31 2019, Stefan Roese wrote:
On 31.01.19 13:29, Mark Brown wrote:
On Thu, Jan 31, 2019 at 11:01:16AM +0100, Stefan Roese wrote:
I'm wondering about the current status of the staging SPI driver
for the MIPS MT7621 (MT7628/88). Are there any plans to pull this
driver into "drivers/spi" soon? If not, what is blocking this
integration?
I've no idea what the status of this driver is, I don't think I've ever
heard of it before - if someone sends a patch I'll review it. I think
staging drivers are supposed to have a TODO in there with known issues?
Yes, thanks. I'll take care of the TODO's and will send a patch to move
it out of staging for review.
Thanks,
Stefan
I had a quick look through..
#define SPI_BPW_MASK(bits) BIT((bits) - 1)
This is an odd use of the BIT macro, as we don't want a bit, we want a
mask.
I'd make that
#define SPI_BPW_MASK(bits) ((1 << (bits)) - 1)
mt7621_spi_reset() contains some magic numbers. Using named
constants would be better.
According to
http://ftp.mqmaker.com/WiTi/Docs/Software/MT7621_ProgrammingGuide_Preliminary_Platform.pdf
the SPI_MASTER register is
Bits:
29-31 rs_slave_select - select the slave SPI device
28 clk_mode is clock divisor is odd, then:
0 - CLK is low longer
1 - CLK is high longer
16-27 rs_clk_sel (N)
SPI clock frequencys is hclk/(N+2)
11-15 cs_dsel_cnt
"De-select time of SPI chip select is configured to occupy
the number of cycles of AHB clock"
10 full-duplex (doens't actually work)
9 Interrupt enable
(I tried using interrupt instead of busy-wait, and flash
reading is much slower)
8 spi_start_sel - Interval between spi_cs_n and spi_sclk (3 or 6)
7 spi_prefetch
6 bidir_mode (doesn't work)
5 cpha (clock phase??)
4 cpol (clock polatity)
3 lsb_first
2 more_buf_mode (4+32 bytes, rather than 4 bytes of buffer)
0-1 serial_mode
0 standard serial
1 dual serial
3 quad serial
4 reserved.
So the mt7621_spi_reset:
selects slave 7, enables more_buf_mode, and disabled full_duplex.
Nothing else jumps out.
If you would take on a final clean up and submission, I would really
appreciate it.
Thanks,
NeilBrown
I'd appreciate a add + remove series as per gregkh's comment, it is real
hard to check if Chuanhong Guo's recent comments/fixes for the broken
"silicon" modes were folded into the code.
John
