No microSD card support about c65gs

[gardners]

Have contacted xesscorp to ask about this problem, and if they are willing to relicense under LGPL.

[gardners]

All the machinery is now there, it just doesn't work. Reading a sector reports an error, but sd_error doesn't seem to get loaded.
Also there are some nasty warnings about CE/CS lines during synthesis.

[gardners]

The above commit (f89a736) fixes some doozy bugs that would certainly have prevented the SD card commands from being accepted. Hopefully it will get it working. There remains an issue where the sector buffer remains mapped, regardless of the value of sector_buffer_mapped. This may be related to the nasty synthesis warnings that need to be addressed.

[gardners]

Okay, we now can read the R1 code from the sd controller, but it is saying $05 = illegal command + idle. This may be because the card I am using is SD not SDHC. Changing mode and trying again.

[gardners]

Will switch from using Xess sdcard module to the apparently freely usable code by Steve Merrifield in an attempt to resolve all of the world of hurt and pain that I have experienced with ISE + Xess code + my code. It is hard to workout where the cause of the problem is. In any case, using the free code will avoid having to wait for Xess to LGPL the code before I can release the complete code or an FPGA binary.

[gardners]

With the above commits it is now possible to read from the SD card, but writing fails. Also need to confirm that CRC byte fix has worked, and that the whole 512 bytes get read in. Then need to debug why writing isn't working.

[gardners]

Okay. After a lot of mucking around, the sdcard controller is merged into master. It almost, but not quite works. Basically reading a sector reads some of it, but misses most of the bytes.

[gardners]

Missed byte was caused by SPI FSM in sdcontroller only running when no fastio accesses were in progress. As kernel and basic currently run from fastio, this bug showed up. The behaviour was added in to prevent dual-port ram complaints raised by xst when using the xess sd controller code. As that investigation showed that the sector buffer could in fact be dual port, we can change this behaviour. In fact, we can avoid dual-port by making the sector buffer only accessible when the busy flag is clear in the sdcontroller status register ($D680 bit 0 for now). This is what commit cb88917 above does. Now building to test if it fixes the problem.

[gardners]

Reading from SDcard now works. Writing is untested, but may work, which can be the subject of another issue.