🚀 go-pugleaf

Now that the curious seem to have turned the discussion to ioctl() calls, I  
would like to take the opportunity to re-post a question of mine about directly  
accessing the Intel 82077 Floppy Disk Controller chip.

For compatibility with a 10-sector per track 3.5" floppy format, I need to send  
a READ command directly to the controller chip with DMA enabled in order to  
read a sector written with an address tag of sector 0.  Reading multiple  
sectors with one call would be nice as well.

I have learned a great deal from the /NextDeveloper/Examples/Floppy code and  
have successfully been able to read sectors 1 through 9 with a higher level  
function {namely ioctl(fd, FDIOCRRW, &rawio)}, but apparently I cannot read  
sector 0 without going to the lowest level because FDIOCRRW is block based and  
assumes that each track starts with sector 1 when it determines the cylinder.

Here is what I have coded so far.  do_ioc() is taken from fdform.c with  
improved error reporting.  I tried sectors 0 and 1, combined with tracks 0 and  
1, for transfers of 1, 9 and 10 sectors without any luck.

| #include <sys/file.h>
| #include <sys/ioctl.h>
| #include <nextdev/fd_extern.h>
| #include <nextdev/disk.h>
| #include <nextdev/dma.h>
| 
| 	bzero(&ioreq, sizeof(struct fd_ioreq));
| 	ioreq.density = FD_DENS_1;
| 	ioreq.timeout = 5000;
| 	ioreq.command = FDCMD_CMD_XFR;
| 	pcmd->mfm = 1;
| 	pcmd->opcode = (read_flag ? FCCMD_READ : FCCMD_WRITE);
| 	pcmd->hds = head;
| 	pcmd->drive_sel = 0;
| 	pcmd->cylinder = cylinder;
| 	pcmd->head = head;
| 	pcmd->sector = sector;
| 	pcmd->sector_size = pfi->sectsize_info.n;
| 	pcmd->eot = EOT;
| 	pcmd->gap_length= pfi->sectsize_info.rw_gap_length;
| 	pcmd->dtl = 0xff;
| 	ioreq.num_cmd_bytes = SIZEOF_RW_CMD;
| 	ioreq.addrs = (caddr_t)addrs;
| 	ioreq.byte_count = DMA_ENDALIGN(int, sector_count * SECT_SIZE);
| 	ioreq.num_stat_bytes = SIZEOF_RW_STAT;
| 	ioreq.flags = (read_flag ? FD_IOF_DMA_RD : FD_IOF_DMA_WR);
| 	rtn = do_ioc(fd, &ioreq);

And here are the results interpreted by my error printing routines (if you can  
answer my question, you might be able to understand these based upon knowledge  
of the Intel controller):

| Accessing floppy disk volume 0
| ...Expected stat byte count = 0x7
|    received stat byte count = 0x0
| ...Unexpected status: a
| FDR_BADPHASE:   controller changed phase unexpectedly
| ...Read/Write Failed
| Status Register 0:
| Head Address 0
| Drive Select 0

Has anyone successully used the ioctl(fd, FDIOCREQ, fdiop) call to read and  
write a floppy?  Even code which uses FDIOCREQ to access a standard 9-sector  
disk would be useful.

Is NeXT listening?  You may be the only resource that can solve this...
-- 
Brian Willoughby	Software Design Engineer, BSEE
BrianW@SoundS.WA.com	SoundSoftware
NeXTmail welcome