Detecting ATA drives while using QEMU always reports the primary slave as present

[sboydlns]

I am try to get drive detection working. I am using QEMU to run my OS.

Even though I have only set up 1 drive in QEMU my drive detection code always reports that the primary slave exists. I can't figure out what I am doing wrong. Here is my drive detection routine. I call this routine 4 times. Once for each combination of I/O port and master/slave. As I mentioned the primary slave is being reported as present and when I dump the fields from the IDENTIFY buffer it seems like it is reporting the same values as the master. i.e. the max LBA is the same for both drives. It reports 25 for both drives which is the size of my raw disk image in sectors.

Any help is appreciated.

Code: Select all

static int detect_disk(int channel, int master) {
static struct SAtaDisk *last_disk = NULL;
int i;
byte select, status, b1, b2;
uint16 bfr[256], w;
struct SAtaDisk *disk;

	select = (master ? 0xa0 : 0xb0);			// Set master / slave
	kprintf("%x\r\n", select);
	// Try to select the drive
	outb(select, channel + ATA_SELECT);
	// 400ns delay to allow channel time to select the drive
	for (i=0; i < 15; i++) {
		status = inb(channel + ATA_STATUS);
	}
	// The ATA specification says that we need to clear these ports to zero
	outw(0, channel + ATA_SECTOR_COUNT);
	outw(0, channel + ATA_LBA_LOW);
	outw(0, channel + ATA_LBA_MID);
	outw(0, channel + ATA_LBA_HIGH);
	// Send the IDENTIFY command
	outb(ATA_IDENTIFY, channel + ATA_COMMAND);
	// Check status. If zero, drive doesn't exist. Otherwise,
	// wait until BUSY clears;
	status = inb(channel + ATA_STATUS);
	if (status == 0) {
		return FALSE;
	}
	while (status & STATUS_BSY) {
		status = inb(channel + ATA_STATUS);
	}
	if (status & STATUS_ERR) {
		return FALSE;
	}
	// Check these ports for zero. If they are not zero the
	// drive is not an ATA drive.
	b1 = inb(channel + ATA_LBA_MID);
	b2 = inb(channel + ATA_LBA_HIGH);
	if ((b1 != 0) || (b2 != 0)) {
		return FALSE;
	}
	// Wait until DRQ or ERR is set
	status = inb(channel + ATA_STATUS);
	while (! (status & (STATUS_DRQ | STATUS_ERR))) {
		status = inb(channel + ATA_STATUS);
	}
	if (status & STATUS_ERR) {
		return FALSE;
	}

	// Read the IDENTIFY data
	for (i = 0; i < 256; i++) {
		bfr[i] = inw(channel + ATA_DATA);
	}
//	snap(&bfr[0], 256);

	// Create a disk control block from the IDENTIFY Data
	disk = kalloc(sizeof(struct SAtaDisk));
	if (! disk) {
		kputs("Out of memory in ata_init\r\n");
		panic();
	}
	if (! low_memory->first_ata) {
		low_memory->first_ata = disk;
	}
	if (last_disk) {
		last_disk->next = disk;
	}
	last_disk = disk;
	disk->channel = channel;
	disk->device = select;
	disk->dma = ATA_IDENT_DMA;
	disk->max_lba = bfr[ATA_IDENT_LBA28SECTORS] + (bfr[ATA_IDENT_LBA28SECTORS + 1] << 16);
	if (bfr[ATA_IDENT_SUPPORTLBA48] & 0x400) {
		disk->flags |= HD_LBA48;
	}
	if (disk->flags & HD_LBA48) {
		disk->max_lba = (uint64)bfr[ATA_IDENT_LBA28SECTORS] +
				        ((uint64)bfr[ATA_IDENT_LBA28SECTORS + 1] << 16) +
				        ((uint64)bfr[ATA_IDENT_LBA28SECTORS + 2] << 32) +
				        ((uint64)bfr[ATA_IDENT_LBA28SECTORS + 3] << 48);
	}
	for (i = 0; i < 40; i +=2) {
		w = bfr[ATA_IDENT_MODEL + (i / 2)];
		disk->model[i] = (w >> 8) & 0xff;
		disk->model[i+1] = w & 0xff;
	}
	bfr[40] = 0;

	return TRUE;
}

[techdude17]

Please edit your post to use the proper code markings, just press the code button while editing.

I've noticed a similar issue to what you're having, and while this is probably not an orthodox solution you can try what I did.

Here is my ATA code for sending ATA_CMD_IDENTIFY:

Code: Select all

int err = 0; // debugging
    int timeout = 0;
    while (1 && timeout < 10000) {
        // ATAPI devices are supposed to set ERR rather than BSY, and then DRQ
        // This is also needed to get CYL_LO and CYL_HI to be reads

        // NOTE: Some ATAPI devices do not set ERR (for some reason). Therefore we do the port check anyways to determine whether the device is ATAPI 
        uint8_t status = ide_read(device, ATA_REG_STATUS);
        if (status & ATA_SR_ERR) {
            // ERR was set
            err = 1;
            break; 
        } else if (!(status & ATA_SR_BSY) && (status & ATA_SR_DRQ)) {
            // DRQ was set and BSY is cleared, likely an ATA device instead
            break;
        }

        timeout++; // Keep waiting...
    }

    // Did it timeout?
    if (timeout >= 10000) {
        LOG_DEVICE(INFO, device, "Timeout while waiting for ATA_CMD_IDENTIFY - assuming dead\n");
        return;
    }

Adding a timeout should fix your code. QEMU seems to send the right bytes regardless of if there's actually a drive there, but if the command times out and BSY still isn't clear I'd abandon ship.

I'm not an expert on this, so it might be helpful to check what Bochs does to confirm your code. Hope this helps!

[Klakap]

After you select drive, you should read status port. If it equals 0xFF, you can be sure that there is no drive and you should not send IDENTIFY command at all. After you send IDENTIFY command, you need to wait for BSY to be clear and DRQ to be set, not only for BSY to be clear as you do in your code.

Also, I would stongly recommend to not use while() cycles without timeout. You should not assume anything about anything in OS development. If for whatever reason your environment will not change bits, your OS will be locked. It is worth few additional lines of code to have protection against such situation.

[sboydlns]

Please edit your post to use the proper code markings, just press the code button while editing.

I clicked the quote button instead of code. My bad, sorry.

[sboydlns]

After you select drive, you should read status port. If it equals 0xFF, you can be sure that there is no drive and you should not send IDENTIFY command at all. After you send IDENTIFY command, you need to wait for BSY to be clear and DRQ to be set, not only for BSY to be clear as you do in your code.

Tried that and it didn't make any difference. What seems to have fixed the problem is adding a check for a status of zero after the drive select. Now when I do my drive selection I find the primary master. The primary slave fails with a status of zero. The secondary master fails with a status of 0x41 (RDY | ERR) and the secondary slave fails with a status of zero.

Also, I would stongly recommend to not use while() cycles without timeout. You should not assume anything about anything in OS development. If for whatever reason your environment will not change bits, your OS will be locked. It is worth few additional lines of code to have protection against such situation.

Good idea. Done.

Thank you