ata: disk_ata_drive: SSD drive doesn't pass check for ata id validity

Hi!

Barebox refuses to deal with OCZ-VERTEX4 ssd sata drive, because:
ata1: ata id invalid

This error is triggered by the following piece of code :

static int ata_id_is_valid(const uint16_t *id)

{
	if ((id[ATA_ID_FIELD_VALID] & 1) == 0) {
		pr_debug("Drive's ID seems invalid\n");
		return -EINVAL;
	}

	return 0;
}

The "id" is a sector-sized block of bytes, received after issuing ATA_CMD_ID_ATA command.

If I remove this check, the drive works fine.

Can someone with better knowledge of ATA standard comment on this issue?

Quoted 20 lines · Peter Mamonov

Hi!

Barebox refuses to deal with OCZ-VERTEX4 ssd sata drive, because:
ata1: ata id invalid

This error is triggered by the following piece of code :

static int ata_id_is_valid(const uint16_t *id)

{
	if ((id[ATA_ID_FIELD_VALID] & 1) == 0) {
		pr_debug("Drive's ID seems invalid\n");
		return -EINVAL;
	}

	return 0;
}

The "id" is a sector-sized block of bytes, received after issuing ATA_CMD_ID_ATA command.

Quoted 4 lines

If I remove this check, the drive works fine.

Can someone with better knowledge of ATA standard comment on this issue?

Juergen, do you have an idea what this check is for? Neither U-Boot nor the Kernel seem to ever check for bit 0 in this field.

Quoted 21 lines · On Friday 30 October 2015 08:31:04 Sascha Hauer wrote:

Quoted 20 lines · Peter Mamonov

Hi!

Barebox refuses to deal with OCZ-VERTEX4 ssd sata drive, because:
ata1: ata id invalid

This error is triggered by the following piece of code :

static int ata_id_is_valid(const uint16_t *id)

{
	if ((id[ATA_ID_FIELD_VALID] & 1) == 0) {
		pr_debug("Drive's ID seems invalid\n");
		return -EINVAL;
	}

	return 0;
}

The "id" is a sector-sized block of bytes, received after issuing ATA_CMD_ID_ATA command.

Quoted 7 lines

Quoted 4 lines

If I remove this check, the drive works fine.

Can someone with better knowledge of ATA standard comment on this issue?

Juergen, do you have an idea what this check is for? Neither U-Boot nor the Kernel seem to ever check for bit 0 in this field.

Seems this bit was meaningful in earlier days: it marked the ID member in the structure returned by the disk as valid (=1) or invalid (=0). In the meantime it seems obsolete. The last occurrence I found is in the Linux kernel in file "drivers/block/mg_disk.c". All other code which checks this register content just check bits 1 and 2. I vote for removing this piece of code.

jbe