drivers: spi_flash: Add Fast Read Dual I/O support

The Fast Read Dual Output and Fast Read Dual I/O commands are
practically identical, the only difference being how the read address is
transferred (saving a whooping 2 bytes which is totally irrelevant for
the amounts of data coreboot tends to read). We originally implemented
Fast Read Dual Output since it's the older command and some older
Winbond chips only supported that one... but it seems that some older
Macronix parts for whatever reason chose to only support Fast Read Dual
I/O instead. So in order to make this work for as many parts as
possible, I guess we'll have to implement both. (Also, the Macronix
device ID situation is utter madness with different chips with different
capabilities often having the same ID, so we basically have to make a
best-effort guess to strike a trade-off between fast speeds and best
chance at supporting all chips. If this turns out to be a problem later,
we may have to add Kconfig overrides for this or resort to SFDP parsing,
although that would defeat the whole point of trying to be fast.)

BUG=b:193486682
TEST=Booted CoachZ (with Dual I/O)

Signed-off-by: Julius Werner <jwerner@chromium.org>
Change-Id: Ia1a20581f251615127f132eadea367b7b66c4709
Reviewed-on: https://review.coreboot.org/c/coreboot/+/56287
Tested-by: build bot (Jenkins) <no-reply@coreboot.org>
Reviewed-by: Furquan Shaikh <furquan@google.com>

6 files changed, 100 insertions, 17 deletions

diff --git a/src/drivers/spi/gigadevice.c b/src/drivers/spi/gigadevice.c
index 15ad9078cf..fc2a405302 100644
--- a/src/drivers/spi/gigadevice.c
+++ b/src/drivers/spi/gigadevice.c
@@ -31,72 +31,84 @@ static const struct spi_flash_part_id flash_table[] = {
 		.id[0]				= 0x4014,
 		.nr_sectors_shift		= 8,
 		.fast_read_dual_output_support	= 1,
+		.fast_read_dual_io_support	= 1,
 	},					/* also GD25Q80B */
 	{
 		/* GD25Q16 */
 		.id[0]				= 0x4015,
 		.nr_sectors_shift		= 9,
 		.fast_read_dual_output_support	= 1,
+		.fast_read_dual_io_support	= 1,
 	},					/* also GD25Q16B */
 	{
 		/* GD25Q32B */
 		.id[0]				= 0x4016,
 		.nr_sectors_shift		= 10,
 		.fast_read_dual_output_support	= 1,
+		.fast_read_dual_io_support	= 1,
 	},					/* also GD25Q32B */
 	{
 		/* GD25Q64 */
 		.id[0]				= 0x4017,
 		.nr_sectors_shift		= 11,
 		.fast_read_dual_output_support	= 1,
+		.fast_read_dual_io_support	= 1,
 	},					/* also GD25Q64B, GD25B64C */
 	{
 		/* GD25Q128 */
 		.id[0]				= 0x4018,
 		.nr_sectors_shift		= 12,
 		.fast_read_dual_output_support	= 1,
+		.fast_read_dual_io_support	= 1,
 	},					/* also GD25Q128B */
 	{
 		/* GD25VQ80C */
 		.id[0]				= 0x4214,
 		.nr_sectors_shift		= 8,
 		.fast_read_dual_output_support	= 1,
+		.fast_read_dual_io_support	= 1,
 	},
 	{
 		/* GD25VQ16C */
 		.id[0]				= 0x4215,
 		.nr_sectors_shift		= 9,
 		.fast_read_dual_output_support	= 1,
+		.fast_read_dual_io_support	= 1,
 	},
 	{
 		/* GD25LQ80 */
 		.id[0]				= 0x6014,
 		.nr_sectors_shift		= 8,
 		.fast_read_dual_output_support	= 1,
+		.fast_read_dual_io_support	= 1,
 	},
 	{
 		/* GD25LQ16 */
 		.id[0]				= 0x6015,
 		.nr_sectors_shift		= 9,
 		.fast_read_dual_output_support	= 1,
+		.fast_read_dual_io_support	= 1,
 	},
 	{
 		/* GD25LQ32 */
 		.id[0]				= 0x6016,
 		.nr_sectors_shift		= 10,
 		.fast_read_dual_output_support	= 1,
+		.fast_read_dual_io_support	= 1,
 	},
 	{
 		/* GD25LQ64C */
 		.id[0]				= 0x6017,
 		.nr_sectors_shift		= 11,
 		.fast_read_dual_output_support	= 1,
+		.fast_read_dual_io_support	= 1,
 	},					/* also GD25LB64C */
 	{
 		/* GD25LQ128 */
 		.id[0]				= 0x6018,
 		.nr_sectors_shift		= 12,
 		.fast_read_dual_output_support	= 1,
+		.fast_read_dual_io_support	= 1,
 	},
 };
 
diff --git a/src/drivers/spi/macronix.c b/src/drivers/spi/macronix.c
index 69e7a9f4c7..b6981e045d 100644
--- a/src/drivers/spi/macronix.c
+++ b/src/drivers/spi/macronix.c
@@ -63,55 +63,75 @@ static const struct spi_flash_part_id flash_table[] = {
 		.id[0] = 0x2415,
 		.nr_sectors_shift = 9,
 	},
+	/*
+	 * NOTE: C225xx JEDEC IDs are basically useless because Macronix keeps
+	 * reusing the same IDs for vastly different chips. 35E versions always
+	 * seem to support Dual I/O but not Dual Output, while 35F versions seem
+	 * to support both, so we only set Dual I/O here to improve our chances
+	 * of compatibility. Since Macronix makes it impossible to search all
+	 * different parts that it recklessly assigned the same IDs to, it's
+	 * hard to know if there may be parts that don't even support Dual I/O
+	 * with these IDs, though (or what we should do if there are).
+	 */
 	{
 		/* MX25L1635E */
 		.id[0] = 0x2515,
 		.nr_sectors_shift = 9,
+		.fast_read_dual_io_support = 1,
 	},
 	{
 		/* MX25U8032E */
 		.id[0] = 0x2534,
 		.nr_sectors_shift = 8,
+		.fast_read_dual_io_support = 1,
 	},
 	{
-		/* MX25U1635E */
+		/* MX25U1635E/MX25U1635F */
 		.id[0] = 0x2535,
 		.nr_sectors_shift = 9,
+		.fast_read_dual_io_support = 1,
 	},
 	{
-		/* MX25U3235E */
+		/* MX25U3235E/MX25U3235F */
 		.id[0] = 0x2536,
 		.nr_sectors_shift = 10,
+		.fast_read_dual_io_support = 1,
 	},
 	{
-		/* MX25U6435F */
+		/* MX25U6435E/MX25U6435F */
 		.id[0] = 0x2537,
 		.nr_sectors_shift = 11,
+		.fast_read_dual_io_support = 1,
 	},
 	{
 		/* MX25U12835F */
 		.id[0] = 0x2538,
 		.nr_sectors_shift = 12,
+		.fast_read_dual_io_support = 1,
 	},
 	{
 		/* MX25U25635F */
 		.id[0] = 0x2539,
 		.nr_sectors_shift = 13,
+		.fast_read_dual_io_support = 1,
 	},
 	{
-		/* MX25U51245G */
+		/* MX25U51235F */
 		.id[0] = 0x253a,
 		.nr_sectors_shift = 14,
+		.fast_read_dual_io_support = 1,
 	},
 	{
 		/* MX25L12855E */
 		.id[0] = 0x2618,
 		.nr_sectors_shift = 12,
+		.fast_read_dual_io_support = 1,
 	},
 	{
 		/* MX25L3235D/MX25L3225D/MX25L3236D/MX25L3237D */
 		.id[0] = 0x5e16,
 		.nr_sectors_shift = 10,
+		.fast_read_dual_io_support = 1,
 	},
 	{
 		/* MX25L6495F */
diff --git a/src/drivers/spi/spi_flash.c b/src/drivers/spi/spi_flash.c
index b674749feb..f3cecd5fc6 100644
--- a/src/drivers/spi/spi_flash.c
+++ b/src/drivers/spi/spi_flash.c
@@ -20,7 +20,7 @@ static void spi_flash_addr(u32 addr, u8 *cmd)
 	cmd[3] = addr >> 0;
 }
 
-static int do_spi_flash_cmd(const struct spi_slave *spi, const void *dout,
+static int do_spi_flash_cmd(const struct spi_slave *spi, const u8 *dout,
 			    size_t bytes_out, void *din, size_t bytes_in)
 {
 	int ret;
@@ -51,8 +51,8 @@ static int do_spi_flash_cmd(const struct spi_slave *spi, const void *dout,
 	return ret;
 }
 
-static int do_dual_read_cmd(const struct spi_slave *spi, const void *dout,
-			    size_t bytes_out, void *din, size_t bytes_in)
+static int do_dual_output_cmd(const struct spi_slave *spi, const u8 *dout,
+			      size_t bytes_out, void *din, size_t bytes_in)
 {
 	int ret;
 
@@ -79,6 +79,31 @@ static int do_dual_read_cmd(const struct spi_slave *spi, const void *dout,
 	return ret;
 }
 
+static int do_dual_io_cmd(const struct spi_slave *spi, const u8 *dout,
+			  size_t bytes_out, void *din, size_t bytes_in)
+{
+	int ret;
+
+	/* Only the very first byte (opcode) is transferred in "single" mode. */
+	struct spi_op vector = { .dout = dout, .bytesout = 1,
+				 .din = NULL, .bytesin = 0 };
+
+	ret = spi_claim_bus(spi);
+	if (ret)
+		return ret;
+
+	ret = spi_xfer_vector(spi, &vector, 1);
+
+	if (!ret)
+		ret = spi->ctrlr->xfer_dual(spi, &dout[1], bytes_out - 1, NULL, 0);
+
+	if (!ret)
+		ret = spi->ctrlr->xfer_dual(spi, NULL, 0, din, bytes_in);
+
+	spi_release_bus(spi);
+	return ret;
+}
+
 int spi_flash_cmd(const struct spi_slave *spi, u8 cmd, void *response, size_t len)
 {
 	int ret = do_spi_flash_cmd(spi, &cmd, sizeof(cmd), response, len);
@@ -119,18 +144,23 @@ int spi_flash_cmd_read(const struct spi_flash *flash, u32 offset,
 {
 	u8 cmd[5];
 	int ret, cmd_len;
-	int (*do_cmd)(const struct spi_slave *spi, const void *din,
+	int (*do_cmd)(const struct spi_slave *spi, const u8 *din,
 		      size_t in_bytes, void *out, size_t out_bytes);
 
 	if (CONFIG(SPI_FLASH_NO_FAST_READ)) {
 		cmd_len = 4;
 		cmd[0] = CMD_READ_ARRAY_SLOW;
 		do_cmd = do_spi_flash_cmd;
-	} else if (flash->flags.dual_spi && flash->spi.ctrlr->xfer_dual) {
+	} else if (flash->flags.dual_io && flash->spi.ctrlr->xfer_dual) {
+		cmd_len = 5;
+		cmd[0] = CMD_READ_FAST_DUAL_IO;
+		cmd[4] = 0;
+		do_cmd = do_dual_io_cmd;
+	} else if (flash->flags.dual_output && flash->spi.ctrlr->xfer_dual) {
 		cmd_len = 5;
 		cmd[0] = CMD_READ_FAST_DUAL_OUTPUT;
 		cmd[4] = 0;
-		do_cmd = do_dual_read_cmd;
+		do_cmd = do_dual_output_cmd;
 	} else {
 		cmd_len = 5;
 		cmd[0] = CMD_READ_ARRAY_FAST;
@@ -352,7 +382,8 @@ static int fill_spi_flash(const struct spi_slave *spi, struct spi_flash *flash,
 	flash->pp_cmd = vi->desc->pp_cmd;
 	flash->wren_cmd = vi->desc->wren_cmd;
 
-	flash->flags.dual_spi = part->fast_read_dual_output_support;
+	flash->flags.dual_output = part->fast_read_dual_output_support;
+	flash->flags.dual_io = part->fast_read_dual_io_support;
 
 	flash->ops = &vi->desc->ops;
 	flash->prot_ops = vi->prot_ops;
@@ -471,8 +502,10 @@ int spi_flash_probe(unsigned int bus, unsigned int cs, struct spi_flash *flash)
 	}
 
 	const char *mode_string = "";
-	if (flash->flags.dual_spi && spi.ctrlr->xfer_dual)
-		mode_string = " (Dual SPI mode)";
+	if (flash->flags.dual_io && spi.ctrlr->xfer_dual)
+		mode_string = " (Dual I/O mode)";
+	else if (flash->flags.dual_output && spi.ctrlr->xfer_dual)
+		mode_string = " (Dual Output mode)";
 	printk(BIOS_INFO,
 	       "SF: Detected %02x %04x with sector size 0x%x, total 0x%x%s\n",
 		flash->vendor, flash->model, flash->sector_size, flash->size, mode_string);
diff --git a/src/drivers/spi/spi_flash_internal.h b/src/drivers/spi/spi_flash_internal.h
index b4d39b3d31..4a7beeab3e 100644
--- a/src/drivers/spi/spi_flash_internal.h
+++ b/src/drivers/spi/spi_flash_internal.h
@@ -15,6 +15,7 @@
 #define CMD_READ_ARRAY_LEGACY		0xe8
 
 #define CMD_READ_FAST_DUAL_OUTPUT	0x3b
+#define CMD_READ_FAST_DUAL_IO		0xbb
 
 #define CMD_READ_STATUS			0x05
 #define CMD_WRITE_ENABLE		0x06
@@ -69,8 +70,9 @@ struct spi_flash_part_id {
 	uint16_t id[2];
 	/* Log based 2 total number of sectors. */
 	uint16_t nr_sectors_shift: 4;
-	uint16_t fast_read_dual_output_support : 1;
-	uint16_t _reserved_for_flags: 3;
+	uint16_t fast_read_dual_output_support : 1;	/*  1-1-2 read */
+	uint16_t fast_read_dual_io_support : 1;		/*  1-2-2 read */
+	uint16_t _reserved_for_flags: 2;
 	/* Block protection. Currently used by Winbond. */
 	uint16_t protection_granularity_shift : 5;
 	uint16_t bp_bits : 3;
diff --git a/src/drivers/spi/winbond.c b/src/drivers/spi/winbond.c
index 9fe0492b75..8f93e3be92 100644
--- a/src/drivers/spi/winbond.c
+++ b/src/drivers/spi/winbond.c
@@ -101,12 +101,14 @@ static const struct spi_flash_part_id flash_table[] = {
 		.id[0]				= 0x4014,
 		.nr_sectors_shift		= 8,
 		.fast_read_dual_output_support	= 1,
+		.fast_read_dual_io_support	= 1,
 	},
 	{
 		/* W25Q16_V */
 		.id[0]				= 0x4015,
 		.nr_sectors_shift		= 9,
 		.fast_read_dual_output_support	= 1,
+		.fast_read_dual_io_support	= 1,
 		.protection_granularity_shift	= 16,
 		.bp_bits			= 3,
 	},
@@ -115,6 +117,7 @@ static const struct spi_flash_part_id flash_table[] = {
 		.id[0]				= 0x6015,
 		.nr_sectors_shift		= 9,
 		.fast_read_dual_output_support	= 1,
+		.fast_read_dual_io_support	= 1,
 		.protection_granularity_shift	= 16,
 		.bp_bits			= 3,
 	},
@@ -123,6 +126,7 @@ static const struct spi_flash_part_id flash_table[] = {
 		.id[0]				= 0x4016,
 		.nr_sectors_shift		= 10,
 		.fast_read_dual_output_support	= 1,
+		.fast_read_dual_io_support	= 1,
 		.protection_granularity_shift	= 16,
 		.bp_bits			= 3,
 	},
@@ -131,6 +135,7 @@ static const struct spi_flash_part_id flash_table[] = {
 		.id[0]				= 0x6016,
 		.nr_sectors_shift		= 10,
 		.fast_read_dual_output_support	= 1,
+		.fast_read_dual_io_support	= 1,
 		.protection_granularity_shift	= 16,
 		.bp_bits			= 3,
 	},
@@ -139,6 +144,7 @@ static const struct spi_flash_part_id flash_table[] = {
 		.id[0]				= 0x4017,
 		.nr_sectors_shift		= 11,
 		.fast_read_dual_output_support	= 1,
+		.fast_read_dual_io_support	= 1,
 		.protection_granularity_shift	= 17,
 		.bp_bits			= 3,
 	},
@@ -147,6 +153,7 @@ static const struct spi_flash_part_id flash_table[] = {
 		.id[0]				= 0x6017,
 		.nr_sectors_shift		= 11,
 		.fast_read_dual_output_support	= 1,
+		.fast_read_dual_io_support	= 1,
 		.protection_granularity_shift	= 17,
 		.bp_bits			= 3,
 	},
@@ -155,6 +162,7 @@ static const struct spi_flash_part_id flash_table[] = {
 		.id[0]				= 0x8017,
 		.nr_sectors_shift		= 11,
 		.fast_read_dual_output_support	= 1,
+		.fast_read_dual_io_support	= 1,
 		.protection_granularity_shift	= 17,
 		.bp_bits			= 3,
 	},
@@ -163,6 +171,7 @@ static const struct spi_flash_part_id flash_table[] = {
 		.id[0]				= 0x4018,
 		.nr_sectors_shift		= 12,
 		.fast_read_dual_output_support	= 1,
+		.fast_read_dual_io_support	= 1,
 		.protection_granularity_shift	= 18,
 		.bp_bits			= 3,
 	},
@@ -171,6 +180,7 @@ static const struct spi_flash_part_id flash_table[] = {
 		.id[0]				= 0x6018,
 		.nr_sectors_shift		= 12,
 		.fast_read_dual_output_support	= 1,
+		.fast_read_dual_io_support	= 1,
 		.protection_granularity_shift	= 18,
 		.bp_bits			= 3,
 	},
@@ -179,6 +189,7 @@ static const struct spi_flash_part_id flash_table[] = {
 		.id[0]				= 0x7018,
 		.nr_sectors_shift		= 12,
 		.fast_read_dual_output_support	= 1,
+		.fast_read_dual_io_support	= 1,
 		.protection_granularity_shift	= 18,
 		.bp_bits			= 3,
 	},
@@ -187,6 +198,7 @@ static const struct spi_flash_part_id flash_table[] = {
 		.id[0]				= 0x8018,
 		.nr_sectors_shift		= 12,
 		.fast_read_dual_output_support	= 1,
+		.fast_read_dual_io_support	= 1,
 		.protection_granularity_shift	= 18,
 		.bp_bits			= 3,
 	},
@@ -195,6 +207,7 @@ static const struct spi_flash_part_id flash_table[] = {
 		.id[0]				= 0x8020,
 		.nr_sectors_shift		= 11,
 		.fast_read_dual_output_support	= 1,
+		.fast_read_dual_io_support	= 1,
 		.protection_granularity_shift	= 16,
 		.bp_bits			= 4,
 	},
@@ -203,6 +216,7 @@ static const struct spi_flash_part_id flash_table[] = {
 		.id[0]				= 0x4019,
 		.nr_sectors_shift		= 13,
 		.fast_read_dual_output_support	= 1,
+		.fast_read_dual_io_support	= 1,
 		.protection_granularity_shift	= 16,
 		.bp_bits			= 4,
 	},
@@ -211,6 +225,7 @@ static const struct spi_flash_part_id flash_table[] = {
 		.id[0]				= 0x7019,
 		.nr_sectors_shift		= 13,
 		.fast_read_dual_output_support	= 1,
+		.fast_read_dual_io_support	= 1,
 		.protection_granularity_shift	= 16,
 		.bp_bits			= 4,
 	},
diff --git a/src/include/spi_flash.h b/src/include/spi_flash.h
index a7f707eb1c..fe77e3ef79 100644
--- a/src/include/spi_flash.h
+++ b/src/include/spi_flash.h
@@ -86,8 +86,9 @@ struct spi_flash {
 	union {
 		u8 raw;
 		struct {
-			u8 dual_spi	: 1;
-			u8 _reserved	: 7;
+			u8 dual_output	: 1;
+			u8 dual_io	: 1;
+			u8 _reserved	: 6;
 		};
 	} flags;
 	u16 model;