On Mon, 30 May 2011 14:08:09 +0200 "Wagner, Helge (GE Intelligent Platforms)" Helge.Wagner@ge.com wrote:
Hi,
i have tried to reprogram a Macronix MX25L6445E flash chip. This worked, but gave me an error message:
ERASE FAILED at 0x0001107f! Expected=0xff, Read=0x00, failed byte count from 0x00010000-0x0001ffff: 0x9c15 ERASE FAILED!
Looking deeper into the datasheet, i saw that the MX25L6445E has the same ID as the MX25L6405, but with (at least) one difference: The MX25L6405 has a block size of 64KByte for erase function 20h, while the MX25L6445E is using 4KByte sector size for the same erase function 20h. As a result the spi_block_erase_20() will fail for the MX25L6445E (while the spi_block_erase_d8() which is used then will succeed).
I have now re-ordered the erase functions for the MX25L6405/MX25L6445E: spi_block_erase_d8 - first spi_block_erase_60 - second spi_block_erase_c7 - third spi_block_erase_20 - last So if one of the d8/60/c7 erase functions are successfull, the spi_block_erase_20 will never be called.
To be sure that the spi_block_erase_20 works for both the MX25L6405 and the MX25L6445E, we could change the block erase size from 64K to 4K. This should work even for the MX25L6405, but with the side effect of the erase taking longer than needed.
Any comments?
Please find my patches included. Signed-off-by: Helge Wagner Helge.Wagner@ge.com
hello helge
thanks for your patch! i have not looked at the datasheet(s), but maybe there is a way to distinguish the two devices (with another id command for example)? if we change the order of eraser functions as your patch suggests we should comment the reason precisely in the code. else we may introduce the previous behavior sometime later e.g. when we refine partial writes. i personally don't see a problem apart from that, but i am not sure if it is safe either and wont break pending patches. so others have to ack it.
Hello Stefan,
i have not looked at the datasheet(s), but maybe there is a way to distinguish the two devices (with another id command for example)?
I have asked Macronix, and they suggest to use the REMS2 function, which is supported on the MX25L6445E (returns 0xC216) and is not supported on the MX25L6405 (according to Macronix, it returns 0x0000 or 0xFFFF for this command).
So now there is no longer a need to re-order the erase functions for the MX25L6405.
Here is my new patch.
Signed-off-by: Helge Wagner helge.wagner@ge.com
Regards, Helge Wagner
diff -u flashrom/chipdrivers.h flashrom-patched/chipdrivers.h --- flashrom/chipdrivers.h 2011-06-27 15:21:57.000000000 +0200 +++ flashrom-patched/chipdrivers.h 2011-06-27 16:29:25.000000000 +0200 @@ -31,6 +31,7 @@ int probe_spi_rems(struct flashchip *flash); int probe_spi_res1(struct flashchip *flash); int probe_spi_res2(struct flashchip *flash); +int probe_spi_mx_2017(struct flashchip *flash); int spi_write_enable(void); int spi_write_disable(void); int spi_block_erase_20(struct flashchip *flash, unsigned int addr, unsigned int blocklen); diff -u flashrom/flashchips.c flashrom-patched/flashchips.c --- flashrom/flashchips.c 2011-06-27 15:21:57.000000000 +0200 +++ flashrom-patched/flashchips.c 2011-06-27 16:26:34.000000000 +0200 @@ -4131,8 +4131,8 @@ .total_size = 8192, .page_size = 256, .feature_bits = FEATURE_WRSR_WREN, - .tested = TEST_OK_PROBE, - .probe = probe_spi_rdid, + .tested = TEST_UNTESTED, + .probe = probe_spi_mx_2017, .probe_timing = TIMING_ZERO, .block_erasers = { @@ -4141,6 +4141,40 @@ .block_erase = spi_block_erase_20, }, { .eraseblocks = { {64 * 1024, 128} }, + .block_erase = spi_block_erase_d8, + }, { + .eraseblocks = { {8 * 1024 * 1024, 1} }, + .block_erase = spi_block_erase_60, + }, { + .eraseblocks = { {8 * 1024 * 1024, 1} }, + .block_erase = spi_block_erase_c7, + } + }, + .unlock = spi_disable_blockprotect, + .write = spi_chip_write_256, + .read = spi_chip_read, + .voltage = {2700, 3600}, + }, + + { + .vendor = "Macronix", + .name = "MX25L6445E", + .bustype = CHIP_BUSTYPE_SPI, + .manufacture_id = MACRONIX_ID, + .model_id = MACRONIX_MX25L6445E, + .total_size = 8192, + .page_size = 256, + .feature_bits = FEATURE_WRSR_WREN, + .tested = TEST_OK_PREW, + .probe = probe_spi_mx_2017, + .probe_timing = TIMING_ZERO, + .block_erasers = + { + { + .eraseblocks = { {4 * 1024, 2048} }, + .block_erase = spi_block_erase_20, + }, { + .eraseblocks = { {64 * 1024, 128} }, .block_erase = spi_block_erase_d8, }, { .eraseblocks = { {8 * 1024 * 1024, 1} }, diff -u flashrom/flashchips.h flashrom-patched/flashchips.h --- flashrom/flashchips.h 2011-06-27 15:21:57.000000000 +0200 +++ flashrom-patched/flashchips.h 2011-06-27 15:46:22.000000000 +0200 @@ -356,7 +356,8 @@ #define MACRONIX_MX25L8005 0x2014 /* Same as MX25V8005 */ #define MACRONIX_MX25L1605 0x2015 /* MX25L1605{,A,D} */ #define MACRONIX_MX25L3205 0x2016 /* MX25L3205{,A} */ -#define MACRONIX_MX25L6405 0x2017 /* MX25L3205{,D} */ +#define MACRONIX_MX25L6405 0x2017 /* MX25L6405{,D} */ +#define MACRONIX_MX25L6445E 0x16 /* MX25L6445E, ID = REMS2 answer */ #define MACRONIX_MX25L12805 0x2018 /* MX25L12805 */ #define MACRONIX_MX25L1635D 0x2415 #define MACRONIX_MX25L1635E 0x2515 /* MX25L1635{E} */ diff -u flashrom/spi.h flashrom-patched/spi.h --- flashrom/spi.h 2011-06-27 15:21:57.000000000 +0200 +++ flashrom-patched/spi.h 2011-06-27 16:29:38.000000000 +0200 @@ -40,6 +40,11 @@ #define JEDEC_REMS_OUTSIZE 0x04 #define JEDEC_REMS_INSIZE 0x02
+/* Read Electronic Manufacturer Signature 2 */ +#define JEDEC_REMS2 0xEF +#define JEDEC_REMS2_OUTSIZE 0x04 +#define JEDEC_REMS2_INSIZE 0x02 + /* Read Electronic Signature */ #define JEDEC_RES 0xab #define JEDEC_RES_OUTSIZE 0x04 diff -u flashrom/spi25.c flashrom-patched/spi25.c --- flashrom/spi25.c 2011-06-27 15:21:57.000000000 +0200 +++ flashrom-patched/spi25.c 2011-06-27 16:30:33.000000000 +0200 @@ -66,6 +66,27 @@ return 0; }
+static int spi_rems2(unsigned char *readarr) +{ + unsigned char cmd[JEDEC_REMS2_OUTSIZE] = { JEDEC_REMS2, 0, 0, 0 }; + uint32_t readaddr; + int ret; + + ret = spi_send_command(sizeof(cmd), JEDEC_REMS2_INSIZE, cmd, readarr); + if (ret == SPI_INVALID_ADDRESS) { + /* Find the lowest even address allowed for reads. */ + readaddr = (spi_get_valid_read_addr() + 1) & ~1; + cmd[1] = (readaddr >> 16) & 0xff, + cmd[2] = (readaddr >> 8) & 0xff, + cmd[3] = (readaddr >> 0) & 0xff, + ret = spi_send_command(sizeof(cmd), JEDEC_REMS2_INSIZE, cmd, readarr); + } + if (ret) + return ret; + msg_cspew("REMS2 returned %02x %02x. ", readarr[0], readarr[1]); + return 0; +} + static int spi_res(unsigned char *readarr, int bytes) { unsigned char cmd[JEDEC_RES_OUTSIZE] = { JEDEC_RES, 0, 0, 0 }; @@ -173,6 +194,49 @@ return probe_spi_rdid_generic(flash, 3); }
+int probe_spi_mx_2017(struct flashchip *flash) +{ + unsigned char readarr[4]; + uint32_t id1; + uint32_t id2; + + if (spi_rdid(readarr, 3)) { + msg_cdbg("\n"); + return 0; + } + id1 = readarr[0]; + id2 = (readarr[1] << 8) | readarr[2]; + + if (id1 == MACRONIX_ID && id2 == MACRONIX_MX25L6405) { + /* We have either MACRONIX_MX25L6405 or MACRONIX_MX25L6445E. + * MACRONIX_MX25L6405 has 64KB for erase function 20h and does not support REMS2 function, + * MACRONIX_MX25L6445E has 4KB for erase function 20h and does support REMS2 function. + * So to differentiate both, we execute REMS2 function (this is suggested by macronix). + */ + if (spi_rems2(readarr)) { + msg_cdbg("\n"); + return 0; + } + id1 = readarr[0]; + id2 = readarr[1]; + if (id1 == MACRONIX_ID && id2 == MACRONIX_MX25L6445E) { + /* MACRONIX_MX25L6445E */ + if (flash->model_id == MACRONIX_MX25L6445E) + return 1; + else + return 0; + } else { + /* MACRONIX_MX25L6405 */ + if (flash->model_id == MACRONIX_MX25L6405) + return 1; + else + return 0; + } + } + /* neither MACRONIX_MX25L6405 nor MACRONIX_MX25L6445E */ + return 0; +} + int probe_spi_rdid4(struct flashchip *flash) { /* Some SPI controllers do not support commands with writecnt=1 and
-----Original Message----- From: Stefan Tauner [mailto:stefan.tauner@student.tuwien.ac.at] Sent: Freitag, 17. Juni 2011 15:49 To: Wagner, Helge (GE Intelligent Platforms) Cc: flashrom@flashrom.org; Carl-Daniel Hailfinger Subject: Re: [flashrom] [PATCH] Macronix MX25L6445E
On Mon, 30 May 2011 14:08:09 +0200 "Wagner, Helge (GE Intelligent Platforms)" Helge.Wagner@ge.com wrote:
Hi,
i have tried to reprogram a Macronix MX25L6445E flash chip. This worked, but gave me an error message:
ERASE FAILED at 0x0001107f! Expected=0xff, Read=0x00, failed byte count from 0x00010000-0x0001ffff: 0x9c15 ERASE FAILED!
Looking deeper into the datasheet, i saw that the MX25L6445E has the same ID as the MX25L6405, but with (at least) one difference: The MX25L6405 has a block size of 64KByte for erase function 20h, while the MX25L6445E is using 4KByte sector size for the same erase function 20h. As a result the spi_block_erase_20() will fail for the MX25L6445E (while the spi_block_erase_d8() which is used then will succeed).
I have now re-ordered the erase functions for the MX25L6405/MX25L6445E: spi_block_erase_d8 - first spi_block_erase_60 - second spi_block_erase_c7 - third spi_block_erase_20 - last So if one of the d8/60/c7 erase functions are successfull, the spi_block_erase_20 will never be called.
To be sure that the spi_block_erase_20 works for both the MX25L6405 and the MX25L6445E, we could change the block erase size from 64K to 4K. This should work even for the MX25L6405, but with the side effect of the erase taking longer than needed.
Any comments?
Please find my patches included. Signed-off-by: Helge Wagner Helge.Wagner@ge.com
hello helge
thanks for your patch! i have not looked at the datasheet(s), but maybe there is a way to distinguish the two devices (with another id command for example)? if we change the order of eraser functions as your patch suggests we should comment the reason precisely in the code. else we may introduce the previous behavior sometime later e.g. when we refine partial writes. i personally don't see a problem apart from that, but i am not sure if it is safe either and wont break pending patches. so others have to ack it. -- Kind regards/Mit freundlichen Grüßen, Stefan Tauner
-
Stefan Tauner -
Wagner, Helge (GE Intelligent Platforms)