mmc: Add support for SDHC cards
PATCH 1/1 mmc: Add support for SDHC cards
Hi all,
Thanks to the generous donation of an SDHC card by John Gilmore, and the surprisingly enlightened decision by the SD Card Association to publish useful specs, I've been able to bash out support for SDHC. The changes are not too profound:
i) Add a card flag indicating the card uses block level addressing and check it in the block driver. As we never took advantage of byte-level addressing, this simply involves skipping the block -> byte translation when sending commands.
ii) The layout of the CSD is changed - a set of fields are discarded to make space for a larger C_SIZE. We did not reference any of the discarded fields except those related to the C_SIZE.
iii) Read and write timeouts are fixed values and not calculated from CSD values.
iv) Before invoking SEND_APP_OP_COND, we must invoke the new SEND_IF_COND to inform the card we support SDHC.
I've done some basic read and write tests and everything seems to work fine but one should obviously use caution in case it eats your data.
Addendum: Similar spec changes were introduced by the MMC Association in version 4.2 to switch to block addressing. We do not have access to the 4.2 spec or even a change summary as we did for 3 -> 4. My guess as to what's changed looks something like the following:
i) A new field is added to the ext_csd to store the capacity that cannot be described using existing csd fields.
ii) A new field is added to the ext_csd to indicate the need for block addressing.
iii) A new write-only field is added to the ext_csd which has to be toggled to tell the card that the host supports block-addressing. But perhaps compatibility requires that a new command be used like for SDHC...
If someone has access to the spec and can shed some light on the subject in a legal way, please let us know. Of course, I don't think there are any MMC HC cards on the market yet, but one was announced last month.
--phil
---
drivers/mmc/mmc.c | 40 +++++++++++++++++++++++++++++++++++----- drivers/mmc/mmc_block.c | 4 +++- include/linux/mmc/card.h | 3 +++ include/linux/mmc/mmc.h | 1 + include/linux/mmc/protocol.h | 13 ++++++++++++- 5 files changed, 54 insertions(+), 7 deletions(-)
drivers/mmc/mmc_block.c +3 -1
@@ -237,7 +237,9 @@ brq.mrq.cmd = &brq.cmd; brq.mrq.data = &brq.data; - brq.cmd.arg = req->sector <<9 ; + brq.cmd.arg = req->sector; +if (!mmc_card_blockaddr (card)) + brq.cmd.arg <<=9 ; brq.cmd.flags =MMC_RSP_R1 |MMC_CMD_ADTC ; brq.data.blksz =1 << md->block_bits; brq.data.blocks = req->nr_sectors >> (md->block_bits -9 );
drivers/mmc/mmc.c +35 -5
@@ -289,7 +289,10 @@else limit_us =100000 ; -if (timeout_us > limit_us) { +/* +* SDHC cards always use these fixed values. +*/ +if (timeout_us > limit_us ||mmc_card_blockaddr (card)) { data->timeout_ns = limit_us *1000 ; data->timeout_clks =0 ; }
@@ -588,12 +591,15 @@if (mmc_card_sd (card)) { csd_struct =UNSTUFF_BITS (resp,126 ,2 ); -if (csd_struct !=0 ) { +if (csd_struct !=0 && csd_struct !=1 ) {printk ("%s: unrecognised CSD structure version %d\n" ,mmc_hostname (card->host), csd_struct);mmc_card_set_bad (card);return ; } +if (csd_struct ==1 ) { +mmc_card_set_blockaddr (card); + } m =UNSTUFF_BITS (resp,115 ,4 ); e =UNSTUFF_BITS (resp,112 ,3 );
@@ -605,9 +611,14 @@ csd->max_dtr = tran_exp[e] * tran_mant[m]; csd->cmdclass =UNSTUFF_BITS (resp,84 ,12 ); - e =UNSTUFF_BITS (resp,47 ,3 ); - m =UNSTUFF_BITS (resp,62 ,12 ); - csd->capacity = (1 + m) << (e +2 ); +if (csd_struct ==0 ) { + e =UNSTUFF_BITS (resp,47 ,3 ); + m =UNSTUFF_BITS (resp,62 ,12 ); + csd->capacity = (1 + m) << (e +2 ); + }else if (csd_struct ==1 ) { + m =UNSTUFF_BITS (resp,48 ,22 ); + csd->capacity = (1 + m) <<10 ; + } csd->read_blkbits =UNSTUFF_BITS (resp,80 ,4 ); csd->read_partial =UNSTUFF_BITS (resp,79 ,1 );
@@ -824,6 +835,25 @@ {struct mmc_command cmd;int i, err =0 ; +static const u8 test_pattern =0xAA ; + +/* +* To support SD 2.0 cards, we must always invoke SD_SEND_IF_COND +* before SD_APP_OP_COND. This command will harmlessly fail for +* SD 1.0 and MMC cards (fortunately including MMC 4 cards). +*/ + cmd.opcode =SD_SEND_IF_COND ; + cmd.arg = ((host->ocr_avail &0xFF8000 ) !=0 ) <<8 | test_pattern; + cmd.flags =MMC_RSP_R7 |MMC_CMD_BCR ; + + err =mmc_wait_for_cmd (host, &cmd,0 ); +if (err ==MMC_ERR_NONE ) { +if ((cmd.resp[0 ] &0xFF ) != test_pattern) { +return MMC_ERR_FAILED ; + }else if (ocr !=0 ) { + ocr |=1 <<30 ; + } + } cmd.opcode =SD_APP_OP_COND ; cmd.arg = ocr;
include/linux/mmc/card.h +3 -0
@@ -71,6 +71,7 @@#define MMC_STATE_SDCARD (1<<3)/* is an SD card */ #define MMC_STATE_READONLY (1<<4)/* card is read-only */ #define MMC_STATE_HIGHSPEED (1<<5)/* card is in high speed mode */ +#define MMC_STATE_BLOCKADDR (1<<6)/* card uses block-addressing */ u32 raw_cid[4 ];/* raw card CID */ u32 raw_csd[4 ];/* raw card CSD */ u32 raw_scr[2 ];/* raw card SCR */
@@ -87,6 +88,7 @@#define mmc_card_sd (c) ((c)->state & MMC_STATE_SDCARD)#define mmc_card_readonly (c) ((c)->state & MMC_STATE_READONLY)#define mmc_card_highspeed (c) ((c)->state & MMC_STATE_HIGHSPEED) +#define mmc_card_blockaddr (c) ((c)->state & MMC_STATE_BLOCKADDR)#define mmc_card_set_present (c) ((c)->state |= MMC_STATE_PRESENT)#define mmc_card_set_dead (c) ((c)->state |= MMC_STATE_DEAD)
@@ -94,6 +96,7 @@#define mmc_card_set_sd (c) ((c)->state |= MMC_STATE_SDCARD)#define mmc_card_set_readonly (c) ((c)->state |= MMC_STATE_READONLY)#define mmc_card_set_highspeed (c) ((c)->state |= MMC_STATE_HIGHSPEED) +#define mmc_card_set_blockaddr (c) ((c)->state |= MMC_STATE_BLOCKADDR)#define mmc_card_name (c) ((c)->cid.prod_name)#define mmc_card_id (c) ((c)->dev.bus_id)
include/linux/mmc/mmc.h +1 -0
@@ -43,6 +43,7 @@#define MMC_RSP_R2 (MMC_RSP_PRESENT|MMC_RSP_136|MMC_RSP_CRC)#define MMC_RSP_R3 (MMC_RSP_PRESENT)#define MMC_RSP_R6 (MMC_RSP_PRESENT|MMC_RSP_CRC) +#define MMC_RSP_R7 (MMC_RSP_PRESENT|MMC_RSP_CRC)#define mmc_resp_type (cmd) ((cmd)->flags & (MMC_RSP_PRESENT|MMC_RSP_136|MMC_RSP_CRC|MMC_RSP_BUSY|MMC_RSP_OPCODE))
include/linux/mmc/protocol.h +12 -1
@@ -79,9 +79,12 @@#define MMC_GEN_CMD 56/* adtc [0] RD/WR R1 */ /* SD commands type argument response */ -/* class 8 */ +/* class 0 */ /* This is basically the same command as for MMC with some quirks. */ #define SD_SEND_RELATIVE_ADDR 3/* bcr R6 */ +#define SD_SEND_IF_COND 8/* bcr [11:0] See below R7 */ + +/* class 10 */ #define SD_SWITCH 6/* adtc [31:0] See below R1 */ /* Application commands */
@@ -115,6 +118,14 @@*/ /* +* SD_SEND_IF_COND argument format: +* +* [31:12] Reserved (0) +* [11:8] Host Voltage Supply Flags +* [7:0] Check Pattern (0xAA) +*/ + +/* MMC status in R1 Type e : error bit
Quoted 6 lines · Philip Langdale
Hi all,
Thanks to the generous donation of an SDHC card by John Gilmore, and the surprisingly enlightened decision by the SD Card Association to publish useful specs, I've been able to bash out support for SDHC. The changes are not too profound:
So I sent that with the wrong From: address (damn you thunderbird!). Please reply to my personal address (this one) and not my work one. :-)
--phil
Quoted 4 lines · Philip Langdale
Hi all,
*snip*
Quoted 2 lines
Signed-off-by Philipl Langdale <[email protected]>
When you have a commit message larger than the patch, you know there is something wrong. ;)
Please skip the part about MMC at least.
Quoted 8 lines
drivers/mmc/mmc.c +1
-1@@ -588,12 +591,15 @@if (mmc_card_sd (card)) { csd_struct =UNSTUFF_BITS (resp,126 ,2 ); -if (csd_struct !=0 ) { +if (csd_struct !=0 && csd_struct !=1 ) { printk("%s: unrecognised CSD structure version %d\n",
drivers/mmc/mmc.c line 605
Quoted 10 lines
drivers/mmc/mmc.c +3
-0mmc_hostname(card->host), csd_struct);mmc_card_set_bad (card);return ; } +if (csd_struct ==1 ) { +mmc_card_set_blockaddr (card); + } m =UNSTUFF_BITS (resp,115 ,4 ); e =UNSTUFF_BITS (resp,112 ,3 );
Actually, the way the spec describes version 2.0 of the CSD leaves my stomach a bit upset. I think we will find cards that put bogus values in the fields, expecting most hosts to use the well defined values. So I'd prefer a switch statement here, where csd_struct == 1 results in hard coded values for many fields.
Quoted 17 lines
drivers/mmc/mmc.c +12
-0@@ -824,6 +835,25 @@ {struct mmc_command cmd;int i, err =0 ; +static const u8 test_pattern =0xAA ; + +/* +* To support SD 2.0 cards, we must always invoke SD_SEND_IF_COND +* before SD_APP_OP_COND. This command will harmlessly fail for +* SD 1.0 and MMC cards (fortunately including MMC 4 cards). +*/ + cmd.opcode =SD_SEND_IF_COND ; + cmd.arg = ((host->ocr_avail &0xFF8000 ) !=0 ) <<8 | test_pattern; + cmd.flags =MMC_RSP_R7 |MMC_CMD_BCR ; + + err =mmc_wait_for_cmd (host, &cmd,0 );
drivers/mmc/mmc.c line 859
Quoted 10 lines
drivers/mmc/mmc.c +7
-0+if (err ==MMC_ERR_NONE ) { +if ((cmd.resp[0 ] &0xFF ) != test_pattern) { +return MMC_ERR_FAILED ; + }else if (ocr !=0 ) { + ocr |=1 <<30 ; + } + } cmd.opcode =SD_APP_OP_COND ; cmd.arg = ocr;
Put this in mmc_setup() with the rest of the initialisation. Hiding in here just confuses things.
Also, please add a comment about why you manipulate the ocr.
include/linux/mmc/mmc.h line 49
Quoted 9 lines
include/linux/mmc/mmc.h +1
-0@@ -43,6 +43,7 @@#define MMC_RSP_R2 (MMC_RSP_PRESENT|MMC_RSP_136|MMC_RSP_CRC)#define MMC_RSP_R3 (MMC_RSP_PRESENT)#define MMC_RSP_R6 (MMC_RSP_PRESENT|MMC_RSP_CRC) +#define MMC_RSP_R7 (MMC_RSP_PRESENT|MMC_RSP_CRC)#define mmc_resp_type (cmd) ((cmd)->flags & (MMC_RSP_PRESENT|MMC_RSP_136|MMC_RSP_CRC|MMC_RSP_BUSY|MMC_RSP_OPCODE))
Wrong. R7 is defined as MMC_RSP_PRESENT|MMC_RSP_CRC|MMC_RSP_OPCODE.
(So is R6 btw, wonder why that is...)
Rgds
Signature · 5 lines
-- Pierre Ossman Linux kernel, MMC maintainer http://www.kernel.org PulseAudio, core developer http://pulseaudio.org rdesktop, core developer http://www.rdesktop.org
I will post the updated diff as a separate follow up.
Quoted 5 lines · Pierre Ossman
When you have a commit message larger than the patch, you know there is something wrong. ;)
Please skip the part about MMC at least.
Heh. I forget that you don't want to manually alter the email. Will do.
Quoted 6 lines
Actually, the way the spec describes version 2.0 of the CSD leaves my stomach a bit upset. I think we will find cards that put bogus values in the fields, expecting most hosts to use the well defined values. So I'd prefer a switch statement here, where
csd_struct== 1 results in hard coded values for many fields.
I'm less pessimistic than you about this, but I have no problems with hard-coding.
Quoted 5 lines
drivers/mmc/mmc.c line 845
Quoted 2 lines
drivers/mmc/mmc.c +1
-0+ cmd.opcode =SD_SEND_IF_COND ;Put this in
mmc_setup()with the rest of the initialisation. Hiding in here just confuses things.
Done. I thought we needed it before both SD_APP_SEND_OP_COND calls but it's only needed before the second one so I've moved it inline into mmc_setup.
Quoted 1 line
Also, please add a comment about why you manipulate the ocr.
Done.
Quoted 4 lines
Wrong. R7 is defined as MMC_RSP_PRESENT|MMC_RSP_CRC|MMC_RSP_OPCODE.
(So is R6 btw, wonder why that is...)
Ah - I knew they were both the same so I made R7 match R6. :-)
I've added OPCODE to both.
Thanks,
--phil
Hi!
Quoted 9 lines
Quoted 6 lines
Hi all,
Thanks to the generous donation of an SDHC card by John Gilmore, and the surprisingly enlightened decision by the SD Card Association to publish useful specs, I've been able to bash out support for SDHC. The changes are not too profound:
So I sent that with the wrong From: address (damn you thunderbird!). Please reply to my personal address (this one) and not my work one. :-)
Would you describe what SDHC is? I know SD flash cards, and IIRC SDIO cards exist, with functionality such as bluetooth...? But SDHC?
Signature · 1 line
Thanks for all the (sleeping) penguins.
Quoted 4 lines · Pavel Machek
Would you describe what SDHC is? I know SD flash cards, and IIRC SDIO cards exist, with functionality such as bluetooth...? But SDHC?
SDHC is short for "Secure Digital High Capacity". It's simply SD flash cards than conform to a new version of the protocol, a version where addressing is done on a sector (512 byte) basis instead of bytes.
Rgds
Signature · 5 lines
-- Pierre Ossman Linux kernel, MMC maintainer http://www.kernel.org PulseAudio, core developer http://pulseaudio.org rdesktop, core developer http://www.rdesktop.org
Quoted 3 lines · Philip Langdale
Heh. I forget that you don't want to manually alter the email. Will do.
I'm rather reluctant to change anything once someone has put a signed-off-by line to it. Different people have different limits as to what they regard a trivial modification.
Quoted 5 lines
Done. I thought we needed it before both
SD_APP_SEND_OP_CONDcalls but it's only needed before the second one so I've moved it inline intommc_setup.
The spec says we need it at both (even though it might not be so in practice). We should follow the spec first and foremost.
Rgds
Signature · 5 lines
-- Pierre Ossman Linux kernel, MMC maintainer http://www.kernel.org PulseAudio, core developer http://pulseaudio.org rdesktop, core developer http://www.rdesktop.org