fix MMC response types
PATCH 1/1 fix MMC response types
Previously the MMC/SD protocol layer didn't provide command and response types to mmc_omap_start_command(). In 2.6.16 it does and the hack that tries to find the correct command/response types based on flags is no longer needed.
This matters particularly for response types because for command types this hack worked fine, while it failed to give correct response type for some SD commands such as CMD3 (in most cases this caused no big problems because the size of the response matched). Regards,
Andrzej
--
balrog 2oo6
Dear Outlook users: Please remove me from your address books
http://www.newsforge.com/article.pl?sid=03/08/21/143258
drivers/mmc/omap.c +27 -28
@@ -185,39 +185,38 @@ mmc_omap_start_command(struct mmc_omap_h host->hw_bus_mode = host->bus_mode; } -if (!(cmd->flags &MMC_RSP_PRESENT )) - resptype =0 ;/* Resp 0 */ - -if (cmd->flags &MMC_RSP_136 ) - resptype =2 ;/* Resp 2 */ -else { -if (host->bus_mode ==MMC_BUSMODE_OPENDRAIN ) - resptype =3 ;/* Resp 3 */ -else - resptype =1 ;/* Resp 1, Resp 1b */ +switch (mmc_resp_type (cmd)) { +case MMC_RSP_NONE : +/* Response type 0 */ +break ; +case MMC_RSP_R1 : +case MMC_RSP_R1B : + resptype =1 ; +break ; +case MMC_RSP_R2 : + resptype =2 ; +break ; +case MMC_RSP_R3 : + resptype =3 ; +break ; +case MMC_RSP_R6 : + resptype =6 ; +break ; } -/* Protocol layer does not provide command type, but our hardware -* needs it! -* any data transfer means adtc type (but that information is not -* in command structure, so we flagged it into host struct.) -* However, telling bc, bcr and ac apart based on response is -* not foolproof: -* CMD0 = bc = resp0 CMD15 = ac = resp0 -* CMD2 = bcr = resp2 CMD10 = ac = resp2 -* -* Resolve to best guess with some exception testing: -* resp0 -> bc, except CMD15 = ac -* rest are ac, except if opendrain -*/ -if (host->data) { +switch (cmd->flags &MMC_CMD_MASK ) { +case MMC_CMD_AC : + cmdtype =OMAP_MMC_CMDTYPE_AC ; +break ; +case MMC_CMD_ADTC : cmdtype =OMAP_MMC_CMDTYPE_ADTC ; - }else if (resptype ==0 && cmd->opcode !=15 ) { +break ; +case MMC_CMD_BC : cmdtype =OMAP_MMC_CMDTYPE_BC ; - }else if (host->bus_mode ==MMC_BUSMODE_OPENDRAIN ) { +break ; +case MMC_CMD_BCR : cmdtype =OMAP_MMC_CMDTYPE_BCR ; - }else { - cmdtype =OMAP_MMC_CMDTYPE_AC ; +break ; } cmdreg = cmd->opcode | (resptype <<8 ) | (cmdtype <<12 );
Quoted 17 lines · ext andrzej zaborowski
Previously the MMC/SD protocol layer didn't provide command and response types to
mmc_omap_start_command(). In 2.6.16 it does and the hack that tries to find the correct command/response types based on flags is no longer needed.
This matters particularly for response types because for command types this hack worked fine, while it failed to give correct response type for some SD commands such as CMD3 (in most cases this caused no big problems because the size of the response matched). Regards,
Andrzej
--
balrog 2oo6Dear Outlook users: Please remove me from your address books
http://www.newsforge.com/article.pl?sid=03/08/21/143258
Quoted 8 lines
------------------------------------------------------------------------
_______________________________________________
Linux-omap-open-source mailing list
[email protected]
http://linux.omap.com/mailman/listinfo/linux-omap-open-source
Hi Andrzej and folks,
Attached to this mail follow my propose of fixing MMC responde types in mmc_omap_start_command().
This is based on the patches of the thread '[RFC] mmc: add OMAP driver' that I sent to LKML mailing-list in order to have integrated the MMC OMAP driver on the mainline kernel.
Comments and suggestions are welcome.
BR,
Carlos.
Signature · 7 lines
Carlos Eduardo Software Engineer Nokia Institute of Technology - INdT Embedded Linux Laboratory - 10LE Phone: +55 92 2126-1079 Mobile: +55 92 8127-1797 E-mail: [email protected]
drivers/mmc/omap.c +18 -26
@@ -49,6 +49,7 @@#include "omap.h" #define DRIVER_NAME "mmci-omap" +#define RSP_TYPE (x) ((x) & ~(MMC_RSP_BUSY|MMC_RSP_OPCODE))#ifdef CONFIG_MMC_DEBUG #define DBG(x...) pr_debug(x)
@@ -185,36 +186,27 @@ mmc_omap_start_command(struct mmc_omap_h host->hw_bus_mode = host->bus_mode; } -if (!(cmd->flags &MMC_RSP_PRESENT )) - resptype =0 ;/* Resp 0 */ - -if (cmd->flags &MMC_RSP_136 ) - resptype =2 ;/* Resp 2 */ -else { -if (host->bus_mode ==MMC_BUSMODE_OPENDRAIN ) - resptype =3 ;/* Resp 3 */ -else - resptype =1 ;/* Resp 1, Resp 1b */ +/* Our hardware needs to know exact type */ +switch (RSP_TYPE (mmc_resp_type (cmd))) { +case RSP_TYPE (MMC_RSP_R1 ): +/* resp 1, resp 1b */ + resptype =1 ; +break ; +case RSP_TYPE (MMC_RSP_R2 ): + resptype =2 ; +break ; +case RSP_TYPE (MMC_RSP_R3 ): + resptype =3 ; +break ; +default : +break ; } -/* Protocol layer does not provide command type, but our hardware -* needs it! -* any data transfer means adtc type (but that information is not -* in command structure, so we flagged it into host struct.) -* However, telling bc, bcr and ac apart based on response is -* not foolproof: -* CMD0 = bc = resp0 CMD15 = ac = resp0 -* CMD2 = bcr = resp2 CMD10 = ac = resp2 -* -* Resolve to best guess with some exception testing: -* resp0 -> bc, except CMD15 = ac -* rest are ac, except if opendrain -*/ -if (host->data) { +if (mmc_cmd_type (cmd) ==MMC_CMD_ADTC ) { cmdtype =OMAP_MMC_CMDTYPE_ADTC ; - }else if (resptype ==0 && cmd->opcode !=15 ) { + }else if (mmc_cmd_type (cmd) ==MMC_CMD_BC ) { cmdtype =OMAP_MMC_CMDTYPE_BC ; - }else if (host->bus_mode ==MMC_BUSMODE_OPENDRAIN ) { + }else if (mmc_cmd_type (cmd) ==MMC_CMD_BCR ) { cmdtype =OMAP_MMC_CMDTYPE_BCR ; }else { cmdtype =OMAP_MMC_CMDTYPE_AC ;
Quoted 29 lines · Carlos Aguiar
Quoted 17 lines · ext andrzej zaborowski
Previously the MMC/SD protocol layer didn't provide command and response types to
mmc_omap_start_command(). In 2.6.16 it does and the hack that tries to find the correct command/response types based on flags is no longer needed.
This matters particularly for response types because for command types this hack worked fine, while it failed to give correct response type for some SD commands such as CMD3 (in most cases this caused no big problems because the size of the response matched). Regards,
Andrzej
--
balrog 2oo6Dear Outlook users: Please remove me from your address books
http://www.newsforge.com/article.pl?sid=03/08/21/143258Quoted 8 lines
------------------------------------------------------------------------
_______________________________________________
Linux-omap-open-source mailing list
[email protected]
http://linux.omap.com/mailman/listinfo/linux-omap-open-sourceHi Andrzej and folks,
Quoted 8 lines
Attached to this mail follow my propose of fixing MMC responde types in
mmc_omap_start_command().This is based on the patches of the thread '[RFC] mmc: add OMAP driver' that I sent to LKML mailing-list in order to have integrated the MMC OMAP driver on the mainline kernel.
Comments and suggestions are welcome.
It looks fine except you forgot about the response type 6. With the current version of mmc_omap_start_command() the hack assumes response type 1 in place of 6 which merely causes the MMC/SD host controller to misinterpret the status flags in the response from the card, but the size of the response matches. However with your patch it would assume no response at all which would most probably cause breakage.
Quoted 5 lines
BR,
Carlos.
Quoted 12 lines
--
Carlos Eduardo
Software Engineer
Nokia Institute of Technology - INdT
Embedded Linux Laboratory - 10LE
Phone: +55 92 2126-1079
Mobile: +55 92 8127-1797
E-mail: [email protected]
Regards,
Andrew
--
balrog 2oo6
Dear Outlook users: Please remove me from your address books
http://www.newsforge.com/article.pl?sid=03/08/21/143258
Quoted 74 lines · ext andrzej zaborowski
Quoted 33 lines · Carlos Aguiar
Quoted 17 lines · ext andrzej zaborowski
Previously the MMC/SD protocol layer didn't provide command and response types to
mmc_omap_start_command(). In 2.6.16 it does and the hack that tries to find the correct command/response types based on flags is no longer needed.
This matters particularly for response types because for command types this hack worked fine, while it failed to give correct response type for some SD commands such as CMD3 (in most cases this caused no big problems because the size of the response matched). Regards,
Andrzej
--
balrog 2oo6Dear Outlook users: Please remove me from your address books
http://www.newsforge.com/article.pl?sid=03/08/21/143258Quoted 10 lines
------------------------------------------------------------------------
_______________________________________________
Linux-omap-open-source mailing list
[email protected]
http://linux.omap.com/mailman/listinfo/linux-omap-open-sourceHi Andrzej and folks,
Quoted 10 lines
Attached to this mail follow my propose of fixing MMC responde types in
mmc_omap_start_command().This is based on the patches of the thread '[RFC] mmc: add OMAP driver' that I sent to LKML mailing-list in order to have integrated the MMC OMAP driver on the mainline kernel.
Comments and suggestions are welcome.
It looks fine except you forgot about the response type 6. With the current version of
mmc_omap_start_command()the hack assumes response type 1 in place of 6 which merely causes the MMC/SD host controller to misinterpret the status flags in the response from the card, but the size of the response matches. However with your patch it would assume no response at all which would most probably cause breakage.Quoted 4 lines
BR,
Carlos.
Quoted 14 lines
--
Carlos Eduardo
Software Engineer
Nokia Institute of Technology - INdT
Embedded Linux Laboratory - 10LE
Phone: +55 92 2126-1079
Mobile: +55 92 8127-1797
E-mail: [email protected]
Quoted 9 lines
Regards,
Andrew
--
balrog 2oo6Dear Outlook users: Please remove me from your address books
http://www.newsforge.com/article.pl?sid=03/08/21/143258
Hi Andrew and folks,
The response type 6 was included. Thanks for your observation.
BR,
Carlos.
Signature · 7 lines
Carlos Eduardo Software Engineer Nokia Institute of Technology - INdT Embedded Linux Laboratory - 10LE Phone: +55 92 2126-1079 Mobile: +55 92 8127-1797 E-mail: [email protected]
drivers/mmc/omap.c +21 -26
@@ -49,6 +49,7 @@#include "omap.h" #define DRIVER_NAME "mmci-omap" +#define RSP_TYPE (x) ((x) & ~(MMC_RSP_BUSY|MMC_RSP_OPCODE))#ifdef CONFIG_MMC_DEBUG #define DBG(x...) pr_debug(x)
@@ -185,36 +186,30 @@ mmc_omap_start_command(struct mmc_omap_h host->hw_bus_mode = host->bus_mode; } -if (!(cmd->flags &MMC_RSP_PRESENT )) - resptype =0 ;/* Resp 0 */ - -if (cmd->flags &MMC_RSP_136 ) - resptype =2 ;/* Resp 2 */ -else { -if (host->bus_mode ==MMC_BUSMODE_OPENDRAIN ) - resptype =3 ;/* Resp 3 */ -else - resptype =1 ;/* Resp 1, Resp 1b */ +/* Our hardware needs to know exact type */ +switch (RSP_TYPE (mmc_resp_type (cmd))) { +case RSP_TYPE (MMC_RSP_R1 ): +/* resp 1, resp 1b */ + resptype =1 ; +break ; +case RSP_TYPE (MMC_RSP_R2 ): + resptype =2 ; +break ; +case RSP_TYPE (MMC_RSP_R3 ): + resptype =3 ; +break ; +case RSP_TYPE (MMC_RSP_R6 ): + resptype =6 ; +break ; +default : +break ; } -/* Protocol layer does not provide command type, but our hardware -* needs it! -* any data transfer means adtc type (but that information is not -* in command structure, so we flagged it into host struct.) -* However, telling bc, bcr and ac apart based on response is -* not foolproof: -* CMD0 = bc = resp0 CMD15 = ac = resp0 -* CMD2 = bcr = resp2 CMD10 = ac = resp2 -* -* Resolve to best guess with some exception testing: -* resp0 -> bc, except CMD15 = ac -* rest are ac, except if opendrain -*/ -if (host->data) { +if (mmc_cmd_type (cmd) ==MMC_CMD_ADTC ) { cmdtype =OMAP_MMC_CMDTYPE_ADTC ; - }else if (resptype ==0 && cmd->opcode !=15 ) { + }else if (mmc_cmd_type (cmd) ==MMC_CMD_BC ) { cmdtype =OMAP_MMC_CMDTYPE_BC ; - }else if (host->bus_mode ==MMC_BUSMODE_OPENDRAIN ) { + }else if (mmc_cmd_type (cmd) ==MMC_CMD_BCR ) { cmdtype =OMAP_MMC_CMDTYPE_BCR ; }else { cmdtype =OMAP_MMC_CMDTYPE_AC ;
So there are planned to do somekind of MMC/SD code merge with the mainline and has it already happened in the current driver version in linux-omap.git?
Just to confuse a little bit more... I have a small patch I received from another ipaq h6300 (omap1510 based) developer. As I do not know anything from the mmc/sd layer, all I can say is that with this patch me and the Deniz (patch author) were able to boot from the SD card to iPAQ h6300. (I tested with 512 mb kingston)
Without this patch we were only able to boot from the MMC cards. Deniz (said that he is not 100 % sure whether all of the changes are actually needed
or would only a subset be enought.
Hopefully this patch is not needed anymore after your updates, but here it is anyway just in case that somebody finds it usefull.
Mika
mmc.c +1 -1
@@ -964,7 +964,7 @@mmc_decode_scr (card); } -mmc_deselect_cards (host); +//mmc_deselect_cards(host); }static unsigned int mmc_calculate_clock (struct mmc_host *host)
omap.c +9 -6
@@ -2,7 +2,7 @@* linux/drivers/media/mmc/omap.c * * Copyright (C) 2004 Nokia Corporation -* Written by Tuukka Tikkanen and Juha Yrj�l� <[email protected]> +* Written by Tuukka Tikkanen and Juha Yrjölä <[email protected]> * Misc hacks here and there by Tony Lindgren <[email protected]> * Other hacks (DMA, SD, etc) by David Brownell *
@@ -231,6 +231,9 @@if (host->data && !(host->data->flags &MMC_DATA_WRITE )) cmdreg |=1 <<15 ; +if (cmd->opcode==0 ) + cmdreg |=1 <<7 ; +clk_enable (host->fclk);OMAP_MMC_WRITE (host->base,CTO ,200 );
@@ -890,7 +893,7 @@ reg &= ~(1 <<5 );OMAP_MMC_WRITE (host->base,SDIO , reg);/* Set maximum timeout */ -OMAP_MMC_WRITE (host->base,CTO ,0xff ); +OMAP_MMC_WRITE (host->base,CTO ,0xfd ); }static inline void set_data_timeout (struct mmc_omap_host *host,struct mmc_request *req)
@@ -901,7 +904,7 @@/* Convert ns to clock cycles by assuming 20MHz frequency * 1 cycle at 20MHz = 500 ns */ - timeout = req->data->timeout_clks + req->data->timeout_ns /500 ; + timeout = (req->data->timeout_clks *10 ) + (req->data->timeout_ns*10 ) /500 ;/* Some cards require more time to do at least the first read operation */ timeout = timeout <<4 ;
@@ -986,7 +989,7 @@mmc_omap_sg_to_buf (host); host->dma_in_use =0 ; } -mod_timer (&host->xfer_timer, jiffies +msecs_to_jiffies (500 )); +mod_timer (&host->xfer_timer, jiffies +msecs_to_jiffies (1500 )); pr_debug("MMC%d: %s %s %s, DTO %d cycles + %d ns, " "%d blocks of %d bytes, %d segments\n",
@@ -1164,7 +1167,7 @@case MMC_POWER_UP :case MMC_POWER_ON :mmc_omap_power (host,1 ); - dsor |=1 <<11 ; + dsor = (120 | (1 <<11 ));break ; }
@@ -1484,4 +1487,4 @@MODULE_DESCRIPTION ("OMAP Multimedia Card driver" );MODULE_LICENSE ("GPL" );MODULE_ALIAS (DRIVER_NAME ); -MODULE_AUTHOR ("Juha Yrj�l�" ); +MODULE_AUTHOR ("Juha Yrjölä" );
Quoted 6 lines
Hopefully this patch is not needed anymore after your updates, but here it is anyway just in case that somebody finds it usefull.
Btw, I tested this more and found out that the only thing really needed to get h6300
to mount SD in addition of MMC card during the boot was this one. (That's on top of the latest omap-git version)
mmc.c +1 -1
@@ -964,7 +964,7 @@mmc_decode_scr (card); } -mmc_deselect_cards (host); +//mmc_deselect_cards(host); }