mmc: enable TRIM/ERASE caps for SDHCI host
PATCH v4 2/3 mmc: enable TRIM/ERASE caps for SDHCI host
SDHCI host controller has TRIM/ERASE capability, enable these caps for erasing purpose.
ERASE command needs R1B response. So for such command with busy signal, before sending command, SDHCI host driver will simply set a maximum value for timeout control register which is safe to use.
--- drivers/mmc/host/sdhci.c | 16 +++++++++++----- 1 files changed, 11 insertions(+), 5 deletions(-)
drivers/mmc/host/sdhci.c +11 -5
@@ -41,7 +41,6 @@static unsigned int debug_quirks =0 ; -static void sdhci_prepare_data (struct sdhci_host *,struct mmc_data *);static void sdhci_finish_data (struct sdhci_host *);static void sdhci_send_command (struct sdhci_host *,struct mmc_command *);
@@ -652,16 +651,23 @@ static void sdhci_set_transfer_irqs(struct sdhci_host *host)sdhci_clear_set_irqs (host, dma_irqs, pio_irqs); } -static void sdhci_prepare_data (struct sdhci_host *host,struct mmc_data *data) +static void sdhci_prepare_data (struct sdhci_host *host,struct mmc_command *cmd) {u8 count;u8 ctrl; +struct mmc_data *data = cmd->data;int ret;WARN_ON (host->data); -if (data ==NULL ) +if (data ==NULL ) { +/* +* set timeout to be maximum value for command with busy signal. +*/ +if (cmd->flags &MMC_RSP_BUSY ) +sdhci_writeb (host,0xE ,SDHCI_TIMEOUT_CONTROL );return ; + }/* Sanity checks */ BUG_ON (data->blksz * data->blocks >524288 );
@@ -921,7 +927,7 @@ static void sdhci_send_command(struct sdhci_host *host, struct mmc_command *cmd) host->cmd = cmd; -sdhci_prepare_data (host, cmd->data); +sdhci_prepare_data (host, cmd);sdhci_writel (host, cmd->arg,SDHCI_ARGUMENT );
@@ -1939,7 +1945,7 @@ int sdhci_add_host(struct sdhci_host *host) mmc->f_min = host->max_clk /SDHCI_MAX_DIV_SPEC_200 ; mmc->f_max = host->max_clk; - mmc->caps |=MMC_CAP_SDIO_IRQ ; + mmc->caps |=MMC_CAP_SDIO_IRQ |MMC_CAP_ERASE ;/* * A controller may support 8-bit width, but the board itself
Signature · 1 line
1.6.6.1
Quoted 8 lines · On Saturday 12 February 2011 07:22:20 Chuanxiao Dong wrote:
SDHCI host controller has TRIM/ERASE capability, enable these caps for erasing purpose.
ERASE command needs R1B response. So for such command with busy signal, before sending command, SDHCI host driver will simply set a maximum value for timeout control register which is safe to use.
Signed-off-by Chuanxiao Dong <[email protected]>
Great stuff! I needed this for a project of mine, but couldn't get it to work reliably without knowing about the timeout logic.
I think, almost no mmc controllers advertice MMC_CAP_ERASE today, which is a real shame, given that it's extremely useful for performance.
Arnd
Excerpts from Chuanxiao Dong's message of Sat Feb 12 07:22:20 +0100 2011:
Quoted 8 lines
SDHCI host controller has TRIM/ERASE capability, enable these caps for erasing purpose.
ERASE command needs R1B response. So for such command with busy signal, before sending command, SDHCI host driver will simply set a maximum value for timeout control register which is safe to use.
Signed-off-by Chuanxiao Dong <[email protected]>
Tested on an OLPC XO-1 using a SanDisk 4GB class 4 card (retail). The BLKDISCARD ioctl (range 0..card size) returned immediately, no timeout issue encountered.
Sascha
Signature · 2 lines
http://sascha.silbe.org/ http://www.infra-silbe.de/
Quoted 7 lines · Chuanxiao Dong
SDHCI host controller has TRIM/ERASE capability, enable these caps for erasing purpose.
ERASE command needs R1B response. So for such command with busy signal, before sending command, SDHCI host driver will simply set a maximum value for timeout control register which is safe to use.
Awesome. I took a look at the eMMC spec, and the following commands also have R1B responses:
1) CMD5 (SLEEP_AWAKE)
2) CMD6 (SWITCH)
3) CMD7 (Select/Deselect - for Disconnected->Programming transitions)
4) CMD12 (STOP_TRANSMISSION - for write case)
5) CMD28 (SET_WRITE_PROT)
6) CMD29 (CLR_WRITE_PROT)
...so this should help with timeouts for those commands as well. Thanks!
A
Quoted 20 lines · Andrei Warkentin
Quoted 7 lines · Chuanxiao Dong
SDHCI host controller has TRIM/ERASE capability, enable these caps for erasing purpose.
ERASE command needs R1B response. So for such command with busy signal, before sending command, SDHCI host driver will simply set a maximum value for timeout control register which is safe to use.
Awesome. I took a look at the eMMC spec, and the following commands also have R1B responses:
1) CMD5 (
SLEEP_AWAKE)
2) CMD6 (SWITCH)
3) CMD7 (Select/Deselect - forDisconnected->Programmingtransitions)
4) CMD12 (STOP_TRANSMISSION- for write case)
5) CMD28 (SET_WRITE_PROT)
6) CMD29 (CLR_WRITE_PROT)
Quoted 4 lines
...so this should help with timeouts for those commands as well. Thanks!
A
On a second thought (and look), I see the following problems with this patch, which are otherwise going to make the MMC code confusing (and invalid).
1) Erase timeout value is defined by spec (eMMC or SD) and already calculated in mmc_set_erase_timeout within core/core.c. You should honor that value instead of picking the max timeout.
2) The trim/erase commands are not the only commands with variable timeouts. For example, certain CMD6 operations take different amounts of time (Partition switch takes ext_csd:PARTITION_SWITCH_TIME) Right now the command structure contains the field "erase_timeout", which curiously enough is used nowhere other than getting set in core/core.c. I propose renaming it to cmd_timeout. If cmd.data is NULL, then cmd_timeout is used to set timeout for busy-type commands.
A
PATCH 1/1 MMC: enable TRIM/ERASE caps for SDHCI host.
SDHCI host controller has TRIM/ERASE capability, enable these caps for erasing purpose.
ERASE command needs R1B response, so fix R1B-type command handling for SDHCI controller. For non-DAT commands using a busy reponse, the cmd->cmd_timeout (in ms) field is used for timeout calculations. cmd->cmd_timeout field is appropriately set to the correct erase timeout in core/core.c.
Based on patch by Chuanxiao Dong <[email protected]>
--- drivers/mmc/core/core.c | 40 ++++++++++++++++++++++------------------ drivers/mmc/host/sdhci.c | 43 +++++++++++++++++++++++++++---------------- include/linux/mmc/core.h | 2 +- 3 files changed, 50 insertions(+), 35 deletions(-)
drivers/mmc/core/core.c +22 -18
@@ -1187,9 +1187,9 @@ void mmc_init_erase(struct mmc_card *card) } } -static void mmc_set_mmc_erase_timeout (struct mmc_card *card, -struct mmc_command *cmd, -unsigned int arg,unsigned int qty) +static unsigned int mmc_mmc_erase_timeout (struct mmc_card *card, +struct mmc_command *cmd, +unsigned int arg,unsigned int qty) {unsigned int erase_timeout;
@@ -1246,38 +1246,42 @@ static void mmc_set_mmc_erase_timeout(struct mmc_card *card,if (mmc_host_is_spi (card->host) && erase_timeout <1000 ) erase_timeout =1000 ; - cmd->erase_timeout = erase_timeout; +return erase_timeout; } -static void mmc_set_sd_erase_timeout (struct mmc_card *card, -struct mmc_command *cmd,unsigned int arg, -unsigned int qty) +static unsigned int mmc_sd_erase_timeout (struct mmc_card *card, +struct mmc_command *cmd,unsigned int arg, +unsigned int qty) { +unsigned int erase_timeout; +if (card->ssr.erase_timeout) {/* Erase timeout specified in SD Status Register (SSR) */ - cmd->erase_timeout = card->ssr.erase_timeout * qty + - card->ssr.erase_offset; + erase_timeout = card->ssr.erase_timeout * qty + + card->ssr.erase_offset; }else {/* * Erase timeout not specified in SD Status Register (SSR) so * use 250ms per write block. */ - cmd->erase_timeout =250 * qty; + erase_timeout =250 * qty; }/* Must not be less than 1 second */ -if (cmd->erase_timeout <1000 ) - cmd->erase_timeout =1000 ; +if (erase_timeout <1000 ) + erase_timeout =1000 ; + +return erase_timeout; } -static void mmc_set_erase_timeout (struct mmc_card *card, -struct mmc_command *cmd,unsigned int arg, -unsigned int qty) +static unsigned int mmc_erase_timeout (struct mmc_card *card, +struct mmc_command *cmd,unsigned int arg, +unsigned int qty) {if (mmc_card_sd (card)) -mmc_set_sd_erase_timeout (card, cmd, arg, qty); +return mmc_sd_erase_timeout (card, cmd, arg, qty);else -mmc_set_mmc_erase_timeout (card, cmd, arg, qty); +return mmc_mmc_erase_timeout (card, cmd, arg, qty); } static int mmc_do_erase(struct mmc_card *card, unsigned int from,
@@ -1351,7 +1355,7 @@ static int mmc_do_erase(struct mmc_card *card, unsigned int from, cmd.opcode =MMC_ERASE ; cmd.arg = arg; cmd.flags =MMC_RSP_SPI_R1B |MMC_RSP_R1B |MMC_CMD_AC ; -mmc_set_erase_timeout (card, &cmd, arg, qty); + cmd.cmd_timeout =mmc_erase_timeout (card, &cmd, arg, qty); err =mmc_wait_for_cmd (card->host, &cmd,0 );if (err) { printk(KERN_ERR "mmc_erase: erase error %d, status %#x\n",
drivers/mmc/host/sdhci.c +27 -16
107 lines not on this page · raw message
include/linux/mmc/core.h +1 -1
@@ -92,7 +92,7 @@ struct mmc_command {* actively failing requests */ -unsigned int erase_timeout;/* in milliseconds */ +unsigned int cmd_timeout;/* in milliseconds */ struct mmc_data *data;/* data segment associated with cmd */ struct mmc_request *mrq;/* associated request */
Signature · 1 line
1.7.0.4
Re: [PATCH] MMC: enable TRIM/ERASE caps for SDHCI host.
Quoted 11 lines · Andrei Warkentin
SDHCI host controller has TRIM/ERASE capability, enable these caps for erasing purpose.
ERASE command needs R1B response, so fix R1B-type command handling for SDHCI controller. For non-DAT commands using a busy reponse, the
cmd->cmd_timeout(in ms) field is used for timeout calculations.cmd->cmd_timeoutfield is appropriately set to the correct erase timeout incore/core.c.Based on patch by Chuanxiao Dong <[email protected]>
Signed-off-by Andrei Warkentin <[email protected]>
+Chuanxiao
Tested on Ricoh SDHCI controller (x64) with Arnd's erase tool (from flashbench) for -
- Toshiba eMMC MMC08G (8Gb device) (with and without a changed version my partitioning patch)
- SDHC SA08G (8Gb uSD in adapter)
Lack of proper R1B handling seems like a pervasive problem for other host controller drivers as well. I can start making patches, but I won't be able to test them since all I've got is SDHCI :(.
A