Hello,
I am using Broadcom northstar2 platform, which uses Arasan eMMC 5.0 controller and micron 4.5.1 eMMC card.
Switched to “rpmb” partition using “mmc dev 0 3” command.
I am trying to use “mmc rpmb counter” command and results is “Transfer data timeout”.
Further investigation, found, mmc_rpmb_get_counter () -à mmc_rpmb_request(), which expects response type of MMC_RSP_R1b.
As part of the rpmb protocol, MMC_CMD_SET_BLOCK_COUNT (23) is set to #1 and MMC_CMD_WRITE_MULTIPLE_BLOCK (25) is send. For multi block write command, data is rpmb data.
As the expected response is set to MMC_RSP_R1b, in sdhci_send_command(), data mask is or’ed with SDHCI_INT_DATA_END (mask |= SDHCI_INT_DATA_END).
After the sdhci_cmd_done() happens, data mask is written to SDHCI_INT_STATUS. I guess this is sending STOP, because of SDHCI_INT_DATA_END. As the actual rpmb data is transferred in sdhci_transfer_data (), which will check for rdy (SDHCI_INT_SPACE_AVAIL| SDHCI_INT_DATA_AVAIL). The SDHCI_INT_STATUS register read value will be always 0x0 and causing for “Transfer data timeout”
Would like to know whether response type for mmc_rpmb_request() can be changed to MMC_RSP_R1 (instead of MMC_RSP_R1b)
Or using SDHCI_QUIRK_BROKEN_R1B quirk in mmc_rpmb_request() like the following or any other alternative ways, please suggest:
--- a/drivers/mmc/rpmb.c
+++ b/drivers/mmc/rpmb.c
@@ -91,6 +91,7 @@ static int mmc_rpmb_request(struct mmc *mmc, const struct s_rpmb *s,
{
struct mmc_cmd cmd = {0};
struct mmc_data data;
+ struct sdhci_host *host = mmc->priv;
int ret;
ret = mmc_set_blockcount(mmc, count, is_rel_write);
@@ -105,6 +106,9 @@ static int mmc_rpmb_request(struct mmc *mmc, const struct s_rpmb *s,
cmd.cmdarg = 0;
cmd.resp_type = MMC_RSP_R1b;
+ if (host->quirks & SDHCI_QUIRK_BROKEN_R1B)
+ cmd.resp_type = MMC_RSP_R1;
+
data.src = (const char *)s;
data.blocks = 1;
data.blocksize = MMC_MAX_BLOCK_LEN;
Thanks,
-Bharat