Krishna Konda <kkonda <at> codeaurora.org> writes: > > On 6/4/2013 4:32 AM, Dirk Behme wrote: > > > > Using some recent Hynix eMMC devices [1] on our Freescale i.MX6 boards > > we get harmless (?), but annoying access timeouts accessing the RPMB > > partition: > > > > mmcblk1rpmb: error -110 transferring data, sector 0, nr 32, cmd response > > 0x900, card status 0xb00 > > mmcblk1rpmb: retrying using single block read > > ... > > > > The output with MMC debug enabled below [2]. > > > > This seems to be harmless because it stops, but is annoying due to some > > auto mounter trying to access all available partitions. > > > The RPMB partition should never be mounted. Its not a regular partition > and does not a file system on it. So my suggestion would to be to modify > the auto mounter to not mount RPMB paritions. > Hi Dirk, Krishna These issues are not caused by udev/auto-mounting, but by the mmc block driver initialisation code. During the mmc initialisation, all the non zero sized partitions (read from ext_csd register) are added in card structures. Then at the time of mmc block driver probing, mmc_blk_alloc_part function is called for each partition which initialises the mmc request queue and and starts mmc queue thread for each partition as shown below root@mx6q35:~# ps x | grep mmc 46 ? S 0:00 [mmcqd/1] 397 ? S 0:00 [mmcqd/2] 400 ? S 0:00 [mmcqd/2boot0] 401 ? S 0:00 [mmcqd/2boot1] 402 ? S 0:00 [mmcqd/2rpmb] mmc queue thread for each partition gets scheduled and calls mmc_blk_issue_rq, and here it sends CMD6 using mmc_blk_part_switch, CMD23-->CMD18 using mmc_blk_issue_rw_rq. When executed for RPMB partition, logs can be seen below [ 109.972564] mmc2: starting CMD6 arg 03b30b01 flags 0000049d --> switch to RPMB area [ 109.972600] sdhci [sdhci_irq()]: *** mmc2 got interrupt: 0x00000001 [ 109.972625] mmc2: req done (CMD6): 0: 00000800 00000000 00000000 00000000 [ 109.977781] mmc2: starting CMD13 arg 00010000 flags 00000195 [ 109.977823] sdhci [sdhci_irq()]: *** mmc2 got interrupt: 0x00000001 [ 109.977849] mmc2: req done (CMD13): 0: 00000900 00000000 00000000 00000000 [ 109.983013] <mmc2: starting CMD23 arg 00000080 flags 00000015> [ 109.983024] mmc2: starting CMD18 arg 00000000 flags 000000b5 [ 109.983037] mmc2: blksz 512 blocks 128 flags 00000200 tsac 100 ms nsac 0 [ 109.983048] mmc2: CMD12 arg 00000000 flags 0000049d [ 109.983083] sdhci [sdhci_irq()]: *** mmc2 got interrupt: 0x00000001 [ 109.983218] sdhci [sdhci_irq()]: *** mmc2 got interrupt: 0x00000001 [ 109.986505] sdhci [sdhci_irq()]: *** mmc2 got interrupt: 0x0000000a [ 109.986518] sdhci [sdhci_data_irq()]: mmc2: DMA base 0x3caf0000, transferred 0x010000 bytes, next 0x3cb00000 [ 109.986672] mmc2: req done (CMD18): 0: 00000900 00000000 00000000 00000000 [ 109.986691] mmc2: 65536 bytes transferred: 0 [ 109.986709] mmc2: (CMD12): 0: 00000000 00000000 00000000 00000000 [ 109.992355] <mmc2: starting CMD23 arg 00000080 flags 00000015> [ 109.992368] mmc2: starting CMD18 arg 00000080 flags 000000b5 [ 109.992381] mmc2: blksz 512 blocks 128 flags 00000200 tsac 100 ms nsac 0 [ 109.992392] mmc2: CMD12 arg 00000000 flags 0000049d [ 109.992424] sdhci [sdhci_irq()]: *** mmc2 got interrupt: 0x00000001 [ 109.992724] sdhci [sdhci_irq()]: *** mmc2 got interrupt: 0x00000001 [ 109.996010] sdhci [sdhci_irq()]: *** mmc2 got interrupt: 0x0000000a [ 109.996025] sdhci [sdhci_data_irq()]: mmc2: DMA base 0x3cb40000, transferred 0x040000 bytes, next 0x3cb80000 [ 109.996160] mmc2: req done (CMD18): 0: 00000900 00000000 00000000 00000000 [ 109.996169] mmc2: 65536 bytes transferred: 0 [ 109.996179] mmc2: (CMD12): 0: 00000000 00000000 00000000 00000000 [ 110.002250] mmc2: starting CMD6 arg 03b30a01 flags 0000049d --> switch to Boot Partition 2 area [ 110.002292] sdhci [sdhci_irq()]: *** mmc2 got interrupt: 0x00000001 [ 110.002318] mmc2: req done (CMD6): 0: 00000800 00000000 00000000 00000000 This is the point where you are facing issues with Hynix devices. A good way to reproduce the issue can be by unbinding/binding the eMMC device with block driver as follows root@mx6q35:~# echo mmc2:0001 > /sys/bus/mmc/drivers/mmcblk/unbind root@mx6q35:~# echo mmc2:0001 > /sys/bus/mmc/drivers/mmcblk/bind Krishna, since RPMB is not a regular partition and a different command sequence(CMD6-->CMD23-->CMD25-->CMD23-->CMD18) as mentioned in JEDEC Standard No. 84-A441, is required to access it, then why mmc initialisation code is using the wrong command sequence(CMD6-->CMD23-->CMD18) to access it? Thanks . Abbas Raza -- To unsubscribe from this list: send the line "unsubscribe linux-mmc" in the body of a message to majordomo@xxxxxxxxxxxxxxx More majordomo info at http://vger.kernel.org/majordomo-info.html
