kernel.git/drivers/mmc/host/alcor.c, branch linux-5.0.y

mmc: alcor: don't write data before command has completed

2019-04-17T06:39:48Z

commit 157c99c5a2956a9ab1ae12de0136a2d8a1b1a307 upstream. The alcor driver is setting up data transfer and submitting the associated MMC command at the same time. While this works most of the time, it occasionally causes problems upon write. In the working case, after setting up the data transfer and submitting the MMC command, an interrupt comes in a moment later with CMD_END and WRITE_BUF_RDY bits set. The data transfer then happens without problem. However, on occasion, the interrupt that arrives at that point only has WRITE_BUF_RDY set. The hardware notifies that it's ready to write data, but the associated MMC command is still running. Regardless, the driver was proceeding to write data immediately, and that would then cause another interrupt indicating data CRC error, and the write would fail. Additionally, the transfer setup function alcor_trigger_data_transfer() was being called 3 times for each write operation, which was confusing and may be contributing to this issue. Solve this by tweaking the driver behaviour to follow the sequence observed in the original ampe_stor vendor driver: 1. When starting request handling, write 0 to DATA_XFER_CTRL 2. Submit the command 3. Wait for CMD_END interrupt and then trigger data transfer 4. For the PIO case, trigger the next step of the data transfer only upon the following DATA_END interrupt, which occurs after the block has been written. I confirmed that the read path still works (DMA & PIO) and also now presents more consistency with the operations performed by ampe_stor. Signed-off-by: Daniel Drake Fixes: c5413ad815a6 ("mmc: add new Alcor Micro Cardreader SD/MMC driver") Cc: stable@vger.kernel.org Signed-off-by: Ulf Hansson Signed-off-by: Greg Kroah-Hartman

mmc: alcor: fix DMA reads

2019-03-27T05:17:51Z

commit 5ea47691bd99e1100707ec63364aff72324e2af4 upstream. Setting max_blk_count to 1 here was causing the mmc block layer to always use the MMC_READ_SINGLE_BLOCK command here, which the driver does not DMA-accelerate. Drop the max_blk_ settings here. The mmc host defaults suffice, along with the max_segs and max_seg_size settings, which I have now documented in more detail. Now each MMC command reads 4 512-byte blocks, using DMA instead of PIO. On my SD card, this increases read performance (measured with dd) from 167kb/sec to 4.6mb/sec. Link: http://lkml.kernel.org/r/CAD8Lp47L5T3jnAjBiPs1cQ+yFA3L6LJtgFvMETnBrY63-Zdi2g@mail.gmail.com Signed-off-by: Daniel Drake Reviewed-by: Oleksij Rempel Fixes: c5413ad815a6 ("mmc: add new Alcor Micro Cardreader SD/MMC driver") Cc: stable@vger.kernel.org Signed-off-by: Ulf Hansson Signed-off-by: Greg Kroah-Hartman

mmc: alcor: remove set but not used variables 'len, clock_out'

2018-12-17T07:26:24Z

Fixes gcc '-Wunused-but-set-variable' warning: drivers/mmc/host/alcor.c: In function 'alcor_data_set_dma': drivers/mmc/host/alcor.c:124:12: warning: variable 'len' set but not used [-Wunused-but-set-variable] drivers/mmc/host/alcor.c:648:15: warning: variable 'clock_out' set but not used [-Wunused-but-set-variable] Signed-off-by: YueHaibing Reviewed-by: Oleksij Rempel Signed-off-by: Ulf Hansson

mmc: add new Alcor Micro Cardreader SD/MMC driver

2018-12-17T07:26:24Z

This driver provides support for Alcor Micro AU6601 and AU6621 SD/MMC controller. Signed-off-by: Oleksij Rempel Signed-off-by: Ulf Hansson