kernel.git/drivers/mmc/core/core.c, branch linux-4.10.y

mmc: core: Further fix thread wake-up

2016-12-20T10:55:04Z

Commit e0097cf5f2f1 ("mmc: queue: Fix queue thread wake-up") did not go far enough. mmc_wait_for_data_req_done() still contains some problems and can be further simplified. First it should not touch context_info->is_waiting_last_req because that is a wake-up control used by the owner of the context. Secondly, it should always return when one of its wake-up conditions is met because, again, that is contolled by the owner of the context. While the current block driver does not have an issue, these problems were exposed during testing of the Software Command Queue patches. Fixes: e0097cf5f2f1 ("mmc: queue: Fix queue thread wake-up") Signed-off-by: Adrian Hunter Tested-by: Harjani Ritesh Signed-off-by: Ulf Hansson

mmc: core: remove BUG_ONs from core.c

2016-12-05T09:31:11Z

BUG_ONs doesn't help anything except for stop the system from running. If it occurs, it implies we should deploy proper error handling for that. So this patch is gonna discard these meaningless BUG_ONs and deploy error handling if needed. Signed-off-by: Shawn Lin Signed-off-by: Ulf Hansson

mmc: queue: Fix queue thread wake-up

2016-12-05T09:31:03Z

The only time the driver sleeps expecting to be woken upon the arrival of a new request, is when the dispatch queue is empty. The only time that it is known whether the dispatch queue is empty is after NULL is returned from blk_fetch_request() while under the queue lock. Recognizing those facts, simplify the synchronization between the queue thread and the request function. A couple of flags tell the request function what to do, and the queue lock and barriers associated with wake-ups ensure synchronization. The result is simpler and allows the removal of the context_info lock. Signed-off-by: Adrian Hunter Reviewed-by: Linus Walleij Reviewed-by: Harjani Ritesh Signed-off-by: Ulf Hansson

mmc: delete is_first_req parameter from pre-request callback

2016-11-29T08:05:27Z

The void (*pre_req) callback in the struct mmc_host_ops vtable is passing an argument "is_first_req" indicating whether this is the first request or not. None of the in-kernel users use this parameter: instead, since they all just do variants of dma_map* they use the DMA cookie to indicate whether a pre* callback has already been done for a request when they decide how to handle it. Delete the parameter from the callback and all users, as it is just pointless cruft. Signed-off-by: Linus Walleij Acked-by: Jaehoon Chung Signed-off-by: Ulf Hansson

mmc: core: Enable __mmc_switch() to change bus speed timing for the host

2016-11-29T08:05:25Z

In cases when a speed mode change is requested for mmc cards, a CMD6 is sent by calling __mmc_switch() during the card initialization. The CMD6 leads to the card entering a busy period. When that is completed, the host must parse the CMD6 status to find out whether the change of the speed mode succeeded. To enable the mmc core to poll the card by using CMD13 to find out when the busy period is completed, it's reasonable to make sure polling is done by having the mmc host and the mmc card, being configured to operate at the same selected bus speed timing. Therefore, let's extend __mmc_switch() to take yet another parameter, which allow its callers to update the bus speed timing of the mmc host. In this way, __mmc_switch() also becomes capable of reading and validating the CMD6 status by sending a CMD13, in cases when that's desired. If __mmc_switch() encounters a failure, we make sure to restores the old bus speed timing for the mmc host, before propagating the error code. Signed-off-by: Ulf Hansson Tested-by: Linus Walleij Acked-by: Adrian Hunter

mmc: core: use enum mmc_blk_status properly

2016-11-29T08:00:51Z

There were several instances of code using the enum mmc_blk_status by arbitrarily converting it to an int and throwing it around to different functions. This makes the code hard to understand to may give rise to strange errors. Especially the function prototype mmc_start_req() had to be modified to take a pointer to an enum mmc_blk_status and the function pointer .err_check() inside struct mmc_async_req needed to return an enum mmc_blk_status. In every case: instead of assigning the block layer error code to an int, use the enum, also change the signature of all functions actually passing this enum to use the enum. To make it possible to use the enum everywhere applicable, move it to so that all code actually using it can also see it. An interesting case was encountered in the MMC test code which did not return a enum mmc_blk_status at all in the .err_check function supposed to check whether asynchronous requests worked or not: instead it returned a normal -ERROR or even the test frameworks internal error codes. The test code would also pass on enum mmc_blk_status codes as error codes inside the test code instead of converting them to the local RESULT_* codes. I have tried to fix all instances properly and run some tests on the result. Cc: Chunyan Zhang Cc: Baolin Wang Signed-off-by: Linus Walleij Signed-off-by: Ulf Hansson

mmc: core: Don't power off the card when starting the host

2016-11-29T08:00:44Z

The MMC_CAP2_NO_PRESCAN_POWERUP was invented to avoid running the power up sequence, mmc_power_up(), during ->probe() of the mmc host driver, but instead defer this to the mmc detect work. This is especially useful for those hosts that suffers from a long initialization time, as this time would otherwise add up to the total boot time. However, due to the introduction of runtime PM of mmc host devices in the mmc core, this behaviour changed a bit. More precisely, it caused the mmc core to runtime resume the host device during ->probe() of the host driver. In cases like the rtsx_usb_sdmmc, runtime resuming the device may be costly and thus affecting the total boot time. To improve this behaviour when using MMC_CAP2_NO_PRESCAN_POWERUP, let's postpone also calling mmc_power_off() when starting the host. This change allows the mmc core to avoid runtime resuming the device, as it don't need to claim the host for that execution path. Cc: Ritesh Raj Sarraf Cc: Alan Stern Signed-off-by: Ulf Hansson

mmc: core: don't try to switch block size for dual rate mode

2016-09-26T19:31:34Z

Per spec, block size should always be 512 bytes for dual rate mode, so any attempts to switch the block size under dual rate mode should be neglected. Signed-off-by: Ziyuan Xu Signed-off-by: Shawn Lin Signed-off-by: Ulf Hansson

mmc: core: Add support for sending commands during data transfer

2016-09-26T19:31:28Z

A host controller driver exposes its capability using caps flag MMC_CAP_CMD_DURING_TFR. A driver with that capability can accept requests that are marked mrq->cap_cmd_during_tfr = true. Then the driver informs the upper layers when the command line is available for further commands by calling mmc_command_done(). Because of that, the driver will not then automatically send STOP commands, and it is the responsibility of the upper layer to send a STOP command if it is required. For requests submitted through the mmc_wait_for_req() interface, the caller sets mrq->cap_cmd_during_tfr = true which causes mmc_wait_for_req() in fact not to wait. The caller can then send commands that do not use the data lines. Finally the caller can wait for the transfer to complete by calling mmc_wait_for_req_done() which is now exported. For requests submitted through the mmc_start_req() interface, the caller again sets mrq->cap_cmd_during_tfr = true, but mmc_start_req() anyway does not wait. The caller can then send commands that do not use the data lines. Finally the caller can wait for the transfer to complete in the normal way i.e. calling mmc_start_req() again. Irrespective of how a cap_cmd_during_tfr request is started, mmc_is_req_done() can be called if the upper layer needs to determine if the request is done. However the appropriate waiting function (either mmc_wait_for_req_done() or mmc_start_req()) must still be called. The implementation consists primarily of a new completion mrq->cmd_completion which notifies when the command line is available for further commands. That completion is completed by mmc_command_done(). When there is an ongoing data transfer, calls to mmc_wait_for_req() will automatically wait on that completion, so the caller does not have to do anything special. Note, in the case of errors, the driver may call mmc_request_done() without calling mmc_command_done() because mmc_request_done() always calls mmc_command_done(). Signed-off-by: Adrian Hunter Signed-off-by: Ulf Hansson

mmc: core: Optimize the mmc erase size alignment

2016-09-26T19:31:26Z

In most cases the 'card->erase_size' is power of 2, then the round_up/down() function is more efficient than '%' operation when the 'card->erase_size' is power of 2. Signed-off-by: Baolin Wang Tested-by: Shawn Lin Signed-off-by: Ulf Hansson