mmc: hsq: Improve random I/O write performance for 4k buffers

By dynamically adjusting the host->hsq_depth, based upon the buffer size
being 4k and that we get at least two I/O write requests in flight, we can
improve the throughput a bit. This is typical for a random I/O write
pattern.

More precisely, by dynamically changing the number of requests in flight
from 2 to 5, we can on some platforms observe ~4-5% increase in throughput.

Signed-off-by: Wenchao Chen <wenchao.chen@unisoc.com>
Link: https://lore.kernel.org/r/20230919074707.25517-3-wenchao.chen@unisoc.com
[Ulf: Re-wrote the commitmsg, minor adjustment to the code - all to clarify.]
Signed-off-by: Ulf Hansson <ulf.hansson@linaro.org>

mmc: core: Allow dynamical updates of the number of requests for hsq

To allow dynamical updates of the current number of used in-flight
requests, let's move away from using a hard-coded value to a use a
corresponding variable in the struct mmc_host.

This can be valuable when optimizing for certain I/O request sequences, as
shown by subsequent changes.

Signed-off-by: Wenchao Chen <wenchao.chen@unisoc.com>
Link: https://lore.kernel.org/r/20230919074707.25517-2-wenchao.chen@unisoc.com
[Ulf: Re-wrote the commitmsg to clarify the change]
Signed-off-by: Ulf Hansson <ulf.hansson@linaro.org>

mmc: mmc-hsq: Use fifo to dispatch mmc_request

Current next_tag selection will cause a large delay in some requests and
destroy the scheduling results of the block scheduling layer. Because the
issued mrq tags cannot ensure that each time is sequential, especially when
the IO load is heavy. In the fio performance test, we found that 4k random
read data was sent to mmc_hsq to start calling request_atomic It takes
nearly 200ms to process the request, while mmc_hsq has processed thousands
of other requests. So we use fifo here to ensure the first in, first out
feature of the request and avoid adding additional delay to the request.

Reviewed-by: Wenchao Chen <wenchao.chen@unisoc.com>
Signed-off-by: Michael Wu <michael@allwinnertech.com>
Link: https://lore.kernel.org/r/20221128093847.22768-1-michael@allwinnertech.com
Signed-off-by: Ulf Hansson <ulf.hansson@linaro.org>

mmc: hsq: Fix data stomping during mmc recovery

The block device uses multiple queues to access emmc. There will be up to 3
requests in the hsq of the host. The current code will check whether there
is a request doing recovery before entering the queue, but it will not check
whether there is a request when the lock is issued. The request is in recovery
mode. If there is a request in recovery, then a read and write request is
initiated at this time, and the conflict between the request and the recovery
request will cause the data to be trampled.

Signed-off-by: Wenchao Chen <wenchao.chen@unisoc.com>
Fixes: 511ce378e16f ("mmc: Add MMC host software queue support")
Cc: stable@vger.kernel.org
Link: https://lore.kernel.org/r/20220916090506.10662-1-wenchao.chen666@gmail.com
Signed-off-by: Ulf Hansson <ulf.hansson@linaro.org>

mmc: host: Introduce the request_atomic() for the host

The SD host controller can process one request in the atomic context if
the card is nonremovable, which means we can submit next request in the
irq hard handler when using the MMC host software queue to reduce the
latency. Thus this patch adds a new API request_atomic() for the host
controller, as well as adding support for host software queue to submit
a request by the new request_atomic() API.

Moreover there is an unusual case that the card is busy when trying to
send a command, and we can not polling the card status in interrupt
context by using request_atomic() to dispatch requests. Thus we should
queue a work to try again in the non-atomic context in case the host
releases the busy signal later.

Suggested-by: Adrian Hunter <adrian.hunter@intel.com>
Acked-by: Adrian Hunter <adrian.hunter@intel.com>
Signed-off-by: Baolin Wang <baolin.wang7@gmail.com>
Link: https://lore.kernel.org/r/a344e27e506cb2329073cbd5cf65e15cc3cbeba9.1586744073.git.baolin.wang7@gmail.com
Signed-off-by: Ulf Hansson <ulf.hansson@linaro.org>

mmc: host: hsq: Add missing MODULE_LICENSE() and MODULE_DESCRIPTION()

Add missing MODULE_LICENSE() and MODULE_DESCRIPTION() in hsq driver to
fix below warning when compiling the hsq as a module.

"WARNING: modpost: missing MODULE_LICENSE() in drivers/mmc/host/mmc_hsq.o".

Reported-by: Stephen Rothwell <sfr@canb.auug.org.au>
Signed-off-by: Baolin Wang <baolin.wang7@gmail.com>
Link: https://lore.kernel.org/r/98ce471185f037fce57520763621590588766381.1582161803.git.baolin.wang7@gmail.com
Signed-off-by: Ulf Hansson <ulf.hansson@linaro.org>

mmc: Add MMC host software queue support

Now the MMC read/write stack will always wait for previous request is
completed by mmc_blk_rw_wait(), before sending a new request to hardware,
or queue a work to complete request, that will bring context switching
overhead and spend some extra time to poll the card for busy completion
for I/O writes via sending CMD13, especially for high I/O per second
rates, to affect the IO performance.

Thus this patch introduces MMC software queue interface based on the
hardware command queue engine's interfaces, which is similar with the
hardware command queue engine's idea, that can remove the context
switching. Moreover we set the default queue depth as 64 for software
queue, which allows more requests to be prepared, merged and inserted
into IO scheduler to improve performance, but we only allow 2 requests
in flight, that is enough to let the irq handler always trigger the
next request without a context switch, as well as avoiding a long latency.

Moreover the host controller should support HW busy detection for I/O
operations when enabling the host software queue. That means, the host
controller must not complete a data transfer request, until after the
card stops signals busy.

From the fio testing data in cover letter, we can see the software
queue can improve some performance with 4K block size, increasing
about 16% for random read, increasing about 90% for random write,
though no obvious improvement for sequential read and write.

Moreover we can expand the software queue interface to support MMC
packed request or packed command in future.

Reviewed-by: Arnd Bergmann <arnd@arndb.de>
Signed-off-by: Baolin Wang <baolin.wang@linaro.org>
Signed-off-by: Baolin Wang <baolin.wang7@gmail.com>
Link: https://lore.kernel.org/r/4409c1586a9b3ed20d57ad2faf6c262fc3ccb6e2.1581478568.git.baolin.wang7@gmail.com
Signed-off-by: Ulf Hansson <ulf.hansson@linaro.org>