miatoll: Override readahead to 128KiB after post-boot

* Qualcomm decided to increase the read-ahead value to 512KiB [1].

 * This value resulted in frequent cases where read-ahead
   would be waiting for new I/O requests and thus ending
   up being synchronous rather than asynchronous.

 * As result, the main goal of Qualcomm's patch [1] which
   is to increase throughput ends up in a regression where
   read throughput is significantly lower than it ever
   would be with the recommended [2] read-ahead value.

 * This patch uses our existing loop in ramdisk to set
   the read-ahead value to 128KiB.

[1] https://source.codeaurora.org/quic/la/kernel/msm-4.19/commit/include/linux/mm.h?h=LA.UM.9.12.r1-09000-SMxx50.0&id=25d04e4ceb7c067907901842c223d62973d73a75
[2] https://git.kernel.org/pub/scm/linux/kernel/git/stable/linux.git/tree/include/linux/mm.h?h=v4.19.160#n2397

Change-Id: I1980d904568fe4f1ce7b676452bcc44e4689f02f
Signed-off-by: Adithya R <gh0strider.2k18.reborn@gmail.com>
This commit is contained in:
Alex Naidis 2017-01-30 14:08:58 +01:00 committed by Aryan Gupta
parent 361812104b
commit a2f3d3c358

View File

@ -83,3 +83,16 @@ on property:vendor.post_boot.parsed=1
write /sys/block/zram0/comp_algorithm lz4
write /sys/block/zram0/max_comp_streams 8
write /proc/sys/vm/page-cluster 0
# Override readahead to 128KiB
write /sys/block/dm-0/queue/read_ahead_kb 128
write /sys/block/dm-1/queue/read_ahead_kb 128
write /sys/block/dm-2/queue/read_ahead_kb 128
write /sys/block/dm-3/queue/read_ahead_kb 128
write /sys/block/mmcblk0/queue/read_ahead_kb 128
write /sys/block/sda/queue/read_ahead_kb 128
write /sys/block/sdb/queue/read_ahead_kb 128
write /sys/block/sdc/queue/read_ahead_kb 128
write /sys/block/sdd/queue/read_ahead_kb 128
write /sys/block/sde/queue/read_ahead_kb 128
write /sys/block/sdf/queue/read_ahead_kb 128