tree eeb9de8c3769f830fdb138b53a39911a97e4e18b
parent 8234c16354a5723aafab7ab24f0b833dc00c926d
author Lorenz Brun <lorenz@monogon.tech> 1658175325 +0000
committer Lorenz Brun <lorenz@monogon.tech> 1658745466 +0000

m/n/c/l/crypt: improve zeroing speed

This improves zeroing speed by a lot by mapping the integrity device
without journal for zeroing. This halves IOPS requirements as well as
making IO mostly sequential (instead of random with the journal).
This also bumps the write block size to 256 times the block size which
experimentally seems to reliably fill up the write queue while not being
too big.
Asynchronous IO (io_uring) doesn't perform better here, at least on not
disks which aren't insanely fast.

Change-Id: Ia8fbed35cd0c8978209fa24dc7615bc71ff1ba2f
Reviewed-on: https://review.monogon.dev/c/monogon/+/840
Tested-by: Jenkins CI
Reviewed-by: Leopold Schabel <leo@monogon.tech>