moby: dont patch CMA at all; just accept the defaults

it seems like things don't actually need this
This commit is contained in:
2024-09-13 00:05:54 +00:00
parent 1971ef5fd4
commit 8b3e413cae

View File

@@ -54,9 +54,14 @@ in
# bumped to 512M on 2023/01 # bumped to 512M on 2023/01
# bumped to 1536M on 2024/05 # bumped to 1536M on 2024/05
# `cat /proc/meminfo` to see CmaTotal/CmaFree if interested in tuning this. # `cat /proc/meminfo` to see CmaTotal/CmaFree if interested in tuning this.
# kernel param mentioned here: <https://cateee.net/lkddb/web-lkddb/CMA_SIZE_PERCENTAGE.html> # kernel param mentioned here: <https://cateee.net/lkddb/web-lkddb/CMA_SIZE_PERCENTAGE.html> (or CONFIG_CMA_SIZE_MBYTES)
# cli arguments described more here: <https://lwn.net/Articles/396707/>
# i think cma mem isn't exclusive -- it can be used as ordinary `malloc`, still. i heard someone suggest the OS default should just be 50% memory to CMA. # i think cma mem isn't exclusive -- it can be used as ordinary `malloc`, still. i heard someone suggest the OS default should just be 50% memory to CMA.
"cma=1536M" # on the other hand the postmarketOS folk seek to remove their usages of CMA altogether: <https://gitlab.com/postmarketOS/pmaports/-/merge_requests/5541#note_2082937081>
# in fact, setting this to `cma=0M` still results in a bootable, graphical environment (as of 2024-09-11).
# "cma=256M"
# "cma=1536M"
# "cma=0M"
# 2023/10/20: potential fix for the lima (GPU) timeout bugs: # 2023/10/20: potential fix for the lima (GPU) timeout bugs:
# - <https://gitlab.com/postmarketOS/pmaports/-/issues/805#note_890467824> # - <https://gitlab.com/postmarketOS/pmaports/-/issues/805#note_890467824>
"lima.sched_timeout_ms=2000" "lima.sched_timeout_ms=2000"