Is KeyDB safe to run on Linux ARM host with PAGESIZE 64k?

[macko99]

Hello,

I intend to run KeyDB on a Kubernetes cluster hosted on an ARM machine with a Linux kernel compiled with a 64k PAGESIZE. To achieve this, I compiled jemalloc to work with larger pages (--with-lg-page=16) and then used it during the KeyDB compilation. However, when attempting to start KeyDB, I encountered a failed assertion in the function linuxMadvFreeForkBugCheck.

After inspecting KeyDB's source code, I determined that the assertion failure was not caused by a potential bug in my kernel but rather by the test implementation itself—it assumed the kernel uses 4k memory pages. I modified the function accordingly, and KeyDB now seems to be working just fine (link to changes for review).

Here’s my question: Is this modified version of KeyDB safe to use in production? Are there any other parts of KeyDB specifically designed to work only with 4k memory pages, or was it just this one test that was not comprehensive?