As expressed in this thread, the intention of F_NOCACHE is to disable caching for accesses via a particular file descriptor and the intention of F_GLOBAL_NOCACHE is to disable caching for accesses via all file descriptors.
- In the thread it is mentioned that F_NOCACHE, then, had an effect similar to F_GLOBAL_NOCACHE and it is intended to be implemented correctly in future systems. Can I get an insight on how the current implementation works?
- If the current implementation of F_NOCACHE is per file descriptor, does that mean, caching due to access via other file descriptors does not affect this file descriptor and will the access be made directly from disk?
- What happens when I set F_NOCACHE 1 for a file descriptor and later someone sets F_GLOBAL_NOCACHE as 0?
- If something is already in cache, read is said to happen via cache even if F_NOCACHE is set to 1. For writing, is the same applicable or is it always write-through?