Maximilian Lorlacks wrote: > Good day, > > It seems that fsync(2) may data after returning EIO once[1]. This > behaviour seems to cause problems with databases such as PostgreSQL > and goes contrary to the man page's description, which says that > "fsync() and fdatasync() should be used by programs that require a > file to be in a known state, for example, in building a simple > transaction facility." > > Are there any plans to change fsync(2) to allow for retries and/or > persistent error reporting?
No, this looks like a bug, but I'm not sure it's getting fixed soon. Keeping a buf with an error in the delayed write list would probably have some serious consequences. When would we ever remove it? For completeness, I'll also add that if the syncer gets to the bad buf before fsync, you might miss the error, but that's only after 30 seconds or so.
