848020 - Do we really have to poison NtFlushBuffersFile?

Status: RESOLVED FIXED

(Core :: XPCOM, defect, P2)

Description

[Rafael Ávila de Espíndola (:espindola) (not reading bugmail)]

Reading the wine source it really looks like NtFlushBuffersFile is just the windows equivalent of flush. So poisoning it would only be necessary if we were trying to prevent data loss in power failures, not firefox calling exit(0).

Am I missing something about NtFlushBuffersFile semantics? Doesn't windows save buffers when a process exits?

Comment 1

[(no longer active)]

I think it does, unless you call TerminateProcess.  It might be worth looking into the differences between ExitProcess and TerminateProcess (I think the first one is supposed to flush all I/O buffers.)

Comment 2

[Rafael Ávila de Espíndola (:espindola) (not reading bugmail)]

Attachment: patch

OK. It looks like that windows does write the data to disk on exit but not on terminate. While poisoning NtFlushBuffersFile would be an interesting project to protect against data loss in a power loss or firefox being killed, that is out of scope of the exit(0) project.

Comment 3

[Rafael Ávila de Espíndola (:espindola) (not reading bugmail)]

https://hg.mozilla.org/integration/mozilla-inbound/rev/fe8a8cad48a4

Comment 4

[Ryan VanderMeulen [:RyanVM]]

https://hg.mozilla.org/mozilla-central/rev/fe8a8cad48a4