Notes on disk flush commands and why they are important.

[Michel Talon]

Matthew Dillon wrote:

>     I don't know if gjournal is filesystem-aware. 

See for example the post by PJD here:
http://markmail.org/message/2hol5ysh6tmzwub2
which mentions that gjournal is minimally filesystem aware, whatever that
means, and that gjournal issues some BIO_FLUSH requests, and when.

> 
>     Scott mentioned barriers vs BIO_FLUSH.  I was already assuming that
>     Jeff's journaling code at least used barriers (i.e. waits for all
>     prior I/O to complete before issuing dependent I/O).  That is
>     mandatory since both NCQ on the device and (potentially)
>     bioqdisksort() (if it is still being used) will reorder write BIOs
>     in-progress.

My understanding of the discussion is that the journalling code doesn't
issue any sort of barrier, and that the author doesn't care of people who
don't use industrial type hardware. So doing it misses at least the case of
problems created by a panic, which happen not so infrequently,
unfortunately.



-- 
Michel Talon