[clug] rsync question

Eyal Lebedinsky eyal at eyal.emu.id.au
Fri Nov 29 22:21:16 MST 2013

Thanks Stephen,

I see. I would have expected the on-disk to not be used when an in-core cached copy exists.
However, the idea that when the cached copy goes away the on-disk copy is read and is wrong,
then it gets rewritten and looks good afterwards. This suggests that file operations use
an in-core structure that is not vulnerable to the 2-second rounding. I probably need to
look into linux internals *every* 20 years...


On 11/30/13 15:52, Stephen Rothwell wrote:
> Hi Eyal,
> On Sat, 30 Nov 2013 15:26:12 +1100 Eyal Lebedinsky <eyal at eyal.emu.id.au> wrote:
>> However this does not explain why a quick rerun finds nothing to send yet later it
>> resends practically everything. The source is static throughout.
> It could depend on when the inodes are flushed from RAM.  While the
> inodes are in RAM, the timestamps will be whatever was last set, but once
> they are flushed to disk and then reread, the 2 second resolution of the
> disk format will change them.

Eyal Lebedinsky	(eyal at eyal.emu.id.au)

More information about the linux mailing list