On Thu, 2010-12-02 at 17:46 +0100, Michael Biebl wrote: > Hi Ben > > On 02.12.2010 07:16, Ben Hutchings wrote: > > >> please consider enabling fanotify in the upcoming 2.6.37(-rcX) packages. > >> > >> Systemd has a readahead implementation based on fanotify which we would > >> like to enable. > > > > There seem to be alternate ways of doing readahead, > > Are you referring to fedora-readahead here which uses the audit mechanism? > Some consider that a misuse of the audit subsystem. > > Then there is ureadahead from Ubuntu, which afaik, requires a kernel patch to work. > > The old readahead project seems to be dead. > > What would be the alternate way to do readahead? Do you have any more specific > information. I don't, I just knew it had been done in other ways. But from what you tell me, they really aren't as good as using fanotify(). > so I'm not yet > > convinced it's worth the cost. (It appears to add about 4-5K code and > > static data, which is nothing on a PC but a bit more significant on an > > ARM system with a fixed-size flash partition for the kernel image.) > > There are plans to use fanotify within tracker, a filesystem indexer and search > engine, which would make it another user. Right. Well, it will definitely be considered. Ben. -- Ben Hutchings Once a job is fouled up, anything done to improve it makes it worse.
Attachment:
signature.asc
Description: This is a digitally signed message part