[dpdk-dev] [PATCH v2 0/4] Clean up EAL runtime data paths

Bruce Richardson bruce.richardson at intel.com
Wed May 9 18:11:42 CEST 2018


On Wed, May 09, 2018 at 04:59:39PM +0100, Van Haaren, Harry wrote:
> > From: dev [mailto:dev-bounces at dpdk.org] On Behalf Of Anatoly Burakov
> > Sent: Monday, April 30, 2018 1:09 PM
> > To: dev at dpdk.org
> > Cc: Richardson, Bruce <bruce.richardson at intel.com>; thomas at monjalon.net
> > Subject: [dpdk-dev] [PATCH v2 0/4] Clean up EAL runtime data paths
> > 
> > As has been suggested [1], all DPDK runtime paths should be put
> > into a single place. This patchset accomplishes exactly that.
> > 
> > If running as root, all files will be put under /var/run/dpdk/<prefix>,
> > otherwise they will be put under $XDG_RUNTIME_PATH/dpdk/<prefix>, or, if
> > that environment variable is not defined, all files will go under
> > /tmp/dpdk/<prefix>.
> > 
> > [1] http://dpdk.org/dev/patchwork/patch/38688/
> > 
> > v2:
> > - Rebase on rc1
> > 
> > Anatoly Burakov (4):
> >   eal: remove unused define
> >   eal: rename function returning hugepage data path
> >   eal: add directory for DPDK runtime data
> >   eal: move all runtime data into DPDK runtime dir
> 
> <snip>
> 
> 
> No full code review, high level comments:
> 
> We have to be careful in changing /var/run/.rte_config, which has always been
> the default DPDK primary application lockfile. This has been used to identify
> if a primary DPDK application is alive (see rte_eal_primary_proc_alive()) and
> possibly the write-lock on this file is checked by other tools/utilities directly
> without any DPDK function call.
> 
> Changing the filepath just before a release isn't a good idea - we should treat
> this as an ABI/API break, as the change will break functionality in other projects
> such as CollectD[1], which (by default ;) rely on the defaults. There is a config
> file for CollectD to manually override the location, but this will cause headaches
> from a usability POV.
> 
> I'm not opposed to the change - particularly as I gather the new memory subsystem
> causes a number of lockfiles to be created - but we must do our due diligence and
> give other projects fair-warning that this change is coming.
> 
> As such, I recommend this patchset in its current form (particularly patches 2,3,4)
> to be deferred past 18.05.
> 
> 
What about if we keep the .rte_config file in the same place but move the
rest? The number of new files causes quite a bit of clutter. We can then
have a deprecation notice for the move in 18.05 and finish the cleanup in
18.08.

/Bruce


More information about the dev mailing list