Re: [PATCH 0/6] Reduce scope address-of-packed-member warning

[Bruce Richardson <bruce.richardson@intel.com>]

On Thu, Oct 17, 2024 at 09:21:48AM -0700, Stephen Hemminger wrote:
> On Thu, 17 Oct 2024 15:22:07 +0100
> Bruce Richardson <bruce.richardson@intel.com> wrote:
> 
> > The warning for address-of-packed-member was being disabled globally in
> > DPDK.  While for drivers which need to access hardware-defined
> > data-structures the use of packed may make sense, for normal libs and
> > applications the use of packed data should be generally avoided.
> > 
> > This patchset initially applies some fixes for cases where we are
> > unnecessarily causing the warning to trigger. Thereafter the 6th patch
> > removes the global enabling of the warning and replaces it with more
> > selective disabling for drivers and for a couple of other components
> > which have not yet been fixed.
> > 
> > Bruce Richardson (6):
> >   ip_frag: remove use of unaligned variable
> >   efd: remove unnecessary packed attributes
> >   bus/ifpga: remove packed attribute
> >   pipeline: remove packed attribute
> >   net: add smaller IPv4 cksum function for simple cases
> >   build: limit scope of packed member warning disabling
> > 
> >  app/test-eventdev/test_pipeline_common.c | 25 +-----------
> >  app/test-pmd/icmpecho.c                  | 23 +----------
> >  app/test-pmd/txonly.c                    | 22 +----------
> >  app/test/packet_burst_generator.c        | 49 +-----------------------
> >  app/test/test_reassembly_perf.c          | 29 +-------------
> >  config/meson.build                       |  1 -
> >  drivers/bus/ifpga/bus_ifpga_driver.h     |  2 +-
> >  drivers/meson.build                      |  9 +++--
> >  examples/ipsec-secgw/meson.build         |  6 +++
> >  lib/efd/rte_efd.c                        |  4 +-
> >  lib/ip_frag/rte_ipv4_reassembly.c        |  4 +-
> >  lib/net/rte_ip.h                         | 33 ++++++++++++++++
> >  lib/pipeline/rte_table_action.c          |  2 +-
> >  lib/vhost/meson.build                    |  5 ++-
> >  14 files changed, 60 insertions(+), 154 deletions(-)
> 
> 
> Any chance to just fix all the places and have it enabled globally?

Not any time soon (at least for me). There are a huge number of drivers
with lots of errors if we turn back on the warning. We should look at all
our common structures first to ensure none of them are unnecessarily
packed.

/Bruce