On Tue, Feb 23, 2016 at 9:01 PM, Pravin B Shelar <pshe...@ovn.org> wrote:
> diff --git a/FAQ.md b/FAQ.md
> index 8bd7ab9..8e0a775 100644
> --- a/FAQ.md
> +++ b/FAQ.md
> @@ -157,6 +157,7 @@ A: The following table lists the Linux kernel versions
> against which the
> | 2.3.x | 2.6.32 to 3.14
> | 2.4.x | 2.6.32 to 4.0
> | 2.5.x | 2.6.32 to 4.3
> +| Master | 3.10 to 4.3
I would just put "2.6.x" instead of "Master" here on the assumption
that that will be the next release. Otherwise, we'll most likely
forget and people will be confused.
There is a fair amount of documentation in INSTALL.md that we can
remove. Specifically, the sections in "Build Requirements" about
CONFIG_BRIDGE and GRE. In addition, in "Building the Sources", many of
the reasons for modprobe failing can be removed as well as the
documentation for the vlan_tso module option.
There's definitely some remaining code, which is now dead on the
kernels that we support (including the vlan_tso option and other code
in gso.h). I think that's fine as long as we get the user facing parts
and a good chunk of the old code. Obviously, we can continue to clean
up and remove pieces which are no longer used over time.
> diff --git a/acinclude.m4 b/acinclude.m4
> index 11c7787..0ae6a81 100644
> --- a/acinclude.m4
> +++ b/acinclude.m4
> - OVS_GREP_IFELSE([$KSRC/include/linux/netdevice.h], [pcpu_sw_netstats])
> + dnl Ubuntu kernel 3.13 has defined this struct but not used for
> netdev->tstats.
> + dnl So check type of tstats.
> + OVS_GREP_IFELSE([$KSRC/include/linux/netdevice.h],
> [pcpu_sw_netstats.*tstats],
> + [OVS_DEFINE([HAVE_PCPU_SW_NETSTATS])])
Is this a bug fix independent of the rest of the series?
> diff --git a/datapath/linux/compat/include/linux/netdevice.h
> b/datapath/linux/compat/include/linux/netdevice.h
> index 19a7b8e..581d7a0 100644
> --- a/datapath/linux/compat/include/linux/netdevice.h
> +++ b/datapath/linux/compat/include/linux/netdevice.h
> @@ -169,13 +169,7 @@ static inline struct net_device
> *netdev_notifier_info_to_dev(void *info)
>
> #include <linux/u64_stats_sync.h>
>
> -struct pcpu_sw_netstats {
> - u64 rx_packets;
> - u64 rx_bytes;
> - u64 tx_packets;
> - u64 tx_bytes;
> - struct u64_stats_sync syncp;
> -};
> +#define pcpu_sw_netstats pcpu_tstats
I guess we can probably drop the #include for u64_stats_sync.h
> diff --git a/datapath/vport-internal_dev.c b/datapath/vport-internal_dev.c
> index 7f21679..ec76398 100644
> --- a/datapath/vport-internal_dev.c
> +++ b/datapath/vport-internal_dev.c
Since we're killing off HAVE_DEV_TSTATS here, I think we should
probably remove the path that sets it too - the checks in acinclude.m4
for rtnl_link_stats64 and the code that sets it in
datapath/linux/compat/include/linux/netdevice.h.
> diff --git a/datapath/vport-netdev.c b/datapath/vport-netdev.c
> index a1df79b..02917b0 100644
> --- a/datapath/vport-netdev.c
> +++ b/datapath/vport-netdev.c
It would be really nice to cleanup ovs_netdev_get_vport() as currently
it's a totally unreadable mess. HAVE_RHEL_OVS_HOOK should be related
to RHEL 6, so I think that all of that code can go, including the
extra netdev_frame_hook().
_______________________________________________
dev mailing list
dev@openvswitch.org
http://openvswitch.org/mailman/listinfo/dev
