I think the right fix here is to eliminate the ‘values’ field and reintroduce
the miniflow_get_values() getter. I’ll post a patch to that effect as part of
the v3 of the rest of the series.
Jarno
> On Jul 16, 2015, at 12:18 PM, Alin Serdean <aserd...@cloudbasesolutions.com>
> wrote:
>
> Won't work with changing [0] to [].
>
> We need to shift the [0] to be the last member of a struct. I have no idea
> about the implications that it will have but something like would work:
>
> diff --git a/lib/classifier.c b/lib/classifier.c
> index 2ed8697..9f58d43 100644
> --- a/lib/classifier.c
> +++ b/lib/classifier.c
> @@ -1351,8 +1351,8 @@ classifier_rule_overlaps(const struct classifier *cls,
> PVECTOR_FOR_EACH_PRIORITY (subtable, target->priority - 1, 2,
> sizeof(struct cls_subtable), &cls->subtables) {
> struct {
> - struct minimask mask;
> uint64_t storage[FLOW_U64S];
> + struct minimask mask;
> } m;
> const struct cls_rule *rule;
>
> diff --git a/lib/dpif-netdev.c b/lib/dpif-netdev.c
> index 9e7343b..054fc48 100644
> --- a/lib/dpif-netdev.c
> +++ b/lib/dpif-netdev.c
> @@ -97,8 +97,8 @@ static struct odp_support dp_netdev_support = {
> struct netdev_flow_key {
> uint32_t hash; /* Hash function differs for different users. */
> uint32_t len; /* Length of the following miniflow (incl. map). */
> - struct miniflow mf;
> uint64_t buf[FLOW_MAX_PACKET_U64S];
> + struct miniflow mf;
> };
>
> /* Exact match cache for frequently used flows
> @@ -133,8 +133,8 @@ struct emc_entry {
> };
>
> struct emc_cache {
> - struct emc_entry entries[EM_FLOW_HASH_ENTRIES];
> int sweep_idx; /* For emc_cache_slow_sweep(). */
> + struct emc_entry entries[EM_FLOW_HASH_ENTRIES];
> };
>
> /* Iterate in the exact match cache through every entry that might contain a
> @@ -166,7 +166,7 @@ static void dpcls_insert(struct dpcls *, struct
> dpcls_rule *,
> const struct netdev_flow_key *mask);
> static void dpcls_remove(struct dpcls *, struct dpcls_rule *);
> static bool dpcls_lookup(const struct dpcls *cls,
> - const struct netdev_flow_key keys[],
> + const struct netdev_flow_key *,
> struct dpcls_rule **rules, size_t cnt);
> ^L
> /* Datapath based on the network device interface from netdev.h.
> diff --git a/lib/flow.h b/lib/flow.h
> index 392cb59..fb412ae 100644
> --- a/lib/flow.h
> +++ b/lib/flow.h
> @@ -398,7 +398,7 @@ BUILD_ASSERT_DECL(FLOW_U64S <= 64);
> * many elements as there are 1-bits in 'map'. */
> struct miniflow {
> uint64_t map;
> - uint64_t values[0];
> + uint64_t values[];
> };
> BUILD_ASSERT_DECL(sizeof(struct miniflow) == sizeof(uint64_t));
>
> But afterwards we will hit
> (https://msdn.microsoft.com/en-us/library/0scy7z2d.aspx):
>
> lib/dpif-netdev.c(137) : error C2233: 'entries' : arrays of objects
> containing zero-size arrays are illegal
> lib/dpif-netdev.c(3337) : error C2233: 'keys' : arrays of objects containing
> zero-size arrays are illegal
> lib/dpif-netdev.c(3881) : error C2233: 'keys' : arrays of objects containing
> zero-size arrays are illegal
> lib/dpif-netdev.c(3883) : warning C4028: formal parameter 4 different from
> declaration
>
> Alin.
>
> -----Mesaj original-----
> De la: Ben Pfaff [mailto:b...@nicira.com]
> Trimis: Thursday, July 16, 2015 10:27 AM
> Către: Alin Serdean
> Cc: Jarno Rajahalme; dev@openvswitch.org
> Subiect: Re: [ovs-dev] [RFC PATCH v2 3/7] flow: Always inline miniflows.
>
> On Thu, Jul 16, 2015 at 01:35:14AM +0000, Alin Serdean wrote:
>> Hi Jarno,
>>
>> This is breaking the windows build
>> (https://ci.appveyor.com/project/blp/ovs/build/1.0.459):
>> lib/classifier.c(1355) : error C2229: struct '<unnamed-tag>' has an illegal
>> zero-sized array.
>>
>> The problem is the following
>> (https://msdn.microsoft.com/en-us/library/0scy7z2d.aspx ):
>> https://github.com/openvswitch/ovs/commit/8fd4792403254f868398a89fb562
>> 5830ee5a08eb#diff-b87c023d7e733f983131f231a672e71eR389
>>
>> Could you please take a look into it?
>
> Usually changing [0] to [] fixes such issues, can you check that that works
> here?
_______________________________________________
dev mailing list
dev@openvswitch.org
http://openvswitch.org/mailman/listinfo/dev
