| Message ID | 20190607232824.23360-1-blp@ovn.org |
|---|---|
| State | Accepted |
| Headers | show |
| Series | [ovs-dev,v2] ofproto-dpif-xlate: Do not use zero-weight buckets in select groups. | expand |
On 6/8/19 1:28 AM, Ben Pfaff wrote: > The OpenFlow specification says that buckets in select groups with a weight > of zero should not be selected, but the ofproto-dpif implementation could > select them in corner cases. This fixes the problem. > > Reported-by: ychen <ychen103103@163.com> > Reported-at: https://mail.openvswitch.org/pipermail/ovs-dev/2019-May/359349.html > Signed-off-by: Ben Pfaff <blp@ovn.org> > --- > v1->v2: Simplify and fix figuring out whether a group is alive. Hi, Ben. I'm looking through old patches after the patchwork cleanup and this one seems to be still valid. It requires a minor rebase, but it's trivial. Patch looks good to me in general and I could apply it. What do you think? It also looks like we will need to backport it. Suggestions on how far it should go are welcome. Best regards, Ilya Maximets. > > ofproto/ofproto-dpif-xlate.c | 18 ++++++++++++------ > 1 file changed, 12 insertions(+), 6 deletions(-) > > diff --git a/ofproto/ofproto-dpif-xlate.c b/ofproto/ofproto-dpif-xlate.c > index 04d69ed06c20..1ace92d22019 100644 > --- a/ofproto/ofproto-dpif-xlate.c > +++ b/ofproto/ofproto-dpif-xlate.c > @@ -1,4 +1,4 @@ > -/* Copyright (c) 2009, 2010, 2011, 2012, 2013, 2014, 2015, 2016, 2017 Nicira, Inc. > +/* Copyright (c) 2009, 2010, 2011, 2012, 2013, 2014, 2015, 2016, 2017, 2019 Nicira, Inc. > * > * Licensed under the Apache License, Version 2.0 (the "License"); > * you may not use this file except in compliance with the License. > @@ -1864,8 +1864,8 @@ group_is_alive(const struct xlate_ctx *ctx, uint32_t group_id, int depth) > #define MAX_LIVENESS_RECURSION 128 /* Arbitrary limit */ > > static bool > -bucket_is_alive(const struct xlate_ctx *ctx, > - struct ofputil_bucket *bucket, int depth) > +bucket_is_alive(const struct xlate_ctx *ctx, const struct group_dpif *group, > + const struct ofputil_bucket *bucket, int depth) > { > if (depth >= MAX_LIVENESS_RECURSION) { > xlate_report_error(ctx, "bucket chaining exceeded %d links", > @@ -1873,6 +1873,12 @@ bucket_is_alive(const struct xlate_ctx *ctx, > return false; > } > > + /* In "select" groups, buckets with weight 0 are not used. > + * In other kinds of groups, weight does not matter. */ > + if (group->up.type == OFPGT11_SELECT && bucket->weight == 0) { > + return false; > + } > + > return (!ofputil_bucket_has_liveness(bucket) > || (bucket->watch_port != OFPP_ANY > && odp_port_is_alive(ctx, bucket->watch_port)) > @@ -1910,7 +1916,7 @@ group_first_live_bucket(const struct xlate_ctx *ctx, > { > struct ofputil_bucket *bucket; > LIST_FOR_EACH (bucket, list_node, &group->up.buckets) { > - if (bucket_is_alive(ctx, bucket, depth)) { > + if (bucket_is_alive(ctx, group, bucket, depth)) { > return bucket; > } > xlate_report_bucket_not_live(ctx, bucket); > @@ -1929,7 +1935,7 @@ group_best_live_bucket(const struct xlate_ctx *ctx, > > struct ofputil_bucket *bucket; > LIST_FOR_EACH (bucket, list_node, &group->up.buckets) { > - if (bucket_is_alive(ctx, bucket, 0)) { > + if (bucket_is_alive(ctx, group, bucket, 0)) { > uint32_t score = > (hash_int(bucket->bucket_id, basis) & 0xffff) * bucket->weight; > if (score >= best_score) { > @@ -4535,7 +4541,7 @@ pick_dp_hash_select_group(struct xlate_ctx *ctx, struct group_dpif *group) > for (int i = 0; i <= hash_mask; i++) { > struct ofputil_bucket *b = > group->hash_map[(dp_hash + i) & hash_mask]; > - if (bucket_is_alive(ctx, b, 0)) { > + if (bucket_is_alive(ctx, group, b, 0)) { > return b; > } > } >
On 9/9/20 20:34, Ilya Maximets wrote: > On 6/8/19 1:28 AM, Ben Pfaff wrote: >> The OpenFlow specification says that buckets in select groups with a weight >> of zero should not be selected, but the ofproto-dpif implementation could >> select them in corner cases. This fixes the problem. >> >> Reported-by: ychen <ychen103103@163.com> >> Reported-at: https://mail.openvswitch.org/pipermail/ovs-dev/2019-May/359349.html >> Signed-off-by: Ben Pfaff <blp@ovn.org> >> --- >> v1->v2: Simplify and fix figuring out whether a group is alive. > > Hi, Ben. I'm looking through old patches after the patchwork cleanup and this > one seems to be still valid. It requires a minor rebase, but it's trivial. > > Patch looks good to me in general and I could apply it. > What do you think? > > It also looks like we will need to backport it. Suggestions on how far it > should go are welcome. Third year's a charm. :) Looking again through the oldest patches in a patchwork, this still seems correct to me. So, applied now. Also backported down to 2.17. Best regards, Ilya Maximets. > > Best regards, Ilya Maximets. > >> >> ofproto/ofproto-dpif-xlate.c | 18 ++++++++++++------ >> 1 file changed, 12 insertions(+), 6 deletions(-) >> >> diff --git a/ofproto/ofproto-dpif-xlate.c b/ofproto/ofproto-dpif-xlate.c >> index 04d69ed06c20..1ace92d22019 100644 >> --- a/ofproto/ofproto-dpif-xlate.c >> +++ b/ofproto/ofproto-dpif-xlate.c >> @@ -1,4 +1,4 @@ >> -/* Copyright (c) 2009, 2010, 2011, 2012, 2013, 2014, 2015, 2016, 2017 Nicira, Inc. >> +/* Copyright (c) 2009, 2010, 2011, 2012, 2013, 2014, 2015, 2016, 2017, 2019 Nicira, Inc. >> * >> * Licensed under the Apache License, Version 2.0 (the "License"); >> * you may not use this file except in compliance with the License. >> @@ -1864,8 +1864,8 @@ group_is_alive(const struct xlate_ctx *ctx, uint32_t group_id, int depth) >> #define MAX_LIVENESS_RECURSION 128 /* Arbitrary limit */ >> >> static bool >> -bucket_is_alive(const struct xlate_ctx *ctx, >> - struct ofputil_bucket *bucket, int depth) >> +bucket_is_alive(const struct xlate_ctx *ctx, const struct group_dpif *group, >> + const struct ofputil_bucket *bucket, int depth) >> { >> if (depth >= MAX_LIVENESS_RECURSION) { >> xlate_report_error(ctx, "bucket chaining exceeded %d links", >> @@ -1873,6 +1873,12 @@ bucket_is_alive(const struct xlate_ctx *ctx, >> return false; >> } >> >> + /* In "select" groups, buckets with weight 0 are not used. >> + * In other kinds of groups, weight does not matter. */ >> + if (group->up.type == OFPGT11_SELECT && bucket->weight == 0) { >> + return false; >> + } >> + >> return (!ofputil_bucket_has_liveness(bucket) >> || (bucket->watch_port != OFPP_ANY >> && odp_port_is_alive(ctx, bucket->watch_port)) >> @@ -1910,7 +1916,7 @@ group_first_live_bucket(const struct xlate_ctx *ctx, >> { >> struct ofputil_bucket *bucket; >> LIST_FOR_EACH (bucket, list_node, &group->up.buckets) { >> - if (bucket_is_alive(ctx, bucket, depth)) { >> + if (bucket_is_alive(ctx, group, bucket, depth)) { >> return bucket; >> } >> xlate_report_bucket_not_live(ctx, bucket); >> @@ -1929,7 +1935,7 @@ group_best_live_bucket(const struct xlate_ctx *ctx, >> >> struct ofputil_bucket *bucket; >> LIST_FOR_EACH (bucket, list_node, &group->up.buckets) { >> - if (bucket_is_alive(ctx, bucket, 0)) { >> + if (bucket_is_alive(ctx, group, bucket, 0)) { >> uint32_t score = >> (hash_int(bucket->bucket_id, basis) & 0xffff) * bucket->weight; >> if (score >= best_score) { >> @@ -4535,7 +4541,7 @@ pick_dp_hash_select_group(struct xlate_ctx *ctx, struct group_dpif *group) >> for (int i = 0; i <= hash_mask; i++) { >> struct ofputil_bucket *b = >> group->hash_map[(dp_hash + i) & hash_mask]; >> - if (bucket_is_alive(ctx, b, 0)) { >> + if (bucket_is_alive(ctx, group, b, 0)) { >> return b; >> } >> } >> >
diff --git a/ofproto/ofproto-dpif-xlate.c b/ofproto/ofproto-dpif-xlate.c index 04d69ed06c20..1ace92d22019 100644 --- a/ofproto/ofproto-dpif-xlate.c +++ b/ofproto/ofproto-dpif-xlate.c @@ -1,4 +1,4 @@ -/* Copyright (c) 2009, 2010, 2011, 2012, 2013, 2014, 2015, 2016, 2017 Nicira, Inc. +/* Copyright (c) 2009, 2010, 2011, 2012, 2013, 2014, 2015, 2016, 2017, 2019 Nicira, Inc. * * Licensed under the Apache License, Version 2.0 (the "License"); * you may not use this file except in compliance with the License. @@ -1864,8 +1864,8 @@ group_is_alive(const struct xlate_ctx *ctx, uint32_t group_id, int depth) #define MAX_LIVENESS_RECURSION 128 /* Arbitrary limit */ static bool -bucket_is_alive(const struct xlate_ctx *ctx, - struct ofputil_bucket *bucket, int depth) +bucket_is_alive(const struct xlate_ctx *ctx, const struct group_dpif *group, + const struct ofputil_bucket *bucket, int depth) { if (depth >= MAX_LIVENESS_RECURSION) { xlate_report_error(ctx, "bucket chaining exceeded %d links", @@ -1873,6 +1873,12 @@ bucket_is_alive(const struct xlate_ctx *ctx, return false; } + /* In "select" groups, buckets with weight 0 are not used. + * In other kinds of groups, weight does not matter. */ + if (group->up.type == OFPGT11_SELECT && bucket->weight == 0) { + return false; + } + return (!ofputil_bucket_has_liveness(bucket) || (bucket->watch_port != OFPP_ANY && odp_port_is_alive(ctx, bucket->watch_port)) @@ -1910,7 +1916,7 @@ group_first_live_bucket(const struct xlate_ctx *ctx, { struct ofputil_bucket *bucket; LIST_FOR_EACH (bucket, list_node, &group->up.buckets) { - if (bucket_is_alive(ctx, bucket, depth)) { + if (bucket_is_alive(ctx, group, bucket, depth)) { return bucket; } xlate_report_bucket_not_live(ctx, bucket); @@ -1929,7 +1935,7 @@ group_best_live_bucket(const struct xlate_ctx *ctx, struct ofputil_bucket *bucket; LIST_FOR_EACH (bucket, list_node, &group->up.buckets) { - if (bucket_is_alive(ctx, bucket, 0)) { + if (bucket_is_alive(ctx, group, bucket, 0)) { uint32_t score = (hash_int(bucket->bucket_id, basis) & 0xffff) * bucket->weight; if (score >= best_score) { @@ -4535,7 +4541,7 @@ pick_dp_hash_select_group(struct xlate_ctx *ctx, struct group_dpif *group) for (int i = 0; i <= hash_mask; i++) { struct ofputil_bucket *b = group->hash_map[(dp_hash + i) & hash_mask]; - if (bucket_is_alive(ctx, b, 0)) { + if (bucket_is_alive(ctx, group, b, 0)) { return b; } }
The OpenFlow specification says that buckets in select groups with a weight of zero should not be selected, but the ofproto-dpif implementation could select them in corner cases. This fixes the problem. Reported-by: ychen <ychen103103@163.com> Reported-at: https://mail.openvswitch.org/pipermail/ovs-dev/2019-May/359349.html Signed-off-by: Ben Pfaff <blp@ovn.org> --- v1->v2: Simplify and fix figuring out whether a group is alive. ofproto/ofproto-dpif-xlate.c | 18 ++++++++++++------ 1 file changed, 12 insertions(+), 6 deletions(-)