On Thu, 9 Jul 2020 00:32:24 +0200 Michal Kubecek wrote:
> On Tue, Jul 07, 2020 at 02:24:29PM -0700, Jakub Kicinski wrote:
> > +
> > +---------------------------------------------+--------+---------------------+
> > + | ``ETHTOOL_A_TUNNEL_INFO_HEADER`` | nested | reply header
> > |
> > +
> > +---------------------------------------------+--------+---------------------+
> > + | ``ETHTOOL_A_TUNNEL_INFO_UDP_PORTS`` | nested | all UDP port
> > tables |
> > +
> > +-+-------------------------------------------+--------+---------------------+
> > + | | ``ETHTOOL_A_TUNNEL_UDP_TABLE`` | nested | one UDP port
> > table |
> > +
> > +-+-+-----------------------------------------+--------+---------------------+
> > + | | | ``ETHTOOL_A_TUNNEL_UDP_TABLE_SIZE`` | u32 | max size of the
> > |
> > + | | | | | table
> > |
> > +
> > +-+-+-----------------------------------------+--------+---------------------+
> > + | | | ``ETHTOOL_A_TUNNEL_UDP_TABLE_TYPES`` | u32 | bitmask of
> > tunnel |
> > + | | | | | types table can
> > hold|
>
> In the code below, this is a bitset, not u32.
I was going back and forth on the type. I'll update the doc to say
bitset, LMK if you prefer u32.
> > +
> > +-+-+-----------------------------------------+--------+---------------------+
> > + | | | ``ETHTOOL_A_TUNNEL_UDP_TABLE_ENTRY`` | nested | offloaded UDP
> > port |
> > +
> > +-+-+-+---------------------------------------+--------+---------------------+
> > + | | | | ``ETHTOOL_A_TUNNEL_UDP_ENTRY_PORT`` | be16 | UDP port
> > |
> > +
> > +-+-+-+---------------------------------------+--------+---------------------+
> > + | | | | ``ETHTOOL_A_TUNNEL_UDP_ENTRY_TYPE`` | u32 | tunnel type
> > |
> > +
> > +-+-+-+---------------------------------------+--------+---------------------+
> > +enum {
> > + ETHTOOL_A_TUNNEL_INFO_UNSPEC,
> > + ETHTOOL_A_TUNNEL_INFO_HEADER, /* nest - _A_HEADER_* */
> > +
> > + ETHTOOL_A_TUNNEL_INFO_UDP_PORTS, /* nest - _UDP_TABLE */
> > +
> > + /* add new constants above here */
> > + __ETHTOOL_A_TUNNEL_INFO_CNT,
> > + ETHTOOL_A_TUNNEL_INFO_MAX = (__ETHTOOL_A_TUNNEL_INFO_CNT - 1)
> > +};
>
> nit: other requests with nested attributes have the enums ordered "from
> inside out", i.e. nested attributes before the nest containing them.
I see! I'll reorder.
> > + ETHTOOL_A_TUNNEL_UDP_TABLE_ENTRY, /* nest - _UDP_ENTRY_*
> > */
> > +
> > + /* add new constants above here */
> > + __ETHTOOL_A_TUNNEL_UDP_TABLE_CNT,
> > + ETHTOOL_A_TUNNEL_UDP_TABLE_MAX = (__ETHTOOL_A_TUNNEL_UDP_TABLE_CNT - 1)
> > +};
> > +
> > +enum {
> > + ETHTOOL_A_TUNNEL_UDP_ENTRY_UNSPEC,
> > +
> > + ETHTOOL_A_TUNNEL_UDP_ENTRY_PORT, /* be16 */
>
> Do we get some benefit from passing the port in network byte order? It
> would be helpful if we expected userspace to copy it e.g. into struct
> sockaddr_in but that doesn't seem to be the case.
I was just following what I believe is a more common pattern.
TC uses be16 AFAIK (flower, act_ct, tunnel), so do the tunnels
(IFLA_GENEVE_PORT, IFLA_VXLAN_PORT). And ethtool flow descriptions.
I can change, but IMHO that'd be more surprising than be16.
> > + ETHTOOL_A_TUNNEL_UDP_ENTRY_TYPE, /* u32 */
> > +
> > + /* add new constants above here */
> > + __ETHTOOL_A_TUNNEL_UDP_ENTRY_CNT,
> > + ETHTOOL_A_TUNNEL_UDP_ENTRY_MAX = (__ETHTOOL_A_TUNNEL_UDP_ENTRY_CNT - 1)
> > +};
> > +
> > +enum {
> > + ETHTOOL_UDP_TUNNEL_TYPE_VXLAN,
> > + ETHTOOL_UDP_TUNNEL_TYPE_GENEVE,
> > + ETHTOOL_UDP_TUNNEL_TYPE_VXLAN_GPE,
> > +
> > + __ETHTOOL_UDP_TUNNEL_TYPE_BIT_CNT
>
> nit: the "BIT" part looks inconsistent (like a leftover form an older
> version where the constants were named differently).
Ah, thanks, missed in rename.
> > +};
> > +
> > /* generic netlink info */
> > #define ETHTOOL_GENL_NAME "ethtool"
> > #define ETHTOOL_GENL_VERSION 1
> [...]
> > diff --git a/net/ethtool/strset.c b/net/ethtool/strset.c
> > index 0eed4e4909ab..5d3f315d4781 100644
> > --- a/net/ethtool/strset.c
> > +++ b/net/ethtool/strset.c
> > @@ -75,6 +75,11 @@ static const struct strset_info info_template[] = {
> > .count = __HWTSTAMP_FILTER_CNT,
> > .strings = ts_rx_filter_names,
> > },
> > + [ETH_SS_UDP_TUNNEL_TYPES] = {
> > + .per_dev = false,
> > + .count = __ETHTOOL_A_TUNNEL_UDP_ENTRY_CNT,
>
> This should be __ETHTOOL_UDP_TUNNEL_TYPE_BIT_CNT (number of strings in
> the set, i.e. number of tunnel types).
:o Ack
> > + .strings = udp_tunnel_type_names,
> > + },
> > };
> >
> > struct strset_req_info {
> > diff --git a/net/ethtool/tunnels.c b/net/ethtool/tunnels.c
> > new file mode 100644
> > index 000000000000..e9e09ea08c6a
> > --- /dev/null
> > +++ b/net/ethtool/tunnels.c
> > @@ -0,0 +1,261 @@
> > +// SPDX-License-Identifier: GPL-2.0-only
> > +
> > +#include <linux/ethtool_netlink.h>
> > +#include <net/udp_tunnel.h>
> > +
> > +#include "bitset.h"
> > +#include "common.h"
> > +#include "netlink.h"
> > +
> > +static const struct nla_policy
> > +ethtool_tunnel_info_policy[ETHTOOL_A_TUNNEL_INFO_MAX + 1] = {
> > + [ETHTOOL_A_TUNNEL_INFO_UNSPEC] = { .type = NLA_REJECT },
> > + [ETHTOOL_A_TUNNEL_INFO_HEADER] = { .type = NLA_NESTED },
> > +};
> > +
> > +static_assert(ETHTOOL_UDP_TUNNEL_TYPE_VXLAN ==
> > ilog2(UDP_TUNNEL_TYPE_VXLAN));
> > +static_assert(ETHTOOL_UDP_TUNNEL_TYPE_GENEVE ==
> > ilog2(UDP_TUNNEL_TYPE_GENEVE));
> > +static_assert(ETHTOOL_UDP_TUNNEL_TYPE_VXLAN_GPE ==
> > + ilog2(UDP_TUNNEL_TYPE_VXLAN_GPE));
> > +
> > +static ssize_t
> > +ethnl_tunnel_info_reply_size(const struct ethnl_req_info *req_base,
> > + struct netlink_ext_ack *extack)
> > +{
> > + bool compact = req_base->flags & ETHTOOL_FLAG_COMPACT_BITSETS;
> > + const struct udp_tunnel_nic_info *info;
> > + unsigned int i;
> > + size_t size;
> > + int ret;
> > +
> > + BUILD_BUG_ON(__ETHTOOL_UDP_TUNNEL_TYPE_BIT_CNT > 32);
> > +
> > + info = req_base->dev->udp_tunnel_nic_info;
> > + if (!info) {
> > + NL_SET_ERR_MSG(extack,
> > + "device does not report tunnel offload info");
> > + return -EOPNOTSUPP;
> > + }
> > +
> > + size = nla_total_size(0); /* _INFO_UDP_PORTS */
> > +
> > + for (i = 0; i < UDP_TUNNEL_NIC_MAX_TABLES; i++) {
> > + if (!info->tables[i].n_entries)
> > + return size;
> > +
> > + size += nla_total_size(0); /* _UDP_TABLE */
> > + size += nla_total_size(sizeof(u32)); /* _UDP_TABLE_SIZE */
> > + ret = ethnl_bitset32_size(&info->tables[i].tunnel_types, NULL,
> > + __ETHTOOL_UDP_TUNNEL_TYPE_BIT_CNT,
> > + udp_tunnel_type_names, compact);
> > + if (ret < 0)
> > + return ret;
> > + size += ret;
> > +
> > + size += udp_tunnel_nic_dump_size(req_base->dev, i);
> > + }
> > +
> > + return size;
> > +}
>
> How big can the message get? Can we be sure the information for one
> device will always fit into a reasonably sized message? Attribute
> ETHTOOL_A_TUNNEL_INFO_UDP_PORTS is limited by 65535 bytes (attribute
> size is u16), can we always fit into this size?
I don't think I've seen any driver with more than 2 tables
or 16 entries total, and they don't seem to be growing in newer
HW (people tend to use standard ports).
188B + 16 * 20B = 508B - so we should be pretty safe with 64k.
