Merge pull request #15250 from opensourcerouting/fix/memory_optimizations

bgpd: Some memory optimizations

Merge pull request #15252 from mjstapp/fix_bitfield_init

lib: init new bitfield word after realloc

Merge pull request #15238 from louis-6wind/bgp-leak-network

bgpd: fix VRF leaking with 'network import-check'

bgpd: Optimize memory usage for bgp_damp_config struct

```
struct bgp_damp_config {
        unsigned int               suppress_value;       /*     0     4 */
        unsigned int               reuse_limit;          /*     4     4 */
        time_t                     max_suppress_time;    /*     8     8 */
        time_t                     half_life;            /*    16     8 */
        unsigned int               reuse_list_size;      /*    24     4 */
        unsigned int               reuse_index_size;     /*    28     4 */
        unsigned int               ceiling;              /*    32     4 */
        unsigned int               decay_rate_per_tick;  /*    36     4 */
        unsigned int               decay_array_size;     /*    40     4 */
        unsigned int               reuse_scale_factor;   /*    44     4 */
        double                     scale_factor;         /*    48     8 */
        double *                   decay_array;          /*    56     8 */
        /* --- cacheline 1 boundary (64 bytes) --- */
        int *                      reuse_index;          /*    64     8 */
        struct bgp_damp_info * *   reuse_list;           /*    72     8 */
        int                        reuse_offset;         /*    80     4 */
        safi_t                     safi;                 /*    84     4 */
        struct bgp_damp_info *     no_reuse_list;        /*    88     8 */
        struct event *             t_reuse;              /*    96     8 */
        afi_t                      afi;                  /*   104     4 */

        /* size: 112, cachelines: 2, members: 19 */
        /* padding: 4 */
        /* last cacheline: 48 bytes */
};   /* saved 8 bytes! */
```

Signed-off-by: Donatas Abraitis <donatas@opensourcerouting.org>

bgpd: Optimize memory usage for bgp_aggregate struct

```
struct bgp_aggregate {
        uint8_t                    summary_only;         /*     0     1 */
        uint8_t                    as_set;               /*     1     1 */
        uint8_t                    origin;               /*     2     1 */
        _Bool                      med_mismatched;       /*     3     1 */
        _Bool                      med_initialized;      /*     4     1 */
        _Bool                      match_med;            /*     5     1 */

        /* XXX 2 bytes hole, try to pack */

        struct {
                char *             name;                 /*     8     8 */
                struct route_map * map;                  /*    16     8 */
        } rmap;                                          /*     8    16 */
        long unsigned int          count;                /*    24     8 */
        long unsigned int          incomplete_origin_count; /*    32     8 */
        long unsigned int          egp_origin_count;     /*    40     8 */
        struct hash *              community_hash;       /*    48     8 */
        struct hash *              ecommunity_hash;      /*    56     8 */
        /* --- cacheline 1 boundary (64 bytes) --- */
        struct hash *              lcommunity_hash;      /*    64     8 */
        struct hash *              aspath_hash;          /*    72     8 */
        struct community *         community;            /*    80     8 */
        struct ecommunity *        ecommunity;           /*    88     8 */
        struct lcommunity *        lcommunity;           /*    96     8 */
        struct aspath *            aspath;               /*   104     8 */
        safi_t                     safi;                 /*   112     4 */
        uint32_t                   med_matched_value;    /*   116     4 */
        char *                     suppress_map_name;    /*   120     8 */
        /* --- cacheline 2 boundary (128 bytes) --- */
        struct route_map *         suppress_map;         /*   128     8 */

        /* size: 136, cachelines: 3, members: 22 */
        /* sum members: 134, holes: 1, sum holes: 2 */
        /* last cacheline: 8 bytes */
};
```

Signed-off-by: Donatas Abraitis <donatas@opensourcerouting.org>

bgpd: Optimize memory usage for bgp_static struct

```
struct bgp_static {
        int                        backdoor;             /*     0     4 */
        uint32_t                   label_index;          /*     4     4 */
        uint8_t                    valid;                /*     8     1 */

        /* XXX 1 byte hole, try to pack */

        uint16_t                   encap_tunneltype;     /*    10     2 */
        uint32_t                   igpmetric;            /*    12     4 */
        struct in_addr             igpnexthop;           /*    16     4 */
        uint32_t                   atomic;               /*    20     4 */
        struct {
                char *             name;                 /*    24     8 */
                struct route_map * map;                  /*    32     8 */
        } rmap;                                          /*    24    16 */
        struct prefix_rd           prd __attribute__((__aligned__(8))); /*    40    16 */
        char *                     prd_pretty;           /*    56     8 */
        /* --- cacheline 1 boundary (64 bytes) --- */
        mpls_label_t               label;                /*    64     4 */

        /* XXX 4 bytes hole, try to pack */

        esi_t *                    eth_s_id;             /*    72     8 */
        struct ethaddr *           router_mac;           /*    80     8 */
        struct prefix              gatewayIp __attribute__((__aligned__(8))); /*    88    56 */

        /* size: 144, cachelines: 3, members: 14 */
        /* sum members: 139, holes: 2, sum holes: 5 */
        /* forced alignments: 2 */
        /* last cacheline: 16 bytes */
} __attribute__((__aligned__(8)));   /* saved 8 bytes! */
```

Signed-off-by: Donatas Abraitis <donatas@opensourcerouting.org>

bgpd: Optimize memory usage for attr struct

```
struct attr {
struct aspath *            aspath;               /*     0     8 */
struct community *         community;            /*     8     8 */
long unsigned int          refcnt;               /*    16     8 */
_uint64_t                  flag;                 /*    24     8 */
struct in_addr             nexthop;              /*    32     4 */
uint32_t                   med;                  /*    36     4 */
uint32_t                   local_pref;           /*    40     4 */
ifindex_t                  nh_ifindex;           /*    44     4 */
uint8_t                    origin;               /*    48     1 */
uint8_t                    es_flags;             /*    49     1 */
uint8_t                    router_flag;          /*    50     1 */
uint8_t                    default_gw;           /*    51     1 */
enum pta_type              pmsi_tnl_type;        /*    52     4 */
uint32_t                   rmap_change_flags;    /*    56     4 */
struct in6_addr            mp_nexthop_global;    /*    60    16 */
/* --- cacheline 1 boundary (64 bytes) was 12 bytes ago --- */
struct in6_addr            mp_nexthop_local;     /*    76    16 */
ifindex_t                  nh_lla_ifindex;       /*    92     4 */
struct ecommunity *        ecommunity;           /*    96     8 */
struct ecommunity *        ipv6_ecommunity;      /*   104     8 */
struct lcommunity *        lcommunity;           /*   112     8 */
struct cluster_list *      cluster1;             /*   120     8 */
/* --- cacheline 2 boundary (128 bytes) --- */
struct transit *           transit;              /*   128     8 */
struct in_addr             mp_nexthop_global_in; /*   136     4 */
struct in_addr             aggregator_addr;      /*   140     4 */
struct in_addr             originator_id;        /*   144     4 */
uint32_t                   weight;               /*   148     4 */
as_t                       aggregator_as;        /*   152     4 */
uint8_t                    mp_nexthop_len;       /*   156     1 */
uint8_t                    mp_nexthop_prefer_global; /*   157     1 */
uint8_t                    sticky;               /*   158     1 */
uint8_t                    distance;             /*   159     1 */
uint16_t                   encap_tunneltype;     /*   160     2 */
uint8_t                    df_alg;               /*   162     1 */

/* XXX 1 byte hole, try to pack */

route_tag_t                tag;                  /*   164     4 */
uint32_t                   label_index;          /*   168     4 */
mpls_label_t               label;                /*   172     4 */
struct bgp_attr_srv6_vpn * srv6_vpn;             /*   176     8 */
struct bgp_attr_srv6_l3vpn * srv6_l3vpn;         /*   184     8 */
/* --- cacheline 3 boundary (192 bytes) --- */
struct bgp_attr_encap_subtlv * encap_subtlvs;    /*   192     8 */
struct bgp_attr_encap_subtlv * vnc_subtlvs;      /*   200     8 */
struct bgp_route_evpn      evpn_overlay;         /*   208    36 */
uint32_t                   mm_seqnum;            /*   244     4 */
uint32_t                   mm_sync_seqnum;       /*   248     4 */
struct ethaddr             rmac;                 /*   252     6 */
/* --- cacheline 4 boundary (256 bytes) was 2 bytes ago --- */
uint16_t                   df_pref;              /*   258     2 */
uint32_t                   rmap_table_id;        /*   260     4 */
uint32_t                   link_bw;              /*   264     4 */
esi_t                      esi;                  /*   268    10 */

/* XXX 2 bytes hole, try to pack */

uint32_t                   srte_color;           /*   280     4 */
uint32_t                   otc;                  /*   284     4 */
enum nexthop_types_t       nh_type;              /*   288     4 */
enum blackhole_type        bh_type;              /*   292     4 */
_uint64_t                  aigp_metric;          /*   296     8 */

/* size: 304, cachelines: 5, members: 53 */
/* sum members: 301, holes: 2, sum holes: 3 */
/* last cacheline: 48 bytes */
};   /* saved 16 bytes! */
```

Signed-off-by: Donatas Abraitis <donatas@opensourcerouting.org>

Merge pull request #15248 from LabNConsulting/chopps/fix-comp-warning

zebra: fix compiler warning about truncation.

lib: init new bitfield word after realloc

Realloc doesn't init - ensure a newly-allocated word is inited.

Signed-off-by: Mark Stapp <mjs@labn.net>

bgpd: Optimize memory usage for bgp_nlri struct

```
struct bgp_nlri {
uint16_t                   afi;                  /*     0     2 */
uint8_t                    safi;                 /*     2     1 */

/* XXX 1 byte hole, try to pack */

bgp_size_t                 length;               /*     4     2 */

/* XXX 2 bytes hole, try to pack */

uint8_t *                  nlri;                 /*     8     8 */

/* size: 16, cachelines: 1, members: 4 */
/* sum members: 13, holes: 2, sum holes: 3 */
/* last cacheline: 16 bytes */
};   /* saved 8 bytes! */
```

Signed-off-by: Donatas Abraitis <donatas@opensourcerouting.org>

zebra: fix compiler warning about truncation.

Signed-off-by: Christian Hopps <chopps@labn.net>

bgpd: Optimize memory usage for bgp_notify struct

```
struct bgp_notify {
uint8_t                    code;                 /*     0     1 */
uint8_t                    subcode;              /*     1     1 */
bgp_size_t                 length;               /*     2     2 */
_Bool                      hard_reset;           /*     4     1 */

/* XXX 3 bytes hole, try to pack */

char *                     data;                 /*     8     8 */
uint8_t *                  raw_data;             /*    16     8 */

/* size: 24, cachelines: 1, members: 6 */
/* sum members: 21, holes: 1, sum holes: 3 */
/* last cacheline: 24 bytes */
};   /* saved 16 bytes! */
```

Signed-off-by: Donatas Abraitis <donatas@opensourcerouting.org>

Merge pull request #15181 from idryzhov/zebra-mgmtd

zebra interfaces NB conversion

topotests: vpnv4 route leaking with import-check

Test vpnv4 route leaking with import-check

Signed-off-by: Louis Scalbert <louis.scalbert@6wind.com>

bgpd: fix VRF leaking with 'network import-check' (4/4)

The following configuration creates an infinite routing leaking loop
because 'rt vpn both' parameters are the same in both VRFs.

> router bgp 5227 vrf r1-cust4
>    no bgp network import-check
>    bgp router-id 192.168.1.1
>    address-family ipv4 unicast
>      network 28.0.0.0/24
>      rd vpn export 10:12
>      rt vpn both 52:100
>      import vpn
>      export vpn
>    exit-address-family
> !
> router bgp 5227 vrf r1-cust5
>    no bgp network import-check
>    bgp router id 192.168.1.1
>    address-family ipv4 unicast
>      network 29.0.0.0/24
>      rd vpn export 10:13
>      rt vpn both 52:100
>      import vpn
>      export vpn
>    exit-address-family

The previous commit has added a routing leak update when a nexthop
update is received from zebra. It indirectly calls
bgp_find_or_add_nexthop() in which a static route triggers a nexthop
cache entry registration that triggers a nexthop update from zebra.

Do not register again the nexthop cache entry if the BGP_STATIC_ROUTE is
already set.

Signed-off-by: Louis Scalbert <louis.scalbert@6wind.com>

bgpd: fix VRF leaking with 'network import-check' (3/4)

If 'bgp network import-check' is defined on the source BGP session,
prefixes that are defined with the network command cannot be leaked to
the other VRFs BGP table even if they are present in the origin VRF RIB
if the 'rt import' statement is defined after the 'network <prefix>'
ones.

When a prefix nexthop is updated, update the prefix route leaking. The
current state of nexthop validation is now stored in the attributes of
the bgp path info. Attributes are compared with the previous ones at
route leaking update so that a nexthop validation change now triggers
the update of destination VRF BGP table.

Signed-off-by: Louis Scalbert <louis.scalbert@6wind.com>

bgpd: fix VRF leaking with 'network import-check' (2/4)

"if not XX else" statements are confusing.

Replace two "if not XX else" statements by "if XX else" to prepare next
commits. The patch is only cosmetic.

Signed-off-by: Louis Scalbert <louis.scalbert@6wind.com>

bgpd: fix VRF leaking with 'network import-check' (1/4)

If 'bgp network import-check' is defined on the source BGP session,
prefixes that are defined with the network command cannot be leaked to
the other VRFs BGP table even if they are present in the origin VRF RIB.

Always validate the nexthop of BGP static routes (i.e. defined with the
network statement) if 'network import-check' is defined on the source
BGP session and the prefix is present in source RIB.

Signed-off-by: Louis Scalbert <louis.scalbert@6wind.com>

Merge pull request #15245 from idryzhov/disable-irdp

zebra, build: disable irdp by default

zebra, build: disable irdp by default

IRDP client (rdisc) was deleted from iputils more than 2 years ago. It's
time to drop IRDP, but first let's stop building and including it in the
packages by default to see if anyone will be complaining.

Signed-off-by: Igor Ryzhov <iryzhov@nfware.com>

lib, mgmtd, zebra: cleanup of zebra conversion to mgmtd

- use `apply_finish` callback when possible to avoid multiple applies per commit
- move table range working to the CLI handler
- remove unnecessary conditional compilation
- remove unnecessary boolean conversion

Signed-off-by: Igor Ryzhov <iryzhov@nfware.com>

zebra: convert to mgmtd

Signed-off-by: Igor Ryzhov <iryzhov@nfware.com>

build: remove mgmtd exception from xref2vtysh

We may actually need to send CLI commands to mgmtd and another daemon at
the same time, for example, if this daemon is not mgmtd-converted. The
only daemon this exception protects is staticd. But we don't actually
need any configuration commands in staticd, so just remove the exception
and don't install unnecessary commands to staticd.

Signed-off-by: Igor Ryzhov <iryzhov@nfware.com>

lib: use shared candidate between vty and mgmtd client

When the daemon is partially mgmtd-converted, it receives configuration
from vty and mgmtmd simultaneosly. This configuration must be
synchronized.

Signed-off-by: Igor Ryzhov <iryzhov@nfware.com>

lib: convert affinity-map to mgmtd

Signed-off-by: Igor Ryzhov <iryzhov@nfware.com>

lib: fix yang_dnode_dup

When duplicating nodes, we should always keep flags, especially the
LYD_NEW flag that indicates not validated data. This allows to select a
new choice's case without the need to explicitly remove the existing one.

Signed-off-by: Igor Ryzhov <iryzhov@nfware.com>

zebra: convert vrf configuration output to NB

Signed-off-by: Igor Ryzhov <iryzhov@nfware.com>

zebra: make vrf netns commands do nothing

These commands don't really provide any functionality. VRF is associated
with netns automatically based on its name, and it's not possible to
associate VRF and netns with different names with these commands:

- When trying to assosiate a VRF with an already existing netns with a
  different name:
  `NS /run/netns/test is already configured with VRF 1(test)`

- When trying to assiciate a VRF with a non-existing netns, so they
  become linked once the netns is created:
  `Invalid pathname for /run/netns/test: No such file or directory`

- When doing "no netns" to unlink the netns and link it back to the same
  VRF:
  `VRF 1 is already configured with VRF test`

- When doing "no netns" to unlink the netns and link it to another VRF:
  `Can not associate NS 4294967295 with NETNS /run/netns/test`

As shown above, not a single usecase is working. We can't remove them
completely to preserve backwards-compatibility, so just make them empty.

The main reason for this change is not to spend a lot of time trying to
figure out how to convert them to northbound.

Signed-off-by: Igor Ryzhov <iryzhov@nfware.com>

zebra: convert table range command to NB

Signed-off-by: Igor Ryzhov <iryzhov@nfware.com>

zebra: convert resolve-via-default commands to NB

Signed-off-by: Igor Ryzhov <iryzhov@nfware.com>

zebra: fix vni NB conversion

- unnecessary command duplication
- usage of oper data during validation
- unnecessary checks for things that can't happen

Signed-off-by: Igor Ryzhov <iryzhov@nfware.com>

zebra: convert route-map delay-timer command to NB

Signed-off-by: Igor Ryzhov <iryzhov@nfware.com>

zebra: convert ip nht commands to NB

Signed-off-by: Igor Ryzhov <iryzhov@nfware.com>

zebra: convert ip protocol commands to NB

Signed-off-by: Igor Ryzhov <iryzhov@nfware.com>

zebra: convert router-id commands to NB

Signed-off-by: Igor Ryzhov <iryzhov@nfware.com>

zebra: convert interface configuration output to NB

Signed-off-by: Igor Ryzhov <iryzhov@nfware.com>

zebra: fix build with --disable-rtadv

Signed-off-by: Igor Ryzhov <iryzhov@nfware.com>

zebra: convert PTM configuration to NB

Signed-off-by: Igor Ryzhov <iryzhov@nfware.com>

lib: add support for "features" when loading YANG modules

Signed-off-by: Igor Ryzhov <iryzhov@nfware.com>

zebra: convert interface ipv6 nd dnssl command to NB

Signed-off-by: Igor Ryzhov <iryzhov@nfware.com>

zebra: convert interface ipv6 nd rdnss command to NB

Signed-off-by: Igor Ryzhov <iryzhov@nfware.com>

zebra: convert interface ipv6 nd prefix command to NB

Signed-off-by: Igor Ryzhov <iryzhov@nfware.com>

zebra: convert interface ipv6 nd router-preference command to NB

Signed-off-by: Igor Ryzhov <iryzhov@nfware.com>

zebra: convert interface ipv6 nd home-agent-lifetime command to NB

Signed-off-by: Igor Ryzhov <iryzhov@nfware.com>

zebra: convert interface ipv6 nd home-agent-preference command to NB

Signed-off-by: Igor Ryzhov <iryzhov@nfware.com>

zebra: convert interface ipv6 nd adv-interval-option command to NB

Signed-off-by: Igor Ryzhov <iryzhov@nfware.com>

zebra: convert interface ipv6 nd home-agent-config-flag command to NB

Signed-off-by: Igor Ryzhov <iryzhov@nfware.com>

zebra: convert interface ipv6 nd ra-fast-retrans command to NB

Signed-off-by: Igor Ryzhov <iryzhov@nfware.com>

zebra: convert interface ipv6 nd ra-hop-limit command to NB

Signed-off-by: Igor Ryzhov <iryzhov@nfware.com>

zebra: convert interface ipv6 nd ra-retrans-interval command to NB

Signed-off-by: Igor Ryzhov <iryzhov@nfware.com>

zebra: convert interface ipv6 nd mtu command to NB

Signed-off-by: Igor Ryzhov <iryzhov@nfware.com>

zebra: convert interface ipv6 nd other-config-flag command to NB

Signed-off-by: Igor Ryzhov <iryzhov@nfware.com>

zebra: convert interface ipv6 nd managed-config-flag command to NB

Signed-off-by: Igor Ryzhov <iryzhov@nfware.com>

zebra: convert interface ipv6 nd reachable-time command to NB

Signed-off-by: Igor Ryzhov <iryzhov@nfware.com>

zebra: convert interface ipv6 nd ra interval/lifetime commands to NB

Signed-off-by: Igor Ryzhov <iryzhov@nfware.com>

zebra: convert interface ipv6 nd suppress-ra command to NB

Signed-off-by: Igor Ryzhov <iryzhov@nfware.com>

zebra: convert interface evpn mh uplink command to NB

Signed-off-by: Igor Ryzhov <iryzhov@nfware.com>

zebra: convert interface evpn mh bypass command to NB

Signed-off-by: Igor Ryzhov <iryzhov@nfware.com>

zebra: convert interface evpn mh es-df-pref command to NB

Signed-off-by: Igor Ryzhov <iryzhov@nfware.com>

zebra: convert interface evpn mh es-id commands to NB

Signed-off-by: Igor Ryzhov <iryzhov@nfware.com>

zebra: convert interface link-params packet-loss command to NB

Signed-off-by: Igor Ryzhov <iryzhov@nfware.com>

lib: fix conversion of yang decimal64 to double

Current denominators are not integers and some of them lose precision
because of that, for example, 1e-6 is actually stored as
9.9999999999999995e-07 and 1-e12 is stored as 9.9999999999999998e-13.
When multiplying by such denominators, we receive incorrect values.
Changing denominators to integers and using division instead of
multiplication improves precision and solves the problem.

Signed-off-by: Igor Ryzhov <iryzhov@nfware.com>

zebra: convert interface link-params delay commands to NB

Signed-off-by: Igor Ryzhov <iryzhov@nfware.com>

zebra: convert interface link-params neighbor command to NB

Signed-off-by: Igor Ryzhov <iryzhov@nfware.com>

zebra: convert interface link-params bandwidth commands to NB

Signed-off-by: Igor Ryzhov <iryzhov@nfware.com>

lib: don't set LP_EXTEND_ADM_GRP by default

It shouldn't be set unless some affinity is configured. NB callbacks
set this flag correctly when necessary.

Signed-off-by: Igor Ryzhov <iryzhov@nfware.com>

zebra: convert interface link-params metric command to NB

Signed-off-by: Igor Ryzhov <iryzhov@nfware.com>

zebra: convert interface link-params command to NB

Make link-params a presence container and activate it when entering the
node. The "enable" command is not necessary anymore but kept hidden for
backwards compatibility.

Signed-off-by: Igor Ryzhov <iryzhov@nfware.com>

zebra: mark interface mpls command as yang

Signed-off-by: Igor Ryzhov <iryzhov@nfware.com>

zebra: convert interface ip/ipv6 address commands to NB

Signed-off-by: Igor Ryzhov <iryzhov@nfware.com>

zebra: convert interface bandwidth command to NB

Signed-off-by: Igor Ryzhov <iryzhov@nfware.com>

zebra: convert interface shutdown command to NB

Replace "shutdown" leaf with "enabled" leaf in frr-zebra YANG module
to make it in line with standard YANG models.

Signed-off-by: Igor Ryzhov <iryzhov@nfware.com>

zebra: convert interface link-detect command to NB

Signed-off-by: Igor Ryzhov <iryzhov@nfware.com>

zebra: convert inteface multicast command to NB

Introduce new "[no] multicast <enable|disable>" command to be able to
remove the configuration. Current "[no] multicast" command cannot be
removed. Current command is hidden but still works for backwards
compatibility.

Signed-off-by: Igor Ryzhov <iryzhov@nfware.com>

tests: fix test_bgp_recursive_route_ebgp_multi_hop

We shouldn't pass "lo" as an interface to create_interface_in_kernel.
It always exists, and this function flushes all configured IP addresses
from it.

Signed-off-by: Igor Ryzhov <iryzhov@nfware.com>

Merge pull request #15243 from opensourcerouting/autoconf-dir-mess

*: fix a pile of directory and/or state retention related issues

Merge pull request #15236 from louis-6wind/fix-label-cmp

bgpd: fix mpls label pointer comparison

Merge pull request #15242 from LabNConsulting/chopps/doc-dev-nb-cleanup

Cleanup the auto-converted northbound doc a little

Merge pull request #15244 from idryzhov/mgmtd-fixes

A couple of mgmtd fixes

mgmtd: fix ripngd oper-data xpaths

Signed-off-by: Igor Ryzhov <iryzhov@nfware.com>

mgmtd: fix get-data transaction hangup

When no data requests were sent to the backends, return immediately,
instead of waiting for a timeout. This can happen if backends providing
the requested data are not connected to mgmtd.

Signed-off-by: Igor Ryzhov <iryzhov@nfware.com>

build: homologize path handling

Use consistent `e_somepath` names for expanded versions of `somepath`.
Also remove all paths from `config.h` and put them into
`lib/config_paths.h` - this is to make more obvious when someone is
doing something probably not quite properly structured.

Signed-off-by: David Lamparter <equinox@opensourcerouting.org>

babeld: remove bogus config path print

This doesn't align with what the code actually loads, remove it.

Signed-off-by: David Lamparter <equinox@opensourcerouting.org>

lib, mgmtd: fix commit history location

Both of these belong in `/var/lib`, not `/var/run`.

Rather hilariously, the history read in
`mgmt_history_read_cmt_record_index` was always failing, because it was
doing a `file_exists(MGMTD_COMMIT_FILE_PATH)` check.  Which is the wrong
macro - it's `.../commit-%s.json`, including the unprocessed `%s`, which
would never exist.

I guess noone ever tried if this actually works.  Cool.

On the plus side, this means I don't have to implement legacy
compatibility for this, since it never worked to begin with.

(SQLite3 DB location is also changed in this commit since it also uses
`DAEMON_DB_DIR`.)

Signed-off-by: David Lamparter <equinox@opensourcerouting.org>

ospf6d: fix GR & auth seqno state location

Unfortunately, `ospf6d` is much worse than `ospfd` and `isisd` regarding
its state saving, due to the existence of the auth trailer code.

Again, this belongs in `/var/lib`, not `/var/run`.

Merge both state files into one, and add reconciliation code for the
auth seqno.

I'm gonna save my comment on the fact that `ospf6_auth_seqno_nvm_delete`
is not in fact used anywhere.  Which is now a warning because it's
`static`.  Well.  It probably should be used somewhere, so leave it in.

Signed-off-by: David Lamparter <equinox@opensourcerouting.org>

ospfd: fix GR state location

This belongs in `/var/lib`, not `/var/run`.

Use library facility to load/save, support previous path as fallback,
and do proper fsync().

Signed-off-by: David Lamparter <equinox@opensourcerouting.org>

isisd: fix overload state location

This belongs in `/var/lib`, not `/var/run`.  Also the filename was
typo'd (`isid-restart.json`).

Change to proper location and fall back to previous in case it's the
first restart after an FRR update from a version with the bugged path.

Signed-off-by: David Lamparter <equinox@opensourcerouting.org>

*: fix `frr_daemon_info` indentation

clang-format doesn't understand FRR_DAEMON_INFO is a long macro where
laying out items semantically makes sense.

(Also use only one `FRR_DAEMON_INFO(` in isisd so editors don't get
confused with the mismatching `( ( )`.

Signed-off-by: David Lamparter <equinox@opensourcerouting.org>

lib: create `frr_daemon_state_{load,save}`

These functions load daemon-specific persistent state from
`/var/lib/frr` and supersede open-coded variants of similar calls in
ospfd, ospf6d and isisd to save GR state and/or sequence numbers.

Unlike the open-coded variants, the save call correctly `fsync()`s the
saved data to ensure disk contents are consistent.

Signed-off-by: David Lamparter <equinox@opensourcerouting.org>

lib: set up `frr_libstatedir`

This needs to be used for persistent state, which currently is misplaced
into `/var/run` / `/run` where it gets deleted across reboots.

Signed-off-by: David Lamparter <equinox@opensourcerouting.org>

build: nuke `LDPD_SOCKET`

This just unnecessarily complicates things by involving autoconf.

Signed-off-by: David Lamparter <equinox@opensourcerouting.org>

build: nuke `ZEBRA_SERV_PATH`

This just unnecessarily complicates things by involving autoconf.

Signed-off-by: David Lamparter <equinox@opensourcerouting.org>

build: nuke `BFDD_CONTROL_SOCKET`

This just unnecessarily complicates things by involving autoconf.

Signed-off-by: David Lamparter <equinox@opensourcerouting.org>

lib, mgmtd: fix wrong mgmtd socket paths

These paths were ignoring the `-N` namespacing option.

Signed-off-by: David Lamparter <equinox@opensourcerouting.org>

*: rename `frr_vtydir` to `frr_runstatedir`

Also remove frr_init_vtydir(), just initialize to default.

Signed-off-by: David Lamparter <equinox@opensourcerouting.org>

build: update packaging & docs for dir changes

`--sysconfdir` and `--localstatedir` now align with general autoconf
practices.

Signed-off-by: David Lamparter <equinox@opensourcerouting.org>

build: untangle sysconfdir & localstatedir

`--sysconfdir` should be `/etc` and `--localstatedir` should be `/var`.
The package-specific subdirectory should be added by configure, not
given by the user, to match established behavior by other packages.

Note that `--bindir`, `--sbindir`, `--libdir` and `--libexecdir` have
different established/expected behavior due to distro specific
multi-arch support.  That's why these are left unchanged.

The reason this is getting fixed now is that we need to use
`--localstatedir` for its actual value to put things in `/var/lib`.  As
it is now, being overloaded for `/run`, the configured `/var` path
becomes inaccessible.

Signed-off-by: David Lamparter <equinox@opensourcerouting.org>

build: use AX_RECURSIVE_EVAL

Replace inline expansion loop.  (Also correctly handles prefix=NONE and
exec_prefix=NONE inside the macro.)

Signed-off-by: David Lamparter <equinox@opensourcerouting.org>

build: add recursion limit for AX_RECURSIVE_EVAL

The autoconf version can just loop forever, abort instead.

Signed-off-by: David Lamparter <equinox@opensourcerouting.org>

build: add ax_recursive_eval.m4

As found in the GNU autoconf archive
https://git.savannah.gnu.org/gitweb/?p=autoconf-archive.git

Signed-off-by: David Lamparter <equinox@opensourcerouting.org>