Line data Source code
1 : #include "fd_netlink_tile_private.h"
2 : #include "../topo/fd_topo.h"
3 : #include "../topo/fd_topob.h"
4 : #include "../metrics/fd_metrics.h"
5 : #include "../waker/fd_waker.h"
6 : #include "../../waltz/ip/fd_fib4_netlink.h"
7 : #include "../../waltz/mib/fd_netdev_netlink.h"
8 : #include "../../waltz/neigh/fd_neigh4_netlink.h"
9 : #include "../../util/pod/fd_pod_format.h"
10 : #include "../../util/log/fd_dtrace.h"
11 : #include "fd_netlink_tile.h"
12 :
13 : #include <errno.h>
14 : #include <net/if.h>
15 : #include <netinet/in.h> /* MSG_DONTWAIT */
16 : #include <sys/epoll.h>
17 : #include <sys/socket.h> /* SOL_{...} */
18 : #include <sys/random.h> /* getrandom */
19 : #include <sys/time.h> /* struct timeval */
20 : #include <time.h> /* CLOCK_REALTIME for seccomp filter */
21 : #include <linux/rtnetlink.h> /* RTM_{...} */
22 :
23 : #define FD_SOCKADDR_IN_SZ sizeof(struct sockaddr_in)
24 : #include <linux/futex.h>
25 : #include "generated/netlink_seccomp.h"
26 :
27 : void
28 : fd_netlink_topo_create( fd_topo_tile_t * netlink_tile,
29 : fd_topo_t * topo,
30 : ulong netlnk_max_routes,
31 : ulong netlnk_max_peer_routes,
32 : ulong netlnk_max_neighbors,
33 0 : char const * bind_interface ) {
34 0 : fd_topo_obj_t * netdev_tbl_obj = fd_topob_obj( topo, "netdev_tbl", "netbase" );
35 0 : fd_topo_obj_t * neigh4_obj = fd_topob_obj( topo, "neigh4_hmap", "netbase" );
36 :
37 0 : fd_topob_tile_uses( topo, netlink_tile, netdev_tbl_obj, FD_SHMEM_JOIN_MODE_READ_WRITE );
38 0 : fd_topob_tile_uses( topo, netlink_tile, neigh4_obj, FD_SHMEM_JOIN_MODE_READ_WRITE );
39 :
40 : /* Configure netdev table */
41 0 : FD_TEST( fd_pod_insertf_ulong( topo->props, NETDEV_MAX, "obj.%lu.dev_max", netdev_tbl_obj->id ) );
42 0 : FD_TEST( fd_pod_insertf_ulong( topo->props, BOND_MASTER_MAX, "obj.%lu.bond_max", netdev_tbl_obj->id ) );
43 :
44 0 : netlink_tile->netlink.route_max = netlnk_max_routes;
45 0 : netlink_tile->netlink.route_peer_max = netlnk_max_peer_routes;
46 :
47 : /* Configure neighbor hashmap */
48 0 : FD_TEST( fd_pod_insertf_ulong( topo->props, netlnk_max_neighbors, "obj.%lu.ele_max", neigh4_obj->id ) );
49 0 : FD_TEST( fd_pod_insertf_ulong( topo->props, 16UL, "obj.%lu.probe_max", neigh4_obj->id ) );
50 0 : ulong neigh4_seed;
51 0 : FD_TEST( 8UL==getrandom( &neigh4_seed, sizeof(ulong), 0 ) );
52 0 : FD_TEST( fd_pod_insertf_ulong( topo->props, neigh4_seed, "obj.%lu.seed", neigh4_obj->id ) );
53 :
54 0 : netlink_tile->netlink.netdev_tbl_obj_id = netdev_tbl_obj->id;
55 0 : memcpy( netlink_tile->netlink.neigh_if, bind_interface, sizeof(netlink_tile->netlink.neigh_if) );
56 0 : netlink_tile->netlink.neigh4_obj_id = neigh4_obj->id;
57 0 : }
58 :
59 : void
60 : fd_netlink_topo_join( fd_topo_t * topo,
61 : fd_topo_tile_t * netlink_tile,
62 0 : fd_topo_tile_t * join_tile ) {
63 0 : fd_topob_tile_uses( topo, join_tile, &topo->objs[ netlink_tile->netlink.netdev_tbl_obj_id ], FD_SHMEM_JOIN_MODE_READ_ONLY );
64 0 : fd_topob_tile_uses( topo, join_tile, &topo->objs[ netlink_tile->netlink.neigh4_obj_id ], FD_SHMEM_JOIN_MODE_READ_ONLY );
65 0 : }
66 :
67 : /* Begin tile methods */
68 :
69 : FD_FN_CONST static inline ulong
70 0 : scratch_align( void ) {
71 0 : return alignof(fd_netlink_tile_ctx_t);
72 0 : }
73 :
74 : FD_FN_PURE static inline ulong
75 0 : scratch_footprint( fd_topo_tile_t const * tile ) {
76 0 : (void)tile;
77 0 : ulong l = FD_LAYOUT_INIT;
78 0 : l = FD_LAYOUT_APPEND( l, alignof(fd_netlink_tile_ctx_t), sizeof(fd_netlink_tile_ctx_t) );
79 0 : return FD_LAYOUT_FINI( l, scratch_align() );
80 0 : }
81 :
82 : static ulong
83 : populate_allowed_seccomp( fd_topo_t const * topo,
84 : fd_topo_tile_t const * tile,
85 : ulong out_cnt,
86 0 : struct sock_filter * out ) {
87 0 : fd_netlink_tile_ctx_t * ctx = fd_topo_obj_laddr( topo, tile->tile_obj_id );
88 0 : FD_TEST( ctx->magic==FD_NETLINK_TILE_CTX_MAGIC );
89 0 : uint epoll_inner_fd = (uint)FD_WAKER_INNER_FD( tile->waker_client_idx );
90 0 : uint epoll_outer_fd = (uint)FD_WAKER_OUTER_FD;
91 :
92 0 : populate_sock_filter_policy_netlink( out_cnt, out, (uint)fd_log_private_logfile_fd(), (uint)ctx->nl_monitor->fd, (uint)ctx->nl_req->fd, (uint)ctx->prober->sock_fd, epoll_inner_fd, epoll_outer_fd );
93 0 : return sock_filter_policy_netlink_instr_cnt;
94 0 : }
95 :
96 : static ulong
97 : populate_allowed_fds( fd_topo_t const * topo,
98 : fd_topo_tile_t const * tile,
99 : ulong out_fds_cnt,
100 0 : int * out_fds ) {
101 0 : fd_netlink_tile_ctx_t * ctx = fd_topo_obj_laddr( topo, tile->tile_obj_id );
102 0 : FD_TEST( ctx->magic==FD_NETLINK_TILE_CTX_MAGIC );
103 :
104 0 : if( FD_UNLIKELY( out_fds_cnt<7UL ) ) FD_LOG_ERR(( "out_fds_cnt too low (%lu)", out_fds_cnt ));
105 :
106 0 : ulong out_cnt = 0UL;
107 0 : out_fds[ out_cnt++ ] = 2; /* stderr */
108 0 : if( FD_LIKELY( -1!=fd_log_private_logfile_fd() ) )
109 0 : out_fds[ out_cnt++ ] = fd_log_private_logfile_fd(); /* logfile */
110 0 : out_fds[ out_cnt++ ] = ctx->nl_monitor->fd;
111 0 : out_fds[ out_cnt++ ] = ctx->nl_req->fd;
112 0 : out_fds[ out_cnt++ ] = ctx->prober->sock_fd;
113 0 : out_fds[ out_cnt++ ] = FD_WAKER_OUTER_FD; /* waker outer epoll fd (rearm) */
114 0 : out_fds[ out_cnt++ ] = FD_WAKER_INNER_FD( tile->waker_client_idx ); /* waker inner epoll fd */
115 0 : return out_cnt;
116 0 : }
117 :
118 : static void
119 : privileged_init( fd_topo_t const * topo,
120 0 : fd_topo_tile_t const * tile ) {
121 0 : if( FD_UNLIKELY( tile->kind_id!=0 ) ) {
122 0 : FD_LOG_ERR(( "Topology contains more than one netlink tile" ));
123 0 : }
124 :
125 0 : uint const neigh_if_idx = if_nametoindex( tile->netlink.neigh_if );
126 0 : if( FD_UNLIKELY( !neigh_if_idx ) ) FD_LOG_ERR(( "if_nametoindex(%.16s) failed (%i-%s)", tile->netlink.neigh_if, errno, fd_io_strerror( errno ) ));
127 :
128 0 : fd_netlink_tile_ctx_t * ctx = fd_topo_obj_laddr( topo, tile->tile_obj_id );
129 0 : fd_memset( ctx, 0, sizeof(fd_netlink_tile_ctx_t) );
130 0 : ctx->magic = FD_NETLINK_TILE_CTX_MAGIC;
131 0 : ctx->neigh4_ifidx = neigh_if_idx;
132 :
133 0 : if( FD_UNLIKELY( !fd_netlink_init( ctx->nl_monitor, 1000U ) ) ) {
134 0 : FD_LOG_ERR(( "Failed to connect to rtnetlink" ));
135 0 : }
136 0 : if( FD_UNLIKELY( !fd_netlink_init( ctx->nl_req, 9000000U ) ) ) {
137 0 : FD_LOG_ERR(( "Failed to connect to rtnetlink" ));
138 0 : }
139 :
140 0 : union {
141 0 : struct sockaddr sa;
142 0 : struct sockaddr_nl sanl;
143 0 : } sa;
144 0 : sa.sanl = (struct sockaddr_nl) {
145 0 : .nl_family = AF_NETLINK,
146 0 : .nl_groups = RTMGRP_LINK | RTMGRP_NEIGH | RTMGRP_IPV4_ROUTE
147 0 : };
148 0 : if( FD_UNLIKELY( 0!=bind( ctx->nl_monitor->fd, &sa.sa, sizeof(struct sockaddr_nl) ) ) ) {
149 0 : FD_LOG_ERR(( "bind(sock,RT_NETLINK,RTMGRP_{LINK,NEIGH,IPV4_ROUTE}) failed (%i-%s)", errno, fd_io_strerror( errno ) ));
150 0 : }
151 :
152 0 : float const max_probes_per_second = 3.f;
153 0 : ulong const max_probe_burst = 128UL;
154 0 : float const probe_delay_seconds = 15.f;
155 0 : fd_neigh4_prober_init( ctx->prober, max_probes_per_second, max_probe_burst, probe_delay_seconds );
156 :
157 0 : ctx->waker_client_idx = tile->waker_client_idx;
158 0 : FD_TEST( ctx->waker_client_idx!=ULONG_MAX );
159 :
160 0 : struct epoll_event ev = { .events = EPOLLIN, .data.fd = ctx->nl_monitor->fd };
161 0 : if( FD_UNLIKELY( -1==epoll_ctl( FD_WAKER_INNER_FD( ctx->waker_client_idx ), EPOLL_CTL_ADD, ctx->nl_monitor->fd, &ev ) ) )
162 0 : FD_LOG_ERR(( "epoll_ctl(ADD,nl_monitor) failed (%i-%s)", errno, fd_io_strerror( errno ) ));
163 0 : }
164 :
165 : static void
166 : unprivileged_init( fd_topo_t const * topo,
167 0 : fd_topo_tile_t const * tile ) {
168 0 : FD_SCRATCH_ALLOC_INIT( l, fd_topo_obj_laddr( topo, tile->tile_obj_id ) );
169 0 : fd_netlink_tile_ctx_t * ctx = FD_SCRATCH_ALLOC_APPEND( l, alignof(fd_netlink_tile_ctx_t), sizeof(fd_netlink_tile_ctx_t) );
170 0 : FD_TEST( ctx->magic==FD_NETLINK_TILE_CTX_MAGIC );
171 :
172 0 : FD_TEST( ctx->waker_client_idx!=ULONG_MAX );
173 0 : ctx->waker_fseq = fd_fseq_join( fd_topo_obj_laddr( topo, tile->waker_fseq_obj_id ) );
174 0 : FD_TEST( ctx->waker_fseq );
175 :
176 0 : FD_TEST( tile->netlink.netdev_tbl_obj_id );
177 0 : FD_TEST( tile->netlink.neigh4_obj_id );
178 0 : ctx->route_max = tile->netlink.route_max;
179 0 : ctx->route_peer_max = tile->netlink.route_peer_max;
180 :
181 0 : FD_TEST( fd_netdev_tbl_join( ctx->netdev_tbl, fd_topo_obj_laddr( topo, tile->netlink.netdev_tbl_obj_id ) ) );
182 :
183 0 : ulong neigh4_obj_id = tile->netlink.neigh4_obj_id;
184 0 : ulong neigh_ele_max = fd_pod_queryf_ulong( topo->props, ULONG_MAX, "obj.%lu.ele_max", neigh4_obj_id );
185 0 : ulong neigh_probe_max = fd_pod_queryf_ulong( topo->props, ULONG_MAX, "obj.%lu.probe_max", neigh4_obj_id );
186 0 : ulong neigh_seed = fd_pod_queryf_ulong( topo->props, ULONG_MAX, "obj.%lu.seed", neigh4_obj_id );
187 0 : FD_TEST( neigh_ele_max!=ULONG_MAX && neigh_probe_max!=ULONG_MAX && neigh_seed!=ULONG_MAX );
188 0 : FD_TEST( fd_neigh4_hmap_join( ctx->neigh4, fd_topo_obj_laddr( topo, neigh4_obj_id ), neigh_ele_max, neigh_probe_max, neigh_seed ) );
189 :
190 0 : FD_TEST( tile->out_cnt==1UL );
191 0 : fd_topo_link_t const * out = &topo->links[ tile->out_link_id[0] ];
192 0 : ctx->out_mem = topo->workspaces[ topo->objs[ out->dcache_obj_id ].wksp_id ].wksp;
193 0 : ctx->out_chunk0 = fd_dcache_compact_chunk0( ctx->out_mem, out->dcache );
194 0 : ctx->out_wmark = fd_dcache_compact_wmark( ctx->out_mem, out->dcache, out->mtu );
195 0 : ctx->out_chunk = ctx->out_chunk0;
196 :
197 :
198 0 : for( ulong i=0UL; i<tile->in_cnt; i++ ) {
199 0 : fd_topo_link_t const * link = &topo->links[ tile->in_link_id[ i ] ];
200 0 : if( FD_UNLIKELY( link->mtu!=0UL ) ) FD_LOG_ERR(( "netlink solicit links must have an MTU of zero" ));
201 0 : }
202 :
203 0 : ctx->action |= FD_NET_TILE_ACTION_LINK_UPDATE;
204 0 : ctx->action |= FD_NET_TILE_ACTION_ROUTE4_UPDATE;
205 0 : ctx->action |= FD_NET_TILE_ACTION_NEIGH_UPDATE;
206 :
207 0 : ctx->update_backoff = (long)( fd_tempo_tick_per_ns( NULL ) * 500e6 ); /* 500ms */
208 0 : }
209 :
210 : /* Begin stem methods */
211 :
212 : static inline void
213 0 : metrics_write( fd_netlink_tile_ctx_t * ctx ) {
214 0 : ctx->metrics.neigh_probe[ FD_METRICS_ENUM_NEIGHBOR_PROBE_RESULT_V_HOST_RATE_LIMITED_IDX ] = ctx->prober->local_rate_limited_cnt;
215 0 : ctx->metrics.neigh_probe[ FD_METRICS_ENUM_NEIGHBOR_PROBE_RESULT_V_GLOBAL_RATE_LIMITED_IDX ] = ctx->prober->global_rate_limited_cnt;
216 0 : FD_MCNT_SET( NETLNK, RX_BUFFER_OVERFLOWED, fd_netlink_enobufs_cnt );
217 0 : FD_MCNT_SET( NETLNK, LINK_TABLE_SYNCED, ctx->metrics.link_full_syncs );
218 0 : FD_MCNT_SET( NETLNK, ROUTE_TABLE_SYNCED, ctx->metrics.route_full_syncs );
219 0 : FD_MCNT_ENUM_COPY( NETLNK, UPDATE_PROCESSED, ctx->metrics.update_cnt );
220 0 : FD_MGAUGE_SET( NETLNK, INTERFACE_COUNT, ctx->netdev_tbl->hdr->dev_cnt );
221 0 : FD_MCNT_ENUM_COPY( NETLNK, NEIGHBOR_PROBE_REQUESTED, ctx->metrics.neigh_probe );
222 0 : }
223 :
224 : static void
225 : iproute_publish( fd_netlink_tile_ctx_t * ctx,
226 : fd_stem_context_t * stem,
227 0 : fd_iproute_msg_t const * msg ) {
228 0 : fd_memcpy( fd_chunk_to_laddr( ctx->out_mem, ctx->out_chunk ), msg, sizeof(*msg) );
229 0 : fd_stem_publish( stem, 0UL, 0UL, ctx->out_chunk, sizeof(*msg), 0UL, 0UL, fd_frag_meta_ts_comp( fd_tickcount() ) );
230 0 : ctx->out_chunk = fd_dcache_compact_next( ctx->out_chunk, sizeof(*msg), ctx->out_chunk0, ctx->out_wmark );
231 0 : }
232 :
233 : /* iproute_dump_begin starts a multipart RTM_GETROUTE request. The
234 : response iterator remains in ctx so subsequent calls can consume it
235 : one netlink message at a time. */
236 :
237 : static int
238 : iproute_dump_begin( fd_netlink_tile_ctx_t * ctx,
239 0 : uint table_id ) {
240 0 : uint seq = ctx->nl_req->seq++;
241 0 : struct {
242 0 : struct nlmsghdr nlh;
243 0 : struct rtmsg rtm;
244 0 : struct rtattr rta;
245 0 : uint table_id;
246 0 : } request = {
247 0 : .nlh = {
248 0 : .nlmsg_type = RTM_GETROUTE,
249 0 : .nlmsg_flags = NLM_F_REQUEST | NLM_F_DUMP,
250 0 : .nlmsg_len = sizeof(request),
251 0 : .nlmsg_seq = seq,
252 0 : },
253 0 : .rtm = { .rtm_family = AF_INET },
254 0 : .rta = {
255 0 : .rta_type = RTA_TABLE,
256 0 : .rta_len = RTA_LENGTH( sizeof(uint) ),
257 0 : },
258 0 : .table_id = table_id,
259 0 : };
260 0 : long send_res = sendto( ctx->nl_req->fd, &request, sizeof(request), 0, NULL, 0 );
261 0 : if( FD_UNLIKELY( send_res<0L ) ) {
262 0 : FD_LOG_WARNING(( "netlink send(%d,RTM_GETROUTE,NLM_F_REQUEST|NLM_F_DUMP) failed (%d-%s)", ctx->nl_req->fd, errno, fd_io_strerror( errno ) ));
263 0 : return FD_FIB_NETLINK_ERR_IO;
264 0 : }
265 0 : if( FD_UNLIKELY( send_res!=(long)sizeof(request) ) ) {
266 0 : FD_LOG_WARNING(( "netlink send(%d,RTM_GETROUTE,NLM_F_REQUEST|NLM_F_DUMP) failed (short write)", ctx->nl_req->fd ));
267 0 : return FD_FIB_NETLINK_ERR_IO;
268 0 : }
269 0 : fd_netlink_iter_init( ctx->dump_iter, ctx->nl_req, ctx->dump_buf, sizeof(ctx->dump_buf) );
270 0 : ctx->dump_active = 1;
271 0 : ctx->dump_advance = 0;
272 0 : ctx->dump_intr = 0;
273 0 : return FD_FIB_NETLINK_SUCCESS;
274 0 : }
275 :
276 : #define FD_NETLINK_DUMP_NEXT_DONE (0)
277 0 : #define FD_NETLINK_DUMP_NEXT_ROUTE (1)
278 0 : #define FD_NETLINK_DUMP_NEXT_PROGRESS (2)
279 :
280 : /* iproute_dump_next consumes at most one netlink message. */
281 :
282 : static int
283 : iproute_dump_next( fd_netlink_tile_ctx_t * ctx,
284 0 : fd_iproute_msg_t * route ) {
285 0 : if( ctx->dump_advance ) {
286 0 : fd_netlink_iter_next( ctx->dump_iter, ctx->nl_req );
287 0 : ctx->dump_advance = 0;
288 0 : }
289 0 : if( fd_netlink_iter_done( ctx->dump_iter ) ) {
290 0 : if( ctx->dump_iter->err==0 ) {
291 0 : struct nlmsghdr const * done = fd_netlink_iter_msg( ctx->dump_iter );
292 0 : if( FD_UNLIKELY( done->nlmsg_flags & NLM_F_DUMP_INTR ) ) ctx->dump_intr = 1;
293 0 : }
294 0 : int err = FD_FIB_NETLINK_SUCCESS;
295 0 : if( FD_UNLIKELY( ctx->dump_iter->err>0 ) ) err = FD_FIB_NETLINK_ERR_IO;
296 0 : else if( FD_UNLIKELY( ctx->dump_intr ) ) err = FD_FIB_NETLINK_ERR_INTR;
297 0 : else if( FD_UNLIKELY( ctx->dump_overflow ) ) err = FD_FIB_NETLINK_ERR_SPACE;
298 0 : ctx->dump_active = 0;
299 0 : return -err;
300 0 : }
301 :
302 0 : struct nlmsghdr const * nlh = fd_netlink_iter_msg( ctx->dump_iter );
303 0 : ctx->dump_advance = 1;
304 0 : if( FD_UNLIKELY( nlh->nlmsg_flags & NLM_F_DUMP_INTR ) ) ctx->dump_intr = 1;
305 0 : if( FD_UNLIKELY( nlh->nlmsg_type==NLMSG_ERROR ) ) {
306 0 : struct nlmsgerr const * err = NLMSG_DATA( nlh );
307 0 : int nl_err = -err->error;
308 0 : FD_LOG_WARNING(( "netlink RTM_GETROUTE,NLM_F_REQUEST|NLM_F_DUMP failed (%d-%s)", nl_err, fd_io_strerror( nl_err ) ));
309 0 : ctx->dump_active = 0;
310 0 : return -FD_FIB_NETLINK_ERR_IO;
311 0 : }
312 0 : if( FD_UNLIKELY( nlh->nlmsg_type!=RTM_NEWROUTE ) ) {
313 0 : FD_LOG_DEBUG(( "unexpected nlmsg_type %u", nlh->nlmsg_type ));
314 0 : return FD_NETLINK_DUMP_NEXT_PROGRESS;
315 0 : }
316 0 : if( FD_UNLIKELY( ctx->dump_overflow ) ) return FD_NETLINK_DUMP_NEXT_PROGRESS;
317 0 : if( !fd_fib4_netlink_translate( nlh, ctx->dump_table_id, route ) ) return FD_NETLINK_DUMP_NEXT_PROGRESS;
318 :
319 0 : ulong table_idx = ctx->dump_table_id==RT_TABLE_LOCAL ? 0UL : 1UL;
320 0 : if( route->prefix==32U && route->dst_addr ) {
321 0 : if( FD_UNLIKELY( ctx->dump_peer_cnt[ table_idx ]>=ctx->route_peer_max ) ) ctx->dump_overflow = 1;
322 0 : else ctx->dump_peer_cnt[ table_idx ]++;
323 0 : } else {
324 : /* fd_fib4 reserves one non-peer slot for its dummy throw route. */
325 0 : if( FD_UNLIKELY( ctx->dump_route_cnt[ table_idx ]+1UL>=ctx->route_max ) ) ctx->dump_overflow = 1;
326 0 : else ctx->dump_route_cnt[ table_idx ]++;
327 0 : }
328 0 : return ctx->dump_overflow ? FD_NETLINK_DUMP_NEXT_PROGRESS : FD_NETLINK_DUMP_NEXT_ROUTE;
329 0 : }
330 :
331 : /* netlink_monitor_read processes at most one netlink message. Returns
332 : 2 if it published a route, 1 if it processed a non-route message, and
333 : 0 if no message was available. */
334 :
335 : static int
336 : netlink_monitor_read( fd_netlink_tile_ctx_t * ctx,
337 : fd_stem_context_t * stem,
338 0 : int flags ) {
339 :
340 0 : if( ctx->monitor_buf_off>=ctx->monitor_buf_sz ) {
341 0 : long msg_sz = recvfrom( ctx->nl_monitor->fd, ctx->monitor_buf, sizeof(ctx->monitor_buf), flags, NULL, NULL );
342 0 : if( msg_sz<=0L ) {
343 0 : if( FD_LIKELY( errno==EAGAIN || errno==EINTR ) ) return 0;
344 0 : if( errno==ENOBUFS ) {
345 0 : fd_netlink_enobufs_cnt++;
346 0 : ctx->action |= FD_NET_TILE_ACTION_ROUTE4_UPDATE;
347 0 : return 0;
348 0 : }
349 0 : FD_LOG_ERR(( "recvfrom(nl_monitor) failed (%i-%s)", errno, fd_io_strerror( errno ) ));
350 0 : }
351 0 : ctx->monitor_buf_sz = msg_sz;
352 0 : ctx->monitor_buf_off = 0L;
353 0 : }
354 :
355 0 : long rem = ctx->monitor_buf_sz-ctx->monitor_buf_off;
356 0 : struct nlmsghdr * nlh = fd_type_pun( ctx->monitor_buf+ctx->monitor_buf_off );
357 0 : if( FD_UNLIKELY( !NLMSG_OK( nlh, rem ) ) ) {
358 0 : FD_LOG_WARNING(( "malformed netlink monitor message" ));
359 0 : ctx->monitor_buf_off = ctx->monitor_buf_sz;
360 0 : return 1;
361 0 : }
362 0 : ctx->monitor_buf_off += (long)NLMSG_ALIGN( nlh->nlmsg_len );
363 0 : FD_DTRACE_PROBE_4( netlink_update, nlh->nlmsg_seq, nlh->nlmsg_type, nlh->nlmsg_len, nlh->nlmsg_flags );
364 0 : switch( nlh->nlmsg_type ) {
365 0 : case RTM_NEWLINK:
366 0 : case RTM_DELLINK:
367 0 : ctx->action |= FD_NET_TILE_ACTION_LINK_UPDATE;
368 0 : ctx->metrics.update_cnt[ FD_METRICS_ENUM_NETLINK_MESSAGE_TYPE_V_LINK_IDX ]++;
369 0 : break;
370 0 : case RTM_NEWROUTE:
371 0 : case RTM_DELROUTE: {
372 0 : fd_iproute_msg_t route;
373 0 : if( fd_fib4_netlink_translate( nlh, RT_TABLE_LOCAL, &route ) ||
374 0 : fd_fib4_netlink_translate( nlh, RT_TABLE_MAIN, &route ) ) {
375 0 : iproute_publish( ctx, stem, &route );
376 0 : ctx->metrics.update_cnt[ FD_METRICS_ENUM_NETLINK_MESSAGE_TYPE_V_IPV4_ROUTE_IDX ]++;
377 0 : return 2;
378 0 : }
379 0 : ctx->metrics.update_cnt[ FD_METRICS_ENUM_NETLINK_MESSAGE_TYPE_V_IPV4_ROUTE_IDX ]++;
380 0 : break;
381 0 : }
382 0 : case RTM_NEWNEIGH:
383 0 : case RTM_DELNEIGH: {
384 0 : fd_neigh4_netlink_ingest_message( ctx->neigh4, nlh, ctx->neigh4_ifidx );
385 0 : ctx->metrics.update_cnt[ FD_METRICS_ENUM_NETLINK_MESSAGE_TYPE_V_NEIGHBOR_IDX ]++;
386 0 : break;
387 0 : }
388 0 : default:
389 0 : FD_LOG_INFO(( "Received unexpected netlink message type %u", nlh->nlmsg_type ));
390 0 : break;
391 0 : }
392 0 : return 1;
393 0 : }
394 :
395 : static void
396 0 : during_housekeeping( fd_netlink_tile_ctx_t * ctx ) {
397 0 : long now = fd_tickcount();
398 0 : if( !ctx->dump_table_id && (ctx->action & FD_NET_TILE_ACTION_LINK_UPDATE) ) {
399 0 : if( now < ctx->link_update_ts ) return;
400 0 : ctx->action &= ~FD_NET_TILE_ACTION_LINK_UPDATE;
401 0 : fd_seqlock_write_lock( &ctx->netdev_tbl->hdr->seqlock );
402 0 : fd_netdev_netlink_load_table( ctx->netdev_tbl, ctx->nl_req );
403 0 : fd_seqlock_write_unlock( &ctx->netdev_tbl->hdr->seqlock );
404 0 : ctx->link_update_ts = now+ctx->update_backoff;
405 0 : ctx->metrics.link_full_syncs++;
406 0 : }
407 0 : if( !ctx->dump_table_id && (ctx->action & FD_NET_TILE_ACTION_NEIGH_UPDATE) ) {
408 0 : ctx->action &= ~FD_NET_TILE_ACTION_NEIGH_UPDATE;
409 0 : fd_neigh4_netlink_request_dump( ctx->nl_req, ctx->neigh4_ifidx );
410 0 : uchar buf[ 4096 ];
411 0 : fd_netlink_iter_t iter[1];
412 0 : for( fd_netlink_iter_init( iter, ctx->nl_req, buf, sizeof(buf) );
413 0 : !fd_netlink_iter_done( iter );
414 0 : fd_netlink_iter_next( iter, ctx->nl_req ) ) {
415 0 : fd_neigh4_netlink_ingest_message( ctx->neigh4, fd_netlink_iter_msg( iter ), ctx->neigh4_ifidx );
416 0 : }
417 0 : }
418 0 : }
419 :
420 : /* after_credit is called once per loop iteration when Stem has at least
421 : one credit available. */
422 :
423 : static void
424 : after_credit( fd_netlink_tile_ctx_t * ctx,
425 : fd_stem_context_t * stem,
426 : int * opt_poll_in,
427 0 : int * charge_busy ) {
428 :
429 0 : long now = fd_tickcount();
430 :
431 0 : if( !ctx->dump_table_id && (ctx->action & FD_NET_TILE_ACTION_ROUTE4_UPDATE) && now>=ctx->route4_update_ts ) {
432 0 : ctx->action &= ~FD_NET_TILE_ACTION_ROUTE4_UPDATE;
433 0 : fd_iproute_msg_t flush = { .op=FD_IPROUTE_OP_FLUSH };
434 0 : iproute_publish( ctx, stem, &flush );
435 0 : fd_memset( ctx->dump_route_cnt, 0, sizeof(ctx->dump_route_cnt) );
436 0 : fd_memset( ctx->dump_peer_cnt, 0, sizeof(ctx->dump_peer_cnt) );
437 0 : ctx->dump_overflow = 0;
438 0 : ctx->dump_table_id = RT_TABLE_LOCAL;
439 0 : *charge_busy = 1;
440 0 : *opt_poll_in = 0;
441 0 : return;
442 0 : }
443 :
444 0 : if( ctx->dump_table_id ) {
445 0 : if( !ctx->dump_active ) {
446 0 : int err = iproute_dump_begin( ctx, ctx->dump_table_id );
447 0 : if( FD_UNLIKELY( err ) ) {
448 0 : ctx->dump_table_id = 0U;
449 0 : ctx->action |= FD_NET_TILE_ACTION_ROUTE4_UPDATE;
450 0 : ctx->route4_update_ts = now+ctx->update_backoff;
451 0 : *charge_busy = 1;
452 0 : return;
453 0 : }
454 0 : }
455 :
456 0 : fd_iproute_msg_t route;
457 0 : int dump_res = iproute_dump_next( ctx, &route );
458 0 : if( dump_res==FD_NETLINK_DUMP_NEXT_ROUTE ) {
459 0 : iproute_publish( ctx, stem, &route );
460 0 : *charge_busy = 1;
461 0 : *opt_poll_in = 0;
462 0 : return;
463 0 : }
464 0 : if( dump_res==FD_NETLINK_DUMP_NEXT_PROGRESS ) {
465 0 : *charge_busy = 1;
466 0 : return;
467 0 : }
468 0 : if( FD_UNLIKELY( dump_res<0 ) ) {
469 0 : if( dump_res==-FD_FIB_NETLINK_ERR_SPACE ) FD_LOG_WARNING(( "routing table exceeds configured netlink route capacity" ));
470 0 : ctx->dump_table_id = 0U;
471 0 : ctx->action |= FD_NET_TILE_ACTION_ROUTE4_UPDATE;
472 0 : ctx->route4_update_ts = now+ctx->update_backoff;
473 0 : *charge_busy = 1;
474 0 : return;
475 0 : }
476 :
477 0 : if( ctx->dump_table_id==RT_TABLE_LOCAL ) {
478 0 : ctx->dump_table_id = RT_TABLE_MAIN;
479 0 : } else {
480 0 : ctx->dump_table_id = 0U;
481 0 : ctx->route4_update_ts = now+ctx->update_backoff;
482 0 : ctx->metrics.route_full_syncs++;
483 0 : }
484 0 : *charge_busy = 1;
485 0 : return;
486 0 : }
487 :
488 0 : int published = 0;
489 0 : if( FD_UNLIKELY( fd_fseq_query( ctx->waker_fseq )==1UL ) ) {
490 0 : fd_fseq_update( ctx->waker_fseq, 0UL );
491 0 : for(;;) {
492 0 : int read_res = netlink_monitor_read( ctx, stem, MSG_DONTWAIT );
493 0 : if( !read_res ) break;
494 0 : *charge_busy = 1;
495 0 : if( read_res==2 ) {
496 0 : published = 1;
497 0 : break;
498 0 : }
499 0 : }
500 0 : fd_waker_client_rearm( ctx->waker_client_idx );
501 0 : ctx->idle_cnt = -1L;
502 0 : }
503 :
504 0 : ctx->idle_cnt++;
505 0 : if( FD_UNLIKELY( !published && ctx->idle_cnt>=128L && !stem->sleep ) )
506 0 : fd_log_sleep( (long)1e6 );
507 :
508 0 : if( FD_UNLIKELY( published ) ) *opt_poll_in = 0;
509 0 : }
510 :
511 : /* after_poll_overrun is called when fd_stem.c was overrun while
512 : checking for new fragments: producers are hot, stay hot. */
513 :
514 : static void
515 0 : after_poll_overrun( fd_netlink_tile_ctx_t * ctx ) {
516 0 : ctx->idle_cnt = -1L;
517 0 : }
518 :
519 : /* after_frag handles a neighbor solicit request */
520 :
521 : static void
522 : after_frag( fd_netlink_tile_ctx_t * ctx,
523 : ulong in_idx,
524 : ulong seq,
525 : ulong sig,
526 : ulong sz,
527 : ulong tsorig,
528 : ulong tspub,
529 0 : fd_stem_context_t * stem ) {
530 0 : (void)in_idx; (void)seq; (void)tsorig; (void)tspub; (void)stem;
531 :
532 0 : ctx->idle_cnt = -1L;
533 0 : long now = fd_tickcount();
534 :
535 : /* Parse request (fully contained in sig field) */
536 :
537 0 : if( FD_UNLIKELY( sz!=0UL ) ) {
538 0 : FD_LOG_WARNING(( "unexpected sz %lu", sz ));
539 0 : }
540 0 : if( FD_UNLIKELY( sig==FD_NETLINK_ROUTE4_SYNC_SIG ) ) {
541 0 : ctx->action |= FD_NET_TILE_ACTION_ROUTE4_UPDATE;
542 0 : return;
543 0 : }
544 0 : uint if_idx = (uint)(sig>>32);
545 0 : uint ip4_addr = (uint)sig;
546 0 : if( FD_UNLIKELY( if_idx!=ctx->neigh4_ifidx ) ) {
547 0 : FD_LOG_ERR(( "received neighbor solicit request for invalid interface index %u", if_idx ));
548 0 : return;
549 0 : }
550 :
551 : /* Drop if the kernel is already working on the request */
552 0 : if( fd_neigh4_hmap_query( ctx->neigh4, &ip4_addr ) ) {
553 0 : ctx->metrics.neigh_probe[ FD_METRICS_ENUM_NEIGHBOR_PROBE_RESULT_V_IN_FLIGHT_IDX ]++;
554 0 : return;
555 0 : }
556 :
557 : /* Insert placeholder (take above branch next time) */
558 :
559 0 : fd_neigh4_entry_t * ele = fd_neigh4_hmap_insert( ctx->neigh4, &ip4_addr );
560 0 : if( FD_UNLIKELY( !ele ) ) {
561 0 : ctx->metrics.neigh_probe[ FD_METRICS_ENUM_NEIGHBOR_PROBE_RESULT_V_TABLE_FULL_IDX ]++;
562 0 : return;
563 0 : }
564 : /* Atomically write the entry, initializing MAC and probe suppression timestamp to 0 */
565 0 : fd_neigh4_entry_t to_insert = (fd_neigh4_entry_t) {
566 0 : .ip4_addr = ip4_addr,
567 0 : .state = FD_NEIGH4_STATE_INCOMPLETE,
568 0 : };
569 0 : fd_neigh4_entry_atomic_st( ele, &to_insert );
570 :
571 : /* Trigger neighbor solicit via netlink */
572 :
573 0 : int probe_res = fd_neigh4_probe_rate_limited( ctx->prober, ele, ip4_addr, now );
574 0 : if( probe_res==0 ) {
575 0 : ctx->metrics.neigh_probe[ FD_METRICS_ENUM_NEIGHBOR_PROBE_RESULT_V_SENT_IDX ]++;
576 0 : } else {
577 0 : fd_neigh4_hmap_remove( ctx->neigh4, ele );
578 0 : if( probe_res>0 ) ctx->metrics.neigh_probe[ FD_METRICS_ENUM_NEIGHBOR_PROBE_RESULT_V_SEND_FAILED_IDX ]++;
579 0 : }
580 :
581 0 : }
582 :
583 0 : #define STEM_BURST (1UL)
584 0 : #define STEM_LAZY ((ulong)13e6) /* 13ms */
585 :
586 0 : #define STEM_CALLBACK_CONTEXT_TYPE fd_netlink_tile_ctx_t
587 0 : #define STEM_CALLBACK_CONTEXT_ALIGN alignof(fd_netlink_tile_ctx_t)
588 :
589 0 : #define STEM_CALLBACK_METRICS_WRITE metrics_write
590 0 : #define STEM_CALLBACK_AFTER_POLL_OVERRUN after_poll_overrun
591 0 : #define STEM_CALLBACK_DURING_HOUSEKEEPING during_housekeeping
592 0 : #define STEM_CALLBACK_AFTER_CREDIT after_credit
593 0 : #define STEM_CALLBACK_AFTER_FRAG after_frag
594 :
595 : #include "../stem/fd_stem.c"
596 :
597 : /* End stem methods */
598 :
599 : fd_topo_run_tile_t fd_tile_netlnk = {
600 : .name = "netlnk",
601 : .populate_allowed_seccomp = populate_allowed_seccomp,
602 : .populate_allowed_fds = populate_allowed_fds,
603 : .scratch_align = scratch_align,
604 : .scratch_footprint = scratch_footprint,
605 : .privileged_init = privileged_init,
606 : .unprivileged_init = unprivileged_init,
607 : .run = stem_run
608 : };
|
