2 * SPDX-License-Identifier: BSD-2-Clause-FreeBSD
4 * Copyright (C) 2011-2014 Matteo Landi
5 * Copyright (C) 2011-2016 Luigi Rizzo
6 * Copyright (C) 2011-2016 Giuseppe Lettieri
7 * Copyright (C) 2011-2016 Vincenzo Maffione
10 * Redistribution and use in source and binary forms, with or without
11 * modification, are permitted provided that the following conditions
13 * 1. Redistributions of source code must retain the above copyright
14 * notice, this list of conditions and the following disclaimer.
15 * 2. Redistributions in binary form must reproduce the above copyright
16 * notice, this list of conditions and the following disclaimer in the
17 * documentation and/or other materials provided with the distribution.
19 * THIS SOFTWARE IS PROVIDED BY THE AUTHOR AND CONTRIBUTORS ``AS IS'' AND
20 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
21 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
22 * ARE DISCLAIMED. IN NO EVENT SHALL THE AUTHOR OR CONTRIBUTORS BE LIABLE
23 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
24 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
25 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
26 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
27 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
28 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
36 * This module supports memory mapped access to network devices,
39 * The module uses a large, memory pool allocated by the kernel
40 * and accessible as mmapped memory by multiple userspace threads/processes.
41 * The memory pool contains packet buffers and "netmap rings",
42 * i.e. user-accessible copies of the interface's queues.
44 * Access to the network card works like this:
45 * 1. a process/thread issues one or more open() on /dev/netmap, to create
46 * select()able file descriptor on which events are reported.
47 * 2. on each descriptor, the process issues an ioctl() to identify
48 * the interface that should report events to the file descriptor.
49 * 3. on each descriptor, the process issues an mmap() request to
50 * map the shared memory region within the process' address space.
51 * The list of interesting queues is indicated by a location in
52 * the shared memory region.
53 * 4. using the functions in the netmap(4) userspace API, a process
54 * can look up the occupation state of a queue, access memory buffers,
55 * and retrieve received packets or enqueue packets to transmit.
56 * 5. using some ioctl()s the process can synchronize the userspace view
57 * of the queue with the actual status in the kernel. This includes both
58 * receiving the notification of new packets, and transmitting new
59 * packets on the output interface.
60 * 6. select() or poll() can be used to wait for events on individual
61 * transmit or receive queues (or all queues for a given interface).
64 SYNCHRONIZATION (USER)
66 The netmap rings and data structures may be shared among multiple
67 user threads or even independent processes.
68 Any synchronization among those threads/processes is delegated
69 to the threads themselves. Only one thread at a time can be in
70 a system call on the same netmap ring. The OS does not enforce
71 this and only guarantees against system crashes in case of
76 Within the kernel, access to the netmap rings is protected as follows:
78 - a spinlock on each ring, to handle producer/consumer races on
79 RX rings attached to the host stack (against multiple host
80 threads writing from the host stack to the same ring),
81 and on 'destination' rings attached to a VALE switch
82 (i.e. RX rings in VALE ports, and TX rings in NIC/host ports)
83 protecting multiple active senders for the same destination)
85 - an atomic variable to guarantee that there is at most one
86 instance of *_*xsync() on the ring at any time.
87 For rings connected to user file
88 descriptors, an atomic_test_and_set() protects this, and the
89 lock on the ring is not actually used.
90 For NIC RX rings connected to a VALE switch, an atomic_test_and_set()
91 is also used to prevent multiple executions (the driver might indeed
92 already guarantee this).
93 For NIC TX rings connected to a VALE switch, the lock arbitrates
94 access to the queue (both when allocating buffers and when pushing
97 - *xsync() should be protected against initializations of the card.
98 On FreeBSD most devices have the reset routine protected by
99 a RING lock (ixgbe, igb, em) or core lock (re). lem is missing
100 the RING protection on rx_reset(), this should be added.
102 On linux there is an external lock on the tx path, which probably
103 also arbitrates access to the reset routine. XXX to be revised
105 - a per-interface core_lock protecting access from the host stack
106 while interfaces may be detached from netmap mode.
107 XXX there should be no need for this lock if we detach the interfaces
108 only while they are down.
113 NMG_LOCK() serializes all modifications to switches and ports.
114 A switch cannot be deleted until all ports are gone.
116 For each switch, an SX lock (RWlock on linux) protects
117 deletion of ports. When configuring or deleting a new port, the
118 lock is acquired in exclusive mode (after holding NMG_LOCK).
119 When forwarding, the lock is acquired in shared mode (without NMG_LOCK).
120 The lock is held throughout the entire forwarding cycle,
121 during which the thread may incur in a page fault.
122 Hence it is important that sleepable shared locks are used.
124 On the rx ring, the per-port lock is grabbed initially to reserve
125 a number of slot in the ring, then the lock is released,
126 packets are copied from source to destination, and then
127 the lock is acquired again and the receive ring is updated.
128 (A similar thing is done on the tx ring for NIC and host stack
129 ports attached to the switch)
134 /* --- internals ----
136 * Roadmap to the code that implements the above.
138 * > 1. a process/thread issues one or more open() on /dev/netmap, to create
139 * > select()able file descriptor on which events are reported.
141 * Internally, we allocate a netmap_priv_d structure, that will be
142 * initialized on ioctl(NIOCREGIF). There is one netmap_priv_d
143 * structure for each open().
146 * FreeBSD: see netmap_open() (netmap_freebsd.c)
147 * linux: see linux_netmap_open() (netmap_linux.c)
149 * > 2. on each descriptor, the process issues an ioctl() to identify
150 * > the interface that should report events to the file descriptor.
152 * Implemented by netmap_ioctl(), NIOCREGIF case, with nmr->nr_cmd==0.
153 * Most important things happen in netmap_get_na() and
154 * netmap_do_regif(), called from there. Additional details can be
155 * found in the comments above those functions.
157 * In all cases, this action creates/takes-a-reference-to a
158 * netmap_*_adapter describing the port, and allocates a netmap_if
159 * and all necessary netmap rings, filling them with netmap buffers.
161 * In this phase, the sync callbacks for each ring are set (these are used
162 * in steps 5 and 6 below). The callbacks depend on the type of adapter.
163 * The adapter creation/initialization code puts them in the
164 * netmap_adapter (fields na->nm_txsync and na->nm_rxsync). Then, they
165 * are copied from there to the netmap_kring's during netmap_do_regif(), by
166 * the nm_krings_create() callback. All the nm_krings_create callbacks
167 * actually call netmap_krings_create() to perform this and the other
168 * common stuff. netmap_krings_create() also takes care of the host rings,
169 * if needed, by setting their sync callbacks appropriately.
171 * Additional actions depend on the kind of netmap_adapter that has been
174 * - netmap_hw_adapter: [netmap.c]
175 * This is a system netdev/ifp with native netmap support.
176 * The ifp is detached from the host stack by redirecting:
177 * - transmissions (from the network stack) to netmap_transmit()
178 * - receive notifications to the nm_notify() callback for
179 * this adapter. The callback is normally netmap_notify(), unless
180 * the ifp is attached to a bridge using bwrap, in which case it
181 * is netmap_bwrap_intr_notify().
183 * - netmap_generic_adapter: [netmap_generic.c]
184 * A system netdev/ifp without native netmap support.
186 * (the decision about native/non native support is taken in
187 * netmap_get_hw_na(), called by netmap_get_na())
189 * - netmap_vp_adapter [netmap_vale.c]
190 * Returned by netmap_get_bdg_na().
191 * This is a persistent or ephemeral VALE port. Ephemeral ports
192 * are created on the fly if they don't already exist, and are
193 * always attached to a bridge.
194 * Persistent VALE ports must must be created separately, and i
195 * then attached like normal NICs. The NIOCREGIF we are examining
196 * will find them only if they had previosly been created and
197 * attached (see VALE_CTL below).
199 * - netmap_pipe_adapter [netmap_pipe.c]
200 * Returned by netmap_get_pipe_na().
201 * Both pipe ends are created, if they didn't already exist.
203 * - netmap_monitor_adapter [netmap_monitor.c]
204 * Returned by netmap_get_monitor_na().
205 * If successful, the nm_sync callbacks of the monitored adapter
206 * will be intercepted by the returned monitor.
208 * - netmap_bwrap_adapter [netmap_vale.c]
209 * Cannot be obtained in this way, see VALE_CTL below
213 * linux: we first go through linux_netmap_ioctl() to
214 * adapt the FreeBSD interface to the linux one.
217 * > 3. on each descriptor, the process issues an mmap() request to
218 * > map the shared memory region within the process' address space.
219 * > The list of interesting queues is indicated by a location in
220 * > the shared memory region.
223 * FreeBSD: netmap_mmap_single (netmap_freebsd.c).
224 * linux: linux_netmap_mmap (netmap_linux.c).
226 * > 4. using the functions in the netmap(4) userspace API, a process
227 * > can look up the occupation state of a queue, access memory buffers,
228 * > and retrieve received packets or enqueue packets to transmit.
230 * these actions do not involve the kernel.
232 * > 5. using some ioctl()s the process can synchronize the userspace view
233 * > of the queue with the actual status in the kernel. This includes both
234 * > receiving the notification of new packets, and transmitting new
235 * > packets on the output interface.
237 * These are implemented in netmap_ioctl(), NIOCTXSYNC and NIOCRXSYNC
238 * cases. They invoke the nm_sync callbacks on the netmap_kring
239 * structures, as initialized in step 2 and maybe later modified
240 * by a monitor. Monitors, however, will always call the original
241 * callback before doing anything else.
244 * > 6. select() or poll() can be used to wait for events on individual
245 * > transmit or receive queues (or all queues for a given interface).
247 * Implemented in netmap_poll(). This will call the same nm_sync()
248 * callbacks as in step 5 above.
251 * linux: we first go through linux_netmap_poll() to adapt
252 * the FreeBSD interface to the linux one.
255 * ---- VALE_CTL -----
257 * VALE switches are controlled by issuing a NIOCREGIF with a non-null
258 * nr_cmd in the nmreq structure. These subcommands are handled by
259 * netmap_bdg_ctl() in netmap_vale.c. Persistent VALE ports are created
260 * and destroyed by issuing the NETMAP_BDG_NEWIF and NETMAP_BDG_DELIF
261 * subcommands, respectively.
263 * Any network interface known to the system (including a persistent VALE
264 * port) can be attached to a VALE switch by issuing the
265 * NETMAP_REQ_VALE_ATTACH command. After the attachment, persistent VALE ports
266 * look exactly like ephemeral VALE ports (as created in step 2 above). The
267 * attachment of other interfaces, instead, requires the creation of a
268 * netmap_bwrap_adapter. Moreover, the attached interface must be put in
269 * netmap mode. This may require the creation of a netmap_generic_adapter if
270 * we have no native support for the interface, or if generic adapters have
271 * been forced by sysctl.
273 * Both persistent VALE ports and bwraps are handled by netmap_get_bdg_na(),
274 * called by nm_bdg_ctl_attach(), and discriminated by the nm_bdg_attach()
275 * callback. In the case of the bwrap, the callback creates the
276 * netmap_bwrap_adapter. The initialization of the bwrap is then
277 * completed by calling netmap_do_regif() on it, in the nm_bdg_ctl()
278 * callback (netmap_bwrap_bdg_ctl in netmap_vale.c).
279 * A generic adapter for the wrapped ifp will be created if needed, when
280 * netmap_get_bdg_na() calls netmap_get_hw_na().
283 * ---- DATAPATHS -----
285 * -= SYSTEM DEVICE WITH NATIVE SUPPORT =-
287 * na == NA(ifp) == netmap_hw_adapter created in DEVICE_netmap_attach()
289 * - tx from netmap userspace:
291 * 1) ioctl(NIOCTXSYNC)/netmap_poll() in process context
292 * kring->nm_sync() == DEVICE_netmap_txsync()
293 * 2) device interrupt handler
294 * na->nm_notify() == netmap_notify()
295 * - rx from netmap userspace:
297 * 1) ioctl(NIOCRXSYNC)/netmap_poll() in process context
298 * kring->nm_sync() == DEVICE_netmap_rxsync()
299 * 2) device interrupt handler
300 * na->nm_notify() == netmap_notify()
301 * - rx from host stack
305 * na->nm_notify == netmap_notify()
306 * 2) ioctl(NIOCRXSYNC)/netmap_poll() in process context
307 * kring->nm_sync() == netmap_rxsync_from_host
308 * netmap_rxsync_from_host(na, NULL, NULL)
310 * ioctl(NIOCTXSYNC)/netmap_poll() in process context
311 * kring->nm_sync() == netmap_txsync_to_host
312 * netmap_txsync_to_host(na)
314 * FreeBSD: na->if_input() == ether_input()
315 * linux: netif_rx() with NM_MAGIC_PRIORITY_RX
318 * -= SYSTEM DEVICE WITH GENERIC SUPPORT =-
320 * na == NA(ifp) == generic_netmap_adapter created in generic_netmap_attach()
322 * - tx from netmap userspace:
324 * 1) ioctl(NIOCTXSYNC)/netmap_poll() in process context
325 * kring->nm_sync() == generic_netmap_txsync()
326 * nm_os_generic_xmit_frame()
327 * linux: dev_queue_xmit() with NM_MAGIC_PRIORITY_TX
328 * ifp->ndo_start_xmit == generic_ndo_start_xmit()
329 * gna->save_start_xmit == orig. dev. start_xmit
330 * FreeBSD: na->if_transmit() == orig. dev if_transmit
331 * 2) generic_mbuf_destructor()
332 * na->nm_notify() == netmap_notify()
333 * - rx from netmap userspace:
334 * 1) ioctl(NIOCRXSYNC)/netmap_poll() in process context
335 * kring->nm_sync() == generic_netmap_rxsync()
338 * generic_rx_handler()
340 * na->nm_notify() == netmap_notify()
341 * - rx from host stack
342 * FreeBSD: same as native
343 * Linux: same as native except:
345 * dev_queue_xmit() without NM_MAGIC_PRIORITY_TX
346 * ifp->ndo_start_xmit == generic_ndo_start_xmit()
348 * na->nm_notify() == netmap_notify()
349 * - tx to host stack (same as native):
357 * ioctl(NIOCTXSYNC)/netmap_poll() in process context
358 * kring->nm_sync() == netmap_vp_txsync()
360 * - system device with native support:
363 * na->nm_notify() == netmap_bwrap_intr_notify(ring_nr != host ring)
364 * kring->nm_sync() == DEVICE_netmap_rxsync()
366 * kring->nm_sync() == DEVICE_netmap_rxsync()
369 * na->nm_notify() == netmap_bwrap_intr_notify(ring_nr == host ring)
370 * kring->nm_sync() == netmap_rxsync_from_host()
373 * - system device with generic support:
374 * from device driver:
375 * generic_rx_handler()
376 * na->nm_notify() == netmap_bwrap_intr_notify(ring_nr != host ring)
377 * kring->nm_sync() == generic_netmap_rxsync()
379 * kring->nm_sync() == generic_netmap_rxsync()
382 * na->nm_notify() == netmap_bwrap_intr_notify(ring_nr == host ring)
383 * kring->nm_sync() == netmap_rxsync_from_host()
386 * (all cases) --> nm_bdg_flush()
387 * dest_na->nm_notify() == (see below)
393 * 1) ioctl(NIOCRXSYNC)/netmap_poll() in process context
394 * kring->nm_sync() == netmap_vp_rxsync()
395 * 2) from nm_bdg_flush()
396 * na->nm_notify() == netmap_notify()
398 * - system device with native support:
400 * na->nm_notify() == netmap_bwrap_notify()
402 * kring->nm_sync() == DEVICE_netmap_txsync()
406 * kring->nm_sync() == netmap_txsync_to_host
407 * netmap_vp_rxsync_locked()
409 * - system device with generic adapter:
411 * na->nm_notify() == netmap_bwrap_notify()
413 * kring->nm_sync() == generic_netmap_txsync()
417 * kring->nm_sync() == netmap_txsync_to_host
423 * OS-specific code that is used only within this file.
424 * Other OS-specific code that must be accessed by drivers
425 * is present in netmap_kern.h
428 #if defined(__FreeBSD__)
429 #include <sys/cdefs.h> /* prerequisite */
430 #include <sys/types.h>
431 #include <sys/errno.h>
432 #include <sys/param.h> /* defines used in kernel.h */
433 #include <sys/kernel.h> /* types used in module initialization */
434 #include <sys/conf.h> /* cdevsw struct, UID, GID */
435 #include <sys/filio.h> /* FIONBIO */
436 #include <sys/sockio.h>
437 #include <sys/socketvar.h> /* struct socket */
438 #include <sys/malloc.h>
439 #include <sys/poll.h>
440 #include <sys/rwlock.h>
441 #include <sys/socket.h> /* sockaddrs */
442 #include <sys/selinfo.h>
443 #include <sys/sysctl.h>
444 #include <sys/jail.h>
445 #include <net/vnet.h>
447 #include <net/if_var.h>
448 #include <net/bpf.h> /* BIOCIMMEDIATE */
449 #include <machine/bus.h> /* bus_dmamap_* */
450 #include <sys/endian.h>
451 #include <sys/refcount.h>
452 #include <net/ethernet.h> /* ETHER_BPF_MTAP */
457 #include "bsd_glue.h"
459 #elif defined(__APPLE__)
461 #warning OSX support is only partial
462 #include "osx_glue.h"
464 #elif defined (_WIN32)
466 #include "win_glue.h"
470 #error Unsupported platform
472 #endif /* unsupported */
477 #include <net/netmap.h>
478 #include <dev/netmap/netmap_kern.h>
479 #include <dev/netmap/netmap_mem2.h>
482 /* user-controlled variables */
484 #ifdef CONFIG_NETMAP_DEBUG
486 #endif /* CONFIG_NETMAP_DEBUG */
488 static int netmap_no_timestamp; /* don't timestamp on rxsync */
489 int netmap_no_pendintr = 1;
490 int netmap_txsync_retry = 2;
491 static int netmap_fwd = 0; /* force transparent forwarding */
494 * netmap_admode selects the netmap mode to use.
495 * Invalid values are reset to NETMAP_ADMODE_BEST
497 enum { NETMAP_ADMODE_BEST = 0, /* use native, fallback to generic */
498 NETMAP_ADMODE_NATIVE, /* either native or none */
499 NETMAP_ADMODE_GENERIC, /* force generic */
500 NETMAP_ADMODE_LAST };
501 static int netmap_admode = NETMAP_ADMODE_BEST;
503 /* netmap_generic_mit controls mitigation of RX notifications for
504 * the generic netmap adapter. The value is a time interval in
506 int netmap_generic_mit = 100*1000;
508 /* We use by default netmap-aware qdiscs with generic netmap adapters,
509 * even if there can be a little performance hit with hardware NICs.
510 * However, using the qdisc is the safer approach, for two reasons:
511 * 1) it prevents non-fifo qdiscs to break the TX notification
512 * scheme, which is based on mbuf destructors when txqdisc is
514 * 2) it makes it possible to transmit over software devices that
515 * change skb->dev, like bridge, veth, ...
517 * Anyway users looking for the best performance should
518 * use native adapters.
521 int netmap_generic_txqdisc = 1;
524 /* Default number of slots and queues for generic adapters. */
525 int netmap_generic_ringsize = 1024;
526 int netmap_generic_rings = 1;
528 /* Non-zero to enable checksum offloading in NIC drivers */
529 int netmap_generic_hwcsum = 0;
531 /* Non-zero if ptnet devices are allowed to use virtio-net headers. */
532 int ptnet_vnet_hdr = 1;
535 * SYSCTL calls are grouped between SYSBEGIN and SYSEND to be emulated
536 * in some other operating systems
540 SYSCTL_DECL(_dev_netmap);
541 SYSCTL_NODE(_dev, OID_AUTO, netmap, CTLFLAG_RW, 0, "Netmap args");
542 SYSCTL_INT(_dev_netmap, OID_AUTO, verbose,
543 CTLFLAG_RW, &netmap_verbose, 0, "Verbose mode");
544 #ifdef CONFIG_NETMAP_DEBUG
545 SYSCTL_INT(_dev_netmap, OID_AUTO, debug,
546 CTLFLAG_RW, &netmap_debug, 0, "Debug messages");
547 #endif /* CONFIG_NETMAP_DEBUG */
548 SYSCTL_INT(_dev_netmap, OID_AUTO, no_timestamp,
549 CTLFLAG_RW, &netmap_no_timestamp, 0, "no_timestamp");
550 SYSCTL_INT(_dev_netmap, OID_AUTO, no_pendintr, CTLFLAG_RW, &netmap_no_pendintr,
551 0, "Always look for new received packets.");
552 SYSCTL_INT(_dev_netmap, OID_AUTO, txsync_retry, CTLFLAG_RW,
553 &netmap_txsync_retry, 0, "Number of txsync loops in bridge's flush.");
555 SYSCTL_INT(_dev_netmap, OID_AUTO, fwd, CTLFLAG_RW, &netmap_fwd, 0,
556 "Force NR_FORWARD mode");
557 SYSCTL_INT(_dev_netmap, OID_AUTO, admode, CTLFLAG_RW, &netmap_admode, 0,
558 "Adapter mode. 0 selects the best option available,"
559 "1 forces native adapter, 2 forces emulated adapter");
560 SYSCTL_INT(_dev_netmap, OID_AUTO, generic_hwcsum, CTLFLAG_RW, &netmap_generic_hwcsum,
561 0, "Hardware checksums. 0 to disable checksum generation by the NIC (default),"
562 "1 to enable checksum generation by the NIC");
563 SYSCTL_INT(_dev_netmap, OID_AUTO, generic_mit, CTLFLAG_RW, &netmap_generic_mit,
564 0, "RX notification interval in nanoseconds");
565 SYSCTL_INT(_dev_netmap, OID_AUTO, generic_ringsize, CTLFLAG_RW,
566 &netmap_generic_ringsize, 0,
567 "Number of per-ring slots for emulated netmap mode");
568 SYSCTL_INT(_dev_netmap, OID_AUTO, generic_rings, CTLFLAG_RW,
569 &netmap_generic_rings, 0,
570 "Number of TX/RX queues for emulated netmap adapters");
572 SYSCTL_INT(_dev_netmap, OID_AUTO, generic_txqdisc, CTLFLAG_RW,
573 &netmap_generic_txqdisc, 0, "Use qdisc for generic adapters");
575 SYSCTL_INT(_dev_netmap, OID_AUTO, ptnet_vnet_hdr, CTLFLAG_RW, &ptnet_vnet_hdr,
576 0, "Allow ptnet devices to use virtio-net headers");
580 NMG_LOCK_T netmap_global_lock;
583 * mark the ring as stopped, and run through the locks
584 * to make sure other users get to see it.
585 * stopped must be either NR_KR_STOPPED (for unbounded stop)
586 * of NR_KR_LOCKED (brief stop for mutual exclusion purposes)
589 netmap_disable_ring(struct netmap_kring *kr, int stopped)
591 nm_kr_stop(kr, stopped);
592 // XXX check if nm_kr_stop is sufficient
593 mtx_lock(&kr->q_lock);
594 mtx_unlock(&kr->q_lock);
598 /* stop or enable a single ring */
600 netmap_set_ring(struct netmap_adapter *na, u_int ring_id, enum txrx t, int stopped)
603 netmap_disable_ring(NMR(na, t)[ring_id], stopped);
605 NMR(na, t)[ring_id]->nkr_stopped = 0;
609 /* stop or enable all the rings of na */
611 netmap_set_all_rings(struct netmap_adapter *na, int stopped)
616 if (!nm_netmap_on(na))
620 for (i = 0; i < netmap_real_rings(na, t); i++) {
621 netmap_set_ring(na, i, t, stopped);
627 * Convenience function used in drivers. Waits for current txsync()s/rxsync()s
628 * to finish and prevents any new one from starting. Call this before turning
629 * netmap mode off, or before removing the hardware rings (e.g., on module
633 netmap_disable_all_rings(struct ifnet *ifp)
635 if (NM_NA_VALID(ifp)) {
636 netmap_set_all_rings(NA(ifp), NM_KR_STOPPED);
641 * Convenience function used in drivers. Re-enables rxsync and txsync on the
642 * adapter's rings In linux drivers, this should be placed near each
646 netmap_enable_all_rings(struct ifnet *ifp)
648 if (NM_NA_VALID(ifp)) {
649 netmap_set_all_rings(NA(ifp), 0 /* enabled */);
654 netmap_make_zombie(struct ifnet *ifp)
656 if (NM_NA_VALID(ifp)) {
657 struct netmap_adapter *na = NA(ifp);
658 netmap_set_all_rings(na, NM_KR_LOCKED);
659 na->na_flags |= NAF_ZOMBIE;
660 netmap_set_all_rings(na, 0);
665 netmap_undo_zombie(struct ifnet *ifp)
667 if (NM_NA_VALID(ifp)) {
668 struct netmap_adapter *na = NA(ifp);
669 if (na->na_flags & NAF_ZOMBIE) {
670 netmap_set_all_rings(na, NM_KR_LOCKED);
671 na->na_flags &= ~NAF_ZOMBIE;
672 netmap_set_all_rings(na, 0);
678 * generic bound_checking function
681 nm_bound_var(u_int *v, u_int dflt, u_int lo, u_int hi, const char *msg)
684 const char *op = NULL;
693 } else if (oldv > hi) {
698 nm_prinf("%s %s to %d (was %d)", op, msg, *v, oldv);
704 * packet-dump function, user-supplied or static buffer.
705 * The destination buffer must be at least 30+4*len
708 nm_dump_buf(char *p, int len, int lim, char *dst)
710 static char _dst[8192];
712 static char hex[] ="0123456789abcdef";
713 char *o; /* output position */
715 #define P_HI(x) hex[((x) & 0xf0)>>4]
716 #define P_LO(x) hex[((x) & 0xf)]
717 #define P_C(x) ((x) >= 0x20 && (x) <= 0x7e ? (x) : '.')
720 if (lim <= 0 || lim > len)
723 sprintf(o, "buf 0x%p len %d lim %d\n", p, len, lim);
725 /* hexdump routine */
726 for (i = 0; i < lim; ) {
727 sprintf(o, "%5d: ", i);
731 for (j=0; j < 16 && i < lim; i++, j++) {
733 o[j*3+1] = P_LO(p[i]);
736 for (j=0; j < 16 && i < lim; i++, j++)
737 o[j + 48] = P_C(p[i]);
750 * Fetch configuration from the device, to cope with dynamic
751 * reconfigurations after loading the module.
753 /* call with NMG_LOCK held */
755 netmap_update_config(struct netmap_adapter *na)
757 struct nm_config_info info;
759 bzero(&info, sizeof(info));
760 if (na->nm_config == NULL ||
761 na->nm_config(na, &info)) {
762 /* take whatever we had at init time */
763 info.num_tx_rings = na->num_tx_rings;
764 info.num_tx_descs = na->num_tx_desc;
765 info.num_rx_rings = na->num_rx_rings;
766 info.num_rx_descs = na->num_rx_desc;
767 info.rx_buf_maxsize = na->rx_buf_maxsize;
770 if (na->num_tx_rings == info.num_tx_rings &&
771 na->num_tx_desc == info.num_tx_descs &&
772 na->num_rx_rings == info.num_rx_rings &&
773 na->num_rx_desc == info.num_rx_descs &&
774 na->rx_buf_maxsize == info.rx_buf_maxsize)
775 return 0; /* nothing changed */
776 if (na->active_fds == 0) {
777 na->num_tx_rings = info.num_tx_rings;
778 na->num_tx_desc = info.num_tx_descs;
779 na->num_rx_rings = info.num_rx_rings;
780 na->num_rx_desc = info.num_rx_descs;
781 na->rx_buf_maxsize = info.rx_buf_maxsize;
783 nm_prinf("configuration changed for %s: txring %d x %d, "
784 "rxring %d x %d, rxbufsz %d",
785 na->name, na->num_tx_rings, na->num_tx_desc,
786 na->num_rx_rings, na->num_rx_desc, na->rx_buf_maxsize);
789 nm_prerr("WARNING: configuration changed for %s while active: "
790 "txring %d x %d, rxring %d x %d, rxbufsz %d",
791 na->name, info.num_tx_rings, info.num_tx_descs,
792 info.num_rx_rings, info.num_rx_descs,
793 info.rx_buf_maxsize);
797 /* nm_sync callbacks for the host rings */
798 static int netmap_txsync_to_host(struct netmap_kring *kring, int flags);
799 static int netmap_rxsync_from_host(struct netmap_kring *kring, int flags);
801 /* create the krings array and initialize the fields common to all adapters.
802 * The array layout is this:
805 * na->tx_rings ----->| | \
806 * | | } na->num_tx_ring
810 * na->rx_rings ----> +----------+
812 * | | } na->num_rx_rings
817 * na->tailroom ----->| | \
818 * | | } tailroom bytes
822 * Note: for compatibility, host krings are created even when not needed.
823 * The tailroom space is currently used by vale ports for allocating leases.
825 /* call with NMG_LOCK held */
827 netmap_krings_create(struct netmap_adapter *na, u_int tailroom)
830 struct netmap_kring *kring;
834 if (na->tx_rings != NULL) {
835 if (netmap_debug & NM_DEBUG_ON)
836 nm_prerr("warning: krings were already created");
840 /* account for the (possibly fake) host rings */
841 n[NR_TX] = netmap_all_rings(na, NR_TX);
842 n[NR_RX] = netmap_all_rings(na, NR_RX);
844 len = (n[NR_TX] + n[NR_RX]) *
845 (sizeof(struct netmap_kring) + sizeof(struct netmap_kring *))
848 na->tx_rings = nm_os_malloc((size_t)len);
849 if (na->tx_rings == NULL) {
850 nm_prerr("Cannot allocate krings");
853 na->rx_rings = na->tx_rings + n[NR_TX];
854 na->tailroom = na->rx_rings + n[NR_RX];
856 /* link the krings in the krings array */
857 kring = (struct netmap_kring *)((char *)na->tailroom + tailroom);
858 for (i = 0; i < n[NR_TX] + n[NR_RX]; i++) {
859 na->tx_rings[i] = kring;
864 * All fields in krings are 0 except the one initialized below.
865 * but better be explicit on important kring fields.
868 ndesc = nma_get_ndesc(na, t);
869 for (i = 0; i < n[t]; i++) {
870 kring = NMR(na, t)[i];
871 bzero(kring, sizeof(*kring));
873 kring->notify_na = na;
876 kring->nkr_num_slots = ndesc;
877 kring->nr_mode = NKR_NETMAP_OFF;
878 kring->nr_pending_mode = NKR_NETMAP_OFF;
879 if (i < nma_get_nrings(na, t)) {
880 kring->nm_sync = (t == NR_TX ? na->nm_txsync : na->nm_rxsync);
882 if (!(na->na_flags & NAF_HOST_RINGS))
883 kring->nr_kflags |= NKR_FAKERING;
884 kring->nm_sync = (t == NR_TX ?
885 netmap_txsync_to_host:
886 netmap_rxsync_from_host);
888 kring->nm_notify = na->nm_notify;
889 kring->rhead = kring->rcur = kring->nr_hwcur = 0;
891 * IMPORTANT: Always keep one slot empty.
893 kring->rtail = kring->nr_hwtail = (t == NR_TX ? ndesc - 1 : 0);
894 snprintf(kring->name, sizeof(kring->name) - 1, "%s %s%d", na->name,
896 ND("ktx %s h %d c %d t %d",
897 kring->name, kring->rhead, kring->rcur, kring->rtail);
898 mtx_init(&kring->q_lock, (t == NR_TX ? "nm_txq_lock" : "nm_rxq_lock"), NULL, MTX_DEF);
899 nm_os_selinfo_init(&kring->si);
901 nm_os_selinfo_init(&na->si[t]);
909 /* undo the actions performed by netmap_krings_create */
910 /* call with NMG_LOCK held */
912 netmap_krings_delete(struct netmap_adapter *na)
914 struct netmap_kring **kring = na->tx_rings;
917 if (na->tx_rings == NULL) {
918 if (netmap_debug & NM_DEBUG_ON)
919 nm_prerr("warning: krings were already deleted");
924 nm_os_selinfo_uninit(&na->si[t]);
926 /* we rely on the krings layout described above */
927 for ( ; kring != na->tailroom; kring++) {
928 mtx_destroy(&(*kring)->q_lock);
929 nm_os_selinfo_uninit(&(*kring)->si);
931 nm_os_free(na->tx_rings);
932 na->tx_rings = na->rx_rings = na->tailroom = NULL;
937 * Destructor for NIC ports. They also have an mbuf queue
938 * on the rings connected to the host so we need to purge
941 /* call with NMG_LOCK held */
943 netmap_hw_krings_delete(struct netmap_adapter *na)
945 u_int lim = netmap_real_rings(na, NR_RX), i;
947 for (i = nma_get_nrings(na, NR_RX); i < lim; i++) {
948 struct mbq *q = &NMR(na, NR_RX)[i]->rx_queue;
949 ND("destroy sw mbq with len %d", mbq_len(q));
953 netmap_krings_delete(na);
957 netmap_mem_drop(struct netmap_adapter *na)
959 int last = netmap_mem_deref(na->nm_mem, na);
960 /* if the native allocator had been overrided on regif,
961 * restore it now and drop the temporary one
963 if (last && na->nm_mem_prev) {
964 netmap_mem_put(na->nm_mem);
965 na->nm_mem = na->nm_mem_prev;
966 na->nm_mem_prev = NULL;
971 * Undo everything that was done in netmap_do_regif(). In particular,
972 * call nm_register(ifp,0) to stop netmap mode on the interface and
973 * revert to normal operation.
975 /* call with NMG_LOCK held */
976 static void netmap_unset_ringid(struct netmap_priv_d *);
977 static void netmap_krings_put(struct netmap_priv_d *);
979 netmap_do_unregif(struct netmap_priv_d *priv)
981 struct netmap_adapter *na = priv->np_na;
985 /* unset nr_pending_mode and possibly release exclusive mode */
986 netmap_krings_put(priv);
989 /* XXX check whether we have to do something with monitor
990 * when rings change nr_mode. */
991 if (na->active_fds <= 0) {
992 /* walk through all the rings and tell any monitor
993 * that the port is going to exit netmap mode
995 netmap_monitor_stop(na);
999 if (na->active_fds <= 0 || nm_kring_pending(priv)) {
1000 na->nm_register(na, 0);
1003 /* delete rings and buffers that are no longer needed */
1004 netmap_mem_rings_delete(na);
1006 if (na->active_fds <= 0) { /* last instance */
1008 * (TO CHECK) We enter here
1009 * when the last reference to this file descriptor goes
1010 * away. This means we cannot have any pending poll()
1011 * or interrupt routine operating on the structure.
1012 * XXX The file may be closed in a thread while
1013 * another thread is using it.
1014 * Linux keeps the file opened until the last reference
1015 * by any outstanding ioctl/poll or mmap is gone.
1016 * FreeBSD does not track mmap()s (but we do) and
1017 * wakes up any sleeping poll(). Need to check what
1018 * happens if the close() occurs while a concurrent
1019 * syscall is running.
1021 if (netmap_debug & NM_DEBUG_ON)
1022 nm_prinf("deleting last instance for %s", na->name);
1024 if (nm_netmap_on(na)) {
1025 nm_prerr("BUG: netmap on while going to delete the krings");
1028 na->nm_krings_delete(na);
1031 /* possibily decrement counter of tx_si/rx_si users */
1032 netmap_unset_ringid(priv);
1033 /* delete the nifp */
1034 netmap_mem_if_delete(na, priv->np_nifp);
1035 /* drop the allocator */
1036 netmap_mem_drop(na);
1037 /* mark the priv as unregistered */
1039 priv->np_nifp = NULL;
1042 struct netmap_priv_d*
1043 netmap_priv_new(void)
1045 struct netmap_priv_d *priv;
1047 priv = nm_os_malloc(sizeof(struct netmap_priv_d));
1056 * Destructor of the netmap_priv_d, called when the fd is closed
1057 * Action: undo all the things done by NIOCREGIF,
1058 * On FreeBSD we need to track whether there are active mmap()s,
1059 * and we use np_active_mmaps for that. On linux, the field is always 0.
1060 * Return: 1 if we can free priv, 0 otherwise.
1063 /* call with NMG_LOCK held */
1065 netmap_priv_delete(struct netmap_priv_d *priv)
1067 struct netmap_adapter *na = priv->np_na;
1069 /* number of active references to this fd */
1070 if (--priv->np_refs > 0) {
1075 netmap_do_unregif(priv);
1077 netmap_unget_na(na, priv->np_ifp);
1078 bzero(priv, sizeof(*priv)); /* for safety */
1083 /* call with NMG_LOCK *not* held */
1085 netmap_dtor(void *data)
1087 struct netmap_priv_d *priv = data;
1090 netmap_priv_delete(priv);
1096 * Handlers for synchronization of the rings from/to the host stack.
1097 * These are associated to a network interface and are just another
1098 * ring pair managed by userspace.
1100 * Netmap also supports transparent forwarding (NS_FORWARD and NR_FORWARD
1103 * - Before releasing buffers on hw RX rings, the application can mark
1104 * them with the NS_FORWARD flag. During the next RXSYNC or poll(), they
1105 * will be forwarded to the host stack, similarly to what happened if
1106 * the application moved them to the host TX ring.
1108 * - Before releasing buffers on the host RX ring, the application can
1109 * mark them with the NS_FORWARD flag. During the next RXSYNC or poll(),
1110 * they will be forwarded to the hw TX rings, saving the application
1111 * from doing the same task in user-space.
1113 * Transparent fowarding can be enabled per-ring, by setting the NR_FORWARD
1114 * flag, or globally with the netmap_fwd sysctl.
1116 * The transfer NIC --> host is relatively easy, just encapsulate
1117 * into mbufs and we are done. The host --> NIC side is slightly
1118 * harder because there might not be room in the tx ring so it
1119 * might take a while before releasing the buffer.
1124 * Pass a whole queue of mbufs to the host stack as coming from 'dst'
1125 * We do not need to lock because the queue is private.
1126 * After this call the queue is empty.
1129 netmap_send_up(struct ifnet *dst, struct mbq *q)
1132 struct mbuf *head = NULL, *prev = NULL;
1134 /* Send packets up, outside the lock; head/prev machinery
1135 * is only useful for Windows. */
1136 while ((m = mbq_dequeue(q)) != NULL) {
1137 if (netmap_debug & NM_DEBUG_HOST)
1138 nm_prinf("sending up pkt %p size %d", m, MBUF_LEN(m));
1139 prev = nm_os_send_up(dst, m, prev);
1144 nm_os_send_up(dst, NULL, head);
1150 * Scan the buffers from hwcur to ring->head, and put a copy of those
1151 * marked NS_FORWARD (or all of them if forced) into a queue of mbufs.
1152 * Drop remaining packets in the unlikely event
1153 * of an mbuf shortage.
1156 netmap_grab_packets(struct netmap_kring *kring, struct mbq *q, int force)
1158 u_int const lim = kring->nkr_num_slots - 1;
1159 u_int const head = kring->rhead;
1161 struct netmap_adapter *na = kring->na;
1163 for (n = kring->nr_hwcur; n != head; n = nm_next(n, lim)) {
1165 struct netmap_slot *slot = &kring->ring->slot[n];
1167 if ((slot->flags & NS_FORWARD) == 0 && !force)
1169 if (slot->len < 14 || slot->len > NETMAP_BUF_SIZE(na)) {
1170 RD(5, "bad pkt at %d len %d", n, slot->len);
1173 slot->flags &= ~NS_FORWARD; // XXX needed ?
1174 /* XXX TODO: adapt to the case of a multisegment packet */
1175 m = m_devget(NMB(na, slot), slot->len, 0, na->ifp, NULL);
1184 _nm_may_forward(struct netmap_kring *kring)
1186 return ((netmap_fwd || kring->ring->flags & NR_FORWARD) &&
1187 kring->na->na_flags & NAF_HOST_RINGS &&
1188 kring->tx == NR_RX);
1192 nm_may_forward_up(struct netmap_kring *kring)
1194 return _nm_may_forward(kring) &&
1195 kring->ring_id != kring->na->num_rx_rings;
1199 nm_may_forward_down(struct netmap_kring *kring, int sync_flags)
1201 return _nm_may_forward(kring) &&
1202 (sync_flags & NAF_CAN_FORWARD_DOWN) &&
1203 kring->ring_id == kring->na->num_rx_rings;
1207 * Send to the NIC rings packets marked NS_FORWARD between
1208 * kring->nr_hwcur and kring->rhead.
1209 * Called under kring->rx_queue.lock on the sw rx ring.
1211 * It can only be called if the user opened all the TX hw rings,
1212 * see NAF_CAN_FORWARD_DOWN flag.
1213 * We can touch the TX netmap rings (slots, head and cur) since
1214 * we are in poll/ioctl system call context, and the application
1215 * is not supposed to touch the ring (using a different thread)
1216 * during the execution of the system call.
1219 netmap_sw_to_nic(struct netmap_adapter *na)
1221 struct netmap_kring *kring = na->rx_rings[na->num_rx_rings];
1222 struct netmap_slot *rxslot = kring->ring->slot;
1223 u_int i, rxcur = kring->nr_hwcur;
1224 u_int const head = kring->rhead;
1225 u_int const src_lim = kring->nkr_num_slots - 1;
1228 /* scan rings to find space, then fill as much as possible */
1229 for (i = 0; i < na->num_tx_rings; i++) {
1230 struct netmap_kring *kdst = na->tx_rings[i];
1231 struct netmap_ring *rdst = kdst->ring;
1232 u_int const dst_lim = kdst->nkr_num_slots - 1;
1234 /* XXX do we trust ring or kring->rcur,rtail ? */
1235 for (; rxcur != head && !nm_ring_empty(rdst);
1236 rxcur = nm_next(rxcur, src_lim) ) {
1237 struct netmap_slot *src, *dst, tmp;
1238 u_int dst_head = rdst->head;
1240 src = &rxslot[rxcur];
1241 if ((src->flags & NS_FORWARD) == 0 && !netmap_fwd)
1246 dst = &rdst->slot[dst_head];
1250 src->buf_idx = dst->buf_idx;
1251 src->flags = NS_BUF_CHANGED;
1253 dst->buf_idx = tmp.buf_idx;
1255 dst->flags = NS_BUF_CHANGED;
1257 rdst->head = rdst->cur = nm_next(dst_head, dst_lim);
1259 /* if (sent) XXX txsync ? it would be just an optimization */
1266 * netmap_txsync_to_host() passes packets up. We are called from a
1267 * system call in user process context, and the only contention
1268 * can be among multiple user threads erroneously calling
1269 * this routine concurrently.
1272 netmap_txsync_to_host(struct netmap_kring *kring, int flags)
1274 struct netmap_adapter *na = kring->na;
1275 u_int const lim = kring->nkr_num_slots - 1;
1276 u_int const head = kring->rhead;
1279 /* Take packets from hwcur to head and pass them up.
1280 * Force hwcur = head since netmap_grab_packets() stops at head
1283 netmap_grab_packets(kring, &q, 1 /* force */);
1284 ND("have %d pkts in queue", mbq_len(&q));
1285 kring->nr_hwcur = head;
1286 kring->nr_hwtail = head + lim;
1287 if (kring->nr_hwtail > lim)
1288 kring->nr_hwtail -= lim + 1;
1290 netmap_send_up(na->ifp, &q);
1296 * rxsync backend for packets coming from the host stack.
1297 * They have been put in kring->rx_queue by netmap_transmit().
1298 * We protect access to the kring using kring->rx_queue.lock
1300 * also moves to the nic hw rings any packet the user has marked
1301 * for transparent-mode forwarding, then sets the NR_FORWARD
1302 * flag in the kring to let the caller push them out
1305 netmap_rxsync_from_host(struct netmap_kring *kring, int flags)
1307 struct netmap_adapter *na = kring->na;
1308 struct netmap_ring *ring = kring->ring;
1310 u_int const lim = kring->nkr_num_slots - 1;
1311 u_int const head = kring->rhead;
1313 struct mbq *q = &kring->rx_queue, fq;
1315 mbq_init(&fq); /* fq holds packets to be freed */
1319 /* First part: import newly received packets */
1321 if (n) { /* grab packets from the queue */
1325 nm_i = kring->nr_hwtail;
1326 stop_i = nm_prev(kring->nr_hwcur, lim);
1327 while ( nm_i != stop_i && (m = mbq_dequeue(q)) != NULL ) {
1328 int len = MBUF_LEN(m);
1329 struct netmap_slot *slot = &ring->slot[nm_i];
1331 m_copydata(m, 0, len, NMB(na, slot));
1332 ND("nm %d len %d", nm_i, len);
1333 if (netmap_debug & NM_DEBUG_HOST)
1334 nm_prinf("%s", nm_dump_buf(NMB(na, slot),len, 128, NULL));
1338 nm_i = nm_next(nm_i, lim);
1339 mbq_enqueue(&fq, m);
1341 kring->nr_hwtail = nm_i;
1345 * Second part: skip past packets that userspace has released.
1347 nm_i = kring->nr_hwcur;
1348 if (nm_i != head) { /* something was released */
1349 if (nm_may_forward_down(kring, flags)) {
1350 ret = netmap_sw_to_nic(na);
1352 kring->nr_kflags |= NR_FORWARD;
1356 kring->nr_hwcur = head;
1368 /* Get a netmap adapter for the port.
1370 * If it is possible to satisfy the request, return 0
1371 * with *na containing the netmap adapter found.
1372 * Otherwise return an error code, with *na containing NULL.
1374 * When the port is attached to a bridge, we always return
1376 * Otherwise, if the port is already bound to a file descriptor,
1377 * then we unconditionally return the existing adapter into *na.
1378 * In all the other cases, we return (into *na) either native,
1379 * generic or NULL, according to the following table:
1382 * active_fds dev.netmap.admode YES NO
1383 * -------------------------------------------------------
1384 * >0 * NA(ifp) NA(ifp)
1386 * 0 NETMAP_ADMODE_BEST NATIVE GENERIC
1387 * 0 NETMAP_ADMODE_NATIVE NATIVE NULL
1388 * 0 NETMAP_ADMODE_GENERIC GENERIC GENERIC
1391 static void netmap_hw_dtor(struct netmap_adapter *); /* needed by NM_IS_NATIVE() */
1393 netmap_get_hw_na(struct ifnet *ifp, struct netmap_mem_d *nmd, struct netmap_adapter **na)
1395 /* generic support */
1396 int i = netmap_admode; /* Take a snapshot. */
1397 struct netmap_adapter *prev_na;
1400 *na = NULL; /* default */
1402 /* reset in case of invalid value */
1403 if (i < NETMAP_ADMODE_BEST || i >= NETMAP_ADMODE_LAST)
1404 i = netmap_admode = NETMAP_ADMODE_BEST;
1406 if (NM_NA_VALID(ifp)) {
1408 /* If an adapter already exists, return it if
1409 * there are active file descriptors or if
1410 * netmap is not forced to use generic
1413 if (NETMAP_OWNED_BY_ANY(prev_na)
1414 || i != NETMAP_ADMODE_GENERIC
1415 || prev_na->na_flags & NAF_FORCE_NATIVE
1417 /* ugly, but we cannot allow an adapter switch
1418 * if some pipe is referring to this one
1420 || prev_na->na_next_pipe > 0
1428 /* If there isn't native support and netmap is not allowed
1429 * to use generic adapters, we cannot satisfy the request.
1431 if (!NM_IS_NATIVE(ifp) && i == NETMAP_ADMODE_NATIVE)
1434 /* Otherwise, create a generic adapter and return it,
1435 * saving the previously used netmap adapter, if any.
1437 * Note that here 'prev_na', if not NULL, MUST be a
1438 * native adapter, and CANNOT be a generic one. This is
1439 * true because generic adapters are created on demand, and
1440 * destroyed when not used anymore. Therefore, if the adapter
1441 * currently attached to an interface 'ifp' is generic, it
1443 * (NA(ifp)->active_fds > 0 || NETMAP_OWNED_BY_KERN(NA(ifp))).
1444 * Consequently, if NA(ifp) is generic, we will enter one of
1445 * the branches above. This ensures that we never override
1446 * a generic adapter with another generic adapter.
1448 error = generic_netmap_attach(ifp);
1455 if (nmd != NULL && !((*na)->na_flags & NAF_MEM_OWNER) &&
1456 (*na)->active_fds == 0 && ((*na)->nm_mem != nmd)) {
1457 (*na)->nm_mem_prev = (*na)->nm_mem;
1458 (*na)->nm_mem = netmap_mem_get(nmd);
1465 * MUST BE CALLED UNDER NMG_LOCK()
1467 * Get a refcounted reference to a netmap adapter attached
1468 * to the interface specified by req.
1469 * This is always called in the execution of an ioctl().
1471 * Return ENXIO if the interface specified by the request does
1472 * not exist, ENOTSUP if netmap is not supported by the interface,
1473 * EBUSY if the interface is already attached to a bridge,
1474 * EINVAL if parameters are invalid, ENOMEM if needed resources
1475 * could not be allocated.
1476 * If successful, hold a reference to the netmap adapter.
1478 * If the interface specified by req is a system one, also keep
1479 * a reference to it and return a valid *ifp.
1482 netmap_get_na(struct nmreq_header *hdr,
1483 struct netmap_adapter **na, struct ifnet **ifp,
1484 struct netmap_mem_d *nmd, int create)
1486 struct nmreq_register *req = (struct nmreq_register *)(uintptr_t)hdr->nr_body;
1488 struct netmap_adapter *ret = NULL;
1491 *na = NULL; /* default return value */
1494 if (hdr->nr_reqtype != NETMAP_REQ_REGISTER) {
1498 if (req->nr_mode == NR_REG_PIPE_MASTER ||
1499 req->nr_mode == NR_REG_PIPE_SLAVE) {
1500 /* Do not accept deprecated pipe modes. */
1501 nm_prerr("Deprecated pipe nr_mode, use xx{yy or xx}yy syntax");
1507 /* if the request contain a memid, try to find the
1508 * corresponding memory region
1510 if (nmd == NULL && req->nr_mem_id) {
1511 nmd = netmap_mem_find(req->nr_mem_id);
1514 /* keep the rereference */
1518 /* We cascade through all possible types of netmap adapter.
1519 * All netmap_get_*_na() functions return an error and an na,
1520 * with the following combinations:
1523 * 0 NULL type doesn't match
1524 * !0 NULL type matches, but na creation/lookup failed
1525 * 0 !NULL type matches and na created/found
1526 * !0 !NULL impossible
1528 error = netmap_get_null_na(hdr, na, nmd, create);
1529 if (error || *na != NULL)
1532 /* try to see if this is a monitor port */
1533 error = netmap_get_monitor_na(hdr, na, nmd, create);
1534 if (error || *na != NULL)
1537 /* try to see if this is a pipe port */
1538 error = netmap_get_pipe_na(hdr, na, nmd, create);
1539 if (error || *na != NULL)
1542 /* try to see if this is a bridge port */
1543 error = netmap_get_vale_na(hdr, na, nmd, create);
1547 if (*na != NULL) /* valid match in netmap_get_bdg_na() */
1551 * This must be a hardware na, lookup the name in the system.
1552 * Note that by hardware we actually mean "it shows up in ifconfig".
1553 * This may still be a tap, a veth/epair, or even a
1554 * persistent VALE port.
1556 *ifp = ifunit_ref(hdr->nr_name);
1562 error = netmap_get_hw_na(*ifp, nmd, &ret);
1567 netmap_adapter_get(ret);
1572 netmap_adapter_put(ret);
1579 netmap_mem_put(nmd);
1584 /* undo netmap_get_na() */
1586 netmap_unget_na(struct netmap_adapter *na, struct ifnet *ifp)
1591 netmap_adapter_put(na);
1595 #define NM_FAIL_ON(t) do { \
1596 if (unlikely(t)) { \
1597 RD(5, "%s: fail '" #t "' " \
1599 "rh %d rc %d rt %d " \
1602 head, cur, ring->tail, \
1603 kring->rhead, kring->rcur, kring->rtail, \
1604 kring->nr_hwcur, kring->nr_hwtail); \
1605 return kring->nkr_num_slots; \
1610 * validate parameters on entry for *_txsync()
1611 * Returns ring->cur if ok, or something >= kring->nkr_num_slots
1614 * rhead, rcur and rtail=hwtail are stored from previous round.
1615 * hwcur is the next packet to send to the ring.
1618 * hwcur <= *rhead <= head <= cur <= tail = *rtail <= hwtail
1620 * hwcur, rhead, rtail and hwtail are reliable
1623 nm_txsync_prologue(struct netmap_kring *kring, struct netmap_ring *ring)
1625 u_int head = ring->head; /* read only once */
1626 u_int cur = ring->cur; /* read only once */
1627 u_int n = kring->nkr_num_slots;
1629 ND(5, "%s kcur %d ktail %d head %d cur %d tail %d",
1631 kring->nr_hwcur, kring->nr_hwtail,
1632 ring->head, ring->cur, ring->tail);
1633 #if 1 /* kernel sanity checks; but we can trust the kring. */
1634 NM_FAIL_ON(kring->nr_hwcur >= n || kring->rhead >= n ||
1635 kring->rtail >= n || kring->nr_hwtail >= n);
1636 #endif /* kernel sanity checks */
1638 * user sanity checks. We only use head,
1639 * A, B, ... are possible positions for head:
1641 * 0 A rhead B rtail C n-1
1642 * 0 D rtail E rhead F n-1
1644 * B, F, D are valid. A, C, E are wrong
1646 if (kring->rtail >= kring->rhead) {
1647 /* want rhead <= head <= rtail */
1648 NM_FAIL_ON(head < kring->rhead || head > kring->rtail);
1649 /* and also head <= cur <= rtail */
1650 NM_FAIL_ON(cur < head || cur > kring->rtail);
1651 } else { /* here rtail < rhead */
1652 /* we need head outside rtail .. rhead */
1653 NM_FAIL_ON(head > kring->rtail && head < kring->rhead);
1655 /* two cases now: head <= rtail or head >= rhead */
1656 if (head <= kring->rtail) {
1657 /* want head <= cur <= rtail */
1658 NM_FAIL_ON(cur < head || cur > kring->rtail);
1659 } else { /* head >= rhead */
1660 /* cur must be outside rtail..head */
1661 NM_FAIL_ON(cur > kring->rtail && cur < head);
1664 if (ring->tail != kring->rtail) {
1665 RD(5, "%s tail overwritten was %d need %d", kring->name,
1666 ring->tail, kring->rtail);
1667 ring->tail = kring->rtail;
1669 kring->rhead = head;
1676 * validate parameters on entry for *_rxsync()
1677 * Returns ring->head if ok, kring->nkr_num_slots on error.
1679 * For a valid configuration,
1680 * hwcur <= head <= cur <= tail <= hwtail
1682 * We only consider head and cur.
1683 * hwcur and hwtail are reliable.
1687 nm_rxsync_prologue(struct netmap_kring *kring, struct netmap_ring *ring)
1689 uint32_t const n = kring->nkr_num_slots;
1692 ND(5,"%s kc %d kt %d h %d c %d t %d",
1694 kring->nr_hwcur, kring->nr_hwtail,
1695 ring->head, ring->cur, ring->tail);
1697 * Before storing the new values, we should check they do not
1698 * move backwards. However:
1699 * - head is not an issue because the previous value is hwcur;
1700 * - cur could in principle go back, however it does not matter
1701 * because we are processing a brand new rxsync()
1703 cur = kring->rcur = ring->cur; /* read only once */
1704 head = kring->rhead = ring->head; /* read only once */
1705 #if 1 /* kernel sanity checks */
1706 NM_FAIL_ON(kring->nr_hwcur >= n || kring->nr_hwtail >= n);
1707 #endif /* kernel sanity checks */
1708 /* user sanity checks */
1709 if (kring->nr_hwtail >= kring->nr_hwcur) {
1710 /* want hwcur <= rhead <= hwtail */
1711 NM_FAIL_ON(head < kring->nr_hwcur || head > kring->nr_hwtail);
1712 /* and also rhead <= rcur <= hwtail */
1713 NM_FAIL_ON(cur < head || cur > kring->nr_hwtail);
1715 /* we need rhead outside hwtail..hwcur */
1716 NM_FAIL_ON(head < kring->nr_hwcur && head > kring->nr_hwtail);
1717 /* two cases now: head <= hwtail or head >= hwcur */
1718 if (head <= kring->nr_hwtail) {
1719 /* want head <= cur <= hwtail */
1720 NM_FAIL_ON(cur < head || cur > kring->nr_hwtail);
1722 /* cur must be outside hwtail..head */
1723 NM_FAIL_ON(cur < head && cur > kring->nr_hwtail);
1726 if (ring->tail != kring->rtail) {
1727 RD(5, "%s tail overwritten was %d need %d",
1729 ring->tail, kring->rtail);
1730 ring->tail = kring->rtail;
1737 * Error routine called when txsync/rxsync detects an error.
1738 * Can't do much more than resetting head = cur = hwcur, tail = hwtail
1739 * Return 1 on reinit.
1741 * This routine is only called by the upper half of the kernel.
1742 * It only reads hwcur (which is changed only by the upper half, too)
1743 * and hwtail (which may be changed by the lower half, but only on
1744 * a tx ring and only to increase it, so any error will be recovered
1745 * on the next call). For the above, we don't strictly need to call
1749 netmap_ring_reinit(struct netmap_kring *kring)
1751 struct netmap_ring *ring = kring->ring;
1752 u_int i, lim = kring->nkr_num_slots - 1;
1755 // XXX KASSERT nm_kr_tryget
1756 RD(10, "called for %s", kring->name);
1757 // XXX probably wrong to trust userspace
1758 kring->rhead = ring->head;
1759 kring->rcur = ring->cur;
1760 kring->rtail = ring->tail;
1762 if (ring->cur > lim)
1764 if (ring->head > lim)
1766 if (ring->tail > lim)
1768 for (i = 0; i <= lim; i++) {
1769 u_int idx = ring->slot[i].buf_idx;
1770 u_int len = ring->slot[i].len;
1771 if (idx < 2 || idx >= kring->na->na_lut.objtotal) {
1772 RD(5, "bad index at slot %d idx %d len %d ", i, idx, len);
1773 ring->slot[i].buf_idx = 0;
1774 ring->slot[i].len = 0;
1775 } else if (len > NETMAP_BUF_SIZE(kring->na)) {
1776 ring->slot[i].len = 0;
1777 RD(5, "bad len at slot %d idx %d len %d", i, idx, len);
1781 RD(10, "total %d errors", errors);
1782 RD(10, "%s reinit, cur %d -> %d tail %d -> %d",
1784 ring->cur, kring->nr_hwcur,
1785 ring->tail, kring->nr_hwtail);
1786 ring->head = kring->rhead = kring->nr_hwcur;
1787 ring->cur = kring->rcur = kring->nr_hwcur;
1788 ring->tail = kring->rtail = kring->nr_hwtail;
1790 return (errors ? 1 : 0);
1793 /* interpret the ringid and flags fields of an nmreq, by translating them
1794 * into a pair of intervals of ring indices:
1796 * [priv->np_txqfirst, priv->np_txqlast) and
1797 * [priv->np_rxqfirst, priv->np_rxqlast)
1801 netmap_interp_ringid(struct netmap_priv_d *priv, uint32_t nr_mode,
1802 uint16_t nr_ringid, uint64_t nr_flags)
1804 struct netmap_adapter *na = priv->np_na;
1805 int excluded_direction[] = { NR_TX_RINGS_ONLY, NR_RX_RINGS_ONLY };
1810 if (nr_flags & excluded_direction[t]) {
1811 priv->np_qfirst[t] = priv->np_qlast[t] = 0;
1815 case NR_REG_ALL_NIC:
1817 priv->np_qfirst[t] = 0;
1818 priv->np_qlast[t] = nma_get_nrings(na, t);
1819 ND("ALL/PIPE: %s %d %d", nm_txrx2str(t),
1820 priv->np_qfirst[t], priv->np_qlast[t]);
1824 if (!(na->na_flags & NAF_HOST_RINGS)) {
1825 nm_prerr("host rings not supported");
1828 priv->np_qfirst[t] = (nr_mode == NR_REG_SW ?
1829 nma_get_nrings(na, t) : 0);
1830 priv->np_qlast[t] = netmap_all_rings(na, t);
1831 ND("%s: %s %d %d", nr_mode == NR_REG_SW ? "SW" : "NIC+SW",
1833 priv->np_qfirst[t], priv->np_qlast[t]);
1835 case NR_REG_ONE_NIC:
1836 if (nr_ringid >= na->num_tx_rings &&
1837 nr_ringid >= na->num_rx_rings) {
1838 nm_prerr("invalid ring id %d", nr_ringid);
1841 /* if not enough rings, use the first one */
1843 if (j >= nma_get_nrings(na, t))
1845 priv->np_qfirst[t] = j;
1846 priv->np_qlast[t] = j + 1;
1847 ND("ONE_NIC: %s %d %d", nm_txrx2str(t),
1848 priv->np_qfirst[t], priv->np_qlast[t]);
1851 nm_prerr("invalid regif type %d", nr_mode);
1855 priv->np_flags = nr_flags;
1857 /* Allow transparent forwarding mode in the host --> nic
1858 * direction only if all the TX hw rings have been opened. */
1859 if (priv->np_qfirst[NR_TX] == 0 &&
1860 priv->np_qlast[NR_TX] >= na->num_tx_rings) {
1861 priv->np_sync_flags |= NAF_CAN_FORWARD_DOWN;
1864 if (netmap_verbose) {
1865 nm_prinf("%s: tx [%d,%d) rx [%d,%d) id %d",
1867 priv->np_qfirst[NR_TX],
1868 priv->np_qlast[NR_TX],
1869 priv->np_qfirst[NR_RX],
1870 priv->np_qlast[NR_RX],
1878 * Set the ring ID. For devices with a single queue, a request
1879 * for all rings is the same as a single ring.
1882 netmap_set_ringid(struct netmap_priv_d *priv, uint32_t nr_mode,
1883 uint16_t nr_ringid, uint64_t nr_flags)
1885 struct netmap_adapter *na = priv->np_na;
1889 error = netmap_interp_ringid(priv, nr_mode, nr_ringid, nr_flags);
1894 priv->np_txpoll = (nr_flags & NR_NO_TX_POLL) ? 0 : 1;
1896 /* optimization: count the users registered for more than
1897 * one ring, which are the ones sleeping on the global queue.
1898 * The default netmap_notify() callback will then
1899 * avoid signaling the global queue if nobody is using it
1902 if (nm_si_user(priv, t))
1909 netmap_unset_ringid(struct netmap_priv_d *priv)
1911 struct netmap_adapter *na = priv->np_na;
1915 if (nm_si_user(priv, t))
1917 priv->np_qfirst[t] = priv->np_qlast[t] = 0;
1920 priv->np_txpoll = 0;
1921 priv->np_kloop_state = 0;
1925 /* Set the nr_pending_mode for the requested rings.
1926 * If requested, also try to get exclusive access to the rings, provided
1927 * the rings we want to bind are not exclusively owned by a previous bind.
1930 netmap_krings_get(struct netmap_priv_d *priv)
1932 struct netmap_adapter *na = priv->np_na;
1934 struct netmap_kring *kring;
1935 int excl = (priv->np_flags & NR_EXCLUSIVE);
1938 if (netmap_debug & NM_DEBUG_ON)
1939 nm_prinf("%s: grabbing tx [%d, %d) rx [%d, %d)",
1941 priv->np_qfirst[NR_TX],
1942 priv->np_qlast[NR_TX],
1943 priv->np_qfirst[NR_RX],
1944 priv->np_qlast[NR_RX]);
1946 /* first round: check that all the requested rings
1947 * are neither alread exclusively owned, nor we
1948 * want exclusive ownership when they are already in use
1951 for (i = priv->np_qfirst[t]; i < priv->np_qlast[t]; i++) {
1952 kring = NMR(na, t)[i];
1953 if ((kring->nr_kflags & NKR_EXCLUSIVE) ||
1954 (kring->users && excl))
1956 ND("ring %s busy", kring->name);
1962 /* second round: increment usage count (possibly marking them
1963 * as exclusive) and set the nr_pending_mode
1966 for (i = priv->np_qfirst[t]; i < priv->np_qlast[t]; i++) {
1967 kring = NMR(na, t)[i];
1970 kring->nr_kflags |= NKR_EXCLUSIVE;
1971 kring->nr_pending_mode = NKR_NETMAP_ON;
1979 /* Undo netmap_krings_get(). This is done by clearing the exclusive mode
1980 * if was asked on regif, and unset the nr_pending_mode if we are the
1981 * last users of the involved rings. */
1983 netmap_krings_put(struct netmap_priv_d *priv)
1985 struct netmap_adapter *na = priv->np_na;
1987 struct netmap_kring *kring;
1988 int excl = (priv->np_flags & NR_EXCLUSIVE);
1991 ND("%s: releasing tx [%d, %d) rx [%d, %d)",
1993 priv->np_qfirst[NR_TX],
1994 priv->np_qlast[NR_TX],
1995 priv->np_qfirst[NR_RX],
1996 priv->np_qlast[MR_RX]);
1999 for (i = priv->np_qfirst[t]; i < priv->np_qlast[t]; i++) {
2000 kring = NMR(na, t)[i];
2002 kring->nr_kflags &= ~NKR_EXCLUSIVE;
2004 if (kring->users == 0)
2005 kring->nr_pending_mode = NKR_NETMAP_OFF;
2011 nm_priv_rx_enabled(struct netmap_priv_d *priv)
2013 return (priv->np_qfirst[NR_RX] != priv->np_qlast[NR_RX]);
2016 /* Validate the CSB entries for both directions (atok and ktoa).
2017 * To be called under NMG_LOCK(). */
2019 netmap_csb_validate(struct netmap_priv_d *priv, struct nmreq_opt_csb *csbo)
2021 struct nm_csb_atok *csb_atok_base =
2022 (struct nm_csb_atok *)(uintptr_t)csbo->csb_atok;
2023 struct nm_csb_ktoa *csb_ktoa_base =
2024 (struct nm_csb_ktoa *)(uintptr_t)csbo->csb_ktoa;
2026 int num_rings[NR_TXRX], tot_rings;
2027 size_t entry_size[2];
2031 if (priv->np_kloop_state & NM_SYNC_KLOOP_RUNNING) {
2032 nm_prerr("Cannot update CSB while kloop is running");
2038 num_rings[t] = priv->np_qlast[t] - priv->np_qfirst[t];
2039 tot_rings += num_rings[t];
2044 if (!(priv->np_flags & NR_EXCLUSIVE)) {
2045 nm_prerr("CSB mode requires NR_EXCLUSIVE");
2049 entry_size[0] = sizeof(*csb_atok_base);
2050 entry_size[1] = sizeof(*csb_ktoa_base);
2051 csb_start[0] = (void *)csb_atok_base;
2052 csb_start[1] = (void *)csb_ktoa_base;
2054 for (i = 0; i < 2; i++) {
2055 /* On Linux we could use access_ok() to simplify
2056 * the validation. However, the advantage of
2057 * this approach is that it works also on
2059 size_t csb_size = tot_rings * entry_size[i];
2063 if ((uintptr_t)csb_start[i] & (entry_size[i]-1)) {
2064 nm_prerr("Unaligned CSB address");
2068 tmp = nm_os_malloc(csb_size);
2072 /* Application --> kernel direction. */
2073 err = copyin(csb_start[i], tmp, csb_size);
2075 /* Kernel --> application direction. */
2076 memset(tmp, 0, csb_size);
2077 err = copyout(tmp, csb_start[i], csb_size);
2081 nm_prerr("Invalid CSB address");
2086 priv->np_csb_atok_base = csb_atok_base;
2087 priv->np_csb_ktoa_base = csb_ktoa_base;
2089 /* Initialize the CSB. */
2091 for (i = 0; i < num_rings[t]; i++) {
2092 struct netmap_kring *kring =
2093 NMR(priv->np_na, t)[i + priv->np_qfirst[t]];
2094 struct nm_csb_atok *csb_atok = csb_atok_base + i;
2095 struct nm_csb_ktoa *csb_ktoa = csb_ktoa_base + i;
2098 csb_atok += num_rings[NR_TX];
2099 csb_ktoa += num_rings[NR_TX];
2102 CSB_WRITE(csb_atok, head, kring->rhead);
2103 CSB_WRITE(csb_atok, cur, kring->rcur);
2104 CSB_WRITE(csb_atok, appl_need_kick, 1);
2105 CSB_WRITE(csb_atok, sync_flags, 1);
2106 CSB_WRITE(csb_ktoa, hwcur, kring->nr_hwcur);
2107 CSB_WRITE(csb_ktoa, hwtail, kring->nr_hwtail);
2108 CSB_WRITE(csb_ktoa, kern_need_kick, 1);
2110 nm_prinf("csb_init for kring %s: head %u, cur %u, "
2111 "hwcur %u, hwtail %u", kring->name,
2112 kring->rhead, kring->rcur, kring->nr_hwcur,
2121 * possibly move the interface to netmap-mode.
2122 * If success it returns a pointer to netmap_if, otherwise NULL.
2123 * This must be called with NMG_LOCK held.
2125 * The following na callbacks are called in the process:
2127 * na->nm_config() [by netmap_update_config]
2128 * (get current number and size of rings)
2130 * We have a generic one for linux (netmap_linux_config).
2131 * The bwrap has to override this, since it has to forward
2132 * the request to the wrapped adapter (netmap_bwrap_config).
2135 * na->nm_krings_create()
2136 * (create and init the krings array)
2138 * One of the following:
2140 * * netmap_hw_krings_create, (hw ports)
2141 * creates the standard layout for the krings
2142 * and adds the mbq (used for the host rings).
2144 * * netmap_vp_krings_create (VALE ports)
2145 * add leases and scratchpads
2147 * * netmap_pipe_krings_create (pipes)
2148 * create the krings and rings of both ends and
2151 * * netmap_monitor_krings_create (monitors)
2152 * avoid allocating the mbq
2154 * * netmap_bwrap_krings_create (bwraps)
2155 * create both the brap krings array,
2156 * the krings array of the wrapped adapter, and
2157 * (if needed) the fake array for the host adapter
2159 * na->nm_register(, 1)
2160 * (put the adapter in netmap mode)
2162 * This may be one of the following:
2164 * * netmap_hw_reg (hw ports)
2165 * checks that the ifp is still there, then calls
2166 * the hardware specific callback;
2168 * * netmap_vp_reg (VALE ports)
2169 * If the port is connected to a bridge,
2170 * set the NAF_NETMAP_ON flag under the
2171 * bridge write lock.
2173 * * netmap_pipe_reg (pipes)
2174 * inform the other pipe end that it is no
2175 * longer responsible for the lifetime of this
2178 * * netmap_monitor_reg (monitors)
2179 * intercept the sync callbacks of the monitored
2182 * * netmap_bwrap_reg (bwraps)
2183 * cross-link the bwrap and hwna rings,
2184 * forward the request to the hwna, override
2185 * the hwna notify callback (to get the frames
2186 * coming from outside go through the bridge).
2191 netmap_do_regif(struct netmap_priv_d *priv, struct netmap_adapter *na,
2192 uint32_t nr_mode, uint16_t nr_ringid, uint64_t nr_flags)
2194 struct netmap_if *nifp = NULL;
2198 priv->np_na = na; /* store the reference */
2199 error = netmap_mem_finalize(na->nm_mem, na);
2203 if (na->active_fds == 0) {
2205 /* cache the allocator info in the na */
2206 error = netmap_mem_get_lut(na->nm_mem, &na->na_lut);
2209 ND("lut %p bufs %u size %u", na->na_lut.lut, na->na_lut.objtotal,
2210 na->na_lut.objsize);
2212 /* ring configuration may have changed, fetch from the card */
2213 netmap_update_config(na);
2216 /* compute the range of tx and rx rings to monitor */
2217 error = netmap_set_ringid(priv, nr_mode, nr_ringid, nr_flags);
2221 if (na->active_fds == 0) {
2223 * If this is the first registration of the adapter,
2224 * perform sanity checks and create the in-kernel view
2225 * of the netmap rings (the netmap krings).
2227 if (na->ifp && nm_priv_rx_enabled(priv)) {
2228 /* This netmap adapter is attached to an ifnet. */
2229 unsigned nbs = NETMAP_BUF_SIZE(na);
2230 unsigned mtu = nm_os_ifnet_mtu(na->ifp);
2232 ND("%s: mtu %d rx_buf_maxsize %d netmap_buf_size %d",
2233 na->name, mtu, na->rx_buf_maxsize, nbs);
2235 if (na->rx_buf_maxsize == 0) {
2236 nm_prerr("%s: error: rx_buf_maxsize == 0", na->name);
2241 if (mtu <= na->rx_buf_maxsize) {
2242 /* The MTU fits a single NIC slot. We only
2243 * Need to check that netmap buffers are
2244 * large enough to hold an MTU. NS_MOREFRAG
2245 * cannot be used in this case. */
2247 nm_prerr("error: netmap buf size (%u) "
2248 "< device MTU (%u)", nbs, mtu);
2253 /* More NIC slots may be needed to receive
2254 * or transmit a single packet. Check that
2255 * the adapter supports NS_MOREFRAG and that
2256 * netmap buffers are large enough to hold
2257 * the maximum per-slot size. */
2258 if (!(na->na_flags & NAF_MOREFRAG)) {
2259 nm_prerr("error: large MTU (%d) needed "
2260 "but %s does not support "
2265 } else if (nbs < na->rx_buf_maxsize) {
2266 nm_prerr("error: using NS_MOREFRAG on "
2267 "%s requires netmap buf size "
2268 ">= %u", na->ifp->if_xname,
2269 na->rx_buf_maxsize);
2273 nm_prinf("info: netmap application on "
2274 "%s needs to support "
2276 "(MTU=%u,netmap_buf_size=%u)",
2277 na->ifp->if_xname, mtu, nbs);
2283 * Depending on the adapter, this may also create
2284 * the netmap rings themselves
2286 error = na->nm_krings_create(na);
2292 /* now the krings must exist and we can check whether some
2293 * previous bind has exclusive ownership on them, and set
2296 error = netmap_krings_get(priv);
2298 goto err_del_krings;
2300 /* create all needed missing netmap rings */
2301 error = netmap_mem_rings_create(na);
2305 /* in all cases, create a new netmap if */
2306 nifp = netmap_mem_if_new(na, priv);
2312 if (nm_kring_pending(priv)) {
2313 /* Some kring is switching mode, tell the adapter to
2315 error = na->nm_register(na, 1);
2320 /* Commit the reference. */
2324 * advertise that the interface is ready by setting np_nifp.
2325 * The barrier is needed because readers (poll, *SYNC and mmap)
2326 * check for priv->np_nifp != NULL without locking
2328 mb(); /* make sure previous writes are visible to all CPUs */
2329 priv->np_nifp = nifp;
2334 netmap_mem_if_delete(na, nifp);
2336 netmap_krings_put(priv);
2337 netmap_mem_rings_delete(na);
2339 if (na->active_fds == 0)
2340 na->nm_krings_delete(na);
2342 if (na->active_fds == 0)
2343 memset(&na->na_lut, 0, sizeof(na->na_lut));
2345 netmap_mem_drop(na);
2353 * update kring and ring at the end of rxsync/txsync.
2356 nm_sync_finalize(struct netmap_kring *kring)
2359 * Update ring tail to what the kernel knows
2360 * After txsync: head/rhead/hwcur might be behind cur/rcur
2363 kring->ring->tail = kring->rtail = kring->nr_hwtail;
2365 ND(5, "%s now hwcur %d hwtail %d head %d cur %d tail %d",
2366 kring->name, kring->nr_hwcur, kring->nr_hwtail,
2367 kring->rhead, kring->rcur, kring->rtail);
2370 /* set ring timestamp */
2372 ring_timestamp_set(struct netmap_ring *ring)
2374 if (netmap_no_timestamp == 0 || ring->flags & NR_TIMESTAMP) {
2375 microtime(&ring->ts);
2379 static int nmreq_copyin(struct nmreq_header *, int);
2380 static int nmreq_copyout(struct nmreq_header *, int);
2381 static int nmreq_checkoptions(struct nmreq_header *);
2384 * ioctl(2) support for the "netmap" device.
2386 * Following a list of accepted commands:
2387 * - NIOCCTRL device control API
2388 * - NIOCTXSYNC sync TX rings
2389 * - NIOCRXSYNC sync RX rings
2390 * - SIOCGIFADDR just for convenience
2391 * - NIOCGINFO deprecated (legacy API)
2392 * - NIOCREGIF deprecated (legacy API)
2394 * Return 0 on success, errno otherwise.
2397 netmap_ioctl(struct netmap_priv_d *priv, u_long cmd, caddr_t data,
2398 struct thread *td, int nr_body_is_user)
2400 struct mbq q; /* packets from RX hw queues to host stack */
2401 struct netmap_adapter *na = NULL;
2402 struct netmap_mem_d *nmd = NULL;
2403 struct ifnet *ifp = NULL;
2405 u_int i, qfirst, qlast;
2406 struct netmap_kring **krings;
2412 struct nmreq_header *hdr = (struct nmreq_header *)data;
2414 if (hdr->nr_version < NETMAP_MIN_API ||
2415 hdr->nr_version > NETMAP_MAX_API) {
2416 nm_prerr("API mismatch: got %d need %d",
2417 hdr->nr_version, NETMAP_API);
2421 /* Make a kernel-space copy of the user-space nr_body.
2422 * For convenince, the nr_body pointer and the pointers
2423 * in the options list will be replaced with their
2424 * kernel-space counterparts. The original pointers are
2425 * saved internally and later restored by nmreq_copyout
2427 error = nmreq_copyin(hdr, nr_body_is_user);
2432 /* Sanitize hdr->nr_name. */
2433 hdr->nr_name[sizeof(hdr->nr_name) - 1] = '\0';
2435 switch (hdr->nr_reqtype) {
2436 case NETMAP_REQ_REGISTER: {
2437 struct nmreq_register *req =
2438 (struct nmreq_register *)(uintptr_t)hdr->nr_body;
2439 struct netmap_if *nifp;
2441 /* Protect access to priv from concurrent requests. */
2444 struct nmreq_option *opt;
2447 if (priv->np_nifp != NULL) { /* thread already registered */
2453 opt = nmreq_findoption((struct nmreq_option *)(uintptr_t)hdr->nr_options,
2454 NETMAP_REQ_OPT_EXTMEM);
2456 struct nmreq_opt_extmem *e =
2457 (struct nmreq_opt_extmem *)opt;
2459 error = nmreq_checkduplicate(opt);
2461 opt->nro_status = error;
2464 nmd = netmap_mem_ext_create(e->nro_usrptr,
2465 &e->nro_info, &error);
2466 opt->nro_status = error;
2470 #endif /* WITH_EXTMEM */
2472 if (nmd == NULL && req->nr_mem_id) {
2473 /* find the allocator and get a reference */
2474 nmd = netmap_mem_find(req->nr_mem_id);
2476 if (netmap_verbose) {
2477 nm_prerr("%s: failed to find mem_id %u",
2478 hdr->nr_name, req->nr_mem_id);
2484 /* find the interface and a reference */
2485 error = netmap_get_na(hdr, &na, &ifp, nmd,
2486 1 /* create */); /* keep reference */
2489 if (NETMAP_OWNED_BY_KERN(na)) {
2494 if (na->virt_hdr_len && !(req->nr_flags & NR_ACCEPT_VNET_HDR)) {
2495 nm_prerr("virt_hdr_len=%d, but application does "
2496 "not accept it", na->virt_hdr_len);
2501 error = netmap_do_regif(priv, na, req->nr_mode,
2502 req->nr_ringid, req->nr_flags);
2503 if (error) { /* reg. failed, release priv and ref */
2507 opt = nmreq_findoption((struct nmreq_option *)(uintptr_t)hdr->nr_options,
2508 NETMAP_REQ_OPT_CSB);
2510 struct nmreq_opt_csb *csbo =
2511 (struct nmreq_opt_csb *)opt;
2512 error = nmreq_checkduplicate(opt);
2514 error = netmap_csb_validate(priv, csbo);
2516 opt->nro_status = error;
2518 netmap_do_unregif(priv);
2523 nifp = priv->np_nifp;
2524 priv->np_td = td; /* for debugging purposes */
2526 /* return the offset of the netmap_if object */
2527 req->nr_rx_rings = na->num_rx_rings;
2528 req->nr_tx_rings = na->num_tx_rings;
2529 req->nr_rx_slots = na->num_rx_desc;
2530 req->nr_tx_slots = na->num_tx_desc;
2531 error = netmap_mem_get_info(na->nm_mem, &req->nr_memsize, &memflags,
2534 netmap_do_unregif(priv);
2537 if (memflags & NETMAP_MEM_PRIVATE) {
2538 *(uint32_t *)(uintptr_t)&nifp->ni_flags |= NI_PRIV_MEM;
2541 priv->np_si[t] = nm_si_user(priv, t) ?
2542 &na->si[t] : &NMR(na, t)[priv->np_qfirst[t]]->si;
2545 if (req->nr_extra_bufs) {
2547 nm_prinf("requested %d extra buffers",
2548 req->nr_extra_bufs);
2549 req->nr_extra_bufs = netmap_extra_alloc(na,
2550 &nifp->ni_bufs_head, req->nr_extra_bufs);
2552 nm_prinf("got %d extra buffers", req->nr_extra_bufs);
2554 req->nr_offset = netmap_mem_if_offset(na->nm_mem, nifp);
2556 error = nmreq_checkoptions(hdr);
2558 netmap_do_unregif(priv);
2562 /* store ifp reference so that priv destructor may release it */
2566 netmap_unget_na(na, ifp);
2568 /* release the reference from netmap_mem_find() or
2569 * netmap_mem_ext_create()
2572 netmap_mem_put(nmd);
2577 case NETMAP_REQ_PORT_INFO_GET: {
2578 struct nmreq_port_info_get *req =
2579 (struct nmreq_port_info_get *)(uintptr_t)hdr->nr_body;
2585 if (hdr->nr_name[0] != '\0') {
2586 /* Build a nmreq_register out of the nmreq_port_info_get,
2587 * so that we can call netmap_get_na(). */
2588 struct nmreq_register regreq;
2589 bzero(®req, sizeof(regreq));
2590 regreq.nr_mode = NR_REG_ALL_NIC;
2591 regreq.nr_tx_slots = req->nr_tx_slots;
2592 regreq.nr_rx_slots = req->nr_rx_slots;
2593 regreq.nr_tx_rings = req->nr_tx_rings;
2594 regreq.nr_rx_rings = req->nr_rx_rings;
2595 regreq.nr_mem_id = req->nr_mem_id;
2597 /* get a refcount */
2598 hdr->nr_reqtype = NETMAP_REQ_REGISTER;
2599 hdr->nr_body = (uintptr_t)®req;
2600 error = netmap_get_na(hdr, &na, &ifp, NULL, 1 /* create */);
2601 hdr->nr_reqtype = NETMAP_REQ_PORT_INFO_GET; /* reset type */
2602 hdr->nr_body = (uintptr_t)req; /* reset nr_body */
2608 nmd = na->nm_mem; /* get memory allocator */
2610 nmd = netmap_mem_find(req->nr_mem_id ? req->nr_mem_id : 1);
2613 nm_prerr("%s: failed to find mem_id %u",
2615 req->nr_mem_id ? req->nr_mem_id : 1);
2621 error = netmap_mem_get_info(nmd, &req->nr_memsize, &memflags,
2625 if (na == NULL) /* only memory info */
2627 netmap_update_config(na);
2628 req->nr_rx_rings = na->num_rx_rings;
2629 req->nr_tx_rings = na->num_tx_rings;
2630 req->nr_rx_slots = na->num_rx_desc;
2631 req->nr_tx_slots = na->num_tx_desc;
2633 netmap_unget_na(na, ifp);
2638 case NETMAP_REQ_VALE_ATTACH: {
2639 error = netmap_vale_attach(hdr, NULL /* userspace request */);
2643 case NETMAP_REQ_VALE_DETACH: {
2644 error = netmap_vale_detach(hdr, NULL /* userspace request */);
2648 case NETMAP_REQ_VALE_LIST: {
2649 error = netmap_vale_list(hdr);
2653 case NETMAP_REQ_PORT_HDR_SET: {
2654 struct nmreq_port_hdr *req =
2655 (struct nmreq_port_hdr *)(uintptr_t)hdr->nr_body;
2656 /* Build a nmreq_register out of the nmreq_port_hdr,
2657 * so that we can call netmap_get_bdg_na(). */
2658 struct nmreq_register regreq;
2659 bzero(®req, sizeof(regreq));
2660 regreq.nr_mode = NR_REG_ALL_NIC;
2662 /* For now we only support virtio-net headers, and only for
2663 * VALE ports, but this may change in future. Valid lengths
2664 * for the virtio-net header are 0 (no header), 10 and 12. */
2665 if (req->nr_hdr_len != 0 &&
2666 req->nr_hdr_len != sizeof(struct nm_vnet_hdr) &&
2667 req->nr_hdr_len != 12) {
2669 nm_prerr("invalid hdr_len %u", req->nr_hdr_len);
2674 hdr->nr_reqtype = NETMAP_REQ_REGISTER;
2675 hdr->nr_body = (uintptr_t)®req;
2676 error = netmap_get_vale_na(hdr, &na, NULL, 0);
2677 hdr->nr_reqtype = NETMAP_REQ_PORT_HDR_SET;
2678 hdr->nr_body = (uintptr_t)req;
2680 struct netmap_vp_adapter *vpna =
2681 (struct netmap_vp_adapter *)na;
2682 na->virt_hdr_len = req->nr_hdr_len;
2683 if (na->virt_hdr_len) {
2684 vpna->mfs = NETMAP_BUF_SIZE(na);
2687 nm_prinf("Using vnet_hdr_len %d for %p", na->virt_hdr_len, na);
2688 netmap_adapter_put(na);
2696 case NETMAP_REQ_PORT_HDR_GET: {
2697 /* Get vnet-header length for this netmap port */
2698 struct nmreq_port_hdr *req =
2699 (struct nmreq_port_hdr *)(uintptr_t)hdr->nr_body;
2700 /* Build a nmreq_register out of the nmreq_port_hdr,
2701 * so that we can call netmap_get_bdg_na(). */
2702 struct nmreq_register regreq;
2705 bzero(®req, sizeof(regreq));
2706 regreq.nr_mode = NR_REG_ALL_NIC;
2708 hdr->nr_reqtype = NETMAP_REQ_REGISTER;
2709 hdr->nr_body = (uintptr_t)®req;
2710 error = netmap_get_na(hdr, &na, &ifp, NULL, 0);
2711 hdr->nr_reqtype = NETMAP_REQ_PORT_HDR_GET;
2712 hdr->nr_body = (uintptr_t)req;
2714 req->nr_hdr_len = na->virt_hdr_len;
2716 netmap_unget_na(na, ifp);
2721 case NETMAP_REQ_VALE_NEWIF: {
2722 error = nm_vi_create(hdr);
2726 case NETMAP_REQ_VALE_DELIF: {
2727 error = nm_vi_destroy(hdr->nr_name);
2731 case NETMAP_REQ_VALE_POLLING_ENABLE:
2732 case NETMAP_REQ_VALE_POLLING_DISABLE: {
2733 error = nm_bdg_polling(hdr);
2736 #endif /* WITH_VALE */
2737 case NETMAP_REQ_POOLS_INFO_GET: {
2738 /* Get information from the memory allocator used for
2740 struct nmreq_pools_info *req =
2741 (struct nmreq_pools_info *)(uintptr_t)hdr->nr_body;
2744 /* Build a nmreq_register out of the nmreq_pools_info,
2745 * so that we can call netmap_get_na(). */
2746 struct nmreq_register regreq;
2747 bzero(®req, sizeof(regreq));
2748 regreq.nr_mem_id = req->nr_mem_id;
2749 regreq.nr_mode = NR_REG_ALL_NIC;
2751 hdr->nr_reqtype = NETMAP_REQ_REGISTER;
2752 hdr->nr_body = (uintptr_t)®req;
2753 error = netmap_get_na(hdr, &na, &ifp, NULL, 1 /* create */);
2754 hdr->nr_reqtype = NETMAP_REQ_POOLS_INFO_GET; /* reset type */
2755 hdr->nr_body = (uintptr_t)req; /* reset nr_body */
2761 nmd = na->nm_mem; /* grab the memory allocator */
2767 /* Finalize the memory allocator, get the pools
2768 * information and release the allocator. */
2769 error = netmap_mem_finalize(nmd, na);
2773 error = netmap_mem_pools_info_get(req, nmd);
2774 netmap_mem_drop(na);
2776 netmap_unget_na(na, ifp);
2781 case NETMAP_REQ_CSB_ENABLE: {
2782 struct nmreq_option *opt;
2784 opt = nmreq_findoption((struct nmreq_option *)(uintptr_t)hdr->nr_options,
2785 NETMAP_REQ_OPT_CSB);
2789 struct nmreq_opt_csb *csbo =
2790 (struct nmreq_opt_csb *)opt;
2791 error = nmreq_checkduplicate(opt);
2794 error = netmap_csb_validate(priv, csbo);
2797 opt->nro_status = error;
2802 case NETMAP_REQ_SYNC_KLOOP_START: {
2803 error = netmap_sync_kloop(priv, hdr);
2807 case NETMAP_REQ_SYNC_KLOOP_STOP: {
2808 error = netmap_sync_kloop_stop(priv);
2817 /* Write back request body to userspace and reset the
2818 * user-space pointer. */
2819 error = nmreq_copyout(hdr, error);
2825 if (unlikely(priv->np_nifp == NULL)) {
2829 mb(); /* make sure following reads are not from cache */
2831 if (unlikely(priv->np_csb_atok_base)) {
2832 nm_prerr("Invalid sync in CSB mode");
2837 na = priv->np_na; /* we have a reference */
2840 t = (cmd == NIOCTXSYNC ? NR_TX : NR_RX);
2841 krings = NMR(na, t);
2842 qfirst = priv->np_qfirst[t];
2843 qlast = priv->np_qlast[t];
2844 sync_flags = priv->np_sync_flags;
2846 for (i = qfirst; i < qlast; i++) {
2847 struct netmap_kring *kring = krings[i];
2848 struct netmap_ring *ring = kring->ring;
2850 if (unlikely(nm_kr_tryget(kring, 1, &error))) {
2851 error = (error ? EIO : 0);
2855 if (cmd == NIOCTXSYNC) {
2856 if (netmap_debug & NM_DEBUG_TXSYNC)
2857 nm_prinf("pre txsync ring %d cur %d hwcur %d",
2860 if (nm_txsync_prologue(kring, ring) >= kring->nkr_num_slots) {
2861 netmap_ring_reinit(kring);
2862 } else if (kring->nm_sync(kring, sync_flags | NAF_FORCE_RECLAIM) == 0) {
2863 nm_sync_finalize(kring);
2865 if (netmap_debug & NM_DEBUG_TXSYNC)
2866 nm_prinf("post txsync ring %d cur %d hwcur %d",
2870 if (nm_rxsync_prologue(kring, ring) >= kring->nkr_num_slots) {
2871 netmap_ring_reinit(kring);
2873 if (nm_may_forward_up(kring)) {
2874 /* transparent forwarding, see netmap_poll() */
2875 netmap_grab_packets(kring, &q, netmap_fwd);
2877 if (kring->nm_sync(kring, sync_flags | NAF_FORCE_READ) == 0) {
2878 nm_sync_finalize(kring);
2880 ring_timestamp_set(ring);
2886 netmap_send_up(na->ifp, &q);
2893 return netmap_ioctl_legacy(priv, cmd, data, td);
2902 nmreq_size_by_type(uint16_t nr_reqtype)
2904 switch (nr_reqtype) {
2905 case NETMAP_REQ_REGISTER:
2906 return sizeof(struct nmreq_register);
2907 case NETMAP_REQ_PORT_INFO_GET:
2908 return sizeof(struct nmreq_port_info_get);
2909 case NETMAP_REQ_VALE_ATTACH:
2910 return sizeof(struct nmreq_vale_attach);
2911 case NETMAP_REQ_VALE_DETACH:
2912 return sizeof(struct nmreq_vale_detach);
2913 case NETMAP_REQ_VALE_LIST:
2914 return sizeof(struct nmreq_vale_list);
2915 case NETMAP_REQ_PORT_HDR_SET:
2916 case NETMAP_REQ_PORT_HDR_GET:
2917 return sizeof(struct nmreq_port_hdr);
2918 case NETMAP_REQ_VALE_NEWIF:
2919 return sizeof(struct nmreq_vale_newif);
2920 case NETMAP_REQ_VALE_DELIF:
2921 case NETMAP_REQ_SYNC_KLOOP_STOP:
2922 case NETMAP_REQ_CSB_ENABLE:
2924 case NETMAP_REQ_VALE_POLLING_ENABLE:
2925 case NETMAP_REQ_VALE_POLLING_DISABLE:
2926 return sizeof(struct nmreq_vale_polling);
2927 case NETMAP_REQ_POOLS_INFO_GET:
2928 return sizeof(struct nmreq_pools_info);
2929 case NETMAP_REQ_SYNC_KLOOP_START:
2930 return sizeof(struct nmreq_sync_kloop_start);
2936 nmreq_opt_size_by_type(uint32_t nro_reqtype, uint64_t nro_size)
2938 size_t rv = sizeof(struct nmreq_option);
2939 #ifdef NETMAP_REQ_OPT_DEBUG
2940 if (nro_reqtype & NETMAP_REQ_OPT_DEBUG)
2941 return (nro_reqtype & ~NETMAP_REQ_OPT_DEBUG);
2942 #endif /* NETMAP_REQ_OPT_DEBUG */
2943 switch (nro_reqtype) {
2945 case NETMAP_REQ_OPT_EXTMEM:
2946 rv = sizeof(struct nmreq_opt_extmem);
2948 #endif /* WITH_EXTMEM */
2949 case NETMAP_REQ_OPT_SYNC_KLOOP_EVENTFDS:
2953 case NETMAP_REQ_OPT_CSB:
2954 rv = sizeof(struct nmreq_opt_csb);
2957 /* subtract the common header */
2958 return rv - sizeof(struct nmreq_option);
2962 nmreq_copyin(struct nmreq_header *hdr, int nr_body_is_user)
2964 size_t rqsz, optsz, bufsz;
2966 char *ker = NULL, *p;
2967 struct nmreq_option **next, *src;
2968 struct nmreq_option buf;
2971 if (hdr->nr_reserved) {
2973 nm_prerr("nr_reserved must be zero");
2977 if (!nr_body_is_user)
2980 hdr->nr_reserved = nr_body_is_user;
2982 /* compute the total size of the buffer */
2983 rqsz = nmreq_size_by_type(hdr->nr_reqtype);
2984 if (rqsz > NETMAP_REQ_MAXSIZE) {
2988 if ((rqsz && hdr->nr_body == (uintptr_t)NULL) ||
2989 (!rqsz && hdr->nr_body != (uintptr_t)NULL)) {
2990 /* Request body expected, but not found; or
2991 * request body found but unexpected. */
2993 nm_prerr("nr_body expected but not found, or vice versa");
2998 bufsz = 2 * sizeof(void *) + rqsz;
3000 for (src = (struct nmreq_option *)(uintptr_t)hdr->nr_options; src;
3001 src = (struct nmreq_option *)(uintptr_t)buf.nro_next)
3003 error = copyin(src, &buf, sizeof(*src));
3006 optsz += sizeof(*src);
3007 optsz += nmreq_opt_size_by_type(buf.nro_reqtype, buf.nro_size);
3008 if (rqsz + optsz > NETMAP_REQ_MAXSIZE) {
3012 bufsz += optsz + sizeof(void *);
3015 ker = nm_os_malloc(bufsz);
3022 /* make a copy of the user pointers */
3023 ptrs = (uint64_t*)p;
3024 *ptrs++ = hdr->nr_body;
3025 *ptrs++ = hdr->nr_options;
3029 error = copyin((void *)(uintptr_t)hdr->nr_body, p, rqsz);
3032 /* overwrite the user pointer with the in-kernel one */
3033 hdr->nr_body = (uintptr_t)p;
3036 /* copy the options */
3037 next = (struct nmreq_option **)&hdr->nr_options;
3040 struct nmreq_option *opt;
3042 /* copy the option header */
3043 ptrs = (uint64_t *)p;
3044 opt = (struct nmreq_option *)(ptrs + 1);
3045 error = copyin(src, opt, sizeof(*src));
3048 /* make a copy of the user next pointer */
3049 *ptrs = opt->nro_next;
3050 /* overwrite the user pointer with the in-kernel one */
3053 /* initialize the option as not supported.
3054 * Recognized options will update this field.
3056 opt->nro_status = EOPNOTSUPP;
3058 p = (char *)(opt + 1);
3060 /* copy the option body */
3061 optsz = nmreq_opt_size_by_type(opt->nro_reqtype,
3064 /* the option body follows the option header */
3065 error = copyin(src + 1, p, optsz);
3071 /* move to next option */
3072 next = (struct nmreq_option **)&opt->nro_next;
3078 ptrs = (uint64_t *)ker;
3079 hdr->nr_body = *ptrs++;
3080 hdr->nr_options = *ptrs++;
3081 hdr->nr_reserved = 0;
3088 nmreq_copyout(struct nmreq_header *hdr, int rerror)
3090 struct nmreq_option *src, *dst;
3091 void *ker = (void *)(uintptr_t)hdr->nr_body, *bufstart;
3096 if (!hdr->nr_reserved)
3099 /* restore the user pointers in the header */
3100 ptrs = (uint64_t *)ker - 2;
3102 hdr->nr_body = *ptrs++;
3103 src = (struct nmreq_option *)(uintptr_t)hdr->nr_options;
3104 hdr->nr_options = *ptrs;
3108 bodysz = nmreq_size_by_type(hdr->nr_reqtype);
3109 error = copyout(ker, (void *)(uintptr_t)hdr->nr_body, bodysz);
3116 /* copy the options */
3117 dst = (struct nmreq_option *)(uintptr_t)hdr->nr_options;
3122 /* restore the user pointer */
3123 next = src->nro_next;
3124 ptrs = (uint64_t *)src - 1;
3125 src->nro_next = *ptrs;
3127 /* always copy the option header */
3128 error = copyout(src, dst, sizeof(*src));
3134 /* copy the option body only if there was no error */
3135 if (!rerror && !src->nro_status) {
3136 optsz = nmreq_opt_size_by_type(src->nro_reqtype,
3139 error = copyout(src + 1, dst + 1, optsz);
3146 src = (struct nmreq_option *)(uintptr_t)next;
3147 dst = (struct nmreq_option *)(uintptr_t)*ptrs;
3152 hdr->nr_reserved = 0;
3153 nm_os_free(bufstart);
3157 struct nmreq_option *
3158 nmreq_findoption(struct nmreq_option *opt, uint16_t reqtype)
3160 for ( ; opt; opt = (struct nmreq_option *)(uintptr_t)opt->nro_next)
3161 if (opt->nro_reqtype == reqtype)
3167 nmreq_checkduplicate(struct nmreq_option *opt) {
3168 uint16_t type = opt->nro_reqtype;
3171 while ((opt = nmreq_findoption((struct nmreq_option *)(uintptr_t)opt->nro_next,
3174 opt->nro_status = EINVAL;
3176 return (dup ? EINVAL : 0);
3180 nmreq_checkoptions(struct nmreq_header *hdr)
3182 struct nmreq_option *opt;
3183 /* return error if there is still any option
3184 * marked as not supported
3187 for (opt = (struct nmreq_option *)(uintptr_t)hdr->nr_options; opt;
3188 opt = (struct nmreq_option *)(uintptr_t)opt->nro_next)
3189 if (opt->nro_status == EOPNOTSUPP)
3196 * select(2) and poll(2) handlers for the "netmap" device.
3198 * Can be called for one or more queues.
3199 * Return true the event mask corresponding to ready events.
3200 * If there are no ready events, do a selrecord on either individual
3201 * selinfo or on the global one.
3202 * Device-dependent parts (locking and sync of tx/rx rings)
3203 * are done through callbacks.
3205 * On linux, arguments are really pwait, the poll table, and 'td' is struct file *
3206 * The first one is remapped to pwait as selrecord() uses the name as an
3210 netmap_poll(struct netmap_priv_d *priv, int events, NM_SELRECORD_T *sr)
3212 struct netmap_adapter *na;
3213 struct netmap_kring *kring;
3214 struct netmap_ring *ring;
3215 u_int i, want[NR_TXRX], revents = 0;
3216 NM_SELINFO_T *si[NR_TXRX];
3217 #define want_tx want[NR_TX]
3218 #define want_rx want[NR_RX]
3219 struct mbq q; /* packets from RX hw queues to host stack */
3222 * In order to avoid nested locks, we need to "double check"
3223 * txsync and rxsync if we decide to do a selrecord().
3224 * retry_tx (and retry_rx, later) prevent looping forever.
3226 int retry_tx = 1, retry_rx = 1;
3228 /* Transparent mode: send_down is 1 if we have found some
3229 * packets to forward (host RX ring --> NIC) during the rx
3230 * scan and we have not sent them down to the NIC yet.
3231 * Transparent mode requires to bind all rings to a single
3235 int sync_flags = priv->np_sync_flags;
3239 if (unlikely(priv->np_nifp == NULL)) {
3242 mb(); /* make sure following reads are not from cache */
3246 if (unlikely(!nm_netmap_on(na)))
3249 if (unlikely(priv->np_csb_atok_base)) {
3250 nm_prerr("Invalid poll in CSB mode");
3254 if (netmap_debug & NM_DEBUG_ON)
3255 nm_prinf("device %s events 0x%x", na->name, events);
3256 want_tx = events & (POLLOUT | POLLWRNORM);
3257 want_rx = events & (POLLIN | POLLRDNORM);
3260 * If the card has more than one queue AND the file descriptor is
3261 * bound to all of them, we sleep on the "global" selinfo, otherwise
3262 * we sleep on individual selinfo (FreeBSD only allows two selinfo's
3263 * per file descriptor).
3264 * The interrupt routine in the driver wake one or the other
3265 * (or both) depending on which clients are active.
3267 * rxsync() is only called if we run out of buffers on a POLLIN.
3268 * txsync() is called if we run out of buffers on POLLOUT, or
3269 * there are pending packets to send. The latter can be disabled
3270 * passing NETMAP_NO_TX_POLL in the NIOCREG call.
3272 si[NR_RX] = nm_si_user(priv, NR_RX) ? &na->si[NR_RX] :
3273 &na->rx_rings[priv->np_qfirst[NR_RX]]->si;
3274 si[NR_TX] = nm_si_user(priv, NR_TX) ? &na->si[NR_TX] :
3275 &na->tx_rings[priv->np_qfirst[NR_TX]]->si;
3279 * We start with a lock free round which is cheap if we have
3280 * slots available. If this fails, then lock and call the sync
3281 * routines. We can't do this on Linux, as the contract says
3282 * that we must call nm_os_selrecord() unconditionally.
3285 enum txrx t = NR_TX;
3286 for (i = priv->np_qfirst[t]; want[t] && i < priv->np_qlast[t]; i++) {
3287 kring = NMR(na, t)[i];
3288 /* XXX compare ring->cur and kring->tail */
3289 if (!nm_ring_empty(kring->ring)) {
3291 want[t] = 0; /* also breaks the loop */
3296 enum txrx t = NR_RX;
3297 want_rx = 0; /* look for a reason to run the handlers */
3298 for (i = priv->np_qfirst[t]; i < priv->np_qlast[t]; i++) {
3299 kring = NMR(na, t)[i];
3300 if (kring->ring->cur == kring->ring->tail /* try fetch new buffers */
3301 || kring->rhead != kring->ring->head /* release buffers */) {
3306 revents |= events & (POLLIN | POLLRDNORM); /* we have data */
3311 /* The selrecord must be unconditional on linux. */
3312 nm_os_selrecord(sr, si[NR_RX]);
3313 nm_os_selrecord(sr, si[NR_TX]);
3317 * If we want to push packets out (priv->np_txpoll) or
3318 * want_tx is still set, we must issue txsync calls
3319 * (on all rings, to avoid that the tx rings stall).
3320 * Fortunately, normal tx mode has np_txpoll set.
3322 if (priv->np_txpoll || want_tx) {
3324 * The first round checks if anyone is ready, if not
3325 * do a selrecord and another round to handle races.
3326 * want_tx goes to 0 if any space is found, and is
3327 * used to skip rings with no pending transmissions.
3330 for (i = priv->np_qfirst[NR_TX]; i < priv->np_qlast[NR_TX]; i++) {
3333 kring = na->tx_rings[i];
3337 * Don't try to txsync this TX ring if we already found some
3338 * space in some of the TX rings (want_tx == 0) and there are no
3339 * TX slots in this ring that need to be flushed to the NIC
3342 if (!send_down && !want_tx && ring->head == kring->nr_hwcur)
3345 if (nm_kr_tryget(kring, 1, &revents))
3348 if (nm_txsync_prologue(kring, ring) >= kring->nkr_num_slots) {
3349 netmap_ring_reinit(kring);
3352 if (kring->nm_sync(kring, sync_flags))
3355 nm_sync_finalize(kring);
3359 * If we found new slots, notify potential
3360 * listeners on the same ring.
3361 * Since we just did a txsync, look at the copies
3362 * of cur,tail in the kring.
3364 found = kring->rcur != kring->rtail;
3366 if (found) { /* notify other listeners */
3370 kring->nm_notify(kring, 0);
3374 /* if there were any packet to forward we must have handled them by now */
3376 if (want_tx && retry_tx && sr) {
3378 nm_os_selrecord(sr, si[NR_TX]);
3386 * If want_rx is still set scan receive rings.
3387 * Do it on all rings because otherwise we starve.
3390 /* two rounds here for race avoidance */
3392 for (i = priv->np_qfirst[NR_RX]; i < priv->np_qlast[NR_RX]; i++) {
3395 kring = na->rx_rings[i];
3398 if (unlikely(nm_kr_tryget(kring, 1, &revents)))
3401 if (nm_rxsync_prologue(kring, ring) >= kring->nkr_num_slots) {
3402 netmap_ring_reinit(kring);
3405 /* now we can use kring->rcur, rtail */
3408 * transparent mode support: collect packets from
3409 * hw rxring(s) that have been released by the user
3411 if (nm_may_forward_up(kring)) {
3412 netmap_grab_packets(kring, &q, netmap_fwd);
3415 /* Clear the NR_FORWARD flag anyway, it may be set by
3416 * the nm_sync() below only on for the host RX ring (see
3417 * netmap_rxsync_from_host()). */
3418 kring->nr_kflags &= ~NR_FORWARD;
3419 if (kring->nm_sync(kring, sync_flags))
3422 nm_sync_finalize(kring);
3423 send_down |= (kring->nr_kflags & NR_FORWARD);
3424 ring_timestamp_set(ring);
3425 found = kring->rcur != kring->rtail;
3431 kring->nm_notify(kring, 0);
3437 if (retry_rx && sr) {
3438 nm_os_selrecord(sr, si[NR_RX]);
3441 if (send_down || retry_rx) {
3444 goto flush_tx; /* and retry_rx */
3451 * Transparent mode: released bufs (i.e. between kring->nr_hwcur and
3452 * ring->head) marked with NS_FORWARD on hw rx rings are passed up
3453 * to the host stack.
3457 netmap_send_up(na->ifp, &q);
3466 nma_intr_enable(struct netmap_adapter *na, int onoff)
3468 bool changed = false;
3473 for (i = 0; i < nma_get_nrings(na, t); i++) {
3474 struct netmap_kring *kring = NMR(na, t)[i];
3475 int on = !(kring->nr_kflags & NKR_NOINTR);
3477 if (!!onoff != !!on) {
3481 kring->nr_kflags &= ~NKR_NOINTR;
3483 kring->nr_kflags |= NKR_NOINTR;
3489 return 0; /* nothing to do */
3493 nm_prerr("Cannot %s interrupts for %s", onoff ? "enable" : "disable",
3498 na->nm_intr(na, onoff);
3504 /*-------------------- driver support routines -------------------*/
3506 /* default notify callback */
3508 netmap_notify(struct netmap_kring *kring, int flags)
3510 struct netmap_adapter *na = kring->notify_na;
3511 enum txrx t = kring->tx;
3513 nm_os_selwakeup(&kring->si);
3514 /* optimization: avoid a wake up on the global
3515 * queue if nobody has registered for more
3518 if (na->si_users[t] > 0)
3519 nm_os_selwakeup(&na->si[t]);
3521 return NM_IRQ_COMPLETED;
3524 /* called by all routines that create netmap_adapters.
3525 * provide some defaults and get a reference to the
3529 netmap_attach_common(struct netmap_adapter *na)
3531 if (!na->rx_buf_maxsize) {
3532 /* Set a conservative default (larger is safer). */
3533 na->rx_buf_maxsize = PAGE_SIZE;
3537 if (na->na_flags & NAF_HOST_RINGS && na->ifp) {
3538 na->if_input = na->ifp->if_input; /* for netmap_send_up */
3540 na->pdev = na; /* make sure netmap_mem_map() is called */
3541 #endif /* __FreeBSD__ */
3542 if (na->na_flags & NAF_HOST_RINGS) {
3543 if (na->num_host_rx_rings == 0)
3544 na->num_host_rx_rings = 1;
3545 if (na->num_host_tx_rings == 0)
3546 na->num_host_tx_rings = 1;
3548 if (na->nm_krings_create == NULL) {
3549 /* we assume that we have been called by a driver,
3550 * since other port types all provide their own
3553 na->nm_krings_create = netmap_hw_krings_create;
3554 na->nm_krings_delete = netmap_hw_krings_delete;
3556 if (na->nm_notify == NULL)
3557 na->nm_notify = netmap_notify;
3560 if (na->nm_mem == NULL) {
3561 /* use the global allocator */
3562 na->nm_mem = netmap_mem_get(&nm_mem);
3565 if (na->nm_bdg_attach == NULL)
3566 /* no special nm_bdg_attach callback. On VALE
3567 * attach, we need to interpose a bwrap
3569 na->nm_bdg_attach = netmap_default_bdg_attach;
3575 /* Wrapper for the register callback provided netmap-enabled
3577 * nm_iszombie(na) means that the driver module has been
3578 * unloaded, so we cannot call into it.
3579 * nm_os_ifnet_lock() must guarantee mutual exclusion with
3583 netmap_hw_reg(struct netmap_adapter *na, int onoff)
3585 struct netmap_hw_adapter *hwna =
3586 (struct netmap_hw_adapter*)na;
3591 if (nm_iszombie(na)) {
3594 } else if (na != NULL) {
3595 na->na_flags &= ~NAF_NETMAP_ON;
3600 error = hwna->nm_hw_register(na, onoff);
3603 nm_os_ifnet_unlock();
3609 netmap_hw_dtor(struct netmap_adapter *na)
3611 if (na->ifp == NULL)
3614 NM_DETACH_NA(na->ifp);
3619 * Allocate a netmap_adapter object, and initialize it from the
3620 * 'arg' passed by the driver on attach.
3621 * We allocate a block of memory of 'size' bytes, which has room
3622 * for struct netmap_adapter plus additional room private to
3624 * Return 0 on success, ENOMEM otherwise.
3627 netmap_attach_ext(struct netmap_adapter *arg, size_t size, int override_reg)
3629 struct netmap_hw_adapter *hwna = NULL;
3630 struct ifnet *ifp = NULL;
3632 if (size < sizeof(struct netmap_hw_adapter)) {
3633 if (netmap_debug & NM_DEBUG_ON)
3634 nm_prerr("Invalid netmap adapter size %d", (int)size);
3638 if (arg == NULL || arg->ifp == NULL) {
3639 if (netmap_debug & NM_DEBUG_ON)
3640 nm_prerr("either arg or arg->ifp is NULL");
3644 if (arg->num_tx_rings == 0 || arg->num_rx_rings == 0) {
3645 if (netmap_debug & NM_DEBUG_ON)
3646 nm_prerr("%s: invalid rings tx %d rx %d",
3647 arg->name, arg->num_tx_rings, arg->num_rx_rings);
3652 if (NM_NA_CLASH(ifp)) {
3653 /* If NA(ifp) is not null but there is no valid netmap
3654 * adapter it means that someone else is using the same
3655 * pointer (e.g. ax25_ptr on linux). This happens for
3656 * instance when also PF_RING is in use. */
3657 nm_prerr("Error: netmap adapter hook is busy");
3661 hwna = nm_os_malloc(size);
3665 hwna->up.na_flags |= NAF_HOST_RINGS | NAF_NATIVE;
3666 strlcpy(hwna->up.name, ifp->if_xname, sizeof(hwna->up.name));
3668 hwna->nm_hw_register = hwna->up.nm_register;
3669 hwna->up.nm_register = netmap_hw_reg;
3671 if (netmap_attach_common(&hwna->up)) {
3675 netmap_adapter_get(&hwna->up);
3677 NM_ATTACH_NA(ifp, &hwna->up);
3679 nm_os_onattach(ifp);
3681 if (arg->nm_dtor == NULL) {
3682 hwna->up.nm_dtor = netmap_hw_dtor;
3685 if_printf(ifp, "netmap queues/slots: TX %d/%d, RX %d/%d\n",
3686 hwna->up.num_tx_rings, hwna->up.num_tx_desc,
3687 hwna->up.num_rx_rings, hwna->up.num_rx_desc);
3691 nm_prerr("fail, arg %p ifp %p na %p", arg, ifp, hwna);
3692 return (hwna ? EINVAL : ENOMEM);
3697 netmap_attach(struct netmap_adapter *arg)
3699 return netmap_attach_ext(arg, sizeof(struct netmap_hw_adapter),
3700 1 /* override nm_reg */);
3705 NM_DBG(netmap_adapter_get)(struct netmap_adapter *na)
3711 refcount_acquire(&na->na_refcount);
3715 /* returns 1 iff the netmap_adapter is destroyed */
3717 NM_DBG(netmap_adapter_put)(struct netmap_adapter *na)
3722 if (!refcount_release(&na->na_refcount))
3728 if (na->tx_rings) { /* XXX should not happen */
3729 if (netmap_debug & NM_DEBUG_ON)
3730 nm_prerr("freeing leftover tx_rings");
3731 na->nm_krings_delete(na);
3733 netmap_pipe_dealloc(na);
3735 netmap_mem_put(na->nm_mem);
3736 bzero(na, sizeof(*na));
3742 /* nm_krings_create callback for all hardware native adapters */
3744 netmap_hw_krings_create(struct netmap_adapter *na)
3746 int ret = netmap_krings_create(na, 0);
3748 /* initialize the mbq for the sw rx ring */
3749 u_int lim = netmap_real_rings(na, NR_RX), i;
3750 for (i = na->num_rx_rings; i < lim; i++) {
3751 mbq_safe_init(&NMR(na, NR_RX)[i]->rx_queue);
3753 ND("initialized sw rx queue %d", na->num_rx_rings);
3761 * Called on module unload by the netmap-enabled drivers
3764 netmap_detach(struct ifnet *ifp)
3766 struct netmap_adapter *na = NA(ifp);
3772 netmap_set_all_rings(na, NM_KR_LOCKED);
3774 * if the netmap adapter is not native, somebody
3775 * changed it, so we can not release it here.
3776 * The NAF_ZOMBIE flag will notify the new owner that
3777 * the driver is gone.
3779 if (!(na->na_flags & NAF_NATIVE) || !netmap_adapter_put(na)) {
3780 na->na_flags |= NAF_ZOMBIE;
3782 /* give active users a chance to notice that NAF_ZOMBIE has been
3783 * turned on, so that they can stop and return an error to userspace.
3784 * Note that this becomes a NOP if there are no active users and,
3785 * therefore, the put() above has deleted the na, since now NA(ifp) is
3788 netmap_enable_all_rings(ifp);
3794 * Intercept packets from the network stack and pass them
3795 * to netmap as incoming packets on the 'software' ring.
3797 * We only store packets in a bounded mbq and then copy them
3798 * in the relevant rxsync routine.
3800 * We rely on the OS to make sure that the ifp and na do not go
3801 * away (typically the caller checks for IFF_DRV_RUNNING or the like).
3802 * In nm_register() or whenever there is a reinitialization,
3803 * we make sure to make the mode change visible here.
3806 netmap_transmit(struct ifnet *ifp, struct mbuf *m)
3808 struct netmap_adapter *na = NA(ifp);
3809 struct netmap_kring *kring, *tx_kring;
3810 u_int len = MBUF_LEN(m);
3811 u_int error = ENOBUFS;
3818 if (i >= na->num_host_rx_rings) {
3819 i = i % na->num_host_rx_rings;
3821 kring = NMR(na, NR_RX)[nma_get_nrings(na, NR_RX) + i];
3823 // XXX [Linux] we do not need this lock
3824 // if we follow the down/configure/up protocol -gl
3825 // mtx_lock(&na->core_lock);
3827 if (!nm_netmap_on(na)) {
3828 nm_prerr("%s not in netmap mode anymore", na->name);
3834 if (txr >= na->num_tx_rings) {
3835 txr %= na->num_tx_rings;
3837 tx_kring = NMR(na, NR_TX)[txr];
3839 if (tx_kring->nr_mode == NKR_NETMAP_OFF) {
3840 return MBUF_TRANSMIT(na, ifp, m);
3843 q = &kring->rx_queue;
3845 // XXX reconsider long packets if we handle fragments
3846 if (len > NETMAP_BUF_SIZE(na)) { /* too long for us */
3847 nm_prerr("%s from_host, drop packet size %d > %d", na->name,
3848 len, NETMAP_BUF_SIZE(na));
3852 if (!netmap_generic_hwcsum) {
3853 if (nm_os_mbuf_has_csum_offld(m)) {
3854 RD(1, "%s drop mbuf that needs checksum offload", na->name);
3859 if (nm_os_mbuf_has_seg_offld(m)) {
3860 RD(1, "%s drop mbuf that needs generic segmentation offload", na->name);
3865 ETHER_BPF_MTAP(ifp, m);
3866 #endif /* __FreeBSD__ */
3868 /* protect against netmap_rxsync_from_host(), netmap_sw_to_nic()
3869 * and maybe other instances of netmap_transmit (the latter
3870 * not possible on Linux).
3871 * We enqueue the mbuf only if we are sure there is going to be
3872 * enough room in the host RX ring, otherwise we drop it.
3876 busy = kring->nr_hwtail - kring->nr_hwcur;
3878 busy += kring->nkr_num_slots;
3879 if (busy + mbq_len(q) >= kring->nkr_num_slots - 1) {
3880 RD(2, "%s full hwcur %d hwtail %d qlen %d", na->name,
3881 kring->nr_hwcur, kring->nr_hwtail, mbq_len(q));
3884 ND(2, "%s %d bufs in queue", na->name, mbq_len(q));
3885 /* notify outside the lock */
3894 /* unconditionally wake up listeners */
3895 kring->nm_notify(kring, 0);
3896 /* this is normally netmap_notify(), but for nics
3897 * connected to a bridge it is netmap_bwrap_intr_notify(),
3898 * that possibly forwards the frames through the switch
3906 * netmap_reset() is called by the driver routines when reinitializing
3907 * a ring. The driver is in charge of locking to protect the kring.
3908 * If native netmap mode is not set just return NULL.
3909 * If native netmap mode is set, in particular, we have to set nr_mode to
3912 struct netmap_slot *
3913 netmap_reset(struct netmap_adapter *na, enum txrx tx, u_int n,
3916 struct netmap_kring *kring;
3919 if (!nm_native_on(na)) {
3920 ND("interface not in native netmap mode");
3921 return NULL; /* nothing to reinitialize */
3924 /* XXX note- in the new scheme, we are not guaranteed to be
3925 * under lock (e.g. when called on a device reset).
3926 * In this case, we should set a flag and do not trust too
3927 * much the values. In practice: TODO
3928 * - set a RESET flag somewhere in the kring
3929 * - do the processing in a conservative way
3930 * - let the *sync() fixup at the end.
3933 if (n >= na->num_tx_rings)
3936 kring = na->tx_rings[n];
3938 if (kring->nr_pending_mode == NKR_NETMAP_OFF) {
3939 kring->nr_mode = NKR_NETMAP_OFF;
3943 // XXX check whether we should use hwcur or rcur
3944 new_hwofs = kring->nr_hwcur - new_cur;
3946 if (n >= na->num_rx_rings)
3948 kring = na->rx_rings[n];
3950 if (kring->nr_pending_mode == NKR_NETMAP_OFF) {
3951 kring->nr_mode = NKR_NETMAP_OFF;
3955 new_hwofs = kring->nr_hwtail - new_cur;
3957 lim = kring->nkr_num_slots - 1;
3958 if (new_hwofs > lim)
3959 new_hwofs -= lim + 1;
3961 /* Always set the new offset value and realign the ring. */
3962 if (netmap_debug & NM_DEBUG_ON)
3963 nm_prinf("%s %s%d hwofs %d -> %d, hwtail %d -> %d",
3965 tx == NR_TX ? "TX" : "RX", n,
3966 kring->nkr_hwofs, new_hwofs,
3968 tx == NR_TX ? lim : kring->nr_hwtail);
3969 kring->nkr_hwofs = new_hwofs;
3971 kring->nr_hwtail = kring->nr_hwcur + lim;
3972 if (kring->nr_hwtail > lim)
3973 kring->nr_hwtail -= lim + 1;
3977 * Wakeup on the individual and global selwait
3978 * We do the wakeup here, but the ring is not yet reconfigured.
3979 * However, we are under lock so there are no races.
3981 kring->nr_mode = NKR_NETMAP_ON;
3982 kring->nm_notify(kring, 0);
3983 return kring->ring->slot;
3988 * Dispatch rx/tx interrupts to the netmap rings.
3990 * "work_done" is non-null on the RX path, NULL for the TX path.
3991 * We rely on the OS to make sure that there is only one active
3992 * instance per queue, and that there is appropriate locking.
3994 * The 'notify' routine depends on what the ring is attached to.
3995 * - for a netmap file descriptor, do a selwakeup on the individual
3996 * waitqueue, plus one on the global one if needed
3997 * (see netmap_notify)
3998 * - for a nic connected to a switch, call the proper forwarding routine
3999 * (see netmap_bwrap_intr_notify)
4002 netmap_common_irq(struct netmap_adapter *na, u_int q, u_int *work_done)
4004 struct netmap_kring *kring;
4005 enum txrx t = (work_done ? NR_RX : NR_TX);
4007 q &= NETMAP_RING_MASK;
4009 if (netmap_debug & (NM_DEBUG_RXINTR|NM_DEBUG_TXINTR)) {
4010 nm_prlim(5, "received %s queue %d", work_done ? "RX" : "TX" , q);
4013 if (q >= nma_get_nrings(na, t))
4014 return NM_IRQ_PASS; // not a physical queue
4016 kring = NMR(na, t)[q];
4018 if (kring->nr_mode == NKR_NETMAP_OFF) {
4023 kring->nr_kflags |= NKR_PENDINTR; // XXX atomic ?
4024 *work_done = 1; /* do not fire napi again */
4027 return kring->nm_notify(kring, 0);
4032 * Default functions to handle rx/tx interrupts from a physical device.
4033 * "work_done" is non-null on the RX path, NULL for the TX path.
4035 * If the card is not in netmap mode, simply return NM_IRQ_PASS,
4036 * so that the caller proceeds with regular processing.
4037 * Otherwise call netmap_common_irq().
4039 * If the card is connected to a netmap file descriptor,
4040 * do a selwakeup on the individual queue, plus one on the global one
4041 * if needed (multiqueue card _and_ there are multiqueue listeners),
4042 * and return NR_IRQ_COMPLETED.
4044 * Finally, if called on rx from an interface connected to a switch,
4045 * calls the proper forwarding routine.
4048 netmap_rx_irq(struct ifnet *ifp, u_int q, u_int *work_done)
4050 struct netmap_adapter *na = NA(ifp);
4053 * XXX emulated netmap mode sets NAF_SKIP_INTR so
4054 * we still use the regular driver even though the previous
4055 * check fails. It is unclear whether we should use
4056 * nm_native_on() here.
4058 if (!nm_netmap_on(na))
4061 if (na->na_flags & NAF_SKIP_INTR) {
4062 ND("use regular interrupt");
4066 return netmap_common_irq(na, q, work_done);
4069 /* set/clear native flags and if_transmit/netdev_ops */
4071 nm_set_native_flags(struct netmap_adapter *na)
4073 struct ifnet *ifp = na->ifp;
4075 /* We do the setup for intercepting packets only if we are the
4076 * first user of this adapapter. */
4077 if (na->active_fds > 0) {
4081 na->na_flags |= NAF_NETMAP_ON;
4083 nm_update_hostrings_mode(na);
4087 nm_clear_native_flags(struct netmap_adapter *na)
4089 struct ifnet *ifp = na->ifp;
4091 /* We undo the setup for intercepting packets only if we are the
4092 * last user of this adapter. */
4093 if (na->active_fds > 0) {
4097 nm_update_hostrings_mode(na);
4100 na->na_flags &= ~NAF_NETMAP_ON;
4104 * Module loader and unloader
4106 * netmap_init() creates the /dev/netmap device and initializes
4107 * all global variables. Returns 0 on success, errno on failure
4108 * (but there is no chance)
4110 * netmap_fini() destroys everything.
4113 static struct cdev *netmap_dev; /* /dev/netmap character device. */
4114 extern struct cdevsw netmap_cdevsw;
4121 destroy_dev(netmap_dev);
4122 /* we assume that there are no longer netmap users */
4124 netmap_uninit_bridges();
4127 nm_prinf("netmap: unloaded module.");
4138 error = netmap_mem_init();
4142 * MAKEDEV_ETERNAL_KLD avoids an expensive check on syscalls
4143 * when the module is compiled in.
4144 * XXX could use make_dev_credv() to get error number
4146 netmap_dev = make_dev_credf(MAKEDEV_ETERNAL_KLD,
4147 &netmap_cdevsw, 0, NULL, UID_ROOT, GID_WHEEL, 0600,
4152 error = netmap_init_bridges();
4157 nm_os_vi_init_index();
4160 error = nm_os_ifnet_init();
4164 nm_prinf("netmap: loaded module");
4168 return (EINVAL); /* may be incorrect */