$OpenBSD: altq.9,v 1.4 2001/07/12 12:41:42 itojun Exp $
Copyright (C) 2004 Max Laier. All rights reserved.
Copyright (C) 2001
Sony Computer Science Laboratories Inc. All rights reserved.
Redistribution and use in source and binary forms, with or without
modification, are permitted provided that the following conditions
are met:
1. Redistributions of source code must retain the above copyright
notice, this list of conditions and the following disclaimer.
2. Redistributions in binary form must reproduce the above copyright
notice, this list of conditions and the following disclaimer in the
documentation and/or other materials provided with the distribution.
THIS SOFTWARE IS PROVIDED BY SONY CSL AND CONTRIBUTORS ``AS IS'' AND
ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
ARE DISCLAIMED. IN NO EVENT SHALL SONY CSL OR CONTRIBUTORS BE LIABLE
FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
SUCH DAMAGE.
$FreeBSD: releng/11.0/share/man/man9/altq.9 286663 2015-08-12 11:56:19Z brueffer $
.Dd August 25, 2004 .Dt ALTQ 9 .Os
.Sh NAME .Nm ALTQ .Nd kernel interfaces for manipulating output queues on network interfaces .Sh SYNOPSIS n sys/types.h n sys/socket.h n net/if.h n net/if_var.h
.Ss Enqueue macros .Fn IFQ_ENQUEUE "struct ifaltq *ifq" "struct mbuf *m" "int error" .Fn IFQ_HANDOFF "struct ifnet *ifp" "struct mbuf *m" "int error" .Fo IFQ_HANDOFF_ADJ .Fa "struct ifnet *ifp" "struct mbuf *m" "int adjust" "int error" .Fc
.Ss Dequeue macros .Fn IFQ_DEQUEUE "struct ifaltq *ifq" "struct mbuf *m" .Fn IFQ_POLL_NOLOCK "struct ifaltq *ifq" "struct mbuf *m" .Fn IFQ_PURGE "struct ifaltq *ifq" .Fn IFQ_IS_EMPTY "struct ifaltq *ifq"
.Ss Driver managed dequeue macros .Fn IFQ_DRV_DEQUEUE "struct ifaltq *ifq" "struct mbuf *m" .Fn IFQ_DRV_PREPEND "struct ifaltq *ifq" "struct mbuf *m" .Fn IFQ_DRV_PURGE "struct ifaltq *ifq" .Fn IFQ_DRV_IS_EMPTY "struct ifaltq *ifq"
.Ss General setup macros .Fn IFQ_SET_MAXLEN "struct ifaltq *ifq" "int len" .Fn IFQ_INC_LEN "struct ifaltq *ifq" .Fn IFQ_DEC_LEN "struct ifaltq *ifq" .Fn IFQ_INC_DROPS "struct ifaltq *ifq" .Fn IFQ_SET_READY "struct ifaltq *ifq" .Sh DESCRIPTION The .Nm system is a framework to manage queuing disciplines on network interfaces. .Nm introduces new macros to manipulate output queues. The output queue macros are used to abstract queue operations and not to touch the internal fields of the output queue structure. The macros are independent from the .Nm implementation, and compatible with the traditional .Vt ifqueue macros for ease of transition.
p .Fn IFQ_ENQUEUE , .Fn IFQ_HANDOFF and .Fn IFQ_HANDOFF_ADJ enqueue a packet .Fa m to the queue .Fa ifq . The underlying queuing discipline may discard the packet. The .Fa error argument is set to 0 on success, or .Er ENOBUFS if the packet is discarded. The packet pointed to by .Fa m will be freed by the device driver on success, or by the queuing discipline on failure, so the caller should not touch .Fa m after enqueuing. .Fn IFQ_HANDOFF and .Fn IFQ_HANDOFF_ADJ combine the enqueue operation with statistic generation and call .Fn if_start upon successful enqueue to initiate the actual send.
p .Fn IFQ_DEQUEUE dequeues a packet from the queue. The dequeued packet is returned in .Fa m , or .Fa m is set to .Dv NULL if no packet is dequeued. The caller must always check .Fa m since a non-empty queue could return .Dv NULL under rate-limiting.
p .Fn IFQ_POLL_NOLOCK returns the next packet without removing it from the queue. The caller must hold the queue mutex when calling .Fn IFQ_POLL_NOLOCK in order to guarantee that a subsequent call to .Fn IFQ_DEQUEUE_NOLOCK dequeues the same packet.
p .Fn IFQ_*_NOLOCK variants (if available) always assume that the caller holds the queue mutex. They can be grabbed with .Fn IFQ_LOCK and released with .Fn IFQ_UNLOCK .
p .Fn IFQ_PURGE discards all the packets in the queue. The purge operation is needed since a non-work conserving queue cannot be emptied by a dequeue loop.
p .Fn IFQ_IS_EMPTY can be used to check if the queue is empty. Note that .Fn IFQ_DEQUEUE could still return .Dv NULL if the queuing discipline is non-work conserving.
p .Fn IFQ_DRV_DEQUEUE moves up to .Fa ifq->ifq_drv_maxlen packets from the queue to the .Dq "driver managed" queue and returns the first one via .Fa m . As for .Fn IFQ_DEQUEUE , .Fa m can be .Dv NULL even for a non-empty queue. Subsequent calls to .Fn IFQ_DRV_DEQUEUE pass the packets from the .Dq "driver managed" queue without obtaining the queue mutex. It is the responsibility of the caller to protect against concurrent access. Enabling .Nm for a given queue sets .Va ifq_drv_maxlen to 0 as the .Dq "bulk dequeue" performed by .Fn IFQ_DRV_DEQUEUE for higher values of .Va ifq_drv_maxlen is adverse to .Nm ALTQ Ns 's internal timing. Note that a driver must not mix .Fn IFQ_DRV_* macros with the default dequeue macros as the default macros do not look at the .Dq "driver managed" queue which might lead to an mbuf leak.
p .Fn IFQ_DRV_PREPEND prepends .Fa m to the .Dq "driver managed" queue from where it will be obtained with the next call to .Fn IFQ_DRV_DEQUEUE .
p .Fn IFQ_DRV_PURGE flushes all packets in the .Dq "driver managed" queue and calls to .Fn IFQ_PURGE afterwards.
p .Fn IFQ_DRV_IS_EMPTY checks for packets in the .Dq "driver managed" part of the queue. If it is empty, it forwards to .Fn IFQ_IS_EMPTY .
p .Fn IFQ_SET_MAXLEN sets the queue length limit to the default FIFO queue. The .Va ifq_drv_maxlen member of the .Vt ifaltq structure controls the length limit of the .Dq "driver managed" queue.
p .Fn IFQ_INC_LEN and .Fn IFQ_DEC_LEN increment or decrement the current queue length in packets. This is mostly for internal purposes.
p .Fn IFQ_INC_DROPS increments the drop counter and is identical to .Fn IF_DROP . It is defined for naming consistency only.
p .Fn IFQ_SET_READY sets a flag to indicate that a driver was converted to use the new macros. .Nm can be enabled only on interfaces with this flag. .Sh COMPATIBILITY .Ss Vt ifaltq structure In order to keep compatibility with the existing code, the new output queue structure .Vt ifaltq has the same fields. The traditional .Fn IF_* macros and the code directly referencing the fields within .Va if_snd still work with .Vt ifaltq . d -literal ##old-style## ##new-style## | struct ifqueue { | struct ifaltq { struct mbuf *ifq_head; | struct mbuf *ifq_head; struct mbuf *ifq_tail; | struct mbuf *ifq_tail; int ifq_len; | int ifq_len; int ifq_maxlen; | int ifq_maxlen; }; | /* driver queue fields */ | ...... | /* altq related fields */ | ...... | }; | .Ed The new structure replaces .Vt "struct ifqueue" in .Vt "struct ifnet" . d -literal ##old-style## ##new-style## | struct ifnet { | struct ifnet { .... | .... | struct ifqueue if_snd; | struct ifaltq if_snd; | .... | .... }; | }; | .Ed The (simplified) new .Fn IFQ_* macros look like: d -literal #define IFQ_DEQUEUE(ifq, m) \e if (ALTQ_IS_ENABLED((ifq)) \e ALTQ_DEQUEUE((ifq), (m)); \e else \e IF_DEQUEUE((ifq), (m)); .Ed .Ss Enqueue operation The semantics of the enqueue operation is changed. In the new style, enqueue and packet drop are combined since they cannot be easily separated in many queuing disciplines. The new enqueue operation corresponds to the following macro that is written with the old macros. d -literal #define IFQ_ENQUEUE(ifq, m, error) \e do { \e if (IF_QFULL((ifq))) { \e m_freem((m)); \e (error) = ENOBUFS; \e IF_DROP(ifq); \e } else { \e IF_ENQUEUE((ifq), (m)); \e (error) = 0; \e } \e } while (0) .Ed
p .Fn IFQ_ENQUEUE does the following:
p l -hyphen -compact t queue a packet, t drop (and free) a packet if the enqueue operation fails. .El
p If the enqueue operation fails, .Fa error is set to .Er ENOBUFS . The .Fa m mbuf is freed by the queuing discipline. The caller should not touch mbuf after calling .Fn IFQ_ENQUEUE so that the caller may need to copy .Va m_pkthdr.len or .Va m_flags field beforehand for statistics. .Fn IFQ_HANDOFF and .Fn IFQ_HANDOFF_ADJ can be used if only default interface statistics and an immediate call to .Fn if_start are desired. The caller should not use .Fn senderr since mbuf was already freed.
p The new style .Fn if_output looks as follows: d -literal ##old-style## ##new-style## | int | int ether_output(ifp, m0, dst, rt0) | ether_output(ifp, m0, dst, rt0) { | { ...... | ...... | | mflags = m->m_flags; | len = m->m_pkthdr.len; s = splimp(); | s = splimp(); if (IF_QFULL(&ifp->if_snd)) { | IFQ_ENQUEUE(&ifp->if_snd, m, | error); IF_DROP(&ifp->if_snd); | if (error != 0) { splx(s); | splx(s); senderr(ENOBUFS); | return (error); } | } IF_ENQUEUE(&ifp->if_snd, m); | ifp->if_obytes += | ifp->if_obytes += len; m->m_pkthdr.len; | if (m->m_flags & M_MCAST) | if (mflags & M_MCAST) ifp->if_omcasts++; | ifp->if_omcasts++; | if ((ifp->if_flags & IFF_OACTIVE) | if ((ifp->if_flags & IFF_OACTIVE) == 0) | == 0) (*ifp->if_start)(ifp); | (*ifp->if_start)(ifp); splx(s); | splx(s); return (error); | return (error); | bad: | bad: if (m) | if (m) m_freem(m); | m_freem(m); return (error); | return (error); } | } | .Ed .Sh HOW TO CONVERT THE EXISTING DRIVERS First, make sure the corresponding .Fn if_output is already converted to the new style.
p Look for .Va if_snd in the driver. Probably, you need to make changes to the lines that include .Va if_snd . .Ss Empty check operation If the code checks .Va ifq_head to see whether the queue is empty or not, use .Fn IFQ_IS_EMPTY . d -literal ##old-style## ##new-style## | if (ifp->if_snd.ifq_head != NULL) | if (!IFQ_IS_EMPTY(&ifp->if_snd)) | .Ed .Fn IFQ_IS_EMPTY only checks if there is any packet stored in the queue. Note that even when .Fn IFQ_IS_EMPTY is .Dv FALSE , .Fn IFQ_DEQUEUE could still return .Dv NULL if the queue is under rate-limiting. .Ss Dequeue operation Replace .Fn IF_DEQUEUE by .Fn IFQ_DEQUEUE . Always check whether the dequeued mbuf is .Dv NULL or not. Note that even when .Fn IFQ_IS_EMPTY is .Dv FALSE , .Fn IFQ_DEQUEUE could return .Dv NULL due to rate-limiting. d -literal ##old-style## ##new-style## | IF_DEQUEUE(&ifp->if_snd, m); | IFQ_DEQUEUE(&ifp->if_snd, m); | if (m == NULL) | return; | .Ed A driver is supposed to call .Fn if_start from transmission complete interrupts in order to trigger the next dequeue. .Ss Poll-and-dequeue operation If the code polls the packet at the head of the queue and actually uses the packet before dequeuing it, use .Fn IFQ_POLL_NOLOCK and .Fn IFQ_DEQUEUE_NOLOCK . d -literal ##old-style## ##new-style## | | IFQ_LOCK(&ifp->if_snd); m = ifp->if_snd.ifq_head; | IFQ_POLL_NOLOCK(&ifp->if_snd, m); if (m != NULL) { | if (m != NULL) { | /* use m to get resources */ | /* use m to get resources */ if (something goes wrong) | if (something goes wrong) | IFQ_UNLOCK(&ifp->if_snd); return; | return; | IF_DEQUEUE(&ifp->if_snd, m); | IFQ_DEQUEUE_NOLOCK(&ifp->if_snd, m); | IFQ_UNLOCK(&ifp->if_snd); | /* kick the hardware */ | /* kick the hardware */ } | } | .Ed It is guaranteed that .Fn IFQ_DEQUEUE_NOLOCK under the same lock as a previous .Fn IFQ_POLL_NOLOCK returns the same packet. Note that they need to be guarded by .Fn IFQ_LOCK . .Ss Eliminating Fn IF_PREPEND If the code uses .Fn IF_PREPEND , you have to eliminate it unless you can use a .Dq "driver managed" queue which allows the use of .Fn IFQ_DRV_PREPEND as a substitute. A common usage of .Fn IF_PREPEND is to cancel the previous dequeue operation. You have to convert the logic into poll-and-dequeue. d -literal ##old-style## ##new-style## | | IFQ_LOCK(&ifp->if_snd); IF_DEQUEUE(&ifp->if_snd, m); | IFQ_POLL_NOLOCK(&ifp->if_snd, m); if (m != NULL) { | if (m != NULL) { | if (something_goes_wrong) { | if (something_goes_wrong) { IF_PREPEND(&ifp->if_snd, m); | IFQ_UNLOCK(&ifp->if_snd); return; | return; } | } | | /* at this point, the driver | * is committed to send this | * packet. | */ | IFQ_DEQUEUE_NOLOCK(&ifp->if_snd, m); | IFQ_UNLOCK(&ifp->if_snd); | /* kick the hardware */ | /* kick the hardware */ } | } | .Ed .Ss Purge operation Use .Fn IFQ_PURGE to empty the queue. Note that a non-work conserving queue cannot be emptied by a dequeue loop. d -literal ##old-style## ##new-style## | while (ifp->if_snd.ifq_head != NULL) {| IFQ_PURGE(&ifp->if_snd); IF_DEQUEUE(&ifp->if_snd, m); | m_freem(m); | } | | .Ed .Ss Conversion using a driver managed queue Convert .Fn IF_* macros to their equivalent .Fn IFQ_DRV_* and employ .Fn IFQ_DRV_IS_EMPTY where appropriate. d -literal ##old-style## ##new-style## | if (ifp->if_snd.ifq_head != NULL) | if (!IFQ_DRV_IS_EMPTY(&ifp->if_snd)) | .Ed Make sure that calls to .Fn IFQ_DRV_DEQUEUE , .Fn IFQ_DRV_PREPEND and .Fn IFQ_DRV_PURGE are protected with a mutex of some kind. .Ss Attach routine Use .Fn IFQ_SET_MAXLEN to set .Va ifq_maxlen to .Fa len . Initialize .Va ifq_drv_maxlen with a sensible value if you plan to use the .Fn IFQ_DRV_* macros. Add .Fn IFQ_SET_READY to show this driver is converted to the new style. (This is used to distinguish new-style drivers.) d -literal ##old-style## ##new-style## | ifp->if_snd.ifq_maxlen = qsize; | IFQ_SET_MAXLEN(&ifp->if_snd, qsize); | ifp->if_snd.ifq_drv_maxlen = qsize; | IFQ_SET_READY(&ifp->if_snd); if_attach(ifp); | if_attach(ifp); | .Ed .Ss Other issues The new macros for statistics: d -literal ##old-style## ##new-style## | IF_DROP(&ifp->if_snd); | IFQ_INC_DROPS(&ifp->if_snd); | ifp->if_snd.ifq_len++; | IFQ_INC_LEN(&ifp->if_snd); | ifp->if_snd.ifq_len--; | IFQ_DEC_LEN(&ifp->if_snd); | .Ed .Sh QUEUING DISCIPLINES Queuing disciplines need to maintain .Fa ifq_len (used by .Fn IFQ_IS_EMPTY ) . Queuing disciplines also need to guarantee that the same mbuf is returned if .Fn IFQ_DEQUEUE is called immediately after .Fn IFQ_POLL . .Sh SEE ALSO .Xr pf 4 , .Xr pf.conf 5 , .Xr pfctl 8 .Sh HISTORY The .Nm system first appeared in March 1997.