mirror/freebsd
1
0
Fork 0
mirror of https://git.FreeBSD.org/src.git synced 2024-12-16 10:20:30 +00:00
Code Issues Releases Activity

Restored Herb's current copyright. The one in the file appeared to have

Browse Source 
been run through indent and additionally doesn't contain clauses 3 and 4
that were apparantly added sometime after the driver was put into FreeBSD.
This commit is contained in:
David Greenman 1995-01-08 10:47:23 +00:00
parent ddfdef1374
commit f7c9e2441f
Notes: svn2git 2020-12-20 02:59:44 +00:00 
svn path=/head/; revision=5436
2 changed files with 56 additions and 488 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

272
sys/dev/ep/if_ep.c
View File

@ -1,249 +1,33 @@
/*
* Copyright (c) 1993 Herb Peyerl <hpeyerl@novatel.ca> All rights reserved.
*
* Copyright (c) 1994 Herb Peyerl <hpeyerl@novatel.ca>
* 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. The name
* of the author may not be used to endorse or promote products derived from
* this software without specific prior written permission
*
* THIS SOFTWARE IS PROVIDED BY THE AUTHOR ``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 THE AUTHOR 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.
*
* From: if_ep.c,v 1.9 1994/01/25 10:46:29 deraadt Exp $ $Id: if_ep.c,v 1.9
* 1994/05/02 22:27:33 ats Exp $
*
* October 26, 1994
*
* Modified by: Andres Vega Garcia
* INRIA - Sophia Antipolis, France
* e-mail: avega@sophia.inria.fr
* finger: avega@pax.inria.fr
*
*
* What is new:
*
* 1) We can recognize more than 1 board.
*
* 2) The problem which used to happen with high trafic is corrected,
* (No more need to 'down' and 'up' the interface).
*
* 3) In the transmission, we use the TX start threshold in a more dynamic
* fashion (IMO the throughput is higher this way).
*
* 4) In the reception, we use the RX early threshold, that parameter is
* adapted as the packets arrive (IMO the throughput is higher this way).
*
* 5) Supports EISA cards.
*
* NB 0: The 32 bits acces is allowed for the EISA configured cards, thoung I
* wasn't able to test the code added.
*
* NB 1: I added the option EP_LOCAL_STATS, it can be temporary as IMO is just
* used while working on this driver and the program that displays this
* information (epstat).
*
*
* Some driver statistics can be viewed with the epstat utility. In order to
* use this, you have to compile if_ep.c with
*
* -DEP_LOCAL_STATS
*
* which can be included in your machine config file (e.g. GENERICAH_EP)
* as an option (option EP_LOCAL_STATS).
*
*
* Modifications since FreeBSD 1.1.5.1 Release:
*
* This explanation concerns the epstart(), epread() and epintr() functions.
*
* =========================================================================
* epstart()
* =========================================================================
*
*
* Let's see what the idea is:
*
*
* Packet |------------------ LEN ---------------------|
*
* A
* CPU |---------------|----------------------------|
*
*
* Card |----------------------------|
*
*
*
* We suppose the Card is able to *write* bytes (send them to the media)
* at the speed S_CARD (bytes/s), and that is faster than the speed of the CPU,
* S_CPU, to write bytes to the TX FIFO, then, we have to write A bytes to the
* FIFO before enableing the transmision in the card. This way both, the card
* and the CPU must finish their writing at the same time.
*
*
* Let TX_RATE = S_CPU / S_CARD, where TX_RATE <= 1
*
* We can find that:
*
* (1) A = LEN * (1 - TX_RATE)
*
*
* Let TX_RATE_R be the *very real* value.
*
* If TX_RATE > TX_RATE_R
* We are supposing the CPU is faster than it really is and
* certainly the card will *finish* before the CPU, having a
* TX Underrun Error, then, in such a case, we have to do:
*
* TX_RATE -= STEP, where STEP is the step at which we
* move TX_RATE
*
* If TX_RATE < TX_RATE_R
* We won't have the TX Underrun Error but it is possible that
* we don't use eficiently the TX START THRESH. feature.
* We prevent this by doing:
*
* TX_RATE += STEP every time we have sent succesfuly
* (without Underrun) a certain number
* of packets.
*
* Now, to avoid dealing with reals I used a FACTOR, then (1) will be
* transformed:
*
* Let tx_rate = FACTOR * TX_RATE, tx_rate is the parameter
* really used.
*
* A = (LEN * (FACTOR - tx_rate)) / FACTOR
*
* Actually FACTOR = 64
*
* (2) A = (LEN * (64 - tx_rate)) >> 6
*
* As I want to have some margin, and
* as I have to write a number multiple of 4:
*
* A = tx_start_threshold = (((LEN * (64 - tx_rate)) >> 6) & ~3) + 16
*
*
* =========================================================================
* epread()
* =========================================================================
*
* I mantain an estimation of the RX packet's average length, and an
* estimation of the RX latency.
*
* Every time I receive a complete packet I compute the average packet's
* length, rx_avg_pkt:
*
* DELTA = LEN - rx_avg_pkt, where LEN is this packet's length
*
* if DELTA > 0
* rx_avg_pkt += AVG_UP * DELTA
* else
* rx_avg_pkt += AVG_DOWN *DELTA
*
*
* AVG_UP < AVG_DOWN
*
* In the first case, I'm interested in being conservative about the
* average packet'length, because if I let it go up *too fast*, a shorter packet
* than expected will probably cause an RX Overrun Error. But if I consider
* the next packet will be smaller than it will, I just will have to wait
* for the packet to complete reception.
*
* In the other case, I'm interested in leting the rx_avg_pkt follow
* the real packet's lenght closer, as it is important not to think the average
* packet is bigger than it realy is. If I don't do that, and the rx_avg_pkt
* goes down *too slow*, I'll find myself thinking the packets are big when they
* are really small and I'll have probably Rx Overrun Errors.
*
* Actualy:
*
* AVG_UP = 1/32
* AVG_DOWN = 1/8
*
*
* Every time I receive an incomplete packet I recompute the RX latency
* (rx_latency).
*
* I know that if rx_latency = 0, when I go read the bytes from the RX
* FIFO, I'll find as many bytes as I programmed in the RX Early Threshold, but
* if rx_latency > 0, I'll find more bytes.
*
* Let CUR_LAT be the RX latency seen by this packet.
*
* CUR_LAT = LEN - rx_early_threshold, where LEN is the number of
* bytes I have just received.
*
* DELTA = CUR_LAT - rx_latency
*
*
* if DELTA >= 0
* rx_latency += LAT_UP * DELTA
* else
* rx_latency += LAT_DOWN * DELTA
*
*
* LAT_UP > LAT_DOWN
*
* In a similar way as for rx average packet's length, I try to be more
* conservative in the more critical case.
*
* In the first case, I have to follow closer the incremets of the RX
* latency, because if I don't, I can find myself thinking that we (CPU) are
* *enough fast* and wait up to the last minute to go read data to find we have
* had RX Overrun Error.
*
* In the other case I must be more conservative to avoid falling in
* the situation I have just described, because if I go down *to fast* I'll
* think we are enough fast and we'll wake up later than due.
*
* Actually:
*
* LAT_UP = 1/4
* LAT_DOWN = 1/32
*
* Finally, I compute the rx_early_threshold for the next packet as:
*
* rx_early_threshold = rx_avg_pkt - rx_latency
*
* But, as I want to have a margin and
* as I have to write a value multiple of 4.
*
*
* rx_early_threshold = (rx_avg_pkt - rx_latency - 16) & ~3
*
*
* But if I have to wait for the rest of an incomplete packet
* from which I have already received CUR_LEN bytes:
*
*
* rx_early_threshold = (rx_avg_pkt-CUR_LEN - rx_latency - 16) & ~3
*
*
* =========================================================================
* epintr()
* =========================================================================
*
* For this function I just tryed to do what is stated in the
* Etherlink III Technical Reference.
*
* It was here where I really solved the problem that used to happen with
* high traffic.
*
*
* Andres
* avega@pax.inria.fr
* 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.
* 3. All advertising materials mentioning features or use of this software
* must display the following acknowledgement:
* This product includes software developed by Herb Peyerl.
* 4. The name of Herb Peyerl may not be used to endorse or promote products
* derived from this software without specific prior written permission.
*
* THIS SOFTWARE IS PROVIDED BY THE AUTHOR ``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 THE AUTHOR 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.
*
* $Id$
*/
#include "ep.h"

272
sys/i386/isa/if_ep.c
View File

@ -1,249 +1,33 @@
/*
* Copyright (c) 1993 Herb Peyerl <hpeyerl@novatel.ca> All rights reserved.
*
* Copyright (c) 1994 Herb Peyerl <hpeyerl@novatel.ca>
* 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. The name
* of the author may not be used to endorse or promote products derived from
* this software without specific prior written permission
*
* THIS SOFTWARE IS PROVIDED BY THE AUTHOR ``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 THE AUTHOR 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.
*
* From: if_ep.c,v 1.9 1994/01/25 10:46:29 deraadt Exp $ $Id: if_ep.c,v 1.9
* 1994/05/02 22:27:33 ats Exp $
*
* October 26, 1994
*
* Modified by: Andres Vega Garcia
* INRIA - Sophia Antipolis, France
* e-mail: avega@sophia.inria.fr
* finger: avega@pax.inria.fr
*
*
* What is new:
*
* 1) We can recognize more than 1 board.
*
* 2) The problem which used to happen with high trafic is corrected,
* (No more need to 'down' and 'up' the interface).
*
* 3) In the transmission, we use the TX start threshold in a more dynamic
* fashion (IMO the throughput is higher this way).
*
* 4) In the reception, we use the RX early threshold, that parameter is
* adapted as the packets arrive (IMO the throughput is higher this way).
*
* 5) Supports EISA cards.
*
* NB 0: The 32 bits acces is allowed for the EISA configured cards, thoung I
* wasn't able to test the code added.
*
* NB 1: I added the option EP_LOCAL_STATS, it can be temporary as IMO is just
* used while working on this driver and the program that displays this
* information (epstat).
*
*
* Some driver statistics can be viewed with the epstat utility. In order to
* use this, you have to compile if_ep.c with
*
* -DEP_LOCAL_STATS
*
* which can be included in your machine config file (e.g. GENERICAH_EP)
* as an option (option EP_LOCAL_STATS).
*
*
* Modifications since FreeBSD 1.1.5.1 Release:
*
* This explanation concerns the epstart(), epread() and epintr() functions.
*
* =========================================================================
* epstart()
* =========================================================================
*
*
* Let's see what the idea is:
*
*
* Packet |------------------ LEN ---------------------|
*
* A
* CPU |---------------|----------------------------|
*
*
* Card |----------------------------|
*
*
*
* We suppose the Card is able to *write* bytes (send them to the media)
* at the speed S_CARD (bytes/s), and that is faster than the speed of the CPU,
* S_CPU, to write bytes to the TX FIFO, then, we have to write A bytes to the
* FIFO before enableing the transmision in the card. This way both, the card
* and the CPU must finish their writing at the same time.
*
*
* Let TX_RATE = S_CPU / S_CARD, where TX_RATE <= 1
*
* We can find that:
*
* (1) A = LEN * (1 - TX_RATE)
*
*
* Let TX_RATE_R be the *very real* value.
*
* If TX_RATE > TX_RATE_R
* We are supposing the CPU is faster than it really is and
* certainly the card will *finish* before the CPU, having a
* TX Underrun Error, then, in such a case, we have to do:
*
* TX_RATE -= STEP, where STEP is the step at which we
* move TX_RATE
*
* If TX_RATE < TX_RATE_R
* We won't have the TX Underrun Error but it is possible that
* we don't use eficiently the TX START THRESH. feature.
* We prevent this by doing:
*
* TX_RATE += STEP every time we have sent succesfuly
* (without Underrun) a certain number
* of packets.
*
* Now, to avoid dealing with reals I used a FACTOR, then (1) will be
* transformed:
*
* Let tx_rate = FACTOR * TX_RATE, tx_rate is the parameter
* really used.
*
* A = (LEN * (FACTOR - tx_rate)) / FACTOR
*
* Actually FACTOR = 64
*
* (2) A = (LEN * (64 - tx_rate)) >> 6
*
* As I want to have some margin, and
* as I have to write a number multiple of 4:
*
* A = tx_start_threshold = (((LEN * (64 - tx_rate)) >> 6) & ~3) + 16
*
*
* =========================================================================
* epread()
* =========================================================================
*
* I mantain an estimation of the RX packet's average length, and an
* estimation of the RX latency.
*
* Every time I receive a complete packet I compute the average packet's
* length, rx_avg_pkt:
*
* DELTA = LEN - rx_avg_pkt, where LEN is this packet's length
*
* if DELTA > 0
* rx_avg_pkt += AVG_UP * DELTA
* else
* rx_avg_pkt += AVG_DOWN *DELTA
*
*
* AVG_UP < AVG_DOWN
*
* In the first case, I'm interested in being conservative about the
* average packet'length, because if I let it go up *too fast*, a shorter packet
* than expected will probably cause an RX Overrun Error. But if I consider
* the next packet will be smaller than it will, I just will have to wait
* for the packet to complete reception.
*
* In the other case, I'm interested in leting the rx_avg_pkt follow
* the real packet's lenght closer, as it is important not to think the average
* packet is bigger than it realy is. If I don't do that, and the rx_avg_pkt
* goes down *too slow*, I'll find myself thinking the packets are big when they
* are really small and I'll have probably Rx Overrun Errors.
*
* Actualy:
*
* AVG_UP = 1/32
* AVG_DOWN = 1/8
*
*
* Every time I receive an incomplete packet I recompute the RX latency
* (rx_latency).
*
* I know that if rx_latency = 0, when I go read the bytes from the RX
* FIFO, I'll find as many bytes as I programmed in the RX Early Threshold, but
* if rx_latency > 0, I'll find more bytes.
*
* Let CUR_LAT be the RX latency seen by this packet.
*
* CUR_LAT = LEN - rx_early_threshold, where LEN is the number of
* bytes I have just received.
*
* DELTA = CUR_LAT - rx_latency
*
*
* if DELTA >= 0
* rx_latency += LAT_UP * DELTA
* else
* rx_latency += LAT_DOWN * DELTA
*
*
* LAT_UP > LAT_DOWN
*
* In a similar way as for rx average packet's length, I try to be more
* conservative in the more critical case.
*
* In the first case, I have to follow closer the incremets of the RX
* latency, because if I don't, I can find myself thinking that we (CPU) are
* *enough fast* and wait up to the last minute to go read data to find we have
* had RX Overrun Error.
*
* In the other case I must be more conservative to avoid falling in
* the situation I have just described, because if I go down *to fast* I'll
* think we are enough fast and we'll wake up later than due.
*
* Actually:
*
* LAT_UP = 1/4
* LAT_DOWN = 1/32
*
* Finally, I compute the rx_early_threshold for the next packet as:
*
* rx_early_threshold = rx_avg_pkt - rx_latency
*
* But, as I want to have a margin and
* as I have to write a value multiple of 4.
*
*
* rx_early_threshold = (rx_avg_pkt - rx_latency - 16) & ~3
*
*
* But if I have to wait for the rest of an incomplete packet
* from which I have already received CUR_LEN bytes:
*
*
* rx_early_threshold = (rx_avg_pkt-CUR_LEN - rx_latency - 16) & ~3
*
*
* =========================================================================
* epintr()
* =========================================================================
*
* For this function I just tryed to do what is stated in the
* Etherlink III Technical Reference.
*
* It was here where I really solved the problem that used to happen with
* high traffic.
*
*
* Andres
* avega@pax.inria.fr
* 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.
* 3. All advertising materials mentioning features or use of this software
* must display the following acknowledgement:
* This product includes software developed by Herb Peyerl.
* 4. The name of Herb Peyerl may not be used to endorse or promote products
* derived from this software without specific prior written permission.
*
* THIS SOFTWARE IS PROVIDED BY THE AUTHOR ``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 THE AUTHOR 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.
*
* $Id$
*/
#include "ep.h"