d/freebsd-skq
1
0
Fork 0
Code Issues Pull Requests Packages Projects Releases Wiki Activity
21c8beb9fd
freebsd-skq/sys/dev/nfe/if_nfe.c

3379 lines
87 KiB
C
Raw Normal View History

Add a pure open source nForce Ethernet driver, under BSDL. This driver was ported from OpenBSD by Shigeaki Tagashira <shigeaki@se.hiroshima-u.ac.jp> and posted at http://www.se.hiroshima-u.ac.jp/~shigeaki/software/freebsd-nfe.html It was additionally cleaned up by me. It is still a work-in-progress and thus is purposefully not in GENERIC. And it conflicts with nve(4), so only one should be loaded.
2006-06-26 23:41:07 +00:00
/* $OpenBSD: if_nfe.c,v 1.54 2006/04/07 12:38:12 jsg Exp $ */
Import the nfe (nForce Ethernet) driver from OpenBSD. This is a pure open source NIC driver with a BSD license. These bits were taken from the OpenBSD public CVS repository on 1-May-06.
2006-06-26 17:25:22 +00:00
/*-
Add a pure open source nForce Ethernet driver, under BSDL. This driver was ported from OpenBSD by Shigeaki Tagashira <shigeaki@se.hiroshima-u.ac.jp> and posted at http://www.se.hiroshima-u.ac.jp/~shigeaki/software/freebsd-nfe.html It was additionally cleaned up by me. It is still a work-in-progress and thus is purposefully not in GENERIC. And it conflicts with nve(4), so only one should be loaded.
2006-06-26 23:41:07 +00:00
* Copyright (c) 2006 Shigeaki Tagashira <shigeaki@se.hiroshima-u.ac.jp>
Import the nfe (nForce Ethernet) driver from OpenBSD. This is a pure open source NIC driver with a BSD license. These bits were taken from the OpenBSD public CVS repository on 1-May-06.
2006-06-26 17:25:22 +00:00
* Copyright (c) 2006 Damien Bergamini <damien.bergamini@free.fr>
* Copyright (c) 2005, 2006 Jonathan Gray <jsg@openbsd.org>
*
* Permission to use, copy, modify, and distribute this software for any
* purpose with or without fee is hereby granted, provided that the above
* copyright notice and this permission notice appear in all copies.
*
* THE SOFTWARE IS PROVIDED "AS IS" AND THE AUTHOR DISCLAIMS ALL WARRANTIES
* WITH REGARD TO THIS SOFTWARE INCLUDING ALL IMPLIED WARRANTIES OF
* MERCHANTABILITY AND FITNESS. IN NO EVENT SHALL THE AUTHOR BE LIABLE FOR
* ANY SPECIAL, DIRECT, INDIRECT, OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES
* WHATSOEVER RESULTING FROM LOSS OF USE, DATA OR PROFITS, WHETHER IN AN
* ACTION OF CONTRACT, NEGLIGENCE OR OTHER TORTIOUS ACTION, ARISING OUT OF
* OR IN CONNECTION WITH THE USE OR PERFORMANCE OF THIS SOFTWARE.
*/
/* Driver for NVIDIA nForce MCP Fast Ethernet and Gigabit Ethernet */
Add a pure open source nForce Ethernet driver, under BSDL. This driver was ported from OpenBSD by Shigeaki Tagashira <shigeaki@se.hiroshima-u.ac.jp> and posted at http://www.se.hiroshima-u.ac.jp/~shigeaki/software/freebsd-nfe.html It was additionally cleaned up by me. It is still a work-in-progress and thus is purposefully not in GENERIC. And it conflicts with nve(4), so only one should be loaded.
2006-06-26 23:41:07 +00:00
#include <sys/cdefs.h>
__FBSDID("$FreeBSD$");
#ifdef HAVE_KERNEL_OPTION_HEADERS
#include "opt_device_polling.h"
#endif
Import the nfe (nForce Ethernet) driver from OpenBSD. This is a pure open source NIC driver with a BSD license. These bits were taken from the OpenBSD public CVS repository on 1-May-06.
2006-06-26 17:25:22 +00:00
#include <sys/param.h>
#include <sys/endian.h>
#include <sys/systm.h>
#include <sys/sockio.h>
#include <sys/mbuf.h>
#include <sys/malloc.h>
Add a pure open source nForce Ethernet driver, under BSDL. This driver was ported from OpenBSD by Shigeaki Tagashira <shigeaki@se.hiroshima-u.ac.jp> and posted at http://www.se.hiroshima-u.ac.jp/~shigeaki/software/freebsd-nfe.html It was additionally cleaned up by me. It is still a work-in-progress and thus is purposefully not in GENERIC. And it conflicts with nve(4), so only one should be loaded.
2006-06-26 23:41:07 +00:00
#include <sys/module.h>
Import the nfe (nForce Ethernet) driver from OpenBSD. This is a pure open source NIC driver with a BSD license. These bits were taken from the OpenBSD public CVS repository on 1-May-06.
2006-06-26 17:25:22 +00:00
#include <sys/kernel.h>
Bring overhauled nfe(4) into tree. o s/printf/device_printf/g o Nuke OpenBSDism. o Nuke NetBSD/OpenBSD specific DMA sync operations.(we don't have a way to sync a single descriptor within a DMA map.) o Remove recursive mutex. o bus_dma(9) clean up. o 40bit DMA address support. o Add protection for Rx map load failure. o Fix a long standing bug for watchdog timeout. [1] o Add additional protections, missing Tx completion interrupt, losing start Tx command, for watchdog timeout. o Switch to taskqueue(9) API to handle interrupts. o Use our own timer for watchdog instead of if_watchdog/if_timer interface. o Advertise VLAN header length/capability correctly to upper layer. o Remove excessive kernel stack consumption in nfe_encap(). o Handle highly fragmented mbuf chains correctly. o Enable etherenet address reprogramming with ifconfig(8). o Add ALTQ/TSO, MSI/MSIX support. o Increased Rx ring to 256 descriptors from 128. o Align Tx/Rx descriptor ring on sizeof(struct nfe_desc64) boundary. o Remove alignment restrictions on Tx/Rx buffers. o Rewritten jumbo frame support code. o Add support for hardware assistend VLAN tag insertion/stripping. o Add support for Tx/Rx flow control based on patches from Peer Chen. [2] o Add a routine that detects whether ethernet address swap routines is required. [3] o Add a workaround that take MAC/PHY out of power down mode. o Add suspend/resume support. o style(9) and code clean up. Special thanks to Shigeaki Tagashira, the original porter of nfe(4), who submitted lots of patches, performed uncountable number of regression tests and maintained nfe(4) for a long time. Without his enthusiastic help and support I could never have completed this overhauling task. The only weak point of nfe(4) compared to nve(4) is instability of manual half-duplex media selection on certain hardwares(auto sensing media type should work for all cases, though). This was a long standing bug of nfe(4) and I still have no idea why it doesn't work on some hardwares. Obtained from: OpenBSD [1] Submitted by: Peer Chen < pchen at nvidia dot com > [2], [3] Reviewed by: Shigeaki Tagashira < shigeaki AT se DOT hiroshima-u DOT ac DOT jp > Tested by: Shigeaki Tagashira, current Discussed with: current Silence from: obrien
2007-06-12 02:16:02 +00:00
#include <sys/queue.h>
Import the nfe (nForce Ethernet) driver from OpenBSD. This is a pure open source NIC driver with a BSD license. These bits were taken from the OpenBSD public CVS repository on 1-May-06.
2006-06-26 17:25:22 +00:00
#include <sys/socket.h>
Bring overhauled nfe(4) into tree. o s/printf/device_printf/g o Nuke OpenBSDism. o Nuke NetBSD/OpenBSD specific DMA sync operations.(we don't have a way to sync a single descriptor within a DMA map.) o Remove recursive mutex. o bus_dma(9) clean up. o 40bit DMA address support. o Add protection for Rx map load failure. o Fix a long standing bug for watchdog timeout. [1] o Add additional protections, missing Tx completion interrupt, losing start Tx command, for watchdog timeout. o Switch to taskqueue(9) API to handle interrupts. o Use our own timer for watchdog instead of if_watchdog/if_timer interface. o Advertise VLAN header length/capability correctly to upper layer. o Remove excessive kernel stack consumption in nfe_encap(). o Handle highly fragmented mbuf chains correctly. o Enable etherenet address reprogramming with ifconfig(8). o Add ALTQ/TSO, MSI/MSIX support. o Increased Rx ring to 256 descriptors from 128. o Align Tx/Rx descriptor ring on sizeof(struct nfe_desc64) boundary. o Remove alignment restrictions on Tx/Rx buffers. o Rewritten jumbo frame support code. o Add support for hardware assistend VLAN tag insertion/stripping. o Add support for Tx/Rx flow control based on patches from Peer Chen. [2] o Add a routine that detects whether ethernet address swap routines is required. [3] o Add a workaround that take MAC/PHY out of power down mode. o Add suspend/resume support. o style(9) and code clean up. Special thanks to Shigeaki Tagashira, the original porter of nfe(4), who submitted lots of patches, performed uncountable number of regression tests and maintained nfe(4) for a long time. Without his enthusiastic help and support I could never have completed this overhauling task. The only weak point of nfe(4) compared to nve(4) is instability of manual half-duplex media selection on certain hardwares(auto sensing media type should work for all cases, though). This was a long standing bug of nfe(4) and I still have no idea why it doesn't work on some hardwares. Obtained from: OpenBSD [1] Submitted by: Peer Chen < pchen at nvidia dot com > [2], [3] Reviewed by: Shigeaki Tagashira < shigeaki AT se DOT hiroshima-u DOT ac DOT jp > Tested by: Shigeaki Tagashira, current Discussed with: current Silence from: obrien
2007-06-12 02:16:02 +00:00
#include <sys/sysctl.h>
Add a pure open source nForce Ethernet driver, under BSDL. This driver was ported from OpenBSD by Shigeaki Tagashira <shigeaki@se.hiroshima-u.ac.jp> and posted at http://www.se.hiroshima-u.ac.jp/~shigeaki/software/freebsd-nfe.html It was additionally cleaned up by me. It is still a work-in-progress and thus is purposefully not in GENERIC. And it conflicts with nve(4), so only one should be loaded.
2006-06-26 23:41:07 +00:00
#include <sys/taskqueue.h>
Import the nfe (nForce Ethernet) driver from OpenBSD. This is a pure open source NIC driver with a BSD license. These bits were taken from the OpenBSD public CVS repository on 1-May-06.
2006-06-26 17:25:22 +00:00
#include <net/if.h>
Add a pure open source nForce Ethernet driver, under BSDL. This driver was ported from OpenBSD by Shigeaki Tagashira <shigeaki@se.hiroshima-u.ac.jp> and posted at http://www.se.hiroshima-u.ac.jp/~shigeaki/software/freebsd-nfe.html It was additionally cleaned up by me. It is still a work-in-progress and thus is purposefully not in GENERIC. And it conflicts with nve(4), so only one should be loaded.
2006-06-26 23:41:07 +00:00
#include <net/if_arp.h>
#include <net/ethernet.h>
Import the nfe (nForce Ethernet) driver from OpenBSD. This is a pure open source NIC driver with a BSD license. These bits were taken from the OpenBSD public CVS repository on 1-May-06.
2006-06-26 17:25:22 +00:00
#include <net/if_dl.h>
#include <net/if_media.h>
#include <net/if_types.h>
#include <net/if_vlan_var.h>
#include <net/bpf.h>
Add a pure open source nForce Ethernet driver, under BSDL. This driver was ported from OpenBSD by Shigeaki Tagashira <shigeaki@se.hiroshima-u.ac.jp> and posted at http://www.se.hiroshima-u.ac.jp/~shigeaki/software/freebsd-nfe.html It was additionally cleaned up by me. It is still a work-in-progress and thus is purposefully not in GENERIC. And it conflicts with nve(4), so only one should be loaded.
2006-06-26 23:41:07 +00:00
#include <machine/bus.h>
#include <machine/resource.h>
#include <sys/bus.h>
#include <sys/rman.h>
Import the nfe (nForce Ethernet) driver from OpenBSD. This is a pure open source NIC driver with a BSD license. These bits were taken from the OpenBSD public CVS repository on 1-May-06.
2006-06-26 17:25:22 +00:00
#include <dev/mii/mii.h>
#include <dev/mii/miivar.h>
#include <dev/pci/pcireg.h>
#include <dev/pci/pcivar.h>
Add a pure open source nForce Ethernet driver, under BSDL. This driver was ported from OpenBSD by Shigeaki Tagashira <shigeaki@se.hiroshima-u.ac.jp> and posted at http://www.se.hiroshima-u.ac.jp/~shigeaki/software/freebsd-nfe.html It was additionally cleaned up by me. It is still a work-in-progress and thus is purposefully not in GENERIC. And it conflicts with nve(4), so only one should be loaded.
2006-06-26 23:41:07 +00:00
#include <dev/nfe/if_nfereg.h>
#include <dev/nfe/if_nfevar.h>
MODULE_DEPEND(nfe, pci, 1, 1, 1);
MODULE_DEPEND(nfe, ether, 1, 1, 1);
MODULE_DEPEND(nfe, miibus, 1, 1, 1);
Bring overhauled nfe(4) into tree. o s/printf/device_printf/g o Nuke OpenBSDism. o Nuke NetBSD/OpenBSD specific DMA sync operations.(we don't have a way to sync a single descriptor within a DMA map.) o Remove recursive mutex. o bus_dma(9) clean up. o 40bit DMA address support. o Add protection for Rx map load failure. o Fix a long standing bug for watchdog timeout. [1] o Add additional protections, missing Tx completion interrupt, losing start Tx command, for watchdog timeout. o Switch to taskqueue(9) API to handle interrupts. o Use our own timer for watchdog instead of if_watchdog/if_timer interface. o Advertise VLAN header length/capability correctly to upper layer. o Remove excessive kernel stack consumption in nfe_encap(). o Handle highly fragmented mbuf chains correctly. o Enable etherenet address reprogramming with ifconfig(8). o Add ALTQ/TSO, MSI/MSIX support. o Increased Rx ring to 256 descriptors from 128. o Align Tx/Rx descriptor ring on sizeof(struct nfe_desc64) boundary. o Remove alignment restrictions on Tx/Rx buffers. o Rewritten jumbo frame support code. o Add support for hardware assistend VLAN tag insertion/stripping. o Add support for Tx/Rx flow control based on patches from Peer Chen. [2] o Add a routine that detects whether ethernet address swap routines is required. [3] o Add a workaround that take MAC/PHY out of power down mode. o Add suspend/resume support. o style(9) and code clean up. Special thanks to Shigeaki Tagashira, the original porter of nfe(4), who submitted lots of patches, performed uncountable number of regression tests and maintained nfe(4) for a long time. Without his enthusiastic help and support I could never have completed this overhauling task. The only weak point of nfe(4) compared to nve(4) is instability of manual half-duplex media selection on certain hardwares(auto sensing media type should work for all cases, though). This was a long standing bug of nfe(4) and I still have no idea why it doesn't work on some hardwares. Obtained from: OpenBSD [1] Submitted by: Peer Chen < pchen at nvidia dot com > [2], [3] Reviewed by: Shigeaki Tagashira < shigeaki AT se DOT hiroshima-u DOT ac DOT jp > Tested by: Shigeaki Tagashira, current Discussed with: current Silence from: obrien
2007-06-12 02:16:02 +00:00
/* "device miibus" required. See GENERIC if you get errors here. */
Add a pure open source nForce Ethernet driver, under BSDL. This driver was ported from OpenBSD by Shigeaki Tagashira <shigeaki@se.hiroshima-u.ac.jp> and posted at http://www.se.hiroshima-u.ac.jp/~shigeaki/software/freebsd-nfe.html It was additionally cleaned up by me. It is still a work-in-progress and thus is purposefully not in GENERIC. And it conflicts with nve(4), so only one should be loaded.
2006-06-26 23:41:07 +00:00
#include "miibus_if.h"
style(9) cleanup.
2006-10-19 10:01:26 +00:00
static int nfe_probe(device_t);
static int nfe_attach(device_t);
static int nfe_detach(device_t);
Bring overhauled nfe(4) into tree. o s/printf/device_printf/g o Nuke OpenBSDism. o Nuke NetBSD/OpenBSD specific DMA sync operations.(we don't have a way to sync a single descriptor within a DMA map.) o Remove recursive mutex. o bus_dma(9) clean up. o 40bit DMA address support. o Add protection for Rx map load failure. o Fix a long standing bug for watchdog timeout. [1] o Add additional protections, missing Tx completion interrupt, losing start Tx command, for watchdog timeout. o Switch to taskqueue(9) API to handle interrupts. o Use our own timer for watchdog instead of if_watchdog/if_timer interface. o Advertise VLAN header length/capability correctly to upper layer. o Remove excessive kernel stack consumption in nfe_encap(). o Handle highly fragmented mbuf chains correctly. o Enable etherenet address reprogramming with ifconfig(8). o Add ALTQ/TSO, MSI/MSIX support. o Increased Rx ring to 256 descriptors from 128. o Align Tx/Rx descriptor ring on sizeof(struct nfe_desc64) boundary. o Remove alignment restrictions on Tx/Rx buffers. o Rewritten jumbo frame support code. o Add support for hardware assistend VLAN tag insertion/stripping. o Add support for Tx/Rx flow control based on patches from Peer Chen. [2] o Add a routine that detects whether ethernet address swap routines is required. [3] o Add a workaround that take MAC/PHY out of power down mode. o Add suspend/resume support. o style(9) and code clean up. Special thanks to Shigeaki Tagashira, the original porter of nfe(4), who submitted lots of patches, performed uncountable number of regression tests and maintained nfe(4) for a long time. Without his enthusiastic help and support I could never have completed this overhauling task. The only weak point of nfe(4) compared to nve(4) is instability of manual half-duplex media selection on certain hardwares(auto sensing media type should work for all cases, though). This was a long standing bug of nfe(4) and I still have no idea why it doesn't work on some hardwares. Obtained from: OpenBSD [1] Submitted by: Peer Chen < pchen at nvidia dot com > [2], [3] Reviewed by: Shigeaki Tagashira < shigeaki AT se DOT hiroshima-u DOT ac DOT jp > Tested by: Shigeaki Tagashira, current Discussed with: current Silence from: obrien
2007-06-12 02:16:02 +00:00
static int nfe_suspend(device_t);
static int nfe_resume(device_t);
Fix function prototype for device_shutdown method.
2007-11-22 02:45:00 +00:00
static int nfe_shutdown(device_t);
P5N32-SLI PREMIUM from ASUSTeK is known to have MSI/MSI-X issue such that nfe(4) does not work with MSI-X. When MSI-X support was introduced, I remember MCP55 controller worked without problems so the issue could be either PCI bridge or BIOS issue. But I also noticed snd_hda(4) disabled MSI on all MCP55 chipset so I'm still not sure this is generic issue of MCP55 chipset. If this was PCI bridge issue we would have added it to a system wide black-list table but it's not clear to me at this moment whether it was caused by either broken BIOS or silicon bug of MCP55 chipset. To workaround the issue, maintain a MSI/MSI-X black-list table in driver and lookup base board manufacturer and product name from the table before attempting to use MSI-X. If driver find an matching entry, nfe(4) will not use MSI/MSI-X and fall back on traditional INTx mode. This approach should be the last resort since it relies on smbios and if another instance of MSI/MSI-X breakage is reported with different maker/product, we may have to get the PCI bridge black-listed instead of adding an new entry. PR: kern/152150
2010-11-14 23:37:43 +00:00
static int nfe_can_use_msix(struct nfe_softc *);
Bring overhauled nfe(4) into tree. o s/printf/device_printf/g o Nuke OpenBSDism. o Nuke NetBSD/OpenBSD specific DMA sync operations.(we don't have a way to sync a single descriptor within a DMA map.) o Remove recursive mutex. o bus_dma(9) clean up. o 40bit DMA address support. o Add protection for Rx map load failure. o Fix a long standing bug for watchdog timeout. [1] o Add additional protections, missing Tx completion interrupt, losing start Tx command, for watchdog timeout. o Switch to taskqueue(9) API to handle interrupts. o Use our own timer for watchdog instead of if_watchdog/if_timer interface. o Advertise VLAN header length/capability correctly to upper layer. o Remove excessive kernel stack consumption in nfe_encap(). o Handle highly fragmented mbuf chains correctly. o Enable etherenet address reprogramming with ifconfig(8). o Add ALTQ/TSO, MSI/MSIX support. o Increased Rx ring to 256 descriptors from 128. o Align Tx/Rx descriptor ring on sizeof(struct nfe_desc64) boundary. o Remove alignment restrictions on Tx/Rx buffers. o Rewritten jumbo frame support code. o Add support for hardware assistend VLAN tag insertion/stripping. o Add support for Tx/Rx flow control based on patches from Peer Chen. [2] o Add a routine that detects whether ethernet address swap routines is required. [3] o Add a workaround that take MAC/PHY out of power down mode. o Add suspend/resume support. o style(9) and code clean up. Special thanks to Shigeaki Tagashira, the original porter of nfe(4), who submitted lots of patches, performed uncountable number of regression tests and maintained nfe(4) for a long time. Without his enthusiastic help and support I could never have completed this overhauling task. The only weak point of nfe(4) compared to nve(4) is instability of manual half-duplex media selection on certain hardwares(auto sensing media type should work for all cases, though). This was a long standing bug of nfe(4) and I still have no idea why it doesn't work on some hardwares. Obtained from: OpenBSD [1] Submitted by: Peer Chen < pchen at nvidia dot com > [2], [3] Reviewed by: Shigeaki Tagashira < shigeaki AT se DOT hiroshima-u DOT ac DOT jp > Tested by: Shigeaki Tagashira, current Discussed with: current Silence from: obrien
2007-06-12 02:16:02 +00:00
static void nfe_power(struct nfe_softc *);
style(9) cleanup.
2006-10-19 10:01:26 +00:00
static int nfe_miibus_readreg(device_t, int, int);
static int nfe_miibus_writereg(device_t, int, int, int);
static void nfe_miibus_statchg(device_t);
Add basic WOL support for MCP ethernet controllers. It seems the controller does not perform automatic switching from 1000Mbps link to 10/100Mbps link when WOL is activated. Implement establishing 10/100Mps link with auto-negotiation in driver. Link status change handler was modified to remove taskqueue based approach since driver now needs synchronous handling for link establishment. Submitted by: Yamagi Burmeister (lists <> yamagi.org ) (initial version) Tested by: Yamagi Burmeister (lists <> yamagi.org ) MFC after: 1 week
2010-11-11 18:08:50 +00:00
static void nfe_mac_config(struct nfe_softc *, struct mii_data *);
Bring overhauled nfe(4) into tree. o s/printf/device_printf/g o Nuke OpenBSDism. o Nuke NetBSD/OpenBSD specific DMA sync operations.(we don't have a way to sync a single descriptor within a DMA map.) o Remove recursive mutex. o bus_dma(9) clean up. o 40bit DMA address support. o Add protection for Rx map load failure. o Fix a long standing bug for watchdog timeout. [1] o Add additional protections, missing Tx completion interrupt, losing start Tx command, for watchdog timeout. o Switch to taskqueue(9) API to handle interrupts. o Use our own timer for watchdog instead of if_watchdog/if_timer interface. o Advertise VLAN header length/capability correctly to upper layer. o Remove excessive kernel stack consumption in nfe_encap(). o Handle highly fragmented mbuf chains correctly. o Enable etherenet address reprogramming with ifconfig(8). o Add ALTQ/TSO, MSI/MSIX support. o Increased Rx ring to 256 descriptors from 128. o Align Tx/Rx descriptor ring on sizeof(struct nfe_desc64) boundary. o Remove alignment restrictions on Tx/Rx buffers. o Rewritten jumbo frame support code. o Add support for hardware assistend VLAN tag insertion/stripping. o Add support for Tx/Rx flow control based on patches from Peer Chen. [2] o Add a routine that detects whether ethernet address swap routines is required. [3] o Add a workaround that take MAC/PHY out of power down mode. o Add suspend/resume support. o style(9) and code clean up. Special thanks to Shigeaki Tagashira, the original porter of nfe(4), who submitted lots of patches, performed uncountable number of regression tests and maintained nfe(4) for a long time. Without his enthusiastic help and support I could never have completed this overhauling task. The only weak point of nfe(4) compared to nve(4) is instability of manual half-duplex media selection on certain hardwares(auto sensing media type should work for all cases, though). This was a long standing bug of nfe(4) and I still have no idea why it doesn't work on some hardwares. Obtained from: OpenBSD [1] Submitted by: Peer Chen < pchen at nvidia dot com > [2], [3] Reviewed by: Shigeaki Tagashira < shigeaki AT se DOT hiroshima-u DOT ac DOT jp > Tested by: Shigeaki Tagashira, current Discussed with: current Silence from: obrien
2007-06-12 02:16:02 +00:00
static void nfe_set_intr(struct nfe_softc *);
static __inline void nfe_enable_intr(struct nfe_softc *);
static __inline void nfe_disable_intr(struct nfe_softc *);
style(9) cleanup.
2006-10-19 10:01:26 +00:00
static int nfe_ioctl(struct ifnet *, u_long, caddr_t);
Bring overhauled nfe(4) into tree. o s/printf/device_printf/g o Nuke OpenBSDism. o Nuke NetBSD/OpenBSD specific DMA sync operations.(we don't have a way to sync a single descriptor within a DMA map.) o Remove recursive mutex. o bus_dma(9) clean up. o 40bit DMA address support. o Add protection for Rx map load failure. o Fix a long standing bug for watchdog timeout. [1] o Add additional protections, missing Tx completion interrupt, losing start Tx command, for watchdog timeout. o Switch to taskqueue(9) API to handle interrupts. o Use our own timer for watchdog instead of if_watchdog/if_timer interface. o Advertise VLAN header length/capability correctly to upper layer. o Remove excessive kernel stack consumption in nfe_encap(). o Handle highly fragmented mbuf chains correctly. o Enable etherenet address reprogramming with ifconfig(8). o Add ALTQ/TSO, MSI/MSIX support. o Increased Rx ring to 256 descriptors from 128. o Align Tx/Rx descriptor ring on sizeof(struct nfe_desc64) boundary. o Remove alignment restrictions on Tx/Rx buffers. o Rewritten jumbo frame support code. o Add support for hardware assistend VLAN tag insertion/stripping. o Add support for Tx/Rx flow control based on patches from Peer Chen. [2] o Add a routine that detects whether ethernet address swap routines is required. [3] o Add a workaround that take MAC/PHY out of power down mode. o Add suspend/resume support. o style(9) and code clean up. Special thanks to Shigeaki Tagashira, the original porter of nfe(4), who submitted lots of patches, performed uncountable number of regression tests and maintained nfe(4) for a long time. Without his enthusiastic help and support I could never have completed this overhauling task. The only weak point of nfe(4) compared to nve(4) is instability of manual half-duplex media selection on certain hardwares(auto sensing media type should work for all cases, though). This was a long standing bug of nfe(4) and I still have no idea why it doesn't work on some hardwares. Obtained from: OpenBSD [1] Submitted by: Peer Chen < pchen at nvidia dot com > [2], [3] Reviewed by: Shigeaki Tagashira < shigeaki AT se DOT hiroshima-u DOT ac DOT jp > Tested by: Shigeaki Tagashira, current Discussed with: current Silence from: obrien
2007-06-12 02:16:02 +00:00
static void nfe_alloc_msix(struct nfe_softc *, int);
static int nfe_intr(void *);
static void nfe_int_task(void *, int);
static __inline void nfe_discard_rxbuf(struct nfe_softc *, int);
static __inline void nfe_discard_jrxbuf(struct nfe_softc *, int);
static int nfe_newbuf(struct nfe_softc *, int);
static int nfe_jnewbuf(struct nfe_softc *, int);
When user_frac in the polling subsystem is low it is going to busy the CPU for too long period than necessary. Additively, interfaces are kept polled (in the tick) even if no more packets are available. In order to avoid such situations a new generic mechanism can be implemented in proactive way, keeping track of the time spent on any packet and fragmenting the time for any tick, stopping the processing as soon as possible. In order to implement such mechanism, the polling handler needs to change, returning the number of packets processed. While the intended logic is not part of this patch, the polling KPI is broken by this commit, adding an int return value and the new flag IFCAP_POLLING_NOCOUNT (which will signal that the return value is meaningless for the installed handler and checking should be skipped). Bump __FreeBSD_version in order to signal such situation. Reviewed by: emaste Sponsored by: Sandvine Incorporated
2009-05-30 15:14:44 +00:00
static int nfe_rxeof(struct nfe_softc *, int, int *);
static int nfe_jrxeof(struct nfe_softc *, int, int *);
Add a pure open source nForce Ethernet driver, under BSDL. This driver was ported from OpenBSD by Shigeaki Tagashira <shigeaki@se.hiroshima-u.ac.jp> and posted at http://www.se.hiroshima-u.ac.jp/~shigeaki/software/freebsd-nfe.html It was additionally cleaned up by me. It is still a work-in-progress and thus is purposefully not in GENERIC. And it conflicts with nve(4), so only one should be loaded.
2006-06-26 23:41:07 +00:00
static void nfe_txeof(struct nfe_softc *);
Bring overhauled nfe(4) into tree. o s/printf/device_printf/g o Nuke OpenBSDism. o Nuke NetBSD/OpenBSD specific DMA sync operations.(we don't have a way to sync a single descriptor within a DMA map.) o Remove recursive mutex. o bus_dma(9) clean up. o 40bit DMA address support. o Add protection for Rx map load failure. o Fix a long standing bug for watchdog timeout. [1] o Add additional protections, missing Tx completion interrupt, losing start Tx command, for watchdog timeout. o Switch to taskqueue(9) API to handle interrupts. o Use our own timer for watchdog instead of if_watchdog/if_timer interface. o Advertise VLAN header length/capability correctly to upper layer. o Remove excessive kernel stack consumption in nfe_encap(). o Handle highly fragmented mbuf chains correctly. o Enable etherenet address reprogramming with ifconfig(8). o Add ALTQ/TSO, MSI/MSIX support. o Increased Rx ring to 256 descriptors from 128. o Align Tx/Rx descriptor ring on sizeof(struct nfe_desc64) boundary. o Remove alignment restrictions on Tx/Rx buffers. o Rewritten jumbo frame support code. o Add support for hardware assistend VLAN tag insertion/stripping. o Add support for Tx/Rx flow control based on patches from Peer Chen. [2] o Add a routine that detects whether ethernet address swap routines is required. [3] o Add a workaround that take MAC/PHY out of power down mode. o Add suspend/resume support. o style(9) and code clean up. Special thanks to Shigeaki Tagashira, the original porter of nfe(4), who submitted lots of patches, performed uncountable number of regression tests and maintained nfe(4) for a long time. Without his enthusiastic help and support I could never have completed this overhauling task. The only weak point of nfe(4) compared to nve(4) is instability of manual half-duplex media selection on certain hardwares(auto sensing media type should work for all cases, though). This was a long standing bug of nfe(4) and I still have no idea why it doesn't work on some hardwares. Obtained from: OpenBSD [1] Submitted by: Peer Chen < pchen at nvidia dot com > [2], [3] Reviewed by: Shigeaki Tagashira < shigeaki AT se DOT hiroshima-u DOT ac DOT jp > Tested by: Shigeaki Tagashira, current Discussed with: current Silence from: obrien
2007-06-12 02:16:02 +00:00
static int nfe_encap(struct nfe_softc *, struct mbuf **);
Add a pure open source nForce Ethernet driver, under BSDL. This driver was ported from OpenBSD by Shigeaki Tagashira <shigeaki@se.hiroshima-u.ac.jp> and posted at http://www.se.hiroshima-u.ac.jp/~shigeaki/software/freebsd-nfe.html It was additionally cleaned up by me. It is still a work-in-progress and thus is purposefully not in GENERIC. And it conflicts with nve(4), so only one should be loaded.
2006-06-26 23:41:07 +00:00
static void nfe_setmulti(struct nfe_softc *);
static void nfe_start(struct ifnet *);
Add a 'locked' variant of the foo_start() routine and call it directly from interrupt handlers and watchdog routines instead of queueing a task to call foo_start(). Reviewed by: yongari MFC after: 1 month
2011-01-03 18:28:30 +00:00
static void nfe_start_locked(struct ifnet *);
Add a pure open source nForce Ethernet driver, under BSDL. This driver was ported from OpenBSD by Shigeaki Tagashira <shigeaki@se.hiroshima-u.ac.jp> and posted at http://www.se.hiroshima-u.ac.jp/~shigeaki/software/freebsd-nfe.html It was additionally cleaned up by me. It is still a work-in-progress and thus is purposefully not in GENERIC. And it conflicts with nve(4), so only one should be loaded.
2006-06-26 23:41:07 +00:00
static void nfe_watchdog(struct ifnet *);
static void nfe_init(void *);
static void nfe_init_locked(void *);
Bring overhauled nfe(4) into tree. o s/printf/device_printf/g o Nuke OpenBSDism. o Nuke NetBSD/OpenBSD specific DMA sync operations.(we don't have a way to sync a single descriptor within a DMA map.) o Remove recursive mutex. o bus_dma(9) clean up. o 40bit DMA address support. o Add protection for Rx map load failure. o Fix a long standing bug for watchdog timeout. [1] o Add additional protections, missing Tx completion interrupt, losing start Tx command, for watchdog timeout. o Switch to taskqueue(9) API to handle interrupts. o Use our own timer for watchdog instead of if_watchdog/if_timer interface. o Advertise VLAN header length/capability correctly to upper layer. o Remove excessive kernel stack consumption in nfe_encap(). o Handle highly fragmented mbuf chains correctly. o Enable etherenet address reprogramming with ifconfig(8). o Add ALTQ/TSO, MSI/MSIX support. o Increased Rx ring to 256 descriptors from 128. o Align Tx/Rx descriptor ring on sizeof(struct nfe_desc64) boundary. o Remove alignment restrictions on Tx/Rx buffers. o Rewritten jumbo frame support code. o Add support for hardware assistend VLAN tag insertion/stripping. o Add support for Tx/Rx flow control based on patches from Peer Chen. [2] o Add a routine that detects whether ethernet address swap routines is required. [3] o Add a workaround that take MAC/PHY out of power down mode. o Add suspend/resume support. o style(9) and code clean up. Special thanks to Shigeaki Tagashira, the original porter of nfe(4), who submitted lots of patches, performed uncountable number of regression tests and maintained nfe(4) for a long time. Without his enthusiastic help and support I could never have completed this overhauling task. The only weak point of nfe(4) compared to nve(4) is instability of manual half-duplex media selection on certain hardwares(auto sensing media type should work for all cases, though). This was a long standing bug of nfe(4) and I still have no idea why it doesn't work on some hardwares. Obtained from: OpenBSD [1] Submitted by: Peer Chen < pchen at nvidia dot com > [2], [3] Reviewed by: Shigeaki Tagashira < shigeaki AT se DOT hiroshima-u DOT ac DOT jp > Tested by: Shigeaki Tagashira, current Discussed with: current Silence from: obrien
2007-06-12 02:16:02 +00:00
static void nfe_stop(struct ifnet *);
Add a pure open source nForce Ethernet driver, under BSDL. This driver was ported from OpenBSD by Shigeaki Tagashira <shigeaki@se.hiroshima-u.ac.jp> and posted at http://www.se.hiroshima-u.ac.jp/~shigeaki/software/freebsd-nfe.html It was additionally cleaned up by me. It is still a work-in-progress and thus is purposefully not in GENERIC. And it conflicts with nve(4), so only one should be loaded.
2006-06-26 23:41:07 +00:00
static int nfe_alloc_rx_ring(struct nfe_softc *, struct nfe_rx_ring *);
Don't fail on device attach if jumbo frame support was unsuccessful. Because nfe(4) hardware doesn't support SG on Rx path, supporting jumbo frame requires very large contiguous kernel memory(i.e. several mega bytes). In case of lack of contiguous kernel memory that allocation request may always fail. However nfe(4) can operate on normal sized MTU frames, so go ahead and just disable jumbo frame support. While I'm here add a new tunable "hw.nfe.jumbo_disable" to disable jumbo frame support. In nfe_poll, make sure to invoke correct Rx handler. Approved by: re (kensmith)
2007-07-24 01:11:00 +00:00
static void nfe_alloc_jrx_ring(struct nfe_softc *, struct nfe_jrx_ring *);
Bring overhauled nfe(4) into tree. o s/printf/device_printf/g o Nuke OpenBSDism. o Nuke NetBSD/OpenBSD specific DMA sync operations.(we don't have a way to sync a single descriptor within a DMA map.) o Remove recursive mutex. o bus_dma(9) clean up. o 40bit DMA address support. o Add protection for Rx map load failure. o Fix a long standing bug for watchdog timeout. [1] o Add additional protections, missing Tx completion interrupt, losing start Tx command, for watchdog timeout. o Switch to taskqueue(9) API to handle interrupts. o Use our own timer for watchdog instead of if_watchdog/if_timer interface. o Advertise VLAN header length/capability correctly to upper layer. o Remove excessive kernel stack consumption in nfe_encap(). o Handle highly fragmented mbuf chains correctly. o Enable etherenet address reprogramming with ifconfig(8). o Add ALTQ/TSO, MSI/MSIX support. o Increased Rx ring to 256 descriptors from 128. o Align Tx/Rx descriptor ring on sizeof(struct nfe_desc64) boundary. o Remove alignment restrictions on Tx/Rx buffers. o Rewritten jumbo frame support code. o Add support for hardware assistend VLAN tag insertion/stripping. o Add support for Tx/Rx flow control based on patches from Peer Chen. [2] o Add a routine that detects whether ethernet address swap routines is required. [3] o Add a workaround that take MAC/PHY out of power down mode. o Add suspend/resume support. o style(9) and code clean up. Special thanks to Shigeaki Tagashira, the original porter of nfe(4), who submitted lots of patches, performed uncountable number of regression tests and maintained nfe(4) for a long time. Without his enthusiastic help and support I could never have completed this overhauling task. The only weak point of nfe(4) compared to nve(4) is instability of manual half-duplex media selection on certain hardwares(auto sensing media type should work for all cases, though). This was a long standing bug of nfe(4) and I still have no idea why it doesn't work on some hardwares. Obtained from: OpenBSD [1] Submitted by: Peer Chen < pchen at nvidia dot com > [2], [3] Reviewed by: Shigeaki Tagashira < shigeaki AT se DOT hiroshima-u DOT ac DOT jp > Tested by: Shigeaki Tagashira, current Discussed with: current Silence from: obrien
2007-06-12 02:16:02 +00:00
static int nfe_init_rx_ring(struct nfe_softc *, struct nfe_rx_ring *);
static int nfe_init_jrx_ring(struct nfe_softc *, struct nfe_jrx_ring *);
Add a pure open source nForce Ethernet driver, under BSDL. This driver was ported from OpenBSD by Shigeaki Tagashira <shigeaki@se.hiroshima-u.ac.jp> and posted at http://www.se.hiroshima-u.ac.jp/~shigeaki/software/freebsd-nfe.html It was additionally cleaned up by me. It is still a work-in-progress and thus is purposefully not in GENERIC. And it conflicts with nve(4), so only one should be loaded.
2006-06-26 23:41:07 +00:00
static void nfe_free_rx_ring(struct nfe_softc *, struct nfe_rx_ring *);
Bring overhauled nfe(4) into tree. o s/printf/device_printf/g o Nuke OpenBSDism. o Nuke NetBSD/OpenBSD specific DMA sync operations.(we don't have a way to sync a single descriptor within a DMA map.) o Remove recursive mutex. o bus_dma(9) clean up. o 40bit DMA address support. o Add protection for Rx map load failure. o Fix a long standing bug for watchdog timeout. [1] o Add additional protections, missing Tx completion interrupt, losing start Tx command, for watchdog timeout. o Switch to taskqueue(9) API to handle interrupts. o Use our own timer for watchdog instead of if_watchdog/if_timer interface. o Advertise VLAN header length/capability correctly to upper layer. o Remove excessive kernel stack consumption in nfe_encap(). o Handle highly fragmented mbuf chains correctly. o Enable etherenet address reprogramming with ifconfig(8). o Add ALTQ/TSO, MSI/MSIX support. o Increased Rx ring to 256 descriptors from 128. o Align Tx/Rx descriptor ring on sizeof(struct nfe_desc64) boundary. o Remove alignment restrictions on Tx/Rx buffers. o Rewritten jumbo frame support code. o Add support for hardware assistend VLAN tag insertion/stripping. o Add support for Tx/Rx flow control based on patches from Peer Chen. [2] o Add a routine that detects whether ethernet address swap routines is required. [3] o Add a workaround that take MAC/PHY out of power down mode. o Add suspend/resume support. o style(9) and code clean up. Special thanks to Shigeaki Tagashira, the original porter of nfe(4), who submitted lots of patches, performed uncountable number of regression tests and maintained nfe(4) for a long time. Without his enthusiastic help and support I could never have completed this overhauling task. The only weak point of nfe(4) compared to nve(4) is instability of manual half-duplex media selection on certain hardwares(auto sensing media type should work for all cases, though). This was a long standing bug of nfe(4) and I still have no idea why it doesn't work on some hardwares. Obtained from: OpenBSD [1] Submitted by: Peer Chen < pchen at nvidia dot com > [2], [3] Reviewed by: Shigeaki Tagashira < shigeaki AT se DOT hiroshima-u DOT ac DOT jp > Tested by: Shigeaki Tagashira, current Discussed with: current Silence from: obrien
2007-06-12 02:16:02 +00:00
static void nfe_free_jrx_ring(struct nfe_softc *, struct nfe_jrx_ring *);
Add a pure open source nForce Ethernet driver, under BSDL. This driver was ported from OpenBSD by Shigeaki Tagashira <shigeaki@se.hiroshima-u.ac.jp> and posted at http://www.se.hiroshima-u.ac.jp/~shigeaki/software/freebsd-nfe.html It was additionally cleaned up by me. It is still a work-in-progress and thus is purposefully not in GENERIC. And it conflicts with nve(4), so only one should be loaded.
2006-06-26 23:41:07 +00:00
static int nfe_alloc_tx_ring(struct nfe_softc *, struct nfe_tx_ring *);
Bring overhauled nfe(4) into tree. o s/printf/device_printf/g o Nuke OpenBSDism. o Nuke NetBSD/OpenBSD specific DMA sync operations.(we don't have a way to sync a single descriptor within a DMA map.) o Remove recursive mutex. o bus_dma(9) clean up. o 40bit DMA address support. o Add protection for Rx map load failure. o Fix a long standing bug for watchdog timeout. [1] o Add additional protections, missing Tx completion interrupt, losing start Tx command, for watchdog timeout. o Switch to taskqueue(9) API to handle interrupts. o Use our own timer for watchdog instead of if_watchdog/if_timer interface. o Advertise VLAN header length/capability correctly to upper layer. o Remove excessive kernel stack consumption in nfe_encap(). o Handle highly fragmented mbuf chains correctly. o Enable etherenet address reprogramming with ifconfig(8). o Add ALTQ/TSO, MSI/MSIX support. o Increased Rx ring to 256 descriptors from 128. o Align Tx/Rx descriptor ring on sizeof(struct nfe_desc64) boundary. o Remove alignment restrictions on Tx/Rx buffers. o Rewritten jumbo frame support code. o Add support for hardware assistend VLAN tag insertion/stripping. o Add support for Tx/Rx flow control based on patches from Peer Chen. [2] o Add a routine that detects whether ethernet address swap routines is required. [3] o Add a workaround that take MAC/PHY out of power down mode. o Add suspend/resume support. o style(9) and code clean up. Special thanks to Shigeaki Tagashira, the original porter of nfe(4), who submitted lots of patches, performed uncountable number of regression tests and maintained nfe(4) for a long time. Without his enthusiastic help and support I could never have completed this overhauling task. The only weak point of nfe(4) compared to nve(4) is instability of manual half-duplex media selection on certain hardwares(auto sensing media type should work for all cases, though). This was a long standing bug of nfe(4) and I still have no idea why it doesn't work on some hardwares. Obtained from: OpenBSD [1] Submitted by: Peer Chen < pchen at nvidia dot com > [2], [3] Reviewed by: Shigeaki Tagashira < shigeaki AT se DOT hiroshima-u DOT ac DOT jp > Tested by: Shigeaki Tagashira, current Discussed with: current Silence from: obrien
2007-06-12 02:16:02 +00:00
static void nfe_init_tx_ring(struct nfe_softc *, struct nfe_tx_ring *);
Add a pure open source nForce Ethernet driver, under BSDL. This driver was ported from OpenBSD by Shigeaki Tagashira <shigeaki@se.hiroshima-u.ac.jp> and posted at http://www.se.hiroshima-u.ac.jp/~shigeaki/software/freebsd-nfe.html It was additionally cleaned up by me. It is still a work-in-progress and thus is purposefully not in GENERIC. And it conflicts with nve(4), so only one should be loaded.
2006-06-26 23:41:07 +00:00
static void nfe_free_tx_ring(struct nfe_softc *, struct nfe_tx_ring *);
static int nfe_ifmedia_upd(struct ifnet *);
static void nfe_ifmedia_sts(struct ifnet *, struct ifmediareq *);
static void nfe_tick(void *);
Bring overhauled nfe(4) into tree. o s/printf/device_printf/g o Nuke OpenBSDism. o Nuke NetBSD/OpenBSD specific DMA sync operations.(we don't have a way to sync a single descriptor within a DMA map.) o Remove recursive mutex. o bus_dma(9) clean up. o 40bit DMA address support. o Add protection for Rx map load failure. o Fix a long standing bug for watchdog timeout. [1] o Add additional protections, missing Tx completion interrupt, losing start Tx command, for watchdog timeout. o Switch to taskqueue(9) API to handle interrupts. o Use our own timer for watchdog instead of if_watchdog/if_timer interface. o Advertise VLAN header length/capability correctly to upper layer. o Remove excessive kernel stack consumption in nfe_encap(). o Handle highly fragmented mbuf chains correctly. o Enable etherenet address reprogramming with ifconfig(8). o Add ALTQ/TSO, MSI/MSIX support. o Increased Rx ring to 256 descriptors from 128. o Align Tx/Rx descriptor ring on sizeof(struct nfe_desc64) boundary. o Remove alignment restrictions on Tx/Rx buffers. o Rewritten jumbo frame support code. o Add support for hardware assistend VLAN tag insertion/stripping. o Add support for Tx/Rx flow control based on patches from Peer Chen. [2] o Add a routine that detects whether ethernet address swap routines is required. [3] o Add a workaround that take MAC/PHY out of power down mode. o Add suspend/resume support. o style(9) and code clean up. Special thanks to Shigeaki Tagashira, the original porter of nfe(4), who submitted lots of patches, performed uncountable number of regression tests and maintained nfe(4) for a long time. Without his enthusiastic help and support I could never have completed this overhauling task. The only weak point of nfe(4) compared to nve(4) is instability of manual half-duplex media selection on certain hardwares(auto sensing media type should work for all cases, though). This was a long standing bug of nfe(4) and I still have no idea why it doesn't work on some hardwares. Obtained from: OpenBSD [1] Submitted by: Peer Chen < pchen at nvidia dot com > [2], [3] Reviewed by: Shigeaki Tagashira < shigeaki AT se DOT hiroshima-u DOT ac DOT jp > Tested by: Shigeaki Tagashira, current Discussed with: current Silence from: obrien
2007-06-12 02:16:02 +00:00
static void nfe_get_macaddr(struct nfe_softc *, uint8_t *);
static void nfe_set_macaddr(struct nfe_softc *, uint8_t *);
static void nfe_dma_map_segs(void *, bus_dma_segment_t *, int, int);
static int sysctl_int_range(SYSCTL_HANDLER_ARGS, int, int);
static int sysctl_hw_nfe_proc_limit(SYSCTL_HANDLER_ARGS);
Add hardware MAC statistics support. Register information was obtained from Linux forcedeth driver. While I'm here move creating a sysctl node for process_limit to function nfe_sysctl_node(). Tested by: "Arno J. Klaassen" < arno <at> heho dot snv dot jussieu dot fr >
2008-10-03 03:58:16 +00:00
static void nfe_sysctl_node(struct nfe_softc *);
static void nfe_stats_clear(struct nfe_softc *);
static void nfe_stats_update(struct nfe_softc *);
Add basic WOL support for MCP ethernet controllers. It seems the controller does not perform automatic switching from 1000Mbps link to 10/100Mbps link when WOL is activated. Implement establishing 10/100Mps link with auto-negotiation in driver. Link status change handler was modified to remove taskqueue based approach since driver now needs synchronous handling for link establishment. Submitted by: Yamagi Burmeister (lists <> yamagi.org ) (initial version) Tested by: Yamagi Burmeister (lists <> yamagi.org ) MFC after: 1 week
2010-11-11 18:08:50 +00:00
static void nfe_set_linkspeed(struct nfe_softc *);
static void nfe_set_wol(struct nfe_softc *);
Import the nfe (nForce Ethernet) driver from OpenBSD. This is a pure open source NIC driver with a BSD license. These bits were taken from the OpenBSD public CVS repository on 1-May-06.
2006-06-26 17:25:22 +00:00
#ifdef NFE_DEBUG
Bring overhauled nfe(4) into tree. o s/printf/device_printf/g o Nuke OpenBSDism. o Nuke NetBSD/OpenBSD specific DMA sync operations.(we don't have a way to sync a single descriptor within a DMA map.) o Remove recursive mutex. o bus_dma(9) clean up. o 40bit DMA address support. o Add protection for Rx map load failure. o Fix a long standing bug for watchdog timeout. [1] o Add additional protections, missing Tx completion interrupt, losing start Tx command, for watchdog timeout. o Switch to taskqueue(9) API to handle interrupts. o Use our own timer for watchdog instead of if_watchdog/if_timer interface. o Advertise VLAN header length/capability correctly to upper layer. o Remove excessive kernel stack consumption in nfe_encap(). o Handle highly fragmented mbuf chains correctly. o Enable etherenet address reprogramming with ifconfig(8). o Add ALTQ/TSO, MSI/MSIX support. o Increased Rx ring to 256 descriptors from 128. o Align Tx/Rx descriptor ring on sizeof(struct nfe_desc64) boundary. o Remove alignment restrictions on Tx/Rx buffers. o Rewritten jumbo frame support code. o Add support for hardware assistend VLAN tag insertion/stripping. o Add support for Tx/Rx flow control based on patches from Peer Chen. [2] o Add a routine that detects whether ethernet address swap routines is required. [3] o Add a workaround that take MAC/PHY out of power down mode. o Add suspend/resume support. o style(9) and code clean up. Special thanks to Shigeaki Tagashira, the original porter of nfe(4), who submitted lots of patches, performed uncountable number of regression tests and maintained nfe(4) for a long time. Without his enthusiastic help and support I could never have completed this overhauling task. The only weak point of nfe(4) compared to nve(4) is instability of manual half-duplex media selection on certain hardwares(auto sensing media type should work for all cases, though). This was a long standing bug of nfe(4) and I still have no idea why it doesn't work on some hardwares. Obtained from: OpenBSD [1] Submitted by: Peer Chen < pchen at nvidia dot com > [2], [3] Reviewed by: Shigeaki Tagashira < shigeaki AT se DOT hiroshima-u DOT ac DOT jp > Tested by: Shigeaki Tagashira, current Discussed with: current Silence from: obrien
2007-06-12 02:16:02 +00:00
static int nfedebug = 0;
#define DPRINTF(sc, ...) do { \
if (nfedebug) \
device_printf((sc)->nfe_dev, __VA_ARGS__); \
} while (0)
#define DPRINTFN(sc, n, ...) do { \
if (nfedebug >= (n)) \
device_printf((sc)->nfe_dev, __VA_ARGS__); \
} while (0)
Import the nfe (nForce Ethernet) driver from OpenBSD. This is a pure open source NIC driver with a BSD license. These bits were taken from the OpenBSD public CVS repository on 1-May-06.
2006-06-26 17:25:22 +00:00
#else
Bring overhauled nfe(4) into tree. o s/printf/device_printf/g o Nuke OpenBSDism. o Nuke NetBSD/OpenBSD specific DMA sync operations.(we don't have a way to sync a single descriptor within a DMA map.) o Remove recursive mutex. o bus_dma(9) clean up. o 40bit DMA address support. o Add protection for Rx map load failure. o Fix a long standing bug for watchdog timeout. [1] o Add additional protections, missing Tx completion interrupt, losing start Tx command, for watchdog timeout. o Switch to taskqueue(9) API to handle interrupts. o Use our own timer for watchdog instead of if_watchdog/if_timer interface. o Advertise VLAN header length/capability correctly to upper layer. o Remove excessive kernel stack consumption in nfe_encap(). o Handle highly fragmented mbuf chains correctly. o Enable etherenet address reprogramming with ifconfig(8). o Add ALTQ/TSO, MSI/MSIX support. o Increased Rx ring to 256 descriptors from 128. o Align Tx/Rx descriptor ring on sizeof(struct nfe_desc64) boundary. o Remove alignment restrictions on Tx/Rx buffers. o Rewritten jumbo frame support code. o Add support for hardware assistend VLAN tag insertion/stripping. o Add support for Tx/Rx flow control based on patches from Peer Chen. [2] o Add a routine that detects whether ethernet address swap routines is required. [3] o Add a workaround that take MAC/PHY out of power down mode. o Add suspend/resume support. o style(9) and code clean up. Special thanks to Shigeaki Tagashira, the original porter of nfe(4), who submitted lots of patches, performed uncountable number of regression tests and maintained nfe(4) for a long time. Without his enthusiastic help and support I could never have completed this overhauling task. The only weak point of nfe(4) compared to nve(4) is instability of manual half-duplex media selection on certain hardwares(auto sensing media type should work for all cases, though). This was a long standing bug of nfe(4) and I still have no idea why it doesn't work on some hardwares. Obtained from: OpenBSD [1] Submitted by: Peer Chen < pchen at nvidia dot com > [2], [3] Reviewed by: Shigeaki Tagashira < shigeaki AT se DOT hiroshima-u DOT ac DOT jp > Tested by: Shigeaki Tagashira, current Discussed with: current Silence from: obrien
2007-06-12 02:16:02 +00:00
#define DPRINTF(sc, ...)
#define DPRINTFN(sc, n, ...)
Import the nfe (nForce Ethernet) driver from OpenBSD. This is a pure open source NIC driver with a BSD license. These bits were taken from the OpenBSD public CVS repository on 1-May-06.
2006-06-26 17:25:22 +00:00
#endif
Add a pure open source nForce Ethernet driver, under BSDL. This driver was ported from OpenBSD by Shigeaki Tagashira <shigeaki@se.hiroshima-u.ac.jp> and posted at http://www.se.hiroshima-u.ac.jp/~shigeaki/software/freebsd-nfe.html It was additionally cleaned up by me. It is still a work-in-progress and thus is purposefully not in GENERIC. And it conflicts with nve(4), so only one should be loaded.
2006-06-26 23:41:07 +00:00
#define NFE_LOCK(_sc) mtx_lock(&(_sc)->nfe_mtx)
#define NFE_UNLOCK(_sc) mtx_unlock(&(_sc)->nfe_mtx)
#define NFE_LOCK_ASSERT(_sc) mtx_assert(&(_sc)->nfe_mtx, MA_OWNED)
Bring overhauled nfe(4) into tree. o s/printf/device_printf/g o Nuke OpenBSDism. o Nuke NetBSD/OpenBSD specific DMA sync operations.(we don't have a way to sync a single descriptor within a DMA map.) o Remove recursive mutex. o bus_dma(9) clean up. o 40bit DMA address support. o Add protection for Rx map load failure. o Fix a long standing bug for watchdog timeout. [1] o Add additional protections, missing Tx completion interrupt, losing start Tx command, for watchdog timeout. o Switch to taskqueue(9) API to handle interrupts. o Use our own timer for watchdog instead of if_watchdog/if_timer interface. o Advertise VLAN header length/capability correctly to upper layer. o Remove excessive kernel stack consumption in nfe_encap(). o Handle highly fragmented mbuf chains correctly. o Enable etherenet address reprogramming with ifconfig(8). o Add ALTQ/TSO, MSI/MSIX support. o Increased Rx ring to 256 descriptors from 128. o Align Tx/Rx descriptor ring on sizeof(struct nfe_desc64) boundary. o Remove alignment restrictions on Tx/Rx buffers. o Rewritten jumbo frame support code. o Add support for hardware assistend VLAN tag insertion/stripping. o Add support for Tx/Rx flow control based on patches from Peer Chen. [2] o Add a routine that detects whether ethernet address swap routines is required. [3] o Add a workaround that take MAC/PHY out of power down mode. o Add suspend/resume support. o style(9) and code clean up. Special thanks to Shigeaki Tagashira, the original porter of nfe(4), who submitted lots of patches, performed uncountable number of regression tests and maintained nfe(4) for a long time. Without his enthusiastic help and support I could never have completed this overhauling task. The only weak point of nfe(4) compared to nve(4) is instability of manual half-duplex media selection on certain hardwares(auto sensing media type should work for all cases, though). This was a long standing bug of nfe(4) and I still have no idea why it doesn't work on some hardwares. Obtained from: OpenBSD [1] Submitted by: Peer Chen < pchen at nvidia dot com > [2], [3] Reviewed by: Shigeaki Tagashira < shigeaki AT se DOT hiroshima-u DOT ac DOT jp > Tested by: Shigeaki Tagashira, current Discussed with: current Silence from: obrien
2007-06-12 02:16:02 +00:00
/* Tunables. */
static int msi_disable = 0;
static int msix_disable = 0;
Don't fail on device attach if jumbo frame support was unsuccessful. Because nfe(4) hardware doesn't support SG on Rx path, supporting jumbo frame requires very large contiguous kernel memory(i.e. several mega bytes). In case of lack of contiguous kernel memory that allocation request may always fail. However nfe(4) can operate on normal sized MTU frames, so go ahead and just disable jumbo frame support. While I'm here add a new tunable "hw.nfe.jumbo_disable" to disable jumbo frame support. In nfe_poll, make sure to invoke correct Rx handler. Approved by: re (kensmith)
2007-07-24 01:11:00 +00:00
static int jumbo_disable = 0;
Bring overhauled nfe(4) into tree. o s/printf/device_printf/g o Nuke OpenBSDism. o Nuke NetBSD/OpenBSD specific DMA sync operations.(we don't have a way to sync a single descriptor within a DMA map.) o Remove recursive mutex. o bus_dma(9) clean up. o 40bit DMA address support. o Add protection for Rx map load failure. o Fix a long standing bug for watchdog timeout. [1] o Add additional protections, missing Tx completion interrupt, losing start Tx command, for watchdog timeout. o Switch to taskqueue(9) API to handle interrupts. o Use our own timer for watchdog instead of if_watchdog/if_timer interface. o Advertise VLAN header length/capability correctly to upper layer. o Remove excessive kernel stack consumption in nfe_encap(). o Handle highly fragmented mbuf chains correctly. o Enable etherenet address reprogramming with ifconfig(8). o Add ALTQ/TSO, MSI/MSIX support. o Increased Rx ring to 256 descriptors from 128. o Align Tx/Rx descriptor ring on sizeof(struct nfe_desc64) boundary. o Remove alignment restrictions on Tx/Rx buffers. o Rewritten jumbo frame support code. o Add support for hardware assistend VLAN tag insertion/stripping. o Add support for Tx/Rx flow control based on patches from Peer Chen. [2] o Add a routine that detects whether ethernet address swap routines is required. [3] o Add a workaround that take MAC/PHY out of power down mode. o Add suspend/resume support. o style(9) and code clean up. Special thanks to Shigeaki Tagashira, the original porter of nfe(4), who submitted lots of patches, performed uncountable number of regression tests and maintained nfe(4) for a long time. Without his enthusiastic help and support I could never have completed this overhauling task. The only weak point of nfe(4) compared to nve(4) is instability of manual half-duplex media selection on certain hardwares(auto sensing media type should work for all cases, though). This was a long standing bug of nfe(4) and I still have no idea why it doesn't work on some hardwares. Obtained from: OpenBSD [1] Submitted by: Peer Chen < pchen at nvidia dot com > [2], [3] Reviewed by: Shigeaki Tagashira < shigeaki AT se DOT hiroshima-u DOT ac DOT jp > Tested by: Shigeaki Tagashira, current Discussed with: current Silence from: obrien
2007-06-12 02:16:02 +00:00
TUNABLE_INT("hw.nfe.msi_disable", &msi_disable);
TUNABLE_INT("hw.nfe.msix_disable", &msix_disable);
Don't fail on device attach if jumbo frame support was unsuccessful. Because nfe(4) hardware doesn't support SG on Rx path, supporting jumbo frame requires very large contiguous kernel memory(i.e. several mega bytes). In case of lack of contiguous kernel memory that allocation request may always fail. However nfe(4) can operate on normal sized MTU frames, so go ahead and just disable jumbo frame support. While I'm here add a new tunable "hw.nfe.jumbo_disable" to disable jumbo frame support. In nfe_poll, make sure to invoke correct Rx handler. Approved by: re (kensmith)
2007-07-24 01:11:00 +00:00
TUNABLE_INT("hw.nfe.jumbo_disable", &jumbo_disable);
Add a pure open source nForce Ethernet driver, under BSDL. This driver was ported from OpenBSD by Shigeaki Tagashira <shigeaki@se.hiroshima-u.ac.jp> and posted at http://www.se.hiroshima-u.ac.jp/~shigeaki/software/freebsd-nfe.html It was additionally cleaned up by me. It is still a work-in-progress and thus is purposefully not in GENERIC. And it conflicts with nve(4), so only one should be loaded.
2006-06-26 23:41:07 +00:00
static device_method_t nfe_methods[] = {
/* Device interface */
DEVMETHOD(device_probe, nfe_probe),
DEVMETHOD(device_attach, nfe_attach),
DEVMETHOD(device_detach, nfe_detach),
Bring overhauled nfe(4) into tree. o s/printf/device_printf/g o Nuke OpenBSDism. o Nuke NetBSD/OpenBSD specific DMA sync operations.(we don't have a way to sync a single descriptor within a DMA map.) o Remove recursive mutex. o bus_dma(9) clean up. o 40bit DMA address support. o Add protection for Rx map load failure. o Fix a long standing bug for watchdog timeout. [1] o Add additional protections, missing Tx completion interrupt, losing start Tx command, for watchdog timeout. o Switch to taskqueue(9) API to handle interrupts. o Use our own timer for watchdog instead of if_watchdog/if_timer interface. o Advertise VLAN header length/capability correctly to upper layer. o Remove excessive kernel stack consumption in nfe_encap(). o Handle highly fragmented mbuf chains correctly. o Enable etherenet address reprogramming with ifconfig(8). o Add ALTQ/TSO, MSI/MSIX support. o Increased Rx ring to 256 descriptors from 128. o Align Tx/Rx descriptor ring on sizeof(struct nfe_desc64) boundary. o Remove alignment restrictions on Tx/Rx buffers. o Rewritten jumbo frame support code. o Add support for hardware assistend VLAN tag insertion/stripping. o Add support for Tx/Rx flow control based on patches from Peer Chen. [2] o Add a routine that detects whether ethernet address swap routines is required. [3] o Add a workaround that take MAC/PHY out of power down mode. o Add suspend/resume support. o style(9) and code clean up. Special thanks to Shigeaki Tagashira, the original porter of nfe(4), who submitted lots of patches, performed uncountable number of regression tests and maintained nfe(4) for a long time. Without his enthusiastic help and support I could never have completed this overhauling task. The only weak point of nfe(4) compared to nve(4) is instability of manual half-duplex media selection on certain hardwares(auto sensing media type should work for all cases, though). This was a long standing bug of nfe(4) and I still have no idea why it doesn't work on some hardwares. Obtained from: OpenBSD [1] Submitted by: Peer Chen < pchen at nvidia dot com > [2], [3] Reviewed by: Shigeaki Tagashira < shigeaki AT se DOT hiroshima-u DOT ac DOT jp > Tested by: Shigeaki Tagashira, current Discussed with: current Silence from: obrien
2007-06-12 02:16:02 +00:00
DEVMETHOD(device_suspend, nfe_suspend),
DEVMETHOD(device_resume, nfe_resume),
Add a pure open source nForce Ethernet driver, under BSDL. This driver was ported from OpenBSD by Shigeaki Tagashira <shigeaki@se.hiroshima-u.ac.jp> and posted at http://www.se.hiroshima-u.ac.jp/~shigeaki/software/freebsd-nfe.html It was additionally cleaned up by me. It is still a work-in-progress and thus is purposefully not in GENERIC. And it conflicts with nve(4), so only one should be loaded.
2006-06-26 23:41:07 +00:00
DEVMETHOD(device_shutdown, nfe_shutdown),
/* bus interface */
DEVMETHOD(bus_print_child, bus_generic_print_child),
DEVMETHOD(bus_driver_added, bus_generic_driver_added),
/* MII interface */
DEVMETHOD(miibus_readreg, nfe_miibus_readreg),
DEVMETHOD(miibus_writereg, nfe_miibus_writereg),
style(9) cleanup.
2006-10-19 10:01:26 +00:00
DEVMETHOD(miibus_statchg, nfe_miibus_statchg),
Add a pure open source nForce Ethernet driver, under BSDL. This driver was ported from OpenBSD by Shigeaki Tagashira <shigeaki@se.hiroshima-u.ac.jp> and posted at http://www.se.hiroshima-u.ac.jp/~shigeaki/software/freebsd-nfe.html It was additionally cleaned up by me. It is still a work-in-progress and thus is purposefully not in GENERIC. And it conflicts with nve(4), so only one should be loaded.
2006-06-26 23:41:07 +00:00
Bring overhauled nfe(4) into tree. o s/printf/device_printf/g o Nuke OpenBSDism. o Nuke NetBSD/OpenBSD specific DMA sync operations.(we don't have a way to sync a single descriptor within a DMA map.) o Remove recursive mutex. o bus_dma(9) clean up. o 40bit DMA address support. o Add protection for Rx map load failure. o Fix a long standing bug for watchdog timeout. [1] o Add additional protections, missing Tx completion interrupt, losing start Tx command, for watchdog timeout. o Switch to taskqueue(9) API to handle interrupts. o Use our own timer for watchdog instead of if_watchdog/if_timer interface. o Advertise VLAN header length/capability correctly to upper layer. o Remove excessive kernel stack consumption in nfe_encap(). o Handle highly fragmented mbuf chains correctly. o Enable etherenet address reprogramming with ifconfig(8). o Add ALTQ/TSO, MSI/MSIX support. o Increased Rx ring to 256 descriptors from 128. o Align Tx/Rx descriptor ring on sizeof(struct nfe_desc64) boundary. o Remove alignment restrictions on Tx/Rx buffers. o Rewritten jumbo frame support code. o Add support for hardware assistend VLAN tag insertion/stripping. o Add support for Tx/Rx flow control based on patches from Peer Chen. [2] o Add a routine that detects whether ethernet address swap routines is required. [3] o Add a workaround that take MAC/PHY out of power down mode. o Add suspend/resume support. o style(9) and code clean up. Special thanks to Shigeaki Tagashira, the original porter of nfe(4), who submitted lots of patches, performed uncountable number of regression tests and maintained nfe(4) for a long time. Without his enthusiastic help and support I could never have completed this overhauling task. The only weak point of nfe(4) compared to nve(4) is instability of manual half-duplex media selection on certain hardwares(auto sensing media type should work for all cases, though). This was a long standing bug of nfe(4) and I still have no idea why it doesn't work on some hardwares. Obtained from: OpenBSD [1] Submitted by: Peer Chen < pchen at nvidia dot com > [2], [3] Reviewed by: Shigeaki Tagashira < shigeaki AT se DOT hiroshima-u DOT ac DOT jp > Tested by: Shigeaki Tagashira, current Discussed with: current Silence from: obrien
2007-06-12 02:16:02 +00:00
{ NULL, NULL }
Import the nfe (nForce Ethernet) driver from OpenBSD. This is a pure open source NIC driver with a BSD license. These bits were taken from the OpenBSD public CVS repository on 1-May-06.
2006-06-26 17:25:22 +00:00
};
Add a pure open source nForce Ethernet driver, under BSDL. This driver was ported from OpenBSD by Shigeaki Tagashira <shigeaki@se.hiroshima-u.ac.jp> and posted at http://www.se.hiroshima-u.ac.jp/~shigeaki/software/freebsd-nfe.html It was additionally cleaned up by me. It is still a work-in-progress and thus is purposefully not in GENERIC. And it conflicts with nve(4), so only one should be loaded.
2006-06-26 23:41:07 +00:00
static driver_t nfe_driver = {
"nfe",
nfe_methods,
sizeof(struct nfe_softc)
};
static devclass_t nfe_devclass;
DRIVER_MODULE(nfe, pci, nfe_driver, nfe_devclass, 0, 0);
DRIVER_MODULE(miibus, nfe, miibus_driver, miibus_devclass, 0, 0);
static struct nfe_type nfe_devs[] = {
{PCI_VENDOR_NVIDIA, PCI_PRODUCT_NVIDIA_NFORCE_LAN,
style(9) cleanup.
2006-10-19 10:01:26 +00:00
"NVIDIA nForce MCP Networking Adapter"},
Add a pure open source nForce Ethernet driver, under BSDL. This driver was ported from OpenBSD by Shigeaki Tagashira <shigeaki@se.hiroshima-u.ac.jp> and posted at http://www.se.hiroshima-u.ac.jp/~shigeaki/software/freebsd-nfe.html It was additionally cleaned up by me. It is still a work-in-progress and thus is purposefully not in GENERIC. And it conflicts with nve(4), so only one should be loaded.
2006-06-26 23:41:07 +00:00
{PCI_VENDOR_NVIDIA, PCI_PRODUCT_NVIDIA_NFORCE2_LAN,
style(9) cleanup.
2006-10-19 10:01:26 +00:00
"NVIDIA nForce2 MCP2 Networking Adapter"},
Add a pure open source nForce Ethernet driver, under BSDL. This driver was ported from OpenBSD by Shigeaki Tagashira <shigeaki@se.hiroshima-u.ac.jp> and posted at http://www.se.hiroshima-u.ac.jp/~shigeaki/software/freebsd-nfe.html It was additionally cleaned up by me. It is still a work-in-progress and thus is purposefully not in GENERIC. And it conflicts with nve(4), so only one should be loaded.
2006-06-26 23:41:07 +00:00
{PCI_VENDOR_NVIDIA, PCI_PRODUCT_NVIDIA_NFORCE2_400_LAN1,
style(9) cleanup.
2006-10-19 10:01:26 +00:00
"NVIDIA nForce2 400 MCP4 Networking Adapter"},
Add a pure open source nForce Ethernet driver, under BSDL. This driver was ported from OpenBSD by Shigeaki Tagashira <shigeaki@se.hiroshima-u.ac.jp> and posted at http://www.se.hiroshima-u.ac.jp/~shigeaki/software/freebsd-nfe.html It was additionally cleaned up by me. It is still a work-in-progress and thus is purposefully not in GENERIC. And it conflicts with nve(4), so only one should be loaded.
2006-06-26 23:41:07 +00:00
{PCI_VENDOR_NVIDIA, PCI_PRODUCT_NVIDIA_NFORCE2_400_LAN2,
style(9) cleanup.
2006-10-19 10:01:26 +00:00
"NVIDIA nForce2 400 MCP5 Networking Adapter"},
Try to be more precise in identifying the various nForce networking adapters.
2006-10-16 16:36:29 +00:00
{PCI_VENDOR_NVIDIA, PCI_PRODUCT_NVIDIA_NFORCE3_LAN1,
style(9) cleanup.
2006-10-19 10:01:26 +00:00
"NVIDIA nForce3 MCP3 Networking Adapter"},
Add a pure open source nForce Ethernet driver, under BSDL. This driver was ported from OpenBSD by Shigeaki Tagashira <shigeaki@se.hiroshima-u.ac.jp> and posted at http://www.se.hiroshima-u.ac.jp/~shigeaki/software/freebsd-nfe.html It was additionally cleaned up by me. It is still a work-in-progress and thus is purposefully not in GENERIC. And it conflicts with nve(4), so only one should be loaded.
2006-06-26 23:41:07 +00:00
{PCI_VENDOR_NVIDIA, PCI_PRODUCT_NVIDIA_NFORCE3_250_LAN,
style(9) cleanup.
2006-10-19 10:01:26 +00:00
"NVIDIA nForce3 250 MCP6 Networking Adapter"},
Add a pure open source nForce Ethernet driver, under BSDL. This driver was ported from OpenBSD by Shigeaki Tagashira <shigeaki@se.hiroshima-u.ac.jp> and posted at http://www.se.hiroshima-u.ac.jp/~shigeaki/software/freebsd-nfe.html It was additionally cleaned up by me. It is still a work-in-progress and thus is purposefully not in GENERIC. And it conflicts with nve(4), so only one should be loaded.
2006-06-26 23:41:07 +00:00
{PCI_VENDOR_NVIDIA, PCI_PRODUCT_NVIDIA_NFORCE3_LAN4,
style(9) cleanup.
2006-10-19 10:01:26 +00:00
"NVIDIA nForce3 MCP7 Networking Adapter"},
Add a pure open source nForce Ethernet driver, under BSDL. This driver was ported from OpenBSD by Shigeaki Tagashira <shigeaki@se.hiroshima-u.ac.jp> and posted at http://www.se.hiroshima-u.ac.jp/~shigeaki/software/freebsd-nfe.html It was additionally cleaned up by me. It is still a work-in-progress and thus is purposefully not in GENERIC. And it conflicts with nve(4), so only one should be loaded.
2006-06-26 23:41:07 +00:00
{PCI_VENDOR_NVIDIA, PCI_PRODUCT_NVIDIA_NFORCE4_LAN1,
style(9) cleanup.
2006-10-19 10:01:26 +00:00
"NVIDIA nForce4 CK804 MCP8 Networking Adapter"},
Add a pure open source nForce Ethernet driver, under BSDL. This driver was ported from OpenBSD by Shigeaki Tagashira <shigeaki@se.hiroshima-u.ac.jp> and posted at http://www.se.hiroshima-u.ac.jp/~shigeaki/software/freebsd-nfe.html It was additionally cleaned up by me. It is still a work-in-progress and thus is purposefully not in GENERIC. And it conflicts with nve(4), so only one should be loaded.
2006-06-26 23:41:07 +00:00
{PCI_VENDOR_NVIDIA, PCI_PRODUCT_NVIDIA_NFORCE4_LAN2,
style(9) cleanup.
2006-10-19 10:01:26 +00:00
"NVIDIA nForce4 CK804 MCP9 Networking Adapter"},
Add a pure open source nForce Ethernet driver, under BSDL. This driver was ported from OpenBSD by Shigeaki Tagashira <shigeaki@se.hiroshima-u.ac.jp> and posted at http://www.se.hiroshima-u.ac.jp/~shigeaki/software/freebsd-nfe.html It was additionally cleaned up by me. It is still a work-in-progress and thus is purposefully not in GENERIC. And it conflicts with nve(4), so only one should be loaded.
2006-06-26 23:41:07 +00:00
{PCI_VENDOR_NVIDIA, PCI_PRODUCT_NVIDIA_MCP04_LAN1,
Bring overhauled nfe(4) into tree. o s/printf/device_printf/g o Nuke OpenBSDism. o Nuke NetBSD/OpenBSD specific DMA sync operations.(we don't have a way to sync a single descriptor within a DMA map.) o Remove recursive mutex. o bus_dma(9) clean up. o 40bit DMA address support. o Add protection for Rx map load failure. o Fix a long standing bug for watchdog timeout. [1] o Add additional protections, missing Tx completion interrupt, losing start Tx command, for watchdog timeout. o Switch to taskqueue(9) API to handle interrupts. o Use our own timer for watchdog instead of if_watchdog/if_timer interface. o Advertise VLAN header length/capability correctly to upper layer. o Remove excessive kernel stack consumption in nfe_encap(). o Handle highly fragmented mbuf chains correctly. o Enable etherenet address reprogramming with ifconfig(8). o Add ALTQ/TSO, MSI/MSIX support. o Increased Rx ring to 256 descriptors from 128. o Align Tx/Rx descriptor ring on sizeof(struct nfe_desc64) boundary. o Remove alignment restrictions on Tx/Rx buffers. o Rewritten jumbo frame support code. o Add support for hardware assistend VLAN tag insertion/stripping. o Add support for Tx/Rx flow control based on patches from Peer Chen. [2] o Add a routine that detects whether ethernet address swap routines is required. [3] o Add a workaround that take MAC/PHY out of power down mode. o Add suspend/resume support. o style(9) and code clean up. Special thanks to Shigeaki Tagashira, the original porter of nfe(4), who submitted lots of patches, performed uncountable number of regression tests and maintained nfe(4) for a long time. Without his enthusiastic help and support I could never have completed this overhauling task. The only weak point of nfe(4) compared to nve(4) is instability of manual half-duplex media selection on certain hardwares(auto sensing media type should work for all cases, though). This was a long standing bug of nfe(4) and I still have no idea why it doesn't work on some hardwares. Obtained from: OpenBSD [1] Submitted by: Peer Chen < pchen at nvidia dot com > [2], [3] Reviewed by: Shigeaki Tagashira < shigeaki AT se DOT hiroshima-u DOT ac DOT jp > Tested by: Shigeaki Tagashira, current Discussed with: current Silence from: obrien
2007-06-12 02:16:02 +00:00
"NVIDIA nForce MCP04 Networking Adapter"}, /* MCP10 */
Add a pure open source nForce Ethernet driver, under BSDL. This driver was ported from OpenBSD by Shigeaki Tagashira <shigeaki@se.hiroshima-u.ac.jp> and posted at http://www.se.hiroshima-u.ac.jp/~shigeaki/software/freebsd-nfe.html It was additionally cleaned up by me. It is still a work-in-progress and thus is purposefully not in GENERIC. And it conflicts with nve(4), so only one should be loaded.
2006-06-26 23:41:07 +00:00
{PCI_VENDOR_NVIDIA, PCI_PRODUCT_NVIDIA_MCP04_LAN2,
Bring overhauled nfe(4) into tree. o s/printf/device_printf/g o Nuke OpenBSDism. o Nuke NetBSD/OpenBSD specific DMA sync operations.(we don't have a way to sync a single descriptor within a DMA map.) o Remove recursive mutex. o bus_dma(9) clean up. o 40bit DMA address support. o Add protection for Rx map load failure. o Fix a long standing bug for watchdog timeout. [1] o Add additional protections, missing Tx completion interrupt, losing start Tx command, for watchdog timeout. o Switch to taskqueue(9) API to handle interrupts. o Use our own timer for watchdog instead of if_watchdog/if_timer interface. o Advertise VLAN header length/capability correctly to upper layer. o Remove excessive kernel stack consumption in nfe_encap(). o Handle highly fragmented mbuf chains correctly. o Enable etherenet address reprogramming with ifconfig(8). o Add ALTQ/TSO, MSI/MSIX support. o Increased Rx ring to 256 descriptors from 128. o Align Tx/Rx descriptor ring on sizeof(struct nfe_desc64) boundary. o Remove alignment restrictions on Tx/Rx buffers. o Rewritten jumbo frame support code. o Add support for hardware assistend VLAN tag insertion/stripping. o Add support for Tx/Rx flow control based on patches from Peer Chen. [2] o Add a routine that detects whether ethernet address swap routines is required. [3] o Add a workaround that take MAC/PHY out of power down mode. o Add suspend/resume support. o style(9) and code clean up. Special thanks to Shigeaki Tagashira, the original porter of nfe(4), who submitted lots of patches, performed uncountable number of regression tests and maintained nfe(4) for a long time. Without his enthusiastic help and support I could never have completed this overhauling task. The only weak point of nfe(4) compared to nve(4) is instability of manual half-duplex media selection on certain hardwares(auto sensing media type should work for all cases, though). This was a long standing bug of nfe(4) and I still have no idea why it doesn't work on some hardwares. Obtained from: OpenBSD [1] Submitted by: Peer Chen < pchen at nvidia dot com > [2], [3] Reviewed by: Shigeaki Tagashira < shigeaki AT se DOT hiroshima-u DOT ac DOT jp > Tested by: Shigeaki Tagashira, current Discussed with: current Silence from: obrien
2007-06-12 02:16:02 +00:00
"NVIDIA nForce MCP04 Networking Adapter"}, /* MCP11 */
Add a pure open source nForce Ethernet driver, under BSDL. This driver was ported from OpenBSD by Shigeaki Tagashira <shigeaki@se.hiroshima-u.ac.jp> and posted at http://www.se.hiroshima-u.ac.jp/~shigeaki/software/freebsd-nfe.html It was additionally cleaned up by me. It is still a work-in-progress and thus is purposefully not in GENERIC. And it conflicts with nve(4), so only one should be loaded.
2006-06-26 23:41:07 +00:00
{PCI_VENDOR_NVIDIA, PCI_PRODUCT_NVIDIA_NFORCE430_LAN1,
style(9) cleanup.
2006-10-19 10:01:26 +00:00
"NVIDIA nForce 430 MCP12 Networking Adapter"},
Add a pure open source nForce Ethernet driver, under BSDL. This driver was ported from OpenBSD by Shigeaki Tagashira <shigeaki@se.hiroshima-u.ac.jp> and posted at http://www.se.hiroshima-u.ac.jp/~shigeaki/software/freebsd-nfe.html It was additionally cleaned up by me. It is still a work-in-progress and thus is purposefully not in GENERIC. And it conflicts with nve(4), so only one should be loaded.
2006-06-26 23:41:07 +00:00
{PCI_VENDOR_NVIDIA, PCI_PRODUCT_NVIDIA_NFORCE430_LAN2,
style(9) cleanup.
2006-10-19 10:01:26 +00:00
"NVIDIA nForce 430 MCP13 Networking Adapter"},
Add a pure open source nForce Ethernet driver, under BSDL. This driver was ported from OpenBSD by Shigeaki Tagashira <shigeaki@se.hiroshima-u.ac.jp> and posted at http://www.se.hiroshima-u.ac.jp/~shigeaki/software/freebsd-nfe.html It was additionally cleaned up by me. It is still a work-in-progress and thus is purposefully not in GENERIC. And it conflicts with nve(4), so only one should be loaded.
2006-06-26 23:41:07 +00:00
{PCI_VENDOR_NVIDIA, PCI_PRODUCT_NVIDIA_MCP55_LAN1,
style(9) cleanup.
2006-10-19 10:01:26 +00:00
"NVIDIA nForce MCP55 Networking Adapter"},
Add a pure open source nForce Ethernet driver, under BSDL. This driver was ported from OpenBSD by Shigeaki Tagashira <shigeaki@se.hiroshima-u.ac.jp> and posted at http://www.se.hiroshima-u.ac.jp/~shigeaki/software/freebsd-nfe.html It was additionally cleaned up by me. It is still a work-in-progress and thus is purposefully not in GENERIC. And it conflicts with nve(4), so only one should be loaded.
2006-06-26 23:41:07 +00:00
{PCI_VENDOR_NVIDIA, PCI_PRODUCT_NVIDIA_MCP55_LAN2,
style(9) cleanup.
2006-10-19 10:01:26 +00:00
"NVIDIA nForce MCP55 Networking Adapter"},
Add support for MCP61 and MCP65 adaptors. Submitted by: Peer Chen at nVIDIA
2006-09-11 07:31:55 +00:00
{PCI_VENDOR_NVIDIA, PCI_PRODUCT_NVIDIA_MCP61_LAN1,
style(9) cleanup.
2006-10-19 10:01:26 +00:00
"NVIDIA nForce MCP61 Networking Adapter"},
Add support for MCP61 and MCP65 adaptors. Submitted by: Peer Chen at nVIDIA
2006-09-11 07:31:55 +00:00
{PCI_VENDOR_NVIDIA, PCI_PRODUCT_NVIDIA_MCP61_LAN2,
style(9) cleanup.
2006-10-19 10:01:26 +00:00
"NVIDIA nForce MCP61 Networking Adapter"},
Add support for MCP61 and MCP65 adaptors. Submitted by: Peer Chen at nVIDIA
2006-09-11 07:31:55 +00:00
{PCI_VENDOR_NVIDIA, PCI_PRODUCT_NVIDIA_MCP61_LAN3,
style(9) cleanup.
2006-10-19 10:01:26 +00:00
"NVIDIA nForce MCP61 Networking Adapter"},
Bring overhauled nfe(4) into tree. o s/printf/device_printf/g o Nuke OpenBSDism. o Nuke NetBSD/OpenBSD specific DMA sync operations.(we don't have a way to sync a single descriptor within a DMA map.) o Remove recursive mutex. o bus_dma(9) clean up. o 40bit DMA address support. o Add protection for Rx map load failure. o Fix a long standing bug for watchdog timeout. [1] o Add additional protections, missing Tx completion interrupt, losing start Tx command, for watchdog timeout. o Switch to taskqueue(9) API to handle interrupts. o Use our own timer for watchdog instead of if_watchdog/if_timer interface. o Advertise VLAN header length/capability correctly to upper layer. o Remove excessive kernel stack consumption in nfe_encap(). o Handle highly fragmented mbuf chains correctly. o Enable etherenet address reprogramming with ifconfig(8). o Add ALTQ/TSO, MSI/MSIX support. o Increased Rx ring to 256 descriptors from 128. o Align Tx/Rx descriptor ring on sizeof(struct nfe_desc64) boundary. o Remove alignment restrictions on Tx/Rx buffers. o Rewritten jumbo frame support code. o Add support for hardware assistend VLAN tag insertion/stripping. o Add support for Tx/Rx flow control based on patches from Peer Chen. [2] o Add a routine that detects whether ethernet address swap routines is required. [3] o Add a workaround that take MAC/PHY out of power down mode. o Add suspend/resume support. o style(9) and code clean up. Special thanks to Shigeaki Tagashira, the original porter of nfe(4), who submitted lots of patches, performed uncountable number of regression tests and maintained nfe(4) for a long time. Without his enthusiastic help and support I could never have completed this overhauling task. The only weak point of nfe(4) compared to nve(4) is instability of manual half-duplex media selection on certain hardwares(auto sensing media type should work for all cases, though). This was a long standing bug of nfe(4) and I still have no idea why it doesn't work on some hardwares. Obtained from: OpenBSD [1] Submitted by: Peer Chen < pchen at nvidia dot com > [2], [3] Reviewed by: Shigeaki Tagashira < shigeaki AT se DOT hiroshima-u DOT ac DOT jp > Tested by: Shigeaki Tagashira, current Discussed with: current Silence from: obrien
2007-06-12 02:16:02 +00:00
{PCI_VENDOR_NVIDIA, PCI_PRODUCT_NVIDIA_MCP61_LAN4,
style(9) cleanup.
2006-10-19 10:01:26 +00:00
"NVIDIA nForce MCP61 Networking Adapter"},
Add support for MCP61 and MCP65 adaptors. Submitted by: Peer Chen at nVIDIA
2006-09-11 07:31:55 +00:00
{PCI_VENDOR_NVIDIA, PCI_PRODUCT_NVIDIA_MCP65_LAN1,
style(9) cleanup.
2006-10-19 10:01:26 +00:00
"NVIDIA nForce MCP65 Networking Adapter"},
Add support for MCP61 and MCP65 adaptors. Submitted by: Peer Chen at nVIDIA
2006-09-11 07:31:55 +00:00
{PCI_VENDOR_NVIDIA, PCI_PRODUCT_NVIDIA_MCP65_LAN2,
style(9) cleanup.
2006-10-19 10:01:26 +00:00
"NVIDIA nForce MCP65 Networking Adapter"},
Add support for MCP61 and MCP65 adaptors. Submitted by: Peer Chen at nVIDIA
2006-09-11 07:31:55 +00:00
{PCI_VENDOR_NVIDIA, PCI_PRODUCT_NVIDIA_MCP65_LAN3,
style(9) cleanup.
2006-10-19 10:01:26 +00:00
"NVIDIA nForce MCP65 Networking Adapter"},
Bring overhauled nfe(4) into tree. o s/printf/device_printf/g o Nuke OpenBSDism. o Nuke NetBSD/OpenBSD specific DMA sync operations.(we don't have a way to sync a single descriptor within a DMA map.) o Remove recursive mutex. o bus_dma(9) clean up. o 40bit DMA address support. o Add protection for Rx map load failure. o Fix a long standing bug for watchdog timeout. [1] o Add additional protections, missing Tx completion interrupt, losing start Tx command, for watchdog timeout. o Switch to taskqueue(9) API to handle interrupts. o Use our own timer for watchdog instead of if_watchdog/if_timer interface. o Advertise VLAN header length/capability correctly to upper layer. o Remove excessive kernel stack consumption in nfe_encap(). o Handle highly fragmented mbuf chains correctly. o Enable etherenet address reprogramming with ifconfig(8). o Add ALTQ/TSO, MSI/MSIX support. o Increased Rx ring to 256 descriptors from 128. o Align Tx/Rx descriptor ring on sizeof(struct nfe_desc64) boundary. o Remove alignment restrictions on Tx/Rx buffers. o Rewritten jumbo frame support code. o Add support for hardware assistend VLAN tag insertion/stripping. o Add support for Tx/Rx flow control based on patches from Peer Chen. [2] o Add a routine that detects whether ethernet address swap routines is required. [3] o Add a workaround that take MAC/PHY out of power down mode. o Add suspend/resume support. o style(9) and code clean up. Special thanks to Shigeaki Tagashira, the original porter of nfe(4), who submitted lots of patches, performed uncountable number of regression tests and maintained nfe(4) for a long time. Without his enthusiastic help and support I could never have completed this overhauling task. The only weak point of nfe(4) compared to nve(4) is instability of manual half-duplex media selection on certain hardwares(auto sensing media type should work for all cases, though). This was a long standing bug of nfe(4) and I still have no idea why it doesn't work on some hardwares. Obtained from: OpenBSD [1] Submitted by: Peer Chen < pchen at nvidia dot com > [2], [3] Reviewed by: Shigeaki Tagashira < shigeaki AT se DOT hiroshima-u DOT ac DOT jp > Tested by: Shigeaki Tagashira, current Discussed with: current Silence from: obrien
2007-06-12 02:16:02 +00:00
{PCI_VENDOR_NVIDIA, PCI_PRODUCT_NVIDIA_MCP65_LAN4,
style(9) cleanup.
2006-10-19 10:01:26 +00:00
"NVIDIA nForce MCP65 Networking Adapter"},
Bring overhauled nfe(4) into tree. o s/printf/device_printf/g o Nuke OpenBSDism. o Nuke NetBSD/OpenBSD specific DMA sync operations.(we don't have a way to sync a single descriptor within a DMA map.) o Remove recursive mutex. o bus_dma(9) clean up. o 40bit DMA address support. o Add protection for Rx map load failure. o Fix a long standing bug for watchdog timeout. [1] o Add additional protections, missing Tx completion interrupt, losing start Tx command, for watchdog timeout. o Switch to taskqueue(9) API to handle interrupts. o Use our own timer for watchdog instead of if_watchdog/if_timer interface. o Advertise VLAN header length/capability correctly to upper layer. o Remove excessive kernel stack consumption in nfe_encap(). o Handle highly fragmented mbuf chains correctly. o Enable etherenet address reprogramming with ifconfig(8). o Add ALTQ/TSO, MSI/MSIX support. o Increased Rx ring to 256 descriptors from 128. o Align Tx/Rx descriptor ring on sizeof(struct nfe_desc64) boundary. o Remove alignment restrictions on Tx/Rx buffers. o Rewritten jumbo frame support code. o Add support for hardware assistend VLAN tag insertion/stripping. o Add support for Tx/Rx flow control based on patches from Peer Chen. [2] o Add a routine that detects whether ethernet address swap routines is required. [3] o Add a workaround that take MAC/PHY out of power down mode. o Add suspend/resume support. o style(9) and code clean up. Special thanks to Shigeaki Tagashira, the original porter of nfe(4), who submitted lots of patches, performed uncountable number of regression tests and maintained nfe(4) for a long time. Without his enthusiastic help and support I could never have completed this overhauling task. The only weak point of nfe(4) compared to nve(4) is instability of manual half-duplex media selection on certain hardwares(auto sensing media type should work for all cases, though). This was a long standing bug of nfe(4) and I still have no idea why it doesn't work on some hardwares. Obtained from: OpenBSD [1] Submitted by: Peer Chen < pchen at nvidia dot com > [2], [3] Reviewed by: Shigeaki Tagashira < shigeaki AT se DOT hiroshima-u DOT ac DOT jp > Tested by: Shigeaki Tagashira, current Discussed with: current Silence from: obrien
2007-06-12 02:16:02 +00:00
{PCI_VENDOR_NVIDIA, PCI_PRODUCT_NVIDIA_MCP67_LAN1,
"NVIDIA nForce MCP67 Networking Adapter"},
{PCI_VENDOR_NVIDIA, PCI_PRODUCT_NVIDIA_MCP67_LAN2,
"NVIDIA nForce MCP67 Networking Adapter"},
{PCI_VENDOR_NVIDIA, PCI_PRODUCT_NVIDIA_MCP67_LAN3,
"NVIDIA nForce MCP67 Networking Adapter"},
{PCI_VENDOR_NVIDIA, PCI_PRODUCT_NVIDIA_MCP67_LAN4,
"NVIDIA nForce MCP67 Networking Adapter"},
Add support for MCP73 chips. Tested by: "Daan Vreeken [PA4DAN]" ( Danovitsch AT vitsch DOT net ) MFC after: 1 week
2008-04-10 01:25:09 +00:00
{PCI_VENDOR_NVIDIA, PCI_PRODUCT_NVIDIA_MCP73_LAN1,
"NVIDIA nForce MCP73 Networking Adapter"},
{PCI_VENDOR_NVIDIA, PCI_PRODUCT_NVIDIA_MCP73_LAN2,
"NVIDIA nForce MCP73 Networking Adapter"},
{PCI_VENDOR_NVIDIA, PCI_PRODUCT_NVIDIA_MCP73_LAN3,
"NVIDIA nForce MCP73 Networking Adapter"},
{PCI_VENDOR_NVIDIA, PCI_PRODUCT_NVIDIA_MCP73_LAN4,
"NVIDIA nForce MCP73 Networking Adapter"},
Add device ids for MCP77/79 and set appropriate support flags. Obtained from: OpenBSD PR: kern/127529
2008-10-01 00:17:54 +00:00
{PCI_VENDOR_NVIDIA, PCI_PRODUCT_NVIDIA_MCP77_LAN1,
"NVIDIA nForce MCP77 Networking Adapter"},
{PCI_VENDOR_NVIDIA, PCI_PRODUCT_NVIDIA_MCP77_LAN2,
"NVIDIA nForce MCP77 Networking Adapter"},
{PCI_VENDOR_NVIDIA, PCI_PRODUCT_NVIDIA_MCP77_LAN3,
"NVIDIA nForce MCP77 Networking Adapter"},
{PCI_VENDOR_NVIDIA, PCI_PRODUCT_NVIDIA_MCP77_LAN4,
"NVIDIA nForce MCP77 Networking Adapter"},
{PCI_VENDOR_NVIDIA, PCI_PRODUCT_NVIDIA_MCP79_LAN1,
"NVIDIA nForce MCP79 Networking Adapter"},
{PCI_VENDOR_NVIDIA, PCI_PRODUCT_NVIDIA_MCP79_LAN2,
"NVIDIA nForce MCP79 Networking Adapter"},
{PCI_VENDOR_NVIDIA, PCI_PRODUCT_NVIDIA_MCP79_LAN3,
"NVIDIA nForce MCP79 Networking Adapter"},
{PCI_VENDOR_NVIDIA, PCI_PRODUCT_NVIDIA_MCP79_LAN4,
"NVIDIA nForce MCP79 Networking Adapter"},
Add a pure open source nForce Ethernet driver, under BSDL. This driver was ported from OpenBSD by Shigeaki Tagashira <shigeaki@se.hiroshima-u.ac.jp> and posted at http://www.se.hiroshima-u.ac.jp/~shigeaki/software/freebsd-nfe.html It was additionally cleaned up by me. It is still a work-in-progress and thus is purposefully not in GENERIC. And it conflicts with nve(4), so only one should be loaded.
2006-06-26 23:41:07 +00:00
{0, 0, NULL}
};
/* Probe for supported hardware ID's */
static int
nfe_probe(device_t dev)
Import the nfe (nForce Ethernet) driver from OpenBSD. This is a pure open source NIC driver with a BSD license. These bits were taken from the OpenBSD public CVS repository on 1-May-06.
2006-06-26 17:25:22 +00:00
{
Add a pure open source nForce Ethernet driver, under BSDL. This driver was ported from OpenBSD by Shigeaki Tagashira <shigeaki@se.hiroshima-u.ac.jp> and posted at http://www.se.hiroshima-u.ac.jp/~shigeaki/software/freebsd-nfe.html It was additionally cleaned up by me. It is still a work-in-progress and thus is purposefully not in GENERIC. And it conflicts with nve(4), so only one should be loaded.
2006-06-26 23:41:07 +00:00
struct nfe_type *t;
t = nfe_devs;
/* Check for matching PCI DEVICE ID's */
while (t->name != NULL) {
if ((pci_get_vendor(dev) == t->vid_id) &&
(pci_get_device(dev) == t->dev_id)) {
device_set_desc(dev, t->name);
Bring overhauled nfe(4) into tree. o s/printf/device_printf/g o Nuke OpenBSDism. o Nuke NetBSD/OpenBSD specific DMA sync operations.(we don't have a way to sync a single descriptor within a DMA map.) o Remove recursive mutex. o bus_dma(9) clean up. o 40bit DMA address support. o Add protection for Rx map load failure. o Fix a long standing bug for watchdog timeout. [1] o Add additional protections, missing Tx completion interrupt, losing start Tx command, for watchdog timeout. o Switch to taskqueue(9) API to handle interrupts. o Use our own timer for watchdog instead of if_watchdog/if_timer interface. o Advertise VLAN header length/capability correctly to upper layer. o Remove excessive kernel stack consumption in nfe_encap(). o Handle highly fragmented mbuf chains correctly. o Enable etherenet address reprogramming with ifconfig(8). o Add ALTQ/TSO, MSI/MSIX support. o Increased Rx ring to 256 descriptors from 128. o Align Tx/Rx descriptor ring on sizeof(struct nfe_desc64) boundary. o Remove alignment restrictions on Tx/Rx buffers. o Rewritten jumbo frame support code. o Add support for hardware assistend VLAN tag insertion/stripping. o Add support for Tx/Rx flow control based on patches from Peer Chen. [2] o Add a routine that detects whether ethernet address swap routines is required. [3] o Add a workaround that take MAC/PHY out of power down mode. o Add suspend/resume support. o style(9) and code clean up. Special thanks to Shigeaki Tagashira, the original porter of nfe(4), who submitted lots of patches, performed uncountable number of regression tests and maintained nfe(4) for a long time. Without his enthusiastic help and support I could never have completed this overhauling task. The only weak point of nfe(4) compared to nve(4) is instability of manual half-duplex media selection on certain hardwares(auto sensing media type should work for all cases, though). This was a long standing bug of nfe(4) and I still have no idea why it doesn't work on some hardwares. Obtained from: OpenBSD [1] Submitted by: Peer Chen < pchen at nvidia dot com > [2], [3] Reviewed by: Shigeaki Tagashira < shigeaki AT se DOT hiroshima-u DOT ac DOT jp > Tested by: Shigeaki Tagashira, current Discussed with: current Silence from: obrien
2007-06-12 02:16:02 +00:00
return (BUS_PROBE_DEFAULT);
Add a pure open source nForce Ethernet driver, under BSDL. This driver was ported from OpenBSD by Shigeaki Tagashira <shigeaki@se.hiroshima-u.ac.jp> and posted at http://www.se.hiroshima-u.ac.jp/~shigeaki/software/freebsd-nfe.html It was additionally cleaned up by me. It is still a work-in-progress and thus is purposefully not in GENERIC. And it conflicts with nve(4), so only one should be loaded.
2006-06-26 23:41:07 +00:00
}
t++;
}
return (ENXIO);
Import the nfe (nForce Ethernet) driver from OpenBSD. This is a pure open source NIC driver with a BSD license. These bits were taken from the OpenBSD public CVS repository on 1-May-06.
2006-06-26 17:25:22 +00:00
}
Bring overhauled nfe(4) into tree. o s/printf/device_printf/g o Nuke OpenBSDism. o Nuke NetBSD/OpenBSD specific DMA sync operations.(we don't have a way to sync a single descriptor within a DMA map.) o Remove recursive mutex. o bus_dma(9) clean up. o 40bit DMA address support. o Add protection for Rx map load failure. o Fix a long standing bug for watchdog timeout. [1] o Add additional protections, missing Tx completion interrupt, losing start Tx command, for watchdog timeout. o Switch to taskqueue(9) API to handle interrupts. o Use our own timer for watchdog instead of if_watchdog/if_timer interface. o Advertise VLAN header length/capability correctly to upper layer. o Remove excessive kernel stack consumption in nfe_encap(). o Handle highly fragmented mbuf chains correctly. o Enable etherenet address reprogramming with ifconfig(8). o Add ALTQ/TSO, MSI/MSIX support. o Increased Rx ring to 256 descriptors from 128. o Align Tx/Rx descriptor ring on sizeof(struct nfe_desc64) boundary. o Remove alignment restrictions on Tx/Rx buffers. o Rewritten jumbo frame support code. o Add support for hardware assistend VLAN tag insertion/stripping. o Add support for Tx/Rx flow control based on patches from Peer Chen. [2] o Add a routine that detects whether ethernet address swap routines is required. [3] o Add a workaround that take MAC/PHY out of power down mode. o Add suspend/resume support. o style(9) and code clean up. Special thanks to Shigeaki Tagashira, the original porter of nfe(4), who submitted lots of patches, performed uncountable number of regression tests and maintained nfe(4) for a long time. Without his enthusiastic help and support I could never have completed this overhauling task. The only weak point of nfe(4) compared to nve(4) is instability of manual half-duplex media selection on certain hardwares(auto sensing media type should work for all cases, though). This was a long standing bug of nfe(4) and I still have no idea why it doesn't work on some hardwares. Obtained from: OpenBSD [1] Submitted by: Peer Chen < pchen at nvidia dot com > [2], [3] Reviewed by: Shigeaki Tagashira < shigeaki AT se DOT hiroshima-u DOT ac DOT jp > Tested by: Shigeaki Tagashira, current Discussed with: current Silence from: obrien
2007-06-12 02:16:02 +00:00
static void
nfe_alloc_msix(struct nfe_softc *sc, int count)
{
int rid;
rid = PCIR_BAR(2);
sc->nfe_msix_res = bus_alloc_resource_any(sc->nfe_dev, SYS_RES_MEMORY,
&rid, RF_ACTIVE);
if (sc->nfe_msix_res == NULL) {
device_printf(sc->nfe_dev,
"couldn't allocate MSIX table resource\n");
return;
}
rid = PCIR_BAR(3);
sc->nfe_msix_pba_res = bus_alloc_resource_any(sc->nfe_dev,
SYS_RES_MEMORY, &rid, RF_ACTIVE);
if (sc->nfe_msix_pba_res == NULL) {
device_printf(sc->nfe_dev,
"couldn't allocate MSIX PBA resource\n");
bus_release_resource(sc->nfe_dev, SYS_RES_MEMORY, PCIR_BAR(2),
sc->nfe_msix_res);
sc->nfe_msix_res = NULL;
return;
}
if (pci_alloc_msix(sc->nfe_dev, &count) == 0) {
if (count == NFE_MSI_MESSAGES) {
if (bootverbose)
device_printf(sc->nfe_dev,
"Using %d MSIX messages\n", count);
sc->nfe_msix = 1;
} else {
if (bootverbose)
device_printf(sc->nfe_dev,
"couldn't allocate MSIX\n");
pci_release_msi(sc->nfe_dev);
bus_release_resource(sc->nfe_dev, SYS_RES_MEMORY,
PCIR_BAR(3), sc->nfe_msix_pba_res);
bus_release_resource(sc->nfe_dev, SYS_RES_MEMORY,
PCIR_BAR(2), sc->nfe_msix_res);
sc->nfe_msix_pba_res = NULL;
sc->nfe_msix_res = NULL;
}
}
}
style(9) cleanup.
2006-10-19 10:01:26 +00:00
Add a pure open source nForce Ethernet driver, under BSDL. This driver was ported from OpenBSD by Shigeaki Tagashira <shigeaki@se.hiroshima-u.ac.jp> and posted at http://www.se.hiroshima-u.ac.jp/~shigeaki/software/freebsd-nfe.html It was additionally cleaned up by me. It is still a work-in-progress and thus is purposefully not in GENERIC. And it conflicts with nve(4), so only one should be loaded.
2006-06-26 23:41:07 +00:00
static int
nfe_attach(device_t dev)
Import the nfe (nForce Ethernet) driver from OpenBSD. This is a pure open source NIC driver with a BSD license. These bits were taken from the OpenBSD public CVS repository on 1-May-06.
2006-06-26 17:25:22 +00:00
{
Add a pure open source nForce Ethernet driver, under BSDL. This driver was ported from OpenBSD by Shigeaki Tagashira <shigeaki@se.hiroshima-u.ac.jp> and posted at http://www.se.hiroshima-u.ac.jp/~shigeaki/software/freebsd-nfe.html It was additionally cleaned up by me. It is still a work-in-progress and thus is purposefully not in GENERIC. And it conflicts with nve(4), so only one should be loaded.
2006-06-26 23:41:07 +00:00
struct nfe_softc *sc;
Import the nfe (nForce Ethernet) driver from OpenBSD. This is a pure open source NIC driver with a BSD license. These bits were taken from the OpenBSD public CVS repository on 1-May-06.
2006-06-26 17:25:22 +00:00
struct ifnet *ifp;
Bring overhauled nfe(4) into tree. o s/printf/device_printf/g o Nuke OpenBSDism. o Nuke NetBSD/OpenBSD specific DMA sync operations.(we don't have a way to sync a single descriptor within a DMA map.) o Remove recursive mutex. o bus_dma(9) clean up. o 40bit DMA address support. o Add protection for Rx map load failure. o Fix a long standing bug for watchdog timeout. [1] o Add additional protections, missing Tx completion interrupt, losing start Tx command, for watchdog timeout. o Switch to taskqueue(9) API to handle interrupts. o Use our own timer for watchdog instead of if_watchdog/if_timer interface. o Advertise VLAN header length/capability correctly to upper layer. o Remove excessive kernel stack consumption in nfe_encap(). o Handle highly fragmented mbuf chains correctly. o Enable etherenet address reprogramming with ifconfig(8). o Add ALTQ/TSO, MSI/MSIX support. o Increased Rx ring to 256 descriptors from 128. o Align Tx/Rx descriptor ring on sizeof(struct nfe_desc64) boundary. o Remove alignment restrictions on Tx/Rx buffers. o Rewritten jumbo frame support code. o Add support for hardware assistend VLAN tag insertion/stripping. o Add support for Tx/Rx flow control based on patches from Peer Chen. [2] o Add a routine that detects whether ethernet address swap routines is required. [3] o Add a workaround that take MAC/PHY out of power down mode. o Add suspend/resume support. o style(9) and code clean up. Special thanks to Shigeaki Tagashira, the original porter of nfe(4), who submitted lots of patches, performed uncountable number of regression tests and maintained nfe(4) for a long time. Without his enthusiastic help and support I could never have completed this overhauling task. The only weak point of nfe(4) compared to nve(4) is instability of manual half-duplex media selection on certain hardwares(auto sensing media type should work for all cases, though). This was a long standing bug of nfe(4) and I still have no idea why it doesn't work on some hardwares. Obtained from: OpenBSD [1] Submitted by: Peer Chen < pchen at nvidia dot com > [2], [3] Reviewed by: Shigeaki Tagashira < shigeaki AT se DOT hiroshima-u DOT ac DOT jp > Tested by: Shigeaki Tagashira, current Discussed with: current Silence from: obrien
2007-06-12 02:16:02 +00:00
bus_addr_t dma_addr_max;
int error = 0, i, msic, reg, rid;
Import the nfe (nForce Ethernet) driver from OpenBSD. This is a pure open source NIC driver with a BSD license. These bits were taken from the OpenBSD public CVS repository on 1-May-06.
2006-06-26 17:25:22 +00:00
Add a pure open source nForce Ethernet driver, under BSDL. This driver was ported from OpenBSD by Shigeaki Tagashira <shigeaki@se.hiroshima-u.ac.jp> and posted at http://www.se.hiroshima-u.ac.jp/~shigeaki/software/freebsd-nfe.html It was additionally cleaned up by me. It is still a work-in-progress and thus is purposefully not in GENERIC. And it conflicts with nve(4), so only one should be loaded.
2006-06-26 23:41:07 +00:00
sc = device_get_softc(dev);
sc->nfe_dev = dev;
Import the nfe (nForce Ethernet) driver from OpenBSD. This is a pure open source NIC driver with a BSD license. These bits were taken from the OpenBSD public CVS repository on 1-May-06.
2006-06-26 17:25:22 +00:00
Add a pure open source nForce Ethernet driver, under BSDL. This driver was ported from OpenBSD by Shigeaki Tagashira <shigeaki@se.hiroshima-u.ac.jp> and posted at http://www.se.hiroshima-u.ac.jp/~shigeaki/software/freebsd-nfe.html It was additionally cleaned up by me. It is still a work-in-progress and thus is purposefully not in GENERIC. And it conflicts with nve(4), so only one should be loaded.
2006-06-26 23:41:07 +00:00
mtx_init(&sc->nfe_mtx, device_get_nameunit(dev), MTX_NETWORK_LOCK,
Bring overhauled nfe(4) into tree. o s/printf/device_printf/g o Nuke OpenBSDism. o Nuke NetBSD/OpenBSD specific DMA sync operations.(we don't have a way to sync a single descriptor within a DMA map.) o Remove recursive mutex. o bus_dma(9) clean up. o 40bit DMA address support. o Add protection for Rx map load failure. o Fix a long standing bug for watchdog timeout. [1] o Add additional protections, missing Tx completion interrupt, losing start Tx command, for watchdog timeout. o Switch to taskqueue(9) API to handle interrupts. o Use our own timer for watchdog instead of if_watchdog/if_timer interface. o Advertise VLAN header length/capability correctly to upper layer. o Remove excessive kernel stack consumption in nfe_encap(). o Handle highly fragmented mbuf chains correctly. o Enable etherenet address reprogramming with ifconfig(8). o Add ALTQ/TSO, MSI/MSIX support. o Increased Rx ring to 256 descriptors from 128. o Align Tx/Rx descriptor ring on sizeof(struct nfe_desc64) boundary. o Remove alignment restrictions on Tx/Rx buffers. o Rewritten jumbo frame support code. o Add support for hardware assistend VLAN tag insertion/stripping. o Add support for Tx/Rx flow control based on patches from Peer Chen. [2] o Add a routine that detects whether ethernet address swap routines is required. [3] o Add a workaround that take MAC/PHY out of power down mode. o Add suspend/resume support. o style(9) and code clean up. Special thanks to Shigeaki Tagashira, the original porter of nfe(4), who submitted lots of patches, performed uncountable number of regression tests and maintained nfe(4) for a long time. Without his enthusiastic help and support I could never have completed this overhauling task. The only weak point of nfe(4) compared to nve(4) is instability of manual half-duplex media selection on certain hardwares(auto sensing media type should work for all cases, though). This was a long standing bug of nfe(4) and I still have no idea why it doesn't work on some hardwares. Obtained from: OpenBSD [1] Submitted by: Peer Chen < pchen at nvidia dot com > [2], [3] Reviewed by: Shigeaki Tagashira < shigeaki AT se DOT hiroshima-u DOT ac DOT jp > Tested by: Shigeaki Tagashira, current Discussed with: current Silence from: obrien
2007-06-12 02:16:02 +00:00
MTX_DEF);
Add a pure open source nForce Ethernet driver, under BSDL. This driver was ported from OpenBSD by Shigeaki Tagashira <shigeaki@se.hiroshima-u.ac.jp> and posted at http://www.se.hiroshima-u.ac.jp/~shigeaki/software/freebsd-nfe.html It was additionally cleaned up by me. It is still a work-in-progress and thus is purposefully not in GENERIC. And it conflicts with nve(4), so only one should be loaded.
2006-06-26 23:41:07 +00:00
callout_init_mtx(&sc->nfe_stat_ch, &sc->nfe_mtx, 0);
pci_enable_busmaster(dev);
style(9) cleanup.
2006-10-19 10:01:26 +00:00
Bring overhauled nfe(4) into tree. o s/printf/device_printf/g o Nuke OpenBSDism. o Nuke NetBSD/OpenBSD specific DMA sync operations.(we don't have a way to sync a single descriptor within a DMA map.) o Remove recursive mutex. o bus_dma(9) clean up. o 40bit DMA address support. o Add protection for Rx map load failure. o Fix a long standing bug for watchdog timeout. [1] o Add additional protections, missing Tx completion interrupt, losing start Tx command, for watchdog timeout. o Switch to taskqueue(9) API to handle interrupts. o Use our own timer for watchdog instead of if_watchdog/if_timer interface. o Advertise VLAN header length/capability correctly to upper layer. o Remove excessive kernel stack consumption in nfe_encap(). o Handle highly fragmented mbuf chains correctly. o Enable etherenet address reprogramming with ifconfig(8). o Add ALTQ/TSO, MSI/MSIX support. o Increased Rx ring to 256 descriptors from 128. o Align Tx/Rx descriptor ring on sizeof(struct nfe_desc64) boundary. o Remove alignment restrictions on Tx/Rx buffers. o Rewritten jumbo frame support code. o Add support for hardware assistend VLAN tag insertion/stripping. o Add support for Tx/Rx flow control based on patches from Peer Chen. [2] o Add a routine that detects whether ethernet address swap routines is required. [3] o Add a workaround that take MAC/PHY out of power down mode. o Add suspend/resume support. o style(9) and code clean up. Special thanks to Shigeaki Tagashira, the original porter of nfe(4), who submitted lots of patches, performed uncountable number of regression tests and maintained nfe(4) for a long time. Without his enthusiastic help and support I could never have completed this overhauling task. The only weak point of nfe(4) compared to nve(4) is instability of manual half-duplex media selection on certain hardwares(auto sensing media type should work for all cases, though). This was a long standing bug of nfe(4) and I still have no idea why it doesn't work on some hardwares. Obtained from: OpenBSD [1] Submitted by: Peer Chen < pchen at nvidia dot com > [2], [3] Reviewed by: Shigeaki Tagashira < shigeaki AT se DOT hiroshima-u DOT ac DOT jp > Tested by: Shigeaki Tagashira, current Discussed with: current Silence from: obrien
2007-06-12 02:16:02 +00:00
rid = PCIR_BAR(0);
sc->nfe_res[0] = bus_alloc_resource_any(dev, SYS_RES_MEMORY, &rid,
RF_ACTIVE);
if (sc->nfe_res[0] == NULL) {
device_printf(dev, "couldn't map memory resources\n");
mtx_destroy(&sc->nfe_mtx);
return (ENXIO);
Import the nfe (nForce Ethernet) driver from OpenBSD. This is a pure open source NIC driver with a BSD license. These bits were taken from the OpenBSD public CVS repository on 1-May-06.
2006-06-26 17:25:22 +00:00
}
Do a sweep of the tree replacing calls to pci_find_extcap() with calls to pci_find_cap() instead.
2011-03-23 13:10:15 +00:00
if (pci_find_cap(dev, PCIY_EXPRESS, &reg) == 0) {
Bring overhauled nfe(4) into tree. o s/printf/device_printf/g o Nuke OpenBSDism. o Nuke NetBSD/OpenBSD specific DMA sync operations.(we don't have a way to sync a single descriptor within a DMA map.) o Remove recursive mutex. o bus_dma(9) clean up. o 40bit DMA address support. o Add protection for Rx map load failure. o Fix a long standing bug for watchdog timeout. [1] o Add additional protections, missing Tx completion interrupt, losing start Tx command, for watchdog timeout. o Switch to taskqueue(9) API to handle interrupts. o Use our own timer for watchdog instead of if_watchdog/if_timer interface. o Advertise VLAN header length/capability correctly to upper layer. o Remove excessive kernel stack consumption in nfe_encap(). o Handle highly fragmented mbuf chains correctly. o Enable etherenet address reprogramming with ifconfig(8). o Add ALTQ/TSO, MSI/MSIX support. o Increased Rx ring to 256 descriptors from 128. o Align Tx/Rx descriptor ring on sizeof(struct nfe_desc64) boundary. o Remove alignment restrictions on Tx/Rx buffers. o Rewritten jumbo frame support code. o Add support for hardware assistend VLAN tag insertion/stripping. o Add support for Tx/Rx flow control based on patches from Peer Chen. [2] o Add a routine that detects whether ethernet address swap routines is required. [3] o Add a workaround that take MAC/PHY out of power down mode. o Add suspend/resume support. o style(9) and code clean up. Special thanks to Shigeaki Tagashira, the original porter of nfe(4), who submitted lots of patches, performed uncountable number of regression tests and maintained nfe(4) for a long time. Without his enthusiastic help and support I could never have completed this overhauling task. The only weak point of nfe(4) compared to nve(4) is instability of manual half-duplex media selection on certain hardwares(auto sensing media type should work for all cases, though). This was a long standing bug of nfe(4) and I still have no idea why it doesn't work on some hardwares. Obtained from: OpenBSD [1] Submitted by: Peer Chen < pchen at nvidia dot com > [2], [3] Reviewed by: Shigeaki Tagashira < shigeaki AT se DOT hiroshima-u DOT ac DOT jp > Tested by: Shigeaki Tagashira, current Discussed with: current Silence from: obrien
2007-06-12 02:16:02 +00:00
uint16_t v, width;
v = pci_read_config(dev, reg + 0x08, 2);
/* Change max. read request size to 4096. */
v &= ~(7 << 12);
v |= (5 << 12);
pci_write_config(dev, reg + 0x08, v, 2);
v = pci_read_config(dev, reg + 0x0c, 2);
/* link capability */
v = (v >> 4) & 0x0f;
width = pci_read_config(dev, reg + 0x12, 2);
/* negotiated link width */
width = (width >> 4) & 0x3f;
if (v != width)
device_printf(sc->nfe_dev,
"warning, negotiated width of link(x%d) != "
"max. width of link(x%d)\n", width, v);
}
Add a pure open source nForce Ethernet driver, under BSDL. This driver was ported from OpenBSD by Shigeaki Tagashira <shigeaki@se.hiroshima-u.ac.jp> and posted at http://www.se.hiroshima-u.ac.jp/~shigeaki/software/freebsd-nfe.html It was additionally cleaned up by me. It is still a work-in-progress and thus is purposefully not in GENERIC. And it conflicts with nve(4), so only one should be loaded.
2006-06-26 23:41:07 +00:00
P5N32-SLI PREMIUM from ASUSTeK is known to have MSI/MSI-X issue such that nfe(4) does not work with MSI-X. When MSI-X support was introduced, I remember MCP55 controller worked without problems so the issue could be either PCI bridge or BIOS issue. But I also noticed snd_hda(4) disabled MSI on all MCP55 chipset so I'm still not sure this is generic issue of MCP55 chipset. If this was PCI bridge issue we would have added it to a system wide black-list table but it's not clear to me at this moment whether it was caused by either broken BIOS or silicon bug of MCP55 chipset. To workaround the issue, maintain a MSI/MSI-X black-list table in driver and lookup base board manufacturer and product name from the table before attempting to use MSI-X. If driver find an matching entry, nfe(4) will not use MSI/MSI-X and fall back on traditional INTx mode. This approach should be the last resort since it relies on smbios and if another instance of MSI/MSI-X breakage is reported with different maker/product, we may have to get the PCI bridge black-listed instead of adding an new entry. PR: kern/152150
2010-11-14 23:37:43 +00:00
if (nfe_can_use_msix(sc) == 0) {
device_printf(sc->nfe_dev,
"MSI/MSI-X capability black-listed, will use INTx\n");
msix_disable = 1;
msi_disable = 1;
}
Add a pure open source nForce Ethernet driver, under BSDL. This driver was ported from OpenBSD by Shigeaki Tagashira <shigeaki@se.hiroshima-u.ac.jp> and posted at http://www.se.hiroshima-u.ac.jp/~shigeaki/software/freebsd-nfe.html It was additionally cleaned up by me. It is still a work-in-progress and thus is purposefully not in GENERIC. And it conflicts with nve(4), so only one should be loaded.
2006-06-26 23:41:07 +00:00
/* Allocate interrupt */
Bring overhauled nfe(4) into tree. o s/printf/device_printf/g o Nuke OpenBSDism. o Nuke NetBSD/OpenBSD specific DMA sync operations.(we don't have a way to sync a single descriptor within a DMA map.) o Remove recursive mutex. o bus_dma(9) clean up. o 40bit DMA address support. o Add protection for Rx map load failure. o Fix a long standing bug for watchdog timeout. [1] o Add additional protections, missing Tx completion interrupt, losing start Tx command, for watchdog timeout. o Switch to taskqueue(9) API to handle interrupts. o Use our own timer for watchdog instead of if_watchdog/if_timer interface. o Advertise VLAN header length/capability correctly to upper layer. o Remove excessive kernel stack consumption in nfe_encap(). o Handle highly fragmented mbuf chains correctly. o Enable etherenet address reprogramming with ifconfig(8). o Add ALTQ/TSO, MSI/MSIX support. o Increased Rx ring to 256 descriptors from 128. o Align Tx/Rx descriptor ring on sizeof(struct nfe_desc64) boundary. o Remove alignment restrictions on Tx/Rx buffers. o Rewritten jumbo frame support code. o Add support for hardware assistend VLAN tag insertion/stripping. o Add support for Tx/Rx flow control based on patches from Peer Chen. [2] o Add a routine that detects whether ethernet address swap routines is required. [3] o Add a workaround that take MAC/PHY out of power down mode. o Add suspend/resume support. o style(9) and code clean up. Special thanks to Shigeaki Tagashira, the original porter of nfe(4), who submitted lots of patches, performed uncountable number of regression tests and maintained nfe(4) for a long time. Without his enthusiastic help and support I could never have completed this overhauling task. The only weak point of nfe(4) compared to nve(4) is instability of manual half-duplex media selection on certain hardwares(auto sensing media type should work for all cases, though). This was a long standing bug of nfe(4) and I still have no idea why it doesn't work on some hardwares. Obtained from: OpenBSD [1] Submitted by: Peer Chen < pchen at nvidia dot com > [2], [3] Reviewed by: Shigeaki Tagashira < shigeaki AT se DOT hiroshima-u DOT ac DOT jp > Tested by: Shigeaki Tagashira, current Discussed with: current Silence from: obrien
2007-06-12 02:16:02 +00:00
if (msix_disable == 0 || msi_disable == 0) {
if (msix_disable == 0 &&
(msic = pci_msix_count(dev)) == NFE_MSI_MESSAGES)
nfe_alloc_msix(sc, msic);
if (msi_disable == 0 && sc->nfe_msix == 0 &&
(msic = pci_msi_count(dev)) == NFE_MSI_MESSAGES &&
pci_alloc_msi(dev, &msic) == 0) {
if (msic == NFE_MSI_MESSAGES) {
if (bootverbose)
device_printf(dev,
"Using %d MSI messages\n", msic);
sc->nfe_msi = 1;
} else
pci_release_msi(dev);
}
}
Import the nfe (nForce Ethernet) driver from OpenBSD. This is a pure open source NIC driver with a BSD license. These bits were taken from the OpenBSD public CVS repository on 1-May-06.
2006-06-26 17:25:22 +00:00
Bring overhauled nfe(4) into tree. o s/printf/device_printf/g o Nuke OpenBSDism. o Nuke NetBSD/OpenBSD specific DMA sync operations.(we don't have a way to sync a single descriptor within a DMA map.) o Remove recursive mutex. o bus_dma(9) clean up. o 40bit DMA address support. o Add protection for Rx map load failure. o Fix a long standing bug for watchdog timeout. [1] o Add additional protections, missing Tx completion interrupt, losing start Tx command, for watchdog timeout. o Switch to taskqueue(9) API to handle interrupts. o Use our own timer for watchdog instead of if_watchdog/if_timer interface. o Advertise VLAN header length/capability correctly to upper layer. o Remove excessive kernel stack consumption in nfe_encap(). o Handle highly fragmented mbuf chains correctly. o Enable etherenet address reprogramming with ifconfig(8). o Add ALTQ/TSO, MSI/MSIX support. o Increased Rx ring to 256 descriptors from 128. o Align Tx/Rx descriptor ring on sizeof(struct nfe_desc64) boundary. o Remove alignment restrictions on Tx/Rx buffers. o Rewritten jumbo frame support code. o Add support for hardware assistend VLAN tag insertion/stripping. o Add support for Tx/Rx flow control based on patches from Peer Chen. [2] o Add a routine that detects whether ethernet address swap routines is required. [3] o Add a workaround that take MAC/PHY out of power down mode. o Add suspend/resume support. o style(9) and code clean up. Special thanks to Shigeaki Tagashira, the original porter of nfe(4), who submitted lots of patches, performed uncountable number of regression tests and maintained nfe(4) for a long time. Without his enthusiastic help and support I could never have completed this overhauling task. The only weak point of nfe(4) compared to nve(4) is instability of manual half-duplex media selection on certain hardwares(auto sensing media type should work for all cases, though). This was a long standing bug of nfe(4) and I still have no idea why it doesn't work on some hardwares. Obtained from: OpenBSD [1] Submitted by: Peer Chen < pchen at nvidia dot com > [2], [3] Reviewed by: Shigeaki Tagashira < shigeaki AT se DOT hiroshima-u DOT ac DOT jp > Tested by: Shigeaki Tagashira, current Discussed with: current Silence from: obrien
2007-06-12 02:16:02 +00:00
if (sc->nfe_msix == 0 && sc->nfe_msi == 0) {
rid = 0;
sc->nfe_irq[0] = bus_alloc_resource_any(dev, SYS_RES_IRQ, &rid,
RF_SHAREABLE | RF_ACTIVE);
if (sc->nfe_irq[0] == NULL) {
device_printf(dev, "couldn't allocate IRQ resources\n");
error = ENXIO;
goto fail;
}
} else {
for (i = 0, rid = 1; i < NFE_MSI_MESSAGES; i++, rid++) {
sc->nfe_irq[i] = bus_alloc_resource_any(dev,
SYS_RES_IRQ, &rid, RF_ACTIVE);
if (sc->nfe_irq[i] == NULL) {
device_printf(dev,
"couldn't allocate IRQ resources for "
"message %d\n", rid);
error = ENXIO;
goto fail;
}
}
/* Map interrupts to vector 0. */
if (sc->nfe_msix != 0) {
NFE_WRITE(sc, NFE_MSIX_MAP0, 0);
NFE_WRITE(sc, NFE_MSIX_MAP1, 0);
} else if (sc->nfe_msi != 0) {
NFE_WRITE(sc, NFE_MSI_MAP0, 0);
NFE_WRITE(sc, NFE_MSI_MAP1, 0);
}
Add a pure open source nForce Ethernet driver, under BSDL. This driver was ported from OpenBSD by Shigeaki Tagashira <shigeaki@se.hiroshima-u.ac.jp> and posted at http://www.se.hiroshima-u.ac.jp/~shigeaki/software/freebsd-nfe.html It was additionally cleaned up by me. It is still a work-in-progress and thus is purposefully not in GENERIC. And it conflicts with nve(4), so only one should be loaded.
2006-06-26 23:41:07 +00:00
}
Import the nfe (nForce Ethernet) driver from OpenBSD. This is a pure open source NIC driver with a BSD license. These bits were taken from the OpenBSD public CVS repository on 1-May-06.
2006-06-26 17:25:22 +00:00
Bring overhauled nfe(4) into tree. o s/printf/device_printf/g o Nuke OpenBSDism. o Nuke NetBSD/OpenBSD specific DMA sync operations.(we don't have a way to sync a single descriptor within a DMA map.) o Remove recursive mutex. o bus_dma(9) clean up. o 40bit DMA address support. o Add protection for Rx map load failure. o Fix a long standing bug for watchdog timeout. [1] o Add additional protections, missing Tx completion interrupt, losing start Tx command, for watchdog timeout. o Switch to taskqueue(9) API to handle interrupts. o Use our own timer for watchdog instead of if_watchdog/if_timer interface. o Advertise VLAN header length/capability correctly to upper layer. o Remove excessive kernel stack consumption in nfe_encap(). o Handle highly fragmented mbuf chains correctly. o Enable etherenet address reprogramming with ifconfig(8). o Add ALTQ/TSO, MSI/MSIX support. o Increased Rx ring to 256 descriptors from 128. o Align Tx/Rx descriptor ring on sizeof(struct nfe_desc64) boundary. o Remove alignment restrictions on Tx/Rx buffers. o Rewritten jumbo frame support code. o Add support for hardware assistend VLAN tag insertion/stripping. o Add support for Tx/Rx flow control based on patches from Peer Chen. [2] o Add a routine that detects whether ethernet address swap routines is required. [3] o Add a workaround that take MAC/PHY out of power down mode. o Add suspend/resume support. o style(9) and code clean up. Special thanks to Shigeaki Tagashira, the original porter of nfe(4), who submitted lots of patches, performed uncountable number of regression tests and maintained nfe(4) for a long time. Without his enthusiastic help and support I could never have completed this overhauling task. The only weak point of nfe(4) compared to nve(4) is instability of manual half-duplex media selection on certain hardwares(auto sensing media type should work for all cases, though). This was a long standing bug of nfe(4) and I still have no idea why it doesn't work on some hardwares. Obtained from: OpenBSD [1] Submitted by: Peer Chen < pchen at nvidia dot com > [2], [3] Reviewed by: Shigeaki Tagashira < shigeaki AT se DOT hiroshima-u DOT ac DOT jp > Tested by: Shigeaki Tagashira, current Discussed with: current Silence from: obrien
2007-06-12 02:16:02 +00:00
/* Set IRQ status/mask register. */
sc->nfe_irq_status = NFE_IRQ_STATUS;
sc->nfe_irq_mask = NFE_IRQ_MASK;
sc->nfe_intrs = NFE_IRQ_WANTED;
sc->nfe_nointrs = 0;
if (sc->nfe_msix != 0) {
sc->nfe_irq_status = NFE_MSIX_IRQ_STATUS;
sc->nfe_nointrs = NFE_IRQ_WANTED;
} else if (sc->nfe_msi != 0) {
sc->nfe_irq_mask = NFE_MSI_IRQ_MASK;
sc->nfe_intrs = NFE_MSI_VECTOR_0_ENABLED;
}
Import the nfe (nForce Ethernet) driver from OpenBSD. This is a pure open source NIC driver with a BSD license. These bits were taken from the OpenBSD public CVS repository on 1-May-06.
2006-06-26 17:25:22 +00:00
Bring overhauled nfe(4) into tree. o s/printf/device_printf/g o Nuke OpenBSDism. o Nuke NetBSD/OpenBSD specific DMA sync operations.(we don't have a way to sync a single descriptor within a DMA map.) o Remove recursive mutex. o bus_dma(9) clean up. o 40bit DMA address support. o Add protection for Rx map load failure. o Fix a long standing bug for watchdog timeout. [1] o Add additional protections, missing Tx completion interrupt, losing start Tx command, for watchdog timeout. o Switch to taskqueue(9) API to handle interrupts. o Use our own timer for watchdog instead of if_watchdog/if_timer interface. o Advertise VLAN header length/capability correctly to upper layer. o Remove excessive kernel stack consumption in nfe_encap(). o Handle highly fragmented mbuf chains correctly. o Enable etherenet address reprogramming with ifconfig(8). o Add ALTQ/TSO, MSI/MSIX support. o Increased Rx ring to 256 descriptors from 128. o Align Tx/Rx descriptor ring on sizeof(struct nfe_desc64) boundary. o Remove alignment restrictions on Tx/Rx buffers. o Rewritten jumbo frame support code. o Add support for hardware assistend VLAN tag insertion/stripping. o Add support for Tx/Rx flow control based on patches from Peer Chen. [2] o Add a routine that detects whether ethernet address swap routines is required. [3] o Add a workaround that take MAC/PHY out of power down mode. o Add suspend/resume support. o style(9) and code clean up. Special thanks to Shigeaki Tagashira, the original porter of nfe(4), who submitted lots of patches, performed uncountable number of regression tests and maintained nfe(4) for a long time. Without his enthusiastic help and support I could never have completed this overhauling task. The only weak point of nfe(4) compared to nve(4) is instability of manual half-duplex media selection on certain hardwares(auto sensing media type should work for all cases, though). This was a long standing bug of nfe(4) and I still have no idea why it doesn't work on some hardwares. Obtained from: OpenBSD [1] Submitted by: Peer Chen < pchen at nvidia dot com > [2], [3] Reviewed by: Shigeaki Tagashira < shigeaki AT se DOT hiroshima-u DOT ac DOT jp > Tested by: Shigeaki Tagashira, current Discussed with: current Silence from: obrien
2007-06-12 02:16:02 +00:00
sc->nfe_devid = pci_get_device(dev);
sc->nfe_revid = pci_get_revid(dev);
Add a pure open source nForce Ethernet driver, under BSDL. This driver was ported from OpenBSD by Shigeaki Tagashira <shigeaki@se.hiroshima-u.ac.jp> and posted at http://www.se.hiroshima-u.ac.jp/~shigeaki/software/freebsd-nfe.html It was additionally cleaned up by me. It is still a work-in-progress and thus is purposefully not in GENERIC. And it conflicts with nve(4), so only one should be loaded.
2006-06-26 23:41:07 +00:00
sc->nfe_flags = 0;
Bring overhauled nfe(4) into tree. o s/printf/device_printf/g o Nuke OpenBSDism. o Nuke NetBSD/OpenBSD specific DMA sync operations.(we don't have a way to sync a single descriptor within a DMA map.) o Remove recursive mutex. o bus_dma(9) clean up. o 40bit DMA address support. o Add protection for Rx map load failure. o Fix a long standing bug for watchdog timeout. [1] o Add additional protections, missing Tx completion interrupt, losing start Tx command, for watchdog timeout. o Switch to taskqueue(9) API to handle interrupts. o Use our own timer for watchdog instead of if_watchdog/if_timer interface. o Advertise VLAN header length/capability correctly to upper layer. o Remove excessive kernel stack consumption in nfe_encap(). o Handle highly fragmented mbuf chains correctly. o Enable etherenet address reprogramming with ifconfig(8). o Add ALTQ/TSO, MSI/MSIX support. o Increased Rx ring to 256 descriptors from 128. o Align Tx/Rx descriptor ring on sizeof(struct nfe_desc64) boundary. o Remove alignment restrictions on Tx/Rx buffers. o Rewritten jumbo frame support code. o Add support for hardware assistend VLAN tag insertion/stripping. o Add support for Tx/Rx flow control based on patches from Peer Chen. [2] o Add a routine that detects whether ethernet address swap routines is required. [3] o Add a workaround that take MAC/PHY out of power down mode. o Add suspend/resume support. o style(9) and code clean up. Special thanks to Shigeaki Tagashira, the original porter of nfe(4), who submitted lots of patches, performed uncountable number of regression tests and maintained nfe(4) for a long time. Without his enthusiastic help and support I could never have completed this overhauling task. The only weak point of nfe(4) compared to nve(4) is instability of manual half-duplex media selection on certain hardwares(auto sensing media type should work for all cases, though). This was a long standing bug of nfe(4) and I still have no idea why it doesn't work on some hardwares. Obtained from: OpenBSD [1] Submitted by: Peer Chen < pchen at nvidia dot com > [2], [3] Reviewed by: Shigeaki Tagashira < shigeaki AT se DOT hiroshima-u DOT ac DOT jp > Tested by: Shigeaki Tagashira, current Discussed with: current Silence from: obrien
2007-06-12 02:16:02 +00:00
switch (sc->nfe_devid) {
Import the nfe (nForce Ethernet) driver from OpenBSD. This is a pure open source NIC driver with a BSD license. These bits were taken from the OpenBSD public CVS repository on 1-May-06.
2006-06-26 17:25:22 +00:00
case PCI_PRODUCT_NVIDIA_NFORCE3_LAN2:
case PCI_PRODUCT_NVIDIA_NFORCE3_LAN3:
case PCI_PRODUCT_NVIDIA_NFORCE3_LAN4:
case PCI_PRODUCT_NVIDIA_NFORCE3_LAN5:
Add a pure open source nForce Ethernet driver, under BSDL. This driver was ported from OpenBSD by Shigeaki Tagashira <shigeaki@se.hiroshima-u.ac.jp> and posted at http://www.se.hiroshima-u.ac.jp/~shigeaki/software/freebsd-nfe.html It was additionally cleaned up by me. It is still a work-in-progress and thus is purposefully not in GENERIC. And it conflicts with nve(4), so only one should be loaded.
2006-06-26 23:41:07 +00:00
sc->nfe_flags |= NFE_JUMBO_SUP | NFE_HW_CSUM;
Import the nfe (nForce Ethernet) driver from OpenBSD. This is a pure open source NIC driver with a BSD license. These bits were taken from the OpenBSD public CVS repository on 1-May-06.
2006-06-26 17:25:22 +00:00
break;
case PCI_PRODUCT_NVIDIA_MCP51_LAN1:
case PCI_PRODUCT_NVIDIA_MCP51_LAN2:
Add hardware MAC statistics support. Register information was obtained from Linux forcedeth driver. While I'm here move creating a sysctl node for process_limit to function nfe_sysctl_node(). Tested by: "Arno J. Klaassen" < arno <at> heho dot snv dot jussieu dot fr >
2008-10-03 03:58:16 +00:00
sc->nfe_flags |= NFE_40BIT_ADDR | NFE_PWR_MGMT | NFE_MIB_V1;
Import the nfe (nForce Ethernet) driver from OpenBSD. This is a pure open source NIC driver with a BSD license. These bits were taken from the OpenBSD public CVS repository on 1-May-06.
2006-06-26 17:25:22 +00:00
break;
case PCI_PRODUCT_NVIDIA_CK804_LAN1:
case PCI_PRODUCT_NVIDIA_CK804_LAN2:
case PCI_PRODUCT_NVIDIA_MCP04_LAN1:
case PCI_PRODUCT_NVIDIA_MCP04_LAN2:
Add hardware MAC statistics support. Register information was obtained from Linux forcedeth driver. While I'm here move creating a sysctl node for process_limit to function nfe_sysctl_node(). Tested by: "Arno J. Klaassen" < arno <at> heho dot snv dot jussieu dot fr >
2008-10-03 03:58:16 +00:00
sc->nfe_flags |= NFE_JUMBO_SUP | NFE_40BIT_ADDR | NFE_HW_CSUM |
NFE_MIB_V1;
Import the nfe (nForce Ethernet) driver from OpenBSD. This is a pure open source NIC driver with a BSD license. These bits were taken from the OpenBSD public CVS repository on 1-May-06.
2006-06-26 17:25:22 +00:00
break;
case PCI_PRODUCT_NVIDIA_MCP55_LAN1:
case PCI_PRODUCT_NVIDIA_MCP55_LAN2:
style(9) cleanup.
2006-10-19 10:01:26 +00:00
sc->nfe_flags |= NFE_JUMBO_SUP | NFE_40BIT_ADDR | NFE_HW_CSUM |
Add hardware MAC statistics support. Register information was obtained from Linux forcedeth driver. While I'm here move creating a sysctl node for process_limit to function nfe_sysctl_node(). Tested by: "Arno J. Klaassen" < arno <at> heho dot snv dot jussieu dot fr >
2008-10-03 03:58:16 +00:00
NFE_HW_VLAN | NFE_PWR_MGMT | NFE_TX_FLOW_CTRL | NFE_MIB_V2;
Import the nfe (nForce Ethernet) driver from OpenBSD. This is a pure open source NIC driver with a BSD license. These bits were taken from the OpenBSD public CVS repository on 1-May-06.
2006-06-26 17:25:22 +00:00
break;
Bring overhauled nfe(4) into tree. o s/printf/device_printf/g o Nuke OpenBSDism. o Nuke NetBSD/OpenBSD specific DMA sync operations.(we don't have a way to sync a single descriptor within a DMA map.) o Remove recursive mutex. o bus_dma(9) clean up. o 40bit DMA address support. o Add protection for Rx map load failure. o Fix a long standing bug for watchdog timeout. [1] o Add additional protections, missing Tx completion interrupt, losing start Tx command, for watchdog timeout. o Switch to taskqueue(9) API to handle interrupts. o Use our own timer for watchdog instead of if_watchdog/if_timer interface. o Advertise VLAN header length/capability correctly to upper layer. o Remove excessive kernel stack consumption in nfe_encap(). o Handle highly fragmented mbuf chains correctly. o Enable etherenet address reprogramming with ifconfig(8). o Add ALTQ/TSO, MSI/MSIX support. o Increased Rx ring to 256 descriptors from 128. o Align Tx/Rx descriptor ring on sizeof(struct nfe_desc64) boundary. o Remove alignment restrictions on Tx/Rx buffers. o Rewritten jumbo frame support code. o Add support for hardware assistend VLAN tag insertion/stripping. o Add support for Tx/Rx flow control based on patches from Peer Chen. [2] o Add a routine that detects whether ethernet address swap routines is required. [3] o Add a workaround that take MAC/PHY out of power down mode. o Add suspend/resume support. o style(9) and code clean up. Special thanks to Shigeaki Tagashira, the original porter of nfe(4), who submitted lots of patches, performed uncountable number of regression tests and maintained nfe(4) for a long time. Without his enthusiastic help and support I could never have completed this overhauling task. The only weak point of nfe(4) compared to nve(4) is instability of manual half-duplex media selection on certain hardwares(auto sensing media type should work for all cases, though). This was a long standing bug of nfe(4) and I still have no idea why it doesn't work on some hardwares. Obtained from: OpenBSD [1] Submitted by: Peer Chen < pchen at nvidia dot com > [2], [3] Reviewed by: Shigeaki Tagashira < shigeaki AT se DOT hiroshima-u DOT ac DOT jp > Tested by: Shigeaki Tagashira, current Discussed with: current Silence from: obrien
2007-06-12 02:16:02 +00:00
Add support for MCP61 and MCP65 adaptors. Submitted by: Peer Chen at nVIDIA
2006-09-11 07:31:55 +00:00
case PCI_PRODUCT_NVIDIA_MCP61_LAN1:
case PCI_PRODUCT_NVIDIA_MCP61_LAN2:
case PCI_PRODUCT_NVIDIA_MCP61_LAN3:
case PCI_PRODUCT_NVIDIA_MCP61_LAN4:
Bring overhauled nfe(4) into tree. o s/printf/device_printf/g o Nuke OpenBSDism. o Nuke NetBSD/OpenBSD specific DMA sync operations.(we don't have a way to sync a single descriptor within a DMA map.) o Remove recursive mutex. o bus_dma(9) clean up. o 40bit DMA address support. o Add protection for Rx map load failure. o Fix a long standing bug for watchdog timeout. [1] o Add additional protections, missing Tx completion interrupt, losing start Tx command, for watchdog timeout. o Switch to taskqueue(9) API to handle interrupts. o Use our own timer for watchdog instead of if_watchdog/if_timer interface. o Advertise VLAN header length/capability correctly to upper layer. o Remove excessive kernel stack consumption in nfe_encap(). o Handle highly fragmented mbuf chains correctly. o Enable etherenet address reprogramming with ifconfig(8). o Add ALTQ/TSO, MSI/MSIX support. o Increased Rx ring to 256 descriptors from 128. o Align Tx/Rx descriptor ring on sizeof(struct nfe_desc64) boundary. o Remove alignment restrictions on Tx/Rx buffers. o Rewritten jumbo frame support code. o Add support for hardware assistend VLAN tag insertion/stripping. o Add support for Tx/Rx flow control based on patches from Peer Chen. [2] o Add a routine that detects whether ethernet address swap routines is required. [3] o Add a workaround that take MAC/PHY out of power down mode. o Add suspend/resume support. o style(9) and code clean up. Special thanks to Shigeaki Tagashira, the original porter of nfe(4), who submitted lots of patches, performed uncountable number of regression tests and maintained nfe(4) for a long time. Without his enthusiastic help and support I could never have completed this overhauling task. The only weak point of nfe(4) compared to nve(4) is instability of manual half-duplex media selection on certain hardwares(auto sensing media type should work for all cases, though). This was a long standing bug of nfe(4) and I still have no idea why it doesn't work on some hardwares. Obtained from: OpenBSD [1] Submitted by: Peer Chen < pchen at nvidia dot com > [2], [3] Reviewed by: Shigeaki Tagashira < shigeaki AT se DOT hiroshima-u DOT ac DOT jp > Tested by: Shigeaki Tagashira, current Discussed with: current Silence from: obrien
2007-06-12 02:16:02 +00:00
case PCI_PRODUCT_NVIDIA_MCP67_LAN1:
case PCI_PRODUCT_NVIDIA_MCP67_LAN2:
case PCI_PRODUCT_NVIDIA_MCP67_LAN3:
case PCI_PRODUCT_NVIDIA_MCP67_LAN4:
Add support for MCP73 chips. Tested by: "Daan Vreeken [PA4DAN]" ( Danovitsch AT vitsch DOT net ) MFC after: 1 week
2008-04-10 01:25:09 +00:00
case PCI_PRODUCT_NVIDIA_MCP73_LAN1:
case PCI_PRODUCT_NVIDIA_MCP73_LAN2:
case PCI_PRODUCT_NVIDIA_MCP73_LAN3:
case PCI_PRODUCT_NVIDIA_MCP73_LAN4:
Bring overhauled nfe(4) into tree. o s/printf/device_printf/g o Nuke OpenBSDism. o Nuke NetBSD/OpenBSD specific DMA sync operations.(we don't have a way to sync a single descriptor within a DMA map.) o Remove recursive mutex. o bus_dma(9) clean up. o 40bit DMA address support. o Add protection for Rx map load failure. o Fix a long standing bug for watchdog timeout. [1] o Add additional protections, missing Tx completion interrupt, losing start Tx command, for watchdog timeout. o Switch to taskqueue(9) API to handle interrupts. o Use our own timer for watchdog instead of if_watchdog/if_timer interface. o Advertise VLAN header length/capability correctly to upper layer. o Remove excessive kernel stack consumption in nfe_encap(). o Handle highly fragmented mbuf chains correctly. o Enable etherenet address reprogramming with ifconfig(8). o Add ALTQ/TSO, MSI/MSIX support. o Increased Rx ring to 256 descriptors from 128. o Align Tx/Rx descriptor ring on sizeof(struct nfe_desc64) boundary. o Remove alignment restrictions on Tx/Rx buffers. o Rewritten jumbo frame support code. o Add support for hardware assistend VLAN tag insertion/stripping. o Add support for Tx/Rx flow control based on patches from Peer Chen. [2] o Add a routine that detects whether ethernet address swap routines is required. [3] o Add a workaround that take MAC/PHY out of power down mode. o Add suspend/resume support. o style(9) and code clean up. Special thanks to Shigeaki Tagashira, the original porter of nfe(4), who submitted lots of patches, performed uncountable number of regression tests and maintained nfe(4) for a long time. Without his enthusiastic help and support I could never have completed this overhauling task. The only weak point of nfe(4) compared to nve(4) is instability of manual half-duplex media selection on certain hardwares(auto sensing media type should work for all cases, though). This was a long standing bug of nfe(4) and I still have no idea why it doesn't work on some hardwares. Obtained from: OpenBSD [1] Submitted by: Peer Chen < pchen at nvidia dot com > [2], [3] Reviewed by: Shigeaki Tagashira < shigeaki AT se DOT hiroshima-u DOT ac DOT jp > Tested by: Shigeaki Tagashira, current Discussed with: current Silence from: obrien
2007-06-12 02:16:02 +00:00
sc->nfe_flags |= NFE_40BIT_ADDR | NFE_PWR_MGMT |
Add hardware MAC statistics support. Register information was obtained from Linux forcedeth driver. While I'm here move creating a sysctl node for process_limit to function nfe_sysctl_node(). Tested by: "Arno J. Klaassen" < arno <at> heho dot snv dot jussieu dot fr >
2008-10-03 03:58:16 +00:00
NFE_CORRECT_MACADDR | NFE_TX_FLOW_CTRL | NFE_MIB_V2;
Add support for MCP61 and MCP65 adaptors. Submitted by: Peer Chen at nVIDIA
2006-09-11 07:31:55 +00:00
break;
Add device ids for MCP77/79 and set appropriate support flags. Obtained from: OpenBSD PR: kern/127529
2008-10-01 00:17:54 +00:00
case PCI_PRODUCT_NVIDIA_MCP77_LAN1:
case PCI_PRODUCT_NVIDIA_MCP77_LAN2:
case PCI_PRODUCT_NVIDIA_MCP77_LAN3:
case PCI_PRODUCT_NVIDIA_MCP77_LAN4:
/* XXX flow control */
sc->nfe_flags |= NFE_40BIT_ADDR | NFE_HW_CSUM | NFE_PWR_MGMT |
Add hardware MAC statistics support. Register information was obtained from Linux forcedeth driver. While I'm here move creating a sysctl node for process_limit to function nfe_sysctl_node(). Tested by: "Arno J. Klaassen" < arno <at> heho dot snv dot jussieu dot fr >
2008-10-03 03:58:16 +00:00
NFE_CORRECT_MACADDR | NFE_MIB_V3;
Add device ids for MCP77/79 and set appropriate support flags. Obtained from: OpenBSD PR: kern/127529
2008-10-01 00:17:54 +00:00
break;
case PCI_PRODUCT_NVIDIA_MCP79_LAN1:
case PCI_PRODUCT_NVIDIA_MCP79_LAN2:
case PCI_PRODUCT_NVIDIA_MCP79_LAN3:
case PCI_PRODUCT_NVIDIA_MCP79_LAN4:
/* XXX flow control */
sc->nfe_flags |= NFE_JUMBO_SUP | NFE_40BIT_ADDR | NFE_HW_CSUM |
Add hardware MAC statistics support. Register information was obtained from Linux forcedeth driver. While I'm here move creating a sysctl node for process_limit to function nfe_sysctl_node(). Tested by: "Arno J. Klaassen" < arno <at> heho dot snv dot jussieu dot fr >
2008-10-03 03:58:16 +00:00
NFE_PWR_MGMT | NFE_CORRECT_MACADDR | NFE_MIB_V3;
Add device ids for MCP77/79 and set appropriate support flags. Obtained from: OpenBSD PR: kern/127529
2008-10-01 00:17:54 +00:00
break;
Add support for MCP61 and MCP65 adaptors. Submitted by: Peer Chen at nVIDIA
2006-09-11 07:31:55 +00:00
case PCI_PRODUCT_NVIDIA_MCP65_LAN1:
case PCI_PRODUCT_NVIDIA_MCP65_LAN2:
case PCI_PRODUCT_NVIDIA_MCP65_LAN3:
case PCI_PRODUCT_NVIDIA_MCP65_LAN4:
Bring overhauled nfe(4) into tree. o s/printf/device_printf/g o Nuke OpenBSDism. o Nuke NetBSD/OpenBSD specific DMA sync operations.(we don't have a way to sync a single descriptor within a DMA map.) o Remove recursive mutex. o bus_dma(9) clean up. o 40bit DMA address support. o Add protection for Rx map load failure. o Fix a long standing bug for watchdog timeout. [1] o Add additional protections, missing Tx completion interrupt, losing start Tx command, for watchdog timeout. o Switch to taskqueue(9) API to handle interrupts. o Use our own timer for watchdog instead of if_watchdog/if_timer interface. o Advertise VLAN header length/capability correctly to upper layer. o Remove excessive kernel stack consumption in nfe_encap(). o Handle highly fragmented mbuf chains correctly. o Enable etherenet address reprogramming with ifconfig(8). o Add ALTQ/TSO, MSI/MSIX support. o Increased Rx ring to 256 descriptors from 128. o Align Tx/Rx descriptor ring on sizeof(struct nfe_desc64) boundary. o Remove alignment restrictions on Tx/Rx buffers. o Rewritten jumbo frame support code. o Add support for hardware assistend VLAN tag insertion/stripping. o Add support for Tx/Rx flow control based on patches from Peer Chen. [2] o Add a routine that detects whether ethernet address swap routines is required. [3] o Add a workaround that take MAC/PHY out of power down mode. o Add suspend/resume support. o style(9) and code clean up. Special thanks to Shigeaki Tagashira, the original porter of nfe(4), who submitted lots of patches, performed uncountable number of regression tests and maintained nfe(4) for a long time. Without his enthusiastic help and support I could never have completed this overhauling task. The only weak point of nfe(4) compared to nve(4) is instability of manual half-duplex media selection on certain hardwares(auto sensing media type should work for all cases, though). This was a long standing bug of nfe(4) and I still have no idea why it doesn't work on some hardwares. Obtained from: OpenBSD [1] Submitted by: Peer Chen < pchen at nvidia dot com > [2], [3] Reviewed by: Shigeaki Tagashira < shigeaki AT se DOT hiroshima-u DOT ac DOT jp > Tested by: Shigeaki Tagashira, current Discussed with: current Silence from: obrien
2007-06-12 02:16:02 +00:00
sc->nfe_flags |= NFE_JUMBO_SUP | NFE_40BIT_ADDR |
Add hardware MAC statistics support. Register information was obtained from Linux forcedeth driver. While I'm here move creating a sysctl node for process_limit to function nfe_sysctl_node(). Tested by: "Arno J. Klaassen" < arno <at> heho dot snv dot jussieu dot fr >
2008-10-03 03:58:16 +00:00
NFE_PWR_MGMT | NFE_CORRECT_MACADDR | NFE_TX_FLOW_CTRL |
NFE_MIB_V2;
style(9) cleanup.
2006-10-19 10:01:26 +00:00
break;
Import the nfe (nForce Ethernet) driver from OpenBSD. This is a pure open source NIC driver with a BSD license. These bits were taken from the OpenBSD public CVS repository on 1-May-06.
2006-06-26 17:25:22 +00:00
}
Bring overhauled nfe(4) into tree. o s/printf/device_printf/g o Nuke OpenBSDism. o Nuke NetBSD/OpenBSD specific DMA sync operations.(we don't have a way to sync a single descriptor within a DMA map.) o Remove recursive mutex. o bus_dma(9) clean up. o 40bit DMA address support. o Add protection for Rx map load failure. o Fix a long standing bug for watchdog timeout. [1] o Add additional protections, missing Tx completion interrupt, losing start Tx command, for watchdog timeout. o Switch to taskqueue(9) API to handle interrupts. o Use our own timer for watchdog instead of if_watchdog/if_timer interface. o Advertise VLAN header length/capability correctly to upper layer. o Remove excessive kernel stack consumption in nfe_encap(). o Handle highly fragmented mbuf chains correctly. o Enable etherenet address reprogramming with ifconfig(8). o Add ALTQ/TSO, MSI/MSIX support. o Increased Rx ring to 256 descriptors from 128. o Align Tx/Rx descriptor ring on sizeof(struct nfe_desc64) boundary. o Remove alignment restrictions on Tx/Rx buffers. o Rewritten jumbo frame support code. o Add support for hardware assistend VLAN tag insertion/stripping. o Add support for Tx/Rx flow control based on patches from Peer Chen. [2] o Add a routine that detects whether ethernet address swap routines is required. [3] o Add a workaround that take MAC/PHY out of power down mode. o Add suspend/resume support. o style(9) and code clean up. Special thanks to Shigeaki Tagashira, the original porter of nfe(4), who submitted lots of patches, performed uncountable number of regression tests and maintained nfe(4) for a long time. Without his enthusiastic help and support I could never have completed this overhauling task. The only weak point of nfe(4) compared to nve(4) is instability of manual half-duplex media selection on certain hardwares(auto sensing media type should work for all cases, though). This was a long standing bug of nfe(4) and I still have no idea why it doesn't work on some hardwares. Obtained from: OpenBSD [1] Submitted by: Peer Chen < pchen at nvidia dot com > [2], [3] Reviewed by: Shigeaki Tagashira < shigeaki AT se DOT hiroshima-u DOT ac DOT jp > Tested by: Shigeaki Tagashira, current Discussed with: current Silence from: obrien
2007-06-12 02:16:02 +00:00
nfe_power(sc);
/* Check for reversed ethernet address */
if ((NFE_READ(sc, NFE_TX_UNK) & NFE_MAC_ADDR_INORDER) != 0)
sc->nfe_flags |= NFE_CORRECT_MACADDR;
nfe_get_macaddr(sc, sc->eaddr);
Add a pure open source nForce Ethernet driver, under BSDL. This driver was ported from OpenBSD by Shigeaki Tagashira <shigeaki@se.hiroshima-u.ac.jp> and posted at http://www.se.hiroshima-u.ac.jp/~shigeaki/software/freebsd-nfe.html It was additionally cleaned up by me. It is still a work-in-progress and thus is purposefully not in GENERIC. And it conflicts with nve(4), so only one should be loaded.
2006-06-26 23:41:07 +00:00
/*
* Allocate the parent bus DMA tag appropriate for PCI.
*/
Bring overhauled nfe(4) into tree. o s/printf/device_printf/g o Nuke OpenBSDism. o Nuke NetBSD/OpenBSD specific DMA sync operations.(we don't have a way to sync a single descriptor within a DMA map.) o Remove recursive mutex. o bus_dma(9) clean up. o 40bit DMA address support. o Add protection for Rx map load failure. o Fix a long standing bug for watchdog timeout. [1] o Add additional protections, missing Tx completion interrupt, losing start Tx command, for watchdog timeout. o Switch to taskqueue(9) API to handle interrupts. o Use our own timer for watchdog instead of if_watchdog/if_timer interface. o Advertise VLAN header length/capability correctly to upper layer. o Remove excessive kernel stack consumption in nfe_encap(). o Handle highly fragmented mbuf chains correctly. o Enable etherenet address reprogramming with ifconfig(8). o Add ALTQ/TSO, MSI/MSIX support. o Increased Rx ring to 256 descriptors from 128. o Align Tx/Rx descriptor ring on sizeof(struct nfe_desc64) boundary. o Remove alignment restrictions on Tx/Rx buffers. o Rewritten jumbo frame support code. o Add support for hardware assistend VLAN tag insertion/stripping. o Add support for Tx/Rx flow control based on patches from Peer Chen. [2] o Add a routine that detects whether ethernet address swap routines is required. [3] o Add a workaround that take MAC/PHY out of power down mode. o Add suspend/resume support. o style(9) and code clean up. Special thanks to Shigeaki Tagashira, the original porter of nfe(4), who submitted lots of patches, performed uncountable number of regression tests and maintained nfe(4) for a long time. Without his enthusiastic help and support I could never have completed this overhauling task. The only weak point of nfe(4) compared to nve(4) is instability of manual half-duplex media selection on certain hardwares(auto sensing media type should work for all cases, though). This was a long standing bug of nfe(4) and I still have no idea why it doesn't work on some hardwares. Obtained from: OpenBSD [1] Submitted by: Peer Chen < pchen at nvidia dot com > [2], [3] Reviewed by: Shigeaki Tagashira < shigeaki AT se DOT hiroshima-u DOT ac DOT jp > Tested by: Shigeaki Tagashira, current Discussed with: current Silence from: obrien
2007-06-12 02:16:02 +00:00
dma_addr_max = BUS_SPACE_MAXADDR_32BIT;
if ((sc->nfe_flags & NFE_40BIT_ADDR) != 0)
dma_addr_max = NFE_DMA_MAXADDR;
error = bus_dma_tag_create(
bus_get_dma_tag(sc->nfe_dev), /* parent */
style(9) cleanup.
2006-10-19 10:01:26 +00:00
1, 0, /* alignment, boundary */
Bring overhauled nfe(4) into tree. o s/printf/device_printf/g o Nuke OpenBSDism. o Nuke NetBSD/OpenBSD specific DMA sync operations.(we don't have a way to sync a single descriptor within a DMA map.) o Remove recursive mutex. o bus_dma(9) clean up. o 40bit DMA address support. o Add protection for Rx map load failure. o Fix a long standing bug for watchdog timeout. [1] o Add additional protections, missing Tx completion interrupt, losing start Tx command, for watchdog timeout. o Switch to taskqueue(9) API to handle interrupts. o Use our own timer for watchdog instead of if_watchdog/if_timer interface. o Advertise VLAN header length/capability correctly to upper layer. o Remove excessive kernel stack consumption in nfe_encap(). o Handle highly fragmented mbuf chains correctly. o Enable etherenet address reprogramming with ifconfig(8). o Add ALTQ/TSO, MSI/MSIX support. o Increased Rx ring to 256 descriptors from 128. o Align Tx/Rx descriptor ring on sizeof(struct nfe_desc64) boundary. o Remove alignment restrictions on Tx/Rx buffers. o Rewritten jumbo frame support code. o Add support for hardware assistend VLAN tag insertion/stripping. o Add support for Tx/Rx flow control based on patches from Peer Chen. [2] o Add a routine that detects whether ethernet address swap routines is required. [3] o Add a workaround that take MAC/PHY out of power down mode. o Add suspend/resume support. o style(9) and code clean up. Special thanks to Shigeaki Tagashira, the original porter of nfe(4), who submitted lots of patches, performed uncountable number of regression tests and maintained nfe(4) for a long time. Without his enthusiastic help and support I could never have completed this overhauling task. The only weak point of nfe(4) compared to nve(4) is instability of manual half-duplex media selection on certain hardwares(auto sensing media type should work for all cases, though). This was a long standing bug of nfe(4) and I still have no idea why it doesn't work on some hardwares. Obtained from: OpenBSD [1] Submitted by: Peer Chen < pchen at nvidia dot com > [2], [3] Reviewed by: Shigeaki Tagashira < shigeaki AT se DOT hiroshima-u DOT ac DOT jp > Tested by: Shigeaki Tagashira, current Discussed with: current Silence from: obrien
2007-06-12 02:16:02 +00:00
dma_addr_max, /* lowaddr */
style(9) cleanup.
2006-10-19 10:01:26 +00:00
BUS_SPACE_MAXADDR, /* highaddr */
NULL, NULL, /* filter, filterarg */
Bring overhauled nfe(4) into tree. o s/printf/device_printf/g o Nuke OpenBSDism. o Nuke NetBSD/OpenBSD specific DMA sync operations.(we don't have a way to sync a single descriptor within a DMA map.) o Remove recursive mutex. o bus_dma(9) clean up. o 40bit DMA address support. o Add protection for Rx map load failure. o Fix a long standing bug for watchdog timeout. [1] o Add additional protections, missing Tx completion interrupt, losing start Tx command, for watchdog timeout. o Switch to taskqueue(9) API to handle interrupts. o Use our own timer for watchdog instead of if_watchdog/if_timer interface. o Advertise VLAN header length/capability correctly to upper layer. o Remove excessive kernel stack consumption in nfe_encap(). o Handle highly fragmented mbuf chains correctly. o Enable etherenet address reprogramming with ifconfig(8). o Add ALTQ/TSO, MSI/MSIX support. o Increased Rx ring to 256 descriptors from 128. o Align Tx/Rx descriptor ring on sizeof(struct nfe_desc64) boundary. o Remove alignment restrictions on Tx/Rx buffers. o Rewritten jumbo frame support code. o Add support for hardware assistend VLAN tag insertion/stripping. o Add support for Tx/Rx flow control based on patches from Peer Chen. [2] o Add a routine that detects whether ethernet address swap routines is required. [3] o Add a workaround that take MAC/PHY out of power down mode. o Add suspend/resume support. o style(9) and code clean up. Special thanks to Shigeaki Tagashira, the original porter of nfe(4), who submitted lots of patches, performed uncountable number of regression tests and maintained nfe(4) for a long time. Without his enthusiastic help and support I could never have completed this overhauling task. The only weak point of nfe(4) compared to nve(4) is instability of manual half-duplex media selection on certain hardwares(auto sensing media type should work for all cases, though). This was a long standing bug of nfe(4) and I still have no idea why it doesn't work on some hardwares. Obtained from: OpenBSD [1] Submitted by: Peer Chen < pchen at nvidia dot com > [2], [3] Reviewed by: Shigeaki Tagashira < shigeaki AT se DOT hiroshima-u DOT ac DOT jp > Tested by: Shigeaki Tagashira, current Discussed with: current Silence from: obrien
2007-06-12 02:16:02 +00:00
BUS_SPACE_MAXSIZE_32BIT, 0, /* maxsize, nsegments */
style(9) cleanup.
2006-10-19 10:01:26 +00:00
BUS_SPACE_MAXSIZE_32BIT, /* maxsegsize */
Bring overhauled nfe(4) into tree. o s/printf/device_printf/g o Nuke OpenBSDism. o Nuke NetBSD/OpenBSD specific DMA sync operations.(we don't have a way to sync a single descriptor within a DMA map.) o Remove recursive mutex. o bus_dma(9) clean up. o 40bit DMA address support. o Add protection for Rx map load failure. o Fix a long standing bug for watchdog timeout. [1] o Add additional protections, missing Tx completion interrupt, losing start Tx command, for watchdog timeout. o Switch to taskqueue(9) API to handle interrupts. o Use our own timer for watchdog instead of if_watchdog/if_timer interface. o Advertise VLAN header length/capability correctly to upper layer. o Remove excessive kernel stack consumption in nfe_encap(). o Handle highly fragmented mbuf chains correctly. o Enable etherenet address reprogramming with ifconfig(8). o Add ALTQ/TSO, MSI/MSIX support. o Increased Rx ring to 256 descriptors from 128. o Align Tx/Rx descriptor ring on sizeof(struct nfe_desc64) boundary. o Remove alignment restrictions on Tx/Rx buffers. o Rewritten jumbo frame support code. o Add support for hardware assistend VLAN tag insertion/stripping. o Add support for Tx/Rx flow control based on patches from Peer Chen. [2] o Add a routine that detects whether ethernet address swap routines is required. [3] o Add a workaround that take MAC/PHY out of power down mode. o Add suspend/resume support. o style(9) and code clean up. Special thanks to Shigeaki Tagashira, the original porter of nfe(4), who submitted lots of patches, performed uncountable number of regression tests and maintained nfe(4) for a long time. Without his enthusiastic help and support I could never have completed this overhauling task. The only weak point of nfe(4) compared to nve(4) is instability of manual half-duplex media selection on certain hardwares(auto sensing media type should work for all cases, though). This was a long standing bug of nfe(4) and I still have no idea why it doesn't work on some hardwares. Obtained from: OpenBSD [1] Submitted by: Peer Chen < pchen at nvidia dot com > [2], [3] Reviewed by: Shigeaki Tagashira < shigeaki AT se DOT hiroshima-u DOT ac DOT jp > Tested by: Shigeaki Tagashira, current Discussed with: current Silence from: obrien
2007-06-12 02:16:02 +00:00
0, /* flags */
style(9) cleanup.
2006-10-19 10:01:26 +00:00
NULL, NULL, /* lockfunc, lockarg */
&sc->nfe_parent_tag);
Add a pure open source nForce Ethernet driver, under BSDL. This driver was ported from OpenBSD by Shigeaki Tagashira <shigeaki@se.hiroshima-u.ac.jp> and posted at http://www.se.hiroshima-u.ac.jp/~shigeaki/software/freebsd-nfe.html It was additionally cleaned up by me. It is still a work-in-progress and thus is purposefully not in GENERIC. And it conflicts with nve(4), so only one should be loaded.
2006-06-26 23:41:07 +00:00
if (error)
goto fail;
Fix the jumbo frame support. Submitted by: Peer Chen at nVIDIA (reworked by me)
2006-11-27 00:16:47 +00:00
ifp = sc->nfe_ifp = if_alloc(IFT_ETHER);
if (ifp == NULL) {
Bring overhauled nfe(4) into tree. o s/printf/device_printf/g o Nuke OpenBSDism. o Nuke NetBSD/OpenBSD specific DMA sync operations.(we don't have a way to sync a single descriptor within a DMA map.) o Remove recursive mutex. o bus_dma(9) clean up. o 40bit DMA address support. o Add protection for Rx map load failure. o Fix a long standing bug for watchdog timeout. [1] o Add additional protections, missing Tx completion interrupt, losing start Tx command, for watchdog timeout. o Switch to taskqueue(9) API to handle interrupts. o Use our own timer for watchdog instead of if_watchdog/if_timer interface. o Advertise VLAN header length/capability correctly to upper layer. o Remove excessive kernel stack consumption in nfe_encap(). o Handle highly fragmented mbuf chains correctly. o Enable etherenet address reprogramming with ifconfig(8). o Add ALTQ/TSO, MSI/MSIX support. o Increased Rx ring to 256 descriptors from 128. o Align Tx/Rx descriptor ring on sizeof(struct nfe_desc64) boundary. o Remove alignment restrictions on Tx/Rx buffers. o Rewritten jumbo frame support code. o Add support for hardware assistend VLAN tag insertion/stripping. o Add support for Tx/Rx flow control based on patches from Peer Chen. [2] o Add a routine that detects whether ethernet address swap routines is required. [3] o Add a workaround that take MAC/PHY out of power down mode. o Add suspend/resume support. o style(9) and code clean up. Special thanks to Shigeaki Tagashira, the original porter of nfe(4), who submitted lots of patches, performed uncountable number of regression tests and maintained nfe(4) for a long time. Without his enthusiastic help and support I could never have completed this overhauling task. The only weak point of nfe(4) compared to nve(4) is instability of manual half-duplex media selection on certain hardwares(auto sensing media type should work for all cases, though). This was a long standing bug of nfe(4) and I still have no idea why it doesn't work on some hardwares. Obtained from: OpenBSD [1] Submitted by: Peer Chen < pchen at nvidia dot com > [2], [3] Reviewed by: Shigeaki Tagashira < shigeaki AT se DOT hiroshima-u DOT ac DOT jp > Tested by: Shigeaki Tagashira, current Discussed with: current Silence from: obrien
2007-06-12 02:16:02 +00:00
device_printf(dev, "can not if_alloc()\n");
Fix the jumbo frame support. Submitted by: Peer Chen at nVIDIA (reworked by me)
2006-11-27 00:16:47 +00:00
error = ENOSPC;
goto fail;
}
Import the nfe (nForce Ethernet) driver from OpenBSD. This is a pure open source NIC driver with a BSD license. These bits were taken from the OpenBSD public CVS repository on 1-May-06.
2006-06-26 17:25:22 +00:00
/*
* Allocate Tx and Rx rings.
*/
Bring overhauled nfe(4) into tree. o s/printf/device_printf/g o Nuke OpenBSDism. o Nuke NetBSD/OpenBSD specific DMA sync operations.(we don't have a way to sync a single descriptor within a DMA map.) o Remove recursive mutex. o bus_dma(9) clean up. o 40bit DMA address support. o Add protection for Rx map load failure. o Fix a long standing bug for watchdog timeout. [1] o Add additional protections, missing Tx completion interrupt, losing start Tx command, for watchdog timeout. o Switch to taskqueue(9) API to handle interrupts. o Use our own timer for watchdog instead of if_watchdog/if_timer interface. o Advertise VLAN header length/capability correctly to upper layer. o Remove excessive kernel stack consumption in nfe_encap(). o Handle highly fragmented mbuf chains correctly. o Enable etherenet address reprogramming with ifconfig(8). o Add ALTQ/TSO, MSI/MSIX support. o Increased Rx ring to 256 descriptors from 128. o Align Tx/Rx descriptor ring on sizeof(struct nfe_desc64) boundary. o Remove alignment restrictions on Tx/Rx buffers. o Rewritten jumbo frame support code. o Add support for hardware assistend VLAN tag insertion/stripping. o Add support for Tx/Rx flow control based on patches from Peer Chen. [2] o Add a routine that detects whether ethernet address swap routines is required. [3] o Add a workaround that take MAC/PHY out of power down mode. o Add suspend/resume support. o style(9) and code clean up. Special thanks to Shigeaki Tagashira, the original porter of nfe(4), who submitted lots of patches, performed uncountable number of regression tests and maintained nfe(4) for a long time. Without his enthusiastic help and support I could never have completed this overhauling task. The only weak point of nfe(4) compared to nve(4) is instability of manual half-duplex media selection on certain hardwares(auto sensing media type should work for all cases, though). This was a long standing bug of nfe(4) and I still have no idea why it doesn't work on some hardwares. Obtained from: OpenBSD [1] Submitted by: Peer Chen < pchen at nvidia dot com > [2], [3] Reviewed by: Shigeaki Tagashira < shigeaki AT se DOT hiroshima-u DOT ac DOT jp > Tested by: Shigeaki Tagashira, current Discussed with: current Silence from: obrien
2007-06-12 02:16:02 +00:00
if ((error = nfe_alloc_tx_ring(sc, &sc->txq)) != 0)
Add a pure open source nForce Ethernet driver, under BSDL. This driver was ported from OpenBSD by Shigeaki Tagashira <shigeaki@se.hiroshima-u.ac.jp> and posted at http://www.se.hiroshima-u.ac.jp/~shigeaki/software/freebsd-nfe.html It was additionally cleaned up by me. It is still a work-in-progress and thus is purposefully not in GENERIC. And it conflicts with nve(4), so only one should be loaded.
2006-06-26 23:41:07 +00:00
goto fail;
Import the nfe (nForce Ethernet) driver from OpenBSD. This is a pure open source NIC driver with a BSD license. These bits were taken from the OpenBSD public CVS repository on 1-May-06.
2006-06-26 17:25:22 +00:00
Bring overhauled nfe(4) into tree. o s/printf/device_printf/g o Nuke OpenBSDism. o Nuke NetBSD/OpenBSD specific DMA sync operations.(we don't have a way to sync a single descriptor within a DMA map.) o Remove recursive mutex. o bus_dma(9) clean up. o 40bit DMA address support. o Add protection for Rx map load failure. o Fix a long standing bug for watchdog timeout. [1] o Add additional protections, missing Tx completion interrupt, losing start Tx command, for watchdog timeout. o Switch to taskqueue(9) API to handle interrupts. o Use our own timer for watchdog instead of if_watchdog/if_timer interface. o Advertise VLAN header length/capability correctly to upper layer. o Remove excessive kernel stack consumption in nfe_encap(). o Handle highly fragmented mbuf chains correctly. o Enable etherenet address reprogramming with ifconfig(8). o Add ALTQ/TSO, MSI/MSIX support. o Increased Rx ring to 256 descriptors from 128. o Align Tx/Rx descriptor ring on sizeof(struct nfe_desc64) boundary. o Remove alignment restrictions on Tx/Rx buffers. o Rewritten jumbo frame support code. o Add support for hardware assistend VLAN tag insertion/stripping. o Add support for Tx/Rx flow control based on patches from Peer Chen. [2] o Add a routine that detects whether ethernet address swap routines is required. [3] o Add a workaround that take MAC/PHY out of power down mode. o Add suspend/resume support. o style(9) and code clean up. Special thanks to Shigeaki Tagashira, the original porter of nfe(4), who submitted lots of patches, performed uncountable number of regression tests and maintained nfe(4) for a long time. Without his enthusiastic help and support I could never have completed this overhauling task. The only weak point of nfe(4) compared to nve(4) is instability of manual half-duplex media selection on certain hardwares(auto sensing media type should work for all cases, though). This was a long standing bug of nfe(4) and I still have no idea why it doesn't work on some hardwares. Obtained from: OpenBSD [1] Submitted by: Peer Chen < pchen at nvidia dot com > [2], [3] Reviewed by: Shigeaki Tagashira < shigeaki AT se DOT hiroshima-u DOT ac DOT jp > Tested by: Shigeaki Tagashira, current Discussed with: current Silence from: obrien
2007-06-12 02:16:02 +00:00
if ((error = nfe_alloc_rx_ring(sc, &sc->rxq)) != 0)
goto fail;
Don't fail on device attach if jumbo frame support was unsuccessful. Because nfe(4) hardware doesn't support SG on Rx path, supporting jumbo frame requires very large contiguous kernel memory(i.e. several mega bytes). In case of lack of contiguous kernel memory that allocation request may always fail. However nfe(4) can operate on normal sized MTU frames, so go ahead and just disable jumbo frame support. While I'm here add a new tunable "hw.nfe.jumbo_disable" to disable jumbo frame support. In nfe_poll, make sure to invoke correct Rx handler. Approved by: re (kensmith)
2007-07-24 01:11:00 +00:00
nfe_alloc_jrx_ring(sc, &sc->jrxq);
Add hardware MAC statistics support. Register information was obtained from Linux forcedeth driver. While I'm here move creating a sysctl node for process_limit to function nfe_sysctl_node(). Tested by: "Arno J. Klaassen" < arno <at> heho dot snv dot jussieu dot fr >
2008-10-03 03:58:16 +00:00
/* Create sysctl node. */
nfe_sysctl_node(sc);
Add a pure open source nForce Ethernet driver, under BSDL. This driver was ported from OpenBSD by Shigeaki Tagashira <shigeaki@se.hiroshima-u.ac.jp> and posted at http://www.se.hiroshima-u.ac.jp/~shigeaki/software/freebsd-nfe.html It was additionally cleaned up by me. It is still a work-in-progress and thus is purposefully not in GENERIC. And it conflicts with nve(4), so only one should be loaded.
2006-06-26 23:41:07 +00:00
Import the nfe (nForce Ethernet) driver from OpenBSD. This is a pure open source NIC driver with a BSD license. These bits were taken from the OpenBSD public CVS repository on 1-May-06.
2006-06-26 17:25:22 +00:00
ifp->if_softc = sc;
Add a pure open source nForce Ethernet driver, under BSDL. This driver was ported from OpenBSD by Shigeaki Tagashira <shigeaki@se.hiroshima-u.ac.jp> and posted at http://www.se.hiroshima-u.ac.jp/~shigeaki/software/freebsd-nfe.html It was additionally cleaned up by me. It is still a work-in-progress and thus is purposefully not in GENERIC. And it conflicts with nve(4), so only one should be loaded.
2006-06-26 23:41:07 +00:00
if_initname(ifp, device_get_name(dev), device_get_unit(dev));
Bring overhauled nfe(4) into tree. o s/printf/device_printf/g o Nuke OpenBSDism. o Nuke NetBSD/OpenBSD specific DMA sync operations.(we don't have a way to sync a single descriptor within a DMA map.) o Remove recursive mutex. o bus_dma(9) clean up. o 40bit DMA address support. o Add protection for Rx map load failure. o Fix a long standing bug for watchdog timeout. [1] o Add additional protections, missing Tx completion interrupt, losing start Tx command, for watchdog timeout. o Switch to taskqueue(9) API to handle interrupts. o Use our own timer for watchdog instead of if_watchdog/if_timer interface. o Advertise VLAN header length/capability correctly to upper layer. o Remove excessive kernel stack consumption in nfe_encap(). o Handle highly fragmented mbuf chains correctly. o Enable etherenet address reprogramming with ifconfig(8). o Add ALTQ/TSO, MSI/MSIX support. o Increased Rx ring to 256 descriptors from 128. o Align Tx/Rx descriptor ring on sizeof(struct nfe_desc64) boundary. o Remove alignment restrictions on Tx/Rx buffers. o Rewritten jumbo frame support code. o Add support for hardware assistend VLAN tag insertion/stripping. o Add support for Tx/Rx flow control based on patches from Peer Chen. [2] o Add a routine that detects whether ethernet address swap routines is required. [3] o Add a workaround that take MAC/PHY out of power down mode. o Add suspend/resume support. o style(9) and code clean up. Special thanks to Shigeaki Tagashira, the original porter of nfe(4), who submitted lots of patches, performed uncountable number of regression tests and maintained nfe(4) for a long time. Without his enthusiastic help and support I could never have completed this overhauling task. The only weak point of nfe(4) compared to nve(4) is instability of manual half-duplex media selection on certain hardwares(auto sensing media type should work for all cases, though). This was a long standing bug of nfe(4) and I still have no idea why it doesn't work on some hardwares. Obtained from: OpenBSD [1] Submitted by: Peer Chen < pchen at nvidia dot com > [2], [3] Reviewed by: Shigeaki Tagashira < shigeaki AT se DOT hiroshima-u DOT ac DOT jp > Tested by: Shigeaki Tagashira, current Discussed with: current Silence from: obrien
2007-06-12 02:16:02 +00:00
ifp->if_mtu = ETHERMTU;
Import the nfe (nForce Ethernet) driver from OpenBSD. This is a pure open source NIC driver with a BSD license. These bits were taken from the OpenBSD public CVS repository on 1-May-06.
2006-06-26 17:25:22 +00:00
ifp->if_flags = IFF_BROADCAST | IFF_SIMPLEX | IFF_MULTICAST;
ifp->if_ioctl = nfe_ioctl;
ifp->if_start = nfe_start;
Bring overhauled nfe(4) into tree. o s/printf/device_printf/g o Nuke OpenBSDism. o Nuke NetBSD/OpenBSD specific DMA sync operations.(we don't have a way to sync a single descriptor within a DMA map.) o Remove recursive mutex. o bus_dma(9) clean up. o 40bit DMA address support. o Add protection for Rx map load failure. o Fix a long standing bug for watchdog timeout. [1] o Add additional protections, missing Tx completion interrupt, losing start Tx command, for watchdog timeout. o Switch to taskqueue(9) API to handle interrupts. o Use our own timer for watchdog instead of if_watchdog/if_timer interface. o Advertise VLAN header length/capability correctly to upper layer. o Remove excessive kernel stack consumption in nfe_encap(). o Handle highly fragmented mbuf chains correctly. o Enable etherenet address reprogramming with ifconfig(8). o Add ALTQ/TSO, MSI/MSIX support. o Increased Rx ring to 256 descriptors from 128. o Align Tx/Rx descriptor ring on sizeof(struct nfe_desc64) boundary. o Remove alignment restrictions on Tx/Rx buffers. o Rewritten jumbo frame support code. o Add support for hardware assistend VLAN tag insertion/stripping. o Add support for Tx/Rx flow control based on patches from Peer Chen. [2] o Add a routine that detects whether ethernet address swap routines is required. [3] o Add a workaround that take MAC/PHY out of power down mode. o Add suspend/resume support. o style(9) and code clean up. Special thanks to Shigeaki Tagashira, the original porter of nfe(4), who submitted lots of patches, performed uncountable number of regression tests and maintained nfe(4) for a long time. Without his enthusiastic help and support I could never have completed this overhauling task. The only weak point of nfe(4) compared to nve(4) is instability of manual half-duplex media selection on certain hardwares(auto sensing media type should work for all cases, though). This was a long standing bug of nfe(4) and I still have no idea why it doesn't work on some hardwares. Obtained from: OpenBSD [1] Submitted by: Peer Chen < pchen at nvidia dot com > [2], [3] Reviewed by: Shigeaki Tagashira < shigeaki AT se DOT hiroshima-u DOT ac DOT jp > Tested by: Shigeaki Tagashira, current Discussed with: current Silence from: obrien
2007-06-12 02:16:02 +00:00
ifp->if_hwassist = 0;
ifp->if_capabilities = 0;
Import the nfe (nForce Ethernet) driver from OpenBSD. This is a pure open source NIC driver with a BSD license. These bits were taken from the OpenBSD public CVS repository on 1-May-06.
2006-06-26 17:25:22 +00:00
ifp->if_init = nfe_init;
Bring overhauled nfe(4) into tree. o s/printf/device_printf/g o Nuke OpenBSDism. o Nuke NetBSD/OpenBSD specific DMA sync operations.(we don't have a way to sync a single descriptor within a DMA map.) o Remove recursive mutex. o bus_dma(9) clean up. o 40bit DMA address support. o Add protection for Rx map load failure. o Fix a long standing bug for watchdog timeout. [1] o Add additional protections, missing Tx completion interrupt, losing start Tx command, for watchdog timeout. o Switch to taskqueue(9) API to handle interrupts. o Use our own timer for watchdog instead of if_watchdog/if_timer interface. o Advertise VLAN header length/capability correctly to upper layer. o Remove excessive kernel stack consumption in nfe_encap(). o Handle highly fragmented mbuf chains correctly. o Enable etherenet address reprogramming with ifconfig(8). o Add ALTQ/TSO, MSI/MSIX support. o Increased Rx ring to 256 descriptors from 128. o Align Tx/Rx descriptor ring on sizeof(struct nfe_desc64) boundary. o Remove alignment restrictions on Tx/Rx buffers. o Rewritten jumbo frame support code. o Add support for hardware assistend VLAN tag insertion/stripping. o Add support for Tx/Rx flow control based on patches from Peer Chen. [2] o Add a routine that detects whether ethernet address swap routines is required. [3] o Add a workaround that take MAC/PHY out of power down mode. o Add suspend/resume support. o style(9) and code clean up. Special thanks to Shigeaki Tagashira, the original porter of nfe(4), who submitted lots of patches, performed uncountable number of regression tests and maintained nfe(4) for a long time. Without his enthusiastic help and support I could never have completed this overhauling task. The only weak point of nfe(4) compared to nve(4) is instability of manual half-duplex media selection on certain hardwares(auto sensing media type should work for all cases, though). This was a long standing bug of nfe(4) and I still have no idea why it doesn't work on some hardwares. Obtained from: OpenBSD [1] Submitted by: Peer Chen < pchen at nvidia dot com > [2], [3] Reviewed by: Shigeaki Tagashira < shigeaki AT se DOT hiroshima-u DOT ac DOT jp > Tested by: Shigeaki Tagashira, current Discussed with: current Silence from: obrien
2007-06-12 02:16:02 +00:00
IFQ_SET_MAXLEN(&ifp->if_snd, NFE_TX_RING_COUNT - 1);
ifp->if_snd.ifq_drv_maxlen = NFE_TX_RING_COUNT - 1;
IFQ_SET_READY(&ifp->if_snd);
Fix the jumbo frame support. Submitted by: Peer Chen at nVIDIA (reworked by me)
2006-11-27 00:16:47 +00:00
Bring overhauled nfe(4) into tree. o s/printf/device_printf/g o Nuke OpenBSDism. o Nuke NetBSD/OpenBSD specific DMA sync operations.(we don't have a way to sync a single descriptor within a DMA map.) o Remove recursive mutex. o bus_dma(9) clean up. o 40bit DMA address support. o Add protection for Rx map load failure. o Fix a long standing bug for watchdog timeout. [1] o Add additional protections, missing Tx completion interrupt, losing start Tx command, for watchdog timeout. o Switch to taskqueue(9) API to handle interrupts. o Use our own timer for watchdog instead of if_watchdog/if_timer interface. o Advertise VLAN header length/capability correctly to upper layer. o Remove excessive kernel stack consumption in nfe_encap(). o Handle highly fragmented mbuf chains correctly. o Enable etherenet address reprogramming with ifconfig(8). o Add ALTQ/TSO, MSI/MSIX support. o Increased Rx ring to 256 descriptors from 128. o Align Tx/Rx descriptor ring on sizeof(struct nfe_desc64) boundary. o Remove alignment restrictions on Tx/Rx buffers. o Rewritten jumbo frame support code. o Add support for hardware assistend VLAN tag insertion/stripping. o Add support for Tx/Rx flow control based on patches from Peer Chen. [2] o Add a routine that detects whether ethernet address swap routines is required. [3] o Add a workaround that take MAC/PHY out of power down mode. o Add suspend/resume support. o style(9) and code clean up. Special thanks to Shigeaki Tagashira, the original porter of nfe(4), who submitted lots of patches, performed uncountable number of regression tests and maintained nfe(4) for a long time. Without his enthusiastic help and support I could never have completed this overhauling task. The only weak point of nfe(4) compared to nve(4) is instability of manual half-duplex media selection on certain hardwares(auto sensing media type should work for all cases, though). This was a long standing bug of nfe(4) and I still have no idea why it doesn't work on some hardwares. Obtained from: OpenBSD [1] Submitted by: Peer Chen < pchen at nvidia dot com > [2], [3] Reviewed by: Shigeaki Tagashira < shigeaki AT se DOT hiroshima-u DOT ac DOT jp > Tested by: Shigeaki Tagashira, current Discussed with: current Silence from: obrien
2007-06-12 02:16:02 +00:00
if (sc->nfe_flags & NFE_HW_CSUM) {
ifp->if_capabilities |= IFCAP_HWCSUM | IFCAP_TSO4;
ifp->if_hwassist |= NFE_CSUM_FEATURES | CSUM_TSO;
}
ifp->if_capenable = ifp->if_capabilities;
Fix the jumbo frame support. Submitted by: Peer Chen at nVIDIA (reworked by me)
2006-11-27 00:16:47 +00:00
Bring overhauled nfe(4) into tree. o s/printf/device_printf/g o Nuke OpenBSDism. o Nuke NetBSD/OpenBSD specific DMA sync operations.(we don't have a way to sync a single descriptor within a DMA map.) o Remove recursive mutex. o bus_dma(9) clean up. o 40bit DMA address support. o Add protection for Rx map load failure. o Fix a long standing bug for watchdog timeout. [1] o Add additional protections, missing Tx completion interrupt, losing start Tx command, for watchdog timeout. o Switch to taskqueue(9) API to handle interrupts. o Use our own timer for watchdog instead of if_watchdog/if_timer interface. o Advertise VLAN header length/capability correctly to upper layer. o Remove excessive kernel stack consumption in nfe_encap(). o Handle highly fragmented mbuf chains correctly. o Enable etherenet address reprogramming with ifconfig(8). o Add ALTQ/TSO, MSI/MSIX support. o Increased Rx ring to 256 descriptors from 128. o Align Tx/Rx descriptor ring on sizeof(struct nfe_desc64) boundary. o Remove alignment restrictions on Tx/Rx buffers. o Rewritten jumbo frame support code. o Add support for hardware assistend VLAN tag insertion/stripping. o Add support for Tx/Rx flow control based on patches from Peer Chen. [2] o Add a routine that detects whether ethernet address swap routines is required. [3] o Add a workaround that take MAC/PHY out of power down mode. o Add suspend/resume support. o style(9) and code clean up. Special thanks to Shigeaki Tagashira, the original porter of nfe(4), who submitted lots of patches, performed uncountable number of regression tests and maintained nfe(4) for a long time. Without his enthusiastic help and support I could never have completed this overhauling task. The only weak point of nfe(4) compared to nve(4) is instability of manual half-duplex media selection on certain hardwares(auto sensing media type should work for all cases, though). This was a long standing bug of nfe(4) and I still have no idea why it doesn't work on some hardwares. Obtained from: OpenBSD [1] Submitted by: Peer Chen < pchen at nvidia dot com > [2], [3] Reviewed by: Shigeaki Tagashira < shigeaki AT se DOT hiroshima-u DOT ac DOT jp > Tested by: Shigeaki Tagashira, current Discussed with: current Silence from: obrien
2007-06-12 02:16:02 +00:00
sc->nfe_framesize = ifp->if_mtu + NFE_RX_HEADERS;
/* VLAN capability setup. */
ifp->if_capabilities |= IFCAP_VLAN_MTU;
if ((sc->nfe_flags & NFE_HW_VLAN) != 0) {
Import the nfe (nForce Ethernet) driver from OpenBSD. This is a pure open source NIC driver with a BSD license. These bits were taken from the OpenBSD public CVS repository on 1-May-06.
2006-06-26 17:25:22 +00:00
ifp->if_capabilities |= IFCAP_VLAN_HWTAGGING;
Bring overhauled nfe(4) into tree. o s/printf/device_printf/g o Nuke OpenBSDism. o Nuke NetBSD/OpenBSD specific DMA sync operations.(we don't have a way to sync a single descriptor within a DMA map.) o Remove recursive mutex. o bus_dma(9) clean up. o 40bit DMA address support. o Add protection for Rx map load failure. o Fix a long standing bug for watchdog timeout. [1] o Add additional protections, missing Tx completion interrupt, losing start Tx command, for watchdog timeout. o Switch to taskqueue(9) API to handle interrupts. o Use our own timer for watchdog instead of if_watchdog/if_timer interface. o Advertise VLAN header length/capability correctly to upper layer. o Remove excessive kernel stack consumption in nfe_encap(). o Handle highly fragmented mbuf chains correctly. o Enable etherenet address reprogramming with ifconfig(8). o Add ALTQ/TSO, MSI/MSIX support. o Increased Rx ring to 256 descriptors from 128. o Align Tx/Rx descriptor ring on sizeof(struct nfe_desc64) boundary. o Remove alignment restrictions on Tx/Rx buffers. o Rewritten jumbo frame support code. o Add support for hardware assistend VLAN tag insertion/stripping. o Add support for Tx/Rx flow control based on patches from Peer Chen. [2] o Add a routine that detects whether ethernet address swap routines is required. [3] o Add a workaround that take MAC/PHY out of power down mode. o Add suspend/resume support. o style(9) and code clean up. Special thanks to Shigeaki Tagashira, the original porter of nfe(4), who submitted lots of patches, performed uncountable number of regression tests and maintained nfe(4) for a long time. Without his enthusiastic help and support I could never have completed this overhauling task. The only weak point of nfe(4) compared to nve(4) is instability of manual half-duplex media selection on certain hardwares(auto sensing media type should work for all cases, though). This was a long standing bug of nfe(4) and I still have no idea why it doesn't work on some hardwares. Obtained from: OpenBSD [1] Submitted by: Peer Chen < pchen at nvidia dot com > [2], [3] Reviewed by: Shigeaki Tagashira < shigeaki AT se DOT hiroshima-u DOT ac DOT jp > Tested by: Shigeaki Tagashira, current Discussed with: current Silence from: obrien
2007-06-12 02:16:02 +00:00
if ((ifp->if_capabilities & IFCAP_HWCSUM) != 0)
MCP55 is the only NVIDIA controller that supports VLAN tag insertion/stripping and it also supports TSO over VLAN. Implement TSO over VLAN support for MCP55 controller. While I'm here clean up SIOCSIFCAP ioctl handler. Since nfe(4) sets ifp capabilities based on various hardware flags in device attach, there is no need to check hardware flags again in SIOCSIFCAP ioctl handler. Also fix a bug which toggled both TX and RX checksum offloading even if user requested either TX or RX checksum configuration change. Tested by: Rob Farmer ( rfarmer <> predatorlabs dot net )
2010-11-17 18:09:02 +00:00
ifp->if_capabilities |= IFCAP_VLAN_HWCSUM |
IFCAP_VLAN_HWTSO;
Import the nfe (nForce Ethernet) driver from OpenBSD. This is a pure open source NIC driver with a BSD license. These bits were taken from the OpenBSD public CVS repository on 1-May-06.
2006-06-26 17:25:22 +00:00
}
Add basic WOL support for MCP ethernet controllers. It seems the controller does not perform automatic switching from 1000Mbps link to 10/100Mbps link when WOL is activated. Implement establishing 10/100Mps link with auto-negotiation in driver. Link status change handler was modified to remove taskqueue based approach since driver now needs synchronous handling for link establishment. Submitted by: Yamagi Burmeister (lists <> yamagi.org ) (initial version) Tested by: Yamagi Burmeister (lists <> yamagi.org ) MFC after: 1 week
2010-11-11 18:08:50 +00:00
Do a sweep of the tree replacing calls to pci_find_extcap() with calls to pci_find_cap() instead.
2011-03-23 13:10:15 +00:00
if (pci_find_cap(dev, PCIY_PMG, &reg) == 0)
Add basic WOL support for MCP ethernet controllers. It seems the controller does not perform automatic switching from 1000Mbps link to 10/100Mbps link when WOL is activated. Implement establishing 10/100Mps link with auto-negotiation in driver. Link status change handler was modified to remove taskqueue based approach since driver now needs synchronous handling for link establishment. Submitted by: Yamagi Burmeister (lists <> yamagi.org ) (initial version) Tested by: Yamagi Burmeister (lists <> yamagi.org ) MFC after: 1 week
2010-11-11 18:08:50 +00:00
ifp->if_capabilities |= IFCAP_WOL_MAGIC;
Add a pure open source nForce Ethernet driver, under BSDL. This driver was ported from OpenBSD by Shigeaki Tagashira <shigeaki@se.hiroshima-u.ac.jp> and posted at http://www.se.hiroshima-u.ac.jp/~shigeaki/software/freebsd-nfe.html It was additionally cleaned up by me. It is still a work-in-progress and thus is purposefully not in GENERIC. And it conflicts with nve(4), so only one should be loaded.
2006-06-26 23:41:07 +00:00
ifp->if_capenable = ifp->if_capabilities;
Bring overhauled nfe(4) into tree. o s/printf/device_printf/g o Nuke OpenBSDism. o Nuke NetBSD/OpenBSD specific DMA sync operations.(we don't have a way to sync a single descriptor within a DMA map.) o Remove recursive mutex. o bus_dma(9) clean up. o 40bit DMA address support. o Add protection for Rx map load failure. o Fix a long standing bug for watchdog timeout. [1] o Add additional protections, missing Tx completion interrupt, losing start Tx command, for watchdog timeout. o Switch to taskqueue(9) API to handle interrupts. o Use our own timer for watchdog instead of if_watchdog/if_timer interface. o Advertise VLAN header length/capability correctly to upper layer. o Remove excessive kernel stack consumption in nfe_encap(). o Handle highly fragmented mbuf chains correctly. o Enable etherenet address reprogramming with ifconfig(8). o Add ALTQ/TSO, MSI/MSIX support. o Increased Rx ring to 256 descriptors from 128. o Align Tx/Rx descriptor ring on sizeof(struct nfe_desc64) boundary. o Remove alignment restrictions on Tx/Rx buffers. o Rewritten jumbo frame support code. o Add support for hardware assistend VLAN tag insertion/stripping. o Add support for Tx/Rx flow control based on patches from Peer Chen. [2] o Add a routine that detects whether ethernet address swap routines is required. [3] o Add a workaround that take MAC/PHY out of power down mode. o Add suspend/resume support. o style(9) and code clean up. Special thanks to Shigeaki Tagashira, the original porter of nfe(4), who submitted lots of patches, performed uncountable number of regression tests and maintained nfe(4) for a long time. Without his enthusiastic help and support I could never have completed this overhauling task. The only weak point of nfe(4) compared to nve(4) is instability of manual half-duplex media selection on certain hardwares(auto sensing media type should work for all cases, though). This was a long standing bug of nfe(4) and I still have no idea why it doesn't work on some hardwares. Obtained from: OpenBSD [1] Submitted by: Peer Chen < pchen at nvidia dot com > [2], [3] Reviewed by: Shigeaki Tagashira < shigeaki AT se DOT hiroshima-u DOT ac DOT jp > Tested by: Shigeaki Tagashira, current Discussed with: current Silence from: obrien
2007-06-12 02:16:02 +00:00
/*
* Tell the upper layer(s) we support long frames.
* Must appear after the call to ether_ifattach() because
* ether_ifattach() sets ifi_hdrlen to the default value.
*/
ifp->if_data.ifi_hdrlen = sizeof(struct ether_vlan_header);
Add a pure open source nForce Ethernet driver, under BSDL. This driver was ported from OpenBSD by Shigeaki Tagashira <shigeaki@se.hiroshima-u.ac.jp> and posted at http://www.se.hiroshima-u.ac.jp/~shigeaki/software/freebsd-nfe.html It was additionally cleaned up by me. It is still a work-in-progress and thus is purposefully not in GENERIC. And it conflicts with nve(4), so only one should be loaded.
2006-06-26 23:41:07 +00:00
#ifdef DEVICE_POLLING
ifp->if_capabilities |= IFCAP_POLLING;
#endif
/* Do MII setup */
Converted the remainder of the NIC drivers to use the mii_attach() introduced in r213878 instead of mii_phy_probe(). Unlike r213893 these are only straight forward conversions though. Reviewed by: yongari
2010-10-15 15:00:30 +00:00
error = mii_attach(dev, &sc->nfe_miibus, ifp, nfe_ifmedia_upd,
o Flesh out the generic IEEE 802.3 annex 31B full duplex flow control support in mii(4): - Merge generic flow control advertisement (which can be enabled by passing by MIIF_DOPAUSE to mii_attach(9)) and parsing support from NetBSD into mii_physubr.c and ukphy_subr.c. Unlike as in NetBSD, IFM_FLOW isn't implemented as a global option via the "don't care mask" but instead as a media specific option this. This has the following advantages: o allows flow control advertisement with autonegotiation to be turned on and off via ifconfig(8) with the default typically being off (though MIIF_FORCEPAUSE has been added causing flow control to be always advertised, allowing to easily MFC this changes for drivers that previously used home-grown support for flow control that behaved that way without breaking POLA) o allows to deal with PHY drivers where flow control advertisement with manual selection doesn't work or at least isn't implemented, like it's the case with brgphy(4), e1000phy(4) and ip1000phy(4), by setting MIIF_NOMANPAUSE o the available combinations of media options are readily available from the `ifconfig -m` output - Add IFM_FLOW to IFM_SHARED_OPTION_DESCRIPTIONS and IFM_ETH_RXPAUSE and IFM_ETH_TXPAUSE to IFM_SUBTYPE_ETHERNET_OPTION_DESCRIPTIONS so these are understood by ifconfig(8). o Make the master/slave support in mii(4) actually usable: - Change IFM_ETH_MASTER from being implemented as a global option via the "don't care mask" to a media specific one as it actually is only applicable to IFM_1000_T to date. - Let mii_phy_setmedia() set GTCR_MAN_MS in IFM_1000_T slave mode to actually configure manually selected slave mode (like we also do in the PHY specific implementations). - Add IFM_ETH_MASTER to IFM_SUBTYPE_ETHERNET_OPTION_DESCRIPTIONS so it is understood by ifconfig(8). o Switch bge(4), bce(4), msk(4), nfe(4) and stge(4) along with brgphy(4), e1000phy(4) and ip1000phy(4) to use the generic flow control support instead of home-grown solutions via IFM_FLAGs. This includes changing these PHY drivers and smcphy(4) to no longer unconditionally advertise support for flow control but only if the selected media has IFM_FLOW set (or MIIF_FORCEPAUSE is set) and implemented for these media variants, i.e. typically only for copper. o Switch brgphy(4), ciphy(4), e1000phy(4) and ip1000phy(4) to report and set IFM_1000_T master mode via IFM_ETH_MASTER instead of via IFF_LINK0 and some IFM_FLAGn. o Switch brgphy(4) to add at least the the supported copper media based on the contents of the BMSR via mii_phy_add_media() instead of hardcoding them. The latter approach seems to have developed historically, besides causing unnecessary code duplication it was also undesirable because brgphy_mii_phy_auto() already based the capability advertisement on the contents of the BMSR though. o Let brgphy(4) set IFM_1000_T master mode on all supported PHY and not just BCM5701. Apparently this was a misinterpretation of a workaround in the Linux tg3 driver; BCM5701 seem to require RGPHY_1000CTL_MSE and BRGPHY_1000CTL_MSC to be set when configuring autonegotiation but this doesn't mean we can't set these as well on other PHYs for manual media selection. o Let ukphy_status() report IFM_1000_T master mode via IFM_ETH_MASTER so IFM_1000_T master mode support now is generally available with all PHY drivers. o Don't let e1000phy(4) set master/slave bits for IFM_1000_SX as it's not applicable there. Reviewed by: yongari (plus additional testing) Obtained from: NetBSD (partially), OpenBSD (partially) MFC after: 2 weeks
2010-11-14 13:26:10 +00:00
nfe_ifmedia_sts, BMSR_DEFCAPMASK, MII_PHY_ANY, MII_OFFSET_ANY,
MIIF_DOPAUSE);
Converted the remainder of the NIC drivers to use the mii_attach() introduced in r213878 instead of mii_phy_probe(). Unlike r213893 these are only straight forward conversions though. Reviewed by: yongari
2010-10-15 15:00:30 +00:00
if (error != 0) {
device_printf(dev, "attaching PHYs failed\n");
Add a pure open source nForce Ethernet driver, under BSDL. This driver was ported from OpenBSD by Shigeaki Tagashira <shigeaki@se.hiroshima-u.ac.jp> and posted at http://www.se.hiroshima-u.ac.jp/~shigeaki/software/freebsd-nfe.html It was additionally cleaned up by me. It is still a work-in-progress and thus is purposefully not in GENERIC. And it conflicts with nve(4), so only one should be loaded.
2006-06-26 23:41:07 +00:00
goto fail;
}
ether_ifattach(ifp, sc->eaddr);
Import the nfe (nForce Ethernet) driver from OpenBSD. This is a pure open source NIC driver with a BSD license. These bits were taken from the OpenBSD public CVS repository on 1-May-06.
2006-06-26 17:25:22 +00:00
Bring overhauled nfe(4) into tree. o s/printf/device_printf/g o Nuke OpenBSDism. o Nuke NetBSD/OpenBSD specific DMA sync operations.(we don't have a way to sync a single descriptor within a DMA map.) o Remove recursive mutex. o bus_dma(9) clean up. o 40bit DMA address support. o Add protection for Rx map load failure. o Fix a long standing bug for watchdog timeout. [1] o Add additional protections, missing Tx completion interrupt, losing start Tx command, for watchdog timeout. o Switch to taskqueue(9) API to handle interrupts. o Use our own timer for watchdog instead of if_watchdog/if_timer interface. o Advertise VLAN header length/capability correctly to upper layer. o Remove excessive kernel stack consumption in nfe_encap(). o Handle highly fragmented mbuf chains correctly. o Enable etherenet address reprogramming with ifconfig(8). o Add ALTQ/TSO, MSI/MSIX support. o Increased Rx ring to 256 descriptors from 128. o Align Tx/Rx descriptor ring on sizeof(struct nfe_desc64) boundary. o Remove alignment restrictions on Tx/Rx buffers. o Rewritten jumbo frame support code. o Add support for hardware assistend VLAN tag insertion/stripping. o Add support for Tx/Rx flow control based on patches from Peer Chen. [2] o Add a routine that detects whether ethernet address swap routines is required. [3] o Add a workaround that take MAC/PHY out of power down mode. o Add suspend/resume support. o style(9) and code clean up. Special thanks to Shigeaki Tagashira, the original porter of nfe(4), who submitted lots of patches, performed uncountable number of regression tests and maintained nfe(4) for a long time. Without his enthusiastic help and support I could never have completed this overhauling task. The only weak point of nfe(4) compared to nve(4) is instability of manual half-duplex media selection on certain hardwares(auto sensing media type should work for all cases, though). This was a long standing bug of nfe(4) and I still have no idea why it doesn't work on some hardwares. Obtained from: OpenBSD [1] Submitted by: Peer Chen < pchen at nvidia dot com > [2], [3] Reviewed by: Shigeaki Tagashira < shigeaki AT se DOT hiroshima-u DOT ac DOT jp > Tested by: Shigeaki Tagashira, current Discussed with: current Silence from: obrien
2007-06-12 02:16:02 +00:00
TASK_INIT(&sc->nfe_int_task, 0, nfe_int_task, sc);
sc->nfe_tq = taskqueue_create_fast("nfe_taskq", M_WAITOK,
taskqueue_thread_enqueue, &sc->nfe_tq);
taskqueue_start_threads(&sc->nfe_tq, 1, PI_NET, "%s taskq",
device_get_nameunit(sc->nfe_dev));
error = 0;
if (sc->nfe_msi == 0 && sc->nfe_msix == 0) {
error = bus_setup_intr(dev, sc->nfe_irq[0],
INTR_TYPE_NET | INTR_MPSAFE, nfe_intr, NULL, sc,
&sc->nfe_intrhand[0]);
} else {
for (i = 0; i < NFE_MSI_MESSAGES; i++) {
error = bus_setup_intr(dev, sc->nfe_irq[i],
INTR_TYPE_NET | INTR_MPSAFE, nfe_intr, NULL, sc,
&sc->nfe_intrhand[i]);
if (error != 0)
break;
}
}
Add a pure open source nForce Ethernet driver, under BSDL. This driver was ported from OpenBSD by Shigeaki Tagashira <shigeaki@se.hiroshima-u.ac.jp> and posted at http://www.se.hiroshima-u.ac.jp/~shigeaki/software/freebsd-nfe.html It was additionally cleaned up by me. It is still a work-in-progress and thus is purposefully not in GENERIC. And it conflicts with nve(4), so only one should be loaded.
2006-06-26 23:41:07 +00:00
if (error) {
Bring overhauled nfe(4) into tree. o s/printf/device_printf/g o Nuke OpenBSDism. o Nuke NetBSD/OpenBSD specific DMA sync operations.(we don't have a way to sync a single descriptor within a DMA map.) o Remove recursive mutex. o bus_dma(9) clean up. o 40bit DMA address support. o Add protection for Rx map load failure. o Fix a long standing bug for watchdog timeout. [1] o Add additional protections, missing Tx completion interrupt, losing start Tx command, for watchdog timeout. o Switch to taskqueue(9) API to handle interrupts. o Use our own timer for watchdog instead of if_watchdog/if_timer interface. o Advertise VLAN header length/capability correctly to upper layer. o Remove excessive kernel stack consumption in nfe_encap(). o Handle highly fragmented mbuf chains correctly. o Enable etherenet address reprogramming with ifconfig(8). o Add ALTQ/TSO, MSI/MSIX support. o Increased Rx ring to 256 descriptors from 128. o Align Tx/Rx descriptor ring on sizeof(struct nfe_desc64) boundary. o Remove alignment restrictions on Tx/Rx buffers. o Rewritten jumbo frame support code. o Add support for hardware assistend VLAN tag insertion/stripping. o Add support for Tx/Rx flow control based on patches from Peer Chen. [2] o Add a routine that detects whether ethernet address swap routines is required. [3] o Add a workaround that take MAC/PHY out of power down mode. o Add suspend/resume support. o style(9) and code clean up. Special thanks to Shigeaki Tagashira, the original porter of nfe(4), who submitted lots of patches, performed uncountable number of regression tests and maintained nfe(4) for a long time. Without his enthusiastic help and support I could never have completed this overhauling task. The only weak point of nfe(4) compared to nve(4) is instability of manual half-duplex media selection on certain hardwares(auto sensing media type should work for all cases, though). This was a long standing bug of nfe(4) and I still have no idea why it doesn't work on some hardwares. Obtained from: OpenBSD [1] Submitted by: Peer Chen < pchen at nvidia dot com > [2], [3] Reviewed by: Shigeaki Tagashira < shigeaki AT se DOT hiroshima-u DOT ac DOT jp > Tested by: Shigeaki Tagashira, current Discussed with: current Silence from: obrien
2007-06-12 02:16:02 +00:00
device_printf(dev, "couldn't set up irq\n");
taskqueue_free(sc->nfe_tq);
sc->nfe_tq = NULL;
Add a pure open source nForce Ethernet driver, under BSDL. This driver was ported from OpenBSD by Shigeaki Tagashira <shigeaki@se.hiroshima-u.ac.jp> and posted at http://www.se.hiroshima-u.ac.jp/~shigeaki/software/freebsd-nfe.html It was additionally cleaned up by me. It is still a work-in-progress and thus is purposefully not in GENERIC. And it conflicts with nve(4), so only one should be loaded.
2006-06-26 23:41:07 +00:00
ether_ifdetach(ifp);
goto fail;
}
fail:
if (error)
nfe_detach(dev);
return (error);
Import the nfe (nForce Ethernet) driver from OpenBSD. This is a pure open source NIC driver with a BSD license. These bits were taken from the OpenBSD public CVS repository on 1-May-06.
2006-06-26 17:25:22 +00:00
}
Add a pure open source nForce Ethernet driver, under BSDL. This driver was ported from OpenBSD by Shigeaki Tagashira <shigeaki@se.hiroshima-u.ac.jp> and posted at http://www.se.hiroshima-u.ac.jp/~shigeaki/software/freebsd-nfe.html It was additionally cleaned up by me. It is still a work-in-progress and thus is purposefully not in GENERIC. And it conflicts with nve(4), so only one should be loaded.
2006-06-26 23:41:07 +00:00
static int
nfe_detach(device_t dev)
Import the nfe (nForce Ethernet) driver from OpenBSD. This is a pure open source NIC driver with a BSD license. These bits were taken from the OpenBSD public CVS repository on 1-May-06.
2006-06-26 17:25:22 +00:00
{
style(9) cleanup.
2006-10-19 10:01:26 +00:00
struct nfe_softc *sc;
struct ifnet *ifp;
Bring overhauled nfe(4) into tree. o s/printf/device_printf/g o Nuke OpenBSDism. o Nuke NetBSD/OpenBSD specific DMA sync operations.(we don't have a way to sync a single descriptor within a DMA map.) o Remove recursive mutex. o bus_dma(9) clean up. o 40bit DMA address support. o Add protection for Rx map load failure. o Fix a long standing bug for watchdog timeout. [1] o Add additional protections, missing Tx completion interrupt, losing start Tx command, for watchdog timeout. o Switch to taskqueue(9) API to handle interrupts. o Use our own timer for watchdog instead of if_watchdog/if_timer interface. o Advertise VLAN header length/capability correctly to upper layer. o Remove excessive kernel stack consumption in nfe_encap(). o Handle highly fragmented mbuf chains correctly. o Enable etherenet address reprogramming with ifconfig(8). o Add ALTQ/TSO, MSI/MSIX support. o Increased Rx ring to 256 descriptors from 128. o Align Tx/Rx descriptor ring on sizeof(struct nfe_desc64) boundary. o Remove alignment restrictions on Tx/Rx buffers. o Rewritten jumbo frame support code. o Add support for hardware assistend VLAN tag insertion/stripping. o Add support for Tx/Rx flow control based on patches from Peer Chen. [2] o Add a routine that detects whether ethernet address swap routines is required. [3] o Add a workaround that take MAC/PHY out of power down mode. o Add suspend/resume support. o style(9) and code clean up. Special thanks to Shigeaki Tagashira, the original porter of nfe(4), who submitted lots of patches, performed uncountable number of regression tests and maintained nfe(4) for a long time. Without his enthusiastic help and support I could never have completed this overhauling task. The only weak point of nfe(4) compared to nve(4) is instability of manual half-duplex media selection on certain hardwares(auto sensing media type should work for all cases, though). This was a long standing bug of nfe(4) and I still have no idea why it doesn't work on some hardwares. Obtained from: OpenBSD [1] Submitted by: Peer Chen < pchen at nvidia dot com > [2], [3] Reviewed by: Shigeaki Tagashira < shigeaki AT se DOT hiroshima-u DOT ac DOT jp > Tested by: Shigeaki Tagashira, current Discussed with: current Silence from: obrien
2007-06-12 02:16:02 +00:00
uint8_t eaddr[ETHER_ADDR_LEN];
int i, rid;
Add a pure open source nForce Ethernet driver, under BSDL. This driver was ported from OpenBSD by Shigeaki Tagashira <shigeaki@se.hiroshima-u.ac.jp> and posted at http://www.se.hiroshima-u.ac.jp/~shigeaki/software/freebsd-nfe.html It was additionally cleaned up by me. It is still a work-in-progress and thus is purposefully not in GENERIC. And it conflicts with nve(4), so only one should be loaded.
2006-06-26 23:41:07 +00:00
sc = device_get_softc(dev);
KASSERT(mtx_initialized(&sc->nfe_mtx), ("nfe mutex not initialized"));
ifp = sc->nfe_ifp;
#ifdef DEVICE_POLLING
Bring overhauled nfe(4) into tree. o s/printf/device_printf/g o Nuke OpenBSDism. o Nuke NetBSD/OpenBSD specific DMA sync operations.(we don't have a way to sync a single descriptor within a DMA map.) o Remove recursive mutex. o bus_dma(9) clean up. o 40bit DMA address support. o Add protection for Rx map load failure. o Fix a long standing bug for watchdog timeout. [1] o Add additional protections, missing Tx completion interrupt, losing start Tx command, for watchdog timeout. o Switch to taskqueue(9) API to handle interrupts. o Use our own timer for watchdog instead of if_watchdog/if_timer interface. o Advertise VLAN header length/capability correctly to upper layer. o Remove excessive kernel stack consumption in nfe_encap(). o Handle highly fragmented mbuf chains correctly. o Enable etherenet address reprogramming with ifconfig(8). o Add ALTQ/TSO, MSI/MSIX support. o Increased Rx ring to 256 descriptors from 128. o Align Tx/Rx descriptor ring on sizeof(struct nfe_desc64) boundary. o Remove alignment restrictions on Tx/Rx buffers. o Rewritten jumbo frame support code. o Add support for hardware assistend VLAN tag insertion/stripping. o Add support for Tx/Rx flow control based on patches from Peer Chen. [2] o Add a routine that detects whether ethernet address swap routines is required. [3] o Add a workaround that take MAC/PHY out of power down mode. o Add suspend/resume support. o style(9) and code clean up. Special thanks to Shigeaki Tagashira, the original porter of nfe(4), who submitted lots of patches, performed uncountable number of regression tests and maintained nfe(4) for a long time. Without his enthusiastic help and support I could never have completed this overhauling task. The only weak point of nfe(4) compared to nve(4) is instability of manual half-duplex media selection on certain hardwares(auto sensing media type should work for all cases, though). This was a long standing bug of nfe(4) and I still have no idea why it doesn't work on some hardwares. Obtained from: OpenBSD [1] Submitted by: Peer Chen < pchen at nvidia dot com > [2], [3] Reviewed by: Shigeaki Tagashira < shigeaki AT se DOT hiroshima-u DOT ac DOT jp > Tested by: Shigeaki Tagashira, current Discussed with: current Silence from: obrien
2007-06-12 02:16:02 +00:00
if (ifp != NULL && ifp->if_capenable & IFCAP_POLLING)
Add a pure open source nForce Ethernet driver, under BSDL. This driver was ported from OpenBSD by Shigeaki Tagashira <shigeaki@se.hiroshima-u.ac.jp> and posted at http://www.se.hiroshima-u.ac.jp/~shigeaki/software/freebsd-nfe.html It was additionally cleaned up by me. It is still a work-in-progress and thus is purposefully not in GENERIC. And it conflicts with nve(4), so only one should be loaded.
2006-06-26 23:41:07 +00:00
ether_poll_deregister(ifp);
#endif
if (device_is_attached(dev)) {
Grrr. new .vimrc file is really giving me trouble.. get the whitespace right.
2006-11-27 00:10:00 +00:00
NFE_LOCK(sc);
Bring overhauled nfe(4) into tree. o s/printf/device_printf/g o Nuke OpenBSDism. o Nuke NetBSD/OpenBSD specific DMA sync operations.(we don't have a way to sync a single descriptor within a DMA map.) o Remove recursive mutex. o bus_dma(9) clean up. o 40bit DMA address support. o Add protection for Rx map load failure. o Fix a long standing bug for watchdog timeout. [1] o Add additional protections, missing Tx completion interrupt, losing start Tx command, for watchdog timeout. o Switch to taskqueue(9) API to handle interrupts. o Use our own timer for watchdog instead of if_watchdog/if_timer interface. o Advertise VLAN header length/capability correctly to upper layer. o Remove excessive kernel stack consumption in nfe_encap(). o Handle highly fragmented mbuf chains correctly. o Enable etherenet address reprogramming with ifconfig(8). o Add ALTQ/TSO, MSI/MSIX support. o Increased Rx ring to 256 descriptors from 128. o Align Tx/Rx descriptor ring on sizeof(struct nfe_desc64) boundary. o Remove alignment restrictions on Tx/Rx buffers. o Rewritten jumbo frame support code. o Add support for hardware assistend VLAN tag insertion/stripping. o Add support for Tx/Rx flow control based on patches from Peer Chen. [2] o Add a routine that detects whether ethernet address swap routines is required. [3] o Add a workaround that take MAC/PHY out of power down mode. o Add suspend/resume support. o style(9) and code clean up. Special thanks to Shigeaki Tagashira, the original porter of nfe(4), who submitted lots of patches, performed uncountable number of regression tests and maintained nfe(4) for a long time. Without his enthusiastic help and support I could never have completed this overhauling task. The only weak point of nfe(4) compared to nve(4) is instability of manual half-duplex media selection on certain hardwares(auto sensing media type should work for all cases, though). This was a long standing bug of nfe(4) and I still have no idea why it doesn't work on some hardwares. Obtained from: OpenBSD [1] Submitted by: Peer Chen < pchen at nvidia dot com > [2], [3] Reviewed by: Shigeaki Tagashira < shigeaki AT se DOT hiroshima-u DOT ac DOT jp > Tested by: Shigeaki Tagashira, current Discussed with: current Silence from: obrien
2007-06-12 02:16:02 +00:00
nfe_stop(ifp);
Add a pure open source nForce Ethernet driver, under BSDL. This driver was ported from OpenBSD by Shigeaki Tagashira <shigeaki@se.hiroshima-u.ac.jp> and posted at http://www.se.hiroshima-u.ac.jp/~shigeaki/software/freebsd-nfe.html It was additionally cleaned up by me. It is still a work-in-progress and thus is purposefully not in GENERIC. And it conflicts with nve(4), so only one should be loaded.
2006-06-26 23:41:07 +00:00
ifp->if_flags &= ~IFF_UP;
Grrr. new .vimrc file is really giving me trouble.. get the whitespace right.
2006-11-27 00:10:00 +00:00
NFE_UNLOCK(sc);
Add a pure open source nForce Ethernet driver, under BSDL. This driver was ported from OpenBSD by Shigeaki Tagashira <shigeaki@se.hiroshima-u.ac.jp> and posted at http://www.se.hiroshima-u.ac.jp/~shigeaki/software/freebsd-nfe.html It was additionally cleaned up by me. It is still a work-in-progress and thus is purposefully not in GENERIC. And it conflicts with nve(4), so only one should be loaded.
2006-06-26 23:41:07 +00:00
callout_drain(&sc->nfe_stat_ch);
ether_ifdetach(ifp);
Import the nfe (nForce Ethernet) driver from OpenBSD. This is a pure open source NIC driver with a BSD license. These bits were taken from the OpenBSD public CVS repository on 1-May-06.
2006-06-26 17:25:22 +00:00
}
Add a pure open source nForce Ethernet driver, under BSDL. This driver was ported from OpenBSD by Shigeaki Tagashira <shigeaki@se.hiroshima-u.ac.jp> and posted at http://www.se.hiroshima-u.ac.jp/~shigeaki/software/freebsd-nfe.html It was additionally cleaned up by me. It is still a work-in-progress and thus is purposefully not in GENERIC. And it conflicts with nve(4), so only one should be loaded.
2006-06-26 23:41:07 +00:00
Bring overhauled nfe(4) into tree. o s/printf/device_printf/g o Nuke OpenBSDism. o Nuke NetBSD/OpenBSD specific DMA sync operations.(we don't have a way to sync a single descriptor within a DMA map.) o Remove recursive mutex. o bus_dma(9) clean up. o 40bit DMA address support. o Add protection for Rx map load failure. o Fix a long standing bug for watchdog timeout. [1] o Add additional protections, missing Tx completion interrupt, losing start Tx command, for watchdog timeout. o Switch to taskqueue(9) API to handle interrupts. o Use our own timer for watchdog instead of if_watchdog/if_timer interface. o Advertise VLAN header length/capability correctly to upper layer. o Remove excessive kernel stack consumption in nfe_encap(). o Handle highly fragmented mbuf chains correctly. o Enable etherenet address reprogramming with ifconfig(8). o Add ALTQ/TSO, MSI/MSIX support. o Increased Rx ring to 256 descriptors from 128. o Align Tx/Rx descriptor ring on sizeof(struct nfe_desc64) boundary. o Remove alignment restrictions on Tx/Rx buffers. o Rewritten jumbo frame support code. o Add support for hardware assistend VLAN tag insertion/stripping. o Add support for Tx/Rx flow control based on patches from Peer Chen. [2] o Add a routine that detects whether ethernet address swap routines is required. [3] o Add a workaround that take MAC/PHY out of power down mode. o Add suspend/resume support. o style(9) and code clean up. Special thanks to Shigeaki Tagashira, the original porter of nfe(4), who submitted lots of patches, performed uncountable number of regression tests and maintained nfe(4) for a long time. Without his enthusiastic help and support I could never have completed this overhauling task. The only weak point of nfe(4) compared to nve(4) is instability of manual half-duplex media selection on certain hardwares(auto sensing media type should work for all cases, though). This was a long standing bug of nfe(4) and I still have no idea why it doesn't work on some hardwares. Obtained from: OpenBSD [1] Submitted by: Peer Chen < pchen at nvidia dot com > [2], [3] Reviewed by: Shigeaki Tagashira < shigeaki AT se DOT hiroshima-u DOT ac DOT jp > Tested by: Shigeaki Tagashira, current Discussed with: current Silence from: obrien
2007-06-12 02:16:02 +00:00
if (ifp) {
/* restore ethernet address */
if ((sc->nfe_flags & NFE_CORRECT_MACADDR) == 0) {
for (i = 0; i < ETHER_ADDR_LEN; i++) {
eaddr[i] = sc->eaddr[5 - i];
}
} else
bcopy(sc->eaddr, eaddr, ETHER_ADDR_LEN);
nfe_set_macaddr(sc, eaddr);
Add a pure open source nForce Ethernet driver, under BSDL. This driver was ported from OpenBSD by Shigeaki Tagashira <shigeaki@se.hiroshima-u.ac.jp> and posted at http://www.se.hiroshima-u.ac.jp/~shigeaki/software/freebsd-nfe.html It was additionally cleaned up by me. It is still a work-in-progress and thus is purposefully not in GENERIC. And it conflicts with nve(4), so only one should be loaded.
2006-06-26 23:41:07 +00:00
if_free(ifp);
Bring overhauled nfe(4) into tree. o s/printf/device_printf/g o Nuke OpenBSDism. o Nuke NetBSD/OpenBSD specific DMA sync operations.(we don't have a way to sync a single descriptor within a DMA map.) o Remove recursive mutex. o bus_dma(9) clean up. o 40bit DMA address support. o Add protection for Rx map load failure. o Fix a long standing bug for watchdog timeout. [1] o Add additional protections, missing Tx completion interrupt, losing start Tx command, for watchdog timeout. o Switch to taskqueue(9) API to handle interrupts. o Use our own timer for watchdog instead of if_watchdog/if_timer interface. o Advertise VLAN header length/capability correctly to upper layer. o Remove excessive kernel stack consumption in nfe_encap(). o Handle highly fragmented mbuf chains correctly. o Enable etherenet address reprogramming with ifconfig(8). o Add ALTQ/TSO, MSI/MSIX support. o Increased Rx ring to 256 descriptors from 128. o Align Tx/Rx descriptor ring on sizeof(struct nfe_desc64) boundary. o Remove alignment restrictions on Tx/Rx buffers. o Rewritten jumbo frame support code. o Add support for hardware assistend VLAN tag insertion/stripping. o Add support for Tx/Rx flow control based on patches from Peer Chen. [2] o Add a routine that detects whether ethernet address swap routines is required. [3] o Add a workaround that take MAC/PHY out of power down mode. o Add suspend/resume support. o style(9) and code clean up. Special thanks to Shigeaki Tagashira, the original porter of nfe(4), who submitted lots of patches, performed uncountable number of regression tests and maintained nfe(4) for a long time. Without his enthusiastic help and support I could never have completed this overhauling task. The only weak point of nfe(4) compared to nve(4) is instability of manual half-duplex media selection on certain hardwares(auto sensing media type should work for all cases, though). This was a long standing bug of nfe(4) and I still have no idea why it doesn't work on some hardwares. Obtained from: OpenBSD [1] Submitted by: Peer Chen < pchen at nvidia dot com > [2], [3] Reviewed by: Shigeaki Tagashira < shigeaki AT se DOT hiroshima-u DOT ac DOT jp > Tested by: Shigeaki Tagashira, current Discussed with: current Silence from: obrien
2007-06-12 02:16:02 +00:00
}
Add a pure open source nForce Ethernet driver, under BSDL. This driver was ported from OpenBSD by Shigeaki Tagashira <shigeaki@se.hiroshima-u.ac.jp> and posted at http://www.se.hiroshima-u.ac.jp/~shigeaki/software/freebsd-nfe.html It was additionally cleaned up by me. It is still a work-in-progress and thus is purposefully not in GENERIC. And it conflicts with nve(4), so only one should be loaded.
2006-06-26 23:41:07 +00:00
if (sc->nfe_miibus)
device_delete_child(dev, sc->nfe_miibus);
bus_generic_detach(dev);
Bring overhauled nfe(4) into tree. o s/printf/device_printf/g o Nuke OpenBSDism. o Nuke NetBSD/OpenBSD specific DMA sync operations.(we don't have a way to sync a single descriptor within a DMA map.) o Remove recursive mutex. o bus_dma(9) clean up. o 40bit DMA address support. o Add protection for Rx map load failure. o Fix a long standing bug for watchdog timeout. [1] o Add additional protections, missing Tx completion interrupt, losing start Tx command, for watchdog timeout. o Switch to taskqueue(9) API to handle interrupts. o Use our own timer for watchdog instead of if_watchdog/if_timer interface. o Advertise VLAN header length/capability correctly to upper layer. o Remove excessive kernel stack consumption in nfe_encap(). o Handle highly fragmented mbuf chains correctly. o Enable etherenet address reprogramming with ifconfig(8). o Add ALTQ/TSO, MSI/MSIX support. o Increased Rx ring to 256 descriptors from 128. o Align Tx/Rx descriptor ring on sizeof(struct nfe_desc64) boundary. o Remove alignment restrictions on Tx/Rx buffers. o Rewritten jumbo frame support code. o Add support for hardware assistend VLAN tag insertion/stripping. o Add support for Tx/Rx flow control based on patches from Peer Chen. [2] o Add a routine that detects whether ethernet address swap routines is required. [3] o Add a workaround that take MAC/PHY out of power down mode. o Add suspend/resume support. o style(9) and code clean up. Special thanks to Shigeaki Tagashira, the original porter of nfe(4), who submitted lots of patches, performed uncountable number of regression tests and maintained nfe(4) for a long time. Without his enthusiastic help and support I could never have completed this overhauling task. The only weak point of nfe(4) compared to nve(4) is instability of manual half-duplex media selection on certain hardwares(auto sensing media type should work for all cases, though). This was a long standing bug of nfe(4) and I still have no idea why it doesn't work on some hardwares. Obtained from: OpenBSD [1] Submitted by: Peer Chen < pchen at nvidia dot com > [2], [3] Reviewed by: Shigeaki Tagashira < shigeaki AT se DOT hiroshima-u DOT ac DOT jp > Tested by: Shigeaki Tagashira, current Discussed with: current Silence from: obrien
2007-06-12 02:16:02 +00:00
if (sc->nfe_tq != NULL) {
taskqueue_drain(sc->nfe_tq, &sc->nfe_int_task);
taskqueue_free(sc->nfe_tq);
sc->nfe_tq = NULL;
}
for (i = 0; i < NFE_MSI_MESSAGES; i++) {
if (sc->nfe_intrhand[i] != NULL) {
bus_teardown_intr(dev, sc->nfe_irq[i],
sc->nfe_intrhand[i]);
sc->nfe_intrhand[i] = NULL;
}
}
Add a pure open source nForce Ethernet driver, under BSDL. This driver was ported from OpenBSD by Shigeaki Tagashira <shigeaki@se.hiroshima-u.ac.jp> and posted at http://www.se.hiroshima-u.ac.jp/~shigeaki/software/freebsd-nfe.html It was additionally cleaned up by me. It is still a work-in-progress and thus is purposefully not in GENERIC. And it conflicts with nve(4), so only one should be loaded.
2006-06-26 23:41:07 +00:00
Bring overhauled nfe(4) into tree. o s/printf/device_printf/g o Nuke OpenBSDism. o Nuke NetBSD/OpenBSD specific DMA sync operations.(we don't have a way to sync a single descriptor within a DMA map.) o Remove recursive mutex. o bus_dma(9) clean up. o 40bit DMA address support. o Add protection for Rx map load failure. o Fix a long standing bug for watchdog timeout. [1] o Add additional protections, missing Tx completion interrupt, losing start Tx command, for watchdog timeout. o Switch to taskqueue(9) API to handle interrupts. o Use our own timer for watchdog instead of if_watchdog/if_timer interface. o Advertise VLAN header length/capability correctly to upper layer. o Remove excessive kernel stack consumption in nfe_encap(). o Handle highly fragmented mbuf chains correctly. o Enable etherenet address reprogramming with ifconfig(8). o Add ALTQ/TSO, MSI/MSIX support. o Increased Rx ring to 256 descriptors from 128. o Align Tx/Rx descriptor ring on sizeof(struct nfe_desc64) boundary. o Remove alignment restrictions on Tx/Rx buffers. o Rewritten jumbo frame support code. o Add support for hardware assistend VLAN tag insertion/stripping. o Add support for Tx/Rx flow control based on patches from Peer Chen. [2] o Add a routine that detects whether ethernet address swap routines is required. [3] o Add a workaround that take MAC/PHY out of power down mode. o Add suspend/resume support. o style(9) and code clean up. Special thanks to Shigeaki Tagashira, the original porter of nfe(4), who submitted lots of patches, performed uncountable number of regression tests and maintained nfe(4) for a long time. Without his enthusiastic help and support I could never have completed this overhauling task. The only weak point of nfe(4) compared to nve(4) is instability of manual half-duplex media selection on certain hardwares(auto sensing media type should work for all cases, though). This was a long standing bug of nfe(4) and I still have no idea why it doesn't work on some hardwares. Obtained from: OpenBSD [1] Submitted by: Peer Chen < pchen at nvidia dot com > [2], [3] Reviewed by: Shigeaki Tagashira < shigeaki AT se DOT hiroshima-u DOT ac DOT jp > Tested by: Shigeaki Tagashira, current Discussed with: current Silence from: obrien
2007-06-12 02:16:02 +00:00
if (sc->nfe_msi == 0 && sc->nfe_msix == 0) {
if (sc->nfe_irq[0] != NULL)
bus_release_resource(dev, SYS_RES_IRQ, 0,
sc->nfe_irq[0]);
} else {
for (i = 0, rid = 1; i < NFE_MSI_MESSAGES; i++, rid++) {
if (sc->nfe_irq[i] != NULL) {
bus_release_resource(dev, SYS_RES_IRQ, rid,
sc->nfe_irq[i]);
sc->nfe_irq[i] = NULL;
}
}
pci_release_msi(dev);
}
if (sc->nfe_msix_pba_res != NULL) {
bus_release_resource(dev, SYS_RES_MEMORY, PCIR_BAR(3),
sc->nfe_msix_pba_res);
sc->nfe_msix_pba_res = NULL;
}
if (sc->nfe_msix_res != NULL) {
bus_release_resource(dev, SYS_RES_MEMORY, PCIR_BAR(2),
sc->nfe_msix_res);
sc->nfe_msix_res = NULL;
}
if (sc->nfe_res[0] != NULL) {
bus_release_resource(dev, SYS_RES_MEMORY, PCIR_BAR(0),
sc->nfe_res[0]);
sc->nfe_res[0] = NULL;
}
Add a pure open source nForce Ethernet driver, under BSDL. This driver was ported from OpenBSD by Shigeaki Tagashira <shigeaki@se.hiroshima-u.ac.jp> and posted at http://www.se.hiroshima-u.ac.jp/~shigeaki/software/freebsd-nfe.html It was additionally cleaned up by me. It is still a work-in-progress and thus is purposefully not in GENERIC. And it conflicts with nve(4), so only one should be loaded.
2006-06-26 23:41:07 +00:00
nfe_free_tx_ring(sc, &sc->txq);
nfe_free_rx_ring(sc, &sc->rxq);
Bring overhauled nfe(4) into tree. o s/printf/device_printf/g o Nuke OpenBSDism. o Nuke NetBSD/OpenBSD specific DMA sync operations.(we don't have a way to sync a single descriptor within a DMA map.) o Remove recursive mutex. o bus_dma(9) clean up. o 40bit DMA address support. o Add protection for Rx map load failure. o Fix a long standing bug for watchdog timeout. [1] o Add additional protections, missing Tx completion interrupt, losing start Tx command, for watchdog timeout. o Switch to taskqueue(9) API to handle interrupts. o Use our own timer for watchdog instead of if_watchdog/if_timer interface. o Advertise VLAN header length/capability correctly to upper layer. o Remove excessive kernel stack consumption in nfe_encap(). o Handle highly fragmented mbuf chains correctly. o Enable etherenet address reprogramming with ifconfig(8). o Add ALTQ/TSO, MSI/MSIX support. o Increased Rx ring to 256 descriptors from 128. o Align Tx/Rx descriptor ring on sizeof(struct nfe_desc64) boundary. o Remove alignment restrictions on Tx/Rx buffers. o Rewritten jumbo frame support code. o Add support for hardware assistend VLAN tag insertion/stripping. o Add support for Tx/Rx flow control based on patches from Peer Chen. [2] o Add a routine that detects whether ethernet address swap routines is required. [3] o Add a workaround that take MAC/PHY out of power down mode. o Add suspend/resume support. o style(9) and code clean up. Special thanks to Shigeaki Tagashira, the original porter of nfe(4), who submitted lots of patches, performed uncountable number of regression tests and maintained nfe(4) for a long time. Without his enthusiastic help and support I could never have completed this overhauling task. The only weak point of nfe(4) compared to nve(4) is instability of manual half-duplex media selection on certain hardwares(auto sensing media type should work for all cases, though). This was a long standing bug of nfe(4) and I still have no idea why it doesn't work on some hardwares. Obtained from: OpenBSD [1] Submitted by: Peer Chen < pchen at nvidia dot com > [2], [3] Reviewed by: Shigeaki Tagashira < shigeaki AT se DOT hiroshima-u DOT ac DOT jp > Tested by: Shigeaki Tagashira, current Discussed with: current Silence from: obrien
2007-06-12 02:16:02 +00:00
nfe_free_jrx_ring(sc, &sc->jrxq);
Add a pure open source nForce Ethernet driver, under BSDL. This driver was ported from OpenBSD by Shigeaki Tagashira <shigeaki@se.hiroshima-u.ac.jp> and posted at http://www.se.hiroshima-u.ac.jp/~shigeaki/software/freebsd-nfe.html It was additionally cleaned up by me. It is still a work-in-progress and thus is purposefully not in GENERIC. And it conflicts with nve(4), so only one should be loaded.
2006-06-26 23:41:07 +00:00
Bring overhauled nfe(4) into tree. o s/printf/device_printf/g o Nuke OpenBSDism. o Nuke NetBSD/OpenBSD specific DMA sync operations.(we don't have a way to sync a single descriptor within a DMA map.) o Remove recursive mutex. o bus_dma(9) clean up. o 40bit DMA address support. o Add protection for Rx map load failure. o Fix a long standing bug for watchdog timeout. [1] o Add additional protections, missing Tx completion interrupt, losing start Tx command, for watchdog timeout. o Switch to taskqueue(9) API to handle interrupts. o Use our own timer for watchdog instead of if_watchdog/if_timer interface. o Advertise VLAN header length/capability correctly to upper layer. o Remove excessive kernel stack consumption in nfe_encap(). o Handle highly fragmented mbuf chains correctly. o Enable etherenet address reprogramming with ifconfig(8). o Add ALTQ/TSO, MSI/MSIX support. o Increased Rx ring to 256 descriptors from 128. o Align Tx/Rx descriptor ring on sizeof(struct nfe_desc64) boundary. o Remove alignment restrictions on Tx/Rx buffers. o Rewritten jumbo frame support code. o Add support for hardware assistend VLAN tag insertion/stripping. o Add support for Tx/Rx flow control based on patches from Peer Chen. [2] o Add a routine that detects whether ethernet address swap routines is required. [3] o Add a workaround that take MAC/PHY out of power down mode. o Add suspend/resume support. o style(9) and code clean up. Special thanks to Shigeaki Tagashira, the original porter of nfe(4), who submitted lots of patches, performed uncountable number of regression tests and maintained nfe(4) for a long time. Without his enthusiastic help and support I could never have completed this overhauling task. The only weak point of nfe(4) compared to nve(4) is instability of manual half-duplex media selection on certain hardwares(auto sensing media type should work for all cases, though). This was a long standing bug of nfe(4) and I still have no idea why it doesn't work on some hardwares. Obtained from: OpenBSD [1] Submitted by: Peer Chen < pchen at nvidia dot com > [2], [3] Reviewed by: Shigeaki Tagashira < shigeaki AT se DOT hiroshima-u DOT ac DOT jp > Tested by: Shigeaki Tagashira, current Discussed with: current Silence from: obrien
2007-06-12 02:16:02 +00:00
if (sc->nfe_parent_tag) {
Add a pure open source nForce Ethernet driver, under BSDL. This driver was ported from OpenBSD by Shigeaki Tagashira <shigeaki@se.hiroshima-u.ac.jp> and posted at http://www.se.hiroshima-u.ac.jp/~shigeaki/software/freebsd-nfe.html It was additionally cleaned up by me. It is still a work-in-progress and thus is purposefully not in GENERIC. And it conflicts with nve(4), so only one should be loaded.
2006-06-26 23:41:07 +00:00
bus_dma_tag_destroy(sc->nfe_parent_tag);
Bring overhauled nfe(4) into tree. o s/printf/device_printf/g o Nuke OpenBSDism. o Nuke NetBSD/OpenBSD specific DMA sync operations.(we don't have a way to sync a single descriptor within a DMA map.) o Remove recursive mutex. o bus_dma(9) clean up. o 40bit DMA address support. o Add protection for Rx map load failure. o Fix a long standing bug for watchdog timeout. [1] o Add additional protections, missing Tx completion interrupt, losing start Tx command, for watchdog timeout. o Switch to taskqueue(9) API to handle interrupts. o Use our own timer for watchdog instead of if_watchdog/if_timer interface. o Advertise VLAN header length/capability correctly to upper layer. o Remove excessive kernel stack consumption in nfe_encap(). o Handle highly fragmented mbuf chains correctly. o Enable etherenet address reprogramming with ifconfig(8). o Add ALTQ/TSO, MSI/MSIX support. o Increased Rx ring to 256 descriptors from 128. o Align Tx/Rx descriptor ring on sizeof(struct nfe_desc64) boundary. o Remove alignment restrictions on Tx/Rx buffers. o Rewritten jumbo frame support code. o Add support for hardware assistend VLAN tag insertion/stripping. o Add support for Tx/Rx flow control based on patches from Peer Chen. [2] o Add a routine that detects whether ethernet address swap routines is required. [3] o Add a workaround that take MAC/PHY out of power down mode. o Add suspend/resume support. o style(9) and code clean up. Special thanks to Shigeaki Tagashira, the original porter of nfe(4), who submitted lots of patches, performed uncountable number of regression tests and maintained nfe(4) for a long time. Without his enthusiastic help and support I could never have completed this overhauling task. The only weak point of nfe(4) compared to nve(4) is instability of manual half-duplex media selection on certain hardwares(auto sensing media type should work for all cases, though). This was a long standing bug of nfe(4) and I still have no idea why it doesn't work on some hardwares. Obtained from: OpenBSD [1] Submitted by: Peer Chen < pchen at nvidia dot com > [2], [3] Reviewed by: Shigeaki Tagashira < shigeaki AT se DOT hiroshima-u DOT ac DOT jp > Tested by: Shigeaki Tagashira, current Discussed with: current Silence from: obrien
2007-06-12 02:16:02 +00:00
sc->nfe_parent_tag = NULL;
}
Add a pure open source nForce Ethernet driver, under BSDL. This driver was ported from OpenBSD by Shigeaki Tagashira <shigeaki@se.hiroshima-u.ac.jp> and posted at http://www.se.hiroshima-u.ac.jp/~shigeaki/software/freebsd-nfe.html It was additionally cleaned up by me. It is still a work-in-progress and thus is purposefully not in GENERIC. And it conflicts with nve(4), so only one should be loaded.
2006-06-26 23:41:07 +00:00
mtx_destroy(&sc->nfe_mtx);
return (0);
Import the nfe (nForce Ethernet) driver from OpenBSD. This is a pure open source NIC driver with a BSD license. These bits were taken from the OpenBSD public CVS repository on 1-May-06.
2006-06-26 17:25:22 +00:00
}
Add a pure open source nForce Ethernet driver, under BSDL. This driver was ported from OpenBSD by Shigeaki Tagashira <shigeaki@se.hiroshima-u.ac.jp> and posted at http://www.se.hiroshima-u.ac.jp/~shigeaki/software/freebsd-nfe.html It was additionally cleaned up by me. It is still a work-in-progress and thus is purposefully not in GENERIC. And it conflicts with nve(4), so only one should be loaded.
2006-06-26 23:41:07 +00:00
Bring overhauled nfe(4) into tree. o s/printf/device_printf/g o Nuke OpenBSDism. o Nuke NetBSD/OpenBSD specific DMA sync operations.(we don't have a way to sync a single descriptor within a DMA map.) o Remove recursive mutex. o bus_dma(9) clean up. o 40bit DMA address support. o Add protection for Rx map load failure. o Fix a long standing bug for watchdog timeout. [1] o Add additional protections, missing Tx completion interrupt, losing start Tx command, for watchdog timeout. o Switch to taskqueue(9) API to handle interrupts. o Use our own timer for watchdog instead of if_watchdog/if_timer interface. o Advertise VLAN header length/capability correctly to upper layer. o Remove excessive kernel stack consumption in nfe_encap(). o Handle highly fragmented mbuf chains correctly. o Enable etherenet address reprogramming with ifconfig(8). o Add ALTQ/TSO, MSI/MSIX support. o Increased Rx ring to 256 descriptors from 128. o Align Tx/Rx descriptor ring on sizeof(struct nfe_desc64) boundary. o Remove alignment restrictions on Tx/Rx buffers. o Rewritten jumbo frame support code. o Add support for hardware assistend VLAN tag insertion/stripping. o Add support for Tx/Rx flow control based on patches from Peer Chen. [2] o Add a routine that detects whether ethernet address swap routines is required. [3] o Add a workaround that take MAC/PHY out of power down mode. o Add suspend/resume support. o style(9) and code clean up. Special thanks to Shigeaki Tagashira, the original porter of nfe(4), who submitted lots of patches, performed uncountable number of regression tests and maintained nfe(4) for a long time. Without his enthusiastic help and support I could never have completed this overhauling task. The only weak point of nfe(4) compared to nve(4) is instability of manual half-duplex media selection on certain hardwares(auto sensing media type should work for all cases, though). This was a long standing bug of nfe(4) and I still have no idea why it doesn't work on some hardwares. Obtained from: OpenBSD [1] Submitted by: Peer Chen < pchen at nvidia dot com > [2], [3] Reviewed by: Shigeaki Tagashira < shigeaki AT se DOT hiroshima-u DOT ac DOT jp > Tested by: Shigeaki Tagashira, current Discussed with: current Silence from: obrien
2007-06-12 02:16:02 +00:00
static int
nfe_suspend(device_t dev)
{
struct nfe_softc *sc;
sc = device_get_softc(dev);
NFE_LOCK(sc);
nfe_stop(sc->nfe_ifp);
Add basic WOL support for MCP ethernet controllers. It seems the controller does not perform automatic switching from 1000Mbps link to 10/100Mbps link when WOL is activated. Implement establishing 10/100Mps link with auto-negotiation in driver. Link status change handler was modified to remove taskqueue based approach since driver now needs synchronous handling for link establishment. Submitted by: Yamagi Burmeister (lists <> yamagi.org ) (initial version) Tested by: Yamagi Burmeister (lists <> yamagi.org ) MFC after: 1 week
2010-11-11 18:08:50 +00:00
nfe_set_wol(sc);
Bring overhauled nfe(4) into tree. o s/printf/device_printf/g o Nuke OpenBSDism. o Nuke NetBSD/OpenBSD specific DMA sync operations.(we don't have a way to sync a single descriptor within a DMA map.) o Remove recursive mutex. o bus_dma(9) clean up. o 40bit DMA address support. o Add protection for Rx map load failure. o Fix a long standing bug for watchdog timeout. [1] o Add additional protections, missing Tx completion interrupt, losing start Tx command, for watchdog timeout. o Switch to taskqueue(9) API to handle interrupts. o Use our own timer for watchdog instead of if_watchdog/if_timer interface. o Advertise VLAN header length/capability correctly to upper layer. o Remove excessive kernel stack consumption in nfe_encap(). o Handle highly fragmented mbuf chains correctly. o Enable etherenet address reprogramming with ifconfig(8). o Add ALTQ/TSO, MSI/MSIX support. o Increased Rx ring to 256 descriptors from 128. o Align Tx/Rx descriptor ring on sizeof(struct nfe_desc64) boundary. o Remove alignment restrictions on Tx/Rx buffers. o Rewritten jumbo frame support code. o Add support for hardware assistend VLAN tag insertion/stripping. o Add support for Tx/Rx flow control based on patches from Peer Chen. [2] o Add a routine that detects whether ethernet address swap routines is required. [3] o Add a workaround that take MAC/PHY out of power down mode. o Add suspend/resume support. o style(9) and code clean up. Special thanks to Shigeaki Tagashira, the original porter of nfe(4), who submitted lots of patches, performed uncountable number of regression tests and maintained nfe(4) for a long time. Without his enthusiastic help and support I could never have completed this overhauling task. The only weak point of nfe(4) compared to nve(4) is instability of manual half-duplex media selection on certain hardwares(auto sensing media type should work for all cases, though). This was a long standing bug of nfe(4) and I still have no idea why it doesn't work on some hardwares. Obtained from: OpenBSD [1] Submitted by: Peer Chen < pchen at nvidia dot com > [2], [3] Reviewed by: Shigeaki Tagashira < shigeaki AT se DOT hiroshima-u DOT ac DOT jp > Tested by: Shigeaki Tagashira, current Discussed with: current Silence from: obrien
2007-06-12 02:16:02 +00:00
sc->nfe_suspended = 1;
NFE_UNLOCK(sc);
return (0);
}
static int
nfe_resume(device_t dev)
{
struct nfe_softc *sc;
struct ifnet *ifp;
sc = device_get_softc(dev);
NFE_LOCK(sc);
Add basic WOL support for MCP ethernet controllers. It seems the controller does not perform automatic switching from 1000Mbps link to 10/100Mbps link when WOL is activated. Implement establishing 10/100Mps link with auto-negotiation in driver. Link status change handler was modified to remove taskqueue based approach since driver now needs synchronous handling for link establishment. Submitted by: Yamagi Burmeister (lists <> yamagi.org ) (initial version) Tested by: Yamagi Burmeister (lists <> yamagi.org ) MFC after: 1 week
2010-11-11 18:08:50 +00:00
nfe_power(sc);
Bring overhauled nfe(4) into tree. o s/printf/device_printf/g o Nuke OpenBSDism. o Nuke NetBSD/OpenBSD specific DMA sync operations.(we don't have a way to sync a single descriptor within a DMA map.) o Remove recursive mutex. o bus_dma(9) clean up. o 40bit DMA address support. o Add protection for Rx map load failure. o Fix a long standing bug for watchdog timeout. [1] o Add additional protections, missing Tx completion interrupt, losing start Tx command, for watchdog timeout. o Switch to taskqueue(9) API to handle interrupts. o Use our own timer for watchdog instead of if_watchdog/if_timer interface. o Advertise VLAN header length/capability correctly to upper layer. o Remove excessive kernel stack consumption in nfe_encap(). o Handle highly fragmented mbuf chains correctly. o Enable etherenet address reprogramming with ifconfig(8). o Add ALTQ/TSO, MSI/MSIX support. o Increased Rx ring to 256 descriptors from 128. o Align Tx/Rx descriptor ring on sizeof(struct nfe_desc64) boundary. o Remove alignment restrictions on Tx/Rx buffers. o Rewritten jumbo frame support code. o Add support for hardware assistend VLAN tag insertion/stripping. o Add support for Tx/Rx flow control based on patches from Peer Chen. [2] o Add a routine that detects whether ethernet address swap routines is required. [3] o Add a workaround that take MAC/PHY out of power down mode. o Add suspend/resume support. o style(9) and code clean up. Special thanks to Shigeaki Tagashira, the original porter of nfe(4), who submitted lots of patches, performed uncountable number of regression tests and maintained nfe(4) for a long time. Without his enthusiastic help and support I could never have completed this overhauling task. The only weak point of nfe(4) compared to nve(4) is instability of manual half-duplex media selection on certain hardwares(auto sensing media type should work for all cases, though). This was a long standing bug of nfe(4) and I still have no idea why it doesn't work on some hardwares. Obtained from: OpenBSD [1] Submitted by: Peer Chen < pchen at nvidia dot com > [2], [3] Reviewed by: Shigeaki Tagashira < shigeaki AT se DOT hiroshima-u DOT ac DOT jp > Tested by: Shigeaki Tagashira, current Discussed with: current Silence from: obrien
2007-06-12 02:16:02 +00:00
ifp = sc->nfe_ifp;
if (ifp->if_flags & IFF_UP)
nfe_init_locked(sc);
sc->nfe_suspended = 0;
NFE_UNLOCK(sc);
return (0);
}
P5N32-SLI PREMIUM from ASUSTeK is known to have MSI/MSI-X issue such that nfe(4) does not work with MSI-X. When MSI-X support was introduced, I remember MCP55 controller worked without problems so the issue could be either PCI bridge or BIOS issue. But I also noticed snd_hda(4) disabled MSI on all MCP55 chipset so I'm still not sure this is generic issue of MCP55 chipset. If this was PCI bridge issue we would have added it to a system wide black-list table but it's not clear to me at this moment whether it was caused by either broken BIOS or silicon bug of MCP55 chipset. To workaround the issue, maintain a MSI/MSI-X black-list table in driver and lookup base board manufacturer and product name from the table before attempting to use MSI-X. If driver find an matching entry, nfe(4) will not use MSI/MSI-X and fall back on traditional INTx mode. This approach should be the last resort since it relies on smbios and if another instance of MSI/MSI-X breakage is reported with different maker/product, we may have to get the PCI bridge black-listed instead of adding an new entry. PR: kern/152150
2010-11-14 23:37:43 +00:00
static int
nfe_can_use_msix(struct nfe_softc *sc)
{
static struct msix_blacklist {
char *maker;
char *product;
} msix_blacklists[] = {
{ "ASUSTeK Computer INC.", "P5N32-SLI PREMIUM" }
};
struct msix_blacklist *mblp;
char *maker, *product;
Plug memory leakage introduced in r215327. Submitted by: jkim
2010-11-15 22:05:08 +00:00
int count, n, use_msix;
P5N32-SLI PREMIUM from ASUSTeK is known to have MSI/MSI-X issue such that nfe(4) does not work with MSI-X. When MSI-X support was introduced, I remember MCP55 controller worked without problems so the issue could be either PCI bridge or BIOS issue. But I also noticed snd_hda(4) disabled MSI on all MCP55 chipset so I'm still not sure this is generic issue of MCP55 chipset. If this was PCI bridge issue we would have added it to a system wide black-list table but it's not clear to me at this moment whether it was caused by either broken BIOS or silicon bug of MCP55 chipset. To workaround the issue, maintain a MSI/MSI-X black-list table in driver and lookup base board manufacturer and product name from the table before attempting to use MSI-X. If driver find an matching entry, nfe(4) will not use MSI/MSI-X and fall back on traditional INTx mode. This approach should be the last resort since it relies on smbios and if another instance of MSI/MSI-X breakage is reported with different maker/product, we may have to get the PCI bridge black-listed instead of adding an new entry. PR: kern/152150
2010-11-14 23:37:43 +00:00
/*
* Search base board manufacturer and product name table
* to see this system has a known MSI/MSI-X issue.
*/
maker = getenv("smbios.planar.maker");
product = getenv("smbios.planar.product");
Plug memory leakage introduced in r215327. Submitted by: jkim
2010-11-15 22:05:08 +00:00
use_msix = 1;
P5N32-SLI PREMIUM from ASUSTeK is known to have MSI/MSI-X issue such that nfe(4) does not work with MSI-X. When MSI-X support was introduced, I remember MCP55 controller worked without problems so the issue could be either PCI bridge or BIOS issue. But I also noticed snd_hda(4) disabled MSI on all MCP55 chipset so I'm still not sure this is generic issue of MCP55 chipset. If this was PCI bridge issue we would have added it to a system wide black-list table but it's not clear to me at this moment whether it was caused by either broken BIOS or silicon bug of MCP55 chipset. To workaround the issue, maintain a MSI/MSI-X black-list table in driver and lookup base board manufacturer and product name from the table before attempting to use MSI-X. If driver find an matching entry, nfe(4) will not use MSI/MSI-X and fall back on traditional INTx mode. This approach should be the last resort since it relies on smbios and if another instance of MSI/MSI-X breakage is reported with different maker/product, we may have to get the PCI bridge black-listed instead of adding an new entry. PR: kern/152150
2010-11-14 23:37:43 +00:00
if (maker != NULL && product != NULL) {
count = sizeof(msix_blacklists) / sizeof(msix_blacklists[0]);
mblp = msix_blacklists;
for (n = 0; n < count; n++) {
if (strcmp(maker, mblp->maker) == 0 &&
Plug memory leakage introduced in r215327. Submitted by: jkim
2010-11-15 22:05:08 +00:00
strcmp(product, mblp->product) == 0) {
use_msix = 0;
break;
}
P5N32-SLI PREMIUM from ASUSTeK is known to have MSI/MSI-X issue such that nfe(4) does not work with MSI-X. When MSI-X support was introduced, I remember MCP55 controller worked without problems so the issue could be either PCI bridge or BIOS issue. But I also noticed snd_hda(4) disabled MSI on all MCP55 chipset so I'm still not sure this is generic issue of MCP55 chipset. If this was PCI bridge issue we would have added it to a system wide black-list table but it's not clear to me at this moment whether it was caused by either broken BIOS or silicon bug of MCP55 chipset. To workaround the issue, maintain a MSI/MSI-X black-list table in driver and lookup base board manufacturer and product name from the table before attempting to use MSI-X. If driver find an matching entry, nfe(4) will not use MSI/MSI-X and fall back on traditional INTx mode. This approach should be the last resort since it relies on smbios and if another instance of MSI/MSI-X breakage is reported with different maker/product, we may have to get the PCI bridge black-listed instead of adding an new entry. PR: kern/152150
2010-11-14 23:37:43 +00:00
mblp++;
}
}
Plug memory leakage introduced in r215327. Submitted by: jkim
2010-11-15 22:05:08 +00:00
if (maker != NULL)
freeenv(maker);
if (product != NULL)
freeenv(product);
P5N32-SLI PREMIUM from ASUSTeK is known to have MSI/MSI-X issue such that nfe(4) does not work with MSI-X. When MSI-X support was introduced, I remember MCP55 controller worked without problems so the issue could be either PCI bridge or BIOS issue. But I also noticed snd_hda(4) disabled MSI on all MCP55 chipset so I'm still not sure this is generic issue of MCP55 chipset. If this was PCI bridge issue we would have added it to a system wide black-list table but it's not clear to me at this moment whether it was caused by either broken BIOS or silicon bug of MCP55 chipset. To workaround the issue, maintain a MSI/MSI-X black-list table in driver and lookup base board manufacturer and product name from the table before attempting to use MSI-X. If driver find an matching entry, nfe(4) will not use MSI/MSI-X and fall back on traditional INTx mode. This approach should be the last resort since it relies on smbios and if another instance of MSI/MSI-X breakage is reported with different maker/product, we may have to get the PCI bridge black-listed instead of adding an new entry. PR: kern/152150
2010-11-14 23:37:43 +00:00
Plug memory leakage introduced in r215327. Submitted by: jkim
2010-11-15 22:05:08 +00:00
return (use_msix);
P5N32-SLI PREMIUM from ASUSTeK is known to have MSI/MSI-X issue such that nfe(4) does not work with MSI-X. When MSI-X support was introduced, I remember MCP55 controller worked without problems so the issue could be either PCI bridge or BIOS issue. But I also noticed snd_hda(4) disabled MSI on all MCP55 chipset so I'm still not sure this is generic issue of MCP55 chipset. If this was PCI bridge issue we would have added it to a system wide black-list table but it's not clear to me at this moment whether it was caused by either broken BIOS or silicon bug of MCP55 chipset. To workaround the issue, maintain a MSI/MSI-X black-list table in driver and lookup base board manufacturer and product name from the table before attempting to use MSI-X. If driver find an matching entry, nfe(4) will not use MSI/MSI-X and fall back on traditional INTx mode. This approach should be the last resort since it relies on smbios and if another instance of MSI/MSI-X breakage is reported with different maker/product, we may have to get the PCI bridge black-listed instead of adding an new entry. PR: kern/152150
2010-11-14 23:37:43 +00:00
}
Bring overhauled nfe(4) into tree. o s/printf/device_printf/g o Nuke OpenBSDism. o Nuke NetBSD/OpenBSD specific DMA sync operations.(we don't have a way to sync a single descriptor within a DMA map.) o Remove recursive mutex. o bus_dma(9) clean up. o 40bit DMA address support. o Add protection for Rx map load failure. o Fix a long standing bug for watchdog timeout. [1] o Add additional protections, missing Tx completion interrupt, losing start Tx command, for watchdog timeout. o Switch to taskqueue(9) API to handle interrupts. o Use our own timer for watchdog instead of if_watchdog/if_timer interface. o Advertise VLAN header length/capability correctly to upper layer. o Remove excessive kernel stack consumption in nfe_encap(). o Handle highly fragmented mbuf chains correctly. o Enable etherenet address reprogramming with ifconfig(8). o Add ALTQ/TSO, MSI/MSIX support. o Increased Rx ring to 256 descriptors from 128. o Align Tx/Rx descriptor ring on sizeof(struct nfe_desc64) boundary. o Remove alignment restrictions on Tx/Rx buffers. o Rewritten jumbo frame support code. o Add support for hardware assistend VLAN tag insertion/stripping. o Add support for Tx/Rx flow control based on patches from Peer Chen. [2] o Add a routine that detects whether ethernet address swap routines is required. [3] o Add a workaround that take MAC/PHY out of power down mode. o Add suspend/resume support. o style(9) and code clean up. Special thanks to Shigeaki Tagashira, the original porter of nfe(4), who submitted lots of patches, performed uncountable number of regression tests and maintained nfe(4) for a long time. Without his enthusiastic help and support I could never have completed this overhauling task. The only weak point of nfe(4) compared to nve(4) is instability of manual half-duplex media selection on certain hardwares(auto sensing media type should work for all cases, though). This was a long standing bug of nfe(4) and I still have no idea why it doesn't work on some hardwares. Obtained from: OpenBSD [1] Submitted by: Peer Chen < pchen at nvidia dot com > [2], [3] Reviewed by: Shigeaki Tagashira < shigeaki AT se DOT hiroshima-u DOT ac DOT jp > Tested by: Shigeaki Tagashira, current Discussed with: current Silence from: obrien
2007-06-12 02:16:02 +00:00
/* Take PHY/NIC out of powerdown, from Linux */
static void
nfe_power(struct nfe_softc *sc)
{
uint32_t pwr;
if ((sc->nfe_flags & NFE_PWR_MGMT) == 0)
return;
NFE_WRITE(sc, NFE_RXTX_CTL, NFE_RXTX_RESET | NFE_RXTX_BIT2);
NFE_WRITE(sc, NFE_MAC_RESET, NFE_MAC_RESET_MAGIC);
DELAY(100);
NFE_WRITE(sc, NFE_MAC_RESET, 0);
DELAY(100);
NFE_WRITE(sc, NFE_RXTX_CTL, NFE_RXTX_BIT2);
pwr = NFE_READ(sc, NFE_PWR2_CTL);
pwr &= ~NFE_PWR2_WAKEUP_MASK;
if (sc->nfe_revid >= 0xa3 &&
(sc->nfe_devid == PCI_PRODUCT_NVIDIA_NFORCE430_LAN1 ||
sc->nfe_devid == PCI_PRODUCT_NVIDIA_NFORCE430_LAN2))
pwr |= NFE_PWR2_REVA3;
NFE_WRITE(sc, NFE_PWR2_CTL, pwr);
}
Add a pure open source nForce Ethernet driver, under BSDL. This driver was ported from OpenBSD by Shigeaki Tagashira <shigeaki@se.hiroshima-u.ac.jp> and posted at http://www.se.hiroshima-u.ac.jp/~shigeaki/software/freebsd-nfe.html It was additionally cleaned up by me. It is still a work-in-progress and thus is purposefully not in GENERIC. And it conflicts with nve(4), so only one should be loaded.
2006-06-26 23:41:07 +00:00
static void
nfe_miibus_statchg(device_t dev)
Bring overhauled nfe(4) into tree. o s/printf/device_printf/g o Nuke OpenBSDism. o Nuke NetBSD/OpenBSD specific DMA sync operations.(we don't have a way to sync a single descriptor within a DMA map.) o Remove recursive mutex. o bus_dma(9) clean up. o 40bit DMA address support. o Add protection for Rx map load failure. o Fix a long standing bug for watchdog timeout. [1] o Add additional protections, missing Tx completion interrupt, losing start Tx command, for watchdog timeout. o Switch to taskqueue(9) API to handle interrupts. o Use our own timer for watchdog instead of if_watchdog/if_timer interface. o Advertise VLAN header length/capability correctly to upper layer. o Remove excessive kernel stack consumption in nfe_encap(). o Handle highly fragmented mbuf chains correctly. o Enable etherenet address reprogramming with ifconfig(8). o Add ALTQ/TSO, MSI/MSIX support. o Increased Rx ring to 256 descriptors from 128. o Align Tx/Rx descriptor ring on sizeof(struct nfe_desc64) boundary. o Remove alignment restrictions on Tx/Rx buffers. o Rewritten jumbo frame support code. o Add support for hardware assistend VLAN tag insertion/stripping. o Add support for Tx/Rx flow control based on patches from Peer Chen. [2] o Add a routine that detects whether ethernet address swap routines is required. [3] o Add a workaround that take MAC/PHY out of power down mode. o Add suspend/resume support. o style(9) and code clean up. Special thanks to Shigeaki Tagashira, the original porter of nfe(4), who submitted lots of patches, performed uncountable number of regression tests and maintained nfe(4) for a long time. Without his enthusiastic help and support I could never have completed this overhauling task. The only weak point of nfe(4) compared to nve(4) is instability of manual half-duplex media selection on certain hardwares(auto sensing media type should work for all cases, though). This was a long standing bug of nfe(4) and I still have no idea why it doesn't work on some hardwares. Obtained from: OpenBSD [1] Submitted by: Peer Chen < pchen at nvidia dot com > [2], [3] Reviewed by: Shigeaki Tagashira < shigeaki AT se DOT hiroshima-u DOT ac DOT jp > Tested by: Shigeaki Tagashira, current Discussed with: current Silence from: obrien
2007-06-12 02:16:02 +00:00
{
struct nfe_softc *sc;
struct mii_data *mii;
struct ifnet *ifp;
Add basic WOL support for MCP ethernet controllers. It seems the controller does not perform automatic switching from 1000Mbps link to 10/100Mbps link when WOL is activated. Implement establishing 10/100Mps link with auto-negotiation in driver. Link status change handler was modified to remove taskqueue based approach since driver now needs synchronous handling for link establishment. Submitted by: Yamagi Burmeister (lists <> yamagi.org ) (initial version) Tested by: Yamagi Burmeister (lists <> yamagi.org ) MFC after: 1 week
2010-11-11 18:08:50 +00:00
uint32_t rxctl, txctl;
Bring overhauled nfe(4) into tree. o s/printf/device_printf/g o Nuke OpenBSDism. o Nuke NetBSD/OpenBSD specific DMA sync operations.(we don't have a way to sync a single descriptor within a DMA map.) o Remove recursive mutex. o bus_dma(9) clean up. o 40bit DMA address support. o Add protection for Rx map load failure. o Fix a long standing bug for watchdog timeout. [1] o Add additional protections, missing Tx completion interrupt, losing start Tx command, for watchdog timeout. o Switch to taskqueue(9) API to handle interrupts. o Use our own timer for watchdog instead of if_watchdog/if_timer interface. o Advertise VLAN header length/capability correctly to upper layer. o Remove excessive kernel stack consumption in nfe_encap(). o Handle highly fragmented mbuf chains correctly. o Enable etherenet address reprogramming with ifconfig(8). o Add ALTQ/TSO, MSI/MSIX support. o Increased Rx ring to 256 descriptors from 128. o Align Tx/Rx descriptor ring on sizeof(struct nfe_desc64) boundary. o Remove alignment restrictions on Tx/Rx buffers. o Rewritten jumbo frame support code. o Add support for hardware assistend VLAN tag insertion/stripping. o Add support for Tx/Rx flow control based on patches from Peer Chen. [2] o Add a routine that detects whether ethernet address swap routines is required. [3] o Add a workaround that take MAC/PHY out of power down mode. o Add suspend/resume support. o style(9) and code clean up. Special thanks to Shigeaki Tagashira, the original porter of nfe(4), who submitted lots of patches, performed uncountable number of regression tests and maintained nfe(4) for a long time. Without his enthusiastic help and support I could never have completed this overhauling task. The only weak point of nfe(4) compared to nve(4) is instability of manual half-duplex media selection on certain hardwares(auto sensing media type should work for all cases, though). This was a long standing bug of nfe(4) and I still have no idea why it doesn't work on some hardwares. Obtained from: OpenBSD [1] Submitted by: Peer Chen < pchen at nvidia dot com > [2], [3] Reviewed by: Shigeaki Tagashira < shigeaki AT se DOT hiroshima-u DOT ac DOT jp > Tested by: Shigeaki Tagashira, current Discussed with: current Silence from: obrien
2007-06-12 02:16:02 +00:00
Add basic WOL support for MCP ethernet controllers. It seems the controller does not perform automatic switching from 1000Mbps link to 10/100Mbps link when WOL is activated. Implement establishing 10/100Mps link with auto-negotiation in driver. Link status change handler was modified to remove taskqueue based approach since driver now needs synchronous handling for link establishment. Submitted by: Yamagi Burmeister (lists <> yamagi.org ) (initial version) Tested by: Yamagi Burmeister (lists <> yamagi.org ) MFC after: 1 week
2010-11-11 18:08:50 +00:00
sc = device_get_softc(dev);
Bring overhauled nfe(4) into tree. o s/printf/device_printf/g o Nuke OpenBSDism. o Nuke NetBSD/OpenBSD specific DMA sync operations.(we don't have a way to sync a single descriptor within a DMA map.) o Remove recursive mutex. o bus_dma(9) clean up. o 40bit DMA address support. o Add protection for Rx map load failure. o Fix a long standing bug for watchdog timeout. [1] o Add additional protections, missing Tx completion interrupt, losing start Tx command, for watchdog timeout. o Switch to taskqueue(9) API to handle interrupts. o Use our own timer for watchdog instead of if_watchdog/if_timer interface. o Advertise VLAN header length/capability correctly to upper layer. o Remove excessive kernel stack consumption in nfe_encap(). o Handle highly fragmented mbuf chains correctly. o Enable etherenet address reprogramming with ifconfig(8). o Add ALTQ/TSO, MSI/MSIX support. o Increased Rx ring to 256 descriptors from 128. o Align Tx/Rx descriptor ring on sizeof(struct nfe_desc64) boundary. o Remove alignment restrictions on Tx/Rx buffers. o Rewritten jumbo frame support code. o Add support for hardware assistend VLAN tag insertion/stripping. o Add support for Tx/Rx flow control based on patches from Peer Chen. [2] o Add a routine that detects whether ethernet address swap routines is required. [3] o Add a workaround that take MAC/PHY out of power down mode. o Add suspend/resume support. o style(9) and code clean up. Special thanks to Shigeaki Tagashira, the original porter of nfe(4), who submitted lots of patches, performed uncountable number of regression tests and maintained nfe(4) for a long time. Without his enthusiastic help and support I could never have completed this overhauling task. The only weak point of nfe(4) compared to nve(4) is instability of manual half-duplex media selection on certain hardwares(auto sensing media type should work for all cases, though). This was a long standing bug of nfe(4) and I still have no idea why it doesn't work on some hardwares. Obtained from: OpenBSD [1] Submitted by: Peer Chen < pchen at nvidia dot com > [2], [3] Reviewed by: Shigeaki Tagashira < shigeaki AT se DOT hiroshima-u DOT ac DOT jp > Tested by: Shigeaki Tagashira, current Discussed with: current Silence from: obrien
2007-06-12 02:16:02 +00:00
Add a pure open source nForce Ethernet driver, under BSDL. This driver was ported from OpenBSD by Shigeaki Tagashira <shigeaki@se.hiroshima-u.ac.jp> and posted at http://www.se.hiroshima-u.ac.jp/~shigeaki/software/freebsd-nfe.html It was additionally cleaned up by me. It is still a work-in-progress and thus is purposefully not in GENERIC. And it conflicts with nve(4), so only one should be loaded.
2006-06-26 23:41:07 +00:00
mii = device_get_softc(sc->nfe_miibus);
Bring overhauled nfe(4) into tree. o s/printf/device_printf/g o Nuke OpenBSDism. o Nuke NetBSD/OpenBSD specific DMA sync operations.(we don't have a way to sync a single descriptor within a DMA map.) o Remove recursive mutex. o bus_dma(9) clean up. o 40bit DMA address support. o Add protection for Rx map load failure. o Fix a long standing bug for watchdog timeout. [1] o Add additional protections, missing Tx completion interrupt, losing start Tx command, for watchdog timeout. o Switch to taskqueue(9) API to handle interrupts. o Use our own timer for watchdog instead of if_watchdog/if_timer interface. o Advertise VLAN header length/capability correctly to upper layer. o Remove excessive kernel stack consumption in nfe_encap(). o Handle highly fragmented mbuf chains correctly. o Enable etherenet address reprogramming with ifconfig(8). o Add ALTQ/TSO, MSI/MSIX support. o Increased Rx ring to 256 descriptors from 128. o Align Tx/Rx descriptor ring on sizeof(struct nfe_desc64) boundary. o Remove alignment restrictions on Tx/Rx buffers. o Rewritten jumbo frame support code. o Add support for hardware assistend VLAN tag insertion/stripping. o Add support for Tx/Rx flow control based on patches from Peer Chen. [2] o Add a routine that detects whether ethernet address swap routines is required. [3] o Add a workaround that take MAC/PHY out of power down mode. o Add suspend/resume support. o style(9) and code clean up. Special thanks to Shigeaki Tagashira, the original porter of nfe(4), who submitted lots of patches, performed uncountable number of regression tests and maintained nfe(4) for a long time. Without his enthusiastic help and support I could never have completed this overhauling task. The only weak point of nfe(4) compared to nve(4) is instability of manual half-duplex media selection on certain hardwares(auto sensing media type should work for all cases, though). This was a long standing bug of nfe(4) and I still have no idea why it doesn't work on some hardwares. Obtained from: OpenBSD [1] Submitted by: Peer Chen < pchen at nvidia dot com > [2], [3] Reviewed by: Shigeaki Tagashira < shigeaki AT se DOT hiroshima-u DOT ac DOT jp > Tested by: Shigeaki Tagashira, current Discussed with: current Silence from: obrien
2007-06-12 02:16:02 +00:00
ifp = sc->nfe_ifp;
Add basic WOL support for MCP ethernet controllers. It seems the controller does not perform automatic switching from 1000Mbps link to 10/100Mbps link when WOL is activated. Implement establishing 10/100Mps link with auto-negotiation in driver. Link status change handler was modified to remove taskqueue based approach since driver now needs synchronous handling for link establishment. Submitted by: Yamagi Burmeister (lists <> yamagi.org ) (initial version) Tested by: Yamagi Burmeister (lists <> yamagi.org ) MFC after: 1 week
2010-11-11 18:08:50 +00:00
sc->nfe_link = 0;
if ((mii->mii_media_status & (IFM_ACTIVE | IFM_AVALID)) ==
(IFM_ACTIVE | IFM_AVALID)) {
switch (IFM_SUBTYPE(mii->mii_media_active)) {
case IFM_10_T:
case IFM_100_TX:
case IFM_1000_T:
Bring overhauled nfe(4) into tree. o s/printf/device_printf/g o Nuke OpenBSDism. o Nuke NetBSD/OpenBSD specific DMA sync operations.(we don't have a way to sync a single descriptor within a DMA map.) o Remove recursive mutex. o bus_dma(9) clean up. o 40bit DMA address support. o Add protection for Rx map load failure. o Fix a long standing bug for watchdog timeout. [1] o Add additional protections, missing Tx completion interrupt, losing start Tx command, for watchdog timeout. o Switch to taskqueue(9) API to handle interrupts. o Use our own timer for watchdog instead of if_watchdog/if_timer interface. o Advertise VLAN header length/capability correctly to upper layer. o Remove excessive kernel stack consumption in nfe_encap(). o Handle highly fragmented mbuf chains correctly. o Enable etherenet address reprogramming with ifconfig(8). o Add ALTQ/TSO, MSI/MSIX support. o Increased Rx ring to 256 descriptors from 128. o Align Tx/Rx descriptor ring on sizeof(struct nfe_desc64) boundary. o Remove alignment restrictions on Tx/Rx buffers. o Rewritten jumbo frame support code. o Add support for hardware assistend VLAN tag insertion/stripping. o Add support for Tx/Rx flow control based on patches from Peer Chen. [2] o Add a routine that detects whether ethernet address swap routines is required. [3] o Add a workaround that take MAC/PHY out of power down mode. o Add suspend/resume support. o style(9) and code clean up. Special thanks to Shigeaki Tagashira, the original porter of nfe(4), who submitted lots of patches, performed uncountable number of regression tests and maintained nfe(4) for a long time. Without his enthusiastic help and support I could never have completed this overhauling task. The only weak point of nfe(4) compared to nve(4) is instability of manual half-duplex media selection on certain hardwares(auto sensing media type should work for all cases, though). This was a long standing bug of nfe(4) and I still have no idea why it doesn't work on some hardwares. Obtained from: OpenBSD [1] Submitted by: Peer Chen < pchen at nvidia dot com > [2], [3] Reviewed by: Shigeaki Tagashira < shigeaki AT se DOT hiroshima-u DOT ac DOT jp > Tested by: Shigeaki Tagashira, current Discussed with: current Silence from: obrien
2007-06-12 02:16:02 +00:00
sc->nfe_link = 1;
Add basic WOL support for MCP ethernet controllers. It seems the controller does not perform automatic switching from 1000Mbps link to 10/100Mbps link when WOL is activated. Implement establishing 10/100Mps link with auto-negotiation in driver. Link status change handler was modified to remove taskqueue based approach since driver now needs synchronous handling for link establishment. Submitted by: Yamagi Burmeister (lists <> yamagi.org ) (initial version) Tested by: Yamagi Burmeister (lists <> yamagi.org ) MFC after: 1 week
2010-11-11 18:08:50 +00:00
break;
default:
break;
}
}
nfe_mac_config(sc, mii);
txctl = NFE_READ(sc, NFE_TX_CTL);
rxctl = NFE_READ(sc, NFE_RX_CTL);
if (sc->nfe_link != 0 && (ifp->if_drv_flags & IFF_DRV_RUNNING) != 0) {
txctl |= NFE_TX_START;
rxctl |= NFE_RX_START;
} else {
txctl &= ~NFE_TX_START;
rxctl &= ~NFE_RX_START;
}
NFE_WRITE(sc, NFE_TX_CTL, txctl);
NFE_WRITE(sc, NFE_RX_CTL, rxctl);
}
static void
nfe_mac_config(struct nfe_softc *sc, struct mii_data *mii)
{
uint32_t link, misc, phy, seed;
uint32_t val;
NFE_LOCK_ASSERT(sc);
Import the nfe (nForce Ethernet) driver from OpenBSD. This is a pure open source NIC driver with a BSD license. These bits were taken from the OpenBSD public CVS repository on 1-May-06.
2006-06-26 17:25:22 +00:00
phy = NFE_READ(sc, NFE_PHY_IFACE);
phy &= ~(NFE_PHY_HDX | NFE_PHY_100TX | NFE_PHY_1000T);
seed = NFE_READ(sc, NFE_RNDSEED);
seed &= ~NFE_SEED_MASK;
Add basic WOL support for MCP ethernet controllers. It seems the controller does not perform automatic switching from 1000Mbps link to 10/100Mbps link when WOL is activated. Implement establishing 10/100Mps link with auto-negotiation in driver. Link status change handler was modified to remove taskqueue based approach since driver now needs synchronous handling for link establishment. Submitted by: Yamagi Burmeister (lists <> yamagi.org ) (initial version) Tested by: Yamagi Burmeister (lists <> yamagi.org ) MFC after: 1 week
2010-11-11 18:08:50 +00:00
misc = NFE_MISC1_MAGIC;
link = NFE_MEDIA_SET;
if ((IFM_OPTIONS(mii->mii_media_active) & IFM_FDX) == 0) {
Import the nfe (nForce Ethernet) driver from OpenBSD. This is a pure open source NIC driver with a BSD license. These bits were taken from the OpenBSD public CVS repository on 1-May-06.
2006-06-26 17:25:22 +00:00
phy |= NFE_PHY_HDX; /* half-duplex */
misc |= NFE_MISC1_HDX;
}
switch (IFM_SUBTYPE(mii->mii_media_active)) {
case IFM_1000_T: /* full-duplex only */
link |= NFE_MEDIA_1000T;
seed |= NFE_SEED_1000T;
phy |= NFE_PHY_1000T;
break;
case IFM_100_TX:
link |= NFE_MEDIA_100TX;
seed |= NFE_SEED_100TX;
phy |= NFE_PHY_100TX;
break;
case IFM_10_T:
link |= NFE_MEDIA_10T;
seed |= NFE_SEED_10T;
break;
}
Bring overhauled nfe(4) into tree. o s/printf/device_printf/g o Nuke OpenBSDism. o Nuke NetBSD/OpenBSD specific DMA sync operations.(we don't have a way to sync a single descriptor within a DMA map.) o Remove recursive mutex. o bus_dma(9) clean up. o 40bit DMA address support. o Add protection for Rx map load failure. o Fix a long standing bug for watchdog timeout. [1] o Add additional protections, missing Tx completion interrupt, losing start Tx command, for watchdog timeout. o Switch to taskqueue(9) API to handle interrupts. o Use our own timer for watchdog instead of if_watchdog/if_timer interface. o Advertise VLAN header length/capability correctly to upper layer. o Remove excessive kernel stack consumption in nfe_encap(). o Handle highly fragmented mbuf chains correctly. o Enable etherenet address reprogramming with ifconfig(8). o Add ALTQ/TSO, MSI/MSIX support. o Increased Rx ring to 256 descriptors from 128. o Align Tx/Rx descriptor ring on sizeof(struct nfe_desc64) boundary. o Remove alignment restrictions on Tx/Rx buffers. o Rewritten jumbo frame support code. o Add support for hardware assistend VLAN tag insertion/stripping. o Add support for Tx/Rx flow control based on patches from Peer Chen. [2] o Add a routine that detects whether ethernet address swap routines is required. [3] o Add a workaround that take MAC/PHY out of power down mode. o Add suspend/resume support. o style(9) and code clean up. Special thanks to Shigeaki Tagashira, the original porter of nfe(4), who submitted lots of patches, performed uncountable number of regression tests and maintained nfe(4) for a long time. Without his enthusiastic help and support I could never have completed this overhauling task. The only weak point of nfe(4) compared to nve(4) is instability of manual half-duplex media selection on certain hardwares(auto sensing media type should work for all cases, though). This was a long standing bug of nfe(4) and I still have no idea why it doesn't work on some hardwares. Obtained from: OpenBSD [1] Submitted by: Peer Chen < pchen at nvidia dot com > [2], [3] Reviewed by: Shigeaki Tagashira < shigeaki AT se DOT hiroshima-u DOT ac DOT jp > Tested by: Shigeaki Tagashira, current Discussed with: current Silence from: obrien
2007-06-12 02:16:02 +00:00
if ((phy & 0x10000000) != 0) {
if (IFM_SUBTYPE(mii->mii_media_active) == IFM_1000_T)
val = NFE_R1_MAGIC_1000;
else
val = NFE_R1_MAGIC_10_100;
} else
val = NFE_R1_MAGIC_DEFAULT;
NFE_WRITE(sc, NFE_SETUP_R1, val);
Import the nfe (nForce Ethernet) driver from OpenBSD. This is a pure open source NIC driver with a BSD license. These bits were taken from the OpenBSD public CVS repository on 1-May-06.
2006-06-26 17:25:22 +00:00
NFE_WRITE(sc, NFE_RNDSEED, seed); /* XXX: gigabit NICs only? */
NFE_WRITE(sc, NFE_PHY_IFACE, phy);
NFE_WRITE(sc, NFE_MISC1, misc);
NFE_WRITE(sc, NFE_LINKSPEED, link);
Bring overhauled nfe(4) into tree. o s/printf/device_printf/g o Nuke OpenBSDism. o Nuke NetBSD/OpenBSD specific DMA sync operations.(we don't have a way to sync a single descriptor within a DMA map.) o Remove recursive mutex. o bus_dma(9) clean up. o 40bit DMA address support. o Add protection for Rx map load failure. o Fix a long standing bug for watchdog timeout. [1] o Add additional protections, missing Tx completion interrupt, losing start Tx command, for watchdog timeout. o Switch to taskqueue(9) API to handle interrupts. o Use our own timer for watchdog instead of if_watchdog/if_timer interface. o Advertise VLAN header length/capability correctly to upper layer. o Remove excessive kernel stack consumption in nfe_encap(). o Handle highly fragmented mbuf chains correctly. o Enable etherenet address reprogramming with ifconfig(8). o Add ALTQ/TSO, MSI/MSIX support. o Increased Rx ring to 256 descriptors from 128. o Align Tx/Rx descriptor ring on sizeof(struct nfe_desc64) boundary. o Remove alignment restrictions on Tx/Rx buffers. o Rewritten jumbo frame support code. o Add support for hardware assistend VLAN tag insertion/stripping. o Add support for Tx/Rx flow control based on patches from Peer Chen. [2] o Add a routine that detects whether ethernet address swap routines is required. [3] o Add a workaround that take MAC/PHY out of power down mode. o Add suspend/resume support. o style(9) and code clean up. Special thanks to Shigeaki Tagashira, the original porter of nfe(4), who submitted lots of patches, performed uncountable number of regression tests and maintained nfe(4) for a long time. Without his enthusiastic help and support I could never have completed this overhauling task. The only weak point of nfe(4) compared to nve(4) is instability of manual half-duplex media selection on certain hardwares(auto sensing media type should work for all cases, though). This was a long standing bug of nfe(4) and I still have no idea why it doesn't work on some hardwares. Obtained from: OpenBSD [1] Submitted by: Peer Chen < pchen at nvidia dot com > [2], [3] Reviewed by: Shigeaki Tagashira < shigeaki AT se DOT hiroshima-u DOT ac DOT jp > Tested by: Shigeaki Tagashira, current Discussed with: current Silence from: obrien
2007-06-12 02:16:02 +00:00
Add basic WOL support for MCP ethernet controllers. It seems the controller does not perform automatic switching from 1000Mbps link to 10/100Mbps link when WOL is activated. Implement establishing 10/100Mps link with auto-negotiation in driver. Link status change handler was modified to remove taskqueue based approach since driver now needs synchronous handling for link establishment. Submitted by: Yamagi Burmeister (lists <> yamagi.org ) (initial version) Tested by: Yamagi Burmeister (lists <> yamagi.org ) MFC after: 1 week
2010-11-11 18:08:50 +00:00
if ((IFM_OPTIONS(mii->mii_media_active) & IFM_FDX) != 0) {
Bring overhauled nfe(4) into tree. o s/printf/device_printf/g o Nuke OpenBSDism. o Nuke NetBSD/OpenBSD specific DMA sync operations.(we don't have a way to sync a single descriptor within a DMA map.) o Remove recursive mutex. o bus_dma(9) clean up. o 40bit DMA address support. o Add protection for Rx map load failure. o Fix a long standing bug for watchdog timeout. [1] o Add additional protections, missing Tx completion interrupt, losing start Tx command, for watchdog timeout. o Switch to taskqueue(9) API to handle interrupts. o Use our own timer for watchdog instead of if_watchdog/if_timer interface. o Advertise VLAN header length/capability correctly to upper layer. o Remove excessive kernel stack consumption in nfe_encap(). o Handle highly fragmented mbuf chains correctly. o Enable etherenet address reprogramming with ifconfig(8). o Add ALTQ/TSO, MSI/MSIX support. o Increased Rx ring to 256 descriptors from 128. o Align Tx/Rx descriptor ring on sizeof(struct nfe_desc64) boundary. o Remove alignment restrictions on Tx/Rx buffers. o Rewritten jumbo frame support code. o Add support for hardware assistend VLAN tag insertion/stripping. o Add support for Tx/Rx flow control based on patches from Peer Chen. [2] o Add a routine that detects whether ethernet address swap routines is required. [3] o Add a workaround that take MAC/PHY out of power down mode. o Add suspend/resume support. o style(9) and code clean up. Special thanks to Shigeaki Tagashira, the original porter of nfe(4), who submitted lots of patches, performed uncountable number of regression tests and maintained nfe(4) for a long time. Without his enthusiastic help and support I could never have completed this overhauling task. The only weak point of nfe(4) compared to nve(4) is instability of manual half-duplex media selection on certain hardwares(auto sensing media type should work for all cases, though). This was a long standing bug of nfe(4) and I still have no idea why it doesn't work on some hardwares. Obtained from: OpenBSD [1] Submitted by: Peer Chen < pchen at nvidia dot com > [2], [3] Reviewed by: Shigeaki Tagashira < shigeaki AT se DOT hiroshima-u DOT ac DOT jp > Tested by: Shigeaki Tagashira, current Discussed with: current Silence from: obrien
2007-06-12 02:16:02 +00:00
/* It seems all hardwares supports Rx pause frames. */
val = NFE_READ(sc, NFE_RXFILTER);
o Flesh out the generic IEEE 802.3 annex 31B full duplex flow control support in mii(4): - Merge generic flow control advertisement (which can be enabled by passing by MIIF_DOPAUSE to mii_attach(9)) and parsing support from NetBSD into mii_physubr.c and ukphy_subr.c. Unlike as in NetBSD, IFM_FLOW isn't implemented as a global option via the "don't care mask" but instead as a media specific option this. This has the following advantages: o allows flow control advertisement with autonegotiation to be turned on and off via ifconfig(8) with the default typically being off (though MIIF_FORCEPAUSE has been added causing flow control to be always advertised, allowing to easily MFC this changes for drivers that previously used home-grown support for flow control that behaved that way without breaking POLA) o allows to deal with PHY drivers where flow control advertisement with manual selection doesn't work or at least isn't implemented, like it's the case with brgphy(4), e1000phy(4) and ip1000phy(4), by setting MIIF_NOMANPAUSE o the available combinations of media options are readily available from the `ifconfig -m` output - Add IFM_FLOW to IFM_SHARED_OPTION_DESCRIPTIONS and IFM_ETH_RXPAUSE and IFM_ETH_TXPAUSE to IFM_SUBTYPE_ETHERNET_OPTION_DESCRIPTIONS so these are understood by ifconfig(8). o Make the master/slave support in mii(4) actually usable: - Change IFM_ETH_MASTER from being implemented as a global option via the "don't care mask" to a media specific one as it actually is only applicable to IFM_1000_T to date. - Let mii_phy_setmedia() set GTCR_MAN_MS in IFM_1000_T slave mode to actually configure manually selected slave mode (like we also do in the PHY specific implementations). - Add IFM_ETH_MASTER to IFM_SUBTYPE_ETHERNET_OPTION_DESCRIPTIONS so it is understood by ifconfig(8). o Switch bge(4), bce(4), msk(4), nfe(4) and stge(4) along with brgphy(4), e1000phy(4) and ip1000phy(4) to use the generic flow control support instead of home-grown solutions via IFM_FLAGs. This includes changing these PHY drivers and smcphy(4) to no longer unconditionally advertise support for flow control but only if the selected media has IFM_FLOW set (or MIIF_FORCEPAUSE is set) and implemented for these media variants, i.e. typically only for copper. o Switch brgphy(4), ciphy(4), e1000phy(4) and ip1000phy(4) to report and set IFM_1000_T master mode via IFM_ETH_MASTER instead of via IFF_LINK0 and some IFM_FLAGn. o Switch brgphy(4) to add at least the the supported copper media based on the contents of the BMSR via mii_phy_add_media() instead of hardcoding them. The latter approach seems to have developed historically, besides causing unnecessary code duplication it was also undesirable because brgphy_mii_phy_auto() already based the capability advertisement on the contents of the BMSR though. o Let brgphy(4) set IFM_1000_T master mode on all supported PHY and not just BCM5701. Apparently this was a misinterpretation of a workaround in the Linux tg3 driver; BCM5701 seem to require RGPHY_1000CTL_MSE and BRGPHY_1000CTL_MSC to be set when configuring autonegotiation but this doesn't mean we can't set these as well on other PHYs for manual media selection. o Let ukphy_status() report IFM_1000_T master mode via IFM_ETH_MASTER so IFM_1000_T master mode support now is generally available with all PHY drivers. o Don't let e1000phy(4) set master/slave bits for IFM_1000_SX as it's not applicable there. Reviewed by: yongari (plus additional testing) Obtained from: NetBSD (partially), OpenBSD (partially) MFC after: 2 weeks
2010-11-14 13:26:10 +00:00
if ((IFM_OPTIONS(mii->mii_media_active) &
IFM_ETH_RXPAUSE) != 0)
Bring overhauled nfe(4) into tree. o s/printf/device_printf/g o Nuke OpenBSDism. o Nuke NetBSD/OpenBSD specific DMA sync operations.(we don't have a way to sync a single descriptor within a DMA map.) o Remove recursive mutex. o bus_dma(9) clean up. o 40bit DMA address support. o Add protection for Rx map load failure. o Fix a long standing bug for watchdog timeout. [1] o Add additional protections, missing Tx completion interrupt, losing start Tx command, for watchdog timeout. o Switch to taskqueue(9) API to handle interrupts. o Use our own timer for watchdog instead of if_watchdog/if_timer interface. o Advertise VLAN header length/capability correctly to upper layer. o Remove excessive kernel stack consumption in nfe_encap(). o Handle highly fragmented mbuf chains correctly. o Enable etherenet address reprogramming with ifconfig(8). o Add ALTQ/TSO, MSI/MSIX support. o Increased Rx ring to 256 descriptors from 128. o Align Tx/Rx descriptor ring on sizeof(struct nfe_desc64) boundary. o Remove alignment restrictions on Tx/Rx buffers. o Rewritten jumbo frame support code. o Add support for hardware assistend VLAN tag insertion/stripping. o Add support for Tx/Rx flow control based on patches from Peer Chen. [2] o Add a routine that detects whether ethernet address swap routines is required. [3] o Add a workaround that take MAC/PHY out of power down mode. o Add suspend/resume support. o style(9) and code clean up. Special thanks to Shigeaki Tagashira, the original porter of nfe(4), who submitted lots of patches, performed uncountable number of regression tests and maintained nfe(4) for a long time. Without his enthusiastic help and support I could never have completed this overhauling task. The only weak point of nfe(4) compared to nve(4) is instability of manual half-duplex media selection on certain hardwares(auto sensing media type should work for all cases, though). This was a long standing bug of nfe(4) and I still have no idea why it doesn't work on some hardwares. Obtained from: OpenBSD [1] Submitted by: Peer Chen < pchen at nvidia dot com > [2], [3] Reviewed by: Shigeaki Tagashira < shigeaki AT se DOT hiroshima-u DOT ac DOT jp > Tested by: Shigeaki Tagashira, current Discussed with: current Silence from: obrien
2007-06-12 02:16:02 +00:00
val |= NFE_PFF_RX_PAUSE;
else
val &= ~NFE_PFF_RX_PAUSE;
NFE_WRITE(sc, NFE_RXFILTER, val);
if ((sc->nfe_flags & NFE_TX_FLOW_CTRL) != 0) {
val = NFE_READ(sc, NFE_MISC1);
Add basic WOL support for MCP ethernet controllers. It seems the controller does not perform automatic switching from 1000Mbps link to 10/100Mbps link when WOL is activated. Implement establishing 10/100Mps link with auto-negotiation in driver. Link status change handler was modified to remove taskqueue based approach since driver now needs synchronous handling for link establishment. Submitted by: Yamagi Burmeister (lists <> yamagi.org ) (initial version) Tested by: Yamagi Burmeister (lists <> yamagi.org ) MFC after: 1 week
2010-11-11 18:08:50 +00:00
if ((IFM_OPTIONS(mii->mii_media_active) &
o Flesh out the generic IEEE 802.3 annex 31B full duplex flow control support in mii(4): - Merge generic flow control advertisement (which can be enabled by passing by MIIF_DOPAUSE to mii_attach(9)) and parsing support from NetBSD into mii_physubr.c and ukphy_subr.c. Unlike as in NetBSD, IFM_FLOW isn't implemented as a global option via the "don't care mask" but instead as a media specific option this. This has the following advantages: o allows flow control advertisement with autonegotiation to be turned on and off via ifconfig(8) with the default typically being off (though MIIF_FORCEPAUSE has been added causing flow control to be always advertised, allowing to easily MFC this changes for drivers that previously used home-grown support for flow control that behaved that way without breaking POLA) o allows to deal with PHY drivers where flow control advertisement with manual selection doesn't work or at least isn't implemented, like it's the case with brgphy(4), e1000phy(4) and ip1000phy(4), by setting MIIF_NOMANPAUSE o the available combinations of media options are readily available from the `ifconfig -m` output - Add IFM_FLOW to IFM_SHARED_OPTION_DESCRIPTIONS and IFM_ETH_RXPAUSE and IFM_ETH_TXPAUSE to IFM_SUBTYPE_ETHERNET_OPTION_DESCRIPTIONS so these are understood by ifconfig(8). o Make the master/slave support in mii(4) actually usable: - Change IFM_ETH_MASTER from being implemented as a global option via the "don't care mask" to a media specific one as it actually is only applicable to IFM_1000_T to date. - Let mii_phy_setmedia() set GTCR_MAN_MS in IFM_1000_T slave mode to actually configure manually selected slave mode (like we also do in the PHY specific implementations). - Add IFM_ETH_MASTER to IFM_SUBTYPE_ETHERNET_OPTION_DESCRIPTIONS so it is understood by ifconfig(8). o Switch bge(4), bce(4), msk(4), nfe(4) and stge(4) along with brgphy(4), e1000phy(4) and ip1000phy(4) to use the generic flow control support instead of home-grown solutions via IFM_FLAGs. This includes changing these PHY drivers and smcphy(4) to no longer unconditionally advertise support for flow control but only if the selected media has IFM_FLOW set (or MIIF_FORCEPAUSE is set) and implemented for these media variants, i.e. typically only for copper. o Switch brgphy(4), ciphy(4), e1000phy(4) and ip1000phy(4) to report and set IFM_1000_T master mode via IFM_ETH_MASTER instead of via IFF_LINK0 and some IFM_FLAGn. o Switch brgphy(4) to add at least the the supported copper media based on the contents of the BMSR via mii_phy_add_media() instead of hardcoding them. The latter approach seems to have developed historically, besides causing unnecessary code duplication it was also undesirable because brgphy_mii_phy_auto() already based the capability advertisement on the contents of the BMSR though. o Let brgphy(4) set IFM_1000_T master mode on all supported PHY and not just BCM5701. Apparently this was a misinterpretation of a workaround in the Linux tg3 driver; BCM5701 seem to require RGPHY_1000CTL_MSE and BRGPHY_1000CTL_MSC to be set when configuring autonegotiation but this doesn't mean we can't set these as well on other PHYs for manual media selection. o Let ukphy_status() report IFM_1000_T master mode via IFM_ETH_MASTER so IFM_1000_T master mode support now is generally available with all PHY drivers. o Don't let e1000phy(4) set master/slave bits for IFM_1000_SX as it's not applicable there. Reviewed by: yongari (plus additional testing) Obtained from: NetBSD (partially), OpenBSD (partially) MFC after: 2 weeks
2010-11-14 13:26:10 +00:00
IFM_ETH_TXPAUSE) != 0) {
Bring overhauled nfe(4) into tree. o s/printf/device_printf/g o Nuke OpenBSDism. o Nuke NetBSD/OpenBSD specific DMA sync operations.(we don't have a way to sync a single descriptor within a DMA map.) o Remove recursive mutex. o bus_dma(9) clean up. o 40bit DMA address support. o Add protection for Rx map load failure. o Fix a long standing bug for watchdog timeout. [1] o Add additional protections, missing Tx completion interrupt, losing start Tx command, for watchdog timeout. o Switch to taskqueue(9) API to handle interrupts. o Use our own timer for watchdog instead of if_watchdog/if_timer interface. o Advertise VLAN header length/capability correctly to upper layer. o Remove excessive kernel stack consumption in nfe_encap(). o Handle highly fragmented mbuf chains correctly. o Enable etherenet address reprogramming with ifconfig(8). o Add ALTQ/TSO, MSI/MSIX support. o Increased Rx ring to 256 descriptors from 128. o Align Tx/Rx descriptor ring on sizeof(struct nfe_desc64) boundary. o Remove alignment restrictions on Tx/Rx buffers. o Rewritten jumbo frame support code. o Add support for hardware assistend VLAN tag insertion/stripping. o Add support for Tx/Rx flow control based on patches from Peer Chen. [2] o Add a routine that detects whether ethernet address swap routines is required. [3] o Add a workaround that take MAC/PHY out of power down mode. o Add suspend/resume support. o style(9) and code clean up. Special thanks to Shigeaki Tagashira, the original porter of nfe(4), who submitted lots of patches, performed uncountable number of regression tests and maintained nfe(4) for a long time. Without his enthusiastic help and support I could never have completed this overhauling task. The only weak point of nfe(4) compared to nve(4) is instability of manual half-duplex media selection on certain hardwares(auto sensing media type should work for all cases, though). This was a long standing bug of nfe(4) and I still have no idea why it doesn't work on some hardwares. Obtained from: OpenBSD [1] Submitted by: Peer Chen < pchen at nvidia dot com > [2], [3] Reviewed by: Shigeaki Tagashira < shigeaki AT se DOT hiroshima-u DOT ac DOT jp > Tested by: Shigeaki Tagashira, current Discussed with: current Silence from: obrien
2007-06-12 02:16:02 +00:00
NFE_WRITE(sc, NFE_TX_PAUSE_FRAME,
NFE_TX_PAUSE_FRAME_ENABLE);
val |= NFE_MISC1_TX_PAUSE;
} else {
val &= ~NFE_MISC1_TX_PAUSE;
NFE_WRITE(sc, NFE_TX_PAUSE_FRAME,
NFE_TX_PAUSE_FRAME_DISABLE);
}
NFE_WRITE(sc, NFE_MISC1, val);
}
} else {
/* disable rx/tx pause frames */
val = NFE_READ(sc, NFE_RXFILTER);
val &= ~NFE_PFF_RX_PAUSE;
NFE_WRITE(sc, NFE_RXFILTER, val);
if ((sc->nfe_flags & NFE_TX_FLOW_CTRL) != 0) {
NFE_WRITE(sc, NFE_TX_PAUSE_FRAME,
NFE_TX_PAUSE_FRAME_DISABLE);
val = NFE_READ(sc, NFE_MISC1);
val &= ~NFE_MISC1_TX_PAUSE;
NFE_WRITE(sc, NFE_MISC1, val);
}
}
Import the nfe (nForce Ethernet) driver from OpenBSD. This is a pure open source NIC driver with a BSD license. These bits were taken from the OpenBSD public CVS repository on 1-May-06.
2006-06-26 17:25:22 +00:00
}
style(9) cleanup.
2006-10-19 10:01:26 +00:00
Add a pure open source nForce Ethernet driver, under BSDL. This driver was ported from OpenBSD by Shigeaki Tagashira <shigeaki@se.hiroshima-u.ac.jp> and posted at http://www.se.hiroshima-u.ac.jp/~shigeaki/software/freebsd-nfe.html It was additionally cleaned up by me. It is still a work-in-progress and thus is purposefully not in GENERIC. And it conflicts with nve(4), so only one should be loaded.
2006-06-26 23:41:07 +00:00
static int
nfe_miibus_readreg(device_t dev, int phy, int reg)
Import the nfe (nForce Ethernet) driver from OpenBSD. This is a pure open source NIC driver with a BSD license. These bits were taken from the OpenBSD public CVS repository on 1-May-06.
2006-06-26 17:25:22 +00:00
{
Add a pure open source nForce Ethernet driver, under BSDL. This driver was ported from OpenBSD by Shigeaki Tagashira <shigeaki@se.hiroshima-u.ac.jp> and posted at http://www.se.hiroshima-u.ac.jp/~shigeaki/software/freebsd-nfe.html It was additionally cleaned up by me. It is still a work-in-progress and thus is purposefully not in GENERIC. And it conflicts with nve(4), so only one should be loaded.
2006-06-26 23:41:07 +00:00
struct nfe_softc *sc = device_get_softc(dev);
Bring overhauled nfe(4) into tree. o s/printf/device_printf/g o Nuke OpenBSDism. o Nuke NetBSD/OpenBSD specific DMA sync operations.(we don't have a way to sync a single descriptor within a DMA map.) o Remove recursive mutex. o bus_dma(9) clean up. o 40bit DMA address support. o Add protection for Rx map load failure. o Fix a long standing bug for watchdog timeout. [1] o Add additional protections, missing Tx completion interrupt, losing start Tx command, for watchdog timeout. o Switch to taskqueue(9) API to handle interrupts. o Use our own timer for watchdog instead of if_watchdog/if_timer interface. o Advertise VLAN header length/capability correctly to upper layer. o Remove excessive kernel stack consumption in nfe_encap(). o Handle highly fragmented mbuf chains correctly. o Enable etherenet address reprogramming with ifconfig(8). o Add ALTQ/TSO, MSI/MSIX support. o Increased Rx ring to 256 descriptors from 128. o Align Tx/Rx descriptor ring on sizeof(struct nfe_desc64) boundary. o Remove alignment restrictions on Tx/Rx buffers. o Rewritten jumbo frame support code. o Add support for hardware assistend VLAN tag insertion/stripping. o Add support for Tx/Rx flow control based on patches from Peer Chen. [2] o Add a routine that detects whether ethernet address swap routines is required. [3] o Add a workaround that take MAC/PHY out of power down mode. o Add suspend/resume support. o style(9) and code clean up. Special thanks to Shigeaki Tagashira, the original porter of nfe(4), who submitted lots of patches, performed uncountable number of regression tests and maintained nfe(4) for a long time. Without his enthusiastic help and support I could never have completed this overhauling task. The only weak point of nfe(4) compared to nve(4) is instability of manual half-duplex media selection on certain hardwares(auto sensing media type should work for all cases, though). This was a long standing bug of nfe(4) and I still have no idea why it doesn't work on some hardwares. Obtained from: OpenBSD [1] Submitted by: Peer Chen < pchen at nvidia dot com > [2], [3] Reviewed by: Shigeaki Tagashira < shigeaki AT se DOT hiroshima-u DOT ac DOT jp > Tested by: Shigeaki Tagashira, current Discussed with: current Silence from: obrien
2007-06-12 02:16:02 +00:00
uint32_t val;
Import the nfe (nForce Ethernet) driver from OpenBSD. This is a pure open source NIC driver with a BSD license. These bits were taken from the OpenBSD public CVS repository on 1-May-06.
2006-06-26 17:25:22 +00:00
int ntries;
NFE_WRITE(sc, NFE_PHY_STATUS, 0xf);
if (NFE_READ(sc, NFE_PHY_CTL) & NFE_PHY_BUSY) {
NFE_WRITE(sc, NFE_PHY_CTL, NFE_PHY_BUSY);
DELAY(100);
}
NFE_WRITE(sc, NFE_PHY_CTL, (phy << NFE_PHYADD_SHIFT) | reg);
Bring overhauled nfe(4) into tree. o s/printf/device_printf/g o Nuke OpenBSDism. o Nuke NetBSD/OpenBSD specific DMA sync operations.(we don't have a way to sync a single descriptor within a DMA map.) o Remove recursive mutex. o bus_dma(9) clean up. o 40bit DMA address support. o Add protection for Rx map load failure. o Fix a long standing bug for watchdog timeout. [1] o Add additional protections, missing Tx completion interrupt, losing start Tx command, for watchdog timeout. o Switch to taskqueue(9) API to handle interrupts. o Use our own timer for watchdog instead of if_watchdog/if_timer interface. o Advertise VLAN header length/capability correctly to upper layer. o Remove excessive kernel stack consumption in nfe_encap(). o Handle highly fragmented mbuf chains correctly. o Enable etherenet address reprogramming with ifconfig(8). o Add ALTQ/TSO, MSI/MSIX support. o Increased Rx ring to 256 descriptors from 128. o Align Tx/Rx descriptor ring on sizeof(struct nfe_desc64) boundary. o Remove alignment restrictions on Tx/Rx buffers. o Rewritten jumbo frame support code. o Add support for hardware assistend VLAN tag insertion/stripping. o Add support for Tx/Rx flow control based on patches from Peer Chen. [2] o Add a routine that detects whether ethernet address swap routines is required. [3] o Add a workaround that take MAC/PHY out of power down mode. o Add suspend/resume support. o style(9) and code clean up. Special thanks to Shigeaki Tagashira, the original porter of nfe(4), who submitted lots of patches, performed uncountable number of regression tests and maintained nfe(4) for a long time. Without his enthusiastic help and support I could never have completed this overhauling task. The only weak point of nfe(4) compared to nve(4) is instability of manual half-duplex media selection on certain hardwares(auto sensing media type should work for all cases, though). This was a long standing bug of nfe(4) and I still have no idea why it doesn't work on some hardwares. Obtained from: OpenBSD [1] Submitted by: Peer Chen < pchen at nvidia dot com > [2], [3] Reviewed by: Shigeaki Tagashira < shigeaki AT se DOT hiroshima-u DOT ac DOT jp > Tested by: Shigeaki Tagashira, current Discussed with: current Silence from: obrien
2007-06-12 02:16:02 +00:00
for (ntries = 0; ntries < NFE_TIMEOUT; ntries++) {
Import the nfe (nForce Ethernet) driver from OpenBSD. This is a pure open source NIC driver with a BSD license. These bits were taken from the OpenBSD public CVS repository on 1-May-06.
2006-06-26 17:25:22 +00:00
DELAY(100);
if (!(NFE_READ(sc, NFE_PHY_CTL) & NFE_PHY_BUSY))
break;
}
Bring overhauled nfe(4) into tree. o s/printf/device_printf/g o Nuke OpenBSDism. o Nuke NetBSD/OpenBSD specific DMA sync operations.(we don't have a way to sync a single descriptor within a DMA map.) o Remove recursive mutex. o bus_dma(9) clean up. o 40bit DMA address support. o Add protection for Rx map load failure. o Fix a long standing bug for watchdog timeout. [1] o Add additional protections, missing Tx completion interrupt, losing start Tx command, for watchdog timeout. o Switch to taskqueue(9) API to handle interrupts. o Use our own timer for watchdog instead of if_watchdog/if_timer interface. o Advertise VLAN header length/capability correctly to upper layer. o Remove excessive kernel stack consumption in nfe_encap(). o Handle highly fragmented mbuf chains correctly. o Enable etherenet address reprogramming with ifconfig(8). o Add ALTQ/TSO, MSI/MSIX support. o Increased Rx ring to 256 descriptors from 128. o Align Tx/Rx descriptor ring on sizeof(struct nfe_desc64) boundary. o Remove alignment restrictions on Tx/Rx buffers. o Rewritten jumbo frame support code. o Add support for hardware assistend VLAN tag insertion/stripping. o Add support for Tx/Rx flow control based on patches from Peer Chen. [2] o Add a routine that detects whether ethernet address swap routines is required. [3] o Add a workaround that take MAC/PHY out of power down mode. o Add suspend/resume support. o style(9) and code clean up. Special thanks to Shigeaki Tagashira, the original porter of nfe(4), who submitted lots of patches, performed uncountable number of regression tests and maintained nfe(4) for a long time. Without his enthusiastic help and support I could never have completed this overhauling task. The only weak point of nfe(4) compared to nve(4) is instability of manual half-duplex media selection on certain hardwares(auto sensing media type should work for all cases, though). This was a long standing bug of nfe(4) and I still have no idea why it doesn't work on some hardwares. Obtained from: OpenBSD [1] Submitted by: Peer Chen < pchen at nvidia dot com > [2], [3] Reviewed by: Shigeaki Tagashira < shigeaki AT se DOT hiroshima-u DOT ac DOT jp > Tested by: Shigeaki Tagashira, current Discussed with: current Silence from: obrien
2007-06-12 02:16:02 +00:00
if (ntries == NFE_TIMEOUT) {
DPRINTFN(sc, 2, "timeout waiting for PHY\n");
Import the nfe (nForce Ethernet) driver from OpenBSD. This is a pure open source NIC driver with a BSD license. These bits were taken from the OpenBSD public CVS repository on 1-May-06.
2006-06-26 17:25:22 +00:00
return 0;
}
if (NFE_READ(sc, NFE_PHY_STATUS) & NFE_PHY_ERROR) {
Bring overhauled nfe(4) into tree. o s/printf/device_printf/g o Nuke OpenBSDism. o Nuke NetBSD/OpenBSD specific DMA sync operations.(we don't have a way to sync a single descriptor within a DMA map.) o Remove recursive mutex. o bus_dma(9) clean up. o 40bit DMA address support. o Add protection for Rx map load failure. o Fix a long standing bug for watchdog timeout. [1] o Add additional protections, missing Tx completion interrupt, losing start Tx command, for watchdog timeout. o Switch to taskqueue(9) API to handle interrupts. o Use our own timer for watchdog instead of if_watchdog/if_timer interface. o Advertise VLAN header length/capability correctly to upper layer. o Remove excessive kernel stack consumption in nfe_encap(). o Handle highly fragmented mbuf chains correctly. o Enable etherenet address reprogramming with ifconfig(8). o Add ALTQ/TSO, MSI/MSIX support. o Increased Rx ring to 256 descriptors from 128. o Align Tx/Rx descriptor ring on sizeof(struct nfe_desc64) boundary. o Remove alignment restrictions on Tx/Rx buffers. o Rewritten jumbo frame support code. o Add support for hardware assistend VLAN tag insertion/stripping. o Add support for Tx/Rx flow control based on patches from Peer Chen. [2] o Add a routine that detects whether ethernet address swap routines is required. [3] o Add a workaround that take MAC/PHY out of power down mode. o Add suspend/resume support. o style(9) and code clean up. Special thanks to Shigeaki Tagashira, the original porter of nfe(4), who submitted lots of patches, performed uncountable number of regression tests and maintained nfe(4) for a long time. Without his enthusiastic help and support I could never have completed this overhauling task. The only weak point of nfe(4) compared to nve(4) is instability of manual half-duplex media selection on certain hardwares(auto sensing media type should work for all cases, though). This was a long standing bug of nfe(4) and I still have no idea why it doesn't work on some hardwares. Obtained from: OpenBSD [1] Submitted by: Peer Chen < pchen at nvidia dot com > [2], [3] Reviewed by: Shigeaki Tagashira < shigeaki AT se DOT hiroshima-u DOT ac DOT jp > Tested by: Shigeaki Tagashira, current Discussed with: current Silence from: obrien
2007-06-12 02:16:02 +00:00
DPRINTFN(sc, 2, "could not read PHY\n");
Import the nfe (nForce Ethernet) driver from OpenBSD. This is a pure open source NIC driver with a BSD license. These bits were taken from the OpenBSD public CVS repository on 1-May-06.
2006-06-26 17:25:22 +00:00
return 0;
}
val = NFE_READ(sc, NFE_PHY_DATA);
if (val != 0xffffffff && val != 0)
sc->mii_phyaddr = phy;
Bring overhauled nfe(4) into tree. o s/printf/device_printf/g o Nuke OpenBSDism. o Nuke NetBSD/OpenBSD specific DMA sync operations.(we don't have a way to sync a single descriptor within a DMA map.) o Remove recursive mutex. o bus_dma(9) clean up. o 40bit DMA address support. o Add protection for Rx map load failure. o Fix a long standing bug for watchdog timeout. [1] o Add additional protections, missing Tx completion interrupt, losing start Tx command, for watchdog timeout. o Switch to taskqueue(9) API to handle interrupts. o Use our own timer for watchdog instead of if_watchdog/if_timer interface. o Advertise VLAN header length/capability correctly to upper layer. o Remove excessive kernel stack consumption in nfe_encap(). o Handle highly fragmented mbuf chains correctly. o Enable etherenet address reprogramming with ifconfig(8). o Add ALTQ/TSO, MSI/MSIX support. o Increased Rx ring to 256 descriptors from 128. o Align Tx/Rx descriptor ring on sizeof(struct nfe_desc64) boundary. o Remove alignment restrictions on Tx/Rx buffers. o Rewritten jumbo frame support code. o Add support for hardware assistend VLAN tag insertion/stripping. o Add support for Tx/Rx flow control based on patches from Peer Chen. [2] o Add a routine that detects whether ethernet address swap routines is required. [3] o Add a workaround that take MAC/PHY out of power down mode. o Add suspend/resume support. o style(9) and code clean up. Special thanks to Shigeaki Tagashira, the original porter of nfe(4), who submitted lots of patches, performed uncountable number of regression tests and maintained nfe(4) for a long time. Without his enthusiastic help and support I could never have completed this overhauling task. The only weak point of nfe(4) compared to nve(4) is instability of manual half-duplex media selection on certain hardwares(auto sensing media type should work for all cases, though). This was a long standing bug of nfe(4) and I still have no idea why it doesn't work on some hardwares. Obtained from: OpenBSD [1] Submitted by: Peer Chen < pchen at nvidia dot com > [2], [3] Reviewed by: Shigeaki Tagashira < shigeaki AT se DOT hiroshima-u DOT ac DOT jp > Tested by: Shigeaki Tagashira, current Discussed with: current Silence from: obrien
2007-06-12 02:16:02 +00:00
DPRINTFN(sc, 2, "mii read phy %d reg 0x%x ret 0x%x\n", phy, reg, val);
Import the nfe (nForce Ethernet) driver from OpenBSD. This is a pure open source NIC driver with a BSD license. These bits were taken from the OpenBSD public CVS repository on 1-May-06.
2006-06-26 17:25:22 +00:00
Bring overhauled nfe(4) into tree. o s/printf/device_printf/g o Nuke OpenBSDism. o Nuke NetBSD/OpenBSD specific DMA sync operations.(we don't have a way to sync a single descriptor within a DMA map.) o Remove recursive mutex. o bus_dma(9) clean up. o 40bit DMA address support. o Add protection for Rx map load failure. o Fix a long standing bug for watchdog timeout. [1] o Add additional protections, missing Tx completion interrupt, losing start Tx command, for watchdog timeout. o Switch to taskqueue(9) API to handle interrupts. o Use our own timer for watchdog instead of if_watchdog/if_timer interface. o Advertise VLAN header length/capability correctly to upper layer. o Remove excessive kernel stack consumption in nfe_encap(). o Handle highly fragmented mbuf chains correctly. o Enable etherenet address reprogramming with ifconfig(8). o Add ALTQ/TSO, MSI/MSIX support. o Increased Rx ring to 256 descriptors from 128. o Align Tx/Rx descriptor ring on sizeof(struct nfe_desc64) boundary. o Remove alignment restrictions on Tx/Rx buffers. o Rewritten jumbo frame support code. o Add support for hardware assistend VLAN tag insertion/stripping. o Add support for Tx/Rx flow control based on patches from Peer Chen. [2] o Add a routine that detects whether ethernet address swap routines is required. [3] o Add a workaround that take MAC/PHY out of power down mode. o Add suspend/resume support. o style(9) and code clean up. Special thanks to Shigeaki Tagashira, the original porter of nfe(4), who submitted lots of patches, performed uncountable number of regression tests and maintained nfe(4) for a long time. Without his enthusiastic help and support I could never have completed this overhauling task. The only weak point of nfe(4) compared to nve(4) is instability of manual half-duplex media selection on certain hardwares(auto sensing media type should work for all cases, though). This was a long standing bug of nfe(4) and I still have no idea why it doesn't work on some hardwares. Obtained from: OpenBSD [1] Submitted by: Peer Chen < pchen at nvidia dot com > [2], [3] Reviewed by: Shigeaki Tagashira < shigeaki AT se DOT hiroshima-u DOT ac DOT jp > Tested by: Shigeaki Tagashira, current Discussed with: current Silence from: obrien
2007-06-12 02:16:02 +00:00
return (val);
Import the nfe (nForce Ethernet) driver from OpenBSD. This is a pure open source NIC driver with a BSD license. These bits were taken from the OpenBSD public CVS repository on 1-May-06.
2006-06-26 17:25:22 +00:00
}
style(9) cleanup.
2006-10-19 10:01:26 +00:00
Add a pure open source nForce Ethernet driver, under BSDL. This driver was ported from OpenBSD by Shigeaki Tagashira <shigeaki@se.hiroshima-u.ac.jp> and posted at http://www.se.hiroshima-u.ac.jp/~shigeaki/software/freebsd-nfe.html It was additionally cleaned up by me. It is still a work-in-progress and thus is purposefully not in GENERIC. And it conflicts with nve(4), so only one should be loaded.
2006-06-26 23:41:07 +00:00
static int
nfe_miibus_writereg(device_t dev, int phy, int reg, int val)
Import the nfe (nForce Ethernet) driver from OpenBSD. This is a pure open source NIC driver with a BSD license. These bits were taken from the OpenBSD public CVS repository on 1-May-06.
2006-06-26 17:25:22 +00:00
{
Add a pure open source nForce Ethernet driver, under BSDL. This driver was ported from OpenBSD by Shigeaki Tagashira <shigeaki@se.hiroshima-u.ac.jp> and posted at http://www.se.hiroshima-u.ac.jp/~shigeaki/software/freebsd-nfe.html It was additionally cleaned up by me. It is still a work-in-progress and thus is purposefully not in GENERIC. And it conflicts with nve(4), so only one should be loaded.
2006-06-26 23:41:07 +00:00
struct nfe_softc *sc = device_get_softc(dev);
Bring overhauled nfe(4) into tree. o s/printf/device_printf/g o Nuke OpenBSDism. o Nuke NetBSD/OpenBSD specific DMA sync operations.(we don't have a way to sync a single descriptor within a DMA map.) o Remove recursive mutex. o bus_dma(9) clean up. o 40bit DMA address support. o Add protection for Rx map load failure. o Fix a long standing bug for watchdog timeout. [1] o Add additional protections, missing Tx completion interrupt, losing start Tx command, for watchdog timeout. o Switch to taskqueue(9) API to handle interrupts. o Use our own timer for watchdog instead of if_watchdog/if_timer interface. o Advertise VLAN header length/capability correctly to upper layer. o Remove excessive kernel stack consumption in nfe_encap(). o Handle highly fragmented mbuf chains correctly. o Enable etherenet address reprogramming with ifconfig(8). o Add ALTQ/TSO, MSI/MSIX support. o Increased Rx ring to 256 descriptors from 128. o Align Tx/Rx descriptor ring on sizeof(struct nfe_desc64) boundary. o Remove alignment restrictions on Tx/Rx buffers. o Rewritten jumbo frame support code. o Add support for hardware assistend VLAN tag insertion/stripping. o Add support for Tx/Rx flow control based on patches from Peer Chen. [2] o Add a routine that detects whether ethernet address swap routines is required. [3] o Add a workaround that take MAC/PHY out of power down mode. o Add suspend/resume support. o style(9) and code clean up. Special thanks to Shigeaki Tagashira, the original porter of nfe(4), who submitted lots of patches, performed uncountable number of regression tests and maintained nfe(4) for a long time. Without his enthusiastic help and support I could never have completed this overhauling task. The only weak point of nfe(4) compared to nve(4) is instability of manual half-duplex media selection on certain hardwares(auto sensing media type should work for all cases, though). This was a long standing bug of nfe(4) and I still have no idea why it doesn't work on some hardwares. Obtained from: OpenBSD [1] Submitted by: Peer Chen < pchen at nvidia dot com > [2], [3] Reviewed by: Shigeaki Tagashira < shigeaki AT se DOT hiroshima-u DOT ac DOT jp > Tested by: Shigeaki Tagashira, current Discussed with: current Silence from: obrien
2007-06-12 02:16:02 +00:00
uint32_t ctl;
style(9) cleanup.
2006-10-19 10:01:26 +00:00
int ntries;
Import the nfe (nForce Ethernet) driver from OpenBSD. This is a pure open source NIC driver with a BSD license. These bits were taken from the OpenBSD public CVS repository on 1-May-06.
2006-06-26 17:25:22 +00:00
NFE_WRITE(sc, NFE_PHY_STATUS, 0xf);
if (NFE_READ(sc, NFE_PHY_CTL) & NFE_PHY_BUSY) {
NFE_WRITE(sc, NFE_PHY_CTL, NFE_PHY_BUSY);
DELAY(100);
}
NFE_WRITE(sc, NFE_PHY_DATA, val);
ctl = NFE_PHY_WRITE | (phy << NFE_PHYADD_SHIFT) | reg;
NFE_WRITE(sc, NFE_PHY_CTL, ctl);
Bring overhauled nfe(4) into tree. o s/printf/device_printf/g o Nuke OpenBSDism. o Nuke NetBSD/OpenBSD specific DMA sync operations.(we don't have a way to sync a single descriptor within a DMA map.) o Remove recursive mutex. o bus_dma(9) clean up. o 40bit DMA address support. o Add protection for Rx map load failure. o Fix a long standing bug for watchdog timeout. [1] o Add additional protections, missing Tx completion interrupt, losing start Tx command, for watchdog timeout. o Switch to taskqueue(9) API to handle interrupts. o Use our own timer for watchdog instead of if_watchdog/if_timer interface. o Advertise VLAN header length/capability correctly to upper layer. o Remove excessive kernel stack consumption in nfe_encap(). o Handle highly fragmented mbuf chains correctly. o Enable etherenet address reprogramming with ifconfig(8). o Add ALTQ/TSO, MSI/MSIX support. o Increased Rx ring to 256 descriptors from 128. o Align Tx/Rx descriptor ring on sizeof(struct nfe_desc64) boundary. o Remove alignment restrictions on Tx/Rx buffers. o Rewritten jumbo frame support code. o Add support for hardware assistend VLAN tag insertion/stripping. o Add support for Tx/Rx flow control based on patches from Peer Chen. [2] o Add a routine that detects whether ethernet address swap routines is required. [3] o Add a workaround that take MAC/PHY out of power down mode. o Add suspend/resume support. o style(9) and code clean up. Special thanks to Shigeaki Tagashira, the original porter of nfe(4), who submitted lots of patches, performed uncountable number of regression tests and maintained nfe(4) for a long time. Without his enthusiastic help and support I could never have completed this overhauling task. The only weak point of nfe(4) compared to nve(4) is instability of manual half-duplex media selection on certain hardwares(auto sensing media type should work for all cases, though). This was a long standing bug of nfe(4) and I still have no idea why it doesn't work on some hardwares. Obtained from: OpenBSD [1] Submitted by: Peer Chen < pchen at nvidia dot com > [2], [3] Reviewed by: Shigeaki Tagashira < shigeaki AT se DOT hiroshima-u DOT ac DOT jp > Tested by: Shigeaki Tagashira, current Discussed with: current Silence from: obrien
2007-06-12 02:16:02 +00:00
for (ntries = 0; ntries < NFE_TIMEOUT; ntries++) {
Import the nfe (nForce Ethernet) driver from OpenBSD. This is a pure open source NIC driver with a BSD license. These bits were taken from the OpenBSD public CVS repository on 1-May-06.
2006-06-26 17:25:22 +00:00
DELAY(100);
if (!(NFE_READ(sc, NFE_PHY_CTL) & NFE_PHY_BUSY))
break;
}
#ifdef NFE_DEBUG
Bring overhauled nfe(4) into tree. o s/printf/device_printf/g o Nuke OpenBSDism. o Nuke NetBSD/OpenBSD specific DMA sync operations.(we don't have a way to sync a single descriptor within a DMA map.) o Remove recursive mutex. o bus_dma(9) clean up. o 40bit DMA address support. o Add protection for Rx map load failure. o Fix a long standing bug for watchdog timeout. [1] o Add additional protections, missing Tx completion interrupt, losing start Tx command, for watchdog timeout. o Switch to taskqueue(9) API to handle interrupts. o Use our own timer for watchdog instead of if_watchdog/if_timer interface. o Advertise VLAN header length/capability correctly to upper layer. o Remove excessive kernel stack consumption in nfe_encap(). o Handle highly fragmented mbuf chains correctly. o Enable etherenet address reprogramming with ifconfig(8). o Add ALTQ/TSO, MSI/MSIX support. o Increased Rx ring to 256 descriptors from 128. o Align Tx/Rx descriptor ring on sizeof(struct nfe_desc64) boundary. o Remove alignment restrictions on Tx/Rx buffers. o Rewritten jumbo frame support code. o Add support for hardware assistend VLAN tag insertion/stripping. o Add support for Tx/Rx flow control based on patches from Peer Chen. [2] o Add a routine that detects whether ethernet address swap routines is required. [3] o Add a workaround that take MAC/PHY out of power down mode. o Add suspend/resume support. o style(9) and code clean up. Special thanks to Shigeaki Tagashira, the original porter of nfe(4), who submitted lots of patches, performed uncountable number of regression tests and maintained nfe(4) for a long time. Without his enthusiastic help and support I could never have completed this overhauling task. The only weak point of nfe(4) compared to nve(4) is instability of manual half-duplex media selection on certain hardwares(auto sensing media type should work for all cases, though). This was a long standing bug of nfe(4) and I still have no idea why it doesn't work on some hardwares. Obtained from: OpenBSD [1] Submitted by: Peer Chen < pchen at nvidia dot com > [2], [3] Reviewed by: Shigeaki Tagashira < shigeaki AT se DOT hiroshima-u DOT ac DOT jp > Tested by: Shigeaki Tagashira, current Discussed with: current Silence from: obrien
2007-06-12 02:16:02 +00:00
if (nfedebug >= 2 && ntries == NFE_TIMEOUT)
device_printf(sc->nfe_dev, "could not write to PHY\n");
Import the nfe (nForce Ethernet) driver from OpenBSD. This is a pure open source NIC driver with a BSD license. These bits were taken from the OpenBSD public CVS repository on 1-May-06.
2006-06-26 17:25:22 +00:00
#endif
Bring overhauled nfe(4) into tree. o s/printf/device_printf/g o Nuke OpenBSDism. o Nuke NetBSD/OpenBSD specific DMA sync operations.(we don't have a way to sync a single descriptor within a DMA map.) o Remove recursive mutex. o bus_dma(9) clean up. o 40bit DMA address support. o Add protection for Rx map load failure. o Fix a long standing bug for watchdog timeout. [1] o Add additional protections, missing Tx completion interrupt, losing start Tx command, for watchdog timeout. o Switch to taskqueue(9) API to handle interrupts. o Use our own timer for watchdog instead of if_watchdog/if_timer interface. o Advertise VLAN header length/capability correctly to upper layer. o Remove excessive kernel stack consumption in nfe_encap(). o Handle highly fragmented mbuf chains correctly. o Enable etherenet address reprogramming with ifconfig(8). o Add ALTQ/TSO, MSI/MSIX support. o Increased Rx ring to 256 descriptors from 128. o Align Tx/Rx descriptor ring on sizeof(struct nfe_desc64) boundary. o Remove alignment restrictions on Tx/Rx buffers. o Rewritten jumbo frame support code. o Add support for hardware assistend VLAN tag insertion/stripping. o Add support for Tx/Rx flow control based on patches from Peer Chen. [2] o Add a routine that detects whether ethernet address swap routines is required. [3] o Add a workaround that take MAC/PHY out of power down mode. o Add suspend/resume support. o style(9) and code clean up. Special thanks to Shigeaki Tagashira, the original porter of nfe(4), who submitted lots of patches, performed uncountable number of regression tests and maintained nfe(4) for a long time. Without his enthusiastic help and support I could never have completed this overhauling task. The only weak point of nfe(4) compared to nve(4) is instability of manual half-duplex media selection on certain hardwares(auto sensing media type should work for all cases, though). This was a long standing bug of nfe(4) and I still have no idea why it doesn't work on some hardwares. Obtained from: OpenBSD [1] Submitted by: Peer Chen < pchen at nvidia dot com > [2], [3] Reviewed by: Shigeaki Tagashira < shigeaki AT se DOT hiroshima-u DOT ac DOT jp > Tested by: Shigeaki Tagashira, current Discussed with: current Silence from: obrien
2007-06-12 02:16:02 +00:00
return (0);
}
struct nfe_dmamap_arg {
bus_addr_t nfe_busaddr;
};
style(9) cleanup.
2006-10-19 10:01:26 +00:00
Add a pure open source nForce Ethernet driver, under BSDL. This driver was ported from OpenBSD by Shigeaki Tagashira <shigeaki@se.hiroshima-u.ac.jp> and posted at http://www.se.hiroshima-u.ac.jp/~shigeaki/software/freebsd-nfe.html It was additionally cleaned up by me. It is still a work-in-progress and thus is purposefully not in GENERIC. And it conflicts with nve(4), so only one should be loaded.
2006-06-26 23:41:07 +00:00
static int
nfe_alloc_rx_ring(struct nfe_softc *sc, struct nfe_rx_ring *ring)
Import the nfe (nForce Ethernet) driver from OpenBSD. This is a pure open source NIC driver with a BSD license. These bits were taken from the OpenBSD public CVS repository on 1-May-06.
2006-06-26 17:25:22 +00:00
{
Bring overhauled nfe(4) into tree. o s/printf/device_printf/g o Nuke OpenBSDism. o Nuke NetBSD/OpenBSD specific DMA sync operations.(we don't have a way to sync a single descriptor within a DMA map.) o Remove recursive mutex. o bus_dma(9) clean up. o 40bit DMA address support. o Add protection for Rx map load failure. o Fix a long standing bug for watchdog timeout. [1] o Add additional protections, missing Tx completion interrupt, losing start Tx command, for watchdog timeout. o Switch to taskqueue(9) API to handle interrupts. o Use our own timer for watchdog instead of if_watchdog/if_timer interface. o Advertise VLAN header length/capability correctly to upper layer. o Remove excessive kernel stack consumption in nfe_encap(). o Handle highly fragmented mbuf chains correctly. o Enable etherenet address reprogramming with ifconfig(8). o Add ALTQ/TSO, MSI/MSIX support. o Increased Rx ring to 256 descriptors from 128. o Align Tx/Rx descriptor ring on sizeof(struct nfe_desc64) boundary. o Remove alignment restrictions on Tx/Rx buffers. o Rewritten jumbo frame support code. o Add support for hardware assistend VLAN tag insertion/stripping. o Add support for Tx/Rx flow control based on patches from Peer Chen. [2] o Add a routine that detects whether ethernet address swap routines is required. [3] o Add a workaround that take MAC/PHY out of power down mode. o Add suspend/resume support. o style(9) and code clean up. Special thanks to Shigeaki Tagashira, the original porter of nfe(4), who submitted lots of patches, performed uncountable number of regression tests and maintained nfe(4) for a long time. Without his enthusiastic help and support I could never have completed this overhauling task. The only weak point of nfe(4) compared to nve(4) is instability of manual half-duplex media selection on certain hardwares(auto sensing media type should work for all cases, though). This was a long standing bug of nfe(4) and I still have no idea why it doesn't work on some hardwares. Obtained from: OpenBSD [1] Submitted by: Peer Chen < pchen at nvidia dot com > [2], [3] Reviewed by: Shigeaki Tagashira < shigeaki AT se DOT hiroshima-u DOT ac DOT jp > Tested by: Shigeaki Tagashira, current Discussed with: current Silence from: obrien
2007-06-12 02:16:02 +00:00
struct nfe_dmamap_arg ctx;
Add a pure open source nForce Ethernet driver, under BSDL. This driver was ported from OpenBSD by Shigeaki Tagashira <shigeaki@se.hiroshima-u.ac.jp> and posted at http://www.se.hiroshima-u.ac.jp/~shigeaki/software/freebsd-nfe.html It was additionally cleaned up by me. It is still a work-in-progress and thus is purposefully not in GENERIC. And it conflicts with nve(4), so only one should be loaded.
2006-06-26 23:41:07 +00:00
struct nfe_rx_data *data;
Bring overhauled nfe(4) into tree. o s/printf/device_printf/g o Nuke OpenBSDism. o Nuke NetBSD/OpenBSD specific DMA sync operations.(we don't have a way to sync a single descriptor within a DMA map.) o Remove recursive mutex. o bus_dma(9) clean up. o 40bit DMA address support. o Add protection for Rx map load failure. o Fix a long standing bug for watchdog timeout. [1] o Add additional protections, missing Tx completion interrupt, losing start Tx command, for watchdog timeout. o Switch to taskqueue(9) API to handle interrupts. o Use our own timer for watchdog instead of if_watchdog/if_timer interface. o Advertise VLAN header length/capability correctly to upper layer. o Remove excessive kernel stack consumption in nfe_encap(). o Handle highly fragmented mbuf chains correctly. o Enable etherenet address reprogramming with ifconfig(8). o Add ALTQ/TSO, MSI/MSIX support. o Increased Rx ring to 256 descriptors from 128. o Align Tx/Rx descriptor ring on sizeof(struct nfe_desc64) boundary. o Remove alignment restrictions on Tx/Rx buffers. o Rewritten jumbo frame support code. o Add support for hardware assistend VLAN tag insertion/stripping. o Add support for Tx/Rx flow control based on patches from Peer Chen. [2] o Add a routine that detects whether ethernet address swap routines is required. [3] o Add a workaround that take MAC/PHY out of power down mode. o Add suspend/resume support. o style(9) and code clean up. Special thanks to Shigeaki Tagashira, the original porter of nfe(4), who submitted lots of patches, performed uncountable number of regression tests and maintained nfe(4) for a long time. Without his enthusiastic help and support I could never have completed this overhauling task. The only weak point of nfe(4) compared to nve(4) is instability of manual half-duplex media selection on certain hardwares(auto sensing media type should work for all cases, though). This was a long standing bug of nfe(4) and I still have no idea why it doesn't work on some hardwares. Obtained from: OpenBSD [1] Submitted by: Peer Chen < pchen at nvidia dot com > [2], [3] Reviewed by: Shigeaki Tagashira < shigeaki AT se DOT hiroshima-u DOT ac DOT jp > Tested by: Shigeaki Tagashira, current Discussed with: current Silence from: obrien
2007-06-12 02:16:02 +00:00
void *desc;
Add a pure open source nForce Ethernet driver, under BSDL. This driver was ported from OpenBSD by Shigeaki Tagashira <shigeaki@se.hiroshima-u.ac.jp> and posted at http://www.se.hiroshima-u.ac.jp/~shigeaki/software/freebsd-nfe.html It was additionally cleaned up by me. It is still a work-in-progress and thus is purposefully not in GENERIC. And it conflicts with nve(4), so only one should be loaded.
2006-06-26 23:41:07 +00:00
int i, error, descsize;
Import the nfe (nForce Ethernet) driver from OpenBSD. This is a pure open source NIC driver with a BSD license. These bits were taken from the OpenBSD public CVS repository on 1-May-06.
2006-06-26 17:25:22 +00:00
Add a pure open source nForce Ethernet driver, under BSDL. This driver was ported from OpenBSD by Shigeaki Tagashira <shigeaki@se.hiroshima-u.ac.jp> and posted at http://www.se.hiroshima-u.ac.jp/~shigeaki/software/freebsd-nfe.html It was additionally cleaned up by me. It is still a work-in-progress and thus is purposefully not in GENERIC. And it conflicts with nve(4), so only one should be loaded.
2006-06-26 23:41:07 +00:00
if (sc->nfe_flags & NFE_40BIT_ADDR) {
Bring overhauled nfe(4) into tree. o s/printf/device_printf/g o Nuke OpenBSDism. o Nuke NetBSD/OpenBSD specific DMA sync operations.(we don't have a way to sync a single descriptor within a DMA map.) o Remove recursive mutex. o bus_dma(9) clean up. o 40bit DMA address support. o Add protection for Rx map load failure. o Fix a long standing bug for watchdog timeout. [1] o Add additional protections, missing Tx completion interrupt, losing start Tx command, for watchdog timeout. o Switch to taskqueue(9) API to handle interrupts. o Use our own timer for watchdog instead of if_watchdog/if_timer interface. o Advertise VLAN header length/capability correctly to upper layer. o Remove excessive kernel stack consumption in nfe_encap(). o Handle highly fragmented mbuf chains correctly. o Enable etherenet address reprogramming with ifconfig(8). o Add ALTQ/TSO, MSI/MSIX support. o Increased Rx ring to 256 descriptors from 128. o Align Tx/Rx descriptor ring on sizeof(struct nfe_desc64) boundary. o Remove alignment restrictions on Tx/Rx buffers. o Rewritten jumbo frame support code. o Add support for hardware assistend VLAN tag insertion/stripping. o Add support for Tx/Rx flow control based on patches from Peer Chen. [2] o Add a routine that detects whether ethernet address swap routines is required. [3] o Add a workaround that take MAC/PHY out of power down mode. o Add suspend/resume support. o style(9) and code clean up. Special thanks to Shigeaki Tagashira, the original porter of nfe(4), who submitted lots of patches, performed uncountable number of regression tests and maintained nfe(4) for a long time. Without his enthusiastic help and support I could never have completed this overhauling task. The only weak point of nfe(4) compared to nve(4) is instability of manual half-duplex media selection on certain hardwares(auto sensing media type should work for all cases, though). This was a long standing bug of nfe(4) and I still have no idea why it doesn't work on some hardwares. Obtained from: OpenBSD [1] Submitted by: Peer Chen < pchen at nvidia dot com > [2], [3] Reviewed by: Shigeaki Tagashira < shigeaki AT se DOT hiroshima-u DOT ac DOT jp > Tested by: Shigeaki Tagashira, current Discussed with: current Silence from: obrien
2007-06-12 02:16:02 +00:00
desc = ring->desc64;
Add a pure open source nForce Ethernet driver, under BSDL. This driver was ported from OpenBSD by Shigeaki Tagashira <shigeaki@se.hiroshima-u.ac.jp> and posted at http://www.se.hiroshima-u.ac.jp/~shigeaki/software/freebsd-nfe.html It was additionally cleaned up by me. It is still a work-in-progress and thus is purposefully not in GENERIC. And it conflicts with nve(4), so only one should be loaded.
2006-06-26 23:41:07 +00:00
descsize = sizeof (struct nfe_desc64);
} else {
Bring overhauled nfe(4) into tree. o s/printf/device_printf/g o Nuke OpenBSDism. o Nuke NetBSD/OpenBSD specific DMA sync operations.(we don't have a way to sync a single descriptor within a DMA map.) o Remove recursive mutex. o bus_dma(9) clean up. o 40bit DMA address support. o Add protection for Rx map load failure. o Fix a long standing bug for watchdog timeout. [1] o Add additional protections, missing Tx completion interrupt, losing start Tx command, for watchdog timeout. o Switch to taskqueue(9) API to handle interrupts. o Use our own timer for watchdog instead of if_watchdog/if_timer interface. o Advertise VLAN header length/capability correctly to upper layer. o Remove excessive kernel stack consumption in nfe_encap(). o Handle highly fragmented mbuf chains correctly. o Enable etherenet address reprogramming with ifconfig(8). o Add ALTQ/TSO, MSI/MSIX support. o Increased Rx ring to 256 descriptors from 128. o Align Tx/Rx descriptor ring on sizeof(struct nfe_desc64) boundary. o Remove alignment restrictions on Tx/Rx buffers. o Rewritten jumbo frame support code. o Add support for hardware assistend VLAN tag insertion/stripping. o Add support for Tx/Rx flow control based on patches from Peer Chen. [2] o Add a routine that detects whether ethernet address swap routines is required. [3] o Add a workaround that take MAC/PHY out of power down mode. o Add suspend/resume support. o style(9) and code clean up. Special thanks to Shigeaki Tagashira, the original porter of nfe(4), who submitted lots of patches, performed uncountable number of regression tests and maintained nfe(4) for a long time. Without his enthusiastic help and support I could never have completed this overhauling task. The only weak point of nfe(4) compared to nve(4) is instability of manual half-duplex media selection on certain hardwares(auto sensing media type should work for all cases, though). This was a long standing bug of nfe(4) and I still have no idea why it doesn't work on some hardwares. Obtained from: OpenBSD [1] Submitted by: Peer Chen < pchen at nvidia dot com > [2], [3] Reviewed by: Shigeaki Tagashira < shigeaki AT se DOT hiroshima-u DOT ac DOT jp > Tested by: Shigeaki Tagashira, current Discussed with: current Silence from: obrien
2007-06-12 02:16:02 +00:00
desc = ring->desc32;
Add a pure open source nForce Ethernet driver, under BSDL. This driver was ported from OpenBSD by Shigeaki Tagashira <shigeaki@se.hiroshima-u.ac.jp> and posted at http://www.se.hiroshima-u.ac.jp/~shigeaki/software/freebsd-nfe.html It was additionally cleaned up by me. It is still a work-in-progress and thus is purposefully not in GENERIC. And it conflicts with nve(4), so only one should be loaded.
2006-06-26 23:41:07 +00:00
descsize = sizeof (struct nfe_desc32);
}
Import the nfe (nForce Ethernet) driver from OpenBSD. This is a pure open source NIC driver with a BSD license. These bits were taken from the OpenBSD public CVS repository on 1-May-06.
2006-06-26 17:25:22 +00:00
Add a pure open source nForce Ethernet driver, under BSDL. This driver was ported from OpenBSD by Shigeaki Tagashira <shigeaki@se.hiroshima-u.ac.jp> and posted at http://www.se.hiroshima-u.ac.jp/~shigeaki/software/freebsd-nfe.html It was additionally cleaned up by me. It is still a work-in-progress and thus is purposefully not in GENERIC. And it conflicts with nve(4), so only one should be loaded.
2006-06-26 23:41:07 +00:00
ring->cur = ring->next = 0;
Import the nfe (nForce Ethernet) driver from OpenBSD. This is a pure open source NIC driver with a BSD license. These bits were taken from the OpenBSD public CVS repository on 1-May-06.
2006-06-26 17:25:22 +00:00
style(9) cleanup.
2006-10-19 10:01:26 +00:00
error = bus_dma_tag_create(sc->nfe_parent_tag,
Bring overhauled nfe(4) into tree. o s/printf/device_printf/g o Nuke OpenBSDism. o Nuke NetBSD/OpenBSD specific DMA sync operations.(we don't have a way to sync a single descriptor within a DMA map.) o Remove recursive mutex. o bus_dma(9) clean up. o 40bit DMA address support. o Add protection for Rx map load failure. o Fix a long standing bug for watchdog timeout. [1] o Add additional protections, missing Tx completion interrupt, losing start Tx command, for watchdog timeout. o Switch to taskqueue(9) API to handle interrupts. o Use our own timer for watchdog instead of if_watchdog/if_timer interface. o Advertise VLAN header length/capability correctly to upper layer. o Remove excessive kernel stack consumption in nfe_encap(). o Handle highly fragmented mbuf chains correctly. o Enable etherenet address reprogramming with ifconfig(8). o Add ALTQ/TSO, MSI/MSIX support. o Increased Rx ring to 256 descriptors from 128. o Align Tx/Rx descriptor ring on sizeof(struct nfe_desc64) boundary. o Remove alignment restrictions on Tx/Rx buffers. o Rewritten jumbo frame support code. o Add support for hardware assistend VLAN tag insertion/stripping. o Add support for Tx/Rx flow control based on patches from Peer Chen. [2] o Add a routine that detects whether ethernet address swap routines is required. [3] o Add a workaround that take MAC/PHY out of power down mode. o Add suspend/resume support. o style(9) and code clean up. Special thanks to Shigeaki Tagashira, the original porter of nfe(4), who submitted lots of patches, performed uncountable number of regression tests and maintained nfe(4) for a long time. Without his enthusiastic help and support I could never have completed this overhauling task. The only weak point of nfe(4) compared to nve(4) is instability of manual half-duplex media selection on certain hardwares(auto sensing media type should work for all cases, though). This was a long standing bug of nfe(4) and I still have no idea why it doesn't work on some hardwares. Obtained from: OpenBSD [1] Submitted by: Peer Chen < pchen at nvidia dot com > [2], [3] Reviewed by: Shigeaki Tagashira < shigeaki AT se DOT hiroshima-u DOT ac DOT jp > Tested by: Shigeaki Tagashira, current Discussed with: current Silence from: obrien
2007-06-12 02:16:02 +00:00
NFE_RING_ALIGN, 0, /* alignment, boundary */
BUS_SPACE_MAXADDR, /* lowaddr */
BUS_SPACE_MAXADDR, /* highaddr */
NULL, NULL, /* filter, filterarg */
NFE_RX_RING_COUNT * descsize, 1, /* maxsize, nsegments */
NFE_RX_RING_COUNT * descsize, /* maxsegsize */
0, /* flags */
NULL, NULL, /* lockfunc, lockarg */
&ring->rx_desc_tag);
Add a pure open source nForce Ethernet driver, under BSDL. This driver was ported from OpenBSD by Shigeaki Tagashira <shigeaki@se.hiroshima-u.ac.jp> and posted at http://www.se.hiroshima-u.ac.jp/~shigeaki/software/freebsd-nfe.html It was additionally cleaned up by me. It is still a work-in-progress and thus is purposefully not in GENERIC. And it conflicts with nve(4), so only one should be loaded.
2006-06-26 23:41:07 +00:00
if (error != 0) {
Bring overhauled nfe(4) into tree. o s/printf/device_printf/g o Nuke OpenBSDism. o Nuke NetBSD/OpenBSD specific DMA sync operations.(we don't have a way to sync a single descriptor within a DMA map.) o Remove recursive mutex. o bus_dma(9) clean up. o 40bit DMA address support. o Add protection for Rx map load failure. o Fix a long standing bug for watchdog timeout. [1] o Add additional protections, missing Tx completion interrupt, losing start Tx command, for watchdog timeout. o Switch to taskqueue(9) API to handle interrupts. o Use our own timer for watchdog instead of if_watchdog/if_timer interface. o Advertise VLAN header length/capability correctly to upper layer. o Remove excessive kernel stack consumption in nfe_encap(). o Handle highly fragmented mbuf chains correctly. o Enable etherenet address reprogramming with ifconfig(8). o Add ALTQ/TSO, MSI/MSIX support. o Increased Rx ring to 256 descriptors from 128. o Align Tx/Rx descriptor ring on sizeof(struct nfe_desc64) boundary. o Remove alignment restrictions on Tx/Rx buffers. o Rewritten jumbo frame support code. o Add support for hardware assistend VLAN tag insertion/stripping. o Add support for Tx/Rx flow control based on patches from Peer Chen. [2] o Add a routine that detects whether ethernet address swap routines is required. [3] o Add a workaround that take MAC/PHY out of power down mode. o Add suspend/resume support. o style(9) and code clean up. Special thanks to Shigeaki Tagashira, the original porter of nfe(4), who submitted lots of patches, performed uncountable number of regression tests and maintained nfe(4) for a long time. Without his enthusiastic help and support I could never have completed this overhauling task. The only weak point of nfe(4) compared to nve(4) is instability of manual half-duplex media selection on certain hardwares(auto sensing media type should work for all cases, though). This was a long standing bug of nfe(4) and I still have no idea why it doesn't work on some hardwares. Obtained from: OpenBSD [1] Submitted by: Peer Chen < pchen at nvidia dot com > [2], [3] Reviewed by: Shigeaki Tagashira < shigeaki AT se DOT hiroshima-u DOT ac DOT jp > Tested by: Shigeaki Tagashira, current Discussed with: current Silence from: obrien
2007-06-12 02:16:02 +00:00
device_printf(sc->nfe_dev, "could not create desc DMA tag\n");
Add a pure open source nForce Ethernet driver, under BSDL. This driver was ported from OpenBSD by Shigeaki Tagashira <shigeaki@se.hiroshima-u.ac.jp> and posted at http://www.se.hiroshima-u.ac.jp/~shigeaki/software/freebsd-nfe.html It was additionally cleaned up by me. It is still a work-in-progress and thus is purposefully not in GENERIC. And it conflicts with nve(4), so only one should be loaded.
2006-06-26 23:41:07 +00:00
goto fail;
Import the nfe (nForce Ethernet) driver from OpenBSD. This is a pure open source NIC driver with a BSD license. These bits were taken from the OpenBSD public CVS repository on 1-May-06.
2006-06-26 17:25:22 +00:00
}
Add a pure open source nForce Ethernet driver, under BSDL. This driver was ported from OpenBSD by Shigeaki Tagashira <shigeaki@se.hiroshima-u.ac.jp> and posted at http://www.se.hiroshima-u.ac.jp/~shigeaki/software/freebsd-nfe.html It was additionally cleaned up by me. It is still a work-in-progress and thus is purposefully not in GENERIC. And it conflicts with nve(4), so only one should be loaded.
2006-06-26 23:41:07 +00:00
/* allocate memory to desc */
Bring overhauled nfe(4) into tree. o s/printf/device_printf/g o Nuke OpenBSDism. o Nuke NetBSD/OpenBSD specific DMA sync operations.(we don't have a way to sync a single descriptor within a DMA map.) o Remove recursive mutex. o bus_dma(9) clean up. o 40bit DMA address support. o Add protection for Rx map load failure. o Fix a long standing bug for watchdog timeout. [1] o Add additional protections, missing Tx completion interrupt, losing start Tx command, for watchdog timeout. o Switch to taskqueue(9) API to handle interrupts. o Use our own timer for watchdog instead of if_watchdog/if_timer interface. o Advertise VLAN header length/capability correctly to upper layer. o Remove excessive kernel stack consumption in nfe_encap(). o Handle highly fragmented mbuf chains correctly. o Enable etherenet address reprogramming with ifconfig(8). o Add ALTQ/TSO, MSI/MSIX support. o Increased Rx ring to 256 descriptors from 128. o Align Tx/Rx descriptor ring on sizeof(struct nfe_desc64) boundary. o Remove alignment restrictions on Tx/Rx buffers. o Rewritten jumbo frame support code. o Add support for hardware assistend VLAN tag insertion/stripping. o Add support for Tx/Rx flow control based on patches from Peer Chen. [2] o Add a routine that detects whether ethernet address swap routines is required. [3] o Add a workaround that take MAC/PHY out of power down mode. o Add suspend/resume support. o style(9) and code clean up. Special thanks to Shigeaki Tagashira, the original porter of nfe(4), who submitted lots of patches, performed uncountable number of regression tests and maintained nfe(4) for a long time. Without his enthusiastic help and support I could never have completed this overhauling task. The only weak point of nfe(4) compared to nve(4) is instability of manual half-duplex media selection on certain hardwares(auto sensing media type should work for all cases, though). This was a long standing bug of nfe(4) and I still have no idea why it doesn't work on some hardwares. Obtained from: OpenBSD [1] Submitted by: Peer Chen < pchen at nvidia dot com > [2], [3] Reviewed by: Shigeaki Tagashira < shigeaki AT se DOT hiroshima-u DOT ac DOT jp > Tested by: Shigeaki Tagashira, current Discussed with: current Silence from: obrien
2007-06-12 02:16:02 +00:00
error = bus_dmamem_alloc(ring->rx_desc_tag, &desc, BUS_DMA_WAITOK |
BUS_DMA_COHERENT | BUS_DMA_ZERO, &ring->rx_desc_map);
Add a pure open source nForce Ethernet driver, under BSDL. This driver was ported from OpenBSD by Shigeaki Tagashira <shigeaki@se.hiroshima-u.ac.jp> and posted at http://www.se.hiroshima-u.ac.jp/~shigeaki/software/freebsd-nfe.html It was additionally cleaned up by me. It is still a work-in-progress and thus is purposefully not in GENERIC. And it conflicts with nve(4), so only one should be loaded.
2006-06-26 23:41:07 +00:00
if (error != 0) {
Bring overhauled nfe(4) into tree. o s/printf/device_printf/g o Nuke OpenBSDism. o Nuke NetBSD/OpenBSD specific DMA sync operations.(we don't have a way to sync a single descriptor within a DMA map.) o Remove recursive mutex. o bus_dma(9) clean up. o 40bit DMA address support. o Add protection for Rx map load failure. o Fix a long standing bug for watchdog timeout. [1] o Add additional protections, missing Tx completion interrupt, losing start Tx command, for watchdog timeout. o Switch to taskqueue(9) API to handle interrupts. o Use our own timer for watchdog instead of if_watchdog/if_timer interface. o Advertise VLAN header length/capability correctly to upper layer. o Remove excessive kernel stack consumption in nfe_encap(). o Handle highly fragmented mbuf chains correctly. o Enable etherenet address reprogramming with ifconfig(8). o Add ALTQ/TSO, MSI/MSIX support. o Increased Rx ring to 256 descriptors from 128. o Align Tx/Rx descriptor ring on sizeof(struct nfe_desc64) boundary. o Remove alignment restrictions on Tx/Rx buffers. o Rewritten jumbo frame support code. o Add support for hardware assistend VLAN tag insertion/stripping. o Add support for Tx/Rx flow control based on patches from Peer Chen. [2] o Add a routine that detects whether ethernet address swap routines is required. [3] o Add a workaround that take MAC/PHY out of power down mode. o Add suspend/resume support. o style(9) and code clean up. Special thanks to Shigeaki Tagashira, the original porter of nfe(4), who submitted lots of patches, performed uncountable number of regression tests and maintained nfe(4) for a long time. Without his enthusiastic help and support I could never have completed this overhauling task. The only weak point of nfe(4) compared to nve(4) is instability of manual half-duplex media selection on certain hardwares(auto sensing media type should work for all cases, though). This was a long standing bug of nfe(4) and I still have no idea why it doesn't work on some hardwares. Obtained from: OpenBSD [1] Submitted by: Peer Chen < pchen at nvidia dot com > [2], [3] Reviewed by: Shigeaki Tagashira < shigeaki AT se DOT hiroshima-u DOT ac DOT jp > Tested by: Shigeaki Tagashira, current Discussed with: current Silence from: obrien
2007-06-12 02:16:02 +00:00
device_printf(sc->nfe_dev, "could not create desc DMA map\n");
Add a pure open source nForce Ethernet driver, under BSDL. This driver was ported from OpenBSD by Shigeaki Tagashira <shigeaki@se.hiroshima-u.ac.jp> and posted at http://www.se.hiroshima-u.ac.jp/~shigeaki/software/freebsd-nfe.html It was additionally cleaned up by me. It is still a work-in-progress and thus is purposefully not in GENERIC. And it conflicts with nve(4), so only one should be loaded.
2006-06-26 23:41:07 +00:00
goto fail;
}
Bring overhauled nfe(4) into tree. o s/printf/device_printf/g o Nuke OpenBSDism. o Nuke NetBSD/OpenBSD specific DMA sync operations.(we don't have a way to sync a single descriptor within a DMA map.) o Remove recursive mutex. o bus_dma(9) clean up. o 40bit DMA address support. o Add protection for Rx map load failure. o Fix a long standing bug for watchdog timeout. [1] o Add additional protections, missing Tx completion interrupt, losing start Tx command, for watchdog timeout. o Switch to taskqueue(9) API to handle interrupts. o Use our own timer for watchdog instead of if_watchdog/if_timer interface. o Advertise VLAN header length/capability correctly to upper layer. o Remove excessive kernel stack consumption in nfe_encap(). o Handle highly fragmented mbuf chains correctly. o Enable etherenet address reprogramming with ifconfig(8). o Add ALTQ/TSO, MSI/MSIX support. o Increased Rx ring to 256 descriptors from 128. o Align Tx/Rx descriptor ring on sizeof(struct nfe_desc64) boundary. o Remove alignment restrictions on Tx/Rx buffers. o Rewritten jumbo frame support code. o Add support for hardware assistend VLAN tag insertion/stripping. o Add support for Tx/Rx flow control based on patches from Peer Chen. [2] o Add a routine that detects whether ethernet address swap routines is required. [3] o Add a workaround that take MAC/PHY out of power down mode. o Add suspend/resume support. o style(9) and code clean up. Special thanks to Shigeaki Tagashira, the original porter of nfe(4), who submitted lots of patches, performed uncountable number of regression tests and maintained nfe(4) for a long time. Without his enthusiastic help and support I could never have completed this overhauling task. The only weak point of nfe(4) compared to nve(4) is instability of manual half-duplex media selection on certain hardwares(auto sensing media type should work for all cases, though). This was a long standing bug of nfe(4) and I still have no idea why it doesn't work on some hardwares. Obtained from: OpenBSD [1] Submitted by: Peer Chen < pchen at nvidia dot com > [2], [3] Reviewed by: Shigeaki Tagashira < shigeaki AT se DOT hiroshima-u DOT ac DOT jp > Tested by: Shigeaki Tagashira, current Discussed with: current Silence from: obrien
2007-06-12 02:16:02 +00:00
if (sc->nfe_flags & NFE_40BIT_ADDR)
ring->desc64 = desc;
else
ring->desc32 = desc;
Import the nfe (nForce Ethernet) driver from OpenBSD. This is a pure open source NIC driver with a BSD license. These bits were taken from the OpenBSD public CVS repository on 1-May-06.
2006-06-26 17:25:22 +00:00
Add a pure open source nForce Ethernet driver, under BSDL. This driver was ported from OpenBSD by Shigeaki Tagashira <shigeaki@se.hiroshima-u.ac.jp> and posted at http://www.se.hiroshima-u.ac.jp/~shigeaki/software/freebsd-nfe.html It was additionally cleaned up by me. It is still a work-in-progress and thus is purposefully not in GENERIC. And it conflicts with nve(4), so only one should be loaded.
2006-06-26 23:41:07 +00:00
/* map desc to device visible address space */
Bring overhauled nfe(4) into tree. o s/printf/device_printf/g o Nuke OpenBSDism. o Nuke NetBSD/OpenBSD specific DMA sync operations.(we don't have a way to sync a single descriptor within a DMA map.) o Remove recursive mutex. o bus_dma(9) clean up. o 40bit DMA address support. o Add protection for Rx map load failure. o Fix a long standing bug for watchdog timeout. [1] o Add additional protections, missing Tx completion interrupt, losing start Tx command, for watchdog timeout. o Switch to taskqueue(9) API to handle interrupts. o Use our own timer for watchdog instead of if_watchdog/if_timer interface. o Advertise VLAN header length/capability correctly to upper layer. o Remove excessive kernel stack consumption in nfe_encap(). o Handle highly fragmented mbuf chains correctly. o Enable etherenet address reprogramming with ifconfig(8). o Add ALTQ/TSO, MSI/MSIX support. o Increased Rx ring to 256 descriptors from 128. o Align Tx/Rx descriptor ring on sizeof(struct nfe_desc64) boundary. o Remove alignment restrictions on Tx/Rx buffers. o Rewritten jumbo frame support code. o Add support for hardware assistend VLAN tag insertion/stripping. o Add support for Tx/Rx flow control based on patches from Peer Chen. [2] o Add a routine that detects whether ethernet address swap routines is required. [3] o Add a workaround that take MAC/PHY out of power down mode. o Add suspend/resume support. o style(9) and code clean up. Special thanks to Shigeaki Tagashira, the original porter of nfe(4), who submitted lots of patches, performed uncountable number of regression tests and maintained nfe(4) for a long time. Without his enthusiastic help and support I could never have completed this overhauling task. The only weak point of nfe(4) compared to nve(4) is instability of manual half-duplex media selection on certain hardwares(auto sensing media type should work for all cases, though). This was a long standing bug of nfe(4) and I still have no idea why it doesn't work on some hardwares. Obtained from: OpenBSD [1] Submitted by: Peer Chen < pchen at nvidia dot com > [2], [3] Reviewed by: Shigeaki Tagashira < shigeaki AT se DOT hiroshima-u DOT ac DOT jp > Tested by: Shigeaki Tagashira, current Discussed with: current Silence from: obrien
2007-06-12 02:16:02 +00:00
ctx.nfe_busaddr = 0;
error = bus_dmamap_load(ring->rx_desc_tag, ring->rx_desc_map, desc,
NFE_RX_RING_COUNT * descsize, nfe_dma_map_segs, &ctx, 0);
if (error != 0) {
device_printf(sc->nfe_dev, "could not load desc DMA map\n");
goto fail;
}
ring->physaddr = ctx.nfe_busaddr;
error = bus_dma_tag_create(sc->nfe_parent_tag,
1, 0, /* alignment, boundary */
BUS_SPACE_MAXADDR, /* lowaddr */
BUS_SPACE_MAXADDR, /* highaddr */
NULL, NULL, /* filter, filterarg */
MCLBYTES, 1, /* maxsize, nsegments */
MCLBYTES, /* maxsegsize */
0, /* flags */
NULL, NULL, /* lockfunc, lockarg */
&ring->rx_data_tag);
Add a pure open source nForce Ethernet driver, under BSDL. This driver was ported from OpenBSD by Shigeaki Tagashira <shigeaki@se.hiroshima-u.ac.jp> and posted at http://www.se.hiroshima-u.ac.jp/~shigeaki/software/freebsd-nfe.html It was additionally cleaned up by me. It is still a work-in-progress and thus is purposefully not in GENERIC. And it conflicts with nve(4), so only one should be loaded.
2006-06-26 23:41:07 +00:00
if (error != 0) {
Bring overhauled nfe(4) into tree. o s/printf/device_printf/g o Nuke OpenBSDism. o Nuke NetBSD/OpenBSD specific DMA sync operations.(we don't have a way to sync a single descriptor within a DMA map.) o Remove recursive mutex. o bus_dma(9) clean up. o 40bit DMA address support. o Add protection for Rx map load failure. o Fix a long standing bug for watchdog timeout. [1] o Add additional protections, missing Tx completion interrupt, losing start Tx command, for watchdog timeout. o Switch to taskqueue(9) API to handle interrupts. o Use our own timer for watchdog instead of if_watchdog/if_timer interface. o Advertise VLAN header length/capability correctly to upper layer. o Remove excessive kernel stack consumption in nfe_encap(). o Handle highly fragmented mbuf chains correctly. o Enable etherenet address reprogramming with ifconfig(8). o Add ALTQ/TSO, MSI/MSIX support. o Increased Rx ring to 256 descriptors from 128. o Align Tx/Rx descriptor ring on sizeof(struct nfe_desc64) boundary. o Remove alignment restrictions on Tx/Rx buffers. o Rewritten jumbo frame support code. o Add support for hardware assistend VLAN tag insertion/stripping. o Add support for Tx/Rx flow control based on patches from Peer Chen. [2] o Add a routine that detects whether ethernet address swap routines is required. [3] o Add a workaround that take MAC/PHY out of power down mode. o Add suspend/resume support. o style(9) and code clean up. Special thanks to Shigeaki Tagashira, the original porter of nfe(4), who submitted lots of patches, performed uncountable number of regression tests and maintained nfe(4) for a long time. Without his enthusiastic help and support I could never have completed this overhauling task. The only weak point of nfe(4) compared to nve(4) is instability of manual half-duplex media selection on certain hardwares(auto sensing media type should work for all cases, though). This was a long standing bug of nfe(4) and I still have no idea why it doesn't work on some hardwares. Obtained from: OpenBSD [1] Submitted by: Peer Chen < pchen at nvidia dot com > [2], [3] Reviewed by: Shigeaki Tagashira < shigeaki AT se DOT hiroshima-u DOT ac DOT jp > Tested by: Shigeaki Tagashira, current Discussed with: current Silence from: obrien
2007-06-12 02:16:02 +00:00
device_printf(sc->nfe_dev, "could not create Rx DMA tag\n");
Add a pure open source nForce Ethernet driver, under BSDL. This driver was ported from OpenBSD by Shigeaki Tagashira <shigeaki@se.hiroshima-u.ac.jp> and posted at http://www.se.hiroshima-u.ac.jp/~shigeaki/software/freebsd-nfe.html It was additionally cleaned up by me. It is still a work-in-progress and thus is purposefully not in GENERIC. And it conflicts with nve(4), so only one should be loaded.
2006-06-26 23:41:07 +00:00
goto fail;
Import the nfe (nForce Ethernet) driver from OpenBSD. This is a pure open source NIC driver with a BSD license. These bits were taken from the OpenBSD public CVS repository on 1-May-06.
2006-06-26 17:25:22 +00:00
}
Bring overhauled nfe(4) into tree. o s/printf/device_printf/g o Nuke OpenBSDism. o Nuke NetBSD/OpenBSD specific DMA sync operations.(we don't have a way to sync a single descriptor within a DMA map.) o Remove recursive mutex. o bus_dma(9) clean up. o 40bit DMA address support. o Add protection for Rx map load failure. o Fix a long standing bug for watchdog timeout. [1] o Add additional protections, missing Tx completion interrupt, losing start Tx command, for watchdog timeout. o Switch to taskqueue(9) API to handle interrupts. o Use our own timer for watchdog instead of if_watchdog/if_timer interface. o Advertise VLAN header length/capability correctly to upper layer. o Remove excessive kernel stack consumption in nfe_encap(). o Handle highly fragmented mbuf chains correctly. o Enable etherenet address reprogramming with ifconfig(8). o Add ALTQ/TSO, MSI/MSIX support. o Increased Rx ring to 256 descriptors from 128. o Align Tx/Rx descriptor ring on sizeof(struct nfe_desc64) boundary. o Remove alignment restrictions on Tx/Rx buffers. o Rewritten jumbo frame support code. o Add support for hardware assistend VLAN tag insertion/stripping. o Add support for Tx/Rx flow control based on patches from Peer Chen. [2] o Add a routine that detects whether ethernet address swap routines is required. [3] o Add a workaround that take MAC/PHY out of power down mode. o Add suspend/resume support. o style(9) and code clean up. Special thanks to Shigeaki Tagashira, the original porter of nfe(4), who submitted lots of patches, performed uncountable number of regression tests and maintained nfe(4) for a long time. Without his enthusiastic help and support I could never have completed this overhauling task. The only weak point of nfe(4) compared to nve(4) is instability of manual half-duplex media selection on certain hardwares(auto sensing media type should work for all cases, though). This was a long standing bug of nfe(4) and I still have no idea why it doesn't work on some hardwares. Obtained from: OpenBSD [1] Submitted by: Peer Chen < pchen at nvidia dot com > [2], [3] Reviewed by: Shigeaki Tagashira < shigeaki AT se DOT hiroshima-u DOT ac DOT jp > Tested by: Shigeaki Tagashira, current Discussed with: current Silence from: obrien
2007-06-12 02:16:02 +00:00
error = bus_dmamap_create(ring->rx_data_tag, 0, &ring->rx_spare_map);
if (error != 0) {
device_printf(sc->nfe_dev,
"could not create Rx DMA spare map\n");
goto fail;
}
Import the nfe (nForce Ethernet) driver from OpenBSD. This is a pure open source NIC driver with a BSD license. These bits were taken from the OpenBSD public CVS repository on 1-May-06.
2006-06-26 17:25:22 +00:00
Add a pure open source nForce Ethernet driver, under BSDL. This driver was ported from OpenBSD by Shigeaki Tagashira <shigeaki@se.hiroshima-u.ac.jp> and posted at http://www.se.hiroshima-u.ac.jp/~shigeaki/software/freebsd-nfe.html It was additionally cleaned up by me. It is still a work-in-progress and thus is purposefully not in GENERIC. And it conflicts with nve(4), so only one should be loaded.
2006-06-26 23:41:07 +00:00
/*
* Pre-allocate Rx buffers and populate Rx ring.
*/
for (i = 0; i < NFE_RX_RING_COUNT; i++) {
data = &sc->rxq.data[i];
Bring overhauled nfe(4) into tree. o s/printf/device_printf/g o Nuke OpenBSDism. o Nuke NetBSD/OpenBSD specific DMA sync operations.(we don't have a way to sync a single descriptor within a DMA map.) o Remove recursive mutex. o bus_dma(9) clean up. o 40bit DMA address support. o Add protection for Rx map load failure. o Fix a long standing bug for watchdog timeout. [1] o Add additional protections, missing Tx completion interrupt, losing start Tx command, for watchdog timeout. o Switch to taskqueue(9) API to handle interrupts. o Use our own timer for watchdog instead of if_watchdog/if_timer interface. o Advertise VLAN header length/capability correctly to upper layer. o Remove excessive kernel stack consumption in nfe_encap(). o Handle highly fragmented mbuf chains correctly. o Enable etherenet address reprogramming with ifconfig(8). o Add ALTQ/TSO, MSI/MSIX support. o Increased Rx ring to 256 descriptors from 128. o Align Tx/Rx descriptor ring on sizeof(struct nfe_desc64) boundary. o Remove alignment restrictions on Tx/Rx buffers. o Rewritten jumbo frame support code. o Add support for hardware assistend VLAN tag insertion/stripping. o Add support for Tx/Rx flow control based on patches from Peer Chen. [2] o Add a routine that detects whether ethernet address swap routines is required. [3] o Add a workaround that take MAC/PHY out of power down mode. o Add suspend/resume support. o style(9) and code clean up. Special thanks to Shigeaki Tagashira, the original porter of nfe(4), who submitted lots of patches, performed uncountable number of regression tests and maintained nfe(4) for a long time. Without his enthusiastic help and support I could never have completed this overhauling task. The only weak point of nfe(4) compared to nve(4) is instability of manual half-duplex media selection on certain hardwares(auto sensing media type should work for all cases, though). This was a long standing bug of nfe(4) and I still have no idea why it doesn't work on some hardwares. Obtained from: OpenBSD [1] Submitted by: Peer Chen < pchen at nvidia dot com > [2], [3] Reviewed by: Shigeaki Tagashira < shigeaki AT se DOT hiroshima-u DOT ac DOT jp > Tested by: Shigeaki Tagashira, current Discussed with: current Silence from: obrien
2007-06-12 02:16:02 +00:00
data->rx_data_map = NULL;
data->m = NULL;
error = bus_dmamap_create(ring->rx_data_tag, 0,
Reformat to KNF after revision 1.12.
2006-11-27 00:23:59 +00:00
&data->rx_data_map);
if (error != 0) {
Bring overhauled nfe(4) into tree. o s/printf/device_printf/g o Nuke OpenBSDism. o Nuke NetBSD/OpenBSD specific DMA sync operations.(we don't have a way to sync a single descriptor within a DMA map.) o Remove recursive mutex. o bus_dma(9) clean up. o 40bit DMA address support. o Add protection for Rx map load failure. o Fix a long standing bug for watchdog timeout. [1] o Add additional protections, missing Tx completion interrupt, losing start Tx command, for watchdog timeout. o Switch to taskqueue(9) API to handle interrupts. o Use our own timer for watchdog instead of if_watchdog/if_timer interface. o Advertise VLAN header length/capability correctly to upper layer. o Remove excessive kernel stack consumption in nfe_encap(). o Handle highly fragmented mbuf chains correctly. o Enable etherenet address reprogramming with ifconfig(8). o Add ALTQ/TSO, MSI/MSIX support. o Increased Rx ring to 256 descriptors from 128. o Align Tx/Rx descriptor ring on sizeof(struct nfe_desc64) boundary. o Remove alignment restrictions on Tx/Rx buffers. o Rewritten jumbo frame support code. o Add support for hardware assistend VLAN tag insertion/stripping. o Add support for Tx/Rx flow control based on patches from Peer Chen. [2] o Add a routine that detects whether ethernet address swap routines is required. [3] o Add a workaround that take MAC/PHY out of power down mode. o Add suspend/resume support. o style(9) and code clean up. Special thanks to Shigeaki Tagashira, the original porter of nfe(4), who submitted lots of patches, performed uncountable number of regression tests and maintained nfe(4) for a long time. Without his enthusiastic help and support I could never have completed this overhauling task. The only weak point of nfe(4) compared to nve(4) is instability of manual half-duplex media selection on certain hardwares(auto sensing media type should work for all cases, though). This was a long standing bug of nfe(4) and I still have no idea why it doesn't work on some hardwares. Obtained from: OpenBSD [1] Submitted by: Peer Chen < pchen at nvidia dot com > [2], [3] Reviewed by: Shigeaki Tagashira < shigeaki AT se DOT hiroshima-u DOT ac DOT jp > Tested by: Shigeaki Tagashira, current Discussed with: current Silence from: obrien
2007-06-12 02:16:02 +00:00
device_printf(sc->nfe_dev,
"could not create Rx DMA map\n");
Reformat to KNF after revision 1.12.
2006-11-27 00:23:59 +00:00
goto fail;
}
Add a pure open source nForce Ethernet driver, under BSDL. This driver was ported from OpenBSD by Shigeaki Tagashira <shigeaki@se.hiroshima-u.ac.jp> and posted at http://www.se.hiroshima-u.ac.jp/~shigeaki/software/freebsd-nfe.html It was additionally cleaned up by me. It is still a work-in-progress and thus is purposefully not in GENERIC. And it conflicts with nve(4), so only one should be loaded.
2006-06-26 23:41:07 +00:00
}
Bring overhauled nfe(4) into tree. o s/printf/device_printf/g o Nuke OpenBSDism. o Nuke NetBSD/OpenBSD specific DMA sync operations.(we don't have a way to sync a single descriptor within a DMA map.) o Remove recursive mutex. o bus_dma(9) clean up. o 40bit DMA address support. o Add protection for Rx map load failure. o Fix a long standing bug for watchdog timeout. [1] o Add additional protections, missing Tx completion interrupt, losing start Tx command, for watchdog timeout. o Switch to taskqueue(9) API to handle interrupts. o Use our own timer for watchdog instead of if_watchdog/if_timer interface. o Advertise VLAN header length/capability correctly to upper layer. o Remove excessive kernel stack consumption in nfe_encap(). o Handle highly fragmented mbuf chains correctly. o Enable etherenet address reprogramming with ifconfig(8). o Add ALTQ/TSO, MSI/MSIX support. o Increased Rx ring to 256 descriptors from 128. o Align Tx/Rx descriptor ring on sizeof(struct nfe_desc64) boundary. o Remove alignment restrictions on Tx/Rx buffers. o Rewritten jumbo frame support code. o Add support for hardware assistend VLAN tag insertion/stripping. o Add support for Tx/Rx flow control based on patches from Peer Chen. [2] o Add a routine that detects whether ethernet address swap routines is required. [3] o Add a workaround that take MAC/PHY out of power down mode. o Add suspend/resume support. o style(9) and code clean up. Special thanks to Shigeaki Tagashira, the original porter of nfe(4), who submitted lots of patches, performed uncountable number of regression tests and maintained nfe(4) for a long time. Without his enthusiastic help and support I could never have completed this overhauling task. The only weak point of nfe(4) compared to nve(4) is instability of manual half-duplex media selection on certain hardwares(auto sensing media type should work for all cases, though). This was a long standing bug of nfe(4) and I still have no idea why it doesn't work on some hardwares. Obtained from: OpenBSD [1] Submitted by: Peer Chen < pchen at nvidia dot com > [2], [3] Reviewed by: Shigeaki Tagashira < shigeaki AT se DOT hiroshima-u DOT ac DOT jp > Tested by: Shigeaki Tagashira, current Discussed with: current Silence from: obrien
2007-06-12 02:16:02 +00:00
fail:
return (error);
Add a pure open source nForce Ethernet driver, under BSDL. This driver was ported from OpenBSD by Shigeaki Tagashira <shigeaki@se.hiroshima-u.ac.jp> and posted at http://www.se.hiroshima-u.ac.jp/~shigeaki/software/freebsd-nfe.html It was additionally cleaned up by me. It is still a work-in-progress and thus is purposefully not in GENERIC. And it conflicts with nve(4), so only one should be loaded.
2006-06-26 23:41:07 +00:00
}
Don't fail on device attach if jumbo frame support was unsuccessful. Because nfe(4) hardware doesn't support SG on Rx path, supporting jumbo frame requires very large contiguous kernel memory(i.e. several mega bytes). In case of lack of contiguous kernel memory that allocation request may always fail. However nfe(4) can operate on normal sized MTU frames, so go ahead and just disable jumbo frame support. While I'm here add a new tunable "hw.nfe.jumbo_disable" to disable jumbo frame support. In nfe_poll, make sure to invoke correct Rx handler. Approved by: re (kensmith)
2007-07-24 01:11:00 +00:00
static void
Bring overhauled nfe(4) into tree. o s/printf/device_printf/g o Nuke OpenBSDism. o Nuke NetBSD/OpenBSD specific DMA sync operations.(we don't have a way to sync a single descriptor within a DMA map.) o Remove recursive mutex. o bus_dma(9) clean up. o 40bit DMA address support. o Add protection for Rx map load failure. o Fix a long standing bug for watchdog timeout. [1] o Add additional protections, missing Tx completion interrupt, losing start Tx command, for watchdog timeout. o Switch to taskqueue(9) API to handle interrupts. o Use our own timer for watchdog instead of if_watchdog/if_timer interface. o Advertise VLAN header length/capability correctly to upper layer. o Remove excessive kernel stack consumption in nfe_encap(). o Handle highly fragmented mbuf chains correctly. o Enable etherenet address reprogramming with ifconfig(8). o Add ALTQ/TSO, MSI/MSIX support. o Increased Rx ring to 256 descriptors from 128. o Align Tx/Rx descriptor ring on sizeof(struct nfe_desc64) boundary. o Remove alignment restrictions on Tx/Rx buffers. o Rewritten jumbo frame support code. o Add support for hardware assistend VLAN tag insertion/stripping. o Add support for Tx/Rx flow control based on patches from Peer Chen. [2] o Add a routine that detects whether ethernet address swap routines is required. [3] o Add a workaround that take MAC/PHY out of power down mode. o Add suspend/resume support. o style(9) and code clean up. Special thanks to Shigeaki Tagashira, the original porter of nfe(4), who submitted lots of patches, performed uncountable number of regression tests and maintained nfe(4) for a long time. Without his enthusiastic help and support I could never have completed this overhauling task. The only weak point of nfe(4) compared to nve(4) is instability of manual half-duplex media selection on certain hardwares(auto sensing media type should work for all cases, though). This was a long standing bug of nfe(4) and I still have no idea why it doesn't work on some hardwares. Obtained from: OpenBSD [1] Submitted by: Peer Chen < pchen at nvidia dot com > [2], [3] Reviewed by: Shigeaki Tagashira < shigeaki AT se DOT hiroshima-u DOT ac DOT jp > Tested by: Shigeaki Tagashira, current Discussed with: current Silence from: obrien
2007-06-12 02:16:02 +00:00
nfe_alloc_jrx_ring(struct nfe_softc *sc, struct nfe_jrx_ring *ring)
Add a pure open source nForce Ethernet driver, under BSDL. This driver was ported from OpenBSD by Shigeaki Tagashira <shigeaki@se.hiroshima-u.ac.jp> and posted at http://www.se.hiroshima-u.ac.jp/~shigeaki/software/freebsd-nfe.html It was additionally cleaned up by me. It is still a work-in-progress and thus is purposefully not in GENERIC. And it conflicts with nve(4), so only one should be loaded.
2006-06-26 23:41:07 +00:00
{
Bring overhauled nfe(4) into tree. o s/printf/device_printf/g o Nuke OpenBSDism. o Nuke NetBSD/OpenBSD specific DMA sync operations.(we don't have a way to sync a single descriptor within a DMA map.) o Remove recursive mutex. o bus_dma(9) clean up. o 40bit DMA address support. o Add protection for Rx map load failure. o Fix a long standing bug for watchdog timeout. [1] o Add additional protections, missing Tx completion interrupt, losing start Tx command, for watchdog timeout. o Switch to taskqueue(9) API to handle interrupts. o Use our own timer for watchdog instead of if_watchdog/if_timer interface. o Advertise VLAN header length/capability correctly to upper layer. o Remove excessive kernel stack consumption in nfe_encap(). o Handle highly fragmented mbuf chains correctly. o Enable etherenet address reprogramming with ifconfig(8). o Add ALTQ/TSO, MSI/MSIX support. o Increased Rx ring to 256 descriptors from 128. o Align Tx/Rx descriptor ring on sizeof(struct nfe_desc64) boundary. o Remove alignment restrictions on Tx/Rx buffers. o Rewritten jumbo frame support code. o Add support for hardware assistend VLAN tag insertion/stripping. o Add support for Tx/Rx flow control based on patches from Peer Chen. [2] o Add a routine that detects whether ethernet address swap routines is required. [3] o Add a workaround that take MAC/PHY out of power down mode. o Add suspend/resume support. o style(9) and code clean up. Special thanks to Shigeaki Tagashira, the original porter of nfe(4), who submitted lots of patches, performed uncountable number of regression tests and maintained nfe(4) for a long time. Without his enthusiastic help and support I could never have completed this overhauling task. The only weak point of nfe(4) compared to nve(4) is instability of manual half-duplex media selection on certain hardwares(auto sensing media type should work for all cases, though). This was a long standing bug of nfe(4) and I still have no idea why it doesn't work on some hardwares. Obtained from: OpenBSD [1] Submitted by: Peer Chen < pchen at nvidia dot com > [2], [3] Reviewed by: Shigeaki Tagashira < shigeaki AT se DOT hiroshima-u DOT ac DOT jp > Tested by: Shigeaki Tagashira, current Discussed with: current Silence from: obrien
2007-06-12 02:16:02 +00:00
struct nfe_dmamap_arg ctx;
Add a pure open source nForce Ethernet driver, under BSDL. This driver was ported from OpenBSD by Shigeaki Tagashira <shigeaki@se.hiroshima-u.ac.jp> and posted at http://www.se.hiroshima-u.ac.jp/~shigeaki/software/freebsd-nfe.html It was additionally cleaned up by me. It is still a work-in-progress and thus is purposefully not in GENERIC. And it conflicts with nve(4), so only one should be loaded.
2006-06-26 23:41:07 +00:00
struct nfe_rx_data *data;
void *desc;
Bring overhauled nfe(4) into tree. o s/printf/device_printf/g o Nuke OpenBSDism. o Nuke NetBSD/OpenBSD specific DMA sync operations.(we don't have a way to sync a single descriptor within a DMA map.) o Remove recursive mutex. o bus_dma(9) clean up. o 40bit DMA address support. o Add protection for Rx map load failure. o Fix a long standing bug for watchdog timeout. [1] o Add additional protections, missing Tx completion interrupt, losing start Tx command, for watchdog timeout. o Switch to taskqueue(9) API to handle interrupts. o Use our own timer for watchdog instead of if_watchdog/if_timer interface. o Advertise VLAN header length/capability correctly to upper layer. o Remove excessive kernel stack consumption in nfe_encap(). o Handle highly fragmented mbuf chains correctly. o Enable etherenet address reprogramming with ifconfig(8). o Add ALTQ/TSO, MSI/MSIX support. o Increased Rx ring to 256 descriptors from 128. o Align Tx/Rx descriptor ring on sizeof(struct nfe_desc64) boundary. o Remove alignment restrictions on Tx/Rx buffers. o Rewritten jumbo frame support code. o Add support for hardware assistend VLAN tag insertion/stripping. o Add support for Tx/Rx flow control based on patches from Peer Chen. [2] o Add a routine that detects whether ethernet address swap routines is required. [3] o Add a workaround that take MAC/PHY out of power down mode. o Add suspend/resume support. o style(9) and code clean up. Special thanks to Shigeaki Tagashira, the original porter of nfe(4), who submitted lots of patches, performed uncountable number of regression tests and maintained nfe(4) for a long time. Without his enthusiastic help and support I could never have completed this overhauling task. The only weak point of nfe(4) compared to nve(4) is instability of manual half-duplex media selection on certain hardwares(auto sensing media type should work for all cases, though). This was a long standing bug of nfe(4) and I still have no idea why it doesn't work on some hardwares. Obtained from: OpenBSD [1] Submitted by: Peer Chen < pchen at nvidia dot com > [2], [3] Reviewed by: Shigeaki Tagashira < shigeaki AT se DOT hiroshima-u DOT ac DOT jp > Tested by: Shigeaki Tagashira, current Discussed with: current Silence from: obrien
2007-06-12 02:16:02 +00:00
int i, error, descsize;
if ((sc->nfe_flags & NFE_JUMBO_SUP) == 0)
Don't fail on device attach if jumbo frame support was unsuccessful. Because nfe(4) hardware doesn't support SG on Rx path, supporting jumbo frame requires very large contiguous kernel memory(i.e. several mega bytes). In case of lack of contiguous kernel memory that allocation request may always fail. However nfe(4) can operate on normal sized MTU frames, so go ahead and just disable jumbo frame support. While I'm here add a new tunable "hw.nfe.jumbo_disable" to disable jumbo frame support. In nfe_poll, make sure to invoke correct Rx handler. Approved by: re (kensmith)
2007-07-24 01:11:00 +00:00
return;
if (jumbo_disable != 0) {
device_printf(sc->nfe_dev, "disabling jumbo frame support\n");
sc->nfe_jumbo_disable = 1;
return;
}
Add a pure open source nForce Ethernet driver, under BSDL. This driver was ported from OpenBSD by Shigeaki Tagashira <shigeaki@se.hiroshima-u.ac.jp> and posted at http://www.se.hiroshima-u.ac.jp/~shigeaki/software/freebsd-nfe.html It was additionally cleaned up by me. It is still a work-in-progress and thus is purposefully not in GENERIC. And it conflicts with nve(4), so only one should be loaded.
2006-06-26 23:41:07 +00:00
if (sc->nfe_flags & NFE_40BIT_ADDR) {
Bring overhauled nfe(4) into tree. o s/printf/device_printf/g o Nuke OpenBSDism. o Nuke NetBSD/OpenBSD specific DMA sync operations.(we don't have a way to sync a single descriptor within a DMA map.) o Remove recursive mutex. o bus_dma(9) clean up. o 40bit DMA address support. o Add protection for Rx map load failure. o Fix a long standing bug for watchdog timeout. [1] o Add additional protections, missing Tx completion interrupt, losing start Tx command, for watchdog timeout. o Switch to taskqueue(9) API to handle interrupts. o Use our own timer for watchdog instead of if_watchdog/if_timer interface. o Advertise VLAN header length/capability correctly to upper layer. o Remove excessive kernel stack consumption in nfe_encap(). o Handle highly fragmented mbuf chains correctly. o Enable etherenet address reprogramming with ifconfig(8). o Add ALTQ/TSO, MSI/MSIX support. o Increased Rx ring to 256 descriptors from 128. o Align Tx/Rx descriptor ring on sizeof(struct nfe_desc64) boundary. o Remove alignment restrictions on Tx/Rx buffers. o Rewritten jumbo frame support code. o Add support for hardware assistend VLAN tag insertion/stripping. o Add support for Tx/Rx flow control based on patches from Peer Chen. [2] o Add a routine that detects whether ethernet address swap routines is required. [3] o Add a workaround that take MAC/PHY out of power down mode. o Add suspend/resume support. o style(9) and code clean up. Special thanks to Shigeaki Tagashira, the original porter of nfe(4), who submitted lots of patches, performed uncountable number of regression tests and maintained nfe(4) for a long time. Without his enthusiastic help and support I could never have completed this overhauling task. The only weak point of nfe(4) compared to nve(4) is instability of manual half-duplex media selection on certain hardwares(auto sensing media type should work for all cases, though). This was a long standing bug of nfe(4) and I still have no idea why it doesn't work on some hardwares. Obtained from: OpenBSD [1] Submitted by: Peer Chen < pchen at nvidia dot com > [2], [3] Reviewed by: Shigeaki Tagashira < shigeaki AT se DOT hiroshima-u DOT ac DOT jp > Tested by: Shigeaki Tagashira, current Discussed with: current Silence from: obrien
2007-06-12 02:16:02 +00:00
desc = ring->jdesc64;
Add a pure open source nForce Ethernet driver, under BSDL. This driver was ported from OpenBSD by Shigeaki Tagashira <shigeaki@se.hiroshima-u.ac.jp> and posted at http://www.se.hiroshima-u.ac.jp/~shigeaki/software/freebsd-nfe.html It was additionally cleaned up by me. It is still a work-in-progress and thus is purposefully not in GENERIC. And it conflicts with nve(4), so only one should be loaded.
2006-06-26 23:41:07 +00:00
descsize = sizeof (struct nfe_desc64);
} else {
Bring overhauled nfe(4) into tree. o s/printf/device_printf/g o Nuke OpenBSDism. o Nuke NetBSD/OpenBSD specific DMA sync operations.(we don't have a way to sync a single descriptor within a DMA map.) o Remove recursive mutex. o bus_dma(9) clean up. o 40bit DMA address support. o Add protection for Rx map load failure. o Fix a long standing bug for watchdog timeout. [1] o Add additional protections, missing Tx completion interrupt, losing start Tx command, for watchdog timeout. o Switch to taskqueue(9) API to handle interrupts. o Use our own timer for watchdog instead of if_watchdog/if_timer interface. o Advertise VLAN header length/capability correctly to upper layer. o Remove excessive kernel stack consumption in nfe_encap(). o Handle highly fragmented mbuf chains correctly. o Enable etherenet address reprogramming with ifconfig(8). o Add ALTQ/TSO, MSI/MSIX support. o Increased Rx ring to 256 descriptors from 128. o Align Tx/Rx descriptor ring on sizeof(struct nfe_desc64) boundary. o Remove alignment restrictions on Tx/Rx buffers. o Rewritten jumbo frame support code. o Add support for hardware assistend VLAN tag insertion/stripping. o Add support for Tx/Rx flow control based on patches from Peer Chen. [2] o Add a routine that detects whether ethernet address swap routines is required. [3] o Add a workaround that take MAC/PHY out of power down mode. o Add suspend/resume support. o style(9) and code clean up. Special thanks to Shigeaki Tagashira, the original porter of nfe(4), who submitted lots of patches, performed uncountable number of regression tests and maintained nfe(4) for a long time. Without his enthusiastic help and support I could never have completed this overhauling task. The only weak point of nfe(4) compared to nve(4) is instability of manual half-duplex media selection on certain hardwares(auto sensing media type should work for all cases, though). This was a long standing bug of nfe(4) and I still have no idea why it doesn't work on some hardwares. Obtained from: OpenBSD [1] Submitted by: Peer Chen < pchen at nvidia dot com > [2], [3] Reviewed by: Shigeaki Tagashira < shigeaki AT se DOT hiroshima-u DOT ac DOT jp > Tested by: Shigeaki Tagashira, current Discussed with: current Silence from: obrien
2007-06-12 02:16:02 +00:00
desc = ring->jdesc32;
Add a pure open source nForce Ethernet driver, under BSDL. This driver was ported from OpenBSD by Shigeaki Tagashira <shigeaki@se.hiroshima-u.ac.jp> and posted at http://www.se.hiroshima-u.ac.jp/~shigeaki/software/freebsd-nfe.html It was additionally cleaned up by me. It is still a work-in-progress and thus is purposefully not in GENERIC. And it conflicts with nve(4), so only one should be loaded.
2006-06-26 23:41:07 +00:00
descsize = sizeof (struct nfe_desc32);
}
Bring overhauled nfe(4) into tree. o s/printf/device_printf/g o Nuke OpenBSDism. o Nuke NetBSD/OpenBSD specific DMA sync operations.(we don't have a way to sync a single descriptor within a DMA map.) o Remove recursive mutex. o bus_dma(9) clean up. o 40bit DMA address support. o Add protection for Rx map load failure. o Fix a long standing bug for watchdog timeout. [1] o Add additional protections, missing Tx completion interrupt, losing start Tx command, for watchdog timeout. o Switch to taskqueue(9) API to handle interrupts. o Use our own timer for watchdog instead of if_watchdog/if_timer interface. o Advertise VLAN header length/capability correctly to upper layer. o Remove excessive kernel stack consumption in nfe_encap(). o Handle highly fragmented mbuf chains correctly. o Enable etherenet address reprogramming with ifconfig(8). o Add ALTQ/TSO, MSI/MSIX support. o Increased Rx ring to 256 descriptors from 128. o Align Tx/Rx descriptor ring on sizeof(struct nfe_desc64) boundary. o Remove alignment restrictions on Tx/Rx buffers. o Rewritten jumbo frame support code. o Add support for hardware assistend VLAN tag insertion/stripping. o Add support for Tx/Rx flow control based on patches from Peer Chen. [2] o Add a routine that detects whether ethernet address swap routines is required. [3] o Add a workaround that take MAC/PHY out of power down mode. o Add suspend/resume support. o style(9) and code clean up. Special thanks to Shigeaki Tagashira, the original porter of nfe(4), who submitted lots of patches, performed uncountable number of regression tests and maintained nfe(4) for a long time. Without his enthusiastic help and support I could never have completed this overhauling task. The only weak point of nfe(4) compared to nve(4) is instability of manual half-duplex media selection on certain hardwares(auto sensing media type should work for all cases, though). This was a long standing bug of nfe(4) and I still have no idea why it doesn't work on some hardwares. Obtained from: OpenBSD [1] Submitted by: Peer Chen < pchen at nvidia dot com > [2], [3] Reviewed by: Shigeaki Tagashira < shigeaki AT se DOT hiroshima-u DOT ac DOT jp > Tested by: Shigeaki Tagashira, current Discussed with: current Silence from: obrien
2007-06-12 02:16:02 +00:00
ring->jcur = ring->jnext = 0;
/* Create DMA tag for jumbo Rx ring. */
error = bus_dma_tag_create(sc->nfe_parent_tag,
NFE_RING_ALIGN, 0, /* alignment, boundary */
BUS_SPACE_MAXADDR, /* lowaddr */
BUS_SPACE_MAXADDR, /* highaddr */
NULL, NULL, /* filter, filterarg */
NFE_JUMBO_RX_RING_COUNT * descsize, /* maxsize */
1, /* nsegments */
NFE_JUMBO_RX_RING_COUNT * descsize, /* maxsegsize */
0, /* flags */
NULL, NULL, /* lockfunc, lockarg */
&ring->jrx_desc_tag);
if (error != 0) {
device_printf(sc->nfe_dev,
"could not create jumbo ring DMA tag\n");
goto fail;
}
/* Create DMA tag for jumbo Rx buffers. */
error = bus_dma_tag_create(sc->nfe_parent_tag,
NVIDIA MCP controllers have no Rx buffer alignment restrictions. Remove PAGE_SIZE alignment used in Rx buffer DMA tag creation. The alignment restriction was used in old local jumbo allocator and nfe(4) switched to UMA backed page allocator for jumbo frame. This change should fix jumbo buffer allocation failure. Reported by: Pascal Braun ( pascal.braun <> continum dot net )
2009-05-25 00:56:01 +00:00
1, 0, /* alignment, boundary */
Bring overhauled nfe(4) into tree. o s/printf/device_printf/g o Nuke OpenBSDism. o Nuke NetBSD/OpenBSD specific DMA sync operations.(we don't have a way to sync a single descriptor within a DMA map.) o Remove recursive mutex. o bus_dma(9) clean up. o 40bit DMA address support. o Add protection for Rx map load failure. o Fix a long standing bug for watchdog timeout. [1] o Add additional protections, missing Tx completion interrupt, losing start Tx command, for watchdog timeout. o Switch to taskqueue(9) API to handle interrupts. o Use our own timer for watchdog instead of if_watchdog/if_timer interface. o Advertise VLAN header length/capability correctly to upper layer. o Remove excessive kernel stack consumption in nfe_encap(). o Handle highly fragmented mbuf chains correctly. o Enable etherenet address reprogramming with ifconfig(8). o Add ALTQ/TSO, MSI/MSIX support. o Increased Rx ring to 256 descriptors from 128. o Align Tx/Rx descriptor ring on sizeof(struct nfe_desc64) boundary. o Remove alignment restrictions on Tx/Rx buffers. o Rewritten jumbo frame support code. o Add support for hardware assistend VLAN tag insertion/stripping. o Add support for Tx/Rx flow control based on patches from Peer Chen. [2] o Add a routine that detects whether ethernet address swap routines is required. [3] o Add a workaround that take MAC/PHY out of power down mode. o Add suspend/resume support. o style(9) and code clean up. Special thanks to Shigeaki Tagashira, the original porter of nfe(4), who submitted lots of patches, performed uncountable number of regression tests and maintained nfe(4) for a long time. Without his enthusiastic help and support I could never have completed this overhauling task. The only weak point of nfe(4) compared to nve(4) is instability of manual half-duplex media selection on certain hardwares(auto sensing media type should work for all cases, though). This was a long standing bug of nfe(4) and I still have no idea why it doesn't work on some hardwares. Obtained from: OpenBSD [1] Submitted by: Peer Chen < pchen at nvidia dot com > [2], [3] Reviewed by: Shigeaki Tagashira < shigeaki AT se DOT hiroshima-u DOT ac DOT jp > Tested by: Shigeaki Tagashira, current Discussed with: current Silence from: obrien
2007-06-12 02:16:02 +00:00
BUS_SPACE_MAXADDR, /* lowaddr */
BUS_SPACE_MAXADDR, /* highaddr */
NULL, NULL, /* filter, filterarg */
Nuke local jumbo allocator and switch to use of UMA backed page allocator for jumbo frame. Also remove unneeded jlist lock which is no longer required to protect jumbo buffers. With these changes jumbo frame performance of nfe(4) was slightly increased and users should not encounter jumbo buffer allocation failure anymore.
2008-03-06 01:47:53 +00:00
MJUM9BYTES, /* maxsize */
Bring overhauled nfe(4) into tree. o s/printf/device_printf/g o Nuke OpenBSDism. o Nuke NetBSD/OpenBSD specific DMA sync operations.(we don't have a way to sync a single descriptor within a DMA map.) o Remove recursive mutex. o bus_dma(9) clean up. o 40bit DMA address support. o Add protection for Rx map load failure. o Fix a long standing bug for watchdog timeout. [1] o Add additional protections, missing Tx completion interrupt, losing start Tx command, for watchdog timeout. o Switch to taskqueue(9) API to handle interrupts. o Use our own timer for watchdog instead of if_watchdog/if_timer interface. o Advertise VLAN header length/capability correctly to upper layer. o Remove excessive kernel stack consumption in nfe_encap(). o Handle highly fragmented mbuf chains correctly. o Enable etherenet address reprogramming with ifconfig(8). o Add ALTQ/TSO, MSI/MSIX support. o Increased Rx ring to 256 descriptors from 128. o Align Tx/Rx descriptor ring on sizeof(struct nfe_desc64) boundary. o Remove alignment restrictions on Tx/Rx buffers. o Rewritten jumbo frame support code. o Add support for hardware assistend VLAN tag insertion/stripping. o Add support for Tx/Rx flow control based on patches from Peer Chen. [2] o Add a routine that detects whether ethernet address swap routines is required. [3] o Add a workaround that take MAC/PHY out of power down mode. o Add suspend/resume support. o style(9) and code clean up. Special thanks to Shigeaki Tagashira, the original porter of nfe(4), who submitted lots of patches, performed uncountable number of regression tests and maintained nfe(4) for a long time. Without his enthusiastic help and support I could never have completed this overhauling task. The only weak point of nfe(4) compared to nve(4) is instability of manual half-duplex media selection on certain hardwares(auto sensing media type should work for all cases, though). This was a long standing bug of nfe(4) and I still have no idea why it doesn't work on some hardwares. Obtained from: OpenBSD [1] Submitted by: Peer Chen < pchen at nvidia dot com > [2], [3] Reviewed by: Shigeaki Tagashira < shigeaki AT se DOT hiroshima-u DOT ac DOT jp > Tested by: Shigeaki Tagashira, current Discussed with: current Silence from: obrien
2007-06-12 02:16:02 +00:00
1, /* nsegments */
Nuke local jumbo allocator and switch to use of UMA backed page allocator for jumbo frame. Also remove unneeded jlist lock which is no longer required to protect jumbo buffers. With these changes jumbo frame performance of nfe(4) was slightly increased and users should not encounter jumbo buffer allocation failure anymore.
2008-03-06 01:47:53 +00:00
MJUM9BYTES, /* maxsegsize */
Bring overhauled nfe(4) into tree. o s/printf/device_printf/g o Nuke OpenBSDism. o Nuke NetBSD/OpenBSD specific DMA sync operations.(we don't have a way to sync a single descriptor within a DMA map.) o Remove recursive mutex. o bus_dma(9) clean up. o 40bit DMA address support. o Add protection for Rx map load failure. o Fix a long standing bug for watchdog timeout. [1] o Add additional protections, missing Tx completion interrupt, losing start Tx command, for watchdog timeout. o Switch to taskqueue(9) API to handle interrupts. o Use our own timer for watchdog instead of if_watchdog/if_timer interface. o Advertise VLAN header length/capability correctly to upper layer. o Remove excessive kernel stack consumption in nfe_encap(). o Handle highly fragmented mbuf chains correctly. o Enable etherenet address reprogramming with ifconfig(8). o Add ALTQ/TSO, MSI/MSIX support. o Increased Rx ring to 256 descriptors from 128. o Align Tx/Rx descriptor ring on sizeof(struct nfe_desc64) boundary. o Remove alignment restrictions on Tx/Rx buffers. o Rewritten jumbo frame support code. o Add support for hardware assistend VLAN tag insertion/stripping. o Add support for Tx/Rx flow control based on patches from Peer Chen. [2] o Add a routine that detects whether ethernet address swap routines is required. [3] o Add a workaround that take MAC/PHY out of power down mode. o Add suspend/resume support. o style(9) and code clean up. Special thanks to Shigeaki Tagashira, the original porter of nfe(4), who submitted lots of patches, performed uncountable number of regression tests and maintained nfe(4) for a long time. Without his enthusiastic help and support I could never have completed this overhauling task. The only weak point of nfe(4) compared to nve(4) is instability of manual half-duplex media selection on certain hardwares(auto sensing media type should work for all cases, though). This was a long standing bug of nfe(4) and I still have no idea why it doesn't work on some hardwares. Obtained from: OpenBSD [1] Submitted by: Peer Chen < pchen at nvidia dot com > [2], [3] Reviewed by: Shigeaki Tagashira < shigeaki AT se DOT hiroshima-u DOT ac DOT jp > Tested by: Shigeaki Tagashira, current Discussed with: current Silence from: obrien
2007-06-12 02:16:02 +00:00
0, /* flags */
NULL, NULL, /* lockfunc, lockarg */
&ring->jrx_data_tag);
if (error != 0) {
device_printf(sc->nfe_dev,
"could not create jumbo Rx buffer DMA tag\n");
goto fail;
}
/* Allocate DMA'able memory and load the DMA map for jumbo Rx ring. */
error = bus_dmamem_alloc(ring->jrx_desc_tag, &desc, BUS_DMA_WAITOK |
BUS_DMA_COHERENT | BUS_DMA_ZERO, &ring->jrx_desc_map);
if (error != 0) {
device_printf(sc->nfe_dev,
"could not allocate DMA'able memory for jumbo Rx ring\n");
goto fail;
}
if (sc->nfe_flags & NFE_40BIT_ADDR)
ring->jdesc64 = desc;
else
ring->jdesc32 = desc;
ctx.nfe_busaddr = 0;
error = bus_dmamap_load(ring->jrx_desc_tag, ring->jrx_desc_map, desc,
NFE_JUMBO_RX_RING_COUNT * descsize, nfe_dma_map_segs, &ctx, 0);
if (error != 0) {
device_printf(sc->nfe_dev,
"could not load DMA'able memory for jumbo Rx ring\n");
goto fail;
}
ring->jphysaddr = ctx.nfe_busaddr;
/* Create DMA maps for jumbo Rx buffers. */
error = bus_dmamap_create(ring->jrx_data_tag, 0, &ring->jrx_spare_map);
if (error != 0) {
device_printf(sc->nfe_dev,
"could not create jumbo Rx DMA spare map\n");
goto fail;
}
for (i = 0; i < NFE_JUMBO_RX_RING_COUNT; i++) {
data = &sc->jrxq.jdata[i];
data->rx_data_map = NULL;
data->m = NULL;
error = bus_dmamap_create(ring->jrx_data_tag, 0,
&data->rx_data_map);
if (error != 0) {
device_printf(sc->nfe_dev,
"could not create jumbo Rx DMA map\n");
goto fail;
}
}
Don't fail on device attach if jumbo frame support was unsuccessful. Because nfe(4) hardware doesn't support SG on Rx path, supporting jumbo frame requires very large contiguous kernel memory(i.e. several mega bytes). In case of lack of contiguous kernel memory that allocation request may always fail. However nfe(4) can operate on normal sized MTU frames, so go ahead and just disable jumbo frame support. While I'm here add a new tunable "hw.nfe.jumbo_disable" to disable jumbo frame support. In nfe_poll, make sure to invoke correct Rx handler. Approved by: re (kensmith)
2007-07-24 01:11:00 +00:00
return;
Bring overhauled nfe(4) into tree. o s/printf/device_printf/g o Nuke OpenBSDism. o Nuke NetBSD/OpenBSD specific DMA sync operations.(we don't have a way to sync a single descriptor within a DMA map.) o Remove recursive mutex. o bus_dma(9) clean up. o 40bit DMA address support. o Add protection for Rx map load failure. o Fix a long standing bug for watchdog timeout. [1] o Add additional protections, missing Tx completion interrupt, losing start Tx command, for watchdog timeout. o Switch to taskqueue(9) API to handle interrupts. o Use our own timer for watchdog instead of if_watchdog/if_timer interface. o Advertise VLAN header length/capability correctly to upper layer. o Remove excessive kernel stack consumption in nfe_encap(). o Handle highly fragmented mbuf chains correctly. o Enable etherenet address reprogramming with ifconfig(8). o Add ALTQ/TSO, MSI/MSIX support. o Increased Rx ring to 256 descriptors from 128. o Align Tx/Rx descriptor ring on sizeof(struct nfe_desc64) boundary. o Remove alignment restrictions on Tx/Rx buffers. o Rewritten jumbo frame support code. o Add support for hardware assistend VLAN tag insertion/stripping. o Add support for Tx/Rx flow control based on patches from Peer Chen. [2] o Add a routine that detects whether ethernet address swap routines is required. [3] o Add a workaround that take MAC/PHY out of power down mode. o Add suspend/resume support. o style(9) and code clean up. Special thanks to Shigeaki Tagashira, the original porter of nfe(4), who submitted lots of patches, performed uncountable number of regression tests and maintained nfe(4) for a long time. Without his enthusiastic help and support I could never have completed this overhauling task. The only weak point of nfe(4) compared to nve(4) is instability of manual half-duplex media selection on certain hardwares(auto sensing media type should work for all cases, though). This was a long standing bug of nfe(4) and I still have no idea why it doesn't work on some hardwares. Obtained from: OpenBSD [1] Submitted by: Peer Chen < pchen at nvidia dot com > [2], [3] Reviewed by: Shigeaki Tagashira < shigeaki AT se DOT hiroshima-u DOT ac DOT jp > Tested by: Shigeaki Tagashira, current Discussed with: current Silence from: obrien
2007-06-12 02:16:02 +00:00
fail:
Don't fail on device attach if jumbo frame support was unsuccessful. Because nfe(4) hardware doesn't support SG on Rx path, supporting jumbo frame requires very large contiguous kernel memory(i.e. several mega bytes). In case of lack of contiguous kernel memory that allocation request may always fail. However nfe(4) can operate on normal sized MTU frames, so go ahead and just disable jumbo frame support. While I'm here add a new tunable "hw.nfe.jumbo_disable" to disable jumbo frame support. In nfe_poll, make sure to invoke correct Rx handler. Approved by: re (kensmith)
2007-07-24 01:11:00 +00:00
/*
* Running without jumbo frame support is ok for most cases
* so don't fail on creating dma tag/map for jumbo frame.
*/
Bring overhauled nfe(4) into tree. o s/printf/device_printf/g o Nuke OpenBSDism. o Nuke NetBSD/OpenBSD specific DMA sync operations.(we don't have a way to sync a single descriptor within a DMA map.) o Remove recursive mutex. o bus_dma(9) clean up. o 40bit DMA address support. o Add protection for Rx map load failure. o Fix a long standing bug for watchdog timeout. [1] o Add additional protections, missing Tx completion interrupt, losing start Tx command, for watchdog timeout. o Switch to taskqueue(9) API to handle interrupts. o Use our own timer for watchdog instead of if_watchdog/if_timer interface. o Advertise VLAN header length/capability correctly to upper layer. o Remove excessive kernel stack consumption in nfe_encap(). o Handle highly fragmented mbuf chains correctly. o Enable etherenet address reprogramming with ifconfig(8). o Add ALTQ/TSO, MSI/MSIX support. o Increased Rx ring to 256 descriptors from 128. o Align Tx/Rx descriptor ring on sizeof(struct nfe_desc64) boundary. o Remove alignment restrictions on Tx/Rx buffers. o Rewritten jumbo frame support code. o Add support for hardware assistend VLAN tag insertion/stripping. o Add support for Tx/Rx flow control based on patches from Peer Chen. [2] o Add a routine that detects whether ethernet address swap routines is required. [3] o Add a workaround that take MAC/PHY out of power down mode. o Add suspend/resume support. o style(9) and code clean up. Special thanks to Shigeaki Tagashira, the original porter of nfe(4), who submitted lots of patches, performed uncountable number of regression tests and maintained nfe(4) for a long time. Without his enthusiastic help and support I could never have completed this overhauling task. The only weak point of nfe(4) compared to nve(4) is instability of manual half-duplex media selection on certain hardwares(auto sensing media type should work for all cases, though). This was a long standing bug of nfe(4) and I still have no idea why it doesn't work on some hardwares. Obtained from: OpenBSD [1] Submitted by: Peer Chen < pchen at nvidia dot com > [2], [3] Reviewed by: Shigeaki Tagashira < shigeaki AT se DOT hiroshima-u DOT ac DOT jp > Tested by: Shigeaki Tagashira, current Discussed with: current Silence from: obrien
2007-06-12 02:16:02 +00:00
nfe_free_jrx_ring(sc, ring);
Don't fail on device attach if jumbo frame support was unsuccessful. Because nfe(4) hardware doesn't support SG on Rx path, supporting jumbo frame requires very large contiguous kernel memory(i.e. several mega bytes). In case of lack of contiguous kernel memory that allocation request may always fail. However nfe(4) can operate on normal sized MTU frames, so go ahead and just disable jumbo frame support. While I'm here add a new tunable "hw.nfe.jumbo_disable" to disable jumbo frame support. In nfe_poll, make sure to invoke correct Rx handler. Approved by: re (kensmith)
2007-07-24 01:11:00 +00:00
device_printf(sc->nfe_dev, "disabling jumbo frame support due to "
"resource shortage\n");
sc->nfe_jumbo_disable = 1;
Bring overhauled nfe(4) into tree. o s/printf/device_printf/g o Nuke OpenBSDism. o Nuke NetBSD/OpenBSD specific DMA sync operations.(we don't have a way to sync a single descriptor within a DMA map.) o Remove recursive mutex. o bus_dma(9) clean up. o 40bit DMA address support. o Add protection for Rx map load failure. o Fix a long standing bug for watchdog timeout. [1] o Add additional protections, missing Tx completion interrupt, losing start Tx command, for watchdog timeout. o Switch to taskqueue(9) API to handle interrupts. o Use our own timer for watchdog instead of if_watchdog/if_timer interface. o Advertise VLAN header length/capability correctly to upper layer. o Remove excessive kernel stack consumption in nfe_encap(). o Handle highly fragmented mbuf chains correctly. o Enable etherenet address reprogramming with ifconfig(8). o Add ALTQ/TSO, MSI/MSIX support. o Increased Rx ring to 256 descriptors from 128. o Align Tx/Rx descriptor ring on sizeof(struct nfe_desc64) boundary. o Remove alignment restrictions on Tx/Rx buffers. o Rewritten jumbo frame support code. o Add support for hardware assistend VLAN tag insertion/stripping. o Add support for Tx/Rx flow control based on patches from Peer Chen. [2] o Add a routine that detects whether ethernet address swap routines is required. [3] o Add a workaround that take MAC/PHY out of power down mode. o Add suspend/resume support. o style(9) and code clean up. Special thanks to Shigeaki Tagashira, the original porter of nfe(4), who submitted lots of patches, performed uncountable number of regression tests and maintained nfe(4) for a long time. Without his enthusiastic help and support I could never have completed this overhauling task. The only weak point of nfe(4) compared to nve(4) is instability of manual half-duplex media selection on certain hardwares(auto sensing media type should work for all cases, though). This was a long standing bug of nfe(4) and I still have no idea why it doesn't work on some hardwares. Obtained from: OpenBSD [1] Submitted by: Peer Chen < pchen at nvidia dot com > [2], [3] Reviewed by: Shigeaki Tagashira < shigeaki AT se DOT hiroshima-u DOT ac DOT jp > Tested by: Shigeaki Tagashira, current Discussed with: current Silence from: obrien
2007-06-12 02:16:02 +00:00
}
static int
nfe_init_rx_ring(struct nfe_softc *sc, struct nfe_rx_ring *ring)
{
void *desc;
size_t descsize;
int i;
ring->cur = ring->next = 0;
if (sc->nfe_flags & NFE_40BIT_ADDR) {
desc = ring->desc64;
descsize = sizeof (struct nfe_desc64);
} else {
desc = ring->desc32;
descsize = sizeof (struct nfe_desc32);
}
bzero(desc, descsize * NFE_RX_RING_COUNT);
for (i = 0; i < NFE_RX_RING_COUNT; i++) {
if (nfe_newbuf(sc, i) != 0)
return (ENOBUFS);
}
bus_dmamap_sync(ring->rx_desc_tag, ring->rx_desc_map,
BUS_DMASYNC_PREREAD | BUS_DMASYNC_PREWRITE);
return (0);
}
static int
nfe_init_jrx_ring(struct nfe_softc *sc, struct nfe_jrx_ring *ring)
{
void *desc;
size_t descsize;
int i;
ring->jcur = ring->jnext = 0;
if (sc->nfe_flags & NFE_40BIT_ADDR) {
desc = ring->jdesc64;
descsize = sizeof (struct nfe_desc64);
} else {
desc = ring->jdesc32;
descsize = sizeof (struct nfe_desc32);
}
Nuke local jumbo allocator and switch to use of UMA backed page allocator for jumbo frame. Also remove unneeded jlist lock which is no longer required to protect jumbo buffers. With these changes jumbo frame performance of nfe(4) was slightly increased and users should not encounter jumbo buffer allocation failure anymore.
2008-03-06 01:47:53 +00:00
bzero(desc, descsize * NFE_JUMBO_RX_RING_COUNT);
Bring overhauled nfe(4) into tree. o s/printf/device_printf/g o Nuke OpenBSDism. o Nuke NetBSD/OpenBSD specific DMA sync operations.(we don't have a way to sync a single descriptor within a DMA map.) o Remove recursive mutex. o bus_dma(9) clean up. o 40bit DMA address support. o Add protection for Rx map load failure. o Fix a long standing bug for watchdog timeout. [1] o Add additional protections, missing Tx completion interrupt, losing start Tx command, for watchdog timeout. o Switch to taskqueue(9) API to handle interrupts. o Use our own timer for watchdog instead of if_watchdog/if_timer interface. o Advertise VLAN header length/capability correctly to upper layer. o Remove excessive kernel stack consumption in nfe_encap(). o Handle highly fragmented mbuf chains correctly. o Enable etherenet address reprogramming with ifconfig(8). o Add ALTQ/TSO, MSI/MSIX support. o Increased Rx ring to 256 descriptors from 128. o Align Tx/Rx descriptor ring on sizeof(struct nfe_desc64) boundary. o Remove alignment restrictions on Tx/Rx buffers. o Rewritten jumbo frame support code. o Add support for hardware assistend VLAN tag insertion/stripping. o Add support for Tx/Rx flow control based on patches from Peer Chen. [2] o Add a routine that detects whether ethernet address swap routines is required. [3] o Add a workaround that take MAC/PHY out of power down mode. o Add suspend/resume support. o style(9) and code clean up. Special thanks to Shigeaki Tagashira, the original porter of nfe(4), who submitted lots of patches, performed uncountable number of regression tests and maintained nfe(4) for a long time. Without his enthusiastic help and support I could never have completed this overhauling task. The only weak point of nfe(4) compared to nve(4) is instability of manual half-duplex media selection on certain hardwares(auto sensing media type should work for all cases, though). This was a long standing bug of nfe(4) and I still have no idea why it doesn't work on some hardwares. Obtained from: OpenBSD [1] Submitted by: Peer Chen < pchen at nvidia dot com > [2], [3] Reviewed by: Shigeaki Tagashira < shigeaki AT se DOT hiroshima-u DOT ac DOT jp > Tested by: Shigeaki Tagashira, current Discussed with: current Silence from: obrien
2007-06-12 02:16:02 +00:00
for (i = 0; i < NFE_JUMBO_RX_RING_COUNT; i++) {
if (nfe_jnewbuf(sc, i) != 0)
return (ENOBUFS);
}
bus_dmamap_sync(ring->jrx_desc_tag, ring->jrx_desc_map,
BUS_DMASYNC_PREREAD | BUS_DMASYNC_PREWRITE);
return (0);
}
static void
nfe_free_rx_ring(struct nfe_softc *sc, struct nfe_rx_ring *ring)
{
struct nfe_rx_data *data;
void *desc;
int i, descsize;
if (sc->nfe_flags & NFE_40BIT_ADDR) {
desc = ring->desc64;
descsize = sizeof (struct nfe_desc64);
} else {
desc = ring->desc32;
descsize = sizeof (struct nfe_desc32);
}
for (i = 0; i < NFE_RX_RING_COUNT; i++) {
data = &ring->data[i];
if (data->rx_data_map != NULL) {
bus_dmamap_destroy(ring->rx_data_tag,
data->rx_data_map);
data->rx_data_map = NULL;
}
if (data->m != NULL) {
m_freem(data->m);
data->m = NULL;
}
}
if (ring->rx_data_tag != NULL) {
if (ring->rx_spare_map != NULL) {
bus_dmamap_destroy(ring->rx_data_tag,
ring->rx_spare_map);
ring->rx_spare_map = NULL;
}
bus_dma_tag_destroy(ring->rx_data_tag);
ring->rx_data_tag = NULL;
}
if (desc != NULL) {
Add a pure open source nForce Ethernet driver, under BSDL. This driver was ported from OpenBSD by Shigeaki Tagashira <shigeaki@se.hiroshima-u.ac.jp> and posted at http://www.se.hiroshima-u.ac.jp/~shigeaki/software/freebsd-nfe.html It was additionally cleaned up by me. It is still a work-in-progress and thus is purposefully not in GENERIC. And it conflicts with nve(4), so only one should be loaded.
2006-06-26 23:41:07 +00:00
bus_dmamap_unload(ring->rx_desc_tag, ring->rx_desc_map);
bus_dmamem_free(ring->rx_desc_tag, desc, ring->rx_desc_map);
Bring overhauled nfe(4) into tree. o s/printf/device_printf/g o Nuke OpenBSDism. o Nuke NetBSD/OpenBSD specific DMA sync operations.(we don't have a way to sync a single descriptor within a DMA map.) o Remove recursive mutex. o bus_dma(9) clean up. o 40bit DMA address support. o Add protection for Rx map load failure. o Fix a long standing bug for watchdog timeout. [1] o Add additional protections, missing Tx completion interrupt, losing start Tx command, for watchdog timeout. o Switch to taskqueue(9) API to handle interrupts. o Use our own timer for watchdog instead of if_watchdog/if_timer interface. o Advertise VLAN header length/capability correctly to upper layer. o Remove excessive kernel stack consumption in nfe_encap(). o Handle highly fragmented mbuf chains correctly. o Enable etherenet address reprogramming with ifconfig(8). o Add ALTQ/TSO, MSI/MSIX support. o Increased Rx ring to 256 descriptors from 128. o Align Tx/Rx descriptor ring on sizeof(struct nfe_desc64) boundary. o Remove alignment restrictions on Tx/Rx buffers. o Rewritten jumbo frame support code. o Add support for hardware assistend VLAN tag insertion/stripping. o Add support for Tx/Rx flow control based on patches from Peer Chen. [2] o Add a routine that detects whether ethernet address swap routines is required. [3] o Add a workaround that take MAC/PHY out of power down mode. o Add suspend/resume support. o style(9) and code clean up. Special thanks to Shigeaki Tagashira, the original porter of nfe(4), who submitted lots of patches, performed uncountable number of regression tests and maintained nfe(4) for a long time. Without his enthusiastic help and support I could never have completed this overhauling task. The only weak point of nfe(4) compared to nve(4) is instability of manual half-duplex media selection on certain hardwares(auto sensing media type should work for all cases, though). This was a long standing bug of nfe(4) and I still have no idea why it doesn't work on some hardwares. Obtained from: OpenBSD [1] Submitted by: Peer Chen < pchen at nvidia dot com > [2], [3] Reviewed by: Shigeaki Tagashira < shigeaki AT se DOT hiroshima-u DOT ac DOT jp > Tested by: Shigeaki Tagashira, current Discussed with: current Silence from: obrien
2007-06-12 02:16:02 +00:00
ring->desc64 = NULL;
ring->desc32 = NULL;
ring->rx_desc_map = NULL;
}
if (ring->rx_desc_tag != NULL) {
Add a pure open source nForce Ethernet driver, under BSDL. This driver was ported from OpenBSD by Shigeaki Tagashira <shigeaki@se.hiroshima-u.ac.jp> and posted at http://www.se.hiroshima-u.ac.jp/~shigeaki/software/freebsd-nfe.html It was additionally cleaned up by me. It is still a work-in-progress and thus is purposefully not in GENERIC. And it conflicts with nve(4), so only one should be loaded.
2006-06-26 23:41:07 +00:00
bus_dma_tag_destroy(ring->rx_desc_tag);
Bring overhauled nfe(4) into tree. o s/printf/device_printf/g o Nuke OpenBSDism. o Nuke NetBSD/OpenBSD specific DMA sync operations.(we don't have a way to sync a single descriptor within a DMA map.) o Remove recursive mutex. o bus_dma(9) clean up. o 40bit DMA address support. o Add protection for Rx map load failure. o Fix a long standing bug for watchdog timeout. [1] o Add additional protections, missing Tx completion interrupt, losing start Tx command, for watchdog timeout. o Switch to taskqueue(9) API to handle interrupts. o Use our own timer for watchdog instead of if_watchdog/if_timer interface. o Advertise VLAN header length/capability correctly to upper layer. o Remove excessive kernel stack consumption in nfe_encap(). o Handle highly fragmented mbuf chains correctly. o Enable etherenet address reprogramming with ifconfig(8). o Add ALTQ/TSO, MSI/MSIX support. o Increased Rx ring to 256 descriptors from 128. o Align Tx/Rx descriptor ring on sizeof(struct nfe_desc64) boundary. o Remove alignment restrictions on Tx/Rx buffers. o Rewritten jumbo frame support code. o Add support for hardware assistend VLAN tag insertion/stripping. o Add support for Tx/Rx flow control based on patches from Peer Chen. [2] o Add a routine that detects whether ethernet address swap routines is required. [3] o Add a workaround that take MAC/PHY out of power down mode. o Add suspend/resume support. o style(9) and code clean up. Special thanks to Shigeaki Tagashira, the original porter of nfe(4), who submitted lots of patches, performed uncountable number of regression tests and maintained nfe(4) for a long time. Without his enthusiastic help and support I could never have completed this overhauling task. The only weak point of nfe(4) compared to nve(4) is instability of manual half-duplex media selection on certain hardwares(auto sensing media type should work for all cases, though). This was a long standing bug of nfe(4) and I still have no idea why it doesn't work on some hardwares. Obtained from: OpenBSD [1] Submitted by: Peer Chen < pchen at nvidia dot com > [2], [3] Reviewed by: Shigeaki Tagashira < shigeaki AT se DOT hiroshima-u DOT ac DOT jp > Tested by: Shigeaki Tagashira, current Discussed with: current Silence from: obrien
2007-06-12 02:16:02 +00:00
ring->rx_desc_tag = NULL;
Add a pure open source nForce Ethernet driver, under BSDL. This driver was ported from OpenBSD by Shigeaki Tagashira <shigeaki@se.hiroshima-u.ac.jp> and posted at http://www.se.hiroshima-u.ac.jp/~shigeaki/software/freebsd-nfe.html It was additionally cleaned up by me. It is still a work-in-progress and thus is purposefully not in GENERIC. And it conflicts with nve(4), so only one should be loaded.
2006-06-26 23:41:07 +00:00
}
Bring overhauled nfe(4) into tree. o s/printf/device_printf/g o Nuke OpenBSDism. o Nuke NetBSD/OpenBSD specific DMA sync operations.(we don't have a way to sync a single descriptor within a DMA map.) o Remove recursive mutex. o bus_dma(9) clean up. o 40bit DMA address support. o Add protection for Rx map load failure. o Fix a long standing bug for watchdog timeout. [1] o Add additional protections, missing Tx completion interrupt, losing start Tx command, for watchdog timeout. o Switch to taskqueue(9) API to handle interrupts. o Use our own timer for watchdog instead of if_watchdog/if_timer interface. o Advertise VLAN header length/capability correctly to upper layer. o Remove excessive kernel stack consumption in nfe_encap(). o Handle highly fragmented mbuf chains correctly. o Enable etherenet address reprogramming with ifconfig(8). o Add ALTQ/TSO, MSI/MSIX support. o Increased Rx ring to 256 descriptors from 128. o Align Tx/Rx descriptor ring on sizeof(struct nfe_desc64) boundary. o Remove alignment restrictions on Tx/Rx buffers. o Rewritten jumbo frame support code. o Add support for hardware assistend VLAN tag insertion/stripping. o Add support for Tx/Rx flow control based on patches from Peer Chen. [2] o Add a routine that detects whether ethernet address swap routines is required. [3] o Add a workaround that take MAC/PHY out of power down mode. o Add suspend/resume support. o style(9) and code clean up. Special thanks to Shigeaki Tagashira, the original porter of nfe(4), who submitted lots of patches, performed uncountable number of regression tests and maintained nfe(4) for a long time. Without his enthusiastic help and support I could never have completed this overhauling task. The only weak point of nfe(4) compared to nve(4) is instability of manual half-duplex media selection on certain hardwares(auto sensing media type should work for all cases, though). This was a long standing bug of nfe(4) and I still have no idea why it doesn't work on some hardwares. Obtained from: OpenBSD [1] Submitted by: Peer Chen < pchen at nvidia dot com > [2], [3] Reviewed by: Shigeaki Tagashira < shigeaki AT se DOT hiroshima-u DOT ac DOT jp > Tested by: Shigeaki Tagashira, current Discussed with: current Silence from: obrien
2007-06-12 02:16:02 +00:00
}
Add a pure open source nForce Ethernet driver, under BSDL. This driver was ported from OpenBSD by Shigeaki Tagashira <shigeaki@se.hiroshima-u.ac.jp> and posted at http://www.se.hiroshima-u.ac.jp/~shigeaki/software/freebsd-nfe.html It was additionally cleaned up by me. It is still a work-in-progress and thus is purposefully not in GENERIC. And it conflicts with nve(4), so only one should be loaded.
2006-06-26 23:41:07 +00:00
Fix the jumbo frame support. Submitted by: Peer Chen at nVIDIA (reworked by me)
2006-11-27 00:16:47 +00:00
Bring overhauled nfe(4) into tree. o s/printf/device_printf/g o Nuke OpenBSDism. o Nuke NetBSD/OpenBSD specific DMA sync operations.(we don't have a way to sync a single descriptor within a DMA map.) o Remove recursive mutex. o bus_dma(9) clean up. o 40bit DMA address support. o Add protection for Rx map load failure. o Fix a long standing bug for watchdog timeout. [1] o Add additional protections, missing Tx completion interrupt, losing start Tx command, for watchdog timeout. o Switch to taskqueue(9) API to handle interrupts. o Use our own timer for watchdog instead of if_watchdog/if_timer interface. o Advertise VLAN header length/capability correctly to upper layer. o Remove excessive kernel stack consumption in nfe_encap(). o Handle highly fragmented mbuf chains correctly. o Enable etherenet address reprogramming with ifconfig(8). o Add ALTQ/TSO, MSI/MSIX support. o Increased Rx ring to 256 descriptors from 128. o Align Tx/Rx descriptor ring on sizeof(struct nfe_desc64) boundary. o Remove alignment restrictions on Tx/Rx buffers. o Rewritten jumbo frame support code. o Add support for hardware assistend VLAN tag insertion/stripping. o Add support for Tx/Rx flow control based on patches from Peer Chen. [2] o Add a routine that detects whether ethernet address swap routines is required. [3] o Add a workaround that take MAC/PHY out of power down mode. o Add suspend/resume support. o style(9) and code clean up. Special thanks to Shigeaki Tagashira, the original porter of nfe(4), who submitted lots of patches, performed uncountable number of regression tests and maintained nfe(4) for a long time. Without his enthusiastic help and support I could never have completed this overhauling task. The only weak point of nfe(4) compared to nve(4) is instability of manual half-duplex media selection on certain hardwares(auto sensing media type should work for all cases, though). This was a long standing bug of nfe(4) and I still have no idea why it doesn't work on some hardwares. Obtained from: OpenBSD [1] Submitted by: Peer Chen < pchen at nvidia dot com > [2], [3] Reviewed by: Shigeaki Tagashira < shigeaki AT se DOT hiroshima-u DOT ac DOT jp > Tested by: Shigeaki Tagashira, current Discussed with: current Silence from: obrien
2007-06-12 02:16:02 +00:00
static void
nfe_free_jrx_ring(struct nfe_softc *sc, struct nfe_jrx_ring *ring)
{
struct nfe_rx_data *data;
void *desc;
int i, descsize;
if ((sc->nfe_flags & NFE_JUMBO_SUP) == 0)
return;
if (sc->nfe_flags & NFE_40BIT_ADDR) {
desc = ring->jdesc64;
descsize = sizeof (struct nfe_desc64);
} else {
desc = ring->jdesc32;
descsize = sizeof (struct nfe_desc32);
}
for (i = 0; i < NFE_JUMBO_RX_RING_COUNT; i++) {
data = &ring->jdata[i];
Reformat to KNF after revision 1.12.
2006-11-27 00:23:59 +00:00
if (data->rx_data_map != NULL) {
Bring overhauled nfe(4) into tree. o s/printf/device_printf/g o Nuke OpenBSDism. o Nuke NetBSD/OpenBSD specific DMA sync operations.(we don't have a way to sync a single descriptor within a DMA map.) o Remove recursive mutex. o bus_dma(9) clean up. o 40bit DMA address support. o Add protection for Rx map load failure. o Fix a long standing bug for watchdog timeout. [1] o Add additional protections, missing Tx completion interrupt, losing start Tx command, for watchdog timeout. o Switch to taskqueue(9) API to handle interrupts. o Use our own timer for watchdog instead of if_watchdog/if_timer interface. o Advertise VLAN header length/capability correctly to upper layer. o Remove excessive kernel stack consumption in nfe_encap(). o Handle highly fragmented mbuf chains correctly. o Enable etherenet address reprogramming with ifconfig(8). o Add ALTQ/TSO, MSI/MSIX support. o Increased Rx ring to 256 descriptors from 128. o Align Tx/Rx descriptor ring on sizeof(struct nfe_desc64) boundary. o Remove alignment restrictions on Tx/Rx buffers. o Rewritten jumbo frame support code. o Add support for hardware assistend VLAN tag insertion/stripping. o Add support for Tx/Rx flow control based on patches from Peer Chen. [2] o Add a routine that detects whether ethernet address swap routines is required. [3] o Add a workaround that take MAC/PHY out of power down mode. o Add suspend/resume support. o style(9) and code clean up. Special thanks to Shigeaki Tagashira, the original porter of nfe(4), who submitted lots of patches, performed uncountable number of regression tests and maintained nfe(4) for a long time. Without his enthusiastic help and support I could never have completed this overhauling task. The only weak point of nfe(4) compared to nve(4) is instability of manual half-duplex media selection on certain hardwares(auto sensing media type should work for all cases, though). This was a long standing bug of nfe(4) and I still have no idea why it doesn't work on some hardwares. Obtained from: OpenBSD [1] Submitted by: Peer Chen < pchen at nvidia dot com > [2], [3] Reviewed by: Shigeaki Tagashira < shigeaki AT se DOT hiroshima-u DOT ac DOT jp > Tested by: Shigeaki Tagashira, current Discussed with: current Silence from: obrien
2007-06-12 02:16:02 +00:00
bus_dmamap_destroy(ring->jrx_data_tag,
Reformat to KNF after revision 1.12.
2006-11-27 00:23:59 +00:00
data->rx_data_map);
Bring overhauled nfe(4) into tree. o s/printf/device_printf/g o Nuke OpenBSDism. o Nuke NetBSD/OpenBSD specific DMA sync operations.(we don't have a way to sync a single descriptor within a DMA map.) o Remove recursive mutex. o bus_dma(9) clean up. o 40bit DMA address support. o Add protection for Rx map load failure. o Fix a long standing bug for watchdog timeout. [1] o Add additional protections, missing Tx completion interrupt, losing start Tx command, for watchdog timeout. o Switch to taskqueue(9) API to handle interrupts. o Use our own timer for watchdog instead of if_watchdog/if_timer interface. o Advertise VLAN header length/capability correctly to upper layer. o Remove excessive kernel stack consumption in nfe_encap(). o Handle highly fragmented mbuf chains correctly. o Enable etherenet address reprogramming with ifconfig(8). o Add ALTQ/TSO, MSI/MSIX support. o Increased Rx ring to 256 descriptors from 128. o Align Tx/Rx descriptor ring on sizeof(struct nfe_desc64) boundary. o Remove alignment restrictions on Tx/Rx buffers. o Rewritten jumbo frame support code. o Add support for hardware assistend VLAN tag insertion/stripping. o Add support for Tx/Rx flow control based on patches from Peer Chen. [2] o Add a routine that detects whether ethernet address swap routines is required. [3] o Add a workaround that take MAC/PHY out of power down mode. o Add suspend/resume support. o style(9) and code clean up. Special thanks to Shigeaki Tagashira, the original porter of nfe(4), who submitted lots of patches, performed uncountable number of regression tests and maintained nfe(4) for a long time. Without his enthusiastic help and support I could never have completed this overhauling task. The only weak point of nfe(4) compared to nve(4) is instability of manual half-duplex media selection on certain hardwares(auto sensing media type should work for all cases, though). This was a long standing bug of nfe(4) and I still have no idea why it doesn't work on some hardwares. Obtained from: OpenBSD [1] Submitted by: Peer Chen < pchen at nvidia dot com > [2], [3] Reviewed by: Shigeaki Tagashira < shigeaki AT se DOT hiroshima-u DOT ac DOT jp > Tested by: Shigeaki Tagashira, current Discussed with: current Silence from: obrien
2007-06-12 02:16:02 +00:00
data->rx_data_map = NULL;
style(9) cleanup.
2006-10-19 10:01:26 +00:00
}
Bring overhauled nfe(4) into tree. o s/printf/device_printf/g o Nuke OpenBSDism. o Nuke NetBSD/OpenBSD specific DMA sync operations.(we don't have a way to sync a single descriptor within a DMA map.) o Remove recursive mutex. o bus_dma(9) clean up. o 40bit DMA address support. o Add protection for Rx map load failure. o Fix a long standing bug for watchdog timeout. [1] o Add additional protections, missing Tx completion interrupt, losing start Tx command, for watchdog timeout. o Switch to taskqueue(9) API to handle interrupts. o Use our own timer for watchdog instead of if_watchdog/if_timer interface. o Advertise VLAN header length/capability correctly to upper layer. o Remove excessive kernel stack consumption in nfe_encap(). o Handle highly fragmented mbuf chains correctly. o Enable etherenet address reprogramming with ifconfig(8). o Add ALTQ/TSO, MSI/MSIX support. o Increased Rx ring to 256 descriptors from 128. o Align Tx/Rx descriptor ring on sizeof(struct nfe_desc64) boundary. o Remove alignment restrictions on Tx/Rx buffers. o Rewritten jumbo frame support code. o Add support for hardware assistend VLAN tag insertion/stripping. o Add support for Tx/Rx flow control based on patches from Peer Chen. [2] o Add a routine that detects whether ethernet address swap routines is required. [3] o Add a workaround that take MAC/PHY out of power down mode. o Add suspend/resume support. o style(9) and code clean up. Special thanks to Shigeaki Tagashira, the original porter of nfe(4), who submitted lots of patches, performed uncountable number of regression tests and maintained nfe(4) for a long time. Without his enthusiastic help and support I could never have completed this overhauling task. The only weak point of nfe(4) compared to nve(4) is instability of manual half-duplex media selection on certain hardwares(auto sensing media type should work for all cases, though). This was a long standing bug of nfe(4) and I still have no idea why it doesn't work on some hardwares. Obtained from: OpenBSD [1] Submitted by: Peer Chen < pchen at nvidia dot com > [2], [3] Reviewed by: Shigeaki Tagashira < shigeaki AT se DOT hiroshima-u DOT ac DOT jp > Tested by: Shigeaki Tagashira, current Discussed with: current Silence from: obrien
2007-06-12 02:16:02 +00:00
if (data->m != NULL) {
Reformat to KNF after revision 1.12.
2006-11-27 00:23:59 +00:00
m_freem(data->m);
Bring overhauled nfe(4) into tree. o s/printf/device_printf/g o Nuke OpenBSDism. o Nuke NetBSD/OpenBSD specific DMA sync operations.(we don't have a way to sync a single descriptor within a DMA map.) o Remove recursive mutex. o bus_dma(9) clean up. o 40bit DMA address support. o Add protection for Rx map load failure. o Fix a long standing bug for watchdog timeout. [1] o Add additional protections, missing Tx completion interrupt, losing start Tx command, for watchdog timeout. o Switch to taskqueue(9) API to handle interrupts. o Use our own timer for watchdog instead of if_watchdog/if_timer interface. o Advertise VLAN header length/capability correctly to upper layer. o Remove excessive kernel stack consumption in nfe_encap(). o Handle highly fragmented mbuf chains correctly. o Enable etherenet address reprogramming with ifconfig(8). o Add ALTQ/TSO, MSI/MSIX support. o Increased Rx ring to 256 descriptors from 128. o Align Tx/Rx descriptor ring on sizeof(struct nfe_desc64) boundary. o Remove alignment restrictions on Tx/Rx buffers. o Rewritten jumbo frame support code. o Add support for hardware assistend VLAN tag insertion/stripping. o Add support for Tx/Rx flow control based on patches from Peer Chen. [2] o Add a routine that detects whether ethernet address swap routines is required. [3] o Add a workaround that take MAC/PHY out of power down mode. o Add suspend/resume support. o style(9) and code clean up. Special thanks to Shigeaki Tagashira, the original porter of nfe(4), who submitted lots of patches, performed uncountable number of regression tests and maintained nfe(4) for a long time. Without his enthusiastic help and support I could never have completed this overhauling task. The only weak point of nfe(4) compared to nve(4) is instability of manual half-duplex media selection on certain hardwares(auto sensing media type should work for all cases, though). This was a long standing bug of nfe(4) and I still have no idea why it doesn't work on some hardwares. Obtained from: OpenBSD [1] Submitted by: Peer Chen < pchen at nvidia dot com > [2], [3] Reviewed by: Shigeaki Tagashira < shigeaki AT se DOT hiroshima-u DOT ac DOT jp > Tested by: Shigeaki Tagashira, current Discussed with: current Silence from: obrien
2007-06-12 02:16:02 +00:00
data->m = NULL;
}
}
if (ring->jrx_data_tag != NULL) {
if (ring->jrx_spare_map != NULL) {
bus_dmamap_destroy(ring->jrx_data_tag,
ring->jrx_spare_map);
ring->jrx_spare_map = NULL;
}
bus_dma_tag_destroy(ring->jrx_data_tag);
ring->jrx_data_tag = NULL;
}
if (desc != NULL) {
bus_dmamap_unload(ring->jrx_desc_tag, ring->jrx_desc_map);
bus_dmamem_free(ring->jrx_desc_tag, desc, ring->jrx_desc_map);
ring->jdesc64 = NULL;
ring->jdesc32 = NULL;
ring->jrx_desc_map = NULL;
}
Nuke local jumbo allocator and switch to use of UMA backed page allocator for jumbo frame. Also remove unneeded jlist lock which is no longer required to protect jumbo buffers. With these changes jumbo frame performance of nfe(4) was slightly increased and users should not encounter jumbo buffer allocation failure anymore.
2008-03-06 01:47:53 +00:00
Bring overhauled nfe(4) into tree. o s/printf/device_printf/g o Nuke OpenBSDism. o Nuke NetBSD/OpenBSD specific DMA sync operations.(we don't have a way to sync a single descriptor within a DMA map.) o Remove recursive mutex. o bus_dma(9) clean up. o 40bit DMA address support. o Add protection for Rx map load failure. o Fix a long standing bug for watchdog timeout. [1] o Add additional protections, missing Tx completion interrupt, losing start Tx command, for watchdog timeout. o Switch to taskqueue(9) API to handle interrupts. o Use our own timer for watchdog instead of if_watchdog/if_timer interface. o Advertise VLAN header length/capability correctly to upper layer. o Remove excessive kernel stack consumption in nfe_encap(). o Handle highly fragmented mbuf chains correctly. o Enable etherenet address reprogramming with ifconfig(8). o Add ALTQ/TSO, MSI/MSIX support. o Increased Rx ring to 256 descriptors from 128. o Align Tx/Rx descriptor ring on sizeof(struct nfe_desc64) boundary. o Remove alignment restrictions on Tx/Rx buffers. o Rewritten jumbo frame support code. o Add support for hardware assistend VLAN tag insertion/stripping. o Add support for Tx/Rx flow control based on patches from Peer Chen. [2] o Add a routine that detects whether ethernet address swap routines is required. [3] o Add a workaround that take MAC/PHY out of power down mode. o Add suspend/resume support. o style(9) and code clean up. Special thanks to Shigeaki Tagashira, the original porter of nfe(4), who submitted lots of patches, performed uncountable number of regression tests and maintained nfe(4) for a long time. Without his enthusiastic help and support I could never have completed this overhauling task. The only weak point of nfe(4) compared to nve(4) is instability of manual half-duplex media selection on certain hardwares(auto sensing media type should work for all cases, though). This was a long standing bug of nfe(4) and I still have no idea why it doesn't work on some hardwares. Obtained from: OpenBSD [1] Submitted by: Peer Chen < pchen at nvidia dot com > [2], [3] Reviewed by: Shigeaki Tagashira < shigeaki AT se DOT hiroshima-u DOT ac DOT jp > Tested by: Shigeaki Tagashira, current Discussed with: current Silence from: obrien
2007-06-12 02:16:02 +00:00
if (ring->jrx_desc_tag != NULL) {
bus_dma_tag_destroy(ring->jrx_desc_tag);
ring->jrx_desc_tag = NULL;
Reformat to KNF after revision 1.12.
2006-11-27 00:23:59 +00:00
}
Add a pure open source nForce Ethernet driver, under BSDL. This driver was ported from OpenBSD by Shigeaki Tagashira <shigeaki@se.hiroshima-u.ac.jp> and posted at http://www.se.hiroshima-u.ac.jp/~shigeaki/software/freebsd-nfe.html It was additionally cleaned up by me. It is still a work-in-progress and thus is purposefully not in GENERIC. And it conflicts with nve(4), so only one should be loaded.
2006-06-26 23:41:07 +00:00
}
style(9) cleanup.
2006-10-19 10:01:26 +00:00
Add a pure open source nForce Ethernet driver, under BSDL. This driver was ported from OpenBSD by Shigeaki Tagashira <shigeaki@se.hiroshima-u.ac.jp> and posted at http://www.se.hiroshima-u.ac.jp/~shigeaki/software/freebsd-nfe.html It was additionally cleaned up by me. It is still a work-in-progress and thus is purposefully not in GENERIC. And it conflicts with nve(4), so only one should be loaded.
2006-06-26 23:41:07 +00:00
static int
nfe_alloc_tx_ring(struct nfe_softc *sc, struct nfe_tx_ring *ring)
{
Bring overhauled nfe(4) into tree. o s/printf/device_printf/g o Nuke OpenBSDism. o Nuke NetBSD/OpenBSD specific DMA sync operations.(we don't have a way to sync a single descriptor within a DMA map.) o Remove recursive mutex. o bus_dma(9) clean up. o 40bit DMA address support. o Add protection for Rx map load failure. o Fix a long standing bug for watchdog timeout. [1] o Add additional protections, missing Tx completion interrupt, losing start Tx command, for watchdog timeout. o Switch to taskqueue(9) API to handle interrupts. o Use our own timer for watchdog instead of if_watchdog/if_timer interface. o Advertise VLAN header length/capability correctly to upper layer. o Remove excessive kernel stack consumption in nfe_encap(). o Handle highly fragmented mbuf chains correctly. o Enable etherenet address reprogramming with ifconfig(8). o Add ALTQ/TSO, MSI/MSIX support. o Increased Rx ring to 256 descriptors from 128. o Align Tx/Rx descriptor ring on sizeof(struct nfe_desc64) boundary. o Remove alignment restrictions on Tx/Rx buffers. o Rewritten jumbo frame support code. o Add support for hardware assistend VLAN tag insertion/stripping. o Add support for Tx/Rx flow control based on patches from Peer Chen. [2] o Add a routine that detects whether ethernet address swap routines is required. [3] o Add a workaround that take MAC/PHY out of power down mode. o Add suspend/resume support. o style(9) and code clean up. Special thanks to Shigeaki Tagashira, the original porter of nfe(4), who submitted lots of patches, performed uncountable number of regression tests and maintained nfe(4) for a long time. Without his enthusiastic help and support I could never have completed this overhauling task. The only weak point of nfe(4) compared to nve(4) is instability of manual half-duplex media selection on certain hardwares(auto sensing media type should work for all cases, though). This was a long standing bug of nfe(4) and I still have no idea why it doesn't work on some hardwares. Obtained from: OpenBSD [1] Submitted by: Peer Chen < pchen at nvidia dot com > [2], [3] Reviewed by: Shigeaki Tagashira < shigeaki AT se DOT hiroshima-u DOT ac DOT jp > Tested by: Shigeaki Tagashira, current Discussed with: current Silence from: obrien
2007-06-12 02:16:02 +00:00
struct nfe_dmamap_arg ctx;
Add a pure open source nForce Ethernet driver, under BSDL. This driver was ported from OpenBSD by Shigeaki Tagashira <shigeaki@se.hiroshima-u.ac.jp> and posted at http://www.se.hiroshima-u.ac.jp/~shigeaki/software/freebsd-nfe.html It was additionally cleaned up by me. It is still a work-in-progress and thus is purposefully not in GENERIC. And it conflicts with nve(4), so only one should be loaded.
2006-06-26 23:41:07 +00:00
int i, error;
Bring overhauled nfe(4) into tree. o s/printf/device_printf/g o Nuke OpenBSDism. o Nuke NetBSD/OpenBSD specific DMA sync operations.(we don't have a way to sync a single descriptor within a DMA map.) o Remove recursive mutex. o bus_dma(9) clean up. o 40bit DMA address support. o Add protection for Rx map load failure. o Fix a long standing bug for watchdog timeout. [1] o Add additional protections, missing Tx completion interrupt, losing start Tx command, for watchdog timeout. o Switch to taskqueue(9) API to handle interrupts. o Use our own timer for watchdog instead of if_watchdog/if_timer interface. o Advertise VLAN header length/capability correctly to upper layer. o Remove excessive kernel stack consumption in nfe_encap(). o Handle highly fragmented mbuf chains correctly. o Enable etherenet address reprogramming with ifconfig(8). o Add ALTQ/TSO, MSI/MSIX support. o Increased Rx ring to 256 descriptors from 128. o Align Tx/Rx descriptor ring on sizeof(struct nfe_desc64) boundary. o Remove alignment restrictions on Tx/Rx buffers. o Rewritten jumbo frame support code. o Add support for hardware assistend VLAN tag insertion/stripping. o Add support for Tx/Rx flow control based on patches from Peer Chen. [2] o Add a routine that detects whether ethernet address swap routines is required. [3] o Add a workaround that take MAC/PHY out of power down mode. o Add suspend/resume support. o style(9) and code clean up. Special thanks to Shigeaki Tagashira, the original porter of nfe(4), who submitted lots of patches, performed uncountable number of regression tests and maintained nfe(4) for a long time. Without his enthusiastic help and support I could never have completed this overhauling task. The only weak point of nfe(4) compared to nve(4) is instability of manual half-duplex media selection on certain hardwares(auto sensing media type should work for all cases, though). This was a long standing bug of nfe(4) and I still have no idea why it doesn't work on some hardwares. Obtained from: OpenBSD [1] Submitted by: Peer Chen < pchen at nvidia dot com > [2], [3] Reviewed by: Shigeaki Tagashira < shigeaki AT se DOT hiroshima-u DOT ac DOT jp > Tested by: Shigeaki Tagashira, current Discussed with: current Silence from: obrien
2007-06-12 02:16:02 +00:00
void *desc;
Add a pure open source nForce Ethernet driver, under BSDL. This driver was ported from OpenBSD by Shigeaki Tagashira <shigeaki@se.hiroshima-u.ac.jp> and posted at http://www.se.hiroshima-u.ac.jp/~shigeaki/software/freebsd-nfe.html It was additionally cleaned up by me. It is still a work-in-progress and thus is purposefully not in GENERIC. And it conflicts with nve(4), so only one should be loaded.
2006-06-26 23:41:07 +00:00
int descsize;
if (sc->nfe_flags & NFE_40BIT_ADDR) {
Bring overhauled nfe(4) into tree. o s/printf/device_printf/g o Nuke OpenBSDism. o Nuke NetBSD/OpenBSD specific DMA sync operations.(we don't have a way to sync a single descriptor within a DMA map.) o Remove recursive mutex. o bus_dma(9) clean up. o 40bit DMA address support. o Add protection for Rx map load failure. o Fix a long standing bug for watchdog timeout. [1] o Add additional protections, missing Tx completion interrupt, losing start Tx command, for watchdog timeout. o Switch to taskqueue(9) API to handle interrupts. o Use our own timer for watchdog instead of if_watchdog/if_timer interface. o Advertise VLAN header length/capability correctly to upper layer. o Remove excessive kernel stack consumption in nfe_encap(). o Handle highly fragmented mbuf chains correctly. o Enable etherenet address reprogramming with ifconfig(8). o Add ALTQ/TSO, MSI/MSIX support. o Increased Rx ring to 256 descriptors from 128. o Align Tx/Rx descriptor ring on sizeof(struct nfe_desc64) boundary. o Remove alignment restrictions on Tx/Rx buffers. o Rewritten jumbo frame support code. o Add support for hardware assistend VLAN tag insertion/stripping. o Add support for Tx/Rx flow control based on patches from Peer Chen. [2] o Add a routine that detects whether ethernet address swap routines is required. [3] o Add a workaround that take MAC/PHY out of power down mode. o Add suspend/resume support. o style(9) and code clean up. Special thanks to Shigeaki Tagashira, the original porter of nfe(4), who submitted lots of patches, performed uncountable number of regression tests and maintained nfe(4) for a long time. Without his enthusiastic help and support I could never have completed this overhauling task. The only weak point of nfe(4) compared to nve(4) is instability of manual half-duplex media selection on certain hardwares(auto sensing media type should work for all cases, though). This was a long standing bug of nfe(4) and I still have no idea why it doesn't work on some hardwares. Obtained from: OpenBSD [1] Submitted by: Peer Chen < pchen at nvidia dot com > [2], [3] Reviewed by: Shigeaki Tagashira < shigeaki AT se DOT hiroshima-u DOT ac DOT jp > Tested by: Shigeaki Tagashira, current Discussed with: current Silence from: obrien
2007-06-12 02:16:02 +00:00
desc = ring->desc64;
Add a pure open source nForce Ethernet driver, under BSDL. This driver was ported from OpenBSD by Shigeaki Tagashira <shigeaki@se.hiroshima-u.ac.jp> and posted at http://www.se.hiroshima-u.ac.jp/~shigeaki/software/freebsd-nfe.html It was additionally cleaned up by me. It is still a work-in-progress and thus is purposefully not in GENERIC. And it conflicts with nve(4), so only one should be loaded.
2006-06-26 23:41:07 +00:00
descsize = sizeof (struct nfe_desc64);
} else {
Bring overhauled nfe(4) into tree. o s/printf/device_printf/g o Nuke OpenBSDism. o Nuke NetBSD/OpenBSD specific DMA sync operations.(we don't have a way to sync a single descriptor within a DMA map.) o Remove recursive mutex. o bus_dma(9) clean up. o 40bit DMA address support. o Add protection for Rx map load failure. o Fix a long standing bug for watchdog timeout. [1] o Add additional protections, missing Tx completion interrupt, losing start Tx command, for watchdog timeout. o Switch to taskqueue(9) API to handle interrupts. o Use our own timer for watchdog instead of if_watchdog/if_timer interface. o Advertise VLAN header length/capability correctly to upper layer. o Remove excessive kernel stack consumption in nfe_encap(). o Handle highly fragmented mbuf chains correctly. o Enable etherenet address reprogramming with ifconfig(8). o Add ALTQ/TSO, MSI/MSIX support. o Increased Rx ring to 256 descriptors from 128. o Align Tx/Rx descriptor ring on sizeof(struct nfe_desc64) boundary. o Remove alignment restrictions on Tx/Rx buffers. o Rewritten jumbo frame support code. o Add support for hardware assistend VLAN tag insertion/stripping. o Add support for Tx/Rx flow control based on patches from Peer Chen. [2] o Add a routine that detects whether ethernet address swap routines is required. [3] o Add a workaround that take MAC/PHY out of power down mode. o Add suspend/resume support. o style(9) and code clean up. Special thanks to Shigeaki Tagashira, the original porter of nfe(4), who submitted lots of patches, performed uncountable number of regression tests and maintained nfe(4) for a long time. Without his enthusiastic help and support I could never have completed this overhauling task. The only weak point of nfe(4) compared to nve(4) is instability of manual half-duplex media selection on certain hardwares(auto sensing media type should work for all cases, though). This was a long standing bug of nfe(4) and I still have no idea why it doesn't work on some hardwares. Obtained from: OpenBSD [1] Submitted by: Peer Chen < pchen at nvidia dot com > [2], [3] Reviewed by: Shigeaki Tagashira < shigeaki AT se DOT hiroshima-u DOT ac DOT jp > Tested by: Shigeaki Tagashira, current Discussed with: current Silence from: obrien
2007-06-12 02:16:02 +00:00
desc = ring->desc32;
Add a pure open source nForce Ethernet driver, under BSDL. This driver was ported from OpenBSD by Shigeaki Tagashira <shigeaki@se.hiroshima-u.ac.jp> and posted at http://www.se.hiroshima-u.ac.jp/~shigeaki/software/freebsd-nfe.html It was additionally cleaned up by me. It is still a work-in-progress and thus is purposefully not in GENERIC. And it conflicts with nve(4), so only one should be loaded.
2006-06-26 23:41:07 +00:00
descsize = sizeof (struct nfe_desc32);
}
ring->queued = 0;
ring->cur = ring->next = 0;
style(9) cleanup.
2006-10-19 10:01:26 +00:00
error = bus_dma_tag_create(sc->nfe_parent_tag,
Bring overhauled nfe(4) into tree. o s/printf/device_printf/g o Nuke OpenBSDism. o Nuke NetBSD/OpenBSD specific DMA sync operations.(we don't have a way to sync a single descriptor within a DMA map.) o Remove recursive mutex. o bus_dma(9) clean up. o 40bit DMA address support. o Add protection for Rx map load failure. o Fix a long standing bug for watchdog timeout. [1] o Add additional protections, missing Tx completion interrupt, losing start Tx command, for watchdog timeout. o Switch to taskqueue(9) API to handle interrupts. o Use our own timer for watchdog instead of if_watchdog/if_timer interface. o Advertise VLAN header length/capability correctly to upper layer. o Remove excessive kernel stack consumption in nfe_encap(). o Handle highly fragmented mbuf chains correctly. o Enable etherenet address reprogramming with ifconfig(8). o Add ALTQ/TSO, MSI/MSIX support. o Increased Rx ring to 256 descriptors from 128. o Align Tx/Rx descriptor ring on sizeof(struct nfe_desc64) boundary. o Remove alignment restrictions on Tx/Rx buffers. o Rewritten jumbo frame support code. o Add support for hardware assistend VLAN tag insertion/stripping. o Add support for Tx/Rx flow control based on patches from Peer Chen. [2] o Add a routine that detects whether ethernet address swap routines is required. [3] o Add a workaround that take MAC/PHY out of power down mode. o Add suspend/resume support. o style(9) and code clean up. Special thanks to Shigeaki Tagashira, the original porter of nfe(4), who submitted lots of patches, performed uncountable number of regression tests and maintained nfe(4) for a long time. Without his enthusiastic help and support I could never have completed this overhauling task. The only weak point of nfe(4) compared to nve(4) is instability of manual half-duplex media selection on certain hardwares(auto sensing media type should work for all cases, though). This was a long standing bug of nfe(4) and I still have no idea why it doesn't work on some hardwares. Obtained from: OpenBSD [1] Submitted by: Peer Chen < pchen at nvidia dot com > [2], [3] Reviewed by: Shigeaki Tagashira < shigeaki AT se DOT hiroshima-u DOT ac DOT jp > Tested by: Shigeaki Tagashira, current Discussed with: current Silence from: obrien
2007-06-12 02:16:02 +00:00
NFE_RING_ALIGN, 0, /* alignment, boundary */
BUS_SPACE_MAXADDR, /* lowaddr */
BUS_SPACE_MAXADDR, /* highaddr */
NULL, NULL, /* filter, filterarg */
NFE_TX_RING_COUNT * descsize, 1, /* maxsize, nsegments */
NFE_TX_RING_COUNT * descsize, /* maxsegsize */
0, /* flags */
NULL, NULL, /* lockfunc, lockarg */
&ring->tx_desc_tag);
Add a pure open source nForce Ethernet driver, under BSDL. This driver was ported from OpenBSD by Shigeaki Tagashira <shigeaki@se.hiroshima-u.ac.jp> and posted at http://www.se.hiroshima-u.ac.jp/~shigeaki/software/freebsd-nfe.html It was additionally cleaned up by me. It is still a work-in-progress and thus is purposefully not in GENERIC. And it conflicts with nve(4), so only one should be loaded.
2006-06-26 23:41:07 +00:00
if (error != 0) {
Bring overhauled nfe(4) into tree. o s/printf/device_printf/g o Nuke OpenBSDism. o Nuke NetBSD/OpenBSD specific DMA sync operations.(we don't have a way to sync a single descriptor within a DMA map.) o Remove recursive mutex. o bus_dma(9) clean up. o 40bit DMA address support. o Add protection for Rx map load failure. o Fix a long standing bug for watchdog timeout. [1] o Add additional protections, missing Tx completion interrupt, losing start Tx command, for watchdog timeout. o Switch to taskqueue(9) API to handle interrupts. o Use our own timer for watchdog instead of if_watchdog/if_timer interface. o Advertise VLAN header length/capability correctly to upper layer. o Remove excessive kernel stack consumption in nfe_encap(). o Handle highly fragmented mbuf chains correctly. o Enable etherenet address reprogramming with ifconfig(8). o Add ALTQ/TSO, MSI/MSIX support. o Increased Rx ring to 256 descriptors from 128. o Align Tx/Rx descriptor ring on sizeof(struct nfe_desc64) boundary. o Remove alignment restrictions on Tx/Rx buffers. o Rewritten jumbo frame support code. o Add support for hardware assistend VLAN tag insertion/stripping. o Add support for Tx/Rx flow control based on patches from Peer Chen. [2] o Add a routine that detects whether ethernet address swap routines is required. [3] o Add a workaround that take MAC/PHY out of power down mode. o Add suspend/resume support. o style(9) and code clean up. Special thanks to Shigeaki Tagashira, the original porter of nfe(4), who submitted lots of patches, performed uncountable number of regression tests and maintained nfe(4) for a long time. Without his enthusiastic help and support I could never have completed this overhauling task. The only weak point of nfe(4) compared to nve(4) is instability of manual half-duplex media selection on certain hardwares(auto sensing media type should work for all cases, though). This was a long standing bug of nfe(4) and I still have no idea why it doesn't work on some hardwares. Obtained from: OpenBSD [1] Submitted by: Peer Chen < pchen at nvidia dot com > [2], [3] Reviewed by: Shigeaki Tagashira < shigeaki AT se DOT hiroshima-u DOT ac DOT jp > Tested by: Shigeaki Tagashira, current Discussed with: current Silence from: obrien
2007-06-12 02:16:02 +00:00
device_printf(sc->nfe_dev, "could not create desc DMA tag\n");
Add a pure open source nForce Ethernet driver, under BSDL. This driver was ported from OpenBSD by Shigeaki Tagashira <shigeaki@se.hiroshima-u.ac.jp> and posted at http://www.se.hiroshima-u.ac.jp/~shigeaki/software/freebsd-nfe.html It was additionally cleaned up by me. It is still a work-in-progress and thus is purposefully not in GENERIC. And it conflicts with nve(4), so only one should be loaded.
2006-06-26 23:41:07 +00:00
goto fail;
}
Bring overhauled nfe(4) into tree. o s/printf/device_printf/g o Nuke OpenBSDism. o Nuke NetBSD/OpenBSD specific DMA sync operations.(we don't have a way to sync a single descriptor within a DMA map.) o Remove recursive mutex. o bus_dma(9) clean up. o 40bit DMA address support. o Add protection for Rx map load failure. o Fix a long standing bug for watchdog timeout. [1] o Add additional protections, missing Tx completion interrupt, losing start Tx command, for watchdog timeout. o Switch to taskqueue(9) API to handle interrupts. o Use our own timer for watchdog instead of if_watchdog/if_timer interface. o Advertise VLAN header length/capability correctly to upper layer. o Remove excessive kernel stack consumption in nfe_encap(). o Handle highly fragmented mbuf chains correctly. o Enable etherenet address reprogramming with ifconfig(8). o Add ALTQ/TSO, MSI/MSIX support. o Increased Rx ring to 256 descriptors from 128. o Align Tx/Rx descriptor ring on sizeof(struct nfe_desc64) boundary. o Remove alignment restrictions on Tx/Rx buffers. o Rewritten jumbo frame support code. o Add support for hardware assistend VLAN tag insertion/stripping. o Add support for Tx/Rx flow control based on patches from Peer Chen. [2] o Add a routine that detects whether ethernet address swap routines is required. [3] o Add a workaround that take MAC/PHY out of power down mode. o Add suspend/resume support. o style(9) and code clean up. Special thanks to Shigeaki Tagashira, the original porter of nfe(4), who submitted lots of patches, performed uncountable number of regression tests and maintained nfe(4) for a long time. Without his enthusiastic help and support I could never have completed this overhauling task. The only weak point of nfe(4) compared to nve(4) is instability of manual half-duplex media selection on certain hardwares(auto sensing media type should work for all cases, though). This was a long standing bug of nfe(4) and I still have no idea why it doesn't work on some hardwares. Obtained from: OpenBSD [1] Submitted by: Peer Chen < pchen at nvidia dot com > [2], [3] Reviewed by: Shigeaki Tagashira < shigeaki AT se DOT hiroshima-u DOT ac DOT jp > Tested by: Shigeaki Tagashira, current Discussed with: current Silence from: obrien
2007-06-12 02:16:02 +00:00
error = bus_dmamem_alloc(ring->tx_desc_tag, &desc, BUS_DMA_WAITOK |
BUS_DMA_COHERENT | BUS_DMA_ZERO, &ring->tx_desc_map);
Add a pure open source nForce Ethernet driver, under BSDL. This driver was ported from OpenBSD by Shigeaki Tagashira <shigeaki@se.hiroshima-u.ac.jp> and posted at http://www.se.hiroshima-u.ac.jp/~shigeaki/software/freebsd-nfe.html It was additionally cleaned up by me. It is still a work-in-progress and thus is purposefully not in GENERIC. And it conflicts with nve(4), so only one should be loaded.
2006-06-26 23:41:07 +00:00
if (error != 0) {
Bring overhauled nfe(4) into tree. o s/printf/device_printf/g o Nuke OpenBSDism. o Nuke NetBSD/OpenBSD specific DMA sync operations.(we don't have a way to sync a single descriptor within a DMA map.) o Remove recursive mutex. o bus_dma(9) clean up. o 40bit DMA address support. o Add protection for Rx map load failure. o Fix a long standing bug for watchdog timeout. [1] o Add additional protections, missing Tx completion interrupt, losing start Tx command, for watchdog timeout. o Switch to taskqueue(9) API to handle interrupts. o Use our own timer for watchdog instead of if_watchdog/if_timer interface. o Advertise VLAN header length/capability correctly to upper layer. o Remove excessive kernel stack consumption in nfe_encap(). o Handle highly fragmented mbuf chains correctly. o Enable etherenet address reprogramming with ifconfig(8). o Add ALTQ/TSO, MSI/MSIX support. o Increased Rx ring to 256 descriptors from 128. o Align Tx/Rx descriptor ring on sizeof(struct nfe_desc64) boundary. o Remove alignment restrictions on Tx/Rx buffers. o Rewritten jumbo frame support code. o Add support for hardware assistend VLAN tag insertion/stripping. o Add support for Tx/Rx flow control based on patches from Peer Chen. [2] o Add a routine that detects whether ethernet address swap routines is required. [3] o Add a workaround that take MAC/PHY out of power down mode. o Add suspend/resume support. o style(9) and code clean up. Special thanks to Shigeaki Tagashira, the original porter of nfe(4), who submitted lots of patches, performed uncountable number of regression tests and maintained nfe(4) for a long time. Without his enthusiastic help and support I could never have completed this overhauling task. The only weak point of nfe(4) compared to nve(4) is instability of manual half-duplex media selection on certain hardwares(auto sensing media type should work for all cases, though). This was a long standing bug of nfe(4) and I still have no idea why it doesn't work on some hardwares. Obtained from: OpenBSD [1] Submitted by: Peer Chen < pchen at nvidia dot com > [2], [3] Reviewed by: Shigeaki Tagashira < shigeaki AT se DOT hiroshima-u DOT ac DOT jp > Tested by: Shigeaki Tagashira, current Discussed with: current Silence from: obrien
2007-06-12 02:16:02 +00:00
device_printf(sc->nfe_dev, "could not create desc DMA map\n");
Add a pure open source nForce Ethernet driver, under BSDL. This driver was ported from OpenBSD by Shigeaki Tagashira <shigeaki@se.hiroshima-u.ac.jp> and posted at http://www.se.hiroshima-u.ac.jp/~shigeaki/software/freebsd-nfe.html It was additionally cleaned up by me. It is still a work-in-progress and thus is purposefully not in GENERIC. And it conflicts with nve(4), so only one should be loaded.
2006-06-26 23:41:07 +00:00
goto fail;
}
Bring overhauled nfe(4) into tree. o s/printf/device_printf/g o Nuke OpenBSDism. o Nuke NetBSD/OpenBSD specific DMA sync operations.(we don't have a way to sync a single descriptor within a DMA map.) o Remove recursive mutex. o bus_dma(9) clean up. o 40bit DMA address support. o Add protection for Rx map load failure. o Fix a long standing bug for watchdog timeout. [1] o Add additional protections, missing Tx completion interrupt, losing start Tx command, for watchdog timeout. o Switch to taskqueue(9) API to handle interrupts. o Use our own timer for watchdog instead of if_watchdog/if_timer interface. o Advertise VLAN header length/capability correctly to upper layer. o Remove excessive kernel stack consumption in nfe_encap(). o Handle highly fragmented mbuf chains correctly. o Enable etherenet address reprogramming with ifconfig(8). o Add ALTQ/TSO, MSI/MSIX support. o Increased Rx ring to 256 descriptors from 128. o Align Tx/Rx descriptor ring on sizeof(struct nfe_desc64) boundary. o Remove alignment restrictions on Tx/Rx buffers. o Rewritten jumbo frame support code. o Add support for hardware assistend VLAN tag insertion/stripping. o Add support for Tx/Rx flow control based on patches from Peer Chen. [2] o Add a routine that detects whether ethernet address swap routines is required. [3] o Add a workaround that take MAC/PHY out of power down mode. o Add suspend/resume support. o style(9) and code clean up. Special thanks to Shigeaki Tagashira, the original porter of nfe(4), who submitted lots of patches, performed uncountable number of regression tests and maintained nfe(4) for a long time. Without his enthusiastic help and support I could never have completed this overhauling task. The only weak point of nfe(4) compared to nve(4) is instability of manual half-duplex media selection on certain hardwares(auto sensing media type should work for all cases, though). This was a long standing bug of nfe(4) and I still have no idea why it doesn't work on some hardwares. Obtained from: OpenBSD [1] Submitted by: Peer Chen < pchen at nvidia dot com > [2], [3] Reviewed by: Shigeaki Tagashira < shigeaki AT se DOT hiroshima-u DOT ac DOT jp > Tested by: Shigeaki Tagashira, current Discussed with: current Silence from: obrien
2007-06-12 02:16:02 +00:00
if (sc->nfe_flags & NFE_40BIT_ADDR)
ring->desc64 = desc;
else
ring->desc32 = desc;
Add a pure open source nForce Ethernet driver, under BSDL. This driver was ported from OpenBSD by Shigeaki Tagashira <shigeaki@se.hiroshima-u.ac.jp> and posted at http://www.se.hiroshima-u.ac.jp/~shigeaki/software/freebsd-nfe.html It was additionally cleaned up by me. It is still a work-in-progress and thus is purposefully not in GENERIC. And it conflicts with nve(4), so only one should be loaded.
2006-06-26 23:41:07 +00:00
Bring overhauled nfe(4) into tree. o s/printf/device_printf/g o Nuke OpenBSDism. o Nuke NetBSD/OpenBSD specific DMA sync operations.(we don't have a way to sync a single descriptor within a DMA map.) o Remove recursive mutex. o bus_dma(9) clean up. o 40bit DMA address support. o Add protection for Rx map load failure. o Fix a long standing bug for watchdog timeout. [1] o Add additional protections, missing Tx completion interrupt, losing start Tx command, for watchdog timeout. o Switch to taskqueue(9) API to handle interrupts. o Use our own timer for watchdog instead of if_watchdog/if_timer interface. o Advertise VLAN header length/capability correctly to upper layer. o Remove excessive kernel stack consumption in nfe_encap(). o Handle highly fragmented mbuf chains correctly. o Enable etherenet address reprogramming with ifconfig(8). o Add ALTQ/TSO, MSI/MSIX support. o Increased Rx ring to 256 descriptors from 128. o Align Tx/Rx descriptor ring on sizeof(struct nfe_desc64) boundary. o Remove alignment restrictions on Tx/Rx buffers. o Rewritten jumbo frame support code. o Add support for hardware assistend VLAN tag insertion/stripping. o Add support for Tx/Rx flow control based on patches from Peer Chen. [2] o Add a routine that detects whether ethernet address swap routines is required. [3] o Add a workaround that take MAC/PHY out of power down mode. o Add suspend/resume support. o style(9) and code clean up. Special thanks to Shigeaki Tagashira, the original porter of nfe(4), who submitted lots of patches, performed uncountable number of regression tests and maintained nfe(4) for a long time. Without his enthusiastic help and support I could never have completed this overhauling task. The only weak point of nfe(4) compared to nve(4) is instability of manual half-duplex media selection on certain hardwares(auto sensing media type should work for all cases, though). This was a long standing bug of nfe(4) and I still have no idea why it doesn't work on some hardwares. Obtained from: OpenBSD [1] Submitted by: Peer Chen < pchen at nvidia dot com > [2], [3] Reviewed by: Shigeaki Tagashira < shigeaki AT se DOT hiroshima-u DOT ac DOT jp > Tested by: Shigeaki Tagashira, current Discussed with: current Silence from: obrien
2007-06-12 02:16:02 +00:00
ctx.nfe_busaddr = 0;
error = bus_dmamap_load(ring->tx_desc_tag, ring->tx_desc_map, desc,
NFE_TX_RING_COUNT * descsize, nfe_dma_map_segs, &ctx, 0);
Add a pure open source nForce Ethernet driver, under BSDL. This driver was ported from OpenBSD by Shigeaki Tagashira <shigeaki@se.hiroshima-u.ac.jp> and posted at http://www.se.hiroshima-u.ac.jp/~shigeaki/software/freebsd-nfe.html It was additionally cleaned up by me. It is still a work-in-progress and thus is purposefully not in GENERIC. And it conflicts with nve(4), so only one should be loaded.
2006-06-26 23:41:07 +00:00
if (error != 0) {
Bring overhauled nfe(4) into tree. o s/printf/device_printf/g o Nuke OpenBSDism. o Nuke NetBSD/OpenBSD specific DMA sync operations.(we don't have a way to sync a single descriptor within a DMA map.) o Remove recursive mutex. o bus_dma(9) clean up. o 40bit DMA address support. o Add protection for Rx map load failure. o Fix a long standing bug for watchdog timeout. [1] o Add additional protections, missing Tx completion interrupt, losing start Tx command, for watchdog timeout. o Switch to taskqueue(9) API to handle interrupts. o Use our own timer for watchdog instead of if_watchdog/if_timer interface. o Advertise VLAN header length/capability correctly to upper layer. o Remove excessive kernel stack consumption in nfe_encap(). o Handle highly fragmented mbuf chains correctly. o Enable etherenet address reprogramming with ifconfig(8). o Add ALTQ/TSO, MSI/MSIX support. o Increased Rx ring to 256 descriptors from 128. o Align Tx/Rx descriptor ring on sizeof(struct nfe_desc64) boundary. o Remove alignment restrictions on Tx/Rx buffers. o Rewritten jumbo frame support code. o Add support for hardware assistend VLAN tag insertion/stripping. o Add support for Tx/Rx flow control based on patches from Peer Chen. [2] o Add a routine that detects whether ethernet address swap routines is required. [3] o Add a workaround that take MAC/PHY out of power down mode. o Add suspend/resume support. o style(9) and code clean up. Special thanks to Shigeaki Tagashira, the original porter of nfe(4), who submitted lots of patches, performed uncountable number of regression tests and maintained nfe(4) for a long time. Without his enthusiastic help and support I could never have completed this overhauling task. The only weak point of nfe(4) compared to nve(4) is instability of manual half-duplex media selection on certain hardwares(auto sensing media type should work for all cases, though). This was a long standing bug of nfe(4) and I still have no idea why it doesn't work on some hardwares. Obtained from: OpenBSD [1] Submitted by: Peer Chen < pchen at nvidia dot com > [2], [3] Reviewed by: Shigeaki Tagashira < shigeaki AT se DOT hiroshima-u DOT ac DOT jp > Tested by: Shigeaki Tagashira, current Discussed with: current Silence from: obrien
2007-06-12 02:16:02 +00:00
device_printf(sc->nfe_dev, "could not load desc DMA map\n");
Add a pure open source nForce Ethernet driver, under BSDL. This driver was ported from OpenBSD by Shigeaki Tagashira <shigeaki@se.hiroshima-u.ac.jp> and posted at http://www.se.hiroshima-u.ac.jp/~shigeaki/software/freebsd-nfe.html It was additionally cleaned up by me. It is still a work-in-progress and thus is purposefully not in GENERIC. And it conflicts with nve(4), so only one should be loaded.
2006-06-26 23:41:07 +00:00
goto fail;
}
Bring overhauled nfe(4) into tree. o s/printf/device_printf/g o Nuke OpenBSDism. o Nuke NetBSD/OpenBSD specific DMA sync operations.(we don't have a way to sync a single descriptor within a DMA map.) o Remove recursive mutex. o bus_dma(9) clean up. o 40bit DMA address support. o Add protection for Rx map load failure. o Fix a long standing bug for watchdog timeout. [1] o Add additional protections, missing Tx completion interrupt, losing start Tx command, for watchdog timeout. o Switch to taskqueue(9) API to handle interrupts. o Use our own timer for watchdog instead of if_watchdog/if_timer interface. o Advertise VLAN header length/capability correctly to upper layer. o Remove excessive kernel stack consumption in nfe_encap(). o Handle highly fragmented mbuf chains correctly. o Enable etherenet address reprogramming with ifconfig(8). o Add ALTQ/TSO, MSI/MSIX support. o Increased Rx ring to 256 descriptors from 128. o Align Tx/Rx descriptor ring on sizeof(struct nfe_desc64) boundary. o Remove alignment restrictions on Tx/Rx buffers. o Rewritten jumbo frame support code. o Add support for hardware assistend VLAN tag insertion/stripping. o Add support for Tx/Rx flow control based on patches from Peer Chen. [2] o Add a routine that detects whether ethernet address swap routines is required. [3] o Add a workaround that take MAC/PHY out of power down mode. o Add suspend/resume support. o style(9) and code clean up. Special thanks to Shigeaki Tagashira, the original porter of nfe(4), who submitted lots of patches, performed uncountable number of regression tests and maintained nfe(4) for a long time. Without his enthusiastic help and support I could never have completed this overhauling task. The only weak point of nfe(4) compared to nve(4) is instability of manual half-duplex media selection on certain hardwares(auto sensing media type should work for all cases, though). This was a long standing bug of nfe(4) and I still have no idea why it doesn't work on some hardwares. Obtained from: OpenBSD [1] Submitted by: Peer Chen < pchen at nvidia dot com > [2], [3] Reviewed by: Shigeaki Tagashira < shigeaki AT se DOT hiroshima-u DOT ac DOT jp > Tested by: Shigeaki Tagashira, current Discussed with: current Silence from: obrien
2007-06-12 02:16:02 +00:00
ring->physaddr = ctx.nfe_busaddr;
Add a pure open source nForce Ethernet driver, under BSDL. This driver was ported from OpenBSD by Shigeaki Tagashira <shigeaki@se.hiroshima-u.ac.jp> and posted at http://www.se.hiroshima-u.ac.jp/~shigeaki/software/freebsd-nfe.html It was additionally cleaned up by me. It is still a work-in-progress and thus is purposefully not in GENERIC. And it conflicts with nve(4), so only one should be loaded.
2006-06-26 23:41:07 +00:00
style(9) cleanup.
2006-10-19 10:01:26 +00:00
error = bus_dma_tag_create(sc->nfe_parent_tag,
Bring overhauled nfe(4) into tree. o s/printf/device_printf/g o Nuke OpenBSDism. o Nuke NetBSD/OpenBSD specific DMA sync operations.(we don't have a way to sync a single descriptor within a DMA map.) o Remove recursive mutex. o bus_dma(9) clean up. o 40bit DMA address support. o Add protection for Rx map load failure. o Fix a long standing bug for watchdog timeout. [1] o Add additional protections, missing Tx completion interrupt, losing start Tx command, for watchdog timeout. o Switch to taskqueue(9) API to handle interrupts. o Use our own timer for watchdog instead of if_watchdog/if_timer interface. o Advertise VLAN header length/capability correctly to upper layer. o Remove excessive kernel stack consumption in nfe_encap(). o Handle highly fragmented mbuf chains correctly. o Enable etherenet address reprogramming with ifconfig(8). o Add ALTQ/TSO, MSI/MSIX support. o Increased Rx ring to 256 descriptors from 128. o Align Tx/Rx descriptor ring on sizeof(struct nfe_desc64) boundary. o Remove alignment restrictions on Tx/Rx buffers. o Rewritten jumbo frame support code. o Add support for hardware assistend VLAN tag insertion/stripping. o Add support for Tx/Rx flow control based on patches from Peer Chen. [2] o Add a routine that detects whether ethernet address swap routines is required. [3] o Add a workaround that take MAC/PHY out of power down mode. o Add suspend/resume support. o style(9) and code clean up. Special thanks to Shigeaki Tagashira, the original porter of nfe(4), who submitted lots of patches, performed uncountable number of regression tests and maintained nfe(4) for a long time. Without his enthusiastic help and support I could never have completed this overhauling task. The only weak point of nfe(4) compared to nve(4) is instability of manual half-duplex media selection on certain hardwares(auto sensing media type should work for all cases, though). This was a long standing bug of nfe(4) and I still have no idea why it doesn't work on some hardwares. Obtained from: OpenBSD [1] Submitted by: Peer Chen < pchen at nvidia dot com > [2], [3] Reviewed by: Shigeaki Tagashira < shigeaki AT se DOT hiroshima-u DOT ac DOT jp > Tested by: Shigeaki Tagashira, current Discussed with: current Silence from: obrien
2007-06-12 02:16:02 +00:00
1, 0,
BUS_SPACE_MAXADDR,
BUS_SPACE_MAXADDR,
NULL, NULL,
Increase a maximum segment size of DMA to 4096. Previously it used MCLBYTES for the segment size but it used too many Tx descriptors in TSO case. While I'm here adjust maximum size of the sum of all segment lengths in a given DMA mapping to 65535, the maximum size, in bytes, of a IP packet.
2007-06-12 02:35:01 +00:00
NFE_TSO_MAXSIZE,
Bring overhauled nfe(4) into tree. o s/printf/device_printf/g o Nuke OpenBSDism. o Nuke NetBSD/OpenBSD specific DMA sync operations.(we don't have a way to sync a single descriptor within a DMA map.) o Remove recursive mutex. o bus_dma(9) clean up. o 40bit DMA address support. o Add protection for Rx map load failure. o Fix a long standing bug for watchdog timeout. [1] o Add additional protections, missing Tx completion interrupt, losing start Tx command, for watchdog timeout. o Switch to taskqueue(9) API to handle interrupts. o Use our own timer for watchdog instead of if_watchdog/if_timer interface. o Advertise VLAN header length/capability correctly to upper layer. o Remove excessive kernel stack consumption in nfe_encap(). o Handle highly fragmented mbuf chains correctly. o Enable etherenet address reprogramming with ifconfig(8). o Add ALTQ/TSO, MSI/MSIX support. o Increased Rx ring to 256 descriptors from 128. o Align Tx/Rx descriptor ring on sizeof(struct nfe_desc64) boundary. o Remove alignment restrictions on Tx/Rx buffers. o Rewritten jumbo frame support code. o Add support for hardware assistend VLAN tag insertion/stripping. o Add support for Tx/Rx flow control based on patches from Peer Chen. [2] o Add a routine that detects whether ethernet address swap routines is required. [3] o Add a workaround that take MAC/PHY out of power down mode. o Add suspend/resume support. o style(9) and code clean up. Special thanks to Shigeaki Tagashira, the original porter of nfe(4), who submitted lots of patches, performed uncountable number of regression tests and maintained nfe(4) for a long time. Without his enthusiastic help and support I could never have completed this overhauling task. The only weak point of nfe(4) compared to nve(4) is instability of manual half-duplex media selection on certain hardwares(auto sensing media type should work for all cases, though). This was a long standing bug of nfe(4) and I still have no idea why it doesn't work on some hardwares. Obtained from: OpenBSD [1] Submitted by: Peer Chen < pchen at nvidia dot com > [2], [3] Reviewed by: Shigeaki Tagashira < shigeaki AT se DOT hiroshima-u DOT ac DOT jp > Tested by: Shigeaki Tagashira, current Discussed with: current Silence from: obrien
2007-06-12 02:16:02 +00:00
NFE_MAX_SCATTER,
Increase a maximum segment size of DMA to 4096. Previously it used MCLBYTES for the segment size but it used too many Tx descriptors in TSO case. While I'm here adjust maximum size of the sum of all segment lengths in a given DMA mapping to 65535, the maximum size, in bytes, of a IP packet.
2007-06-12 02:35:01 +00:00
NFE_TSO_MAXSGSIZE,
Bring overhauled nfe(4) into tree. o s/printf/device_printf/g o Nuke OpenBSDism. o Nuke NetBSD/OpenBSD specific DMA sync operations.(we don't have a way to sync a single descriptor within a DMA map.) o Remove recursive mutex. o bus_dma(9) clean up. o 40bit DMA address support. o Add protection for Rx map load failure. o Fix a long standing bug for watchdog timeout. [1] o Add additional protections, missing Tx completion interrupt, losing start Tx command, for watchdog timeout. o Switch to taskqueue(9) API to handle interrupts. o Use our own timer for watchdog instead of if_watchdog/if_timer interface. o Advertise VLAN header length/capability correctly to upper layer. o Remove excessive kernel stack consumption in nfe_encap(). o Handle highly fragmented mbuf chains correctly. o Enable etherenet address reprogramming with ifconfig(8). o Add ALTQ/TSO, MSI/MSIX support. o Increased Rx ring to 256 descriptors from 128. o Align Tx/Rx descriptor ring on sizeof(struct nfe_desc64) boundary. o Remove alignment restrictions on Tx/Rx buffers. o Rewritten jumbo frame support code. o Add support for hardware assistend VLAN tag insertion/stripping. o Add support for Tx/Rx flow control based on patches from Peer Chen. [2] o Add a routine that detects whether ethernet address swap routines is required. [3] o Add a workaround that take MAC/PHY out of power down mode. o Add suspend/resume support. o style(9) and code clean up. Special thanks to Shigeaki Tagashira, the original porter of nfe(4), who submitted lots of patches, performed uncountable number of regression tests and maintained nfe(4) for a long time. Without his enthusiastic help and support I could never have completed this overhauling task. The only weak point of nfe(4) compared to nve(4) is instability of manual half-duplex media selection on certain hardwares(auto sensing media type should work for all cases, though). This was a long standing bug of nfe(4) and I still have no idea why it doesn't work on some hardwares. Obtained from: OpenBSD [1] Submitted by: Peer Chen < pchen at nvidia dot com > [2], [3] Reviewed by: Shigeaki Tagashira < shigeaki AT se DOT hiroshima-u DOT ac DOT jp > Tested by: Shigeaki Tagashira, current Discussed with: current Silence from: obrien
2007-06-12 02:16:02 +00:00
0,
NULL, NULL,
&ring->tx_data_tag);
Add a pure open source nForce Ethernet driver, under BSDL. This driver was ported from OpenBSD by Shigeaki Tagashira <shigeaki@se.hiroshima-u.ac.jp> and posted at http://www.se.hiroshima-u.ac.jp/~shigeaki/software/freebsd-nfe.html It was additionally cleaned up by me. It is still a work-in-progress and thus is purposefully not in GENERIC. And it conflicts with nve(4), so only one should be loaded.
2006-06-26 23:41:07 +00:00
if (error != 0) {
Bring overhauled nfe(4) into tree. o s/printf/device_printf/g o Nuke OpenBSDism. o Nuke NetBSD/OpenBSD specific DMA sync operations.(we don't have a way to sync a single descriptor within a DMA map.) o Remove recursive mutex. o bus_dma(9) clean up. o 40bit DMA address support. o Add protection for Rx map load failure. o Fix a long standing bug for watchdog timeout. [1] o Add additional protections, missing Tx completion interrupt, losing start Tx command, for watchdog timeout. o Switch to taskqueue(9) API to handle interrupts. o Use our own timer for watchdog instead of if_watchdog/if_timer interface. o Advertise VLAN header length/capability correctly to upper layer. o Remove excessive kernel stack consumption in nfe_encap(). o Handle highly fragmented mbuf chains correctly. o Enable etherenet address reprogramming with ifconfig(8). o Add ALTQ/TSO, MSI/MSIX support. o Increased Rx ring to 256 descriptors from 128. o Align Tx/Rx descriptor ring on sizeof(struct nfe_desc64) boundary. o Remove alignment restrictions on Tx/Rx buffers. o Rewritten jumbo frame support code. o Add support for hardware assistend VLAN tag insertion/stripping. o Add support for Tx/Rx flow control based on patches from Peer Chen. [2] o Add a routine that detects whether ethernet address swap routines is required. [3] o Add a workaround that take MAC/PHY out of power down mode. o Add suspend/resume support. o style(9) and code clean up. Special thanks to Shigeaki Tagashira, the original porter of nfe(4), who submitted lots of patches, performed uncountable number of regression tests and maintained nfe(4) for a long time. Without his enthusiastic help and support I could never have completed this overhauling task. The only weak point of nfe(4) compared to nve(4) is instability of manual half-duplex media selection on certain hardwares(auto sensing media type should work for all cases, though). This was a long standing bug of nfe(4) and I still have no idea why it doesn't work on some hardwares. Obtained from: OpenBSD [1] Submitted by: Peer Chen < pchen at nvidia dot com > [2], [3] Reviewed by: Shigeaki Tagashira < shigeaki AT se DOT hiroshima-u DOT ac DOT jp > Tested by: Shigeaki Tagashira, current Discussed with: current Silence from: obrien
2007-06-12 02:16:02 +00:00
device_printf(sc->nfe_dev, "could not create Tx DMA tag\n");
goto fail;
Add a pure open source nForce Ethernet driver, under BSDL. This driver was ported from OpenBSD by Shigeaki Tagashira <shigeaki@se.hiroshima-u.ac.jp> and posted at http://www.se.hiroshima-u.ac.jp/~shigeaki/software/freebsd-nfe.html It was additionally cleaned up by me. It is still a work-in-progress and thus is purposefully not in GENERIC. And it conflicts with nve(4), so only one should be loaded.
2006-06-26 23:41:07 +00:00
}
for (i = 0; i < NFE_TX_RING_COUNT; i++) {
style(9) cleanup.
2006-10-19 10:01:26 +00:00
error = bus_dmamap_create(ring->tx_data_tag, 0,
&ring->data[i].tx_data_map);
Add a pure open source nForce Ethernet driver, under BSDL. This driver was ported from OpenBSD by Shigeaki Tagashira <shigeaki@se.hiroshima-u.ac.jp> and posted at http://www.se.hiroshima-u.ac.jp/~shigeaki/software/freebsd-nfe.html It was additionally cleaned up by me. It is still a work-in-progress and thus is purposefully not in GENERIC. And it conflicts with nve(4), so only one should be loaded.
2006-06-26 23:41:07 +00:00
if (error != 0) {
Bring overhauled nfe(4) into tree. o s/printf/device_printf/g o Nuke OpenBSDism. o Nuke NetBSD/OpenBSD specific DMA sync operations.(we don't have a way to sync a single descriptor within a DMA map.) o Remove recursive mutex. o bus_dma(9) clean up. o 40bit DMA address support. o Add protection for Rx map load failure. o Fix a long standing bug for watchdog timeout. [1] o Add additional protections, missing Tx completion interrupt, losing start Tx command, for watchdog timeout. o Switch to taskqueue(9) API to handle interrupts. o Use our own timer for watchdog instead of if_watchdog/if_timer interface. o Advertise VLAN header length/capability correctly to upper layer. o Remove excessive kernel stack consumption in nfe_encap(). o Handle highly fragmented mbuf chains correctly. o Enable etherenet address reprogramming with ifconfig(8). o Add ALTQ/TSO, MSI/MSIX support. o Increased Rx ring to 256 descriptors from 128. o Align Tx/Rx descriptor ring on sizeof(struct nfe_desc64) boundary. o Remove alignment restrictions on Tx/Rx buffers. o Rewritten jumbo frame support code. o Add support for hardware assistend VLAN tag insertion/stripping. o Add support for Tx/Rx flow control based on patches from Peer Chen. [2] o Add a routine that detects whether ethernet address swap routines is required. [3] o Add a workaround that take MAC/PHY out of power down mode. o Add suspend/resume support. o style(9) and code clean up. Special thanks to Shigeaki Tagashira, the original porter of nfe(4), who submitted lots of patches, performed uncountable number of regression tests and maintained nfe(4) for a long time. Without his enthusiastic help and support I could never have completed this overhauling task. The only weak point of nfe(4) compared to nve(4) is instability of manual half-duplex media selection on certain hardwares(auto sensing media type should work for all cases, though). This was a long standing bug of nfe(4) and I still have no idea why it doesn't work on some hardwares. Obtained from: OpenBSD [1] Submitted by: Peer Chen < pchen at nvidia dot com > [2], [3] Reviewed by: Shigeaki Tagashira < shigeaki AT se DOT hiroshima-u DOT ac DOT jp > Tested by: Shigeaki Tagashira, current Discussed with: current Silence from: obrien
2007-06-12 02:16:02 +00:00
device_printf(sc->nfe_dev,
"could not create Tx DMA map\n");
Add a pure open source nForce Ethernet driver, under BSDL. This driver was ported from OpenBSD by Shigeaki Tagashira <shigeaki@se.hiroshima-u.ac.jp> and posted at http://www.se.hiroshima-u.ac.jp/~shigeaki/software/freebsd-nfe.html It was additionally cleaned up by me. It is still a work-in-progress and thus is purposefully not in GENERIC. And it conflicts with nve(4), so only one should be loaded.
2006-06-26 23:41:07 +00:00
goto fail;
}
}
Bring overhauled nfe(4) into tree. o s/printf/device_printf/g o Nuke OpenBSDism. o Nuke NetBSD/OpenBSD specific DMA sync operations.(we don't have a way to sync a single descriptor within a DMA map.) o Remove recursive mutex. o bus_dma(9) clean up. o 40bit DMA address support. o Add protection for Rx map load failure. o Fix a long standing bug for watchdog timeout. [1] o Add additional protections, missing Tx completion interrupt, losing start Tx command, for watchdog timeout. o Switch to taskqueue(9) API to handle interrupts. o Use our own timer for watchdog instead of if_watchdog/if_timer interface. o Advertise VLAN header length/capability correctly to upper layer. o Remove excessive kernel stack consumption in nfe_encap(). o Handle highly fragmented mbuf chains correctly. o Enable etherenet address reprogramming with ifconfig(8). o Add ALTQ/TSO, MSI/MSIX support. o Increased Rx ring to 256 descriptors from 128. o Align Tx/Rx descriptor ring on sizeof(struct nfe_desc64) boundary. o Remove alignment restrictions on Tx/Rx buffers. o Rewritten jumbo frame support code. o Add support for hardware assistend VLAN tag insertion/stripping. o Add support for Tx/Rx flow control based on patches from Peer Chen. [2] o Add a routine that detects whether ethernet address swap routines is required. [3] o Add a workaround that take MAC/PHY out of power down mode. o Add suspend/resume support. o style(9) and code clean up. Special thanks to Shigeaki Tagashira, the original porter of nfe(4), who submitted lots of patches, performed uncountable number of regression tests and maintained nfe(4) for a long time. Without his enthusiastic help and support I could never have completed this overhauling task. The only weak point of nfe(4) compared to nve(4) is instability of manual half-duplex media selection on certain hardwares(auto sensing media type should work for all cases, though). This was a long standing bug of nfe(4) and I still have no idea why it doesn't work on some hardwares. Obtained from: OpenBSD [1] Submitted by: Peer Chen < pchen at nvidia dot com > [2], [3] Reviewed by: Shigeaki Tagashira < shigeaki AT se DOT hiroshima-u DOT ac DOT jp > Tested by: Shigeaki Tagashira, current Discussed with: current Silence from: obrien
2007-06-12 02:16:02 +00:00
fail:
return (error);
Add a pure open source nForce Ethernet driver, under BSDL. This driver was ported from OpenBSD by Shigeaki Tagashira <shigeaki@se.hiroshima-u.ac.jp> and posted at http://www.se.hiroshima-u.ac.jp/~shigeaki/software/freebsd-nfe.html It was additionally cleaned up by me. It is still a work-in-progress and thus is purposefully not in GENERIC. And it conflicts with nve(4), so only one should be loaded.
2006-06-26 23:41:07 +00:00
}
static void
Bring overhauled nfe(4) into tree. o s/printf/device_printf/g o Nuke OpenBSDism. o Nuke NetBSD/OpenBSD specific DMA sync operations.(we don't have a way to sync a single descriptor within a DMA map.) o Remove recursive mutex. o bus_dma(9) clean up. o 40bit DMA address support. o Add protection for Rx map load failure. o Fix a long standing bug for watchdog timeout. [1] o Add additional protections, missing Tx completion interrupt, losing start Tx command, for watchdog timeout. o Switch to taskqueue(9) API to handle interrupts. o Use our own timer for watchdog instead of if_watchdog/if_timer interface. o Advertise VLAN header length/capability correctly to upper layer. o Remove excessive kernel stack consumption in nfe_encap(). o Handle highly fragmented mbuf chains correctly. o Enable etherenet address reprogramming with ifconfig(8). o Add ALTQ/TSO, MSI/MSIX support. o Increased Rx ring to 256 descriptors from 128. o Align Tx/Rx descriptor ring on sizeof(struct nfe_desc64) boundary. o Remove alignment restrictions on Tx/Rx buffers. o Rewritten jumbo frame support code. o Add support for hardware assistend VLAN tag insertion/stripping. o Add support for Tx/Rx flow control based on patches from Peer Chen. [2] o Add a routine that detects whether ethernet address swap routines is required. [3] o Add a workaround that take MAC/PHY out of power down mode. o Add suspend/resume support. o style(9) and code clean up. Special thanks to Shigeaki Tagashira, the original porter of nfe(4), who submitted lots of patches, performed uncountable number of regression tests and maintained nfe(4) for a long time. Without his enthusiastic help and support I could never have completed this overhauling task. The only weak point of nfe(4) compared to nve(4) is instability of manual half-duplex media selection on certain hardwares(auto sensing media type should work for all cases, though). This was a long standing bug of nfe(4) and I still have no idea why it doesn't work on some hardwares. Obtained from: OpenBSD [1] Submitted by: Peer Chen < pchen at nvidia dot com > [2], [3] Reviewed by: Shigeaki Tagashira < shigeaki AT se DOT hiroshima-u DOT ac DOT jp > Tested by: Shigeaki Tagashira, current Discussed with: current Silence from: obrien
2007-06-12 02:16:02 +00:00
nfe_init_tx_ring(struct nfe_softc *sc, struct nfe_tx_ring *ring)
Add a pure open source nForce Ethernet driver, under BSDL. This driver was ported from OpenBSD by Shigeaki Tagashira <shigeaki@se.hiroshima-u.ac.jp> and posted at http://www.se.hiroshima-u.ac.jp/~shigeaki/software/freebsd-nfe.html It was additionally cleaned up by me. It is still a work-in-progress and thus is purposefully not in GENERIC. And it conflicts with nve(4), so only one should be loaded.
2006-06-26 23:41:07 +00:00
{
Bring overhauled nfe(4) into tree. o s/printf/device_printf/g o Nuke OpenBSDism. o Nuke NetBSD/OpenBSD specific DMA sync operations.(we don't have a way to sync a single descriptor within a DMA map.) o Remove recursive mutex. o bus_dma(9) clean up. o 40bit DMA address support. o Add protection for Rx map load failure. o Fix a long standing bug for watchdog timeout. [1] o Add additional protections, missing Tx completion interrupt, losing start Tx command, for watchdog timeout. o Switch to taskqueue(9) API to handle interrupts. o Use our own timer for watchdog instead of if_watchdog/if_timer interface. o Advertise VLAN header length/capability correctly to upper layer. o Remove excessive kernel stack consumption in nfe_encap(). o Handle highly fragmented mbuf chains correctly. o Enable etherenet address reprogramming with ifconfig(8). o Add ALTQ/TSO, MSI/MSIX support. o Increased Rx ring to 256 descriptors from 128. o Align Tx/Rx descriptor ring on sizeof(struct nfe_desc64) boundary. o Remove alignment restrictions on Tx/Rx buffers. o Rewritten jumbo frame support code. o Add support for hardware assistend VLAN tag insertion/stripping. o Add support for Tx/Rx flow control based on patches from Peer Chen. [2] o Add a routine that detects whether ethernet address swap routines is required. [3] o Add a workaround that take MAC/PHY out of power down mode. o Add suspend/resume support. o style(9) and code clean up. Special thanks to Shigeaki Tagashira, the original porter of nfe(4), who submitted lots of patches, performed uncountable number of regression tests and maintained nfe(4) for a long time. Without his enthusiastic help and support I could never have completed this overhauling task. The only weak point of nfe(4) compared to nve(4) is instability of manual half-duplex media selection on certain hardwares(auto sensing media type should work for all cases, though). This was a long standing bug of nfe(4) and I still have no idea why it doesn't work on some hardwares. Obtained from: OpenBSD [1] Submitted by: Peer Chen < pchen at nvidia dot com > [2], [3] Reviewed by: Shigeaki Tagashira < shigeaki AT se DOT hiroshima-u DOT ac DOT jp > Tested by: Shigeaki Tagashira, current Discussed with: current Silence from: obrien
2007-06-12 02:16:02 +00:00
void *desc;
size_t descsize;
Add a pure open source nForce Ethernet driver, under BSDL. This driver was ported from OpenBSD by Shigeaki Tagashira <shigeaki@se.hiroshima-u.ac.jp> and posted at http://www.se.hiroshima-u.ac.jp/~shigeaki/software/freebsd-nfe.html It was additionally cleaned up by me. It is still a work-in-progress and thus is purposefully not in GENERIC. And it conflicts with nve(4), so only one should be loaded.
2006-06-26 23:41:07 +00:00
Bring overhauled nfe(4) into tree. o s/printf/device_printf/g o Nuke OpenBSDism. o Nuke NetBSD/OpenBSD specific DMA sync operations.(we don't have a way to sync a single descriptor within a DMA map.) o Remove recursive mutex. o bus_dma(9) clean up. o 40bit DMA address support. o Add protection for Rx map load failure. o Fix a long standing bug for watchdog timeout. [1] o Add additional protections, missing Tx completion interrupt, losing start Tx command, for watchdog timeout. o Switch to taskqueue(9) API to handle interrupts. o Use our own timer for watchdog instead of if_watchdog/if_timer interface. o Advertise VLAN header length/capability correctly to upper layer. o Remove excessive kernel stack consumption in nfe_encap(). o Handle highly fragmented mbuf chains correctly. o Enable etherenet address reprogramming with ifconfig(8). o Add ALTQ/TSO, MSI/MSIX support. o Increased Rx ring to 256 descriptors from 128. o Align Tx/Rx descriptor ring on sizeof(struct nfe_desc64) boundary. o Remove alignment restrictions on Tx/Rx buffers. o Rewritten jumbo frame support code. o Add support for hardware assistend VLAN tag insertion/stripping. o Add support for Tx/Rx flow control based on patches from Peer Chen. [2] o Add a routine that detects whether ethernet address swap routines is required. [3] o Add a workaround that take MAC/PHY out of power down mode. o Add suspend/resume support. o style(9) and code clean up. Special thanks to Shigeaki Tagashira, the original porter of nfe(4), who submitted lots of patches, performed uncountable number of regression tests and maintained nfe(4) for a long time. Without his enthusiastic help and support I could never have completed this overhauling task. The only weak point of nfe(4) compared to nve(4) is instability of manual half-duplex media selection on certain hardwares(auto sensing media type should work for all cases, though). This was a long standing bug of nfe(4) and I still have no idea why it doesn't work on some hardwares. Obtained from: OpenBSD [1] Submitted by: Peer Chen < pchen at nvidia dot com > [2], [3] Reviewed by: Shigeaki Tagashira < shigeaki AT se DOT hiroshima-u DOT ac DOT jp > Tested by: Shigeaki Tagashira, current Discussed with: current Silence from: obrien
2007-06-12 02:16:02 +00:00
sc->nfe_force_tx = 0;
ring->queued = 0;
ring->cur = ring->next = 0;
if (sc->nfe_flags & NFE_40BIT_ADDR) {
desc = ring->desc64;
descsize = sizeof (struct nfe_desc64);
} else {
desc = ring->desc32;
descsize = sizeof (struct nfe_desc32);
Add a pure open source nForce Ethernet driver, under BSDL. This driver was ported from OpenBSD by Shigeaki Tagashira <shigeaki@se.hiroshima-u.ac.jp> and posted at http://www.se.hiroshima-u.ac.jp/~shigeaki/software/freebsd-nfe.html It was additionally cleaned up by me. It is still a work-in-progress and thus is purposefully not in GENERIC. And it conflicts with nve(4), so only one should be loaded.
2006-06-26 23:41:07 +00:00
}
Bring overhauled nfe(4) into tree. o s/printf/device_printf/g o Nuke OpenBSDism. o Nuke NetBSD/OpenBSD specific DMA sync operations.(we don't have a way to sync a single descriptor within a DMA map.) o Remove recursive mutex. o bus_dma(9) clean up. o 40bit DMA address support. o Add protection for Rx map load failure. o Fix a long standing bug for watchdog timeout. [1] o Add additional protections, missing Tx completion interrupt, losing start Tx command, for watchdog timeout. o Switch to taskqueue(9) API to handle interrupts. o Use our own timer for watchdog instead of if_watchdog/if_timer interface. o Advertise VLAN header length/capability correctly to upper layer. o Remove excessive kernel stack consumption in nfe_encap(). o Handle highly fragmented mbuf chains correctly. o Enable etherenet address reprogramming with ifconfig(8). o Add ALTQ/TSO, MSI/MSIX support. o Increased Rx ring to 256 descriptors from 128. o Align Tx/Rx descriptor ring on sizeof(struct nfe_desc64) boundary. o Remove alignment restrictions on Tx/Rx buffers. o Rewritten jumbo frame support code. o Add support for hardware assistend VLAN tag insertion/stripping. o Add support for Tx/Rx flow control based on patches from Peer Chen. [2] o Add a routine that detects whether ethernet address swap routines is required. [3] o Add a workaround that take MAC/PHY out of power down mode. o Add suspend/resume support. o style(9) and code clean up. Special thanks to Shigeaki Tagashira, the original porter of nfe(4), who submitted lots of patches, performed uncountable number of regression tests and maintained nfe(4) for a long time. Without his enthusiastic help and support I could never have completed this overhauling task. The only weak point of nfe(4) compared to nve(4) is instability of manual half-duplex media selection on certain hardwares(auto sensing media type should work for all cases, though). This was a long standing bug of nfe(4) and I still have no idea why it doesn't work on some hardwares. Obtained from: OpenBSD [1] Submitted by: Peer Chen < pchen at nvidia dot com > [2], [3] Reviewed by: Shigeaki Tagashira < shigeaki AT se DOT hiroshima-u DOT ac DOT jp > Tested by: Shigeaki Tagashira, current Discussed with: current Silence from: obrien
2007-06-12 02:16:02 +00:00
bzero(desc, descsize * NFE_TX_RING_COUNT);
Add a pure open source nForce Ethernet driver, under BSDL. This driver was ported from OpenBSD by Shigeaki Tagashira <shigeaki@se.hiroshima-u.ac.jp> and posted at http://www.se.hiroshima-u.ac.jp/~shigeaki/software/freebsd-nfe.html It was additionally cleaned up by me. It is still a work-in-progress and thus is purposefully not in GENERIC. And it conflicts with nve(4), so only one should be loaded.
2006-06-26 23:41:07 +00:00
style(9) cleanup.
2006-10-19 10:01:26 +00:00
bus_dmamap_sync(ring->tx_desc_tag, ring->tx_desc_map,
Bring overhauled nfe(4) into tree. o s/printf/device_printf/g o Nuke OpenBSDism. o Nuke NetBSD/OpenBSD specific DMA sync operations.(we don't have a way to sync a single descriptor within a DMA map.) o Remove recursive mutex. o bus_dma(9) clean up. o 40bit DMA address support. o Add protection for Rx map load failure. o Fix a long standing bug for watchdog timeout. [1] o Add additional protections, missing Tx completion interrupt, losing start Tx command, for watchdog timeout. o Switch to taskqueue(9) API to handle interrupts. o Use our own timer for watchdog instead of if_watchdog/if_timer interface. o Advertise VLAN header length/capability correctly to upper layer. o Remove excessive kernel stack consumption in nfe_encap(). o Handle highly fragmented mbuf chains correctly. o Enable etherenet address reprogramming with ifconfig(8). o Add ALTQ/TSO, MSI/MSIX support. o Increased Rx ring to 256 descriptors from 128. o Align Tx/Rx descriptor ring on sizeof(struct nfe_desc64) boundary. o Remove alignment restrictions on Tx/Rx buffers. o Rewritten jumbo frame support code. o Add support for hardware assistend VLAN tag insertion/stripping. o Add support for Tx/Rx flow control based on patches from Peer Chen. [2] o Add a routine that detects whether ethernet address swap routines is required. [3] o Add a workaround that take MAC/PHY out of power down mode. o Add suspend/resume support. o style(9) and code clean up. Special thanks to Shigeaki Tagashira, the original porter of nfe(4), who submitted lots of patches, performed uncountable number of regression tests and maintained nfe(4) for a long time. Without his enthusiastic help and support I could never have completed this overhauling task. The only weak point of nfe(4) compared to nve(4) is instability of manual half-duplex media selection on certain hardwares(auto sensing media type should work for all cases, though). This was a long standing bug of nfe(4) and I still have no idea why it doesn't work on some hardwares. Obtained from: OpenBSD [1] Submitted by: Peer Chen < pchen at nvidia dot com > [2], [3] Reviewed by: Shigeaki Tagashira < shigeaki AT se DOT hiroshima-u DOT ac DOT jp > Tested by: Shigeaki Tagashira, current Discussed with: current Silence from: obrien
2007-06-12 02:16:02 +00:00
BUS_DMASYNC_PREREAD | BUS_DMASYNC_PREWRITE);
Add a pure open source nForce Ethernet driver, under BSDL. This driver was ported from OpenBSD by Shigeaki Tagashira <shigeaki@se.hiroshima-u.ac.jp> and posted at http://www.se.hiroshima-u.ac.jp/~shigeaki/software/freebsd-nfe.html It was additionally cleaned up by me. It is still a work-in-progress and thus is purposefully not in GENERIC. And it conflicts with nve(4), so only one should be loaded.
2006-06-26 23:41:07 +00:00
}
style(9) cleanup.
2006-10-19 10:01:26 +00:00
Add a pure open source nForce Ethernet driver, under BSDL. This driver was ported from OpenBSD by Shigeaki Tagashira <shigeaki@se.hiroshima-u.ac.jp> and posted at http://www.se.hiroshima-u.ac.jp/~shigeaki/software/freebsd-nfe.html It was additionally cleaned up by me. It is still a work-in-progress and thus is purposefully not in GENERIC. And it conflicts with nve(4), so only one should be loaded.
2006-06-26 23:41:07 +00:00
static void
nfe_free_tx_ring(struct nfe_softc *sc, struct nfe_tx_ring *ring)
{
struct nfe_tx_data *data;
void *desc;
int i, descsize;
if (sc->nfe_flags & NFE_40BIT_ADDR) {
desc = ring->desc64;
descsize = sizeof (struct nfe_desc64);
} else {
desc = ring->desc32;
descsize = sizeof (struct nfe_desc32);
}
for (i = 0; i < NFE_TX_RING_COUNT; i++) {
data = &ring->data[i];
if (data->m != NULL) {
Bring overhauled nfe(4) into tree. o s/printf/device_printf/g o Nuke OpenBSDism. o Nuke NetBSD/OpenBSD specific DMA sync operations.(we don't have a way to sync a single descriptor within a DMA map.) o Remove recursive mutex. o bus_dma(9) clean up. o 40bit DMA address support. o Add protection for Rx map load failure. o Fix a long standing bug for watchdog timeout. [1] o Add additional protections, missing Tx completion interrupt, losing start Tx command, for watchdog timeout. o Switch to taskqueue(9) API to handle interrupts. o Use our own timer for watchdog instead of if_watchdog/if_timer interface. o Advertise VLAN header length/capability correctly to upper layer. o Remove excessive kernel stack consumption in nfe_encap(). o Handle highly fragmented mbuf chains correctly. o Enable etherenet address reprogramming with ifconfig(8). o Add ALTQ/TSO, MSI/MSIX support. o Increased Rx ring to 256 descriptors from 128. o Align Tx/Rx descriptor ring on sizeof(struct nfe_desc64) boundary. o Remove alignment restrictions on Tx/Rx buffers. o Rewritten jumbo frame support code. o Add support for hardware assistend VLAN tag insertion/stripping. o Add support for Tx/Rx flow control based on patches from Peer Chen. [2] o Add a routine that detects whether ethernet address swap routines is required. [3] o Add a workaround that take MAC/PHY out of power down mode. o Add suspend/resume support. o style(9) and code clean up. Special thanks to Shigeaki Tagashira, the original porter of nfe(4), who submitted lots of patches, performed uncountable number of regression tests and maintained nfe(4) for a long time. Without his enthusiastic help and support I could never have completed this overhauling task. The only weak point of nfe(4) compared to nve(4) is instability of manual half-duplex media selection on certain hardwares(auto sensing media type should work for all cases, though). This was a long standing bug of nfe(4) and I still have no idea why it doesn't work on some hardwares. Obtained from: OpenBSD [1] Submitted by: Peer Chen < pchen at nvidia dot com > [2], [3] Reviewed by: Shigeaki Tagashira < shigeaki AT se DOT hiroshima-u DOT ac DOT jp > Tested by: Shigeaki Tagashira, current Discussed with: current Silence from: obrien
2007-06-12 02:16:02 +00:00
bus_dmamap_sync(ring->tx_data_tag, data->tx_data_map,
style(9) cleanup.
2006-10-19 10:01:26 +00:00
BUS_DMASYNC_POSTWRITE);
Bring overhauled nfe(4) into tree. o s/printf/device_printf/g o Nuke OpenBSDism. o Nuke NetBSD/OpenBSD specific DMA sync operations.(we don't have a way to sync a single descriptor within a DMA map.) o Remove recursive mutex. o bus_dma(9) clean up. o 40bit DMA address support. o Add protection for Rx map load failure. o Fix a long standing bug for watchdog timeout. [1] o Add additional protections, missing Tx completion interrupt, losing start Tx command, for watchdog timeout. o Switch to taskqueue(9) API to handle interrupts. o Use our own timer for watchdog instead of if_watchdog/if_timer interface. o Advertise VLAN header length/capability correctly to upper layer. o Remove excessive kernel stack consumption in nfe_encap(). o Handle highly fragmented mbuf chains correctly. o Enable etherenet address reprogramming with ifconfig(8). o Add ALTQ/TSO, MSI/MSIX support. o Increased Rx ring to 256 descriptors from 128. o Align Tx/Rx descriptor ring on sizeof(struct nfe_desc64) boundary. o Remove alignment restrictions on Tx/Rx buffers. o Rewritten jumbo frame support code. o Add support for hardware assistend VLAN tag insertion/stripping. o Add support for Tx/Rx flow control based on patches from Peer Chen. [2] o Add a routine that detects whether ethernet address swap routines is required. [3] o Add a workaround that take MAC/PHY out of power down mode. o Add suspend/resume support. o style(9) and code clean up. Special thanks to Shigeaki Tagashira, the original porter of nfe(4), who submitted lots of patches, performed uncountable number of regression tests and maintained nfe(4) for a long time. Without his enthusiastic help and support I could never have completed this overhauling task. The only weak point of nfe(4) compared to nve(4) is instability of manual half-duplex media selection on certain hardwares(auto sensing media type should work for all cases, though). This was a long standing bug of nfe(4) and I still have no idea why it doesn't work on some hardwares. Obtained from: OpenBSD [1] Submitted by: Peer Chen < pchen at nvidia dot com > [2], [3] Reviewed by: Shigeaki Tagashira < shigeaki AT se DOT hiroshima-u DOT ac DOT jp > Tested by: Shigeaki Tagashira, current Discussed with: current Silence from: obrien
2007-06-12 02:16:02 +00:00
bus_dmamap_unload(ring->tx_data_tag, data->tx_data_map);
Add a pure open source nForce Ethernet driver, under BSDL. This driver was ported from OpenBSD by Shigeaki Tagashira <shigeaki@se.hiroshima-u.ac.jp> and posted at http://www.se.hiroshima-u.ac.jp/~shigeaki/software/freebsd-nfe.html It was additionally cleaned up by me. It is still a work-in-progress and thus is purposefully not in GENERIC. And it conflicts with nve(4), so only one should be loaded.
2006-06-26 23:41:07 +00:00
m_freem(data->m);
Bring overhauled nfe(4) into tree. o s/printf/device_printf/g o Nuke OpenBSDism. o Nuke NetBSD/OpenBSD specific DMA sync operations.(we don't have a way to sync a single descriptor within a DMA map.) o Remove recursive mutex. o bus_dma(9) clean up. o 40bit DMA address support. o Add protection for Rx map load failure. o Fix a long standing bug for watchdog timeout. [1] o Add additional protections, missing Tx completion interrupt, losing start Tx command, for watchdog timeout. o Switch to taskqueue(9) API to handle interrupts. o Use our own timer for watchdog instead of if_watchdog/if_timer interface. o Advertise VLAN header length/capability correctly to upper layer. o Remove excessive kernel stack consumption in nfe_encap(). o Handle highly fragmented mbuf chains correctly. o Enable etherenet address reprogramming with ifconfig(8). o Add ALTQ/TSO, MSI/MSIX support. o Increased Rx ring to 256 descriptors from 128. o Align Tx/Rx descriptor ring on sizeof(struct nfe_desc64) boundary. o Remove alignment restrictions on Tx/Rx buffers. o Rewritten jumbo frame support code. o Add support for hardware assistend VLAN tag insertion/stripping. o Add support for Tx/Rx flow control based on patches from Peer Chen. [2] o Add a routine that detects whether ethernet address swap routines is required. [3] o Add a workaround that take MAC/PHY out of power down mode. o Add suspend/resume support. o style(9) and code clean up. Special thanks to Shigeaki Tagashira, the original porter of nfe(4), who submitted lots of patches, performed uncountable number of regression tests and maintained nfe(4) for a long time. Without his enthusiastic help and support I could never have completed this overhauling task. The only weak point of nfe(4) compared to nve(4) is instability of manual half-duplex media selection on certain hardwares(auto sensing media type should work for all cases, though). This was a long standing bug of nfe(4) and I still have no idea why it doesn't work on some hardwares. Obtained from: OpenBSD [1] Submitted by: Peer Chen < pchen at nvidia dot com > [2], [3] Reviewed by: Shigeaki Tagashira < shigeaki AT se DOT hiroshima-u DOT ac DOT jp > Tested by: Shigeaki Tagashira, current Discussed with: current Silence from: obrien
2007-06-12 02:16:02 +00:00
data->m = NULL;
}
if (data->tx_data_map != NULL) {
bus_dmamap_destroy(ring->tx_data_tag,
data->tx_data_map);
data->tx_data_map = NULL;
Add a pure open source nForce Ethernet driver, under BSDL. This driver was ported from OpenBSD by Shigeaki Tagashira <shigeaki@se.hiroshima-u.ac.jp> and posted at http://www.se.hiroshima-u.ac.jp/~shigeaki/software/freebsd-nfe.html It was additionally cleaned up by me. It is still a work-in-progress and thus is purposefully not in GENERIC. And it conflicts with nve(4), so only one should be loaded.
2006-06-26 23:41:07 +00:00
}
}
Bring overhauled nfe(4) into tree. o s/printf/device_printf/g o Nuke OpenBSDism. o Nuke NetBSD/OpenBSD specific DMA sync operations.(we don't have a way to sync a single descriptor within a DMA map.) o Remove recursive mutex. o bus_dma(9) clean up. o 40bit DMA address support. o Add protection for Rx map load failure. o Fix a long standing bug for watchdog timeout. [1] o Add additional protections, missing Tx completion interrupt, losing start Tx command, for watchdog timeout. o Switch to taskqueue(9) API to handle interrupts. o Use our own timer for watchdog instead of if_watchdog/if_timer interface. o Advertise VLAN header length/capability correctly to upper layer. o Remove excessive kernel stack consumption in nfe_encap(). o Handle highly fragmented mbuf chains correctly. o Enable etherenet address reprogramming with ifconfig(8). o Add ALTQ/TSO, MSI/MSIX support. o Increased Rx ring to 256 descriptors from 128. o Align Tx/Rx descriptor ring on sizeof(struct nfe_desc64) boundary. o Remove alignment restrictions on Tx/Rx buffers. o Rewritten jumbo frame support code. o Add support for hardware assistend VLAN tag insertion/stripping. o Add support for Tx/Rx flow control based on patches from Peer Chen. [2] o Add a routine that detects whether ethernet address swap routines is required. [3] o Add a workaround that take MAC/PHY out of power down mode. o Add suspend/resume support. o style(9) and code clean up. Special thanks to Shigeaki Tagashira, the original porter of nfe(4), who submitted lots of patches, performed uncountable number of regression tests and maintained nfe(4) for a long time. Without his enthusiastic help and support I could never have completed this overhauling task. The only weak point of nfe(4) compared to nve(4) is instability of manual half-duplex media selection on certain hardwares(auto sensing media type should work for all cases, though). This was a long standing bug of nfe(4) and I still have no idea why it doesn't work on some hardwares. Obtained from: OpenBSD [1] Submitted by: Peer Chen < pchen at nvidia dot com > [2], [3] Reviewed by: Shigeaki Tagashira < shigeaki AT se DOT hiroshima-u DOT ac DOT jp > Tested by: Shigeaki Tagashira, current Discussed with: current Silence from: obrien
2007-06-12 02:16:02 +00:00
if (ring->tx_data_tag != NULL) {
bus_dma_tag_destroy(ring->tx_data_tag);
ring->tx_data_tag = NULL;
Add a pure open source nForce Ethernet driver, under BSDL. This driver was ported from OpenBSD by Shigeaki Tagashira <shigeaki@se.hiroshima-u.ac.jp> and posted at http://www.se.hiroshima-u.ac.jp/~shigeaki/software/freebsd-nfe.html It was additionally cleaned up by me. It is still a work-in-progress and thus is purposefully not in GENERIC. And it conflicts with nve(4), so only one should be loaded.
2006-06-26 23:41:07 +00:00
}
Bring overhauled nfe(4) into tree. o s/printf/device_printf/g o Nuke OpenBSDism. o Nuke NetBSD/OpenBSD specific DMA sync operations.(we don't have a way to sync a single descriptor within a DMA map.) o Remove recursive mutex. o bus_dma(9) clean up. o 40bit DMA address support. o Add protection for Rx map load failure. o Fix a long standing bug for watchdog timeout. [1] o Add additional protections, missing Tx completion interrupt, losing start Tx command, for watchdog timeout. o Switch to taskqueue(9) API to handle interrupts. o Use our own timer for watchdog instead of if_watchdog/if_timer interface. o Advertise VLAN header length/capability correctly to upper layer. o Remove excessive kernel stack consumption in nfe_encap(). o Handle highly fragmented mbuf chains correctly. o Enable etherenet address reprogramming with ifconfig(8). o Add ALTQ/TSO, MSI/MSIX support. o Increased Rx ring to 256 descriptors from 128. o Align Tx/Rx descriptor ring on sizeof(struct nfe_desc64) boundary. o Remove alignment restrictions on Tx/Rx buffers. o Rewritten jumbo frame support code. o Add support for hardware assistend VLAN tag insertion/stripping. o Add support for Tx/Rx flow control based on patches from Peer Chen. [2] o Add a routine that detects whether ethernet address swap routines is required. [3] o Add a workaround that take MAC/PHY out of power down mode. o Add suspend/resume support. o style(9) and code clean up. Special thanks to Shigeaki Tagashira, the original porter of nfe(4), who submitted lots of patches, performed uncountable number of regression tests and maintained nfe(4) for a long time. Without his enthusiastic help and support I could never have completed this overhauling task. The only weak point of nfe(4) compared to nve(4) is instability of manual half-duplex media selection on certain hardwares(auto sensing media type should work for all cases, though). This was a long standing bug of nfe(4) and I still have no idea why it doesn't work on some hardwares. Obtained from: OpenBSD [1] Submitted by: Peer Chen < pchen at nvidia dot com > [2], [3] Reviewed by: Shigeaki Tagashira < shigeaki AT se DOT hiroshima-u DOT ac DOT jp > Tested by: Shigeaki Tagashira, current Discussed with: current Silence from: obrien
2007-06-12 02:16:02 +00:00
if (desc != NULL) {
bus_dmamap_sync(ring->tx_desc_tag, ring->tx_desc_map,
BUS_DMASYNC_POSTWRITE);
bus_dmamap_unload(ring->tx_desc_tag, ring->tx_desc_map);
bus_dmamem_free(ring->tx_desc_tag, desc, ring->tx_desc_map);
ring->desc64 = NULL;
ring->desc32 = NULL;
ring->tx_desc_map = NULL;
bus_dma_tag_destroy(ring->tx_desc_tag);
ring->tx_desc_tag = NULL;
}
Add a pure open source nForce Ethernet driver, under BSDL. This driver was ported from OpenBSD by Shigeaki Tagashira <shigeaki@se.hiroshima-u.ac.jp> and posted at http://www.se.hiroshima-u.ac.jp/~shigeaki/software/freebsd-nfe.html It was additionally cleaned up by me. It is still a work-in-progress and thus is purposefully not in GENERIC. And it conflicts with nve(4), so only one should be loaded.
2006-06-26 23:41:07 +00:00
}
#ifdef DEVICE_POLLING
static poll_handler_t nfe_poll;
style(9) cleanup.
2006-10-19 10:01:26 +00:00
When user_frac in the polling subsystem is low it is going to busy the CPU for too long period than necessary. Additively, interfaces are kept polled (in the tick) even if no more packets are available. In order to avoid such situations a new generic mechanism can be implemented in proactive way, keeping track of the time spent on any packet and fragmenting the time for any tick, stopping the processing as soon as possible. In order to implement such mechanism, the polling handler needs to change, returning the number of packets processed. While the intended logic is not part of this patch, the polling KPI is broken by this commit, adding an int return value and the new flag IFCAP_POLLING_NOCOUNT (which will signal that the return value is meaningless for the installed handler and checking should be skipped). Bump __FreeBSD_version in order to signal such situation. Reviewed by: emaste Sponsored by: Sandvine Incorporated
2009-05-30 15:14:44 +00:00
static int
Add a pure open source nForce Ethernet driver, under BSDL. This driver was ported from OpenBSD by Shigeaki Tagashira <shigeaki@se.hiroshima-u.ac.jp> and posted at http://www.se.hiroshima-u.ac.jp/~shigeaki/software/freebsd-nfe.html It was additionally cleaned up by me. It is still a work-in-progress and thus is purposefully not in GENERIC. And it conflicts with nve(4), so only one should be loaded.
2006-06-26 23:41:07 +00:00
nfe_poll(struct ifnet *ifp, enum poll_cmd cmd, int count)
{
Missed two nits in rev 1.7.
2006-11-17 16:49:40 +00:00
struct nfe_softc *sc = ifp->if_softc;
Bring overhauled nfe(4) into tree. o s/printf/device_printf/g o Nuke OpenBSDism. o Nuke NetBSD/OpenBSD specific DMA sync operations.(we don't have a way to sync a single descriptor within a DMA map.) o Remove recursive mutex. o bus_dma(9) clean up. o 40bit DMA address support. o Add protection for Rx map load failure. o Fix a long standing bug for watchdog timeout. [1] o Add additional protections, missing Tx completion interrupt, losing start Tx command, for watchdog timeout. o Switch to taskqueue(9) API to handle interrupts. o Use our own timer for watchdog instead of if_watchdog/if_timer interface. o Advertise VLAN header length/capability correctly to upper layer. o Remove excessive kernel stack consumption in nfe_encap(). o Handle highly fragmented mbuf chains correctly. o Enable etherenet address reprogramming with ifconfig(8). o Add ALTQ/TSO, MSI/MSIX support. o Increased Rx ring to 256 descriptors from 128. o Align Tx/Rx descriptor ring on sizeof(struct nfe_desc64) boundary. o Remove alignment restrictions on Tx/Rx buffers. o Rewritten jumbo frame support code. o Add support for hardware assistend VLAN tag insertion/stripping. o Add support for Tx/Rx flow control based on patches from Peer Chen. [2] o Add a routine that detects whether ethernet address swap routines is required. [3] o Add a workaround that take MAC/PHY out of power down mode. o Add suspend/resume support. o style(9) and code clean up. Special thanks to Shigeaki Tagashira, the original porter of nfe(4), who submitted lots of patches, performed uncountable number of regression tests and maintained nfe(4) for a long time. Without his enthusiastic help and support I could never have completed this overhauling task. The only weak point of nfe(4) compared to nve(4) is instability of manual half-duplex media selection on certain hardwares(auto sensing media type should work for all cases, though). This was a long standing bug of nfe(4) and I still have no idea why it doesn't work on some hardwares. Obtained from: OpenBSD [1] Submitted by: Peer Chen < pchen at nvidia dot com > [2], [3] Reviewed by: Shigeaki Tagashira < shigeaki AT se DOT hiroshima-u DOT ac DOT jp > Tested by: Shigeaki Tagashira, current Discussed with: current Silence from: obrien
2007-06-12 02:16:02 +00:00
uint32_t r;
When user_frac in the polling subsystem is low it is going to busy the CPU for too long period than necessary. Additively, interfaces are kept polled (in the tick) even if no more packets are available. In order to avoid such situations a new generic mechanism can be implemented in proactive way, keeping track of the time spent on any packet and fragmenting the time for any tick, stopping the processing as soon as possible. In order to implement such mechanism, the polling handler needs to change, returning the number of packets processed. While the intended logic is not part of this patch, the polling KPI is broken by this commit, adding an int return value and the new flag IFCAP_POLLING_NOCOUNT (which will signal that the return value is meaningless for the installed handler and checking should be skipped). Bump __FreeBSD_version in order to signal such situation. Reviewed by: emaste Sponsored by: Sandvine Incorporated
2009-05-30 15:14:44 +00:00
int rx_npkts = 0;
Add a pure open source nForce Ethernet driver, under BSDL. This driver was ported from OpenBSD by Shigeaki Tagashira <shigeaki@se.hiroshima-u.ac.jp> and posted at http://www.se.hiroshima-u.ac.jp/~shigeaki/software/freebsd-nfe.html It was additionally cleaned up by me. It is still a work-in-progress and thus is purposefully not in GENERIC. And it conflicts with nve(4), so only one should be loaded.
2006-06-26 23:41:07 +00:00
NFE_LOCK(sc);
if (!(ifp->if_drv_flags & IFF_DRV_RUNNING)) {
Bring overhauled nfe(4) into tree. o s/printf/device_printf/g o Nuke OpenBSDism. o Nuke NetBSD/OpenBSD specific DMA sync operations.(we don't have a way to sync a single descriptor within a DMA map.) o Remove recursive mutex. o bus_dma(9) clean up. o 40bit DMA address support. o Add protection for Rx map load failure. o Fix a long standing bug for watchdog timeout. [1] o Add additional protections, missing Tx completion interrupt, losing start Tx command, for watchdog timeout. o Switch to taskqueue(9) API to handle interrupts. o Use our own timer for watchdog instead of if_watchdog/if_timer interface. o Advertise VLAN header length/capability correctly to upper layer. o Remove excessive kernel stack consumption in nfe_encap(). o Handle highly fragmented mbuf chains correctly. o Enable etherenet address reprogramming with ifconfig(8). o Add ALTQ/TSO, MSI/MSIX support. o Increased Rx ring to 256 descriptors from 128. o Align Tx/Rx descriptor ring on sizeof(struct nfe_desc64) boundary. o Remove alignment restrictions on Tx/Rx buffers. o Rewritten jumbo frame support code. o Add support for hardware assistend VLAN tag insertion/stripping. o Add support for Tx/Rx flow control based on patches from Peer Chen. [2] o Add a routine that detects whether ethernet address swap routines is required. [3] o Add a workaround that take MAC/PHY out of power down mode. o Add suspend/resume support. o style(9) and code clean up. Special thanks to Shigeaki Tagashira, the original porter of nfe(4), who submitted lots of patches, performed uncountable number of regression tests and maintained nfe(4) for a long time. Without his enthusiastic help and support I could never have completed this overhauling task. The only weak point of nfe(4) compared to nve(4) is instability of manual half-duplex media selection on certain hardwares(auto sensing media type should work for all cases, though). This was a long standing bug of nfe(4) and I still have no idea why it doesn't work on some hardwares. Obtained from: OpenBSD [1] Submitted by: Peer Chen < pchen at nvidia dot com > [2], [3] Reviewed by: Shigeaki Tagashira < shigeaki AT se DOT hiroshima-u DOT ac DOT jp > Tested by: Shigeaki Tagashira, current Discussed with: current Silence from: obrien
2007-06-12 02:16:02 +00:00
NFE_UNLOCK(sc);
When user_frac in the polling subsystem is low it is going to busy the CPU for too long period than necessary. Additively, interfaces are kept polled (in the tick) even if no more packets are available. In order to avoid such situations a new generic mechanism can be implemented in proactive way, keeping track of the time spent on any packet and fragmenting the time for any tick, stopping the processing as soon as possible. In order to implement such mechanism, the polling handler needs to change, returning the number of packets processed. While the intended logic is not part of this patch, the polling KPI is broken by this commit, adding an int return value and the new flag IFCAP_POLLING_NOCOUNT (which will signal that the return value is meaningless for the installed handler and checking should be skipped). Bump __FreeBSD_version in order to signal such situation. Reviewed by: emaste Sponsored by: Sandvine Incorporated
2009-05-30 15:14:44 +00:00
return (rx_npkts);
Add a pure open source nForce Ethernet driver, under BSDL. This driver was ported from OpenBSD by Shigeaki Tagashira <shigeaki@se.hiroshima-u.ac.jp> and posted at http://www.se.hiroshima-u.ac.jp/~shigeaki/software/freebsd-nfe.html It was additionally cleaned up by me. It is still a work-in-progress and thus is purposefully not in GENERIC. And it conflicts with nve(4), so only one should be loaded.
2006-06-26 23:41:07 +00:00
}
Don't fail on device attach if jumbo frame support was unsuccessful. Because nfe(4) hardware doesn't support SG on Rx path, supporting jumbo frame requires very large contiguous kernel memory(i.e. several mega bytes). In case of lack of contiguous kernel memory that allocation request may always fail. However nfe(4) can operate on normal sized MTU frames, so go ahead and just disable jumbo frame support. While I'm here add a new tunable "hw.nfe.jumbo_disable" to disable jumbo frame support. In nfe_poll, make sure to invoke correct Rx handler. Approved by: re (kensmith)
2007-07-24 01:11:00 +00:00
if (sc->nfe_framesize > MCLBYTES - ETHER_HDR_LEN)
When user_frac in the polling subsystem is low it is going to busy the CPU for too long period than necessary. Additively, interfaces are kept polled (in the tick) even if no more packets are available. In order to avoid such situations a new generic mechanism can be implemented in proactive way, keeping track of the time spent on any packet and fragmenting the time for any tick, stopping the processing as soon as possible. In order to implement such mechanism, the polling handler needs to change, returning the number of packets processed. While the intended logic is not part of this patch, the polling KPI is broken by this commit, adding an int return value and the new flag IFCAP_POLLING_NOCOUNT (which will signal that the return value is meaningless for the installed handler and checking should be skipped). Bump __FreeBSD_version in order to signal such situation. Reviewed by: emaste Sponsored by: Sandvine Incorporated
2009-05-30 15:14:44 +00:00
rx_npkts = nfe_jrxeof(sc, count, &rx_npkts);
Don't fail on device attach if jumbo frame support was unsuccessful. Because nfe(4) hardware doesn't support SG on Rx path, supporting jumbo frame requires very large contiguous kernel memory(i.e. several mega bytes). In case of lack of contiguous kernel memory that allocation request may always fail. However nfe(4) can operate on normal sized MTU frames, so go ahead and just disable jumbo frame support. While I'm here add a new tunable "hw.nfe.jumbo_disable" to disable jumbo frame support. In nfe_poll, make sure to invoke correct Rx handler. Approved by: re (kensmith)
2007-07-24 01:11:00 +00:00
else
When user_frac in the polling subsystem is low it is going to busy the CPU for too long period than necessary. Additively, interfaces are kept polled (in the tick) even if no more packets are available. In order to avoid such situations a new generic mechanism can be implemented in proactive way, keeping track of the time spent on any packet and fragmenting the time for any tick, stopping the processing as soon as possible. In order to implement such mechanism, the polling handler needs to change, returning the number of packets processed. While the intended logic is not part of this patch, the polling KPI is broken by this commit, adding an int return value and the new flag IFCAP_POLLING_NOCOUNT (which will signal that the return value is meaningless for the installed handler and checking should be skipped). Bump __FreeBSD_version in order to signal such situation. Reviewed by: emaste Sponsored by: Sandvine Incorporated
2009-05-30 15:14:44 +00:00
rx_npkts = nfe_rxeof(sc, count, &rx_npkts);
Add a pure open source nForce Ethernet driver, under BSDL. This driver was ported from OpenBSD by Shigeaki Tagashira <shigeaki@se.hiroshima-u.ac.jp> and posted at http://www.se.hiroshima-u.ac.jp/~shigeaki/software/freebsd-nfe.html It was additionally cleaned up by me. It is still a work-in-progress and thus is purposefully not in GENERIC. And it conflicts with nve(4), so only one should be loaded.
2006-06-26 23:41:07 +00:00
nfe_txeof(sc);
if (!IFQ_DRV_IS_EMPTY(&ifp->if_snd))
Add a 'locked' variant of the foo_start() routine and call it directly from interrupt handlers and watchdog routines instead of queueing a task to call foo_start(). Reviewed by: yongari MFC after: 1 month
2011-01-03 18:28:30 +00:00
nfe_start_locked(ifp);
Add a pure open source nForce Ethernet driver, under BSDL. This driver was ported from OpenBSD by Shigeaki Tagashira <shigeaki@se.hiroshima-u.ac.jp> and posted at http://www.se.hiroshima-u.ac.jp/~shigeaki/software/freebsd-nfe.html It was additionally cleaned up by me. It is still a work-in-progress and thus is purposefully not in GENERIC. And it conflicts with nve(4), so only one should be loaded.
2006-06-26 23:41:07 +00:00
if (cmd == POLL_AND_CHECK_STATUS) {
Bring overhauled nfe(4) into tree. o s/printf/device_printf/g o Nuke OpenBSDism. o Nuke NetBSD/OpenBSD specific DMA sync operations.(we don't have a way to sync a single descriptor within a DMA map.) o Remove recursive mutex. o bus_dma(9) clean up. o 40bit DMA address support. o Add protection for Rx map load failure. o Fix a long standing bug for watchdog timeout. [1] o Add additional protections, missing Tx completion interrupt, losing start Tx command, for watchdog timeout. o Switch to taskqueue(9) API to handle interrupts. o Use our own timer for watchdog instead of if_watchdog/if_timer interface. o Advertise VLAN header length/capability correctly to upper layer. o Remove excessive kernel stack consumption in nfe_encap(). o Handle highly fragmented mbuf chains correctly. o Enable etherenet address reprogramming with ifconfig(8). o Add ALTQ/TSO, MSI/MSIX support. o Increased Rx ring to 256 descriptors from 128. o Align Tx/Rx descriptor ring on sizeof(struct nfe_desc64) boundary. o Remove alignment restrictions on Tx/Rx buffers. o Rewritten jumbo frame support code. o Add support for hardware assistend VLAN tag insertion/stripping. o Add support for Tx/Rx flow control based on patches from Peer Chen. [2] o Add a routine that detects whether ethernet address swap routines is required. [3] o Add a workaround that take MAC/PHY out of power down mode. o Add suspend/resume support. o style(9) and code clean up. Special thanks to Shigeaki Tagashira, the original porter of nfe(4), who submitted lots of patches, performed uncountable number of regression tests and maintained nfe(4) for a long time. Without his enthusiastic help and support I could never have completed this overhauling task. The only weak point of nfe(4) compared to nve(4) is instability of manual half-duplex media selection on certain hardwares(auto sensing media type should work for all cases, though). This was a long standing bug of nfe(4) and I still have no idea why it doesn't work on some hardwares. Obtained from: OpenBSD [1] Submitted by: Peer Chen < pchen at nvidia dot com > [2], [3] Reviewed by: Shigeaki Tagashira < shigeaki AT se DOT hiroshima-u DOT ac DOT jp > Tested by: Shigeaki Tagashira, current Discussed with: current Silence from: obrien
2007-06-12 02:16:02 +00:00
if ((r = NFE_READ(sc, sc->nfe_irq_status)) == 0) {
NFE_UNLOCK(sc);
When user_frac in the polling subsystem is low it is going to busy the CPU for too long period than necessary. Additively, interfaces are kept polled (in the tick) even if no more packets are available. In order to avoid such situations a new generic mechanism can be implemented in proactive way, keeping track of the time spent on any packet and fragmenting the time for any tick, stopping the processing as soon as possible. In order to implement such mechanism, the polling handler needs to change, returning the number of packets processed. While the intended logic is not part of this patch, the polling KPI is broken by this commit, adding an int return value and the new flag IFCAP_POLLING_NOCOUNT (which will signal that the return value is meaningless for the installed handler and checking should be skipped). Bump __FreeBSD_version in order to signal such situation. Reviewed by: emaste Sponsored by: Sandvine Incorporated
2009-05-30 15:14:44 +00:00
return (rx_npkts);
style(9) cleanup.
2006-10-19 10:01:26 +00:00
}
Bring overhauled nfe(4) into tree. o s/printf/device_printf/g o Nuke OpenBSDism. o Nuke NetBSD/OpenBSD specific DMA sync operations.(we don't have a way to sync a single descriptor within a DMA map.) o Remove recursive mutex. o bus_dma(9) clean up. o 40bit DMA address support. o Add protection for Rx map load failure. o Fix a long standing bug for watchdog timeout. [1] o Add additional protections, missing Tx completion interrupt, losing start Tx command, for watchdog timeout. o Switch to taskqueue(9) API to handle interrupts. o Use our own timer for watchdog instead of if_watchdog/if_timer interface. o Advertise VLAN header length/capability correctly to upper layer. o Remove excessive kernel stack consumption in nfe_encap(). o Handle highly fragmented mbuf chains correctly. o Enable etherenet address reprogramming with ifconfig(8). o Add ALTQ/TSO, MSI/MSIX support. o Increased Rx ring to 256 descriptors from 128. o Align Tx/Rx descriptor ring on sizeof(struct nfe_desc64) boundary. o Remove alignment restrictions on Tx/Rx buffers. o Rewritten jumbo frame support code. o Add support for hardware assistend VLAN tag insertion/stripping. o Add support for Tx/Rx flow control based on patches from Peer Chen. [2] o Add a routine that detects whether ethernet address swap routines is required. [3] o Add a workaround that take MAC/PHY out of power down mode. o Add suspend/resume support. o style(9) and code clean up. Special thanks to Shigeaki Tagashira, the original porter of nfe(4), who submitted lots of patches, performed uncountable number of regression tests and maintained nfe(4) for a long time. Without his enthusiastic help and support I could never have completed this overhauling task. The only weak point of nfe(4) compared to nve(4) is instability of manual half-duplex media selection on certain hardwares(auto sensing media type should work for all cases, though). This was a long standing bug of nfe(4) and I still have no idea why it doesn't work on some hardwares. Obtained from: OpenBSD [1] Submitted by: Peer Chen < pchen at nvidia dot com > [2], [3] Reviewed by: Shigeaki Tagashira < shigeaki AT se DOT hiroshima-u DOT ac DOT jp > Tested by: Shigeaki Tagashira, current Discussed with: current Silence from: obrien
2007-06-12 02:16:02 +00:00
NFE_WRITE(sc, sc->nfe_irq_status, r);
Add a pure open source nForce Ethernet driver, under BSDL. This driver was ported from OpenBSD by Shigeaki Tagashira <shigeaki@se.hiroshima-u.ac.jp> and posted at http://www.se.hiroshima-u.ac.jp/~shigeaki/software/freebsd-nfe.html It was additionally cleaned up by me. It is still a work-in-progress and thus is purposefully not in GENERIC. And it conflicts with nve(4), so only one should be loaded.
2006-06-26 23:41:07 +00:00
style(9) cleanup.
2006-10-19 10:01:26 +00:00
if (r & NFE_IRQ_LINK) {
NFE_READ(sc, NFE_PHY_STATUS);
NFE_WRITE(sc, NFE_PHY_STATUS, 0xf);
Bring overhauled nfe(4) into tree. o s/printf/device_printf/g o Nuke OpenBSDism. o Nuke NetBSD/OpenBSD specific DMA sync operations.(we don't have a way to sync a single descriptor within a DMA map.) o Remove recursive mutex. o bus_dma(9) clean up. o 40bit DMA address support. o Add protection for Rx map load failure. o Fix a long standing bug for watchdog timeout. [1] o Add additional protections, missing Tx completion interrupt, losing start Tx command, for watchdog timeout. o Switch to taskqueue(9) API to handle interrupts. o Use our own timer for watchdog instead of if_watchdog/if_timer interface. o Advertise VLAN header length/capability correctly to upper layer. o Remove excessive kernel stack consumption in nfe_encap(). o Handle highly fragmented mbuf chains correctly. o Enable etherenet address reprogramming with ifconfig(8). o Add ALTQ/TSO, MSI/MSIX support. o Increased Rx ring to 256 descriptors from 128. o Align Tx/Rx descriptor ring on sizeof(struct nfe_desc64) boundary. o Remove alignment restrictions on Tx/Rx buffers. o Rewritten jumbo frame support code. o Add support for hardware assistend VLAN tag insertion/stripping. o Add support for Tx/Rx flow control based on patches from Peer Chen. [2] o Add a routine that detects whether ethernet address swap routines is required. [3] o Add a workaround that take MAC/PHY out of power down mode. o Add suspend/resume support. o style(9) and code clean up. Special thanks to Shigeaki Tagashira, the original porter of nfe(4), who submitted lots of patches, performed uncountable number of regression tests and maintained nfe(4) for a long time. Without his enthusiastic help and support I could never have completed this overhauling task. The only weak point of nfe(4) compared to nve(4) is instability of manual half-duplex media selection on certain hardwares(auto sensing media type should work for all cases, though). This was a long standing bug of nfe(4) and I still have no idea why it doesn't work on some hardwares. Obtained from: OpenBSD [1] Submitted by: Peer Chen < pchen at nvidia dot com > [2], [3] Reviewed by: Shigeaki Tagashira < shigeaki AT se DOT hiroshima-u DOT ac DOT jp > Tested by: Shigeaki Tagashira, current Discussed with: current Silence from: obrien
2007-06-12 02:16:02 +00:00
DPRINTF(sc, "link state changed\n");
style(9) cleanup.
2006-10-19 10:01:26 +00:00
}
Add a pure open source nForce Ethernet driver, under BSDL. This driver was ported from OpenBSD by Shigeaki Tagashira <shigeaki@se.hiroshima-u.ac.jp> and posted at http://www.se.hiroshima-u.ac.jp/~shigeaki/software/freebsd-nfe.html It was additionally cleaned up by me. It is still a work-in-progress and thus is purposefully not in GENERIC. And it conflicts with nve(4), so only one should be loaded.
2006-06-26 23:41:07 +00:00
}
Bring overhauled nfe(4) into tree. o s/printf/device_printf/g o Nuke OpenBSDism. o Nuke NetBSD/OpenBSD specific DMA sync operations.(we don't have a way to sync a single descriptor within a DMA map.) o Remove recursive mutex. o bus_dma(9) clean up. o 40bit DMA address support. o Add protection for Rx map load failure. o Fix a long standing bug for watchdog timeout. [1] o Add additional protections, missing Tx completion interrupt, losing start Tx command, for watchdog timeout. o Switch to taskqueue(9) API to handle interrupts. o Use our own timer for watchdog instead of if_watchdog/if_timer interface. o Advertise VLAN header length/capability correctly to upper layer. o Remove excessive kernel stack consumption in nfe_encap(). o Handle highly fragmented mbuf chains correctly. o Enable etherenet address reprogramming with ifconfig(8). o Add ALTQ/TSO, MSI/MSIX support. o Increased Rx ring to 256 descriptors from 128. o Align Tx/Rx descriptor ring on sizeof(struct nfe_desc64) boundary. o Remove alignment restrictions on Tx/Rx buffers. o Rewritten jumbo frame support code. o Add support for hardware assistend VLAN tag insertion/stripping. o Add support for Tx/Rx flow control based on patches from Peer Chen. [2] o Add a routine that detects whether ethernet address swap routines is required. [3] o Add a workaround that take MAC/PHY out of power down mode. o Add suspend/resume support. o style(9) and code clean up. Special thanks to Shigeaki Tagashira, the original porter of nfe(4), who submitted lots of patches, performed uncountable number of regression tests and maintained nfe(4) for a long time. Without his enthusiastic help and support I could never have completed this overhauling task. The only weak point of nfe(4) compared to nve(4) is instability of manual half-duplex media selection on certain hardwares(auto sensing media type should work for all cases, though). This was a long standing bug of nfe(4) and I still have no idea why it doesn't work on some hardwares. Obtained from: OpenBSD [1] Submitted by: Peer Chen < pchen at nvidia dot com > [2], [3] Reviewed by: Shigeaki Tagashira < shigeaki AT se DOT hiroshima-u DOT ac DOT jp > Tested by: Shigeaki Tagashira, current Discussed with: current Silence from: obrien
2007-06-12 02:16:02 +00:00
NFE_UNLOCK(sc);
When user_frac in the polling subsystem is low it is going to busy the CPU for too long period than necessary. Additively, interfaces are kept polled (in the tick) even if no more packets are available. In order to avoid such situations a new generic mechanism can be implemented in proactive way, keeping track of the time spent on any packet and fragmenting the time for any tick, stopping the processing as soon as possible. In order to implement such mechanism, the polling handler needs to change, returning the number of packets processed. While the intended logic is not part of this patch, the polling KPI is broken by this commit, adding an int return value and the new flag IFCAP_POLLING_NOCOUNT (which will signal that the return value is meaningless for the installed handler and checking should be skipped). Bump __FreeBSD_version in order to signal such situation. Reviewed by: emaste Sponsored by: Sandvine Incorporated
2009-05-30 15:14:44 +00:00
return (rx_npkts);
Add a pure open source nForce Ethernet driver, under BSDL. This driver was ported from OpenBSD by Shigeaki Tagashira <shigeaki@se.hiroshima-u.ac.jp> and posted at http://www.se.hiroshima-u.ac.jp/~shigeaki/software/freebsd-nfe.html It was additionally cleaned up by me. It is still a work-in-progress and thus is purposefully not in GENERIC. And it conflicts with nve(4), so only one should be loaded.
2006-06-26 23:41:07 +00:00
}
#endif /* DEVICE_POLLING */
Bring overhauled nfe(4) into tree. o s/printf/device_printf/g o Nuke OpenBSDism. o Nuke NetBSD/OpenBSD specific DMA sync operations.(we don't have a way to sync a single descriptor within a DMA map.) o Remove recursive mutex. o bus_dma(9) clean up. o 40bit DMA address support. o Add protection for Rx map load failure. o Fix a long standing bug for watchdog timeout. [1] o Add additional protections, missing Tx completion interrupt, losing start Tx command, for watchdog timeout. o Switch to taskqueue(9) API to handle interrupts. o Use our own timer for watchdog instead of if_watchdog/if_timer interface. o Advertise VLAN header length/capability correctly to upper layer. o Remove excessive kernel stack consumption in nfe_encap(). o Handle highly fragmented mbuf chains correctly. o Enable etherenet address reprogramming with ifconfig(8). o Add ALTQ/TSO, MSI/MSIX support. o Increased Rx ring to 256 descriptors from 128. o Align Tx/Rx descriptor ring on sizeof(struct nfe_desc64) boundary. o Remove alignment restrictions on Tx/Rx buffers. o Rewritten jumbo frame support code. o Add support for hardware assistend VLAN tag insertion/stripping. o Add support for Tx/Rx flow control based on patches from Peer Chen. [2] o Add a routine that detects whether ethernet address swap routines is required. [3] o Add a workaround that take MAC/PHY out of power down mode. o Add suspend/resume support. o style(9) and code clean up. Special thanks to Shigeaki Tagashira, the original porter of nfe(4), who submitted lots of patches, performed uncountable number of regression tests and maintained nfe(4) for a long time. Without his enthusiastic help and support I could never have completed this overhauling task. The only weak point of nfe(4) compared to nve(4) is instability of manual half-duplex media selection on certain hardwares(auto sensing media type should work for all cases, though). This was a long standing bug of nfe(4) and I still have no idea why it doesn't work on some hardwares. Obtained from: OpenBSD [1] Submitted by: Peer Chen < pchen at nvidia dot com > [2], [3] Reviewed by: Shigeaki Tagashira < shigeaki AT se DOT hiroshima-u DOT ac DOT jp > Tested by: Shigeaki Tagashira, current Discussed with: current Silence from: obrien
2007-06-12 02:16:02 +00:00
static void
nfe_set_intr(struct nfe_softc *sc)
{
if (sc->nfe_msi != 0)
NFE_WRITE(sc, NFE_IRQ_MASK, NFE_IRQ_WANTED);
}
/* In MSIX, a write to mask reegisters behaves as XOR. */
static __inline void
nfe_enable_intr(struct nfe_softc *sc)
{
if (sc->nfe_msix != 0) {
/* XXX Should have a better way to enable interrupts! */
if (NFE_READ(sc, sc->nfe_irq_mask) == 0)
NFE_WRITE(sc, sc->nfe_irq_mask, sc->nfe_intrs);
} else
NFE_WRITE(sc, sc->nfe_irq_mask, sc->nfe_intrs);
}
static __inline void
nfe_disable_intr(struct nfe_softc *sc)
{
if (sc->nfe_msix != 0) {
/* XXX Should have a better way to disable interrupts! */
if (NFE_READ(sc, sc->nfe_irq_mask) != 0)
NFE_WRITE(sc, sc->nfe_irq_mask, sc->nfe_nointrs);
} else
NFE_WRITE(sc, sc->nfe_irq_mask, sc->nfe_nointrs);
}
Add a pure open source nForce Ethernet driver, under BSDL. This driver was ported from OpenBSD by Shigeaki Tagashira <shigeaki@se.hiroshima-u.ac.jp> and posted at http://www.se.hiroshima-u.ac.jp/~shigeaki/software/freebsd-nfe.html It was additionally cleaned up by me. It is still a work-in-progress and thus is purposefully not in GENERIC. And it conflicts with nve(4), so only one should be loaded.
2006-06-26 23:41:07 +00:00
static int
nfe_ioctl(struct ifnet *ifp, u_long cmd, caddr_t data)
{
Bring overhauled nfe(4) into tree. o s/printf/device_printf/g o Nuke OpenBSDism. o Nuke NetBSD/OpenBSD specific DMA sync operations.(we don't have a way to sync a single descriptor within a DMA map.) o Remove recursive mutex. o bus_dma(9) clean up. o 40bit DMA address support. o Add protection for Rx map load failure. o Fix a long standing bug for watchdog timeout. [1] o Add additional protections, missing Tx completion interrupt, losing start Tx command, for watchdog timeout. o Switch to taskqueue(9) API to handle interrupts. o Use our own timer for watchdog instead of if_watchdog/if_timer interface. o Advertise VLAN header length/capability correctly to upper layer. o Remove excessive kernel stack consumption in nfe_encap(). o Handle highly fragmented mbuf chains correctly. o Enable etherenet address reprogramming with ifconfig(8). o Add ALTQ/TSO, MSI/MSIX support. o Increased Rx ring to 256 descriptors from 128. o Align Tx/Rx descriptor ring on sizeof(struct nfe_desc64) boundary. o Remove alignment restrictions on Tx/Rx buffers. o Rewritten jumbo frame support code. o Add support for hardware assistend VLAN tag insertion/stripping. o Add support for Tx/Rx flow control based on patches from Peer Chen. [2] o Add a routine that detects whether ethernet address swap routines is required. [3] o Add a workaround that take MAC/PHY out of power down mode. o Add suspend/resume support. o style(9) and code clean up. Special thanks to Shigeaki Tagashira, the original porter of nfe(4), who submitted lots of patches, performed uncountable number of regression tests and maintained nfe(4) for a long time. Without his enthusiastic help and support I could never have completed this overhauling task. The only weak point of nfe(4) compared to nve(4) is instability of manual half-duplex media selection on certain hardwares(auto sensing media type should work for all cases, though). This was a long standing bug of nfe(4) and I still have no idea why it doesn't work on some hardwares. Obtained from: OpenBSD [1] Submitted by: Peer Chen < pchen at nvidia dot com > [2], [3] Reviewed by: Shigeaki Tagashira < shigeaki AT se DOT hiroshima-u DOT ac DOT jp > Tested by: Shigeaki Tagashira, current Discussed with: current Silence from: obrien
2007-06-12 02:16:02 +00:00
struct nfe_softc *sc;
struct ifreq *ifr;
style(9) cleanup.
2006-10-19 10:01:26 +00:00
struct mii_data *mii;
Bring overhauled nfe(4) into tree. o s/printf/device_printf/g o Nuke OpenBSDism. o Nuke NetBSD/OpenBSD specific DMA sync operations.(we don't have a way to sync a single descriptor within a DMA map.) o Remove recursive mutex. o bus_dma(9) clean up. o 40bit DMA address support. o Add protection for Rx map load failure. o Fix a long standing bug for watchdog timeout. [1] o Add additional protections, missing Tx completion interrupt, losing start Tx command, for watchdog timeout. o Switch to taskqueue(9) API to handle interrupts. o Use our own timer for watchdog instead of if_watchdog/if_timer interface. o Advertise VLAN header length/capability correctly to upper layer. o Remove excessive kernel stack consumption in nfe_encap(). o Handle highly fragmented mbuf chains correctly. o Enable etherenet address reprogramming with ifconfig(8). o Add ALTQ/TSO, MSI/MSIX support. o Increased Rx ring to 256 descriptors from 128. o Align Tx/Rx descriptor ring on sizeof(struct nfe_desc64) boundary. o Remove alignment restrictions on Tx/Rx buffers. o Rewritten jumbo frame support code. o Add support for hardware assistend VLAN tag insertion/stripping. o Add support for Tx/Rx flow control based on patches from Peer Chen. [2] o Add a routine that detects whether ethernet address swap routines is required. [3] o Add a workaround that take MAC/PHY out of power down mode. o Add suspend/resume support. o style(9) and code clean up. Special thanks to Shigeaki Tagashira, the original porter of nfe(4), who submitted lots of patches, performed uncountable number of regression tests and maintained nfe(4) for a long time. Without his enthusiastic help and support I could never have completed this overhauling task. The only weak point of nfe(4) compared to nve(4) is instability of manual half-duplex media selection on certain hardwares(auto sensing media type should work for all cases, though). This was a long standing bug of nfe(4) and I still have no idea why it doesn't work on some hardwares. Obtained from: OpenBSD [1] Submitted by: Peer Chen < pchen at nvidia dot com > [2], [3] Reviewed by: Shigeaki Tagashira < shigeaki AT se DOT hiroshima-u DOT ac DOT jp > Tested by: Shigeaki Tagashira, current Discussed with: current Silence from: obrien
2007-06-12 02:16:02 +00:00
int error, init, mask;
Add a pure open source nForce Ethernet driver, under BSDL. This driver was ported from OpenBSD by Shigeaki Tagashira <shigeaki@se.hiroshima-u.ac.jp> and posted at http://www.se.hiroshima-u.ac.jp/~shigeaki/software/freebsd-nfe.html It was additionally cleaned up by me. It is still a work-in-progress and thus is purposefully not in GENERIC. And it conflicts with nve(4), so only one should be loaded.
2006-06-26 23:41:07 +00:00
Bring overhauled nfe(4) into tree. o s/printf/device_printf/g o Nuke OpenBSDism. o Nuke NetBSD/OpenBSD specific DMA sync operations.(we don't have a way to sync a single descriptor within a DMA map.) o Remove recursive mutex. o bus_dma(9) clean up. o 40bit DMA address support. o Add protection for Rx map load failure. o Fix a long standing bug for watchdog timeout. [1] o Add additional protections, missing Tx completion interrupt, losing start Tx command, for watchdog timeout. o Switch to taskqueue(9) API to handle interrupts. o Use our own timer for watchdog instead of if_watchdog/if_timer interface. o Advertise VLAN header length/capability correctly to upper layer. o Remove excessive kernel stack consumption in nfe_encap(). o Handle highly fragmented mbuf chains correctly. o Enable etherenet address reprogramming with ifconfig(8). o Add ALTQ/TSO, MSI/MSIX support. o Increased Rx ring to 256 descriptors from 128. o Align Tx/Rx descriptor ring on sizeof(struct nfe_desc64) boundary. o Remove alignment restrictions on Tx/Rx buffers. o Rewritten jumbo frame support code. o Add support for hardware assistend VLAN tag insertion/stripping. o Add support for Tx/Rx flow control based on patches from Peer Chen. [2] o Add a routine that detects whether ethernet address swap routines is required. [3] o Add a workaround that take MAC/PHY out of power down mode. o Add suspend/resume support. o style(9) and code clean up. Special thanks to Shigeaki Tagashira, the original porter of nfe(4), who submitted lots of patches, performed uncountable number of regression tests and maintained nfe(4) for a long time. Without his enthusiastic help and support I could never have completed this overhauling task. The only weak point of nfe(4) compared to nve(4) is instability of manual half-duplex media selection on certain hardwares(auto sensing media type should work for all cases, though). This was a long standing bug of nfe(4) and I still have no idea why it doesn't work on some hardwares. Obtained from: OpenBSD [1] Submitted by: Peer Chen < pchen at nvidia dot com > [2], [3] Reviewed by: Shigeaki Tagashira < shigeaki AT se DOT hiroshima-u DOT ac DOT jp > Tested by: Shigeaki Tagashira, current Discussed with: current Silence from: obrien
2007-06-12 02:16:02 +00:00
sc = ifp->if_softc;
ifr = (struct ifreq *) data;
error = 0;
init = 0;
Add a pure open source nForce Ethernet driver, under BSDL. This driver was ported from OpenBSD by Shigeaki Tagashira <shigeaki@se.hiroshima-u.ac.jp> and posted at http://www.se.hiroshima-u.ac.jp/~shigeaki/software/freebsd-nfe.html It was additionally cleaned up by me. It is still a work-in-progress and thus is purposefully not in GENERIC. And it conflicts with nve(4), so only one should be loaded.
2006-06-26 23:41:07 +00:00
switch (cmd) {
case SIOCSIFMTU:
Bring overhauled nfe(4) into tree. o s/printf/device_printf/g o Nuke OpenBSDism. o Nuke NetBSD/OpenBSD specific DMA sync operations.(we don't have a way to sync a single descriptor within a DMA map.) o Remove recursive mutex. o bus_dma(9) clean up. o 40bit DMA address support. o Add protection for Rx map load failure. o Fix a long standing bug for watchdog timeout. [1] o Add additional protections, missing Tx completion interrupt, losing start Tx command, for watchdog timeout. o Switch to taskqueue(9) API to handle interrupts. o Use our own timer for watchdog instead of if_watchdog/if_timer interface. o Advertise VLAN header length/capability correctly to upper layer. o Remove excessive kernel stack consumption in nfe_encap(). o Handle highly fragmented mbuf chains correctly. o Enable etherenet address reprogramming with ifconfig(8). o Add ALTQ/TSO, MSI/MSIX support. o Increased Rx ring to 256 descriptors from 128. o Align Tx/Rx descriptor ring on sizeof(struct nfe_desc64) boundary. o Remove alignment restrictions on Tx/Rx buffers. o Rewritten jumbo frame support code. o Add support for hardware assistend VLAN tag insertion/stripping. o Add support for Tx/Rx flow control based on patches from Peer Chen. [2] o Add a routine that detects whether ethernet address swap routines is required. [3] o Add a workaround that take MAC/PHY out of power down mode. o Add suspend/resume support. o style(9) and code clean up. Special thanks to Shigeaki Tagashira, the original porter of nfe(4), who submitted lots of patches, performed uncountable number of regression tests and maintained nfe(4) for a long time. Without his enthusiastic help and support I could never have completed this overhauling task. The only weak point of nfe(4) compared to nve(4) is instability of manual half-duplex media selection on certain hardwares(auto sensing media type should work for all cases, though). This was a long standing bug of nfe(4) and I still have no idea why it doesn't work on some hardwares. Obtained from: OpenBSD [1] Submitted by: Peer Chen < pchen at nvidia dot com > [2], [3] Reviewed by: Shigeaki Tagashira < shigeaki AT se DOT hiroshima-u DOT ac DOT jp > Tested by: Shigeaki Tagashira, current Discussed with: current Silence from: obrien
2007-06-12 02:16:02 +00:00
if (ifr->ifr_mtu < ETHERMIN || ifr->ifr_mtu > NFE_JUMBO_MTU)
Fix the jumbo frame support. Submitted by: Peer Chen at nVIDIA (reworked by me)
2006-11-27 00:16:47 +00:00
error = EINVAL;
Bring overhauled nfe(4) into tree. o s/printf/device_printf/g o Nuke OpenBSDism. o Nuke NetBSD/OpenBSD specific DMA sync operations.(we don't have a way to sync a single descriptor within a DMA map.) o Remove recursive mutex. o bus_dma(9) clean up. o 40bit DMA address support. o Add protection for Rx map load failure. o Fix a long standing bug for watchdog timeout. [1] o Add additional protections, missing Tx completion interrupt, losing start Tx command, for watchdog timeout. o Switch to taskqueue(9) API to handle interrupts. o Use our own timer for watchdog instead of if_watchdog/if_timer interface. o Advertise VLAN header length/capability correctly to upper layer. o Remove excessive kernel stack consumption in nfe_encap(). o Handle highly fragmented mbuf chains correctly. o Enable etherenet address reprogramming with ifconfig(8). o Add ALTQ/TSO, MSI/MSIX support. o Increased Rx ring to 256 descriptors from 128. o Align Tx/Rx descriptor ring on sizeof(struct nfe_desc64) boundary. o Remove alignment restrictions on Tx/Rx buffers. o Rewritten jumbo frame support code. o Add support for hardware assistend VLAN tag insertion/stripping. o Add support for Tx/Rx flow control based on patches from Peer Chen. [2] o Add a routine that detects whether ethernet address swap routines is required. [3] o Add a workaround that take MAC/PHY out of power down mode. o Add suspend/resume support. o style(9) and code clean up. Special thanks to Shigeaki Tagashira, the original porter of nfe(4), who submitted lots of patches, performed uncountable number of regression tests and maintained nfe(4) for a long time. Without his enthusiastic help and support I could never have completed this overhauling task. The only weak point of nfe(4) compared to nve(4) is instability of manual half-duplex media selection on certain hardwares(auto sensing media type should work for all cases, though). This was a long standing bug of nfe(4) and I still have no idea why it doesn't work on some hardwares. Obtained from: OpenBSD [1] Submitted by: Peer Chen < pchen at nvidia dot com > [2], [3] Reviewed by: Shigeaki Tagashira < shigeaki AT se DOT hiroshima-u DOT ac DOT jp > Tested by: Shigeaki Tagashira, current Discussed with: current Silence from: obrien
2007-06-12 02:16:02 +00:00
else if (ifp->if_mtu != ifr->ifr_mtu) {
Don't fail on device attach if jumbo frame support was unsuccessful. Because nfe(4) hardware doesn't support SG on Rx path, supporting jumbo frame requires very large contiguous kernel memory(i.e. several mega bytes). In case of lack of contiguous kernel memory that allocation request may always fail. However nfe(4) can operate on normal sized MTU frames, so go ahead and just disable jumbo frame support. While I'm here add a new tunable "hw.nfe.jumbo_disable" to disable jumbo frame support. In nfe_poll, make sure to invoke correct Rx handler. Approved by: re (kensmith)
2007-07-24 01:11:00 +00:00
if ((((sc->nfe_flags & NFE_JUMBO_SUP) == 0) ||
(sc->nfe_jumbo_disable != 0)) &&
Bring overhauled nfe(4) into tree. o s/printf/device_printf/g o Nuke OpenBSDism. o Nuke NetBSD/OpenBSD specific DMA sync operations.(we don't have a way to sync a single descriptor within a DMA map.) o Remove recursive mutex. o bus_dma(9) clean up. o 40bit DMA address support. o Add protection for Rx map load failure. o Fix a long standing bug for watchdog timeout. [1] o Add additional protections, missing Tx completion interrupt, losing start Tx command, for watchdog timeout. o Switch to taskqueue(9) API to handle interrupts. o Use our own timer for watchdog instead of if_watchdog/if_timer interface. o Advertise VLAN header length/capability correctly to upper layer. o Remove excessive kernel stack consumption in nfe_encap(). o Handle highly fragmented mbuf chains correctly. o Enable etherenet address reprogramming with ifconfig(8). o Add ALTQ/TSO, MSI/MSIX support. o Increased Rx ring to 256 descriptors from 128. o Align Tx/Rx descriptor ring on sizeof(struct nfe_desc64) boundary. o Remove alignment restrictions on Tx/Rx buffers. o Rewritten jumbo frame support code. o Add support for hardware assistend VLAN tag insertion/stripping. o Add support for Tx/Rx flow control based on patches from Peer Chen. [2] o Add a routine that detects whether ethernet address swap routines is required. [3] o Add a workaround that take MAC/PHY out of power down mode. o Add suspend/resume support. o style(9) and code clean up. Special thanks to Shigeaki Tagashira, the original porter of nfe(4), who submitted lots of patches, performed uncountable number of regression tests and maintained nfe(4) for a long time. Without his enthusiastic help and support I could never have completed this overhauling task. The only weak point of nfe(4) compared to nve(4) is instability of manual half-duplex media selection on certain hardwares(auto sensing media type should work for all cases, though). This was a long standing bug of nfe(4) and I still have no idea why it doesn't work on some hardwares. Obtained from: OpenBSD [1] Submitted by: Peer Chen < pchen at nvidia dot com > [2], [3] Reviewed by: Shigeaki Tagashira < shigeaki AT se DOT hiroshima-u DOT ac DOT jp > Tested by: Shigeaki Tagashira, current Discussed with: current Silence from: obrien
2007-06-12 02:16:02 +00:00
ifr->ifr_mtu > ETHERMTU)
error = EINVAL;
else {
NFE_LOCK(sc);
ifp->if_mtu = ifr->ifr_mtu;
Reinitialize driver when MTU is changed and driver is running. Reported by: Huang, Yusheng ( yusheng.huang <> bluecoat com )
2011-01-24 17:53:58 +00:00
if (ifp->if_drv_flags & IFF_DRV_RUNNING) {
ifp->if_drv_flags &= ~IFF_DRV_RUNNING;
Bring overhauled nfe(4) into tree. o s/printf/device_printf/g o Nuke OpenBSDism. o Nuke NetBSD/OpenBSD specific DMA sync operations.(we don't have a way to sync a single descriptor within a DMA map.) o Remove recursive mutex. o bus_dma(9) clean up. o 40bit DMA address support. o Add protection for Rx map load failure. o Fix a long standing bug for watchdog timeout. [1] o Add additional protections, missing Tx completion interrupt, losing start Tx command, for watchdog timeout. o Switch to taskqueue(9) API to handle interrupts. o Use our own timer for watchdog instead of if_watchdog/if_timer interface. o Advertise VLAN header length/capability correctly to upper layer. o Remove excessive kernel stack consumption in nfe_encap(). o Handle highly fragmented mbuf chains correctly. o Enable etherenet address reprogramming with ifconfig(8). o Add ALTQ/TSO, MSI/MSIX support. o Increased Rx ring to 256 descriptors from 128. o Align Tx/Rx descriptor ring on sizeof(struct nfe_desc64) boundary. o Remove alignment restrictions on Tx/Rx buffers. o Rewritten jumbo frame support code. o Add support for hardware assistend VLAN tag insertion/stripping. o Add support for Tx/Rx flow control based on patches from Peer Chen. [2] o Add a routine that detects whether ethernet address swap routines is required. [3] o Add a workaround that take MAC/PHY out of power down mode. o Add suspend/resume support. o style(9) and code clean up. Special thanks to Shigeaki Tagashira, the original porter of nfe(4), who submitted lots of patches, performed uncountable number of regression tests and maintained nfe(4) for a long time. Without his enthusiastic help and support I could never have completed this overhauling task. The only weak point of nfe(4) compared to nve(4) is instability of manual half-duplex media selection on certain hardwares(auto sensing media type should work for all cases, though). This was a long standing bug of nfe(4) and I still have no idea why it doesn't work on some hardwares. Obtained from: OpenBSD [1] Submitted by: Peer Chen < pchen at nvidia dot com > [2], [3] Reviewed by: Shigeaki Tagashira < shigeaki AT se DOT hiroshima-u DOT ac DOT jp > Tested by: Shigeaki Tagashira, current Discussed with: current Silence from: obrien
2007-06-12 02:16:02 +00:00
nfe_init_locked(sc);
Reinitialize driver when MTU is changed and driver is running. Reported by: Huang, Yusheng ( yusheng.huang <> bluecoat com )
2011-01-24 17:53:58 +00:00
}
Bring overhauled nfe(4) into tree. o s/printf/device_printf/g o Nuke OpenBSDism. o Nuke NetBSD/OpenBSD specific DMA sync operations.(we don't have a way to sync a single descriptor within a DMA map.) o Remove recursive mutex. o bus_dma(9) clean up. o 40bit DMA address support. o Add protection for Rx map load failure. o Fix a long standing bug for watchdog timeout. [1] o Add additional protections, missing Tx completion interrupt, losing start Tx command, for watchdog timeout. o Switch to taskqueue(9) API to handle interrupts. o Use our own timer for watchdog instead of if_watchdog/if_timer interface. o Advertise VLAN header length/capability correctly to upper layer. o Remove excessive kernel stack consumption in nfe_encap(). o Handle highly fragmented mbuf chains correctly. o Enable etherenet address reprogramming with ifconfig(8). o Add ALTQ/TSO, MSI/MSIX support. o Increased Rx ring to 256 descriptors from 128. o Align Tx/Rx descriptor ring on sizeof(struct nfe_desc64) boundary. o Remove alignment restrictions on Tx/Rx buffers. o Rewritten jumbo frame support code. o Add support for hardware assistend VLAN tag insertion/stripping. o Add support for Tx/Rx flow control based on patches from Peer Chen. [2] o Add a routine that detects whether ethernet address swap routines is required. [3] o Add a workaround that take MAC/PHY out of power down mode. o Add suspend/resume support. o style(9) and code clean up. Special thanks to Shigeaki Tagashira, the original porter of nfe(4), who submitted lots of patches, performed uncountable number of regression tests and maintained nfe(4) for a long time. Without his enthusiastic help and support I could never have completed this overhauling task. The only weak point of nfe(4) compared to nve(4) is instability of manual half-duplex media selection on certain hardwares(auto sensing media type should work for all cases, though). This was a long standing bug of nfe(4) and I still have no idea why it doesn't work on some hardwares. Obtained from: OpenBSD [1] Submitted by: Peer Chen < pchen at nvidia dot com > [2], [3] Reviewed by: Shigeaki Tagashira < shigeaki AT se DOT hiroshima-u DOT ac DOT jp > Tested by: Shigeaki Tagashira, current Discussed with: current Silence from: obrien
2007-06-12 02:16:02 +00:00
NFE_UNLOCK(sc);
Fix the jumbo frame support. Submitted by: Peer Chen at nVIDIA (reworked by me)
2006-11-27 00:16:47 +00:00
}
Add a pure open source nForce Ethernet driver, under BSDL. This driver was ported from OpenBSD by Shigeaki Tagashira <shigeaki@se.hiroshima-u.ac.jp> and posted at http://www.se.hiroshima-u.ac.jp/~shigeaki/software/freebsd-nfe.html It was additionally cleaned up by me. It is still a work-in-progress and thus is purposefully not in GENERIC. And it conflicts with nve(4), so only one should be loaded.
2006-06-26 23:41:07 +00:00
}
break;
case SIOCSIFFLAGS:
NFE_LOCK(sc);
if (ifp->if_flags & IFF_UP) {
/*
* If only the PROMISC or ALLMULTI flag changes, then
* don't do a full re-init of the chip, just update
* the Rx filter.
*/
if ((ifp->if_drv_flags & IFF_DRV_RUNNING) &&
((ifp->if_flags ^ sc->nfe_if_flags) &
(IFF_ALLMULTI | IFF_PROMISC)) != 0)
nfe_setmulti(sc);
else
nfe_init_locked(sc);
} else {
if (ifp->if_drv_flags & IFF_DRV_RUNNING)
Bring overhauled nfe(4) into tree. o s/printf/device_printf/g o Nuke OpenBSDism. o Nuke NetBSD/OpenBSD specific DMA sync operations.(we don't have a way to sync a single descriptor within a DMA map.) o Remove recursive mutex. o bus_dma(9) clean up. o 40bit DMA address support. o Add protection for Rx map load failure. o Fix a long standing bug for watchdog timeout. [1] o Add additional protections, missing Tx completion interrupt, losing start Tx command, for watchdog timeout. o Switch to taskqueue(9) API to handle interrupts. o Use our own timer for watchdog instead of if_watchdog/if_timer interface. o Advertise VLAN header length/capability correctly to upper layer. o Remove excessive kernel stack consumption in nfe_encap(). o Handle highly fragmented mbuf chains correctly. o Enable etherenet address reprogramming with ifconfig(8). o Add ALTQ/TSO, MSI/MSIX support. o Increased Rx ring to 256 descriptors from 128. o Align Tx/Rx descriptor ring on sizeof(struct nfe_desc64) boundary. o Remove alignment restrictions on Tx/Rx buffers. o Rewritten jumbo frame support code. o Add support for hardware assistend VLAN tag insertion/stripping. o Add support for Tx/Rx flow control based on patches from Peer Chen. [2] o Add a routine that detects whether ethernet address swap routines is required. [3] o Add a workaround that take MAC/PHY out of power down mode. o Add suspend/resume support. o style(9) and code clean up. Special thanks to Shigeaki Tagashira, the original porter of nfe(4), who submitted lots of patches, performed uncountable number of regression tests and maintained nfe(4) for a long time. Without his enthusiastic help and support I could never have completed this overhauling task. The only weak point of nfe(4) compared to nve(4) is instability of manual half-duplex media selection on certain hardwares(auto sensing media type should work for all cases, though). This was a long standing bug of nfe(4) and I still have no idea why it doesn't work on some hardwares. Obtained from: OpenBSD [1] Submitted by: Peer Chen < pchen at nvidia dot com > [2], [3] Reviewed by: Shigeaki Tagashira < shigeaki AT se DOT hiroshima-u DOT ac DOT jp > Tested by: Shigeaki Tagashira, current Discussed with: current Silence from: obrien
2007-06-12 02:16:02 +00:00
nfe_stop(ifp);
Add a pure open source nForce Ethernet driver, under BSDL. This driver was ported from OpenBSD by Shigeaki Tagashira <shigeaki@se.hiroshima-u.ac.jp> and posted at http://www.se.hiroshima-u.ac.jp/~shigeaki/software/freebsd-nfe.html It was additionally cleaned up by me. It is still a work-in-progress and thus is purposefully not in GENERIC. And it conflicts with nve(4), so only one should be loaded.
2006-06-26 23:41:07 +00:00
}
sc->nfe_if_flags = ifp->if_flags;
NFE_UNLOCK(sc);
error = 0;
break;
case SIOCADDMULTI:
case SIOCDELMULTI:
Bring overhauled nfe(4) into tree. o s/printf/device_printf/g o Nuke OpenBSDism. o Nuke NetBSD/OpenBSD specific DMA sync operations.(we don't have a way to sync a single descriptor within a DMA map.) o Remove recursive mutex. o bus_dma(9) clean up. o 40bit DMA address support. o Add protection for Rx map load failure. o Fix a long standing bug for watchdog timeout. [1] o Add additional protections, missing Tx completion interrupt, losing start Tx command, for watchdog timeout. o Switch to taskqueue(9) API to handle interrupts. o Use our own timer for watchdog instead of if_watchdog/if_timer interface. o Advertise VLAN header length/capability correctly to upper layer. o Remove excessive kernel stack consumption in nfe_encap(). o Handle highly fragmented mbuf chains correctly. o Enable etherenet address reprogramming with ifconfig(8). o Add ALTQ/TSO, MSI/MSIX support. o Increased Rx ring to 256 descriptors from 128. o Align Tx/Rx descriptor ring on sizeof(struct nfe_desc64) boundary. o Remove alignment restrictions on Tx/Rx buffers. o Rewritten jumbo frame support code. o Add support for hardware assistend VLAN tag insertion/stripping. o Add support for Tx/Rx flow control based on patches from Peer Chen. [2] o Add a routine that detects whether ethernet address swap routines is required. [3] o Add a workaround that take MAC/PHY out of power down mode. o Add suspend/resume support. o style(9) and code clean up. Special thanks to Shigeaki Tagashira, the original porter of nfe(4), who submitted lots of patches, performed uncountable number of regression tests and maintained nfe(4) for a long time. Without his enthusiastic help and support I could never have completed this overhauling task. The only weak point of nfe(4) compared to nve(4) is instability of manual half-duplex media selection on certain hardwares(auto sensing media type should work for all cases, though). This was a long standing bug of nfe(4) and I still have no idea why it doesn't work on some hardwares. Obtained from: OpenBSD [1] Submitted by: Peer Chen < pchen at nvidia dot com > [2], [3] Reviewed by: Shigeaki Tagashira < shigeaki AT se DOT hiroshima-u DOT ac DOT jp > Tested by: Shigeaki Tagashira, current Discussed with: current Silence from: obrien
2007-06-12 02:16:02 +00:00
if ((ifp->if_drv_flags & IFF_DRV_RUNNING) != 0) {
Add a pure open source nForce Ethernet driver, under BSDL. This driver was ported from OpenBSD by Shigeaki Tagashira <shigeaki@se.hiroshima-u.ac.jp> and posted at http://www.se.hiroshima-u.ac.jp/~shigeaki/software/freebsd-nfe.html It was additionally cleaned up by me. It is still a work-in-progress and thus is purposefully not in GENERIC. And it conflicts with nve(4), so only one should be loaded.
2006-06-26 23:41:07 +00:00
NFE_LOCK(sc);
nfe_setmulti(sc);
NFE_UNLOCK(sc);
error = 0;
}
break;
case SIOCSIFMEDIA:
case SIOCGIFMEDIA:
mii = device_get_softc(sc->nfe_miibus);
error = ifmedia_ioctl(ifp, ifr, &mii->mii_media, cmd);
break;
case SIOCSIFCAP:
Bring overhauled nfe(4) into tree. o s/printf/device_printf/g o Nuke OpenBSDism. o Nuke NetBSD/OpenBSD specific DMA sync operations.(we don't have a way to sync a single descriptor within a DMA map.) o Remove recursive mutex. o bus_dma(9) clean up. o 40bit DMA address support. o Add protection for Rx map load failure. o Fix a long standing bug for watchdog timeout. [1] o Add additional protections, missing Tx completion interrupt, losing start Tx command, for watchdog timeout. o Switch to taskqueue(9) API to handle interrupts. o Use our own timer for watchdog instead of if_watchdog/if_timer interface. o Advertise VLAN header length/capability correctly to upper layer. o Remove excessive kernel stack consumption in nfe_encap(). o Handle highly fragmented mbuf chains correctly. o Enable etherenet address reprogramming with ifconfig(8). o Add ALTQ/TSO, MSI/MSIX support. o Increased Rx ring to 256 descriptors from 128. o Align Tx/Rx descriptor ring on sizeof(struct nfe_desc64) boundary. o Remove alignment restrictions on Tx/Rx buffers. o Rewritten jumbo frame support code. o Add support for hardware assistend VLAN tag insertion/stripping. o Add support for Tx/Rx flow control based on patches from Peer Chen. [2] o Add a routine that detects whether ethernet address swap routines is required. [3] o Add a workaround that take MAC/PHY out of power down mode. o Add suspend/resume support. o style(9) and code clean up. Special thanks to Shigeaki Tagashira, the original porter of nfe(4), who submitted lots of patches, performed uncountable number of regression tests and maintained nfe(4) for a long time. Without his enthusiastic help and support I could never have completed this overhauling task. The only weak point of nfe(4) compared to nve(4) is instability of manual half-duplex media selection on certain hardwares(auto sensing media type should work for all cases, though). This was a long standing bug of nfe(4) and I still have no idea why it doesn't work on some hardwares. Obtained from: OpenBSD [1] Submitted by: Peer Chen < pchen at nvidia dot com > [2], [3] Reviewed by: Shigeaki Tagashira < shigeaki AT se DOT hiroshima-u DOT ac DOT jp > Tested by: Shigeaki Tagashira, current Discussed with: current Silence from: obrien
2007-06-12 02:16:02 +00:00
mask = ifr->ifr_reqcap ^ ifp->if_capenable;
Add a pure open source nForce Ethernet driver, under BSDL. This driver was ported from OpenBSD by Shigeaki Tagashira <shigeaki@se.hiroshima-u.ac.jp> and posted at http://www.se.hiroshima-u.ac.jp/~shigeaki/software/freebsd-nfe.html It was additionally cleaned up by me. It is still a work-in-progress and thus is purposefully not in GENERIC. And it conflicts with nve(4), so only one should be loaded.
2006-06-26 23:41:07 +00:00
#ifdef DEVICE_POLLING
Bring overhauled nfe(4) into tree. o s/printf/device_printf/g o Nuke OpenBSDism. o Nuke NetBSD/OpenBSD specific DMA sync operations.(we don't have a way to sync a single descriptor within a DMA map.) o Remove recursive mutex. o bus_dma(9) clean up. o 40bit DMA address support. o Add protection for Rx map load failure. o Fix a long standing bug for watchdog timeout. [1] o Add additional protections, missing Tx completion interrupt, losing start Tx command, for watchdog timeout. o Switch to taskqueue(9) API to handle interrupts. o Use our own timer for watchdog instead of if_watchdog/if_timer interface. o Advertise VLAN header length/capability correctly to upper layer. o Remove excessive kernel stack consumption in nfe_encap(). o Handle highly fragmented mbuf chains correctly. o Enable etherenet address reprogramming with ifconfig(8). o Add ALTQ/TSO, MSI/MSIX support. o Increased Rx ring to 256 descriptors from 128. o Align Tx/Rx descriptor ring on sizeof(struct nfe_desc64) boundary. o Remove alignment restrictions on Tx/Rx buffers. o Rewritten jumbo frame support code. o Add support for hardware assistend VLAN tag insertion/stripping. o Add support for Tx/Rx flow control based on patches from Peer Chen. [2] o Add a routine that detects whether ethernet address swap routines is required. [3] o Add a workaround that take MAC/PHY out of power down mode. o Add suspend/resume support. o style(9) and code clean up. Special thanks to Shigeaki Tagashira, the original porter of nfe(4), who submitted lots of patches, performed uncountable number of regression tests and maintained nfe(4) for a long time. Without his enthusiastic help and support I could never have completed this overhauling task. The only weak point of nfe(4) compared to nve(4) is instability of manual half-duplex media selection on certain hardwares(auto sensing media type should work for all cases, though). This was a long standing bug of nfe(4) and I still have no idea why it doesn't work on some hardwares. Obtained from: OpenBSD [1] Submitted by: Peer Chen < pchen at nvidia dot com > [2], [3] Reviewed by: Shigeaki Tagashira < shigeaki AT se DOT hiroshima-u DOT ac DOT jp > Tested by: Shigeaki Tagashira, current Discussed with: current Silence from: obrien
2007-06-12 02:16:02 +00:00
if ((mask & IFCAP_POLLING) != 0) {
if ((ifr->ifr_reqcap & IFCAP_POLLING) != 0) {
Add a pure open source nForce Ethernet driver, under BSDL. This driver was ported from OpenBSD by Shigeaki Tagashira <shigeaki@se.hiroshima-u.ac.jp> and posted at http://www.se.hiroshima-u.ac.jp/~shigeaki/software/freebsd-nfe.html It was additionally cleaned up by me. It is still a work-in-progress and thus is purposefully not in GENERIC. And it conflicts with nve(4), so only one should be loaded.
2006-06-26 23:41:07 +00:00
error = ether_poll_register(nfe_poll, ifp);
if (error)
Bring overhauled nfe(4) into tree. o s/printf/device_printf/g o Nuke OpenBSDism. o Nuke NetBSD/OpenBSD specific DMA sync operations.(we don't have a way to sync a single descriptor within a DMA map.) o Remove recursive mutex. o bus_dma(9) clean up. o 40bit DMA address support. o Add protection for Rx map load failure. o Fix a long standing bug for watchdog timeout. [1] o Add additional protections, missing Tx completion interrupt, losing start Tx command, for watchdog timeout. o Switch to taskqueue(9) API to handle interrupts. o Use our own timer for watchdog instead of if_watchdog/if_timer interface. o Advertise VLAN header length/capability correctly to upper layer. o Remove excessive kernel stack consumption in nfe_encap(). o Handle highly fragmented mbuf chains correctly. o Enable etherenet address reprogramming with ifconfig(8). o Add ALTQ/TSO, MSI/MSIX support. o Increased Rx ring to 256 descriptors from 128. o Align Tx/Rx descriptor ring on sizeof(struct nfe_desc64) boundary. o Remove alignment restrictions on Tx/Rx buffers. o Rewritten jumbo frame support code. o Add support for hardware assistend VLAN tag insertion/stripping. o Add support for Tx/Rx flow control based on patches from Peer Chen. [2] o Add a routine that detects whether ethernet address swap routines is required. [3] o Add a workaround that take MAC/PHY out of power down mode. o Add suspend/resume support. o style(9) and code clean up. Special thanks to Shigeaki Tagashira, the original porter of nfe(4), who submitted lots of patches, performed uncountable number of regression tests and maintained nfe(4) for a long time. Without his enthusiastic help and support I could never have completed this overhauling task. The only weak point of nfe(4) compared to nve(4) is instability of manual half-duplex media selection on certain hardwares(auto sensing media type should work for all cases, though). This was a long standing bug of nfe(4) and I still have no idea why it doesn't work on some hardwares. Obtained from: OpenBSD [1] Submitted by: Peer Chen < pchen at nvidia dot com > [2], [3] Reviewed by: Shigeaki Tagashira < shigeaki AT se DOT hiroshima-u DOT ac DOT jp > Tested by: Shigeaki Tagashira, current Discussed with: current Silence from: obrien
2007-06-12 02:16:02 +00:00
break;
Add a pure open source nForce Ethernet driver, under BSDL. This driver was ported from OpenBSD by Shigeaki Tagashira <shigeaki@se.hiroshima-u.ac.jp> and posted at http://www.se.hiroshima-u.ac.jp/~shigeaki/software/freebsd-nfe.html It was additionally cleaned up by me. It is still a work-in-progress and thus is purposefully not in GENERIC. And it conflicts with nve(4), so only one should be loaded.
2006-06-26 23:41:07 +00:00
NFE_LOCK(sc);
Bring overhauled nfe(4) into tree. o s/printf/device_printf/g o Nuke OpenBSDism. o Nuke NetBSD/OpenBSD specific DMA sync operations.(we don't have a way to sync a single descriptor within a DMA map.) o Remove recursive mutex. o bus_dma(9) clean up. o 40bit DMA address support. o Add protection for Rx map load failure. o Fix a long standing bug for watchdog timeout. [1] o Add additional protections, missing Tx completion interrupt, losing start Tx command, for watchdog timeout. o Switch to taskqueue(9) API to handle interrupts. o Use our own timer for watchdog instead of if_watchdog/if_timer interface. o Advertise VLAN header length/capability correctly to upper layer. o Remove excessive kernel stack consumption in nfe_encap(). o Handle highly fragmented mbuf chains correctly. o Enable etherenet address reprogramming with ifconfig(8). o Add ALTQ/TSO, MSI/MSIX support. o Increased Rx ring to 256 descriptors from 128. o Align Tx/Rx descriptor ring on sizeof(struct nfe_desc64) boundary. o Remove alignment restrictions on Tx/Rx buffers. o Rewritten jumbo frame support code. o Add support for hardware assistend VLAN tag insertion/stripping. o Add support for Tx/Rx flow control based on patches from Peer Chen. [2] o Add a routine that detects whether ethernet address swap routines is required. [3] o Add a workaround that take MAC/PHY out of power down mode. o Add suspend/resume support. o style(9) and code clean up. Special thanks to Shigeaki Tagashira, the original porter of nfe(4), who submitted lots of patches, performed uncountable number of regression tests and maintained nfe(4) for a long time. Without his enthusiastic help and support I could never have completed this overhauling task. The only weak point of nfe(4) compared to nve(4) is instability of manual half-duplex media selection on certain hardwares(auto sensing media type should work for all cases, though). This was a long standing bug of nfe(4) and I still have no idea why it doesn't work on some hardwares. Obtained from: OpenBSD [1] Submitted by: Peer Chen < pchen at nvidia dot com > [2], [3] Reviewed by: Shigeaki Tagashira < shigeaki AT se DOT hiroshima-u DOT ac DOT jp > Tested by: Shigeaki Tagashira, current Discussed with: current Silence from: obrien
2007-06-12 02:16:02 +00:00
nfe_disable_intr(sc);
style(9) cleanup.
2006-10-19 10:01:26 +00:00
ifp->if_capenable |= IFCAP_POLLING;
Add a pure open source nForce Ethernet driver, under BSDL. This driver was ported from OpenBSD by Shigeaki Tagashira <shigeaki@se.hiroshima-u.ac.jp> and posted at http://www.se.hiroshima-u.ac.jp/~shigeaki/software/freebsd-nfe.html It was additionally cleaned up by me. It is still a work-in-progress and thus is purposefully not in GENERIC. And it conflicts with nve(4), so only one should be loaded.
2006-06-26 23:41:07 +00:00
NFE_UNLOCK(sc);
} else {
error = ether_poll_deregister(ifp);
/* Enable interrupt even in error case */
NFE_LOCK(sc);
Bring overhauled nfe(4) into tree. o s/printf/device_printf/g o Nuke OpenBSDism. o Nuke NetBSD/OpenBSD specific DMA sync operations.(we don't have a way to sync a single descriptor within a DMA map.) o Remove recursive mutex. o bus_dma(9) clean up. o 40bit DMA address support. o Add protection for Rx map load failure. o Fix a long standing bug for watchdog timeout. [1] o Add additional protections, missing Tx completion interrupt, losing start Tx command, for watchdog timeout. o Switch to taskqueue(9) API to handle interrupts. o Use our own timer for watchdog instead of if_watchdog/if_timer interface. o Advertise VLAN header length/capability correctly to upper layer. o Remove excessive kernel stack consumption in nfe_encap(). o Handle highly fragmented mbuf chains correctly. o Enable etherenet address reprogramming with ifconfig(8). o Add ALTQ/TSO, MSI/MSIX support. o Increased Rx ring to 256 descriptors from 128. o Align Tx/Rx descriptor ring on sizeof(struct nfe_desc64) boundary. o Remove alignment restrictions on Tx/Rx buffers. o Rewritten jumbo frame support code. o Add support for hardware assistend VLAN tag insertion/stripping. o Add support for Tx/Rx flow control based on patches from Peer Chen. [2] o Add a routine that detects whether ethernet address swap routines is required. [3] o Add a workaround that take MAC/PHY out of power down mode. o Add suspend/resume support. o style(9) and code clean up. Special thanks to Shigeaki Tagashira, the original porter of nfe(4), who submitted lots of patches, performed uncountable number of regression tests and maintained nfe(4) for a long time. Without his enthusiastic help and support I could never have completed this overhauling task. The only weak point of nfe(4) compared to nve(4) is instability of manual half-duplex media selection on certain hardwares(auto sensing media type should work for all cases, though). This was a long standing bug of nfe(4) and I still have no idea why it doesn't work on some hardwares. Obtained from: OpenBSD [1] Submitted by: Peer Chen < pchen at nvidia dot com > [2], [3] Reviewed by: Shigeaki Tagashira < shigeaki AT se DOT hiroshima-u DOT ac DOT jp > Tested by: Shigeaki Tagashira, current Discussed with: current Silence from: obrien
2007-06-12 02:16:02 +00:00
nfe_enable_intr(sc);
Add a pure open source nForce Ethernet driver, under BSDL. This driver was ported from OpenBSD by Shigeaki Tagashira <shigeaki@se.hiroshima-u.ac.jp> and posted at http://www.se.hiroshima-u.ac.jp/~shigeaki/software/freebsd-nfe.html It was additionally cleaned up by me. It is still a work-in-progress and thus is purposefully not in GENERIC. And it conflicts with nve(4), so only one should be loaded.
2006-06-26 23:41:07 +00:00
ifp->if_capenable &= ~IFCAP_POLLING;
NFE_UNLOCK(sc);
}
}
style(9) cleanup.
2006-10-19 10:01:26 +00:00
#endif /* DEVICE_POLLING */
Add basic WOL support for MCP ethernet controllers. It seems the controller does not perform automatic switching from 1000Mbps link to 10/100Mbps link when WOL is activated. Implement establishing 10/100Mps link with auto-negotiation in driver. Link status change handler was modified to remove taskqueue based approach since driver now needs synchronous handling for link establishment. Submitted by: Yamagi Burmeister (lists <> yamagi.org ) (initial version) Tested by: Yamagi Burmeister (lists <> yamagi.org ) MFC after: 1 week
2010-11-11 18:08:50 +00:00
if ((mask & IFCAP_WOL_MAGIC) != 0 &&
(ifp->if_capabilities & IFCAP_WOL_MAGIC) != 0)
ifp->if_capenable ^= IFCAP_WOL_MAGIC;
MCP55 is the only NVIDIA controller that supports VLAN tag insertion/stripping and it also supports TSO over VLAN. Implement TSO over VLAN support for MCP55 controller. While I'm here clean up SIOCSIFCAP ioctl handler. Since nfe(4) sets ifp capabilities based on various hardware flags in device attach, there is no need to check hardware flags again in SIOCSIFCAP ioctl handler. Also fix a bug which toggled both TX and RX checksum offloading even if user requested either TX or RX checksum configuration change. Tested by: Rob Farmer ( rfarmer <> predatorlabs dot net )
2010-11-17 18:09:02 +00:00
if ((mask & IFCAP_TXCSUM) != 0 &&
(ifp->if_capabilities & IFCAP_TXCSUM) != 0) {
ifp->if_capenable ^= IFCAP_TXCSUM;
if ((ifp->if_capenable & IFCAP_TXCSUM) != 0)
Bring overhauled nfe(4) into tree. o s/printf/device_printf/g o Nuke OpenBSDism. o Nuke NetBSD/OpenBSD specific DMA sync operations.(we don't have a way to sync a single descriptor within a DMA map.) o Remove recursive mutex. o bus_dma(9) clean up. o 40bit DMA address support. o Add protection for Rx map load failure. o Fix a long standing bug for watchdog timeout. [1] o Add additional protections, missing Tx completion interrupt, losing start Tx command, for watchdog timeout. o Switch to taskqueue(9) API to handle interrupts. o Use our own timer for watchdog instead of if_watchdog/if_timer interface. o Advertise VLAN header length/capability correctly to upper layer. o Remove excessive kernel stack consumption in nfe_encap(). o Handle highly fragmented mbuf chains correctly. o Enable etherenet address reprogramming with ifconfig(8). o Add ALTQ/TSO, MSI/MSIX support. o Increased Rx ring to 256 descriptors from 128. o Align Tx/Rx descriptor ring on sizeof(struct nfe_desc64) boundary. o Remove alignment restrictions on Tx/Rx buffers. o Rewritten jumbo frame support code. o Add support for hardware assistend VLAN tag insertion/stripping. o Add support for Tx/Rx flow control based on patches from Peer Chen. [2] o Add a routine that detects whether ethernet address swap routines is required. [3] o Add a workaround that take MAC/PHY out of power down mode. o Add suspend/resume support. o style(9) and code clean up. Special thanks to Shigeaki Tagashira, the original porter of nfe(4), who submitted lots of patches, performed uncountable number of regression tests and maintained nfe(4) for a long time. Without his enthusiastic help and support I could never have completed this overhauling task. The only weak point of nfe(4) compared to nve(4) is instability of manual half-duplex media selection on certain hardwares(auto sensing media type should work for all cases, though). This was a long standing bug of nfe(4) and I still have no idea why it doesn't work on some hardwares. Obtained from: OpenBSD [1] Submitted by: Peer Chen < pchen at nvidia dot com > [2], [3] Reviewed by: Shigeaki Tagashira < shigeaki AT se DOT hiroshima-u DOT ac DOT jp > Tested by: Shigeaki Tagashira, current Discussed with: current Silence from: obrien
2007-06-12 02:16:02 +00:00
ifp->if_hwassist |= NFE_CSUM_FEATURES;
Add a pure open source nForce Ethernet driver, under BSDL. This driver was ported from OpenBSD by Shigeaki Tagashira <shigeaki@se.hiroshima-u.ac.jp> and posted at http://www.se.hiroshima-u.ac.jp/~shigeaki/software/freebsd-nfe.html It was additionally cleaned up by me. It is still a work-in-progress and thus is purposefully not in GENERIC. And it conflicts with nve(4), so only one should be loaded.
2006-06-26 23:41:07 +00:00
else
Bring overhauled nfe(4) into tree. o s/printf/device_printf/g o Nuke OpenBSDism. o Nuke NetBSD/OpenBSD specific DMA sync operations.(we don't have a way to sync a single descriptor within a DMA map.) o Remove recursive mutex. o bus_dma(9) clean up. o 40bit DMA address support. o Add protection for Rx map load failure. o Fix a long standing bug for watchdog timeout. [1] o Add additional protections, missing Tx completion interrupt, losing start Tx command, for watchdog timeout. o Switch to taskqueue(9) API to handle interrupts. o Use our own timer for watchdog instead of if_watchdog/if_timer interface. o Advertise VLAN header length/capability correctly to upper layer. o Remove excessive kernel stack consumption in nfe_encap(). o Handle highly fragmented mbuf chains correctly. o Enable etherenet address reprogramming with ifconfig(8). o Add ALTQ/TSO, MSI/MSIX support. o Increased Rx ring to 256 descriptors from 128. o Align Tx/Rx descriptor ring on sizeof(struct nfe_desc64) boundary. o Remove alignment restrictions on Tx/Rx buffers. o Rewritten jumbo frame support code. o Add support for hardware assistend VLAN tag insertion/stripping. o Add support for Tx/Rx flow control based on patches from Peer Chen. [2] o Add a routine that detects whether ethernet address swap routines is required. [3] o Add a workaround that take MAC/PHY out of power down mode. o Add suspend/resume support. o style(9) and code clean up. Special thanks to Shigeaki Tagashira, the original porter of nfe(4), who submitted lots of patches, performed uncountable number of regression tests and maintained nfe(4) for a long time. Without his enthusiastic help and support I could never have completed this overhauling task. The only weak point of nfe(4) compared to nve(4) is instability of manual half-duplex media selection on certain hardwares(auto sensing media type should work for all cases, though). This was a long standing bug of nfe(4) and I still have no idea why it doesn't work on some hardwares. Obtained from: OpenBSD [1] Submitted by: Peer Chen < pchen at nvidia dot com > [2], [3] Reviewed by: Shigeaki Tagashira < shigeaki AT se DOT hiroshima-u DOT ac DOT jp > Tested by: Shigeaki Tagashira, current Discussed with: current Silence from: obrien
2007-06-12 02:16:02 +00:00
ifp->if_hwassist &= ~NFE_CSUM_FEATURES;
MCP55 is the only NVIDIA controller that supports VLAN tag insertion/stripping and it also supports TSO over VLAN. Implement TSO over VLAN support for MCP55 controller. While I'm here clean up SIOCSIFCAP ioctl handler. Since nfe(4) sets ifp capabilities based on various hardware flags in device attach, there is no need to check hardware flags again in SIOCSIFCAP ioctl handler. Also fix a bug which toggled both TX and RX checksum offloading even if user requested either TX or RX checksum configuration change. Tested by: Rob Farmer ( rfarmer <> predatorlabs dot net )
2010-11-17 18:09:02 +00:00
}
if ((mask & IFCAP_RXCSUM) != 0 &&
(ifp->if_capabilities & IFCAP_RXCSUM) != 0) {
ifp->if_capenable ^= IFCAP_RXCSUM;
Bring overhauled nfe(4) into tree. o s/printf/device_printf/g o Nuke OpenBSDism. o Nuke NetBSD/OpenBSD specific DMA sync operations.(we don't have a way to sync a single descriptor within a DMA map.) o Remove recursive mutex. o bus_dma(9) clean up. o 40bit DMA address support. o Add protection for Rx map load failure. o Fix a long standing bug for watchdog timeout. [1] o Add additional protections, missing Tx completion interrupt, losing start Tx command, for watchdog timeout. o Switch to taskqueue(9) API to handle interrupts. o Use our own timer for watchdog instead of if_watchdog/if_timer interface. o Advertise VLAN header length/capability correctly to upper layer. o Remove excessive kernel stack consumption in nfe_encap(). o Handle highly fragmented mbuf chains correctly. o Enable etherenet address reprogramming with ifconfig(8). o Add ALTQ/TSO, MSI/MSIX support. o Increased Rx ring to 256 descriptors from 128. o Align Tx/Rx descriptor ring on sizeof(struct nfe_desc64) boundary. o Remove alignment restrictions on Tx/Rx buffers. o Rewritten jumbo frame support code. o Add support for hardware assistend VLAN tag insertion/stripping. o Add support for Tx/Rx flow control based on patches from Peer Chen. [2] o Add a routine that detects whether ethernet address swap routines is required. [3] o Add a workaround that take MAC/PHY out of power down mode. o Add suspend/resume support. o style(9) and code clean up. Special thanks to Shigeaki Tagashira, the original porter of nfe(4), who submitted lots of patches, performed uncountable number of regression tests and maintained nfe(4) for a long time. Without his enthusiastic help and support I could never have completed this overhauling task. The only weak point of nfe(4) compared to nve(4) is instability of manual half-duplex media selection on certain hardwares(auto sensing media type should work for all cases, though). This was a long standing bug of nfe(4) and I still have no idea why it doesn't work on some hardwares. Obtained from: OpenBSD [1] Submitted by: Peer Chen < pchen at nvidia dot com > [2], [3] Reviewed by: Shigeaki Tagashira < shigeaki AT se DOT hiroshima-u DOT ac DOT jp > Tested by: Shigeaki Tagashira, current Discussed with: current Silence from: obrien
2007-06-12 02:16:02 +00:00
init++;
Add a pure open source nForce Ethernet driver, under BSDL. This driver was ported from OpenBSD by Shigeaki Tagashira <shigeaki@se.hiroshima-u.ac.jp> and posted at http://www.se.hiroshima-u.ac.jp/~shigeaki/software/freebsd-nfe.html It was additionally cleaned up by me. It is still a work-in-progress and thus is purposefully not in GENERIC. And it conflicts with nve(4), so only one should be loaded.
2006-06-26 23:41:07 +00:00
}
MCP55 is the only NVIDIA controller that supports VLAN tag insertion/stripping and it also supports TSO over VLAN. Implement TSO over VLAN support for MCP55 controller. While I'm here clean up SIOCSIFCAP ioctl handler. Since nfe(4) sets ifp capabilities based on various hardware flags in device attach, there is no need to check hardware flags again in SIOCSIFCAP ioctl handler. Also fix a bug which toggled both TX and RX checksum offloading even if user requested either TX or RX checksum configuration change. Tested by: Rob Farmer ( rfarmer <> predatorlabs dot net )
2010-11-17 18:09:02 +00:00
if ((mask & IFCAP_TSO4) != 0 &&
(ifp->if_capabilities & IFCAP_TSO4) != 0) {
ifp->if_capenable ^= IFCAP_TSO4;
if ((IFCAP_TSO4 & ifp->if_capenable) != 0)
ifp->if_hwassist |= CSUM_TSO;
else
ifp->if_hwassist &= ~CSUM_TSO;
}
if ((mask & IFCAP_VLAN_HWTSO) != 0 &&
(ifp->if_capabilities & IFCAP_VLAN_HWTSO) != 0)
ifp->if_capenable ^= IFCAP_VLAN_HWTSO;
if ((mask & IFCAP_VLAN_HWTAGGING) != 0 &&
(ifp->if_capabilities & IFCAP_VLAN_HWTAGGING) != 0) {
Bring overhauled nfe(4) into tree. o s/printf/device_printf/g o Nuke OpenBSDism. o Nuke NetBSD/OpenBSD specific DMA sync operations.(we don't have a way to sync a single descriptor within a DMA map.) o Remove recursive mutex. o bus_dma(9) clean up. o 40bit DMA address support. o Add protection for Rx map load failure. o Fix a long standing bug for watchdog timeout. [1] o Add additional protections, missing Tx completion interrupt, losing start Tx command, for watchdog timeout. o Switch to taskqueue(9) API to handle interrupts. o Use our own timer for watchdog instead of if_watchdog/if_timer interface. o Advertise VLAN header length/capability correctly to upper layer. o Remove excessive kernel stack consumption in nfe_encap(). o Handle highly fragmented mbuf chains correctly. o Enable etherenet address reprogramming with ifconfig(8). o Add ALTQ/TSO, MSI/MSIX support. o Increased Rx ring to 256 descriptors from 128. o Align Tx/Rx descriptor ring on sizeof(struct nfe_desc64) boundary. o Remove alignment restrictions on Tx/Rx buffers. o Rewritten jumbo frame support code. o Add support for hardware assistend VLAN tag insertion/stripping. o Add support for Tx/Rx flow control based on patches from Peer Chen. [2] o Add a routine that detects whether ethernet address swap routines is required. [3] o Add a workaround that take MAC/PHY out of power down mode. o Add suspend/resume support. o style(9) and code clean up. Special thanks to Shigeaki Tagashira, the original porter of nfe(4), who submitted lots of patches, performed uncountable number of regression tests and maintained nfe(4) for a long time. Without his enthusiastic help and support I could never have completed this overhauling task. The only weak point of nfe(4) compared to nve(4) is instability of manual half-duplex media selection on certain hardwares(auto sensing media type should work for all cases, though). This was a long standing bug of nfe(4) and I still have no idea why it doesn't work on some hardwares. Obtained from: OpenBSD [1] Submitted by: Peer Chen < pchen at nvidia dot com > [2], [3] Reviewed by: Shigeaki Tagashira < shigeaki AT se DOT hiroshima-u DOT ac DOT jp > Tested by: Shigeaki Tagashira, current Discussed with: current Silence from: obrien
2007-06-12 02:16:02 +00:00
ifp->if_capenable ^= IFCAP_VLAN_HWTAGGING;
MCP55 is the only NVIDIA controller that supports VLAN tag insertion/stripping and it also supports TSO over VLAN. Implement TSO over VLAN support for MCP55 controller. While I'm here clean up SIOCSIFCAP ioctl handler. Since nfe(4) sets ifp capabilities based on various hardware flags in device attach, there is no need to check hardware flags again in SIOCSIFCAP ioctl handler. Also fix a bug which toggled both TX and RX checksum offloading even if user requested either TX or RX checksum configuration change. Tested by: Rob Farmer ( rfarmer <> predatorlabs dot net )
2010-11-17 18:09:02 +00:00
if ((ifp->if_capenable & IFCAP_VLAN_HWTAGGING) == 0)
ifp->if_capenable &= ~IFCAP_VLAN_HWTSO;
Bring overhauled nfe(4) into tree. o s/printf/device_printf/g o Nuke OpenBSDism. o Nuke NetBSD/OpenBSD specific DMA sync operations.(we don't have a way to sync a single descriptor within a DMA map.) o Remove recursive mutex. o bus_dma(9) clean up. o 40bit DMA address support. o Add protection for Rx map load failure. o Fix a long standing bug for watchdog timeout. [1] o Add additional protections, missing Tx completion interrupt, losing start Tx command, for watchdog timeout. o Switch to taskqueue(9) API to handle interrupts. o Use our own timer for watchdog instead of if_watchdog/if_timer interface. o Advertise VLAN header length/capability correctly to upper layer. o Remove excessive kernel stack consumption in nfe_encap(). o Handle highly fragmented mbuf chains correctly. o Enable etherenet address reprogramming with ifconfig(8). o Add ALTQ/TSO, MSI/MSIX support. o Increased Rx ring to 256 descriptors from 128. o Align Tx/Rx descriptor ring on sizeof(struct nfe_desc64) boundary. o Remove alignment restrictions on Tx/Rx buffers. o Rewritten jumbo frame support code. o Add support for hardware assistend VLAN tag insertion/stripping. o Add support for Tx/Rx flow control based on patches from Peer Chen. [2] o Add a routine that detects whether ethernet address swap routines is required. [3] o Add a workaround that take MAC/PHY out of power down mode. o Add suspend/resume support. o style(9) and code clean up. Special thanks to Shigeaki Tagashira, the original porter of nfe(4), who submitted lots of patches, performed uncountable number of regression tests and maintained nfe(4) for a long time. Without his enthusiastic help and support I could never have completed this overhauling task. The only weak point of nfe(4) compared to nve(4) is instability of manual half-duplex media selection on certain hardwares(auto sensing media type should work for all cases, though). This was a long standing bug of nfe(4) and I still have no idea why it doesn't work on some hardwares. Obtained from: OpenBSD [1] Submitted by: Peer Chen < pchen at nvidia dot com > [2], [3] Reviewed by: Shigeaki Tagashira < shigeaki AT se DOT hiroshima-u DOT ac DOT jp > Tested by: Shigeaki Tagashira, current Discussed with: current Silence from: obrien
2007-06-12 02:16:02 +00:00
init++;
}
/*
* XXX
* It seems that VLAN stripping requires Rx checksum offload.
* Unfortunately FreeBSD has no way to disable only Rx side
* VLAN stripping. So when we know Rx checksum offload is
* disabled turn entire hardware VLAN assist off.
*/
MCP55 is the only NVIDIA controller that supports VLAN tag insertion/stripping and it also supports TSO over VLAN. Implement TSO over VLAN support for MCP55 controller. While I'm here clean up SIOCSIFCAP ioctl handler. Since nfe(4) sets ifp capabilities based on various hardware flags in device attach, there is no need to check hardware flags again in SIOCSIFCAP ioctl handler. Also fix a bug which toggled both TX and RX checksum offloading even if user requested either TX or RX checksum configuration change. Tested by: Rob Farmer ( rfarmer <> predatorlabs dot net )
2010-11-17 18:09:02 +00:00
if ((ifp->if_capenable & IFCAP_RXCSUM) == 0) {
if ((ifp->if_capenable & IFCAP_VLAN_HWTAGGING) != 0)
init++;
ifp->if_capenable &= ~(IFCAP_VLAN_HWTAGGING |
IFCAP_VLAN_HWTSO);
Bring overhauled nfe(4) into tree. o s/printf/device_printf/g o Nuke OpenBSDism. o Nuke NetBSD/OpenBSD specific DMA sync operations.(we don't have a way to sync a single descriptor within a DMA map.) o Remove recursive mutex. o bus_dma(9) clean up. o 40bit DMA address support. o Add protection for Rx map load failure. o Fix a long standing bug for watchdog timeout. [1] o Add additional protections, missing Tx completion interrupt, losing start Tx command, for watchdog timeout. o Switch to taskqueue(9) API to handle interrupts. o Use our own timer for watchdog instead of if_watchdog/if_timer interface. o Advertise VLAN header length/capability correctly to upper layer. o Remove excessive kernel stack consumption in nfe_encap(). o Handle highly fragmented mbuf chains correctly. o Enable etherenet address reprogramming with ifconfig(8). o Add ALTQ/TSO, MSI/MSIX support. o Increased Rx ring to 256 descriptors from 128. o Align Tx/Rx descriptor ring on sizeof(struct nfe_desc64) boundary. o Remove alignment restrictions on Tx/Rx buffers. o Rewritten jumbo frame support code. o Add support for hardware assistend VLAN tag insertion/stripping. o Add support for Tx/Rx flow control based on patches from Peer Chen. [2] o Add a routine that detects whether ethernet address swap routines is required. [3] o Add a workaround that take MAC/PHY out of power down mode. o Add suspend/resume support. o style(9) and code clean up. Special thanks to Shigeaki Tagashira, the original porter of nfe(4), who submitted lots of patches, performed uncountable number of regression tests and maintained nfe(4) for a long time. Without his enthusiastic help and support I could never have completed this overhauling task. The only weak point of nfe(4) compared to nve(4) is instability of manual half-duplex media selection on certain hardwares(auto sensing media type should work for all cases, though). This was a long standing bug of nfe(4) and I still have no idea why it doesn't work on some hardwares. Obtained from: OpenBSD [1] Submitted by: Peer Chen < pchen at nvidia dot com > [2], [3] Reviewed by: Shigeaki Tagashira < shigeaki AT se DOT hiroshima-u DOT ac DOT jp > Tested by: Shigeaki Tagashira, current Discussed with: current Silence from: obrien
2007-06-12 02:16:02 +00:00
}
if (init > 0 && (ifp->if_drv_flags & IFF_DRV_RUNNING) != 0) {
ifp->if_drv_flags &= ~IFF_DRV_RUNNING;
Fix TX/RX checksum offload. Submitted by: Peer Chen at nVIDIA
2006-11-27 04:47:27 +00:00
nfe_init(sc);
Bring overhauled nfe(4) into tree. o s/printf/device_printf/g o Nuke OpenBSDism. o Nuke NetBSD/OpenBSD specific DMA sync operations.(we don't have a way to sync a single descriptor within a DMA map.) o Remove recursive mutex. o bus_dma(9) clean up. o 40bit DMA address support. o Add protection for Rx map load failure. o Fix a long standing bug for watchdog timeout. [1] o Add additional protections, missing Tx completion interrupt, losing start Tx command, for watchdog timeout. o Switch to taskqueue(9) API to handle interrupts. o Use our own timer for watchdog instead of if_watchdog/if_timer interface. o Advertise VLAN header length/capability correctly to upper layer. o Remove excessive kernel stack consumption in nfe_encap(). o Handle highly fragmented mbuf chains correctly. o Enable etherenet address reprogramming with ifconfig(8). o Add ALTQ/TSO, MSI/MSIX support. o Increased Rx ring to 256 descriptors from 128. o Align Tx/Rx descriptor ring on sizeof(struct nfe_desc64) boundary. o Remove alignment restrictions on Tx/Rx buffers. o Rewritten jumbo frame support code. o Add support for hardware assistend VLAN tag insertion/stripping. o Add support for Tx/Rx flow control based on patches from Peer Chen. [2] o Add a routine that detects whether ethernet address swap routines is required. [3] o Add a workaround that take MAC/PHY out of power down mode. o Add suspend/resume support. o style(9) and code clean up. Special thanks to Shigeaki Tagashira, the original porter of nfe(4), who submitted lots of patches, performed uncountable number of regression tests and maintained nfe(4) for a long time. Without his enthusiastic help and support I could never have completed this overhauling task. The only weak point of nfe(4) compared to nve(4) is instability of manual half-duplex media selection on certain hardwares(auto sensing media type should work for all cases, though). This was a long standing bug of nfe(4) and I still have no idea why it doesn't work on some hardwares. Obtained from: OpenBSD [1] Submitted by: Peer Chen < pchen at nvidia dot com > [2], [3] Reviewed by: Shigeaki Tagashira < shigeaki AT se DOT hiroshima-u DOT ac DOT jp > Tested by: Shigeaki Tagashira, current Discussed with: current Silence from: obrien
2007-06-12 02:16:02 +00:00
}
MCP55 is the only NVIDIA controller that supports VLAN tag insertion/stripping and it also supports TSO over VLAN. Implement TSO over VLAN support for MCP55 controller. While I'm here clean up SIOCSIFCAP ioctl handler. Since nfe(4) sets ifp capabilities based on various hardware flags in device attach, there is no need to check hardware flags again in SIOCSIFCAP ioctl handler. Also fix a bug which toggled both TX and RX checksum offloading even if user requested either TX or RX checksum configuration change. Tested by: Rob Farmer ( rfarmer <> predatorlabs dot net )
2010-11-17 18:09:02 +00:00
VLAN_CAPABILITIES(ifp);
Add a pure open source nForce Ethernet driver, under BSDL. This driver was ported from OpenBSD by Shigeaki Tagashira <shigeaki@se.hiroshima-u.ac.jp> and posted at http://www.se.hiroshima-u.ac.jp/~shigeaki/software/freebsd-nfe.html It was additionally cleaned up by me. It is still a work-in-progress and thus is purposefully not in GENERIC. And it conflicts with nve(4), so only one should be loaded.
2006-06-26 23:41:07 +00:00
break;
default:
error = ether_ioctl(ifp, cmd, data);
break;
}
Bring overhauled nfe(4) into tree. o s/printf/device_printf/g o Nuke OpenBSDism. o Nuke NetBSD/OpenBSD specific DMA sync operations.(we don't have a way to sync a single descriptor within a DMA map.) o Remove recursive mutex. o bus_dma(9) clean up. o 40bit DMA address support. o Add protection for Rx map load failure. o Fix a long standing bug for watchdog timeout. [1] o Add additional protections, missing Tx completion interrupt, losing start Tx command, for watchdog timeout. o Switch to taskqueue(9) API to handle interrupts. o Use our own timer for watchdog instead of if_watchdog/if_timer interface. o Advertise VLAN header length/capability correctly to upper layer. o Remove excessive kernel stack consumption in nfe_encap(). o Handle highly fragmented mbuf chains correctly. o Enable etherenet address reprogramming with ifconfig(8). o Add ALTQ/TSO, MSI/MSIX support. o Increased Rx ring to 256 descriptors from 128. o Align Tx/Rx descriptor ring on sizeof(struct nfe_desc64) boundary. o Remove alignment restrictions on Tx/Rx buffers. o Rewritten jumbo frame support code. o Add support for hardware assistend VLAN tag insertion/stripping. o Add support for Tx/Rx flow control based on patches from Peer Chen. [2] o Add a routine that detects whether ethernet address swap routines is required. [3] o Add a workaround that take MAC/PHY out of power down mode. o Add suspend/resume support. o style(9) and code clean up. Special thanks to Shigeaki Tagashira, the original porter of nfe(4), who submitted lots of patches, performed uncountable number of regression tests and maintained nfe(4) for a long time. Without his enthusiastic help and support I could never have completed this overhauling task. The only weak point of nfe(4) compared to nve(4) is instability of manual half-duplex media selection on certain hardwares(auto sensing media type should work for all cases, though). This was a long standing bug of nfe(4) and I still have no idea why it doesn't work on some hardwares. Obtained from: OpenBSD [1] Submitted by: Peer Chen < pchen at nvidia dot com > [2], [3] Reviewed by: Shigeaki Tagashira < shigeaki AT se DOT hiroshima-u DOT ac DOT jp > Tested by: Shigeaki Tagashira, current Discussed with: current Silence from: obrien
2007-06-12 02:16:02 +00:00
return (error);
Add a pure open source nForce Ethernet driver, under BSDL. This driver was ported from OpenBSD by Shigeaki Tagashira <shigeaki@se.hiroshima-u.ac.jp> and posted at http://www.se.hiroshima-u.ac.jp/~shigeaki/software/freebsd-nfe.html It was additionally cleaned up by me. It is still a work-in-progress and thus is purposefully not in GENERIC. And it conflicts with nve(4), so only one should be loaded.
2006-06-26 23:41:07 +00:00
}
Bring overhauled nfe(4) into tree. o s/printf/device_printf/g o Nuke OpenBSDism. o Nuke NetBSD/OpenBSD specific DMA sync operations.(we don't have a way to sync a single descriptor within a DMA map.) o Remove recursive mutex. o bus_dma(9) clean up. o 40bit DMA address support. o Add protection for Rx map load failure. o Fix a long standing bug for watchdog timeout. [1] o Add additional protections, missing Tx completion interrupt, losing start Tx command, for watchdog timeout. o Switch to taskqueue(9) API to handle interrupts. o Use our own timer for watchdog instead of if_watchdog/if_timer interface. o Advertise VLAN header length/capability correctly to upper layer. o Remove excessive kernel stack consumption in nfe_encap(). o Handle highly fragmented mbuf chains correctly. o Enable etherenet address reprogramming with ifconfig(8). o Add ALTQ/TSO, MSI/MSIX support. o Increased Rx ring to 256 descriptors from 128. o Align Tx/Rx descriptor ring on sizeof(struct nfe_desc64) boundary. o Remove alignment restrictions on Tx/Rx buffers. o Rewritten jumbo frame support code. o Add support for hardware assistend VLAN tag insertion/stripping. o Add support for Tx/Rx flow control based on patches from Peer Chen. [2] o Add a routine that detects whether ethernet address swap routines is required. [3] o Add a workaround that take MAC/PHY out of power down mode. o Add suspend/resume support. o style(9) and code clean up. Special thanks to Shigeaki Tagashira, the original porter of nfe(4), who submitted lots of patches, performed uncountable number of regression tests and maintained nfe(4) for a long time. Without his enthusiastic help and support I could never have completed this overhauling task. The only weak point of nfe(4) compared to nve(4) is instability of manual half-duplex media selection on certain hardwares(auto sensing media type should work for all cases, though). This was a long standing bug of nfe(4) and I still have no idea why it doesn't work on some hardwares. Obtained from: OpenBSD [1] Submitted by: Peer Chen < pchen at nvidia dot com > [2], [3] Reviewed by: Shigeaki Tagashira < shigeaki AT se DOT hiroshima-u DOT ac DOT jp > Tested by: Shigeaki Tagashira, current Discussed with: current Silence from: obrien
2007-06-12 02:16:02 +00:00
static int
style(9) cleanup.
2006-10-19 10:01:26 +00:00
nfe_intr(void *arg)
Bring overhauled nfe(4) into tree. o s/printf/device_printf/g o Nuke OpenBSDism. o Nuke NetBSD/OpenBSD specific DMA sync operations.(we don't have a way to sync a single descriptor within a DMA map.) o Remove recursive mutex. o bus_dma(9) clean up. o 40bit DMA address support. o Add protection for Rx map load failure. o Fix a long standing bug for watchdog timeout. [1] o Add additional protections, missing Tx completion interrupt, losing start Tx command, for watchdog timeout. o Switch to taskqueue(9) API to handle interrupts. o Use our own timer for watchdog instead of if_watchdog/if_timer interface. o Advertise VLAN header length/capability correctly to upper layer. o Remove excessive kernel stack consumption in nfe_encap(). o Handle highly fragmented mbuf chains correctly. o Enable etherenet address reprogramming with ifconfig(8). o Add ALTQ/TSO, MSI/MSIX support. o Increased Rx ring to 256 descriptors from 128. o Align Tx/Rx descriptor ring on sizeof(struct nfe_desc64) boundary. o Remove alignment restrictions on Tx/Rx buffers. o Rewritten jumbo frame support code. o Add support for hardware assistend VLAN tag insertion/stripping. o Add support for Tx/Rx flow control based on patches from Peer Chen. [2] o Add a routine that detects whether ethernet address swap routines is required. [3] o Add a workaround that take MAC/PHY out of power down mode. o Add suspend/resume support. o style(9) and code clean up. Special thanks to Shigeaki Tagashira, the original porter of nfe(4), who submitted lots of patches, performed uncountable number of regression tests and maintained nfe(4) for a long time. Without his enthusiastic help and support I could never have completed this overhauling task. The only weak point of nfe(4) compared to nve(4) is instability of manual half-duplex media selection on certain hardwares(auto sensing media type should work for all cases, though). This was a long standing bug of nfe(4) and I still have no idea why it doesn't work on some hardwares. Obtained from: OpenBSD [1] Submitted by: Peer Chen < pchen at nvidia dot com > [2], [3] Reviewed by: Shigeaki Tagashira < shigeaki AT se DOT hiroshima-u DOT ac DOT jp > Tested by: Shigeaki Tagashira, current Discussed with: current Silence from: obrien
2007-06-12 02:16:02 +00:00
{
struct nfe_softc *sc;
uint32_t status;
sc = (struct nfe_softc *)arg;
status = NFE_READ(sc, sc->nfe_irq_status);
if (status == 0 || status == 0xffffffff)
return (FILTER_STRAY);
nfe_disable_intr(sc);
use the private task q thread instead of the shared system thread Reviewed by: yongari MFC after: 1 week
2007-11-16 15:51:47 +00:00
taskqueue_enqueue_fast(sc->nfe_tq, &sc->nfe_int_task);
Bring overhauled nfe(4) into tree. o s/printf/device_printf/g o Nuke OpenBSDism. o Nuke NetBSD/OpenBSD specific DMA sync operations.(we don't have a way to sync a single descriptor within a DMA map.) o Remove recursive mutex. o bus_dma(9) clean up. o 40bit DMA address support. o Add protection for Rx map load failure. o Fix a long standing bug for watchdog timeout. [1] o Add additional protections, missing Tx completion interrupt, losing start Tx command, for watchdog timeout. o Switch to taskqueue(9) API to handle interrupts. o Use our own timer for watchdog instead of if_watchdog/if_timer interface. o Advertise VLAN header length/capability correctly to upper layer. o Remove excessive kernel stack consumption in nfe_encap(). o Handle highly fragmented mbuf chains correctly. o Enable etherenet address reprogramming with ifconfig(8). o Add ALTQ/TSO, MSI/MSIX support. o Increased Rx ring to 256 descriptors from 128. o Align Tx/Rx descriptor ring on sizeof(struct nfe_desc64) boundary. o Remove alignment restrictions on Tx/Rx buffers. o Rewritten jumbo frame support code. o Add support for hardware assistend VLAN tag insertion/stripping. o Add support for Tx/Rx flow control based on patches from Peer Chen. [2] o Add a routine that detects whether ethernet address swap routines is required. [3] o Add a workaround that take MAC/PHY out of power down mode. o Add suspend/resume support. o style(9) and code clean up. Special thanks to Shigeaki Tagashira, the original porter of nfe(4), who submitted lots of patches, performed uncountable number of regression tests and maintained nfe(4) for a long time. Without his enthusiastic help and support I could never have completed this overhauling task. The only weak point of nfe(4) compared to nve(4) is instability of manual half-duplex media selection on certain hardwares(auto sensing media type should work for all cases, though). This was a long standing bug of nfe(4) and I still have no idea why it doesn't work on some hardwares. Obtained from: OpenBSD [1] Submitted by: Peer Chen < pchen at nvidia dot com > [2], [3] Reviewed by: Shigeaki Tagashira < shigeaki AT se DOT hiroshima-u DOT ac DOT jp > Tested by: Shigeaki Tagashira, current Discussed with: current Silence from: obrien
2007-06-12 02:16:02 +00:00
return (FILTER_HANDLED);
}
static void
nfe_int_task(void *arg, int pending)
Add a pure open source nForce Ethernet driver, under BSDL. This driver was ported from OpenBSD by Shigeaki Tagashira <shigeaki@se.hiroshima-u.ac.jp> and posted at http://www.se.hiroshima-u.ac.jp/~shigeaki/software/freebsd-nfe.html It was additionally cleaned up by me. It is still a work-in-progress and thus is purposefully not in GENERIC. And it conflicts with nve(4), so only one should be loaded.
2006-06-26 23:41:07 +00:00
{
struct nfe_softc *sc = arg;
struct ifnet *ifp = sc->nfe_ifp;
Bring overhauled nfe(4) into tree. o s/printf/device_printf/g o Nuke OpenBSDism. o Nuke NetBSD/OpenBSD specific DMA sync operations.(we don't have a way to sync a single descriptor within a DMA map.) o Remove recursive mutex. o bus_dma(9) clean up. o 40bit DMA address support. o Add protection for Rx map load failure. o Fix a long standing bug for watchdog timeout. [1] o Add additional protections, missing Tx completion interrupt, losing start Tx command, for watchdog timeout. o Switch to taskqueue(9) API to handle interrupts. o Use our own timer for watchdog instead of if_watchdog/if_timer interface. o Advertise VLAN header length/capability correctly to upper layer. o Remove excessive kernel stack consumption in nfe_encap(). o Handle highly fragmented mbuf chains correctly. o Enable etherenet address reprogramming with ifconfig(8). o Add ALTQ/TSO, MSI/MSIX support. o Increased Rx ring to 256 descriptors from 128. o Align Tx/Rx descriptor ring on sizeof(struct nfe_desc64) boundary. o Remove alignment restrictions on Tx/Rx buffers. o Rewritten jumbo frame support code. o Add support for hardware assistend VLAN tag insertion/stripping. o Add support for Tx/Rx flow control based on patches from Peer Chen. [2] o Add a routine that detects whether ethernet address swap routines is required. [3] o Add a workaround that take MAC/PHY out of power down mode. o Add suspend/resume support. o style(9) and code clean up. Special thanks to Shigeaki Tagashira, the original porter of nfe(4), who submitted lots of patches, performed uncountable number of regression tests and maintained nfe(4) for a long time. Without his enthusiastic help and support I could never have completed this overhauling task. The only weak point of nfe(4) compared to nve(4) is instability of manual half-duplex media selection on certain hardwares(auto sensing media type should work for all cases, though). This was a long standing bug of nfe(4) and I still have no idea why it doesn't work on some hardwares. Obtained from: OpenBSD [1] Submitted by: Peer Chen < pchen at nvidia dot com > [2], [3] Reviewed by: Shigeaki Tagashira < shigeaki AT se DOT hiroshima-u DOT ac DOT jp > Tested by: Shigeaki Tagashira, current Discussed with: current Silence from: obrien
2007-06-12 02:16:02 +00:00
uint32_t r;
int domore;
Add a pure open source nForce Ethernet driver, under BSDL. This driver was ported from OpenBSD by Shigeaki Tagashira <shigeaki@se.hiroshima-u.ac.jp> and posted at http://www.se.hiroshima-u.ac.jp/~shigeaki/software/freebsd-nfe.html It was additionally cleaned up by me. It is still a work-in-progress and thus is purposefully not in GENERIC. And it conflicts with nve(4), so only one should be loaded.
2006-06-26 23:41:07 +00:00
style(9) cleanup.
2006-10-19 10:01:26 +00:00
NFE_LOCK(sc);
Add a pure open source nForce Ethernet driver, under BSDL. This driver was ported from OpenBSD by Shigeaki Tagashira <shigeaki@se.hiroshima-u.ac.jp> and posted at http://www.se.hiroshima-u.ac.jp/~shigeaki/software/freebsd-nfe.html It was additionally cleaned up by me. It is still a work-in-progress and thus is purposefully not in GENERIC. And it conflicts with nve(4), so only one should be loaded.
2006-06-26 23:41:07 +00:00
Bring overhauled nfe(4) into tree. o s/printf/device_printf/g o Nuke OpenBSDism. o Nuke NetBSD/OpenBSD specific DMA sync operations.(we don't have a way to sync a single descriptor within a DMA map.) o Remove recursive mutex. o bus_dma(9) clean up. o 40bit DMA address support. o Add protection for Rx map load failure. o Fix a long standing bug for watchdog timeout. [1] o Add additional protections, missing Tx completion interrupt, losing start Tx command, for watchdog timeout. o Switch to taskqueue(9) API to handle interrupts. o Use our own timer for watchdog instead of if_watchdog/if_timer interface. o Advertise VLAN header length/capability correctly to upper layer. o Remove excessive kernel stack consumption in nfe_encap(). o Handle highly fragmented mbuf chains correctly. o Enable etherenet address reprogramming with ifconfig(8). o Add ALTQ/TSO, MSI/MSIX support. o Increased Rx ring to 256 descriptors from 128. o Align Tx/Rx descriptor ring on sizeof(struct nfe_desc64) boundary. o Remove alignment restrictions on Tx/Rx buffers. o Rewritten jumbo frame support code. o Add support for hardware assistend VLAN tag insertion/stripping. o Add support for Tx/Rx flow control based on patches from Peer Chen. [2] o Add a routine that detects whether ethernet address swap routines is required. [3] o Add a workaround that take MAC/PHY out of power down mode. o Add suspend/resume support. o style(9) and code clean up. Special thanks to Shigeaki Tagashira, the original porter of nfe(4), who submitted lots of patches, performed uncountable number of regression tests and maintained nfe(4) for a long time. Without his enthusiastic help and support I could never have completed this overhauling task. The only weak point of nfe(4) compared to nve(4) is instability of manual half-duplex media selection on certain hardwares(auto sensing media type should work for all cases, though). This was a long standing bug of nfe(4) and I still have no idea why it doesn't work on some hardwares. Obtained from: OpenBSD [1] Submitted by: Peer Chen < pchen at nvidia dot com > [2], [3] Reviewed by: Shigeaki Tagashira < shigeaki AT se DOT hiroshima-u DOT ac DOT jp > Tested by: Shigeaki Tagashira, current Discussed with: current Silence from: obrien
2007-06-12 02:16:02 +00:00
if ((r = NFE_READ(sc, sc->nfe_irq_status)) == 0) {
nfe_enable_intr(sc);
NFE_UNLOCK(sc);
return; /* not for us */
}
NFE_WRITE(sc, sc->nfe_irq_status, r);
DPRINTFN(sc, 5, "nfe_intr: interrupt register %x\n", r);
Add a pure open source nForce Ethernet driver, under BSDL. This driver was ported from OpenBSD by Shigeaki Tagashira <shigeaki@se.hiroshima-u.ac.jp> and posted at http://www.se.hiroshima-u.ac.jp/~shigeaki/software/freebsd-nfe.html It was additionally cleaned up by me. It is still a work-in-progress and thus is purposefully not in GENERIC. And it conflicts with nve(4), so only one should be loaded.
2006-06-26 23:41:07 +00:00
#ifdef DEVICE_POLLING
if (ifp->if_capenable & IFCAP_POLLING) {
NFE_UNLOCK(sc);
return;
}
#endif
if (r & NFE_IRQ_LINK) {
NFE_READ(sc, NFE_PHY_STATUS);
NFE_WRITE(sc, NFE_PHY_STATUS, 0xf);
Bring overhauled nfe(4) into tree. o s/printf/device_printf/g o Nuke OpenBSDism. o Nuke NetBSD/OpenBSD specific DMA sync operations.(we don't have a way to sync a single descriptor within a DMA map.) o Remove recursive mutex. o bus_dma(9) clean up. o 40bit DMA address support. o Add protection for Rx map load failure. o Fix a long standing bug for watchdog timeout. [1] o Add additional protections, missing Tx completion interrupt, losing start Tx command, for watchdog timeout. o Switch to taskqueue(9) API to handle interrupts. o Use our own timer for watchdog instead of if_watchdog/if_timer interface. o Advertise VLAN header length/capability correctly to upper layer. o Remove excessive kernel stack consumption in nfe_encap(). o Handle highly fragmented mbuf chains correctly. o Enable etherenet address reprogramming with ifconfig(8). o Add ALTQ/TSO, MSI/MSIX support. o Increased Rx ring to 256 descriptors from 128. o Align Tx/Rx descriptor ring on sizeof(struct nfe_desc64) boundary. o Remove alignment restrictions on Tx/Rx buffers. o Rewritten jumbo frame support code. o Add support for hardware assistend VLAN tag insertion/stripping. o Add support for Tx/Rx flow control based on patches from Peer Chen. [2] o Add a routine that detects whether ethernet address swap routines is required. [3] o Add a workaround that take MAC/PHY out of power down mode. o Add suspend/resume support. o style(9) and code clean up. Special thanks to Shigeaki Tagashira, the original porter of nfe(4), who submitted lots of patches, performed uncountable number of regression tests and maintained nfe(4) for a long time. Without his enthusiastic help and support I could never have completed this overhauling task. The only weak point of nfe(4) compared to nve(4) is instability of manual half-duplex media selection on certain hardwares(auto sensing media type should work for all cases, though). This was a long standing bug of nfe(4) and I still have no idea why it doesn't work on some hardwares. Obtained from: OpenBSD [1] Submitted by: Peer Chen < pchen at nvidia dot com > [2], [3] Reviewed by: Shigeaki Tagashira < shigeaki AT se DOT hiroshima-u DOT ac DOT jp > Tested by: Shigeaki Tagashira, current Discussed with: current Silence from: obrien
2007-06-12 02:16:02 +00:00
DPRINTF(sc, "link state changed\n");
Add a pure open source nForce Ethernet driver, under BSDL. This driver was ported from OpenBSD by Shigeaki Tagashira <shigeaki@se.hiroshima-u.ac.jp> and posted at http://www.se.hiroshima-u.ac.jp/~shigeaki/software/freebsd-nfe.html It was additionally cleaned up by me. It is still a work-in-progress and thus is purposefully not in GENERIC. And it conflicts with nve(4), so only one should be loaded.
2006-06-26 23:41:07 +00:00
}
During boot(before setting IP address) PHY can generate link state change interrupt if the link is established with link parter. However interrupt handler didn't acknowledge the interrupt if nfe(4) was not running at the time of interrupt delivery. This caused endless interrupt generation. Fix the bug by acknowledging the interrupt regardless of running state of the driver. PR: kern/116295 Submitted by: Mark Derbyshire (mark At taom dot com) Approved by: re (kensmith)
2007-09-14 05:12:25 +00:00
if ((ifp->if_drv_flags & IFF_DRV_RUNNING) == 0) {
NFE_UNLOCK(sc);
If driver is not running, disable interrupts and do not try to process received frames. Previously it was possible to handle RX interrupts even if controller is not fully initialized. This resulted in non-working driver after system is up and running. Reported by: hselasky Tested by: hselasky
2011-05-31 18:45:15 +00:00
nfe_disable_intr(sc);
During boot(before setting IP address) PHY can generate link state change interrupt if the link is established with link parter. However interrupt handler didn't acknowledge the interrupt if nfe(4) was not running at the time of interrupt delivery. This caused endless interrupt generation. Fix the bug by acknowledging the interrupt regardless of running state of the driver. PR: kern/116295 Submitted by: Mark Derbyshire (mark At taom dot com) Approved by: re (kensmith)
2007-09-14 05:12:25 +00:00
return;
}
Bring overhauled nfe(4) into tree. o s/printf/device_printf/g o Nuke OpenBSDism. o Nuke NetBSD/OpenBSD specific DMA sync operations.(we don't have a way to sync a single descriptor within a DMA map.) o Remove recursive mutex. o bus_dma(9) clean up. o 40bit DMA address support. o Add protection for Rx map load failure. o Fix a long standing bug for watchdog timeout. [1] o Add additional protections, missing Tx completion interrupt, losing start Tx command, for watchdog timeout. o Switch to taskqueue(9) API to handle interrupts. o Use our own timer for watchdog instead of if_watchdog/if_timer interface. o Advertise VLAN header length/capability correctly to upper layer. o Remove excessive kernel stack consumption in nfe_encap(). o Handle highly fragmented mbuf chains correctly. o Enable etherenet address reprogramming with ifconfig(8). o Add ALTQ/TSO, MSI/MSIX support. o Increased Rx ring to 256 descriptors from 128. o Align Tx/Rx descriptor ring on sizeof(struct nfe_desc64) boundary. o Remove alignment restrictions on Tx/Rx buffers. o Rewritten jumbo frame support code. o Add support for hardware assistend VLAN tag insertion/stripping. o Add support for Tx/Rx flow control based on patches from Peer Chen. [2] o Add a routine that detects whether ethernet address swap routines is required. [3] o Add a workaround that take MAC/PHY out of power down mode. o Add suspend/resume support. o style(9) and code clean up. Special thanks to Shigeaki Tagashira, the original porter of nfe(4), who submitted lots of patches, performed uncountable number of regression tests and maintained nfe(4) for a long time. Without his enthusiastic help and support I could never have completed this overhauling task. The only weak point of nfe(4) compared to nve(4) is instability of manual half-duplex media selection on certain hardwares(auto sensing media type should work for all cases, though). This was a long standing bug of nfe(4) and I still have no idea why it doesn't work on some hardwares. Obtained from: OpenBSD [1] Submitted by: Peer Chen < pchen at nvidia dot com > [2], [3] Reviewed by: Shigeaki Tagashira < shigeaki AT se DOT hiroshima-u DOT ac DOT jp > Tested by: Shigeaki Tagashira, current Discussed with: current Silence from: obrien
2007-06-12 02:16:02 +00:00
domore = 0;
/* check Rx ring */
if (sc->nfe_framesize > MCLBYTES - ETHER_HDR_LEN)
When user_frac in the polling subsystem is low it is going to busy the CPU for too long period than necessary. Additively, interfaces are kept polled (in the tick) even if no more packets are available. In order to avoid such situations a new generic mechanism can be implemented in proactive way, keeping track of the time spent on any packet and fragmenting the time for any tick, stopping the processing as soon as possible. In order to implement such mechanism, the polling handler needs to change, returning the number of packets processed. While the intended logic is not part of this patch, the polling KPI is broken by this commit, adding an int return value and the new flag IFCAP_POLLING_NOCOUNT (which will signal that the return value is meaningless for the installed handler and checking should be skipped). Bump __FreeBSD_version in order to signal such situation. Reviewed by: emaste Sponsored by: Sandvine Incorporated
2009-05-30 15:14:44 +00:00
domore = nfe_jrxeof(sc, sc->nfe_process_limit, NULL);
Bring overhauled nfe(4) into tree. o s/printf/device_printf/g o Nuke OpenBSDism. o Nuke NetBSD/OpenBSD specific DMA sync operations.(we don't have a way to sync a single descriptor within a DMA map.) o Remove recursive mutex. o bus_dma(9) clean up. o 40bit DMA address support. o Add protection for Rx map load failure. o Fix a long standing bug for watchdog timeout. [1] o Add additional protections, missing Tx completion interrupt, losing start Tx command, for watchdog timeout. o Switch to taskqueue(9) API to handle interrupts. o Use our own timer for watchdog instead of if_watchdog/if_timer interface. o Advertise VLAN header length/capability correctly to upper layer. o Remove excessive kernel stack consumption in nfe_encap(). o Handle highly fragmented mbuf chains correctly. o Enable etherenet address reprogramming with ifconfig(8). o Add ALTQ/TSO, MSI/MSIX support. o Increased Rx ring to 256 descriptors from 128. o Align Tx/Rx descriptor ring on sizeof(struct nfe_desc64) boundary. o Remove alignment restrictions on Tx/Rx buffers. o Rewritten jumbo frame support code. o Add support for hardware assistend VLAN tag insertion/stripping. o Add support for Tx/Rx flow control based on patches from Peer Chen. [2] o Add a routine that detects whether ethernet address swap routines is required. [3] o Add a workaround that take MAC/PHY out of power down mode. o Add suspend/resume support. o style(9) and code clean up. Special thanks to Shigeaki Tagashira, the original porter of nfe(4), who submitted lots of patches, performed uncountable number of regression tests and maintained nfe(4) for a long time. Without his enthusiastic help and support I could never have completed this overhauling task. The only weak point of nfe(4) compared to nve(4) is instability of manual half-duplex media selection on certain hardwares(auto sensing media type should work for all cases, though). This was a long standing bug of nfe(4) and I still have no idea why it doesn't work on some hardwares. Obtained from: OpenBSD [1] Submitted by: Peer Chen < pchen at nvidia dot com > [2], [3] Reviewed by: Shigeaki Tagashira < shigeaki AT se DOT hiroshima-u DOT ac DOT jp > Tested by: Shigeaki Tagashira, current Discussed with: current Silence from: obrien
2007-06-12 02:16:02 +00:00
else
When user_frac in the polling subsystem is low it is going to busy the CPU for too long period than necessary. Additively, interfaces are kept polled (in the tick) even if no more packets are available. In order to avoid such situations a new generic mechanism can be implemented in proactive way, keeping track of the time spent on any packet and fragmenting the time for any tick, stopping the processing as soon as possible. In order to implement such mechanism, the polling handler needs to change, returning the number of packets processed. While the intended logic is not part of this patch, the polling KPI is broken by this commit, adding an int return value and the new flag IFCAP_POLLING_NOCOUNT (which will signal that the return value is meaningless for the installed handler and checking should be skipped). Bump __FreeBSD_version in order to signal such situation. Reviewed by: emaste Sponsored by: Sandvine Incorporated
2009-05-30 15:14:44 +00:00
domore = nfe_rxeof(sc, sc->nfe_process_limit, NULL);
Bring overhauled nfe(4) into tree. o s/printf/device_printf/g o Nuke OpenBSDism. o Nuke NetBSD/OpenBSD specific DMA sync operations.(we don't have a way to sync a single descriptor within a DMA map.) o Remove recursive mutex. o bus_dma(9) clean up. o 40bit DMA address support. o Add protection for Rx map load failure. o Fix a long standing bug for watchdog timeout. [1] o Add additional protections, missing Tx completion interrupt, losing start Tx command, for watchdog timeout. o Switch to taskqueue(9) API to handle interrupts. o Use our own timer for watchdog instead of if_watchdog/if_timer interface. o Advertise VLAN header length/capability correctly to upper layer. o Remove excessive kernel stack consumption in nfe_encap(). o Handle highly fragmented mbuf chains correctly. o Enable etherenet address reprogramming with ifconfig(8). o Add ALTQ/TSO, MSI/MSIX support. o Increased Rx ring to 256 descriptors from 128. o Align Tx/Rx descriptor ring on sizeof(struct nfe_desc64) boundary. o Remove alignment restrictions on Tx/Rx buffers. o Rewritten jumbo frame support code. o Add support for hardware assistend VLAN tag insertion/stripping. o Add support for Tx/Rx flow control based on patches from Peer Chen. [2] o Add a routine that detects whether ethernet address swap routines is required. [3] o Add a workaround that take MAC/PHY out of power down mode. o Add suspend/resume support. o style(9) and code clean up. Special thanks to Shigeaki Tagashira, the original porter of nfe(4), who submitted lots of patches, performed uncountable number of regression tests and maintained nfe(4) for a long time. Without his enthusiastic help and support I could never have completed this overhauling task. The only weak point of nfe(4) compared to nve(4) is instability of manual half-duplex media selection on certain hardwares(auto sensing media type should work for all cases, though). This was a long standing bug of nfe(4) and I still have no idea why it doesn't work on some hardwares. Obtained from: OpenBSD [1] Submitted by: Peer Chen < pchen at nvidia dot com > [2], [3] Reviewed by: Shigeaki Tagashira < shigeaki AT se DOT hiroshima-u DOT ac DOT jp > Tested by: Shigeaki Tagashira, current Discussed with: current Silence from: obrien
2007-06-12 02:16:02 +00:00
/* check Tx ring */
nfe_txeof(sc);
if (!IFQ_DRV_IS_EMPTY(&ifp->if_snd))
Add a 'locked' variant of the foo_start() routine and call it directly from interrupt handlers and watchdog routines instead of queueing a task to call foo_start(). Reviewed by: yongari MFC after: 1 month
2011-01-03 18:28:30 +00:00
nfe_start_locked(ifp);
Bring overhauled nfe(4) into tree. o s/printf/device_printf/g o Nuke OpenBSDism. o Nuke NetBSD/OpenBSD specific DMA sync operations.(we don't have a way to sync a single descriptor within a DMA map.) o Remove recursive mutex. o bus_dma(9) clean up. o 40bit DMA address support. o Add protection for Rx map load failure. o Fix a long standing bug for watchdog timeout. [1] o Add additional protections, missing Tx completion interrupt, losing start Tx command, for watchdog timeout. o Switch to taskqueue(9) API to handle interrupts. o Use our own timer for watchdog instead of if_watchdog/if_timer interface. o Advertise VLAN header length/capability correctly to upper layer. o Remove excessive kernel stack consumption in nfe_encap(). o Handle highly fragmented mbuf chains correctly. o Enable etherenet address reprogramming with ifconfig(8). o Add ALTQ/TSO, MSI/MSIX support. o Increased Rx ring to 256 descriptors from 128. o Align Tx/Rx descriptor ring on sizeof(struct nfe_desc64) boundary. o Remove alignment restrictions on Tx/Rx buffers. o Rewritten jumbo frame support code. o Add support for hardware assistend VLAN tag insertion/stripping. o Add support for Tx/Rx flow control based on patches from Peer Chen. [2] o Add a routine that detects whether ethernet address swap routines is required. [3] o Add a workaround that take MAC/PHY out of power down mode. o Add suspend/resume support. o style(9) and code clean up. Special thanks to Shigeaki Tagashira, the original porter of nfe(4), who submitted lots of patches, performed uncountable number of regression tests and maintained nfe(4) for a long time. Without his enthusiastic help and support I could never have completed this overhauling task. The only weak point of nfe(4) compared to nve(4) is instability of manual half-duplex media selection on certain hardwares(auto sensing media type should work for all cases, though). This was a long standing bug of nfe(4) and I still have no idea why it doesn't work on some hardwares. Obtained from: OpenBSD [1] Submitted by: Peer Chen < pchen at nvidia dot com > [2], [3] Reviewed by: Shigeaki Tagashira < shigeaki AT se DOT hiroshima-u DOT ac DOT jp > Tested by: Shigeaki Tagashira, current Discussed with: current Silence from: obrien
2007-06-12 02:16:02 +00:00
NFE_UNLOCK(sc);
if (domore || (NFE_READ(sc, sc->nfe_irq_status) != 0)) {
use the private task q thread instead of the shared system thread Reviewed by: yongari MFC after: 1 week
2007-11-16 15:51:47 +00:00
taskqueue_enqueue_fast(sc->nfe_tq, &sc->nfe_int_task);
Bring overhauled nfe(4) into tree. o s/printf/device_printf/g o Nuke OpenBSDism. o Nuke NetBSD/OpenBSD specific DMA sync operations.(we don't have a way to sync a single descriptor within a DMA map.) o Remove recursive mutex. o bus_dma(9) clean up. o 40bit DMA address support. o Add protection for Rx map load failure. o Fix a long standing bug for watchdog timeout. [1] o Add additional protections, missing Tx completion interrupt, losing start Tx command, for watchdog timeout. o Switch to taskqueue(9) API to handle interrupts. o Use our own timer for watchdog instead of if_watchdog/if_timer interface. o Advertise VLAN header length/capability correctly to upper layer. o Remove excessive kernel stack consumption in nfe_encap(). o Handle highly fragmented mbuf chains correctly. o Enable etherenet address reprogramming with ifconfig(8). o Add ALTQ/TSO, MSI/MSIX support. o Increased Rx ring to 256 descriptors from 128. o Align Tx/Rx descriptor ring on sizeof(struct nfe_desc64) boundary. o Remove alignment restrictions on Tx/Rx buffers. o Rewritten jumbo frame support code. o Add support for hardware assistend VLAN tag insertion/stripping. o Add support for Tx/Rx flow control based on patches from Peer Chen. [2] o Add a routine that detects whether ethernet address swap routines is required. [3] o Add a workaround that take MAC/PHY out of power down mode. o Add suspend/resume support. o style(9) and code clean up. Special thanks to Shigeaki Tagashira, the original porter of nfe(4), who submitted lots of patches, performed uncountable number of regression tests and maintained nfe(4) for a long time. Without his enthusiastic help and support I could never have completed this overhauling task. The only weak point of nfe(4) compared to nve(4) is instability of manual half-duplex media selection on certain hardwares(auto sensing media type should work for all cases, though). This was a long standing bug of nfe(4) and I still have no idea why it doesn't work on some hardwares. Obtained from: OpenBSD [1] Submitted by: Peer Chen < pchen at nvidia dot com > [2], [3] Reviewed by: Shigeaki Tagashira < shigeaki AT se DOT hiroshima-u DOT ac DOT jp > Tested by: Shigeaki Tagashira, current Discussed with: current Silence from: obrien
2007-06-12 02:16:02 +00:00
return;
Add a pure open source nForce Ethernet driver, under BSDL. This driver was ported from OpenBSD by Shigeaki Tagashira <shigeaki@se.hiroshima-u.ac.jp> and posted at http://www.se.hiroshima-u.ac.jp/~shigeaki/software/freebsd-nfe.html It was additionally cleaned up by me. It is still a work-in-progress and thus is purposefully not in GENERIC. And it conflicts with nve(4), so only one should be loaded.
2006-06-26 23:41:07 +00:00
}
Bring overhauled nfe(4) into tree. o s/printf/device_printf/g o Nuke OpenBSDism. o Nuke NetBSD/OpenBSD specific DMA sync operations.(we don't have a way to sync a single descriptor within a DMA map.) o Remove recursive mutex. o bus_dma(9) clean up. o 40bit DMA address support. o Add protection for Rx map load failure. o Fix a long standing bug for watchdog timeout. [1] o Add additional protections, missing Tx completion interrupt, losing start Tx command, for watchdog timeout. o Switch to taskqueue(9) API to handle interrupts. o Use our own timer for watchdog instead of if_watchdog/if_timer interface. o Advertise VLAN header length/capability correctly to upper layer. o Remove excessive kernel stack consumption in nfe_encap(). o Handle highly fragmented mbuf chains correctly. o Enable etherenet address reprogramming with ifconfig(8). o Add ALTQ/TSO, MSI/MSIX support. o Increased Rx ring to 256 descriptors from 128. o Align Tx/Rx descriptor ring on sizeof(struct nfe_desc64) boundary. o Remove alignment restrictions on Tx/Rx buffers. o Rewritten jumbo frame support code. o Add support for hardware assistend VLAN tag insertion/stripping. o Add support for Tx/Rx flow control based on patches from Peer Chen. [2] o Add a routine that detects whether ethernet address swap routines is required. [3] o Add a workaround that take MAC/PHY out of power down mode. o Add suspend/resume support. o style(9) and code clean up. Special thanks to Shigeaki Tagashira, the original porter of nfe(4), who submitted lots of patches, performed uncountable number of regression tests and maintained nfe(4) for a long time. Without his enthusiastic help and support I could never have completed this overhauling task. The only weak point of nfe(4) compared to nve(4) is instability of manual half-duplex media selection on certain hardwares(auto sensing media type should work for all cases, though). This was a long standing bug of nfe(4) and I still have no idea why it doesn't work on some hardwares. Obtained from: OpenBSD [1] Submitted by: Peer Chen < pchen at nvidia dot com > [2], [3] Reviewed by: Shigeaki Tagashira < shigeaki AT se DOT hiroshima-u DOT ac DOT jp > Tested by: Shigeaki Tagashira, current Discussed with: current Silence from: obrien
2007-06-12 02:16:02 +00:00
/* Reenable interrupts. */
nfe_enable_intr(sc);
}
Add a pure open source nForce Ethernet driver, under BSDL. This driver was ported from OpenBSD by Shigeaki Tagashira <shigeaki@se.hiroshima-u.ac.jp> and posted at http://www.se.hiroshima-u.ac.jp/~shigeaki/software/freebsd-nfe.html It was additionally cleaned up by me. It is still a work-in-progress and thus is purposefully not in GENERIC. And it conflicts with nve(4), so only one should be loaded.
2006-06-26 23:41:07 +00:00
Bring overhauled nfe(4) into tree. o s/printf/device_printf/g o Nuke OpenBSDism. o Nuke NetBSD/OpenBSD specific DMA sync operations.(we don't have a way to sync a single descriptor within a DMA map.) o Remove recursive mutex. o bus_dma(9) clean up. o 40bit DMA address support. o Add protection for Rx map load failure. o Fix a long standing bug for watchdog timeout. [1] o Add additional protections, missing Tx completion interrupt, losing start Tx command, for watchdog timeout. o Switch to taskqueue(9) API to handle interrupts. o Use our own timer for watchdog instead of if_watchdog/if_timer interface. o Advertise VLAN header length/capability correctly to upper layer. o Remove excessive kernel stack consumption in nfe_encap(). o Handle highly fragmented mbuf chains correctly. o Enable etherenet address reprogramming with ifconfig(8). o Add ALTQ/TSO, MSI/MSIX support. o Increased Rx ring to 256 descriptors from 128. o Align Tx/Rx descriptor ring on sizeof(struct nfe_desc64) boundary. o Remove alignment restrictions on Tx/Rx buffers. o Rewritten jumbo frame support code. o Add support for hardware assistend VLAN tag insertion/stripping. o Add support for Tx/Rx flow control based on patches from Peer Chen. [2] o Add a routine that detects whether ethernet address swap routines is required. [3] o Add a workaround that take MAC/PHY out of power down mode. o Add suspend/resume support. o style(9) and code clean up. Special thanks to Shigeaki Tagashira, the original porter of nfe(4), who submitted lots of patches, performed uncountable number of regression tests and maintained nfe(4) for a long time. Without his enthusiastic help and support I could never have completed this overhauling task. The only weak point of nfe(4) compared to nve(4) is instability of manual half-duplex media selection on certain hardwares(auto sensing media type should work for all cases, though). This was a long standing bug of nfe(4) and I still have no idea why it doesn't work on some hardwares. Obtained from: OpenBSD [1] Submitted by: Peer Chen < pchen at nvidia dot com > [2], [3] Reviewed by: Shigeaki Tagashira < shigeaki AT se DOT hiroshima-u DOT ac DOT jp > Tested by: Shigeaki Tagashira, current Discussed with: current Silence from: obrien
2007-06-12 02:16:02 +00:00
static __inline void
nfe_discard_rxbuf(struct nfe_softc *sc, int idx)
{
struct nfe_desc32 *desc32;
struct nfe_desc64 *desc64;
struct nfe_rx_data *data;
struct mbuf *m;
data = &sc->rxq.data[idx];
m = data->m;
Add a pure open source nForce Ethernet driver, under BSDL. This driver was ported from OpenBSD by Shigeaki Tagashira <shigeaki@se.hiroshima-u.ac.jp> and posted at http://www.se.hiroshima-u.ac.jp/~shigeaki/software/freebsd-nfe.html It was additionally cleaned up by me. It is still a work-in-progress and thus is purposefully not in GENERIC. And it conflicts with nve(4), so only one should be loaded.
2006-06-26 23:41:07 +00:00
Bring overhauled nfe(4) into tree. o s/printf/device_printf/g o Nuke OpenBSDism. o Nuke NetBSD/OpenBSD specific DMA sync operations.(we don't have a way to sync a single descriptor within a DMA map.) o Remove recursive mutex. o bus_dma(9) clean up. o 40bit DMA address support. o Add protection for Rx map load failure. o Fix a long standing bug for watchdog timeout. [1] o Add additional protections, missing Tx completion interrupt, losing start Tx command, for watchdog timeout. o Switch to taskqueue(9) API to handle interrupts. o Use our own timer for watchdog instead of if_watchdog/if_timer interface. o Advertise VLAN header length/capability correctly to upper layer. o Remove excessive kernel stack consumption in nfe_encap(). o Handle highly fragmented mbuf chains correctly. o Enable etherenet address reprogramming with ifconfig(8). o Add ALTQ/TSO, MSI/MSIX support. o Increased Rx ring to 256 descriptors from 128. o Align Tx/Rx descriptor ring on sizeof(struct nfe_desc64) boundary. o Remove alignment restrictions on Tx/Rx buffers. o Rewritten jumbo frame support code. o Add support for hardware assistend VLAN tag insertion/stripping. o Add support for Tx/Rx flow control based on patches from Peer Chen. [2] o Add a routine that detects whether ethernet address swap routines is required. [3] o Add a workaround that take MAC/PHY out of power down mode. o Add suspend/resume support. o style(9) and code clean up. Special thanks to Shigeaki Tagashira, the original porter of nfe(4), who submitted lots of patches, performed uncountable number of regression tests and maintained nfe(4) for a long time. Without his enthusiastic help and support I could never have completed this overhauling task. The only weak point of nfe(4) compared to nve(4) is instability of manual half-duplex media selection on certain hardwares(auto sensing media type should work for all cases, though). This was a long standing bug of nfe(4) and I still have no idea why it doesn't work on some hardwares. Obtained from: OpenBSD [1] Submitted by: Peer Chen < pchen at nvidia dot com > [2], [3] Reviewed by: Shigeaki Tagashira < shigeaki AT se DOT hiroshima-u DOT ac DOT jp > Tested by: Shigeaki Tagashira, current Discussed with: current Silence from: obrien
2007-06-12 02:16:02 +00:00
if (sc->nfe_flags & NFE_40BIT_ADDR) {
desc64 = &sc->rxq.desc64[idx];
/* VLAN packet may have overwritten it. */
desc64->physaddr[0] = htole32(NFE_ADDR_HI(data->paddr));
desc64->physaddr[1] = htole32(NFE_ADDR_LO(data->paddr));
desc64->length = htole16(m->m_len);
desc64->flags = htole16(NFE_RX_READY);
} else {
desc32 = &sc->rxq.desc32[idx];
desc32->length = htole16(m->m_len);
desc32->flags = htole16(NFE_RX_READY);
}
Add a pure open source nForce Ethernet driver, under BSDL. This driver was ported from OpenBSD by Shigeaki Tagashira <shigeaki@se.hiroshima-u.ac.jp> and posted at http://www.se.hiroshima-u.ac.jp/~shigeaki/software/freebsd-nfe.html It was additionally cleaned up by me. It is still a work-in-progress and thus is purposefully not in GENERIC. And it conflicts with nve(4), so only one should be loaded.
2006-06-26 23:41:07 +00:00
}
style(9) cleanup.
2006-10-19 10:01:26 +00:00
Bring overhauled nfe(4) into tree. o s/printf/device_printf/g o Nuke OpenBSDism. o Nuke NetBSD/OpenBSD specific DMA sync operations.(we don't have a way to sync a single descriptor within a DMA map.) o Remove recursive mutex. o bus_dma(9) clean up. o 40bit DMA address support. o Add protection for Rx map load failure. o Fix a long standing bug for watchdog timeout. [1] o Add additional protections, missing Tx completion interrupt, losing start Tx command, for watchdog timeout. o Switch to taskqueue(9) API to handle interrupts. o Use our own timer for watchdog instead of if_watchdog/if_timer interface. o Advertise VLAN header length/capability correctly to upper layer. o Remove excessive kernel stack consumption in nfe_encap(). o Handle highly fragmented mbuf chains correctly. o Enable etherenet address reprogramming with ifconfig(8). o Add ALTQ/TSO, MSI/MSIX support. o Increased Rx ring to 256 descriptors from 128. o Align Tx/Rx descriptor ring on sizeof(struct nfe_desc64) boundary. o Remove alignment restrictions on Tx/Rx buffers. o Rewritten jumbo frame support code. o Add support for hardware assistend VLAN tag insertion/stripping. o Add support for Tx/Rx flow control based on patches from Peer Chen. [2] o Add a routine that detects whether ethernet address swap routines is required. [3] o Add a workaround that take MAC/PHY out of power down mode. o Add suspend/resume support. o style(9) and code clean up. Special thanks to Shigeaki Tagashira, the original porter of nfe(4), who submitted lots of patches, performed uncountable number of regression tests and maintained nfe(4) for a long time. Without his enthusiastic help and support I could never have completed this overhauling task. The only weak point of nfe(4) compared to nve(4) is instability of manual half-duplex media selection on certain hardwares(auto sensing media type should work for all cases, though). This was a long standing bug of nfe(4) and I still have no idea why it doesn't work on some hardwares. Obtained from: OpenBSD [1] Submitted by: Peer Chen < pchen at nvidia dot com > [2], [3] Reviewed by: Shigeaki Tagashira < shigeaki AT se DOT hiroshima-u DOT ac DOT jp > Tested by: Shigeaki Tagashira, current Discussed with: current Silence from: obrien
2007-06-12 02:16:02 +00:00
static __inline void
nfe_discard_jrxbuf(struct nfe_softc *sc, int idx)
Add a pure open source nForce Ethernet driver, under BSDL. This driver was ported from OpenBSD by Shigeaki Tagashira <shigeaki@se.hiroshima-u.ac.jp> and posted at http://www.se.hiroshima-u.ac.jp/~shigeaki/software/freebsd-nfe.html It was additionally cleaned up by me. It is still a work-in-progress and thus is purposefully not in GENERIC. And it conflicts with nve(4), so only one should be loaded.
2006-06-26 23:41:07 +00:00
{
Bring overhauled nfe(4) into tree. o s/printf/device_printf/g o Nuke OpenBSDism. o Nuke NetBSD/OpenBSD specific DMA sync operations.(we don't have a way to sync a single descriptor within a DMA map.) o Remove recursive mutex. o bus_dma(9) clean up. o 40bit DMA address support. o Add protection for Rx map load failure. o Fix a long standing bug for watchdog timeout. [1] o Add additional protections, missing Tx completion interrupt, losing start Tx command, for watchdog timeout. o Switch to taskqueue(9) API to handle interrupts. o Use our own timer for watchdog instead of if_watchdog/if_timer interface. o Advertise VLAN header length/capability correctly to upper layer. o Remove excessive kernel stack consumption in nfe_encap(). o Handle highly fragmented mbuf chains correctly. o Enable etherenet address reprogramming with ifconfig(8). o Add ALTQ/TSO, MSI/MSIX support. o Increased Rx ring to 256 descriptors from 128. o Align Tx/Rx descriptor ring on sizeof(struct nfe_desc64) boundary. o Remove alignment restrictions on Tx/Rx buffers. o Rewritten jumbo frame support code. o Add support for hardware assistend VLAN tag insertion/stripping. o Add support for Tx/Rx flow control based on patches from Peer Chen. [2] o Add a routine that detects whether ethernet address swap routines is required. [3] o Add a workaround that take MAC/PHY out of power down mode. o Add suspend/resume support. o style(9) and code clean up. Special thanks to Shigeaki Tagashira, the original porter of nfe(4), who submitted lots of patches, performed uncountable number of regression tests and maintained nfe(4) for a long time. Without his enthusiastic help and support I could never have completed this overhauling task. The only weak point of nfe(4) compared to nve(4) is instability of manual half-duplex media selection on certain hardwares(auto sensing media type should work for all cases, though). This was a long standing bug of nfe(4) and I still have no idea why it doesn't work on some hardwares. Obtained from: OpenBSD [1] Submitted by: Peer Chen < pchen at nvidia dot com > [2], [3] Reviewed by: Shigeaki Tagashira < shigeaki AT se DOT hiroshima-u DOT ac DOT jp > Tested by: Shigeaki Tagashira, current Discussed with: current Silence from: obrien
2007-06-12 02:16:02 +00:00
struct nfe_desc32 *desc32;
struct nfe_desc64 *desc64;
struct nfe_rx_data *data;
struct mbuf *m;
data = &sc->jrxq.jdata[idx];
m = data->m;
style(9) cleanup.
2006-10-19 10:01:26 +00:00
Bring overhauled nfe(4) into tree. o s/printf/device_printf/g o Nuke OpenBSDism. o Nuke NetBSD/OpenBSD specific DMA sync operations.(we don't have a way to sync a single descriptor within a DMA map.) o Remove recursive mutex. o bus_dma(9) clean up. o 40bit DMA address support. o Add protection for Rx map load failure. o Fix a long standing bug for watchdog timeout. [1] o Add additional protections, missing Tx completion interrupt, losing start Tx command, for watchdog timeout. o Switch to taskqueue(9) API to handle interrupts. o Use our own timer for watchdog instead of if_watchdog/if_timer interface. o Advertise VLAN header length/capability correctly to upper layer. o Remove excessive kernel stack consumption in nfe_encap(). o Handle highly fragmented mbuf chains correctly. o Enable etherenet address reprogramming with ifconfig(8). o Add ALTQ/TSO, MSI/MSIX support. o Increased Rx ring to 256 descriptors from 128. o Align Tx/Rx descriptor ring on sizeof(struct nfe_desc64) boundary. o Remove alignment restrictions on Tx/Rx buffers. o Rewritten jumbo frame support code. o Add support for hardware assistend VLAN tag insertion/stripping. o Add support for Tx/Rx flow control based on patches from Peer Chen. [2] o Add a routine that detects whether ethernet address swap routines is required. [3] o Add a workaround that take MAC/PHY out of power down mode. o Add suspend/resume support. o style(9) and code clean up. Special thanks to Shigeaki Tagashira, the original porter of nfe(4), who submitted lots of patches, performed uncountable number of regression tests and maintained nfe(4) for a long time. Without his enthusiastic help and support I could never have completed this overhauling task. The only weak point of nfe(4) compared to nve(4) is instability of manual half-duplex media selection on certain hardwares(auto sensing media type should work for all cases, though). This was a long standing bug of nfe(4) and I still have no idea why it doesn't work on some hardwares. Obtained from: OpenBSD [1] Submitted by: Peer Chen < pchen at nvidia dot com > [2], [3] Reviewed by: Shigeaki Tagashira < shigeaki AT se DOT hiroshima-u DOT ac DOT jp > Tested by: Shigeaki Tagashira, current Discussed with: current Silence from: obrien
2007-06-12 02:16:02 +00:00
if (sc->nfe_flags & NFE_40BIT_ADDR) {
desc64 = &sc->jrxq.jdesc64[idx];
/* VLAN packet may have overwritten it. */
desc64->physaddr[0] = htole32(NFE_ADDR_HI(data->paddr));
desc64->physaddr[1] = htole32(NFE_ADDR_LO(data->paddr));
desc64->length = htole16(m->m_len);
desc64->flags = htole16(NFE_RX_READY);
} else {
desc32 = &sc->jrxq.jdesc32[idx];
desc32->length = htole16(m->m_len);
desc32->flags = htole16(NFE_RX_READY);
}
Add a pure open source nForce Ethernet driver, under BSDL. This driver was ported from OpenBSD by Shigeaki Tagashira <shigeaki@se.hiroshima-u.ac.jp> and posted at http://www.se.hiroshima-u.ac.jp/~shigeaki/software/freebsd-nfe.html It was additionally cleaned up by me. It is still a work-in-progress and thus is purposefully not in GENERIC. And it conflicts with nve(4), so only one should be loaded.
2006-06-26 23:41:07 +00:00
}
style(9) cleanup.
2006-10-19 10:01:26 +00:00
Bring overhauled nfe(4) into tree. o s/printf/device_printf/g o Nuke OpenBSDism. o Nuke NetBSD/OpenBSD specific DMA sync operations.(we don't have a way to sync a single descriptor within a DMA map.) o Remove recursive mutex. o bus_dma(9) clean up. o 40bit DMA address support. o Add protection for Rx map load failure. o Fix a long standing bug for watchdog timeout. [1] o Add additional protections, missing Tx completion interrupt, losing start Tx command, for watchdog timeout. o Switch to taskqueue(9) API to handle interrupts. o Use our own timer for watchdog instead of if_watchdog/if_timer interface. o Advertise VLAN header length/capability correctly to upper layer. o Remove excessive kernel stack consumption in nfe_encap(). o Handle highly fragmented mbuf chains correctly. o Enable etherenet address reprogramming with ifconfig(8). o Add ALTQ/TSO, MSI/MSIX support. o Increased Rx ring to 256 descriptors from 128. o Align Tx/Rx descriptor ring on sizeof(struct nfe_desc64) boundary. o Remove alignment restrictions on Tx/Rx buffers. o Rewritten jumbo frame support code. o Add support for hardware assistend VLAN tag insertion/stripping. o Add support for Tx/Rx flow control based on patches from Peer Chen. [2] o Add a routine that detects whether ethernet address swap routines is required. [3] o Add a workaround that take MAC/PHY out of power down mode. o Add suspend/resume support. o style(9) and code clean up. Special thanks to Shigeaki Tagashira, the original porter of nfe(4), who submitted lots of patches, performed uncountable number of regression tests and maintained nfe(4) for a long time. Without his enthusiastic help and support I could never have completed this overhauling task. The only weak point of nfe(4) compared to nve(4) is instability of manual half-duplex media selection on certain hardwares(auto sensing media type should work for all cases, though). This was a long standing bug of nfe(4) and I still have no idea why it doesn't work on some hardwares. Obtained from: OpenBSD [1] Submitted by: Peer Chen < pchen at nvidia dot com > [2], [3] Reviewed by: Shigeaki Tagashira < shigeaki AT se DOT hiroshima-u DOT ac DOT jp > Tested by: Shigeaki Tagashira, current Discussed with: current Silence from: obrien
2007-06-12 02:16:02 +00:00
static int
nfe_newbuf(struct nfe_softc *sc, int idx)
Add a pure open source nForce Ethernet driver, under BSDL. This driver was ported from OpenBSD by Shigeaki Tagashira <shigeaki@se.hiroshima-u.ac.jp> and posted at http://www.se.hiroshima-u.ac.jp/~shigeaki/software/freebsd-nfe.html It was additionally cleaned up by me. It is still a work-in-progress and thus is purposefully not in GENERIC. And it conflicts with nve(4), so only one should be loaded.
2006-06-26 23:41:07 +00:00
{
Bring overhauled nfe(4) into tree. o s/printf/device_printf/g o Nuke OpenBSDism. o Nuke NetBSD/OpenBSD specific DMA sync operations.(we don't have a way to sync a single descriptor within a DMA map.) o Remove recursive mutex. o bus_dma(9) clean up. o 40bit DMA address support. o Add protection for Rx map load failure. o Fix a long standing bug for watchdog timeout. [1] o Add additional protections, missing Tx completion interrupt, losing start Tx command, for watchdog timeout. o Switch to taskqueue(9) API to handle interrupts. o Use our own timer for watchdog instead of if_watchdog/if_timer interface. o Advertise VLAN header length/capability correctly to upper layer. o Remove excessive kernel stack consumption in nfe_encap(). o Handle highly fragmented mbuf chains correctly. o Enable etherenet address reprogramming with ifconfig(8). o Add ALTQ/TSO, MSI/MSIX support. o Increased Rx ring to 256 descriptors from 128. o Align Tx/Rx descriptor ring on sizeof(struct nfe_desc64) boundary. o Remove alignment restrictions on Tx/Rx buffers. o Rewritten jumbo frame support code. o Add support for hardware assistend VLAN tag insertion/stripping. o Add support for Tx/Rx flow control based on patches from Peer Chen. [2] o Add a routine that detects whether ethernet address swap routines is required. [3] o Add a workaround that take MAC/PHY out of power down mode. o Add suspend/resume support. o style(9) and code clean up. Special thanks to Shigeaki Tagashira, the original porter of nfe(4), who submitted lots of patches, performed uncountable number of regression tests and maintained nfe(4) for a long time. Without his enthusiastic help and support I could never have completed this overhauling task. The only weak point of nfe(4) compared to nve(4) is instability of manual half-duplex media selection on certain hardwares(auto sensing media type should work for all cases, though). This was a long standing bug of nfe(4) and I still have no idea why it doesn't work on some hardwares. Obtained from: OpenBSD [1] Submitted by: Peer Chen < pchen at nvidia dot com > [2], [3] Reviewed by: Shigeaki Tagashira < shigeaki AT se DOT hiroshima-u DOT ac DOT jp > Tested by: Shigeaki Tagashira, current Discussed with: current Silence from: obrien
2007-06-12 02:16:02 +00:00
struct nfe_rx_data *data;
struct nfe_desc32 *desc32;
struct nfe_desc64 *desc64;
struct mbuf *m;
bus_dma_segment_t segs[1];
bus_dmamap_t map;
int nsegs;
style(9) cleanup.
2006-10-19 10:01:26 +00:00
Bring overhauled nfe(4) into tree. o s/printf/device_printf/g o Nuke OpenBSDism. o Nuke NetBSD/OpenBSD specific DMA sync operations.(we don't have a way to sync a single descriptor within a DMA map.) o Remove recursive mutex. o bus_dma(9) clean up. o 40bit DMA address support. o Add protection for Rx map load failure. o Fix a long standing bug for watchdog timeout. [1] o Add additional protections, missing Tx completion interrupt, losing start Tx command, for watchdog timeout. o Switch to taskqueue(9) API to handle interrupts. o Use our own timer for watchdog instead of if_watchdog/if_timer interface. o Advertise VLAN header length/capability correctly to upper layer. o Remove excessive kernel stack consumption in nfe_encap(). o Handle highly fragmented mbuf chains correctly. o Enable etherenet address reprogramming with ifconfig(8). o Add ALTQ/TSO, MSI/MSIX support. o Increased Rx ring to 256 descriptors from 128. o Align Tx/Rx descriptor ring on sizeof(struct nfe_desc64) boundary. o Remove alignment restrictions on Tx/Rx buffers. o Rewritten jumbo frame support code. o Add support for hardware assistend VLAN tag insertion/stripping. o Add support for Tx/Rx flow control based on patches from Peer Chen. [2] o Add a routine that detects whether ethernet address swap routines is required. [3] o Add a workaround that take MAC/PHY out of power down mode. o Add suspend/resume support. o style(9) and code clean up. Special thanks to Shigeaki Tagashira, the original porter of nfe(4), who submitted lots of patches, performed uncountable number of regression tests and maintained nfe(4) for a long time. Without his enthusiastic help and support I could never have completed this overhauling task. The only weak point of nfe(4) compared to nve(4) is instability of manual half-duplex media selection on certain hardwares(auto sensing media type should work for all cases, though). This was a long standing bug of nfe(4) and I still have no idea why it doesn't work on some hardwares. Obtained from: OpenBSD [1] Submitted by: Peer Chen < pchen at nvidia dot com > [2], [3] Reviewed by: Shigeaki Tagashira < shigeaki AT se DOT hiroshima-u DOT ac DOT jp > Tested by: Shigeaki Tagashira, current Discussed with: current Silence from: obrien
2007-06-12 02:16:02 +00:00
m = m_getcl(M_DONTWAIT, MT_DATA, M_PKTHDR);
if (m == NULL)
return (ENOBUFS);
m->m_len = m->m_pkthdr.len = MCLBYTES;
m_adj(m, ETHER_ALIGN);
if (bus_dmamap_load_mbuf_sg(sc->rxq.rx_data_tag, sc->rxq.rx_spare_map,
m, segs, &nsegs, BUS_DMA_NOWAIT) != 0) {
m_freem(m);
return (ENOBUFS);
}
KASSERT(nsegs == 1, ("%s: %d segments returned!", __func__, nsegs));
data = &sc->rxq.data[idx];
if (data->m != NULL) {
bus_dmamap_sync(sc->rxq.rx_data_tag, data->rx_data_map,
BUS_DMASYNC_POSTREAD);
bus_dmamap_unload(sc->rxq.rx_data_tag, data->rx_data_map);
}
map = data->rx_data_map;
data->rx_data_map = sc->rxq.rx_spare_map;
sc->rxq.rx_spare_map = map;
bus_dmamap_sync(sc->rxq.rx_data_tag, data->rx_data_map,
BUS_DMASYNC_PREREAD);
data->paddr = segs[0].ds_addr;
data->m = m;
/* update mapping address in h/w descriptor */
if (sc->nfe_flags & NFE_40BIT_ADDR) {
desc64 = &sc->rxq.desc64[idx];
desc64->physaddr[0] = htole32(NFE_ADDR_HI(segs[0].ds_addr));
desc64->physaddr[1] = htole32(NFE_ADDR_LO(segs[0].ds_addr));
desc64->length = htole16(segs[0].ds_len);
desc64->flags = htole16(NFE_RX_READY);
} else {
desc32 = &sc->rxq.desc32[idx];
desc32->physaddr = htole32(NFE_ADDR_LO(segs[0].ds_addr));
desc32->length = htole16(segs[0].ds_len);
desc32->flags = htole16(NFE_RX_READY);
}
return (0);
Add a pure open source nForce Ethernet driver, under BSDL. This driver was ported from OpenBSD by Shigeaki Tagashira <shigeaki@se.hiroshima-u.ac.jp> and posted at http://www.se.hiroshima-u.ac.jp/~shigeaki/software/freebsd-nfe.html It was additionally cleaned up by me. It is still a work-in-progress and thus is purposefully not in GENERIC. And it conflicts with nve(4), so only one should be loaded.
2006-06-26 23:41:07 +00:00
}
style(9) cleanup.
2006-10-19 10:01:26 +00:00
Bring overhauled nfe(4) into tree. o s/printf/device_printf/g o Nuke OpenBSDism. o Nuke NetBSD/OpenBSD specific DMA sync operations.(we don't have a way to sync a single descriptor within a DMA map.) o Remove recursive mutex. o bus_dma(9) clean up. o 40bit DMA address support. o Add protection for Rx map load failure. o Fix a long standing bug for watchdog timeout. [1] o Add additional protections, missing Tx completion interrupt, losing start Tx command, for watchdog timeout. o Switch to taskqueue(9) API to handle interrupts. o Use our own timer for watchdog instead of if_watchdog/if_timer interface. o Advertise VLAN header length/capability correctly to upper layer. o Remove excessive kernel stack consumption in nfe_encap(). o Handle highly fragmented mbuf chains correctly. o Enable etherenet address reprogramming with ifconfig(8). o Add ALTQ/TSO, MSI/MSIX support. o Increased Rx ring to 256 descriptors from 128. o Align Tx/Rx descriptor ring on sizeof(struct nfe_desc64) boundary. o Remove alignment restrictions on Tx/Rx buffers. o Rewritten jumbo frame support code. o Add support for hardware assistend VLAN tag insertion/stripping. o Add support for Tx/Rx flow control based on patches from Peer Chen. [2] o Add a routine that detects whether ethernet address swap routines is required. [3] o Add a workaround that take MAC/PHY out of power down mode. o Add suspend/resume support. o style(9) and code clean up. Special thanks to Shigeaki Tagashira, the original porter of nfe(4), who submitted lots of patches, performed uncountable number of regression tests and maintained nfe(4) for a long time. Without his enthusiastic help and support I could never have completed this overhauling task. The only weak point of nfe(4) compared to nve(4) is instability of manual half-duplex media selection on certain hardwares(auto sensing media type should work for all cases, though). This was a long standing bug of nfe(4) and I still have no idea why it doesn't work on some hardwares. Obtained from: OpenBSD [1] Submitted by: Peer Chen < pchen at nvidia dot com > [2], [3] Reviewed by: Shigeaki Tagashira < shigeaki AT se DOT hiroshima-u DOT ac DOT jp > Tested by: Shigeaki Tagashira, current Discussed with: current Silence from: obrien
2007-06-12 02:16:02 +00:00
static int
nfe_jnewbuf(struct nfe_softc *sc, int idx)
Add a pure open source nForce Ethernet driver, under BSDL. This driver was ported from OpenBSD by Shigeaki Tagashira <shigeaki@se.hiroshima-u.ac.jp> and posted at http://www.se.hiroshima-u.ac.jp/~shigeaki/software/freebsd-nfe.html It was additionally cleaned up by me. It is still a work-in-progress and thus is purposefully not in GENERIC. And it conflicts with nve(4), so only one should be loaded.
2006-06-26 23:41:07 +00:00
{
Bring overhauled nfe(4) into tree. o s/printf/device_printf/g o Nuke OpenBSDism. o Nuke NetBSD/OpenBSD specific DMA sync operations.(we don't have a way to sync a single descriptor within a DMA map.) o Remove recursive mutex. o bus_dma(9) clean up. o 40bit DMA address support. o Add protection for Rx map load failure. o Fix a long standing bug for watchdog timeout. [1] o Add additional protections, missing Tx completion interrupt, losing start Tx command, for watchdog timeout. o Switch to taskqueue(9) API to handle interrupts. o Use our own timer for watchdog instead of if_watchdog/if_timer interface. o Advertise VLAN header length/capability correctly to upper layer. o Remove excessive kernel stack consumption in nfe_encap(). o Handle highly fragmented mbuf chains correctly. o Enable etherenet address reprogramming with ifconfig(8). o Add ALTQ/TSO, MSI/MSIX support. o Increased Rx ring to 256 descriptors from 128. o Align Tx/Rx descriptor ring on sizeof(struct nfe_desc64) boundary. o Remove alignment restrictions on Tx/Rx buffers. o Rewritten jumbo frame support code. o Add support for hardware assistend VLAN tag insertion/stripping. o Add support for Tx/Rx flow control based on patches from Peer Chen. [2] o Add a routine that detects whether ethernet address swap routines is required. [3] o Add a workaround that take MAC/PHY out of power down mode. o Add suspend/resume support. o style(9) and code clean up. Special thanks to Shigeaki Tagashira, the original porter of nfe(4), who submitted lots of patches, performed uncountable number of regression tests and maintained nfe(4) for a long time. Without his enthusiastic help and support I could never have completed this overhauling task. The only weak point of nfe(4) compared to nve(4) is instability of manual half-duplex media selection on certain hardwares(auto sensing media type should work for all cases, though). This was a long standing bug of nfe(4) and I still have no idea why it doesn't work on some hardwares. Obtained from: OpenBSD [1] Submitted by: Peer Chen < pchen at nvidia dot com > [2], [3] Reviewed by: Shigeaki Tagashira < shigeaki AT se DOT hiroshima-u DOT ac DOT jp > Tested by: Shigeaki Tagashira, current Discussed with: current Silence from: obrien
2007-06-12 02:16:02 +00:00
struct nfe_rx_data *data;
struct nfe_desc32 *desc32;
struct nfe_desc64 *desc64;
struct mbuf *m;
bus_dma_segment_t segs[1];
bus_dmamap_t map;
int nsegs;
Nuke local jumbo allocator and switch to use of UMA backed page allocator for jumbo frame. Also remove unneeded jlist lock which is no longer required to protect jumbo buffers. With these changes jumbo frame performance of nfe(4) was slightly increased and users should not encounter jumbo buffer allocation failure anymore.
2008-03-06 01:47:53 +00:00
m = m_getjcl(M_DONTWAIT, MT_DATA, M_PKTHDR, MJUM9BYTES);
Bring overhauled nfe(4) into tree. o s/printf/device_printf/g o Nuke OpenBSDism. o Nuke NetBSD/OpenBSD specific DMA sync operations.(we don't have a way to sync a single descriptor within a DMA map.) o Remove recursive mutex. o bus_dma(9) clean up. o 40bit DMA address support. o Add protection for Rx map load failure. o Fix a long standing bug for watchdog timeout. [1] o Add additional protections, missing Tx completion interrupt, losing start Tx command, for watchdog timeout. o Switch to taskqueue(9) API to handle interrupts. o Use our own timer for watchdog instead of if_watchdog/if_timer interface. o Advertise VLAN header length/capability correctly to upper layer. o Remove excessive kernel stack consumption in nfe_encap(). o Handle highly fragmented mbuf chains correctly. o Enable etherenet address reprogramming with ifconfig(8). o Add ALTQ/TSO, MSI/MSIX support. o Increased Rx ring to 256 descriptors from 128. o Align Tx/Rx descriptor ring on sizeof(struct nfe_desc64) boundary. o Remove alignment restrictions on Tx/Rx buffers. o Rewritten jumbo frame support code. o Add support for hardware assistend VLAN tag insertion/stripping. o Add support for Tx/Rx flow control based on patches from Peer Chen. [2] o Add a routine that detects whether ethernet address swap routines is required. [3] o Add a workaround that take MAC/PHY out of power down mode. o Add suspend/resume support. o style(9) and code clean up. Special thanks to Shigeaki Tagashira, the original porter of nfe(4), who submitted lots of patches, performed uncountable number of regression tests and maintained nfe(4) for a long time. Without his enthusiastic help and support I could never have completed this overhauling task. The only weak point of nfe(4) compared to nve(4) is instability of manual half-duplex media selection on certain hardwares(auto sensing media type should work for all cases, though). This was a long standing bug of nfe(4) and I still have no idea why it doesn't work on some hardwares. Obtained from: OpenBSD [1] Submitted by: Peer Chen < pchen at nvidia dot com > [2], [3] Reviewed by: Shigeaki Tagashira < shigeaki AT se DOT hiroshima-u DOT ac DOT jp > Tested by: Shigeaki Tagashira, current Discussed with: current Silence from: obrien
2007-06-12 02:16:02 +00:00
if (m == NULL)
return (ENOBUFS);
if ((m->m_flags & M_EXT) == 0) {
m_freem(m);
return (ENOBUFS);
}
Nuke local jumbo allocator and switch to use of UMA backed page allocator for jumbo frame. Also remove unneeded jlist lock which is no longer required to protect jumbo buffers. With these changes jumbo frame performance of nfe(4) was slightly increased and users should not encounter jumbo buffer allocation failure anymore.
2008-03-06 01:47:53 +00:00
m->m_pkthdr.len = m->m_len = MJUM9BYTES;
Bring overhauled nfe(4) into tree. o s/printf/device_printf/g o Nuke OpenBSDism. o Nuke NetBSD/OpenBSD specific DMA sync operations.(we don't have a way to sync a single descriptor within a DMA map.) o Remove recursive mutex. o bus_dma(9) clean up. o 40bit DMA address support. o Add protection for Rx map load failure. o Fix a long standing bug for watchdog timeout. [1] o Add additional protections, missing Tx completion interrupt, losing start Tx command, for watchdog timeout. o Switch to taskqueue(9) API to handle interrupts. o Use our own timer for watchdog instead of if_watchdog/if_timer interface. o Advertise VLAN header length/capability correctly to upper layer. o Remove excessive kernel stack consumption in nfe_encap(). o Handle highly fragmented mbuf chains correctly. o Enable etherenet address reprogramming with ifconfig(8). o Add ALTQ/TSO, MSI/MSIX support. o Increased Rx ring to 256 descriptors from 128. o Align Tx/Rx descriptor ring on sizeof(struct nfe_desc64) boundary. o Remove alignment restrictions on Tx/Rx buffers. o Rewritten jumbo frame support code. o Add support for hardware assistend VLAN tag insertion/stripping. o Add support for Tx/Rx flow control based on patches from Peer Chen. [2] o Add a routine that detects whether ethernet address swap routines is required. [3] o Add a workaround that take MAC/PHY out of power down mode. o Add suspend/resume support. o style(9) and code clean up. Special thanks to Shigeaki Tagashira, the original porter of nfe(4), who submitted lots of patches, performed uncountable number of regression tests and maintained nfe(4) for a long time. Without his enthusiastic help and support I could never have completed this overhauling task. The only weak point of nfe(4) compared to nve(4) is instability of manual half-duplex media selection on certain hardwares(auto sensing media type should work for all cases, though). This was a long standing bug of nfe(4) and I still have no idea why it doesn't work on some hardwares. Obtained from: OpenBSD [1] Submitted by: Peer Chen < pchen at nvidia dot com > [2], [3] Reviewed by: Shigeaki Tagashira < shigeaki AT se DOT hiroshima-u DOT ac DOT jp > Tested by: Shigeaki Tagashira, current Discussed with: current Silence from: obrien
2007-06-12 02:16:02 +00:00
m_adj(m, ETHER_ALIGN);
if (bus_dmamap_load_mbuf_sg(sc->jrxq.jrx_data_tag,
sc->jrxq.jrx_spare_map, m, segs, &nsegs, BUS_DMA_NOWAIT) != 0) {
m_freem(m);
return (ENOBUFS);
}
KASSERT(nsegs == 1, ("%s: %d segments returned!", __func__, nsegs));
data = &sc->jrxq.jdata[idx];
if (data->m != NULL) {
bus_dmamap_sync(sc->jrxq.jrx_data_tag, data->rx_data_map,
BUS_DMASYNC_POSTREAD);
bus_dmamap_unload(sc->jrxq.jrx_data_tag, data->rx_data_map);
}
map = data->rx_data_map;
data->rx_data_map = sc->jrxq.jrx_spare_map;
sc->jrxq.jrx_spare_map = map;
bus_dmamap_sync(sc->jrxq.jrx_data_tag, data->rx_data_map,
BUS_DMASYNC_PREREAD);
data->paddr = segs[0].ds_addr;
data->m = m;
/* update mapping address in h/w descriptor */
if (sc->nfe_flags & NFE_40BIT_ADDR) {
desc64 = &sc->jrxq.jdesc64[idx];
desc64->physaddr[0] = htole32(NFE_ADDR_HI(segs[0].ds_addr));
desc64->physaddr[1] = htole32(NFE_ADDR_LO(segs[0].ds_addr));
desc64->length = htole16(segs[0].ds_len);
desc64->flags = htole16(NFE_RX_READY);
} else {
desc32 = &sc->jrxq.jdesc32[idx];
desc32->physaddr = htole32(NFE_ADDR_LO(segs[0].ds_addr));
desc32->length = htole16(segs[0].ds_len);
desc32->flags = htole16(NFE_RX_READY);
}
style(9) cleanup.
2006-10-19 10:01:26 +00:00
Bring overhauled nfe(4) into tree. o s/printf/device_printf/g o Nuke OpenBSDism. o Nuke NetBSD/OpenBSD specific DMA sync operations.(we don't have a way to sync a single descriptor within a DMA map.) o Remove recursive mutex. o bus_dma(9) clean up. o 40bit DMA address support. o Add protection for Rx map load failure. o Fix a long standing bug for watchdog timeout. [1] o Add additional protections, missing Tx completion interrupt, losing start Tx command, for watchdog timeout. o Switch to taskqueue(9) API to handle interrupts. o Use our own timer for watchdog instead of if_watchdog/if_timer interface. o Advertise VLAN header length/capability correctly to upper layer. o Remove excessive kernel stack consumption in nfe_encap(). o Handle highly fragmented mbuf chains correctly. o Enable etherenet address reprogramming with ifconfig(8). o Add ALTQ/TSO, MSI/MSIX support. o Increased Rx ring to 256 descriptors from 128. o Align Tx/Rx descriptor ring on sizeof(struct nfe_desc64) boundary. o Remove alignment restrictions on Tx/Rx buffers. o Rewritten jumbo frame support code. o Add support for hardware assistend VLAN tag insertion/stripping. o Add support for Tx/Rx flow control based on patches from Peer Chen. [2] o Add a routine that detects whether ethernet address swap routines is required. [3] o Add a workaround that take MAC/PHY out of power down mode. o Add suspend/resume support. o style(9) and code clean up. Special thanks to Shigeaki Tagashira, the original porter of nfe(4), who submitted lots of patches, performed uncountable number of regression tests and maintained nfe(4) for a long time. Without his enthusiastic help and support I could never have completed this overhauling task. The only weak point of nfe(4) compared to nve(4) is instability of manual half-duplex media selection on certain hardwares(auto sensing media type should work for all cases, though). This was a long standing bug of nfe(4) and I still have no idea why it doesn't work on some hardwares. Obtained from: OpenBSD [1] Submitted by: Peer Chen < pchen at nvidia dot com > [2], [3] Reviewed by: Shigeaki Tagashira < shigeaki AT se DOT hiroshima-u DOT ac DOT jp > Tested by: Shigeaki Tagashira, current Discussed with: current Silence from: obrien
2007-06-12 02:16:02 +00:00
return (0);
Add a pure open source nForce Ethernet driver, under BSDL. This driver was ported from OpenBSD by Shigeaki Tagashira <shigeaki@se.hiroshima-u.ac.jp> and posted at http://www.se.hiroshima-u.ac.jp/~shigeaki/software/freebsd-nfe.html It was additionally cleaned up by me. It is still a work-in-progress and thus is purposefully not in GENERIC. And it conflicts with nve(4), so only one should be loaded.
2006-06-26 23:41:07 +00:00
}
style(9) cleanup.
2006-10-19 10:01:26 +00:00
Bring overhauled nfe(4) into tree. o s/printf/device_printf/g o Nuke OpenBSDism. o Nuke NetBSD/OpenBSD specific DMA sync operations.(we don't have a way to sync a single descriptor within a DMA map.) o Remove recursive mutex. o bus_dma(9) clean up. o 40bit DMA address support. o Add protection for Rx map load failure. o Fix a long standing bug for watchdog timeout. [1] o Add additional protections, missing Tx completion interrupt, losing start Tx command, for watchdog timeout. o Switch to taskqueue(9) API to handle interrupts. o Use our own timer for watchdog instead of if_watchdog/if_timer interface. o Advertise VLAN header length/capability correctly to upper layer. o Remove excessive kernel stack consumption in nfe_encap(). o Handle highly fragmented mbuf chains correctly. o Enable etherenet address reprogramming with ifconfig(8). o Add ALTQ/TSO, MSI/MSIX support. o Increased Rx ring to 256 descriptors from 128. o Align Tx/Rx descriptor ring on sizeof(struct nfe_desc64) boundary. o Remove alignment restrictions on Tx/Rx buffers. o Rewritten jumbo frame support code. o Add support for hardware assistend VLAN tag insertion/stripping. o Add support for Tx/Rx flow control based on patches from Peer Chen. [2] o Add a routine that detects whether ethernet address swap routines is required. [3] o Add a workaround that take MAC/PHY out of power down mode. o Add suspend/resume support. o style(9) and code clean up. Special thanks to Shigeaki Tagashira, the original porter of nfe(4), who submitted lots of patches, performed uncountable number of regression tests and maintained nfe(4) for a long time. Without his enthusiastic help and support I could never have completed this overhauling task. The only weak point of nfe(4) compared to nve(4) is instability of manual half-duplex media selection on certain hardwares(auto sensing media type should work for all cases, though). This was a long standing bug of nfe(4) and I still have no idea why it doesn't work on some hardwares. Obtained from: OpenBSD [1] Submitted by: Peer Chen < pchen at nvidia dot com > [2], [3] Reviewed by: Shigeaki Tagashira < shigeaki AT se DOT hiroshima-u DOT ac DOT jp > Tested by: Shigeaki Tagashira, current Discussed with: current Silence from: obrien
2007-06-12 02:16:02 +00:00
static int
When user_frac in the polling subsystem is low it is going to busy the CPU for too long period than necessary. Additively, interfaces are kept polled (in the tick) even if no more packets are available. In order to avoid such situations a new generic mechanism can be implemented in proactive way, keeping track of the time spent on any packet and fragmenting the time for any tick, stopping the processing as soon as possible. In order to implement such mechanism, the polling handler needs to change, returning the number of packets processed. While the intended logic is not part of this patch, the polling KPI is broken by this commit, adding an int return value and the new flag IFCAP_POLLING_NOCOUNT (which will signal that the return value is meaningless for the installed handler and checking should be skipped). Bump __FreeBSD_version in order to signal such situation. Reviewed by: emaste Sponsored by: Sandvine Incorporated
2009-05-30 15:14:44 +00:00
nfe_rxeof(struct nfe_softc *sc, int count, int *rx_npktsp)
Add a pure open source nForce Ethernet driver, under BSDL. This driver was ported from OpenBSD by Shigeaki Tagashira <shigeaki@se.hiroshima-u.ac.jp> and posted at http://www.se.hiroshima-u.ac.jp/~shigeaki/software/freebsd-nfe.html It was additionally cleaned up by me. It is still a work-in-progress and thus is purposefully not in GENERIC. And it conflicts with nve(4), so only one should be loaded.
2006-06-26 23:41:07 +00:00
{
Bring overhauled nfe(4) into tree. o s/printf/device_printf/g o Nuke OpenBSDism. o Nuke NetBSD/OpenBSD specific DMA sync operations.(we don't have a way to sync a single descriptor within a DMA map.) o Remove recursive mutex. o bus_dma(9) clean up. o 40bit DMA address support. o Add protection for Rx map load failure. o Fix a long standing bug for watchdog timeout. [1] o Add additional protections, missing Tx completion interrupt, losing start Tx command, for watchdog timeout. o Switch to taskqueue(9) API to handle interrupts. o Use our own timer for watchdog instead of if_watchdog/if_timer interface. o Advertise VLAN header length/capability correctly to upper layer. o Remove excessive kernel stack consumption in nfe_encap(). o Handle highly fragmented mbuf chains correctly. o Enable etherenet address reprogramming with ifconfig(8). o Add ALTQ/TSO, MSI/MSIX support. o Increased Rx ring to 256 descriptors from 128. o Align Tx/Rx descriptor ring on sizeof(struct nfe_desc64) boundary. o Remove alignment restrictions on Tx/Rx buffers. o Rewritten jumbo frame support code. o Add support for hardware assistend VLAN tag insertion/stripping. o Add support for Tx/Rx flow control based on patches from Peer Chen. [2] o Add a routine that detects whether ethernet address swap routines is required. [3] o Add a workaround that take MAC/PHY out of power down mode. o Add suspend/resume support. o style(9) and code clean up. Special thanks to Shigeaki Tagashira, the original porter of nfe(4), who submitted lots of patches, performed uncountable number of regression tests and maintained nfe(4) for a long time. Without his enthusiastic help and support I could never have completed this overhauling task. The only weak point of nfe(4) compared to nve(4) is instability of manual half-duplex media selection on certain hardwares(auto sensing media type should work for all cases, though). This was a long standing bug of nfe(4) and I still have no idea why it doesn't work on some hardwares. Obtained from: OpenBSD [1] Submitted by: Peer Chen < pchen at nvidia dot com > [2], [3] Reviewed by: Shigeaki Tagashira < shigeaki AT se DOT hiroshima-u DOT ac DOT jp > Tested by: Shigeaki Tagashira, current Discussed with: current Silence from: obrien
2007-06-12 02:16:02 +00:00
struct ifnet *ifp = sc->nfe_ifp;
struct nfe_desc32 *desc32;
struct nfe_desc64 *desc64;
struct nfe_rx_data *data;
struct mbuf *m;
uint16_t flags;
When user_frac in the polling subsystem is low it is going to busy the CPU for too long period than necessary. Additively, interfaces are kept polled (in the tick) even if no more packets are available. In order to avoid such situations a new generic mechanism can be implemented in proactive way, keeping track of the time spent on any packet and fragmenting the time for any tick, stopping the processing as soon as possible. In order to implement such mechanism, the polling handler needs to change, returning the number of packets processed. While the intended logic is not part of this patch, the polling KPI is broken by this commit, adding an int return value and the new flag IFCAP_POLLING_NOCOUNT (which will signal that the return value is meaningless for the installed handler and checking should be skipped). Bump __FreeBSD_version in order to signal such situation. Reviewed by: emaste Sponsored by: Sandvine Incorporated
2009-05-30 15:14:44 +00:00
int len, prog, rx_npkts;
Bring overhauled nfe(4) into tree. o s/printf/device_printf/g o Nuke OpenBSDism. o Nuke NetBSD/OpenBSD specific DMA sync operations.(we don't have a way to sync a single descriptor within a DMA map.) o Remove recursive mutex. o bus_dma(9) clean up. o 40bit DMA address support. o Add protection for Rx map load failure. o Fix a long standing bug for watchdog timeout. [1] o Add additional protections, missing Tx completion interrupt, losing start Tx command, for watchdog timeout. o Switch to taskqueue(9) API to handle interrupts. o Use our own timer for watchdog instead of if_watchdog/if_timer interface. o Advertise VLAN header length/capability correctly to upper layer. o Remove excessive kernel stack consumption in nfe_encap(). o Handle highly fragmented mbuf chains correctly. o Enable etherenet address reprogramming with ifconfig(8). o Add ALTQ/TSO, MSI/MSIX support. o Increased Rx ring to 256 descriptors from 128. o Align Tx/Rx descriptor ring on sizeof(struct nfe_desc64) boundary. o Remove alignment restrictions on Tx/Rx buffers. o Rewritten jumbo frame support code. o Add support for hardware assistend VLAN tag insertion/stripping. o Add support for Tx/Rx flow control based on patches from Peer Chen. [2] o Add a routine that detects whether ethernet address swap routines is required. [3] o Add a workaround that take MAC/PHY out of power down mode. o Add suspend/resume support. o style(9) and code clean up. Special thanks to Shigeaki Tagashira, the original porter of nfe(4), who submitted lots of patches, performed uncountable number of regression tests and maintained nfe(4) for a long time. Without his enthusiastic help and support I could never have completed this overhauling task. The only weak point of nfe(4) compared to nve(4) is instability of manual half-duplex media selection on certain hardwares(auto sensing media type should work for all cases, though). This was a long standing bug of nfe(4) and I still have no idea why it doesn't work on some hardwares. Obtained from: OpenBSD [1] Submitted by: Peer Chen < pchen at nvidia dot com > [2], [3] Reviewed by: Shigeaki Tagashira < shigeaki AT se DOT hiroshima-u DOT ac DOT jp > Tested by: Shigeaki Tagashira, current Discussed with: current Silence from: obrien
2007-06-12 02:16:02 +00:00
uint32_t vtag = 0;
When user_frac in the polling subsystem is low it is going to busy the CPU for too long period than necessary. Additively, interfaces are kept polled (in the tick) even if no more packets are available. In order to avoid such situations a new generic mechanism can be implemented in proactive way, keeping track of the time spent on any packet and fragmenting the time for any tick, stopping the processing as soon as possible. In order to implement such mechanism, the polling handler needs to change, returning the number of packets processed. While the intended logic is not part of this patch, the polling KPI is broken by this commit, adding an int return value and the new flag IFCAP_POLLING_NOCOUNT (which will signal that the return value is meaningless for the installed handler and checking should be skipped). Bump __FreeBSD_version in order to signal such situation. Reviewed by: emaste Sponsored by: Sandvine Incorporated
2009-05-30 15:14:44 +00:00
rx_npkts = 0;
Bring overhauled nfe(4) into tree. o s/printf/device_printf/g o Nuke OpenBSDism. o Nuke NetBSD/OpenBSD specific DMA sync operations.(we don't have a way to sync a single descriptor within a DMA map.) o Remove recursive mutex. o bus_dma(9) clean up. o 40bit DMA address support. o Add protection for Rx map load failure. o Fix a long standing bug for watchdog timeout. [1] o Add additional protections, missing Tx completion interrupt, losing start Tx command, for watchdog timeout. o Switch to taskqueue(9) API to handle interrupts. o Use our own timer for watchdog instead of if_watchdog/if_timer interface. o Advertise VLAN header length/capability correctly to upper layer. o Remove excessive kernel stack consumption in nfe_encap(). o Handle highly fragmented mbuf chains correctly. o Enable etherenet address reprogramming with ifconfig(8). o Add ALTQ/TSO, MSI/MSIX support. o Increased Rx ring to 256 descriptors from 128. o Align Tx/Rx descriptor ring on sizeof(struct nfe_desc64) boundary. o Remove alignment restrictions on Tx/Rx buffers. o Rewritten jumbo frame support code. o Add support for hardware assistend VLAN tag insertion/stripping. o Add support for Tx/Rx flow control based on patches from Peer Chen. [2] o Add a routine that detects whether ethernet address swap routines is required. [3] o Add a workaround that take MAC/PHY out of power down mode. o Add suspend/resume support. o style(9) and code clean up. Special thanks to Shigeaki Tagashira, the original porter of nfe(4), who submitted lots of patches, performed uncountable number of regression tests and maintained nfe(4) for a long time. Without his enthusiastic help and support I could never have completed this overhauling task. The only weak point of nfe(4) compared to nve(4) is instability of manual half-duplex media selection on certain hardwares(auto sensing media type should work for all cases, though). This was a long standing bug of nfe(4) and I still have no idea why it doesn't work on some hardwares. Obtained from: OpenBSD [1] Submitted by: Peer Chen < pchen at nvidia dot com > [2], [3] Reviewed by: Shigeaki Tagashira < shigeaki AT se DOT hiroshima-u DOT ac DOT jp > Tested by: Shigeaki Tagashira, current Discussed with: current Silence from: obrien
2007-06-12 02:16:02 +00:00
NFE_LOCK_ASSERT(sc);
bus_dmamap_sync(sc->rxq.rx_desc_tag, sc->rxq.rx_desc_map,
BUS_DMASYNC_POSTREAD);
for (prog = 0;;NFE_INC(sc->rxq.cur, NFE_RX_RING_COUNT), vtag = 0) {
if (count <= 0)
break;
count--;
data = &sc->rxq.data[sc->rxq.cur];
if (sc->nfe_flags & NFE_40BIT_ADDR) {
desc64 = &sc->rxq.desc64[sc->rxq.cur];
vtag = le32toh(desc64->physaddr[1]);
flags = le16toh(desc64->flags);
len = le16toh(desc64->length) & NFE_RX_LEN_MASK;
} else {
desc32 = &sc->rxq.desc32[sc->rxq.cur];
flags = le16toh(desc32->flags);
len = le16toh(desc32->length) & NFE_RX_LEN_MASK;
}
if (flags & NFE_RX_READY)
break;
prog++;
if ((sc->nfe_flags & (NFE_JUMBO_SUP | NFE_40BIT_ADDR)) == 0) {
if (!(flags & NFE_RX_VALID_V1)) {
ifp->if_ierrors++;
nfe_discard_rxbuf(sc, sc->rxq.cur);
continue;
}
if ((flags & NFE_RX_FIXME_V1) == NFE_RX_FIXME_V1) {
flags &= ~NFE_RX_ERROR;
len--; /* fix buffer length */
}
} else {
if (!(flags & NFE_RX_VALID_V2)) {
ifp->if_ierrors++;
nfe_discard_rxbuf(sc, sc->rxq.cur);
continue;
}
if ((flags & NFE_RX_FIXME_V2) == NFE_RX_FIXME_V2) {
flags &= ~NFE_RX_ERROR;
len--; /* fix buffer length */
}
}
if (flags & NFE_RX_ERROR) {
ifp->if_ierrors++;
nfe_discard_rxbuf(sc, sc->rxq.cur);
continue;
}
style(9) cleanup.
2006-10-19 10:01:26 +00:00
Bring overhauled nfe(4) into tree. o s/printf/device_printf/g o Nuke OpenBSDism. o Nuke NetBSD/OpenBSD specific DMA sync operations.(we don't have a way to sync a single descriptor within a DMA map.) o Remove recursive mutex. o bus_dma(9) clean up. o 40bit DMA address support. o Add protection for Rx map load failure. o Fix a long standing bug for watchdog timeout. [1] o Add additional protections, missing Tx completion interrupt, losing start Tx command, for watchdog timeout. o Switch to taskqueue(9) API to handle interrupts. o Use our own timer for watchdog instead of if_watchdog/if_timer interface. o Advertise VLAN header length/capability correctly to upper layer. o Remove excessive kernel stack consumption in nfe_encap(). o Handle highly fragmented mbuf chains correctly. o Enable etherenet address reprogramming with ifconfig(8). o Add ALTQ/TSO, MSI/MSIX support. o Increased Rx ring to 256 descriptors from 128. o Align Tx/Rx descriptor ring on sizeof(struct nfe_desc64) boundary. o Remove alignment restrictions on Tx/Rx buffers. o Rewritten jumbo frame support code. o Add support for hardware assistend VLAN tag insertion/stripping. o Add support for Tx/Rx flow control based on patches from Peer Chen. [2] o Add a routine that detects whether ethernet address swap routines is required. [3] o Add a workaround that take MAC/PHY out of power down mode. o Add suspend/resume support. o style(9) and code clean up. Special thanks to Shigeaki Tagashira, the original porter of nfe(4), who submitted lots of patches, performed uncountable number of regression tests and maintained nfe(4) for a long time. Without his enthusiastic help and support I could never have completed this overhauling task. The only weak point of nfe(4) compared to nve(4) is instability of manual half-duplex media selection on certain hardwares(auto sensing media type should work for all cases, though). This was a long standing bug of nfe(4) and I still have no idea why it doesn't work on some hardwares. Obtained from: OpenBSD [1] Submitted by: Peer Chen < pchen at nvidia dot com > [2], [3] Reviewed by: Shigeaki Tagashira < shigeaki AT se DOT hiroshima-u DOT ac DOT jp > Tested by: Shigeaki Tagashira, current Discussed with: current Silence from: obrien
2007-06-12 02:16:02 +00:00
m = data->m;
if (nfe_newbuf(sc, sc->rxq.cur) != 0) {
ifp->if_iqdrops++;
nfe_discard_rxbuf(sc, sc->rxq.cur);
continue;
}
Add a pure open source nForce Ethernet driver, under BSDL. This driver was ported from OpenBSD by Shigeaki Tagashira <shigeaki@se.hiroshima-u.ac.jp> and posted at http://www.se.hiroshima-u.ac.jp/~shigeaki/software/freebsd-nfe.html It was additionally cleaned up by me. It is still a work-in-progress and thus is purposefully not in GENERIC. And it conflicts with nve(4), so only one should be loaded.
2006-06-26 23:41:07 +00:00
Bring overhauled nfe(4) into tree. o s/printf/device_printf/g o Nuke OpenBSDism. o Nuke NetBSD/OpenBSD specific DMA sync operations.(we don't have a way to sync a single descriptor within a DMA map.) o Remove recursive mutex. o bus_dma(9) clean up. o 40bit DMA address support. o Add protection for Rx map load failure. o Fix a long standing bug for watchdog timeout. [1] o Add additional protections, missing Tx completion interrupt, losing start Tx command, for watchdog timeout. o Switch to taskqueue(9) API to handle interrupts. o Use our own timer for watchdog instead of if_watchdog/if_timer interface. o Advertise VLAN header length/capability correctly to upper layer. o Remove excessive kernel stack consumption in nfe_encap(). o Handle highly fragmented mbuf chains correctly. o Enable etherenet address reprogramming with ifconfig(8). o Add ALTQ/TSO, MSI/MSIX support. o Increased Rx ring to 256 descriptors from 128. o Align Tx/Rx descriptor ring on sizeof(struct nfe_desc64) boundary. o Remove alignment restrictions on Tx/Rx buffers. o Rewritten jumbo frame support code. o Add support for hardware assistend VLAN tag insertion/stripping. o Add support for Tx/Rx flow control based on patches from Peer Chen. [2] o Add a routine that detects whether ethernet address swap routines is required. [3] o Add a workaround that take MAC/PHY out of power down mode. o Add suspend/resume support. o style(9) and code clean up. Special thanks to Shigeaki Tagashira, the original porter of nfe(4), who submitted lots of patches, performed uncountable number of regression tests and maintained nfe(4) for a long time. Without his enthusiastic help and support I could never have completed this overhauling task. The only weak point of nfe(4) compared to nve(4) is instability of manual half-duplex media selection on certain hardwares(auto sensing media type should work for all cases, though). This was a long standing bug of nfe(4) and I still have no idea why it doesn't work on some hardwares. Obtained from: OpenBSD [1] Submitted by: Peer Chen < pchen at nvidia dot com > [2], [3] Reviewed by: Shigeaki Tagashira < shigeaki AT se DOT hiroshima-u DOT ac DOT jp > Tested by: Shigeaki Tagashira, current Discussed with: current Silence from: obrien
2007-06-12 02:16:02 +00:00
if ((vtag & NFE_RX_VTAG) != 0 &&
(ifp->if_capenable & IFCAP_VLAN_HWTAGGING) != 0) {
m->m_pkthdr.ether_vtag = vtag & 0xffff;
m->m_flags |= M_VLANTAG;
}
style(9) cleanup.
2006-10-19 10:01:26 +00:00
Bring overhauled nfe(4) into tree. o s/printf/device_printf/g o Nuke OpenBSDism. o Nuke NetBSD/OpenBSD specific DMA sync operations.(we don't have a way to sync a single descriptor within a DMA map.) o Remove recursive mutex. o bus_dma(9) clean up. o 40bit DMA address support. o Add protection for Rx map load failure. o Fix a long standing bug for watchdog timeout. [1] o Add additional protections, missing Tx completion interrupt, losing start Tx command, for watchdog timeout. o Switch to taskqueue(9) API to handle interrupts. o Use our own timer for watchdog instead of if_watchdog/if_timer interface. o Advertise VLAN header length/capability correctly to upper layer. o Remove excessive kernel stack consumption in nfe_encap(). o Handle highly fragmented mbuf chains correctly. o Enable etherenet address reprogramming with ifconfig(8). o Add ALTQ/TSO, MSI/MSIX support. o Increased Rx ring to 256 descriptors from 128. o Align Tx/Rx descriptor ring on sizeof(struct nfe_desc64) boundary. o Remove alignment restrictions on Tx/Rx buffers. o Rewritten jumbo frame support code. o Add support for hardware assistend VLAN tag insertion/stripping. o Add support for Tx/Rx flow control based on patches from Peer Chen. [2] o Add a routine that detects whether ethernet address swap routines is required. [3] o Add a workaround that take MAC/PHY out of power down mode. o Add suspend/resume support. o style(9) and code clean up. Special thanks to Shigeaki Tagashira, the original porter of nfe(4), who submitted lots of patches, performed uncountable number of regression tests and maintained nfe(4) for a long time. Without his enthusiastic help and support I could never have completed this overhauling task. The only weak point of nfe(4) compared to nve(4) is instability of manual half-duplex media selection on certain hardwares(auto sensing media type should work for all cases, though). This was a long standing bug of nfe(4) and I still have no idea why it doesn't work on some hardwares. Obtained from: OpenBSD [1] Submitted by: Peer Chen < pchen at nvidia dot com > [2], [3] Reviewed by: Shigeaki Tagashira < shigeaki AT se DOT hiroshima-u DOT ac DOT jp > Tested by: Shigeaki Tagashira, current Discussed with: current Silence from: obrien
2007-06-12 02:16:02 +00:00
m->m_pkthdr.len = m->m_len = len;
m->m_pkthdr.rcvif = ifp;
style(9) cleanup.
2006-10-19 10:01:26 +00:00
Bring overhauled nfe(4) into tree. o s/printf/device_printf/g o Nuke OpenBSDism. o Nuke NetBSD/OpenBSD specific DMA sync operations.(we don't have a way to sync a single descriptor within a DMA map.) o Remove recursive mutex. o bus_dma(9) clean up. o 40bit DMA address support. o Add protection for Rx map load failure. o Fix a long standing bug for watchdog timeout. [1] o Add additional protections, missing Tx completion interrupt, losing start Tx command, for watchdog timeout. o Switch to taskqueue(9) API to handle interrupts. o Use our own timer for watchdog instead of if_watchdog/if_timer interface. o Advertise VLAN header length/capability correctly to upper layer. o Remove excessive kernel stack consumption in nfe_encap(). o Handle highly fragmented mbuf chains correctly. o Enable etherenet address reprogramming with ifconfig(8). o Add ALTQ/TSO, MSI/MSIX support. o Increased Rx ring to 256 descriptors from 128. o Align Tx/Rx descriptor ring on sizeof(struct nfe_desc64) boundary. o Remove alignment restrictions on Tx/Rx buffers. o Rewritten jumbo frame support code. o Add support for hardware assistend VLAN tag insertion/stripping. o Add support for Tx/Rx flow control based on patches from Peer Chen. [2] o Add a routine that detects whether ethernet address swap routines is required. [3] o Add a workaround that take MAC/PHY out of power down mode. o Add suspend/resume support. o style(9) and code clean up. Special thanks to Shigeaki Tagashira, the original porter of nfe(4), who submitted lots of patches, performed uncountable number of regression tests and maintained nfe(4) for a long time. Without his enthusiastic help and support I could never have completed this overhauling task. The only weak point of nfe(4) compared to nve(4) is instability of manual half-duplex media selection on certain hardwares(auto sensing media type should work for all cases, though). This was a long standing bug of nfe(4) and I still have no idea why it doesn't work on some hardwares. Obtained from: OpenBSD [1] Submitted by: Peer Chen < pchen at nvidia dot com > [2], [3] Reviewed by: Shigeaki Tagashira < shigeaki AT se DOT hiroshima-u DOT ac DOT jp > Tested by: Shigeaki Tagashira, current Discussed with: current Silence from: obrien
2007-06-12 02:16:02 +00:00
if ((ifp->if_capenable & IFCAP_RXCSUM) != 0) {
if ((flags & NFE_RX_IP_CSUMOK) != 0) {
m->m_pkthdr.csum_flags |= CSUM_IP_CHECKED;
m->m_pkthdr.csum_flags |= CSUM_IP_VALID;
if ((flags & NFE_RX_TCP_CSUMOK) != 0 ||
(flags & NFE_RX_UDP_CSUMOK) != 0) {
m->m_pkthdr.csum_flags |=
CSUM_DATA_VALID | CSUM_PSEUDO_HDR;
m->m_pkthdr.csum_data = 0xffff;
}
}
}
Add a pure open source nForce Ethernet driver, under BSDL. This driver was ported from OpenBSD by Shigeaki Tagashira <shigeaki@se.hiroshima-u.ac.jp> and posted at http://www.se.hiroshima-u.ac.jp/~shigeaki/software/freebsd-nfe.html It was additionally cleaned up by me. It is still a work-in-progress and thus is purposefully not in GENERIC. And it conflicts with nve(4), so only one should be loaded.
2006-06-26 23:41:07 +00:00
Bring overhauled nfe(4) into tree. o s/printf/device_printf/g o Nuke OpenBSDism. o Nuke NetBSD/OpenBSD specific DMA sync operations.(we don't have a way to sync a single descriptor within a DMA map.) o Remove recursive mutex. o bus_dma(9) clean up. o 40bit DMA address support. o Add protection for Rx map load failure. o Fix a long standing bug for watchdog timeout. [1] o Add additional protections, missing Tx completion interrupt, losing start Tx command, for watchdog timeout. o Switch to taskqueue(9) API to handle interrupts. o Use our own timer for watchdog instead of if_watchdog/if_timer interface. o Advertise VLAN header length/capability correctly to upper layer. o Remove excessive kernel stack consumption in nfe_encap(). o Handle highly fragmented mbuf chains correctly. o Enable etherenet address reprogramming with ifconfig(8). o Add ALTQ/TSO, MSI/MSIX support. o Increased Rx ring to 256 descriptors from 128. o Align Tx/Rx descriptor ring on sizeof(struct nfe_desc64) boundary. o Remove alignment restrictions on Tx/Rx buffers. o Rewritten jumbo frame support code. o Add support for hardware assistend VLAN tag insertion/stripping. o Add support for Tx/Rx flow control based on patches from Peer Chen. [2] o Add a routine that detects whether ethernet address swap routines is required. [3] o Add a workaround that take MAC/PHY out of power down mode. o Add suspend/resume support. o style(9) and code clean up. Special thanks to Shigeaki Tagashira, the original porter of nfe(4), who submitted lots of patches, performed uncountable number of regression tests and maintained nfe(4) for a long time. Without his enthusiastic help and support I could never have completed this overhauling task. The only weak point of nfe(4) compared to nve(4) is instability of manual half-duplex media selection on certain hardwares(auto sensing media type should work for all cases, though). This was a long standing bug of nfe(4) and I still have no idea why it doesn't work on some hardwares. Obtained from: OpenBSD [1] Submitted by: Peer Chen < pchen at nvidia dot com > [2], [3] Reviewed by: Shigeaki Tagashira < shigeaki AT se DOT hiroshima-u DOT ac DOT jp > Tested by: Shigeaki Tagashira, current Discussed with: current Silence from: obrien
2007-06-12 02:16:02 +00:00
ifp->if_ipackets++;
style(9) cleanup.
2006-10-19 10:01:26 +00:00
Bring overhauled nfe(4) into tree. o s/printf/device_printf/g o Nuke OpenBSDism. o Nuke NetBSD/OpenBSD specific DMA sync operations.(we don't have a way to sync a single descriptor within a DMA map.) o Remove recursive mutex. o bus_dma(9) clean up. o 40bit DMA address support. o Add protection for Rx map load failure. o Fix a long standing bug for watchdog timeout. [1] o Add additional protections, missing Tx completion interrupt, losing start Tx command, for watchdog timeout. o Switch to taskqueue(9) API to handle interrupts. o Use our own timer for watchdog instead of if_watchdog/if_timer interface. o Advertise VLAN header length/capability correctly to upper layer. o Remove excessive kernel stack consumption in nfe_encap(). o Handle highly fragmented mbuf chains correctly. o Enable etherenet address reprogramming with ifconfig(8). o Add ALTQ/TSO, MSI/MSIX support. o Increased Rx ring to 256 descriptors from 128. o Align Tx/Rx descriptor ring on sizeof(struct nfe_desc64) boundary. o Remove alignment restrictions on Tx/Rx buffers. o Rewritten jumbo frame support code. o Add support for hardware assistend VLAN tag insertion/stripping. o Add support for Tx/Rx flow control based on patches from Peer Chen. [2] o Add a routine that detects whether ethernet address swap routines is required. [3] o Add a workaround that take MAC/PHY out of power down mode. o Add suspend/resume support. o style(9) and code clean up. Special thanks to Shigeaki Tagashira, the original porter of nfe(4), who submitted lots of patches, performed uncountable number of regression tests and maintained nfe(4) for a long time. Without his enthusiastic help and support I could never have completed this overhauling task. The only weak point of nfe(4) compared to nve(4) is instability of manual half-duplex media selection on certain hardwares(auto sensing media type should work for all cases, though). This was a long standing bug of nfe(4) and I still have no idea why it doesn't work on some hardwares. Obtained from: OpenBSD [1] Submitted by: Peer Chen < pchen at nvidia dot com > [2], [3] Reviewed by: Shigeaki Tagashira < shigeaki AT se DOT hiroshima-u DOT ac DOT jp > Tested by: Shigeaki Tagashira, current Discussed with: current Silence from: obrien
2007-06-12 02:16:02 +00:00
NFE_UNLOCK(sc);
(*ifp->if_input)(ifp, m);
NFE_LOCK(sc);
When user_frac in the polling subsystem is low it is going to busy the CPU for too long period than necessary. Additively, interfaces are kept polled (in the tick) even if no more packets are available. In order to avoid such situations a new generic mechanism can be implemented in proactive way, keeping track of the time spent on any packet and fragmenting the time for any tick, stopping the processing as soon as possible. In order to implement such mechanism, the polling handler needs to change, returning the number of packets processed. While the intended logic is not part of this patch, the polling KPI is broken by this commit, adding an int return value and the new flag IFCAP_POLLING_NOCOUNT (which will signal that the return value is meaningless for the installed handler and checking should be skipped). Bump __FreeBSD_version in order to signal such situation. Reviewed by: emaste Sponsored by: Sandvine Incorporated
2009-05-30 15:14:44 +00:00
rx_npkts++;
Bring overhauled nfe(4) into tree. o s/printf/device_printf/g o Nuke OpenBSDism. o Nuke NetBSD/OpenBSD specific DMA sync operations.(we don't have a way to sync a single descriptor within a DMA map.) o Remove recursive mutex. o bus_dma(9) clean up. o 40bit DMA address support. o Add protection for Rx map load failure. o Fix a long standing bug for watchdog timeout. [1] o Add additional protections, missing Tx completion interrupt, losing start Tx command, for watchdog timeout. o Switch to taskqueue(9) API to handle interrupts. o Use our own timer for watchdog instead of if_watchdog/if_timer interface. o Advertise VLAN header length/capability correctly to upper layer. o Remove excessive kernel stack consumption in nfe_encap(). o Handle highly fragmented mbuf chains correctly. o Enable etherenet address reprogramming with ifconfig(8). o Add ALTQ/TSO, MSI/MSIX support. o Increased Rx ring to 256 descriptors from 128. o Align Tx/Rx descriptor ring on sizeof(struct nfe_desc64) boundary. o Remove alignment restrictions on Tx/Rx buffers. o Rewritten jumbo frame support code. o Add support for hardware assistend VLAN tag insertion/stripping. o Add support for Tx/Rx flow control based on patches from Peer Chen. [2] o Add a routine that detects whether ethernet address swap routines is required. [3] o Add a workaround that take MAC/PHY out of power down mode. o Add suspend/resume support. o style(9) and code clean up. Special thanks to Shigeaki Tagashira, the original porter of nfe(4), who submitted lots of patches, performed uncountable number of regression tests and maintained nfe(4) for a long time. Without his enthusiastic help and support I could never have completed this overhauling task. The only weak point of nfe(4) compared to nve(4) is instability of manual half-duplex media selection on certain hardwares(auto sensing media type should work for all cases, though). This was a long standing bug of nfe(4) and I still have no idea why it doesn't work on some hardwares. Obtained from: OpenBSD [1] Submitted by: Peer Chen < pchen at nvidia dot com > [2], [3] Reviewed by: Shigeaki Tagashira < shigeaki AT se DOT hiroshima-u DOT ac DOT jp > Tested by: Shigeaki Tagashira, current Discussed with: current Silence from: obrien
2007-06-12 02:16:02 +00:00
}
if (prog > 0)
bus_dmamap_sync(sc->rxq.rx_desc_tag, sc->rxq.rx_desc_map,
BUS_DMASYNC_PREREAD | BUS_DMASYNC_PREWRITE);
Add a pure open source nForce Ethernet driver, under BSDL. This driver was ported from OpenBSD by Shigeaki Tagashira <shigeaki@se.hiroshima-u.ac.jp> and posted at http://www.se.hiroshima-u.ac.jp/~shigeaki/software/freebsd-nfe.html It was additionally cleaned up by me. It is still a work-in-progress and thus is purposefully not in GENERIC. And it conflicts with nve(4), so only one should be loaded.
2006-06-26 23:41:07 +00:00
When user_frac in the polling subsystem is low it is going to busy the CPU for too long period than necessary. Additively, interfaces are kept polled (in the tick) even if no more packets are available. In order to avoid such situations a new generic mechanism can be implemented in proactive way, keeping track of the time spent on any packet and fragmenting the time for any tick, stopping the processing as soon as possible. In order to implement such mechanism, the polling handler needs to change, returning the number of packets processed. While the intended logic is not part of this patch, the polling KPI is broken by this commit, adding an int return value and the new flag IFCAP_POLLING_NOCOUNT (which will signal that the return value is meaningless for the installed handler and checking should be skipped). Bump __FreeBSD_version in order to signal such situation. Reviewed by: emaste Sponsored by: Sandvine Incorporated
2009-05-30 15:14:44 +00:00
if (rx_npktsp != NULL)
*rx_npktsp = rx_npkts;
Bring overhauled nfe(4) into tree. o s/printf/device_printf/g o Nuke OpenBSDism. o Nuke NetBSD/OpenBSD specific DMA sync operations.(we don't have a way to sync a single descriptor within a DMA map.) o Remove recursive mutex. o bus_dma(9) clean up. o 40bit DMA address support. o Add protection for Rx map load failure. o Fix a long standing bug for watchdog timeout. [1] o Add additional protections, missing Tx completion interrupt, losing start Tx command, for watchdog timeout. o Switch to taskqueue(9) API to handle interrupts. o Use our own timer for watchdog instead of if_watchdog/if_timer interface. o Advertise VLAN header length/capability correctly to upper layer. o Remove excessive kernel stack consumption in nfe_encap(). o Handle highly fragmented mbuf chains correctly. o Enable etherenet address reprogramming with ifconfig(8). o Add ALTQ/TSO, MSI/MSIX support. o Increased Rx ring to 256 descriptors from 128. o Align Tx/Rx descriptor ring on sizeof(struct nfe_desc64) boundary. o Remove alignment restrictions on Tx/Rx buffers. o Rewritten jumbo frame support code. o Add support for hardware assistend VLAN tag insertion/stripping. o Add support for Tx/Rx flow control based on patches from Peer Chen. [2] o Add a routine that detects whether ethernet address swap routines is required. [3] o Add a workaround that take MAC/PHY out of power down mode. o Add suspend/resume support. o style(9) and code clean up. Special thanks to Shigeaki Tagashira, the original porter of nfe(4), who submitted lots of patches, performed uncountable number of regression tests and maintained nfe(4) for a long time. Without his enthusiastic help and support I could never have completed this overhauling task. The only weak point of nfe(4) compared to nve(4) is instability of manual half-duplex media selection on certain hardwares(auto sensing media type should work for all cases, though). This was a long standing bug of nfe(4) and I still have no idea why it doesn't work on some hardwares. Obtained from: OpenBSD [1] Submitted by: Peer Chen < pchen at nvidia dot com > [2], [3] Reviewed by: Shigeaki Tagashira < shigeaki AT se DOT hiroshima-u DOT ac DOT jp > Tested by: Shigeaki Tagashira, current Discussed with: current Silence from: obrien
2007-06-12 02:16:02 +00:00
return (count > 0 ? 0 : EAGAIN);
Add a pure open source nForce Ethernet driver, under BSDL. This driver was ported from OpenBSD by Shigeaki Tagashira <shigeaki@se.hiroshima-u.ac.jp> and posted at http://www.se.hiroshima-u.ac.jp/~shigeaki/software/freebsd-nfe.html It was additionally cleaned up by me. It is still a work-in-progress and thus is purposefully not in GENERIC. And it conflicts with nve(4), so only one should be loaded.
2006-06-26 23:41:07 +00:00
}
style(9) cleanup.
2006-10-19 10:01:26 +00:00
Bring overhauled nfe(4) into tree. o s/printf/device_printf/g o Nuke OpenBSDism. o Nuke NetBSD/OpenBSD specific DMA sync operations.(we don't have a way to sync a single descriptor within a DMA map.) o Remove recursive mutex. o bus_dma(9) clean up. o 40bit DMA address support. o Add protection for Rx map load failure. o Fix a long standing bug for watchdog timeout. [1] o Add additional protections, missing Tx completion interrupt, losing start Tx command, for watchdog timeout. o Switch to taskqueue(9) API to handle interrupts. o Use our own timer for watchdog instead of if_watchdog/if_timer interface. o Advertise VLAN header length/capability correctly to upper layer. o Remove excessive kernel stack consumption in nfe_encap(). o Handle highly fragmented mbuf chains correctly. o Enable etherenet address reprogramming with ifconfig(8). o Add ALTQ/TSO, MSI/MSIX support. o Increased Rx ring to 256 descriptors from 128. o Align Tx/Rx descriptor ring on sizeof(struct nfe_desc64) boundary. o Remove alignment restrictions on Tx/Rx buffers. o Rewritten jumbo frame support code. o Add support for hardware assistend VLAN tag insertion/stripping. o Add support for Tx/Rx flow control based on patches from Peer Chen. [2] o Add a routine that detects whether ethernet address swap routines is required. [3] o Add a workaround that take MAC/PHY out of power down mode. o Add suspend/resume support. o style(9) and code clean up. Special thanks to Shigeaki Tagashira, the original porter of nfe(4), who submitted lots of patches, performed uncountable number of regression tests and maintained nfe(4) for a long time. Without his enthusiastic help and support I could never have completed this overhauling task. The only weak point of nfe(4) compared to nve(4) is instability of manual half-duplex media selection on certain hardwares(auto sensing media type should work for all cases, though). This was a long standing bug of nfe(4) and I still have no idea why it doesn't work on some hardwares. Obtained from: OpenBSD [1] Submitted by: Peer Chen < pchen at nvidia dot com > [2], [3] Reviewed by: Shigeaki Tagashira < shigeaki AT se DOT hiroshima-u DOT ac DOT jp > Tested by: Shigeaki Tagashira, current Discussed with: current Silence from: obrien
2007-06-12 02:16:02 +00:00
static int
When user_frac in the polling subsystem is low it is going to busy the CPU for too long period than necessary. Additively, interfaces are kept polled (in the tick) even if no more packets are available. In order to avoid such situations a new generic mechanism can be implemented in proactive way, keeping track of the time spent on any packet and fragmenting the time for any tick, stopping the processing as soon as possible. In order to implement such mechanism, the polling handler needs to change, returning the number of packets processed. While the intended logic is not part of this patch, the polling KPI is broken by this commit, adding an int return value and the new flag IFCAP_POLLING_NOCOUNT (which will signal that the return value is meaningless for the installed handler and checking should be skipped). Bump __FreeBSD_version in order to signal such situation. Reviewed by: emaste Sponsored by: Sandvine Incorporated
2009-05-30 15:14:44 +00:00
nfe_jrxeof(struct nfe_softc *sc, int count, int *rx_npktsp)
Add a pure open source nForce Ethernet driver, under BSDL. This driver was ported from OpenBSD by Shigeaki Tagashira <shigeaki@se.hiroshima-u.ac.jp> and posted at http://www.se.hiroshima-u.ac.jp/~shigeaki/software/freebsd-nfe.html It was additionally cleaned up by me. It is still a work-in-progress and thus is purposefully not in GENERIC. And it conflicts with nve(4), so only one should be loaded.
2006-06-26 23:41:07 +00:00
{
struct ifnet *ifp = sc->nfe_ifp;
Bring overhauled nfe(4) into tree. o s/printf/device_printf/g o Nuke OpenBSDism. o Nuke NetBSD/OpenBSD specific DMA sync operations.(we don't have a way to sync a single descriptor within a DMA map.) o Remove recursive mutex. o bus_dma(9) clean up. o 40bit DMA address support. o Add protection for Rx map load failure. o Fix a long standing bug for watchdog timeout. [1] o Add additional protections, missing Tx completion interrupt, losing start Tx command, for watchdog timeout. o Switch to taskqueue(9) API to handle interrupts. o Use our own timer for watchdog instead of if_watchdog/if_timer interface. o Advertise VLAN header length/capability correctly to upper layer. o Remove excessive kernel stack consumption in nfe_encap(). o Handle highly fragmented mbuf chains correctly. o Enable etherenet address reprogramming with ifconfig(8). o Add ALTQ/TSO, MSI/MSIX support. o Increased Rx ring to 256 descriptors from 128. o Align Tx/Rx descriptor ring on sizeof(struct nfe_desc64) boundary. o Remove alignment restrictions on Tx/Rx buffers. o Rewritten jumbo frame support code. o Add support for hardware assistend VLAN tag insertion/stripping. o Add support for Tx/Rx flow control based on patches from Peer Chen. [2] o Add a routine that detects whether ethernet address swap routines is required. [3] o Add a workaround that take MAC/PHY out of power down mode. o Add suspend/resume support. o style(9) and code clean up. Special thanks to Shigeaki Tagashira, the original porter of nfe(4), who submitted lots of patches, performed uncountable number of regression tests and maintained nfe(4) for a long time. Without his enthusiastic help and support I could never have completed this overhauling task. The only weak point of nfe(4) compared to nve(4) is instability of manual half-duplex media selection on certain hardwares(auto sensing media type should work for all cases, though). This was a long standing bug of nfe(4) and I still have no idea why it doesn't work on some hardwares. Obtained from: OpenBSD [1] Submitted by: Peer Chen < pchen at nvidia dot com > [2], [3] Reviewed by: Shigeaki Tagashira < shigeaki AT se DOT hiroshima-u DOT ac DOT jp > Tested by: Shigeaki Tagashira, current Discussed with: current Silence from: obrien
2007-06-12 02:16:02 +00:00
struct nfe_desc32 *desc32;
struct nfe_desc64 *desc64;
Add a pure open source nForce Ethernet driver, under BSDL. This driver was ported from OpenBSD by Shigeaki Tagashira <shigeaki@se.hiroshima-u.ac.jp> and posted at http://www.se.hiroshima-u.ac.jp/~shigeaki/software/freebsd-nfe.html It was additionally cleaned up by me. It is still a work-in-progress and thus is purposefully not in GENERIC. And it conflicts with nve(4), so only one should be loaded.
2006-06-26 23:41:07 +00:00
struct nfe_rx_data *data;
Bring overhauled nfe(4) into tree. o s/printf/device_printf/g o Nuke OpenBSDism. o Nuke NetBSD/OpenBSD specific DMA sync operations.(we don't have a way to sync a single descriptor within a DMA map.) o Remove recursive mutex. o bus_dma(9) clean up. o 40bit DMA address support. o Add protection for Rx map load failure. o Fix a long standing bug for watchdog timeout. [1] o Add additional protections, missing Tx completion interrupt, losing start Tx command, for watchdog timeout. o Switch to taskqueue(9) API to handle interrupts. o Use our own timer for watchdog instead of if_watchdog/if_timer interface. o Advertise VLAN header length/capability correctly to upper layer. o Remove excessive kernel stack consumption in nfe_encap(). o Handle highly fragmented mbuf chains correctly. o Enable etherenet address reprogramming with ifconfig(8). o Add ALTQ/TSO, MSI/MSIX support. o Increased Rx ring to 256 descriptors from 128. o Align Tx/Rx descriptor ring on sizeof(struct nfe_desc64) boundary. o Remove alignment restrictions on Tx/Rx buffers. o Rewritten jumbo frame support code. o Add support for hardware assistend VLAN tag insertion/stripping. o Add support for Tx/Rx flow control based on patches from Peer Chen. [2] o Add a routine that detects whether ethernet address swap routines is required. [3] o Add a workaround that take MAC/PHY out of power down mode. o Add suspend/resume support. o style(9) and code clean up. Special thanks to Shigeaki Tagashira, the original porter of nfe(4), who submitted lots of patches, performed uncountable number of regression tests and maintained nfe(4) for a long time. Without his enthusiastic help and support I could never have completed this overhauling task. The only weak point of nfe(4) compared to nve(4) is instability of manual half-duplex media selection on certain hardwares(auto sensing media type should work for all cases, though). This was a long standing bug of nfe(4) and I still have no idea why it doesn't work on some hardwares. Obtained from: OpenBSD [1] Submitted by: Peer Chen < pchen at nvidia dot com > [2], [3] Reviewed by: Shigeaki Tagashira < shigeaki AT se DOT hiroshima-u DOT ac DOT jp > Tested by: Shigeaki Tagashira, current Discussed with: current Silence from: obrien
2007-06-12 02:16:02 +00:00
struct mbuf *m;
uint16_t flags;
When user_frac in the polling subsystem is low it is going to busy the CPU for too long period than necessary. Additively, interfaces are kept polled (in the tick) even if no more packets are available. In order to avoid such situations a new generic mechanism can be implemented in proactive way, keeping track of the time spent on any packet and fragmenting the time for any tick, stopping the processing as soon as possible. In order to implement such mechanism, the polling handler needs to change, returning the number of packets processed. While the intended logic is not part of this patch, the polling KPI is broken by this commit, adding an int return value and the new flag IFCAP_POLLING_NOCOUNT (which will signal that the return value is meaningless for the installed handler and checking should be skipped). Bump __FreeBSD_version in order to signal such situation. Reviewed by: emaste Sponsored by: Sandvine Incorporated
2009-05-30 15:14:44 +00:00
int len, prog, rx_npkts;
Bring overhauled nfe(4) into tree. o s/printf/device_printf/g o Nuke OpenBSDism. o Nuke NetBSD/OpenBSD specific DMA sync operations.(we don't have a way to sync a single descriptor within a DMA map.) o Remove recursive mutex. o bus_dma(9) clean up. o 40bit DMA address support. o Add protection for Rx map load failure. o Fix a long standing bug for watchdog timeout. [1] o Add additional protections, missing Tx completion interrupt, losing start Tx command, for watchdog timeout. o Switch to taskqueue(9) API to handle interrupts. o Use our own timer for watchdog instead of if_watchdog/if_timer interface. o Advertise VLAN header length/capability correctly to upper layer. o Remove excessive kernel stack consumption in nfe_encap(). o Handle highly fragmented mbuf chains correctly. o Enable etherenet address reprogramming with ifconfig(8). o Add ALTQ/TSO, MSI/MSIX support. o Increased Rx ring to 256 descriptors from 128. o Align Tx/Rx descriptor ring on sizeof(struct nfe_desc64) boundary. o Remove alignment restrictions on Tx/Rx buffers. o Rewritten jumbo frame support code. o Add support for hardware assistend VLAN tag insertion/stripping. o Add support for Tx/Rx flow control based on patches from Peer Chen. [2] o Add a routine that detects whether ethernet address swap routines is required. [3] o Add a workaround that take MAC/PHY out of power down mode. o Add suspend/resume support. o style(9) and code clean up. Special thanks to Shigeaki Tagashira, the original porter of nfe(4), who submitted lots of patches, performed uncountable number of regression tests and maintained nfe(4) for a long time. Without his enthusiastic help and support I could never have completed this overhauling task. The only weak point of nfe(4) compared to nve(4) is instability of manual half-duplex media selection on certain hardwares(auto sensing media type should work for all cases, though). This was a long standing bug of nfe(4) and I still have no idea why it doesn't work on some hardwares. Obtained from: OpenBSD [1] Submitted by: Peer Chen < pchen at nvidia dot com > [2], [3] Reviewed by: Shigeaki Tagashira < shigeaki AT se DOT hiroshima-u DOT ac DOT jp > Tested by: Shigeaki Tagashira, current Discussed with: current Silence from: obrien
2007-06-12 02:16:02 +00:00
uint32_t vtag = 0;
Add a pure open source nForce Ethernet driver, under BSDL. This driver was ported from OpenBSD by Shigeaki Tagashira <shigeaki@se.hiroshima-u.ac.jp> and posted at http://www.se.hiroshima-u.ac.jp/~shigeaki/software/freebsd-nfe.html It was additionally cleaned up by me. It is still a work-in-progress and thus is purposefully not in GENERIC. And it conflicts with nve(4), so only one should be loaded.
2006-06-26 23:41:07 +00:00
When user_frac in the polling subsystem is low it is going to busy the CPU for too long period than necessary. Additively, interfaces are kept polled (in the tick) even if no more packets are available. In order to avoid such situations a new generic mechanism can be implemented in proactive way, keeping track of the time spent on any packet and fragmenting the time for any tick, stopping the processing as soon as possible. In order to implement such mechanism, the polling handler needs to change, returning the number of packets processed. While the intended logic is not part of this patch, the polling KPI is broken by this commit, adding an int return value and the new flag IFCAP_POLLING_NOCOUNT (which will signal that the return value is meaningless for the installed handler and checking should be skipped). Bump __FreeBSD_version in order to signal such situation. Reviewed by: emaste Sponsored by: Sandvine Incorporated
2009-05-30 15:14:44 +00:00
rx_npkts = 0;
Add a pure open source nForce Ethernet driver, under BSDL. This driver was ported from OpenBSD by Shigeaki Tagashira <shigeaki@se.hiroshima-u.ac.jp> and posted at http://www.se.hiroshima-u.ac.jp/~shigeaki/software/freebsd-nfe.html It was additionally cleaned up by me. It is still a work-in-progress and thus is purposefully not in GENERIC. And it conflicts with nve(4), so only one should be loaded.
2006-06-26 23:41:07 +00:00
NFE_LOCK_ASSERT(sc);
Import the nfe (nForce Ethernet) driver from OpenBSD. This is a pure open source NIC driver with a BSD license. These bits were taken from the OpenBSD public CVS repository on 1-May-06.
2006-06-26 17:25:22 +00:00
Bring overhauled nfe(4) into tree. o s/printf/device_printf/g o Nuke OpenBSDism. o Nuke NetBSD/OpenBSD specific DMA sync operations.(we don't have a way to sync a single descriptor within a DMA map.) o Remove recursive mutex. o bus_dma(9) clean up. o 40bit DMA address support. o Add protection for Rx map load failure. o Fix a long standing bug for watchdog timeout. [1] o Add additional protections, missing Tx completion interrupt, losing start Tx command, for watchdog timeout. o Switch to taskqueue(9) API to handle interrupts. o Use our own timer for watchdog instead of if_watchdog/if_timer interface. o Advertise VLAN header length/capability correctly to upper layer. o Remove excessive kernel stack consumption in nfe_encap(). o Handle highly fragmented mbuf chains correctly. o Enable etherenet address reprogramming with ifconfig(8). o Add ALTQ/TSO, MSI/MSIX support. o Increased Rx ring to 256 descriptors from 128. o Align Tx/Rx descriptor ring on sizeof(struct nfe_desc64) boundary. o Remove alignment restrictions on Tx/Rx buffers. o Rewritten jumbo frame support code. o Add support for hardware assistend VLAN tag insertion/stripping. o Add support for Tx/Rx flow control based on patches from Peer Chen. [2] o Add a routine that detects whether ethernet address swap routines is required. [3] o Add a workaround that take MAC/PHY out of power down mode. o Add suspend/resume support. o style(9) and code clean up. Special thanks to Shigeaki Tagashira, the original porter of nfe(4), who submitted lots of patches, performed uncountable number of regression tests and maintained nfe(4) for a long time. Without his enthusiastic help and support I could never have completed this overhauling task. The only weak point of nfe(4) compared to nve(4) is instability of manual half-duplex media selection on certain hardwares(auto sensing media type should work for all cases, though). This was a long standing bug of nfe(4) and I still have no idea why it doesn't work on some hardwares. Obtained from: OpenBSD [1] Submitted by: Peer Chen < pchen at nvidia dot com > [2], [3] Reviewed by: Shigeaki Tagashira < shigeaki AT se DOT hiroshima-u DOT ac DOT jp > Tested by: Shigeaki Tagashira, current Discussed with: current Silence from: obrien
2007-06-12 02:16:02 +00:00
bus_dmamap_sync(sc->jrxq.jrx_desc_tag, sc->jrxq.jrx_desc_map,
BUS_DMASYNC_POSTREAD);
Add a pure open source nForce Ethernet driver, under BSDL. This driver was ported from OpenBSD by Shigeaki Tagashira <shigeaki@se.hiroshima-u.ac.jp> and posted at http://www.se.hiroshima-u.ac.jp/~shigeaki/software/freebsd-nfe.html It was additionally cleaned up by me. It is still a work-in-progress and thus is purposefully not in GENERIC. And it conflicts with nve(4), so only one should be loaded.
2006-06-26 23:41:07 +00:00
Bring overhauled nfe(4) into tree. o s/printf/device_printf/g o Nuke OpenBSDism. o Nuke NetBSD/OpenBSD specific DMA sync operations.(we don't have a way to sync a single descriptor within a DMA map.) o Remove recursive mutex. o bus_dma(9) clean up. o 40bit DMA address support. o Add protection for Rx map load failure. o Fix a long standing bug for watchdog timeout. [1] o Add additional protections, missing Tx completion interrupt, losing start Tx command, for watchdog timeout. o Switch to taskqueue(9) API to handle interrupts. o Use our own timer for watchdog instead of if_watchdog/if_timer interface. o Advertise VLAN header length/capability correctly to upper layer. o Remove excessive kernel stack consumption in nfe_encap(). o Handle highly fragmented mbuf chains correctly. o Enable etherenet address reprogramming with ifconfig(8). o Add ALTQ/TSO, MSI/MSIX support. o Increased Rx ring to 256 descriptors from 128. o Align Tx/Rx descriptor ring on sizeof(struct nfe_desc64) boundary. o Remove alignment restrictions on Tx/Rx buffers. o Rewritten jumbo frame support code. o Add support for hardware assistend VLAN tag insertion/stripping. o Add support for Tx/Rx flow control based on patches from Peer Chen. [2] o Add a routine that detects whether ethernet address swap routines is required. [3] o Add a workaround that take MAC/PHY out of power down mode. o Add suspend/resume support. o style(9) and code clean up. Special thanks to Shigeaki Tagashira, the original porter of nfe(4), who submitted lots of patches, performed uncountable number of regression tests and maintained nfe(4) for a long time. Without his enthusiastic help and support I could never have completed this overhauling task. The only weak point of nfe(4) compared to nve(4) is instability of manual half-duplex media selection on certain hardwares(auto sensing media type should work for all cases, though). This was a long standing bug of nfe(4) and I still have no idea why it doesn't work on some hardwares. Obtained from: OpenBSD [1] Submitted by: Peer Chen < pchen at nvidia dot com > [2], [3] Reviewed by: Shigeaki Tagashira < shigeaki AT se DOT hiroshima-u DOT ac DOT jp > Tested by: Shigeaki Tagashira, current Discussed with: current Silence from: obrien
2007-06-12 02:16:02 +00:00
for (prog = 0;;NFE_INC(sc->jrxq.jcur, NFE_JUMBO_RX_RING_COUNT),
vtag = 0) {
if (count <= 0)
break;
count--;
Add a pure open source nForce Ethernet driver, under BSDL. This driver was ported from OpenBSD by Shigeaki Tagashira <shigeaki@se.hiroshima-u.ac.jp> and posted at http://www.se.hiroshima-u.ac.jp/~shigeaki/software/freebsd-nfe.html It was additionally cleaned up by me. It is still a work-in-progress and thus is purposefully not in GENERIC. And it conflicts with nve(4), so only one should be loaded.
2006-06-26 23:41:07 +00:00
Bring overhauled nfe(4) into tree. o s/printf/device_printf/g o Nuke OpenBSDism. o Nuke NetBSD/OpenBSD specific DMA sync operations.(we don't have a way to sync a single descriptor within a DMA map.) o Remove recursive mutex. o bus_dma(9) clean up. o 40bit DMA address support. o Add protection for Rx map load failure. o Fix a long standing bug for watchdog timeout. [1] o Add additional protections, missing Tx completion interrupt, losing start Tx command, for watchdog timeout. o Switch to taskqueue(9) API to handle interrupts. o Use our own timer for watchdog instead of if_watchdog/if_timer interface. o Advertise VLAN header length/capability correctly to upper layer. o Remove excessive kernel stack consumption in nfe_encap(). o Handle highly fragmented mbuf chains correctly. o Enable etherenet address reprogramming with ifconfig(8). o Add ALTQ/TSO, MSI/MSIX support. o Increased Rx ring to 256 descriptors from 128. o Align Tx/Rx descriptor ring on sizeof(struct nfe_desc64) boundary. o Remove alignment restrictions on Tx/Rx buffers. o Rewritten jumbo frame support code. o Add support for hardware assistend VLAN tag insertion/stripping. o Add support for Tx/Rx flow control based on patches from Peer Chen. [2] o Add a routine that detects whether ethernet address swap routines is required. [3] o Add a workaround that take MAC/PHY out of power down mode. o Add suspend/resume support. o style(9) and code clean up. Special thanks to Shigeaki Tagashira, the original porter of nfe(4), who submitted lots of patches, performed uncountable number of regression tests and maintained nfe(4) for a long time. Without his enthusiastic help and support I could never have completed this overhauling task. The only weak point of nfe(4) compared to nve(4) is instability of manual half-duplex media selection on certain hardwares(auto sensing media type should work for all cases, though). This was a long standing bug of nfe(4) and I still have no idea why it doesn't work on some hardwares. Obtained from: OpenBSD [1] Submitted by: Peer Chen < pchen at nvidia dot com > [2], [3] Reviewed by: Shigeaki Tagashira < shigeaki AT se DOT hiroshima-u DOT ac DOT jp > Tested by: Shigeaki Tagashira, current Discussed with: current Silence from: obrien
2007-06-12 02:16:02 +00:00
data = &sc->jrxq.jdata[sc->jrxq.jcur];
Import the nfe (nForce Ethernet) driver from OpenBSD. This is a pure open source NIC driver with a BSD license. These bits were taken from the OpenBSD public CVS repository on 1-May-06.
2006-06-26 17:25:22 +00:00
Add a pure open source nForce Ethernet driver, under BSDL. This driver was ported from OpenBSD by Shigeaki Tagashira <shigeaki@se.hiroshima-u.ac.jp> and posted at http://www.se.hiroshima-u.ac.jp/~shigeaki/software/freebsd-nfe.html It was additionally cleaned up by me. It is still a work-in-progress and thus is purposefully not in GENERIC. And it conflicts with nve(4), so only one should be loaded.
2006-06-26 23:41:07 +00:00
if (sc->nfe_flags & NFE_40BIT_ADDR) {
Bring overhauled nfe(4) into tree. o s/printf/device_printf/g o Nuke OpenBSDism. o Nuke NetBSD/OpenBSD specific DMA sync operations.(we don't have a way to sync a single descriptor within a DMA map.) o Remove recursive mutex. o bus_dma(9) clean up. o 40bit DMA address support. o Add protection for Rx map load failure. o Fix a long standing bug for watchdog timeout. [1] o Add additional protections, missing Tx completion interrupt, losing start Tx command, for watchdog timeout. o Switch to taskqueue(9) API to handle interrupts. o Use our own timer for watchdog instead of if_watchdog/if_timer interface. o Advertise VLAN header length/capability correctly to upper layer. o Remove excessive kernel stack consumption in nfe_encap(). o Handle highly fragmented mbuf chains correctly. o Enable etherenet address reprogramming with ifconfig(8). o Add ALTQ/TSO, MSI/MSIX support. o Increased Rx ring to 256 descriptors from 128. o Align Tx/Rx descriptor ring on sizeof(struct nfe_desc64) boundary. o Remove alignment restrictions on Tx/Rx buffers. o Rewritten jumbo frame support code. o Add support for hardware assistend VLAN tag insertion/stripping. o Add support for Tx/Rx flow control based on patches from Peer Chen. [2] o Add a routine that detects whether ethernet address swap routines is required. [3] o Add a workaround that take MAC/PHY out of power down mode. o Add suspend/resume support. o style(9) and code clean up. Special thanks to Shigeaki Tagashira, the original porter of nfe(4), who submitted lots of patches, performed uncountable number of regression tests and maintained nfe(4) for a long time. Without his enthusiastic help and support I could never have completed this overhauling task. The only weak point of nfe(4) compared to nve(4) is instability of manual half-duplex media selection on certain hardwares(auto sensing media type should work for all cases, though). This was a long standing bug of nfe(4) and I still have no idea why it doesn't work on some hardwares. Obtained from: OpenBSD [1] Submitted by: Peer Chen < pchen at nvidia dot com > [2], [3] Reviewed by: Shigeaki Tagashira < shigeaki AT se DOT hiroshima-u DOT ac DOT jp > Tested by: Shigeaki Tagashira, current Discussed with: current Silence from: obrien
2007-06-12 02:16:02 +00:00
desc64 = &sc->jrxq.jdesc64[sc->jrxq.jcur];
vtag = le32toh(desc64->physaddr[1]);
flags = le16toh(desc64->flags);
len = le16toh(desc64->length) & NFE_RX_LEN_MASK;
Import the nfe (nForce Ethernet) driver from OpenBSD. This is a pure open source NIC driver with a BSD license. These bits were taken from the OpenBSD public CVS repository on 1-May-06.
2006-06-26 17:25:22 +00:00
} else {
Bring overhauled nfe(4) into tree. o s/printf/device_printf/g o Nuke OpenBSDism. o Nuke NetBSD/OpenBSD specific DMA sync operations.(we don't have a way to sync a single descriptor within a DMA map.) o Remove recursive mutex. o bus_dma(9) clean up. o 40bit DMA address support. o Add protection for Rx map load failure. o Fix a long standing bug for watchdog timeout. [1] o Add additional protections, missing Tx completion interrupt, losing start Tx command, for watchdog timeout. o Switch to taskqueue(9) API to handle interrupts. o Use our own timer for watchdog instead of if_watchdog/if_timer interface. o Advertise VLAN header length/capability correctly to upper layer. o Remove excessive kernel stack consumption in nfe_encap(). o Handle highly fragmented mbuf chains correctly. o Enable etherenet address reprogramming with ifconfig(8). o Add ALTQ/TSO, MSI/MSIX support. o Increased Rx ring to 256 descriptors from 128. o Align Tx/Rx descriptor ring on sizeof(struct nfe_desc64) boundary. o Remove alignment restrictions on Tx/Rx buffers. o Rewritten jumbo frame support code. o Add support for hardware assistend VLAN tag insertion/stripping. o Add support for Tx/Rx flow control based on patches from Peer Chen. [2] o Add a routine that detects whether ethernet address swap routines is required. [3] o Add a workaround that take MAC/PHY out of power down mode. o Add suspend/resume support. o style(9) and code clean up. Special thanks to Shigeaki Tagashira, the original porter of nfe(4), who submitted lots of patches, performed uncountable number of regression tests and maintained nfe(4) for a long time. Without his enthusiastic help and support I could never have completed this overhauling task. The only weak point of nfe(4) compared to nve(4) is instability of manual half-duplex media selection on certain hardwares(auto sensing media type should work for all cases, though). This was a long standing bug of nfe(4) and I still have no idea why it doesn't work on some hardwares. Obtained from: OpenBSD [1] Submitted by: Peer Chen < pchen at nvidia dot com > [2], [3] Reviewed by: Shigeaki Tagashira < shigeaki AT se DOT hiroshima-u DOT ac DOT jp > Tested by: Shigeaki Tagashira, current Discussed with: current Silence from: obrien
2007-06-12 02:16:02 +00:00
desc32 = &sc->jrxq.jdesc32[sc->jrxq.jcur];
flags = le16toh(desc32->flags);
len = le16toh(desc32->length) & NFE_RX_LEN_MASK;
Import the nfe (nForce Ethernet) driver from OpenBSD. This is a pure open source NIC driver with a BSD license. These bits were taken from the OpenBSD public CVS repository on 1-May-06.
2006-06-26 17:25:22 +00:00
}
if (flags & NFE_RX_READY)
break;
Bring overhauled nfe(4) into tree. o s/printf/device_printf/g o Nuke OpenBSDism. o Nuke NetBSD/OpenBSD specific DMA sync operations.(we don't have a way to sync a single descriptor within a DMA map.) o Remove recursive mutex. o bus_dma(9) clean up. o 40bit DMA address support. o Add protection for Rx map load failure. o Fix a long standing bug for watchdog timeout. [1] o Add additional protections, missing Tx completion interrupt, losing start Tx command, for watchdog timeout. o Switch to taskqueue(9) API to handle interrupts. o Use our own timer for watchdog instead of if_watchdog/if_timer interface. o Advertise VLAN header length/capability correctly to upper layer. o Remove excessive kernel stack consumption in nfe_encap(). o Handle highly fragmented mbuf chains correctly. o Enable etherenet address reprogramming with ifconfig(8). o Add ALTQ/TSO, MSI/MSIX support. o Increased Rx ring to 256 descriptors from 128. o Align Tx/Rx descriptor ring on sizeof(struct nfe_desc64) boundary. o Remove alignment restrictions on Tx/Rx buffers. o Rewritten jumbo frame support code. o Add support for hardware assistend VLAN tag insertion/stripping. o Add support for Tx/Rx flow control based on patches from Peer Chen. [2] o Add a routine that detects whether ethernet address swap routines is required. [3] o Add a workaround that take MAC/PHY out of power down mode. o Add suspend/resume support. o style(9) and code clean up. Special thanks to Shigeaki Tagashira, the original porter of nfe(4), who submitted lots of patches, performed uncountable number of regression tests and maintained nfe(4) for a long time. Without his enthusiastic help and support I could never have completed this overhauling task. The only weak point of nfe(4) compared to nve(4) is instability of manual half-duplex media selection on certain hardwares(auto sensing media type should work for all cases, though). This was a long standing bug of nfe(4) and I still have no idea why it doesn't work on some hardwares. Obtained from: OpenBSD [1] Submitted by: Peer Chen < pchen at nvidia dot com > [2], [3] Reviewed by: Shigeaki Tagashira < shigeaki AT se DOT hiroshima-u DOT ac DOT jp > Tested by: Shigeaki Tagashira, current Discussed with: current Silence from: obrien
2007-06-12 02:16:02 +00:00
prog++;
Add a pure open source nForce Ethernet driver, under BSDL. This driver was ported from OpenBSD by Shigeaki Tagashira <shigeaki@se.hiroshima-u.ac.jp> and posted at http://www.se.hiroshima-u.ac.jp/~shigeaki/software/freebsd-nfe.html It was additionally cleaned up by me. It is still a work-in-progress and thus is purposefully not in GENERIC. And it conflicts with nve(4), so only one should be loaded.
2006-06-26 23:41:07 +00:00
if ((sc->nfe_flags & (NFE_JUMBO_SUP | NFE_40BIT_ADDR)) == 0) {
Bring overhauled nfe(4) into tree. o s/printf/device_printf/g o Nuke OpenBSDism. o Nuke NetBSD/OpenBSD specific DMA sync operations.(we don't have a way to sync a single descriptor within a DMA map.) o Remove recursive mutex. o bus_dma(9) clean up. o 40bit DMA address support. o Add protection for Rx map load failure. o Fix a long standing bug for watchdog timeout. [1] o Add additional protections, missing Tx completion interrupt, losing start Tx command, for watchdog timeout. o Switch to taskqueue(9) API to handle interrupts. o Use our own timer for watchdog instead of if_watchdog/if_timer interface. o Advertise VLAN header length/capability correctly to upper layer. o Remove excessive kernel stack consumption in nfe_encap(). o Handle highly fragmented mbuf chains correctly. o Enable etherenet address reprogramming with ifconfig(8). o Add ALTQ/TSO, MSI/MSIX support. o Increased Rx ring to 256 descriptors from 128. o Align Tx/Rx descriptor ring on sizeof(struct nfe_desc64) boundary. o Remove alignment restrictions on Tx/Rx buffers. o Rewritten jumbo frame support code. o Add support for hardware assistend VLAN tag insertion/stripping. o Add support for Tx/Rx flow control based on patches from Peer Chen. [2] o Add a routine that detects whether ethernet address swap routines is required. [3] o Add a workaround that take MAC/PHY out of power down mode. o Add suspend/resume support. o style(9) and code clean up. Special thanks to Shigeaki Tagashira, the original porter of nfe(4), who submitted lots of patches, performed uncountable number of regression tests and maintained nfe(4) for a long time. Without his enthusiastic help and support I could never have completed this overhauling task. The only weak point of nfe(4) compared to nve(4) is instability of manual half-duplex media selection on certain hardwares(auto sensing media type should work for all cases, though). This was a long standing bug of nfe(4) and I still have no idea why it doesn't work on some hardwares. Obtained from: OpenBSD [1] Submitted by: Peer Chen < pchen at nvidia dot com > [2], [3] Reviewed by: Shigeaki Tagashira < shigeaki AT se DOT hiroshima-u DOT ac DOT jp > Tested by: Shigeaki Tagashira, current Discussed with: current Silence from: obrien
2007-06-12 02:16:02 +00:00
if (!(flags & NFE_RX_VALID_V1)) {
ifp->if_ierrors++;
nfe_discard_jrxbuf(sc, sc->jrxq.jcur);
continue;
}
Import the nfe (nForce Ethernet) driver from OpenBSD. This is a pure open source NIC driver with a BSD license. These bits were taken from the OpenBSD public CVS repository on 1-May-06.
2006-06-26 17:25:22 +00:00
if ((flags & NFE_RX_FIXME_V1) == NFE_RX_FIXME_V1) {
flags &= ~NFE_RX_ERROR;
len--; /* fix buffer length */
}
} else {
Bring overhauled nfe(4) into tree. o s/printf/device_printf/g o Nuke OpenBSDism. o Nuke NetBSD/OpenBSD specific DMA sync operations.(we don't have a way to sync a single descriptor within a DMA map.) o Remove recursive mutex. o bus_dma(9) clean up. o 40bit DMA address support. o Add protection for Rx map load failure. o Fix a long standing bug for watchdog timeout. [1] o Add additional protections, missing Tx completion interrupt, losing start Tx command, for watchdog timeout. o Switch to taskqueue(9) API to handle interrupts. o Use our own timer for watchdog instead of if_watchdog/if_timer interface. o Advertise VLAN header length/capability correctly to upper layer. o Remove excessive kernel stack consumption in nfe_encap(). o Handle highly fragmented mbuf chains correctly. o Enable etherenet address reprogramming with ifconfig(8). o Add ALTQ/TSO, MSI/MSIX support. o Increased Rx ring to 256 descriptors from 128. o Align Tx/Rx descriptor ring on sizeof(struct nfe_desc64) boundary. o Remove alignment restrictions on Tx/Rx buffers. o Rewritten jumbo frame support code. o Add support for hardware assistend VLAN tag insertion/stripping. o Add support for Tx/Rx flow control based on patches from Peer Chen. [2] o Add a routine that detects whether ethernet address swap routines is required. [3] o Add a workaround that take MAC/PHY out of power down mode. o Add suspend/resume support. o style(9) and code clean up. Special thanks to Shigeaki Tagashira, the original porter of nfe(4), who submitted lots of patches, performed uncountable number of regression tests and maintained nfe(4) for a long time. Without his enthusiastic help and support I could never have completed this overhauling task. The only weak point of nfe(4) compared to nve(4) is instability of manual half-duplex media selection on certain hardwares(auto sensing media type should work for all cases, though). This was a long standing bug of nfe(4) and I still have no idea why it doesn't work on some hardwares. Obtained from: OpenBSD [1] Submitted by: Peer Chen < pchen at nvidia dot com > [2], [3] Reviewed by: Shigeaki Tagashira < shigeaki AT se DOT hiroshima-u DOT ac DOT jp > Tested by: Shigeaki Tagashira, current Discussed with: current Silence from: obrien
2007-06-12 02:16:02 +00:00
if (!(flags & NFE_RX_VALID_V2)) {
ifp->if_ierrors++;
nfe_discard_jrxbuf(sc, sc->jrxq.jcur);
continue;
}
Import the nfe (nForce Ethernet) driver from OpenBSD. This is a pure open source NIC driver with a BSD license. These bits were taken from the OpenBSD public CVS repository on 1-May-06.
2006-06-26 17:25:22 +00:00
if ((flags & NFE_RX_FIXME_V2) == NFE_RX_FIXME_V2) {
flags &= ~NFE_RX_ERROR;
len--; /* fix buffer length */
}
}
if (flags & NFE_RX_ERROR) {
ifp->if_ierrors++;
Bring overhauled nfe(4) into tree. o s/printf/device_printf/g o Nuke OpenBSDism. o Nuke NetBSD/OpenBSD specific DMA sync operations.(we don't have a way to sync a single descriptor within a DMA map.) o Remove recursive mutex. o bus_dma(9) clean up. o 40bit DMA address support. o Add protection for Rx map load failure. o Fix a long standing bug for watchdog timeout. [1] o Add additional protections, missing Tx completion interrupt, losing start Tx command, for watchdog timeout. o Switch to taskqueue(9) API to handle interrupts. o Use our own timer for watchdog instead of if_watchdog/if_timer interface. o Advertise VLAN header length/capability correctly to upper layer. o Remove excessive kernel stack consumption in nfe_encap(). o Handle highly fragmented mbuf chains correctly. o Enable etherenet address reprogramming with ifconfig(8). o Add ALTQ/TSO, MSI/MSIX support. o Increased Rx ring to 256 descriptors from 128. o Align Tx/Rx descriptor ring on sizeof(struct nfe_desc64) boundary. o Remove alignment restrictions on Tx/Rx buffers. o Rewritten jumbo frame support code. o Add support for hardware assistend VLAN tag insertion/stripping. o Add support for Tx/Rx flow control based on patches from Peer Chen. [2] o Add a routine that detects whether ethernet address swap routines is required. [3] o Add a workaround that take MAC/PHY out of power down mode. o Add suspend/resume support. o style(9) and code clean up. Special thanks to Shigeaki Tagashira, the original porter of nfe(4), who submitted lots of patches, performed uncountable number of regression tests and maintained nfe(4) for a long time. Without his enthusiastic help and support I could never have completed this overhauling task. The only weak point of nfe(4) compared to nve(4) is instability of manual half-duplex media selection on certain hardwares(auto sensing media type should work for all cases, though). This was a long standing bug of nfe(4) and I still have no idea why it doesn't work on some hardwares. Obtained from: OpenBSD [1] Submitted by: Peer Chen < pchen at nvidia dot com > [2], [3] Reviewed by: Shigeaki Tagashira < shigeaki AT se DOT hiroshima-u DOT ac DOT jp > Tested by: Shigeaki Tagashira, current Discussed with: current Silence from: obrien
2007-06-12 02:16:02 +00:00
nfe_discard_jrxbuf(sc, sc->jrxq.jcur);
continue;
Import the nfe (nForce Ethernet) driver from OpenBSD. This is a pure open source NIC driver with a BSD license. These bits were taken from the OpenBSD public CVS repository on 1-May-06.
2006-06-26 17:25:22 +00:00
}
Bring overhauled nfe(4) into tree. o s/printf/device_printf/g o Nuke OpenBSDism. o Nuke NetBSD/OpenBSD specific DMA sync operations.(we don't have a way to sync a single descriptor within a DMA map.) o Remove recursive mutex. o bus_dma(9) clean up. o 40bit DMA address support. o Add protection for Rx map load failure. o Fix a long standing bug for watchdog timeout. [1] o Add additional protections, missing Tx completion interrupt, losing start Tx command, for watchdog timeout. o Switch to taskqueue(9) API to handle interrupts. o Use our own timer for watchdog instead of if_watchdog/if_timer interface. o Advertise VLAN header length/capability correctly to upper layer. o Remove excessive kernel stack consumption in nfe_encap(). o Handle highly fragmented mbuf chains correctly. o Enable etherenet address reprogramming with ifconfig(8). o Add ALTQ/TSO, MSI/MSIX support. o Increased Rx ring to 256 descriptors from 128. o Align Tx/Rx descriptor ring on sizeof(struct nfe_desc64) boundary. o Remove alignment restrictions on Tx/Rx buffers. o Rewritten jumbo frame support code. o Add support for hardware assistend VLAN tag insertion/stripping. o Add support for Tx/Rx flow control based on patches from Peer Chen. [2] o Add a routine that detects whether ethernet address swap routines is required. [3] o Add a workaround that take MAC/PHY out of power down mode. o Add suspend/resume support. o style(9) and code clean up. Special thanks to Shigeaki Tagashira, the original porter of nfe(4), who submitted lots of patches, performed uncountable number of regression tests and maintained nfe(4) for a long time. Without his enthusiastic help and support I could never have completed this overhauling task. The only weak point of nfe(4) compared to nve(4) is instability of manual half-duplex media selection on certain hardwares(auto sensing media type should work for all cases, though). This was a long standing bug of nfe(4) and I still have no idea why it doesn't work on some hardwares. Obtained from: OpenBSD [1] Submitted by: Peer Chen < pchen at nvidia dot com > [2], [3] Reviewed by: Shigeaki Tagashira < shigeaki AT se DOT hiroshima-u DOT ac DOT jp > Tested by: Shigeaki Tagashira, current Discussed with: current Silence from: obrien
2007-06-12 02:16:02 +00:00
m = data->m;
if (nfe_jnewbuf(sc, sc->jrxq.jcur) != 0) {
ifp->if_iqdrops++;
nfe_discard_jrxbuf(sc, sc->jrxq.jcur);
continue;
Reformat to KNF after revision 1.12.
2006-11-27 00:23:59 +00:00
}
Bring overhauled nfe(4) into tree. o s/printf/device_printf/g o Nuke OpenBSDism. o Nuke NetBSD/OpenBSD specific DMA sync operations.(we don't have a way to sync a single descriptor within a DMA map.) o Remove recursive mutex. o bus_dma(9) clean up. o 40bit DMA address support. o Add protection for Rx map load failure. o Fix a long standing bug for watchdog timeout. [1] o Add additional protections, missing Tx completion interrupt, losing start Tx command, for watchdog timeout. o Switch to taskqueue(9) API to handle interrupts. o Use our own timer for watchdog instead of if_watchdog/if_timer interface. o Advertise VLAN header length/capability correctly to upper layer. o Remove excessive kernel stack consumption in nfe_encap(). o Handle highly fragmented mbuf chains correctly. o Enable etherenet address reprogramming with ifconfig(8). o Add ALTQ/TSO, MSI/MSIX support. o Increased Rx ring to 256 descriptors from 128. o Align Tx/Rx descriptor ring on sizeof(struct nfe_desc64) boundary. o Remove alignment restrictions on Tx/Rx buffers. o Rewritten jumbo frame support code. o Add support for hardware assistend VLAN tag insertion/stripping. o Add support for Tx/Rx flow control based on patches from Peer Chen. [2] o Add a routine that detects whether ethernet address swap routines is required. [3] o Add a workaround that take MAC/PHY out of power down mode. o Add suspend/resume support. o style(9) and code clean up. Special thanks to Shigeaki Tagashira, the original porter of nfe(4), who submitted lots of patches, performed uncountable number of regression tests and maintained nfe(4) for a long time. Without his enthusiastic help and support I could never have completed this overhauling task. The only weak point of nfe(4) compared to nve(4) is instability of manual half-duplex media selection on certain hardwares(auto sensing media type should work for all cases, though). This was a long standing bug of nfe(4) and I still have no idea why it doesn't work on some hardwares. Obtained from: OpenBSD [1] Submitted by: Peer Chen < pchen at nvidia dot com > [2], [3] Reviewed by: Shigeaki Tagashira < shigeaki AT se DOT hiroshima-u DOT ac DOT jp > Tested by: Shigeaki Tagashira, current Discussed with: current Silence from: obrien
2007-06-12 02:16:02 +00:00
if ((vtag & NFE_RX_VTAG) != 0 &&
(ifp->if_capenable & IFCAP_VLAN_HWTAGGING) != 0) {
m->m_pkthdr.ether_vtag = vtag & 0xffff;
m->m_flags |= M_VLANTAG;
Reformat to KNF after revision 1.12.
2006-11-27 00:23:59 +00:00
}
Import the nfe (nForce Ethernet) driver from OpenBSD. This is a pure open source NIC driver with a BSD license. These bits were taken from the OpenBSD public CVS repository on 1-May-06.
2006-06-26 17:25:22 +00:00
m->m_pkthdr.len = m->m_len = len;
m->m_pkthdr.rcvif = ifp;
Bring overhauled nfe(4) into tree. o s/printf/device_printf/g o Nuke OpenBSDism. o Nuke NetBSD/OpenBSD specific DMA sync operations.(we don't have a way to sync a single descriptor within a DMA map.) o Remove recursive mutex. o bus_dma(9) clean up. o 40bit DMA address support. o Add protection for Rx map load failure. o Fix a long standing bug for watchdog timeout. [1] o Add additional protections, missing Tx completion interrupt, losing start Tx command, for watchdog timeout. o Switch to taskqueue(9) API to handle interrupts. o Use our own timer for watchdog instead of if_watchdog/if_timer interface. o Advertise VLAN header length/capability correctly to upper layer. o Remove excessive kernel stack consumption in nfe_encap(). o Handle highly fragmented mbuf chains correctly. o Enable etherenet address reprogramming with ifconfig(8). o Add ALTQ/TSO, MSI/MSIX support. o Increased Rx ring to 256 descriptors from 128. o Align Tx/Rx descriptor ring on sizeof(struct nfe_desc64) boundary. o Remove alignment restrictions on Tx/Rx buffers. o Rewritten jumbo frame support code. o Add support for hardware assistend VLAN tag insertion/stripping. o Add support for Tx/Rx flow control based on patches from Peer Chen. [2] o Add a routine that detects whether ethernet address swap routines is required. [3] o Add a workaround that take MAC/PHY out of power down mode. o Add suspend/resume support. o style(9) and code clean up. Special thanks to Shigeaki Tagashira, the original porter of nfe(4), who submitted lots of patches, performed uncountable number of regression tests and maintained nfe(4) for a long time. Without his enthusiastic help and support I could never have completed this overhauling task. The only weak point of nfe(4) compared to nve(4) is instability of manual half-duplex media selection on certain hardwares(auto sensing media type should work for all cases, though). This was a long standing bug of nfe(4) and I still have no idea why it doesn't work on some hardwares. Obtained from: OpenBSD [1] Submitted by: Peer Chen < pchen at nvidia dot com > [2], [3] Reviewed by: Shigeaki Tagashira < shigeaki AT se DOT hiroshima-u DOT ac DOT jp > Tested by: Shigeaki Tagashira, current Discussed with: current Silence from: obrien
2007-06-12 02:16:02 +00:00
if ((ifp->if_capenable & IFCAP_RXCSUM) != 0) {
if ((flags & NFE_RX_IP_CSUMOK) != 0) {
m->m_pkthdr.csum_flags |= CSUM_IP_CHECKED;
Add a pure open source nForce Ethernet driver, under BSDL. This driver was ported from OpenBSD by Shigeaki Tagashira <shigeaki@se.hiroshima-u.ac.jp> and posted at http://www.se.hiroshima-u.ac.jp/~shigeaki/software/freebsd-nfe.html It was additionally cleaned up by me. It is still a work-in-progress and thus is purposefully not in GENERIC. And it conflicts with nve(4), so only one should be loaded.
2006-06-26 23:41:07 +00:00
m->m_pkthdr.csum_flags |= CSUM_IP_VALID;
Bring overhauled nfe(4) into tree. o s/printf/device_printf/g o Nuke OpenBSDism. o Nuke NetBSD/OpenBSD specific DMA sync operations.(we don't have a way to sync a single descriptor within a DMA map.) o Remove recursive mutex. o bus_dma(9) clean up. o 40bit DMA address support. o Add protection for Rx map load failure. o Fix a long standing bug for watchdog timeout. [1] o Add additional protections, missing Tx completion interrupt, losing start Tx command, for watchdog timeout. o Switch to taskqueue(9) API to handle interrupts. o Use our own timer for watchdog instead of if_watchdog/if_timer interface. o Advertise VLAN header length/capability correctly to upper layer. o Remove excessive kernel stack consumption in nfe_encap(). o Handle highly fragmented mbuf chains correctly. o Enable etherenet address reprogramming with ifconfig(8). o Add ALTQ/TSO, MSI/MSIX support. o Increased Rx ring to 256 descriptors from 128. o Align Tx/Rx descriptor ring on sizeof(struct nfe_desc64) boundary. o Remove alignment restrictions on Tx/Rx buffers. o Rewritten jumbo frame support code. o Add support for hardware assistend VLAN tag insertion/stripping. o Add support for Tx/Rx flow control based on patches from Peer Chen. [2] o Add a routine that detects whether ethernet address swap routines is required. [3] o Add a workaround that take MAC/PHY out of power down mode. o Add suspend/resume support. o style(9) and code clean up. Special thanks to Shigeaki Tagashira, the original porter of nfe(4), who submitted lots of patches, performed uncountable number of regression tests and maintained nfe(4) for a long time. Without his enthusiastic help and support I could never have completed this overhauling task. The only weak point of nfe(4) compared to nve(4) is instability of manual half-duplex media selection on certain hardwares(auto sensing media type should work for all cases, though). This was a long standing bug of nfe(4) and I still have no idea why it doesn't work on some hardwares. Obtained from: OpenBSD [1] Submitted by: Peer Chen < pchen at nvidia dot com > [2], [3] Reviewed by: Shigeaki Tagashira < shigeaki AT se DOT hiroshima-u DOT ac DOT jp > Tested by: Shigeaki Tagashira, current Discussed with: current Silence from: obrien
2007-06-12 02:16:02 +00:00
if ((flags & NFE_RX_TCP_CSUMOK) != 0 ||
(flags & NFE_RX_UDP_CSUMOK) != 0) {
m->m_pkthdr.csum_flags |=
CSUM_DATA_VALID | CSUM_PSEUDO_HDR;
m->m_pkthdr.csum_data = 0xffff;
}
Add a pure open source nForce Ethernet driver, under BSDL. This driver was ported from OpenBSD by Shigeaki Tagashira <shigeaki@se.hiroshima-u.ac.jp> and posted at http://www.se.hiroshima-u.ac.jp/~shigeaki/software/freebsd-nfe.html It was additionally cleaned up by me. It is still a work-in-progress and thus is purposefully not in GENERIC. And it conflicts with nve(4), so only one should be loaded.
2006-06-26 23:41:07 +00:00
}
Import the nfe (nForce Ethernet) driver from OpenBSD. This is a pure open source NIC driver with a BSD license. These bits were taken from the OpenBSD public CVS repository on 1-May-06.
2006-06-26 17:25:22 +00:00
}
ifp->if_ipackets++;
Add a pure open source nForce Ethernet driver, under BSDL. This driver was ported from OpenBSD by Shigeaki Tagashira <shigeaki@se.hiroshima-u.ac.jp> and posted at http://www.se.hiroshima-u.ac.jp/~shigeaki/software/freebsd-nfe.html It was additionally cleaned up by me. It is still a work-in-progress and thus is purposefully not in GENERIC. And it conflicts with nve(4), so only one should be loaded.
2006-06-26 23:41:07 +00:00
NFE_UNLOCK(sc);
(*ifp->if_input)(ifp, m);
NFE_LOCK(sc);
When user_frac in the polling subsystem is low it is going to busy the CPU for too long period than necessary. Additively, interfaces are kept polled (in the tick) even if no more packets are available. In order to avoid such situations a new generic mechanism can be implemented in proactive way, keeping track of the time spent on any packet and fragmenting the time for any tick, stopping the processing as soon as possible. In order to implement such mechanism, the polling handler needs to change, returning the number of packets processed. While the intended logic is not part of this patch, the polling KPI is broken by this commit, adding an int return value and the new flag IFCAP_POLLING_NOCOUNT (which will signal that the return value is meaningless for the installed handler and checking should be skipped). Bump __FreeBSD_version in order to signal such situation. Reviewed by: emaste Sponsored by: Sandvine Incorporated
2009-05-30 15:14:44 +00:00
rx_npkts++;
Bring overhauled nfe(4) into tree. o s/printf/device_printf/g o Nuke OpenBSDism. o Nuke NetBSD/OpenBSD specific DMA sync operations.(we don't have a way to sync a single descriptor within a DMA map.) o Remove recursive mutex. o bus_dma(9) clean up. o 40bit DMA address support. o Add protection for Rx map load failure. o Fix a long standing bug for watchdog timeout. [1] o Add additional protections, missing Tx completion interrupt, losing start Tx command, for watchdog timeout. o Switch to taskqueue(9) API to handle interrupts. o Use our own timer for watchdog instead of if_watchdog/if_timer interface. o Advertise VLAN header length/capability correctly to upper layer. o Remove excessive kernel stack consumption in nfe_encap(). o Handle highly fragmented mbuf chains correctly. o Enable etherenet address reprogramming with ifconfig(8). o Add ALTQ/TSO, MSI/MSIX support. o Increased Rx ring to 256 descriptors from 128. o Align Tx/Rx descriptor ring on sizeof(struct nfe_desc64) boundary. o Remove alignment restrictions on Tx/Rx buffers. o Rewritten jumbo frame support code. o Add support for hardware assistend VLAN tag insertion/stripping. o Add support for Tx/Rx flow control based on patches from Peer Chen. [2] o Add a routine that detects whether ethernet address swap routines is required. [3] o Add a workaround that take MAC/PHY out of power down mode. o Add suspend/resume support. o style(9) and code clean up. Special thanks to Shigeaki Tagashira, the original porter of nfe(4), who submitted lots of patches, performed uncountable number of regression tests and maintained nfe(4) for a long time. Without his enthusiastic help and support I could never have completed this overhauling task. The only weak point of nfe(4) compared to nve(4) is instability of manual half-duplex media selection on certain hardwares(auto sensing media type should work for all cases, though). This was a long standing bug of nfe(4) and I still have no idea why it doesn't work on some hardwares. Obtained from: OpenBSD [1] Submitted by: Peer Chen < pchen at nvidia dot com > [2], [3] Reviewed by: Shigeaki Tagashira < shigeaki AT se DOT hiroshima-u DOT ac DOT jp > Tested by: Shigeaki Tagashira, current Discussed with: current Silence from: obrien
2007-06-12 02:16:02 +00:00
}
Import the nfe (nForce Ethernet) driver from OpenBSD. This is a pure open source NIC driver with a BSD license. These bits were taken from the OpenBSD public CVS repository on 1-May-06.
2006-06-26 17:25:22 +00:00
Bring overhauled nfe(4) into tree. o s/printf/device_printf/g o Nuke OpenBSDism. o Nuke NetBSD/OpenBSD specific DMA sync operations.(we don't have a way to sync a single descriptor within a DMA map.) o Remove recursive mutex. o bus_dma(9) clean up. o 40bit DMA address support. o Add protection for Rx map load failure. o Fix a long standing bug for watchdog timeout. [1] o Add additional protections, missing Tx completion interrupt, losing start Tx command, for watchdog timeout. o Switch to taskqueue(9) API to handle interrupts. o Use our own timer for watchdog instead of if_watchdog/if_timer interface. o Advertise VLAN header length/capability correctly to upper layer. o Remove excessive kernel stack consumption in nfe_encap(). o Handle highly fragmented mbuf chains correctly. o Enable etherenet address reprogramming with ifconfig(8). o Add ALTQ/TSO, MSI/MSIX support. o Increased Rx ring to 256 descriptors from 128. o Align Tx/Rx descriptor ring on sizeof(struct nfe_desc64) boundary. o Remove alignment restrictions on Tx/Rx buffers. o Rewritten jumbo frame support code. o Add support for hardware assistend VLAN tag insertion/stripping. o Add support for Tx/Rx flow control based on patches from Peer Chen. [2] o Add a routine that detects whether ethernet address swap routines is required. [3] o Add a workaround that take MAC/PHY out of power down mode. o Add suspend/resume support. o style(9) and code clean up. Special thanks to Shigeaki Tagashira, the original porter of nfe(4), who submitted lots of patches, performed uncountable number of regression tests and maintained nfe(4) for a long time. Without his enthusiastic help and support I could never have completed this overhauling task. The only weak point of nfe(4) compared to nve(4) is instability of manual half-duplex media selection on certain hardwares(auto sensing media type should work for all cases, though). This was a long standing bug of nfe(4) and I still have no idea why it doesn't work on some hardwares. Obtained from: OpenBSD [1] Submitted by: Peer Chen < pchen at nvidia dot com > [2], [3] Reviewed by: Shigeaki Tagashira < shigeaki AT se DOT hiroshima-u DOT ac DOT jp > Tested by: Shigeaki Tagashira, current Discussed with: current Silence from: obrien
2007-06-12 02:16:02 +00:00
if (prog > 0)
bus_dmamap_sync(sc->jrxq.jrx_desc_tag, sc->jrxq.jrx_desc_map,
BUS_DMASYNC_PREREAD | BUS_DMASYNC_PREWRITE);
Import the nfe (nForce Ethernet) driver from OpenBSD. This is a pure open source NIC driver with a BSD license. These bits were taken from the OpenBSD public CVS repository on 1-May-06.
2006-06-26 17:25:22 +00:00
When user_frac in the polling subsystem is low it is going to busy the CPU for too long period than necessary. Additively, interfaces are kept polled (in the tick) even if no more packets are available. In order to avoid such situations a new generic mechanism can be implemented in proactive way, keeping track of the time spent on any packet and fragmenting the time for any tick, stopping the processing as soon as possible. In order to implement such mechanism, the polling handler needs to change, returning the number of packets processed. While the intended logic is not part of this patch, the polling KPI is broken by this commit, adding an int return value and the new flag IFCAP_POLLING_NOCOUNT (which will signal that the return value is meaningless for the installed handler and checking should be skipped). Bump __FreeBSD_version in order to signal such situation. Reviewed by: emaste Sponsored by: Sandvine Incorporated
2009-05-30 15:14:44 +00:00
if (rx_npktsp != NULL)
*rx_npktsp = rx_npkts;
Bring overhauled nfe(4) into tree. o s/printf/device_printf/g o Nuke OpenBSDism. o Nuke NetBSD/OpenBSD specific DMA sync operations.(we don't have a way to sync a single descriptor within a DMA map.) o Remove recursive mutex. o bus_dma(9) clean up. o 40bit DMA address support. o Add protection for Rx map load failure. o Fix a long standing bug for watchdog timeout. [1] o Add additional protections, missing Tx completion interrupt, losing start Tx command, for watchdog timeout. o Switch to taskqueue(9) API to handle interrupts. o Use our own timer for watchdog instead of if_watchdog/if_timer interface. o Advertise VLAN header length/capability correctly to upper layer. o Remove excessive kernel stack consumption in nfe_encap(). o Handle highly fragmented mbuf chains correctly. o Enable etherenet address reprogramming with ifconfig(8). o Add ALTQ/TSO, MSI/MSIX support. o Increased Rx ring to 256 descriptors from 128. o Align Tx/Rx descriptor ring on sizeof(struct nfe_desc64) boundary. o Remove alignment restrictions on Tx/Rx buffers. o Rewritten jumbo frame support code. o Add support for hardware assistend VLAN tag insertion/stripping. o Add support for Tx/Rx flow control based on patches from Peer Chen. [2] o Add a routine that detects whether ethernet address swap routines is required. [3] o Add a workaround that take MAC/PHY out of power down mode. o Add suspend/resume support. o style(9) and code clean up. Special thanks to Shigeaki Tagashira, the original porter of nfe(4), who submitted lots of patches, performed uncountable number of regression tests and maintained nfe(4) for a long time. Without his enthusiastic help and support I could never have completed this overhauling task. The only weak point of nfe(4) compared to nve(4) is instability of manual half-duplex media selection on certain hardwares(auto sensing media type should work for all cases, though). This was a long standing bug of nfe(4) and I still have no idea why it doesn't work on some hardwares. Obtained from: OpenBSD [1] Submitted by: Peer Chen < pchen at nvidia dot com > [2], [3] Reviewed by: Shigeaki Tagashira < shigeaki AT se DOT hiroshima-u DOT ac DOT jp > Tested by: Shigeaki Tagashira, current Discussed with: current Silence from: obrien
2007-06-12 02:16:02 +00:00
return (count > 0 ? 0 : EAGAIN);
Import the nfe (nForce Ethernet) driver from OpenBSD. This is a pure open source NIC driver with a BSD license. These bits were taken from the OpenBSD public CVS repository on 1-May-06.
2006-06-26 17:25:22 +00:00
}
style(9) cleanup.
2006-10-19 10:01:26 +00:00
static void
nfe_txeof(struct nfe_softc *sc)
Import the nfe (nForce Ethernet) driver from OpenBSD. This is a pure open source NIC driver with a BSD license. These bits were taken from the OpenBSD public CVS repository on 1-May-06.
2006-06-26 17:25:22 +00:00
{
Add a pure open source nForce Ethernet driver, under BSDL. This driver was ported from OpenBSD by Shigeaki Tagashira <shigeaki@se.hiroshima-u.ac.jp> and posted at http://www.se.hiroshima-u.ac.jp/~shigeaki/software/freebsd-nfe.html It was additionally cleaned up by me. It is still a work-in-progress and thus is purposefully not in GENERIC. And it conflicts with nve(4), so only one should be loaded.
2006-06-26 23:41:07 +00:00
struct ifnet *ifp = sc->nfe_ifp;
Import the nfe (nForce Ethernet) driver from OpenBSD. This is a pure open source NIC driver with a BSD license. These bits were taken from the OpenBSD public CVS repository on 1-May-06.
2006-06-26 17:25:22 +00:00
struct nfe_desc32 *desc32;
struct nfe_desc64 *desc64;
struct nfe_tx_data *data = NULL;
Bring overhauled nfe(4) into tree. o s/printf/device_printf/g o Nuke OpenBSDism. o Nuke NetBSD/OpenBSD specific DMA sync operations.(we don't have a way to sync a single descriptor within a DMA map.) o Remove recursive mutex. o bus_dma(9) clean up. o 40bit DMA address support. o Add protection for Rx map load failure. o Fix a long standing bug for watchdog timeout. [1] o Add additional protections, missing Tx completion interrupt, losing start Tx command, for watchdog timeout. o Switch to taskqueue(9) API to handle interrupts. o Use our own timer for watchdog instead of if_watchdog/if_timer interface. o Advertise VLAN header length/capability correctly to upper layer. o Remove excessive kernel stack consumption in nfe_encap(). o Handle highly fragmented mbuf chains correctly. o Enable etherenet address reprogramming with ifconfig(8). o Add ALTQ/TSO, MSI/MSIX support. o Increased Rx ring to 256 descriptors from 128. o Align Tx/Rx descriptor ring on sizeof(struct nfe_desc64) boundary. o Remove alignment restrictions on Tx/Rx buffers. o Rewritten jumbo frame support code. o Add support for hardware assistend VLAN tag insertion/stripping. o Add support for Tx/Rx flow control based on patches from Peer Chen. [2] o Add a routine that detects whether ethernet address swap routines is required. [3] o Add a workaround that take MAC/PHY out of power down mode. o Add suspend/resume support. o style(9) and code clean up. Special thanks to Shigeaki Tagashira, the original porter of nfe(4), who submitted lots of patches, performed uncountable number of regression tests and maintained nfe(4) for a long time. Without his enthusiastic help and support I could never have completed this overhauling task. The only weak point of nfe(4) compared to nve(4) is instability of manual half-duplex media selection on certain hardwares(auto sensing media type should work for all cases, though). This was a long standing bug of nfe(4) and I still have no idea why it doesn't work on some hardwares. Obtained from: OpenBSD [1] Submitted by: Peer Chen < pchen at nvidia dot com > [2], [3] Reviewed by: Shigeaki Tagashira < shigeaki AT se DOT hiroshima-u DOT ac DOT jp > Tested by: Shigeaki Tagashira, current Discussed with: current Silence from: obrien
2007-06-12 02:16:02 +00:00
uint16_t flags;
int cons, prog;
Add a pure open source nForce Ethernet driver, under BSDL. This driver was ported from OpenBSD by Shigeaki Tagashira <shigeaki@se.hiroshima-u.ac.jp> and posted at http://www.se.hiroshima-u.ac.jp/~shigeaki/software/freebsd-nfe.html It was additionally cleaned up by me. It is still a work-in-progress and thus is purposefully not in GENERIC. And it conflicts with nve(4), so only one should be loaded.
2006-06-26 23:41:07 +00:00
NFE_LOCK_ASSERT(sc);
Import the nfe (nForce Ethernet) driver from OpenBSD. This is a pure open source NIC driver with a BSD license. These bits were taken from the OpenBSD public CVS repository on 1-May-06.
2006-06-26 17:25:22 +00:00
Bring overhauled nfe(4) into tree. o s/printf/device_printf/g o Nuke OpenBSDism. o Nuke NetBSD/OpenBSD specific DMA sync operations.(we don't have a way to sync a single descriptor within a DMA map.) o Remove recursive mutex. o bus_dma(9) clean up. o 40bit DMA address support. o Add protection for Rx map load failure. o Fix a long standing bug for watchdog timeout. [1] o Add additional protections, missing Tx completion interrupt, losing start Tx command, for watchdog timeout. o Switch to taskqueue(9) API to handle interrupts. o Use our own timer for watchdog instead of if_watchdog/if_timer interface. o Advertise VLAN header length/capability correctly to upper layer. o Remove excessive kernel stack consumption in nfe_encap(). o Handle highly fragmented mbuf chains correctly. o Enable etherenet address reprogramming with ifconfig(8). o Add ALTQ/TSO, MSI/MSIX support. o Increased Rx ring to 256 descriptors from 128. o Align Tx/Rx descriptor ring on sizeof(struct nfe_desc64) boundary. o Remove alignment restrictions on Tx/Rx buffers. o Rewritten jumbo frame support code. o Add support for hardware assistend VLAN tag insertion/stripping. o Add support for Tx/Rx flow control based on patches from Peer Chen. [2] o Add a routine that detects whether ethernet address swap routines is required. [3] o Add a workaround that take MAC/PHY out of power down mode. o Add suspend/resume support. o style(9) and code clean up. Special thanks to Shigeaki Tagashira, the original porter of nfe(4), who submitted lots of patches, performed uncountable number of regression tests and maintained nfe(4) for a long time. Without his enthusiastic help and support I could never have completed this overhauling task. The only weak point of nfe(4) compared to nve(4) is instability of manual half-duplex media selection on certain hardwares(auto sensing media type should work for all cases, though). This was a long standing bug of nfe(4) and I still have no idea why it doesn't work on some hardwares. Obtained from: OpenBSD [1] Submitted by: Peer Chen < pchen at nvidia dot com > [2], [3] Reviewed by: Shigeaki Tagashira < shigeaki AT se DOT hiroshima-u DOT ac DOT jp > Tested by: Shigeaki Tagashira, current Discussed with: current Silence from: obrien
2007-06-12 02:16:02 +00:00
bus_dmamap_sync(sc->txq.tx_desc_tag, sc->txq.tx_desc_map,
BUS_DMASYNC_POSTREAD);
Import the nfe (nForce Ethernet) driver from OpenBSD. This is a pure open source NIC driver with a BSD license. These bits were taken from the OpenBSD public CVS repository on 1-May-06.
2006-06-26 17:25:22 +00:00
Bring overhauled nfe(4) into tree. o s/printf/device_printf/g o Nuke OpenBSDism. o Nuke NetBSD/OpenBSD specific DMA sync operations.(we don't have a way to sync a single descriptor within a DMA map.) o Remove recursive mutex. o bus_dma(9) clean up. o 40bit DMA address support. o Add protection for Rx map load failure. o Fix a long standing bug for watchdog timeout. [1] o Add additional protections, missing Tx completion interrupt, losing start Tx command, for watchdog timeout. o Switch to taskqueue(9) API to handle interrupts. o Use our own timer for watchdog instead of if_watchdog/if_timer interface. o Advertise VLAN header length/capability correctly to upper layer. o Remove excessive kernel stack consumption in nfe_encap(). o Handle highly fragmented mbuf chains correctly. o Enable etherenet address reprogramming with ifconfig(8). o Add ALTQ/TSO, MSI/MSIX support. o Increased Rx ring to 256 descriptors from 128. o Align Tx/Rx descriptor ring on sizeof(struct nfe_desc64) boundary. o Remove alignment restrictions on Tx/Rx buffers. o Rewritten jumbo frame support code. o Add support for hardware assistend VLAN tag insertion/stripping. o Add support for Tx/Rx flow control based on patches from Peer Chen. [2] o Add a routine that detects whether ethernet address swap routines is required. [3] o Add a workaround that take MAC/PHY out of power down mode. o Add suspend/resume support. o style(9) and code clean up. Special thanks to Shigeaki Tagashira, the original porter of nfe(4), who submitted lots of patches, performed uncountable number of regression tests and maintained nfe(4) for a long time. Without his enthusiastic help and support I could never have completed this overhauling task. The only weak point of nfe(4) compared to nve(4) is instability of manual half-duplex media selection on certain hardwares(auto sensing media type should work for all cases, though). This was a long standing bug of nfe(4) and I still have no idea why it doesn't work on some hardwares. Obtained from: OpenBSD [1] Submitted by: Peer Chen < pchen at nvidia dot com > [2], [3] Reviewed by: Shigeaki Tagashira < shigeaki AT se DOT hiroshima-u DOT ac DOT jp > Tested by: Shigeaki Tagashira, current Discussed with: current Silence from: obrien
2007-06-12 02:16:02 +00:00
prog = 0;
for (cons = sc->txq.next; cons != sc->txq.cur;
NFE_INC(cons, NFE_TX_RING_COUNT)) {
if (sc->nfe_flags & NFE_40BIT_ADDR) {
desc64 = &sc->txq.desc64[cons];
flags = le16toh(desc64->flags);
Import the nfe (nForce Ethernet) driver from OpenBSD. This is a pure open source NIC driver with a BSD license. These bits were taken from the OpenBSD public CVS repository on 1-May-06.
2006-06-26 17:25:22 +00:00
} else {
Bring overhauled nfe(4) into tree. o s/printf/device_printf/g o Nuke OpenBSDism. o Nuke NetBSD/OpenBSD specific DMA sync operations.(we don't have a way to sync a single descriptor within a DMA map.) o Remove recursive mutex. o bus_dma(9) clean up. o 40bit DMA address support. o Add protection for Rx map load failure. o Fix a long standing bug for watchdog timeout. [1] o Add additional protections, missing Tx completion interrupt, losing start Tx command, for watchdog timeout. o Switch to taskqueue(9) API to handle interrupts. o Use our own timer for watchdog instead of if_watchdog/if_timer interface. o Advertise VLAN header length/capability correctly to upper layer. o Remove excessive kernel stack consumption in nfe_encap(). o Handle highly fragmented mbuf chains correctly. o Enable etherenet address reprogramming with ifconfig(8). o Add ALTQ/TSO, MSI/MSIX support. o Increased Rx ring to 256 descriptors from 128. o Align Tx/Rx descriptor ring on sizeof(struct nfe_desc64) boundary. o Remove alignment restrictions on Tx/Rx buffers. o Rewritten jumbo frame support code. o Add support for hardware assistend VLAN tag insertion/stripping. o Add support for Tx/Rx flow control based on patches from Peer Chen. [2] o Add a routine that detects whether ethernet address swap routines is required. [3] o Add a workaround that take MAC/PHY out of power down mode. o Add suspend/resume support. o style(9) and code clean up. Special thanks to Shigeaki Tagashira, the original porter of nfe(4), who submitted lots of patches, performed uncountable number of regression tests and maintained nfe(4) for a long time. Without his enthusiastic help and support I could never have completed this overhauling task. The only weak point of nfe(4) compared to nve(4) is instability of manual half-duplex media selection on certain hardwares(auto sensing media type should work for all cases, though). This was a long standing bug of nfe(4) and I still have no idea why it doesn't work on some hardwares. Obtained from: OpenBSD [1] Submitted by: Peer Chen < pchen at nvidia dot com > [2], [3] Reviewed by: Shigeaki Tagashira < shigeaki AT se DOT hiroshima-u DOT ac DOT jp > Tested by: Shigeaki Tagashira, current Discussed with: current Silence from: obrien
2007-06-12 02:16:02 +00:00
desc32 = &sc->txq.desc32[cons];
flags = le16toh(desc32->flags);
Import the nfe (nForce Ethernet) driver from OpenBSD. This is a pure open source NIC driver with a BSD license. These bits were taken from the OpenBSD public CVS repository on 1-May-06.
2006-06-26 17:25:22 +00:00
}
if (flags & NFE_TX_VALID)
break;
Bring overhauled nfe(4) into tree. o s/printf/device_printf/g o Nuke OpenBSDism. o Nuke NetBSD/OpenBSD specific DMA sync operations.(we don't have a way to sync a single descriptor within a DMA map.) o Remove recursive mutex. o bus_dma(9) clean up. o 40bit DMA address support. o Add protection for Rx map load failure. o Fix a long standing bug for watchdog timeout. [1] o Add additional protections, missing Tx completion interrupt, losing start Tx command, for watchdog timeout. o Switch to taskqueue(9) API to handle interrupts. o Use our own timer for watchdog instead of if_watchdog/if_timer interface. o Advertise VLAN header length/capability correctly to upper layer. o Remove excessive kernel stack consumption in nfe_encap(). o Handle highly fragmented mbuf chains correctly. o Enable etherenet address reprogramming with ifconfig(8). o Add ALTQ/TSO, MSI/MSIX support. o Increased Rx ring to 256 descriptors from 128. o Align Tx/Rx descriptor ring on sizeof(struct nfe_desc64) boundary. o Remove alignment restrictions on Tx/Rx buffers. o Rewritten jumbo frame support code. o Add support for hardware assistend VLAN tag insertion/stripping. o Add support for Tx/Rx flow control based on patches from Peer Chen. [2] o Add a routine that detects whether ethernet address swap routines is required. [3] o Add a workaround that take MAC/PHY out of power down mode. o Add suspend/resume support. o style(9) and code clean up. Special thanks to Shigeaki Tagashira, the original porter of nfe(4), who submitted lots of patches, performed uncountable number of regression tests and maintained nfe(4) for a long time. Without his enthusiastic help and support I could never have completed this overhauling task. The only weak point of nfe(4) compared to nve(4) is instability of manual half-duplex media selection on certain hardwares(auto sensing media type should work for all cases, though). This was a long standing bug of nfe(4) and I still have no idea why it doesn't work on some hardwares. Obtained from: OpenBSD [1] Submitted by: Peer Chen < pchen at nvidia dot com > [2], [3] Reviewed by: Shigeaki Tagashira < shigeaki AT se DOT hiroshima-u DOT ac DOT jp > Tested by: Shigeaki Tagashira, current Discussed with: current Silence from: obrien
2007-06-12 02:16:02 +00:00
prog++;
sc->txq.queued--;
data = &sc->txq.data[cons];
Import the nfe (nForce Ethernet) driver from OpenBSD. This is a pure open source NIC driver with a BSD license. These bits were taken from the OpenBSD public CVS repository on 1-May-06.
2006-06-26 17:25:22 +00:00
Add a pure open source nForce Ethernet driver, under BSDL. This driver was ported from OpenBSD by Shigeaki Tagashira <shigeaki@se.hiroshima-u.ac.jp> and posted at http://www.se.hiroshima-u.ac.jp/~shigeaki/software/freebsd-nfe.html It was additionally cleaned up by me. It is still a work-in-progress and thus is purposefully not in GENERIC. And it conflicts with nve(4), so only one should be loaded.
2006-06-26 23:41:07 +00:00
if ((sc->nfe_flags & (NFE_JUMBO_SUP | NFE_40BIT_ADDR)) == 0) {
Bring overhauled nfe(4) into tree. o s/printf/device_printf/g o Nuke OpenBSDism. o Nuke NetBSD/OpenBSD specific DMA sync operations.(we don't have a way to sync a single descriptor within a DMA map.) o Remove recursive mutex. o bus_dma(9) clean up. o 40bit DMA address support. o Add protection for Rx map load failure. o Fix a long standing bug for watchdog timeout. [1] o Add additional protections, missing Tx completion interrupt, losing start Tx command, for watchdog timeout. o Switch to taskqueue(9) API to handle interrupts. o Use our own timer for watchdog instead of if_watchdog/if_timer interface. o Advertise VLAN header length/capability correctly to upper layer. o Remove excessive kernel stack consumption in nfe_encap(). o Handle highly fragmented mbuf chains correctly. o Enable etherenet address reprogramming with ifconfig(8). o Add ALTQ/TSO, MSI/MSIX support. o Increased Rx ring to 256 descriptors from 128. o Align Tx/Rx descriptor ring on sizeof(struct nfe_desc64) boundary. o Remove alignment restrictions on Tx/Rx buffers. o Rewritten jumbo frame support code. o Add support for hardware assistend VLAN tag insertion/stripping. o Add support for Tx/Rx flow control based on patches from Peer Chen. [2] o Add a routine that detects whether ethernet address swap routines is required. [3] o Add a workaround that take MAC/PHY out of power down mode. o Add suspend/resume support. o style(9) and code clean up. Special thanks to Shigeaki Tagashira, the original porter of nfe(4), who submitted lots of patches, performed uncountable number of regression tests and maintained nfe(4) for a long time. Without his enthusiastic help and support I could never have completed this overhauling task. The only weak point of nfe(4) compared to nve(4) is instability of manual half-duplex media selection on certain hardwares(auto sensing media type should work for all cases, though). This was a long standing bug of nfe(4) and I still have no idea why it doesn't work on some hardwares. Obtained from: OpenBSD [1] Submitted by: Peer Chen < pchen at nvidia dot com > [2], [3] Reviewed by: Shigeaki Tagashira < shigeaki AT se DOT hiroshima-u DOT ac DOT jp > Tested by: Shigeaki Tagashira, current Discussed with: current Silence from: obrien
2007-06-12 02:16:02 +00:00
if ((flags & NFE_TX_LASTFRAG_V1) == 0)
continue;
Import the nfe (nForce Ethernet) driver from OpenBSD. This is a pure open source NIC driver with a BSD license. These bits were taken from the OpenBSD public CVS repository on 1-May-06.
2006-06-26 17:25:22 +00:00
if ((flags & NFE_TX_ERROR_V1) != 0) {
Bring overhauled nfe(4) into tree. o s/printf/device_printf/g o Nuke OpenBSDism. o Nuke NetBSD/OpenBSD specific DMA sync operations.(we don't have a way to sync a single descriptor within a DMA map.) o Remove recursive mutex. o bus_dma(9) clean up. o 40bit DMA address support. o Add protection for Rx map load failure. o Fix a long standing bug for watchdog timeout. [1] o Add additional protections, missing Tx completion interrupt, losing start Tx command, for watchdog timeout. o Switch to taskqueue(9) API to handle interrupts. o Use our own timer for watchdog instead of if_watchdog/if_timer interface. o Advertise VLAN header length/capability correctly to upper layer. o Remove excessive kernel stack consumption in nfe_encap(). o Handle highly fragmented mbuf chains correctly. o Enable etherenet address reprogramming with ifconfig(8). o Add ALTQ/TSO, MSI/MSIX support. o Increased Rx ring to 256 descriptors from 128. o Align Tx/Rx descriptor ring on sizeof(struct nfe_desc64) boundary. o Remove alignment restrictions on Tx/Rx buffers. o Rewritten jumbo frame support code. o Add support for hardware assistend VLAN tag insertion/stripping. o Add support for Tx/Rx flow control based on patches from Peer Chen. [2] o Add a routine that detects whether ethernet address swap routines is required. [3] o Add a workaround that take MAC/PHY out of power down mode. o Add suspend/resume support. o style(9) and code clean up. Special thanks to Shigeaki Tagashira, the original porter of nfe(4), who submitted lots of patches, performed uncountable number of regression tests and maintained nfe(4) for a long time. Without his enthusiastic help and support I could never have completed this overhauling task. The only weak point of nfe(4) compared to nve(4) is instability of manual half-duplex media selection on certain hardwares(auto sensing media type should work for all cases, though). This was a long standing bug of nfe(4) and I still have no idea why it doesn't work on some hardwares. Obtained from: OpenBSD [1] Submitted by: Peer Chen < pchen at nvidia dot com > [2], [3] Reviewed by: Shigeaki Tagashira < shigeaki AT se DOT hiroshima-u DOT ac DOT jp > Tested by: Shigeaki Tagashira, current Discussed with: current Silence from: obrien
2007-06-12 02:16:02 +00:00
device_printf(sc->nfe_dev,
"tx v1 error 0x%4b\n", flags, NFE_V1_TXERR);
Add a pure open source nForce Ethernet driver, under BSDL. This driver was ported from OpenBSD by Shigeaki Tagashira <shigeaki@se.hiroshima-u.ac.jp> and posted at http://www.se.hiroshima-u.ac.jp/~shigeaki/software/freebsd-nfe.html It was additionally cleaned up by me. It is still a work-in-progress and thus is purposefully not in GENERIC. And it conflicts with nve(4), so only one should be loaded.
2006-06-26 23:41:07 +00:00
Import the nfe (nForce Ethernet) driver from OpenBSD. This is a pure open source NIC driver with a BSD license. These bits were taken from the OpenBSD public CVS repository on 1-May-06.
2006-06-26 17:25:22 +00:00
ifp->if_oerrors++;
} else
ifp->if_opackets++;
} else {
Bring overhauled nfe(4) into tree. o s/printf/device_printf/g o Nuke OpenBSDism. o Nuke NetBSD/OpenBSD specific DMA sync operations.(we don't have a way to sync a single descriptor within a DMA map.) o Remove recursive mutex. o bus_dma(9) clean up. o 40bit DMA address support. o Add protection for Rx map load failure. o Fix a long standing bug for watchdog timeout. [1] o Add additional protections, missing Tx completion interrupt, losing start Tx command, for watchdog timeout. o Switch to taskqueue(9) API to handle interrupts. o Use our own timer for watchdog instead of if_watchdog/if_timer interface. o Advertise VLAN header length/capability correctly to upper layer. o Remove excessive kernel stack consumption in nfe_encap(). o Handle highly fragmented mbuf chains correctly. o Enable etherenet address reprogramming with ifconfig(8). o Add ALTQ/TSO, MSI/MSIX support. o Increased Rx ring to 256 descriptors from 128. o Align Tx/Rx descriptor ring on sizeof(struct nfe_desc64) boundary. o Remove alignment restrictions on Tx/Rx buffers. o Rewritten jumbo frame support code. o Add support for hardware assistend VLAN tag insertion/stripping. o Add support for Tx/Rx flow control based on patches from Peer Chen. [2] o Add a routine that detects whether ethernet address swap routines is required. [3] o Add a workaround that take MAC/PHY out of power down mode. o Add suspend/resume support. o style(9) and code clean up. Special thanks to Shigeaki Tagashira, the original porter of nfe(4), who submitted lots of patches, performed uncountable number of regression tests and maintained nfe(4) for a long time. Without his enthusiastic help and support I could never have completed this overhauling task. The only weak point of nfe(4) compared to nve(4) is instability of manual half-duplex media selection on certain hardwares(auto sensing media type should work for all cases, though). This was a long standing bug of nfe(4) and I still have no idea why it doesn't work on some hardwares. Obtained from: OpenBSD [1] Submitted by: Peer Chen < pchen at nvidia dot com > [2], [3] Reviewed by: Shigeaki Tagashira < shigeaki AT se DOT hiroshima-u DOT ac DOT jp > Tested by: Shigeaki Tagashira, current Discussed with: current Silence from: obrien
2007-06-12 02:16:02 +00:00
if ((flags & NFE_TX_LASTFRAG_V2) == 0)
continue;
Import the nfe (nForce Ethernet) driver from OpenBSD. This is a pure open source NIC driver with a BSD license. These bits were taken from the OpenBSD public CVS repository on 1-May-06.
2006-06-26 17:25:22 +00:00
if ((flags & NFE_TX_ERROR_V2) != 0) {
Bring overhauled nfe(4) into tree. o s/printf/device_printf/g o Nuke OpenBSDism. o Nuke NetBSD/OpenBSD specific DMA sync operations.(we don't have a way to sync a single descriptor within a DMA map.) o Remove recursive mutex. o bus_dma(9) clean up. o 40bit DMA address support. o Add protection for Rx map load failure. o Fix a long standing bug for watchdog timeout. [1] o Add additional protections, missing Tx completion interrupt, losing start Tx command, for watchdog timeout. o Switch to taskqueue(9) API to handle interrupts. o Use our own timer for watchdog instead of if_watchdog/if_timer interface. o Advertise VLAN header length/capability correctly to upper layer. o Remove excessive kernel stack consumption in nfe_encap(). o Handle highly fragmented mbuf chains correctly. o Enable etherenet address reprogramming with ifconfig(8). o Add ALTQ/TSO, MSI/MSIX support. o Increased Rx ring to 256 descriptors from 128. o Align Tx/Rx descriptor ring on sizeof(struct nfe_desc64) boundary. o Remove alignment restrictions on Tx/Rx buffers. o Rewritten jumbo frame support code. o Add support for hardware assistend VLAN tag insertion/stripping. o Add support for Tx/Rx flow control based on patches from Peer Chen. [2] o Add a routine that detects whether ethernet address swap routines is required. [3] o Add a workaround that take MAC/PHY out of power down mode. o Add suspend/resume support. o style(9) and code clean up. Special thanks to Shigeaki Tagashira, the original porter of nfe(4), who submitted lots of patches, performed uncountable number of regression tests and maintained nfe(4) for a long time. Without his enthusiastic help and support I could never have completed this overhauling task. The only weak point of nfe(4) compared to nve(4) is instability of manual half-duplex media selection on certain hardwares(auto sensing media type should work for all cases, though). This was a long standing bug of nfe(4) and I still have no idea why it doesn't work on some hardwares. Obtained from: OpenBSD [1] Submitted by: Peer Chen < pchen at nvidia dot com > [2], [3] Reviewed by: Shigeaki Tagashira < shigeaki AT se DOT hiroshima-u DOT ac DOT jp > Tested by: Shigeaki Tagashira, current Discussed with: current Silence from: obrien
2007-06-12 02:16:02 +00:00
device_printf(sc->nfe_dev,
"tx v2 error 0x%4b\n", flags, NFE_V2_TXERR);
Import the nfe (nForce Ethernet) driver from OpenBSD. This is a pure open source NIC driver with a BSD license. These bits were taken from the OpenBSD public CVS repository on 1-May-06.
2006-06-26 17:25:22 +00:00
ifp->if_oerrors++;
} else
ifp->if_opackets++;
}
/* last fragment of the mbuf chain transmitted */
Bring overhauled nfe(4) into tree. o s/printf/device_printf/g o Nuke OpenBSDism. o Nuke NetBSD/OpenBSD specific DMA sync operations.(we don't have a way to sync a single descriptor within a DMA map.) o Remove recursive mutex. o bus_dma(9) clean up. o 40bit DMA address support. o Add protection for Rx map load failure. o Fix a long standing bug for watchdog timeout. [1] o Add additional protections, missing Tx completion interrupt, losing start Tx command, for watchdog timeout. o Switch to taskqueue(9) API to handle interrupts. o Use our own timer for watchdog instead of if_watchdog/if_timer interface. o Advertise VLAN header length/capability correctly to upper layer. o Remove excessive kernel stack consumption in nfe_encap(). o Handle highly fragmented mbuf chains correctly. o Enable etherenet address reprogramming with ifconfig(8). o Add ALTQ/TSO, MSI/MSIX support. o Increased Rx ring to 256 descriptors from 128. o Align Tx/Rx descriptor ring on sizeof(struct nfe_desc64) boundary. o Remove alignment restrictions on Tx/Rx buffers. o Rewritten jumbo frame support code. o Add support for hardware assistend VLAN tag insertion/stripping. o Add support for Tx/Rx flow control based on patches from Peer Chen. [2] o Add a routine that detects whether ethernet address swap routines is required. [3] o Add a workaround that take MAC/PHY out of power down mode. o Add suspend/resume support. o style(9) and code clean up. Special thanks to Shigeaki Tagashira, the original porter of nfe(4), who submitted lots of patches, performed uncountable number of regression tests and maintained nfe(4) for a long time. Without his enthusiastic help and support I could never have completed this overhauling task. The only weak point of nfe(4) compared to nve(4) is instability of manual half-duplex media selection on certain hardwares(auto sensing media type should work for all cases, though). This was a long standing bug of nfe(4) and I still have no idea why it doesn't work on some hardwares. Obtained from: OpenBSD [1] Submitted by: Peer Chen < pchen at nvidia dot com > [2], [3] Reviewed by: Shigeaki Tagashira < shigeaki AT se DOT hiroshima-u DOT ac DOT jp > Tested by: Shigeaki Tagashira, current Discussed with: current Silence from: obrien
2007-06-12 02:16:02 +00:00
KASSERT(data->m != NULL, ("%s: freeing NULL mbuf!", __func__));
bus_dmamap_sync(sc->txq.tx_data_tag, data->tx_data_map,
Add a pure open source nForce Ethernet driver, under BSDL. This driver was ported from OpenBSD by Shigeaki Tagashira <shigeaki@se.hiroshima-u.ac.jp> and posted at http://www.se.hiroshima-u.ac.jp/~shigeaki/software/freebsd-nfe.html It was additionally cleaned up by me. It is still a work-in-progress and thus is purposefully not in GENERIC. And it conflicts with nve(4), so only one should be loaded.
2006-06-26 23:41:07 +00:00
BUS_DMASYNC_POSTWRITE);
Bring overhauled nfe(4) into tree. o s/printf/device_printf/g o Nuke OpenBSDism. o Nuke NetBSD/OpenBSD specific DMA sync operations.(we don't have a way to sync a single descriptor within a DMA map.) o Remove recursive mutex. o bus_dma(9) clean up. o 40bit DMA address support. o Add protection for Rx map load failure. o Fix a long standing bug for watchdog timeout. [1] o Add additional protections, missing Tx completion interrupt, losing start Tx command, for watchdog timeout. o Switch to taskqueue(9) API to handle interrupts. o Use our own timer for watchdog instead of if_watchdog/if_timer interface. o Advertise VLAN header length/capability correctly to upper layer. o Remove excessive kernel stack consumption in nfe_encap(). o Handle highly fragmented mbuf chains correctly. o Enable etherenet address reprogramming with ifconfig(8). o Add ALTQ/TSO, MSI/MSIX support. o Increased Rx ring to 256 descriptors from 128. o Align Tx/Rx descriptor ring on sizeof(struct nfe_desc64) boundary. o Remove alignment restrictions on Tx/Rx buffers. o Rewritten jumbo frame support code. o Add support for hardware assistend VLAN tag insertion/stripping. o Add support for Tx/Rx flow control based on patches from Peer Chen. [2] o Add a routine that detects whether ethernet address swap routines is required. [3] o Add a workaround that take MAC/PHY out of power down mode. o Add suspend/resume support. o style(9) and code clean up. Special thanks to Shigeaki Tagashira, the original porter of nfe(4), who submitted lots of patches, performed uncountable number of regression tests and maintained nfe(4) for a long time. Without his enthusiastic help and support I could never have completed this overhauling task. The only weak point of nfe(4) compared to nve(4) is instability of manual half-duplex media selection on certain hardwares(auto sensing media type should work for all cases, though). This was a long standing bug of nfe(4) and I still have no idea why it doesn't work on some hardwares. Obtained from: OpenBSD [1] Submitted by: Peer Chen < pchen at nvidia dot com > [2], [3] Reviewed by: Shigeaki Tagashira < shigeaki AT se DOT hiroshima-u DOT ac DOT jp > Tested by: Shigeaki Tagashira, current Discussed with: current Silence from: obrien
2007-06-12 02:16:02 +00:00
bus_dmamap_unload(sc->txq.tx_data_tag, data->tx_data_map);
Import the nfe (nForce Ethernet) driver from OpenBSD. This is a pure open source NIC driver with a BSD license. These bits were taken from the OpenBSD public CVS repository on 1-May-06.
2006-06-26 17:25:22 +00:00
m_freem(data->m);
data->m = NULL;
}
Bring overhauled nfe(4) into tree. o s/printf/device_printf/g o Nuke OpenBSDism. o Nuke NetBSD/OpenBSD specific DMA sync operations.(we don't have a way to sync a single descriptor within a DMA map.) o Remove recursive mutex. o bus_dma(9) clean up. o 40bit DMA address support. o Add protection for Rx map load failure. o Fix a long standing bug for watchdog timeout. [1] o Add additional protections, missing Tx completion interrupt, losing start Tx command, for watchdog timeout. o Switch to taskqueue(9) API to handle interrupts. o Use our own timer for watchdog instead of if_watchdog/if_timer interface. o Advertise VLAN header length/capability correctly to upper layer. o Remove excessive kernel stack consumption in nfe_encap(). o Handle highly fragmented mbuf chains correctly. o Enable etherenet address reprogramming with ifconfig(8). o Add ALTQ/TSO, MSI/MSIX support. o Increased Rx ring to 256 descriptors from 128. o Align Tx/Rx descriptor ring on sizeof(struct nfe_desc64) boundary. o Remove alignment restrictions on Tx/Rx buffers. o Rewritten jumbo frame support code. o Add support for hardware assistend VLAN tag insertion/stripping. o Add support for Tx/Rx flow control based on patches from Peer Chen. [2] o Add a routine that detects whether ethernet address swap routines is required. [3] o Add a workaround that take MAC/PHY out of power down mode. o Add suspend/resume support. o style(9) and code clean up. Special thanks to Shigeaki Tagashira, the original porter of nfe(4), who submitted lots of patches, performed uncountable number of regression tests and maintained nfe(4) for a long time. Without his enthusiastic help and support I could never have completed this overhauling task. The only weak point of nfe(4) compared to nve(4) is instability of manual half-duplex media selection on certain hardwares(auto sensing media type should work for all cases, though). This was a long standing bug of nfe(4) and I still have no idea why it doesn't work on some hardwares. Obtained from: OpenBSD [1] Submitted by: Peer Chen < pchen at nvidia dot com > [2], [3] Reviewed by: Shigeaki Tagashira < shigeaki AT se DOT hiroshima-u DOT ac DOT jp > Tested by: Shigeaki Tagashira, current Discussed with: current Silence from: obrien
2007-06-12 02:16:02 +00:00
if (prog > 0) {
sc->nfe_force_tx = 0;
sc->txq.next = cons;
Add a pure open source nForce Ethernet driver, under BSDL. This driver was ported from OpenBSD by Shigeaki Tagashira <shigeaki@se.hiroshima-u.ac.jp> and posted at http://www.se.hiroshima-u.ac.jp/~shigeaki/software/freebsd-nfe.html It was additionally cleaned up by me. It is still a work-in-progress and thus is purposefully not in GENERIC. And it conflicts with nve(4), so only one should be loaded.
2006-06-26 23:41:07 +00:00
ifp->if_drv_flags &= ~IFF_DRV_OACTIVE;
Bring overhauled nfe(4) into tree. o s/printf/device_printf/g o Nuke OpenBSDism. o Nuke NetBSD/OpenBSD specific DMA sync operations.(we don't have a way to sync a single descriptor within a DMA map.) o Remove recursive mutex. o bus_dma(9) clean up. o 40bit DMA address support. o Add protection for Rx map load failure. o Fix a long standing bug for watchdog timeout. [1] o Add additional protections, missing Tx completion interrupt, losing start Tx command, for watchdog timeout. o Switch to taskqueue(9) API to handle interrupts. o Use our own timer for watchdog instead of if_watchdog/if_timer interface. o Advertise VLAN header length/capability correctly to upper layer. o Remove excessive kernel stack consumption in nfe_encap(). o Handle highly fragmented mbuf chains correctly. o Enable etherenet address reprogramming with ifconfig(8). o Add ALTQ/TSO, MSI/MSIX support. o Increased Rx ring to 256 descriptors from 128. o Align Tx/Rx descriptor ring on sizeof(struct nfe_desc64) boundary. o Remove alignment restrictions on Tx/Rx buffers. o Rewritten jumbo frame support code. o Add support for hardware assistend VLAN tag insertion/stripping. o Add support for Tx/Rx flow control based on patches from Peer Chen. [2] o Add a routine that detects whether ethernet address swap routines is required. [3] o Add a workaround that take MAC/PHY out of power down mode. o Add suspend/resume support. o style(9) and code clean up. Special thanks to Shigeaki Tagashira, the original porter of nfe(4), who submitted lots of patches, performed uncountable number of regression tests and maintained nfe(4) for a long time. Without his enthusiastic help and support I could never have completed this overhauling task. The only weak point of nfe(4) compared to nve(4) is instability of manual half-duplex media selection on certain hardwares(auto sensing media type should work for all cases, though). This was a long standing bug of nfe(4) and I still have no idea why it doesn't work on some hardwares. Obtained from: OpenBSD [1] Submitted by: Peer Chen < pchen at nvidia dot com > [2], [3] Reviewed by: Shigeaki Tagashira < shigeaki AT se DOT hiroshima-u DOT ac DOT jp > Tested by: Shigeaki Tagashira, current Discussed with: current Silence from: obrien
2007-06-12 02:16:02 +00:00
if (sc->txq.queued == 0)
sc->nfe_watchdog_timer = 0;
}
}
style(9) cleanup.
2006-10-19 10:01:26 +00:00
static int
Bring overhauled nfe(4) into tree. o s/printf/device_printf/g o Nuke OpenBSDism. o Nuke NetBSD/OpenBSD specific DMA sync operations.(we don't have a way to sync a single descriptor within a DMA map.) o Remove recursive mutex. o bus_dma(9) clean up. o 40bit DMA address support. o Add protection for Rx map load failure. o Fix a long standing bug for watchdog timeout. [1] o Add additional protections, missing Tx completion interrupt, losing start Tx command, for watchdog timeout. o Switch to taskqueue(9) API to handle interrupts. o Use our own timer for watchdog instead of if_watchdog/if_timer interface. o Advertise VLAN header length/capability correctly to upper layer. o Remove excessive kernel stack consumption in nfe_encap(). o Handle highly fragmented mbuf chains correctly. o Enable etherenet address reprogramming with ifconfig(8). o Add ALTQ/TSO, MSI/MSIX support. o Increased Rx ring to 256 descriptors from 128. o Align Tx/Rx descriptor ring on sizeof(struct nfe_desc64) boundary. o Remove alignment restrictions on Tx/Rx buffers. o Rewritten jumbo frame support code. o Add support for hardware assistend VLAN tag insertion/stripping. o Add support for Tx/Rx flow control based on patches from Peer Chen. [2] o Add a routine that detects whether ethernet address swap routines is required. [3] o Add a workaround that take MAC/PHY out of power down mode. o Add suspend/resume support. o style(9) and code clean up. Special thanks to Shigeaki Tagashira, the original porter of nfe(4), who submitted lots of patches, performed uncountable number of regression tests and maintained nfe(4) for a long time. Without his enthusiastic help and support I could never have completed this overhauling task. The only weak point of nfe(4) compared to nve(4) is instability of manual half-duplex media selection on certain hardwares(auto sensing media type should work for all cases, though). This was a long standing bug of nfe(4) and I still have no idea why it doesn't work on some hardwares. Obtained from: OpenBSD [1] Submitted by: Peer Chen < pchen at nvidia dot com > [2], [3] Reviewed by: Shigeaki Tagashira < shigeaki AT se DOT hiroshima-u DOT ac DOT jp > Tested by: Shigeaki Tagashira, current Discussed with: current Silence from: obrien
2007-06-12 02:16:02 +00:00
nfe_encap(struct nfe_softc *sc, struct mbuf **m_head)
Import the nfe (nForce Ethernet) driver from OpenBSD. This is a pure open source NIC driver with a BSD license. These bits were taken from the OpenBSD public CVS repository on 1-May-06.
2006-06-26 17:25:22 +00:00
{
Bring overhauled nfe(4) into tree. o s/printf/device_printf/g o Nuke OpenBSDism. o Nuke NetBSD/OpenBSD specific DMA sync operations.(we don't have a way to sync a single descriptor within a DMA map.) o Remove recursive mutex. o bus_dma(9) clean up. o 40bit DMA address support. o Add protection for Rx map load failure. o Fix a long standing bug for watchdog timeout. [1] o Add additional protections, missing Tx completion interrupt, losing start Tx command, for watchdog timeout. o Switch to taskqueue(9) API to handle interrupts. o Use our own timer for watchdog instead of if_watchdog/if_timer interface. o Advertise VLAN header length/capability correctly to upper layer. o Remove excessive kernel stack consumption in nfe_encap(). o Handle highly fragmented mbuf chains correctly. o Enable etherenet address reprogramming with ifconfig(8). o Add ALTQ/TSO, MSI/MSIX support. o Increased Rx ring to 256 descriptors from 128. o Align Tx/Rx descriptor ring on sizeof(struct nfe_desc64) boundary. o Remove alignment restrictions on Tx/Rx buffers. o Rewritten jumbo frame support code. o Add support for hardware assistend VLAN tag insertion/stripping. o Add support for Tx/Rx flow control based on patches from Peer Chen. [2] o Add a routine that detects whether ethernet address swap routines is required. [3] o Add a workaround that take MAC/PHY out of power down mode. o Add suspend/resume support. o style(9) and code clean up. Special thanks to Shigeaki Tagashira, the original porter of nfe(4), who submitted lots of patches, performed uncountable number of regression tests and maintained nfe(4) for a long time. Without his enthusiastic help and support I could never have completed this overhauling task. The only weak point of nfe(4) compared to nve(4) is instability of manual half-duplex media selection on certain hardwares(auto sensing media type should work for all cases, though). This was a long standing bug of nfe(4) and I still have no idea why it doesn't work on some hardwares. Obtained from: OpenBSD [1] Submitted by: Peer Chen < pchen at nvidia dot com > [2], [3] Reviewed by: Shigeaki Tagashira < shigeaki AT se DOT hiroshima-u DOT ac DOT jp > Tested by: Shigeaki Tagashira, current Discussed with: current Silence from: obrien
2007-06-12 02:16:02 +00:00
struct nfe_desc32 *desc32 = NULL;
struct nfe_desc64 *desc64 = NULL;
Import the nfe (nForce Ethernet) driver from OpenBSD. This is a pure open source NIC driver with a BSD license. These bits were taken from the OpenBSD public CVS repository on 1-May-06.
2006-06-26 17:25:22 +00:00
bus_dmamap_t map;
Add a pure open source nForce Ethernet driver, under BSDL. This driver was ported from OpenBSD by Shigeaki Tagashira <shigeaki@se.hiroshima-u.ac.jp> and posted at http://www.se.hiroshima-u.ac.jp/~shigeaki/software/freebsd-nfe.html It was additionally cleaned up by me. It is still a work-in-progress and thus is purposefully not in GENERIC. And it conflicts with nve(4), so only one should be loaded.
2006-06-26 23:41:07 +00:00
bus_dma_segment_t segs[NFE_MAX_SCATTER];
Bring overhauled nfe(4) into tree. o s/printf/device_printf/g o Nuke OpenBSDism. o Nuke NetBSD/OpenBSD specific DMA sync operations.(we don't have a way to sync a single descriptor within a DMA map.) o Remove recursive mutex. o bus_dma(9) clean up. o 40bit DMA address support. o Add protection for Rx map load failure. o Fix a long standing bug for watchdog timeout. [1] o Add additional protections, missing Tx completion interrupt, losing start Tx command, for watchdog timeout. o Switch to taskqueue(9) API to handle interrupts. o Use our own timer for watchdog instead of if_watchdog/if_timer interface. o Advertise VLAN header length/capability correctly to upper layer. o Remove excessive kernel stack consumption in nfe_encap(). o Handle highly fragmented mbuf chains correctly. o Enable etherenet address reprogramming with ifconfig(8). o Add ALTQ/TSO, MSI/MSIX support. o Increased Rx ring to 256 descriptors from 128. o Align Tx/Rx descriptor ring on sizeof(struct nfe_desc64) boundary. o Remove alignment restrictions on Tx/Rx buffers. o Rewritten jumbo frame support code. o Add support for hardware assistend VLAN tag insertion/stripping. o Add support for Tx/Rx flow control based on patches from Peer Chen. [2] o Add a routine that detects whether ethernet address swap routines is required. [3] o Add a workaround that take MAC/PHY out of power down mode. o Add suspend/resume support. o style(9) and code clean up. Special thanks to Shigeaki Tagashira, the original porter of nfe(4), who submitted lots of patches, performed uncountable number of regression tests and maintained nfe(4) for a long time. Without his enthusiastic help and support I could never have completed this overhauling task. The only weak point of nfe(4) compared to nve(4) is instability of manual half-duplex media selection on certain hardwares(auto sensing media type should work for all cases, though). This was a long standing bug of nfe(4) and I still have no idea why it doesn't work on some hardwares. Obtained from: OpenBSD [1] Submitted by: Peer Chen < pchen at nvidia dot com > [2], [3] Reviewed by: Shigeaki Tagashira < shigeaki AT se DOT hiroshima-u DOT ac DOT jp > Tested by: Shigeaki Tagashira, current Discussed with: current Silence from: obrien
2007-06-12 02:16:02 +00:00
int error, i, nsegs, prod, si;
uint32_t tso_segsz;
uint16_t cflags, flags;
struct mbuf *m;
Import the nfe (nForce Ethernet) driver from OpenBSD. This is a pure open source NIC driver with a BSD license. These bits were taken from the OpenBSD public CVS repository on 1-May-06.
2006-06-26 17:25:22 +00:00
Bring overhauled nfe(4) into tree. o s/printf/device_printf/g o Nuke OpenBSDism. o Nuke NetBSD/OpenBSD specific DMA sync operations.(we don't have a way to sync a single descriptor within a DMA map.) o Remove recursive mutex. o bus_dma(9) clean up. o 40bit DMA address support. o Add protection for Rx map load failure. o Fix a long standing bug for watchdog timeout. [1] o Add additional protections, missing Tx completion interrupt, losing start Tx command, for watchdog timeout. o Switch to taskqueue(9) API to handle interrupts. o Use our own timer for watchdog instead of if_watchdog/if_timer interface. o Advertise VLAN header length/capability correctly to upper layer. o Remove excessive kernel stack consumption in nfe_encap(). o Handle highly fragmented mbuf chains correctly. o Enable etherenet address reprogramming with ifconfig(8). o Add ALTQ/TSO, MSI/MSIX support. o Increased Rx ring to 256 descriptors from 128. o Align Tx/Rx descriptor ring on sizeof(struct nfe_desc64) boundary. o Remove alignment restrictions on Tx/Rx buffers. o Rewritten jumbo frame support code. o Add support for hardware assistend VLAN tag insertion/stripping. o Add support for Tx/Rx flow control based on patches from Peer Chen. [2] o Add a routine that detects whether ethernet address swap routines is required. [3] o Add a workaround that take MAC/PHY out of power down mode. o Add suspend/resume support. o style(9) and code clean up. Special thanks to Shigeaki Tagashira, the original porter of nfe(4), who submitted lots of patches, performed uncountable number of regression tests and maintained nfe(4) for a long time. Without his enthusiastic help and support I could never have completed this overhauling task. The only weak point of nfe(4) compared to nve(4) is instability of manual half-duplex media selection on certain hardwares(auto sensing media type should work for all cases, though). This was a long standing bug of nfe(4) and I still have no idea why it doesn't work on some hardwares. Obtained from: OpenBSD [1] Submitted by: Peer Chen < pchen at nvidia dot com > [2], [3] Reviewed by: Shigeaki Tagashira < shigeaki AT se DOT hiroshima-u DOT ac DOT jp > Tested by: Shigeaki Tagashira, current Discussed with: current Silence from: obrien
2007-06-12 02:16:02 +00:00
prod = si = sc->txq.cur;
map = sc->txq.data[prod].tx_data_map;
Add a pure open source nForce Ethernet driver, under BSDL. This driver was ported from OpenBSD by Shigeaki Tagashira <shigeaki@se.hiroshima-u.ac.jp> and posted at http://www.se.hiroshima-u.ac.jp/~shigeaki/software/freebsd-nfe.html It was additionally cleaned up by me. It is still a work-in-progress and thus is purposefully not in GENERIC. And it conflicts with nve(4), so only one should be loaded.
2006-06-26 23:41:07 +00:00
Bring overhauled nfe(4) into tree. o s/printf/device_printf/g o Nuke OpenBSDism. o Nuke NetBSD/OpenBSD specific DMA sync operations.(we don't have a way to sync a single descriptor within a DMA map.) o Remove recursive mutex. o bus_dma(9) clean up. o 40bit DMA address support. o Add protection for Rx map load failure. o Fix a long standing bug for watchdog timeout. [1] o Add additional protections, missing Tx completion interrupt, losing start Tx command, for watchdog timeout. o Switch to taskqueue(9) API to handle interrupts. o Use our own timer for watchdog instead of if_watchdog/if_timer interface. o Advertise VLAN header length/capability correctly to upper layer. o Remove excessive kernel stack consumption in nfe_encap(). o Handle highly fragmented mbuf chains correctly. o Enable etherenet address reprogramming with ifconfig(8). o Add ALTQ/TSO, MSI/MSIX support. o Increased Rx ring to 256 descriptors from 128. o Align Tx/Rx descriptor ring on sizeof(struct nfe_desc64) boundary. o Remove alignment restrictions on Tx/Rx buffers. o Rewritten jumbo frame support code. o Add support for hardware assistend VLAN tag insertion/stripping. o Add support for Tx/Rx flow control based on patches from Peer Chen. [2] o Add a routine that detects whether ethernet address swap routines is required. [3] o Add a workaround that take MAC/PHY out of power down mode. o Add suspend/resume support. o style(9) and code clean up. Special thanks to Shigeaki Tagashira, the original porter of nfe(4), who submitted lots of patches, performed uncountable number of regression tests and maintained nfe(4) for a long time. Without his enthusiastic help and support I could never have completed this overhauling task. The only weak point of nfe(4) compared to nve(4) is instability of manual half-duplex media selection on certain hardwares(auto sensing media type should work for all cases, though). This was a long standing bug of nfe(4) and I still have no idea why it doesn't work on some hardwares. Obtained from: OpenBSD [1] Submitted by: Peer Chen < pchen at nvidia dot com > [2], [3] Reviewed by: Shigeaki Tagashira < shigeaki AT se DOT hiroshima-u DOT ac DOT jp > Tested by: Shigeaki Tagashira, current Discussed with: current Silence from: obrien
2007-06-12 02:16:02 +00:00
error = bus_dmamap_load_mbuf_sg(sc->txq.tx_data_tag, map, *m_head, segs,
Add a pure open source nForce Ethernet driver, under BSDL. This driver was ported from OpenBSD by Shigeaki Tagashira <shigeaki@se.hiroshima-u.ac.jp> and posted at http://www.se.hiroshima-u.ac.jp/~shigeaki/software/freebsd-nfe.html It was additionally cleaned up by me. It is still a work-in-progress and thus is purposefully not in GENERIC. And it conflicts with nve(4), so only one should be loaded.
2006-06-26 23:41:07 +00:00
&nsegs, BUS_DMA_NOWAIT);
Bring overhauled nfe(4) into tree. o s/printf/device_printf/g o Nuke OpenBSDism. o Nuke NetBSD/OpenBSD specific DMA sync operations.(we don't have a way to sync a single descriptor within a DMA map.) o Remove recursive mutex. o bus_dma(9) clean up. o 40bit DMA address support. o Add protection for Rx map load failure. o Fix a long standing bug for watchdog timeout. [1] o Add additional protections, missing Tx completion interrupt, losing start Tx command, for watchdog timeout. o Switch to taskqueue(9) API to handle interrupts. o Use our own timer for watchdog instead of if_watchdog/if_timer interface. o Advertise VLAN header length/capability correctly to upper layer. o Remove excessive kernel stack consumption in nfe_encap(). o Handle highly fragmented mbuf chains correctly. o Enable etherenet address reprogramming with ifconfig(8). o Add ALTQ/TSO, MSI/MSIX support. o Increased Rx ring to 256 descriptors from 128. o Align Tx/Rx descriptor ring on sizeof(struct nfe_desc64) boundary. o Remove alignment restrictions on Tx/Rx buffers. o Rewritten jumbo frame support code. o Add support for hardware assistend VLAN tag insertion/stripping. o Add support for Tx/Rx flow control based on patches from Peer Chen. [2] o Add a routine that detects whether ethernet address swap routines is required. [3] o Add a workaround that take MAC/PHY out of power down mode. o Add suspend/resume support. o style(9) and code clean up. Special thanks to Shigeaki Tagashira, the original porter of nfe(4), who submitted lots of patches, performed uncountable number of regression tests and maintained nfe(4) for a long time. Without his enthusiastic help and support I could never have completed this overhauling task. The only weak point of nfe(4) compared to nve(4) is instability of manual half-duplex media selection on certain hardwares(auto sensing media type should work for all cases, though). This was a long standing bug of nfe(4) and I still have no idea why it doesn't work on some hardwares. Obtained from: OpenBSD [1] Submitted by: Peer Chen < pchen at nvidia dot com > [2], [3] Reviewed by: Shigeaki Tagashira < shigeaki AT se DOT hiroshima-u DOT ac DOT jp > Tested by: Shigeaki Tagashira, current Discussed with: current Silence from: obrien
2007-06-12 02:16:02 +00:00
if (error == EFBIG) {
- Retire npe_defrag(), gem_defrag(), msk_defrag(), nfe_defrag(), and re_defrag() and use m_collapse() instead. - Replace a reference to ath_defrag() in a comment in if_wpi.c with m_collapse().
2008-01-17 23:37:47 +00:00
m = m_collapse(*m_head, M_DONTWAIT, NFE_MAX_SCATTER);
Bring overhauled nfe(4) into tree. o s/printf/device_printf/g o Nuke OpenBSDism. o Nuke NetBSD/OpenBSD specific DMA sync operations.(we don't have a way to sync a single descriptor within a DMA map.) o Remove recursive mutex. o bus_dma(9) clean up. o 40bit DMA address support. o Add protection for Rx map load failure. o Fix a long standing bug for watchdog timeout. [1] o Add additional protections, missing Tx completion interrupt, losing start Tx command, for watchdog timeout. o Switch to taskqueue(9) API to handle interrupts. o Use our own timer for watchdog instead of if_watchdog/if_timer interface. o Advertise VLAN header length/capability correctly to upper layer. o Remove excessive kernel stack consumption in nfe_encap(). o Handle highly fragmented mbuf chains correctly. o Enable etherenet address reprogramming with ifconfig(8). o Add ALTQ/TSO, MSI/MSIX support. o Increased Rx ring to 256 descriptors from 128. o Align Tx/Rx descriptor ring on sizeof(struct nfe_desc64) boundary. o Remove alignment restrictions on Tx/Rx buffers. o Rewritten jumbo frame support code. o Add support for hardware assistend VLAN tag insertion/stripping. o Add support for Tx/Rx flow control based on patches from Peer Chen. [2] o Add a routine that detects whether ethernet address swap routines is required. [3] o Add a workaround that take MAC/PHY out of power down mode. o Add suspend/resume support. o style(9) and code clean up. Special thanks to Shigeaki Tagashira, the original porter of nfe(4), who submitted lots of patches, performed uncountable number of regression tests and maintained nfe(4) for a long time. Without his enthusiastic help and support I could never have completed this overhauling task. The only weak point of nfe(4) compared to nve(4) is instability of manual half-duplex media selection on certain hardwares(auto sensing media type should work for all cases, though). This was a long standing bug of nfe(4) and I still have no idea why it doesn't work on some hardwares. Obtained from: OpenBSD [1] Submitted by: Peer Chen < pchen at nvidia dot com > [2], [3] Reviewed by: Shigeaki Tagashira < shigeaki AT se DOT hiroshima-u DOT ac DOT jp > Tested by: Shigeaki Tagashira, current Discussed with: current Silence from: obrien
2007-06-12 02:16:02 +00:00
if (m == NULL) {
m_freem(*m_head);
*m_head = NULL;
return (ENOBUFS);
}
*m_head = m;
error = bus_dmamap_load_mbuf_sg(sc->txq.tx_data_tag, map,
*m_head, segs, &nsegs, BUS_DMA_NOWAIT);
if (error != 0) {
m_freem(*m_head);
*m_head = NULL;
return (ENOBUFS);
}
} else if (error != 0)
return (error);
if (nsegs == 0) {
m_freem(*m_head);
*m_head = NULL;
return (EIO);
Import the nfe (nForce Ethernet) driver from OpenBSD. This is a pure open source NIC driver with a BSD license. These bits were taken from the OpenBSD public CVS repository on 1-May-06.
2006-06-26 17:25:22 +00:00
}
Bring overhauled nfe(4) into tree. o s/printf/device_printf/g o Nuke OpenBSDism. o Nuke NetBSD/OpenBSD specific DMA sync operations.(we don't have a way to sync a single descriptor within a DMA map.) o Remove recursive mutex. o bus_dma(9) clean up. o 40bit DMA address support. o Add protection for Rx map load failure. o Fix a long standing bug for watchdog timeout. [1] o Add additional protections, missing Tx completion interrupt, losing start Tx command, for watchdog timeout. o Switch to taskqueue(9) API to handle interrupts. o Use our own timer for watchdog instead of if_watchdog/if_timer interface. o Advertise VLAN header length/capability correctly to upper layer. o Remove excessive kernel stack consumption in nfe_encap(). o Handle highly fragmented mbuf chains correctly. o Enable etherenet address reprogramming with ifconfig(8). o Add ALTQ/TSO, MSI/MSIX support. o Increased Rx ring to 256 descriptors from 128. o Align Tx/Rx descriptor ring on sizeof(struct nfe_desc64) boundary. o Remove alignment restrictions on Tx/Rx buffers. o Rewritten jumbo frame support code. o Add support for hardware assistend VLAN tag insertion/stripping. o Add support for Tx/Rx flow control based on patches from Peer Chen. [2] o Add a routine that detects whether ethernet address swap routines is required. [3] o Add a workaround that take MAC/PHY out of power down mode. o Add suspend/resume support. o style(9) and code clean up. Special thanks to Shigeaki Tagashira, the original porter of nfe(4), who submitted lots of patches, performed uncountable number of regression tests and maintained nfe(4) for a long time. Without his enthusiastic help and support I could never have completed this overhauling task. The only weak point of nfe(4) compared to nve(4) is instability of manual half-duplex media selection on certain hardwares(auto sensing media type should work for all cases, though). This was a long standing bug of nfe(4) and I still have no idea why it doesn't work on some hardwares. Obtained from: OpenBSD [1] Submitted by: Peer Chen < pchen at nvidia dot com > [2], [3] Reviewed by: Shigeaki Tagashira < shigeaki AT se DOT hiroshima-u DOT ac DOT jp > Tested by: Shigeaki Tagashira, current Discussed with: current Silence from: obrien
2007-06-12 02:16:02 +00:00
if (sc->txq.queued + nsegs >= NFE_TX_RING_COUNT - 2) {
Add a pure open source nForce Ethernet driver, under BSDL. This driver was ported from OpenBSD by Shigeaki Tagashira <shigeaki@se.hiroshima-u.ac.jp> and posted at http://www.se.hiroshima-u.ac.jp/~shigeaki/software/freebsd-nfe.html It was additionally cleaned up by me. It is still a work-in-progress and thus is purposefully not in GENERIC. And it conflicts with nve(4), so only one should be loaded.
2006-06-26 23:41:07 +00:00
bus_dmamap_unload(sc->txq.tx_data_tag, map);
Bring overhauled nfe(4) into tree. o s/printf/device_printf/g o Nuke OpenBSDism. o Nuke NetBSD/OpenBSD specific DMA sync operations.(we don't have a way to sync a single descriptor within a DMA map.) o Remove recursive mutex. o bus_dma(9) clean up. o 40bit DMA address support. o Add protection for Rx map load failure. o Fix a long standing bug for watchdog timeout. [1] o Add additional protections, missing Tx completion interrupt, losing start Tx command, for watchdog timeout. o Switch to taskqueue(9) API to handle interrupts. o Use our own timer for watchdog instead of if_watchdog/if_timer interface. o Advertise VLAN header length/capability correctly to upper layer. o Remove excessive kernel stack consumption in nfe_encap(). o Handle highly fragmented mbuf chains correctly. o Enable etherenet address reprogramming with ifconfig(8). o Add ALTQ/TSO, MSI/MSIX support. o Increased Rx ring to 256 descriptors from 128. o Align Tx/Rx descriptor ring on sizeof(struct nfe_desc64) boundary. o Remove alignment restrictions on Tx/Rx buffers. o Rewritten jumbo frame support code. o Add support for hardware assistend VLAN tag insertion/stripping. o Add support for Tx/Rx flow control based on patches from Peer Chen. [2] o Add a routine that detects whether ethernet address swap routines is required. [3] o Add a workaround that take MAC/PHY out of power down mode. o Add suspend/resume support. o style(9) and code clean up. Special thanks to Shigeaki Tagashira, the original porter of nfe(4), who submitted lots of patches, performed uncountable number of regression tests and maintained nfe(4) for a long time. Without his enthusiastic help and support I could never have completed this overhauling task. The only weak point of nfe(4) compared to nve(4) is instability of manual half-duplex media selection on certain hardwares(auto sensing media type should work for all cases, though). This was a long standing bug of nfe(4) and I still have no idea why it doesn't work on some hardwares. Obtained from: OpenBSD [1] Submitted by: Peer Chen < pchen at nvidia dot com > [2], [3] Reviewed by: Shigeaki Tagashira < shigeaki AT se DOT hiroshima-u DOT ac DOT jp > Tested by: Shigeaki Tagashira, current Discussed with: current Silence from: obrien
2007-06-12 02:16:02 +00:00
return (ENOBUFS);
Import the nfe (nForce Ethernet) driver from OpenBSD. This is a pure open source NIC driver with a BSD license. These bits were taken from the OpenBSD public CVS repository on 1-May-06.
2006-06-26 17:25:22 +00:00
}
Bring overhauled nfe(4) into tree. o s/printf/device_printf/g o Nuke OpenBSDism. o Nuke NetBSD/OpenBSD specific DMA sync operations.(we don't have a way to sync a single descriptor within a DMA map.) o Remove recursive mutex. o bus_dma(9) clean up. o 40bit DMA address support. o Add protection for Rx map load failure. o Fix a long standing bug for watchdog timeout. [1] o Add additional protections, missing Tx completion interrupt, losing start Tx command, for watchdog timeout. o Switch to taskqueue(9) API to handle interrupts. o Use our own timer for watchdog instead of if_watchdog/if_timer interface. o Advertise VLAN header length/capability correctly to upper layer. o Remove excessive kernel stack consumption in nfe_encap(). o Handle highly fragmented mbuf chains correctly. o Enable etherenet address reprogramming with ifconfig(8). o Add ALTQ/TSO, MSI/MSIX support. o Increased Rx ring to 256 descriptors from 128. o Align Tx/Rx descriptor ring on sizeof(struct nfe_desc64) boundary. o Remove alignment restrictions on Tx/Rx buffers. o Rewritten jumbo frame support code. o Add support for hardware assistend VLAN tag insertion/stripping. o Add support for Tx/Rx flow control based on patches from Peer Chen. [2] o Add a routine that detects whether ethernet address swap routines is required. [3] o Add a workaround that take MAC/PHY out of power down mode. o Add suspend/resume support. o style(9) and code clean up. Special thanks to Shigeaki Tagashira, the original porter of nfe(4), who submitted lots of patches, performed uncountable number of regression tests and maintained nfe(4) for a long time. Without his enthusiastic help and support I could never have completed this overhauling task. The only weak point of nfe(4) compared to nve(4) is instability of manual half-duplex media selection on certain hardwares(auto sensing media type should work for all cases, though). This was a long standing bug of nfe(4) and I still have no idea why it doesn't work on some hardwares. Obtained from: OpenBSD [1] Submitted by: Peer Chen < pchen at nvidia dot com > [2], [3] Reviewed by: Shigeaki Tagashira < shigeaki AT se DOT hiroshima-u DOT ac DOT jp > Tested by: Shigeaki Tagashira, current Discussed with: current Silence from: obrien
2007-06-12 02:16:02 +00:00
m = *m_head;
cflags = flags = 0;
tso_segsz = 0;
With r206844, CSUM_TCP is also set for CSUM_TSO case. Modify drivers to take into account for the change. Basically CSUM_TSO should be checked before checking CSUM_TCP.
2010-04-19 22:10:40 +00:00
if ((m->m_pkthdr.csum_flags & CSUM_TSO) != 0) {
tso_segsz = (uint32_t)m->m_pkthdr.tso_segsz <<
NFE_TX_TSO_SHIFT;
cflags &= ~(NFE_TX_IP_CSUM | NFE_TX_TCP_UDP_CSUM);
cflags |= NFE_TX_TSO;
} else if ((m->m_pkthdr.csum_flags & NFE_CSUM_FEATURES) != 0) {
Bring overhauled nfe(4) into tree. o s/printf/device_printf/g o Nuke OpenBSDism. o Nuke NetBSD/OpenBSD specific DMA sync operations.(we don't have a way to sync a single descriptor within a DMA map.) o Remove recursive mutex. o bus_dma(9) clean up. o 40bit DMA address support. o Add protection for Rx map load failure. o Fix a long standing bug for watchdog timeout. [1] o Add additional protections, missing Tx completion interrupt, losing start Tx command, for watchdog timeout. o Switch to taskqueue(9) API to handle interrupts. o Use our own timer for watchdog instead of if_watchdog/if_timer interface. o Advertise VLAN header length/capability correctly to upper layer. o Remove excessive kernel stack consumption in nfe_encap(). o Handle highly fragmented mbuf chains correctly. o Enable etherenet address reprogramming with ifconfig(8). o Add ALTQ/TSO, MSI/MSIX support. o Increased Rx ring to 256 descriptors from 128. o Align Tx/Rx descriptor ring on sizeof(struct nfe_desc64) boundary. o Remove alignment restrictions on Tx/Rx buffers. o Rewritten jumbo frame support code. o Add support for hardware assistend VLAN tag insertion/stripping. o Add support for Tx/Rx flow control based on patches from Peer Chen. [2] o Add a routine that detects whether ethernet address swap routines is required. [3] o Add a workaround that take MAC/PHY out of power down mode. o Add suspend/resume support. o style(9) and code clean up. Special thanks to Shigeaki Tagashira, the original porter of nfe(4), who submitted lots of patches, performed uncountable number of regression tests and maintained nfe(4) for a long time. Without his enthusiastic help and support I could never have completed this overhauling task. The only weak point of nfe(4) compared to nve(4) is instability of manual half-duplex media selection on certain hardwares(auto sensing media type should work for all cases, though). This was a long standing bug of nfe(4) and I still have no idea why it doesn't work on some hardwares. Obtained from: OpenBSD [1] Submitted by: Peer Chen < pchen at nvidia dot com > [2], [3] Reviewed by: Shigeaki Tagashira < shigeaki AT se DOT hiroshima-u DOT ac DOT jp > Tested by: Shigeaki Tagashira, current Discussed with: current Silence from: obrien
2007-06-12 02:16:02 +00:00
if ((m->m_pkthdr.csum_flags & CSUM_IP) != 0)
cflags |= NFE_TX_IP_CSUM;
if ((m->m_pkthdr.csum_flags & CSUM_TCP) != 0)
cflags |= NFE_TX_TCP_UDP_CSUM;
if ((m->m_pkthdr.csum_flags & CSUM_UDP) != 0)
cflags |= NFE_TX_TCP_UDP_CSUM;
}
Import the nfe (nForce Ethernet) driver from OpenBSD. This is a pure open source NIC driver with a BSD license. These bits were taken from the OpenBSD public CVS repository on 1-May-06.
2006-06-26 17:25:22 +00:00
Add a pure open source nForce Ethernet driver, under BSDL. This driver was ported from OpenBSD by Shigeaki Tagashira <shigeaki@se.hiroshima-u.ac.jp> and posted at http://www.se.hiroshima-u.ac.jp/~shigeaki/software/freebsd-nfe.html It was additionally cleaned up by me. It is still a work-in-progress and thus is purposefully not in GENERIC. And it conflicts with nve(4), so only one should be loaded.
2006-06-26 23:41:07 +00:00
for (i = 0; i < nsegs; i++) {
if (sc->nfe_flags & NFE_40BIT_ADDR) {
Bring overhauled nfe(4) into tree. o s/printf/device_printf/g o Nuke OpenBSDism. o Nuke NetBSD/OpenBSD specific DMA sync operations.(we don't have a way to sync a single descriptor within a DMA map.) o Remove recursive mutex. o bus_dma(9) clean up. o 40bit DMA address support. o Add protection for Rx map load failure. o Fix a long standing bug for watchdog timeout. [1] o Add additional protections, missing Tx completion interrupt, losing start Tx command, for watchdog timeout. o Switch to taskqueue(9) API to handle interrupts. o Use our own timer for watchdog instead of if_watchdog/if_timer interface. o Advertise VLAN header length/capability correctly to upper layer. o Remove excessive kernel stack consumption in nfe_encap(). o Handle highly fragmented mbuf chains correctly. o Enable etherenet address reprogramming with ifconfig(8). o Add ALTQ/TSO, MSI/MSIX support. o Increased Rx ring to 256 descriptors from 128. o Align Tx/Rx descriptor ring on sizeof(struct nfe_desc64) boundary. o Remove alignment restrictions on Tx/Rx buffers. o Rewritten jumbo frame support code. o Add support for hardware assistend VLAN tag insertion/stripping. o Add support for Tx/Rx flow control based on patches from Peer Chen. [2] o Add a routine that detects whether ethernet address swap routines is required. [3] o Add a workaround that take MAC/PHY out of power down mode. o Add suspend/resume support. o style(9) and code clean up. Special thanks to Shigeaki Tagashira, the original porter of nfe(4), who submitted lots of patches, performed uncountable number of regression tests and maintained nfe(4) for a long time. Without his enthusiastic help and support I could never have completed this overhauling task. The only weak point of nfe(4) compared to nve(4) is instability of manual half-duplex media selection on certain hardwares(auto sensing media type should work for all cases, though). This was a long standing bug of nfe(4) and I still have no idea why it doesn't work on some hardwares. Obtained from: OpenBSD [1] Submitted by: Peer Chen < pchen at nvidia dot com > [2], [3] Reviewed by: Shigeaki Tagashira < shigeaki AT se DOT hiroshima-u DOT ac DOT jp > Tested by: Shigeaki Tagashira, current Discussed with: current Silence from: obrien
2007-06-12 02:16:02 +00:00
desc64 = &sc->txq.desc64[prod];
desc64->physaddr[0] =
htole32(NFE_ADDR_HI(segs[i].ds_addr));
desc64->physaddr[1] =
htole32(NFE_ADDR_LO(segs[i].ds_addr));
desc64->vtag = 0;
Add a pure open source nForce Ethernet driver, under BSDL. This driver was ported from OpenBSD by Shigeaki Tagashira <shigeaki@se.hiroshima-u.ac.jp> and posted at http://www.se.hiroshima-u.ac.jp/~shigeaki/software/freebsd-nfe.html It was additionally cleaned up by me. It is still a work-in-progress and thus is purposefully not in GENERIC. And it conflicts with nve(4), so only one should be loaded.
2006-06-26 23:41:07 +00:00
desc64->length = htole16(segs[i].ds_len - 1);
Import the nfe (nForce Ethernet) driver from OpenBSD. This is a pure open source NIC driver with a BSD license. These bits were taken from the OpenBSD public CVS repository on 1-May-06.
2006-06-26 17:25:22 +00:00
desc64->flags = htole16(flags);
} else {
Bring overhauled nfe(4) into tree. o s/printf/device_printf/g o Nuke OpenBSDism. o Nuke NetBSD/OpenBSD specific DMA sync operations.(we don't have a way to sync a single descriptor within a DMA map.) o Remove recursive mutex. o bus_dma(9) clean up. o 40bit DMA address support. o Add protection for Rx map load failure. o Fix a long standing bug for watchdog timeout. [1] o Add additional protections, missing Tx completion interrupt, losing start Tx command, for watchdog timeout. o Switch to taskqueue(9) API to handle interrupts. o Use our own timer for watchdog instead of if_watchdog/if_timer interface. o Advertise VLAN header length/capability correctly to upper layer. o Remove excessive kernel stack consumption in nfe_encap(). o Handle highly fragmented mbuf chains correctly. o Enable etherenet address reprogramming with ifconfig(8). o Add ALTQ/TSO, MSI/MSIX support. o Increased Rx ring to 256 descriptors from 128. o Align Tx/Rx descriptor ring on sizeof(struct nfe_desc64) boundary. o Remove alignment restrictions on Tx/Rx buffers. o Rewritten jumbo frame support code. o Add support for hardware assistend VLAN tag insertion/stripping. o Add support for Tx/Rx flow control based on patches from Peer Chen. [2] o Add a routine that detects whether ethernet address swap routines is required. [3] o Add a workaround that take MAC/PHY out of power down mode. o Add suspend/resume support. o style(9) and code clean up. Special thanks to Shigeaki Tagashira, the original porter of nfe(4), who submitted lots of patches, performed uncountable number of regression tests and maintained nfe(4) for a long time. Without his enthusiastic help and support I could never have completed this overhauling task. The only weak point of nfe(4) compared to nve(4) is instability of manual half-duplex media selection on certain hardwares(auto sensing media type should work for all cases, though). This was a long standing bug of nfe(4) and I still have no idea why it doesn't work on some hardwares. Obtained from: OpenBSD [1] Submitted by: Peer Chen < pchen at nvidia dot com > [2], [3] Reviewed by: Shigeaki Tagashira < shigeaki AT se DOT hiroshima-u DOT ac DOT jp > Tested by: Shigeaki Tagashira, current Discussed with: current Silence from: obrien
2007-06-12 02:16:02 +00:00
desc32 = &sc->txq.desc32[prod];
desc32->physaddr =
htole32(NFE_ADDR_LO(segs[i].ds_addr));
Add a pure open source nForce Ethernet driver, under BSDL. This driver was ported from OpenBSD by Shigeaki Tagashira <shigeaki@se.hiroshima-u.ac.jp> and posted at http://www.se.hiroshima-u.ac.jp/~shigeaki/software/freebsd-nfe.html It was additionally cleaned up by me. It is still a work-in-progress and thus is purposefully not in GENERIC. And it conflicts with nve(4), so only one should be loaded.
2006-06-26 23:41:07 +00:00
desc32->length = htole16(segs[i].ds_len - 1);
Import the nfe (nForce Ethernet) driver from OpenBSD. This is a pure open source NIC driver with a BSD license. These bits were taken from the OpenBSD public CVS repository on 1-May-06.
2006-06-26 17:25:22 +00:00
desc32->flags = htole16(flags);
}
Bring overhauled nfe(4) into tree. o s/printf/device_printf/g o Nuke OpenBSDism. o Nuke NetBSD/OpenBSD specific DMA sync operations.(we don't have a way to sync a single descriptor within a DMA map.) o Remove recursive mutex. o bus_dma(9) clean up. o 40bit DMA address support. o Add protection for Rx map load failure. o Fix a long standing bug for watchdog timeout. [1] o Add additional protections, missing Tx completion interrupt, losing start Tx command, for watchdog timeout. o Switch to taskqueue(9) API to handle interrupts. o Use our own timer for watchdog instead of if_watchdog/if_timer interface. o Advertise VLAN header length/capability correctly to upper layer. o Remove excessive kernel stack consumption in nfe_encap(). o Handle highly fragmented mbuf chains correctly. o Enable etherenet address reprogramming with ifconfig(8). o Add ALTQ/TSO, MSI/MSIX support. o Increased Rx ring to 256 descriptors from 128. o Align Tx/Rx descriptor ring on sizeof(struct nfe_desc64) boundary. o Remove alignment restrictions on Tx/Rx buffers. o Rewritten jumbo frame support code. o Add support for hardware assistend VLAN tag insertion/stripping. o Add support for Tx/Rx flow control based on patches from Peer Chen. [2] o Add a routine that detects whether ethernet address swap routines is required. [3] o Add a workaround that take MAC/PHY out of power down mode. o Add suspend/resume support. o style(9) and code clean up. Special thanks to Shigeaki Tagashira, the original porter of nfe(4), who submitted lots of patches, performed uncountable number of regression tests and maintained nfe(4) for a long time. Without his enthusiastic help and support I could never have completed this overhauling task. The only weak point of nfe(4) compared to nve(4) is instability of manual half-duplex media selection on certain hardwares(auto sensing media type should work for all cases, though). This was a long standing bug of nfe(4) and I still have no idea why it doesn't work on some hardwares. Obtained from: OpenBSD [1] Submitted by: Peer Chen < pchen at nvidia dot com > [2], [3] Reviewed by: Shigeaki Tagashira < shigeaki AT se DOT hiroshima-u DOT ac DOT jp > Tested by: Shigeaki Tagashira, current Discussed with: current Silence from: obrien
2007-06-12 02:16:02 +00:00
/*
* Setting of the valid bit in the first descriptor is
* deferred until the whole chain is fully setup.
*/
flags |= NFE_TX_VALID;
style(9) cleanup.
2006-10-19 10:01:26 +00:00
Import the nfe (nForce Ethernet) driver from OpenBSD. This is a pure open source NIC driver with a BSD license. These bits were taken from the OpenBSD public CVS repository on 1-May-06.
2006-06-26 17:25:22 +00:00
sc->txq.queued++;
Bring overhauled nfe(4) into tree. o s/printf/device_printf/g o Nuke OpenBSDism. o Nuke NetBSD/OpenBSD specific DMA sync operations.(we don't have a way to sync a single descriptor within a DMA map.) o Remove recursive mutex. o bus_dma(9) clean up. o 40bit DMA address support. o Add protection for Rx map load failure. o Fix a long standing bug for watchdog timeout. [1] o Add additional protections, missing Tx completion interrupt, losing start Tx command, for watchdog timeout. o Switch to taskqueue(9) API to handle interrupts. o Use our own timer for watchdog instead of if_watchdog/if_timer interface. o Advertise VLAN header length/capability correctly to upper layer. o Remove excessive kernel stack consumption in nfe_encap(). o Handle highly fragmented mbuf chains correctly. o Enable etherenet address reprogramming with ifconfig(8). o Add ALTQ/TSO, MSI/MSIX support. o Increased Rx ring to 256 descriptors from 128. o Align Tx/Rx descriptor ring on sizeof(struct nfe_desc64) boundary. o Remove alignment restrictions on Tx/Rx buffers. o Rewritten jumbo frame support code. o Add support for hardware assistend VLAN tag insertion/stripping. o Add support for Tx/Rx flow control based on patches from Peer Chen. [2] o Add a routine that detects whether ethernet address swap routines is required. [3] o Add a workaround that take MAC/PHY out of power down mode. o Add suspend/resume support. o style(9) and code clean up. Special thanks to Shigeaki Tagashira, the original porter of nfe(4), who submitted lots of patches, performed uncountable number of regression tests and maintained nfe(4) for a long time. Without his enthusiastic help and support I could never have completed this overhauling task. The only weak point of nfe(4) compared to nve(4) is instability of manual half-duplex media selection on certain hardwares(auto sensing media type should work for all cases, though). This was a long standing bug of nfe(4) and I still have no idea why it doesn't work on some hardwares. Obtained from: OpenBSD [1] Submitted by: Peer Chen < pchen at nvidia dot com > [2], [3] Reviewed by: Shigeaki Tagashira < shigeaki AT se DOT hiroshima-u DOT ac DOT jp > Tested by: Shigeaki Tagashira, current Discussed with: current Silence from: obrien
2007-06-12 02:16:02 +00:00
NFE_INC(prod, NFE_TX_RING_COUNT);
Import the nfe (nForce Ethernet) driver from OpenBSD. This is a pure open source NIC driver with a BSD license. These bits were taken from the OpenBSD public CVS repository on 1-May-06.
2006-06-26 17:25:22 +00:00
}
Bring overhauled nfe(4) into tree. o s/printf/device_printf/g o Nuke OpenBSDism. o Nuke NetBSD/OpenBSD specific DMA sync operations.(we don't have a way to sync a single descriptor within a DMA map.) o Remove recursive mutex. o bus_dma(9) clean up. o 40bit DMA address support. o Add protection for Rx map load failure. o Fix a long standing bug for watchdog timeout. [1] o Add additional protections, missing Tx completion interrupt, losing start Tx command, for watchdog timeout. o Switch to taskqueue(9) API to handle interrupts. o Use our own timer for watchdog instead of if_watchdog/if_timer interface. o Advertise VLAN header length/capability correctly to upper layer. o Remove excessive kernel stack consumption in nfe_encap(). o Handle highly fragmented mbuf chains correctly. o Enable etherenet address reprogramming with ifconfig(8). o Add ALTQ/TSO, MSI/MSIX support. o Increased Rx ring to 256 descriptors from 128. o Align Tx/Rx descriptor ring on sizeof(struct nfe_desc64) boundary. o Remove alignment restrictions on Tx/Rx buffers. o Rewritten jumbo frame support code. o Add support for hardware assistend VLAN tag insertion/stripping. o Add support for Tx/Rx flow control based on patches from Peer Chen. [2] o Add a routine that detects whether ethernet address swap routines is required. [3] o Add a workaround that take MAC/PHY out of power down mode. o Add suspend/resume support. o style(9) and code clean up. Special thanks to Shigeaki Tagashira, the original porter of nfe(4), who submitted lots of patches, performed uncountable number of regression tests and maintained nfe(4) for a long time. Without his enthusiastic help and support I could never have completed this overhauling task. The only weak point of nfe(4) compared to nve(4) is instability of manual half-duplex media selection on certain hardwares(auto sensing media type should work for all cases, though). This was a long standing bug of nfe(4) and I still have no idea why it doesn't work on some hardwares. Obtained from: OpenBSD [1] Submitted by: Peer Chen < pchen at nvidia dot com > [2], [3] Reviewed by: Shigeaki Tagashira < shigeaki AT se DOT hiroshima-u DOT ac DOT jp > Tested by: Shigeaki Tagashira, current Discussed with: current Silence from: obrien
2007-06-12 02:16:02 +00:00
/*
* the whole mbuf chain has been DMA mapped, fix last/first descriptor.
* csum flags, vtag and TSO belong to the first fragment only.
*/
Add a pure open source nForce Ethernet driver, under BSDL. This driver was ported from OpenBSD by Shigeaki Tagashira <shigeaki@se.hiroshima-u.ac.jp> and posted at http://www.se.hiroshima-u.ac.jp/~shigeaki/software/freebsd-nfe.html It was additionally cleaned up by me. It is still a work-in-progress and thus is purposefully not in GENERIC. And it conflicts with nve(4), so only one should be loaded.
2006-06-26 23:41:07 +00:00
if (sc->nfe_flags & NFE_40BIT_ADDR) {
Bring overhauled nfe(4) into tree. o s/printf/device_printf/g o Nuke OpenBSDism. o Nuke NetBSD/OpenBSD specific DMA sync operations.(we don't have a way to sync a single descriptor within a DMA map.) o Remove recursive mutex. o bus_dma(9) clean up. o 40bit DMA address support. o Add protection for Rx map load failure. o Fix a long standing bug for watchdog timeout. [1] o Add additional protections, missing Tx completion interrupt, losing start Tx command, for watchdog timeout. o Switch to taskqueue(9) API to handle interrupts. o Use our own timer for watchdog instead of if_watchdog/if_timer interface. o Advertise VLAN header length/capability correctly to upper layer. o Remove excessive kernel stack consumption in nfe_encap(). o Handle highly fragmented mbuf chains correctly. o Enable etherenet address reprogramming with ifconfig(8). o Add ALTQ/TSO, MSI/MSIX support. o Increased Rx ring to 256 descriptors from 128. o Align Tx/Rx descriptor ring on sizeof(struct nfe_desc64) boundary. o Remove alignment restrictions on Tx/Rx buffers. o Rewritten jumbo frame support code. o Add support for hardware assistend VLAN tag insertion/stripping. o Add support for Tx/Rx flow control based on patches from Peer Chen. [2] o Add a routine that detects whether ethernet address swap routines is required. [3] o Add a workaround that take MAC/PHY out of power down mode. o Add suspend/resume support. o style(9) and code clean up. Special thanks to Shigeaki Tagashira, the original porter of nfe(4), who submitted lots of patches, performed uncountable number of regression tests and maintained nfe(4) for a long time. Without his enthusiastic help and support I could never have completed this overhauling task. The only weak point of nfe(4) compared to nve(4) is instability of manual half-duplex media selection on certain hardwares(auto sensing media type should work for all cases, though). This was a long standing bug of nfe(4) and I still have no idea why it doesn't work on some hardwares. Obtained from: OpenBSD [1] Submitted by: Peer Chen < pchen at nvidia dot com > [2], [3] Reviewed by: Shigeaki Tagashira < shigeaki AT se DOT hiroshima-u DOT ac DOT jp > Tested by: Shigeaki Tagashira, current Discussed with: current Silence from: obrien
2007-06-12 02:16:02 +00:00
desc64->flags |= htole16(NFE_TX_LASTFRAG_V2);
desc64 = &sc->txq.desc64[si];
if ((m->m_flags & M_VLANTAG) != 0)
desc64->vtag = htole32(NFE_TX_VTAG |
m->m_pkthdr.ether_vtag);
if (tso_segsz != 0) {
/*
* XXX
* The following indicates the descriptor element
* is a 32bit quantity.
*/
desc64->length |= htole16((uint16_t)tso_segsz);
desc64->flags |= htole16(tso_segsz >> 16);
}
/*
* finally, set the valid/checksum/TSO bit in the first
* descriptor.
*/
desc64->flags |= htole16(NFE_TX_VALID | cflags);
Import the nfe (nForce Ethernet) driver from OpenBSD. This is a pure open source NIC driver with a BSD license. These bits were taken from the OpenBSD public CVS repository on 1-May-06.
2006-06-26 17:25:22 +00:00
} else {
Add a pure open source nForce Ethernet driver, under BSDL. This driver was ported from OpenBSD by Shigeaki Tagashira <shigeaki@se.hiroshima-u.ac.jp> and posted at http://www.se.hiroshima-u.ac.jp/~shigeaki/software/freebsd-nfe.html It was additionally cleaned up by me. It is still a work-in-progress and thus is purposefully not in GENERIC. And it conflicts with nve(4), so only one should be loaded.
2006-06-26 23:41:07 +00:00
if (sc->nfe_flags & NFE_JUMBO_SUP)
Bring overhauled nfe(4) into tree. o s/printf/device_printf/g o Nuke OpenBSDism. o Nuke NetBSD/OpenBSD specific DMA sync operations.(we don't have a way to sync a single descriptor within a DMA map.) o Remove recursive mutex. o bus_dma(9) clean up. o 40bit DMA address support. o Add protection for Rx map load failure. o Fix a long standing bug for watchdog timeout. [1] o Add additional protections, missing Tx completion interrupt, losing start Tx command, for watchdog timeout. o Switch to taskqueue(9) API to handle interrupts. o Use our own timer for watchdog instead of if_watchdog/if_timer interface. o Advertise VLAN header length/capability correctly to upper layer. o Remove excessive kernel stack consumption in nfe_encap(). o Handle highly fragmented mbuf chains correctly. o Enable etherenet address reprogramming with ifconfig(8). o Add ALTQ/TSO, MSI/MSIX support. o Increased Rx ring to 256 descriptors from 128. o Align Tx/Rx descriptor ring on sizeof(struct nfe_desc64) boundary. o Remove alignment restrictions on Tx/Rx buffers. o Rewritten jumbo frame support code. o Add support for hardware assistend VLAN tag insertion/stripping. o Add support for Tx/Rx flow control based on patches from Peer Chen. [2] o Add a routine that detects whether ethernet address swap routines is required. [3] o Add a workaround that take MAC/PHY out of power down mode. o Add suspend/resume support. o style(9) and code clean up. Special thanks to Shigeaki Tagashira, the original porter of nfe(4), who submitted lots of patches, performed uncountable number of regression tests and maintained nfe(4) for a long time. Without his enthusiastic help and support I could never have completed this overhauling task. The only weak point of nfe(4) compared to nve(4) is instability of manual half-duplex media selection on certain hardwares(auto sensing media type should work for all cases, though). This was a long standing bug of nfe(4) and I still have no idea why it doesn't work on some hardwares. Obtained from: OpenBSD [1] Submitted by: Peer Chen < pchen at nvidia dot com > [2], [3] Reviewed by: Shigeaki Tagashira < shigeaki AT se DOT hiroshima-u DOT ac DOT jp > Tested by: Shigeaki Tagashira, current Discussed with: current Silence from: obrien
2007-06-12 02:16:02 +00:00
desc32->flags |= htole16(NFE_TX_LASTFRAG_V2);
Import the nfe (nForce Ethernet) driver from OpenBSD. This is a pure open source NIC driver with a BSD license. These bits were taken from the OpenBSD public CVS repository on 1-May-06.
2006-06-26 17:25:22 +00:00
else
Bring overhauled nfe(4) into tree. o s/printf/device_printf/g o Nuke OpenBSDism. o Nuke NetBSD/OpenBSD specific DMA sync operations.(we don't have a way to sync a single descriptor within a DMA map.) o Remove recursive mutex. o bus_dma(9) clean up. o 40bit DMA address support. o Add protection for Rx map load failure. o Fix a long standing bug for watchdog timeout. [1] o Add additional protections, missing Tx completion interrupt, losing start Tx command, for watchdog timeout. o Switch to taskqueue(9) API to handle interrupts. o Use our own timer for watchdog instead of if_watchdog/if_timer interface. o Advertise VLAN header length/capability correctly to upper layer. o Remove excessive kernel stack consumption in nfe_encap(). o Handle highly fragmented mbuf chains correctly. o Enable etherenet address reprogramming with ifconfig(8). o Add ALTQ/TSO, MSI/MSIX support. o Increased Rx ring to 256 descriptors from 128. o Align Tx/Rx descriptor ring on sizeof(struct nfe_desc64) boundary. o Remove alignment restrictions on Tx/Rx buffers. o Rewritten jumbo frame support code. o Add support for hardware assistend VLAN tag insertion/stripping. o Add support for Tx/Rx flow control based on patches from Peer Chen. [2] o Add a routine that detects whether ethernet address swap routines is required. [3] o Add a workaround that take MAC/PHY out of power down mode. o Add suspend/resume support. o style(9) and code clean up. Special thanks to Shigeaki Tagashira, the original porter of nfe(4), who submitted lots of patches, performed uncountable number of regression tests and maintained nfe(4) for a long time. Without his enthusiastic help and support I could never have completed this overhauling task. The only weak point of nfe(4) compared to nve(4) is instability of manual half-duplex media selection on certain hardwares(auto sensing media type should work for all cases, though). This was a long standing bug of nfe(4) and I still have no idea why it doesn't work on some hardwares. Obtained from: OpenBSD [1] Submitted by: Peer Chen < pchen at nvidia dot com > [2], [3] Reviewed by: Shigeaki Tagashira < shigeaki AT se DOT hiroshima-u DOT ac DOT jp > Tested by: Shigeaki Tagashira, current Discussed with: current Silence from: obrien
2007-06-12 02:16:02 +00:00
desc32->flags |= htole16(NFE_TX_LASTFRAG_V1);
desc32 = &sc->txq.desc32[si];
if (tso_segsz != 0) {
/*
* XXX
* The following indicates the descriptor element
* is a 32bit quantity.
*/
desc32->length |= htole16((uint16_t)tso_segsz);
desc32->flags |= htole16(tso_segsz >> 16);
}
/*
* finally, set the valid/checksum/TSO bit in the first
* descriptor.
*/
desc32->flags |= htole16(NFE_TX_VALID | cflags);
Import the nfe (nForce Ethernet) driver from OpenBSD. This is a pure open source NIC driver with a BSD license. These bits were taken from the OpenBSD public CVS repository on 1-May-06.
2006-06-26 17:25:22 +00:00
}
Bring overhauled nfe(4) into tree. o s/printf/device_printf/g o Nuke OpenBSDism. o Nuke NetBSD/OpenBSD specific DMA sync operations.(we don't have a way to sync a single descriptor within a DMA map.) o Remove recursive mutex. o bus_dma(9) clean up. o 40bit DMA address support. o Add protection for Rx map load failure. o Fix a long standing bug for watchdog timeout. [1] o Add additional protections, missing Tx completion interrupt, losing start Tx command, for watchdog timeout. o Switch to taskqueue(9) API to handle interrupts. o Use our own timer for watchdog instead of if_watchdog/if_timer interface. o Advertise VLAN header length/capability correctly to upper layer. o Remove excessive kernel stack consumption in nfe_encap(). o Handle highly fragmented mbuf chains correctly. o Enable etherenet address reprogramming with ifconfig(8). o Add ALTQ/TSO, MSI/MSIX support. o Increased Rx ring to 256 descriptors from 128. o Align Tx/Rx descriptor ring on sizeof(struct nfe_desc64) boundary. o Remove alignment restrictions on Tx/Rx buffers. o Rewritten jumbo frame support code. o Add support for hardware assistend VLAN tag insertion/stripping. o Add support for Tx/Rx flow control based on patches from Peer Chen. [2] o Add a routine that detects whether ethernet address swap routines is required. [3] o Add a workaround that take MAC/PHY out of power down mode. o Add suspend/resume support. o style(9) and code clean up. Special thanks to Shigeaki Tagashira, the original porter of nfe(4), who submitted lots of patches, performed uncountable number of regression tests and maintained nfe(4) for a long time. Without his enthusiastic help and support I could never have completed this overhauling task. The only weak point of nfe(4) compared to nve(4) is instability of manual half-duplex media selection on certain hardwares(auto sensing media type should work for all cases, though). This was a long standing bug of nfe(4) and I still have no idea why it doesn't work on some hardwares. Obtained from: OpenBSD [1] Submitted by: Peer Chen < pchen at nvidia dot com > [2], [3] Reviewed by: Shigeaki Tagashira < shigeaki AT se DOT hiroshima-u DOT ac DOT jp > Tested by: Shigeaki Tagashira, current Discussed with: current Silence from: obrien
2007-06-12 02:16:02 +00:00
sc->txq.cur = prod;
prod = (prod + NFE_TX_RING_COUNT - 1) % NFE_TX_RING_COUNT;
sc->txq.data[si].tx_data_map = sc->txq.data[prod].tx_data_map;
sc->txq.data[prod].tx_data_map = map;
sc->txq.data[prod].m = m;
Add a pure open source nForce Ethernet driver, under BSDL. This driver was ported from OpenBSD by Shigeaki Tagashira <shigeaki@se.hiroshima-u.ac.jp> and posted at http://www.se.hiroshima-u.ac.jp/~shigeaki/software/freebsd-nfe.html It was additionally cleaned up by me. It is still a work-in-progress and thus is purposefully not in GENERIC. And it conflicts with nve(4), so only one should be loaded.
2006-06-26 23:41:07 +00:00
bus_dmamap_sync(sc->txq.tx_data_tag, map, BUS_DMASYNC_PREWRITE);
Bring overhauled nfe(4) into tree. o s/printf/device_printf/g o Nuke OpenBSDism. o Nuke NetBSD/OpenBSD specific DMA sync operations.(we don't have a way to sync a single descriptor within a DMA map.) o Remove recursive mutex. o bus_dma(9) clean up. o 40bit DMA address support. o Add protection for Rx map load failure. o Fix a long standing bug for watchdog timeout. [1] o Add additional protections, missing Tx completion interrupt, losing start Tx command, for watchdog timeout. o Switch to taskqueue(9) API to handle interrupts. o Use our own timer for watchdog instead of if_watchdog/if_timer interface. o Advertise VLAN header length/capability correctly to upper layer. o Remove excessive kernel stack consumption in nfe_encap(). o Handle highly fragmented mbuf chains correctly. o Enable etherenet address reprogramming with ifconfig(8). o Add ALTQ/TSO, MSI/MSIX support. o Increased Rx ring to 256 descriptors from 128. o Align Tx/Rx descriptor ring on sizeof(struct nfe_desc64) boundary. o Remove alignment restrictions on Tx/Rx buffers. o Rewritten jumbo frame support code. o Add support for hardware assistend VLAN tag insertion/stripping. o Add support for Tx/Rx flow control based on patches from Peer Chen. [2] o Add a routine that detects whether ethernet address swap routines is required. [3] o Add a workaround that take MAC/PHY out of power down mode. o Add suspend/resume support. o style(9) and code clean up. Special thanks to Shigeaki Tagashira, the original porter of nfe(4), who submitted lots of patches, performed uncountable number of regression tests and maintained nfe(4) for a long time. Without his enthusiastic help and support I could never have completed this overhauling task. The only weak point of nfe(4) compared to nve(4) is instability of manual half-duplex media selection on certain hardwares(auto sensing media type should work for all cases, though). This was a long standing bug of nfe(4) and I still have no idea why it doesn't work on some hardwares. Obtained from: OpenBSD [1] Submitted by: Peer Chen < pchen at nvidia dot com > [2], [3] Reviewed by: Shigeaki Tagashira < shigeaki AT se DOT hiroshima-u DOT ac DOT jp > Tested by: Shigeaki Tagashira, current Discussed with: current Silence from: obrien
2007-06-12 02:16:02 +00:00
return (0);
Add a pure open source nForce Ethernet driver, under BSDL. This driver was ported from OpenBSD by Shigeaki Tagashira <shigeaki@se.hiroshima-u.ac.jp> and posted at http://www.se.hiroshima-u.ac.jp/~shigeaki/software/freebsd-nfe.html It was additionally cleaned up by me. It is still a work-in-progress and thus is purposefully not in GENERIC. And it conflicts with nve(4), so only one should be loaded.
2006-06-26 23:41:07 +00:00
}
style(9) cleanup.
2006-10-19 10:01:26 +00:00
static void
nfe_setmulti(struct nfe_softc *sc)
Add a pure open source nForce Ethernet driver, under BSDL. This driver was ported from OpenBSD by Shigeaki Tagashira <shigeaki@se.hiroshima-u.ac.jp> and posted at http://www.se.hiroshima-u.ac.jp/~shigeaki/software/freebsd-nfe.html It was additionally cleaned up by me. It is still a work-in-progress and thus is purposefully not in GENERIC. And it conflicts with nve(4), so only one should be loaded.
2006-06-26 23:41:07 +00:00
{
struct ifnet *ifp = sc->nfe_ifp;
style(9) cleanup.
2006-10-19 10:01:26 +00:00
struct ifmultiaddr *ifma;
Add a pure open source nForce Ethernet driver, under BSDL. This driver was ported from OpenBSD by Shigeaki Tagashira <shigeaki@se.hiroshima-u.ac.jp> and posted at http://www.se.hiroshima-u.ac.jp/~shigeaki/software/freebsd-nfe.html It was additionally cleaned up by me. It is still a work-in-progress and thus is purposefully not in GENERIC. And it conflicts with nve(4), so only one should be loaded.
2006-06-26 23:41:07 +00:00
int i;
Bring overhauled nfe(4) into tree. o s/printf/device_printf/g o Nuke OpenBSDism. o Nuke NetBSD/OpenBSD specific DMA sync operations.(we don't have a way to sync a single descriptor within a DMA map.) o Remove recursive mutex. o bus_dma(9) clean up. o 40bit DMA address support. o Add protection for Rx map load failure. o Fix a long standing bug for watchdog timeout. [1] o Add additional protections, missing Tx completion interrupt, losing start Tx command, for watchdog timeout. o Switch to taskqueue(9) API to handle interrupts. o Use our own timer for watchdog instead of if_watchdog/if_timer interface. o Advertise VLAN header length/capability correctly to upper layer. o Remove excessive kernel stack consumption in nfe_encap(). o Handle highly fragmented mbuf chains correctly. o Enable etherenet address reprogramming with ifconfig(8). o Add ALTQ/TSO, MSI/MSIX support. o Increased Rx ring to 256 descriptors from 128. o Align Tx/Rx descriptor ring on sizeof(struct nfe_desc64) boundary. o Remove alignment restrictions on Tx/Rx buffers. o Rewritten jumbo frame support code. o Add support for hardware assistend VLAN tag insertion/stripping. o Add support for Tx/Rx flow control based on patches from Peer Chen. [2] o Add a routine that detects whether ethernet address swap routines is required. [3] o Add a workaround that take MAC/PHY out of power down mode. o Add suspend/resume support. o style(9) and code clean up. Special thanks to Shigeaki Tagashira, the original porter of nfe(4), who submitted lots of patches, performed uncountable number of regression tests and maintained nfe(4) for a long time. Without his enthusiastic help and support I could never have completed this overhauling task. The only weak point of nfe(4) compared to nve(4) is instability of manual half-duplex media selection on certain hardwares(auto sensing media type should work for all cases, though). This was a long standing bug of nfe(4) and I still have no idea why it doesn't work on some hardwares. Obtained from: OpenBSD [1] Submitted by: Peer Chen < pchen at nvidia dot com > [2], [3] Reviewed by: Shigeaki Tagashira < shigeaki AT se DOT hiroshima-u DOT ac DOT jp > Tested by: Shigeaki Tagashira, current Discussed with: current Silence from: obrien
2007-06-12 02:16:02 +00:00
uint32_t filter;
uint8_t addr[ETHER_ADDR_LEN], mask[ETHER_ADDR_LEN];
uint8_t etherbroadcastaddr[ETHER_ADDR_LEN] = {
style(9) cleanup.
2006-10-19 10:01:26 +00:00
0xff, 0xff, 0xff, 0xff, 0xff, 0xff
};
Add a pure open source nForce Ethernet driver, under BSDL. This driver was ported from OpenBSD by Shigeaki Tagashira <shigeaki@se.hiroshima-u.ac.jp> and posted at http://www.se.hiroshima-u.ac.jp/~shigeaki/software/freebsd-nfe.html It was additionally cleaned up by me. It is still a work-in-progress and thus is purposefully not in GENERIC. And it conflicts with nve(4), so only one should be loaded.
2006-06-26 23:41:07 +00:00
NFE_LOCK_ASSERT(sc);
if ((ifp->if_flags & (IFF_ALLMULTI | IFF_PROMISC)) != 0) {
bzero(addr, ETHER_ADDR_LEN);
bzero(mask, ETHER_ADDR_LEN);
goto done;
}
bcopy(etherbroadcastaddr, addr, ETHER_ADDR_LEN);
bcopy(etherbroadcastaddr, mask, ETHER_ADDR_LEN);
Use if_maddr_rlock()/if_maddr_runlock() rather than IF_ADDR_LOCK()/ IF_ADDR_UNLOCK() across network device drivers when accessing the per-interface multicast address list, if_multiaddrs. This will allow us to change the locking strategy without affecting our driver programming interface or binary interface. For two wireless drivers, remove unnecessary locking, since they don't actually access the multicast address list. Approved by: re (kib) MFC after: 6 weeks
2009-06-26 11:45:06 +00:00
if_maddr_rlock(ifp);
Add a pure open source nForce Ethernet driver, under BSDL. This driver was ported from OpenBSD by Shigeaki Tagashira <shigeaki@se.hiroshima-u.ac.jp> and posted at http://www.se.hiroshima-u.ac.jp/~shigeaki/software/freebsd-nfe.html It was additionally cleaned up by me. It is still a work-in-progress and thus is purposefully not in GENERIC. And it conflicts with nve(4), so only one should be loaded.
2006-06-26 23:41:07 +00:00
TAILQ_FOREACH(ifma, &ifp->if_multiaddrs, ifma_link) {
u_char *addrp;
if (ifma->ifma_addr->sa_family != AF_LINK)
continue;
addrp = LLADDR((struct sockaddr_dl *) ifma->ifma_addr);
for (i = 0; i < ETHER_ADDR_LEN; i++) {
u_int8_t mcaddr = addrp[i];
addr[i] &= mcaddr;
mask[i] &= ~mcaddr;
}
}
Use if_maddr_rlock()/if_maddr_runlock() rather than IF_ADDR_LOCK()/ IF_ADDR_UNLOCK() across network device drivers when accessing the per-interface multicast address list, if_multiaddrs. This will allow us to change the locking strategy without affecting our driver programming interface or binary interface. For two wireless drivers, remove unnecessary locking, since they don't actually access the multicast address list. Approved by: re (kib) MFC after: 6 weeks
2009-06-26 11:45:06 +00:00
if_maddr_runlock(ifp);
Add a pure open source nForce Ethernet driver, under BSDL. This driver was ported from OpenBSD by Shigeaki Tagashira <shigeaki@se.hiroshima-u.ac.jp> and posted at http://www.se.hiroshima-u.ac.jp/~shigeaki/software/freebsd-nfe.html It was additionally cleaned up by me. It is still a work-in-progress and thus is purposefully not in GENERIC. And it conflicts with nve(4), so only one should be loaded.
2006-06-26 23:41:07 +00:00
for (i = 0; i < ETHER_ADDR_LEN; i++) {
mask[i] |= addr[i];
}
done:
addr[0] |= 0x01; /* make sure multicast bit is set */
NFE_WRITE(sc, NFE_MULTIADDR_HI,
addr[3] << 24 | addr[2] << 16 | addr[1] << 8 | addr[0]);
NFE_WRITE(sc, NFE_MULTIADDR_LO,
addr[5] << 8 | addr[4]);
NFE_WRITE(sc, NFE_MULTIMASK_HI,
mask[3] << 24 | mask[2] << 16 | mask[1] << 8 | mask[0]);
NFE_WRITE(sc, NFE_MULTIMASK_LO,
mask[5] << 8 | mask[4]);
Bring overhauled nfe(4) into tree. o s/printf/device_printf/g o Nuke OpenBSDism. o Nuke NetBSD/OpenBSD specific DMA sync operations.(we don't have a way to sync a single descriptor within a DMA map.) o Remove recursive mutex. o bus_dma(9) clean up. o 40bit DMA address support. o Add protection for Rx map load failure. o Fix a long standing bug for watchdog timeout. [1] o Add additional protections, missing Tx completion interrupt, losing start Tx command, for watchdog timeout. o Switch to taskqueue(9) API to handle interrupts. o Use our own timer for watchdog instead of if_watchdog/if_timer interface. o Advertise VLAN header length/capability correctly to upper layer. o Remove excessive kernel stack consumption in nfe_encap(). o Handle highly fragmented mbuf chains correctly. o Enable etherenet address reprogramming with ifconfig(8). o Add ALTQ/TSO, MSI/MSIX support. o Increased Rx ring to 256 descriptors from 128. o Align Tx/Rx descriptor ring on sizeof(struct nfe_desc64) boundary. o Remove alignment restrictions on Tx/Rx buffers. o Rewritten jumbo frame support code. o Add support for hardware assistend VLAN tag insertion/stripping. o Add support for Tx/Rx flow control based on patches from Peer Chen. [2] o Add a routine that detects whether ethernet address swap routines is required. [3] o Add a workaround that take MAC/PHY out of power down mode. o Add suspend/resume support. o style(9) and code clean up. Special thanks to Shigeaki Tagashira, the original porter of nfe(4), who submitted lots of patches, performed uncountable number of regression tests and maintained nfe(4) for a long time. Without his enthusiastic help and support I could never have completed this overhauling task. The only weak point of nfe(4) compared to nve(4) is instability of manual half-duplex media selection on certain hardwares(auto sensing media type should work for all cases, though). This was a long standing bug of nfe(4) and I still have no idea why it doesn't work on some hardwares. Obtained from: OpenBSD [1] Submitted by: Peer Chen < pchen at nvidia dot com > [2], [3] Reviewed by: Shigeaki Tagashira < shigeaki AT se DOT hiroshima-u DOT ac DOT jp > Tested by: Shigeaki Tagashira, current Discussed with: current Silence from: obrien
2007-06-12 02:16:02 +00:00
filter = NFE_READ(sc, NFE_RXFILTER);
filter &= NFE_PFF_RX_PAUSE;
filter |= NFE_RXFILTER_MAGIC;
filter |= (ifp->if_flags & IFF_PROMISC) ? NFE_PFF_PROMISC : NFE_PFF_U2M;
Add a pure open source nForce Ethernet driver, under BSDL. This driver was ported from OpenBSD by Shigeaki Tagashira <shigeaki@se.hiroshima-u.ac.jp> and posted at http://www.se.hiroshima-u.ac.jp/~shigeaki/software/freebsd-nfe.html It was additionally cleaned up by me. It is still a work-in-progress and thus is purposefully not in GENERIC. And it conflicts with nve(4), so only one should be loaded.
2006-06-26 23:41:07 +00:00
NFE_WRITE(sc, NFE_RXFILTER, filter);
}
Import the nfe (nForce Ethernet) driver from OpenBSD. This is a pure open source NIC driver with a BSD license. These bits were taken from the OpenBSD public CVS repository on 1-May-06.
2006-06-26 17:25:22 +00:00
style(9) cleanup.
2006-10-19 10:01:26 +00:00
static void
Add a 'locked' variant of the foo_start() routine and call it directly from interrupt handlers and watchdog routines instead of queueing a task to call foo_start(). Reviewed by: yongari MFC after: 1 month
2011-01-03 18:28:30 +00:00
nfe_start(struct ifnet *ifp)
Add a pure open source nForce Ethernet driver, under BSDL. This driver was ported from OpenBSD by Shigeaki Tagashira <shigeaki@se.hiroshima-u.ac.jp> and posted at http://www.se.hiroshima-u.ac.jp/~shigeaki/software/freebsd-nfe.html It was additionally cleaned up by me. It is still a work-in-progress and thus is purposefully not in GENERIC. And it conflicts with nve(4), so only one should be loaded.
2006-06-26 23:41:07 +00:00
{
Add a 'locked' variant of the foo_start() routine and call it directly from interrupt handlers and watchdog routines instead of queueing a task to call foo_start(). Reviewed by: yongari MFC after: 1 month
2011-01-03 18:28:30 +00:00
struct nfe_softc *sc = ifp->if_softc;
Import the nfe (nForce Ethernet) driver from OpenBSD. This is a pure open source NIC driver with a BSD license. These bits were taken from the OpenBSD public CVS repository on 1-May-06.
2006-06-26 17:25:22 +00:00
Add a 'locked' variant of the foo_start() routine and call it directly from interrupt handlers and watchdog routines instead of queueing a task to call foo_start(). Reviewed by: yongari MFC after: 1 month
2011-01-03 18:28:30 +00:00
NFE_LOCK(sc);
nfe_start_locked(ifp);
NFE_UNLOCK(sc);
Import the nfe (nForce Ethernet) driver from OpenBSD. This is a pure open source NIC driver with a BSD license. These bits were taken from the OpenBSD public CVS repository on 1-May-06.
2006-06-26 17:25:22 +00:00
}
style(9) cleanup.
2006-10-19 10:01:26 +00:00
static void
Add a 'locked' variant of the foo_start() routine and call it directly from interrupt handlers and watchdog routines instead of queueing a task to call foo_start(). Reviewed by: yongari MFC after: 1 month
2011-01-03 18:28:30 +00:00
nfe_start_locked(struct ifnet *ifp)
Import the nfe (nForce Ethernet) driver from OpenBSD. This is a pure open source NIC driver with a BSD license. These bits were taken from the OpenBSD public CVS repository on 1-May-06.
2006-06-26 17:25:22 +00:00
{
struct nfe_softc *sc = ifp->if_softc;
struct mbuf *m0;
Bring overhauled nfe(4) into tree. o s/printf/device_printf/g o Nuke OpenBSDism. o Nuke NetBSD/OpenBSD specific DMA sync operations.(we don't have a way to sync a single descriptor within a DMA map.) o Remove recursive mutex. o bus_dma(9) clean up. o 40bit DMA address support. o Add protection for Rx map load failure. o Fix a long standing bug for watchdog timeout. [1] o Add additional protections, missing Tx completion interrupt, losing start Tx command, for watchdog timeout. o Switch to taskqueue(9) API to handle interrupts. o Use our own timer for watchdog instead of if_watchdog/if_timer interface. o Advertise VLAN header length/capability correctly to upper layer. o Remove excessive kernel stack consumption in nfe_encap(). o Handle highly fragmented mbuf chains correctly. o Enable etherenet address reprogramming with ifconfig(8). o Add ALTQ/TSO, MSI/MSIX support. o Increased Rx ring to 256 descriptors from 128. o Align Tx/Rx descriptor ring on sizeof(struct nfe_desc64) boundary. o Remove alignment restrictions on Tx/Rx buffers. o Rewritten jumbo frame support code. o Add support for hardware assistend VLAN tag insertion/stripping. o Add support for Tx/Rx flow control based on patches from Peer Chen. [2] o Add a routine that detects whether ethernet address swap routines is required. [3] o Add a workaround that take MAC/PHY out of power down mode. o Add suspend/resume support. o style(9) and code clean up. Special thanks to Shigeaki Tagashira, the original porter of nfe(4), who submitted lots of patches, performed uncountable number of regression tests and maintained nfe(4) for a long time. Without his enthusiastic help and support I could never have completed this overhauling task. The only weak point of nfe(4) compared to nve(4) is instability of manual half-duplex media selection on certain hardwares(auto sensing media type should work for all cases, though). This was a long standing bug of nfe(4) and I still have no idea why it doesn't work on some hardwares. Obtained from: OpenBSD [1] Submitted by: Peer Chen < pchen at nvidia dot com > [2], [3] Reviewed by: Shigeaki Tagashira < shigeaki AT se DOT hiroshima-u DOT ac DOT jp > Tested by: Shigeaki Tagashira, current Discussed with: current Silence from: obrien
2007-06-12 02:16:02 +00:00
int enq;
Add a 'locked' variant of the foo_start() routine and call it directly from interrupt handlers and watchdog routines instead of queueing a task to call foo_start(). Reviewed by: yongari MFC after: 1 month
2011-01-03 18:28:30 +00:00
NFE_LOCK_ASSERT(sc);
Import the nfe (nForce Ethernet) driver from OpenBSD. This is a pure open source NIC driver with a BSD license. These bits were taken from the OpenBSD public CVS repository on 1-May-06.
2006-06-26 17:25:22 +00:00
Bring overhauled nfe(4) into tree. o s/printf/device_printf/g o Nuke OpenBSDism. o Nuke NetBSD/OpenBSD specific DMA sync operations.(we don't have a way to sync a single descriptor within a DMA map.) o Remove recursive mutex. o bus_dma(9) clean up. o 40bit DMA address support. o Add protection for Rx map load failure. o Fix a long standing bug for watchdog timeout. [1] o Add additional protections, missing Tx completion interrupt, losing start Tx command, for watchdog timeout. o Switch to taskqueue(9) API to handle interrupts. o Use our own timer for watchdog instead of if_watchdog/if_timer interface. o Advertise VLAN header length/capability correctly to upper layer. o Remove excessive kernel stack consumption in nfe_encap(). o Handle highly fragmented mbuf chains correctly. o Enable etherenet address reprogramming with ifconfig(8). o Add ALTQ/TSO, MSI/MSIX support. o Increased Rx ring to 256 descriptors from 128. o Align Tx/Rx descriptor ring on sizeof(struct nfe_desc64) boundary. o Remove alignment restrictions on Tx/Rx buffers. o Rewritten jumbo frame support code. o Add support for hardware assistend VLAN tag insertion/stripping. o Add support for Tx/Rx flow control based on patches from Peer Chen. [2] o Add a routine that detects whether ethernet address swap routines is required. [3] o Add a workaround that take MAC/PHY out of power down mode. o Add suspend/resume support. o style(9) and code clean up. Special thanks to Shigeaki Tagashira, the original porter of nfe(4), who submitted lots of patches, performed uncountable number of regression tests and maintained nfe(4) for a long time. Without his enthusiastic help and support I could never have completed this overhauling task. The only weak point of nfe(4) compared to nve(4) is instability of manual half-duplex media selection on certain hardwares(auto sensing media type should work for all cases, though). This was a long standing bug of nfe(4) and I still have no idea why it doesn't work on some hardwares. Obtained from: OpenBSD [1] Submitted by: Peer Chen < pchen at nvidia dot com > [2], [3] Reviewed by: Shigeaki Tagashira < shigeaki AT se DOT hiroshima-u DOT ac DOT jp > Tested by: Shigeaki Tagashira, current Discussed with: current Silence from: obrien
2007-06-12 02:16:02 +00:00
if ((ifp->if_drv_flags & (IFF_DRV_RUNNING | IFF_DRV_OACTIVE)) !=
Add a 'locked' variant of the foo_start() routine and call it directly from interrupt handlers and watchdog routines instead of queueing a task to call foo_start(). Reviewed by: yongari MFC after: 1 month
2011-01-03 18:28:30 +00:00
IFF_DRV_RUNNING || sc->nfe_link == 0)
Add a pure open source nForce Ethernet driver, under BSDL. This driver was ported from OpenBSD by Shigeaki Tagashira <shigeaki@se.hiroshima-u.ac.jp> and posted at http://www.se.hiroshima-u.ac.jp/~shigeaki/software/freebsd-nfe.html It was additionally cleaned up by me. It is still a work-in-progress and thus is purposefully not in GENERIC. And it conflicts with nve(4), so only one should be loaded.
2006-06-26 23:41:07 +00:00
return;
Bring overhauled nfe(4) into tree. o s/printf/device_printf/g o Nuke OpenBSDism. o Nuke NetBSD/OpenBSD specific DMA sync operations.(we don't have a way to sync a single descriptor within a DMA map.) o Remove recursive mutex. o bus_dma(9) clean up. o 40bit DMA address support. o Add protection for Rx map load failure. o Fix a long standing bug for watchdog timeout. [1] o Add additional protections, missing Tx completion interrupt, losing start Tx command, for watchdog timeout. o Switch to taskqueue(9) API to handle interrupts. o Use our own timer for watchdog instead of if_watchdog/if_timer interface. o Advertise VLAN header length/capability correctly to upper layer. o Remove excessive kernel stack consumption in nfe_encap(). o Handle highly fragmented mbuf chains correctly. o Enable etherenet address reprogramming with ifconfig(8). o Add ALTQ/TSO, MSI/MSIX support. o Increased Rx ring to 256 descriptors from 128. o Align Tx/Rx descriptor ring on sizeof(struct nfe_desc64) boundary. o Remove alignment restrictions on Tx/Rx buffers. o Rewritten jumbo frame support code. o Add support for hardware assistend VLAN tag insertion/stripping. o Add support for Tx/Rx flow control based on patches from Peer Chen. [2] o Add a routine that detects whether ethernet address swap routines is required. [3] o Add a workaround that take MAC/PHY out of power down mode. o Add suspend/resume support. o style(9) and code clean up. Special thanks to Shigeaki Tagashira, the original porter of nfe(4), who submitted lots of patches, performed uncountable number of regression tests and maintained nfe(4) for a long time. Without his enthusiastic help and support I could never have completed this overhauling task. The only weak point of nfe(4) compared to nve(4) is instability of manual half-duplex media selection on certain hardwares(auto sensing media type should work for all cases, though). This was a long standing bug of nfe(4) and I still have no idea why it doesn't work on some hardwares. Obtained from: OpenBSD [1] Submitted by: Peer Chen < pchen at nvidia dot com > [2], [3] Reviewed by: Shigeaki Tagashira < shigeaki AT se DOT hiroshima-u DOT ac DOT jp > Tested by: Shigeaki Tagashira, current Discussed with: current Silence from: obrien
2007-06-12 02:16:02 +00:00
for (enq = 0; !IFQ_DRV_IS_EMPTY(&ifp->if_snd);) {
IFQ_DRV_DEQUEUE(&ifp->if_snd, m0);
Import the nfe (nForce Ethernet) driver from OpenBSD. This is a pure open source NIC driver with a BSD license. These bits were taken from the OpenBSD public CVS repository on 1-May-06.
2006-06-26 17:25:22 +00:00
if (m0 == NULL)
break;
Bring overhauled nfe(4) into tree. o s/printf/device_printf/g o Nuke OpenBSDism. o Nuke NetBSD/OpenBSD specific DMA sync operations.(we don't have a way to sync a single descriptor within a DMA map.) o Remove recursive mutex. o bus_dma(9) clean up. o 40bit DMA address support. o Add protection for Rx map load failure. o Fix a long standing bug for watchdog timeout. [1] o Add additional protections, missing Tx completion interrupt, losing start Tx command, for watchdog timeout. o Switch to taskqueue(9) API to handle interrupts. o Use our own timer for watchdog instead of if_watchdog/if_timer interface. o Advertise VLAN header length/capability correctly to upper layer. o Remove excessive kernel stack consumption in nfe_encap(). o Handle highly fragmented mbuf chains correctly. o Enable etherenet address reprogramming with ifconfig(8). o Add ALTQ/TSO, MSI/MSIX support. o Increased Rx ring to 256 descriptors from 128. o Align Tx/Rx descriptor ring on sizeof(struct nfe_desc64) boundary. o Remove alignment restrictions on Tx/Rx buffers. o Rewritten jumbo frame support code. o Add support for hardware assistend VLAN tag insertion/stripping. o Add support for Tx/Rx flow control based on patches from Peer Chen. [2] o Add a routine that detects whether ethernet address swap routines is required. [3] o Add a workaround that take MAC/PHY out of power down mode. o Add suspend/resume support. o style(9) and code clean up. Special thanks to Shigeaki Tagashira, the original porter of nfe(4), who submitted lots of patches, performed uncountable number of regression tests and maintained nfe(4) for a long time. Without his enthusiastic help and support I could never have completed this overhauling task. The only weak point of nfe(4) compared to nve(4) is instability of manual half-duplex media selection on certain hardwares(auto sensing media type should work for all cases, though). This was a long standing bug of nfe(4) and I still have no idea why it doesn't work on some hardwares. Obtained from: OpenBSD [1] Submitted by: Peer Chen < pchen at nvidia dot com > [2], [3] Reviewed by: Shigeaki Tagashira < shigeaki AT se DOT hiroshima-u DOT ac DOT jp > Tested by: Shigeaki Tagashira, current Discussed with: current Silence from: obrien
2007-06-12 02:16:02 +00:00
if (nfe_encap(sc, &m0) != 0) {
if (m0 == NULL)
break;
IFQ_DRV_PREPEND(&ifp->if_snd, m0);
Add a pure open source nForce Ethernet driver, under BSDL. This driver was ported from OpenBSD by Shigeaki Tagashira <shigeaki@se.hiroshima-u.ac.jp> and posted at http://www.se.hiroshima-u.ac.jp/~shigeaki/software/freebsd-nfe.html It was additionally cleaned up by me. It is still a work-in-progress and thus is purposefully not in GENERIC. And it conflicts with nve(4), so only one should be loaded.
2006-06-26 23:41:07 +00:00
ifp->if_drv_flags |= IFF_DRV_OACTIVE;
Import the nfe (nForce Ethernet) driver from OpenBSD. This is a pure open source NIC driver with a BSD license. These bits were taken from the OpenBSD public CVS repository on 1-May-06.
2006-06-26 17:25:22 +00:00
break;
}
Bring overhauled nfe(4) into tree. o s/printf/device_printf/g o Nuke OpenBSDism. o Nuke NetBSD/OpenBSD specific DMA sync operations.(we don't have a way to sync a single descriptor within a DMA map.) o Remove recursive mutex. o bus_dma(9) clean up. o 40bit DMA address support. o Add protection for Rx map load failure. o Fix a long standing bug for watchdog timeout. [1] o Add additional protections, missing Tx completion interrupt, losing start Tx command, for watchdog timeout. o Switch to taskqueue(9) API to handle interrupts. o Use our own timer for watchdog instead of if_watchdog/if_timer interface. o Advertise VLAN header length/capability correctly to upper layer. o Remove excessive kernel stack consumption in nfe_encap(). o Handle highly fragmented mbuf chains correctly. o Enable etherenet address reprogramming with ifconfig(8). o Add ALTQ/TSO, MSI/MSIX support. o Increased Rx ring to 256 descriptors from 128. o Align Tx/Rx descriptor ring on sizeof(struct nfe_desc64) boundary. o Remove alignment restrictions on Tx/Rx buffers. o Rewritten jumbo frame support code. o Add support for hardware assistend VLAN tag insertion/stripping. o Add support for Tx/Rx flow control based on patches from Peer Chen. [2] o Add a routine that detects whether ethernet address swap routines is required. [3] o Add a workaround that take MAC/PHY out of power down mode. o Add suspend/resume support. o style(9) and code clean up. Special thanks to Shigeaki Tagashira, the original porter of nfe(4), who submitted lots of patches, performed uncountable number of regression tests and maintained nfe(4) for a long time. Without his enthusiastic help and support I could never have completed this overhauling task. The only weak point of nfe(4) compared to nve(4) is instability of manual half-duplex media selection on certain hardwares(auto sensing media type should work for all cases, though). This was a long standing bug of nfe(4) and I still have no idea why it doesn't work on some hardwares. Obtained from: OpenBSD [1] Submitted by: Peer Chen < pchen at nvidia dot com > [2], [3] Reviewed by: Shigeaki Tagashira < shigeaki AT se DOT hiroshima-u DOT ac DOT jp > Tested by: Shigeaki Tagashira, current Discussed with: current Silence from: obrien
2007-06-12 02:16:02 +00:00
enq++;
Catch up the rest of the drivers with the ether_vlan_mtap modifications. If these drivers are setting M_VLANTAG because they are stripping the layer 2 802.1Q headers, then they need to be re-inserting them so any bpf(4) peers can properly decode them. It should be noted that this is compiled tested only. MFC after: 3 weeks
2007-03-04 03:38:08 +00:00
ETHER_BPF_MTAP(ifp, m0);
Import the nfe (nForce Ethernet) driver from OpenBSD. This is a pure open source NIC driver with a BSD license. These bits were taken from the OpenBSD public CVS repository on 1-May-06.
2006-06-26 17:25:22 +00:00
}
Bring overhauled nfe(4) into tree. o s/printf/device_printf/g o Nuke OpenBSDism. o Nuke NetBSD/OpenBSD specific DMA sync operations.(we don't have a way to sync a single descriptor within a DMA map.) o Remove recursive mutex. o bus_dma(9) clean up. o 40bit DMA address support. o Add protection for Rx map load failure. o Fix a long standing bug for watchdog timeout. [1] o Add additional protections, missing Tx completion interrupt, losing start Tx command, for watchdog timeout. o Switch to taskqueue(9) API to handle interrupts. o Use our own timer for watchdog instead of if_watchdog/if_timer interface. o Advertise VLAN header length/capability correctly to upper layer. o Remove excessive kernel stack consumption in nfe_encap(). o Handle highly fragmented mbuf chains correctly. o Enable etherenet address reprogramming with ifconfig(8). o Add ALTQ/TSO, MSI/MSIX support. o Increased Rx ring to 256 descriptors from 128. o Align Tx/Rx descriptor ring on sizeof(struct nfe_desc64) boundary. o Remove alignment restrictions on Tx/Rx buffers. o Rewritten jumbo frame support code. o Add support for hardware assistend VLAN tag insertion/stripping. o Add support for Tx/Rx flow control based on patches from Peer Chen. [2] o Add a routine that detects whether ethernet address swap routines is required. [3] o Add a workaround that take MAC/PHY out of power down mode. o Add suspend/resume support. o style(9) and code clean up. Special thanks to Shigeaki Tagashira, the original porter of nfe(4), who submitted lots of patches, performed uncountable number of regression tests and maintained nfe(4) for a long time. Without his enthusiastic help and support I could never have completed this overhauling task. The only weak point of nfe(4) compared to nve(4) is instability of manual half-duplex media selection on certain hardwares(auto sensing media type should work for all cases, though). This was a long standing bug of nfe(4) and I still have no idea why it doesn't work on some hardwares. Obtained from: OpenBSD [1] Submitted by: Peer Chen < pchen at nvidia dot com > [2], [3] Reviewed by: Shigeaki Tagashira < shigeaki AT se DOT hiroshima-u DOT ac DOT jp > Tested by: Shigeaki Tagashira, current Discussed with: current Silence from: obrien
2007-06-12 02:16:02 +00:00
if (enq > 0) {
bus_dmamap_sync(sc->txq.tx_desc_tag, sc->txq.tx_desc_map,
BUS_DMASYNC_PREREAD | BUS_DMASYNC_PREWRITE);
Import the nfe (nForce Ethernet) driver from OpenBSD. This is a pure open source NIC driver with a BSD license. These bits were taken from the OpenBSD public CVS repository on 1-May-06.
2006-06-26 17:25:22 +00:00
Bring overhauled nfe(4) into tree. o s/printf/device_printf/g o Nuke OpenBSDism. o Nuke NetBSD/OpenBSD specific DMA sync operations.(we don't have a way to sync a single descriptor within a DMA map.) o Remove recursive mutex. o bus_dma(9) clean up. o 40bit DMA address support. o Add protection for Rx map load failure. o Fix a long standing bug for watchdog timeout. [1] o Add additional protections, missing Tx completion interrupt, losing start Tx command, for watchdog timeout. o Switch to taskqueue(9) API to handle interrupts. o Use our own timer for watchdog instead of if_watchdog/if_timer interface. o Advertise VLAN header length/capability correctly to upper layer. o Remove excessive kernel stack consumption in nfe_encap(). o Handle highly fragmented mbuf chains correctly. o Enable etherenet address reprogramming with ifconfig(8). o Add ALTQ/TSO, MSI/MSIX support. o Increased Rx ring to 256 descriptors from 128. o Align Tx/Rx descriptor ring on sizeof(struct nfe_desc64) boundary. o Remove alignment restrictions on Tx/Rx buffers. o Rewritten jumbo frame support code. o Add support for hardware assistend VLAN tag insertion/stripping. o Add support for Tx/Rx flow control based on patches from Peer Chen. [2] o Add a routine that detects whether ethernet address swap routines is required. [3] o Add a workaround that take MAC/PHY out of power down mode. o Add suspend/resume support. o style(9) and code clean up. Special thanks to Shigeaki Tagashira, the original porter of nfe(4), who submitted lots of patches, performed uncountable number of regression tests and maintained nfe(4) for a long time. Without his enthusiastic help and support I could never have completed this overhauling task. The only weak point of nfe(4) compared to nve(4) is instability of manual half-duplex media selection on certain hardwares(auto sensing media type should work for all cases, though). This was a long standing bug of nfe(4) and I still have no idea why it doesn't work on some hardwares. Obtained from: OpenBSD [1] Submitted by: Peer Chen < pchen at nvidia dot com > [2], [3] Reviewed by: Shigeaki Tagashira < shigeaki AT se DOT hiroshima-u DOT ac DOT jp > Tested by: Shigeaki Tagashira, current Discussed with: current Silence from: obrien
2007-06-12 02:16:02 +00:00
/* kick Tx */
NFE_WRITE(sc, NFE_RXTX_CTL, NFE_RXTX_KICKTX | sc->rxtxctl);
Import the nfe (nForce Ethernet) driver from OpenBSD. This is a pure open source NIC driver with a BSD license. These bits were taken from the OpenBSD public CVS repository on 1-May-06.
2006-06-26 17:25:22 +00:00
Bring overhauled nfe(4) into tree. o s/printf/device_printf/g o Nuke OpenBSDism. o Nuke NetBSD/OpenBSD specific DMA sync operations.(we don't have a way to sync a single descriptor within a DMA map.) o Remove recursive mutex. o bus_dma(9) clean up. o 40bit DMA address support. o Add protection for Rx map load failure. o Fix a long standing bug for watchdog timeout. [1] o Add additional protections, missing Tx completion interrupt, losing start Tx command, for watchdog timeout. o Switch to taskqueue(9) API to handle interrupts. o Use our own timer for watchdog instead of if_watchdog/if_timer interface. o Advertise VLAN header length/capability correctly to upper layer. o Remove excessive kernel stack consumption in nfe_encap(). o Handle highly fragmented mbuf chains correctly. o Enable etherenet address reprogramming with ifconfig(8). o Add ALTQ/TSO, MSI/MSIX support. o Increased Rx ring to 256 descriptors from 128. o Align Tx/Rx descriptor ring on sizeof(struct nfe_desc64) boundary. o Remove alignment restrictions on Tx/Rx buffers. o Rewritten jumbo frame support code. o Add support for hardware assistend VLAN tag insertion/stripping. o Add support for Tx/Rx flow control based on patches from Peer Chen. [2] o Add a routine that detects whether ethernet address swap routines is required. [3] o Add a workaround that take MAC/PHY out of power down mode. o Add suspend/resume support. o style(9) and code clean up. Special thanks to Shigeaki Tagashira, the original porter of nfe(4), who submitted lots of patches, performed uncountable number of regression tests and maintained nfe(4) for a long time. Without his enthusiastic help and support I could never have completed this overhauling task. The only weak point of nfe(4) compared to nve(4) is instability of manual half-duplex media selection on certain hardwares(auto sensing media type should work for all cases, though). This was a long standing bug of nfe(4) and I still have no idea why it doesn't work on some hardwares. Obtained from: OpenBSD [1] Submitted by: Peer Chen < pchen at nvidia dot com > [2], [3] Reviewed by: Shigeaki Tagashira < shigeaki AT se DOT hiroshima-u DOT ac DOT jp > Tested by: Shigeaki Tagashira, current Discussed with: current Silence from: obrien
2007-06-12 02:16:02 +00:00
/*
* Set a timeout in case the chip goes out to lunch.
*/
sc->nfe_watchdog_timer = 5;
}
Import the nfe (nForce Ethernet) driver from OpenBSD. This is a pure open source NIC driver with a BSD license. These bits were taken from the OpenBSD public CVS repository on 1-May-06.
2006-06-26 17:25:22 +00:00
}
style(9) cleanup.
2006-10-19 10:01:26 +00:00
static void
nfe_watchdog(struct ifnet *ifp)
Import the nfe (nForce Ethernet) driver from OpenBSD. This is a pure open source NIC driver with a BSD license. These bits were taken from the OpenBSD public CVS repository on 1-May-06.
2006-06-26 17:25:22 +00:00
{
struct nfe_softc *sc = ifp->if_softc;
Bring overhauled nfe(4) into tree. o s/printf/device_printf/g o Nuke OpenBSDism. o Nuke NetBSD/OpenBSD specific DMA sync operations.(we don't have a way to sync a single descriptor within a DMA map.) o Remove recursive mutex. o bus_dma(9) clean up. o 40bit DMA address support. o Add protection for Rx map load failure. o Fix a long standing bug for watchdog timeout. [1] o Add additional protections, missing Tx completion interrupt, losing start Tx command, for watchdog timeout. o Switch to taskqueue(9) API to handle interrupts. o Use our own timer for watchdog instead of if_watchdog/if_timer interface. o Advertise VLAN header length/capability correctly to upper layer. o Remove excessive kernel stack consumption in nfe_encap(). o Handle highly fragmented mbuf chains correctly. o Enable etherenet address reprogramming with ifconfig(8). o Add ALTQ/TSO, MSI/MSIX support. o Increased Rx ring to 256 descriptors from 128. o Align Tx/Rx descriptor ring on sizeof(struct nfe_desc64) boundary. o Remove alignment restrictions on Tx/Rx buffers. o Rewritten jumbo frame support code. o Add support for hardware assistend VLAN tag insertion/stripping. o Add support for Tx/Rx flow control based on patches from Peer Chen. [2] o Add a routine that detects whether ethernet address swap routines is required. [3] o Add a workaround that take MAC/PHY out of power down mode. o Add suspend/resume support. o style(9) and code clean up. Special thanks to Shigeaki Tagashira, the original porter of nfe(4), who submitted lots of patches, performed uncountable number of regression tests and maintained nfe(4) for a long time. Without his enthusiastic help and support I could never have completed this overhauling task. The only weak point of nfe(4) compared to nve(4) is instability of manual half-duplex media selection on certain hardwares(auto sensing media type should work for all cases, though). This was a long standing bug of nfe(4) and I still have no idea why it doesn't work on some hardwares. Obtained from: OpenBSD [1] Submitted by: Peer Chen < pchen at nvidia dot com > [2], [3] Reviewed by: Shigeaki Tagashira < shigeaki AT se DOT hiroshima-u DOT ac DOT jp > Tested by: Shigeaki Tagashira, current Discussed with: current Silence from: obrien
2007-06-12 02:16:02 +00:00
if (sc->nfe_watchdog_timer == 0 || --sc->nfe_watchdog_timer)
return;
/* Check if we've lost Tx completion interrupt. */
nfe_txeof(sc);
if (sc->txq.queued == 0) {
if_printf(ifp, "watchdog timeout (missed Tx interrupts) "
"-- recovering\n");
if (!IFQ_DRV_IS_EMPTY(&ifp->if_snd))
Add a 'locked' variant of the foo_start() routine and call it directly from interrupt handlers and watchdog routines instead of queueing a task to call foo_start(). Reviewed by: yongari MFC after: 1 month
2011-01-03 18:28:30 +00:00
nfe_start_locked(ifp);
Bring overhauled nfe(4) into tree. o s/printf/device_printf/g o Nuke OpenBSDism. o Nuke NetBSD/OpenBSD specific DMA sync operations.(we don't have a way to sync a single descriptor within a DMA map.) o Remove recursive mutex. o bus_dma(9) clean up. o 40bit DMA address support. o Add protection for Rx map load failure. o Fix a long standing bug for watchdog timeout. [1] o Add additional protections, missing Tx completion interrupt, losing start Tx command, for watchdog timeout. o Switch to taskqueue(9) API to handle interrupts. o Use our own timer for watchdog instead of if_watchdog/if_timer interface. o Advertise VLAN header length/capability correctly to upper layer. o Remove excessive kernel stack consumption in nfe_encap(). o Handle highly fragmented mbuf chains correctly. o Enable etherenet address reprogramming with ifconfig(8). o Add ALTQ/TSO, MSI/MSIX support. o Increased Rx ring to 256 descriptors from 128. o Align Tx/Rx descriptor ring on sizeof(struct nfe_desc64) boundary. o Remove alignment restrictions on Tx/Rx buffers. o Rewritten jumbo frame support code. o Add support for hardware assistend VLAN tag insertion/stripping. o Add support for Tx/Rx flow control based on patches from Peer Chen. [2] o Add a routine that detects whether ethernet address swap routines is required. [3] o Add a workaround that take MAC/PHY out of power down mode. o Add suspend/resume support. o style(9) and code clean up. Special thanks to Shigeaki Tagashira, the original porter of nfe(4), who submitted lots of patches, performed uncountable number of regression tests and maintained nfe(4) for a long time. Without his enthusiastic help and support I could never have completed this overhauling task. The only weak point of nfe(4) compared to nve(4) is instability of manual half-duplex media selection on certain hardwares(auto sensing media type should work for all cases, though). This was a long standing bug of nfe(4) and I still have no idea why it doesn't work on some hardwares. Obtained from: OpenBSD [1] Submitted by: Peer Chen < pchen at nvidia dot com > [2], [3] Reviewed by: Shigeaki Tagashira < shigeaki AT se DOT hiroshima-u DOT ac DOT jp > Tested by: Shigeaki Tagashira, current Discussed with: current Silence from: obrien
2007-06-12 02:16:02 +00:00
return;
}
/* Check if we've lost start Tx command. */
sc->nfe_force_tx++;
if (sc->nfe_force_tx <= 3) {
/*
* If this is the case for watchdog timeout, the following
* code should go to nfe_txeof().
*/
NFE_WRITE(sc, NFE_RXTX_CTL, NFE_RXTX_KICKTX | sc->rxtxctl);
return;
}
sc->nfe_force_tx = 0;
if_printf(ifp, "watchdog timeout\n");
Import the nfe (nForce Ethernet) driver from OpenBSD. This is a pure open source NIC driver with a BSD license. These bits were taken from the OpenBSD public CVS repository on 1-May-06.
2006-06-26 17:25:22 +00:00
Add a pure open source nForce Ethernet driver, under BSDL. This driver was ported from OpenBSD by Shigeaki Tagashira <shigeaki@se.hiroshima-u.ac.jp> and posted at http://www.se.hiroshima-u.ac.jp/~shigeaki/software/freebsd-nfe.html It was additionally cleaned up by me. It is still a work-in-progress and thus is purposefully not in GENERIC. And it conflicts with nve(4), so only one should be loaded.
2006-06-26 23:41:07 +00:00
ifp->if_drv_flags &= ~IFF_DRV_RUNNING;
Import the nfe (nForce Ethernet) driver from OpenBSD. This is a pure open source NIC driver with a BSD license. These bits were taken from the OpenBSD public CVS repository on 1-May-06.
2006-06-26 17:25:22 +00:00
ifp->if_oerrors++;
Bring overhauled nfe(4) into tree. o s/printf/device_printf/g o Nuke OpenBSDism. o Nuke NetBSD/OpenBSD specific DMA sync operations.(we don't have a way to sync a single descriptor within a DMA map.) o Remove recursive mutex. o bus_dma(9) clean up. o 40bit DMA address support. o Add protection for Rx map load failure. o Fix a long standing bug for watchdog timeout. [1] o Add additional protections, missing Tx completion interrupt, losing start Tx command, for watchdog timeout. o Switch to taskqueue(9) API to handle interrupts. o Use our own timer for watchdog instead of if_watchdog/if_timer interface. o Advertise VLAN header length/capability correctly to upper layer. o Remove excessive kernel stack consumption in nfe_encap(). o Handle highly fragmented mbuf chains correctly. o Enable etherenet address reprogramming with ifconfig(8). o Add ALTQ/TSO, MSI/MSIX support. o Increased Rx ring to 256 descriptors from 128. o Align Tx/Rx descriptor ring on sizeof(struct nfe_desc64) boundary. o Remove alignment restrictions on Tx/Rx buffers. o Rewritten jumbo frame support code. o Add support for hardware assistend VLAN tag insertion/stripping. o Add support for Tx/Rx flow control based on patches from Peer Chen. [2] o Add a routine that detects whether ethernet address swap routines is required. [3] o Add a workaround that take MAC/PHY out of power down mode. o Add suspend/resume support. o style(9) and code clean up. Special thanks to Shigeaki Tagashira, the original porter of nfe(4), who submitted lots of patches, performed uncountable number of regression tests and maintained nfe(4) for a long time. Without his enthusiastic help and support I could never have completed this overhauling task. The only weak point of nfe(4) compared to nve(4) is instability of manual half-duplex media selection on certain hardwares(auto sensing media type should work for all cases, though). This was a long standing bug of nfe(4) and I still have no idea why it doesn't work on some hardwares. Obtained from: OpenBSD [1] Submitted by: Peer Chen < pchen at nvidia dot com > [2], [3] Reviewed by: Shigeaki Tagashira < shigeaki AT se DOT hiroshima-u DOT ac DOT jp > Tested by: Shigeaki Tagashira, current Discussed with: current Silence from: obrien
2007-06-12 02:16:02 +00:00
nfe_init_locked(sc);
Import the nfe (nForce Ethernet) driver from OpenBSD. This is a pure open source NIC driver with a BSD license. These bits were taken from the OpenBSD public CVS repository on 1-May-06.
2006-06-26 17:25:22 +00:00
}
style(9) cleanup.
2006-10-19 10:01:26 +00:00
static void
nfe_init(void *xsc)
Import the nfe (nForce Ethernet) driver from OpenBSD. This is a pure open source NIC driver with a BSD license. These bits were taken from the OpenBSD public CVS repository on 1-May-06.
2006-06-26 17:25:22 +00:00
{
Add a pure open source nForce Ethernet driver, under BSDL. This driver was ported from OpenBSD by Shigeaki Tagashira <shigeaki@se.hiroshima-u.ac.jp> and posted at http://www.se.hiroshima-u.ac.jp/~shigeaki/software/freebsd-nfe.html It was additionally cleaned up by me. It is still a work-in-progress and thus is purposefully not in GENERIC. And it conflicts with nve(4), so only one should be loaded.
2006-06-26 23:41:07 +00:00
struct nfe_softc *sc = xsc;
NFE_LOCK(sc);
nfe_init_locked(sc);
NFE_UNLOCK(sc);
}
style(9) cleanup.
2006-10-19 10:01:26 +00:00
static void
nfe_init_locked(void *xsc)
Add a pure open source nForce Ethernet driver, under BSDL. This driver was ported from OpenBSD by Shigeaki Tagashira <shigeaki@se.hiroshima-u.ac.jp> and posted at http://www.se.hiroshima-u.ac.jp/~shigeaki/software/freebsd-nfe.html It was additionally cleaned up by me. It is still a work-in-progress and thus is purposefully not in GENERIC. And it conflicts with nve(4), so only one should be loaded.
2006-06-26 23:41:07 +00:00
{
struct nfe_softc *sc = xsc;
struct ifnet *ifp = sc->nfe_ifp;
struct mii_data *mii;
Bring overhauled nfe(4) into tree. o s/printf/device_printf/g o Nuke OpenBSDism. o Nuke NetBSD/OpenBSD specific DMA sync operations.(we don't have a way to sync a single descriptor within a DMA map.) o Remove recursive mutex. o bus_dma(9) clean up. o 40bit DMA address support. o Add protection for Rx map load failure. o Fix a long standing bug for watchdog timeout. [1] o Add additional protections, missing Tx completion interrupt, losing start Tx command, for watchdog timeout. o Switch to taskqueue(9) API to handle interrupts. o Use our own timer for watchdog instead of if_watchdog/if_timer interface. o Advertise VLAN header length/capability correctly to upper layer. o Remove excessive kernel stack consumption in nfe_encap(). o Handle highly fragmented mbuf chains correctly. o Enable etherenet address reprogramming with ifconfig(8). o Add ALTQ/TSO, MSI/MSIX support. o Increased Rx ring to 256 descriptors from 128. o Align Tx/Rx descriptor ring on sizeof(struct nfe_desc64) boundary. o Remove alignment restrictions on Tx/Rx buffers. o Rewritten jumbo frame support code. o Add support for hardware assistend VLAN tag insertion/stripping. o Add support for Tx/Rx flow control based on patches from Peer Chen. [2] o Add a routine that detects whether ethernet address swap routines is required. [3] o Add a workaround that take MAC/PHY out of power down mode. o Add suspend/resume support. o style(9) and code clean up. Special thanks to Shigeaki Tagashira, the original porter of nfe(4), who submitted lots of patches, performed uncountable number of regression tests and maintained nfe(4) for a long time. Without his enthusiastic help and support I could never have completed this overhauling task. The only weak point of nfe(4) compared to nve(4) is instability of manual half-duplex media selection on certain hardwares(auto sensing media type should work for all cases, though). This was a long standing bug of nfe(4) and I still have no idea why it doesn't work on some hardwares. Obtained from: OpenBSD [1] Submitted by: Peer Chen < pchen at nvidia dot com > [2], [3] Reviewed by: Shigeaki Tagashira < shigeaki AT se DOT hiroshima-u DOT ac DOT jp > Tested by: Shigeaki Tagashira, current Discussed with: current Silence from: obrien
2007-06-12 02:16:02 +00:00
uint32_t val;
int error;
Add a pure open source nForce Ethernet driver, under BSDL. This driver was ported from OpenBSD by Shigeaki Tagashira <shigeaki@se.hiroshima-u.ac.jp> and posted at http://www.se.hiroshima-u.ac.jp/~shigeaki/software/freebsd-nfe.html It was additionally cleaned up by me. It is still a work-in-progress and thus is purposefully not in GENERIC. And it conflicts with nve(4), so only one should be loaded.
2006-06-26 23:41:07 +00:00
NFE_LOCK_ASSERT(sc);
mii = device_get_softc(sc->nfe_miibus);
Bring overhauled nfe(4) into tree. o s/printf/device_printf/g o Nuke OpenBSDism. o Nuke NetBSD/OpenBSD specific DMA sync operations.(we don't have a way to sync a single descriptor within a DMA map.) o Remove recursive mutex. o bus_dma(9) clean up. o 40bit DMA address support. o Add protection for Rx map load failure. o Fix a long standing bug for watchdog timeout. [1] o Add additional protections, missing Tx completion interrupt, losing start Tx command, for watchdog timeout. o Switch to taskqueue(9) API to handle interrupts. o Use our own timer for watchdog instead of if_watchdog/if_timer interface. o Advertise VLAN header length/capability correctly to upper layer. o Remove excessive kernel stack consumption in nfe_encap(). o Handle highly fragmented mbuf chains correctly. o Enable etherenet address reprogramming with ifconfig(8). o Add ALTQ/TSO, MSI/MSIX support. o Increased Rx ring to 256 descriptors from 128. o Align Tx/Rx descriptor ring on sizeof(struct nfe_desc64) boundary. o Remove alignment restrictions on Tx/Rx buffers. o Rewritten jumbo frame support code. o Add support for hardware assistend VLAN tag insertion/stripping. o Add support for Tx/Rx flow control based on patches from Peer Chen. [2] o Add a routine that detects whether ethernet address swap routines is required. [3] o Add a workaround that take MAC/PHY out of power down mode. o Add suspend/resume support. o style(9) and code clean up. Special thanks to Shigeaki Tagashira, the original porter of nfe(4), who submitted lots of patches, performed uncountable number of regression tests and maintained nfe(4) for a long time. Without his enthusiastic help and support I could never have completed this overhauling task. The only weak point of nfe(4) compared to nve(4) is instability of manual half-duplex media selection on certain hardwares(auto sensing media type should work for all cases, though). This was a long standing bug of nfe(4) and I still have no idea why it doesn't work on some hardwares. Obtained from: OpenBSD [1] Submitted by: Peer Chen < pchen at nvidia dot com > [2], [3] Reviewed by: Shigeaki Tagashira < shigeaki AT se DOT hiroshima-u DOT ac DOT jp > Tested by: Shigeaki Tagashira, current Discussed with: current Silence from: obrien
2007-06-12 02:16:02 +00:00
if (ifp->if_drv_flags & IFF_DRV_RUNNING)
Add a pure open source nForce Ethernet driver, under BSDL. This driver was ported from OpenBSD by Shigeaki Tagashira <shigeaki@se.hiroshima-u.ac.jp> and posted at http://www.se.hiroshima-u.ac.jp/~shigeaki/software/freebsd-nfe.html It was additionally cleaned up by me. It is still a work-in-progress and thus is purposefully not in GENERIC. And it conflicts with nve(4), so only one should be loaded.
2006-06-26 23:41:07 +00:00
return;
Import the nfe (nForce Ethernet) driver from OpenBSD. This is a pure open source NIC driver with a BSD license. These bits were taken from the OpenBSD public CVS repository on 1-May-06.
2006-06-26 17:25:22 +00:00
Bring overhauled nfe(4) into tree. o s/printf/device_printf/g o Nuke OpenBSDism. o Nuke NetBSD/OpenBSD specific DMA sync operations.(we don't have a way to sync a single descriptor within a DMA map.) o Remove recursive mutex. o bus_dma(9) clean up. o 40bit DMA address support. o Add protection for Rx map load failure. o Fix a long standing bug for watchdog timeout. [1] o Add additional protections, missing Tx completion interrupt, losing start Tx command, for watchdog timeout. o Switch to taskqueue(9) API to handle interrupts. o Use our own timer for watchdog instead of if_watchdog/if_timer interface. o Advertise VLAN header length/capability correctly to upper layer. o Remove excessive kernel stack consumption in nfe_encap(). o Handle highly fragmented mbuf chains correctly. o Enable etherenet address reprogramming with ifconfig(8). o Add ALTQ/TSO, MSI/MSIX support. o Increased Rx ring to 256 descriptors from 128. o Align Tx/Rx descriptor ring on sizeof(struct nfe_desc64) boundary. o Remove alignment restrictions on Tx/Rx buffers. o Rewritten jumbo frame support code. o Add support for hardware assistend VLAN tag insertion/stripping. o Add support for Tx/Rx flow control based on patches from Peer Chen. [2] o Add a routine that detects whether ethernet address swap routines is required. [3] o Add a workaround that take MAC/PHY out of power down mode. o Add suspend/resume support. o style(9) and code clean up. Special thanks to Shigeaki Tagashira, the original porter of nfe(4), who submitted lots of patches, performed uncountable number of regression tests and maintained nfe(4) for a long time. Without his enthusiastic help and support I could never have completed this overhauling task. The only weak point of nfe(4) compared to nve(4) is instability of manual half-duplex media selection on certain hardwares(auto sensing media type should work for all cases, though). This was a long standing bug of nfe(4) and I still have no idea why it doesn't work on some hardwares. Obtained from: OpenBSD [1] Submitted by: Peer Chen < pchen at nvidia dot com > [2], [3] Reviewed by: Shigeaki Tagashira < shigeaki AT se DOT hiroshima-u DOT ac DOT jp > Tested by: Shigeaki Tagashira, current Discussed with: current Silence from: obrien
2007-06-12 02:16:02 +00:00
nfe_stop(ifp);
sc->nfe_framesize = ifp->if_mtu + NFE_RX_HEADERS;
nfe_init_tx_ring(sc, &sc->txq);
if (sc->nfe_framesize > (MCLBYTES - ETHER_HDR_LEN))
error = nfe_init_jrx_ring(sc, &sc->jrxq);
else
error = nfe_init_rx_ring(sc, &sc->rxq);
if (error != 0) {
device_printf(sc->nfe_dev,
"initialization failed: no memory for rx buffers\n");
nfe_stop(ifp);
return;
}
Import the nfe (nForce Ethernet) driver from OpenBSD. This is a pure open source NIC driver with a BSD license. These bits were taken from the OpenBSD public CVS repository on 1-May-06.
2006-06-26 17:25:22 +00:00
Bring overhauled nfe(4) into tree. o s/printf/device_printf/g o Nuke OpenBSDism. o Nuke NetBSD/OpenBSD specific DMA sync operations.(we don't have a way to sync a single descriptor within a DMA map.) o Remove recursive mutex. o bus_dma(9) clean up. o 40bit DMA address support. o Add protection for Rx map load failure. o Fix a long standing bug for watchdog timeout. [1] o Add additional protections, missing Tx completion interrupt, losing start Tx command, for watchdog timeout. o Switch to taskqueue(9) API to handle interrupts. o Use our own timer for watchdog instead of if_watchdog/if_timer interface. o Advertise VLAN header length/capability correctly to upper layer. o Remove excessive kernel stack consumption in nfe_encap(). o Handle highly fragmented mbuf chains correctly. o Enable etherenet address reprogramming with ifconfig(8). o Add ALTQ/TSO, MSI/MSIX support. o Increased Rx ring to 256 descriptors from 128. o Align Tx/Rx descriptor ring on sizeof(struct nfe_desc64) boundary. o Remove alignment restrictions on Tx/Rx buffers. o Rewritten jumbo frame support code. o Add support for hardware assistend VLAN tag insertion/stripping. o Add support for Tx/Rx flow control based on patches from Peer Chen. [2] o Add a routine that detects whether ethernet address swap routines is required. [3] o Add a workaround that take MAC/PHY out of power down mode. o Add suspend/resume support. o style(9) and code clean up. Special thanks to Shigeaki Tagashira, the original porter of nfe(4), who submitted lots of patches, performed uncountable number of regression tests and maintained nfe(4) for a long time. Without his enthusiastic help and support I could never have completed this overhauling task. The only weak point of nfe(4) compared to nve(4) is instability of manual half-duplex media selection on certain hardwares(auto sensing media type should work for all cases, though). This was a long standing bug of nfe(4) and I still have no idea why it doesn't work on some hardwares. Obtained from: OpenBSD [1] Submitted by: Peer Chen < pchen at nvidia dot com > [2], [3] Reviewed by: Shigeaki Tagashira < shigeaki AT se DOT hiroshima-u DOT ac DOT jp > Tested by: Shigeaki Tagashira, current Discussed with: current Silence from: obrien
2007-06-12 02:16:02 +00:00
val = 0;
if ((sc->nfe_flags & NFE_CORRECT_MACADDR) != 0)
val |= NFE_MAC_ADDR_INORDER;
NFE_WRITE(sc, NFE_TX_UNK, val);
Import the nfe (nForce Ethernet) driver from OpenBSD. This is a pure open source NIC driver with a BSD license. These bits were taken from the OpenBSD public CVS repository on 1-May-06.
2006-06-26 17:25:22 +00:00
NFE_WRITE(sc, NFE_STATUS, 0);
Bring overhauled nfe(4) into tree. o s/printf/device_printf/g o Nuke OpenBSDism. o Nuke NetBSD/OpenBSD specific DMA sync operations.(we don't have a way to sync a single descriptor within a DMA map.) o Remove recursive mutex. o bus_dma(9) clean up. o 40bit DMA address support. o Add protection for Rx map load failure. o Fix a long standing bug for watchdog timeout. [1] o Add additional protections, missing Tx completion interrupt, losing start Tx command, for watchdog timeout. o Switch to taskqueue(9) API to handle interrupts. o Use our own timer for watchdog instead of if_watchdog/if_timer interface. o Advertise VLAN header length/capability correctly to upper layer. o Remove excessive kernel stack consumption in nfe_encap(). o Handle highly fragmented mbuf chains correctly. o Enable etherenet address reprogramming with ifconfig(8). o Add ALTQ/TSO, MSI/MSIX support. o Increased Rx ring to 256 descriptors from 128. o Align Tx/Rx descriptor ring on sizeof(struct nfe_desc64) boundary. o Remove alignment restrictions on Tx/Rx buffers. o Rewritten jumbo frame support code. o Add support for hardware assistend VLAN tag insertion/stripping. o Add support for Tx/Rx flow control based on patches from Peer Chen. [2] o Add a routine that detects whether ethernet address swap routines is required. [3] o Add a workaround that take MAC/PHY out of power down mode. o Add suspend/resume support. o style(9) and code clean up. Special thanks to Shigeaki Tagashira, the original porter of nfe(4), who submitted lots of patches, performed uncountable number of regression tests and maintained nfe(4) for a long time. Without his enthusiastic help and support I could never have completed this overhauling task. The only weak point of nfe(4) compared to nve(4) is instability of manual half-duplex media selection on certain hardwares(auto sensing media type should work for all cases, though). This was a long standing bug of nfe(4) and I still have no idea why it doesn't work on some hardwares. Obtained from: OpenBSD [1] Submitted by: Peer Chen < pchen at nvidia dot com > [2], [3] Reviewed by: Shigeaki Tagashira < shigeaki AT se DOT hiroshima-u DOT ac DOT jp > Tested by: Shigeaki Tagashira, current Discussed with: current Silence from: obrien
2007-06-12 02:16:02 +00:00
if ((sc->nfe_flags & NFE_TX_FLOW_CTRL) != 0)
NFE_WRITE(sc, NFE_TX_PAUSE_FRAME, NFE_TX_PAUSE_FRAME_DISABLE);
Import the nfe (nForce Ethernet) driver from OpenBSD. This is a pure open source NIC driver with a BSD license. These bits were taken from the OpenBSD public CVS repository on 1-May-06.
2006-06-26 17:25:22 +00:00
sc->rxtxctl = NFE_RXTX_BIT2;
Add a pure open source nForce Ethernet driver, under BSDL. This driver was ported from OpenBSD by Shigeaki Tagashira <shigeaki@se.hiroshima-u.ac.jp> and posted at http://www.se.hiroshima-u.ac.jp/~shigeaki/software/freebsd-nfe.html It was additionally cleaned up by me. It is still a work-in-progress and thus is purposefully not in GENERIC. And it conflicts with nve(4), so only one should be loaded.
2006-06-26 23:41:07 +00:00
if (sc->nfe_flags & NFE_40BIT_ADDR)
Import the nfe (nForce Ethernet) driver from OpenBSD. This is a pure open source NIC driver with a BSD license. These bits were taken from the OpenBSD public CVS repository on 1-May-06.
2006-06-26 17:25:22 +00:00
sc->rxtxctl |= NFE_RXTX_V3MAGIC;
Add a pure open source nForce Ethernet driver, under BSDL. This driver was ported from OpenBSD by Shigeaki Tagashira <shigeaki@se.hiroshima-u.ac.jp> and posted at http://www.se.hiroshima-u.ac.jp/~shigeaki/software/freebsd-nfe.html It was additionally cleaned up by me. It is still a work-in-progress and thus is purposefully not in GENERIC. And it conflicts with nve(4), so only one should be loaded.
2006-06-26 23:41:07 +00:00
else if (sc->nfe_flags & NFE_JUMBO_SUP)
Import the nfe (nForce Ethernet) driver from OpenBSD. This is a pure open source NIC driver with a BSD license. These bits were taken from the OpenBSD public CVS repository on 1-May-06.
2006-06-26 17:25:22 +00:00
sc->rxtxctl |= NFE_RXTX_V2MAGIC;
Fix TX/RX checksum offload. Submitted by: Peer Chen at nVIDIA
2006-11-27 04:47:27 +00:00
Bring overhauled nfe(4) into tree. o s/printf/device_printf/g o Nuke OpenBSDism. o Nuke NetBSD/OpenBSD specific DMA sync operations.(we don't have a way to sync a single descriptor within a DMA map.) o Remove recursive mutex. o bus_dma(9) clean up. o 40bit DMA address support. o Add protection for Rx map load failure. o Fix a long standing bug for watchdog timeout. [1] o Add additional protections, missing Tx completion interrupt, losing start Tx command, for watchdog timeout. o Switch to taskqueue(9) API to handle interrupts. o Use our own timer for watchdog instead of if_watchdog/if_timer interface. o Advertise VLAN header length/capability correctly to upper layer. o Remove excessive kernel stack consumption in nfe_encap(). o Handle highly fragmented mbuf chains correctly. o Enable etherenet address reprogramming with ifconfig(8). o Add ALTQ/TSO, MSI/MSIX support. o Increased Rx ring to 256 descriptors from 128. o Align Tx/Rx descriptor ring on sizeof(struct nfe_desc64) boundary. o Remove alignment restrictions on Tx/Rx buffers. o Rewritten jumbo frame support code. o Add support for hardware assistend VLAN tag insertion/stripping. o Add support for Tx/Rx flow control based on patches from Peer Chen. [2] o Add a routine that detects whether ethernet address swap routines is required. [3] o Add a workaround that take MAC/PHY out of power down mode. o Add suspend/resume support. o style(9) and code clean up. Special thanks to Shigeaki Tagashira, the original porter of nfe(4), who submitted lots of patches, performed uncountable number of regression tests and maintained nfe(4) for a long time. Without his enthusiastic help and support I could never have completed this overhauling task. The only weak point of nfe(4) compared to nve(4) is instability of manual half-duplex media selection on certain hardwares(auto sensing media type should work for all cases, though). This was a long standing bug of nfe(4) and I still have no idea why it doesn't work on some hardwares. Obtained from: OpenBSD [1] Submitted by: Peer Chen < pchen at nvidia dot com > [2], [3] Reviewed by: Shigeaki Tagashira < shigeaki AT se DOT hiroshima-u DOT ac DOT jp > Tested by: Shigeaki Tagashira, current Discussed with: current Silence from: obrien
2007-06-12 02:16:02 +00:00
if ((ifp->if_capenable & IFCAP_RXCSUM) != 0)
Import the nfe (nForce Ethernet) driver from OpenBSD. This is a pure open source NIC driver with a BSD license. These bits were taken from the OpenBSD public CVS repository on 1-May-06.
2006-06-26 17:25:22 +00:00
sc->rxtxctl |= NFE_RXTX_RXCSUM;
Bring overhauled nfe(4) into tree. o s/printf/device_printf/g o Nuke OpenBSDism. o Nuke NetBSD/OpenBSD specific DMA sync operations.(we don't have a way to sync a single descriptor within a DMA map.) o Remove recursive mutex. o bus_dma(9) clean up. o 40bit DMA address support. o Add protection for Rx map load failure. o Fix a long standing bug for watchdog timeout. [1] o Add additional protections, missing Tx completion interrupt, losing start Tx command, for watchdog timeout. o Switch to taskqueue(9) API to handle interrupts. o Use our own timer for watchdog instead of if_watchdog/if_timer interface. o Advertise VLAN header length/capability correctly to upper layer. o Remove excessive kernel stack consumption in nfe_encap(). o Handle highly fragmented mbuf chains correctly. o Enable etherenet address reprogramming with ifconfig(8). o Add ALTQ/TSO, MSI/MSIX support. o Increased Rx ring to 256 descriptors from 128. o Align Tx/Rx descriptor ring on sizeof(struct nfe_desc64) boundary. o Remove alignment restrictions on Tx/Rx buffers. o Rewritten jumbo frame support code. o Add support for hardware assistend VLAN tag insertion/stripping. o Add support for Tx/Rx flow control based on patches from Peer Chen. [2] o Add a routine that detects whether ethernet address swap routines is required. [3] o Add a workaround that take MAC/PHY out of power down mode. o Add suspend/resume support. o style(9) and code clean up. Special thanks to Shigeaki Tagashira, the original porter of nfe(4), who submitted lots of patches, performed uncountable number of regression tests and maintained nfe(4) for a long time. Without his enthusiastic help and support I could never have completed this overhauling task. The only weak point of nfe(4) compared to nve(4) is instability of manual half-duplex media selection on certain hardwares(auto sensing media type should work for all cases, though). This was a long standing bug of nfe(4) and I still have no idea why it doesn't work on some hardwares. Obtained from: OpenBSD [1] Submitted by: Peer Chen < pchen at nvidia dot com > [2], [3] Reviewed by: Shigeaki Tagashira < shigeaki AT se DOT hiroshima-u DOT ac DOT jp > Tested by: Shigeaki Tagashira, current Discussed with: current Silence from: obrien
2007-06-12 02:16:02 +00:00
if ((ifp->if_capenable & IFCAP_VLAN_HWTAGGING) != 0)
sc->rxtxctl |= NFE_RXTX_VTAG_INSERT | NFE_RXTX_VTAG_STRIP;
Add a pure open source nForce Ethernet driver, under BSDL. This driver was ported from OpenBSD by Shigeaki Tagashira <shigeaki@se.hiroshima-u.ac.jp> and posted at http://www.se.hiroshima-u.ac.jp/~shigeaki/software/freebsd-nfe.html It was additionally cleaned up by me. It is still a work-in-progress and thus is purposefully not in GENERIC. And it conflicts with nve(4), so only one should be loaded.
2006-06-26 23:41:07 +00:00
Import the nfe (nForce Ethernet) driver from OpenBSD. This is a pure open source NIC driver with a BSD license. These bits were taken from the OpenBSD public CVS repository on 1-May-06.
2006-06-26 17:25:22 +00:00
NFE_WRITE(sc, NFE_RXTX_CTL, NFE_RXTX_RESET | sc->rxtxctl);
DELAY(10);
NFE_WRITE(sc, NFE_RXTX_CTL, sc->rxtxctl);
Bring overhauled nfe(4) into tree. o s/printf/device_printf/g o Nuke OpenBSDism. o Nuke NetBSD/OpenBSD specific DMA sync operations.(we don't have a way to sync a single descriptor within a DMA map.) o Remove recursive mutex. o bus_dma(9) clean up. o 40bit DMA address support. o Add protection for Rx map load failure. o Fix a long standing bug for watchdog timeout. [1] o Add additional protections, missing Tx completion interrupt, losing start Tx command, for watchdog timeout. o Switch to taskqueue(9) API to handle interrupts. o Use our own timer for watchdog instead of if_watchdog/if_timer interface. o Advertise VLAN header length/capability correctly to upper layer. o Remove excessive kernel stack consumption in nfe_encap(). o Handle highly fragmented mbuf chains correctly. o Enable etherenet address reprogramming with ifconfig(8). o Add ALTQ/TSO, MSI/MSIX support. o Increased Rx ring to 256 descriptors from 128. o Align Tx/Rx descriptor ring on sizeof(struct nfe_desc64) boundary. o Remove alignment restrictions on Tx/Rx buffers. o Rewritten jumbo frame support code. o Add support for hardware assistend VLAN tag insertion/stripping. o Add support for Tx/Rx flow control based on patches from Peer Chen. [2] o Add a routine that detects whether ethernet address swap routines is required. [3] o Add a workaround that take MAC/PHY out of power down mode. o Add suspend/resume support. o style(9) and code clean up. Special thanks to Shigeaki Tagashira, the original porter of nfe(4), who submitted lots of patches, performed uncountable number of regression tests and maintained nfe(4) for a long time. Without his enthusiastic help and support I could never have completed this overhauling task. The only weak point of nfe(4) compared to nve(4) is instability of manual half-duplex media selection on certain hardwares(auto sensing media type should work for all cases, though). This was a long standing bug of nfe(4) and I still have no idea why it doesn't work on some hardwares. Obtained from: OpenBSD [1] Submitted by: Peer Chen < pchen at nvidia dot com > [2], [3] Reviewed by: Shigeaki Tagashira < shigeaki AT se DOT hiroshima-u DOT ac DOT jp > Tested by: Shigeaki Tagashira, current Discussed with: current Silence from: obrien
2007-06-12 02:16:02 +00:00
if ((ifp->if_capenable & IFCAP_VLAN_HWTAGGING) != 0)
Import the nfe (nForce Ethernet) driver from OpenBSD. This is a pure open source NIC driver with a BSD license. These bits were taken from the OpenBSD public CVS repository on 1-May-06.
2006-06-26 17:25:22 +00:00
NFE_WRITE(sc, NFE_VTAG_CTL, NFE_VTAG_ENABLE);
Bring overhauled nfe(4) into tree. o s/printf/device_printf/g o Nuke OpenBSDism. o Nuke NetBSD/OpenBSD specific DMA sync operations.(we don't have a way to sync a single descriptor within a DMA map.) o Remove recursive mutex. o bus_dma(9) clean up. o 40bit DMA address support. o Add protection for Rx map load failure. o Fix a long standing bug for watchdog timeout. [1] o Add additional protections, missing Tx completion interrupt, losing start Tx command, for watchdog timeout. o Switch to taskqueue(9) API to handle interrupts. o Use our own timer for watchdog instead of if_watchdog/if_timer interface. o Advertise VLAN header length/capability correctly to upper layer. o Remove excessive kernel stack consumption in nfe_encap(). o Handle highly fragmented mbuf chains correctly. o Enable etherenet address reprogramming with ifconfig(8). o Add ALTQ/TSO, MSI/MSIX support. o Increased Rx ring to 256 descriptors from 128. o Align Tx/Rx descriptor ring on sizeof(struct nfe_desc64) boundary. o Remove alignment restrictions on Tx/Rx buffers. o Rewritten jumbo frame support code. o Add support for hardware assistend VLAN tag insertion/stripping. o Add support for Tx/Rx flow control based on patches from Peer Chen. [2] o Add a routine that detects whether ethernet address swap routines is required. [3] o Add a workaround that take MAC/PHY out of power down mode. o Add suspend/resume support. o style(9) and code clean up. Special thanks to Shigeaki Tagashira, the original porter of nfe(4), who submitted lots of patches, performed uncountable number of regression tests and maintained nfe(4) for a long time. Without his enthusiastic help and support I could never have completed this overhauling task. The only weak point of nfe(4) compared to nve(4) is instability of manual half-duplex media selection on certain hardwares(auto sensing media type should work for all cases, though). This was a long standing bug of nfe(4) and I still have no idea why it doesn't work on some hardwares. Obtained from: OpenBSD [1] Submitted by: Peer Chen < pchen at nvidia dot com > [2], [3] Reviewed by: Shigeaki Tagashira < shigeaki AT se DOT hiroshima-u DOT ac DOT jp > Tested by: Shigeaki Tagashira, current Discussed with: current Silence from: obrien
2007-06-12 02:16:02 +00:00
else
NFE_WRITE(sc, NFE_VTAG_CTL, 0);
Import the nfe (nForce Ethernet) driver from OpenBSD. This is a pure open source NIC driver with a BSD license. These bits were taken from the OpenBSD public CVS repository on 1-May-06.
2006-06-26 17:25:22 +00:00
NFE_WRITE(sc, NFE_SETUP_R6, 0);
Add a pure open source nForce Ethernet driver, under BSDL. This driver was ported from OpenBSD by Shigeaki Tagashira <shigeaki@se.hiroshima-u.ac.jp> and posted at http://www.se.hiroshima-u.ac.jp/~shigeaki/software/freebsd-nfe.html It was additionally cleaned up by me. It is still a work-in-progress and thus is purposefully not in GENERIC. And it conflicts with nve(4), so only one should be loaded.
2006-06-26 23:41:07 +00:00
/* set MAC address */
Bring overhauled nfe(4) into tree. o s/printf/device_printf/g o Nuke OpenBSDism. o Nuke NetBSD/OpenBSD specific DMA sync operations.(we don't have a way to sync a single descriptor within a DMA map.) o Remove recursive mutex. o bus_dma(9) clean up. o 40bit DMA address support. o Add protection for Rx map load failure. o Fix a long standing bug for watchdog timeout. [1] o Add additional protections, missing Tx completion interrupt, losing start Tx command, for watchdog timeout. o Switch to taskqueue(9) API to handle interrupts. o Use our own timer for watchdog instead of if_watchdog/if_timer interface. o Advertise VLAN header length/capability correctly to upper layer. o Remove excessive kernel stack consumption in nfe_encap(). o Handle highly fragmented mbuf chains correctly. o Enable etherenet address reprogramming with ifconfig(8). o Add ALTQ/TSO, MSI/MSIX support. o Increased Rx ring to 256 descriptors from 128. o Align Tx/Rx descriptor ring on sizeof(struct nfe_desc64) boundary. o Remove alignment restrictions on Tx/Rx buffers. o Rewritten jumbo frame support code. o Add support for hardware assistend VLAN tag insertion/stripping. o Add support for Tx/Rx flow control based on patches from Peer Chen. [2] o Add a routine that detects whether ethernet address swap routines is required. [3] o Add a workaround that take MAC/PHY out of power down mode. o Add suspend/resume support. o style(9) and code clean up. Special thanks to Shigeaki Tagashira, the original porter of nfe(4), who submitted lots of patches, performed uncountable number of regression tests and maintained nfe(4) for a long time. Without his enthusiastic help and support I could never have completed this overhauling task. The only weak point of nfe(4) compared to nve(4) is instability of manual half-duplex media selection on certain hardwares(auto sensing media type should work for all cases, though). This was a long standing bug of nfe(4) and I still have no idea why it doesn't work on some hardwares. Obtained from: OpenBSD [1] Submitted by: Peer Chen < pchen at nvidia dot com > [2], [3] Reviewed by: Shigeaki Tagashira < shigeaki AT se DOT hiroshima-u DOT ac DOT jp > Tested by: Shigeaki Tagashira, current Discussed with: current Silence from: obrien
2007-06-12 02:16:02 +00:00
nfe_set_macaddr(sc, IF_LLADDR(ifp));
Import the nfe (nForce Ethernet) driver from OpenBSD. This is a pure open source NIC driver with a BSD license. These bits were taken from the OpenBSD public CVS repository on 1-May-06.
2006-06-26 17:25:22 +00:00
Add a pure open source nForce Ethernet driver, under BSDL. This driver was ported from OpenBSD by Shigeaki Tagashira <shigeaki@se.hiroshima-u.ac.jp> and posted at http://www.se.hiroshima-u.ac.jp/~shigeaki/software/freebsd-nfe.html It was additionally cleaned up by me. It is still a work-in-progress and thus is purposefully not in GENERIC. And it conflicts with nve(4), so only one should be loaded.
2006-06-26 23:41:07 +00:00
/* tell MAC where rings are in memory */
Bring overhauled nfe(4) into tree. o s/printf/device_printf/g o Nuke OpenBSDism. o Nuke NetBSD/OpenBSD specific DMA sync operations.(we don't have a way to sync a single descriptor within a DMA map.) o Remove recursive mutex. o bus_dma(9) clean up. o 40bit DMA address support. o Add protection for Rx map load failure. o Fix a long standing bug for watchdog timeout. [1] o Add additional protections, missing Tx completion interrupt, losing start Tx command, for watchdog timeout. o Switch to taskqueue(9) API to handle interrupts. o Use our own timer for watchdog instead of if_watchdog/if_timer interface. o Advertise VLAN header length/capability correctly to upper layer. o Remove excessive kernel stack consumption in nfe_encap(). o Handle highly fragmented mbuf chains correctly. o Enable etherenet address reprogramming with ifconfig(8). o Add ALTQ/TSO, MSI/MSIX support. o Increased Rx ring to 256 descriptors from 128. o Align Tx/Rx descriptor ring on sizeof(struct nfe_desc64) boundary. o Remove alignment restrictions on Tx/Rx buffers. o Rewritten jumbo frame support code. o Add support for hardware assistend VLAN tag insertion/stripping. o Add support for Tx/Rx flow control based on patches from Peer Chen. [2] o Add a routine that detects whether ethernet address swap routines is required. [3] o Add a workaround that take MAC/PHY out of power down mode. o Add suspend/resume support. o style(9) and code clean up. Special thanks to Shigeaki Tagashira, the original porter of nfe(4), who submitted lots of patches, performed uncountable number of regression tests and maintained nfe(4) for a long time. Without his enthusiastic help and support I could never have completed this overhauling task. The only weak point of nfe(4) compared to nve(4) is instability of manual half-duplex media selection on certain hardwares(auto sensing media type should work for all cases, though). This was a long standing bug of nfe(4) and I still have no idea why it doesn't work on some hardwares. Obtained from: OpenBSD [1] Submitted by: Peer Chen < pchen at nvidia dot com > [2], [3] Reviewed by: Shigeaki Tagashira < shigeaki AT se DOT hiroshima-u DOT ac DOT jp > Tested by: Shigeaki Tagashira, current Discussed with: current Silence from: obrien
2007-06-12 02:16:02 +00:00
if (sc->nfe_framesize > MCLBYTES - ETHER_HDR_LEN) {
NFE_WRITE(sc, NFE_RX_RING_ADDR_HI,
NFE_ADDR_HI(sc->jrxq.jphysaddr));
NFE_WRITE(sc, NFE_RX_RING_ADDR_LO,
NFE_ADDR_LO(sc->jrxq.jphysaddr));
} else {
NFE_WRITE(sc, NFE_RX_RING_ADDR_HI,
NFE_ADDR_HI(sc->rxq.physaddr));
NFE_WRITE(sc, NFE_RX_RING_ADDR_LO,
NFE_ADDR_LO(sc->rxq.physaddr));
}
NFE_WRITE(sc, NFE_TX_RING_ADDR_HI, NFE_ADDR_HI(sc->txq.physaddr));
NFE_WRITE(sc, NFE_TX_RING_ADDR_LO, NFE_ADDR_LO(sc->txq.physaddr));
Import the nfe (nForce Ethernet) driver from OpenBSD. This is a pure open source NIC driver with a BSD license. These bits were taken from the OpenBSD public CVS repository on 1-May-06.
2006-06-26 17:25:22 +00:00
Add a pure open source nForce Ethernet driver, under BSDL. This driver was ported from OpenBSD by Shigeaki Tagashira <shigeaki@se.hiroshima-u.ac.jp> and posted at http://www.se.hiroshima-u.ac.jp/~shigeaki/software/freebsd-nfe.html It was additionally cleaned up by me. It is still a work-in-progress and thus is purposefully not in GENERIC. And it conflicts with nve(4), so only one should be loaded.
2006-06-26 23:41:07 +00:00
NFE_WRITE(sc, NFE_RING_SIZE,
(NFE_RX_RING_COUNT - 1) << 16 |
(NFE_TX_RING_COUNT - 1));
Import the nfe (nForce Ethernet) driver from OpenBSD. This is a pure open source NIC driver with a BSD license. These bits were taken from the OpenBSD public CVS repository on 1-May-06.
2006-06-26 17:25:22 +00:00
Bring overhauled nfe(4) into tree. o s/printf/device_printf/g o Nuke OpenBSDism. o Nuke NetBSD/OpenBSD specific DMA sync operations.(we don't have a way to sync a single descriptor within a DMA map.) o Remove recursive mutex. o bus_dma(9) clean up. o 40bit DMA address support. o Add protection for Rx map load failure. o Fix a long standing bug for watchdog timeout. [1] o Add additional protections, missing Tx completion interrupt, losing start Tx command, for watchdog timeout. o Switch to taskqueue(9) API to handle interrupts. o Use our own timer for watchdog instead of if_watchdog/if_timer interface. o Advertise VLAN header length/capability correctly to upper layer. o Remove excessive kernel stack consumption in nfe_encap(). o Handle highly fragmented mbuf chains correctly. o Enable etherenet address reprogramming with ifconfig(8). o Add ALTQ/TSO, MSI/MSIX support. o Increased Rx ring to 256 descriptors from 128. o Align Tx/Rx descriptor ring on sizeof(struct nfe_desc64) boundary. o Remove alignment restrictions on Tx/Rx buffers. o Rewritten jumbo frame support code. o Add support for hardware assistend VLAN tag insertion/stripping. o Add support for Tx/Rx flow control based on patches from Peer Chen. [2] o Add a routine that detects whether ethernet address swap routines is required. [3] o Add a workaround that take MAC/PHY out of power down mode. o Add suspend/resume support. o style(9) and code clean up. Special thanks to Shigeaki Tagashira, the original porter of nfe(4), who submitted lots of patches, performed uncountable number of regression tests and maintained nfe(4) for a long time. Without his enthusiastic help and support I could never have completed this overhauling task. The only weak point of nfe(4) compared to nve(4) is instability of manual half-duplex media selection on certain hardwares(auto sensing media type should work for all cases, though). This was a long standing bug of nfe(4) and I still have no idea why it doesn't work on some hardwares. Obtained from: OpenBSD [1] Submitted by: Peer Chen < pchen at nvidia dot com > [2], [3] Reviewed by: Shigeaki Tagashira < shigeaki AT se DOT hiroshima-u DOT ac DOT jp > Tested by: Shigeaki Tagashira, current Discussed with: current Silence from: obrien
2007-06-12 02:16:02 +00:00
NFE_WRITE(sc, NFE_RXBUFSZ, sc->nfe_framesize);
Import the nfe (nForce Ethernet) driver from OpenBSD. This is a pure open source NIC driver with a BSD license. These bits were taken from the OpenBSD public CVS repository on 1-May-06.
2006-06-26 17:25:22 +00:00
Add a pure open source nForce Ethernet driver, under BSDL. This driver was ported from OpenBSD by Shigeaki Tagashira <shigeaki@se.hiroshima-u.ac.jp> and posted at http://www.se.hiroshima-u.ac.jp/~shigeaki/software/freebsd-nfe.html It was additionally cleaned up by me. It is still a work-in-progress and thus is purposefully not in GENERIC. And it conflicts with nve(4), so only one should be loaded.
2006-06-26 23:41:07 +00:00
/* force MAC to wakeup */
Bring overhauled nfe(4) into tree. o s/printf/device_printf/g o Nuke OpenBSDism. o Nuke NetBSD/OpenBSD specific DMA sync operations.(we don't have a way to sync a single descriptor within a DMA map.) o Remove recursive mutex. o bus_dma(9) clean up. o 40bit DMA address support. o Add protection for Rx map load failure. o Fix a long standing bug for watchdog timeout. [1] o Add additional protections, missing Tx completion interrupt, losing start Tx command, for watchdog timeout. o Switch to taskqueue(9) API to handle interrupts. o Use our own timer for watchdog instead of if_watchdog/if_timer interface. o Advertise VLAN header length/capability correctly to upper layer. o Remove excessive kernel stack consumption in nfe_encap(). o Handle highly fragmented mbuf chains correctly. o Enable etherenet address reprogramming with ifconfig(8). o Add ALTQ/TSO, MSI/MSIX support. o Increased Rx ring to 256 descriptors from 128. o Align Tx/Rx descriptor ring on sizeof(struct nfe_desc64) boundary. o Remove alignment restrictions on Tx/Rx buffers. o Rewritten jumbo frame support code. o Add support for hardware assistend VLAN tag insertion/stripping. o Add support for Tx/Rx flow control based on patches from Peer Chen. [2] o Add a routine that detects whether ethernet address swap routines is required. [3] o Add a workaround that take MAC/PHY out of power down mode. o Add suspend/resume support. o style(9) and code clean up. Special thanks to Shigeaki Tagashira, the original porter of nfe(4), who submitted lots of patches, performed uncountable number of regression tests and maintained nfe(4) for a long time. Without his enthusiastic help and support I could never have completed this overhauling task. The only weak point of nfe(4) compared to nve(4) is instability of manual half-duplex media selection on certain hardwares(auto sensing media type should work for all cases, though). This was a long standing bug of nfe(4) and I still have no idea why it doesn't work on some hardwares. Obtained from: OpenBSD [1] Submitted by: Peer Chen < pchen at nvidia dot com > [2], [3] Reviewed by: Shigeaki Tagashira < shigeaki AT se DOT hiroshima-u DOT ac DOT jp > Tested by: Shigeaki Tagashira, current Discussed with: current Silence from: obrien
2007-06-12 02:16:02 +00:00
val = NFE_READ(sc, NFE_PWR_STATE);
if ((val & NFE_PWR_WAKEUP) == 0)
NFE_WRITE(sc, NFE_PWR_STATE, val | NFE_PWR_WAKEUP);
Add a pure open source nForce Ethernet driver, under BSDL. This driver was ported from OpenBSD by Shigeaki Tagashira <shigeaki@se.hiroshima-u.ac.jp> and posted at http://www.se.hiroshima-u.ac.jp/~shigeaki/software/freebsd-nfe.html It was additionally cleaned up by me. It is still a work-in-progress and thus is purposefully not in GENERIC. And it conflicts with nve(4), so only one should be loaded.
2006-06-26 23:41:07 +00:00
DELAY(10);
Bring overhauled nfe(4) into tree. o s/printf/device_printf/g o Nuke OpenBSDism. o Nuke NetBSD/OpenBSD specific DMA sync operations.(we don't have a way to sync a single descriptor within a DMA map.) o Remove recursive mutex. o bus_dma(9) clean up. o 40bit DMA address support. o Add protection for Rx map load failure. o Fix a long standing bug for watchdog timeout. [1] o Add additional protections, missing Tx completion interrupt, losing start Tx command, for watchdog timeout. o Switch to taskqueue(9) API to handle interrupts. o Use our own timer for watchdog instead of if_watchdog/if_timer interface. o Advertise VLAN header length/capability correctly to upper layer. o Remove excessive kernel stack consumption in nfe_encap(). o Handle highly fragmented mbuf chains correctly. o Enable etherenet address reprogramming with ifconfig(8). o Add ALTQ/TSO, MSI/MSIX support. o Increased Rx ring to 256 descriptors from 128. o Align Tx/Rx descriptor ring on sizeof(struct nfe_desc64) boundary. o Remove alignment restrictions on Tx/Rx buffers. o Rewritten jumbo frame support code. o Add support for hardware assistend VLAN tag insertion/stripping. o Add support for Tx/Rx flow control based on patches from Peer Chen. [2] o Add a routine that detects whether ethernet address swap routines is required. [3] o Add a workaround that take MAC/PHY out of power down mode. o Add suspend/resume support. o style(9) and code clean up. Special thanks to Shigeaki Tagashira, the original porter of nfe(4), who submitted lots of patches, performed uncountable number of regression tests and maintained nfe(4) for a long time. Without his enthusiastic help and support I could never have completed this overhauling task. The only weak point of nfe(4) compared to nve(4) is instability of manual half-duplex media selection on certain hardwares(auto sensing media type should work for all cases, though). This was a long standing bug of nfe(4) and I still have no idea why it doesn't work on some hardwares. Obtained from: OpenBSD [1] Submitted by: Peer Chen < pchen at nvidia dot com > [2], [3] Reviewed by: Shigeaki Tagashira < shigeaki AT se DOT hiroshima-u DOT ac DOT jp > Tested by: Shigeaki Tagashira, current Discussed with: current Silence from: obrien
2007-06-12 02:16:02 +00:00
val = NFE_READ(sc, NFE_PWR_STATE);
NFE_WRITE(sc, NFE_PWR_STATE, val | NFE_PWR_VALID);
Import the nfe (nForce Ethernet) driver from OpenBSD. This is a pure open source NIC driver with a BSD license. These bits were taken from the OpenBSD public CVS repository on 1-May-06.
2006-06-26 17:25:22 +00:00
Add a pure open source nForce Ethernet driver, under BSDL. This driver was ported from OpenBSD by Shigeaki Tagashira <shigeaki@se.hiroshima-u.ac.jp> and posted at http://www.se.hiroshima-u.ac.jp/~shigeaki/software/freebsd-nfe.html It was additionally cleaned up by me. It is still a work-in-progress and thus is purposefully not in GENERIC. And it conflicts with nve(4), so only one should be loaded.
2006-06-26 23:41:07 +00:00
#if 1
/* configure interrupts coalescing/mitigation */
NFE_WRITE(sc, NFE_IMTIMER, NFE_IM_DEFAULT);
#else
/* no interrupt mitigation: one interrupt per packet */
NFE_WRITE(sc, NFE_IMTIMER, 970);
#endif
Import the nfe (nForce Ethernet) driver from OpenBSD. This is a pure open source NIC driver with a BSD license. These bits were taken from the OpenBSD public CVS repository on 1-May-06.
2006-06-26 17:25:22 +00:00
Bring overhauled nfe(4) into tree. o s/printf/device_printf/g o Nuke OpenBSDism. o Nuke NetBSD/OpenBSD specific DMA sync operations.(we don't have a way to sync a single descriptor within a DMA map.) o Remove recursive mutex. o bus_dma(9) clean up. o 40bit DMA address support. o Add protection for Rx map load failure. o Fix a long standing bug for watchdog timeout. [1] o Add additional protections, missing Tx completion interrupt, losing start Tx command, for watchdog timeout. o Switch to taskqueue(9) API to handle interrupts. o Use our own timer for watchdog instead of if_watchdog/if_timer interface. o Advertise VLAN header length/capability correctly to upper layer. o Remove excessive kernel stack consumption in nfe_encap(). o Handle highly fragmented mbuf chains correctly. o Enable etherenet address reprogramming with ifconfig(8). o Add ALTQ/TSO, MSI/MSIX support. o Increased Rx ring to 256 descriptors from 128. o Align Tx/Rx descriptor ring on sizeof(struct nfe_desc64) boundary. o Remove alignment restrictions on Tx/Rx buffers. o Rewritten jumbo frame support code. o Add support for hardware assistend VLAN tag insertion/stripping. o Add support for Tx/Rx flow control based on patches from Peer Chen. [2] o Add a routine that detects whether ethernet address swap routines is required. [3] o Add a workaround that take MAC/PHY out of power down mode. o Add suspend/resume support. o style(9) and code clean up. Special thanks to Shigeaki Tagashira, the original porter of nfe(4), who submitted lots of patches, performed uncountable number of regression tests and maintained nfe(4) for a long time. Without his enthusiastic help and support I could never have completed this overhauling task. The only weak point of nfe(4) compared to nve(4) is instability of manual half-duplex media selection on certain hardwares(auto sensing media type should work for all cases, though). This was a long standing bug of nfe(4) and I still have no idea why it doesn't work on some hardwares. Obtained from: OpenBSD [1] Submitted by: Peer Chen < pchen at nvidia dot com > [2], [3] Reviewed by: Shigeaki Tagashira < shigeaki AT se DOT hiroshima-u DOT ac DOT jp > Tested by: Shigeaki Tagashira, current Discussed with: current Silence from: obrien
2007-06-12 02:16:02 +00:00
NFE_WRITE(sc, NFE_SETUP_R1, NFE_R1_MAGIC_10_100);
Add a pure open source nForce Ethernet driver, under BSDL. This driver was ported from OpenBSD by Shigeaki Tagashira <shigeaki@se.hiroshima-u.ac.jp> and posted at http://www.se.hiroshima-u.ac.jp/~shigeaki/software/freebsd-nfe.html It was additionally cleaned up by me. It is still a work-in-progress and thus is purposefully not in GENERIC. And it conflicts with nve(4), so only one should be loaded.
2006-06-26 23:41:07 +00:00
NFE_WRITE(sc, NFE_SETUP_R2, NFE_R2_MAGIC);
NFE_WRITE(sc, NFE_SETUP_R6, NFE_R6_MAGIC);
Import the nfe (nForce Ethernet) driver from OpenBSD. This is a pure open source NIC driver with a BSD license. These bits were taken from the OpenBSD public CVS repository on 1-May-06.
2006-06-26 17:25:22 +00:00
Add a pure open source nForce Ethernet driver, under BSDL. This driver was ported from OpenBSD by Shigeaki Tagashira <shigeaki@se.hiroshima-u.ac.jp> and posted at http://www.se.hiroshima-u.ac.jp/~shigeaki/software/freebsd-nfe.html It was additionally cleaned up by me. It is still a work-in-progress and thus is purposefully not in GENERIC. And it conflicts with nve(4), so only one should be loaded.
2006-06-26 23:41:07 +00:00
/* update MAC knowledge of PHY; generates a NFE_IRQ_LINK interrupt */
NFE_WRITE(sc, NFE_STATUS, sc->mii_phyaddr << 24 | NFE_STATUS_MAGIC);
Import the nfe (nForce Ethernet) driver from OpenBSD. This is a pure open source NIC driver with a BSD license. These bits were taken from the OpenBSD public CVS repository on 1-May-06.
2006-06-26 17:25:22 +00:00
Add a pure open source nForce Ethernet driver, under BSDL. This driver was ported from OpenBSD by Shigeaki Tagashira <shigeaki@se.hiroshima-u.ac.jp> and posted at http://www.se.hiroshima-u.ac.jp/~shigeaki/software/freebsd-nfe.html It was additionally cleaned up by me. It is still a work-in-progress and thus is purposefully not in GENERIC. And it conflicts with nve(4), so only one should be loaded.
2006-06-26 23:41:07 +00:00
NFE_WRITE(sc, NFE_SETUP_R4, NFE_R4_MAGIC);
Add basic WOL support for MCP ethernet controllers. It seems the controller does not perform automatic switching from 1000Mbps link to 10/100Mbps link when WOL is activated. Implement establishing 10/100Mps link with auto-negotiation in driver. Link status change handler was modified to remove taskqueue based approach since driver now needs synchronous handling for link establishment. Submitted by: Yamagi Burmeister (lists <> yamagi.org ) (initial version) Tested by: Yamagi Burmeister (lists <> yamagi.org ) MFC after: 1 week
2010-11-11 18:08:50 +00:00
/* Disable WOL. */
NFE_WRITE(sc, NFE_WOL_CTL, 0);
Import the nfe (nForce Ethernet) driver from OpenBSD. This is a pure open source NIC driver with a BSD license. These bits were taken from the OpenBSD public CVS repository on 1-May-06.
2006-06-26 17:25:22 +00:00
Add a pure open source nForce Ethernet driver, under BSDL. This driver was ported from OpenBSD by Shigeaki Tagashira <shigeaki@se.hiroshima-u.ac.jp> and posted at http://www.se.hiroshima-u.ac.jp/~shigeaki/software/freebsd-nfe.html It was additionally cleaned up by me. It is still a work-in-progress and thus is purposefully not in GENERIC. And it conflicts with nve(4), so only one should be loaded.
2006-06-26 23:41:07 +00:00
sc->rxtxctl &= ~NFE_RXTX_BIT2;
NFE_WRITE(sc, NFE_RXTX_CTL, sc->rxtxctl);
DELAY(10);
NFE_WRITE(sc, NFE_RXTX_CTL, NFE_RXTX_BIT1 | sc->rxtxctl);
Import the nfe (nForce Ethernet) driver from OpenBSD. This is a pure open source NIC driver with a BSD license. These bits were taken from the OpenBSD public CVS repository on 1-May-06.
2006-06-26 17:25:22 +00:00
Add a pure open source nForce Ethernet driver, under BSDL. This driver was ported from OpenBSD by Shigeaki Tagashira <shigeaki@se.hiroshima-u.ac.jp> and posted at http://www.se.hiroshima-u.ac.jp/~shigeaki/software/freebsd-nfe.html It was additionally cleaned up by me. It is still a work-in-progress and thus is purposefully not in GENERIC. And it conflicts with nve(4), so only one should be loaded.
2006-06-26 23:41:07 +00:00
/* set Rx filter */
nfe_setmulti(sc);
Import the nfe (nForce Ethernet) driver from OpenBSD. This is a pure open source NIC driver with a BSD license. These bits were taken from the OpenBSD public CVS repository on 1-May-06.
2006-06-26 17:25:22 +00:00
Add a pure open source nForce Ethernet driver, under BSDL. This driver was ported from OpenBSD by Shigeaki Tagashira <shigeaki@se.hiroshima-u.ac.jp> and posted at http://www.se.hiroshima-u.ac.jp/~shigeaki/software/freebsd-nfe.html It was additionally cleaned up by me. It is still a work-in-progress and thus is purposefully not in GENERIC. And it conflicts with nve(4), so only one should be loaded.
2006-06-26 23:41:07 +00:00
/* enable Rx */
NFE_WRITE(sc, NFE_RX_CTL, NFE_RX_START);
Import the nfe (nForce Ethernet) driver from OpenBSD. This is a pure open source NIC driver with a BSD license. These bits were taken from the OpenBSD public CVS repository on 1-May-06.
2006-06-26 17:25:22 +00:00
Add a pure open source nForce Ethernet driver, under BSDL. This driver was ported from OpenBSD by Shigeaki Tagashira <shigeaki@se.hiroshima-u.ac.jp> and posted at http://www.se.hiroshima-u.ac.jp/~shigeaki/software/freebsd-nfe.html It was additionally cleaned up by me. It is still a work-in-progress and thus is purposefully not in GENERIC. And it conflicts with nve(4), so only one should be loaded.
2006-06-26 23:41:07 +00:00
/* enable Tx */
NFE_WRITE(sc, NFE_TX_CTL, NFE_TX_START);
Import the nfe (nForce Ethernet) driver from OpenBSD. This is a pure open source NIC driver with a BSD license. These bits were taken from the OpenBSD public CVS repository on 1-May-06.
2006-06-26 17:25:22 +00:00
Add a pure open source nForce Ethernet driver, under BSDL. This driver was ported from OpenBSD by Shigeaki Tagashira <shigeaki@se.hiroshima-u.ac.jp> and posted at http://www.se.hiroshima-u.ac.jp/~shigeaki/software/freebsd-nfe.html It was additionally cleaned up by me. It is still a work-in-progress and thus is purposefully not in GENERIC. And it conflicts with nve(4), so only one should be loaded.
2006-06-26 23:41:07 +00:00
NFE_WRITE(sc, NFE_PHY_STATUS, 0xf);
Import the nfe (nForce Ethernet) driver from OpenBSD. This is a pure open source NIC driver with a BSD license. These bits were taken from the OpenBSD public CVS repository on 1-May-06.
2006-06-26 17:25:22 +00:00
Add hardware MAC statistics support. Register information was obtained from Linux forcedeth driver. While I'm here move creating a sysctl node for process_limit to function nfe_sysctl_node(). Tested by: "Arno J. Klaassen" < arno <at> heho dot snv dot jussieu dot fr >
2008-10-03 03:58:16 +00:00
/* Clear hardware stats. */
nfe_stats_clear(sc);
Add a pure open source nForce Ethernet driver, under BSDL. This driver was ported from OpenBSD by Shigeaki Tagashira <shigeaki@se.hiroshima-u.ac.jp> and posted at http://www.se.hiroshima-u.ac.jp/~shigeaki/software/freebsd-nfe.html It was additionally cleaned up by me. It is still a work-in-progress and thus is purposefully not in GENERIC. And it conflicts with nve(4), so only one should be loaded.
2006-06-26 23:41:07 +00:00
#ifdef DEVICE_POLLING
if (ifp->if_capenable & IFCAP_POLLING)
Bring overhauled nfe(4) into tree. o s/printf/device_printf/g o Nuke OpenBSDism. o Nuke NetBSD/OpenBSD specific DMA sync operations.(we don't have a way to sync a single descriptor within a DMA map.) o Remove recursive mutex. o bus_dma(9) clean up. o 40bit DMA address support. o Add protection for Rx map load failure. o Fix a long standing bug for watchdog timeout. [1] o Add additional protections, missing Tx completion interrupt, losing start Tx command, for watchdog timeout. o Switch to taskqueue(9) API to handle interrupts. o Use our own timer for watchdog instead of if_watchdog/if_timer interface. o Advertise VLAN header length/capability correctly to upper layer. o Remove excessive kernel stack consumption in nfe_encap(). o Handle highly fragmented mbuf chains correctly. o Enable etherenet address reprogramming with ifconfig(8). o Add ALTQ/TSO, MSI/MSIX support. o Increased Rx ring to 256 descriptors from 128. o Align Tx/Rx descriptor ring on sizeof(struct nfe_desc64) boundary. o Remove alignment restrictions on Tx/Rx buffers. o Rewritten jumbo frame support code. o Add support for hardware assistend VLAN tag insertion/stripping. o Add support for Tx/Rx flow control based on patches from Peer Chen. [2] o Add a routine that detects whether ethernet address swap routines is required. [3] o Add a workaround that take MAC/PHY out of power down mode. o Add suspend/resume support. o style(9) and code clean up. Special thanks to Shigeaki Tagashira, the original porter of nfe(4), who submitted lots of patches, performed uncountable number of regression tests and maintained nfe(4) for a long time. Without his enthusiastic help and support I could never have completed this overhauling task. The only weak point of nfe(4) compared to nve(4) is instability of manual half-duplex media selection on certain hardwares(auto sensing media type should work for all cases, though). This was a long standing bug of nfe(4) and I still have no idea why it doesn't work on some hardwares. Obtained from: OpenBSD [1] Submitted by: Peer Chen < pchen at nvidia dot com > [2], [3] Reviewed by: Shigeaki Tagashira < shigeaki AT se DOT hiroshima-u DOT ac DOT jp > Tested by: Shigeaki Tagashira, current Discussed with: current Silence from: obrien
2007-06-12 02:16:02 +00:00
nfe_disable_intr(sc);
Add a pure open source nForce Ethernet driver, under BSDL. This driver was ported from OpenBSD by Shigeaki Tagashira <shigeaki@se.hiroshima-u.ac.jp> and posted at http://www.se.hiroshima-u.ac.jp/~shigeaki/software/freebsd-nfe.html It was additionally cleaned up by me. It is still a work-in-progress and thus is purposefully not in GENERIC. And it conflicts with nve(4), so only one should be loaded.
2006-06-26 23:41:07 +00:00
else
#endif
Bring overhauled nfe(4) into tree. o s/printf/device_printf/g o Nuke OpenBSDism. o Nuke NetBSD/OpenBSD specific DMA sync operations.(we don't have a way to sync a single descriptor within a DMA map.) o Remove recursive mutex. o bus_dma(9) clean up. o 40bit DMA address support. o Add protection for Rx map load failure. o Fix a long standing bug for watchdog timeout. [1] o Add additional protections, missing Tx completion interrupt, losing start Tx command, for watchdog timeout. o Switch to taskqueue(9) API to handle interrupts. o Use our own timer for watchdog instead of if_watchdog/if_timer interface. o Advertise VLAN header length/capability correctly to upper layer. o Remove excessive kernel stack consumption in nfe_encap(). o Handle highly fragmented mbuf chains correctly. o Enable etherenet address reprogramming with ifconfig(8). o Add ALTQ/TSO, MSI/MSIX support. o Increased Rx ring to 256 descriptors from 128. o Align Tx/Rx descriptor ring on sizeof(struct nfe_desc64) boundary. o Remove alignment restrictions on Tx/Rx buffers. o Rewritten jumbo frame support code. o Add support for hardware assistend VLAN tag insertion/stripping. o Add support for Tx/Rx flow control based on patches from Peer Chen. [2] o Add a routine that detects whether ethernet address swap routines is required. [3] o Add a workaround that take MAC/PHY out of power down mode. o Add suspend/resume support. o style(9) and code clean up. Special thanks to Shigeaki Tagashira, the original porter of nfe(4), who submitted lots of patches, performed uncountable number of regression tests and maintained nfe(4) for a long time. Without his enthusiastic help and support I could never have completed this overhauling task. The only weak point of nfe(4) compared to nve(4) is instability of manual half-duplex media selection on certain hardwares(auto sensing media type should work for all cases, though). This was a long standing bug of nfe(4) and I still have no idea why it doesn't work on some hardwares. Obtained from: OpenBSD [1] Submitted by: Peer Chen < pchen at nvidia dot com > [2], [3] Reviewed by: Shigeaki Tagashira < shigeaki AT se DOT hiroshima-u DOT ac DOT jp > Tested by: Shigeaki Tagashira, current Discussed with: current Silence from: obrien
2007-06-12 02:16:02 +00:00
nfe_set_intr(sc);
nfe_enable_intr(sc); /* enable interrupts */
Import the nfe (nForce Ethernet) driver from OpenBSD. This is a pure open source NIC driver with a BSD license. These bits were taken from the OpenBSD public CVS repository on 1-May-06.
2006-06-26 17:25:22 +00:00
Add a pure open source nForce Ethernet driver, under BSDL. This driver was ported from OpenBSD by Shigeaki Tagashira <shigeaki@se.hiroshima-u.ac.jp> and posted at http://www.se.hiroshima-u.ac.jp/~shigeaki/software/freebsd-nfe.html It was additionally cleaned up by me. It is still a work-in-progress and thus is purposefully not in GENERIC. And it conflicts with nve(4), so only one should be loaded.
2006-06-26 23:41:07 +00:00
ifp->if_drv_flags |= IFF_DRV_RUNNING;
ifp->if_drv_flags &= ~IFF_DRV_OACTIVE;
Import the nfe (nForce Ethernet) driver from OpenBSD. This is a pure open source NIC driver with a BSD license. These bits were taken from the OpenBSD public CVS repository on 1-May-06.
2006-06-26 17:25:22 +00:00
Add a pure open source nForce Ethernet driver, under BSDL. This driver was ported from OpenBSD by Shigeaki Tagashira <shigeaki@se.hiroshima-u.ac.jp> and posted at http://www.se.hiroshima-u.ac.jp/~shigeaki/software/freebsd-nfe.html It was additionally cleaned up by me. It is still a work-in-progress and thus is purposefully not in GENERIC. And it conflicts with nve(4), so only one should be loaded.
2006-06-26 23:41:07 +00:00
sc->nfe_link = 0;
Bring overhauled nfe(4) into tree. o s/printf/device_printf/g o Nuke OpenBSDism. o Nuke NetBSD/OpenBSD specific DMA sync operations.(we don't have a way to sync a single descriptor within a DMA map.) o Remove recursive mutex. o bus_dma(9) clean up. o 40bit DMA address support. o Add protection for Rx map load failure. o Fix a long standing bug for watchdog timeout. [1] o Add additional protections, missing Tx completion interrupt, losing start Tx command, for watchdog timeout. o Switch to taskqueue(9) API to handle interrupts. o Use our own timer for watchdog instead of if_watchdog/if_timer interface. o Advertise VLAN header length/capability correctly to upper layer. o Remove excessive kernel stack consumption in nfe_encap(). o Handle highly fragmented mbuf chains correctly. o Enable etherenet address reprogramming with ifconfig(8). o Add ALTQ/TSO, MSI/MSIX support. o Increased Rx ring to 256 descriptors from 128. o Align Tx/Rx descriptor ring on sizeof(struct nfe_desc64) boundary. o Remove alignment restrictions on Tx/Rx buffers. o Rewritten jumbo frame support code. o Add support for hardware assistend VLAN tag insertion/stripping. o Add support for Tx/Rx flow control based on patches from Peer Chen. [2] o Add a routine that detects whether ethernet address swap routines is required. [3] o Add a workaround that take MAC/PHY out of power down mode. o Add suspend/resume support. o style(9) and code clean up. Special thanks to Shigeaki Tagashira, the original porter of nfe(4), who submitted lots of patches, performed uncountable number of regression tests and maintained nfe(4) for a long time. Without his enthusiastic help and support I could never have completed this overhauling task. The only weak point of nfe(4) compared to nve(4) is instability of manual half-duplex media selection on certain hardwares(auto sensing media type should work for all cases, though). This was a long standing bug of nfe(4) and I still have no idea why it doesn't work on some hardwares. Obtained from: OpenBSD [1] Submitted by: Peer Chen < pchen at nvidia dot com > [2], [3] Reviewed by: Shigeaki Tagashira < shigeaki AT se DOT hiroshima-u DOT ac DOT jp > Tested by: Shigeaki Tagashira, current Discussed with: current Silence from: obrien
2007-06-12 02:16:02 +00:00
mii_mediachg(mii);
Import the nfe (nForce Ethernet) driver from OpenBSD. This is a pure open source NIC driver with a BSD license. These bits were taken from the OpenBSD public CVS repository on 1-May-06.
2006-06-26 17:25:22 +00:00
Bring overhauled nfe(4) into tree. o s/printf/device_printf/g o Nuke OpenBSDism. o Nuke NetBSD/OpenBSD specific DMA sync operations.(we don't have a way to sync a single descriptor within a DMA map.) o Remove recursive mutex. o bus_dma(9) clean up. o 40bit DMA address support. o Add protection for Rx map load failure. o Fix a long standing bug for watchdog timeout. [1] o Add additional protections, missing Tx completion interrupt, losing start Tx command, for watchdog timeout. o Switch to taskqueue(9) API to handle interrupts. o Use our own timer for watchdog instead of if_watchdog/if_timer interface. o Advertise VLAN header length/capability correctly to upper layer. o Remove excessive kernel stack consumption in nfe_encap(). o Handle highly fragmented mbuf chains correctly. o Enable etherenet address reprogramming with ifconfig(8). o Add ALTQ/TSO, MSI/MSIX support. o Increased Rx ring to 256 descriptors from 128. o Align Tx/Rx descriptor ring on sizeof(struct nfe_desc64) boundary. o Remove alignment restrictions on Tx/Rx buffers. o Rewritten jumbo frame support code. o Add support for hardware assistend VLAN tag insertion/stripping. o Add support for Tx/Rx flow control based on patches from Peer Chen. [2] o Add a routine that detects whether ethernet address swap routines is required. [3] o Add a workaround that take MAC/PHY out of power down mode. o Add suspend/resume support. o style(9) and code clean up. Special thanks to Shigeaki Tagashira, the original porter of nfe(4), who submitted lots of patches, performed uncountable number of regression tests and maintained nfe(4) for a long time. Without his enthusiastic help and support I could never have completed this overhauling task. The only weak point of nfe(4) compared to nve(4) is instability of manual half-duplex media selection on certain hardwares(auto sensing media type should work for all cases, though). This was a long standing bug of nfe(4) and I still have no idea why it doesn't work on some hardwares. Obtained from: OpenBSD [1] Submitted by: Peer Chen < pchen at nvidia dot com > [2], [3] Reviewed by: Shigeaki Tagashira < shigeaki AT se DOT hiroshima-u DOT ac DOT jp > Tested by: Shigeaki Tagashira, current Discussed with: current Silence from: obrien
2007-06-12 02:16:02 +00:00
callout_reset(&sc->nfe_stat_ch, hz, nfe_tick, sc);
Import the nfe (nForce Ethernet) driver from OpenBSD. This is a pure open source NIC driver with a BSD license. These bits were taken from the OpenBSD public CVS repository on 1-May-06.
2006-06-26 17:25:22 +00:00
}
style(9) cleanup.
2006-10-19 10:01:26 +00:00
static void
Bring overhauled nfe(4) into tree. o s/printf/device_printf/g o Nuke OpenBSDism. o Nuke NetBSD/OpenBSD specific DMA sync operations.(we don't have a way to sync a single descriptor within a DMA map.) o Remove recursive mutex. o bus_dma(9) clean up. o 40bit DMA address support. o Add protection for Rx map load failure. o Fix a long standing bug for watchdog timeout. [1] o Add additional protections, missing Tx completion interrupt, losing start Tx command, for watchdog timeout. o Switch to taskqueue(9) API to handle interrupts. o Use our own timer for watchdog instead of if_watchdog/if_timer interface. o Advertise VLAN header length/capability correctly to upper layer. o Remove excessive kernel stack consumption in nfe_encap(). o Handle highly fragmented mbuf chains correctly. o Enable etherenet address reprogramming with ifconfig(8). o Add ALTQ/TSO, MSI/MSIX support. o Increased Rx ring to 256 descriptors from 128. o Align Tx/Rx descriptor ring on sizeof(struct nfe_desc64) boundary. o Remove alignment restrictions on Tx/Rx buffers. o Rewritten jumbo frame support code. o Add support for hardware assistend VLAN tag insertion/stripping. o Add support for Tx/Rx flow control based on patches from Peer Chen. [2] o Add a routine that detects whether ethernet address swap routines is required. [3] o Add a workaround that take MAC/PHY out of power down mode. o Add suspend/resume support. o style(9) and code clean up. Special thanks to Shigeaki Tagashira, the original porter of nfe(4), who submitted lots of patches, performed uncountable number of regression tests and maintained nfe(4) for a long time. Without his enthusiastic help and support I could never have completed this overhauling task. The only weak point of nfe(4) compared to nve(4) is instability of manual half-duplex media selection on certain hardwares(auto sensing media type should work for all cases, though). This was a long standing bug of nfe(4) and I still have no idea why it doesn't work on some hardwares. Obtained from: OpenBSD [1] Submitted by: Peer Chen < pchen at nvidia dot com > [2], [3] Reviewed by: Shigeaki Tagashira < shigeaki AT se DOT hiroshima-u DOT ac DOT jp > Tested by: Shigeaki Tagashira, current Discussed with: current Silence from: obrien
2007-06-12 02:16:02 +00:00
nfe_stop(struct ifnet *ifp)
Import the nfe (nForce Ethernet) driver from OpenBSD. This is a pure open source NIC driver with a BSD license. These bits were taken from the OpenBSD public CVS repository on 1-May-06.
2006-06-26 17:25:22 +00:00
{
Add a pure open source nForce Ethernet driver, under BSDL. This driver was ported from OpenBSD by Shigeaki Tagashira <shigeaki@se.hiroshima-u.ac.jp> and posted at http://www.se.hiroshima-u.ac.jp/~shigeaki/software/freebsd-nfe.html It was additionally cleaned up by me. It is still a work-in-progress and thus is purposefully not in GENERIC. And it conflicts with nve(4), so only one should be loaded.
2006-06-26 23:41:07 +00:00
struct nfe_softc *sc = ifp->if_softc;
Bring overhauled nfe(4) into tree. o s/printf/device_printf/g o Nuke OpenBSDism. o Nuke NetBSD/OpenBSD specific DMA sync operations.(we don't have a way to sync a single descriptor within a DMA map.) o Remove recursive mutex. o bus_dma(9) clean up. o 40bit DMA address support. o Add protection for Rx map load failure. o Fix a long standing bug for watchdog timeout. [1] o Add additional protections, missing Tx completion interrupt, losing start Tx command, for watchdog timeout. o Switch to taskqueue(9) API to handle interrupts. o Use our own timer for watchdog instead of if_watchdog/if_timer interface. o Advertise VLAN header length/capability correctly to upper layer. o Remove excessive kernel stack consumption in nfe_encap(). o Handle highly fragmented mbuf chains correctly. o Enable etherenet address reprogramming with ifconfig(8). o Add ALTQ/TSO, MSI/MSIX support. o Increased Rx ring to 256 descriptors from 128. o Align Tx/Rx descriptor ring on sizeof(struct nfe_desc64) boundary. o Remove alignment restrictions on Tx/Rx buffers. o Rewritten jumbo frame support code. o Add support for hardware assistend VLAN tag insertion/stripping. o Add support for Tx/Rx flow control based on patches from Peer Chen. [2] o Add a routine that detects whether ethernet address swap routines is required. [3] o Add a workaround that take MAC/PHY out of power down mode. o Add suspend/resume support. o style(9) and code clean up. Special thanks to Shigeaki Tagashira, the original porter of nfe(4), who submitted lots of patches, performed uncountable number of regression tests and maintained nfe(4) for a long time. Without his enthusiastic help and support I could never have completed this overhauling task. The only weak point of nfe(4) compared to nve(4) is instability of manual half-duplex media selection on certain hardwares(auto sensing media type should work for all cases, though). This was a long standing bug of nfe(4) and I still have no idea why it doesn't work on some hardwares. Obtained from: OpenBSD [1] Submitted by: Peer Chen < pchen at nvidia dot com > [2], [3] Reviewed by: Shigeaki Tagashira < shigeaki AT se DOT hiroshima-u DOT ac DOT jp > Tested by: Shigeaki Tagashira, current Discussed with: current Silence from: obrien
2007-06-12 02:16:02 +00:00
struct nfe_rx_ring *rx_ring;
struct nfe_jrx_ring *jrx_ring;
struct nfe_tx_ring *tx_ring;
struct nfe_rx_data *rdata;
struct nfe_tx_data *tdata;
int i;
Import the nfe (nForce Ethernet) driver from OpenBSD. This is a pure open source NIC driver with a BSD license. These bits were taken from the OpenBSD public CVS repository on 1-May-06.
2006-06-26 17:25:22 +00:00
Add a pure open source nForce Ethernet driver, under BSDL. This driver was ported from OpenBSD by Shigeaki Tagashira <shigeaki@se.hiroshima-u.ac.jp> and posted at http://www.se.hiroshima-u.ac.jp/~shigeaki/software/freebsd-nfe.html It was additionally cleaned up by me. It is still a work-in-progress and thus is purposefully not in GENERIC. And it conflicts with nve(4), so only one should be loaded.
2006-06-26 23:41:07 +00:00
NFE_LOCK_ASSERT(sc);
Import the nfe (nForce Ethernet) driver from OpenBSD. This is a pure open source NIC driver with a BSD license. These bits were taken from the OpenBSD public CVS repository on 1-May-06.
2006-06-26 17:25:22 +00:00
Bring overhauled nfe(4) into tree. o s/printf/device_printf/g o Nuke OpenBSDism. o Nuke NetBSD/OpenBSD specific DMA sync operations.(we don't have a way to sync a single descriptor within a DMA map.) o Remove recursive mutex. o bus_dma(9) clean up. o 40bit DMA address support. o Add protection for Rx map load failure. o Fix a long standing bug for watchdog timeout. [1] o Add additional protections, missing Tx completion interrupt, losing start Tx command, for watchdog timeout. o Switch to taskqueue(9) API to handle interrupts. o Use our own timer for watchdog instead of if_watchdog/if_timer interface. o Advertise VLAN header length/capability correctly to upper layer. o Remove excessive kernel stack consumption in nfe_encap(). o Handle highly fragmented mbuf chains correctly. o Enable etherenet address reprogramming with ifconfig(8). o Add ALTQ/TSO, MSI/MSIX support. o Increased Rx ring to 256 descriptors from 128. o Align Tx/Rx descriptor ring on sizeof(struct nfe_desc64) boundary. o Remove alignment restrictions on Tx/Rx buffers. o Rewritten jumbo frame support code. o Add support for hardware assistend VLAN tag insertion/stripping. o Add support for Tx/Rx flow control based on patches from Peer Chen. [2] o Add a routine that detects whether ethernet address swap routines is required. [3] o Add a workaround that take MAC/PHY out of power down mode. o Add suspend/resume support. o style(9) and code clean up. Special thanks to Shigeaki Tagashira, the original porter of nfe(4), who submitted lots of patches, performed uncountable number of regression tests and maintained nfe(4) for a long time. Without his enthusiastic help and support I could never have completed this overhauling task. The only weak point of nfe(4) compared to nve(4) is instability of manual half-duplex media selection on certain hardwares(auto sensing media type should work for all cases, though). This was a long standing bug of nfe(4) and I still have no idea why it doesn't work on some hardwares. Obtained from: OpenBSD [1] Submitted by: Peer Chen < pchen at nvidia dot com > [2], [3] Reviewed by: Shigeaki Tagashira < shigeaki AT se DOT hiroshima-u DOT ac DOT jp > Tested by: Shigeaki Tagashira, current Discussed with: current Silence from: obrien
2007-06-12 02:16:02 +00:00
sc->nfe_watchdog_timer = 0;
Add a pure open source nForce Ethernet driver, under BSDL. This driver was ported from OpenBSD by Shigeaki Tagashira <shigeaki@se.hiroshima-u.ac.jp> and posted at http://www.se.hiroshima-u.ac.jp/~shigeaki/software/freebsd-nfe.html It was additionally cleaned up by me. It is still a work-in-progress and thus is purposefully not in GENERIC. And it conflicts with nve(4), so only one should be loaded.
2006-06-26 23:41:07 +00:00
ifp->if_drv_flags &= ~(IFF_DRV_RUNNING | IFF_DRV_OACTIVE);
Import the nfe (nForce Ethernet) driver from OpenBSD. This is a pure open source NIC driver with a BSD license. These bits were taken from the OpenBSD public CVS repository on 1-May-06.
2006-06-26 17:25:22 +00:00
Add a pure open source nForce Ethernet driver, under BSDL. This driver was ported from OpenBSD by Shigeaki Tagashira <shigeaki@se.hiroshima-u.ac.jp> and posted at http://www.se.hiroshima-u.ac.jp/~shigeaki/software/freebsd-nfe.html It was additionally cleaned up by me. It is still a work-in-progress and thus is purposefully not in GENERIC. And it conflicts with nve(4), so only one should be loaded.
2006-06-26 23:41:07 +00:00
callout_stop(&sc->nfe_stat_ch);
Import the nfe (nForce Ethernet) driver from OpenBSD. This is a pure open source NIC driver with a BSD license. These bits were taken from the OpenBSD public CVS repository on 1-May-06.
2006-06-26 17:25:22 +00:00
Add a pure open source nForce Ethernet driver, under BSDL. This driver was ported from OpenBSD by Shigeaki Tagashira <shigeaki@se.hiroshima-u.ac.jp> and posted at http://www.se.hiroshima-u.ac.jp/~shigeaki/software/freebsd-nfe.html It was additionally cleaned up by me. It is still a work-in-progress and thus is purposefully not in GENERIC. And it conflicts with nve(4), so only one should be loaded.
2006-06-26 23:41:07 +00:00
/* abort Tx */
NFE_WRITE(sc, NFE_TX_CTL, 0);
Import the nfe (nForce Ethernet) driver from OpenBSD. This is a pure open source NIC driver with a BSD license. These bits were taken from the OpenBSD public CVS repository on 1-May-06.
2006-06-26 17:25:22 +00:00
Add a pure open source nForce Ethernet driver, under BSDL. This driver was ported from OpenBSD by Shigeaki Tagashira <shigeaki@se.hiroshima-u.ac.jp> and posted at http://www.se.hiroshima-u.ac.jp/~shigeaki/software/freebsd-nfe.html It was additionally cleaned up by me. It is still a work-in-progress and thus is purposefully not in GENERIC. And it conflicts with nve(4), so only one should be loaded.
2006-06-26 23:41:07 +00:00
/* disable Rx */
NFE_WRITE(sc, NFE_RX_CTL, 0);
Import the nfe (nForce Ethernet) driver from OpenBSD. This is a pure open source NIC driver with a BSD license. These bits were taken from the OpenBSD public CVS repository on 1-May-06.
2006-06-26 17:25:22 +00:00
Add a pure open source nForce Ethernet driver, under BSDL. This driver was ported from OpenBSD by Shigeaki Tagashira <shigeaki@se.hiroshima-u.ac.jp> and posted at http://www.se.hiroshima-u.ac.jp/~shigeaki/software/freebsd-nfe.html It was additionally cleaned up by me. It is still a work-in-progress and thus is purposefully not in GENERIC. And it conflicts with nve(4), so only one should be loaded.
2006-06-26 23:41:07 +00:00
/* disable interrupts */
Bring overhauled nfe(4) into tree. o s/printf/device_printf/g o Nuke OpenBSDism. o Nuke NetBSD/OpenBSD specific DMA sync operations.(we don't have a way to sync a single descriptor within a DMA map.) o Remove recursive mutex. o bus_dma(9) clean up. o 40bit DMA address support. o Add protection for Rx map load failure. o Fix a long standing bug for watchdog timeout. [1] o Add additional protections, missing Tx completion interrupt, losing start Tx command, for watchdog timeout. o Switch to taskqueue(9) API to handle interrupts. o Use our own timer for watchdog instead of if_watchdog/if_timer interface. o Advertise VLAN header length/capability correctly to upper layer. o Remove excessive kernel stack consumption in nfe_encap(). o Handle highly fragmented mbuf chains correctly. o Enable etherenet address reprogramming with ifconfig(8). o Add ALTQ/TSO, MSI/MSIX support. o Increased Rx ring to 256 descriptors from 128. o Align Tx/Rx descriptor ring on sizeof(struct nfe_desc64) boundary. o Remove alignment restrictions on Tx/Rx buffers. o Rewritten jumbo frame support code. o Add support for hardware assistend VLAN tag insertion/stripping. o Add support for Tx/Rx flow control based on patches from Peer Chen. [2] o Add a routine that detects whether ethernet address swap routines is required. [3] o Add a workaround that take MAC/PHY out of power down mode. o Add suspend/resume support. o style(9) and code clean up. Special thanks to Shigeaki Tagashira, the original porter of nfe(4), who submitted lots of patches, performed uncountable number of regression tests and maintained nfe(4) for a long time. Without his enthusiastic help and support I could never have completed this overhauling task. The only weak point of nfe(4) compared to nve(4) is instability of manual half-duplex media selection on certain hardwares(auto sensing media type should work for all cases, though). This was a long standing bug of nfe(4) and I still have no idea why it doesn't work on some hardwares. Obtained from: OpenBSD [1] Submitted by: Peer Chen < pchen at nvidia dot com > [2], [3] Reviewed by: Shigeaki Tagashira < shigeaki AT se DOT hiroshima-u DOT ac DOT jp > Tested by: Shigeaki Tagashira, current Discussed with: current Silence from: obrien
2007-06-12 02:16:02 +00:00
nfe_disable_intr(sc);
Import the nfe (nForce Ethernet) driver from OpenBSD. This is a pure open source NIC driver with a BSD license. These bits were taken from the OpenBSD public CVS repository on 1-May-06.
2006-06-26 17:25:22 +00:00
Add a pure open source nForce Ethernet driver, under BSDL. This driver was ported from OpenBSD by Shigeaki Tagashira <shigeaki@se.hiroshima-u.ac.jp> and posted at http://www.se.hiroshima-u.ac.jp/~shigeaki/software/freebsd-nfe.html It was additionally cleaned up by me. It is still a work-in-progress and thus is purposefully not in GENERIC. And it conflicts with nve(4), so only one should be loaded.
2006-06-26 23:41:07 +00:00
sc->nfe_link = 0;
Import the nfe (nForce Ethernet) driver from OpenBSD. This is a pure open source NIC driver with a BSD license. These bits were taken from the OpenBSD public CVS repository on 1-May-06.
2006-06-26 17:25:22 +00:00
Bring overhauled nfe(4) into tree. o s/printf/device_printf/g o Nuke OpenBSDism. o Nuke NetBSD/OpenBSD specific DMA sync operations.(we don't have a way to sync a single descriptor within a DMA map.) o Remove recursive mutex. o bus_dma(9) clean up. o 40bit DMA address support. o Add protection for Rx map load failure. o Fix a long standing bug for watchdog timeout. [1] o Add additional protections, missing Tx completion interrupt, losing start Tx command, for watchdog timeout. o Switch to taskqueue(9) API to handle interrupts. o Use our own timer for watchdog instead of if_watchdog/if_timer interface. o Advertise VLAN header length/capability correctly to upper layer. o Remove excessive kernel stack consumption in nfe_encap(). o Handle highly fragmented mbuf chains correctly. o Enable etherenet address reprogramming with ifconfig(8). o Add ALTQ/TSO, MSI/MSIX support. o Increased Rx ring to 256 descriptors from 128. o Align Tx/Rx descriptor ring on sizeof(struct nfe_desc64) boundary. o Remove alignment restrictions on Tx/Rx buffers. o Rewritten jumbo frame support code. o Add support for hardware assistend VLAN tag insertion/stripping. o Add support for Tx/Rx flow control based on patches from Peer Chen. [2] o Add a routine that detects whether ethernet address swap routines is required. [3] o Add a workaround that take MAC/PHY out of power down mode. o Add suspend/resume support. o style(9) and code clean up. Special thanks to Shigeaki Tagashira, the original porter of nfe(4), who submitted lots of patches, performed uncountable number of regression tests and maintained nfe(4) for a long time. Without his enthusiastic help and support I could never have completed this overhauling task. The only weak point of nfe(4) compared to nve(4) is instability of manual half-duplex media selection on certain hardwares(auto sensing media type should work for all cases, though). This was a long standing bug of nfe(4) and I still have no idea why it doesn't work on some hardwares. Obtained from: OpenBSD [1] Submitted by: Peer Chen < pchen at nvidia dot com > [2], [3] Reviewed by: Shigeaki Tagashira < shigeaki AT se DOT hiroshima-u DOT ac DOT jp > Tested by: Shigeaki Tagashira, current Discussed with: current Silence from: obrien
2007-06-12 02:16:02 +00:00
/* free Rx and Tx mbufs still in the queues. */
rx_ring = &sc->rxq;
for (i = 0; i < NFE_RX_RING_COUNT; i++) {
rdata = &rx_ring->data[i];
if (rdata->m != NULL) {
bus_dmamap_sync(rx_ring->rx_data_tag,
rdata->rx_data_map, BUS_DMASYNC_POSTREAD);
bus_dmamap_unload(rx_ring->rx_data_tag,
rdata->rx_data_map);
m_freem(rdata->m);
rdata->m = NULL;
}
}
style(9) cleanup.
2006-10-19 10:01:26 +00:00
Bring overhauled nfe(4) into tree. o s/printf/device_printf/g o Nuke OpenBSDism. o Nuke NetBSD/OpenBSD specific DMA sync operations.(we don't have a way to sync a single descriptor within a DMA map.) o Remove recursive mutex. o bus_dma(9) clean up. o 40bit DMA address support. o Add protection for Rx map load failure. o Fix a long standing bug for watchdog timeout. [1] o Add additional protections, missing Tx completion interrupt, losing start Tx command, for watchdog timeout. o Switch to taskqueue(9) API to handle interrupts. o Use our own timer for watchdog instead of if_watchdog/if_timer interface. o Advertise VLAN header length/capability correctly to upper layer. o Remove excessive kernel stack consumption in nfe_encap(). o Handle highly fragmented mbuf chains correctly. o Enable etherenet address reprogramming with ifconfig(8). o Add ALTQ/TSO, MSI/MSIX support. o Increased Rx ring to 256 descriptors from 128. o Align Tx/Rx descriptor ring on sizeof(struct nfe_desc64) boundary. o Remove alignment restrictions on Tx/Rx buffers. o Rewritten jumbo frame support code. o Add support for hardware assistend VLAN tag insertion/stripping. o Add support for Tx/Rx flow control based on patches from Peer Chen. [2] o Add a routine that detects whether ethernet address swap routines is required. [3] o Add a workaround that take MAC/PHY out of power down mode. o Add suspend/resume support. o style(9) and code clean up. Special thanks to Shigeaki Tagashira, the original porter of nfe(4), who submitted lots of patches, performed uncountable number of regression tests and maintained nfe(4) for a long time. Without his enthusiastic help and support I could never have completed this overhauling task. The only weak point of nfe(4) compared to nve(4) is instability of manual half-duplex media selection on certain hardwares(auto sensing media type should work for all cases, though). This was a long standing bug of nfe(4) and I still have no idea why it doesn't work on some hardwares. Obtained from: OpenBSD [1] Submitted by: Peer Chen < pchen at nvidia dot com > [2], [3] Reviewed by: Shigeaki Tagashira < shigeaki AT se DOT hiroshima-u DOT ac DOT jp > Tested by: Shigeaki Tagashira, current Discussed with: current Silence from: obrien
2007-06-12 02:16:02 +00:00
if ((sc->nfe_flags & NFE_JUMBO_SUP) != 0) {
jrx_ring = &sc->jrxq;
for (i = 0; i < NFE_JUMBO_RX_RING_COUNT; i++) {
rdata = &jrx_ring->jdata[i];
if (rdata->m != NULL) {
bus_dmamap_sync(jrx_ring->jrx_data_tag,
rdata->rx_data_map, BUS_DMASYNC_POSTREAD);
bus_dmamap_unload(jrx_ring->jrx_data_tag,
rdata->rx_data_map);
m_freem(rdata->m);
rdata->m = NULL;
}
}
}
Add a pure open source nForce Ethernet driver, under BSDL. This driver was ported from OpenBSD by Shigeaki Tagashira <shigeaki@se.hiroshima-u.ac.jp> and posted at http://www.se.hiroshima-u.ac.jp/~shigeaki/software/freebsd-nfe.html It was additionally cleaned up by me. It is still a work-in-progress and thus is purposefully not in GENERIC. And it conflicts with nve(4), so only one should be loaded.
2006-06-26 23:41:07 +00:00
Bring overhauled nfe(4) into tree. o s/printf/device_printf/g o Nuke OpenBSDism. o Nuke NetBSD/OpenBSD specific DMA sync operations.(we don't have a way to sync a single descriptor within a DMA map.) o Remove recursive mutex. o bus_dma(9) clean up. o 40bit DMA address support. o Add protection for Rx map load failure. o Fix a long standing bug for watchdog timeout. [1] o Add additional protections, missing Tx completion interrupt, losing start Tx command, for watchdog timeout. o Switch to taskqueue(9) API to handle interrupts. o Use our own timer for watchdog instead of if_watchdog/if_timer interface. o Advertise VLAN header length/capability correctly to upper layer. o Remove excessive kernel stack consumption in nfe_encap(). o Handle highly fragmented mbuf chains correctly. o Enable etherenet address reprogramming with ifconfig(8). o Add ALTQ/TSO, MSI/MSIX support. o Increased Rx ring to 256 descriptors from 128. o Align Tx/Rx descriptor ring on sizeof(struct nfe_desc64) boundary. o Remove alignment restrictions on Tx/Rx buffers. o Rewritten jumbo frame support code. o Add support for hardware assistend VLAN tag insertion/stripping. o Add support for Tx/Rx flow control based on patches from Peer Chen. [2] o Add a routine that detects whether ethernet address swap routines is required. [3] o Add a workaround that take MAC/PHY out of power down mode. o Add suspend/resume support. o style(9) and code clean up. Special thanks to Shigeaki Tagashira, the original porter of nfe(4), who submitted lots of patches, performed uncountable number of regression tests and maintained nfe(4) for a long time. Without his enthusiastic help and support I could never have completed this overhauling task. The only weak point of nfe(4) compared to nve(4) is instability of manual half-duplex media selection on certain hardwares(auto sensing media type should work for all cases, though). This was a long standing bug of nfe(4) and I still have no idea why it doesn't work on some hardwares. Obtained from: OpenBSD [1] Submitted by: Peer Chen < pchen at nvidia dot com > [2], [3] Reviewed by: Shigeaki Tagashira < shigeaki AT se DOT hiroshima-u DOT ac DOT jp > Tested by: Shigeaki Tagashira, current Discussed with: current Silence from: obrien
2007-06-12 02:16:02 +00:00
tx_ring = &sc->txq;
for (i = 0; i < NFE_RX_RING_COUNT; i++) {
tdata = &tx_ring->data[i];
if (tdata->m != NULL) {
bus_dmamap_sync(tx_ring->tx_data_tag,
tdata->tx_data_map, BUS_DMASYNC_POSTWRITE);
bus_dmamap_unload(tx_ring->tx_data_tag,
tdata->tx_data_map);
m_freem(tdata->m);
tdata->m = NULL;
}
}
Add hardware MAC statistics support. Register information was obtained from Linux forcedeth driver. While I'm here move creating a sysctl node for process_limit to function nfe_sysctl_node(). Tested by: "Arno J. Klaassen" < arno <at> heho dot snv dot jussieu dot fr >
2008-10-03 03:58:16 +00:00
/* Update hardware stats. */
nfe_stats_update(sc);
Import the nfe (nForce Ethernet) driver from OpenBSD. This is a pure open source NIC driver with a BSD license. These bits were taken from the OpenBSD public CVS repository on 1-May-06.
2006-06-26 17:25:22 +00:00
}
style(9) cleanup.
2006-10-19 10:01:26 +00:00
static int
Bring overhauled nfe(4) into tree. o s/printf/device_printf/g o Nuke OpenBSDism. o Nuke NetBSD/OpenBSD specific DMA sync operations.(we don't have a way to sync a single descriptor within a DMA map.) o Remove recursive mutex. o bus_dma(9) clean up. o 40bit DMA address support. o Add protection for Rx map load failure. o Fix a long standing bug for watchdog timeout. [1] o Add additional protections, missing Tx completion interrupt, losing start Tx command, for watchdog timeout. o Switch to taskqueue(9) API to handle interrupts. o Use our own timer for watchdog instead of if_watchdog/if_timer interface. o Advertise VLAN header length/capability correctly to upper layer. o Remove excessive kernel stack consumption in nfe_encap(). o Handle highly fragmented mbuf chains correctly. o Enable etherenet address reprogramming with ifconfig(8). o Add ALTQ/TSO, MSI/MSIX support. o Increased Rx ring to 256 descriptors from 128. o Align Tx/Rx descriptor ring on sizeof(struct nfe_desc64) boundary. o Remove alignment restrictions on Tx/Rx buffers. o Rewritten jumbo frame support code. o Add support for hardware assistend VLAN tag insertion/stripping. o Add support for Tx/Rx flow control based on patches from Peer Chen. [2] o Add a routine that detects whether ethernet address swap routines is required. [3] o Add a workaround that take MAC/PHY out of power down mode. o Add suspend/resume support. o style(9) and code clean up. Special thanks to Shigeaki Tagashira, the original porter of nfe(4), who submitted lots of patches, performed uncountable number of regression tests and maintained nfe(4) for a long time. Without his enthusiastic help and support I could never have completed this overhauling task. The only weak point of nfe(4) compared to nve(4) is instability of manual half-duplex media selection on certain hardwares(auto sensing media type should work for all cases, though). This was a long standing bug of nfe(4) and I still have no idea why it doesn't work on some hardwares. Obtained from: OpenBSD [1] Submitted by: Peer Chen < pchen at nvidia dot com > [2], [3] Reviewed by: Shigeaki Tagashira < shigeaki AT se DOT hiroshima-u DOT ac DOT jp > Tested by: Shigeaki Tagashira, current Discussed with: current Silence from: obrien
2007-06-12 02:16:02 +00:00
nfe_ifmedia_upd(struct ifnet *ifp)
Import the nfe (nForce Ethernet) driver from OpenBSD. This is a pure open source NIC driver with a BSD license. These bits were taken from the OpenBSD public CVS repository on 1-May-06.
2006-06-26 17:25:22 +00:00
{
style(9) cleanup.
2006-10-19 10:01:26 +00:00
struct nfe_softc *sc = ifp->if_softc;
struct mii_data *mii;
Import the nfe (nForce Ethernet) driver from OpenBSD. This is a pure open source NIC driver with a BSD license. These bits were taken from the OpenBSD public CVS repository on 1-May-06.
2006-06-26 17:25:22 +00:00
Bring overhauled nfe(4) into tree. o s/printf/device_printf/g o Nuke OpenBSDism. o Nuke NetBSD/OpenBSD specific DMA sync operations.(we don't have a way to sync a single descriptor within a DMA map.) o Remove recursive mutex. o bus_dma(9) clean up. o 40bit DMA address support. o Add protection for Rx map load failure. o Fix a long standing bug for watchdog timeout. [1] o Add additional protections, missing Tx completion interrupt, losing start Tx command, for watchdog timeout. o Switch to taskqueue(9) API to handle interrupts. o Use our own timer for watchdog instead of if_watchdog/if_timer interface. o Advertise VLAN header length/capability correctly to upper layer. o Remove excessive kernel stack consumption in nfe_encap(). o Handle highly fragmented mbuf chains correctly. o Enable etherenet address reprogramming with ifconfig(8). o Add ALTQ/TSO, MSI/MSIX support. o Increased Rx ring to 256 descriptors from 128. o Align Tx/Rx descriptor ring on sizeof(struct nfe_desc64) boundary. o Remove alignment restrictions on Tx/Rx buffers. o Rewritten jumbo frame support code. o Add support for hardware assistend VLAN tag insertion/stripping. o Add support for Tx/Rx flow control based on patches from Peer Chen. [2] o Add a routine that detects whether ethernet address swap routines is required. [3] o Add a workaround that take MAC/PHY out of power down mode. o Add suspend/resume support. o style(9) and code clean up. Special thanks to Shigeaki Tagashira, the original porter of nfe(4), who submitted lots of patches, performed uncountable number of regression tests and maintained nfe(4) for a long time. Without his enthusiastic help and support I could never have completed this overhauling task. The only weak point of nfe(4) compared to nve(4) is instability of manual half-duplex media selection on certain hardwares(auto sensing media type should work for all cases, though). This was a long standing bug of nfe(4) and I still have no idea why it doesn't work on some hardwares. Obtained from: OpenBSD [1] Submitted by: Peer Chen < pchen at nvidia dot com > [2], [3] Reviewed by: Shigeaki Tagashira < shigeaki AT se DOT hiroshima-u DOT ac DOT jp > Tested by: Shigeaki Tagashira, current Discussed with: current Silence from: obrien
2007-06-12 02:16:02 +00:00
NFE_LOCK(sc);
Add a pure open source nForce Ethernet driver, under BSDL. This driver was ported from OpenBSD by Shigeaki Tagashira <shigeaki@se.hiroshima-u.ac.jp> and posted at http://www.se.hiroshima-u.ac.jp/~shigeaki/software/freebsd-nfe.html It was additionally cleaned up by me. It is still a work-in-progress and thus is purposefully not in GENERIC. And it conflicts with nve(4), so only one should be loaded.
2006-06-26 23:41:07 +00:00
mii = device_get_softc(sc->nfe_miibus);
mii_mediachg(mii);
Bring overhauled nfe(4) into tree. o s/printf/device_printf/g o Nuke OpenBSDism. o Nuke NetBSD/OpenBSD specific DMA sync operations.(we don't have a way to sync a single descriptor within a DMA map.) o Remove recursive mutex. o bus_dma(9) clean up. o 40bit DMA address support. o Add protection for Rx map load failure. o Fix a long standing bug for watchdog timeout. [1] o Add additional protections, missing Tx completion interrupt, losing start Tx command, for watchdog timeout. o Switch to taskqueue(9) API to handle interrupts. o Use our own timer for watchdog instead of if_watchdog/if_timer interface. o Advertise VLAN header length/capability correctly to upper layer. o Remove excessive kernel stack consumption in nfe_encap(). o Handle highly fragmented mbuf chains correctly. o Enable etherenet address reprogramming with ifconfig(8). o Add ALTQ/TSO, MSI/MSIX support. o Increased Rx ring to 256 descriptors from 128. o Align Tx/Rx descriptor ring on sizeof(struct nfe_desc64) boundary. o Remove alignment restrictions on Tx/Rx buffers. o Rewritten jumbo frame support code. o Add support for hardware assistend VLAN tag insertion/stripping. o Add support for Tx/Rx flow control based on patches from Peer Chen. [2] o Add a routine that detects whether ethernet address swap routines is required. [3] o Add a workaround that take MAC/PHY out of power down mode. o Add suspend/resume support. o style(9) and code clean up. Special thanks to Shigeaki Tagashira, the original porter of nfe(4), who submitted lots of patches, performed uncountable number of regression tests and maintained nfe(4) for a long time. Without his enthusiastic help and support I could never have completed this overhauling task. The only weak point of nfe(4) compared to nve(4) is instability of manual half-duplex media selection on certain hardwares(auto sensing media type should work for all cases, though). This was a long standing bug of nfe(4) and I still have no idea why it doesn't work on some hardwares. Obtained from: OpenBSD [1] Submitted by: Peer Chen < pchen at nvidia dot com > [2], [3] Reviewed by: Shigeaki Tagashira < shigeaki AT se DOT hiroshima-u DOT ac DOT jp > Tested by: Shigeaki Tagashira, current Discussed with: current Silence from: obrien
2007-06-12 02:16:02 +00:00
NFE_UNLOCK(sc);
Import the nfe (nForce Ethernet) driver from OpenBSD. This is a pure open source NIC driver with a BSD license. These bits were taken from the OpenBSD public CVS repository on 1-May-06.
2006-06-26 17:25:22 +00:00
Add a pure open source nForce Ethernet driver, under BSDL. This driver was ported from OpenBSD by Shigeaki Tagashira <shigeaki@se.hiroshima-u.ac.jp> and posted at http://www.se.hiroshima-u.ac.jp/~shigeaki/software/freebsd-nfe.html It was additionally cleaned up by me. It is still a work-in-progress and thus is purposefully not in GENERIC. And it conflicts with nve(4), so only one should be loaded.
2006-06-26 23:41:07 +00:00
return (0);
Import the nfe (nForce Ethernet) driver from OpenBSD. This is a pure open source NIC driver with a BSD license. These bits were taken from the OpenBSD public CVS repository on 1-May-06.
2006-06-26 17:25:22 +00:00
}
style(9) cleanup.
2006-10-19 10:01:26 +00:00
static void
nfe_ifmedia_sts(struct ifnet *ifp, struct ifmediareq *ifmr)
Import the nfe (nForce Ethernet) driver from OpenBSD. This is a pure open source NIC driver with a BSD license. These bits were taken from the OpenBSD public CVS repository on 1-May-06.
2006-06-26 17:25:22 +00:00
{
style(9) cleanup.
2006-10-19 10:01:26 +00:00
struct nfe_softc *sc;
struct mii_data *mii;
Import the nfe (nForce Ethernet) driver from OpenBSD. This is a pure open source NIC driver with a BSD license. These bits were taken from the OpenBSD public CVS repository on 1-May-06.
2006-06-26 17:25:22 +00:00
Add a pure open source nForce Ethernet driver, under BSDL. This driver was ported from OpenBSD by Shigeaki Tagashira <shigeaki@se.hiroshima-u.ac.jp> and posted at http://www.se.hiroshima-u.ac.jp/~shigeaki/software/freebsd-nfe.html It was additionally cleaned up by me. It is still a work-in-progress and thus is purposefully not in GENERIC. And it conflicts with nve(4), so only one should be loaded.
2006-06-26 23:41:07 +00:00
sc = ifp->if_softc;
Import the nfe (nForce Ethernet) driver from OpenBSD. This is a pure open source NIC driver with a BSD license. These bits were taken from the OpenBSD public CVS repository on 1-May-06.
2006-06-26 17:25:22 +00:00
Add a pure open source nForce Ethernet driver, under BSDL. This driver was ported from OpenBSD by Shigeaki Tagashira <shigeaki@se.hiroshima-u.ac.jp> and posted at http://www.se.hiroshima-u.ac.jp/~shigeaki/software/freebsd-nfe.html It was additionally cleaned up by me. It is still a work-in-progress and thus is purposefully not in GENERIC. And it conflicts with nve(4), so only one should be loaded.
2006-06-26 23:41:07 +00:00
NFE_LOCK(sc);
mii = device_get_softc(sc->nfe_miibus);
mii_pollstat(mii);
Import the nfe (nForce Ethernet) driver from OpenBSD. This is a pure open source NIC driver with a BSD license. These bits were taken from the OpenBSD public CVS repository on 1-May-06.
2006-06-26 17:25:22 +00:00
Add a pure open source nForce Ethernet driver, under BSDL. This driver was ported from OpenBSD by Shigeaki Tagashira <shigeaki@se.hiroshima-u.ac.jp> and posted at http://www.se.hiroshima-u.ac.jp/~shigeaki/software/freebsd-nfe.html It was additionally cleaned up by me. It is still a work-in-progress and thus is purposefully not in GENERIC. And it conflicts with nve(4), so only one should be loaded.
2006-06-26 23:41:07 +00:00
ifmr->ifm_active = mii->mii_media_active;
ifmr->ifm_status = mii->mii_media_status;
Close a race where SIOCGIFMEDIA ioctl get inconsistent link status. Because driver is accessing a common MII structure in mii_pollstat(), updating user supplied structure should be done before dropping a driver lock. Reported by: Karim (fodillemlinkarimi <> gmail dot com)
2011-10-17 19:49:00 +00:00
NFE_UNLOCK(sc);
Import the nfe (nForce Ethernet) driver from OpenBSD. This is a pure open source NIC driver with a BSD license. These bits were taken from the OpenBSD public CVS repository on 1-May-06.
2006-06-26 17:25:22 +00:00
}
Add a pure open source nForce Ethernet driver, under BSDL. This driver was ported from OpenBSD by Shigeaki Tagashira <shigeaki@se.hiroshima-u.ac.jp> and posted at http://www.se.hiroshima-u.ac.jp/~shigeaki/software/freebsd-nfe.html It was additionally cleaned up by me. It is still a work-in-progress and thus is purposefully not in GENERIC. And it conflicts with nve(4), so only one should be loaded.
2006-06-26 23:41:07 +00:00
style(9) cleanup.
2006-10-19 10:01:26 +00:00
void
Bring overhauled nfe(4) into tree. o s/printf/device_printf/g o Nuke OpenBSDism. o Nuke NetBSD/OpenBSD specific DMA sync operations.(we don't have a way to sync a single descriptor within a DMA map.) o Remove recursive mutex. o bus_dma(9) clean up. o 40bit DMA address support. o Add protection for Rx map load failure. o Fix a long standing bug for watchdog timeout. [1] o Add additional protections, missing Tx completion interrupt, losing start Tx command, for watchdog timeout. o Switch to taskqueue(9) API to handle interrupts. o Use our own timer for watchdog instead of if_watchdog/if_timer interface. o Advertise VLAN header length/capability correctly to upper layer. o Remove excessive kernel stack consumption in nfe_encap(). o Handle highly fragmented mbuf chains correctly. o Enable etherenet address reprogramming with ifconfig(8). o Add ALTQ/TSO, MSI/MSIX support. o Increased Rx ring to 256 descriptors from 128. o Align Tx/Rx descriptor ring on sizeof(struct nfe_desc64) boundary. o Remove alignment restrictions on Tx/Rx buffers. o Rewritten jumbo frame support code. o Add support for hardware assistend VLAN tag insertion/stripping. o Add support for Tx/Rx flow control based on patches from Peer Chen. [2] o Add a routine that detects whether ethernet address swap routines is required. [3] o Add a workaround that take MAC/PHY out of power down mode. o Add suspend/resume support. o style(9) and code clean up. Special thanks to Shigeaki Tagashira, the original porter of nfe(4), who submitted lots of patches, performed uncountable number of regression tests and maintained nfe(4) for a long time. Without his enthusiastic help and support I could never have completed this overhauling task. The only weak point of nfe(4) compared to nve(4) is instability of manual half-duplex media selection on certain hardwares(auto sensing media type should work for all cases, though). This was a long standing bug of nfe(4) and I still have no idea why it doesn't work on some hardwares. Obtained from: OpenBSD [1] Submitted by: Peer Chen < pchen at nvidia dot com > [2], [3] Reviewed by: Shigeaki Tagashira < shigeaki AT se DOT hiroshima-u DOT ac DOT jp > Tested by: Shigeaki Tagashira, current Discussed with: current Silence from: obrien
2007-06-12 02:16:02 +00:00
nfe_tick(void *xsc)
Import the nfe (nForce Ethernet) driver from OpenBSD. This is a pure open source NIC driver with a BSD license. These bits were taken from the OpenBSD public CVS repository on 1-May-06.
2006-06-26 17:25:22 +00:00
{
style(9) cleanup.
2006-10-19 10:01:26 +00:00
struct nfe_softc *sc;
struct mii_data *mii;
struct ifnet *ifp;
Import the nfe (nForce Ethernet) driver from OpenBSD. This is a pure open source NIC driver with a BSD license. These bits were taken from the OpenBSD public CVS repository on 1-May-06.
2006-06-26 17:25:22 +00:00
Bring overhauled nfe(4) into tree. o s/printf/device_printf/g o Nuke OpenBSDism. o Nuke NetBSD/OpenBSD specific DMA sync operations.(we don't have a way to sync a single descriptor within a DMA map.) o Remove recursive mutex. o bus_dma(9) clean up. o 40bit DMA address support. o Add protection for Rx map load failure. o Fix a long standing bug for watchdog timeout. [1] o Add additional protections, missing Tx completion interrupt, losing start Tx command, for watchdog timeout. o Switch to taskqueue(9) API to handle interrupts. o Use our own timer for watchdog instead of if_watchdog/if_timer interface. o Advertise VLAN header length/capability correctly to upper layer. o Remove excessive kernel stack consumption in nfe_encap(). o Handle highly fragmented mbuf chains correctly. o Enable etherenet address reprogramming with ifconfig(8). o Add ALTQ/TSO, MSI/MSIX support. o Increased Rx ring to 256 descriptors from 128. o Align Tx/Rx descriptor ring on sizeof(struct nfe_desc64) boundary. o Remove alignment restrictions on Tx/Rx buffers. o Rewritten jumbo frame support code. o Add support for hardware assistend VLAN tag insertion/stripping. o Add support for Tx/Rx flow control based on patches from Peer Chen. [2] o Add a routine that detects whether ethernet address swap routines is required. [3] o Add a workaround that take MAC/PHY out of power down mode. o Add suspend/resume support. o style(9) and code clean up. Special thanks to Shigeaki Tagashira, the original porter of nfe(4), who submitted lots of patches, performed uncountable number of regression tests and maintained nfe(4) for a long time. Without his enthusiastic help and support I could never have completed this overhauling task. The only weak point of nfe(4) compared to nve(4) is instability of manual half-duplex media selection on certain hardwares(auto sensing media type should work for all cases, though). This was a long standing bug of nfe(4) and I still have no idea why it doesn't work on some hardwares. Obtained from: OpenBSD [1] Submitted by: Peer Chen < pchen at nvidia dot com > [2], [3] Reviewed by: Shigeaki Tagashira < shigeaki AT se DOT hiroshima-u DOT ac DOT jp > Tested by: Shigeaki Tagashira, current Discussed with: current Silence from: obrien
2007-06-12 02:16:02 +00:00
sc = (struct nfe_softc *)xsc;
Import the nfe (nForce Ethernet) driver from OpenBSD. This is a pure open source NIC driver with a BSD license. These bits were taken from the OpenBSD public CVS repository on 1-May-06.
2006-06-26 17:25:22 +00:00
style(9) cleanup.
2006-10-19 10:01:26 +00:00
NFE_LOCK_ASSERT(sc);
Import the nfe (nForce Ethernet) driver from OpenBSD. This is a pure open source NIC driver with a BSD license. These bits were taken from the OpenBSD public CVS repository on 1-May-06.
2006-06-26 17:25:22 +00:00
Add a pure open source nForce Ethernet driver, under BSDL. This driver was ported from OpenBSD by Shigeaki Tagashira <shigeaki@se.hiroshima-u.ac.jp> and posted at http://www.se.hiroshima-u.ac.jp/~shigeaki/software/freebsd-nfe.html It was additionally cleaned up by me. It is still a work-in-progress and thus is purposefully not in GENERIC. And it conflicts with nve(4), so only one should be loaded.
2006-06-26 23:41:07 +00:00
ifp = sc->nfe_ifp;
Import the nfe (nForce Ethernet) driver from OpenBSD. This is a pure open source NIC driver with a BSD license. These bits were taken from the OpenBSD public CVS repository on 1-May-06.
2006-06-26 17:25:22 +00:00
Add a pure open source nForce Ethernet driver, under BSDL. This driver was ported from OpenBSD by Shigeaki Tagashira <shigeaki@se.hiroshima-u.ac.jp> and posted at http://www.se.hiroshima-u.ac.jp/~shigeaki/software/freebsd-nfe.html It was additionally cleaned up by me. It is still a work-in-progress and thus is purposefully not in GENERIC. And it conflicts with nve(4), so only one should be loaded.
2006-06-26 23:41:07 +00:00
mii = device_get_softc(sc->nfe_miibus);
mii_tick(mii);
Add hardware MAC statistics support. Register information was obtained from Linux forcedeth driver. While I'm here move creating a sysctl node for process_limit to function nfe_sysctl_node(). Tested by: "Arno J. Klaassen" < arno <at> heho dot snv dot jussieu dot fr >
2008-10-03 03:58:16 +00:00
nfe_stats_update(sc);
Bring overhauled nfe(4) into tree. o s/printf/device_printf/g o Nuke OpenBSDism. o Nuke NetBSD/OpenBSD specific DMA sync operations.(we don't have a way to sync a single descriptor within a DMA map.) o Remove recursive mutex. o bus_dma(9) clean up. o 40bit DMA address support. o Add protection for Rx map load failure. o Fix a long standing bug for watchdog timeout. [1] o Add additional protections, missing Tx completion interrupt, losing start Tx command, for watchdog timeout. o Switch to taskqueue(9) API to handle interrupts. o Use our own timer for watchdog instead of if_watchdog/if_timer interface. o Advertise VLAN header length/capability correctly to upper layer. o Remove excessive kernel stack consumption in nfe_encap(). o Handle highly fragmented mbuf chains correctly. o Enable etherenet address reprogramming with ifconfig(8). o Add ALTQ/TSO, MSI/MSIX support. o Increased Rx ring to 256 descriptors from 128. o Align Tx/Rx descriptor ring on sizeof(struct nfe_desc64) boundary. o Remove alignment restrictions on Tx/Rx buffers. o Rewritten jumbo frame support code. o Add support for hardware assistend VLAN tag insertion/stripping. o Add support for Tx/Rx flow control based on patches from Peer Chen. [2] o Add a routine that detects whether ethernet address swap routines is required. [3] o Add a workaround that take MAC/PHY out of power down mode. o Add suspend/resume support. o style(9) and code clean up. Special thanks to Shigeaki Tagashira, the original porter of nfe(4), who submitted lots of patches, performed uncountable number of regression tests and maintained nfe(4) for a long time. Without his enthusiastic help and support I could never have completed this overhauling task. The only weak point of nfe(4) compared to nve(4) is instability of manual half-duplex media selection on certain hardwares(auto sensing media type should work for all cases, though). This was a long standing bug of nfe(4) and I still have no idea why it doesn't work on some hardwares. Obtained from: OpenBSD [1] Submitted by: Peer Chen < pchen at nvidia dot com > [2], [3] Reviewed by: Shigeaki Tagashira < shigeaki AT se DOT hiroshima-u DOT ac DOT jp > Tested by: Shigeaki Tagashira, current Discussed with: current Silence from: obrien
2007-06-12 02:16:02 +00:00
nfe_watchdog(ifp);
Add a pure open source nForce Ethernet driver, under BSDL. This driver was ported from OpenBSD by Shigeaki Tagashira <shigeaki@se.hiroshima-u.ac.jp> and posted at http://www.se.hiroshima-u.ac.jp/~shigeaki/software/freebsd-nfe.html It was additionally cleaned up by me. It is still a work-in-progress and thus is purposefully not in GENERIC. And it conflicts with nve(4), so only one should be loaded.
2006-06-26 23:41:07 +00:00
callout_reset(&sc->nfe_stat_ch, hz, nfe_tick, sc);
}
Import the nfe (nForce Ethernet) driver from OpenBSD. This is a pure open source NIC driver with a BSD license. These bits were taken from the OpenBSD public CVS repository on 1-May-06.
2006-06-26 17:25:22 +00:00
Fix function prototype for device_shutdown method.
2007-11-22 02:45:00 +00:00
static int
style(9) cleanup.
2006-10-19 10:01:26 +00:00
nfe_shutdown(device_t dev)
Add a pure open source nForce Ethernet driver, under BSDL. This driver was ported from OpenBSD by Shigeaki Tagashira <shigeaki@se.hiroshima-u.ac.jp> and posted at http://www.se.hiroshima-u.ac.jp/~shigeaki/software/freebsd-nfe.html It was additionally cleaned up by me. It is still a work-in-progress and thus is purposefully not in GENERIC. And it conflicts with nve(4), so only one should be loaded.
2006-06-26 23:41:07 +00:00
{
Add basic WOL support for MCP ethernet controllers. It seems the controller does not perform automatic switching from 1000Mbps link to 10/100Mbps link when WOL is activated. Implement establishing 10/100Mps link with auto-negotiation in driver. Link status change handler was modified to remove taskqueue based approach since driver now needs synchronous handling for link establishment. Submitted by: Yamagi Burmeister (lists <> yamagi.org ) (initial version) Tested by: Yamagi Burmeister (lists <> yamagi.org ) MFC after: 1 week
2010-11-11 18:08:50 +00:00
return (nfe_suspend(dev));
Import the nfe (nForce Ethernet) driver from OpenBSD. This is a pure open source NIC driver with a BSD license. These bits were taken from the OpenBSD public CVS repository on 1-May-06.
2006-06-26 17:25:22 +00:00
}
Add a pure open source nForce Ethernet driver, under BSDL. This driver was ported from OpenBSD by Shigeaki Tagashira <shigeaki@se.hiroshima-u.ac.jp> and posted at http://www.se.hiroshima-u.ac.jp/~shigeaki/software/freebsd-nfe.html It was additionally cleaned up by me. It is still a work-in-progress and thus is purposefully not in GENERIC. And it conflicts with nve(4), so only one should be loaded.
2006-06-26 23:41:07 +00:00
style(9) cleanup.
2006-10-19 10:01:26 +00:00
static void
Bring overhauled nfe(4) into tree. o s/printf/device_printf/g o Nuke OpenBSDism. o Nuke NetBSD/OpenBSD specific DMA sync operations.(we don't have a way to sync a single descriptor within a DMA map.) o Remove recursive mutex. o bus_dma(9) clean up. o 40bit DMA address support. o Add protection for Rx map load failure. o Fix a long standing bug for watchdog timeout. [1] o Add additional protections, missing Tx completion interrupt, losing start Tx command, for watchdog timeout. o Switch to taskqueue(9) API to handle interrupts. o Use our own timer for watchdog instead of if_watchdog/if_timer interface. o Advertise VLAN header length/capability correctly to upper layer. o Remove excessive kernel stack consumption in nfe_encap(). o Handle highly fragmented mbuf chains correctly. o Enable etherenet address reprogramming with ifconfig(8). o Add ALTQ/TSO, MSI/MSIX support. o Increased Rx ring to 256 descriptors from 128. o Align Tx/Rx descriptor ring on sizeof(struct nfe_desc64) boundary. o Remove alignment restrictions on Tx/Rx buffers. o Rewritten jumbo frame support code. o Add support for hardware assistend VLAN tag insertion/stripping. o Add support for Tx/Rx flow control based on patches from Peer Chen. [2] o Add a routine that detects whether ethernet address swap routines is required. [3] o Add a workaround that take MAC/PHY out of power down mode. o Add suspend/resume support. o style(9) and code clean up. Special thanks to Shigeaki Tagashira, the original porter of nfe(4), who submitted lots of patches, performed uncountable number of regression tests and maintained nfe(4) for a long time. Without his enthusiastic help and support I could never have completed this overhauling task. The only weak point of nfe(4) compared to nve(4) is instability of manual half-duplex media selection on certain hardwares(auto sensing media type should work for all cases, though). This was a long standing bug of nfe(4) and I still have no idea why it doesn't work on some hardwares. Obtained from: OpenBSD [1] Submitted by: Peer Chen < pchen at nvidia dot com > [2], [3] Reviewed by: Shigeaki Tagashira < shigeaki AT se DOT hiroshima-u DOT ac DOT jp > Tested by: Shigeaki Tagashira, current Discussed with: current Silence from: obrien
2007-06-12 02:16:02 +00:00
nfe_get_macaddr(struct nfe_softc *sc, uint8_t *addr)
Import the nfe (nForce Ethernet) driver from OpenBSD. This is a pure open source NIC driver with a BSD license. These bits were taken from the OpenBSD public CVS repository on 1-May-06.
2006-06-26 17:25:22 +00:00
{
Bring overhauled nfe(4) into tree. o s/printf/device_printf/g o Nuke OpenBSDism. o Nuke NetBSD/OpenBSD specific DMA sync operations.(we don't have a way to sync a single descriptor within a DMA map.) o Remove recursive mutex. o bus_dma(9) clean up. o 40bit DMA address support. o Add protection for Rx map load failure. o Fix a long standing bug for watchdog timeout. [1] o Add additional protections, missing Tx completion interrupt, losing start Tx command, for watchdog timeout. o Switch to taskqueue(9) API to handle interrupts. o Use our own timer for watchdog instead of if_watchdog/if_timer interface. o Advertise VLAN header length/capability correctly to upper layer. o Remove excessive kernel stack consumption in nfe_encap(). o Handle highly fragmented mbuf chains correctly. o Enable etherenet address reprogramming with ifconfig(8). o Add ALTQ/TSO, MSI/MSIX support. o Increased Rx ring to 256 descriptors from 128. o Align Tx/Rx descriptor ring on sizeof(struct nfe_desc64) boundary. o Remove alignment restrictions on Tx/Rx buffers. o Rewritten jumbo frame support code. o Add support for hardware assistend VLAN tag insertion/stripping. o Add support for Tx/Rx flow control based on patches from Peer Chen. [2] o Add a routine that detects whether ethernet address swap routines is required. [3] o Add a workaround that take MAC/PHY out of power down mode. o Add suspend/resume support. o style(9) and code clean up. Special thanks to Shigeaki Tagashira, the original porter of nfe(4), who submitted lots of patches, performed uncountable number of regression tests and maintained nfe(4) for a long time. Without his enthusiastic help and support I could never have completed this overhauling task. The only weak point of nfe(4) compared to nve(4) is instability of manual half-duplex media selection on certain hardwares(auto sensing media type should work for all cases, though). This was a long standing bug of nfe(4) and I still have no idea why it doesn't work on some hardwares. Obtained from: OpenBSD [1] Submitted by: Peer Chen < pchen at nvidia dot com > [2], [3] Reviewed by: Shigeaki Tagashira < shigeaki AT se DOT hiroshima-u DOT ac DOT jp > Tested by: Shigeaki Tagashira, current Discussed with: current Silence from: obrien
2007-06-12 02:16:02 +00:00
uint32_t val;
if ((sc->nfe_flags & NFE_CORRECT_MACADDR) == 0) {
val = NFE_READ(sc, NFE_MACADDR_LO);
addr[0] = (val >> 8) & 0xff;
addr[1] = (val & 0xff);
val = NFE_READ(sc, NFE_MACADDR_HI);
addr[2] = (val >> 24) & 0xff;
addr[3] = (val >> 16) & 0xff;
addr[4] = (val >> 8) & 0xff;
addr[5] = (val & 0xff);
} else {
val = NFE_READ(sc, NFE_MACADDR_LO);
addr[5] = (val >> 8) & 0xff;
addr[4] = (val & 0xff);
Import the nfe (nForce Ethernet) driver from OpenBSD. This is a pure open source NIC driver with a BSD license. These bits were taken from the OpenBSD public CVS repository on 1-May-06.
2006-06-26 17:25:22 +00:00
Bring overhauled nfe(4) into tree. o s/printf/device_printf/g o Nuke OpenBSDism. o Nuke NetBSD/OpenBSD specific DMA sync operations.(we don't have a way to sync a single descriptor within a DMA map.) o Remove recursive mutex. o bus_dma(9) clean up. o 40bit DMA address support. o Add protection for Rx map load failure. o Fix a long standing bug for watchdog timeout. [1] o Add additional protections, missing Tx completion interrupt, losing start Tx command, for watchdog timeout. o Switch to taskqueue(9) API to handle interrupts. o Use our own timer for watchdog instead of if_watchdog/if_timer interface. o Advertise VLAN header length/capability correctly to upper layer. o Remove excessive kernel stack consumption in nfe_encap(). o Handle highly fragmented mbuf chains correctly. o Enable etherenet address reprogramming with ifconfig(8). o Add ALTQ/TSO, MSI/MSIX support. o Increased Rx ring to 256 descriptors from 128. o Align Tx/Rx descriptor ring on sizeof(struct nfe_desc64) boundary. o Remove alignment restrictions on Tx/Rx buffers. o Rewritten jumbo frame support code. o Add support for hardware assistend VLAN tag insertion/stripping. o Add support for Tx/Rx flow control based on patches from Peer Chen. [2] o Add a routine that detects whether ethernet address swap routines is required. [3] o Add a workaround that take MAC/PHY out of power down mode. o Add suspend/resume support. o style(9) and code clean up. Special thanks to Shigeaki Tagashira, the original porter of nfe(4), who submitted lots of patches, performed uncountable number of regression tests and maintained nfe(4) for a long time. Without his enthusiastic help and support I could never have completed this overhauling task. The only weak point of nfe(4) compared to nve(4) is instability of manual half-duplex media selection on certain hardwares(auto sensing media type should work for all cases, though). This was a long standing bug of nfe(4) and I still have no idea why it doesn't work on some hardwares. Obtained from: OpenBSD [1] Submitted by: Peer Chen < pchen at nvidia dot com > [2], [3] Reviewed by: Shigeaki Tagashira < shigeaki AT se DOT hiroshima-u DOT ac DOT jp > Tested by: Shigeaki Tagashira, current Discussed with: current Silence from: obrien
2007-06-12 02:16:02 +00:00
val = NFE_READ(sc, NFE_MACADDR_HI);
addr[3] = (val >> 24) & 0xff;
addr[2] = (val >> 16) & 0xff;
addr[1] = (val >> 8) & 0xff;
addr[0] = (val & 0xff);
}
Import the nfe (nForce Ethernet) driver from OpenBSD. This is a pure open source NIC driver with a BSD license. These bits were taken from the OpenBSD public CVS repository on 1-May-06.
2006-06-26 17:25:22 +00:00
}
style(9) cleanup.
2006-10-19 10:01:26 +00:00
static void
Bring overhauled nfe(4) into tree. o s/printf/device_printf/g o Nuke OpenBSDism. o Nuke NetBSD/OpenBSD specific DMA sync operations.(we don't have a way to sync a single descriptor within a DMA map.) o Remove recursive mutex. o bus_dma(9) clean up. o 40bit DMA address support. o Add protection for Rx map load failure. o Fix a long standing bug for watchdog timeout. [1] o Add additional protections, missing Tx completion interrupt, losing start Tx command, for watchdog timeout. o Switch to taskqueue(9) API to handle interrupts. o Use our own timer for watchdog instead of if_watchdog/if_timer interface. o Advertise VLAN header length/capability correctly to upper layer. o Remove excessive kernel stack consumption in nfe_encap(). o Handle highly fragmented mbuf chains correctly. o Enable etherenet address reprogramming with ifconfig(8). o Add ALTQ/TSO, MSI/MSIX support. o Increased Rx ring to 256 descriptors from 128. o Align Tx/Rx descriptor ring on sizeof(struct nfe_desc64) boundary. o Remove alignment restrictions on Tx/Rx buffers. o Rewritten jumbo frame support code. o Add support for hardware assistend VLAN tag insertion/stripping. o Add support for Tx/Rx flow control based on patches from Peer Chen. [2] o Add a routine that detects whether ethernet address swap routines is required. [3] o Add a workaround that take MAC/PHY out of power down mode. o Add suspend/resume support. o style(9) and code clean up. Special thanks to Shigeaki Tagashira, the original porter of nfe(4), who submitted lots of patches, performed uncountable number of regression tests and maintained nfe(4) for a long time. Without his enthusiastic help and support I could never have completed this overhauling task. The only weak point of nfe(4) compared to nve(4) is instability of manual half-duplex media selection on certain hardwares(auto sensing media type should work for all cases, though). This was a long standing bug of nfe(4) and I still have no idea why it doesn't work on some hardwares. Obtained from: OpenBSD [1] Submitted by: Peer Chen < pchen at nvidia dot com > [2], [3] Reviewed by: Shigeaki Tagashira < shigeaki AT se DOT hiroshima-u DOT ac DOT jp > Tested by: Shigeaki Tagashira, current Discussed with: current Silence from: obrien
2007-06-12 02:16:02 +00:00
nfe_set_macaddr(struct nfe_softc *sc, uint8_t *addr)
Import the nfe (nForce Ethernet) driver from OpenBSD. This is a pure open source NIC driver with a BSD license. These bits were taken from the OpenBSD public CVS repository on 1-May-06.
2006-06-26 17:25:22 +00:00
{
Add a pure open source nForce Ethernet driver, under BSDL. This driver was ported from OpenBSD by Shigeaki Tagashira <shigeaki@se.hiroshima-u.ac.jp> and posted at http://www.se.hiroshima-u.ac.jp/~shigeaki/software/freebsd-nfe.html It was additionally cleaned up by me. It is still a work-in-progress and thus is purposefully not in GENERIC. And it conflicts with nve(4), so only one should be loaded.
2006-06-26 23:41:07 +00:00
NFE_WRITE(sc, NFE_MACADDR_LO, addr[5] << 8 | addr[4]);
NFE_WRITE(sc, NFE_MACADDR_HI, addr[3] << 24 | addr[2] << 16 |
addr[1] << 8 | addr[0]);
Import the nfe (nForce Ethernet) driver from OpenBSD. This is a pure open source NIC driver with a BSD license. These bits were taken from the OpenBSD public CVS repository on 1-May-06.
2006-06-26 17:25:22 +00:00
}
style(9) cleanup.
2006-10-19 10:01:26 +00:00
Add a pure open source nForce Ethernet driver, under BSDL. This driver was ported from OpenBSD by Shigeaki Tagashira <shigeaki@se.hiroshima-u.ac.jp> and posted at http://www.se.hiroshima-u.ac.jp/~shigeaki/software/freebsd-nfe.html It was additionally cleaned up by me. It is still a work-in-progress and thus is purposefully not in GENERIC. And it conflicts with nve(4), so only one should be loaded.
2006-06-26 23:41:07 +00:00
/*
* Map a single buffer address.
*/
static void
Bring overhauled nfe(4) into tree. o s/printf/device_printf/g o Nuke OpenBSDism. o Nuke NetBSD/OpenBSD specific DMA sync operations.(we don't have a way to sync a single descriptor within a DMA map.) o Remove recursive mutex. o bus_dma(9) clean up. o 40bit DMA address support. o Add protection for Rx map load failure. o Fix a long standing bug for watchdog timeout. [1] o Add additional protections, missing Tx completion interrupt, losing start Tx command, for watchdog timeout. o Switch to taskqueue(9) API to handle interrupts. o Use our own timer for watchdog instead of if_watchdog/if_timer interface. o Advertise VLAN header length/capability correctly to upper layer. o Remove excessive kernel stack consumption in nfe_encap(). o Handle highly fragmented mbuf chains correctly. o Enable etherenet address reprogramming with ifconfig(8). o Add ALTQ/TSO, MSI/MSIX support. o Increased Rx ring to 256 descriptors from 128. o Align Tx/Rx descriptor ring on sizeof(struct nfe_desc64) boundary. o Remove alignment restrictions on Tx/Rx buffers. o Rewritten jumbo frame support code. o Add support for hardware assistend VLAN tag insertion/stripping. o Add support for Tx/Rx flow control based on patches from Peer Chen. [2] o Add a routine that detects whether ethernet address swap routines is required. [3] o Add a workaround that take MAC/PHY out of power down mode. o Add suspend/resume support. o style(9) and code clean up. Special thanks to Shigeaki Tagashira, the original porter of nfe(4), who submitted lots of patches, performed uncountable number of regression tests and maintained nfe(4) for a long time. Without his enthusiastic help and support I could never have completed this overhauling task. The only weak point of nfe(4) compared to nve(4) is instability of manual half-duplex media selection on certain hardwares(auto sensing media type should work for all cases, though). This was a long standing bug of nfe(4) and I still have no idea why it doesn't work on some hardwares. Obtained from: OpenBSD [1] Submitted by: Peer Chen < pchen at nvidia dot com > [2], [3] Reviewed by: Shigeaki Tagashira < shigeaki AT se DOT hiroshima-u DOT ac DOT jp > Tested by: Shigeaki Tagashira, current Discussed with: current Silence from: obrien
2007-06-12 02:16:02 +00:00
nfe_dma_map_segs(void *arg, bus_dma_segment_t *segs, int nseg, int error)
Import the nfe (nForce Ethernet) driver from OpenBSD. This is a pure open source NIC driver with a BSD license. These bits were taken from the OpenBSD public CVS repository on 1-May-06.
2006-06-26 17:25:22 +00:00
{
Bring overhauled nfe(4) into tree. o s/printf/device_printf/g o Nuke OpenBSDism. o Nuke NetBSD/OpenBSD specific DMA sync operations.(we don't have a way to sync a single descriptor within a DMA map.) o Remove recursive mutex. o bus_dma(9) clean up. o 40bit DMA address support. o Add protection for Rx map load failure. o Fix a long standing bug for watchdog timeout. [1] o Add additional protections, missing Tx completion interrupt, losing start Tx command, for watchdog timeout. o Switch to taskqueue(9) API to handle interrupts. o Use our own timer for watchdog instead of if_watchdog/if_timer interface. o Advertise VLAN header length/capability correctly to upper layer. o Remove excessive kernel stack consumption in nfe_encap(). o Handle highly fragmented mbuf chains correctly. o Enable etherenet address reprogramming with ifconfig(8). o Add ALTQ/TSO, MSI/MSIX support. o Increased Rx ring to 256 descriptors from 128. o Align Tx/Rx descriptor ring on sizeof(struct nfe_desc64) boundary. o Remove alignment restrictions on Tx/Rx buffers. o Rewritten jumbo frame support code. o Add support for hardware assistend VLAN tag insertion/stripping. o Add support for Tx/Rx flow control based on patches from Peer Chen. [2] o Add a routine that detects whether ethernet address swap routines is required. [3] o Add a workaround that take MAC/PHY out of power down mode. o Add suspend/resume support. o style(9) and code clean up. Special thanks to Shigeaki Tagashira, the original porter of nfe(4), who submitted lots of patches, performed uncountable number of regression tests and maintained nfe(4) for a long time. Without his enthusiastic help and support I could never have completed this overhauling task. The only weak point of nfe(4) compared to nve(4) is instability of manual half-duplex media selection on certain hardwares(auto sensing media type should work for all cases, though). This was a long standing bug of nfe(4) and I still have no idea why it doesn't work on some hardwares. Obtained from: OpenBSD [1] Submitted by: Peer Chen < pchen at nvidia dot com > [2], [3] Reviewed by: Shigeaki Tagashira < shigeaki AT se DOT hiroshima-u DOT ac DOT jp > Tested by: Shigeaki Tagashira, current Discussed with: current Silence from: obrien
2007-06-12 02:16:02 +00:00
struct nfe_dmamap_arg *ctx;
Import the nfe (nForce Ethernet) driver from OpenBSD. This is a pure open source NIC driver with a BSD license. These bits were taken from the OpenBSD public CVS repository on 1-May-06.
2006-06-26 17:25:22 +00:00
Bring overhauled nfe(4) into tree. o s/printf/device_printf/g o Nuke OpenBSDism. o Nuke NetBSD/OpenBSD specific DMA sync operations.(we don't have a way to sync a single descriptor within a DMA map.) o Remove recursive mutex. o bus_dma(9) clean up. o 40bit DMA address support. o Add protection for Rx map load failure. o Fix a long standing bug for watchdog timeout. [1] o Add additional protections, missing Tx completion interrupt, losing start Tx command, for watchdog timeout. o Switch to taskqueue(9) API to handle interrupts. o Use our own timer for watchdog instead of if_watchdog/if_timer interface. o Advertise VLAN header length/capability correctly to upper layer. o Remove excessive kernel stack consumption in nfe_encap(). o Handle highly fragmented mbuf chains correctly. o Enable etherenet address reprogramming with ifconfig(8). o Add ALTQ/TSO, MSI/MSIX support. o Increased Rx ring to 256 descriptors from 128. o Align Tx/Rx descriptor ring on sizeof(struct nfe_desc64) boundary. o Remove alignment restrictions on Tx/Rx buffers. o Rewritten jumbo frame support code. o Add support for hardware assistend VLAN tag insertion/stripping. o Add support for Tx/Rx flow control based on patches from Peer Chen. [2] o Add a routine that detects whether ethernet address swap routines is required. [3] o Add a workaround that take MAC/PHY out of power down mode. o Add suspend/resume support. o style(9) and code clean up. Special thanks to Shigeaki Tagashira, the original porter of nfe(4), who submitted lots of patches, performed uncountable number of regression tests and maintained nfe(4) for a long time. Without his enthusiastic help and support I could never have completed this overhauling task. The only weak point of nfe(4) compared to nve(4) is instability of manual half-duplex media selection on certain hardwares(auto sensing media type should work for all cases, though). This was a long standing bug of nfe(4) and I still have no idea why it doesn't work on some hardwares. Obtained from: OpenBSD [1] Submitted by: Peer Chen < pchen at nvidia dot com > [2], [3] Reviewed by: Shigeaki Tagashira < shigeaki AT se DOT hiroshima-u DOT ac DOT jp > Tested by: Shigeaki Tagashira, current Discussed with: current Silence from: obrien
2007-06-12 02:16:02 +00:00
if (error != 0)
Add a pure open source nForce Ethernet driver, under BSDL. This driver was ported from OpenBSD by Shigeaki Tagashira <shigeaki@se.hiroshima-u.ac.jp> and posted at http://www.se.hiroshima-u.ac.jp/~shigeaki/software/freebsd-nfe.html It was additionally cleaned up by me. It is still a work-in-progress and thus is purposefully not in GENERIC. And it conflicts with nve(4), so only one should be loaded.
2006-06-26 23:41:07 +00:00
return;
KASSERT(nseg == 1, ("too many DMA segments, %d should be 1", nseg));
Bring overhauled nfe(4) into tree. o s/printf/device_printf/g o Nuke OpenBSDism. o Nuke NetBSD/OpenBSD specific DMA sync operations.(we don't have a way to sync a single descriptor within a DMA map.) o Remove recursive mutex. o bus_dma(9) clean up. o 40bit DMA address support. o Add protection for Rx map load failure. o Fix a long standing bug for watchdog timeout. [1] o Add additional protections, missing Tx completion interrupt, losing start Tx command, for watchdog timeout. o Switch to taskqueue(9) API to handle interrupts. o Use our own timer for watchdog instead of if_watchdog/if_timer interface. o Advertise VLAN header length/capability correctly to upper layer. o Remove excessive kernel stack consumption in nfe_encap(). o Handle highly fragmented mbuf chains correctly. o Enable etherenet address reprogramming with ifconfig(8). o Add ALTQ/TSO, MSI/MSIX support. o Increased Rx ring to 256 descriptors from 128. o Align Tx/Rx descriptor ring on sizeof(struct nfe_desc64) boundary. o Remove alignment restrictions on Tx/Rx buffers. o Rewritten jumbo frame support code. o Add support for hardware assistend VLAN tag insertion/stripping. o Add support for Tx/Rx flow control based on patches from Peer Chen. [2] o Add a routine that detects whether ethernet address swap routines is required. [3] o Add a workaround that take MAC/PHY out of power down mode. o Add suspend/resume support. o style(9) and code clean up. Special thanks to Shigeaki Tagashira, the original porter of nfe(4), who submitted lots of patches, performed uncountable number of regression tests and maintained nfe(4) for a long time. Without his enthusiastic help and support I could never have completed this overhauling task. The only weak point of nfe(4) compared to nve(4) is instability of manual half-duplex media selection on certain hardwares(auto sensing media type should work for all cases, though). This was a long standing bug of nfe(4) and I still have no idea why it doesn't work on some hardwares. Obtained from: OpenBSD [1] Submitted by: Peer Chen < pchen at nvidia dot com > [2], [3] Reviewed by: Shigeaki Tagashira < shigeaki AT se DOT hiroshima-u DOT ac DOT jp > Tested by: Shigeaki Tagashira, current Discussed with: current Silence from: obrien
2007-06-12 02:16:02 +00:00
ctx = (struct nfe_dmamap_arg *)arg;
ctx->nfe_busaddr = segs[0].ds_addr;
}
static int
sysctl_int_range(SYSCTL_HANDLER_ARGS, int low, int high)
{
int error, value;
if (!arg1)
return (EINVAL);
value = *(int *)arg1;
error = sysctl_handle_int(oidp, &value, 0, req);
if (error || !req->newptr)
return (error);
if (value < low || value > high)
return (EINVAL);
*(int *)arg1 = value;
return (0);
}
static int
sysctl_hw_nfe_proc_limit(SYSCTL_HANDLER_ARGS)
{
Import the nfe (nForce Ethernet) driver from OpenBSD. This is a pure open source NIC driver with a BSD license. These bits were taken from the OpenBSD public CVS repository on 1-May-06.
2006-06-26 17:25:22 +00:00
Bring overhauled nfe(4) into tree. o s/printf/device_printf/g o Nuke OpenBSDism. o Nuke NetBSD/OpenBSD specific DMA sync operations.(we don't have a way to sync a single descriptor within a DMA map.) o Remove recursive mutex. o bus_dma(9) clean up. o 40bit DMA address support. o Add protection for Rx map load failure. o Fix a long standing bug for watchdog timeout. [1] o Add additional protections, missing Tx completion interrupt, losing start Tx command, for watchdog timeout. o Switch to taskqueue(9) API to handle interrupts. o Use our own timer for watchdog instead of if_watchdog/if_timer interface. o Advertise VLAN header length/capability correctly to upper layer. o Remove excessive kernel stack consumption in nfe_encap(). o Handle highly fragmented mbuf chains correctly. o Enable etherenet address reprogramming with ifconfig(8). o Add ALTQ/TSO, MSI/MSIX support. o Increased Rx ring to 256 descriptors from 128. o Align Tx/Rx descriptor ring on sizeof(struct nfe_desc64) boundary. o Remove alignment restrictions on Tx/Rx buffers. o Rewritten jumbo frame support code. o Add support for hardware assistend VLAN tag insertion/stripping. o Add support for Tx/Rx flow control based on patches from Peer Chen. [2] o Add a routine that detects whether ethernet address swap routines is required. [3] o Add a workaround that take MAC/PHY out of power down mode. o Add suspend/resume support. o style(9) and code clean up. Special thanks to Shigeaki Tagashira, the original porter of nfe(4), who submitted lots of patches, performed uncountable number of regression tests and maintained nfe(4) for a long time. Without his enthusiastic help and support I could never have completed this overhauling task. The only weak point of nfe(4) compared to nve(4) is instability of manual half-duplex media selection on certain hardwares(auto sensing media type should work for all cases, though). This was a long standing bug of nfe(4) and I still have no idea why it doesn't work on some hardwares. Obtained from: OpenBSD [1] Submitted by: Peer Chen < pchen at nvidia dot com > [2], [3] Reviewed by: Shigeaki Tagashira < shigeaki AT se DOT hiroshima-u DOT ac DOT jp > Tested by: Shigeaki Tagashira, current Discussed with: current Silence from: obrien
2007-06-12 02:16:02 +00:00
return (sysctl_int_range(oidp, arg1, arg2, req, NFE_PROC_MIN,
NFE_PROC_MAX));
Import the nfe (nForce Ethernet) driver from OpenBSD. This is a pure open source NIC driver with a BSD license. These bits were taken from the OpenBSD public CVS repository on 1-May-06.
2006-06-26 17:25:22 +00:00
}
Add hardware MAC statistics support. Register information was obtained from Linux forcedeth driver. While I'm here move creating a sysctl node for process_limit to function nfe_sysctl_node(). Tested by: "Arno J. Klaassen" < arno <at> heho dot snv dot jussieu dot fr >
2008-10-03 03:58:16 +00:00
#define NFE_SYSCTL_STAT_ADD32(c, h, n, p, d) \
SYSCTL_ADD_UINT(c, h, OID_AUTO, n, CTLFLAG_RD, p, 0, d)
#define NFE_SYSCTL_STAT_ADD64(c, h, n, p, d) \
sysctl(9) cleanup checkpoint: amd64 GENERIC builds cleanly. Commit the rest of the devices.
2011-01-12 19:53:56 +00:00
SYSCTL_ADD_UQUAD(c, h, OID_AUTO, n, CTLFLAG_RD, p, d)
Add hardware MAC statistics support. Register information was obtained from Linux forcedeth driver. While I'm here move creating a sysctl node for process_limit to function nfe_sysctl_node(). Tested by: "Arno J. Klaassen" < arno <at> heho dot snv dot jussieu dot fr >
2008-10-03 03:58:16 +00:00
static void
nfe_sysctl_node(struct nfe_softc *sc)
{
struct sysctl_ctx_list *ctx;
struct sysctl_oid_list *child, *parent;
struct sysctl_oid *tree;
struct nfe_hw_stats *stats;
int error;
stats = &sc->nfe_stats;
ctx = device_get_sysctl_ctx(sc->nfe_dev);
child = SYSCTL_CHILDREN(device_get_sysctl_tree(sc->nfe_dev));
SYSCTL_ADD_PROC(ctx, child,
OID_AUTO, "process_limit", CTLTYPE_INT | CTLFLAG_RW,
&sc->nfe_process_limit, 0, sysctl_hw_nfe_proc_limit, "I",
"max number of Rx events to process");
sc->nfe_process_limit = NFE_PROC_DEFAULT;
error = resource_int_value(device_get_name(sc->nfe_dev),
device_get_unit(sc->nfe_dev), "process_limit",
&sc->nfe_process_limit);
if (error == 0) {
if (sc->nfe_process_limit < NFE_PROC_MIN ||
sc->nfe_process_limit > NFE_PROC_MAX) {
device_printf(sc->nfe_dev,
"process_limit value out of range; "
"using default: %d\n", NFE_PROC_DEFAULT);
sc->nfe_process_limit = NFE_PROC_DEFAULT;
}
}
if ((sc->nfe_flags & (NFE_MIB_V1 | NFE_MIB_V2 | NFE_MIB_V3)) == 0)
return;
tree = SYSCTL_ADD_NODE(ctx, child, OID_AUTO, "stats", CTLFLAG_RD,
NULL, "NFE statistics");
parent = SYSCTL_CHILDREN(tree);
/* Rx statistics. */
tree = SYSCTL_ADD_NODE(ctx, parent, OID_AUTO, "rx", CTLFLAG_RD,
NULL, "Rx MAC statistics");
child = SYSCTL_CHILDREN(tree);
NFE_SYSCTL_STAT_ADD32(ctx, child, "frame_errors",
&stats->rx_frame_errors, "Framing Errors");
NFE_SYSCTL_STAT_ADD32(ctx, child, "extra_bytes",
&stats->rx_extra_bytes, "Extra Bytes");
NFE_SYSCTL_STAT_ADD32(ctx, child, "late_cols",
&stats->rx_late_cols, "Late Collisions");
NFE_SYSCTL_STAT_ADD32(ctx, child, "runts",
&stats->rx_runts, "Runts");
NFE_SYSCTL_STAT_ADD32(ctx, child, "jumbos",
&stats->rx_jumbos, "Jumbos");
NFE_SYSCTL_STAT_ADD32(ctx, child, "fifo_overuns",
&stats->rx_fifo_overuns, "FIFO Overruns");
NFE_SYSCTL_STAT_ADD32(ctx, child, "crc_errors",
&stats->rx_crc_errors, "CRC Errors");
NFE_SYSCTL_STAT_ADD32(ctx, child, "fae",
&stats->rx_fae, "Frame Alignment Errors");
NFE_SYSCTL_STAT_ADD32(ctx, child, "len_errors",
&stats->rx_len_errors, "Length Errors");
NFE_SYSCTL_STAT_ADD32(ctx, child, "unicast",
&stats->rx_unicast, "Unicast Frames");
NFE_SYSCTL_STAT_ADD32(ctx, child, "multicast",
&stats->rx_multicast, "Multicast Frames");
Fix a typo. Submitted by: clemens fischer via Shigeaki Tagashira
2008-12-20 00:04:04 +00:00
NFE_SYSCTL_STAT_ADD32(ctx, child, "broadcast",
Add hardware MAC statistics support. Register information was obtained from Linux forcedeth driver. While I'm here move creating a sysctl node for process_limit to function nfe_sysctl_node(). Tested by: "Arno J. Klaassen" < arno <at> heho dot snv dot jussieu dot fr >
2008-10-03 03:58:16 +00:00
&stats->rx_broadcast, "Broadcast Frames");
if ((sc->nfe_flags & NFE_MIB_V2) != 0) {
NFE_SYSCTL_STAT_ADD64(ctx, child, "octets",
&stats->rx_octets, "Octets");
NFE_SYSCTL_STAT_ADD32(ctx, child, "pause",
&stats->rx_pause, "Pause frames");
NFE_SYSCTL_STAT_ADD32(ctx, child, "drops",
&stats->rx_drops, "Drop frames");
}
/* Tx statistics. */
tree = SYSCTL_ADD_NODE(ctx, parent, OID_AUTO, "tx", CTLFLAG_RD,
NULL, "Tx MAC statistics");
child = SYSCTL_CHILDREN(tree);
NFE_SYSCTL_STAT_ADD64(ctx, child, "octets",
&stats->tx_octets, "Octets");
NFE_SYSCTL_STAT_ADD32(ctx, child, "zero_rexmits",
&stats->tx_zero_rexmits, "Zero Retransmits");
NFE_SYSCTL_STAT_ADD32(ctx, child, "one_rexmits",
&stats->tx_one_rexmits, "One Retransmits");
NFE_SYSCTL_STAT_ADD32(ctx, child, "multi_rexmits",
&stats->tx_multi_rexmits, "Multiple Retransmits");
NFE_SYSCTL_STAT_ADD32(ctx, child, "late_cols",
&stats->tx_late_cols, "Late Collisions");
NFE_SYSCTL_STAT_ADD32(ctx, child, "fifo_underuns",
&stats->tx_fifo_underuns, "FIFO Underruns");
NFE_SYSCTL_STAT_ADD32(ctx, child, "carrier_losts",
&stats->tx_carrier_losts, "Carrier Losts");
NFE_SYSCTL_STAT_ADD32(ctx, child, "excess_deferrals",
&stats->tx_excess_deferals, "Excess Deferrals");
NFE_SYSCTL_STAT_ADD32(ctx, child, "retry_errors",
&stats->tx_retry_errors, "Retry Errors");
if ((sc->nfe_flags & NFE_MIB_V2) != 0) {
NFE_SYSCTL_STAT_ADD32(ctx, child, "deferrals",
&stats->tx_deferals, "Deferrals");
NFE_SYSCTL_STAT_ADD32(ctx, child, "frames",
&stats->tx_frames, "Frames");
NFE_SYSCTL_STAT_ADD32(ctx, child, "pause",
&stats->tx_pause, "Pause Frames");
}
if ((sc->nfe_flags & NFE_MIB_V3) != 0) {
NFE_SYSCTL_STAT_ADD32(ctx, child, "unicast",
&stats->tx_deferals, "Unicast Frames");
NFE_SYSCTL_STAT_ADD32(ctx, child, "multicast",
&stats->tx_frames, "Multicast Frames");
NFE_SYSCTL_STAT_ADD32(ctx, child, "broadcast",
&stats->tx_pause, "Broadcast Frames");
}
}
#undef NFE_SYSCTL_STAT_ADD32
#undef NFE_SYSCTL_STAT_ADD64
static void
nfe_stats_clear(struct nfe_softc *sc)
{
int i, mib_cnt;
if ((sc->nfe_flags & NFE_MIB_V1) != 0)
mib_cnt = NFE_NUM_MIB_STATV1;
else if ((sc->nfe_flags & (NFE_MIB_V2 | NFE_MIB_V3)) != 0)
mib_cnt = NFE_NUM_MIB_STATV2;
else
return;
for (i = 0; i < mib_cnt; i += sizeof(uint32_t))
NFE_READ(sc, NFE_TX_OCTET + i);
if ((sc->nfe_flags & NFE_MIB_V3) != 0) {
NFE_READ(sc, NFE_TX_UNICAST);
NFE_READ(sc, NFE_TX_MULTICAST);
NFE_READ(sc, NFE_TX_BROADCAST);
}
}
static void
nfe_stats_update(struct nfe_softc *sc)
{
struct nfe_hw_stats *stats;
NFE_LOCK_ASSERT(sc);
if ((sc->nfe_flags & (NFE_MIB_V1 | NFE_MIB_V2 | NFE_MIB_V3)) == 0)
return;
stats = &sc->nfe_stats;
stats->tx_octets += NFE_READ(sc, NFE_TX_OCTET);
stats->tx_zero_rexmits += NFE_READ(sc, NFE_TX_ZERO_REXMIT);
stats->tx_one_rexmits += NFE_READ(sc, NFE_TX_ONE_REXMIT);
stats->tx_multi_rexmits += NFE_READ(sc, NFE_TX_MULTI_REXMIT);
stats->tx_late_cols += NFE_READ(sc, NFE_TX_LATE_COL);
stats->tx_fifo_underuns += NFE_READ(sc, NFE_TX_FIFO_UNDERUN);
stats->tx_carrier_losts += NFE_READ(sc, NFE_TX_CARRIER_LOST);
stats->tx_excess_deferals += NFE_READ(sc, NFE_TX_EXCESS_DEFERRAL);
stats->tx_retry_errors += NFE_READ(sc, NFE_TX_RETRY_ERROR);
stats->rx_frame_errors += NFE_READ(sc, NFE_RX_FRAME_ERROR);
stats->rx_extra_bytes += NFE_READ(sc, NFE_RX_EXTRA_BYTES);
stats->rx_late_cols += NFE_READ(sc, NFE_RX_LATE_COL);
stats->rx_runts += NFE_READ(sc, NFE_RX_RUNT);
stats->rx_jumbos += NFE_READ(sc, NFE_RX_JUMBO);
stats->rx_fifo_overuns += NFE_READ(sc, NFE_RX_FIFO_OVERUN);
stats->rx_crc_errors += NFE_READ(sc, NFE_RX_CRC_ERROR);
stats->rx_fae += NFE_READ(sc, NFE_RX_FAE);
stats->rx_len_errors += NFE_READ(sc, NFE_RX_LEN_ERROR);
stats->rx_unicast += NFE_READ(sc, NFE_RX_UNICAST);
stats->rx_multicast += NFE_READ(sc, NFE_RX_MULTICAST);
stats->rx_broadcast += NFE_READ(sc, NFE_RX_BROADCAST);
if ((sc->nfe_flags & NFE_MIB_V2) != 0) {
stats->tx_deferals += NFE_READ(sc, NFE_TX_DEFERAL);
stats->tx_frames += NFE_READ(sc, NFE_TX_FRAME);
stats->rx_octets += NFE_READ(sc, NFE_RX_OCTET);
stats->tx_pause += NFE_READ(sc, NFE_TX_PAUSE);
stats->rx_pause += NFE_READ(sc, NFE_RX_PAUSE);
stats->rx_drops += NFE_READ(sc, NFE_RX_DROP);
}
if ((sc->nfe_flags & NFE_MIB_V3) != 0) {
stats->tx_unicast += NFE_READ(sc, NFE_TX_UNICAST);
stats->tx_multicast += NFE_READ(sc, NFE_TX_MULTICAST);
stats->rx_broadcast += NFE_READ(sc, NFE_TX_BROADCAST);
}
}
Add basic WOL support for MCP ethernet controllers. It seems the controller does not perform automatic switching from 1000Mbps link to 10/100Mbps link when WOL is activated. Implement establishing 10/100Mps link with auto-negotiation in driver. Link status change handler was modified to remove taskqueue based approach since driver now needs synchronous handling for link establishment. Submitted by: Yamagi Burmeister (lists <> yamagi.org ) (initial version) Tested by: Yamagi Burmeister (lists <> yamagi.org ) MFC after: 1 week
2010-11-11 18:08:50 +00:00
static void
nfe_set_linkspeed(struct nfe_softc *sc)
{
struct mii_softc *miisc;
struct mii_data *mii;
int aneg, i, phyno;
NFE_LOCK_ASSERT(sc);
mii = device_get_softc(sc->nfe_miibus);
mii_pollstat(mii);
aneg = 0;
if ((mii->mii_media_status & (IFM_ACTIVE | IFM_AVALID)) ==
(IFM_ACTIVE | IFM_AVALID)) {
switch IFM_SUBTYPE(mii->mii_media_active) {
case IFM_10_T:
case IFM_100_TX:
return;
case IFM_1000_T:
aneg++;
break;
default:
break;
}
}
- Remove attempts to implement setting of BMCR_LOOP/MIIF_NOLOOP (reporting IFM_LOOP based on BMCR_LOOP is left in place though as it might provide useful for debugging). For most mii(4) drivers it was unclear whether the PHYs driven by them actually support loopback or not. Moreover, typically loopback mode also needs to be activated on the MAC, which none of the Ethernet drivers using mii(4) implements. Given that loopback media has no real use (and obviously hardly had a chance to actually work) besides for driver development (which just loopback mode should be sufficient for though, i.e one doesn't necessary need support for loopback media) support for it is just dropped as both NetBSD and OpenBSD already did quite some time ago. - Let mii_phy_add_media() also announce the support of IFM_NONE. - Restructure the PHY entry points to use a structure of entry points instead of discrete function pointers, and extend this to include a "reset" entry point. Make sure any PHY-specific reset routine is always used, and provide one for lxtphy(4) which disables MII interrupts (as is done for a few other PHYs we have drivers for). This includes changing NIC drivers which previously just called the generic mii_phy_reset() to now actually call the PHY-specific reset routine, which might be crucial in some cases. While at it, the redundant checks in these NIC drivers for mii->mii_instance not being zero before calling the reset routines were removed because as soon as one PHY driver attaches mii->mii_instance is incremented and we hardly can end up in their media change callbacks etc if no PHY driver has attached as mii_attach() would have failed in that case and not attach a miibus(4) instance. Consequently, NIC drivers now no longer should call mii_phy_reset() directly, so it was removed from EXPORT_SYMS. - Add a mii_phy_dev_attach() as a companion helper to mii_phy_dev_probe(). The purpose of that function is to perform the common steps to attach a PHY driver instance and to hook it up to the miibus(4) instance and to optionally also handle the probing, addition and initialization of the supported media. So all a PHY driver without any special requirements has to do in its bus attach method is to call mii_phy_dev_attach() along with PHY-specific MIIF_* flags, a pointer to its PHY functions and the add_media set to one. All PHY drivers were updated to take advantage of mii_phy_dev_attach() as appropriate. Along with these changes the capability mask was added to the mii_softc structure so PHY drivers taking advantage of mii_phy_dev_attach() but still handling media on their own do not need to fiddle with the MII attach arguments anyway. - Keep track of the PHY offset in the mii_softc structure. This is done for compatibility with NetBSD/OpenBSD. - Keep track of the PHY's OUI, model and revision in the mii_softc structure. Several PHY drivers require this information also after attaching and previously had to wrap their own softc around mii_softc. NetBSD/OpenBSD also keep track of the model and revision on their mii_softc structure. All PHY drivers were updated to take advantage as appropriate. - Convert the mebers of the MII data structure to unsigned where appropriate. This is partly inspired by NetBSD/OpenBSD. - According to IEEE 802.3-2002 the bits actually have to be reversed when mapping an OUI to the MII ID registers. All PHY drivers and miidevs where changed as necessary. Actually this now again allows to largely share miidevs with NetBSD, which fixed this problem already 9 years ago. Consequently miidevs was synced as far as possible. - Add MIIF_NOMANPAUSE and mii_phy_flowstatus() calls to drivers that weren't explicitly converted to support flow control before. It's unclear whether flow control actually works with these but typically it should and their net behavior should be more correct with these changes in place than without if the MAC driver sets MIIF_DOPAUSE. Obtained from: NetBSD (partially) Reviewed by: yongari (earlier version), silence on arch@ and net@
2011-05-03 19:51:29 +00:00
miisc = LIST_FIRST(&mii->mii_phys);
phyno = miisc->mii_phy;
LIST_FOREACH(miisc, &mii->mii_phys, mii_list)
PHY_RESET(miisc);
Add basic WOL support for MCP ethernet controllers. It seems the controller does not perform automatic switching from 1000Mbps link to 10/100Mbps link when WOL is activated. Implement establishing 10/100Mps link with auto-negotiation in driver. Link status change handler was modified to remove taskqueue based approach since driver now needs synchronous handling for link establishment. Submitted by: Yamagi Burmeister (lists <> yamagi.org ) (initial version) Tested by: Yamagi Burmeister (lists <> yamagi.org ) MFC after: 1 week
2010-11-11 18:08:50 +00:00
nfe_miibus_writereg(sc->nfe_dev, phyno, MII_100T2CR, 0);
nfe_miibus_writereg(sc->nfe_dev, phyno,
MII_ANAR, ANAR_TX_FD | ANAR_TX | ANAR_10_FD | ANAR_10 | ANAR_CSMA);
nfe_miibus_writereg(sc->nfe_dev, phyno,
MII_BMCR, BMCR_RESET | BMCR_AUTOEN | BMCR_STARTNEG);
DELAY(1000);
if (aneg != 0) {
/*
* Poll link state until nfe(4) get a 10/100Mbps link.
*/
for (i = 0; i < MII_ANEGTICKS_GIGE; i++) {
mii_pollstat(mii);
if ((mii->mii_media_status & (IFM_ACTIVE | IFM_AVALID))
== (IFM_ACTIVE | IFM_AVALID)) {
switch (IFM_SUBTYPE(mii->mii_media_active)) {
case IFM_10_T:
case IFM_100_TX:
nfe_mac_config(sc, mii);
return;
default:
break;
}
}
NFE_UNLOCK(sc);
pause("nfelnk", hz);
NFE_LOCK(sc);
}
if (i == MII_ANEGTICKS_GIGE)
device_printf(sc->nfe_dev,
"establishing a link failed, WOL may not work!");
}
/*
* No link, force MAC to have 100Mbps, full-duplex link.
* This is the last resort and may/may not work.
*/
mii->mii_media_status = IFM_AVALID | IFM_ACTIVE;
mii->mii_media_active = IFM_ETHER | IFM_100_TX | IFM_FDX;
nfe_mac_config(sc, mii);
}
static void
nfe_set_wol(struct nfe_softc *sc)
{
struct ifnet *ifp;
uint32_t wolctl;
int pmc;
uint16_t pmstat;
NFE_LOCK_ASSERT(sc);
Do a sweep of the tree replacing calls to pci_find_extcap() with calls to pci_find_cap() instead.
2011-03-23 13:10:15 +00:00
if (pci_find_cap(sc->nfe_dev, PCIY_PMG, &pmc) != 0)
Add basic WOL support for MCP ethernet controllers. It seems the controller does not perform automatic switching from 1000Mbps link to 10/100Mbps link when WOL is activated. Implement establishing 10/100Mps link with auto-negotiation in driver. Link status change handler was modified to remove taskqueue based approach since driver now needs synchronous handling for link establishment. Submitted by: Yamagi Burmeister (lists <> yamagi.org ) (initial version) Tested by: Yamagi Burmeister (lists <> yamagi.org ) MFC after: 1 week
2010-11-11 18:08:50 +00:00
return;
ifp = sc->nfe_ifp;
if ((ifp->if_capenable & IFCAP_WOL_MAGIC) != 0)
wolctl = NFE_WOL_MAGIC;
else
wolctl = 0;
NFE_WRITE(sc, NFE_WOL_CTL, wolctl);
if ((ifp->if_capenable & IFCAP_WOL_MAGIC) != 0) {
nfe_set_linkspeed(sc);
if ((sc->nfe_flags & NFE_PWR_MGMT) != 0)
NFE_WRITE(sc, NFE_PWR2_CTL,
NFE_READ(sc, NFE_PWR2_CTL) & ~NFE_PWR2_GATE_CLOCKS);
/* Enable RX. */
NFE_WRITE(sc, NFE_RX_RING_ADDR_HI, 0);
NFE_WRITE(sc, NFE_RX_RING_ADDR_LO, 0);
NFE_WRITE(sc, NFE_RX_CTL, NFE_READ(sc, NFE_RX_CTL) |
NFE_RX_START);
}
/* Request PME if WOL is requested. */
pmstat = pci_read_config(sc->nfe_dev, pmc + PCIR_POWER_STATUS, 2);
pmstat &= ~(PCIM_PSTAT_PME | PCIM_PSTAT_PMEENABLE);
if ((ifp->if_capenable & IFCAP_WOL) != 0)
pmstat |= PCIM_PSTAT_PME | PCIM_PSTAT_PMEENABLE;
pci_write_config(sc->nfe_dev, pmc + PCIR_POWER_STATUS, pmstat, 2);
}
Reference in New Issue Copy Permalink