![[BACK]](/icons/back.gif){kind=icon}
Up to [cvs.NetBSD.org] / src / sys / dev / ic
Request diff between arbitrary revisions
Keyword substitution: kv
Default branch: MAIN
Fix a problem noted by Taylor...
touching if_flags is forbidden in this context (no IFNET_LOCK guaranteed)
sc_promisc should be cached when if_flags changes, not when
SIOCADDMULTI/SIOCDELMULTI runs
by caching if_flags in sc_if_flags via a ifflags_cb.
Remove unused struct eqos_softc member
Pull up following revision(s) (requested by riastradh in ticket #453): sys/dev/ic/dwc_eqos_var.h: revision 1.7 sys/dev/ic/dwc_eqos_var.h: revision 1.8 sys/dev/ic/dwc_eqos_var.h: revision 1.9 sys/dev/ic/dwc_eqos.c: revision 1.30 sys/dev/ic/dwc_eqos.c: revision 1.31 sys/dev/ic/dwc_eqos.c: revision 1.32 sys/dev/ic/dwc_eqos.c: revision 1.33 eqos(4): Wait for callout to halt and make sure it stays halted. eqos(4): Don't touch if_flags in tx path. Can't touch this without IFNET_LOCK. eqos(4): Fix locking around multicast filter updates. - Can't touch if_flags without IFNET_LOCK. - Can't take IFNET_LOCK in SIOCADDMULTI/SIOCDELMULTI path. Instead, cache IFF_PROMISC and IFF_ALLMULTI on if_init under a lock we can take in this path. XXX Is IFF_ALLMULTI relevant any more? Hasn't it been moved to ethercom flags? XXX Should not take sc_lock around if_init/stop -- IFNET_LOCK is enough. Should narrow scope of sc_lock to be just tick/mii/multi stuff. eqos(4): Fix multicast filter updates. 1. Don't touch the obsolete IFF_ALLMULTI. 2. Set ETHER_F_ALLMULTI if we're accepting all multicast addresses. 3. If any multicast entry range is not a single address, accept all multicast addresses.
Pull up following revision(s) (requested by msaitoh in ticket #446):
sys/dev/pci/if_eqos_pci.c: revision 1.3
sys/arch/i386/conf/GENERIC: revision 1.1251
sys/arch/i386/conf/GENERIC: revision 1.1252
sys/arch/amd64/conf/GENERIC: revision 1.607
sys/arch/amd64/conf/GENERIC: revision 1.608
sys/dev/ic/dwc_eqos.c: revision 1.20
sys/dev/ic/dwc_eqos.c: revision 1.21
share/man/man4/eqos.4: revision 1.2
sys/dev/ic/dwc_eqos.c: revision 1.22
sys/dev/ic/dwc_eqos_reg.h: revision 1.7
sys/dev/ic/dwc_eqos.c: revision 1.23
sys/dev/ic/dwc_eqos_reg.h: revision 1.8
sys/dev/ic/dwc_eqos.c: revision 1.24
sys/dev/ic/dwc_eqos.c: revision 1.25
sys/dev/ic/dwc_eqos.c: revision 1.26
sys/dev/ic/dwc_eqos.c: revision 1.27
sys/dev/ic/dwc_eqos_var.h: revision 1.5
sys/dev/ic/dwc_eqos.c: revision 1.28
sys/dev/ic/dwc_eqos_var.h: revision 1.6
sys/dev/ic/dwc_eqos.c: revision 1.29
sys/dev/ic/dwc_eqos.c: revision 1.18
sys/dev/ic/dwc_eqos.c: revision 1.19
sys/dev/pci/files.pci: revision 1.448
sys/dev/pci/if_eqos_pci.c: revision 1.1
sys/dev/pci/if_eqos_pci.c: revision 1.2
eqos(4): Fix definition of GMAC_MAC_HW_FEATURE1_RXFIFOSIZE.
eqos(4): Fix a bug that the MAC address is swapped.
Don't swap the MAC address in eqos_get_eaddr(). Other OSes except FreeBSD
(which was based on NetBSD's) don't swap it. With this change, my own
OnLogic Helix 330's MAC address becomes correct. The OUI is 84:8b:cd:4d.
It's owned by Logic Supply and they were acquired by OnLogic.
On Quartz64 with UEFI, the MAC address is wrongly set and the multicast
bit might be set. To do workaround, clear the bit if it's set.
eqos(4): Add missing clock range.
eqos(4): Accept if snpsver == 0x52. Tested with Intel Elkhart Lake.
TODO:
Multiqueue support.
Add watchdog timer.
Add detach function.
eqos(4): Add initial support for Intel Elkhart Lake internal Ethernet devices.
- Only tested on PSE SGMII 1G Ethernet MAC with MaxLinear GPY115.
- I don't know why dmat64 doesn't work. eqos_attach() have a special
code if EQOS_HW_FEATURE_ADDR64_32BIT(sc) is true, but it seems it doesn't
work.
- TODO:
Multiqueue support.
Detach support.
eqos(4): Fix compile error for arch that sizeof(bus_size_t) == 4 (i386).
Trailing whitespace
eqos(4): Disable eqos(4) by default because it's not stable on x86.
eqos(4): KNF. No functional change.
eqos(4): Add and modify some DPRINTF()s.
eqos(4): Add sysctls for debugging.
eqos(4): Use EQOS_TXLOCK() more to be stable.
Fix a bug that sc_tx.{cur,next,queued} become inconsitent.
Use txlock when accessing TX data.
eqos(4): Set TX/RX DMA burst length to improve performance.
eqos(4): Set flow control correctly.
eqos_pci: Limit to 32bit DMA only for PSE devices.
eqos(4): Fix typo in comment.
eqos(4): Fix locking around multicast filter updates. - Can't touch if_flags without IFNET_LOCK. - Can't take IFNET_LOCK in SIOCADDMULTI/SIOCDELMULTI path. Instead, cache IFF_PROMISC and IFF_ALLMULTI on if_init under a lock we can take in this path. XXX Is IFF_ALLMULTI relevant any more? Hasn't it been moved to ethercom flags? XXX Should not take sc_lock around if_init/stop -- IFNET_LOCK is enough. Should narrow scope of sc_lock to be just tick/mii/multi stuff.
eqos(4): Don't touch if_flags in tx path. Can't touch this without IFNET_LOCK.
eqos(4): Wait for callout to halt and make sure it stays halted.
eqos(4): Set TX/RX DMA burst length to improve performance.
eqos(4): Add sysctls for debugging.
jumboframe support for eqos(4) Tested up to mtu=9000. Hardware may allow up to mtu=16364 (frame length=16382), but this has not been tested. - Separate calls to eqos_setup_rxdesc() from eqos_setup_rxbuf(). if m_getcl() fails, discard the packets received at that time and reuse for the next buffer. - Restore m_adj(m, ETHER_ALIGN) with limited conditions. Only if MCLBYTES is greater than 2050, so it is not normally aligned.
eqos: handle the GMAC_MTL_INTERRUPT_STATUS register having something drain a couple of registers that want either a read or a write-1-to- clear bit, and keep track of how many happen via evcnt. i had this trigger one time, but not since adding instrumentation to see exactly it was saying (the GMAC_MTL_INTERRUPT_STATUS_Q0IS bit was set, and it requires some handling now implemented.) ok jmcneill
eqos: interrupt evcnt and minor fix there's basic interrupt evcnt, which is the parent for the rx and tx interrupts, and 3 status interrupts, one of which has 7 more subtypes (watchdog timeout, carrier missing/lost, etc.) as these cases are evcnt counted now, make some debugging default off. avoid removing bits from dma_status when rx/tx is handled, so that later check of dma_status for non-zero does not trip. the two bits in dma_status removed are never checked again besides the test that may have failed (but probably doesn't as at least one other bit in two other variables will be set.) ok jmcneill
Add driver for DesignWare Ethernet Quality-of-Service controller.