o Header file cleanup.
o bus_dma(9) conversion.
- Removed all consumers of vtophys(9) and converted to use
bus_dma(9).
- 64bit DMA support was disabled because DP83821 is not capable
of handling the DMA request. 64bit DMA request on DP83820
requires different descriptor structures and it's hard to
dynamically change descriptor format at run time so I disabled
it. Note, this is the same behavior as previous one but
previously nge(4) didn't explicitly disable 64bit mode on
DP83820.
- Added Tx/Rx descriptor ring alignment requirements(8 bytes
alignment).
- Limit maximum number of Tx DMA segments to 16. In fact,
controller does not seem to have limitations on number of Tx
DMA segments but 16 should be enough for most cases and
m_collapse(9) will handle highly fragmented frames without
consuming a lot of CPU cycles.
- Added Rx buffer alignment requirements(8 bytes alignment). This
means driver should fixup received frames to align on 16bits
boundary on strict-alignment architectures.
- Nuked driver private data structure in descriptor ring.
- Added endianness support code in Tx/Rx descriptor access.
o Prefer faster memory mapped register access to I/O mapped access.
Added fall-back mechanism to use alternative register access.
The hardware supports both memory and I/O mapped access.
o Added suspend/resume methods but it wasn't tested as controller I
have does not support PCI PME.
o Removed swap argument in nge_read_eeprom() since endianness
should be handled after reading EEPROM.
o Implemented experimental 802.3x full-duplex flow-control. ATM
it was commented out but will be activated after we have generic
flow-control framework in mii(4) layer.
o Rearranged promiscuous mode settings and simplified logic.
o Always disable Rx filter prior to changing Rx filter functions as
indicated in DP83820/DP83821 datasheet.
o Added an explicit DELAY in timeout loop of nge_reset().
o Added a sysctl variable dev.nge.%d.int_holdoff to control
interrupt moderation. Valid ranges are 1 to 255(default 1) in
units of 100us. The actual delivery of interrupt would be delayed
based on the sysctl value. The interface has to be brought down
and up again before a change takes effect. With proper tuning
value, users do not need to resort to polling(4) anymore.
o Added ALTQ(4) support.
o Added missing IFCAP_VLAN_HWCSUM as nge(4) can offload Tx/Rx
checksum calculation on VLAN tagged frames as well as VLAN tag
insertion/stripping. Also add IFCAP_VLAN_MTU capability as nge(4)
can handle VLAN tagged oversized frames.
o Fixed media header length for VLAN.
o Rearranged nge_detach routine such that it's now used for general
clean-up routine.
o Enabled MWI.
o Accessing EEPROM takes very long time so read 6 bytes ethernet
address with one call instead of 3 separate accesses.
o Don't set if_mtu in device attach, it's already set in
ether_ifattach().
o Don't do any special things for TBI interface. Remove TBI
specific media handling in the driver and have gentbi(4) handle
it. Add glue code to read/write TBI PHY registers in miibus
method. This change removes a lot of PHY handling code in driver
and now its functionality is handled by mii(4).
o Alignment fixup code is now applied only for strict-alignment
architectures. Previously the code was applied for all
architectures except i386. With this change amd64 will get
instant Rx performance boost.
o When driver fails to allocate a new mbuf, update if_qdrops so
users can see what was wrong in Rx path.
o Added a workaround for a hardware bug which resulted in short
VLAN tagged frames(e.g. ARP) was rejected as if runt frame was
received. With this workaround nge(4) now accepts the short VLAN
tagged frame and nge(4) can take full advantage of hardware VLAN
tag stripping. I have no idea how this bug wasn't known so far,
without the workaround nge(4) may never work on VLAN
environments.
o Fixed Rx checksum offload logic such that it now honors active
interface capability configured with ifconfig(8).
o In nge_start()/nge_txencap(), always leave at least one free
descriptor as indicated in datasheet. Without this the hardware
would be confused with ring descriptor structure(e.g. no clue
for the end of descriptor ring).
o Removed dead-code that checks interrupts on PHY hardware. The
code was designed to detect link state changes but it was
disabled as driving nge_tick clock would break auto-negotiation
timer. This code is no longer needed as nge(4) now uses mii(4)
and link state change handling is done with mii callback.
o Rearranged ethernet address programming logic such that it works
on strict-alignment architectures.
o Added IFCAP_VLAN_HWTAGGING/IFCAP_VLAN_HWCSUM handler in
nge_ioctl() such that the functionality is configurable with
ifconfig(8). DP83820/DP83821 can do checksum offload for VLAN
tagged frames so enable Tx/Rx checksum offload for VLAN
interfaces.
o Simplified IFCAP_POLLING selection logic in nge_ioctl().
o Fixed module unload panic when bpf listeners are active.
o Tx/Rx descriptor ring address uses 64bit DMA address for
readability. High address part of DMA would be 0 as nge(4)
disabled 64bit DMA transfers so it's ok for DP83821.
o Removed volatile keyword in softc as bus_dmamap_sync(9) should
take care of this.
o Removed extra driver private structures in descriptor ring. These
extra elements are not part of descriptor structure. Embedding
private driver structure into descriptor ring is not good idea
as its size may be different on 32bit/64bit architectures.
o Added miibus_linkchg method handler to catch link state changes.
o Removed unneeded nge_ifmedia in softc. All TBI access is handled
in gentbi(4). There is no difference between TBI and non-TBI case
now.
o Removed "gigabit link up" message handling in nge_tick. Link
state change notification is already performed by mii(4) and
checking link state by accessing PHY registers in periodic timer
handler of driver is wrong. All link state and speed/duplex
monitoring should be handled in PHY driver.
o Use our own timer for watchdog instead of if_watchdog/if_timer
interface.
o Added hardware MAC statistics counter, users canget current MAC
statistics from dev.nge.%d.stats sysctl node(%d is unit number of
a device).
o Removed unused macros, NGE_LASTDESC, NGE_MODE, NGE_OWNDESC,
NGE_RXBYTES.
o Increased number of Tx/Rx descriptors from 128 to 256. From my
experience on gigabit ethernet controllers, number of descriptors
should be 256 or higher to get an optimal performance on gigabit
link.
o Increased jumbo frame length to 9022 bytes to cope with other
gigabit ethernet drivers. Experimentation shows no problems with
9022 bytes.
o Removed unused member variables in softc.
o Switched from bus_space_{read|write}_4 to bus_{read|write}_4.
o Added support for WOL.
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
m_pkthdr.ether_vlan. The presence of the M_VLANTAG flag on the mbuf
signifies the presence and validity of its content.
Drivers that support hardware VLAN tag stripping fill in the received
VLAN tag (containing both vlan and priority information) into the
ether_vtag mbuf packet header field:
m->m_pkthdr.ether_vtag = vlan_id; /* ntohs()? */
m->m_flags |= M_VLANTAG;
to mark the packet m with the specified VLAN tag.
On output the driver should check the mbuf for the M_VLANTAG flag to
see if a VLAN tag is present and valid:
if (m->m_flags & M_VLANTAG) {
... = m->m_pkthdr.ether_vtag; /* htons()? */
... pass tag to hardware ...
}
VLAN tags are stored in host byte order. Byte swapping may be necessary.
(Note: This driver conversion was mechanic and did not add or remove any
byte swapping in the drivers.)
Remove zone_mtag_vlan UMA zone and MTAG_VLAN definition. No more tag
memory allocation have to be done.
Reviewed by: thompsa, yar
Sponsored by: TCP/IP Optimization Fundraise 2005
if_watchdog, etc., or in functions used only in these methods.
In all other functions in the driver use device_printf().
- Use __func__ instead of typing function name.
Submitted by: Alex Lyashkov <umka sevcity.net>
case if memory allocation failed.
- Remove fourth argument from VLAN_INPUT_TAG(), that was used
incorrectly in almost all drivers. Indicate failure with
mbuf value of NULL.
In collaboration with: yongari, ru, sam
rather than in ifindex_table[]; all (except one) accesses are
through ifp anyway. IF_LLADDR() works faster, and all (except
one) ifaddr_byindex() users were converted to use ifp->if_addr.
- Stop storing a (pointer to) Ethernet address in "struct arpcom",
and drop the IFP2ENADDR() macro; all users have been converted
to use IF_LLADDR() instead.
- Use device_printf() and if_printf() and remove nge_unit.
- Use callout_init_mtx() and remove nge_tick_locked() as nge_tick() is now
always called with the driver lock held.
- Use M_ZERO to contigmalloc() when allocating nge_ldata. It was possible
for the random garbage to be used in certain cases otherwise.
- Cleanup attach error handling including no longer leaking nge_ldata.
- Add locking to the ifmedia callouts.
- Lock accesses to if_hwassist and if_capenable in nge_ioctl().
Submitted by: Yuriy N. Shkandybin jura at networks dot ru (1, 3, 4)
Tested by: Yuriy N. Shkandybin jura at networks dot ru
MFC after: 3 days
opt_device_polling.h
- Include opt_device_polling.h into appropriate files.
- Embrace with HAVE_KERNEL_OPTION_HEADERS the include in the files that
can be compiled as loadable modules.
Reviewed by: bde
o Axe poll in trap.
o Axe IFF_POLLING flag from if_flags.
o Rework revision 1.21 (Giant removal), in such a way that
poll_mtx is not dropped during call to polling handler.
This fixes problem with idle polling.
o Make registration and deregistration from polling in a
functional way, insted of next tick/interrupt.
o Obsolete kern.polling.enable. Polling is turned on/off
with ifconfig.
Detailed kern_poll.c changes:
- Remove polling handler flags, introduced in 1.21. The are not
needed now.
- Forget and do not check if_flags, if_capenable and if_drv_flags.
- Call all registered polling handlers unconditionally.
- Do not drop poll_mtx, when entering polling handlers.
- In ether_poll() NET_LOCK_GIANT prior to locking poll_mtx.
- In netisr_poll() axe the block, where polling code asks drivers
to unregister.
- In netisr_poll() and ether_poll() do polling always, if any
handlers are present.
- In ether_poll_[de]register() remove a lot of error hiding code. Assert
that arguments are correct, instead.
- In ether_poll_[de]register() use standard return values in case of
error or success.
- Introduce poll_switch() that is a sysctl handler for kern.polling.enable.
poll_switch() goes through interface list and enabled/disables polling.
A message that kern.polling.enable is deprecated is printed.
Detailed driver changes:
- On attach driver announces IFCAP_POLLING in if_capabilities, but
not in if_capenable.
- On detach driver calls ether_poll_deregister() if polling is enabled.
- In polling handler driver obtains its lock and checks IFF_DRV_RUNNING
flag. If there is no, then unlocks and returns.
- In ioctl handler driver checks for IFCAP_POLLING flag requested to
be set or cleared. Driver first calls ether_poll_[de]register(), then
obtains driver lock and [dis/en]ables interrupts.
- In interrupt handler driver checks IFCAP_POLLING flag in if_capenable.
If present, then returns.This is important to protect from spurious
interrupts.
Reviewed by: ru, sam, jhb
could get an interrupt after we free the ifp, and the interrupt
handler depended on the ifp being still alive, this could, in theory,
cause a crash. Eliminate this possibility by moving the if_free to
after the bus_teardown_intr() call.
IFF_DRV_RUNNING, as well as the move from ifnet.if_flags to
ifnet.if_drv_flags. Device drivers are now responsible for
synchronizing access to these flags, as they are in if_drv_flags. This
helps prevent races between the network stack and device driver in
maintaining the interface flags field.
Many __FreeBSD__ and __FreeBSD_version checks maintained and continued;
some less so.
Reviewed by: pjd, bz
MFC after: 7 days
over iteration of their multicast address lists when synchronizing the
hardware address filter with the network stack-maintained list.
Problem reported by: Ed Maste (emaste at phaedrus dot sandvine dot ca>
MFC after: 1 week
struct ifnet or the layer 2 common structure it was embedded in have
been replaced with a struct ifnet pointer to be filled by a call to the
new function, if_alloc(). The layer 2 common structure is also allocated
via if_alloc() based on the interface type. It is hung off the new
struct ifnet member, if_l2com.
This change removes the size of these structures from the kernel ABI and
will allow us to better manage them as interfaces come and go.
Other changes of note:
- Struct arpcom is no longer referenced in normal interface code.
Instead the Ethernet address is accessed via the IFP2ENADDR() macro.
To enforce this ac_enaddr has been renamed to _ac_enaddr.
- The second argument to ether_ifattach is now always the mac address
from driver private storage rather than sometimes being ac_enaddr.
Reviewed by: sobomax, sam
problems:
1) Add locking for SMP, code provided by Alan Cox
2) While testing Alan's patches, I observed serious problems with
the jumbo buffer allocation code (machine crashed twice), so I gutted
it and rewrote the receive handler to use multiple chained descriptors.
Each RX descriptor gets a single 2K cluster, and the chip will fill in
as many as it needs to hold the complete packet.
User reports that this corrects the data corruption issues previously
observed and discussed on -current.
Note that this driver still needs to be hit with the busdma stick.
I intend to inflict said beating in the near future.
MFC after: 1 week
the interface as IFF_NEEDSGIANT so if_start is run holding Giant.
Note: this driver does declare and occasionally reference mutexes,
but I believe not nearly enough to provide safety.
