yongari [Thu, 14 Jan 2010 22:26:52 +0000 (22:26 +0000)]
MFC r200950,200955,200965-200966,201767-201768
r200950:
Implement RX interrupt moderation using one-shot timer interrupt.
Unlike TX interrupt, ST201 does not provide any mechanism to
suppress RX interrupts. ste(4) can generate more than 70k RX
interrupts under heavy RX traffics such that these excessive
interrupts make system useless to process other useful things.
Maybe this was the major reason why polling support code was
introduced to ste(4).
The STE_COUNTDOWN register provides a programmable counter that
will generate an interrupt upon its expiration. We program
STE_DMACTL register to use 3.2us clock rate to drive the counter
register. Whenever ste(4) serves RX interrupt, the driver rearm
the timer to expire after STE_IM_RX_TIMER_DEFAULT time and disables
further generation of RX interrupts. This trick seems to work well
and ste(4) generates less than 8k RX interrupts even under 64 bytes
UDP torture test. Combined with TX interrupts, the total number of
interrupts are less than 10k which looks reasonable on heavily
loaded controller.
The default RX interrupt moderation time is 150us. Users can change
the value at any time with dev.ste.%d.int_rx_mod sysctl node.
Setting it 0 effectively disables the RX interrupt moderation
feature. Now we have both TX/RX interrupt moderation code so remove
loop of interrupt handler which resulted in sub-optimal performance
as well as more register accesses.
r200955:
Add suspend/resume support as well as basic WOL.
While I'm here simplify SIOCSIFCAP handler.
r200965:
Update if_iqdrops in case of RX buffer allocation failure.
r201767:
Fix EEPROM access code to return data in host byte order.
EEPROM on ST201 always returns 16bits data with little endian
format so conversion to host order is required.
This change fixes inversed ethernet address on sparc64.
r201768:
Make sure to store dma address of RX buffer in little endian form.
This fixes the last bug which keeps ste(4) from working on sparc64.
yongari [Thu, 14 Jan 2010 22:15:51 +0000 (22:15 +0000)]
MFC r200904-200908,200910-200913
r200904:
Don't reinitialize controller if driver is already running. This
reduces number of link state UP/DOWN changes.
r200905:
Reimplement controller reset. Datasheet says full reset takes about
1ms. Since we switched to memory register mapping make sure to
flush PCI posted write by reading the register again.
While I'm here add additional delays in loop while driver waits the
completion of the reset.
r200906:
Overhaul RX filter programming.
o Let RX filter handler program promiscuous/multicast filter as
well as broadcasting.
o Remove unnecessary register access.
o Simplify ioctl handler and have set_rxfilter to handle
IFF_PROMISC and IFF_ALLMULTI change instead of directly
programming the controller.
o Removed unnecessary error variable reinitialization in ioctl
handler.
o Add IFF_DRV_RUNNING check before programming multicast filter.
o Configure maximum allowed frame length before enabling MAC.
Datasheet didn't say the exact ordering of programming sequence
but it looks more natural to set maximum allowed frame length
first prior to enabling controller.
r200907:
Don't report link status if driver is not running.
r200908:
Report the correct result of mii_mediachg(). Previously it always
used to return success without respect to the result.
While I'm here use mii_mediachg() in ste_init_locked which allows
driver to use currently configured media. ste_ifmedia_upd() is
supposed to be called whenever user changes current media settings.
r200910:
Implement hardware MAC statistics counter support. The counters
could be accessed with dev.ste.0.stats sysctl node.
r200911:
Remove unused duplicated register definition. It seems the
definition was made to access STE_ASICCTL register as 16bits but
ste(4) always access the register as 32bits so it was never used
before.
r200912:
Correct STE_COUNTDOWN register offset. The datasheet was wrong.
r200913:
We don't need to generate DMA complete interrupt for every
transmitted frames. So request interrupt for every 16th frames. Due
to the limitation of hardware we can't suppress the interrupt as
driver should have to check TX status register. The TX status
register can store up to 31 TX status so driver can't send more
than 31 frames without reading TX status register.
With this change controller would not generate TX completion
interrupt for every frame, so reclaim transmitted frames in
ste_tick().
r200854:
Add minimal dealy while ste(4) is waiting for the end of active DMA
cycle.
r200856:
Introduce sc_flags member variable and use it to keep track of
link state and PHY related information.
Remove ste_link and ste_one_phy variable of softc as it's not used
anymore.
While I'm here add IFF_DRV_RUNNING check in ste_start_locked().
r200865:
Reimplement miibus_statchg method. Don't rely on link state change
interrupt. If we want to use link state change interrupt ste(4)
should also implement auto-negotiation complete handler as well as
various PHY access handling. Now link state change is handled by
mii(4) polling so it will automatically update link state UP/DOWN
events which in turn make ste(4) usable with lagg(4).
r199559 added a private timer to drive watchdog and the timer also
used to drive MAC statistics update. Because the MAC statistics
update is called whenever statistics counter reaches near-full, it
drove watchdog timer too fast such that it caused false watchdog
timeouts under heavy TX traffic conditions.
Fix the regression by separating ste_stats_update() from driving
watchdog timer and introduce a new function ste_tick() that handles
periodic job such as driving watchdog, MAC statistics update and
link state check etc.
While I'm here clear armed watchdog timer in ste_stop().
r200873:
Instead of relying on hard resetting of controller to stop
receiving incoming traffics, try harder to gracefully stop active
DMA cycles and then stop MACs. This is the way what datasheet
recommends and seems to work reliably. Resetting controller while
active DMAs are in progress is bad thing as we can't predict how
DMAs touche allocated TX/RX buffers. This change ensures controller
stop state before attempting to release allocated TX/RX buffers.
Also update MAC statistics which could have been updated during the
wait time of MAC stop.
While I'm here remove unnecessary controller resets in various
location. ste(4) no longer relies on hard controller reset to stop
controller and resetting controller also clears all configured
settings which makes it hard to implement WOL in near future.
Now resetting a controller is performed in ste_init_locked().
r200875:
Prefer memory space register mapping over io space. If memory space
mapping fails fall back to old io space mapping.
While I'm here use PCIR_BAR macro.
r200877:
Prefer bus_write_{1,2,4}/bus_read_{1,2,4} to
bus_space_write_{1,2,4}/bus_space_read_{1,2,4}.
Remove unused ste_bhandle and ste_btag in softc.
r200884:
Reimplement Tx status error handler as recommended by datasheet.
If ste(4) encounter TX underrun or excessive collisions the TX MAC
of controller is stalled so driver should wake it up again. TX
underrun requires increasing TX threshold value to minimize
further TX underruns. Previously ste(4) used to reset controller
to recover from TX underrun, excessive collision and reclaiming
error. However datasheet says only TX underrun requires resetting
entire controller. So implement ste_restart_tx() that restarts TX
MAC and do not perform full reset except TX underrun case.
Now ste(4) uses CSR_READ_2 instead of CSR_READ_1 to read
STE_TX_STATUS register. This way ste(4) will also read frame id
value and we can write the same value back to STE_TX_FRAMEID
register instead of overwriting it to 0. The datasheet was wrong
in write back of STE_TX_STATUS so add some comments why we do so.
Also always invoke ste_txeoc() after ste_txeof() in ste_poll as
without reading TX status register can stall TX MAC.
yongari [Thu, 14 Jan 2010 21:45:39 +0000 (21:45 +0000)]
MFC r200853:
Add bus_dma(9) and endianness support to ste(4).
o Sorted includes and added missing header files.
o Added basic endianness support. In theory ste(4) should work on
any architectures.
o Remove the use of contigmalloc(9), contigfree(9) and vtophys(9).
o Added 8 byte alignment limitation of TX/RX descriptor.
o Added 1 byte alignment requirement for TX/RX buffers.
o ste(4) controllers does not support DAC. Limit DMA address space
to be within 32bit address.
o Added spare DMA map to gracefully recover from DMA map failure.
o Removed dead code for checking STE_RXSTAT_DMADONE bit. The bit
was already checked in each iteration of loop so it can't be true.
o Added second argument count to ste_rxeof(). It is used to limit
number of iterations done in RX handler. ATM polling is the only
consumer.
o Removed ste_rxeoc() which was added to address RX stuck issue
(cvs rev 1.66). Unlike TX descriptors, ST201 supports chaining
descriptors to form a ring for RX descriptors. If RX descriptor
chaining is not supported it's possible for controller to stop
receiving incoming frames once controller pass the end of RX
descriptor which in turn requires driver post new RX
descriptors to receive more frames. For TX descriptors which
does not support chaning, we exactly do manual chaining in
driver by concatenating new descriptors to the end of previous
TX chain.
Maybe the workaround was borrowed from other drivers that does
not support RX descriptor chaining, which is not valid for ST201
controllers. I still have no idea how this address RX stuck
issue and I can't reproduce the RX stuck issue on DFE-550TX
controller.
o Removed hw.ste_rxsyncs sysctl as the workaround was removed.
o TX/RX side bus_dmamap_load_mbuf_sg(9) support.
o Reimplemented optimized ste_encap().
o Simplified TX logic of ste_start_locked().
o Added comments for TFD/RFD requirements.
o Increased number of RX descriptors to 128 from 64. 128 gave much
better performance than 64 under high network loads.
yongari [Thu, 14 Jan 2010 20:38:40 +0000 (20:38 +0000)]
Partial merge r199559:
- Add a private timer to drive the transmit watchdog instead of using
if_watchdog and if_timer.
- Fix some issues in detach for sn(4), ste(4), and ti(4). Primarily this
means calling ether_ifdetach() before anything else.
rmacklem [Thu, 14 Jan 2010 17:35:07 +0000 (17:35 +0000)]
MFC: r201345
Fix the experimental NFS client so that it can create Unix
domain sockets on an NFSv4 mount point. It was generating
incorrect XDR in the request for this case.
marius [Wed, 13 Jan 2010 21:23:27 +0000 (21:23 +0000)]
MFC: r200938
- Don't check for a valid interrupt controller on every interrupt
in intr_execute_handlers(). If we managed to get here without an
associated interrupt controller we have way bigger problems.
While at it predict stray vector interrupts as false as they are
rather unlikely.
- Don't blindly call the clear function of an interrupt controller
when adding a handler in inthand_add() as interrupt controllers
like the one driven by upa(4) are auto-clearing and thus provide
NULL instead.
marius [Wed, 13 Jan 2010 21:19:46 +0000 (21:19 +0000)]
MFC: r200925
- By re-arranging the code in OF_decode_addr() somewhat and accepting
a bit of a detour we can just iterate through the banks array instead
of having to calculate every offset. This change is inspired by the
powerpc version of this function.
- Add support for the JBus to EBus bridges which hang off of nexus(4).
marius [Wed, 13 Jan 2010 21:16:07 +0000 (21:16 +0000)]
MFC: r200923
- Add support for the IOMMUs of Fire JBus to PCIe and Oberon Uranus
to PCIe bridges.
- Add support for talking the PROM mappings over to the kernel IOTSB
just like we do with the kernel TSB in order to allow OFW drivers
to continue to work.
- Change some members, parameters and variables to unsigned where
more appropriate.
marius [Wed, 13 Jan 2010 21:12:27 +0000 (21:12 +0000)]
MFC: r200921
- Add quirk handling for ALi M5229, mainly setting the magic "force
enable IDE I/O" bit which prevents data access traps with revision
0xc8 in Fire-based machines when pci(4) enables PCIM_CMD_PORTEN.
- Like for sun4v also don't add the PCI side of host-PCIe bridges to
the bus on sun4u as they don't have configuration space implement
there either.
marius [Wed, 13 Jan 2010 21:03:04 +0000 (21:03 +0000)]
MFC: r200917
Hook ebus(4) and isa(4) up to the sun4v LINT build in order to
ensure that their compilation doesn't break as they are expected
to work as-is now (but aren't actually run-time tested).
marius [Wed, 13 Jan 2010 20:51:21 +0000 (20:51 +0000)]
MFC: r200915
Don't probe the bq4802 variant found in Ultra 25 and 45 for now as
this chip isn't MC146818 compatible and requires different handlers
(but which I can't test due to lack of such hardware).
marius [Wed, 13 Jan 2010 20:40:49 +0000 (20:40 +0000)]
MFC: r200914
Don't use an out register to hold the vector number across the call
of the interrupt handler in intr_fast() as the handler might clobber
it (no in-tree handler currently does but an upcoming one will).
While at it, tidy the register usage in the interrupt counting code.
marius [Wed, 13 Jan 2010 20:32:54 +0000 (20:32 +0000)]
MFC: r200879
- Add support for the JBus to EBus bridges which hang off of nexus(4)
and are found in sun4u and sun4v machines based on the Fire ASIC.
- Initialize the configuration space of the PCI to EBus variant the
same way as OpenSolaris does.
marius [Wed, 13 Jan 2010 20:07:45 +0000 (20:07 +0000)]
MFC: r200878
- Add macros for the states of the interrupt clear registers.
- Change INTMAP_VEC() to take an INO as its second argument rather
than an INR. The former is what I actually intended with this
macro and how it's currently used.
marius [Wed, 13 Jan 2010 20:03:24 +0000 (20:03 +0000)]
MFC: r200874
Enroll these drivers in multipass probing. The motivation behind this
is that the JBus to EBus bridges share the interrupt controller of a
sibling JBus to PCIe bridge (at least as far as the OFW device tree
is concerned, in reality they are part of the same chip) so we have to
probe and attach the latter first. That happens to be also the case
due to the fact that the JBus to PCIe bridges appear first in the OFW
device tree but it doesn't hurt to ensure the right order.
marius [Wed, 13 Jan 2010 19:55:51 +0000 (19:55 +0000)]
MFC: 197164
Factor out the duplicated macro for the device type used in the
OFW device tree for PCI bridges and add a new one for PCI Express.
While at it, take advantage of the former for the rman(9) work-
around in jbusppm(4).
jhb [Wed, 13 Jan 2010 18:12:21 +0000 (18:12 +0000)]
MFC 199607, 200797, 201270, 201669:
Use pthread_once() to initialize the thread-local storage for localtime()
and gmtime() and _once() to initialize gmt state rather than home-rolled
versions using pthread mutex locks.
jhb [Tue, 12 Jan 2010 19:55:07 +0000 (19:55 +0000)]
MFC 201215:
Add support for configuring vlan(4) interfaces as child devices similar to
wlan(4) interfaces. vlan(4) interfaces are listed via a new 'vlans_<IF>'
variable. If a vlan interface is a number, then that number is treated as
the vlan tag for the interface and the interface will be named '<IF>.<tag>'.
Otherwise, the vlan tag must be provided via a vlan parameter in a
'create_args_<vlan>' variable.
While I'm here, fix a few nits in rc.conf(5) and mention create_args_<IF> in
the description of cloned_interfaces.
brooks [Tue, 12 Jan 2010 06:11:36 +0000 (06:11 +0000)]
MFC r201954:
Update the comment on printing group membership to reflect that fact
that each group the process is a member of is printed rather than
an entry for each group the user could be a member of.
qingli [Tue, 12 Jan 2010 00:04:13 +0000 (00:04 +0000)]
MFC r201544
An existing incomplete ARP entry would expire a subsequent
statically configured entry of the same host. This bug was
due to the expiration timer was not cancelled when installing
the static entry. Since there exist a potential race condition
with respect to timer cancellation, simply check for the
LLE_STATIC bit inside the expiration function instead of
cancelling the active timer.
gallatin [Mon, 11 Jan 2010 20:32:51 +0000 (20:32 +0000)]
Sync mxge(4) with head:
r197391: Add support for TX throttling
r198250: Move mxge(4)'s NIC watchdog reset handler from
a callout to a taskqueue
r198303: Make mxge do a better job recovering from NIC h/w faults
r200845: Don't take the driver mutex in mxge_tick()
r201758: Remove extraneous semicolons
rmacklem [Mon, 11 Jan 2010 19:30:23 +0000 (19:30 +0000)]
MFC: r201029
When porting the experimental nfs subsystem to the FreeBSD8 krpc,
I added 3 functions that were already in the experimental client
under different names. This patch deletes the functions in the
experimental client and renames the calls to use the other set.
(This is just removal of duplicated code and does not fix any bug.)
jhb [Mon, 11 Jan 2010 18:42:29 +0000 (18:42 +0000)]
MFC 201522:
Expose the upper 256 ptys in the default devfs rules. I should have updated
this when expanding the old pty(4) driver to use 512 ptys by default. This
is more important for 7.x.
delphij [Mon, 11 Jan 2010 02:31:00 +0000 (02:31 +0000)]
MFC r201143:
Apply OpenSolaris revision 8021:b8fe9660eb2d which brings our zpool
to version 14, making it possible for zpools created on OpenSolaris
2009.06 be used on FreeBSD.
jilles [Sun, 10 Jan 2010 16:58:12 +0000 (16:58 +0000)]
MFC r210428: sh: Add a regression test that tries out all arithmetic ops.
The gaps in the numbering sequence cannot really be avoided, as HEAD changes
are not necessarily merged in the same order as they are made (if they are
merged at all).
fabient [Sun, 10 Jan 2010 16:04:32 +0000 (16:04 +0000)]
MFC 201023:
* Support the L1D_CACHE_LD event on Core2 processors.
* Correct a group of typos: for Core2 programmable events, check
user supplied umask values against the correct event descriptor
field.
yongari [Sat, 9 Jan 2010 01:17:31 +0000 (01:17 +0000)]
MFC r200693:
Make sure to enable Next Page bit for IP1001. Otherwise the PHY
fails to re-establishe 1000baseT link after downgrading to
10/100Mbps link.
yongari [Sat, 9 Jan 2010 00:36:28 +0000 (00:36 +0000)]
MFC r200641,200676
r200641:
Document newly added loader tunable and sysctl variables.
o hw.vge.msi_disable
o dev.vge.%d.int_holdoff
o dev.vge.%d.rx_coal_pkt
o dev.vge.%d.tx_coal_pkt
r200676:
Document more VIA Velocity family controllers to vge(4). Previously
it mentioned only VT6122. While I'm here remove the mention of
VT3119 which seems to be VIA's internal model name and VT3119
wouldn't be available to end users.
yongari [Sat, 9 Jan 2010 00:26:57 +0000 (00:26 +0000)]
MFC r200696,200740,200756,200758-200759,200972
r200696:
Add rudimentary WOL support. While I'm here remove enabling
busmastering/memory address in resume path. Bus driver will handle
that.
r200740:
Swap VGE_TXQTIMER and VGE_RXQTIMER register definition. Pending
timer for Tx queue is at 0x3E.
r200756:
Correct fragment bit definition in comments.
r200758:
VT6130 datasheet was wrong. If VT6130 receive a jumbo frame the
controller will split the jumbo frame into multiple RX buffers.
However it seems the hardware always dma the frame to 8 bytes
boundary for the split frames. Only the first part of the fragment
can have 4 byte alignment and subsequent buffers should be 8 bytes
aligned. Change RX buffer the alignment requirement to 8 bytes from
4 bytes.
r200759:
Disable jumbo frame support for PCIe VT6130/VT6132 controllers.
Quite contrary to VT6130 datasheet which says it supports up to 8K
jumbo frame, VT6130 does not seem to send jumbo frame that is
larger than 4K in length. Trying to send a frame that is larger
than 4K cause TX MAC hang.
Even though it's possible to allow 4K jumbo frame for VT6130, I
think it's meaningless to allow 4K jumbo frame. I'm not sure VT6132
also has the same limitation but I guess it uses the same MAC of
VT6130.
yongari [Sat, 9 Jan 2010 00:17:43 +0000 (00:17 +0000)]
MFC r200638:
Implement interrupt moderation scheme supported by VT61xx
controllers. TX/RX interrupt mitigation is controlled by
VGE_TXSUPPTHR and VGE_RXSUPPTHR register. These registers suppress
generation of interrupts until the programmed frames counter equals
to the registers. VT61xx also supports interrupt hold off timer
register. If this interrupt hold off timer is active all interrupts
would be disabled until the timer reaches to 0. The timer value is
reloaded whenever VGE_ISR register written. The timer resolution is
about 20us.
Previously vge(4) used single shot timer to reduce Tx completion
interrupts. This required VGE_CRS1 register access in Tx
start/completion handler to rearm new timeout value and it did not
show satisfactory result(more than 50k interrupts under load). Rx
interrupts was not moderated at all such that vge(4) used to
generate too many interrupts which in turn made polling(4) better
approach under high network load.
This change activates all interrupt moderation mechanism and
initial values were tuned to generate interrupt less than 8k per
second. That number of interrupts wouldn't add additional packet
latencies compared to polling(4). These interrupt parameters could
be changed with sysctl.
dev.vge.%d.int_holdoff
dev.vge.%d.rx_coal_pkt
dev.vge.%d.tx_coal_pkt
Interface has be brought down and up again before change take
effect.
With interrupt moderation there is no more need to loop in
interrupt handler. This loop always added one more register access.
While I'm here remove dead code which tried to implement subset of
interrupt moderation.
yongari [Sat, 9 Jan 2010 00:12:59 +0000 (00:12 +0000)]
MFC r200616-200617,200635,200639,200644
r200616:
Add new flag VGE_FLAG_SUSPENDED to mark suspended state and
remove suspended member in softc.
r200617:
Add "Velocity" to probe message which will make it clearer which
ethernet controller was recognized. VIA consistently calls
"Velocity" family for gigabit ethernet controllers. For fast
ethernet controllers they uses "Rhine" family(vr(4) controllers))
and vr(4) already shows "Rhine" in probe message.
yongari [Sat, 9 Jan 2010 00:02:40 +0000 (00:02 +0000)]
MFC r200551-200552,200555,200558,200609,200613
r200551:
Whenever link state change interrupt is raised, vge_tick() is
called and vge(4) used to drive auto-negotiation timer(mii_tick) in
vge_tick(). Therefore the mii_tick was not called for every hz such
that auto-negotiation complete was never handled in vge(4).
Use mii_pollstat to extract current negotiated speed/duplex instead
of mii_tick. The latter is valid only for auto-negotiation case.
While I'm here change the confusing function name vge_tick() to
vge_link_statchg().
r200552:
Report media change result to caller instead of returning success
without regard to the result.
r200555:
Don't report current link status if interface is not UP.
If interface is not UP, the current link status wouldn't
reflect the negotiated status.
r200558:
Tell upper layer vge(4) supports long frames. This should be done
after ether_ifattach(), as ether_ifattach() initializes it with
ETHER_HDR_LEN.
While I'm here remove setting if_mtu, it's already handled in
ether_ifattach().
r200609:
All vge(4) controllers support RX/TX checksum offloading for VLAN
tagged frames so add checksum offloading capabilities. Also add
missing VLAN hardware tagging control in ioctl handler and let
upper stack know current VLAN capabilities.
r200613:
Rewrite RX filter setup and simplify code.
Now promiscuous mode and multicast handling is performed in single
function, vge_rxfilter().
yongari [Fri, 8 Jan 2010 22:26:24 +0000 (22:26 +0000)]
MFC r200538,200540-200541,200543,200545,200548
r200538:
Introduce vge_flags member in softc. The vge_flags member will
record device specific bits. Remove vge_link and use vge_flags.
While here, move clearing link state before mii_mediachg() as
mii_mediachg() may affect link state.
r200540:
Save PHY address by reading VGE_MIICFG register. For PCIe
controllers(VT613x), we assume the PHY address is 1.
Use the saved PHY address in MII register access routines and
remove accessing VGE_MIICFG register.
While I'm here save PCI express capability register which will be
used in near future.
r200541:
Add MSI support for VT613x controllers.
r200543:
Increase output queue size from 64 to 255.
r200545:
We don't have to reload EEPROM in vge_reset(). Because vge_reset()
is called in vge_init_lock(), vge(4) always used to reload EEPROM.
Also add more comment why vge(4) clears VGE_CHIPCFG0_PACPI bit.
While I'm here add missing new line in vge_reset().
yongari [Fri, 8 Jan 2010 22:08:18 +0000 (22:08 +0000)]
MFC r200526-200527,200529,200531-200533,200536
r200526:
Use PCIR_BAR instead of hard-coded value.
r200527:
Fix typo in register definition.
r200529:
Clear VGE_TXDESC_Q bit for transmitted frames. The VGE_TXDESC_Q bit
seems to work like a tag that indicates 'not list end' of queued
frames. Without having a VGE_TXDESC_Q bit indicates 'list end'. So
the last frame of multiple queued frames has no VGE_TXDESC_Q bit.
The hardware has peculiar behavior for VGE_TXDESC_Q bit handling.
If the VGE_TXDESC_Q bit of descriptor was set the controller would
fetch next descriptor. However if next descriptor's OWN bit was
cleared but VGE_TXDESC_Q was set, it could confuse controller.
Clearing VGE_TXDESC_Q bit for transmitted frames ensure correct
behavior.
yongari [Fri, 8 Jan 2010 21:37:16 +0000 (21:37 +0000)]
MFC r200525:
Overhaul bus_dma(9) usage and fix various things.
o Separate TX/RX buffer DMA tag from TX/RX descriptor ring DMA tag.
o Separate RX buffer DMA tag from common buffer DMA tag. RX DMA
tag has different restriction compared to TX DMA tag.
o Add 40bit DMA address support.
o Adjust TX/RX descriptor ring alignment to 64 bytes from 256
bytes as documented in datasheet.
o Added check to ensure TX/RX ring reside within a 4GB boundary.
Since TX/RX ring shares the same high address register they
should have the same high address.
o TX/RX side bus_dmamap_load_mbuf_sg(9) support.
o Add lock assertion to vge_setmulti().
o Add RX spare DMA map to recover from DMA map load failure.
o Add optimized RX buffer handler, vge_discard_rxbuf which is
activated when vge(4) sees bad frames.
o Don't blindly update VGE_RXDESC_RESIDUECNT register. Datasheet
says the register should be updated only when number of
available RX descriptors are multiple of 4.
o Use __NO_STRICT_ALIGNMENT instead of defining VGE_FIXUP_RX which
is only set for i386 architecture. Previously vge(4) also
performed expensive copy operation to align IP header on amd64.
This change should give RX performance boost on amd64
architecture.
o Don't reinitialize controller if driver is already running. This
should reduce number of link state flipping.
o Since vge(4) drops a driver lock before passing received frame
to upper layer, make sure vge(4) is still running after
re-acquiring driver lock.
o Add second argument count to vge_rxeof(). The argument will
limit number of packets could be processed in RX handler.
o Rearrange vge_rxeof() not to allocate RX buffer if received
frame was bad packet.
o Removed if_printf that prints DMA map failure. This type of
message shouldn't be used in fast path of driver.
o Reduce number of allowed TX buffer fragments to 6 from 7. A TX
descriptor allows 7 fragments of a frame. However the CMZ field
of descriptor has just 3bits and the controller wants to see
fragment + 1 in the field. So if we have 7 fragments the field
value would be 0 which seems to cause unexpected results under
certain conditions. This change should fix occasional TX hang
observed on vge(4).
o Simplify vge_stat_locked() and add number of available TX
descriptor check.
o vge(4) controllers lack padding short frames. Make sure to fill
zero for the padded bytes. This closes unintended information
disclosure.
o Don't set VGE_TDCTL_JUMBO flag. Datasheet is not clear whether
this bit should be set by driver or write-back status bit after
transmission. At least vendor's driver does not set this bit so
remove it. Without this bit vge(4) still can send jumbo frames.
o Don't start driver when vge(4) know there are not enough RX
buffers.
o Remove volatile keyword in RX descriptor structure. This should
be handled by bus_dma(9).
o Collapse two 16bits member of TX/RX descriptor into single 32bits
member.
o Reduce number of RX descriptors to 252 from 256. The
VGE_RXDESCNUM is 16bits register but only lower 8bits are valid.
So the maximum number of RX descriptors would be 255. However
the number of should be multiple of 4 as controller wants to
update 4 RX descriptors at a time. This limits the maximum
number of RX descriptor to be 252.
yongari [Fri, 8 Jan 2010 21:15:09 +0000 (21:15 +0000)]
MFC r198987,199414,199543,200422
Partial merge r198987:
Use device_printf() and if_printf() instead of printf() with an explicit
unit number and remove 'unit' members from softc.
Partial merge r199414:
Use the bus_*() routines rather than bus_space_*() for register operations.
r199543:
Several fixes to this driver:
- Overhaul the locking to avoid recursion and add missing locking in a few
places.
- Don't schedule a task to call vge_start() from contexts that are safe to
call vge_start() directly. Just invoke the routine directly instead
(this is what all of the other NIC drivers I am familiar with do). Note
that vge(4) does not use an interrupt filter handler which is the primary
reason some other drivers use tasks.
- Add a new private timer to drive the watchdog timer instead of using
if_watchdog and if_timer.
- Fixup detach by calling ether_ifdetach() before stopping the interface.
r200422:
Remove driver lock assertion in MII register access. This change
was made in r199543 to remove MTX_RECURSE. These routines can be
called in device attach phase(e.g. mii_phy_probe()) so checking
assertion here is not right as caller does not hold a driver lock.
rmacklem [Fri, 8 Jan 2010 20:25:59 +0000 (20:25 +0000)]
MFC: r200999
Modify the experimental server so that it uses VOP_ACCESSX().
This is necessary in order to enable NFSv4 ACL support. The
argument to nfsvno_accchk() was changed to an accmode_t and
the function nfsrv_aclaccess() was no longer needed and,
therefore, deleted.
netchild [Fri, 8 Jan 2010 09:59:13 +0000 (09:59 +0000)]
MFC r197816:
---snip---
Prevent paging pressure from draining arc too much
- always drain arc if above arc_c_max - never drain arc if arc is below
arc_c_max
---snip---
bms [Thu, 7 Jan 2010 14:15:34 +0000 (14:15 +0000)]
MFC r200871:
Use ALLOW_NEW_SOURCES and BLOCK_OLD_SOURCES to signal a join or leave
with SSM MLDv2 by default.
This is current practice and complies with RFC 4604, as well as being
required by production IPv6 networks in Japan.
The behaviour may be disabled by setting the net.inet6.mld.use_allow
sysctl/tunable to 0.
yongari [Thu, 7 Jan 2010 00:55:07 +0000 (00:55 +0000)]
MFC r200088,200227-200228,200246,200264,201446
r200088:
Add workaround to overcome hardware limitation which allows only a
single outstanding DMA read operation. Most controllers targeted to
client with PCIe bus interface(e.g. BCM5761) may have this
limitation. All controllers for servers does not have this
limitation.
Collapsing mbuf chains to reduce number of memory reads before
transmitting was most effective way to workaround this. I got about
940Mbps from 850Mbps with mbuf collapsing on BCM5761. However it
takes a lot of CPU cycles to collapse mbuf chains so add tunable to
control the number of allowed TX buffers before collapsing. The
default value is 0 which effectively disables the forced collapsing.
For most cases 2 would yield best performance(about 930Mbps)
without much sacrificing CPU cycles.
Note the collapsing is only activated when the controller is on
PCIe bus and the frame does not need TSO operation. TSO does not
seem to suffer from the hardware limitation because the payload
size is much bigger than normal IP datagram.
Thanks to davidch@ who told me the limitation of client controllers
and actually gave possible workarounds to mitigate the limitation.
r200227:
Remove PHY isolate/power down code in bge_stop(). The isolation
handler in brgphy(4) does not exist and brgphy(4) just resets the
PHY and returns EINVAL as it has no isolation handler. I also agree
on Marius's opinion that stop handler of every NIC driver seems to
be the wrong place for implementing PHY isolate/power down.
If we need PHY isolate/power down it should be implemented in
brgphy(4) and users should administratively down the PHY.
r200228:
Don't access jumbo frame related registers if controller lacks the
feature. These registers are reserved on controllers that have no
support for jumbo frame.
Only BCM5700 has mini ring so do not poke mini ring related
registers if controller is not BCM5700.
r200246:
Partially revert r200228. For mini RCB case, bge(4) still have to
disable mini ring withtout regard to mini ring support.
r200264:
Create sysctl node(dev.bge.%d.focred_collapse) instead of
hw.bge.forced_collapse. hw.bge.forced_collapse affects all bge(4)
controllers on system which may not desirable behavior of the
sysctl node. Also allow the sysctl node could be modified at any
time.
r201446:
Fix regression introduced in r198318. BCM5754/BCM5754M uses the
same ASIC ID of BCM5758 such that r198318 incorecctly enabled TSO
on BCM5754.BCM5754M controllers. BCM5754/BCM5754M needs a special
firmware to enable TSO and bge(4) does not support firmware based
TSO.
r199670:
Fix two long standing bugs on bge(4). Most pre BCM5755 controllers
have a DMA bug when buffer address crosses a multiple of the 4GB
boundary(e.g. 4GB, 8GB, 12GB etc). Limit DMA address to be within
4GB address for these controllers. The second DMA bug limits DMA
address to be within 40bit address space. This bug applies to
BCM5714 and BCM5715 and 5708(bce(4) controller). This is not
actually a MAC controller bug but an issue with the embedded PCIe
to PCI-X bridge in the device. So for BCM5714/BCM5715 controllers
also limit the DMA address to be within 40bit address space.
Special thanks to davidch@ who gave me detailed errata information.
I think this change will fix long standing bge(4) instability
issues on systems with more than 4GB memory.
r199671:
Implement TSO for BCM5755 or newer controllers. Some controllers
seem to require a special firmware to use TSO. But the firmware is
not available to FreeBSD and Linux claims that the TSO performed by
the firmware is slower than hardware based TSO. Moreover the
firmware based TSO has one known bug which can't handle TSO if
ethernet header + IP/TCP header is greater than 80 bytes. The
workaround for the TSO bug exist but it seems it's too expensive
than not using TSO at all. Some hardwares also have the TSO bug so
limit the TSO to the controllers that are not affected TSO issues
(e.g. 5755 or higher).
While I'm here set VLAN tag bit to all descriptors that belengs to
a frame instead of the first descriptor of a frame. The datasheet
is not clear how to handle VLAN tag bit but it worked either way in
my testing. This makes it simplify TSO configuration a little bit.
Big thanks to davidch@ who sent me detailed TSO information.
Without this I was not able to implement it.
r199674:
Add missing function prototype in r199671.
r199679:
Reduce status block size DMAed by controller. bge(4) uses single
Tx/Rx/Rx return ring such that large part of status block was not
used at all. All bge(4) controllers except BCM5700 AX/BX has a
feature to control the size of status block. So use minimum status
block size allowed in controller. This reduces number of DMAed
status block size to 32 bytes from 80 bytes.
r199761:
BGE_FLAG_40BIT_BUG should be set before creating DMA tags.
r199807:
Make sure one shot MSI is enabled.
r199808:
Fix typo which inversed the logic which in turn disabled MSI.
yongari [Wed, 6 Jan 2010 23:42:15 +0000 (23:42 +0000)]
MFC r199667-199668
r199667:
Cache Rx producer/Tx consumer index as soon as we know status block
update and then clear status block. Previously it used to access
these index without synchronization which may cause problems when
bounce buffers are used. Also add missing bus_dmamap_sync(9) in
polling handler. Since we now update status block in driver, adjust
bus_dmamap_sync(9) for status block.
r199668:
For MSI case, interrupt is not shared and we don't need to force
PCI flush to get correct status block update. Add an optimized
interrupt handler that is activated for MSI case. Actual interrupt
handling is done by taskqueue such that the handler does not
require driver lock for Rx path. The MSI capable bge(4) controllers
automatically disables further interrupt once it enters interrupt
state so we don't need PIO access to disable interrupt in interrupt
handler.
yongari [Wed, 6 Jan 2010 23:34:53 +0000 (23:34 +0000)]
MFC 199663-199666
r199663:
Due to newly added PCIe capabilities fallback code for finding the
PCIe capability did not work right on recent controllers. Remove
FreeBSD 6.x support code.
r199664:
Use capability pointer to access PCIe registers rather than
directly access them at fixed address. While I'm here don't touch
other bits of PCIe device control register except max payload size.
r199665:
Controller does not write Rx descriptors, remove BUS_DMASYNC_PREREAD.
r199666:
Rearrange bge_start_locked to see we can send more frames by
checking IFF_DRV_RUNNING and IFF_DRV_OACTIVE flags. Also if we
have less than 16 free send BDs set IFF_DRV_OACTIVE and try it
later. Previously bge(4) used to reserve 16 free send BDs after
loading dma maps but hardware just need one reserved send BD. If
prouder index has the same value of consumer index it means the Tx
queue is empty.
While I'm here check IFQ_DRV_IS_EMPTY first to save one lock
operation.
yongari [Wed, 6 Jan 2010 23:26:09 +0000 (23:26 +0000)]
MFC r199065,199115-199116,199153,199661-199662
r199065:
Correct disabling checksum offloading for BCM5700 B0.
r199115:
Add missing bus_dmamap_sync(9) before issuing kick command.
r199116:
Zero out Tx/Rx descriptors before using them. Also add missing
bus_dmamap_sync(9) after Tx descriptor initialization.
r199153:
Controller does not update Tx descriptors(send BDs) after sending
frames so remove unnecessary BUS_DMASYNC_PREREAD and
BUS_DMASYNC_POSTREAD of bus_dmamap_sync(9).
r199009:
bge(4) already switched to use UMA backed page allocator and local
memory allocator for jumbo frame was removed long time ago. Remove
no more used macros.
r199010:
Do bus_dmamap_sync call only if frame size is greater than
standard buffer size. If controller is not capable of handling
jumbo frame, interface MTU couldn't be larger than standard MTU
which in turn the received should be fit in standard buffer. This
fixes bus_dmamap_sync call for jumbo ring is called even if
interface is configured to use standard MTU.
Also if total frame size could be fit into standard buffer don't
use jumbo buffers.
r199011:
Reimplement Rx buffer allocation to handle dma map load failure.
Introduce two spare dma maps for standard buffer and jumbo buffer
respectively. If loading a dma map failed reuse previously loaded
dma map. This should fix unloaded dma map is used in case of dma
map load failure. Also don't blindly unload dma map and defer
dma map sync and unloading operation until we know dma map for new
buffer is successfully loaded. This change saves unnecessary dma
load/unload operation. Previously bge(4) tried to reuse mbuf
with unloaded dma map which is really bad thing in bus_dma(9)
perspective.
While I'm here update if_iqdrops if we can't allocate Rx buffers.
r199014:
Fix I mssied in r199011. Rx ring index also should be updated.
If we fill Rx ring full instead of half we can simplify this logic
but this requires more experimentation.
r199020:
Tell upper layer we support long frames. ether_ifattach()
initializes it to ETHER_HDR_LEN so we have to override it after
calling ether_ifattch().
While I'm here remove setting if_mtu value, it's initialized in
ether_ifattach().
r199035:
Don't count input errors twice, we always read input errors from
MAC in bge_tick. Previously it used to show more number of input
errors. I noticed actual input errors were less than 8% even for
64 bytes UDP frames generated by netperf.
Since we always access BGE_RXLP_LOCSTAT_IFIN_DROPS register in
bge_tick, remove useless code protected by #ifdef notyet.
r199036:
Count number of inbound packets which were chosen to be discarded
as input errors. Also count out of receive BDs as input errors.
r199054:
Partially revert r199035.
Revision 1.158 says only lower ten bits of
BGE_RXLP_LOCSTAT_IFIN_DROPS register is valid. For BCM5761 case it
seems the controller maintains 16bits value for the register.
However 16bits are still too small to count all dropped packets
happened in a second. To get a correct counter we have to read the
register in bge_rxeof() which would be too expensive.
yongari [Wed, 6 Jan 2010 22:45:49 +0000 (22:45 +0000)]
MFC r198923-198924,198927-198928
r198923:
Use correct dma tag for jumbo buffer.
r198924:
Covert bge_newbuf_std to use bus_dmamap_load_mbuf_sg(9). Note,
bge_newbuf_std still has a bug for handling dma map load failure
under high network load. Just reusing mbuf is not enough as driver
already unloaded the dma map of the mbuf. Graceful recovery needs
more work.
Ideally we can just update dma address part of a Rx descriptor
because the controller never overwrite the Rx descriptor. This
requires some Rx initialization code changes and it would be done
later after fixing other incorrect bus_dma(9) usages.
r198927:
Remove common DMA tag used for TX/RX mbufs and create Tx DMA tag
and Rx DMA tag separately. Previously it used a common mbuf DMA tag
for both Tx and Rx path but Rx buffer(standard ring case) should
have a single DMA segment and maximum buffer size of the segment
should be less than or equal to MCLBYTES. This change also make it
possible to add TSO with minor changes.
r198928:
Make bge_newbuf_std()/bge_newbuf_jumbo() returns actual error code
for buffer allocation. If driver know we are out of Rx buffers let
controller stop. This should fix panic when interface is run even
if it had no configured Rx buffers.
gavin [Wed, 6 Jan 2010 20:40:41 +0000 (20:40 +0000)]
MFC r200820:
Support the tablet in (at least) the Toshiba Portege M200 Tablet PC.
This device only appears on the ACPI bus, so isn't caught by the current
entry for it in the uart(4) ISA attachment.
jkim [Wed, 6 Jan 2010 20:28:47 +0000 (20:28 +0000)]
MFC: r200251
- Try pre-allocating all FIBs upfront. Previously we tried pre-allocating
128 FIBs first and allocated more later if necessary. Remove now unused
definitions from the header file[1].
- Force sequential bus scanning. It seems parallel scanning is in fact
slower and causes more harm than good[1]. Adjust a comment to reflect that.
qingli [Tue, 5 Jan 2010 22:33:10 +0000 (22:33 +0000)]
MFC r201285
Consolidate the route message generation code for when address
aliases were added or deleted. The announced route entry for
an address alias is no longer empty because this empty route
entry was causing some route daemon to fail and exit abnormally.
qingli [Tue, 5 Jan 2010 22:28:23 +0000 (22:28 +0000)]
MFC r201284
Multiple IPv6 addresses of the same prefix can be installed on the
same interface. The first address will install the prefix route into
the kernel routing table and that prefix will be marked as on-link.
Without RADIX_MPATH enabled, the other address aliases of the same
prefix will update the prefix reference count but no other routes
will be installed. Consequently the prefixes associated with these
addresses would not be marked as on-link. As such, incoming packets
destined to these address aliases will fail the ND6 on-link check
on input. This patch fixes the above problem by searching the kernel
routing table and try to find an on-link prefix on the given interface.
qingli [Tue, 5 Jan 2010 22:14:55 +0000 (22:14 +0000)]
MFC r201282, r201543
r201282
-------
The proxy arp entries could not be added into the system over the
IFF_POINTOPOINT link types. The reason was due to the routing
entry returned from the kernel covering the remote end is of an
interface type that does not support ARP. This patch fixes this
problem by providing a hint to the kernel routing code, which
indicates the prefix route instead of the PPP host route should
be returned to the caller. Since a host route to the local end
point is also added into the routing table, and there could be
multiple such instantiations due to multiple PPP links can be
created with the same local end IP address, this patch also fixes
the loopback route installation failure problem observed prior to
this patch. The reference count of loopback route to local end would
be either incremented or decremented. The first instantiation would
create the entry and the last removal would delete the route entry.
r201543
-------
The IFA_RTSELF address flag marks a loopback route has been installed
for the interface address. This marker is necessary to properly support
PPP types of links where multiple links can have the same local end
IP address. The IFA_RTSELF flag bit maps to the RTF_HOST value, which
was combined into the route flag bits during prefix installation in
IPv6. This inclusion causing the prefix route to be unusable. This
patch fixes this bug by excluding the IFA_RTSELF flag during route
installation.
projects / FreeBSD / stable / 8.git / log