stashed away in ath_node.
As much as I tried to stuff that behind the ATH_NODE lock, unfortunately
the locking is just too plain hairy (for me! And I wrote it!) to do
cleanly. Hence using atomics here instead of a lock. The ATH_NODE lock
just isn't currently used anywhere besides the rate control updates.
If in the future everything gets migrated back to using a single ATH_NODE
lock or a single global ATH_TX lock (ie, a single TX lock for all TX and
TX completion) then fine, I'll remove the atomics.
it run out of multiple concurrent contexts.
Right now the ath(4) TX processing is a bit hairy. Specifically:
* It was running out of ath_start(), which could occur from multiple
concurrent sending processes (as if_start() can be started from multiple
sending threads nowdays.. sigh)
* during RX if fast frames are enabled (so not really at the moment, not
until I fix this particular feature again..)
* during ath_reset() - so anything which calls that
* during ath_tx_proc*() in the ath taskqueue - ie, TX is attempted again
after TX completion, as there's now hopefully some ath_bufs available.
* Then, the ic_raw_xmit() method can queue raw frames for transmission
at any time, from any net80211 TX context. Ew.
This has caused packet ordering issues in the past - specifically,
there's absolutely no guarantee that preemption won't occuring _during_
ath_start() by the TX completion processing, which will call ath_start()
again. It's a mess - 802.11 really, really wants things to be in
sequence or things go all kinds of loopy.
So:
* create a new task struct for TX'ing;
* make the if_start method simply queue the task on the ath taskqueue;
* make ath_start() just be called by the new TX task;
* make ath_tx_kick() just schedule the ath TX task, rather than directly
calling ath_start().
Now yes, this means that I've taken a step backwards in terms of
concurrency - TX -and- RX now occur in the same single-task taskqueue.
But there's nothing stopping me from separating out the TX / TX completion
code into a separate taskqueue which runs in parallel with the RX path,
if that ends up being appropriate for some platforms.
This fixes the CCMP/seqno concurrency issues that creep up when you
transmit large amounts of uni-directional UDP traffic (>200MBit) on a
FreeBSD STA -> AP, as now there's only one TX context no matter what's
going on (TX completion->retry/software queue,
userland->net80211->ath_start(), TX completion -> ath_start());
but it won't fix any concurrency issues between raw transmitted frames
and non-raw transmitted frames (eg EAPOL frames on TID 16 and any other
TID 16 multicast traffic that gets put on the CABQ.) That is going to
require a bunch more re-architecture before it's feasible to fix.
In any case, this is a big step towards making the majority of the TX
path locking irrelevant, as now almost all TX activity occurs in the
taskqueue.
Phew.
Right now processing a full 512 frame queue takes quite a while (measured
on the order of milliseconds.) Because of this, the TX processing ends up
sometimes preempting the taskqueue:
* userland sends a frame
* it goes in through net80211 and out to ath_start()
* ath_start() will end up either direct dispatching or software queuing a
frame.
If TX had to wait for RX to finish, it would add quite a few ms of
additional latency to the packet transmission. This in the past has
caused issues with TCP throughput.
Now, as part of my attempt to bring sanity to the TX/RX paths, the first
step is to make the RX processing happen in smaller 'parts'. That way
when TX is pushed into the ath taskqueue, there won't be so much latency
in the way of things.
The bigger scale change (which will come much later) is to actually
process the frames in the ath_intr taskqueue but process _frames_ in
the ath driver taskqueue. That would reduce the latency between
processing and requeuing new descriptors. But that'll come later.
The actual work:
* Add ATH_RX_MAX at 128 (static for now);
* break out of the processing loop if npkts reaches ATH_RX_MAX;
* if we processed ATH_RX_MAX or more frames during the processing loop,
immediately reschedule another RX taskqueue run. This will handle
the further frames in the taskqueue.
This should have very minimal impact on the general throughput case,
unless the scheduler is being very very strange or the ath taskqueue
ends up spending a lot of time on non-RX operations (such as TX
completion.)
and Sierra Wireless MC8790V. Also implement the .ucom_poll method.
Note: This makes it possible to use lqr/echo in ppp.conf. And it
resolves ppp hanging during the PPp> phase.
Reviewed by: hps
MFC after: 1 week
- Disable the support for the second channel on twin-channel EISA cards as
the current incarnation can't possibly work correctly (it hasn't worked
since switching to new-bus where new-bus allocates the softc). If anyone
bothers to test this again it can be fixed properly and brought back.
- Use device_printf() and device_get_nameunit() instead of adv_name().
- Remove use of explicit bus space handles and tags.
- Use PCI bus accessors and helper routines rather than accessing
config registers directly.
- Handle failures from adv_attach().
Tested by: no one (hope it works)
virtqueue: Fix non-indirect virtqueues
We really must walk the entire descriptor chain in order
to append the to be free'd chain to the existing free
chain.
Submitted by: Bryan Venteicher (bryanv@daemoninthecloset.org)
Reported by: cognet
This change will enable IPMI access on 5717/5718/5719/5720 and 5761
controllers. Because ASF is not available when APE firmware is
present, bge_allow_asf tunable is ignored when driver detects APE
firmware. Also bge(4) no longer performs two resets(one blind
reset and the other reset with firmware in mind) in device attach.
Now bge(4) performs a reset with enough information in bge_reset().
The APE firmware also needs special handling to make suspend/resume
work but it was not implemented yet.
With this change, bge(4) should work on any 5717/5718/5719/5720
controllers. Special thanks to Mike Hibler at Emulab who setup
remote debugging on Dell R820. Without his help I couldn't be able
to address several issues happened on Dell Rx20 systems. And many
thanks to Broadcom for continuing to support FreeBSD!
Submitted by: davidch (initial version)
H/W donated by: Broadcom
Tested by: many
Tested on: Del R820/R720/R620/R420/R320 and HP Proliant DL 360 G8
handling(jumbo, wire speed etc) in brgphy_reset(). Touching
BRGPHY_MII_AUXCTL register seems to confuse APE firmware such that
it couldn't establish a link.
BGE_PCI_PCISTATE register before issuing global reset. After
issuing reset, it reads BGE_PCI_PCISTATE register again and
compares the saved register value and current value. It was used to
know whether the global reset operation was completed or not.
Unfortunately, this logic caused several issues on recent BCM5717/
5718/5719 and BCM5720 controllers. It seems APE firmware accesses
some registers while global reset is in progress such that reading
BGE_PCI_PCISTATE register after reset does not yield old pre-reset
state value. This resulted in consuming too much time in global
reset and sometimes it couldn't successfully complete reset.
The BGE_MISCCFG_RESET_CORE_CLOCKS of BGE_MISC_CFG register is
self-clearing bit so driver is able to know the reset completion.
But the core-lock reset will disable indirect/flat/standard access
modes such that driver cannot poll BGE_MISCCFG_RESET_CORE_CLOCKS
bit of BGE_MISC_CFG register. So just wait enough time for
core-clock reset to complete.
Data sheet says driver should wait 100us for PCI/PCI-X devices and
100ms for PCIe devices. I chose 1ms for PCI/PCI-X since this value
was used for many years in bge(4). For PCIe devices, use 100ms as
recommended by data sheet.
bge_chipinit() also cleared BGE_MAC_MODE register which shall clear
firmware configured mode information. I think this will result in
losing ASF/IPMI link in device attachment. Let bge_reset() honor
firmware configured BGE_MAC_MODE register and don't announce driver
is UP in bge_reset(). Firmware should have control over driver until
it's fully initialized by driver.
While I'm here, enable workaround for PCI-X BCM5704 A0 in
bge_reset(). This will prevent internal arbitration logic from
switching to the other DMA engine after a retry cycle.
set not-NULL during SIM registration and set to UMASS_GONE on destruction.
Debug messages there look broken for at least 9 years, as they dereference
softc value that was just checked to be equal to NULL.
- Remove magic pointer value UMASS_GONE and use simple NULL instead.
Found by: Clang Static Analyzer
While here, change ISCI_LED to ISCI_PHY since conceptually the hardware
ties the LEDs to a phy and the LEDs for a given phy cannot be controlled
independently.
Submitted by: Paul Maulberger <Paul.Maulberger at gmx.de> (with modifications)
This lets userspace read arbitrary information from the SFP+ modules
etc. on this bus.
Reading multiple bytes in the same transaction isn't possible right now.
I'll update the driver once the chip's firmware supports this.
MFC after: 3 days
that requires 10ms delay after device reset. Because that code was
there from day 1, I guess it was added to give enough settlement
time after updating BGE_MAC_MODE register.
The recommended delay time for BGE_MAC_MODE after updating is 40us
and it was already done in r241219.
The VCPU(Virtual CPU) of BCM5906 is used to provide a mechanism to
control the bootcode execution and to pick up configuration data
stored inside the EEPROM.
The bootcode of BCM5906 will check the BGE_VCPU_STATUS_DRV_RESET
bit to decide which booting procedure to choose.
Data sheet indicates the VCPU of BCM5906 should set
BGE_VCPU_STATUS_DRV_RESET bit *before* VCPU reset or global reset.
water mark to 256 bytes. Otherwise controller will encounter DMA
write under run errors and would result in RX DMA hang. If the
maximum payload size is 128 bytes, the water mark is set to 128
bytes as usual.
While here, set maximum read request size to 2048 for BCM5719/BCM5720.
For other PCIe devices, use 4096. And reprogram the maximum read
request size whenever device reset is performed.
make analog input loopback and dual-stream playback work by enabling
signal mixing by nid 22, as it should be according to info returned by
the CODEC. Otherwise pin nid 28 receives only signal from DAC nid 16.
PR: kern/169124
MFC after: 1 week
the ATH_TXQ_* macros.
* Introduce the new macros;
* rename the TID queue and TID filtered frame queue so the compiler
tells me I'm using the wrong macro.
These should correspond 1:1 to the existing code.
to 32k swamped the controller causing firmware hangs. Instead, round
requests smaller than 64k up to the next power of 2 as a general rule.
To handle the one known special case of a command that accepts a 12k
buffer returning a 24k-ish reply, round requests between 8k and 16k up
to 32k rather than 16k. The result is that commands less than 8k should
now be rounded up to a smaller size (either 4k or 8k) rather than 32k.
PR: kern/155658
Tested by: Andreas Longwitz
MFC after: 1 week
