- Rework the entire pcm_channel structure:
* Remove rarely used link placeholder, instead, make each pcm_channel
as head/link of each own/each other. Unlock - Lock sequence due to
sleep malloc has been reduced.
* Implement "busy" queue which will contain list of busy/active
channels. This greatly reduce locking contention for example while
servicing interrupt for hardware with many channels or when virtual
channels reach its 256 peak channels.
- So I heard you like v chan ... O RLY?
Welcome to Virtual **Record** Channels (vrec, rec vchans, vchans for
recording, Rec-Chan, you decide), the ultimate solutions for your
nagging O_RDWR full-duplex wannabe (note: flash plugins) monopolizing
single record channel causing EBUSY. Vrec works exactly like Vchans
(or, should I rename it to "Vplay" :) , except that it operates on the
opposite direction (recording). Up to 256 vrecs (like vchans) are
possible.
Notes:
* Relocate dev.pcm.%d.{vchans,vchanformat,vchanrate} to each of its
respective node/direction:
dev.pcm.%d.play.* for "play" (cdev = dsp%d.vp%d)
dev.pcm.%d.rec.* for "record" (cdev = dsp%d.vr%d)
* Don't expect that it will magically give you ability to split
"recording source" (eg: 1 channel for cdrom, 1 channel for mic,
etc). Just admit that you only have a *single* recording source /
channel. Please bug your hardware vendor instead :)
- Bump maxautovchans from 4 to 16. For a full-fledged multimedia
desktop/workstation with too many soundservers installed (esound,
artsd, jackd, pulse/polypaudio, ding-dong pling plong mudkip fuh fuh,
etc), 4 seems inadequate. There will be no memory penalty here, since
virtual channels are allocate only by demand.
- Nuke/Rework the entire statically created cdev entries. Everything is
clonable through snd own clone manager which designed to withstand many
kind of abusive devfs droids such as:
* while : ; do /bin/test -e /dev/dsp ; done
* jot 16777216 0 | while read x ; do ls /dev/dsp0.$x ; done
* hundreds (could be thousands) concurrent threads/process opening
"/dev/dsp" (previously, this might result EBUSY even with just
3 contesting threads/procs).
o Reusable clone objects (instead of creating new one like there's no
tomorrow) after certain expiration deadline. The clone allocator will
decide whether to reuse, share, or creating new clone.
o Automatic garbage collector.
- Dynamic unit magic allocator. Maximum attached soundcards can be tuned
using tunable "hw.snd.maxunit" (Default to 512). Minimum is 16, and
maximum is 2048.
- ..other fixes, mostly related to concurrency issues.
joel@ will do the manpage updates on sound(4).
Have fun.
Change the VOP_OPEN(), vn_open() vnode operation and d_fdopen() cdev operation
argument from being file descriptor index into the pointer to struct file.
Proposed and reviewed by: jhb
Reviewed by: daichi (unionfs)
Approved by: re (kensmith)
- Coverity Prevent(tm) CID 1906 a bogus use of bzero where unneeded.
- ICH8 systems autoneg to 100 rather than 1000, this can also be
seen in 82573, the logic was backwards.
- On new 82575 quadports half duplex tx speed is slow... this was due
to overwriting TCTL reg rather than adding bits.
OpenBSD's if_ral.c.
I didn't make the LINKSYS4 -> CISCOLINKSYS name change, nor did I
include the RALINK RT2573 that's supported by the rum(4) driver. I
didn't merge any code changes either.
race seen on smp laptops when suspending where the rx task can be
entered after the interface is detach'd.
NB: use of taskqueue_drain while holding the softc mutex is problematic
Submitted by: ambrisko
MFC after: 1 month
by the subsequent mix_setdevs() and friends.
- Minor style(9) declaration arrangement nit.
Requested by: joeld
Submitted by: pluknet <pluknet@gmail.com>
- Double the number of descriptors that a single call to send can use
- Quadruple the number of descriptors that can be reclaimed per pass
- only run reclaim twice per second
- increase coalesce timer from 3.5us to 5us
fix printf warning on 64-bit platforms
Neither me nor Ariff have access to any of this hardware, so all tests
have been made by Konstantin and Artem. Commit message mostly written
by Konstantin.
envy24:
- Add test code to support rear line-in input on 'Terratec DMX 6fire'
audio card. This code is also intended to be used in the future for
support of cards, that have I2C-to-GPIO expanders wired between the
control line of the audio codec and the Envy24, however such cards
are too complex and i can't add that support without hardware sample
of such board, i've already tried and failed.
envy24ht:
- Add support for 'AudioTrak Prodigy HD2'.
- Add support for 'AudioTrak Prodigy 7.1 XT'.
- Add support for 'ESI Juli@' (Works ok, DAC volume is hard-coded for
the time being, so 'mixer vol ...' doesn't work, only 'mixer pcm
...' works). [1]
- Fix bug in the init data for M-Audio Revolution 5.1, that
results in distorted sound.
- Add software volume control (now 'mixer pcm' works, thanks to Ariff).
- Add support for more samples rates - 176.4kHz and 192kHz.
- Fix problem with the 192kHz samples rate playback when 24.576MHz
crystal is used on the board instead of 49.152MHz crystal.
spicds:
- Add support for Asahi Kasei flagship DAC - AK4396 (used in AudioTrak
Prodigy HD2).
Submitted by: Konstantin Dimitrov <kosio.dimitrov@gmail.com>
Tested by: Artem Antonov [1]
Reviewed by: ariff
the card, panic explicitly if EN_DEBUG is enabled. In the (default)
case of !EN_DEBUG, the driver resets the card. Probably this case
shouldn't exist at all.
SD Simplified specification, as well as other SD and SDIO
implemenations I've examined, suggest this disclaimer may be required.
It is unclear to me exactly what the license would be for, or why it
might be required. Err on the side of caution and include this
disclaimer so anybody deploying this code can judge for themselves. I
have no further unformation about the details.
clusters. This helps quite a bit on my low end machines (improves
performance by about 300Kpps when being blasted by a hardware
packet generator).
- Include one extended f/w counter forgotten in earlier commit
Sponsored by: Myricom Inc.
- upgrade to reflect state of 1.0.0.86
- move from firmware rev 3.2 to 4.0.0
- import driver bits for offload functionality
- remove binary distribution clause from top level files as it
runs counter to the intent of purely supporting the hardware
MFC after: 3 days
back in a simulated resume instead of entering the requested suspend state.
This helps in testing drivers separately from the acpi suspend code. To
test your drivers, set debug.acpi.suspend_bounce=1 and then run
acpiconf -s3 (or 4).
MFC after: 1 day
device's, not the bridge's, softc to be used to check the
PCIB_DISABLE_MSI flag. This resulted in randomly allowing
or denying MSI interrupts based on whatever value the driver
happened to store at sizeof(device_t) bytes into its softc.
I noticed this when I stopped getting MSI interrupts
after slighly re-arranging mxge's softc yesterday.
- Remove code to use the special wc_fifo. It has been disabled by default
in our other drivers as it actually slows down transmit by a small amount
- Dynamically determine the amount of space required for the rx_done
ring rather than hardcoding it.
- Compute the number of tx descriptors we are willing to transmit per
frame as the minimum of 128 or 1/4 the tx ring size.
- Fix a typo in the tx dma tag setup which could lead to unnecessary
defragging of TSO packets (and potentially even dropping TSO packets
due to EFBIG being returned).
- Add a counter to keep track of how many times we've needed to
defragment a frame. It should always be zero.
- Export new extended f/w counters via sysctl
Sponsored by: Myricom, Inc.
- Update to the latest (1.4.18) f/w. This f/w introduces a new
receive mode which allows us to use FreeBSD's physically discontinuous
MJUM9BYTES clusters.
- Switch the driver from chaining MJUMPAGESIZE clusters to using
MJUM9BYTES clusters to avoid mbuf chaining overheads. Due to this
change, people running obsolete f/w images will be limited to an MTU of
PAGE_SIZE - 16.
- Add (disabled by default) support for Large Receive Offload.
Sponsored by: Myricom, Inc.
adapter list still capable, but only PCI-E adapters are now enabled.
The user can enable older PCI-X or PCI adapters using ifconfig.
Secondly, Arthur Hartwig pointed out my MSI change was not working
correctly, changed to something that now does. Thanks Arthur.
There was also a fundamental bug in the 82575 MSIX code, the MSIX
registers had to be mapped, opps :)
Rubber-stamped by: Pdeuskar
the power_nodriver tunable is off. pci_cfg_save() already checks the
tunable internally, and no other callers of pci_cfg_save() check the
tunable.
Reviewed by: imp
- Updated firmware to latest release (v3.4.8) to fix TSO + jumbo frame lockup
- Added MSI (hw.bce.msi_enable) and TSO (hw.bce.tso_enable) sysctls
- Fixed kernel panic when MSI is used and module is unloaded
- Added several new debug routines
- Removed slack space for RX/TX chains since it only covers sloppy coding
- Fixed a potential problem when programming jumbo MTU size in hardware
- Various other comment changes
MFC after: 4 weeks
because on at least my dc based cards there's garbage in there. The
recent changes in the resource code appears to have unmasked this
problem... At least dc now probes/attaches better than it did before.
Also, we no longer need to write to the cfg for the other registers.
legacy codepath match the 82575, without this we were seeing bridging
fail on 82546 adapters. Secondly, I have limited TSO to PCI Express
adapters, I meant to do this and it got dropped in the earlier delta.
Next, I am dropping in the latest shared code from our development
team, consensus was that this should be done frequently, so I am :)
Approved by: pdeuskar
This patch does the following:
- Remove un-necessary code that is not even compiling into the driver
under TW_OSL_NON_DMA_MEM_ALLOC_PER_REQUEST defines.
- Remove bundled firmware image and associated "files" entry for tw_cl_fwimg.c
- Remove bundled firmware flashing routines. We now have tw_update userspace
FreeBSD controller flash utility.
- Fix driver crash on load due to shared interrupt.
- Fix 2 lock leaks for Giant lock.
- Fix CCB leak.
- Add support for 9650SE controllers.
Many thanks to 3Ware/AMCC for continuing to support FreeBSD.
time workaround for problems with 82571 adapters and LAAs, one port
getting reset can cause the other to have its RAR[0] also reset,
thus overwriting an LAA. This fix works around it by also keeping
the address in the last array member.
The other bug is specific to the new 575 adapter, its transmit code
logic in handling hwassists was too crude, it broken when doing
bridges. I am much happier with the new logic,we may want to change
the legacy path at some point to something similar.
Reviewed by: pdeuskar
Approved by: pdeuskar
- Update to latest (1.4.17) firmware.
- Use the new MXGEFW_CMD_UNALIGNED_TEST (added in firmare 1.4.16) to
have the firmware tell us if the PCIe chipset supports aligned PCIe
completions.
- Hard to maintain, and frequently out of date whitelist of PCIe
chipsets known to produce aligned completions removed, as it has been
replaced in its role of selecting the correct firmware to run by the
use of MXGEFW_CMD_UNALIGNED_TEST.
- Break the dma test out of mxge_reset() and into its own function
(mxge_dma_test()) so it can be used by both the normal DMA test, and
to run the unaligned test.
- Improved support for enabling ECRCs
Sponsored by: Myricom Inc.
Seems to work on RELENG_4 through -current and also on sparc64
now. There may still be some issues with the auto attach/detach
code to sort out.
MFC after: 3 days
- http://www.intel.com/design/chipsets/specupdt/245051.htm
AC97 Soft Audio and Soft Modem Master Abort Errata
Issue:
Use of either soft audio or soft modem on an Intel® 82443MX PCISet
based platform running a 100 MHz Processor System Bus and an AC97 codec
may result in failures. The system continues to function normally while
the AC97 hardware may not resume and may require a cold-boot to
recover. As a result of the failure, the Master Abort Status bit will
be set in the audio or modem function PCI header space.
Workaround:
Force uncacheable DMA on both BDL and pcm buffers.
Tested by: Emil Holmstr|m <emil@linux.se>
- Remove explicit call to pmap_change_attr(), since we now have proper
and functional definition of BUS_DMA_NOCACHE.
- Enable PCI(e) bus snooping for non i386/amd64 as an alternative for
uncacheable DMA.
- Codecs changes:
* Analag Device -> Analog Devices, AD1988.
* New codec: VIA VT1708 and VT1709, Realtek ALC262, ALC861-VD and
ALC885.
* Various fixups for Conexant Waikiki, fix recording (read: microphone)
on various Analog Devices codecs due to vendor BIOS mess, various
quirks for several ASUS laptops/boards.
- Fix connection list handling, closely following the specification to
handle range of nids.
- Basic Jack sense polling infrastructure for possible hardwares with
broken unsolicited response interrupt.
Ideas/Submitted/Tested by: Andriy Gapon <avg@icyb.net.ua>,
#freebsd-azalia, many.
Updated copyright date to 2007.
Tested with BCM5706 A3.
Added ID for BCM5708 B2.
Removed unused driver version string.
Modified BCE_PRINTF macro to automatically fill-in the sc pointer.
Fixed a kernel panic when the driver was loaded as a module from the
command-line because the MII bus pointer was null (i.e. the MII bus
hadn't been enumerated yet).
Added fix proposed by Vladimir Ivanov <wawa@yandex-team.ru> to prevent
driver state corruption when releasing the lock during the ISR in
bce_rx_intr() to send packets up the stack.
Added new TX chain and register read sysctl interfaces for debugging.
Cleaned up formatting for various other debug routines.
Added a new statistic maintained by firmware which tracks the number
of received packets dropped because no receive buffers are available.
correct network drivers with respect to busmaster DMA, go over it
with at duster to make other aspects of it a role model:
Eliminate the pci specific softc, it serves no rational purpose.
Use convenience resource allocation/deallocation functions to save
code and errorhandling.
Switch from bus_space_{read|write}_%u() to bus_{read|write}_%u()
functions and forget about tags and handles, the resource will know
about those, should they be needed. This also eliminates a number
of inconsistently named local variables.
in comments for .c and .h files respectively. Jack may want to clean up
style or other aspects once he's up and about again, but this gets the
kernel compiling.
shared code infrastructure that is family specific and
modular. There is also support for our latest gigabit
nic, the 82575 that is MSI/X and multiqueue capable.
The new shared code changes some interfaces to the core
code but testing at Intel has been going on for months,
it is fairly stable.
I have attempted to be careful in retaining any fixes that
CURRENT had and we did not, I apologize in advance if any
thing gets clobbered, I'm sure I'll hear about it :)
Approved by pdeuskar
- Simplify the amount of work that has be done for each architecture by
pushing more of the truly MI code down into the PCI bus driver.
- Don't bind MSI-X indicies to IRQs so that we can allow a driver to map
multiple MSI-X messages into a single IRQ when handling a message
shortage.
The changes include:
- Add a new pcib_if method: PCIB_MAP_MSI() which is called by the PCI bus
to calculate the address and data values for a given MSI/MSI-X IRQ.
The x86 nexus drivers map this into a call to a new 'msi_map()' function
in msi.c that does the mapping.
- Retire the pcib_if method PCIB_REMAP_MSIX() and remove the 'index'
parameter from PCIB_ALLOC_MSIX(). MD code no longer has any knowledge
of the MSI-X index for a given MSI-X IRQ.
- The PCI bus driver now stores more MSI-X state in a child's ivars.
Specifically, it now stores an array of IRQs (called "message vectors" in
the code) that have associated address and data values, and a small
virtual version of the MSI-X table that specifies the message vector
that a given MSI-X table entry uses. Sparse mappings are permitted in
the virtual table.
- The PCI bus driver now configures the MSI and MSI-X address/data
registers directly via custom bus_setup_intr() and bus_teardown_intr()
methods. pci_setup_intr() invokes PCIB_MAP_MSI() to determine the
address and data values for a given message as needed. The MD code
no longer has to call back down into the PCI bus code to set these
values from the nexus' bus_setup_intr() handler.
- The PCI bus code provides a callout (pci_remap_msi_irq()) that the MD
code can call to force the PCI bus to re-invoke PCIB_MAP_MSI() to get
new values of the address and data fields for a given IRQ. The x86
MSI code uses this when an MSI IRQ is moved to a different CPU, requiring
a new value of the 'address' field.
- The x86 MSI psuedo-driver loses a lot of code, and in fact the separate
MSI/MSI-X pseudo-PICs are collapsed down into a single MSI PIC driver
since the only remaining diff between the two is a substring in a
bootverbose printf.
- The PCI bus driver will now restore MSI-X state (including programming
entries in the MSI-X table) on device resume.
- The interface for pci_remap_msix() has changed. Instead of accepting
indices for the allocated vectors, it accepts a mini-virtual table
(with a new length parameter). This table is an array of u_ints, where
each value specifies which allocated message vector to use for the
corresponding MSI-X message. A vector of 0 forces a message to not
have an associated IRQ. The device may choose to only use some of the
IRQs assigned, in which case the unused IRQs must be at the "end" and
will be released back to the system. This allows a driver to use the
same remap table for different shortage values. For example, if a driver
wants 4 messages, it can use the same remap table (which only uses the
first two messages) for the cases when it only gets 2 or 3 messages and
in the latter case the PCI bus will release the 3rd IRQ back to the
system.
MFC after: 1 month
pointers. A structure is more readable and less error-prone. It
also avoids problems when a function pointer doesn't have the
same width as a void pointer.
sc->mii_anegticks according to whether the respective BGE chip
supports Fast Ethernet only or also Gigabit Ethernet.
- At least the BGE chips I've tested with wedge when isolating them
so document this as the reason for setting MIIF_NOISOLATE and
remove the unused (and partially even #ifdef'ed out) isolation
related code. Add code that panics if we encounter a non-zero MII
instance as generally there's no way a PHY requiring MIIF_NOISOLATE
can be handled gracefully in a multi-PHY configuration (it's ok for
the internal PHY of single-PHY-only-NIC to not support isolation
though).
- Additionally set MIIF_NOLOOP as loopback doesn't seem to work
either and remove the #ifdef'ed out code for adding respective
media. The MIIF_NOLOOP flag currently triggers nothing but
hopefully will be respected by mii_phy_setmedia() later on.
Reviewed by: jkim, yongari
MFC after: 1 month
Blade 2500, Fire V210 and probably some other sparc64 machines.
These chips are typically not fitted with an EEPROM which means
that we have to obtain the MAC address via OFW and that some chip
tests will just always fail.
These changes are based on the respective code found in OpenBSD
with some additional info obtained from OpenSolaris and some style
suggestions by jkim@. They also have the desired side-effect of
respecting the 'local-mac-address?' system configuration variable
for the affected BGEs.
- In bge_attach() factor out calling bge_release_resources() before
going to the fail label into the fail label as well as replace a
magic 6 with ETHER_ADDR_LEN.
Reviewed by: yongari (before style changes), jkim
- Wake up DMA engine after adding a new receive buffer.
- Skip buffers which have unknown state after error.
- More rigid error detection.
MFC after: 1 week
as some combinations of chipset, controller and target do not behave
correctly when DMA is enabled for other commands.
PR: kern/103602
MFC after: 2 weeks
were never freed, but the big ring was freed twice.
-Don't supply rx hw csums for frames which are padded beyond the
length specified in the ip header. If the padding is non-zero,
the hw csum will be incorrect for such frames.
Sponsored by: Myricom
This workaround the problem in Parallels/VMWare where the emulated drivers are
slower, especially with ATA_FLUSHCACHE. The problem appears much more
frequently with ZFS which use it a lot more.
Approved: sos, pjd
specific request and thus should first try to be allocated from the
sys_resource pool. This avoids using the sys_resource pool for wildcard
requests that have bounded ranges coming from cbb(4) and Host-PCI pcib(4)
drivers.
Tested by: Andrea Bittau <a.bittau of cs.ucl.ac.uk fame>
Sleuthing by: Andrea Bittau as well
that the MSI mapping window is fixed at 0xfee00000 and the capability
does not include two more dwords used to program the address. Supporting
this mostly results in quieting spurious warnings during boot about
non-default MSI mapping windows.
- HT 2.00b also added a new HT capability type, so support that in pciconf.
MFC after: 3 days
Tested by: jmg
It seems that valid pause frames(Tx flow control) cause GMAC to hang
such that it resulted in watchdog timeout. As a work around don't
flush Rx MAC FIFO if we've received pause frames.
Tested by: Harald Schmalzbauer (h DOT schmalzbauer AT omnisec DOT de)
Under certain circumtances, if TSO is active, Yukon II generates
corrupted IP packets. All corrupted IP packets I noticed were the the
last segmented packet in a TSO request. The corrupted packet resulted
in retransmission of the damaged packet which in turn decreased network
performance dramatically.
Unfortunately it seems that there is no way to workaround this bug
as TSO is completely handled in hardware. Disable TSO until we find a
working workaround or a new silicon revision that doesn't have this
hardware bug.
supports software encrypt/decrypt.
The nuked code itself is quite problematic, as pointed out by sam@ ---
wk->wk_keyix should be replaced by the loop count.
Tested with WEP/TKIP/CCMP/no-protection.
Approved by: sam@ (mentor)
Noticed by: Hans Petter Selasky <hselasky@c2i.net>
and new SCBs were allocated on demand later if needed. This has two
problems. First, allocating SCBs involves allocating contiguous memory,
and if memory is exhausted then the VM will try to page out to satisfy
the request, leading to recursion and deadlock. The second problem is
that it can cause lock order reversals due to parts of the VM still being
under Giant.
Fix the problem be allocating the full pool at driver attach, when it is
safe to do so.
and should only be applied on certain specific card / vendor, hence the
addition of ac97_getsubvendor().
- Fix low volume issue on several MSI laptops through ALC655 quirk.
Reported/Tested by: Christian Mueller
<raptor-freebsd-multimedia@xpls.de>
MFC after: 1 week
- For ural(4):
o Fix node leakage in ural_start(), if ural_tx_mgt() fails.
o Fix mbuf leakage in ural_tx_{mgt,data}(), if usbd_transfer() fails.
o In ural_tx_{mgt,data}(), set ural_tx_data.{m,ni} to NULL, if
usbd_transfer() fails, so they will not be freed again in ural_stop().
Approved by: sam (mentor)
Group mutexes used in hwpmc(4) into 3 "types" in the sense of
witness(4):
- leaf spin mutexes---only one of these should be held at a time,
so these mutexes are specified as belonging to a single witness
type "pmc-leaf".
- `struct pmc_owner' descriptors are protected by a spin mutex of
witness type "pmc-owner-proc". Since we call wakeup_one() while
holding these mutexes, the witness type of these mutexes needs
to dominate that of "sleepq chain" mutexes.
- logger threads use a sleep mutex, of type "pmc-sleep".
Submitted by: wkoszek (earlier patch)
Previously whenever PROMISC mode turned on/off link renegotiation
occurs and it could resulted in network unavailability for serveral
seconds.(Depending on switch STP settings it could last several tens
seconds.)
Reported by: Prokofiev S.P. < proks AT logos DOT uptel DOT net >
Tested by: Prokofiev S.P. < proks AT logos DOT uptel DOT net >
The 6105M and 6102 does not have the DWORD alignment problem, so
don't m_defrag() every packet in the transmit path for those.
More stringent usage of tx-descriptor ring and its flags.
Tested on 6102 and 6105M, other chips may also be able to run
without the m_defrag() but I have neither hardware nor docs to
find out.
Sponsored by: Soekris Engineering
use to synchornize and protect all data objects that are used for that
SIM. Drivers that are not yet MPSAFE register Giant and operate as
usual. RIght now, no drivers are MPSAFE, though a few will be changed
in the coming week as this work settles down.
The driver API has changed, so all CAM drivers will need to be recompiled.
The userland API has not changed, so tools like camcontrol do not need to
be recompiled.
implement robust version of m_collapse
add support for sf_buf
add fix for m_iovappend
add calls to m_sanity under INVARIANTS
fix m_freem_vec to correctly travese the mbuf iovec chain
Yukon II generated corrupted TCP checksum for short TCP packets
that's less than 60 bytes in size(e.g. window probe packet, pure ACK
packet etc). Padding the frame with zeros to make the frame minimum
ethernet frame size didn't work at all. Instead of dropping Tx
checksum offload support we calculate TCP checksum with S/W method
when we encounter short TCP frames.
Fortunately it seems that short UDP datagrams appear to be handled
correctly by Yukon II.
While I'm here simplify ethernet/VLAN header size calculation logic.
PR: 111384
anymore. Previously it tried to access interrupt register to disable
interrupts which could result in hang if the hardware was not
properly initialized by system BIOS/ACPI.
Tested by: Benjamin Hansmann (benjamin.hansmann AT rub dot de)
MFC after: 3 days
and flags with an sxlock. This leads to a significant and measurable
performance improvement as a result of access to shared locking for
frequent lookup operations, reduced general overhead, and reduced overhead
in the event of contention. All of these are imported for threaded
applications where simultaneous access to a shared file descriptor array
occurs frequently. Kris has reported 2x-4x transaction rate improvements
on 8-core MySQL benchmarks; smaller improvements can be expected for many
workloads as a result of reduced overhead.
- Generally eliminate the distinction between "fast" and regular
acquisisition of the filedesc lock; the plan is that they will now all
be fast. Change all locking instances to either shared or exclusive
locks.
- Correct a bug (pointed out by kib) in fdfree() where previously msleep()
was called without the mutex held; sx_sleep() is now always called with
the sxlock held exclusively.
- Universally hold the struct file lock over changes to struct file,
rather than the filedesc lock or no lock. Always update the f_ops
field last. A further memory barrier is required here in the future
(discussed with jhb).
- Improve locking and reference management in linux_at(), which fails to
properly acquire vnode references before using vnode pointers. Annotate
improper use of vn_fullpath(), which will be replaced at a future date.
In fcntl(), we conservatively acquire an exclusive lock, even though in
some cases a shared lock may be sufficient, which should be revisited.
The dropping of the filedesc lock in fdgrowtable() is no longer required
as the sxlock can be held over the sleep operation; we should consider
removing that (pointed out by attilio).
Tested by: kris
Discussed with: jhb, kris, attilio, jeff
Defer mbuf allocation and initialization until after data has already been
received in a cluster
This reduces cpu utilization somewhat, but it only improves the rx path.
Recent changes to TCP appear to make us rate limited by the TX path.
This is the first step in reducing mbuf management overhead for manipulating
clusters.
MFC after: 3 days
been defragged and had their headers in the same cluster as their
payload would be fed to the NIC in header-sized chunks, and would
likely exceed the number of available transmit descriptors.
- If a TSO frame exceeds the number of available transmit descriptors,
don't leak busdmma resources when freeing it.
Sponsored by: Myricom Inc.
in the putc() method. Likewise, in the getc() method, don't check for
received characters with an interval defined in terms of the baudrate.
In both cases it works equally well to implement a fixed delay. More
importantly, it avoids calculating a delay that's roughly 1/10th the
time it takes to send/receive a character. The calculation is costly
and happens for every character sent or received, affecting low-level
console or debug port performance significantly. Secondly, when the
RCLK is not available or unreliable, the delays could disrupt normal
operation.
The fixed delay is 1/10th the time it takes to send a character at
230400 bps.
it obtained through the uart_class structure. This allows us
to declare the uart_class structure as weak and as such allows
us to reference it even when it's not compiled-in.
It also allows is to get the uart_ops structure by name, which
makes it possible to implement the dt tag handling in uart_getenv().
The side-effect of all this is that we're using the uart_class
structure more consistently which means that we now also have
access to the size of the bus space block needed by the hardware
when we map the bus space, eliminating any hardcoding.
execution should help us avoiding potential deadlock and illegal locking
while sleeping in various mixer -> usb calls. To enable it, use
hint.uaudio.%d.async="1" or sysctl dev.uaudio.%d.async=1. Default is
disable, to remain compatible with old behaviour (with slight risk of
potential deadlock).
When the linux port changes were imported which split the
target command list to be separate from the initiator command
list and the handle format changed to encode a type in the handle
the implications to the function isp_handle_index (which only
the NetBSD/OpenBSD/FreeBSD ports use) were overlooked.
The fault is twofold: first, the index into the DMA maps
in isp_pci is wrong because a target command handle with
the type bit left in place caused a bad index (and panic)
into dma map. Secondly, the assumption of the array
of DMA maps in either PCS or SBUS attachment structures is
that there is a linear mapping between handle index and
DMA map index. This can no longer be true if there are
overlapping index spaces for initiator mode and target
mode commands.
These changes bandaid around the problem by forcing us
to not have simultaneous dual roles and doing the appropriate
masking to make sure things are indexed correctly. A longer
term fix is being devloped.
obtaining and releasing shared and exclusive locks. The algorithms for
manipulating the lock cookie are very similar to that rwlocks. This patch
also adds support for exclusive locks using the same algorithm as mutexes.
A new sx_init_flags() function has been added so that optional flags can be
specified to alter a given locks behavior. The flags include SX_DUPOK,
SX_NOWITNESS, SX_NOPROFILE, and SX_QUITE which are all identical in nature
to the similar flags for mutexes.
Adaptive spinning on select locks may be enabled by enabling the
ADAPTIVE_SX kernel option. Only locks initialized with the SX_ADAPTIVESPIN
flag via sx_init_flags() will adaptively spin.
The common cases for sx_slock(), sx_sunlock(), sx_xlock(), and sx_xunlock()
are now performed inline in non-debug kernels. As a result, <sys/sx.h> now
requires <sys/lock.h> to be included prior to <sys/sx.h>.
The new kernel option SX_NOINLINE can be used to disable the aforementioned
inlining in non-debug kernels.
The size of struct sx has changed, so the kernel ABI is probably greatly
disturbed.
MFC after: 1 month
Submitted by: attilio
Tested by: kris, pjd
that the driver clock is identical to the processor or bus clock.
This is the case for the PowerQUICC processor. When the clock is
high enough, overflows happen in the calculation of the time it
takes to send 1/10 of a character, used in delay loops. Fix the
overflows so as to fix bugs in the delay loops that can cause either
insufficient delays or excessive delays.
