Saves about a 1us on average from the interrupt handler on one test
system.
Signed-off-by: Shaun Ruffell <sruffell@digium.com>
Acked-by: Michael Spiceland <mspiceland@digium.com>
Acked-by: Russ Meyerriecks <rmeyerriecks@digium.com>
git-svn-id: http://svn.asterisk.org/svn/dahdi/linux/trunk@10239 a0bf4364-ded3-4de4-8d8a-66a801d63aff
Eliminates the need to store a copy of the flags and variety from the
global devtype.
Signed-off-by: Shaun Ruffell <sruffell@digium.com>
Acked-by: Michael Spiceland <mspiceland@digium.com>
Acked-by: Russ Meyerriecks <rmeyerriecks@digium.com>
git-svn-id: http://svn.asterisk.org/svn/dahdi/linux/trunk@10238 a0bf4364-ded3-4de4-8d8a-66a801d63aff
'memaddr' and 'memlen' is already cached as part of the underlying pci
device so the wct4xxp driver does not need to cache it again. 'canary',
'passno', 'master', and 'oct_rw_count' are unused.
In t4_span 'irqmisses' was incremented, but never used anywhere, and there is
already the irqmisses on the span itself.
Signed-off-by: Shaun Ruffell <sruffell@digium.com>
Acked-by: Michael Spiceland <mspiceland@digium.com>
Acked-by: Russ Meyerriecks <rmeyerriecks@digium.com>
git-svn-id: http://svn.asterisk.org/svn/dahdi/linux/trunk@10237 a0bf4364-ded3-4de4-8d8a-66a801d63aff
I was unable to measure a performance change with prefetching.
Signed-off-by: Shaun Ruffell <sruffell@digium.com>
Acked-by: Michael Spiceland <mspiceland@digium.com>
Acked-by: Russ Meyerriecks <rmeyerriecks@digium.com>
git-svn-id: http://svn.asterisk.org/svn/dahdi/linux/trunk@10236 a0bf4364-ded3-4de4-8d8a-66a801d63aff
We are already in hardirq context and can therefore save the cli/sti
call.
Signed-off-by: Shaun Ruffell <sruffell@digium.com>
Acked-by: Michael Spiceland <mspiceland@digium.com>
Acked-by: Russ Meyerriecks <rmeyerriecks@digium.com>
git-svn-id: http://svn.asterisk.org/svn/dahdi/linux/trunk@10235 a0bf4364-ded3-4de4-8d8a-66a801d63aff
Speeds up these calls primarily by eliminating unnecessary flushes
of writes to the PCI bus.
Before:
7.095 us | __t4_framer_in();
5.835 us | __t4_framer_out();
7.122 us | __t4_framer_in();
7.071 us | __t4_framer_in();
7.059 us | __t4_framer_in();
5.859 us | __t4_framer_out();
7.076 us | __t4_framer_in();
5.852 us | __t4_framer_out();
7.124 us | __t4_framer_in();
7.080 us | __t4_framer_in();
After:
1.694 us | __t4_framer_out();
1.686 us | __t4_framer_out();
1.695 us | __t4_framer_out();
3.182 us | __t4_framer_in();
3.283 us | __t4_framer_in();
2.889 us | __t4_framer_in();
2.942 us | __t4_framer_in();
2.951 us | __t4_framer_in();
2.906 us | __t4_framer_in();
Signed-off-by: Shaun Ruffell <sruffell@digium.com>
Acked-by: Michael Spiceland <mspiceland@digium.com>
Acked-by: Russ Meyerriecks <rmeyerriecks@digium.com>
git-svn-id: http://svn.asterisk.org/svn/dahdi/linux/trunk@10234 a0bf4364-ded3-4de4-8d8a-66a801d63aff
The 'pedanticpci' module parameter, which is always on by default,
inserts extra reads from the card in order to ensure that all writes are
properly flushed through any PCI bridges which may post the writes. The
side effect is that this takes more CPU time for registers reads and
writes, especially to the framer registers.
It is never recommended to run with pedanticpci set to 0, so I'm
removing it as a module parameter so that the default case does not take
a performance hit checking for whether the parameter is set or not.
Signed-off-by: Shaun Ruffell <sruffell@digium.com>
Acked-by: Michael Spiceland <mspiceland@digium.com>
Acked-by: Russ Meyerriecks <rmeyerriecks@digium.com>
git-svn-id: http://svn.asterisk.org/svn/dahdi/linux/trunk@10233 a0bf4364-ded3-4de4-8d8a-66a801d63aff
Also has the nice side effect of eliminating a comparison from the
interrupt handler.
Signed-off-by: Shaun Ruffell <sruffell@digium.com>
Acked-by: Michael Spiceland <mspiceland@digium.com>
Acked-by: Russ Meyerriecks <rmeyerriecks@digium.com>
git-svn-id: http://svn.asterisk.org/svn/dahdi/linux/trunk@10232 a0bf4364-ded3-4de4-8d8a-66a801d63aff
The overhead of reading the framer registers is significant and can
result in latency bumps / data drops on heavily loaded systems.
Instead of checking all spans every millisecond when they are in
alarm we will instead only check every 100 ms.
On one test system, dropped the % CPU time spent in hard interrupt
context from 10% per TDM4XX when all four spans are in alarm to
closer to 2%.
Signed-off-by: Shaun Ruffell <sruffell@digium.com>
Acked-by: Michael Spiceland <mspiceland@digium.com>
Acked-by: Russ Meyerriecks <rmeyerriecks@digium.com>
git-svn-id: http://svn.asterisk.org/svn/dahdi/linux/trunk@10231 a0bf4364-ded3-4de4-8d8a-66a801d63aff
This is the "normal" condition and can be lumped with the other
informational messages. Otherwise, just this one message might go to the
console depending on the system configuration which can be confusing.
Signed-off-by: Shaun Ruffell <sruffell@digium.com>
Acked-by: Michael Spiceland <mspiceland@digium.com>
Acked-by: Russ Meyerriecks <rmeyerriecks@digium.com>
git-svn-id: http://svn.asterisk.org/svn/dahdi/linux/trunk@10230 a0bf4364-ded3-4de4-8d8a-66a801d63aff
Commit r9750 "wct4xxp: Reduce the memory footprint of the hardware
echocanceler.", as part of reducing the non-pageable memory required to
support the VPMOCT064 and VPMOCT128, disabled caching of some hardware
echocan registers. This resulted in more physical reads to the echo
canceler. These new read transactions exposed an existing issue where
sometimes reads could be turned into writes which put a channel into an
unintended state preventing Asterisk from detecting any DTMF.
This issue is resolved by ensuring that the write signal to the Octasic
part is explicitly cleared between when the address is presented on the
bus and when the read and chip select signals are asserted. The cost is
an increase in the average time to enable / disable echo cancellation by
about 5 us on one Intel Xeon X3220 test machine (~250ns increase per
read from the Octasic part and 20 reads to enable / disable a channel).
This commit resolves a behavioral regression first introduced in 2.5.0
and 2.4.1 which could take many calls before revealing itself. This
change only affects cards with a VPMOCT128 or VPMOCT064 installed.
Signed-off-by: Shaun Ruffell <sruffell@digium.com>
Acked-by: Doug Bailey <dbailey@digium.com>
git-svn-id: http://svn.asterisk.org/svn/dahdi/linux/trunk@10220 a0bf4364-ded3-4de4-8d8a-66a801d63aff
These members are not used anywhere and are now gone.
Signed-off-by: Shaun Ruffell <sruffell@digium.com>
git-svn-id: http://svn.asterisk.org/svn/dahdi/linux/trunk@10204 a0bf4364-ded3-4de4-8d8a-66a801d63aff
The logic loops through the static cards[] array to determine timing,
but the subloop was based off the current card's numspans member.
This could cause a null dereference in the case where two cards of
different span densities were connected via timing cables.
Reported-by: Doug Bailey <dbailey@digium.com>
Signed-off-by: Russ Meyerriecks <rmeyerriecks@digium.com>
git-svn-id: http://svn.asterisk.org/svn/dahdi/linux/trunk@10140 a0bf4364-ded3-4de4-8d8a-66a801d63aff
The dual and quad span cards have a rotary switch onboard which controls
the order that cards serviced by this driver are registered with the
core of DAHDI. This commit adds a module parameter 'ignore_rotary'
which, when set to 1, causes the driver to ignore the position of the
rotary switch and only consider the physical slot when registering with
DAHDI.
Ignoring the rotary switch settings also permits the PCI device to be
bound and unbound from the driver at runtime since registration with
DAHDI no longer only happens when the module is first initialized.
By default, the rotary switch will still be used to determine
registration order. This commit does not change the default behavior.
Signed-off-by: Shaun Ruffell <sruffell@digium.com>
git-svn-id: http://svn.asterisk.org/svn/dahdi/linux/trunk@10016 a0bf4364-ded3-4de4-8d8a-66a801d63aff
This defines a NULL value for the name of an echocan as invalid. This
will allow dahdi_genconf to probe for the presence of a hardware echocan
on a channel by trying to attach one. If there is not a hardware echocan
available DAHDI_ATTACH_ECHOCAN ioctl will return -EINVAL if 'hwec' was
specified as the name of the echo canceler now.
Signed-off-by: Shaun Ruffell <sruffell@digium.com>
Acked-by: Tzafrir Cohen <tzafrir.cohen@xorcom.com>
git-svn-id: http://svn.asterisk.org/svn/dahdi/linux/trunk@9943 a0bf4364-ded3-4de4-8d8a-66a801d63aff
This change ensures that the dahdi_span_ops callbacks are not called
with any spinlocks held, and that the module is pinned in memory, and
also passes the struct file * pointer to the callbacks.
Passing the file pointer to the callbacks allows the board drivers to
check any flags on the file descriptor used to configure the
span/channel. The intent here is to allow dahdi_config to open the
/dev/dahdi/ctl file in a non-blocking mode in case there is a lengthy
processes that needs to happen as part of configuration.
Signed-off-by: Shaun Ruffell <sruffell@digium.com>
Acked-by: Tzafrir Cohen <tzafrir.cohen@xorcom.com>
git-svn-id: http://svn.asterisk.org/svn/dahdi/linux/trunk@9940 a0bf4364-ded3-4de4-8d8a-66a801d63aff
This allows dahdi-base to know whether or not there is a VPM attached to
the module as soon as it's registered as opposed to waiting for start
span. This will simplify dahdi_genconf's task of creating a valid
configuration file.
Signed-off-by: Shaun Ruffell <sruffell@digium.com>
git-svn-id: http://svn.asterisk.org/svn/dahdi/linux/trunk@9933 a0bf4364-ded3-4de4-8d8a-66a801d63aff
If we always set maintstat to DAHDI_MAINT_NONE, dahdi_base will lose
track of what it thinks the current state of the span is.
For example, if you run
$ dahdi_maint -s 1 --loopback localhost
When t4_clear_maint is called, the current maintenance mode state,
'maintstat', is set to DAHDI_MAINT_NONE. So the next time you call:
$ dahdi_maint -s 1 --loopback off
dahdi-base.c will believe that the user is trying to set the maintenance
state from DAHDI_MAINT_NONE to DAHDI_MAINT_NONE and will not actually do
anything.
Signed-off-by: Shaun Ruffell <sruffell@digium.com>
Acked-by: Russ Meyerriecks <rmeyerriecks@digium.com>
git-svn-id: http://svn.asterisk.org/svn/dahdi/linux/trunk@9932 a0bf4364-ded3-4de4-8d8a-66a801d63aff
Do not allow the interrupt handler or another CPU to change the value
between when we get the initial value and when we write the modified
value.
Also includes a minor formatting fix where braces were not aligned, and
remove 'inline' from t4_framer_in and t4_framer_out definitions.
Signed-off-by: Shaun Ruffell <sruffell@digium.com>
Acked-by: Russ Meyerriecks <rmeyerriecks@digium.com>
git-svn-id: http://svn.asterisk.org/svn/dahdi/linux/trunk@9931 a0bf4364-ded3-4de4-8d8a-66a801d63aff
Turn this option on for the entire wct4xxp driver in order to quiet the
warnings in the oct612x source files. These files are from a vendor drop and
the goal is to limit the deviations from the vendor if possible.
Signed-off-by: Shaun Ruffell <sruffell@digium.com>
Acked-by: Tzafrir Cohen <tzafrir.cohen@xorcom.com>
git-svn-id: http://svn.asterisk.org/svn/dahdi/linux/trunk@9930 a0bf4364-ded3-4de4-8d8a-66a801d63aff
gcc-4.6 now warns about variables that are set but never used. Clean up
unused variables everywhere except the oct612x subdirectory.
The oct612x should go in a separate patch in case that needs to be
pulled out into a separate project again.
Signed-off-by: Shaun Ruffell <sruffell@digium.com>
Acked-by: Tzafrir Cohen <tzafrir.cohen@xorcom.com>
git-svn-id: http://svn.asterisk.org/svn/dahdi/linux/trunk@9929 a0bf4364-ded3-4de4-8d8a-66a801d63aff
These drivers (all DAHDI drivers) should explicitly block suspend if they are
loaded.
Signed-off-by: Shaun Ruffell <sruffell@digium.com>
Acked-by: Michael Spiceland <mspiceland@digium.com>
Acked-by: Kinsey Moore <kmoore@digium.com>
git-svn-id: http://svn.asterisk.org/svn/dahdi/linux/trunk@9882 a0bf4364-ded3-4de4-8d8a-66a801d63aff
Makes the driver a little more kexec friendly when sharing an interrupt
line with another device.
Signed-off-by: Shaun Ruffell <sruffell@digium.com>
Acked-by: Michael Spiceland <mspiceland@digium.com>
Acked-by: Kinsey Moore <kmoore@digium.com>
git-svn-id: http://svn.asterisk.org/svn/dahdi/linux/trunk@9881 a0bf4364-ded3-4de4-8d8a-66a801d63aff
Since the oct6100 API consumes less stack there is no longer a danger of
overflowing our stack during load.
Signed-off-by: Shaun Ruffell <sruffell@digium.com>
git-svn-id: http://svn.asterisk.org/svn/dahdi/linux/trunk@9752 a0bf4364-ded3-4de4-8d8a-66a801d63aff
This saves ~300K of kernel memory for each quad or dual span VPM. Due
to the fact that this change disables caching of the NLP words (among
other things), the time to disable the echocans appears to have
increased by ~1ms.
Before this change:
=======================================================================
0) ! 356.498 us | vpm450m_setecmode(); <--- disable
0) ! 387.762 us | vpm450m_setecmode(); <--- enable
0) ! 429.839 us | vpm450m_setecmode(); <--- disable
]# echo 1 > /proc/sys/vm/drop_caches && free -k
total used free shared buffers cached
Mem: 2005352 228368 1776984 0 132
5540
-/+ buffers/cache: 222696 1782656
Swap: 983036 0 983036
After this change:
=======================================================================
0) ! 1109.515 us | vpm450m_setecmode(); <--- disable
0) ! 339.017 us | vpm450m_setecmode(); <--- enable
0) ! 1431.460 us | vpm450m_setecmode(); <--- disable
]# echo 1 > /proc/sys/vm/drop_caches && free -k
total used free shared buffers cached
Mem: 2005352 228080 1777272 0 112
5484
-/+ buffers/cache: 222484 1782868
Swap: 983036 0 983036
Signed-off-by: Shaun Ruffell <sruffell@digium.com>
git-svn-id: http://svn.asterisk.org/svn/dahdi/linux/trunk@9750 a0bf4364-ded3-4de4-8d8a-66a801d63aff
_t4_remove_one is now used during module initialization so it should not
be placed in the exit section of the module.
Signed-off-by: Shaun Ruffell <sruffell@digium.com>
git-svn-id: http://svn.asterisk.org/svn/dahdi/linux/trunk@9650 a0bf4364-ded3-4de4-8d8a-66a801d63aff
If dahdi_register fails, we would like the error to propagate to the
user who ran modprobe.
Signed-off-by: Shaun Ruffell <sruffell@digium.com>
git-svn-id: http://svn.asterisk.org/svn/dahdi/linux/trunk@9636 a0bf4364-ded3-4de4-8d8a-66a801d63aff
Extended reset is needed primarily with the PCI express version of the
dual and quad-span cards. Enable it by default for those cards and
allow it to be forced on or off globally for the driver as a compile
time option.
The options to force it should be able to come out if there aren't any
further reports that the compile time option needs to be set.
DAHDI-773
Signed-off-by: Shaun Ruffell <sruffell@digium.com>
git-svn-id: http://svn.asterisk.org/svn/dahdi/linux/trunk@9635 a0bf4364-ded3-4de4-8d8a-66a801d63aff
There is already a safe string copying function in all the kernels DAHDI
currently supports.
Signed-off-by: Shaun Ruffell <sruffell@digium.com>
Acked-by: Kinsey Moore <kmoore@digium.com>
git-svn-id: http://svn.asterisk.org/svn/dahdi/linux/trunk@9585 a0bf4364-ded3-4de4-8d8a-66a801d63aff
2.6.25 added the DEFINE_PCI_DEVICE_TABLE macro to make sure that the
pci_device_id tables are put into the correct section in the binary.
Convert all the places where the tables were defined to use them.
This is linux-2.6 commit where the change went in along with the
rationale: 90a1ba0c5e39eeea278f263c28ae02166c5911c8
Signed-off-by: Shaun Ruffell <sruffell@digium.com>
Acked-by: Kinsey Moore <kmoore@digium.com>
git-svn-id: http://svn.asterisk.org/svn/dahdi/linux/trunk@9584 a0bf4364-ded3-4de4-8d8a-66a801d63aff
This is a very slight performance improvement. Eliminating the need to
save the IRQ flags is probably more of a boost than grabbing and
releasing the reglock.
Signed-off-by: Shaun Ruffell <sruffell@digium.com>
git-svn-id: http://svn.asterisk.org/svn/dahdi/linux/trunk@9558 a0bf4364-ded3-4de4-8d8a-66a801d63aff
Changed the echocan factory name to a function (get_name) called to get the
name. This allows a factory to return a different name when being called in
reference to a channel such as in the case of hardware echo cancellers. To
further accommodate this change for HWEC, a new echocan_name function was
added to the span ops struct and is used in hwec_factory in dahdi-base for
all cards that support hardware echo cancellation.
Signed-off-by: Kinsey Moore <kmoore@digium.com>
Acked-by: Shaun Ruffell <sruffell@digium.com>
git-svn-id: http://svn.asterisk.org/svn/dahdi/linux/trunk@9524 a0bf4364-ded3-4de4-8d8a-66a801d63aff
This field was redundant and was only used in places where the factory's name could be used.
Signed-off-by: Kinsey Moore <kmoore@digium.com>
Acked-by: Shaun Ruffell <sruffell@digium.com>
git-svn-id: http://svn.asterisk.org/svn/dahdi/linux/trunk@9523 a0bf4364-ded3-4de4-8d8a-66a801d63aff
The loopup/loopdown T1 maintenance feature in the single through quad-span
drivers now function properly, according to AT&T TR54016, by sending a
full rate pattern down the line. T1.403 ESF/Data Link patterns are not
supported currently.
Also grouped all the maint reset register clears under a single irq lock
to crank the performance up past 11.
Signed-off-by: Russ Meyerriecks <rmeyerriecks@digium.com>
Acked-by: Shaun Ruffell <sruffell@digium.com>
Acked-by: Kinsey Moore <kmoore@digium.com>
git-svn-id: http://svn.asterisk.org/svn/dahdi/linux/trunk@9516 a0bf4364-ded3-4de4-8d8a-66a801d63aff
DAHDI_MAINT_LOOPSTOP is being removed due to the redundancy with
DAHDI_MAINT_NONE. Also removing some timing logic, as amount of time a
loopup or loopdown signal is held on the line, is now defined in
userspace with dahdi_maint.
Signed-off-by: Russ Meyerriecks <rmeyerriecks@digium.com>
Acked-by: Shaun Ruffell <sruffell@digium.com>
Acked-by: Kinsey Moore <kmoore@digium.com>
git-svn-id: http://svn.asterisk.org/svn/dahdi/linux/trunk@9515 a0bf4364-ded3-4de4-8d8a-66a801d63aff
The dahdi_span->maintq wait queue was very old and not being used so it
has been removed.
Signed-off-by: Russ Meyerriecks <rmeyerriecks@digium.com>
Acked-by: Shaun Ruffell <sruffell@digium.com>
Acked-by: Tzafrir Cohen <tzafrir.cohen@xorcom..com>
git-svn-id: http://svn.asterisk.org/svn/dahdi/linux/trunk@9514 a0bf4364-ded3-4de4-8d8a-66a801d63aff
Renamed the NMF workaround message to the more informational
crc4-multiframe (mis)alignment message.
git-svn-id: http://svn.asterisk.org/svn/dahdi/linux/trunk@9461 a0bf4364-ded3-4de4-8d8a-66a801d63aff
Push all tests for the DAHDI_FLAGBIT_NETDEV flag behind a
'dahdi_have_netdev' function so if CONFIG_DAHDI_NET is not defined the
compiler can just remove all the flag tests. Also, makes sure that the
bit is checked / set atomically.
(closes issue #9379)
Signed-off-by: Shaun Ruffell <sruffell@digium.com>
Acked-by: Kinsey Moore <kmoore@digium.com>
Acked-by: Russ Meyerriecks <rmeyerriecks@digium.com>
git-svn-id: http://svn.asterisk.org/svn/dahdi/linux/trunk@9444 a0bf4364-ded3-4de4-8d8a-66a801d63aff
The check for DEFINE_SPINLOCK was spread throughout the source tree. If
not defined we can just define it in inlucde/dahdi/kernel.h. Now
include/dahdi/kernel.h is the only place that references
SPIN_LOCK_UNLOCKED (which breaks lockdep checking if DEFINE_SPINLOCK is
otherwise defined in the kernel).
Signed-off-by: Shaun Ruffell <sruffell@digium.com>
Acked-by: Kinsey Moore <kmoore@digium.com>
Acked-by: Russ Meyerriecks <rmeyerriecks@digium.com>
Acked-by: Tzafrir Cohen <tzafrir.cohen@xorcom.com>
Review: https://reviewboard.asterisk.org/r/940/
git-svn-id: http://svn.asterisk.org/svn/dahdi/linux/trunk@9411 a0bf4364-ded3-4de4-8d8a-66a801d63aff
The registers on the device are already accessed with readl/writel and
the readchunk and writechunk are mapped into coherent DMA region. The
contents of those buffers should not be changing in the middle of any
transmit/receive prep call.
Signed-off-by: Shaun Ruffell <sruffell@digium.com>
Acked-by: Russ Meyerriecks <rmeyerriecks@digium.com>
git-svn-id: http://svn.asterisk.org/svn/dahdi/linux/trunk@9400 a0bf4364-ded3-4de4-8d8a-66a801d63aff
Normally you can see RBS bit states in dahdi_tool, but you might also
want them logged to dmesg for troubleshooting.
(issue #18025)
Signed-off-by: Shaun Ruffell <sruffell@digium.com>
Acked-by: Russ Meyerriecks <rmeyerriecks@digium.com>
git-svn-id: http://svn.asterisk.org/svn/dahdi/linux/trunk@9399 a0bf4364-ded3-4de4-8d8a-66a801d63aff
dahdi_tool was incorrectly reporting all spans to be in local timing
mode. This is because the driver tracks which span it's timing syncs to,
within the card local struct "wc". We have to manually go through and
copy timing updates to the span local structs because this is what
dahdi_tool actually reads.
dahdi-526
git-svn-id: http://svn.asterisk.org/svn/dahdi/linux/trunk@9312 a0bf4364-ded3-4de4-8d8a-66a801d63aff
This needs some more testing before it's on by default. If the card is
otherwise functioning, these messages may be confusing to the user. If
the card is not functioning, the driver can be reloaded with debug to
check for this condition.
Signed-off-by: Shaun Ruffell <sruffell@digium.com>
git-svn-id: http://svn.asterisk.org/svn/dahdi/linux/trunk@9205 a0bf4364-ded3-4de4-8d8a-66a801d63aff
