xpp: a massive backport from DAHDI. From Xorcom branch-rel-6839-r6908 .

Sun Mar  1 2009 Oron Peled <oron@actcom.co.il> - xpp.r6795
  * Fix cases where the command_queue overflowed during initialization.
    - Also add a 'command_queue_length' parameter to xpp.ko
  * More migrations to sysfs:
    - Add a 'transport' attribute to our astribank devices which
      points to the usb device we use. E.g:
       /sys/bus/astribanks/devices/xbus-00/transport is symlinked to
       ../../../../../../devices/pci0000:00/0000:00:10.4/usb5/5-4
    - Move /proc/xpp/XBUS-??/XPD-??/span to
      /sys/bus/xpds/devices/??:?:?/span
    - Migrate from /proc/xpp/sync to:
      /sys/bus/astribanks/drivers/xppdrv/sync
    - New 'offhook' attribute in:
      /sys/bus/xpds/devices/??:?:?/offhook
  * PRI: change the "timing" priority to match the convention used by
         other PRI cards -- I.e: lower numbers (not 0) have higher
	 priority.
  * FXO: 
    - Power denial: create two module parameters instead of hard-coded
      constants (power_denial_safezone, power_denial_minlen).
      For sites that get non-standard power-denial signals from central
      office on offhook.
    - Don't hangup on power-denial, just notify Dahdi and wait for
    - Fix caller-id detection for the case central office sends it before
      first ring without any indication before.
      Asterisk's desicion.
Mon, Dec  8 2008 Oron Peled <oron@actcom.co.il> - xpp.r6430
  * PRI:
    - Match our span clocking priorities (in system.conf)
      to Digium -- this is a reversal of the previous state.
      Now lower numbers (greater than 0) are better.
    - Synchronization fixes for PRI ports other than 0.
    - Fix T1 CRC for some countries (e.g: China).
  * FXS: fix bug in VMWI detection if using old asterisk
    which does not provide ZT_VMWI ioctl().
  * FXO:
    - Improve caller_id_style module parameter. This provide a workaround
      for countries that send this information without any notification
      (reverse polarity, ring, etc.)
    - Don't force on-hook upon power-denial. So, loopstart devices
      would ignore these as expected.
  * Implement a flow-control to prevent user space (init_card_* scripts)
    from pressuring our command queue.


git-svn-id: http://svn.digium.com/svn/zaptel/branches/1.4@4631 5390a7c7-147a-4af0-8ec9-7488f05a26cb

1 files changed, 10 insertions, 1 deletions

diff --git a/kernel/xpp/utils/genzaptelconf b/kernel/xpp/utils/genzaptelconf
index 163f595..5f28c50 100755
--- a/kernel/xpp/utils/genzaptelconf
+++ b/kernel/xpp/utils/genzaptelconf
@@ -79,6 +79,11 @@ SPAN_GROUP_BASE=10
 # and other devices).
 brint_overlap=no
 
+# Wait for all the Astribanks to load.
+# If you don't have latest zaptel, you may need to set this to
+# 'waitfor_xpds_internal'
+WAITFOR_XPDS=${WAITFOR_XPDS:-/usr/share/zaptel/waitfor_xpds}
+
 # a temporary directory to store whatever we need to remember.
 #
 # The main loop of genconf is run in a sub-process.
@@ -265,6 +270,10 @@ zap_reg_xpp() {
 	done
 }
 
+waitfor_xpds_internal() {
+	cat /proc/xpp/XBUS-[0-9]*/waitfor_xpds 2>/dev/null || true
+}
+
 # Initialize the Xorcom Astribank (xpp/)
 xpp_startup() {
 	# do nothing if the module xpp was not loaded, or if no 
@@ -273,7 +282,7 @@ xpp_startup() {
 	if ! grep -q 'STATUS=connected' /proc/xpp/xbuses; then return 0; fi
 
 	echo "Waiting for Astribank devices to initialize:"
-	cat /proc/xpp/XBUS-[0-9]*/waitfor_xpds 2>/dev/null || true
+	$WAITFOR_XPDS
 	
 	# overriding locales for the above two, as perl can be noisy
 	# when locales are missing.