diff options
author | Shaun Ruffell <sruffell@digium.com> | 2009-03-11 14:51:11 +0000 |
---|---|---|
committer | Shaun Ruffell <sruffell@digium.com> | 2009-03-11 14:51:11 +0000 |
commit | c90f59429e964b1a14ffae775957cf515d204fe1 (patch) | |
tree | 1a1babbe0d5f3923c25032644a52e6d4eb66fc27 | |
parent | 5e7cabc0425d1c8821b36a84e51d753e58568c39 (diff) |
Allow 10ms for voicebus hardware to settle after reset.
The voicebus hardware needs more time to settle after a reset. The short settle
time explains why there was frequently one IRQ miss reported in the proc file
for the spans.
Reported by: jsloan
git-svn-id: http://svn.asterisk.org/svn/dahdi/linux/trunk@6126 a0bf4364-ded3-4de4-8d8a-66a801d63aff
-rw-r--r-- | drivers/dahdi/voicebus.c | 12 |
1 files changed, 2 insertions, 10 deletions
diff --git a/drivers/dahdi/voicebus.c b/drivers/dahdi/voicebus.c index b184a1a..b34f8a6 100644 --- a/drivers/dahdi/voicebus.c +++ b/drivers/dahdi/voicebus.c @@ -6,16 +6,12 @@ * Matthew Fredrickson <creslin@digium.com>, and * Michael Spiceland <mspiceland@digium.com> * - * Copyright (C) 2007-2008 Digium, Inc. + * Copyright (C) 2007-2009 Digium, Inc. * * All rights reserved. * VoiceBus is a registered trademark of Digium. * - * \todo Make the client drivers back out gracefully when presented with a - * signal. - * \todo Modify clients to sleep with timeout when waiting for interrupt. - * \todo Check on a 64-bit CPU / Kernel */ /* @@ -655,7 +651,7 @@ vb_reset_interface(struct voicebus *vb) * The calls to setsdi above toggle the reset line of the CPLD. Wait * here to give the CPLD time to stabilize after reset. */ - mdelay(1); + msleep(10); return ((reg&0x7) == 0x4) ? 0 : -EIO; } @@ -890,10 +886,6 @@ voicebus_start(struct voicebus *vb) */ /* - * NOTE: The very first buffer after coming out of reset is used to - * prime the pump and is lost. So we do not want the client driver to - * prepare it, since it will never see the corresponding receive - * buffer. * NOTE: handle_transmit is normally only called in the context of the * deferred processing thread. Since the deferred processing thread * is known to not be running at this point, it is safe to call the |