Re: 2.6.16-rc2-mm1 -- BUG: warning at drivers/ieee1394/ohci1394.c:235/get_phy_reg()

[Stefan Richter <stefanr@s5r6.in-berlin.de>]

Lee Revell wrote:
> On Sat, 2006-02-11 at 01:31 +0100, Stefan Richter wrote:
>>Andrew Morton wrote:
>>>That's a -mm-only warning telling you that get_phy_reg() is doing a
>>>one-millisecond-or-more busywait while local interrupts are disabled.
>>
>>Same with set_phy_reg, ohci_soft_reset, ohci_hw_csr_reg.
[...]
> In fact I'm pretty sure we have seen reports on the linux-audio-user
> list of apps reporting underruns when a 1394 drive is accessed.

These might have different causes. The 1394 storage driver, sbp2, 
performs quite a lot of protocol handling in IRQ or soft IRQ context. I 
have been planning to move this off into process context -- not really 
because of potential latency issues but for sake of more reliable 
protocol handling. I hope I get to it RSN.

AFAICS the offending ohci1394 functions mentioned above are never called 
during normal access to an SBP-2 device. They only happen when a host 
adapter is initialized or shut down or whenever a 1394 device is plugged 
in or out (actually whenever self ID reception is completed).
-- 
Stefan Richter
-=====-=-==- --=- -=-==
http://arcgraph.de/sr/