PING: Someone on the core team. (Modem Problem)

[John Baldwin]

On Monday 19 February 2007 12:20, M. Warner Losh wrote:
> In message: <45D57B29.2050408 at pacbell.net>
>             Daniel Rudy <dr2867 at pacbell.net> writes:
> : At about the time of 2/15/2007 12:14 AM, Peter Jeremy stated the 
following:
> : > On 2007-Feb-14 00:27:36 -0800, Daniel Rudy <dr2867 at pacbell.net> wrote:
> : >> Changing the slot did help.  I moved it from slot 3 to slot 1.  But, 
now
> : >> it's dropping characters with a port speed of 57600, and I am also
> : >> getting irq overrun errors from the kernel too now.
> : > 
> : > This is not good.  Interrupt latency is a bit of a sore point but the
> : > FIFO trigger level is 8 bytes so getting SILO overflows implies a
> : > latency of >1.38msec.  Does sio4 report as [FAST] in the dmesg?
> : 
> : No, it does not report as fast.  I changed the /boot/device.hints file
> : and commented out all the sio devices, since all the serial ports on the
> : system are disabled anyways.  This is the only one that is up and running.
> : 
> : Here is the verbose dmesg from a reboot that I did just now:
> : 
> : sio0: Reserved 0x100 bytes for rid 0x14 type 4 at 0xd400
> : sio0: configured irq 17 not in bitmap of probed irqs 0
> : sio0: port may not be enabled
> : sio0: irq maps: 0xcb9 0xcb9 0xcb9 0xcb9
> : sio0: <Lucent kermit based PCI Modem> port
> : 0xd400-0xd4ff,0xd800-0xd8ff,0xdc00-0xdc07 mem 0xeb105000-0xeb1050ff
> :  irq 17 at device 9.0 on pci0
> : sio0: type 16550A
> : 
> : What I don't understand is why it keeps saying that "configured irq 17
> : not in bitmap of probed irqs 0"...  What exactly does that mean?
> 
> IRQ 17 was configured for the sio device.  The sio probe provokes an
> interrupt and then reads the interrupt mask.  When it read it, the
> mask was '0' meaning nothing was interrupting...

Yeah, but the mask is meaningless for non-ISA IRQs as the mask is the
mask from the 8259As, but IRQ 17 isn't routed via those, so sio complains
about a missing interrupt even though it is working fine.

Also, in 6.2, SIO should always be able to probe a FAST since 6.1 and later 
have the changes that let FAST and non-FAST share IRQs.

-- 
John Baldwin