When my PCI a/d converter is placed in a slot such that it shares its interrupt with
the driver for the onboard ethernet controller, I see interrupt latencies of up to 900 microseconds.
When it is moved to another slot such that it doesnt share its interrupt with the
ethernet controller driver the latencies return to normal RTP published specs.
This is with 6.2NC. I dont know how to determine exactly which ethernet driver is running
(pidin or sin shows only ‘io-net’). A ‘pci -v’ says the controller has device id "9200h,
3C905C-TX’.
The machine is a Dell Optiplex GX400.
Art Hays <> avhays@nih.gov> > wrote in article <aftpv5$rh3$> 1@inn.qnx.com> >…
[…]
This is with 6.2NC. I dont know how to determine exactly which ethernet driver is running
(pidin or sin shows only ‘io-net’).
pidin arg
???
Sorry, it doesn’t tell you about driver… but you could ‘pidin mem’ and see appropriate
devn-xxxx.so in list of shared objects of io-net (exactly after all threads of io-net)
Sorry, it doesn’t tell you about driver… but you could ‘pidin mem’ and see appropriate
devn-xxxx.so in list of shared objects of io-net (exactly after all threads of io-net)
Eduard.
ed1k at ukr dot net
nicinfo
or, more specifically
nicinfo /dev/io-net/en0
RealTek 8139 Ethernet Controller
Physical Node ID … 00A0D2 11317F
Current Physical Node ID … 00A0D2 11317F
Media Rate … 100.00 Mb/s half-duplex UTP
MTU … 1514
Lan … 0
I/O Port Range … 0xEC00 → 0xECFF
Hardware Interrupt … 0xB
Promiscuous … Disabled
Multicast … Enabled
I would appreciate some more info about the effects of this problem so I can
implement a workaround:
1.) I assume that the ethernet driver on interrupt 11 is masking any further interrupts
on 11 from being answered. What I see on a scope is that INTA# from my a/d is being asserted
for almost 1 millisecond before it is answered. Is this assumption correct? (this machine has
no ISA slots).
2.) When the a/d slot is changed so that it is assigned interrupt 10 I never see this long latency.
So I
assume that the ethernet interrupt routine doesnt lock out interrupt 10 from being answered. Is
this
correct?
3.) Since interrupt 10 has a higher priority than interrupt 11, does this mean that all interrupts
with lower priority than 11 would also be blocked by the ethernet driver?
When my PCI a/d converter is placed in a slot such that it shares its interrupt with
the driver for the onboard ethernet controller, I see interrupt latencies of up to 900
microseconds.
When it is moved to another slot such that it doesnt share its interrupt with the
ethernet controller driver the latencies return to normal RTP published specs.
This is with 6.2NC. I dont know how to determine exactly which ethernet driver is running
(pidin or sin shows only ‘io-net’). A ‘pci -v’ says the controller has device id "9200h,
3C905C-TX’.
The machine is a Dell Optiplex GX400.