Re: Silly IRQ allocation on Dell 1950

From: Tai-hwa Liang <avatar_at_mmlab.cse.yzu.edu.tw>
Date: Sun, 5 Nov 2006 10:09:58 +0800 (CST)
On Fri, 3 Nov 2006, Danny Braniss wrote:
>> This is typical Dell, and it gets even worse if you have a laptop.
>> Imagine every PCI device being on the sole interrupt line that is routed
>> on the motherboard.  Growing MSI support would get around this for bce
>> and many other devices.
>>
>> Scott
>>
> why blame only Dell?, this is from a top of the line(?) IBM
>
> ibm-x3650> vmstat -i
> interrupt                          total       rate
> irq4: sio0                            15          0
> irq15: ata1                           47          0
> irq16: bce0                        28747        290
> irq17: bce1 aac0                      13          0 <-----------------
> irq23: uhci0 uhci+                     4          0
> cpu0: timer                       193265       1952
> cpu1: timer                       193080       1950
> cpu2: timer                       191979       1939
> cpu3: timer                       191978       1939
> Total                             799128       8072
>
> with all the hipe on virtualization, I'm wandering if the day
> will come and we will have virtual irqs ...

   Though not bce related, same here on an old Tyan Tiger MPX board:

# vmstat -i
interrupt                          total       rate
irq1: atkbd0                        4584          0
irq6: fdc0                             9          0
irq14: ata0                     10540911         21
irq15: ata1                     10540714         21
irq17: fxp0                     85976410        175
irq19: fxp1 xl0+               161770331        330 <-----
cpu0: timer                     97988242        200
cpu1: timer                     97988235        200
Total                          464809436        948

   I've tried to move the two fxp cards to different PCI slots but still got
the same result: one of them seems to "love" the onboard xl0. ;)

   Swapping one of the fxp with a rl could be a workaround.  But to replace
fxp with rl?  Hmm...

-- 
Cheers,
Tai-hwa Liang
Received on Sun Nov 05 2006 - 02:03:37 UTC

This archive was generated by hypermail 2.4.0 : Wed May 19 2021 - 11:39:02 UTC