Interrupt storm on ASUS M2N-E (was: amd64/101873: Freebsd amd64 hangs on booting sata drive?)

Frank Reppin wrote:

> From: Frank Reppin <frank_at_undermydesk.org>
> To: bug-followup_at_FreeBSD.org,  habeeb_at_cfl.rr.com
> Cc:  
> Subject: Re: amd64/101873: Freebsd amd64 hangs on booting sata drive?
> 
>  I've had the same experience with my brand new and shiny
>  amd64 equipmnent...:
>  
>  Mainboard -> ASUS M2N-E (nforce 570 - *no* SLI)
>  RAM       -> DDR2 800 CL5 made by MDT (1.8V ... 2x512)
>  CPU       -> AMD64 4200+ AM2 EE
>  GPU       -> ASUS Top Silent 7600GS
>  HD        -> 2x SATA2 Seagate 160GB drives
>  
>  ... it won't boot the official 6.1 REL but booting/installing from:
>  
>    7.0-CURRENT-200608-amd64-disc1.iso
>  
>  fixed this issue for me.

Lucky man :-)

I got to the same (temporary) conclusion before realizing that, as soon 
as the SATA disks are probed, the system is plagued by an interrupt 
storm that render it almost unusable:

#vmstat -i
interrupt                          total       rate
irq1: atkbd0                           6          0
irq6: fdc0                             1          0
irq14: ata0                           47          0
irq16: fwohci0                         1          0
irq17: fxp0                         1809          0
irq21: ohci0+                          2          0
irq22: ehci0                           1          0
irq23: atapci1                 716308228     159605
cpu0: timer                      8974123       1999
Total                          725284218     161605

Booting amd64 or i386 gives the same results. On a different boot 
atapci1 was on irq20 (but I cannot remember if it was amd64 or i386), 
but the storm was unchanged. Now the PC has a i386 minimal installation.

I also tried kldload-ing if_nfe (the on-board ethernet is a Marvell): 
the driver attaches, but the responsiveness of the system deteriorates 
visibly, and I ended up with a corrupted root partition trying to build 
world with both /usr/src and /usr/obj mounted from NFS.

How could I solve/work around this issue?
I put on my home server the ACPI dump files, boot messages (verbose) and 
pciconf -lv, just in case some good soul can shed some light...

    http://stable.commit.it/m2n-e/

TIA
Angelo Turetta
Modena - Italy