Re: Disk Issues?

[Erik Berls <eberls%gmail.com@localhost>]

I added 'extras = "-s2"' to the config file.

Explicitly selecting option 1 still had the hangs.  (Started on
extract of man.tgz on a full install.)
Also went directly into sysinst, almost like the extras line was ignored.
Again, frequent stallouts of the extract, but always recovering.


Explicitly selecting option 2 in the NetBSD bootstrap (disable ACPI)
got me a supervisor fault:
warning: no /dev/console
panic: cnopen: no console device
fatal breakpoint trap in supervisor mode
trap type 1 code 0 rip ffffffff804f1575 cs 8 rflags 246 cr2
7f7ffde3d188 cpl 0 rsp ffff8000243d67f0
Stopped in pid 2.1 (sh) at      ffffffff804f1575:     leave


Explicitly selecting option 3 in the NetBSD bootstrap (disable ACPI
and SMP) got a hang after:
sysbeep0 at pcppi0
fdc0 at isa0 port 0x3f0-0x3f7 irq 6 drq 2
attimer0: attached to pcppi0

I could break in and got:
fatal breakpoint trap in supervisor mode
trap type 1 code 0 rip ffffffff804f1575 cs 8 rflags 202 cr2  0 cpl 6
rsp ffff80002395a9c8
Stopped in pid 0.9 (system) at 0xffffffff804f1575:     leave
db{0}>


Note: I've upgraded the dom0 kernel to 5.0.1 for the above.
using amd64cd-5.0.1.iso for the install image gets the same affect:
man.tgz stalls a lot, xcomp.tgz stalls at the end, xserver.tgz stalls
in the middle


Any ideas on what I should poke at?

-=erik.


On Tue, Jul 28, 2009 at 8:57 AM, Manuel 
Bouyer<bouyer%antioche.eu.org@localhost> wrote:
> On Wed, Jul 22, 2009 at 04:00:18PM -0700, Erik Berls wrote:
>> Hi,
>>
>> A quick scan of the list didn't turn anything up, but I would have
>> otherwise thought that people would have noticed this:
>>
>> I'm seeing issues with *disk* IO under HVM that doesn't exist under PV.
>>
>> Setup:
>> Hardware:
>> - Dual core AMD64
>> - 4G RAM
>> - 250G HD  (with a local mount point called /n/xen)
>> - dual ethernet
>>
>> dom0:
>> - netbsd/amd64 5.0_RC2 XEN3_DOM0
>> - 256M allocation
>>
>> domU - HVM
>> - netbsd/amd64 5.0_RC2 GENERIC
>> - 2048M allocation
>> - file:/n/xen/IMG/dev0,ioemu:hda,w  phy:/dev/wd0e,hba,w
>> file:/n/xen/NetBSD5/amd64cd.iso,ioemu:hdc,r
>>
>> domU - PV
>> - netbsd/amd64 5.0_RC2 XEN3_DOMU
>> - 2048M allocation
>> - file:/n/xen/IMG/dev0,0x03,w  phy:/dev/wd0e,0x04,w
>>
>>
>> Under HVM:
>> * On boot I get lost interrupts for the phy: drive(s) [originally the
>> first drive was a phy: as well].
>> * I can pull a disklabel off the drive, but attempting to mount
>> results in timeouts and eventual failure.
>> * file: access randomly hangs;  extracting sets on the install paused
>> midway through and only completed a couple of 10s of _minutes_ later.
>> (sysinst reported -stalled-)
>
> Looks like interrupt routing issues. Did you try disabling ACPI in the
> HVM domU ?
>
> --
> Manuel Bouyer, LIP6, Universite Paris VI.           
> Manuel.Bouyer%lip6.fr@localhost
>     NetBSD: 26 ans d'experience feront toujours la difference
> --
>



-- 
"Exactly two things have made airline travel safer since 9/11:
reinforcement of cockpit doors, and passengers who now know that they
may have to fight back. Everything else ... is security theater."
--Bruce Schneier