[Date Prev][Date Next] [Thread Prev][Thread Next] [Date Index] [Thread Index]

Bug#622622: Bug#678236: linux-image-2.6.32-5-amd64: Disk IO error when running debian 6.0.x with buslogic virtual disk and 4G+ mem in a VMware VM



Hi Ben,

I just tried Debian 6.0.2 i386. The installer also chooses "amd64" kernel instead of "i686-bigmem" kernel
When I install a VM with 6G memory.
I guess bug 622622 was only fixed in CD image, but not the DVD image.
I have found both "amd64" and "i686-bigmem" kernel are in the first DVD ISO image.

root@pek2-lab-08-dhcp176:/media/cdrom0# cat /etc/debian_version
6.0.2
root@pek2-lab-08-dhcp176:/media/cdrom0# find . -name  linux-image-*
./pool/main/l/linux-2.6/linux-image-2.6.32-5-486_2.6.32-35_i386.deb
./pool/main/l/linux-2.6/linux-image-2.6.32-5-686-bigmem_2.6.32-35_i386.deb
./pool/main/l/linux-2.6/linux-image-2.6.32-5-686_2.6.32-35_i386.deb
./pool/main/l/linux-2.6/linux-image-2.6.32-5-amd64_2.6.32-35_i386.deb


Regards,
Peter Cao

----- Original Message -----
From: "Ben Hutchings" <ben@decadent.org.uk>
To: "Peter Cao" <pengzhencao@vmware.com>
Cc: "chunmei(Tracy) Huang" <cmhuang@vmware.com>, "Arvind Kumar" <arkumar@vmware.com>, 678236@bugs.debian.org
Sent: Friday, June 22, 2012 12:38:19 PM
Subject: Re: Bug#678236: linux-image-2.6.32-5-amd64: Disk IO error when running debian 6.0.x with buslogic virtual disk and 4G+ mem in a VMware VM

On Thu, 2012-06-21 at 01:28 -0700, Peter Cao wrote:
> Hi Ben,
> 
> 
> ----- Original Message -----
> From: "Ben Hutchings" <ben@decadent.org.uk>
> To: "Peter Cao" <pengzhencao@vmware.com>, 678236@bugs.debian.org
> Cc: "chunmei(Tracy) Huang" <cmhuang@vmware.com>, "Arvind Kumar" <arkumar@vmware.com>
> Sent: Thursday, June 21, 2012 12:34:07 PM
> Subject: Re: Bug#678236: linux-image-2.6.32-5-amd64: Disk IO error when running debian 6.0.x with buslogic virtual disk and 4G+ mem in a VMware VM
> 
> 
> >We offer both but the installer is supposed to pick the '686-bigmem'
> >configuration by default.
> 
> No. The installer picks amd64 kernel by default for configuration with 4G+ mem.
> And it  normal "i686" kernel if mem<4G.

Oh, now I remember this bug: <http://bugs.debian.org/622622>.  It only
affected installation without a network mirror enabled, so far as I
know.  It was supposed to be fixed in point release 6.0.2, so I don't
understand why you would still see it.

> >It would not surprise me if BusLogic was broken on 64-bit systems, but
> >presumably you would have noticed that in other 64-bit Linux guests.
> >The userland architecture (i386 vs amd64) should not make any
> >difference.
> 
> This issue can not be repro on real amd64 debian(amd64 debian kernel+ amd64 userland, installed with amd64 ISO).
> And It can not be repro  on ubuntu10.10 PAE kernel with 4G+ memory
> 
> >What I meant is that if the VMware tools do not configure a Debian 6.0
> >guest to have a PV-SCSI adapter by default, this is a poor default and
> >that is a bug (independent of any bugs in the guest kernel or driver).
> 
> The default configuration will use lsilogic disk controller for debian.

OK, so if the default works then this isn't such a disaster.

> We do not use pvscsi disk controller by default for all guest OS.
> I am not the owner for creating VM configurations but I guess this is a balance for performance and best compatibility.
> 
> >> We can use pvscsi for best performance but our customer may need the
> >> buslogic virtual adapter so we have to support this configuration.
> 
> >Why would they specifically need that when they use a guest and host
> >that support PV-SCSI?
> 
> We just support this configuration in our product and it is the user's choice to use which controller.
> PVSCSI should have best performance with vmware tools installed, but not every user install tools, though we highly recommended end user to install
> vmware tools for best IO/mem performance.
> 
> Ben, do you have any suggestion on how to debug the buslogic driver?

Sorry, I am not at all familiar with BusLogic or SCSI drivers in
general.

Ben.

-- 
Ben Hutchings
Every program is either trivial or else contains at least one bug



Reply to: