qemu-system-ppc64 easily cause file corruption

[wzis]

the qemu-system-ppc64 is used to run RHEL5.9 for IBM Power on RHEL 5.3.
Previously I was using QEMU 1.5.x for several months with no problem. But after the RHEL 5.3 host damaged, and rebuilt, now I tried both QEMU 1.6.2 and QEMU 1.7.0, found both can easily cause file corruptions. Symptoms:

* using scp to transfer a tar.bz file from the RHEL 5.3 host to the RHEL 5.9 PPC VM, found the size is correct, but the content is corrupted from the middle of the 90+ MB file. re-transfer again, got a correct file. But after untar, found some extracted files corrupted.
The extracted file corruption happened several times, and also the filesystem in the VM had corrupted several times, had to restore the boot image to recover.

[Paolo Bonzini (bonzini)]

Hi, any chance you could try the latest snapshot from git? It has a lot of PPC64 changes. Also, please pass your QEMU command line.

[wzis (wzis)]

I used the following way to start the VM:

#!/bin/bash
ifconfig -a|grep tap0 >/dev/null 2>&1 || qemu-ifup tap0

qemu-system-ppc64 -hda ppcrhel5.img -cpu POWER7 -machine type=pseries,usb=off -m 768 -nographic -net nic -net tap,ifname=tap0,script=no
qemu-ifdown tap0

[wzis (wzis)]

I found a CentOS 6.4 machine which has QEMU 1.5.3, and copied the ppcrhel5.img to that machine, and tested, found it's reliable, and no such file and filesystem easy corruption issue. So, it's the 1.6.2 and 1.7.0 have the problem.

[wzis (wzis)]

I can't test from git: on one box with git, it failed for the pixman missing. and the RHEL 5.3 doesn't have git.

[Peter Maydell (pmaydell)]

On 10 March 2014 10:00, wzis <email address hidden> wrote:
> I can't test from git: on one box with git, it failed for the pixman
> missing

Did you try the git submodule command that configure
suggests when it can't find pixman? That will pull
in and build a local copy of pixman for QEMU.

thanks
-- PMM

[Paolo Bonzini (bonzini)]

Also, RHEL5 does have git in the EPEL repository.

[wzis (wzis)]

On my RHEL 5.3 box, there are too many problems to get git work. And on the CentOS 6.4 it has many problem to get pixman, so I just can't test the latest qemu unless you also put the pixman in as 1.7.0 does.

> Date: Mon, 10 Mar 2014 10:53:48 +0000
> From: <email address hidden>
> To: <email address hidden>
> Subject: [Bug 1289898] Re: qemu-system-ppc64 easily cause file corruption
>
> Also, RHEL5 does have git in the EPEL repository.
>
> --
> You received this bug notification because you are subscribed to the bug
> report.
> https://bugs.launchpad.net/bugs/1289898
>
> Title:
> qemu-system-ppc64 easily cause file corruption
>
> Status in QEMU:
> New
>
> Bug description:
> the qemu-system-ppc64 is used to run RHEL5.9 for IBM Power on RHEL 5.3.
> Previously I was using QEMU 1.5.x for several months with no problem. But after the RHEL 5.3 host damaged, and rebuilt, now I tried both QEMU 1.6.2 and QEMU 1.7.0, found both can easily cause file corruptions. Symptoms:
>
> * using scp to transfer a tar.bz file from the RHEL 5.3 host to the RHEL 5.9 PPC VM, found the size is correct, but the content is corrupted from the middle of the 90+ MB file. re-transfer again, got a correct file. But after untar, found some extracted files corrupted.
> The extracted file corruption happened several times, and also the filesystem in the VM had corrupted several times, had to restore the boot image to recover.
>
> To manage notifications about this bug go to:
> https://bugs.launchpad.net/qemu/+bug/1289898/+subscriptions

[wzis (wzis)]

Yes, I tried that, but it failed due to some php component missing.

> Date: Mon, 10 Mar 2014 10:45:34 +0000
> From: <email address hidden>
> To: <email address hidden>
> Subject: Re: [Qemu-devel] [Bug 1289898] Re: qemu-system-ppc64 easily cause file corruption
>
> On 10 March 2014 10:00, wzis <email address hidden> wrote:
> > I can't test from git: on one box with git, it failed for the pixman
> > missing
>
> Did you try the git submodule command that configure
> suggests when it can't find pixman? That will pull
> in and build a local copy of pixman for QEMU.
>
> thanks
> -- PMM
>
> --
> You received this bug notification because you are subscribed to the bug
> report.
> https://bugs.launchpad.net/bugs/1289898
>
> Title:
> qemu-system-ppc64 easily cause file corruption
>
> Status in QEMU:
> New
>
> Bug description:
> the qemu-system-ppc64 is used to run RHEL5.9 for IBM Power on RHEL 5.3.
> Previously I was using QEMU 1.5.x for several months with no problem. But after the RHEL 5.3 host damaged, and rebuilt, now I tried both QEMU 1.6.2 and QEMU 1.7.0, found both can easily cause file corruptions. Symptoms:
>
> * using scp to transfer a tar.bz file from the RHEL 5.3 host to the RHEL 5.9 PPC VM, found the size is correct, but the content is corrupted from the middle of the 90+ MB file. re-transfer again, got a correct file. But after untar, found some extracted files corrupted.
> The extracted file corruption happened several times, and also the filesystem in the VM had corrupted several times, had to restore the boot image to recover.
>
> To manage notifications about this bug go to:
> https://bugs.launchpad.net/qemu/+bug/1289898/+subscriptions

[wzis (wzis)]

and lots of dependency issue .

[wzis (wzis)]

After got qemu 2.3.0, I tested on CentOS6.6, the qemu-system-ppc64 also causes filesystem corruption for the big endian RHEL5.9 for power. So this should confirm the qemu for ppc has bug from at least 1.7 onward.

[wzis (wzis)]

And seems no developer cares about this issue.

[Thomas Huth (th-huth)]

Is this the same issue as bug https://bugs.launchpad.net/qemu/+bug/1463812 ? If so, I think we should close one of the two tickets - no need to track an issue twice.