sata link hard resets on linux host when starting winxp guest with sata controller

[Thomas]

Hi !

I'm running Vbox 2.1.4 on debian linux kernel 2.6.28.9 with a SI3114 Sata controller. As soon as I start my WinXp guest with Sata controller config I experience quite frequent sata link hard resets showing in my hosts's syslog.

kernel: ata1: hard resetting link kernel: ata1: SATA link up 1.5 Gbps (SStatus 113 SControl 310) kernel: ata1.00: configured for UDMA/33 kernel: ata1: EH complete

There is no effect visible on the guest except for performance degradation due to the sata link resets and of course if the host crashes due to that.

Any idea what might be the reason or how to further investigate that problem ? I'd exclude the hardware bits as I've all swapped them with no effect and now finally came to notice the coincidence with the Virtualbox guest running or not.

There's no logfiles or coredumps to provide as the host system either keeps running or just stops responding to anything.

Regards, Thomas

[Thomas]

Further details: I'm running a two-disk software mirror (md) as system disk and a four-disk software raid5 as data disk. The latter one contains the Virtualbox instances and installation. The sata resets only appear with the system disk / mirror and seem to be away when I set the secondary disk faulty, remove it from the array and reboot with a single system disk only. I've got a bunch of identical disks and have swapped the system disks so I would rather rule out a hardware failure on one of the system disks.

[Frank Mehnert]

Perhaps some timing problem? Dose your CPU has VT-x or AMD-V and, if so, could you try toggling the VM option and see if that makes a difference? I cannot imagine how the guest should trigger this problem on the host. The same should happen if you choose the legacy IDE controller emulation for the guest.

[Thomas]

Don't have VT-x or AMV-V. I'll setup a second machine to test whether switching to IDE mode makes any difference. It's just when booting the VMs the resets are very frequent - on the other hand that might just make visible an existing problem on the host. Thanks!

[Thomas]

I guess the SATA issue is to be solved on the host. Nevertheless I experience on of my virtual machine (WindowsXP) just freezing quite regularly. I can still connect via VRDP but the screen is frozen. That happens quite regularly. It might have to do with the SATA issue on the host.

Having a close look at the logfiles I got notice of the following lines I'd really appreciate to get an explanation for to understand whether they state a problem or are just informational:

This one appears six times when booting the virtual machine: Guest Log: BIOS: int13_harddisk: function 02, parameters out of range 0000/0000/0001!

These ones appear once when booting: 82:54:49.298 PATM: Disabling IDT 3b patch handler 847a8dd4 82:54:49.301 PATM: patmR3RefreshPatch: succeeded to refresh patch at 847a8dd4 82:54:49.621 Guest adapter information contains unsupported type 5. The block has been skipped. 82:55:05.985 PATM: Disable block at 83b55c14 - write 83b55d02-83b55d06

These ones appear from time to time and I wonder if that indicates some problem with tickless timer on the host's linux kernel: 83:03:30.294 PIT: mode=2 count=0x4ad (1197) - 996.81 Hz (ch=0) 83:03:30.331 PIT: mode=2 count=0x2ead (11949) - 99.85 Hz (ch=0) 83:03:33.033 PIT: mode=2 count=0x4ad (1197) - 996.81 Hz (ch=0) 83:03:33.052 PIT: mode=2 count=0x2ead (11949) - 99.85 Hz (ch=0) 83:03:39.550 PIT: mode=2 count=0x4ad (1197) - 996.81 Hz (ch=0) 83:03:39.552 PIT: mode=2 count=0x2ead (11949) - 99.85 Hz (ch=0)

[Frank Mehnert]

Please reopen if still relevant with VBox 3.1.4.