Re: Can't boot new 4.4 kernel with IOMMU enabled

[Joerg Roedel <joro-zLv9SwRftAIdnm+yROfE0A@public.gmane.org>]

On Thu, Jan 14, 2016 at 09:04:08AM -0500, Mark Hounschell wrote:
> OK, I verified the 2 disks are hanging off the Marvel chips. 
> Here is the dmesg from booting with your patch and the 2 
> disks not showing up.

Does it boot with iommu=soft?

> These appear to be the 2 failing disks
> #dmesg | grep -i fail 
> [    1.899572] ata8.00: failed to IDENTIFY (INIT_DEV_PARAMS failed, err_mask=0x80)
> [    6.898365] ata7.00: failed to IDENTIFY (INIT_DEV_PARAMS failed, err_mask=0x80)
> [    7.369344] ata8.00: failed to IDENTIFY (INIT_DEV_PARAMS failed, err_mask=0x80)
> [   12.363887] ata7.00: failed to IDENTIFY (INIT_DEV_PARAMS failed, err_mask=0x80)
> [   12.842392] ata8.00: failed to IDENTIFY (INIT_DEV_PARAMS failed, err_mask=0x80)
> [   17.836405] ata7.00: failed to IDENTIFY (INIT_DEV_PARAMS failed, err_mask=0x80)

The err_mask indicates that the device fetched an invalid command, so
likely it read some garbage from memory it shouldn't read it from.
Unfortunatly there are no IO_PAGE_FAULTs which could give us any clue.

> Here are the pci-ids of the Marvel chips
> #lspci | grep -i marvel
> 07:00.0 SATA controller: Marvell Technology Group Ltd. 88SE9172 SATA 6Gb/s Controller (rev 12)
> 13:00.0 SATA controller: Marvell Technology Group Ltd. 88SE9172 SATA 6Gb/s Controller (rev 12)

Your disks seem to be connected to the SATA controler at 13:00.0, which
is in an iommu group with devices 00:15.*, 00:11.0 and 00:12.0.

For testing you could try to connect the disks to the SATA controler at
07:00.0. This one is alone in its group.

Otherwise it would be good if you could bisect the issue to find the
broken commit.



	Joerg