Re: Commit dd6d87 breaks serial console

["Roger Pau Monné" <roger.pau@citrix.com>]

On 15/07/14 03:24, Zhang, Yang Z wrote:
> Roger Pau Monné wrote on 2014-07-15:
>> On 14/07/14 18:25, Andrew Cooper wrote:
>>> On 14/07/14 17:07, Roger Pau Monné wrote:
>>>> Hello,
>>>>
>>>> The following commit:
>>>>
>>>> commit dd6d87a4fcc8350c752e389abb3bcd8479c37555
>>>> Author: Jan Beulich <jbeulich@suse.com>
>>>> Date:   Fri Jun 20 10:26:37 2014 +0200
>>>>
>>>>     VT-d: drop redundant calls to invalidate_sync()
>>>>     
>>>>     The call tree iommu_flush_iec_index() -> __iommu_flush_iec() already
>>>>     invokes invalidate_sync(). Removing the superfluous instances at once
>>>>     allows the function to become static.
>>>>     
>>>>     Signed-off-by: Jan Beulich <jbeulich@suse.com>
>>>>     Reviewed-by: Andrew Cooper <andrew.cooper3@citrix.com>
>>>>     Acked-by: Yang Zhang <yang.z.zhang@intel.com>
>>>> Breaks serial console on my box, this is all the output I get with
>>>> this commit applied:
>>>
>>> Are you sure you have bisected this correctly?
>>>
>>> Unless there is some hardware side-effect of issuing two qinvalidate
>>> wait descriptors back-to-back, the change still looks to be good.
>>> Alternately, are we missing a hardware erratum for your system?
>>
>> Just for the record, setting iommu=no-intremap solves the problem, so
>> I'm quite sure I've bisected it right. This is on a Xeon W3550, I can
>> provide whatever other hardware information is needed.
> 
> It is so strange. The patch is really simple. Does Xen still alive when serial console breaks? Also, can you try to add invalidate_sync one by one to figure which one caused the problem?

Yes, Xen seems to work fine (apart from the serial console), I can
access Dom0 using ssh and create guests.

The invalidate_sync that prevents this problem is the one in
ioapic_rte_to_remap_entry, I have successfully removed the other
invalidate_sync in msi_msg_to_remap_entry without problems.

Roger.