Re: ATA failure regression in kernel 4.2

[Alexander Holler <holler@ahsoftware.de>]

Am 31.07.2015 um 12:42 schrieb Jiang Liu:
>
>
> On 2015/7/31 18:38, Alexander Holler wrote:
>> Am 28.07.2015 um 20:37 schrieb Alexander Holler:
>>
>>> Same problem here with the AMD SATA controller (1022:7801). It failed to
>>> identify the second disk when using 4.2-rc4.
>>>
>>> nointremap helped, nothing else tested
>>
>> A quick look at the change history and two tests are suggesting that the
>> right one already is in cc.
>>
>> commit
>>
>> f7fa7ae x86/irq: Avoid memory allocation in __assign_irq_vector()
>>
>> fails while
>>
>> af87bae x86/htirq: Use new irqdomain interfaces to allocate/free IRQ
>>
>> works. Inbetween almost all changes are from Jiang Liu.
>>
>> Haven't bisected it up to now but might do it later.
> Hi Alexander,
> 	Really thanks for you help. I have debugged this regression
> for a whole day without any progress:( Everything seems correct to me,
> and I have no hardware to reproduce and debug it. Still need more time:(
 > Thanks!
 > Gerry
 >

There's nothing to excuse. It wasn't a flame, just a determination. I'm 
sorry if it might have sounded like a flame. Only people which are doing 
nothing don't make failures, and doing nothing often is a failure 
itself. But I'm drifting aways.

I've now bisected those commits I've had already identified by other 
means down to the commit 52f518a3a7c2f80551a38d38be28bc9f335e713c 
"x86/MSI: Use hierarchical irqdomains to manage MSI interrupts".

That is the one where 4.2-rc* fails here first.

Regards,

Alexander Holler