Re: [Xen-devel] [PATCH v2 0/1] xen/device-tree: Do not remap IRQs for secondary IRQ controllers

[Wei Liu <wei.liu2@xxxxxxxxxx>]

On Mon, May 16, 2016 at 05:03:54PM +0200, Edgar E. Iglesias wrote:
> From: "Edgar E. Iglesias" <edgar.iglesias@xxxxxxxxxx>
> 
> I'm sending this as a v2 considering that I previously posted a diff
> in email discussions.
> 
> This patch fixes an DT problem with the ZynqMP PCIe node.
> Today, we try to remap IRQs regardless of if they are directly
> connected to the primary controller or not. If they are indirectly
> connected via secondary IRQ controllers, we abort the boot with
> an error.
> 
> The ZynqMP PCIe DTS bindings were upstreamed to Linux after Xen 4.6, so
> this issue is not a regression.
> PCIe is also not generally a critical feature of current ZynqMP boards,
> i.e the boards are functional without PCIe although for some users PCIe
> may be more or less critical.
> 
> As I see it we have two options:
> 
> 1. A safe option is to disable the PCIe node in the ZynqMP platform.
>    We can then fix this with calm for 4.8.
>    + It will fix the dom0 boot.
>    + Very safe and unintrusive.
>    - PCIe will not be functional.
> 
> 2. Apply this patch to 4.7
>    + It will fix the dom0 boot.
>    + PCIe will be functional.
>    - Intrusive fix. Although the fix really only affects a case that
>      otherwise would have resulted in an aborted boot of dom0.
> 
> I'm happy to submit a patch for option nr 1 if you guys feel that's
> safer/better at this point.
> 

I'm a bit wary that this patch touches common code because the issue you
describe seems to be board specific. It's also unclear why this would
make PCIe function on ZynqMP from the commit message. 

I admit I don't know much about ARM so I will leave the further review
and judgement to ARM maintainers. I don't think OSSTest will pick up bug
in this patch FWIW.

Wei.

> Best regards,
> Edgar
> 
> ChangeLog:
> 
> v1 -> v2:
> * Rephrased comment no longer mentioning bus-bridges.
> * Fixed coding style issues.
> * Added a dt_dprintk to log skipped IRQs.
> 
> 
> Edgar E. Iglesias (1):
>   xen/device-tree: Do not remap IRQs for secondary IRQ controllers
> 
>  xen/common/device_tree.c | 16 ++++++++++++++++
>  1 file changed, 16 insertions(+)
> 
> -- 
> 2.5.0
> 

_______________________________________________
Xen-devel mailing list
Xen-devel@xxxxxxxxxxxxx
http://lists.xen.org/xen-devel