Re: [Xen-devel] [PATCH v7 00/28] Add ITS support

[Vijay Kilari <vijay.kilari@xxxxxxxxx>]

On Fri, Sep 18, 2015 at 7:21 PM, Julien Grall <julien.grall@xxxxxxxxxx> wrote:
> On 18/09/15 14:08, vijay.kilari@xxxxxxxxx wrote:
>> From: Vijaya Kumar K <Vijaya.Kumar@xxxxxxxxxxxxxxxxxx>
>>
>> This is based on DraftG version
>> http://xenbits.xen.org/people/ianc/vits/draftG.pdf
>>
>> Following major features are supported
>>  - GICv3 ITS support for arm64 platform
>>  - Only Dom0 is supported. For DomU pci passthrough feature
>>    is required.
>>
>> Changes in v7:
>>
>>  - Rebased to latest staging branch.
>>  - Compiled all the patches individually for both arm32 and arm64
>>  - Patch xen-arm-ITS-implement-hw_irq_controller-for-LPIs.patch
>>    split into two patches.
>>       - xen-arm-ITS-implement-hw_irq_controller-for-LPIs.patch and
>>       - xen-arm-ITS-Plumbing-hw_irq_controller-for-LPIs.patch.
>>  - Moved patch xen-arm-ITS-Add-GITS-registers-emulation.patch
>>    before vITS compilation.
>>  - Merged xen-arm-ITS-Enable-virtual-ITS-driver.patch with
>>    patch xen-arm-ITS-Export-ITS-info-to-Virtual-ITS.patch.
>>  - Dropped patch xen-arm-ITS-Add-32-bit-access-to-GICR_TYPER.patch.
>>    This is only minor change. Hence merged with
>>    xen-arm-ITS-Add-GICR-register-emulation.patch.
>>  - Dropped patch xen-arm-ITS-Introduce-helper-to-get-number-of-event-.patch
>>  - Dropped gic_lpi_supported helper function.
>>  - Changed LPI property table usage and handling.
>>  - Added support to pass nr_lpis from bootargs.
>>  - Added support to enable/disable its support from bootargs.
>>  - Fixed issues around freeing of its_device on error.
>>  - Introduced vits.h file
>>  - Dropped xen-dt-Handle-correctly-node-with-interrupt-map-in-d.patch
>>    from this series as it is already merged.
>>
>>   Could not share code via Github. Network is very slow.
>>   Please, let me know if required.
>
> One of the major reason to provide a branch is to know exactly your
> baseline.
>
> In this case, your baseline is not staging because I'm not able to apply
> cleanly this branch when there was no {,v}gic change at all in the last
> 2 months.
>
> I bet it's staging + my vgic series + some of your modification to this
> series.

  Yes it depends on your vgic series. New branch 'staging_its_v7" is
pushed to github

https://github.com/vijaykilari/its_v6.git

>
> In general, you should mention what is the baseline, if you not staging,
> so people can apply your series to try it and understand the changes.
>
> So please provide a git branch with this series and required patch applied.
>
>> Changes in v6:
>>
>>  - Rebased to latest staging branch.
>>  - Compiled all the patches individually for both arm32 and arm64
>>  - Split the patch "xen/arm: ITS: Allocate irq descriptors for LPIs" into 
>> two.
>>    One for allocating LPI irq_desc and other patch for allocating 
>> pending_irq desc
>>    for LPIs
>>  - Following new patches are introduced
>>      1) xen/arm: Rename NR_IRQs and vgic_num_irqs helper function
>>      2) xen/arm: ITS: Introduce msi_desc for LPIs
>>      3) xen/arm: Move vgic locking inside get_irq_priority callback
>>      4) xen/arm: ITS: Store LPIs allocated and IRQ ID bits per domain
>>      5) xen/arm: ITS: Introduce helper to get number of event IDs
>>      6) xen/arm: ITS: Add virtual ITS availability check helper
>>      7) xen/arm: ITS: Add 32-bit access to GICR_TYPER)
>>      8) xen/arm: ITS: Allocate pending_lpi descriptors for LPIs
>>
>>  - Based on below patch set
>>      http://lists.xen.org/archives/html/xen-devel/2015-08/msg00168.html
>>
>>  Some Major TODOs:
>>    1) Avoid making vits_process_cmd() static in later point of time
>>    2) How to handle LPI that does not have LPI config table entry.
>>    3) Enable/disable ITS to Dom0
>
> The TODO in the middle of the changelog is a bit strange. Is the current
> TODO list or the one of the previous version?

These are previous version TODOs that are under changes in v6.

>
> I was also expected to see in the TODO  some item mentioned by Ian on
> the previous version [1] such as:
>
>    C. The issue regarding what to do with physical LPIs which arrive
> before the guest has mapped the corresponding event in the vits or which
> is masked

yes, I will add this TODO to v7.

>
> Regards,
>
> [1]
> http://lists.xenproject.org/archives/html/xen-devel/2015-09/msg01134.html
>
> --
> Julien Grall

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