ABI change proposal before the 2.6.22 freeze

ABI change proposal before the 2.6.22 freeze

[Gregory Haskins]

(assuming that hasn't happened already?)

One thing that would make my life easier going forward would be if we could separate out the interrupt related state from the [gs]et_sregs() calls.  What I propose is that we make two new calls (say [gs]et_pic).  For "level-0" mode this will be saving/restoring the 256-bit "userint" array.  For other modes, it will be saving whatever is relevant for the APIC (registers, etc).  The userspace code would be aware of which mode it is operating under anyway, since it has to enable the feature.

The benefit here is that this is a simple change that we can slip in now before the ABI is frozen.  IIUC, we have already been changing the ABI in the trunk recently anyway.  I am just thinking that the LAPIC work I am doing probably wont make it in before the freeze, but this small change surely could.

Thoughts?  I can put together a patch if there is acceptance of the idea.

-Greg

-------------------------------------------------------------------------
This SF.net email is sponsored by DB2 Express
Download DB2 Express C - the FREE version of DB2 express and take
control of your XML. No limits. Just data. Click to get it now.
http://sourceforge.net/powerbar/db2/

Re: ABI change proposal before the 2.6.22 freeze

[Dor Laor]

>(assuming that hasn't happened already?)

You're right on time..

>
>One thing that would make my life easier going forward would be if we
could
>separate out the interrupt related state from the [gs]et_sregs() calls.
>What I propose is that we make two new calls (say [gs]et_pic).  For
"level-
>0" mode this will be saving/restoring the 256-bit "userint" array.  For
>other modes, it will be saving whatever is relevant for the APIC
>(registers, etc).  The userspace code would be aware of which mode it
is
>operating under anyway, since it has to enable the feature.
>
>The benefit here is that this is a simple change that we can slip in
now
>before the ABI is frozen.  IIUC, we have already been changing the ABI
in
>the trunk recently anyway.  I am just thinking that the LAPIC work I am
>doing probably wont make it in before the freeze, but this small change
>surely could.

Sounds logical and helpful.


>Thoughts?  I can put together a patch if there is acceptance of the
idea.

- In that case, the kvm_run should be checked too, the apic_base, cr8
are synced there. I believe that when running with level-1, there won't
be a need of them no more. The reason we added them there is to stop
calling [gs]et_pic every run ioctl.
- A Patch is always welcomed ;)

>-Greg
>
>-----------------------------------------------------------------------
--
>This SF.net email is sponsored by DB2 Express
>Download DB2 Express C - the FREE version of DB2 express and take
>control of your XML. No limits. Just data. Click to get it now.
>http://sourceforge.net/powerbar/db2/
>_______________________________________________
>kvm-devel mailing list
>kvm-devel-5NWGOfrQmneRv+LV9MX5uipxlwaOVQ5f@public.gmane.org
>https://lists.sourceforge.net/lists/listinfo/kvm-devel

-------------------------------------------------------------------------
This SF.net email is sponsored by DB2 Express
Download DB2 Express C - the FREE version of DB2 express and take
control of your XML. No limits. Just data. Click to get it now.
http://sourceforge.net/powerbar/db2/

Re: ABI change proposal before the 2.6.22 freeze

[Avi Kivity]

Gregory Haskins wrote:
> (assuming that hasn't happened already?)
>
> One thing that would make my life easier going forward would be if we could separate out the interrupt related state from the [gs]et_sregs() calls.  What I propose is that we make two new calls (say [gs]et_pic).  For "level-0" mode this will be saving/restoring the 256-bit "userint" array.  For other modes, it will be saving whatever is relevant for the APIC (registers, etc).  The userspace code would be aware of which mode it is operating under anyway, since it has to enable the feature.
>   

While we can slip in such a change to mainline, we're committed to 
supporting the current ABI for Fedora 7, so there's not much point.

However, I don't see an issue.  You could specify that when the feature 
is enabled, the irq state in get/set sregs is ignored, and the new 
functions are to be used.

-- 
error compiling committee.c: too many arguments to function


-------------------------------------------------------------------------
This SF.net email is sponsored by DB2 Express
Download DB2 Express C - the FREE version of DB2 express and take
control of your XML. No limits. Just data. Click to get it now.
http://sourceforge.net/powerbar/db2/