[PATCH RFC] pass write value to in_range pointers

2009-06-22 Thread Michael S. Tsirkin
It seems that a lot of complexity and trickiness with iosignalfd is handling the group/item relationship, which comes about because kvm does not currently let a device on the bus claim a write transaction based on the value written. This could be greatly simplified if the value written was passed

Re: [PATCH RFC] pass write value to in_range pointers

2009-06-22 Thread Gregory Haskins
Michael S. Tsirkin wrote: > It seems that a lot of complexity and trickiness with iosignalfd is > handling the group/item relationship, which comes about because kvm does > not currently let a device on the bus claim a write transaction based on the > value written. This could be greatly simplifie

Re: [PATCH RFC] pass write value to in_range pointers

2009-06-22 Thread Gregory Haskins
Michael S. Tsirkin wrote: > On Mon, Jun 22, 2009 at 11:45:00AM -0400, Gregory Haskins wrote: > >> Michael S. Tsirkin wrote: >> >>> It seems that a lot of complexity and trickiness with iosignalfd is >>> handling the group/item relationship, which comes about because kvm does >>> not current

Re: [PATCH RFC] pass write value to in_range pointers

2009-06-22 Thread Michael S. Tsirkin
On Mon, Jun 22, 2009 at 11:45:00AM -0400, Gregory Haskins wrote: > Michael S. Tsirkin wrote: > > It seems that a lot of complexity and trickiness with iosignalfd is > > handling the group/item relationship, which comes about because kvm does > > not currently let a device on the bus claim a write t

Re: [PATCH RFC] pass write value to in_range pointers

2009-06-22 Thread Michael S. Tsirkin
On Mon, Jun 22, 2009 at 12:29:10PM -0400, Gregory Haskins wrote: > Michael S. Tsirkin wrote: > > On Mon, Jun 22, 2009 at 11:45:00AM -0400, Gregory Haskins wrote: > > > >> Michael S. Tsirkin wrote: > >> > >>> It seems that a lot of complexity and trickiness with iosignalfd is > >>> handling

Re: [PATCH RFC] pass write value to in_range pointers

2009-06-22 Thread Gregory Haskins
Michael S. Tsirkin wrote: > On Mon, Jun 22, 2009 at 12:29:10PM -0400, Gregory Haskins wrote: > >> Michael S. Tsirkin wrote: >> >>> On Mon, Jun 22, 2009 at 11:45:00AM -0400, Gregory Haskins wrote: >>> >>> Michael S. Tsirkin wrote: > It seems that a

Re: [PATCH RFC] pass write value to in_range pointers

2009-06-23 Thread Avi Kivity
On 06/22/2009 07:08 PM, Michael S. Tsirkin wrote: On Mon, Jun 22, 2009 at 11:45:00AM -0400, Gregory Haskins wrote: Michael S. Tsirkin wrote: It seems that a lot of complexity and trickiness with iosignalfd is handling the group/item relationship, which comes about because kvm does not

Re: [PATCH RFC] pass write value to in_range pointers

2009-06-23 Thread Avi Kivity
On 06/22/2009 07:29 PM, Gregory Haskins wrote: We actually already have aliasing: is_write flag is used for this purpose. Yes, but read/write address aliasing is not the same thing is multi-match data aliasing. Besides, your proposal also breaks some of the natural relationship models (e

Re: [PATCH RFC] pass write value to in_range pointers

2009-06-23 Thread Gregory Haskins
Avi Kivity wrote: > On 06/22/2009 07:08 PM, Michael S. Tsirkin wrote: >> On Mon, Jun 22, 2009 at 11:45:00AM -0400, Gregory Haskins wrote: >> >>> Michael S. Tsirkin wrote: >>> It seems that a lot of complexity and trickiness with iosignalfd is handling the group/item relationship,

Re: [PATCH RFC] pass write value to in_range pointers

2009-06-23 Thread Michael S. Tsirkin
On Tue, Jun 23, 2009 at 12:04:06AM -0400, Gregory Haskins wrote: > >> It will also need to support > >> multiple matches. > >> > > > > What, signal many fds on the same address/value pair? > > I see this as a bug. Why is this a good thing to support? > > Just increases the chance of leaking t

Re: [PATCH RFC] pass write value to in_range pointers

2009-06-23 Thread Michael S. Tsirkin
On Tue, Jun 23, 2009 at 07:41:12AM -0400, Gregory Haskins wrote: > Avi Kivity wrote: > > On 06/22/2009 07:08 PM, Michael S. Tsirkin wrote: > >> On Mon, Jun 22, 2009 at 11:45:00AM -0400, Gregory Haskins wrote: > >> > >>> Michael S. Tsirkin wrote: > >>> > It seems that a lot of complexit

Re: [PATCH RFC] pass write value to in_range pointers

2009-06-23 Thread Avi Kivity
On 06/23/2009 07:04 AM, Gregory Haskins wrote: Well, for one its not very clear what the benefit of the read/write aliasing even is. ;) Apparently coalesced_mmio uses it, but even so I doubt that is for the purposes of having one device do reads while another does writes. I could be wrong, thou

Re: [PATCH RFC] pass write value to in_range pointers

2009-06-23 Thread Gregory Haskins
Michael S. Tsirkin wrote: > On Tue, Jun 23, 2009 at 12:04:06AM -0400, Gregory Haskins wrote: > It will also need to support multiple matches. >>> What, signal many fds on the same address/value pair? >>> I see this as a bug. Why is this a good thing to support?

Re: [PATCH RFC] pass write value to in_range pointers

2009-06-23 Thread Avi Kivity
On 06/23/2009 02:44 PM, Michael S. Tsirkin wrote: On Tue, Jun 23, 2009 at 12:04:06AM -0400, Gregory Haskins wrote: It will also need to support multiple matches. What, signal many fds on the same address/value pair? I see this as a bug. Why is this a good thing to support? Just

Re: [PATCH RFC] pass write value to in_range pointers

2009-06-23 Thread Gregory Haskins
Avi Kivity wrote: > On 06/23/2009 02:44 PM, Michael S. Tsirkin wrote: >> On Tue, Jun 23, 2009 at 12:04:06AM -0400, Gregory Haskins wrote: >> > It will also need to support > multiple matches. > > What, signal many fds on the same address/value pair? I see th

Re: [PATCH RFC] pass write value to in_range pointers

2009-06-23 Thread Avi Kivity
On 06/23/2009 03:01 PM, Gregory Haskins wrote: Ok, so for now I will just crank up the io_bus array, and we can address scale another day. Can I just drop patch 2/3 and let the io_bus govern the limit? So long as we have a runtime-discoverable limit, yes. -- error compiling committee.c: t