Re: [Xen-devel] [PATCH v4 5/8] hvmloader: Correct bug in low mmio region accounting

[Ian Jackson <Ian.Jackson@xxxxxxxxxxxxx>]

Jan Beulich writes ("Re: [Xen-devel] [PATCH v4 5/8] hvmloader: Correct bug in 
low mmio region accounting"):
> On 21.06.13 at 13:19, Ian Jackson <Ian.Jackson@xxxxxxxxxxxxx> wrote:
> > But for the first device I think it may be possible for resource->base
> > not to be a multiple of the bar_sz, and in that case it might be that
> > the precalculation thinks it will fit when the actual placement
> > calculation doesn't.
> > 
> > Do you think this is possible ?
> 
> This is possible only from an abstract perspective, not in reality:
> PCI_MEM_START being 0x{f,e,c,8}0000000, PCI_MEM_END being
> 0xfc000000, and allocations starting with the biggest BARs
> (where you already correctly noted that BARs are always a power
> of 2 in size), the current base address can be misaligned only
> when the BAR size is too large to fit anyway. In which case it'll
> go into the space above 4Gb, and to that range the precalculation
> doesn't apply.

Ah.  Right.  Err, OK.  I'm convinced by this argument.

It's not a good reflection on the clarity of this code, though.
Perhaps, George, you could mention this issue in a comment or the
commit message.

But anyway, this, and 6/8,

Acked-by: Ian Jackson <ian.jackson@xxxxxxxxxxxxx>

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