[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: [Xen-devel] [PATCH 0/5] x86: EPT/MTRR interaction adjustments and cleanup

> -----Original Message-----
> From: Jan Beulich [mailto:JBeulich@xxxxxxxx]
> Sent: Tuesday, February 25, 2014 6:21 PM
> To: xen-devel
> Cc: Xu, Dongxiao; Dong, Eddie; Nakajima, Jun; Zhang, Yang Z; Keir Fraser
> Subject: [PATCH 0/5] x86: EPT/MTRR interaction adjustments and cleanup
> 1: x86/hvm: refine the judgment on IDENT_PT for EMT
> 2: x86/HVM: fix memory type merging in epte_get_entry_emt()
> 3: x86/HVM: consolidate passthrough handling in epte_get_entry_emt()
> 4: x86/HVM: use manifest constants / enumerators for memory types
> 5: x86/HVM: adjust data definitions in mtrr.c

For 1-3 (related to logic change), they look good to me.


> With this series in place (or actually the first three patches thereof,
> as the rest is cleanup), apart from the need to fully drop the
> dependency on HVM_PARAM_IDENT_PT (see the discussion started
> at http://lists.xenproject.org/archives/html/xen-devel/2014-02/msg02150.html)
> the other main question is whether the dependency on iommu_snoop
> is really correct: I don't see why the IOMMU's snooping capability
> would affect the cachability of memory accesses - especially in the
> GPU passthrough case, RAM pages may need mapping as UC/WC
> if the GPU is permitted direct access to them - uniformly using WB
> here seems to be calling for problems.
> Signed-off-by: Jan Beulich <jbeulich@xxxxxxxx>

Xen-devel mailing list



Lists.xenproject.org is hosted with RackSpace, monitoring our
servers 24x7x365 and backed by RackSpace's Fanatical Support®.