Konrad Rzeszutek Wilk
2013-Feb-15 16:26 UTC
What went in Linux 3.6, 3.7 and 3.8 from Xen standpoint.
Hey, I realized I hadn''t done my usual ''here is what goes in'' for about half-a-year. So catching up and doing it all at once. v3.6: - Fix a lot of bugs: Systems with MP BIOS failing, Systems with ACPI NUMA failing, FLR in xen-pciback leaving the devices unusable, 32-bit PCI sounds cards in dom0 not working, fix crashes when using acpidump, fix crashes with CONFIG_MAX_DOMAIN_PAGES=512 - Make the P2M interaction on MMIO ranges use less memory during booting (aka, Reuse existing P2M leafs). - Simplification and cleanups in the code base. Coverity fixes - Performance optimizations by caching TLS and GDT descriptors - Performance optimizations in PTE page manipulations - Xen MCE driver added (to see MCE events that Xen hypervisor gets) - Xen PCPU driver added (to online/offline physical CPUs via dom0) v3.7: - Initial support for ARM working under Xen as both guest and initial domain. - Security fixes. - Fix RCU warning, add fallback code for old hypervisors, fix memory leaks in gntdev driver, fix some pvops calls failing, Fixes in xen-[kbd|fb|blk|net|hvc]-backend to deal with CLOSED transition - Allow xen/privcmd to use v2 of MMAPBATCH command (and fixes for it) - Support Xen backends to work with paged out grants (meaning work with HVM guests that have its memory paged out) - Performance optimization in xen/privcmd for migrating guests. - Performance improvements when doing kdump for PVonHVM guests. - Xen DBGP driver added (USB EHCI debug driver) - FLR support in xen-pciback. - Support wildcards in xen-pciback.hide=(*) argument parsing. - Xen EFI support, and keyboard shift status flag. - Late usage of Xen-SWIOTLB allowing PV PCI passthrough guest to boot without ''iommu=soft'' as an argument and late initialization of SWIOTLB. - Support more than 128GB in a PV guest. - Cleanups in the initial pagetable creation. v3.8: - Persistent feature grant in xen-block system allowing greater performance. - More fixes in the Xen-pciback for wildcard parsing - Xen Processor Aggregator Device (PAD) added. - Optimizations for xen/privcmd for ARM and PVH via new hypercall (add_to_physmap_range) - Xen ARM can use the balloon driver - Fixes for vcpu onlining/offlining, grant table initialization, parsing of cpu onlining/offlining values, checks in xen-pciback, locking fixes in gtndev, fix stack corruptions, fix xen_iret checks. xen-pciback DoSing dom0 with messages, fix mmap batch ioctl error path. - Further enh to allow PVHVM backend drivers (so moving dom0 functionality in guests)
Jan Beulich
2013-Feb-15 16:47 UTC
Re: What went in Linux 3.6, 3.7 and 3.8 from Xen standpoint.
>>> On 15.02.13 at 17:26, Konrad Rzeszutek Wilk <konrad.wilk@oracle.com> wrote: > v3.7: > - Initial support for ARM working under Xen as both guest and initial > domain. > - Security fixes. > - Fix RCU warning, add fallback code for old hypervisors, fix memory leaks in > gntdev driver, fix some pvops calls failing, Fixes in > xen-[kbd|fb|blk|net|hvc]-backend to deal with CLOSED transition > - Allow xen/privcmd to use v2 of MMAPBATCH command (and fixes for it) > - Support Xen backends to work with paged out grants (meaning work with HVM > guests that have its memory paged out) > - Performance optimization in xen/privcmd for migrating guests. > - Performance improvements when doing kdump for PVonHVM guests. > - Xen DBGP driver added (USB EHCI debug driver) > - FLR support in xen-pciback. > - Support wildcards in xen-pciback.hide=(*) argument parsing. > - Xen EFI support,Where?> and keyboard shift status flag. > - Late usage of Xen-SWIOTLB allowing PV PCI passthrough guest to boot without > ''iommu=soft'' as an argument and late initialization of SWIOTLB. > - Support more than 128GB in a PV guest. > - Cleanups in the initial pagetable creation.Jan
Konrad Rzeszutek Wilk
2013-Feb-15 17:30 UTC
Re: What went in Linux 3.6, 3.7 and 3.8 from Xen standpoint.
On Fri, Feb 15, 2013 at 04:47:47PM +0000, Jan Beulich wrote:> >>> On 15.02.13 at 17:26, Konrad Rzeszutek Wilk <konrad.wilk@oracle.com> wrote: > > v3.7: > > - Initial support for ARM working under Xen as both guest and initial > > domain. > > - Security fixes. > > - Fix RCU warning, add fallback code for old hypervisors, fix memory leaks in > > gntdev driver, fix some pvops calls failing, Fixes in > > xen-[kbd|fb|blk|net|hvc]-backend to deal with CLOSED transition > > - Allow xen/privcmd to use v2 of MMAPBATCH command (and fixes for it) > > - Support Xen backends to work with paged out grants (meaning work with HVM > > guests that have its memory paged out) > > - Performance optimization in xen/privcmd for migrating guests. > > - Performance improvements when doing kdump for PVonHVM guests. > > - Xen DBGP driver added (USB EHCI debug driver) > > - FLR support in xen-pciback. > > - Support wildcards in xen-pciback.hide=(*) argument parsing. > > - Xen EFI support, > > Where?That should have said Xen VESA EFI support. Thanks for spotting that.> > > and keyboard shift status flag. > > - Late usage of Xen-SWIOTLB allowing PV PCI passthrough guest to boot without > > ''iommu=soft'' as an argument and late initialization of SWIOTLB. > > - Support more than 128GB in a PV guest. > > - Cleanups in the initial pagetable creation. > > Jan >