Re: conf/files.xen changes proposal

[Jean-Yves Migeon <jeanyves.migeon%free.fr@localhost>]

Le 03/07/2013 14:18, rudolf a écrit :
> Jean-Yves Migeon wrote:
> > Le 27/06/13 20:46, rudolf a écrit :
> > > #   compile  XEN3_DOM0_LEN6_9/cpu_ucode_amd.o
> > > /usr/B/tools/bin/x86_64--netbsd-gcc -mcmodel=kernel -mno-red-zone
> > > -ffreestanding -fno-zero-initialized-in-bss -g -O2
> > > -fno-omit-frame-pointer -g -fstack-protector -Wstack-protector 
> > > --param
> > > ssp-buffer-size=1 -fno-strict-aliasing -std=gnu99 -Werror -Wall
> > > -Wno-main -Wno-format-zero-length -Wpointer-arith 
> > > -Wmissing-prototypes
> > > -Wstrict-prototypes -Wold-style-definition -Wswitch -Wshadow 
> > > -Wcast-qual
> > > -Wwrite-strings -Wno-unreachable-code -Wno-pointer-sign 
> > > -Wno-attributes
> > > -Wextra -Wno-unused-parameter -Wold-style-definition 
> > > -Wno-sign-compare
> > > -march=core2 --sysroot=/usr/B/obj/destdir.amd64 -Damd64 -Dx86_64 
> > > -I.
> > > -I/usr/B/obj/sys/arch/amd64/compile/XEN3_DOM0_LEN6_9/xen-ma
> > > -I/usr/src/sys/../common/include -I/usr/src/sys/arch -I/usr/src/sys
> > > -nostdinc -DMSGBUFSIZE=24576 -DMAXUSERS=32 -D_KERNEL -D_KERNEL_OPT
> > > -std=gnu99 -I/usr/src/sys/lib/libkern/../../../common/lib/libc/quad
> > > -I/usr/src/sys/lib/libkern/../../../common/lib/libc/string
> > >
> > > -I/usr/src/sys/lib/libkern/../../../common/lib/libc/arch/x86_64/string
> > > -I/usr/src/sys/dist/ipf 
> > > -I/usr/src/sys/external/bsd/acpica/dist/include
> > > -c /usr/src/sys/arch/x86/x86/cpu_ucode_amd.c
> > > cc1: warnings being treated as errors
> > > In file included from 
> > > /usr/src/sys/arch/x86/x86/cpu_ucode_amd.c:46:0:
> > > ./x86/cpu_ucode.h:40:32: error: 'struct cpu_ucode_softc' declared 
> > > inside
> > > parameter list
> > > ./x86/cpu_ucode.h:40:32: error: its scope is only this definition 
> > > or
> > > declaration, which is probably not what you want
> >
> > There is definitely something wrong here: cpu_ucode should not pull 
> > out
> > amd cpu_ucode for Xen, but I cannot reproduce that on my host.
>
> I can reproduce the failure with CVS tag netbsd-6-1-RELEASE.

Okay I'll give it a look with your patch.

> > Removing "dom0ops" from the config flags has side effects.
>
> I don't know the code nor the problematic around the feature, so I
> don't see any theoretical reason why the CPU_UCODE feature should be
> included in every XEN_DOM0 kernel.

This feature is needed in case you want to upload microcode for a CPU; 
bug fixing is the main reason. However the hypervisor "sits" between the 
domain and the hardware, you cannot talk to the CPU directly. Xen 
implements the "update_microcode" hypercall for that, but your patch 
probably raises an issue where an #ifdef is likely missing when you 
comment out the option.

This operation requires privileges (modifying microcode has security 
consequences), so it is under dom0's responsibility to do it. For 
evident reasons, domU does not have it.

--
Jean-Yves Migeon