Message ID | 20130208121743.GD14526@pd.tnic |
---|---|
State | New |
Headers | show |
Hi, Am 08.02.2013 13:17, schrieb Borislav Petkov: > On Fri, Feb 08, 2013 at 12:38:03PM +0100, Andreas Färber wrote: >> Am 08.02.2013 10:30, schrieb Borislav Petkov: >>> .ext2_features = (PPRO_FEATURES & CPUID_EXT2_AMD_ALIASES) | >>> CPUID_EXT2_NX, >>> .ext3_features = CPUID_EXT3_LAHF_LM, >> >> Otherwise if someone can ack (or if you can point me to a manual), this >> looks like a good bugfix for v1.4. > > Right, I don't know what v1.4 is It's the QEMU release that we're about to make in a week. :-) http://wiki.qemu.org/Planning/1.4 > but this still needs Richard's patchset > enabling MOVBE dynamic translation in qemu to go in first before > enabling MOVBE for the n270 model. OK, so this goes through Richard's queue then, no objections. Cheers, Andreas
Il 08/02/2013 14:44, Andreas Färber ha scritto: >>> >> Otherwise if someone can ack (or if you can point me to a manual), this >>> >> looks like a good bugfix for v1.4. >> > >> > Right, I don't know what v1.4 is > It's the QEMU release that we're about to make in a week. :-) > http://wiki.qemu.org/Planning/1.4 > >> > but this still needs Richard's patchset >> > enabling MOVBE dynamic translation in qemu to go in first before >> > enabling MOVBE for the n270 model. > OK, so this goes through Richard's queue then, no objections. I'm not sure I understand the relationship between QEMU CPUID bits and TCG/KVM, but perhaps this could go in 1.4 for KVM. Does TCG have a way to mask bits that aren't supported in the translator? For example AVX that is enabled by SandyBridge. Paolo
On Fri, Feb 08, 2013 at 02:59:25PM +0100, Paolo Bonzini wrote: > Il 08/02/2013 14:44, Andreas Färber ha scritto: > >>> >> Otherwise if someone can ack (or if you can point me to a manual), this > >>> >> looks like a good bugfix for v1.4. > >> > > >> > Right, I don't know what v1.4 is > > It's the QEMU release that we're about to make in a week. :-) > > http://wiki.qemu.org/Planning/1.4 > > > >> > but this still needs Richard's patchset > >> > enabling MOVBE dynamic translation in qemu to go in first before > >> > enabling MOVBE for the n270 model. > > OK, so this goes through Richard's queue then, no objections. > > I'm not sure I understand the relationship between QEMU CPUID bits and > TCG/KVM, but perhaps this could go in 1.4 for KVM. > > Does TCG have a way to mask bits that aren't supported in the > translator? For example AVX that is enabled by SandyBridge. TCG mode automatically masks the bits not supported by TCG, see how x86_cpu_realizefn() uses TCG_FEATURES/TCG_EXT_FEATURES/etc.
Il 08/02/2013 16:10, Eduardo Habkost ha scritto: > On Fri, Feb 08, 2013 at 02:59:25PM +0100, Paolo Bonzini wrote: >> Il 08/02/2013 14:44, Andreas Färber ha scritto: >>>>>>> Otherwise if someone can ack (or if you can point me to a manual), this >>>>>>> looks like a good bugfix for v1.4. >>>>> >>>>> Right, I don't know what v1.4 is >>> It's the QEMU release that we're about to make in a week. :-) >>> http://wiki.qemu.org/Planning/1.4 >>> >>>>> but this still needs Richard's patchset >>>>> enabling MOVBE dynamic translation in qemu to go in first before >>>>> enabling MOVBE for the n270 model. >>> OK, so this goes through Richard's queue then, no objections. >> >> I'm not sure I understand the relationship between QEMU CPUID bits and >> TCG/KVM, but perhaps this could go in 1.4 for KVM. >> >> Does TCG have a way to mask bits that aren't supported in the >> translator? For example AVX that is enabled by SandyBridge. > > TCG mode automatically masks the bits not supported by TCG, see how > x86_cpu_realizefn() uses TCG_FEATURES/TCG_EXT_FEATURES/etc. Uh, but n270 doesn't support vmx. :( Too bad. Then this patch indeed is not needed in 1.4. Paolo
diff --git a/target-i386/cpu.c b/target-i386/cpu.c index 4a516e1f9e25..36a133462a8d 100644 --- a/target-i386/cpu.c +++ b/target-i386/cpu.c @@ -620,7 +620,8 @@ static x86_def_t builtin_x86_defs[] = { CPUID_ACPI | CPUID_SS | CPUID_HT | CPUID_TM | CPUID_PBE, /* Some CPUs got no CPUID_SEP */ .ext_features = CPUID_EXT_SSE3 | CPUID_EXT_MONITOR | CPUID_EXT_SSSE3 | - CPUID_EXT_DSCPL | CPUID_EXT_EST | CPUID_EXT_TM2 | CPUID_EXT_XTPR, + CPUID_EXT_DSCPL | CPUID_EXT_EST | CPUID_EXT_TM2 | CPUID_EXT_XTPR | + CPUID_EXT_MOVBE, .ext2_features = (PPRO_FEATURES & CPUID_EXT2_AMD_ALIASES) | CPUID_EXT2_NX, .ext3_features = CPUID_EXT3_LAHF_LM,