Re: [kvm-devel] [RFC] Expose host cpuid to guest
On Fri, 23 Nov 2007 12:00:00 am Avi Kivity wrote: > Mike Lampard wrote: > > With current kvm-git (commit 51727a110220681f6f43b005d069e28c58f5d151) > > (userspace is current to commit 6a385c9539f9746d7ff51ef34c064c3eba86448b) > > and the userspace portion of this patch I cannot boot a 64 bit guest > > (Mandriva 2008 x64) on my AMD x2 without -cpu host. The guest kernel > > says "Your CPU does not support long mode. Use a 32bit distribution.". > > Works fine without the userspace portion of the patch. erm.. sorry, seems the patch mis-applied @ my end, I just re-applied the userspace patch and the above problem is no longer, 1 down, 1 to go: > > > > With -cpu host, the 64bit kernel starts to boot then panics almost > > immediately: > > Code: 0f 30 e9 ed fc ff ff b8 66 00 00 00 f0 0f ab 45 08 e9 e2 fa > > RIP [ identify_cpu+0x609/0x700 > > RSP > > Kernel panic - not syncing: Attempted to kill the idle task! > > > > In the host log I see: cpu0 unhandled wrmsr: 0xc0010015, which might not > > have anything to do with this patch of course :) > > A listing/disassembly of the function will help show where the problem is. yes, of course.. I'll post it as soon as I can. Thanks for the help Mike - This SF.net email is sponsored by: Microsoft Defy all challenges. Microsoft(R) Visual Studio 2005. http://clk.atdmt.com/MRT/go/vse012070mrt/direct/01/ ___ kvm-devel mailing list kvm-devel@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/kvm-devel
Re: [kvm-devel] [RFC] Expose host cpuid to guest
On Thu, Nov 22, 2007 at 11:50:14PM +1030, Mike Lampard wrote: > On Thu, 22 Nov 2007 11:31:18 pm Dan Kenigsberg wrote: > > On Thu, Nov 22, 2007 at 10:12:04AM +1030, Mike Lampard wrote: > > > On Wed, 21 Nov 2007 09:36:31 pm Dan Kenigsberg wrote: > > > > These patches expose host CPU features (that are known to work under > > > > KVM) to guests. It makes a couple of benchmarks run faster, and > > > > generally gives kvm's user better info on its host. > > > > > > > > The kernel-space patch adds KVM_GET_SUPPORTED_CPUID ioctl to obtain the > > > > table of cpuid functions supported by the host. The user-space patch > > > > allows fine-tuning this table from the command-line. > > > > > > > > I had to define struct kvm_cpuid2, KVM_SET_CPUID2 etc., because cpuid > > > > functions are a little more complex than just function-value pairs. > > > > > > > > Dan. > > > > > > With current kvm-git (commit 51727a110220681f6f43b005d069e28c58f5d151) > > > (userspace is current to commit 6a385c9539f9746d7ff51ef34c064c3eba86448b) > > > and the userspace portion of this patch I cannot boot a 64 bit guest > > > (Mandriva 2008 x64) on my AMD x2 without -cpu host. The guest kernel > > > says "Your CPU does not support long mode. Use a 32bit distribution.". > > > Works fine without the userspace portion of the patch. > > > > Thanks for your report. I'm probably not exposing one(?) of AMD's x86_64 > > required cpuid features. However, I fail to reproduce this on Dual-Core > > AMD Opteron(tm) Processor 2210 running 2.6.18-8.1.15.el5. What is your > > host? > > 64bit 2.6.23.8 on an AMD Athlon X2 4400. > cpuid flags: fpu vme de pse tsc msr pae mce cx8 apic sep mtrr pge mca cmov > pat > pse36 clflush mmx fxsr sse sse2 ht syscall nx mmxext fxsr_opt rdtscp lm > 3dnowext 3dnow rep_good pni cx16 lahf_lm cmp_legacy svm extapic cr8_legacy > misalignsse > > If you need further info, just yell. I didnt want to clutter up the > mailinglist with the full output of cpuid, but will if it helps you. Since I'm currently out of ideas, would you try it with -no-kvm? Since current Qemu cvs has a -cpu userspace patch, it might be worthwhile to test that, too. Dan. - This SF.net email is sponsored by: Microsoft Defy all challenges. Microsoft(R) Visual Studio 2005. http://clk.atdmt.com/MRT/go/vse012070mrt/direct/01/ ___ kvm-devel mailing list kvm-devel@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/kvm-devel
Re: [kvm-devel] [RFC] Expose host cpuid to guest
Mike Lampard wrote: > With current kvm-git (commit 51727a110220681f6f43b005d069e28c58f5d151) > (userspace is current to commit 6a385c9539f9746d7ff51ef34c064c3eba86448b) and > the userspace portion of this patch I cannot boot a 64 bit guest (Mandriva > 2008 x64) on my AMD x2 without -cpu host. The guest kernel says "Your CPU > does not support long mode. Use a 32bit distribution.". Works fine without > the userspace portion of the patch. > > With -cpu host, the 64bit kernel starts to boot then panics almost > immediately: > Code: 0f 30 e9 ed fc ff ff b8 66 00 00 00 f0 0f ab 45 08 e9 e2 fa > RIP [ identify_cpu+0x609/0x700 > RSP > Kernel panic - not syncing: Attempted to kill the idle task! > > In the host log I see: cpu0 unhandled wrmsr: 0xc0010015, which might not have > anything to do with this patch of course :) > A listing/disassembly of the function will help show where the problem is. -- error compiling committee.c: too many arguments to function - This SF.net email is sponsored by: Microsoft Defy all challenges. Microsoft(R) Visual Studio 2005. http://clk.atdmt.com/MRT/go/vse012070mrt/direct/01/ ___ kvm-devel mailing list kvm-devel@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/kvm-devel
Re: [kvm-devel] [RFC] Expose host cpuid to guest
On Thu, 22 Nov 2007 11:31:18 pm Dan Kenigsberg wrote: > On Thu, Nov 22, 2007 at 10:12:04AM +1030, Mike Lampard wrote: > > On Wed, 21 Nov 2007 09:36:31 pm Dan Kenigsberg wrote: > > > These patches expose host CPU features (that are known to work under > > > KVM) to guests. It makes a couple of benchmarks run faster, and > > > generally gives kvm's user better info on its host. > > > > > > The kernel-space patch adds KVM_GET_SUPPORTED_CPUID ioctl to obtain the > > > table of cpuid functions supported by the host. The user-space patch > > > allows fine-tuning this table from the command-line. > > > > > > I had to define struct kvm_cpuid2, KVM_SET_CPUID2 etc., because cpuid > > > functions are a little more complex than just function-value pairs. > > > > > > Dan. > > > > With current kvm-git (commit 51727a110220681f6f43b005d069e28c58f5d151) > > (userspace is current to commit 6a385c9539f9746d7ff51ef34c064c3eba86448b) > > and the userspace portion of this patch I cannot boot a 64 bit guest > > (Mandriva 2008 x64) on my AMD x2 without -cpu host. The guest kernel > > says "Your CPU does not support long mode. Use a 32bit distribution.". > > Works fine without the userspace portion of the patch. > > Thanks for your report. I'm probably not exposing one(?) of AMD's x86_64 > required cpuid features. However, I fail to reproduce this on Dual-Core > AMD Opteron(tm) Processor 2210 running 2.6.18-8.1.15.el5. What is your > host? 64bit 2.6.23.8 on an AMD Athlon X2 4400. cpuid flags: fpu vme de pse tsc msr pae mce cx8 apic sep mtrr pge mca cmov pat pse36 clflush mmx fxsr sse sse2 ht syscall nx mmxext fxsr_opt rdtscp lm 3dnowext 3dnow rep_good pni cx16 lahf_lm cmp_legacy svm extapic cr8_legacy misalignsse If you need further info, just yell. I didnt want to clutter up the mailinglist with the full output of cpuid, but will if it helps you. Mike - This SF.net email is sponsored by: Microsoft Defy all challenges. Microsoft(R) Visual Studio 2005. http://clk.atdmt.com/MRT/go/vse012070mrt/direct/01/ ___ kvm-devel mailing list kvm-devel@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/kvm-devel
Re: [kvm-devel] [RFC] Expose host cpuid to guest
On Thu, Nov 22, 2007 at 10:12:04AM +1030, Mike Lampard wrote: > On Wed, 21 Nov 2007 09:36:31 pm Dan Kenigsberg wrote: > > These patches expose host CPU features (that are known to work under > > KVM) to guests. It makes a couple of benchmarks run faster, and > > generally gives kvm's user better info on its host. > > > > The kernel-space patch adds KVM_GET_SUPPORTED_CPUID ioctl to obtain the > > table of cpuid functions supported by the host. The user-space patch > > allows fine-tuning this table from the command-line. > > > > I had to define struct kvm_cpuid2, KVM_SET_CPUID2 etc., because cpuid > > functions are a little more complex than just function-value pairs. > > > > Dan. > > With current kvm-git (commit 51727a110220681f6f43b005d069e28c58f5d151) > (userspace is current to commit 6a385c9539f9746d7ff51ef34c064c3eba86448b) and > the userspace portion of this patch I cannot boot a 64 bit guest (Mandriva > 2008 x64) on my AMD x2 without -cpu host. The guest kernel says "Your CPU > does not support long mode. Use a 32bit distribution.". Works fine without > the userspace portion of the patch. Thanks for your report. I'm probably not exposing one(?) of AMD's x86_64 required cpuid features. However, I fail to reproduce this on Dual-Core AMD Opteron(tm) Processor 2210 running 2.6.18-8.1.15.el5. What is your host? Dan. - This SF.net email is sponsored by: Microsoft Defy all challenges. Microsoft(R) Visual Studio 2005. http://clk.atdmt.com/MRT/go/vse012070mrt/direct/01/ ___ kvm-devel mailing list kvm-devel@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/kvm-devel
Re: [kvm-devel] [RFC] Expose host cpuid to guest
On Wed, 21 Nov 2007 09:36:31 pm Dan Kenigsberg wrote: > These patches expose host CPU features (that are known to work under > KVM) to guests. It makes a couple of benchmarks run faster, and > generally gives kvm's user better info on its host. > > The kernel-space patch adds KVM_GET_SUPPORTED_CPUID ioctl to obtain the > table of cpuid functions supported by the host. The user-space patch > allows fine-tuning this table from the command-line. > > I had to define struct kvm_cpuid2, KVM_SET_CPUID2 etc., because cpuid > functions are a little more complex than just function-value pairs. > > Dan. With current kvm-git (commit 51727a110220681f6f43b005d069e28c58f5d151) (userspace is current to commit 6a385c9539f9746d7ff51ef34c064c3eba86448b) and the userspace portion of this patch I cannot boot a 64 bit guest (Mandriva 2008 x64) on my AMD x2 without -cpu host. The guest kernel says "Your CPU does not support long mode. Use a 32bit distribution.". Works fine without the userspace portion of the patch. With -cpu host, the 64bit kernel starts to boot then panics almost immediately: Code: 0f 30 e9 ed fc ff ff b8 66 00 00 00 f0 0f ab 45 08 e9 e2 fa RIP [ identify_cpu+0x609/0x700 RSP Kernel panic - not syncing: Attempted to kill the idle task! In the host log I see: cpu0 unhandled wrmsr: 0xc0010015, which might not have anything to do with this patch of course :) Mike - This SF.net email is sponsored by: Microsoft Defy all challenges. Microsoft(R) Visual Studio 2005. http://clk.atdmt.com/MRT/go/vse012070mrt/direct/01/ ___ kvm-devel mailing list kvm-devel@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/kvm-devel
Re: [kvm-devel] [RFC] Expose host cpuid to guest
Dan Kenigsberg wrote: > On Wed, Nov 21, 2007 at 01:52:15PM +0200, Avi Kivity wrote: > >> Dan Kenigsberg wrote: >> >>> These patches expose host CPU features (that are known to work under >>> KVM) to guests. It makes a couple of benchmarks run faster, and >>> generally gives kvm's user better info on its host. >>> >>> The kernel-space patch adds KVM_GET_SUPPORTED_CPUID ioctl to obtain the >>> table of cpuid functions supported by the host. The user-space patch >>> allows fine-tuning this table from the command-line. >>> >>> I had to define struct kvm_cpuid2, KVM_SET_CPUID2 etc., because cpuid >>> functions are a little more complex than just function-value pairs. >>> commit e9775d0a16097cfb71779cb2fb985fb3e5040dc8 >>> Author: Dan Kenigsberg <[EMAIL PROTECTED]> >>> Date: Sun Nov 18 13:55:26 2007 +0200 >>> >>> Support -cpu host option. Negotiate cpuid table with userspace. >>> > > I believe the attached two patches apply all your comments. > > Applied kernel patch. Will wait with userspace until next qemu merge (which is pending qemu-cvs regressions). -- error compiling committee.c: too many arguments to function - This SF.net email is sponsored by: Microsoft Defy all challenges. Microsoft(R) Visual Studio 2005. http://clk.atdmt.com/MRT/go/vse012070mrt/direct/01/ ___ kvm-devel mailing list kvm-devel@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/kvm-devel
Re: [kvm-devel] [RFC] Expose host cpuid to guest
On Wed, Nov 21, 2007 at 01:52:15PM +0200, Avi Kivity wrote:
> Dan Kenigsberg wrote:
>> These patches expose host CPU features (that are known to work under
>> KVM) to guests. It makes a couple of benchmarks run faster, and
>> generally gives kvm's user better info on its host.
>>
>> The kernel-space patch adds KVM_GET_SUPPORTED_CPUID ioctl to obtain the
>> table of cpuid functions supported by the host. The user-space patch
>> allows fine-tuning this table from the command-line.
>>
>> I had to define struct kvm_cpuid2, KVM_SET_CPUID2 etc., because cpuid
>> functions are a little more complex than just function-value pairs.
>> commit e9775d0a16097cfb71779cb2fb985fb3e5040dc8
>> Author: Dan Kenigsberg <[EMAIL PROTECTED]>
>> Date: Sun Nov 18 13:55:26 2007 +0200
>>
>> Support -cpu host option. Negotiate cpuid table with userspace.
I believe the attached two patches apply all your comments.
Dan.
commit 53c12a2602642ed01488ae382d23d350e5464a71
Author: Dan Kenigsberg <[EMAIL PROTECTED]>
Date: Wed Nov 21 16:03:55 2007 +0200
Expose host CPU features to guests. a new KVM_GET_SUPPORTED_CPUID ioctl is
used
to obtain the table of cpuid functions supported by the host & kvm. The
user-space may fine-tune this table and use the KVM_SET_CPUID2 ioctl to set
the
table seen by guests.
This patch exposes CPU features that are hitherto hidden (e.g. SSE, cache
structure) and makes a couple of benchmarks faster.
The table of cpuid entries and kvm_emulate_cpuid() had to be changed because
for a couple of cpuid functions, edx - and not only eax - is significant.
Worse, function 2 is stateful; its value may change between consequtive
calls.
Signed-off-by: Dan Kenigsberg <[EMAIL PROTECTED]>
diff --git a/drivers/kvm/x86.c b/drivers/kvm/x86.c
index b42c9d4..c14967a 100644
--- a/drivers/kvm/x86.c
+++ b/drivers/kvm/x86.c
@@ -626,6 +626,7 @@ int kvm_dev_ioctl_check_extension(long ext)
case KVM_CAP_MMU_SHADOW_CACHE_CONTROL:
case KVM_CAP_USER_MEMORY:
case KVM_CAP_SET_TSS_ADDR:
+ case KVM_CAP_EXT_CPUID:
r = 1;
break;
default:
@@ -688,13 +689,18 @@ void kvm_arch_vcpu_put(struct kvm_vcpu *vcpu)
kvm_put_guest_fpu(vcpu);
}
+static int is_efer_nx(void) {
+ u64 efer;
+
+ rdmsrl(MSR_EFER, efer);
+ return efer & EFER_NX;
+}
+
static void cpuid_fix_nx_cap(struct kvm_vcpu *vcpu)
{
- u64 efer;
int i;
- struct kvm_cpuid_entry *e, *entry;
+ struct kvm_cpuid_entry2 *e, *entry;
- rdmsrl(MSR_EFER, efer);
entry = NULL;
for (i = 0; i < vcpu->cpuid_nent; ++i) {
e = &vcpu->cpuid_entries[i];
@@ -703,16 +709,56 @@ static void cpuid_fix_nx_cap(struct kvm_vcpu *vcpu)
break;
}
}
- if (entry && (entry->edx & (1 << 20)) && !(efer & EFER_NX)) {
+ if (entry && (entry->edx & (1 << 20)) && !is_efer_nx()) {
entry->edx &= ~(1 << 20);
printk(KERN_INFO "kvm: guest NX capability removed\n");
}
}
+/* when an old userspace process fills a new kernel module */
static int kvm_vcpu_ioctl_set_cpuid(struct kvm_vcpu *vcpu,
struct kvm_cpuid *cpuid,
struct kvm_cpuid_entry __user *entries)
{
+ int r, i;
+ struct kvm_cpuid_entry *cpuid_entries;
+
+ r = -E2BIG;
+ if (cpuid->nent > KVM_MAX_CPUID_ENTRIES)
+ goto out;
+ r = -ENOMEM;
+ cpuid_entries = vmalloc(sizeof(struct kvm_cpuid_entry) * cpuid->nent);
+ if (!cpuid_entries)
+ goto out;
+ r = -EFAULT;
+ if (copy_from_user(cpuid_entries, entries,
+ cpuid->nent * sizeof(struct kvm_cpuid_entry)))
+ goto out_free;
+ for (i = 0; i < cpuid->nent; i++) {
+ vcpu->cpuid_entries[i].function = cpuid_entries[i].function;
+ vcpu->cpuid_entries[i].eax = cpuid_entries[i].eax;
+ vcpu->cpuid_entries[i].ebx = cpuid_entries[i].ebx;
+ vcpu->cpuid_entries[i].ecx = cpuid_entries[i].ecx;
+ vcpu->cpuid_entries[i].edx = cpuid_entries[i].edx;
+ vcpu->cpuid_entries[i].index = 0;
+ vcpu->cpuid_entries[i].flags = 0;
+ vcpu->cpuid_entries[i].padding[0] = 0;
+ vcpu->cpuid_entries[i].padding[1] = 0;
+ vcpu->cpuid_entries[i].padding[2] = 0;
+ }
+ vcpu->cpuid_nent = cpuid->nent;
+ r = 0;
+
+out_free:
+ vfree(cpuid_entries);
+out:
+ return r;
+}
+
+static int kvm_vcpu_ioctl_set_cpuid2(struct kvm_vcpu *vcpu,
+ struct kvm_cpuid2 *cpuid,
+ struct kvm_cpuid_entry2 __user *entries)
+{
int r;
r = -E2BIG;
@@ -720,7 +766,7 @@ static int kvm_vcpu_ioctl_set_cpuid(struct kvm_vcpu *vcpu,
g
Re: [kvm-devel] [RFC] Expose host cpuid to guest
Dan Kenigsberg wrote:
> These patches expose host CPU features (that are known to work under
> KVM) to guests. It makes a couple of benchmarks run faster, and
> generally gives kvm's user better info on its host.
>
> The kernel-space patch adds KVM_GET_SUPPORTED_CPUID ioctl to obtain the
> table of cpuid functions supported by the host. The user-space patch
> allows fine-tuning this table from the command-line.
>
> I had to define struct kvm_cpuid2, KVM_SET_CPUID2 etc., because cpuid
> functions are a little more complex than just function-value pairs.
>
> commit e9775d0a16097cfb71779cb2fb985fb3e5040dc8
> Author: Dan Kenigsberg <[EMAIL PROTECTED]>
> Date: Sun Nov 18 13:55:26 2007 +0200
>
> Support -cpu host option. Negotiate cpuid table with userspace.
>
The kernel doesn't have a -cpu option. The description needs to be more
descriptive (motivation, special cases in cpuid).
>
>
> +static int is_efer_nx(void) {
> + u64 efer;
>
blank line
> + rdmsrl(MSR_EFER, efer);
> + return efer & EFER_NX;
> +}
> +
>
>
> +static int kvm_vcpu_ioctl_get_cpuid2(struct kvm_vcpu *vcpu,
> + struct kvm_cpuid2 *cpuid,
> + struct kvm_cpuid_entry2 __user *entries)
> +{
> + int r;
> +
> + r = -E2BIG;
> + if (cpuid->nent < vcpu->cpuid_nent)
> + goto out;
> + r = -EFAULT;
> + if (copy_to_user(entries, &vcpu->cpuid_entries,
> +vcpu->cpuid_nent * sizeof(struct kvm_cpuid_entry2)))
> + goto out;
> + return 0;
> +
> +out:
> +cpuid->nent = vcpu->cpuid_nent;
>
whitespace damage here
> + return r;
> +
> +static void do_cpuid_ent(struct kvm_cpuid_entry2 *entry, u32 function,
> + u32 index, int *nent, int maxnent)
> +{
> + const __u32 kvm_supported_word0_x86_features = bit(X86_FEATURE_FPU) |
> + bit(X86_FEATURE_VME) | bit(X86_FEATURE_DE) |
> + bit(X86_FEATURE_PSE) | bit(X86_FEATURE_TSC) |
> + bit(X86_FEATURE_MSR) | bit(X86_FEATURE_PAE) |
> + bit(X86_FEATURE_CX8) | bit(X86_FEATURE_APIC) |
> + bit(X86_FEATURE_SEP) | bit(X86_FEATURE_PGE) |
> + bit(X86_FEATURE_CMOV) | bit(X86_FEATURE_PSE36) |
> + bit(X86_FEATURE_CLFLSH) | bit(X86_FEATURE_MMX) |
> + bit(X86_FEATURE_FXSR) | bit(X86_FEATURE_XMM) |
> + bit(X86_FEATURE_XMM2) | bit(X86_FEATURE_SELFSNOOP);
>
u32, not __u32.
> + const __u32 kvm_supported_word1_x86_features = bit(X86_FEATURE_FPU) |
> + bit(X86_FEATURE_VME) | bit(X86_FEATURE_DE) |
> + bit(X86_FEATURE_PSE) | bit(X86_FEATURE_TSC) |
> + bit(X86_FEATURE_MSR) | bit(X86_FEATURE_PAE) |
> + bit(X86_FEATURE_CX8) | bit(X86_FEATURE_APIC) |
> + bit(X86_FEATURE_PGE) |
> + bit(X86_FEATURE_CMOV) | bit(X86_FEATURE_PSE36) |
> + bit(X86_FEATURE_MMX) | bit(X86_FEATURE_FXSR) |
> + bit(X86_FEATURE_SYSCALL) |
> + (bit(X86_FEATURE_NX) && is_efer_nx()) |
> +#ifdef CONFIG_X86_64
> + bit(X86_FEATURE_LM) |
> +#endif
> + /* TODO: make sure the following features are
> + * safe for SVM guests */
> + bit(X86_FEATURE_MMXEXT) |
> + bit(X86_FEATURE_RDTSCP) | bit(X86_FEATURE_3DNOWEXT) |
> + bit(X86_FEATURE_3DNOW);
>
rdtscp isn't, I believe.
> + const __u32 kvm_supported_word3_x86_features =
> + bit(X86_FEATURE_XMM3) | bit(X86_FEATURE_CX16);
> + const __u32 kvm_supported_word6_x86_features =
> + bit(X86_FEATURE_LAHF_LM) | bit(X86_FEATURE_CMP_LEGACY);
> +
> + /* all func 2 cpuid_count() should be called on the same cpu */
> + if (function==2)
> + get_cpu();
>
Avoid the special case, just to it unconditionally.
> + do_cpuid_1_ent(entry, function, index);
> + ++*nent;
> +
> + switch (function) {
> + case 0:
> + entry->eax = min(entry->eax, (u32)0xb);
> + break;
> + case 1:
> + entry->edx &= kvm_supported_word0_x86_features;
> + entry->ecx &= kvm_supported_word3_x86_features;
> + break;
> + /* function 2 entries are STATEFUL. That is, repeated cpuid commands
> + * may return different values. This forces us to get_cpu() before
> + * issuing the first command, and also to emulate this annoying behavior
> + * in kvm_emulate_cpuid() using KVM_CPUID_FLAG_STATE_READ_NEXT */
> + case 2: {
> + int t, times = entry->eax & 0xff;
>
Indent this normally relative to other entries.
> +
> + entry->flags |= KVM_CPUID_FLAG_STATEFUL_FUNC;
> + for (t = 1; t < times && *nent < maxnent; ++t) {
> + do_cpuid_1_ent(&entry[t], function, 0);
> + entry[t].flags |= KVM_CPUID_FLAG_STATEFUL_FUNC;
> +
Re: [kvm-devel] [RFC] Expose host cpuid to guest
* Dan Kenigsberg wrote: > These patches expose host CPU features (that are known to work under > KVM) to guests. It makes a couple of benchmarks run faster, and > generally gives kvm's user better info on its host. > > The kernel-space patch adds KVM_GET_SUPPORTED_CPUID ioctl to obtain the > table of cpuid functions supported by the host. The user-space patch > allows fine-tuning this table from the command-line. > > I had to define struct kvm_cpuid2, KVM_SET_CPUID2 etc., because cpuid > functions are a little more complex than just function-value pairs. > > Dan. Hey Dan, were those amd-specific things sorted out? - This SF.net email is sponsored by: Microsoft Defy all challenges. Microsoft(R) Visual Studio 2005. http://clk.atdmt.com/MRT/go/vse012070mrt/direct/01/ ___ kvm-devel mailing list kvm-devel@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/kvm-devel
