2012/7/31 Stefan Hajnoczi <stefa...@gmail.com>:
> On Wed, Jul 25, 2012 at 12:11 PM, <riegama...@gmail.com> wrote:
>> From: Dunrong Huang <riegama...@gmail.com>
>>
>> The VCPU count limit in kernel now is 254, defined by KVM_MAX_VCPUS
>> in kernel's header files. But the count limit in QEMU is 255,
>> so QEMU will failed to start if user passes "-enable-kvm" and "-smp 255"
>> to it.
>>
>> Exit QEMU with an error if KVM is enabled and number of SMP cpus requested
>> exceeds KVM_MAX_VCPUS.
>>
>> Signed-off-by: Dunrong Huang <riegama...@gmail.com>
>> ---
>> v1 -> v2:
>> Checking if the number of smp cpus requested exceeds KVM limit count
>> if and only if kvm is enabled.
>>
>> vl.c | 11 +++++++++++
>> 1 files changed, 11 insertions(+), 0 deletions(-)
>>
>> diff --git a/vl.c b/vl.c
>> index 8904db1..cdd1c96 100644
>> --- a/vl.c
>> +++ b/vl.c
>> @@ -169,6 +169,11 @@ int main(int argc, char **argv)
>>
>> #define MAX_VIRTIO_CONSOLES 1
>>
>> +/* KVM_MAX_VCPUS defined in kernel's header files */
>> +#ifndef KVM_MAX_VCPUS
>> +#define KVM_MAX_VCPUS 254
>> +#endif
>> +
>> static const char *data_dir;
>> const char *bios_name = NULL;
>> enum vga_retrace_method vga_retrace_method = VGA_RETRACE_DUMB;
>> @@ -3348,6 +3353,12 @@ int main(int argc, char **argv, char **envp)
>>
>> configure_accelerator();
>>
>> + if (kvm_enabled() && smp_cpus > KVM_MAX_VCPUS) {
>> + fprintf(stderr, "Number of SMP cpus requested (%d) exceeds max cpus
>> "
>> + "supported by KVM (%d)\n", smp_cpus, KVM_MAX_VCPUS);
>> + exit(1);
>> + }
>
> After a little discussion on IRC, two points emerged:
> 1. Use KVM_CAP_MAX_VCPUS to query the max number of vcpus at runtime.
> 2. We should fail gracefully when ioctl() fails.
>
> In other words, using the KVM_MAX_VCPUS value from userspace isn't a
> good idea. Imagine what happens when the user upgrades their kernel
> without recompiling QEMU. If the KVM_MAX_VCPUS value increased in the
> kernel QEMU would not know.
>
> Please either drop this patch completely or at least using
> KVM_CAP_MAX_VCPUS so we fetch the maximum value at runtime.
>
> Stefan
I see. Thanks for your comments
I will send a patch based the disscussion we talked on IRC
--
Best Regards,
Dunrong Huang
