[PATCH] x86: make the dom0_max_vcpus option more flexible

[PATCH] x86: make the dom0_max_vcpus option more flexible

[David Vrabel]

From: David Vrabel <david.vrabel@citrix.com>

The dom0_max_vcpus command line option only allows the exact number of
VCPUs for dom0 to be set.  It is not possible to say "up to N VCPUs
but no more than the number physically present."

Add min: and max: prefixes to the option to set a minimum number of
VCPUs, and a maximum which does not exceed the number of PCPUs.

For example, with "dom0_max_vcpus=min:4,max:8":

    PCPUs  Dom0 VCPUs
     2      4
     4      4
     6      6
     8      8
    10      8

The existing behaviour of "dom0_max_vcpus=N" still works as before.

Signed-off-by: David Vrabel <david.vrabel@citrix.com>
---
 docs/misc/xen-command-line.markdown |   29 +++++++++++++++++++++++++++--
 xen/arch/x86/domain_build.c         |   23 ++++++++++++++++++++++-
 2 files changed, 49 insertions(+), 3 deletions(-)

diff --git a/docs/misc/xen-command-line.markdown b/docs/misc/xen-command-line.markdown
index a6195f2..5f0c2cd 100644
--- a/docs/misc/xen-command-line.markdown
+++ b/docs/misc/xen-command-line.markdown
@@ -272,10 +272,35 @@ Specify the bit width of the DMA heap.
 
 ### dom0\_ioports\_disable
 ### dom0\_max\_vcpus
+
+Either:
+
 > `= <integer>`
 
-Specify the maximum number of vcpus to give to dom0.  This defaults
-to the number of pcpus on the host.
+The maximum number of VCPUs to give to dom0.  This number of VCPUs can
+be more than the number of PCPUs on the host.  The default is the
+number of PCPUs.
+
+Or:
+
+> `= List of ( min:<integer> | max:<integer> )`
+
+With the `min:` option dom0 will have at least this minimum number of
+VCPUs (default: 1).  This may be more than the number of PCPUs on the
+host.
+
+With the `max:` option dom0 will have a VCPUs for each PCPUs but no
+more than this maximum number (default: unlimited).
+
+For example, with `dom0_max_vcpus=min:4,max:8`:
+
+     Number of
+  PCPUs | Dom0 VCPUs
+   2    |  4
+   4    |  4
+   6    |  6
+   8    |  8
+  10    |  8
 
 ### dom0\_mem (ia64)
 > `= <size>`
diff --git a/xen/arch/x86/domain_build.c b/xen/arch/x86/domain_build.c
index b3c5d4c..5407f8d 100644
--- a/xen/arch/x86/domain_build.c
+++ b/xen/arch/x86/domain_build.c
@@ -83,7 +83,24 @@ static void __init parse_dom0_mem(const char *s)
 custom_param("dom0_mem", parse_dom0_mem);
 
 static unsigned int __initdata opt_dom0_max_vcpus;
-integer_param("dom0_max_vcpus", opt_dom0_max_vcpus);
+static unsigned int __initdata opt_dom0_max_vcpus_min = 1;
+static unsigned int __initdata opt_dom0_max_vcpus_max = UINT_MAX;
+
+static void __init parse_dom0_max_vcpus(const char *s)
+{
+    do {
+        if ( !strncmp(s, "min:", 4) )
+            opt_dom0_max_vcpus_min = simple_strtoul(s+4, &s, 0);
+        else if ( !strncmp(s, "max:", 4) )
+            opt_dom0_max_vcpus_max = simple_strtoul(s+4, &s, 0);
+        else
+            opt_dom0_max_vcpus = simple_strtoul(s, &s, 0);
+        if ( *s != ',' )
+            break;
+    } while ( *s++ == ',' );
+        
+}
+custom_param("dom0_max_vcpus", parse_dom0_max_vcpus);
 
 struct vcpu *__init alloc_dom0_vcpu0(void)
 {
@@ -91,6 +108,10 @@ struct vcpu *__init alloc_dom0_vcpu0(void)
         opt_dom0_max_vcpus = num_cpupool_cpus(cpupool0);
     if ( opt_dom0_max_vcpus > MAX_VIRT_CPUS )
         opt_dom0_max_vcpus = MAX_VIRT_CPUS;
+    if ( opt_dom0_max_vcpus_min > opt_dom0_max_vcpus )
+        opt_dom0_max_vcpus = opt_dom0_max_vcpus_min;
+    if ( opt_dom0_max_vcpus_max < opt_dom0_max_vcpus )
+        opt_dom0_max_vcpus = opt_dom0_max_vcpus_max;
 
     dom0->vcpu = xzalloc_array(struct vcpu *, opt_dom0_max_vcpus);
     if ( !dom0->vcpu )
-- 
1.7.2.5

Re: [PATCH] x86: make the dom0_max_vcpus option more flexible

[Jan Beulich]

>>> On 04.05.12 at 18:01, David Vrabel <david.vrabel@citrix.com> wrote:
> From: David Vrabel <david.vrabel@citrix.com>
> 
> The dom0_max_vcpus command line option only allows the exact number of
> VCPUs for dom0 to be set.  It is not possible to say "up to N VCPUs
> but no more than the number physically present."
> 
> Add min: and max: prefixes to the option to set a minimum number of
> VCPUs, and a maximum which does not exceed the number of PCPUs.
> 
> For example, with "dom0_max_vcpus=min:4,max:8":

Both "...max...=min:..." and "...max...=max:" look pretty odd to me;
how about simply allowing a range along with a simple number (since
negative values make no sense, omitting either side of the range would
be supportable if necessary.

>     PCPUs  Dom0 VCPUs
>      2      4
>      4      4
>      6      6
>      8      8
>     10      8
> 
> The existing behaviour of "dom0_max_vcpus=N" still works as before.
> 
> Signed-off-by: David Vrabel <david.vrabel@citrix.com>
> ---
>  docs/misc/xen-command-line.markdown |   29 +++++++++++++++++++++++++++--
>  xen/arch/x86/domain_build.c         |   23 ++++++++++++++++++++++-
>  2 files changed, 49 insertions(+), 3 deletions(-)
> 
> diff --git a/docs/misc/xen-command-line.markdown 
> b/docs/misc/xen-command-line.markdown
> index a6195f2..5f0c2cd 100644
> --- a/docs/misc/xen-command-line.markdown
> +++ b/docs/misc/xen-command-line.markdown
> @@ -272,10 +272,35 @@ Specify the bit width of the DMA heap.
>  
>  ### dom0\_ioports\_disable
>  ### dom0\_max\_vcpus
> +
> +Either:
> +
>  > `= <integer>`
>  
> -Specify the maximum number of vcpus to give to dom0.  This defaults
> -to the number of pcpus on the host.
> +The maximum number of VCPUs to give to dom0.  This number of VCPUs can
> +be more than the number of PCPUs on the host.  The default is the
> +number of PCPUs.
> +
> +Or:
> +
> +> `= List of ( min:<integer> | max:<integer> )`
> +
> +With the `min:` option dom0 will have at least this minimum number of
> +VCPUs (default: 1).  This may be more than the number of PCPUs on the
> +host.
> +
> +With the `max:` option dom0 will have a VCPUs for each PCPUs but no
> +more than this maximum number (default: unlimited).
> +
> +For example, with `dom0_max_vcpus=min:4,max:8`:
> +
> +     Number of
> +  PCPUs | Dom0 VCPUs
> +   2    |  4
> +   4    |  4
> +   6    |  6
> +   8    |  8
> +  10    |  8
>  
>  ### dom0\_mem (ia64)
>  > `= <size>`
> diff --git a/xen/arch/x86/domain_build.c b/xen/arch/x86/domain_build.c
> index b3c5d4c..5407f8d 100644
> --- a/xen/arch/x86/domain_build.c
> +++ b/xen/arch/x86/domain_build.c
> @@ -83,7 +83,24 @@ static void __init parse_dom0_mem(const char *s)
>  custom_param("dom0_mem", parse_dom0_mem);
>  
>  static unsigned int __initdata opt_dom0_max_vcpus;
> -integer_param("dom0_max_vcpus", opt_dom0_max_vcpus);
> +static unsigned int __initdata opt_dom0_max_vcpus_min = 1;
> +static unsigned int __initdata opt_dom0_max_vcpus_max = UINT_MAX;
> +
> +static void __init parse_dom0_max_vcpus(const char *s)
> +{
> +    do {
> +        if ( !strncmp(s, "min:", 4) )
> +            opt_dom0_max_vcpus_min = simple_strtoul(s+4, &s, 0);
> +        else if ( !strncmp(s, "max:", 4) )
> +            opt_dom0_max_vcpus_max = simple_strtoul(s+4, &s, 0);
> +        else
> +            opt_dom0_max_vcpus = simple_strtoul(s, &s, 0);
> +        if ( *s != ',' )
> +            break;
> +    } while ( *s++ == ',' );
> +        
> +}
> +custom_param("dom0_max_vcpus", parse_dom0_max_vcpus);
>  
>  struct vcpu *__init alloc_dom0_vcpu0(void)
>  {
> @@ -91,6 +108,10 @@ struct vcpu *__init alloc_dom0_vcpu0(void)
>          opt_dom0_max_vcpus = num_cpupool_cpus(cpupool0);
>      if ( opt_dom0_max_vcpus > MAX_VIRT_CPUS )
>          opt_dom0_max_vcpus = MAX_VIRT_CPUS;
> +    if ( opt_dom0_max_vcpus_min > opt_dom0_max_vcpus )
> +        opt_dom0_max_vcpus = opt_dom0_max_vcpus_min;

Enlarging the value after the MAX_VIRT_CPUS range check must
not be done. You probably simply want to move your addition up
two lines.

> +    if ( opt_dom0_max_vcpus_max < opt_dom0_max_vcpus )
> +        opt_dom0_max_vcpus = opt_dom0_max_vcpus_max;

But please avoid ...=max: (number lost for some reason) rendering
the box unbootable (I'd say a maximum of zero should be interpreted
as 1).

Jan

>  
>      dom0->vcpu = xzalloc_array(struct vcpu *, opt_dom0_max_vcpus);
>      if ( !dom0->vcpu )
> -- 
> 1.7.2.5

Re: [PATCH] x86: make the dom0_max_vcpus option more flexible

[David Vrabel]

On 04/05/12 17:12, Jan Beulich wrote:
>>>> On 04.05.12 at 18:01, David Vrabel <david.vrabel@citrix.com> wrote:
>> From: David Vrabel <david.vrabel@citrix.com>
>>
>> The dom0_max_vcpus command line option only allows the exact number of
>> VCPUs for dom0 to be set.  It is not possible to say "up to N VCPUs
>> but no more than the number physically present."
>>
>> Add min: and max: prefixes to the option to set a minimum number of
>> VCPUs, and a maximum which does not exceed the number of PCPUs.
>>
>> For example, with "dom0_max_vcpus=min:4,max:8":
> 
> Both "...max...=min:..." and "...max...=max:" look pretty odd to me;
> how about simply allowing a range along with a simple number (since
> negative values make no sense, omitting either side of the range would
> be supportable if necessary.

I was copying the way dom0_mem worked but yeah, it's not very pretty.

Is dom0_max_vcpus=<min>-<max> (e.g., dom0_max_vcpus=4-8)  what you were
thinking of?

Using a single value would have to set both <min> and <max> or the
behaviour of the option changes (i.e., =N is the same as =N-N).

David

Re: [PATCH] x86: make the dom0_max_vcpus option more flexible

[David Vrabel]

On 04/05/12 17:26, David Vrabel wrote:
> On 04/05/12 17:12, Jan Beulich wrote:
>>>>> On 04.05.12 at 18:01, David Vrabel <david.vrabel@citrix.com> wrote:
>>> From: David Vrabel <david.vrabel@citrix.com>
>>>
>>> The dom0_max_vcpus command line option only allows the exact number of
>>> VCPUs for dom0 to be set.  It is not possible to say "up to N VCPUs
>>> but no more than the number physically present."
>>>
>>> Add min: and max: prefixes to the option to set a minimum number of
>>> VCPUs, and a maximum which does not exceed the number of PCPUs.
>>>
>>> For example, with "dom0_max_vcpus=min:4,max:8":
>>
>> Both "...max...=min:..." and "...max...=max:" look pretty odd to me;
>> how about simply allowing a range along with a simple number (since
>> negative values make no sense, omitting either side of the range would
>> be supportable if necessary.
> 
> I was copying the way dom0_mem worked but yeah, it's not very pretty.
> 
> Is dom0_max_vcpus=<min>-<max> (e.g., dom0_max_vcpus=4-8)  what you were
> thinking of?
> 
> Using a single value would have to set both <min> and <max> or the
> behaviour of the option changes (i.e., =N is the same as =N-N).

This is what I ended up with.

8<------------------------------
>From af1543965db76ab81139de7f072a7c4daf61157f Mon Sep 17 00:00:00 2001
From: David Vrabel <david.vrabel@citrix.com>
Date: Fri, 4 May 2012 16:09:52 +0100
Subject: [PATCH] x86: make the dom0_max_vcpus option more flexible

The dom0_max_vcpus command line option only allows the exact number of
VCPUs for dom0 to be set.  It is not possible to say "up to N VCPUs
but no more than the number physically present."

Allow a range for the option to set a minimum number of VCPUs, and a
maximum which does not exceed the number of PCPUs.

For example, with "dom0_max_vcpus=4-8":

    PCPUs  Dom0 VCPUs
     2      4
     4      4
     6      6
     8      8
    10      8

Existing command lines with "dom0_max_vcpus=N" still work as before.

Signed-off-by: David Vrabel <david.vrabel@citrix.com>
---
 docs/misc/xen-command-line.markdown |   29 +++++++++++++++++++++--
 xen/arch/x86/domain_build.c         |   43 +++++++++++++++++++++++++---------
 2 files changed, 57 insertions(+), 15 deletions(-)

diff --git a/docs/misc/xen-command-line.markdown b/docs/misc/xen-command-line.markdown
index a6195f2..4e4f713 100644
--- a/docs/misc/xen-command-line.markdown
+++ b/docs/misc/xen-command-line.markdown
@@ -272,10 +272,33 @@ Specify the bit width of the DMA heap.
 
 ### dom0\_ioports\_disable
 ### dom0\_max\_vcpus
-> `= <integer>`
 
-Specify the maximum number of vcpus to give to dom0.  This defaults
-to the number of pcpus on the host.
+Either:
+
+> `= <integer>`.
+
+The number of VCPUs to give to dom0.  This number of VCPUs can be more
+than the number of PCPUs on the host.  The default is the number of
+PCPUs.
+
+Or:
+
+> `= <min>-<max>` where `<min>` and `<max>` are integers.
+
+Gives dom0 a number of VCPUs equal to the number of PCPUs, but always
+at least `<min>` and no more than `<max>`.  Using `<min>` may give
+more VCPUs than PCPUs.  `<min>` or `<max>` may be omitted and the
+defaults of 1 and unlimited respectively are used instead.
+
+For example, with `dom0_max_vcpus=4-8`:
+
+     Number of
+  PCPUs | Dom0 VCPUs
+   2    |  4
+   4    |  4
+   6    |  6
+   8    |  8
+  10    |  8
 
 ### dom0\_mem (ia64)
 > `= <size>`
diff --git a/xen/arch/x86/domain_build.c b/xen/arch/x86/domain_build.c
index b3c5d4c..686b626 100644
--- a/xen/arch/x86/domain_build.c
+++ b/xen/arch/x86/domain_build.c
@@ -82,20 +82,39 @@ static void __init parse_dom0_mem(const char *s)
 }
 custom_param("dom0_mem", parse_dom0_mem);
 
-static unsigned int __initdata opt_dom0_max_vcpus;
-integer_param("dom0_max_vcpus", opt_dom0_max_vcpus);
+static unsigned int __initdata opt_dom0_max_vcpus_min = 1;
+static unsigned int __initdata opt_dom0_max_vcpus_max = UINT_MAX;
+
+static void __init parse_dom0_max_vcpus(const char *s)
+{
+    if (*s == '-')              /* -M */
+        opt_dom0_max_vcpus_max = simple_strtoul(s + 1, &s, 0);
+    else {                      /* N, N-, or N-M */
+        opt_dom0_max_vcpus_min = simple_strtoul(s, &s, 0);
+        if (*s++ == '\0')       /* N */
+            opt_dom0_max_vcpus_max = opt_dom0_max_vcpus_min;
+        else if (*s != '\0')    /* N-M */
+            opt_dom0_max_vcpus_max = simple_strtoul(s, &s, 0);
+    }
+}
+custom_param("dom0_max_vcpus", parse_dom0_max_vcpus);
 
 struct vcpu *__init alloc_dom0_vcpu0(void)
 {
-    if ( opt_dom0_max_vcpus == 0 )
-        opt_dom0_max_vcpus = num_cpupool_cpus(cpupool0);
-    if ( opt_dom0_max_vcpus > MAX_VIRT_CPUS )
-        opt_dom0_max_vcpus = MAX_VIRT_CPUS;
+    unsigned max_vcpus;
+
+    max_vcpus = num_cpupool_cpus(cpupool0);
+    if ( opt_dom0_max_vcpus_min > max_vcpus )
+        max_vcpus = opt_dom0_max_vcpus_min;
+    if ( opt_dom0_max_vcpus_max < max_vcpus )
+        max_vcpus = opt_dom0_max_vcpus_max;
+    if ( max_vcpus > MAX_VIRT_CPUS )
+        max_vcpus = MAX_VIRT_CPUS;
 
-    dom0->vcpu = xzalloc_array(struct vcpu *, opt_dom0_max_vcpus);
+    dom0->vcpu = xzalloc_array(struct vcpu *, max_vcpus);
     if ( !dom0->vcpu )
         return NULL;
-    dom0->max_vcpus = opt_dom0_max_vcpus;
+    dom0->max_vcpus = max_vcpus;
 
     return alloc_vcpu(dom0, 0, 0);
 }
@@ -185,11 +204,11 @@ static unsigned long __init compute_dom0_nr_pages(
     unsigned long max_pages = dom0_max_nrpages;
 
     /* Reserve memory for further dom0 vcpu-struct allocations... */
-    avail -= (opt_dom0_max_vcpus - 1UL)
+    avail -= (d->max_vcpus - 1UL)
              << get_order_from_bytes(sizeof(struct vcpu));
     /* ...and compat_l4's, if needed. */
     if ( is_pv_32on64_domain(d) )
-        avail -= opt_dom0_max_vcpus - 1;
+        avail -= d->max_vcpus - 1;
 
     /* Reserve memory for iommu_dom0_init() (rough estimate). */
     if ( iommu_enabled )
@@ -883,10 +902,10 @@ int __init construct_dom0(
     for ( i = 0; i < XEN_LEGACY_MAX_VCPUS; i++ )
         shared_info(d, vcpu_info[i].evtchn_upcall_mask) = 1;
 
-    printk("Dom0 has maximum %u VCPUs\n", opt_dom0_max_vcpus);
+    printk("Dom0 has maximum %u VCPUs\n", d->max_vcpus);
 
     cpu = cpumask_first(cpupool0->cpu_valid);
-    for ( i = 1; i < opt_dom0_max_vcpus; i++ )
+    for ( i = 1; i < d->max_vcpus; i++ )
     {
         cpu = cpumask_cycle(cpu, cpupool0->cpu_valid);
         (void)alloc_vcpu(d, i, cpu);
-- 
1.7.2.5

Re: [PATCH] x86: make the dom0_max_vcpus option more flexible

[Jan Beulich]

>>> On 04.05.12 at 20:18, David Vrabel <david.vrabel@citrix.com> wrote:
> On 04/05/12 17:26, David Vrabel wrote:
>> On 04/05/12 17:12, Jan Beulich wrote:
>>>>>> On 04.05.12 at 18:01, David Vrabel <david.vrabel@citrix.com> wrote:
>>>> From: David Vrabel <david.vrabel@citrix.com>
>>>>
>>>> The dom0_max_vcpus command line option only allows the exact number of
>>>> VCPUs for dom0 to be set.  It is not possible to say "up to N VCPUs
>>>> but no more than the number physically present."
>>>>
>>>> Add min: and max: prefixes to the option to set a minimum number of
>>>> VCPUs, and a maximum which does not exceed the number of PCPUs.
>>>>
>>>> For example, with "dom0_max_vcpus=min:4,max:8":
>>>
>>> Both "...max...=min:..." and "...max...=max:" look pretty odd to me;
>>> how about simply allowing a range along with a simple number (since
>>> negative values make no sense, omitting either side of the range would
>>> be supportable if necessary.
>> 
>> I was copying the way dom0_mem worked but yeah, it's not very pretty.
>> 
>> Is dom0_max_vcpus=<min>-<max> (e.g., dom0_max_vcpus=4-8)  what you were
>> thinking of?
>> 
>> Using a single value would have to set both <min> and <max> or the
>> behaviour of the option changes (i.e., =N is the same as =N-N).
> 
> This is what I ended up with.
> 
> 8<------------------------------
> From af1543965db76ab81139de7f072a7c4daf61157f Mon Sep 17 00:00:00 2001
> From: David Vrabel <david.vrabel@citrix.com>
> Date: Fri, 4 May 2012 16:09:52 +0100
> Subject: [PATCH] x86: make the dom0_max_vcpus option more flexible
> 
> The dom0_max_vcpus command line option only allows the exact number of
> VCPUs for dom0 to be set.  It is not possible to say "up to N VCPUs
> but no more than the number physically present."
> 
> Allow a range for the option to set a minimum number of VCPUs, and a
> maximum which does not exceed the number of PCPUs.
> 
> For example, with "dom0_max_vcpus=4-8":
> 
>     PCPUs  Dom0 VCPUs
>      2      4
>      4      4
>      6      6
>      8      8
>     10      8
> 
> Existing command lines with "dom0_max_vcpus=N" still work as before.
> 
> Signed-off-by: David Vrabel <david.vrabel@citrix.com>

Acked-by: Jan Beulich <jbeulich@suse.com>

But I'm not sure whether this qualifies for going in for 4.2...

Jan

> ---
>  docs/misc/xen-command-line.markdown |   29 +++++++++++++++++++++--
>  xen/arch/x86/domain_build.c         |   43 +++++++++++++++++++++++++---------
>  2 files changed, 57 insertions(+), 15 deletions(-)
> 
> diff --git a/docs/misc/xen-command-line.markdown 
> b/docs/misc/xen-command-line.markdown
> index a6195f2..4e4f713 100644
> --- a/docs/misc/xen-command-line.markdown
> +++ b/docs/misc/xen-command-line.markdown
> @@ -272,10 +272,33 @@ Specify the bit width of the DMA heap.
>  
>  ### dom0\_ioports\_disable
>  ### dom0\_max\_vcpus
> -> `= <integer>`
>  
> -Specify the maximum number of vcpus to give to dom0.  This defaults
> -to the number of pcpus on the host.
> +Either:
> +
> +> `= <integer>`.
> +
> +The number of VCPUs to give to dom0.  This number of VCPUs can be more
> +than the number of PCPUs on the host.  The default is the number of
> +PCPUs.
> +
> +Or:
> +
> +> `= <min>-<max>` where `<min>` and `<max>` are integers.
> +
> +Gives dom0 a number of VCPUs equal to the number of PCPUs, but always
> +at least `<min>` and no more than `<max>`.  Using `<min>` may give
> +more VCPUs than PCPUs.  `<min>` or `<max>` may be omitted and the
> +defaults of 1 and unlimited respectively are used instead.
> +
> +For example, with `dom0_max_vcpus=4-8`:
> +
> +     Number of
> +  PCPUs | Dom0 VCPUs
> +   2    |  4
> +   4    |  4
> +   6    |  6
> +   8    |  8
> +  10    |  8
>  
>  ### dom0\_mem (ia64)
>  > `= <size>`
> diff --git a/xen/arch/x86/domain_build.c b/xen/arch/x86/domain_build.c
> index b3c5d4c..686b626 100644
> --- a/xen/arch/x86/domain_build.c
> +++ b/xen/arch/x86/domain_build.c
> @@ -82,20 +82,39 @@ static void __init parse_dom0_mem(const char *s)
>  }
>  custom_param("dom0_mem", parse_dom0_mem);
>  
> -static unsigned int __initdata opt_dom0_max_vcpus;
> -integer_param("dom0_max_vcpus", opt_dom0_max_vcpus);
> +static unsigned int __initdata opt_dom0_max_vcpus_min = 1;
> +static unsigned int __initdata opt_dom0_max_vcpus_max = UINT_MAX;
> +
> +static void __init parse_dom0_max_vcpus(const char *s)
> +{
> +    if (*s == '-')              /* -M */
> +        opt_dom0_max_vcpus_max = simple_strtoul(s + 1, &s, 0);
> +    else {                      /* N, N-, or N-M */
> +        opt_dom0_max_vcpus_min = simple_strtoul(s, &s, 0);
> +        if (*s++ == '\0')       /* N */
> +            opt_dom0_max_vcpus_max = opt_dom0_max_vcpus_min;
> +        else if (*s != '\0')    /* N-M */
> +            opt_dom0_max_vcpus_max = simple_strtoul(s, &s, 0);
> +    }
> +}
> +custom_param("dom0_max_vcpus", parse_dom0_max_vcpus);
>  
>  struct vcpu *__init alloc_dom0_vcpu0(void)
>  {
> -    if ( opt_dom0_max_vcpus == 0 )
> -        opt_dom0_max_vcpus = num_cpupool_cpus(cpupool0);
> -    if ( opt_dom0_max_vcpus > MAX_VIRT_CPUS )
> -        opt_dom0_max_vcpus = MAX_VIRT_CPUS;
> +    unsigned max_vcpus;
> +
> +    max_vcpus = num_cpupool_cpus(cpupool0);
> +    if ( opt_dom0_max_vcpus_min > max_vcpus )
> +        max_vcpus = opt_dom0_max_vcpus_min;
> +    if ( opt_dom0_max_vcpus_max < max_vcpus )
> +        max_vcpus = opt_dom0_max_vcpus_max;
> +    if ( max_vcpus > MAX_VIRT_CPUS )
> +        max_vcpus = MAX_VIRT_CPUS;
>  
> -    dom0->vcpu = xzalloc_array(struct vcpu *, opt_dom0_max_vcpus);
> +    dom0->vcpu = xzalloc_array(struct vcpu *, max_vcpus);
>      if ( !dom0->vcpu )
>          return NULL;
> -    dom0->max_vcpus = opt_dom0_max_vcpus;
> +    dom0->max_vcpus = max_vcpus;
>  
>      return alloc_vcpu(dom0, 0, 0);
>  }
> @@ -185,11 +204,11 @@ static unsigned long __init compute_dom0_nr_pages(
>      unsigned long max_pages = dom0_max_nrpages;
>  
>      /* Reserve memory for further dom0 vcpu-struct allocations... */
> -    avail -= (opt_dom0_max_vcpus - 1UL)
> +    avail -= (d->max_vcpus - 1UL)
>               << get_order_from_bytes(sizeof(struct vcpu));
>      /* ...and compat_l4's, if needed. */
>      if ( is_pv_32on64_domain(d) )
> -        avail -= opt_dom0_max_vcpus - 1;
> +        avail -= d->max_vcpus - 1;
>  
>      /* Reserve memory for iommu_dom0_init() (rough estimate). */
>      if ( iommu_enabled )
> @@ -883,10 +902,10 @@ int __init construct_dom0(
>      for ( i = 0; i < XEN_LEGACY_MAX_VCPUS; i++ )
>          shared_info(d, vcpu_info[i].evtchn_upcall_mask) = 1;
>  
> -    printk("Dom0 has maximum %u VCPUs\n", opt_dom0_max_vcpus);
> +    printk("Dom0 has maximum %u VCPUs\n", d->max_vcpus);
>  
>      cpu = cpumask_first(cpupool0->cpu_valid);
> -    for ( i = 1; i < opt_dom0_max_vcpus; i++ )
> +    for ( i = 1; i < d->max_vcpus; i++ )
>      {
>          cpu = cpumask_cycle(cpu, cpupool0->cpu_valid);
>          (void)alloc_vcpu(d, i, cpu);
> -- 
> 1.7.2.5

Re: [PATCH] x86: make the dom0_max_vcpus option more flexible

[David Vrabel]

On 07/05/12 09:19, Jan Beulich wrote:
>> Subject: [PATCH] x86: make the dom0_max_vcpus option more flexible
>>
>> The dom0_max_vcpus command line option only allows the exact number of
>> VCPUs for dom0 to be set.  It is not possible to say "up to N VCPUs
>> but no more than the number physically present."
>>
>> Allow a range for the option to set a minimum number of VCPUs, and a
>> maximum which does not exceed the number of PCPUs.
>>
>> For example, with "dom0_max_vcpus=4-8":
>>
>>     PCPUs  Dom0 VCPUs
>>      2      4
>>      4      4
>>      6      6
>>      8      8
>>     10      8
>>
>> Existing command lines with "dom0_max_vcpus=N" still work as before.
>>
>> Signed-off-by: David Vrabel <david.vrabel@citrix.com>
> 
> Acked-by: Jan Beulich <jbeulich@suse.com>
> 
> But I'm not sure whether this qualifies for going in for 4.2...

I don't think it's a 4.2 candidate.  I posted it now we need this
functionality in XenServer and I didn't want to change the command line
in a way that would be incompatible with upstream.

David

Re: [PATCH] x86: make the dom0_max_vcpus option more flexible

[Jan Beulich]

>>> On 04.05.12 at 18:26, David Vrabel <david.vrabel@citrix.com> wrote:
> On 04/05/12 17:12, Jan Beulich wrote:
>>>>> On 04.05.12 at 18:01, David Vrabel <david.vrabel@citrix.com> wrote:
>>> From: David Vrabel <david.vrabel@citrix.com>
>>>
>>> The dom0_max_vcpus command line option only allows the exact number of
>>> VCPUs for dom0 to be set.  It is not possible to say "up to N VCPUs
>>> but no more than the number physically present."
>>>
>>> Add min: and max: prefixes to the option to set a minimum number of
>>> VCPUs, and a maximum which does not exceed the number of PCPUs.
>>>
>>> For example, with "dom0_max_vcpus=min:4,max:8":
>> 
>> Both "...max...=min:..." and "...max...=max:" look pretty odd to me;
>> how about simply allowing a range along with a simple number (since
>> negative values make no sense, omitting either side of the range would
>> be supportable if necessary.
> 
> I was copying the way dom0_mem worked but yeah, it's not very pretty.
> 
> Is dom0_max_vcpus=<min>-<max> (e.g., dom0_max_vcpus=4-8)  what you were
> thinking of?

Yes.

> Using a single value would have to set both <min> and <max> or the
> behaviour of the option changes (i.e., =N is the same as =N-N).

That sounds plausible.

Jan

[PATCH] x86: make the dom0_max_vcpus option more flexible

[David Vrabel]

From: David Vrabel <david.vrabel@citrix.com>

The dom0_max_vcpus command line option only allows the exact number of
VCPUs for dom0 to be set.  It is not possible to say "up to N VCPUs
but no more than the number physically present."

Allow a range for the option to set a minimum number of VCPUs, and a
maximum which does not exceed the number of PCPUs.

For example, with "dom0_max_vcpus=4-8":

    PCPUs  Dom0 VCPUs
     2      4
     4      4
     6      6
     8      8
    10      8

Existing command lines with "dom0_max_vcpus=N" still work as before
(and are equivalent to dom0_max_vcpus=N-N).

Signed-off-by: David Vrabel <david.vrabel@citrix.com>
Acked-by: Jan Beulich <jbeulich@suse.com>
---
Changes since v2:
  - none, repost for Xen 4.3
---
 docs/misc/xen-command-line.markdown |   29 +++++++++++++++++++++--
 xen/arch/x86/domain_build.c         |   43 +++++++++++++++++++++++++---------
 2 files changed, 57 insertions(+), 15 deletions(-)

diff --git a/docs/misc/xen-command-line.markdown b/docs/misc/xen-command-line.markdown
index 6599931..97a2bb2 100644
--- a/docs/misc/xen-command-line.markdown
+++ b/docs/misc/xen-command-line.markdown
@@ -383,10 +383,33 @@ Specify the bit width of the DMA heap.
 Specify a list of IO ports to be excluded from dom0 access.
 
 ### dom0\_max\_vcpus
-> `= <integer>`
 
-Specify the maximum number of vcpus to give to dom0.  This defaults
-to the number of pcpus on the host.
+Either:
+
+> `= <integer>`.
+
+The number of VCPUs to give to dom0.  This number of VCPUs can be more
+than the number of PCPUs on the host.  The default is the number of
+PCPUs.
+
+Or:
+
+> `= <min>-<max>` where `<min>` and `<max>` are integers.
+
+Gives dom0 a number of VCPUs equal to the number of PCPUs, but always
+at least `<min>` and no more than `<max>`.  Using `<min>` may give
+more VCPUs than PCPUs.  `<min>` or `<max>` may be omitted and the
+defaults of 1 and unlimited respectively are used instead.
+
+For example, with `dom0_max_vcpus=4-8`:
+
+     Number of
+  PCPUs | Dom0 VCPUs
+   2    |  4
+   4    |  4
+   6    |  6
+   8    |  8
+  10    |  8
 
 ### dom0\_mem
 > `= List of ( min:<size> | max:<size> | <size> )`
diff --git a/xen/arch/x86/domain_build.c b/xen/arch/x86/domain_build.c
index b3c5d4c..686b626 100644
--- a/xen/arch/x86/domain_build.c
+++ b/xen/arch/x86/domain_build.c
@@ -82,20 +82,39 @@ static void __init parse_dom0_mem(const char *s)
 }
 custom_param("dom0_mem", parse_dom0_mem);
 
-static unsigned int __initdata opt_dom0_max_vcpus;
-integer_param("dom0_max_vcpus", opt_dom0_max_vcpus);
+static unsigned int __initdata opt_dom0_max_vcpus_min = 1;
+static unsigned int __initdata opt_dom0_max_vcpus_max = UINT_MAX;
+
+static void __init parse_dom0_max_vcpus(const char *s)
+{
+    if (*s == '-')              /* -M */
+        opt_dom0_max_vcpus_max = simple_strtoul(s + 1, &s, 0);
+    else {                      /* N, N-, or N-M */
+        opt_dom0_max_vcpus_min = simple_strtoul(s, &s, 0);
+        if (*s++ == '\0')       /* N */
+            opt_dom0_max_vcpus_max = opt_dom0_max_vcpus_min;
+        else if (*s != '\0')    /* N-M */
+            opt_dom0_max_vcpus_max = simple_strtoul(s, &s, 0);
+    }
+}
+custom_param("dom0_max_vcpus", parse_dom0_max_vcpus);
 
 struct vcpu *__init alloc_dom0_vcpu0(void)
 {
-    if ( opt_dom0_max_vcpus == 0 )
-        opt_dom0_max_vcpus = num_cpupool_cpus(cpupool0);
-    if ( opt_dom0_max_vcpus > MAX_VIRT_CPUS )
-        opt_dom0_max_vcpus = MAX_VIRT_CPUS;
+    unsigned max_vcpus;
+
+    max_vcpus = num_cpupool_cpus(cpupool0);
+    if ( opt_dom0_max_vcpus_min > max_vcpus )
+        max_vcpus = opt_dom0_max_vcpus_min;
+    if ( opt_dom0_max_vcpus_max < max_vcpus )
+        max_vcpus = opt_dom0_max_vcpus_max;
+    if ( max_vcpus > MAX_VIRT_CPUS )
+        max_vcpus = MAX_VIRT_CPUS;
 
-    dom0->vcpu = xzalloc_array(struct vcpu *, opt_dom0_max_vcpus);
+    dom0->vcpu = xzalloc_array(struct vcpu *, max_vcpus);
     if ( !dom0->vcpu )
         return NULL;
-    dom0->max_vcpus = opt_dom0_max_vcpus;
+    dom0->max_vcpus = max_vcpus;
 
     return alloc_vcpu(dom0, 0, 0);
 }
@@ -185,11 +204,11 @@ static unsigned long __init compute_dom0_nr_pages(
     unsigned long max_pages = dom0_max_nrpages;
 
     /* Reserve memory for further dom0 vcpu-struct allocations... */
-    avail -= (opt_dom0_max_vcpus - 1UL)
+    avail -= (d->max_vcpus - 1UL)
              << get_order_from_bytes(sizeof(struct vcpu));
     /* ...and compat_l4's, if needed. */
     if ( is_pv_32on64_domain(d) )
-        avail -= opt_dom0_max_vcpus - 1;
+        avail -= d->max_vcpus - 1;
 
     /* Reserve memory for iommu_dom0_init() (rough estimate). */
     if ( iommu_enabled )
@@ -883,10 +902,10 @@ int __init construct_dom0(
     for ( i = 0; i < XEN_LEGACY_MAX_VCPUS; i++ )
         shared_info(d, vcpu_info[i].evtchn_upcall_mask) = 1;
 
-    printk("Dom0 has maximum %u VCPUs\n", opt_dom0_max_vcpus);
+    printk("Dom0 has maximum %u VCPUs\n", d->max_vcpus);
 
     cpu = cpumask_first(cpupool0->cpu_valid);
-    for ( i = 1; i < opt_dom0_max_vcpus; i++ )
+    for ( i = 1; i < d->max_vcpus; i++ )
     {
         cpu = cpumask_cycle(cpu, cpupool0->cpu_valid);
         (void)alloc_vcpu(d, i, cpu);
-- 
1.7.2.5

Re: [PATCH] x86: make the dom0_max_vcpus option more flexible

[Jan Beulich]

>>> On 10.09.12 at 19:26, David Vrabel <david.vrabel@citrix.com> wrote:
> From: David Vrabel <david.vrabel@citrix.com>
> 
> The dom0_max_vcpus command line option only allows the exact number of
> VCPUs for dom0 to be set.  It is not possible to say "up to N VCPUs
> but no more than the number physically present."
> 
> Allow a range for the option to set a minimum number of VCPUs, and a
> maximum which does not exceed the number of PCPUs.
> 
> For example, with "dom0_max_vcpus=4-8":
> 
>     PCPUs  Dom0 VCPUs
>      2      4
>      4      4
>      6      6
>      8      8
>     10      8
> 
> Existing command lines with "dom0_max_vcpus=N" still work as before
> (and are equivalent to dom0_max_vcpus=N-N).
> 
> Signed-off-by: David Vrabel <david.vrabel@citrix.com>
> Acked-by: Jan Beulich <jbeulich@suse.com>
> ---
> Changes since v2:
>   - none, repost for Xen 4.3

Thanks for resending - it didn't get lost, just didn't get around
to apply it yet as I first wanted to have it in my local tree for
a short while.

Jan