[PATCH v2 0/1] spapr.c: set a 'kvm-type' default value instead of relying on NULL

[PATCH v2 0/1] spapr.c: set a 'kvm-type' default value instead of relying on NULL

[Daniel Henrique Barboza]

changes from v1, all proposed by Greg:
* Removed comment at the start of spapr_get_kvm_type()
* Added a one liner return in spapr_get_kvm_type()
* use g_ascii_strcasecmp() in all str comparisons in spapr_kvm_type()
* Exposed the 'auto' default kvm-type mode in the docs
* changed HV and PR to lower-case in the docs
v1 link: https://lists.gnu.org/archive/html/qemu-devel/2020-12/msg01677.html


This patch addresses an issue that happens with the pseries machine after
testing Paolo's patch [1]:

$ sudo ./ppc64-softmmu/qemu-system-ppc64 -nographic -nodefaults -machine pseries --enable-kvm
qemu-system-ppc64: Unknown kvm-type specified '' 

The reason lies on how qemu_opt_get() and object_property_get_str() works
when there is no 'kvm-type' specified. We were conting on receiving NULL
for kvm-type, but the latter will use a blank string "". Instead on relying
on NULL, let's expose the already existing 'auto' kvm-type mode to the users
and use that as default.

[1] https://lists.gnu.org/archive/html/qemu-devel/2020-12/msg00471.html

Daniel Henrique Barboza (1):
  spapr.c: set a 'kvm-type' default value instead of relying on NULL

 hw/ppc/spapr.c | 15 ++++++++++-----
 1 file changed, 10 insertions(+), 5 deletions(-)

-- 
2.26.2

[PATCH v2 1/1] spapr.c: set a 'kvm-type' default value instead of relying on NULL

[Daniel Henrique Barboza]

spapr_kvm_type() is considering 'vm_type=NULL' as a valid input, where
the function returns 0. This is relying on the current QEMU machine
options handling logic, where the absence of the 'kvm-type' option
will be reflected as 'vm_type=NULL' in this function.

This is not robust, and will break if QEMU options code decides to propagate
something else in the case mentioned above (e.g. an empty string instead
of NULL).

Let's avoid this entirely by setting a non-NULL default value in case of
no user input for 'kvm-type'. spapr_kvm_type() was changed to handle 3 fixed
values of kvm-type: "auto", "hv", and "pr", with "auto" being the default
if no kvm-type was set by the user. This allows us to always be predictable
regardless of any enhancements/changes made in QEMU options mechanics.

While we're at it, let's also document in 'kvm-type' description the
already existing default mode, now named 'auto'. The information provided
about it is based on how the pseries kernel handles the KVM_CREATE_VM
ioctl(), where the default value '0' makes the kernel choose an available
KVM module to use, giving precedence to kvm_hv. This logic is described in
the kernel source file arch/powerpc/kvm/powerpc.c, function kvm_arch_init_vm().

Signed-off-by: Daniel Henrique Barboza <danielhb413@gmail.com>
---
 hw/ppc/spapr.c | 15 ++++++++++-----
 1 file changed, 10 insertions(+), 5 deletions(-)

diff --git a/hw/ppc/spapr.c b/hw/ppc/spapr.c
index b7e0894019..f097f4ea30 100644
--- a/hw/ppc/spapr.c
+++ b/hw/ppc/spapr.c
@@ -3021,17 +3021,18 @@ static void spapr_machine_init(MachineState *machine)
     qemu_cond_init(&spapr->fwnmi_machine_check_interlock_cond);
 }
 
+#define DEFAULT_KVM_TYPE "auto"
 static int spapr_kvm_type(MachineState *machine, const char *vm_type)
 {
-    if (!vm_type) {
+    if (!g_ascii_strcasecmp(vm_type, DEFAULT_KVM_TYPE)) {
         return 0;
     }
 
-    if (!strcmp(vm_type, "HV")) {
+    if (!g_ascii_strcasecmp(vm_type, "hv")) {
         return 1;
     }
 
-    if (!strcmp(vm_type, "PR")) {
+    if (!g_ascii_strcasecmp(vm_type, "pr")) {
         return 2;
     }
 
@@ -3131,7 +3132,7 @@ static char *spapr_get_kvm_type(Object *obj, Error **errp)
 {
     SpaprMachineState *spapr = SPAPR_MACHINE(obj);
 
-    return g_strdup(spapr->kvm_type);
+    return g_strdup(spapr->kvm_type ? spapr->kvm_type : DEFAULT_KVM_TYPE);
 }
 
 static void spapr_set_kvm_type(Object *obj, const char *value, Error **errp)
@@ -3273,7 +3274,11 @@ static void spapr_instance_init(Object *obj)
     object_property_add_str(obj, "kvm-type",
                             spapr_get_kvm_type, spapr_set_kvm_type);
     object_property_set_description(obj, "kvm-type",
-                                    "Specifies the KVM virtualization mode (HV, PR)");
+                                    "Specifies the KVM virtualization mode (auto,"
+                                    " hv, pr). Defaults to 'auto'. This mode will use"
+                                    " any available KVM module loaded in the host,"
+                                    " where kvm_hv takes precedence if both kvm_hv and"
+                                    " kvm_pr are loaded.");
     object_property_add_bool(obj, "modern-hotplug-events",
                             spapr_get_modern_hotplug_events,
                             spapr_set_modern_hotplug_events);
-- 
2.26.2

Re: [PATCH v2 1/1] spapr.c: set a 'kvm-type' default value instead of relying on NULL

[Greg Kurz]

On Thu, 10 Dec 2020 10:07:21 -0300
Daniel Henrique Barboza <danielhb413@gmail.com> wrote:

> spapr_kvm_type() is considering 'vm_type=NULL' as a valid input, where
> the function returns 0. This is relying on the current QEMU machine
> options handling logic, where the absence of the 'kvm-type' option
> will be reflected as 'vm_type=NULL' in this function.
> 
> This is not robust, and will break if QEMU options code decides to propagate
> something else in the case mentioned above (e.g. an empty string instead
> of NULL).
> 
> Let's avoid this entirely by setting a non-NULL default value in case of
> no user input for 'kvm-type'. spapr_kvm_type() was changed to handle 3 fixed
> values of kvm-type: "auto", "hv", and "pr", with "auto" being the default
> if no kvm-type was set by the user. This allows us to always be predictable
> regardless of any enhancements/changes made in QEMU options mechanics.
> 
> While we're at it, let's also document in 'kvm-type' description the
> already existing default mode, now named 'auto'. The information provided
> about it is based on how the pseries kernel handles the KVM_CREATE_VM
> ioctl(), where the default value '0' makes the kernel choose an available
> KVM module to use, giving precedence to kvm_hv. This logic is described in
> the kernel source file arch/powerpc/kvm/powerpc.c, function kvm_arch_init_vm().
> 
> Signed-off-by: Daniel Henrique Barboza <danielhb413@gmail.com>
> ---
>  hw/ppc/spapr.c | 15 ++++++++++-----
>  1 file changed, 10 insertions(+), 5 deletions(-)
> 
> diff --git a/hw/ppc/spapr.c b/hw/ppc/spapr.c
> index b7e0894019..f097f4ea30 100644
> --- a/hw/ppc/spapr.c
> +++ b/hw/ppc/spapr.c
> @@ -3021,17 +3021,18 @@ static void spapr_machine_init(MachineState *machine)
>      qemu_cond_init(&spapr->fwnmi_machine_check_interlock_cond);
>  }
>  
> +#define DEFAULT_KVM_TYPE "auto"
>  static int spapr_kvm_type(MachineState *machine, const char *vm_type)
>  {
> -    if (!vm_type) {
> +    if (!g_ascii_strcasecmp(vm_type, DEFAULT_KVM_TYPE)) {

strcmp() would be better here : we don't want to support an 
already existing "AUTO" value like for the other ones.

And, of course, add a comment to explain why the two other
ones are handled differently.

>          return 0;
>      }
>  
> -    if (!strcmp(vm_type, "HV")) {
> +    if (!g_ascii_strcasecmp(vm_type, "hv")) {
>          return 1;
>      }
>  
> -    if (!strcmp(vm_type, "PR")) {
> +    if (!g_ascii_strcasecmp(vm_type, "pr")) {
>          return 2;
>      }
>  
> @@ -3131,7 +3132,7 @@ static char *spapr_get_kvm_type(Object *obj, Error **errp)
>  {
>      SpaprMachineState *spapr = SPAPR_MACHINE(obj);
>  
> -    return g_strdup(spapr->kvm_type);
> +    return g_strdup(spapr->kvm_type ? spapr->kvm_type : DEFAULT_KVM_TYPE);

As in Paolo's diff, it seems better to simply initialize spapr->kvm_type ...

>  }
>  
>  static void spapr_set_kvm_type(Object *obj, const char *value, Error **errp)
> @@ -3273,7 +3274,11 @@ static void spapr_instance_init(Object *obj)

... here.

    spapr->kvm_type = g_strdup(DEFAULT_KVM_TYPE);

>      object_property_add_str(obj, "kvm-type",
>                              spapr_get_kvm_type, spapr_set_kvm_type);
>      object_property_set_description(obj, "kvm-type",
> -                                    "Specifies the KVM virtualization mode (HV, PR)");
> +                                    "Specifies the KVM virtualization mode (auto,"
> +                                    " hv, pr). Defaults to 'auto'. This mode will use"
> +                                    " any available KVM module loaded in the host,"
> +                                    " where kvm_hv takes precedence if both kvm_hv and"
> +                                    " kvm_pr are loaded.");
>      object_property_add_bool(obj, "modern-hotplug-events",
>                              spapr_get_modern_hotplug_events,
>                              spapr_set_modern_hotplug_events);

Re: [PATCH v2 1/1] spapr.c: set a 'kvm-type' default value instead of relying on NULL

[Greg Kurz]

On Thu, 10 Dec 2020 15:01:20 +0100
Greg Kurz <groug@kaod.org> wrote:

> On Thu, 10 Dec 2020 10:07:21 -0300
> Daniel Henrique Barboza <danielhb413@gmail.com> wrote:
> 
> > spapr_kvm_type() is considering 'vm_type=NULL' as a valid input, where
> > the function returns 0. This is relying on the current QEMU machine
> > options handling logic, where the absence of the 'kvm-type' option
> > will be reflected as 'vm_type=NULL' in this function.
> > 
> > This is not robust, and will break if QEMU options code decides to propagate
> > something else in the case mentioned above (e.g. an empty string instead
> > of NULL).
> > 
> > Let's avoid this entirely by setting a non-NULL default value in case of
> > no user input for 'kvm-type'. spapr_kvm_type() was changed to handle 3 fixed
> > values of kvm-type: "auto", "hv", and "pr", with "auto" being the default
> > if no kvm-type was set by the user. This allows us to always be predictable
> > regardless of any enhancements/changes made in QEMU options mechanics.
> > 
> > While we're at it, let's also document in 'kvm-type' description the
> > already existing default mode, now named 'auto'. The information provided
> > about it is based on how the pseries kernel handles the KVM_CREATE_VM
> > ioctl(), where the default value '0' makes the kernel choose an available
> > KVM module to use, giving precedence to kvm_hv. This logic is described in
> > the kernel source file arch/powerpc/kvm/powerpc.c, function kvm_arch_init_vm().
> > 
> > Signed-off-by: Daniel Henrique Barboza <danielhb413@gmail.com>
> > ---
> >  hw/ppc/spapr.c | 15 ++++++++++-----
> >  1 file changed, 10 insertions(+), 5 deletions(-)
> > 
> > diff --git a/hw/ppc/spapr.c b/hw/ppc/spapr.c
> > index b7e0894019..f097f4ea30 100644
> > --- a/hw/ppc/spapr.c
> > +++ b/hw/ppc/spapr.c
> > @@ -3021,17 +3021,18 @@ static void spapr_machine_init(MachineState *machine)
> >      qemu_cond_init(&spapr->fwnmi_machine_check_interlock_cond);
> >  }
> >  
> > +#define DEFAULT_KVM_TYPE "auto"
> >  static int spapr_kvm_type(MachineState *machine, const char *vm_type)
> >  {
> > -    if (!vm_type) {
> > +    if (!g_ascii_strcasecmp(vm_type, DEFAULT_KVM_TYPE)) {
> 
> strcmp() would be better here : we don't want to support an 
> already existing "AUTO" value like for the other ones.
> 

And you should also keep the check on vm_type since it is NULL
if the kvm-type wasn't passed to the command line:

qemu-system-ppc64: GLib: g_ascii_strcasecmp: assertion 's1 != NULL' failed

> And, of course, add a comment to explain why the two other
> ones are handled differently.
> 
> >          return 0;
> >      }
> >  
> > -    if (!strcmp(vm_type, "HV")) {
> > +    if (!g_ascii_strcasecmp(vm_type, "hv")) {
> >          return 1;
> >      }
> >  
> > -    if (!strcmp(vm_type, "PR")) {
> > +    if (!g_ascii_strcasecmp(vm_type, "pr")) {
> >          return 2;
> >      }
> >  
> > @@ -3131,7 +3132,7 @@ static char *spapr_get_kvm_type(Object *obj, Error **errp)
> >  {
> >      SpaprMachineState *spapr = SPAPR_MACHINE(obj);
> >  
> > -    return g_strdup(spapr->kvm_type);
> > +    return g_strdup(spapr->kvm_type ? spapr->kvm_type : DEFAULT_KVM_TYPE);
> 
> As in Paolo's diff, it seems better to simply initialize spapr->kvm_type ...
> 
> >  }
> >  
> >  static void spapr_set_kvm_type(Object *obj, const char *value, Error **errp)
> > @@ -3273,7 +3274,11 @@ static void spapr_instance_init(Object *obj)
> 
> ... here.
> 
>     spapr->kvm_type = g_strdup(DEFAULT_KVM_TYPE);
> 
> >      object_property_add_str(obj, "kvm-type",
> >                              spapr_get_kvm_type, spapr_set_kvm_type);
> >      object_property_set_description(obj, "kvm-type",
> > -                                    "Specifies the KVM virtualization mode (HV, PR)");
> > +                                    "Specifies the KVM virtualization mode (auto,"
> > +                                    " hv, pr). Defaults to 'auto'. This mode will use"
> > +                                    " any available KVM module loaded in the host,"
> > +                                    " where kvm_hv takes precedence if both kvm_hv and"
> > +                                    " kvm_pr are loaded.");
> >      object_property_add_bool(obj, "modern-hotplug-events",
> >                              spapr_get_modern_hotplug_events,
> >                              spapr_set_modern_hotplug_events);
> 
> 

Re: [PATCH v2 1/1] spapr.c: set a 'kvm-type' default value instead of relying on NULL

[Daniel Henrique Barboza]

On 12/10/20 11:14 AM, Greg Kurz wrote:
> On Thu, 10 Dec 2020 15:01:20 +0100
> Greg Kurz <groug@kaod.org> wrote:
> 
>> On Thu, 10 Dec 2020 10:07:21 -0300
>> Daniel Henrique Barboza <danielhb413@gmail.com> wrote:
>>
>>> spapr_kvm_type() is considering 'vm_type=NULL' as a valid input, where
>>> the function returns 0. This is relying on the current QEMU machine
>>> options handling logic, where the absence of the 'kvm-type' option
>>> will be reflected as 'vm_type=NULL' in this function.
>>>
>>> This is not robust, and will break if QEMU options code decides to propagate
>>> something else in the case mentioned above (e.g. an empty string instead
>>> of NULL).
>>>
>>> Let's avoid this entirely by setting a non-NULL default value in case of
>>> no user input for 'kvm-type'. spapr_kvm_type() was changed to handle 3 fixed
>>> values of kvm-type: "auto", "hv", and "pr", with "auto" being the default
>>> if no kvm-type was set by the user. This allows us to always be predictable
>>> regardless of any enhancements/changes made in QEMU options mechanics.
>>>
>>> While we're at it, let's also document in 'kvm-type' description the
>>> already existing default mode, now named 'auto'. The information provided
>>> about it is based on how the pseries kernel handles the KVM_CREATE_VM
>>> ioctl(), where the default value '0' makes the kernel choose an available
>>> KVM module to use, giving precedence to kvm_hv. This logic is described in
>>> the kernel source file arch/powerpc/kvm/powerpc.c, function kvm_arch_init_vm().
>>>
>>> Signed-off-by: Daniel Henrique Barboza <danielhb413@gmail.com>
>>> ---
>>>   hw/ppc/spapr.c | 15 ++++++++++-----
>>>   1 file changed, 10 insertions(+), 5 deletions(-)
>>>
>>> diff --git a/hw/ppc/spapr.c b/hw/ppc/spapr.c
>>> index b7e0894019..f097f4ea30 100644
>>> --- a/hw/ppc/spapr.c
>>> +++ b/hw/ppc/spapr.c
>>> @@ -3021,17 +3021,18 @@ static void spapr_machine_init(MachineState *machine)
>>>       qemu_cond_init(&spapr->fwnmi_machine_check_interlock_cond);
>>>   }
>>>   
>>> +#define DEFAULT_KVM_TYPE "auto"
>>>   static int spapr_kvm_type(MachineState *machine, const char *vm_type)
>>>   {
>>> -    if (!vm_type) {
>>> +    if (!g_ascii_strcasecmp(vm_type, DEFAULT_KVM_TYPE)) {
>>
>> strcmp() would be better here : we don't want to support an
>> already existing "AUTO" value like for the other ones.
>>
> 
> And you should also keep the check on vm_type since it is NULL
> if the kvm-type wasn't passed to the command line:
> 
> qemu-system-ppc64: GLib: g_ascii_strcasecmp: assertion 's1 != NULL' failed


Hm, I tested this case and didn't hit that. Perhaphs because I am testing this
on top of Paolo's patches instead of master.

I'll keep the NULL check as a fallback to 'auto' just to be safe then.


DHB

> 
>> And, of course, add a comment to explain why the two other
>> ones are handled differently.
>>
>>>           return 0;
>>>       }
>>>   
>>> -    if (!strcmp(vm_type, "HV")) {
>>> +    if (!g_ascii_strcasecmp(vm_type, "hv")) {
>>>           return 1;
>>>       }
>>>   
>>> -    if (!strcmp(vm_type, "PR")) {
>>> +    if (!g_ascii_strcasecmp(vm_type, "pr")) {
>>>           return 2;
>>>       }
>>>   
>>> @@ -3131,7 +3132,7 @@ static char *spapr_get_kvm_type(Object *obj, Error **errp)
>>>   {
>>>       SpaprMachineState *spapr = SPAPR_MACHINE(obj);
>>>   
>>> -    return g_strdup(spapr->kvm_type);
>>> +    return g_strdup(spapr->kvm_type ? spapr->kvm_type : DEFAULT_KVM_TYPE);
>>
>> As in Paolo's diff, it seems better to simply initialize spapr->kvm_type ...
>>
>>>   }
>>>   
>>>   static void spapr_set_kvm_type(Object *obj, const char *value, Error **errp)
>>> @@ -3273,7 +3274,11 @@ static void spapr_instance_init(Object *obj)
>>
>> ... here.
>>
>>      spapr->kvm_type = g_strdup(DEFAULT_KVM_TYPE);
>>
>>>       object_property_add_str(obj, "kvm-type",
>>>                               spapr_get_kvm_type, spapr_set_kvm_type);
>>>       object_property_set_description(obj, "kvm-type",
>>> -                                    "Specifies the KVM virtualization mode (HV, PR)");
>>> +                                    "Specifies the KVM virtualization mode (auto,"
>>> +                                    " hv, pr). Defaults to 'auto'. This mode will use"
>>> +                                    " any available KVM module loaded in the host,"
>>> +                                    " where kvm_hv takes precedence if both kvm_hv and"
>>> +                                    " kvm_pr are loaded.");
>>>       object_property_add_bool(obj, "modern-hotplug-events",
>>>                               spapr_get_modern_hotplug_events,
>>>                               spapr_set_modern_hotplug_events);
>>
>>
>