[Qemu-devel] [PATCH] i386/kvm: Fix build with -m32

[Qemu-devel] [PATCH] i386/kvm: Fix build with -m32

[Max Reitz]

find_next_bit() takes a pointer of type "const unsigned long *", but the
first argument passed here is a "uint64_t *".  These types are
incompatible when compiling qemu with -m32.

Just cast it to "const void *", find_next_bit() works fine with any type
on little-endian hosts (which x86 is).

Fixes: c686193072a47032d83cb4e131dc49ae30f9e5d
Signed-off-by: Max Reitz <mreitz@redhat.com>
---
 target/i386/kvm.c | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/target/i386/kvm.c b/target/i386/kvm.c
index e4b4f5756a..1b5f3b1c00 100644
--- a/target/i386/kvm.c
+++ b/target/i386/kvm.c
@@ -1050,7 +1050,7 @@ static int hv_cpuid_check_and_set(CPUState *cs, struct kvm_cpuid2 *cpuid,
     }
 
     deps = kvm_hyperv_properties[feature].dependencies;
-    while ((dep_feat = find_next_bit(&deps, 64, dep_feat)) < 64) {
+    while ((dep_feat = find_next_bit((const void *)&deps, 64, dep_feat)) < 64) {
         if (!(hyperv_feat_enabled(cpu, dep_feat))) {
                 fprintf(stderr,
                         "Hyper-V %s requires Hyper-V %s\n",
-- 
2.21.0

Re: [Qemu-devel] [PATCH] i386/kvm: Fix build with -m32

[Eduardo Habkost]

On Mon, Jun 24, 2019 at 09:02:14PM +0200, Max Reitz wrote:
> find_next_bit() takes a pointer of type "const unsigned long *", but the
> first argument passed here is a "uint64_t *".  These types are
> incompatible when compiling qemu with -m32.
> 
> Just cast it to "const void *", find_next_bit() works fine with any type
> on little-endian hosts (which x86 is).
> 
> Fixes: c686193072a47032d83cb4e131dc49ae30f9e5d
> Signed-off-by: Max Reitz <mreitz@redhat.com>

Why not declare kvm_hyperv_properties.dependencies with the right
type for bitmaps, using
  unsigned long dependencies[BITS_TO_LONGS(64)]
?

> ---
>  target/i386/kvm.c | 2 +-
>  1 file changed, 1 insertion(+), 1 deletion(-)
> 
> diff --git a/target/i386/kvm.c b/target/i386/kvm.c
> index e4b4f5756a..1b5f3b1c00 100644
> --- a/target/i386/kvm.c
> +++ b/target/i386/kvm.c
> @@ -1050,7 +1050,7 @@ static int hv_cpuid_check_and_set(CPUState *cs, struct kvm_cpuid2 *cpuid,
>      }
>  
>      deps = kvm_hyperv_properties[feature].dependencies;
> -    while ((dep_feat = find_next_bit(&deps, 64, dep_feat)) < 64) {
> +    while ((dep_feat = find_next_bit((const void *)&deps, 64, dep_feat)) < 64) {
>          if (!(hyperv_feat_enabled(cpu, dep_feat))) {
>                  fprintf(stderr,
>                          "Hyper-V %s requires Hyper-V %s\n",
> -- 
> 2.21.0
> 

-- 
Eduardo

Re: [Qemu-devel] [PATCH] i386/kvm: Fix build with -m32

[Max Reitz]

On 24.06.19 21:21, Eduardo Habkost wrote:
> On Mon, Jun 24, 2019 at 09:02:14PM +0200, Max Reitz wrote:
>> find_next_bit() takes a pointer of type "const unsigned long *", but the
>> first argument passed here is a "uint64_t *".  These types are
>> incompatible when compiling qemu with -m32.
>>
>> Just cast it to "const void *", find_next_bit() works fine with any type
>> on little-endian hosts (which x86 is).
>>
>> Fixes: c686193072a47032d83cb4e131dc49ae30f9e5d
>> Signed-off-by: Max Reitz <mreitz@redhat.com>
> 
> Why not declare kvm_hyperv_properties.dependencies with the right
> type for bitmaps, using
>   unsigned long dependencies[BITS_TO_LONGS(64)]
> ?

How would you (statically) initialize that field, then?

I cannot imagine a reasonable static way that does not invoke the same
“The host must be little-endian, so it’s OK” assumption.

The better question is perhaps, why not use ffsll().  Hm.  I don’t know,
maybe I should?

Max

Re: [Qemu-devel] [PATCH] i386/kvm: Fix build with -m32

[Max Reitz]

On 24.06.19 21:26, Max Reitz wrote:
> On 24.06.19 21:21, Eduardo Habkost wrote:
>> On Mon, Jun 24, 2019 at 09:02:14PM +0200, Max Reitz wrote:
>>> find_next_bit() takes a pointer of type "const unsigned long *", but the
>>> first argument passed here is a "uint64_t *".  These types are
>>> incompatible when compiling qemu with -m32.
>>>
>>> Just cast it to "const void *", find_next_bit() works fine with any type
>>> on little-endian hosts (which x86 is).
>>>
>>> Fixes: c686193072a47032d83cb4e131dc49ae30f9e5d
>>> Signed-off-by: Max Reitz <mreitz@redhat.com>
>>
>> Why not declare kvm_hyperv_properties.dependencies with the right
>> type for bitmaps, using
>>   unsigned long dependencies[BITS_TO_LONGS(64)]
>> ?
> 
> How would you (statically) initialize that field, then?
> 
> I cannot imagine a reasonable static way that does not invoke the same
> “The host must be little-endian, so it’s OK” assumption.

Sorry, brain fart.  That’s not the problem because in either case, the
lower index will receive the lower-indexed bits.

But we’d still have to deal with the fact that it could either be one or
two indices, which doesn’t seem nice to initialize either.

Max

> The better question is perhaps, why not use ffsll().  Hm.  I don’t know,
> maybe I should?
> 
> Max
> 

Re: [Qemu-devel] [PATCH] i386/kvm: Fix build with -m32

[Eduardo Habkost]

On Mon, Jun 24, 2019 at 09:30:26PM +0200, Max Reitz wrote:
> On 24.06.19 21:26, Max Reitz wrote:
> > On 24.06.19 21:21, Eduardo Habkost wrote:
> >> On Mon, Jun 24, 2019 at 09:02:14PM +0200, Max Reitz wrote:
> >>> find_next_bit() takes a pointer of type "const unsigned long *", but the
> >>> first argument passed here is a "uint64_t *".  These types are
> >>> incompatible when compiling qemu with -m32.
> >>>
> >>> Just cast it to "const void *", find_next_bit() works fine with any type
> >>> on little-endian hosts (which x86 is).
> >>>
> >>> Fixes: c686193072a47032d83cb4e131dc49ae30f9e5d
> >>> Signed-off-by: Max Reitz <mreitz@redhat.com>
> >>
> >> Why not declare kvm_hyperv_properties.dependencies with the right
> >> type for bitmaps, using
> >>   unsigned long dependencies[BITS_TO_LONGS(64)]
> >> ?
> > 
> > How would you (statically) initialize that field, then?
> > 
> > I cannot imagine a reasonable static way that does not invoke the same
> > “The host must be little-endian, so it’s OK” assumption.
> 
> Sorry, brain fart.  That’s not the problem because in either case, the
> lower index will receive the lower-indexed bits.
> 
> But we’d still have to deal with the fact that it could either be one or
> two indices, which doesn’t seem nice to initialize either.

Right, a uint64_t field is more convenient to initialize.

> 
> Max
> 
> > The better question is perhaps, why not use ffsll().  Hm.  I don’t know,
> > maybe I should?

uint64_t + ffsll() seems simple and appropriate.

-- 
Eduardo