[PATCH] Account for Fedora kernels with backported vzalloc

[PATCH] Account for Fedora kernels with backported vzalloc

[Bernhard Kohl]

kvm-kmod-2.6.38.6 does not compile on Fedora kernels >= 2.6.35.11.

Signed-off-by: Bernhard Kohl <bernhard.kohl@nsn.com>
---
 configure                     |    8 ++++++++
 external-module-compat-comm.h |    5 ++++-
 2 files changed, 12 insertions(+), 1 deletions(-)

diff --git a/configure b/configure
index a40c39f..b91587c 100755
--- a/configure
+++ b/configure
@@ -153,6 +153,13 @@ if [ -n "$no_uname" -a "$want_module" ]; then
 	depmod_version=$kernel_version
 fi
 
+# Check if it is a Fedora kernel, e.g. 2.6.35.13-91.fc14.x86_64
+if echo "$kernel_version" | grep -qE '[0-9.-]+\.fc[0-9]+\..+'; then
+    config_fedora_kernel="#define CONFIG_FEDORA_KERNEL 1"
+else
+    config_fedora_kernel="#undef CONFIG_FEDORA_KERNEL"
+fi
+
 rm -f include/asm include-compat/asm
 mkdir -p include
 ln -sf asm-"$karch" include/asm
@@ -175,4 +182,5 @@ EOF
 
 cat <<EOF > kvm-kmod-config.h
 #define KERNEL_EXTRAVERSION $kernel_extraversion
+$config_fedora_kernel
 EOF
diff --git a/external-module-compat-comm.h b/external-module-compat-comm.h
index 6b6bddb..b7d9b25 100644
--- a/external-module-compat-comm.h
+++ b/external-module-compat-comm.h
@@ -881,7 +881,10 @@ static inline u32 hash_32(u32 val, unsigned int bits)
 #define __rcu
 #endif
 
-#if LINUX_VERSION_CODE < KERNEL_VERSION(2,6,37)
+#if LINUX_VERSION_CODE < KERNEL_VERSION(2,6,37) && \
+    (!defined(CONFIG_FEDORA_KERNEL) || \
+     LINUX_VERSION_CODE == KERNEL_VERSION(2,6,35) && KERNEL_EXTRAVERSION < 11 || \
+     LINUX_VERSION_CODE < KERNEL_VERSION(2,6,35))
 #include <linux/vmalloc.h>
 static inline void *vzalloc(unsigned long size)
 {
-- 
1.7.4.4

Re: [PATCH] Account for Fedora kernels with backported vzalloc

[Jan Kiszka]

On 2011-05-17 16:55, Bernhard Kohl wrote:
> kvm-kmod-2.6.38.6 does not compile on Fedora kernels >= 2.6.35.11.
> 
> Signed-off-by: Bernhard Kohl <bernhard.kohl@nsn.com>
> ---
>  configure                     |    8 ++++++++
>  external-module-compat-comm.h |    5 ++++-
>  2 files changed, 12 insertions(+), 1 deletions(-)
> 
> diff --git a/configure b/configure
> index a40c39f..b91587c 100755
> --- a/configure
> +++ b/configure
> @@ -153,6 +153,13 @@ if [ -n "$no_uname" -a "$want_module" ]; then
>  	depmod_version=$kernel_version
>  fi
>  
> +# Check if it is a Fedora kernel, e.g. 2.6.35.13-91.fc14.x86_64
> +if echo "$kernel_version" | grep -qE '[0-9.-]+\.fc[0-9]+\..+'; then
> +    config_fedora_kernel="#define CONFIG_FEDORA_KERNEL 1"
> +else
> +    config_fedora_kernel="#undef CONFIG_FEDORA_KERNEL"
> +fi
> +

Is there no define set by fedora kernels themselves?

>  rm -f include/asm include-compat/asm
>  mkdir -p include
>  ln -sf asm-"$karch" include/asm
> @@ -175,4 +182,5 @@ EOF
>  
>  cat <<EOF > kvm-kmod-config.h
>  #define KERNEL_EXTRAVERSION $kernel_extraversion
> +$config_fedora_kernel
>  EOF
> diff --git a/external-module-compat-comm.h b/external-module-compat-comm.h
> index 6b6bddb..b7d9b25 100644
> --- a/external-module-compat-comm.h
> +++ b/external-module-compat-comm.h
> @@ -881,7 +881,10 @@ static inline u32 hash_32(u32 val, unsigned int bits)
>  #define __rcu
>  #endif
>  
> -#if LINUX_VERSION_CODE < KERNEL_VERSION(2,6,37)
> +#if LINUX_VERSION_CODE < KERNEL_VERSION(2,6,37) && \
> +    (!defined(CONFIG_FEDORA_KERNEL) || \
> +     LINUX_VERSION_CODE == KERNEL_VERSION(2,6,35) && KERNEL_EXTRAVERSION < 11 || \
> +     LINUX_VERSION_CODE < KERNEL_VERSION(2,6,35))
>  #include <linux/vmalloc.h>
>  static inline void *vzalloc(unsigned long size)
>  {

Is this intentionally or accidentally targeting Fedora 2.6.35.x only?
Because the commit log may imply some 2.6.36-fc could carry it as well
(if they exist).

Thanks,
Jan

-- 
Siemens AG, Corporate Technology, CT T DE IT 1
Corporate Competence Center Embedded Linux

RE: [PATCH] Account for Fedora kernels with backported vzalloc

[Kohl, Bernhard (NSN - DE/Munich)]

> -----Original Message-----
> From: ext Jan Kiszka [mailto:jan.kiszka@siemens.com]
> Sent: Tuesday, May 17, 2011 5:27 PM
> To: Kohl, Bernhard (NSN - DE/Munich)
> Cc: kvm@vger.kernel.org
> Subject: Re: [PATCH] Account for Fedora kernels with backported
vzalloc
> 
> On 2011-05-17 16:55, Bernhard Kohl wrote:
> > kvm-kmod-2.6.38.6 does not compile on Fedora kernels >= 2.6.35.11.
> >
> > Signed-off-by: Bernhard Kohl <bernhard.kohl@nsn.com>
> > ---
> >  configure                     |    8 ++++++++
> >  external-module-compat-comm.h |    5 ++++-
> >  2 files changed, 12 insertions(+), 1 deletions(-)
> >
> > diff --git a/configure b/configure
> > index a40c39f..b91587c 100755
> > --- a/configure
> > +++ b/configure
> > @@ -153,6 +153,13 @@ if [ -n "$no_uname" -a "$want_module" ]; then
> >  	depmod_version=$kernel_version
> >  fi
> >
> > +# Check if it is a Fedora kernel, e.g. 2.6.35.13-91.fc14.x86_64
> > +if echo "$kernel_version" | grep -qE '[0-9.-]+\.fc[0-9]+\..+'; then
> > +    config_fedora_kernel="#define CONFIG_FEDORA_KERNEL 1"
> > +else
> > +    config_fedora_kernel="#undef CONFIG_FEDORA_KERNEL"
> > +fi
> > +
> 
> Is there no define set by fedora kernels themselves?

No, there is nothing in the include/generated/autoconf.h
to identify a Fedora kernel. I assume this is the place
where CONFIG_SUSE_KERNEL comes from.

> 
> >  rm -f include/asm include-compat/asm
> >  mkdir -p include
> >  ln -sf asm-"$karch" include/asm
> > @@ -175,4 +182,5 @@ EOF
> >
> >  cat <<EOF > kvm-kmod-config.h
> >  #define KERNEL_EXTRAVERSION $kernel_extraversion
> > +$config_fedora_kernel
> >  EOF
> > diff --git a/external-module-compat-comm.h b/external-module-compat-
> comm.h
> > index 6b6bddb..b7d9b25 100644
> > --- a/external-module-compat-comm.h
> > +++ b/external-module-compat-comm.h
> > @@ -881,7 +881,10 @@ static inline u32 hash_32(u32 val, unsigned int
> bits)
> >  #define __rcu
> >  #endif
> >
> > -#if LINUX_VERSION_CODE < KERNEL_VERSION(2,6,37)
> > +#if LINUX_VERSION_CODE < KERNEL_VERSION(2,6,37) && \
> > +    (!defined(CONFIG_FEDORA_KERNEL) || \
> > +     LINUX_VERSION_CODE == KERNEL_VERSION(2,6,35) &&
> KERNEL_EXTRAVERSION < 11 || \
> > +     LINUX_VERSION_CODE < KERNEL_VERSION(2,6,35))
> >  #include <linux/vmalloc.h>
> >  static inline void *vzalloc(unsigned long size)
> >  {
> 
> Is this intentionally or accidentally targeting Fedora 2.6.35.x only?
> Because the commit log may imply some 2.6.36-fc could carry it as well
> (if they exist).

I assume Fedora 14 will stay with kernel 2.6.35 and Fedora 15 starts
with kernel 2.6.38 where the problem does not exist anyway.

> 
> Thanks,
> Jan
> 
> --
> Siemens AG, Corporate Technology, CT T DE IT 1
> Corporate Competence Center Embedded Linux

Re: [PATCH] Account for Fedora kernels with backported vzalloc

[Jan Kiszka]

On 2011-05-17 17:44, Kohl, Bernhard (NSN - DE/Munich) wrote:
>> -----Original Message-----
>> From: ext Jan Kiszka [mailto:jan.kiszka@siemens.com]
>> Sent: Tuesday, May 17, 2011 5:27 PM
>> To: Kohl, Bernhard (NSN - DE/Munich)
>> Cc: kvm@vger.kernel.org
>> Subject: Re: [PATCH] Account for Fedora kernels with backported
> vzalloc
>>
>> On 2011-05-17 16:55, Bernhard Kohl wrote:
>>> kvm-kmod-2.6.38.6 does not compile on Fedora kernels >= 2.6.35.11.
>>>
>>> Signed-off-by: Bernhard Kohl <bernhard.kohl@nsn.com>
>>> ---
>>>  configure                     |    8 ++++++++
>>>  external-module-compat-comm.h |    5 ++++-
>>>  2 files changed, 12 insertions(+), 1 deletions(-)
>>>
>>> diff --git a/configure b/configure
>>> index a40c39f..b91587c 100755
>>> --- a/configure
>>> +++ b/configure
>>> @@ -153,6 +153,13 @@ if [ -n "$no_uname" -a "$want_module" ]; then
>>>  	depmod_version=$kernel_version
>>>  fi
>>>
>>> +# Check if it is a Fedora kernel, e.g. 2.6.35.13-91.fc14.x86_64
>>> +if echo "$kernel_version" | grep -qE '[0-9.-]+\.fc[0-9]+\..+'; then
>>> +    config_fedora_kernel="#define CONFIG_FEDORA_KERNEL 1"
>>> +else
>>> +    config_fedora_kernel="#undef CONFIG_FEDORA_KERNEL"
>>> +fi
>>> +
>>
>> Is there no define set by fedora kernels themselves?
> 
> No, there is nothing in the include/generated/autoconf.h
> to identify a Fedora kernel. I assume this is the place
> where CONFIG_SUSE_KERNEL comes from.
> 
>>
>>>  rm -f include/asm include-compat/asm
>>>  mkdir -p include
>>>  ln -sf asm-"$karch" include/asm
>>> @@ -175,4 +182,5 @@ EOF
>>>
>>>  cat <<EOF > kvm-kmod-config.h
>>>  #define KERNEL_EXTRAVERSION $kernel_extraversion
>>> +$config_fedora_kernel
>>>  EOF
>>> diff --git a/external-module-compat-comm.h b/external-module-compat-
>> comm.h
>>> index 6b6bddb..b7d9b25 100644
>>> --- a/external-module-compat-comm.h
>>> +++ b/external-module-compat-comm.h
>>> @@ -881,7 +881,10 @@ static inline u32 hash_32(u32 val, unsigned int
>> bits)
>>>  #define __rcu
>>>  #endif
>>>
>>> -#if LINUX_VERSION_CODE < KERNEL_VERSION(2,6,37)
>>> +#if LINUX_VERSION_CODE < KERNEL_VERSION(2,6,37) && \
>>> +    (!defined(CONFIG_FEDORA_KERNEL) || \
>>> +     LINUX_VERSION_CODE == KERNEL_VERSION(2,6,35) &&
>> KERNEL_EXTRAVERSION < 11 || \
>>> +     LINUX_VERSION_CODE < KERNEL_VERSION(2,6,35))
>>>  #include <linux/vmalloc.h>
>>>  static inline void *vzalloc(unsigned long size)
>>>  {
>>
>> Is this intentionally or accidentally targeting Fedora 2.6.35.x only?
>> Because the commit log may imply some 2.6.36-fc could carry it as well
>> (if they exist).
> 
> I assume Fedora 14 will stay with kernel 2.6.35 and Fedora 15 starts
> with kernel 2.6.38 where the problem does not exist anyway.
> 

Ok, fine. I've merge it, just with a small reformatting.

Jan

-- 
Siemens AG, Corporate Technology, CT T DE IT 1
Corporate Competence Center Embedded Linux

Re: [PATCH] Account for Fedora kernels with backported vzalloc

[Avi Kivity]

On 05/17/2011 05:55 PM, Bernhard Kohl wrote:
> kvm-kmod-2.6.38.6 does not compile on Fedora kernels>= 2.6.35.11.

> index a40c39f..b91587c 100755
> --- a/configure
> +++ b/configure
> @@ -153,6 +153,13 @@ if [ -n "$no_uname" -a "$want_module" ]; then
>   	depmod_version=$kernel_version
>   fi
>
> +# Check if it is a Fedora kernel, e.g. 2.6.35.13-91.fc14.x86_64
> +if echo "$kernel_version" | grep -qE '[0-9.-]+\.fc[0-9]+\..+'; then
> +    config_fedora_kernel="#define CONFIG_FEDORA_KERNEL 1"
> +else
> +    config_fedora_kernel="#undef CONFIG_FEDORA_KERNEL"
> +fi
> +

Curious, why are you targetting Fedora kernels at all?  They have a 
really short shelf life.  I though kvm-kmod was for people using longer 
term kernels like enterprise distros or long lived embedded projects.

For Fedora, you can use the native kvm, or install a rawhide kernel, or 
an upstream kernel with 'make localmodconfig'.

-- 
error compiling committee.c: too many arguments to function

RE: [PATCH] Account for Fedora kernels with backported vzalloc

[Kohl, Bernhard (NSN - DE/Munich)]

> -----Original Message-----
> From: ext Avi Kivity [mailto:avi@redhat.com]
> Sent: Tuesday, May 17, 2011 7:30 PM
> To: Kohl, Bernhard (NSN - DE/Munich)
> Cc: kvm@vger.kernel.org; jan.kiszka@siemens.com
> Subject: Re: [PATCH] Account for Fedora kernels with backported
vzalloc
> 
> On 05/17/2011 05:55 PM, Bernhard Kohl wrote:
> > kvm-kmod-2.6.38.6 does not compile on Fedora kernels>= 2.6.35.11.
> 
> > index a40c39f..b91587c 100755
> > --- a/configure
> > +++ b/configure
> > @@ -153,6 +153,13 @@ if [ -n "$no_uname" -a "$want_module" ]; then
> >   	depmod_version=$kernel_version
> >   fi
> >
> > +# Check if it is a Fedora kernel, e.g. 2.6.35.13-91.fc14.x86_64
> > +if echo "$kernel_version" | grep -qE '[0-9.-]+\.fc[0-9]+\..+'; then
> > +    config_fedora_kernel="#define CONFIG_FEDORA_KERNEL 1"
> > +else
> > +    config_fedora_kernel="#undef CONFIG_FEDORA_KERNEL"
> > +fi
> > +
> 
> Curious, why are you targetting Fedora kernels at all?  They have a
> really short shelf life.  I though kvm-kmod was for people using
longer
> term kernels like enterprise distros or long lived embedded projects.

Here at NSN a couple of developers and testers are using Fedora +
kvm-kmod
to run some of our systems (with in-house developed OS) on KVM. This
works
also without kvm-kmod since kernel 2.6.35 (Fedora 14). The number of
users
(which are not Linux experts at all) is growing since over a year.
Usually
I prepare and test new versions of kvm-kmod and then it's easy for these
guys to install it on their machines. So I always keep track to be
up-to-date with the upstream kvm-kmod.

The next candidate for kvm-kmod usage here might be your patch set
"KVM in-guest performance monitoring", but I think we need a complete
kernel for this as major parts of these patches are outside KVM. Some
people here are eager for this new feature. Accidentally my colleague
Thomas started investigating the performance counter topic 2 weeks ago.

Nevertheless I hope that kvm-kmod will live for a while. It is quite
useful for our work.

> 
> For Fedora, you can use the native kvm, or install a rawhide kernel,
or
> an upstream kernel with 'make localmodconfig'.

Thanks for this hint. I did not yet try localmodconfig.

Bernhard

> 
> --
> error compiling committee.c: too many arguments to function

Re: [PATCH] Account for Fedora kernels with backported vzalloc

[Avi Kivity]

On 05/18/2011 12:11 PM, Kohl, Bernhard (NSN - DE/Munich) wrote:
> >
> >  Curious, why are you targetting Fedora kernels at all?  They have a
> >  really short shelf life.  I though kvm-kmod was for people using
> longer
> >  term kernels like enterprise distros or long lived embedded projects.
>
> Here at NSN a couple of developers and testers are using Fedora +
> kvm-kmod
> to run some of our systems (with in-house developed OS) on KVM. This
> works
> also without kvm-kmod since kernel 2.6.35 (Fedora 14). The number of
> users
> (which are not Linux experts at all) is growing since over a year.
> Usually
> I prepare and test new versions of kvm-kmod and then it's easy for these
> guys to install it on their machines. So I always keep track to be
> up-to-date with the upstream kvm-kmod.
>
> The next candidate for kvm-kmod usage here might be your patch set
> "KVM in-guest performance monitoring", but I think we need a complete
> kernel for this as major parts of these patches are outside KVM. Some
> people here are eager for this new feature. Accidentally my colleague
> Thomas started investigating the performance counter topic 2 weeks ago.
>
> Nevertheless I hope that kvm-kmod will live for a while. It is quite
> useful for our work.

Yes, I can see how it helps your workflow.  I guess the only viable 
alternative would be to prepare a full kernel rpm, but that is a lot 
more work.

Regarding the PMU, the patchset modifies the core perf_event code and 
exports some symbols.  We might be able to work around these issues but 
this isn't certain.

-- 
error compiling committee.c: too many arguments to function