[Qemu-devel] [PATCH] exec: silence hugetlbfs warning under qtest

[Qemu-devel] [PATCH] exec: silence hugetlbfs warning under qtest

[marcandre.lureau]

From: Marc-André Lureau <marcandre.lureau@redhat.com>

vhost-user-test prints a warning. A test should not need to run on
hugetlbfs, let's silence the warning under qtest. Unfortunately, the
condition can't check on qtest_enabled() or qtest_driver() since they
are initialized later.

Signed-off-by: Marc-André Lureau <marcandre.lureau@redhat.com>
---
 exec.c | 3 ++-
 1 file changed, 2 insertions(+), 1 deletion(-)

diff --git a/exec.c b/exec.c
index 8af2570..d9c231d 100644
--- a/exec.c
+++ b/exec.c
@@ -1194,8 +1194,9 @@ static long gethugepagesize(const char *path, Error **errp)
         return 0;
     }
 
-    if (fs.f_type != HUGETLBFS_MAGIC)
+    if (fs.f_type != HUGETLBFS_MAGIC && !getenv("QTEST_QEMU_BINARY")) {
         fprintf(stderr, "Warning: path not on HugeTLBFS: %s\n", path);
+    }
 
     return fs.f_bsize;
 }
-- 
2.4.3

Re: [Qemu-devel] [PATCH] exec: silence hugetlbfs warning under qtest

[Michael S. Tsirkin]

On Tue, Oct 27, 2015 at 05:29:43PM +0100, marcandre.lureau@redhat.com wrote:
> From: Marc-André Lureau <marcandre.lureau@redhat.com>
> 
> vhost-user-test prints a warning. A test should not need to run on
> hugetlbfs, let's silence the warning under qtest. Unfortunately, the
> condition can't check on qtest_enabled() or qtest_driver() since they
> are initialized later.
> 
> Signed-off-by: Marc-André Lureau <marcandre.lureau@redhat.com>

Let's move the warning to somewhere later then.

> ---
>  exec.c | 3 ++-
>  1 file changed, 2 insertions(+), 1 deletion(-)
> 
> diff --git a/exec.c b/exec.c
> index 8af2570..d9c231d 100644
> --- a/exec.c
> +++ b/exec.c
> @@ -1194,8 +1194,9 @@ static long gethugepagesize(const char *path, Error **errp)
>          return 0;
>      }
>  
> -    if (fs.f_type != HUGETLBFS_MAGIC)
> +    if (fs.f_type != HUGETLBFS_MAGIC && !getenv("QTEST_QEMU_BINARY")) {
>          fprintf(stderr, "Warning: path not on HugeTLBFS: %s\n", path);
> +    }
>  
>      return fs.f_bsize;
>  }
> -- 
> 2.4.3

Re: [Qemu-devel] [PATCH] exec: silence hugetlbfs warning under qtest

[Michael S. Tsirkin]

On Tue, Oct 27, 2015 at 05:29:43PM +0100, marcandre.lureau@redhat.com wrote:
> From: Marc-André Lureau <marcandre.lureau@redhat.com>
> 
> vhost-user-test prints a warning. A test should not need to run on
> hugetlbfs, let's silence the warning under qtest. Unfortunately, the
> condition can't check on qtest_enabled() or qtest_driver() since they
> are initialized later.
> 
> Signed-off-by: Marc-André Lureau <marcandre.lureau@redhat.com>

I guess we need a command line option to suppress the warning.
Seems a lot of work for very little benefit, but that's
the best I could come up with.

> ---
>  exec.c | 3 ++-
>  1 file changed, 2 insertions(+), 1 deletion(-)
> 
> diff --git a/exec.c b/exec.c
> index 8af2570..d9c231d 100644
> --- a/exec.c
> +++ b/exec.c
> @@ -1194,8 +1194,9 @@ static long gethugepagesize(const char *path, Error **errp)
>          return 0;
>      }
>  
> -    if (fs.f_type != HUGETLBFS_MAGIC)
> +    if (fs.f_type != HUGETLBFS_MAGIC && !getenv("QTEST_QEMU_BINARY")) {
>          fprintf(stderr, "Warning: path not on HugeTLBFS: %s\n", path);
> +    }
>  
>      return fs.f_bsize;
>  }
> -- 
> 2.4.3

Re: [Qemu-devel] [PATCH] exec: silence hugetlbfs warning under qtest

[Michael S. Tsirkin]

On Tue, Oct 27, 2015 at 05:29:43PM +0100, marcandre.lureau@redhat.com wrote:
> From: Marc-André Lureau <marcandre.lureau@redhat.com>
> 
> vhost-user-test prints a warning. A test should not need to run on
> hugetlbfs, let's silence the warning under qtest. Unfortunately, the
> condition can't check on qtest_enabled() or qtest_driver() since they
> are initialized later.
> 
> Signed-off-by: Marc-André Lureau <marcandre.lureau@redhat.com>

Any idea what's the best way to address this?
Is poking at environment like this appropriate?
Maybe a command line flag to silence the warning?

> ---
>  exec.c | 3 ++-
>  1 file changed, 2 insertions(+), 1 deletion(-)
> 
> diff --git a/exec.c b/exec.c
> index 8af2570..d9c231d 100644
> --- a/exec.c
> +++ b/exec.c
> @@ -1194,8 +1194,9 @@ static long gethugepagesize(const char *path, Error **errp)
>          return 0;
>      }
>  
> -    if (fs.f_type != HUGETLBFS_MAGIC)
> +    if (fs.f_type != HUGETLBFS_MAGIC && !getenv("QTEST_QEMU_BINARY")) {
>          fprintf(stderr, "Warning: path not on HugeTLBFS: %s\n", path);
> +    }
>  
>      return fs.f_bsize;
>  }
> -- 
> 2.4.3

Re: [Qemu-devel] [PATCH] exec: silence hugetlbfs warning under qtest

[Markus Armbruster]

"Michael S. Tsirkin" <mst@redhat.com> writes:

> On Tue, Oct 27, 2015 at 05:29:43PM +0100, marcandre.lureau@redhat.com wrote:
>> From: Marc-André Lureau <marcandre.lureau@redhat.com>
>> 
>> vhost-user-test prints a warning. A test should not need to run on
>> hugetlbfs, let's silence the warning under qtest. Unfortunately, the
>> condition can't check on qtest_enabled() or qtest_driver() since they
>> are initialized later.
>> 
>> Signed-off-by: Marc-André Lureau <marcandre.lureau@redhat.com>
>
> Any idea what's the best way to address this?

Can we reorder things so that qtest_enabled() can be used?

> Is poking at environment like this appropriate?

I'm afraid it isn't.  QTEST_QEMU_BINARY can be in the environment even
when you're not running under qtest.

> Maybe a command line flag to silence the warning?

There's -qtest and -qtest-log.  I guess keying on those would be less
unclean than QTEST_QEMU_BINARY.

>> ---
>>  exec.c | 3 ++-
>>  1 file changed, 2 insertions(+), 1 deletion(-)
>> 
>> diff --git a/exec.c b/exec.c
>> index 8af2570..d9c231d 100644
>> --- a/exec.c
>> +++ b/exec.c
>> @@ -1194,8 +1194,9 @@ static long gethugepagesize(const char *path, Error **errp)
>>          return 0;
>>      }
>>  
>> -    if (fs.f_type != HUGETLBFS_MAGIC)
>> +    if (fs.f_type != HUGETLBFS_MAGIC && !getenv("QTEST_QEMU_BINARY")) {
>>          fprintf(stderr, "Warning: path not on HugeTLBFS: %s\n", path);
>> +    }
>>  
>>      return fs.f_bsize;
>>  }
>> -- 
>> 2.4.3

Re: [Qemu-devel] [PATCH] exec: silence hugetlbfs warning under qtest

[Michael S. Tsirkin]

On Tue, Oct 27, 2015 at 05:29:43PM +0100, marcandre.lureau@redhat.com wrote:
> From: Marc-André Lureau <marcandre.lureau@redhat.com>
> 
> vhost-user-test prints a warning. A test should not need to run on
> hugetlbfs, let's silence the warning under qtest. Unfortunately, the
> condition can't check on qtest_enabled() or qtest_driver() since they
> are initialized later.
> 
> Signed-off-by: Marc-André Lureau <marcandre.lureau@redhat.com>

OK so far this is the best we could come up with, so I'll queue this up
for after 2.5 unless someone objects.

> ---
>  exec.c | 3 ++-
>  1 file changed, 2 insertions(+), 1 deletion(-)
> 
> diff --git a/exec.c b/exec.c
> index 8af2570..d9c231d 100644
> --- a/exec.c
> +++ b/exec.c
> @@ -1194,8 +1194,9 @@ static long gethugepagesize(const char *path, Error **errp)
>          return 0;
>      }
>  
> -    if (fs.f_type != HUGETLBFS_MAGIC)
> +    if (fs.f_type != HUGETLBFS_MAGIC && !getenv("QTEST_QEMU_BINARY")) {
>          fprintf(stderr, "Warning: path not on HugeTLBFS: %s\n", path);
> +    }
>  
>      return fs.f_bsize;
>  }
> -- 
> 2.4.3

Re: [Qemu-devel] [PATCH] exec: silence hugetlbfs warning under qtest

[Michael S. Tsirkin]

On Tue, Nov 24, 2015 at 03:30:23PM +0100, Paolo Bonzini wrote:
> 
> 
> On 24/11/2015 15:25, Michael S. Tsirkin wrote:
> > I'm worried that things go slow and people don't have a way to find out
> > why, and documentation isn't the first place people look for when facing
> > a performance issue.
> > 
> > At the moment we call MADV_HUGEPAGE unconditionally and unfortunately
> > there's no way to report it's not having the intended effect.
> > 
> > Maybe we want a "don't enable hugepages" flag.
> > That could also have the effect of suppressing the warnings.
> 
> THP is anyway slower than hugetlbfs.  If people really want performance,
> they need to use hugetlbfs.  For everything else I'd be surprised if
> they even notice.
> 
> Paolo

If you want to post and merge a patch removing this warning completely,
and revert the original one, that's fine with me.

-- 
MST

Re: [Qemu-devel] [PATCH] exec: silence hugetlbfs warning under qtest

[Markus Armbruster]

"Daniel P. Berrange" <berrange@redhat.com> writes:

> On Tue, Nov 24, 2015 at 04:10:54PM +0200, Michael S. Tsirkin wrote:
>> On Mon, Nov 23, 2015 at 06:10:08PM +0000, Daniel P. Berrange wrote:
>> > On Mon, Nov 23, 2015 at 07:01:33PM +0100, Marc-André Lureau wrote:
>> > > On Mon, Nov 23, 2015 at 6:40 PM, Paolo Bonzini
>> > > <pbonzini@redhat.com> wrote:
>> > > > Before: object-initial, chardev, qtest, object-late (not in the patch)
>> > > >
>> > > > After: chardev, qtest, object-initial, object-late (not in the patch)
>> > > >
>> > > > Objects must be initialized before chardev (except rng-egd) since in the
>> > > > future chardev will need to use objects, in particular secret objects.
>> > > > Was the swap intentional?
>> > > 
>> > > Yes, without the swap, qtest was not initialized before memory
>> > > is allocated.
>> > > 
>> > > The alternative I could think of is to check the QTEST_QEMU_BINARY
>> > > variable:
>> > > http://lists.nongnu.org/archive/html/qemu-devel/2015-11/msg01527.html
>> > 
>> > Why do we not simply delete the warning message about the path not
>> > being on hugetlbfs ? ie, why does QEMU try to force a policy that
>> > a memory-file backend has to be on hugetlbfs, as opposed to on
>> > a plain tmpfs ?  I've previously had user request that we allow
>> > use of plain tmpfs, because they want to use vhost-user without
>> > also using hugepages, and that could be done with plain tmpfs.
>> 
>> Because THP does not work on any other filesystem,
>> so many workloads are much slower.
>> That's why it's a warning, not an error.
>
> AFAICT this warning message is not in a codepath that is specific to
> use of THP. This is just generic code for allocating guest memory
> backed by a file, which does not have any assumption / prerequisite
> that THP is wanted or enabled. So adding warnings that are specifically
> related to THP is inappropriate.

If a certain workload needs hugepages to perform, it should ask for them
explicitly.  Any means we provide for that purpose should error out when
it can provide them.  Guessing what the user meant and warning him when
we're giving him something else is generally not a good idea.

Whether -object memory-backend-file is a means to explicitly ask for
hugepages is less than clear.  Its file comment says "for hugetlbfs",
but it works just fine with files on other file systems.

Aside: we have *three* copies of gethugepagesize() in the tree: one that
warns in exec.c, one in target-ppc/kvm.c and a third one in
ivshmem-server.c.

> The fact that THP only works with a hugetlbfs path is merely a
> documentation item to record against the command line option for
> -mem-path.

Re: [Qemu-devel] [PATCH] exec: silence hugetlbfs warning under qtest

[Paolo Bonzini]

On 24/11/2015 15:25, Michael S. Tsirkin wrote:
> I'm worried that things go slow and people don't have a way to find out
> why, and documentation isn't the first place people look for when facing
> a performance issue.
> 
> At the moment we call MADV_HUGEPAGE unconditionally and unfortunately
> there's no way to report it's not having the intended effect.
> 
> Maybe we want a "don't enable hugepages" flag.
> That could also have the effect of suppressing the warnings.

THP is anyway slower than hugetlbfs.  If people really want performance,
they need to use hugetlbfs.  For everything else I'd be surprised if
they even notice.

Paolo

Re: [Qemu-devel] [PATCH] exec: silence hugetlbfs warning under qtest

[Daniel P. Berrange]

On Tue, Nov 24, 2015 at 04:25:40PM +0200, Michael S. Tsirkin wrote:
> On Tue, Nov 24, 2015 at 02:19:36PM +0000, Daniel P. Berrange wrote:
> > On Tue, Nov 24, 2015 at 04:10:54PM +0200, Michael S. Tsirkin wrote:
> > > On Mon, Nov 23, 2015 at 06:10:08PM +0000, Daniel P. Berrange wrote:
> > > > On Mon, Nov 23, 2015 at 07:01:33PM +0100, Marc-André Lureau wrote:
> > > > > On Mon, Nov 23, 2015 at 6:40 PM, Paolo Bonzini <pbonzini@redhat.com> wrote:
> > > > > > Before: object-initial, chardev, qtest, object-late (not in the patch)
> > > > > >
> > > > > > After: chardev, qtest, object-initial, object-late (not in the patch)
> > > > > >
> > > > > > Objects must be initialized before chardev (except rng-egd) since in the
> > > > > > future chardev will need to use objects, in particular secret objects.
> > > > > > Was the swap intentional?
> > > > > 
> > > > > Yes, without the swap, qtest was not initialized before memory is allocated.
> > > > > 
> > > > > The alternative I could think of is to check the QTEST_QEMU_BINARY
> > > > > variable: http://lists.nongnu.org/archive/html/qemu-devel/2015-11/msg01527.html
> > > > 
> > > > Why do we not simply delete the warning message about the path not
> > > > being on hugetlbfs ? ie, why does QEMU try to force a policy that
> > > > a memory-file backend has to be on hugetlbfs, as opposed to on
> > > > a plain tmpfs ?  I've previously had user request that we allow
> > > > use of plain tmpfs, because they want to use vhost-user without
> > > > also using hugepages, and that could be done with plain tmpfs.
> > > 
> > > Because THP does not work on any other filesystem,
> > > so many workloads are much slower.
> > > That's why it's a warning, not an error.
> > 
> > AFAICT this warning message is not in a codepath that is specific to
> > use of THP. This is just generic code for allocating guest memory
> > backed by a file, which does not have any assumption / prerequisite
> > that THP is wanted or enabled. So adding warnings that are specifically
> > related to THP is inappropriate.
> > 
> > The fact that THP only works with a hugetlbfs path is merely a
> > documentation item to record against the command line option for
> > -mem-path.
> 
> I'm worried that things go slow and people don't have a way to find out
> why, and documentation isn't the first place people look for when facing
> a performance issue.

If you are launching QEMU directly that I think it reasonable to expect
people know the difference between giving a guest a hugetlbfs path
vs a regular tmpfs path, and can specifically choose with they prefer.

If launching via libvirt, the error scenario you're concerned about is
impossible, as libvirt will be choosing the hugetlbfs path to match
the requested huge page size from teh guest XML.

Regards,
Daniel
-- 
|: http://berrange.com      -o-    http://www.flickr.com/photos/dberrange/ :|
|: http://libvirt.org              -o-             http://virt-manager.org :|
|: http://autobuild.org       -o-         http://search.cpan.org/~danberr/ :|
|: http://entangle-photo.org       -o-       http://live.gnome.org/gtk-vnc :|

Re: [Qemu-devel] [PATCH] exec: silence hugetlbfs warning under qtest

[Michael S. Tsirkin]

On Tue, Nov 24, 2015 at 02:19:36PM +0000, Daniel P. Berrange wrote:
> On Tue, Nov 24, 2015 at 04:10:54PM +0200, Michael S. Tsirkin wrote:
> > On Mon, Nov 23, 2015 at 06:10:08PM +0000, Daniel P. Berrange wrote:
> > > On Mon, Nov 23, 2015 at 07:01:33PM +0100, Marc-André Lureau wrote:
> > > > On Mon, Nov 23, 2015 at 6:40 PM, Paolo Bonzini <pbonzini@redhat.com> wrote:
> > > > > Before: object-initial, chardev, qtest, object-late (not in the patch)
> > > > >
> > > > > After: chardev, qtest, object-initial, object-late (not in the patch)
> > > > >
> > > > > Objects must be initialized before chardev (except rng-egd) since in the
> > > > > future chardev will need to use objects, in particular secret objects.
> > > > > Was the swap intentional?
> > > > 
> > > > Yes, without the swap, qtest was not initialized before memory is allocated.
> > > > 
> > > > The alternative I could think of is to check the QTEST_QEMU_BINARY
> > > > variable: http://lists.nongnu.org/archive/html/qemu-devel/2015-11/msg01527.html
> > > 
> > > Why do we not simply delete the warning message about the path not
> > > being on hugetlbfs ? ie, why does QEMU try to force a policy that
> > > a memory-file backend has to be on hugetlbfs, as opposed to on
> > > a plain tmpfs ?  I've previously had user request that we allow
> > > use of plain tmpfs, because they want to use vhost-user without
> > > also using hugepages, and that could be done with plain tmpfs.
> > 
> > Because THP does not work on any other filesystem,
> > so many workloads are much slower.
> > That's why it's a warning, not an error.
> 
> AFAICT this warning message is not in a codepath that is specific to
> use of THP. This is just generic code for allocating guest memory
> backed by a file, which does not have any assumption / prerequisite
> that THP is wanted or enabled. So adding warnings that are specifically
> related to THP is inappropriate.
> 
> The fact that THP only works with a hugetlbfs path is merely a
> documentation item to record against the command line option for
> -mem-path.

I'm worried that things go slow and people don't have a way to find out
why, and documentation isn't the first place people look for when facing
a performance issue.

At the moment we call MADV_HUGEPAGE unconditionally and unfortunately
there's no way to report it's not having the intended effect.

Maybe we want a "don't enable hugepages" flag.
That could also have the effect of suppressing the warnings.

> 
> Regards,
> Daniel
> -- 
> |: http://berrange.com      -o-    http://www.flickr.com/photos/dberrange/ :|
> |: http://libvirt.org              -o-             http://virt-manager.org :|
> |: http://autobuild.org       -o-         http://search.cpan.org/~danberr/ :|
> |: http://entangle-photo.org       -o-       http://live.gnome.org/gtk-vnc :|

Re: [Qemu-devel] [PATCH] exec: silence hugetlbfs warning under qtest

[Daniel P. Berrange]

On Tue, Nov 24, 2015 at 04:10:54PM +0200, Michael S. Tsirkin wrote:
> On Mon, Nov 23, 2015 at 06:10:08PM +0000, Daniel P. Berrange wrote:
> > On Mon, Nov 23, 2015 at 07:01:33PM +0100, Marc-André Lureau wrote:
> > > On Mon, Nov 23, 2015 at 6:40 PM, Paolo Bonzini <pbonzini@redhat.com> wrote:
> > > > Before: object-initial, chardev, qtest, object-late (not in the patch)
> > > >
> > > > After: chardev, qtest, object-initial, object-late (not in the patch)
> > > >
> > > > Objects must be initialized before chardev (except rng-egd) since in the
> > > > future chardev will need to use objects, in particular secret objects.
> > > > Was the swap intentional?
> > > 
> > > Yes, without the swap, qtest was not initialized before memory is allocated.
> > > 
> > > The alternative I could think of is to check the QTEST_QEMU_BINARY
> > > variable: http://lists.nongnu.org/archive/html/qemu-devel/2015-11/msg01527.html
> > 
> > Why do we not simply delete the warning message about the path not
> > being on hugetlbfs ? ie, why does QEMU try to force a policy that
> > a memory-file backend has to be on hugetlbfs, as opposed to on
> > a plain tmpfs ?  I've previously had user request that we allow
> > use of plain tmpfs, because they want to use vhost-user without
> > also using hugepages, and that could be done with plain tmpfs.
> 
> Because THP does not work on any other filesystem,
> so many workloads are much slower.
> That's why it's a warning, not an error.

AFAICT this warning message is not in a codepath that is specific to
use of THP. This is just generic code for allocating guest memory
backed by a file, which does not have any assumption / prerequisite
that THP is wanted or enabled. So adding warnings that are specifically
related to THP is inappropriate.

The fact that THP only works with a hugetlbfs path is merely a
documentation item to record against the command line option for
-mem-path.


Regards,
Daniel
-- 
|: http://berrange.com      -o-    http://www.flickr.com/photos/dberrange/ :|
|: http://libvirt.org              -o-             http://virt-manager.org :|
|: http://autobuild.org       -o-         http://search.cpan.org/~danberr/ :|
|: http://entangle-photo.org       -o-       http://live.gnome.org/gtk-vnc :|

Re: [Qemu-devel] [PATCH] exec: silence hugetlbfs warning under qtest

[Michael S. Tsirkin]

On Mon, Nov 23, 2015 at 07:01:33PM +0100, Marc-André Lureau wrote:
> On Mon, Nov 23, 2015 at 6:40 PM, Paolo Bonzini <pbonzini@redhat.com> wrote:
> > Before: object-initial, chardev, qtest, object-late (not in the patch)
> >
> > After: chardev, qtest, object-initial, object-late (not in the patch)
> >
> > Objects must be initialized before chardev (except rng-egd) since in the
> > future chardev will need to use objects, in particular secret objects.
> > Was the swap intentional?
> 
> Yes, without the swap, qtest was not initialized before memory is allocated.
> 
> The alternative I could think of is to check the QTEST_QEMU_BINARY
> variable: http://lists.nongnu.org/archive/html/qemu-devel/2015-11/msg01527.html

Better add a custom flag just for running tests.
environment setting are too fragile.

> 
> -- 
> Marc-André Lureau

Re: [Qemu-devel] [PATCH] exec: silence hugetlbfs warning under qtest

[Michael S. Tsirkin]

On Mon, Nov 23, 2015 at 06:10:08PM +0000, Daniel P. Berrange wrote:
> On Mon, Nov 23, 2015 at 07:01:33PM +0100, Marc-André Lureau wrote:
> > On Mon, Nov 23, 2015 at 6:40 PM, Paolo Bonzini <pbonzini@redhat.com> wrote:
> > > Before: object-initial, chardev, qtest, object-late (not in the patch)
> > >
> > > After: chardev, qtest, object-initial, object-late (not in the patch)
> > >
> > > Objects must be initialized before chardev (except rng-egd) since in the
> > > future chardev will need to use objects, in particular secret objects.
> > > Was the swap intentional?
> > 
> > Yes, without the swap, qtest was not initialized before memory is allocated.
> > 
> > The alternative I could think of is to check the QTEST_QEMU_BINARY
> > variable: http://lists.nongnu.org/archive/html/qemu-devel/2015-11/msg01527.html
> 
> Why do we not simply delete the warning message about the path not
> being on hugetlbfs ? ie, why does QEMU try to force a policy that
> a memory-file backend has to be on hugetlbfs, as opposed to on
> a plain tmpfs ?  I've previously had user request that we allow
> use of plain tmpfs, because they want to use vhost-user without
> also using hugepages, and that could be done with plain tmpfs.
> 
> Regards,
> Daniel

Because THP does not work on any other filesystem,
so many workloads are much slower.
That's why it's a warning, not an error.

> -- 
> |: http://berrange.com      -o-    http://www.flickr.com/photos/dberrange/ :|
> |: http://libvirt.org              -o-             http://virt-manager.org :|
> |: http://autobuild.org       -o-         http://search.cpan.org/~danberr/ :|
> |: http://entangle-photo.org       -o-       http://live.gnome.org/gtk-vnc :|

Re: [Qemu-devel] [PATCH] exec: silence hugetlbfs warning under qtest

[Daniel P. Berrange]

On Mon, Nov 23, 2015 at 07:01:33PM +0100, Marc-André Lureau wrote:
> On Mon, Nov 23, 2015 at 6:40 PM, Paolo Bonzini <pbonzini@redhat.com> wrote:
> > Before: object-initial, chardev, qtest, object-late (not in the patch)
> >
> > After: chardev, qtest, object-initial, object-late (not in the patch)
> >
> > Objects must be initialized before chardev (except rng-egd) since in the
> > future chardev will need to use objects, in particular secret objects.
> > Was the swap intentional?
> 
> Yes, without the swap, qtest was not initialized before memory is allocated.
> 
> The alternative I could think of is to check the QTEST_QEMU_BINARY
> variable: http://lists.nongnu.org/archive/html/qemu-devel/2015-11/msg01527.html

Why do we not simply delete the warning message about the path not
being on hugetlbfs ? ie, why does QEMU try to force a policy that
a memory-file backend has to be on hugetlbfs, as opposed to on
a plain tmpfs ?  I've previously had user request that we allow
use of plain tmpfs, because they want to use vhost-user without
also using hugepages, and that could be done with plain tmpfs.

Regards,
Daniel
-- 
|: http://berrange.com      -o-    http://www.flickr.com/photos/dberrange/ :|
|: http://libvirt.org              -o-             http://virt-manager.org :|
|: http://autobuild.org       -o-         http://search.cpan.org/~danberr/ :|
|: http://entangle-photo.org       -o-       http://live.gnome.org/gtk-vnc :|

Re: [Qemu-devel] [PATCH] exec: silence hugetlbfs warning under qtest

[Marc-André Lureau]

On Mon, Nov 23, 2015 at 6:40 PM, Paolo Bonzini <pbonzini@redhat.com> wrote:
> Before: object-initial, chardev, qtest, object-late (not in the patch)
>
> After: chardev, qtest, object-initial, object-late (not in the patch)
>
> Objects must be initialized before chardev (except rng-egd) since in the
> future chardev will need to use objects, in particular secret objects.
> Was the swap intentional?

Yes, without the swap, qtest was not initialized before memory is allocated.

The alternative I could think of is to check the QTEST_QEMU_BINARY
variable: http://lists.nongnu.org/archive/html/qemu-devel/2015-11/msg01527.html


-- 
Marc-André Lureau

Re: [Qemu-devel] [PATCH] exec: silence hugetlbfs warning under qtest

[Daniel P. Berrange]

On Mon, Nov 23, 2015 at 05:46:06PM +0000, Daniel P. Berrange wrote:
> On Mon, Nov 23, 2015 at 06:40:58PM +0100, Paolo Bonzini wrote:
> > 
> > 
> > On 18/11/2015 21:49, Michael S. Tsirkin wrote:
> > > On Wed, Nov 18, 2015 at 10:02:58AM +0100, marcandre.lureau@redhat.com wrote:
> > >> From: Marc-André Lureau <marcandre.lureau@redhat.com>
> > >>
> > >> vhost-user-test prints a warning. A test should not need to run on
> > >> hugetlbfs, let's silence the warning under qtest. The
> > >> condition can't check on qtest_enabled() since vhost-user-test actually
> > >> doesn't use qtest accel. However, qtest_driver() can be used, if
> > >> qtest_init() is called early enough. For that reason, move chardev and
> > >> qtest initialization early.
> > >>
> > >> Signed-off-by: Marc-André Lureau <marcandre.lureau@redhat.com>
> > > 
> > > Seems ok, and let us bring back the 2.4 test that was
> > > dropped in 2.5, but I'd like an ack on this from someone else.
> > 
> > I wonder if we need the warning at all, but this patch is okay.
> > 
> > Anyhow:
> > 
> > 
> > 
> > -    if (qemu_opts_foreach(qemu_find_opts("object"),
> > -                          object_create,
> > -                          object_create_initial, NULL)) {
> > +    if (qemu_opts_foreach(qemu_find_opts("chardev"),
> > +                          chardev_init_func, NULL, NULL)) {
> >          exit(1);
> >      }
> > 
> > -    if (qemu_opts_foreach(qemu_find_opts("chardev"),
> > -                          chardev_init_func, NULL, NULL)) {
> > +    if (qtest_chrdev) {
> > +        Error *local_err = NULL;
> > +        qtest_init(qtest_chrdev, qtest_log, &local_err);
> > +        if (local_err) {
> > +            error_report_err(local_err);
> > +            exit(1);
> > +        }
> > +    }
> > +
> > +    if (qemu_opts_foreach(qemu_find_opts("object"),
> > +                          object_create,
> > +                          object_create_initial, NULL)) {
> >          exit(1);
> >      }
> > 
> > 
> > Before: object-initial, chardev, qtest, object-late (not in the patch)
> > 
> > After: chardev, qtest, object-initial, object-late (not in the patch)
> > 
> > Objects must be initialized before chardev (except rng-egd) since in the
> > future chardev will need to use objects, in particular secret objects.

BTW, it was TLS credentials object that prompted my initial re-ordering,
but you're right that secret objects will need creation before chardevs
too, since TLS credentials will need secrets.

Regards,
Daniel
-- 
|: http://berrange.com      -o-    http://www.flickr.com/photos/dberrange/ :|
|: http://libvirt.org              -o-             http://virt-manager.org :|
|: http://autobuild.org       -o-         http://search.cpan.org/~danberr/ :|
|: http://entangle-photo.org       -o-       http://live.gnome.org/gtk-vnc :|

Re: [Qemu-devel] [PATCH] exec: silence hugetlbfs warning under qtest

[Daniel P. Berrange]

On Mon, Nov 23, 2015 at 06:40:58PM +0100, Paolo Bonzini wrote:
> 
> 
> On 18/11/2015 21:49, Michael S. Tsirkin wrote:
> > On Wed, Nov 18, 2015 at 10:02:58AM +0100, marcandre.lureau@redhat.com wrote:
> >> From: Marc-André Lureau <marcandre.lureau@redhat.com>
> >>
> >> vhost-user-test prints a warning. A test should not need to run on
> >> hugetlbfs, let's silence the warning under qtest. The
> >> condition can't check on qtest_enabled() since vhost-user-test actually
> >> doesn't use qtest accel. However, qtest_driver() can be used, if
> >> qtest_init() is called early enough. For that reason, move chardev and
> >> qtest initialization early.
> >>
> >> Signed-off-by: Marc-André Lureau <marcandre.lureau@redhat.com>
> > 
> > Seems ok, and let us bring back the 2.4 test that was
> > dropped in 2.5, but I'd like an ack on this from someone else.
> 
> I wonder if we need the warning at all, but this patch is okay.
> 
> Anyhow:
> 
> 
> 
> -    if (qemu_opts_foreach(qemu_find_opts("object"),
> -                          object_create,
> -                          object_create_initial, NULL)) {
> +    if (qemu_opts_foreach(qemu_find_opts("chardev"),
> +                          chardev_init_func, NULL, NULL)) {
>          exit(1);
>      }
> 
> -    if (qemu_opts_foreach(qemu_find_opts("chardev"),
> -                          chardev_init_func, NULL, NULL)) {
> +    if (qtest_chrdev) {
> +        Error *local_err = NULL;
> +        qtest_init(qtest_chrdev, qtest_log, &local_err);
> +        if (local_err) {
> +            error_report_err(local_err);
> +            exit(1);
> +        }
> +    }
> +
> +    if (qemu_opts_foreach(qemu_find_opts("object"),
> +                          object_create,
> +                          object_create_initial, NULL)) {
>          exit(1);
>      }
> 
> 
> Before: object-initial, chardev, qtest, object-late (not in the patch)
> 
> After: chardev, qtest, object-initial, object-late (not in the patch)
> 
> Objects must be initialized before chardev (except rng-egd) since in the
> future chardev will need to use objects, in particular secret objects.
> Was the swap intentional?

Yeah, that is essentially reverting my patch to split up object
creation:

  commit f08f9271bfe3f19a5eb3d7a2f48532065304d5c8
  Author: Daniel P. Berrange <berrange@redhat.com>
  Date:   Wed May 13 17:14:04 2015 +0100

    vl: Create (most) objects before creating chardev backends

So I'd rather we did not do that. IMHO having the goal that all
objects be created first is the right long term direction, if
we view the future as being 100% object based. Hence most of the
object types are created on object-initial, and only a few things
in object-late.

Regards,
Daniel
-- 
|: http://berrange.com      -o-    http://www.flickr.com/photos/dberrange/ :|
|: http://libvirt.org              -o-             http://virt-manager.org :|
|: http://autobuild.org       -o-         http://search.cpan.org/~danberr/ :|
|: http://entangle-photo.org       -o-       http://live.gnome.org/gtk-vnc :|

Re: [Qemu-devel] [PATCH] exec: silence hugetlbfs warning under qtest

[Paolo Bonzini]

On 18/11/2015 21:49, Michael S. Tsirkin wrote:
> On Wed, Nov 18, 2015 at 10:02:58AM +0100, marcandre.lureau@redhat.com wrote:
>> From: Marc-André Lureau <marcandre.lureau@redhat.com>
>>
>> vhost-user-test prints a warning. A test should not need to run on
>> hugetlbfs, let's silence the warning under qtest. The
>> condition can't check on qtest_enabled() since vhost-user-test actually
>> doesn't use qtest accel. However, qtest_driver() can be used, if
>> qtest_init() is called early enough. For that reason, move chardev and
>> qtest initialization early.
>>
>> Signed-off-by: Marc-André Lureau <marcandre.lureau@redhat.com>
> 
> Seems ok, and let us bring back the 2.4 test that was
> dropped in 2.5, but I'd like an ack on this from someone else.

I wonder if we need the warning at all, but this patch is okay.

Anyhow:



-    if (qemu_opts_foreach(qemu_find_opts("object"),
-                          object_create,
-                          object_create_initial, NULL)) {
+    if (qemu_opts_foreach(qemu_find_opts("chardev"),
+                          chardev_init_func, NULL, NULL)) {
         exit(1);
     }

-    if (qemu_opts_foreach(qemu_find_opts("chardev"),
-                          chardev_init_func, NULL, NULL)) {
+    if (qtest_chrdev) {
+        Error *local_err = NULL;
+        qtest_init(qtest_chrdev, qtest_log, &local_err);
+        if (local_err) {
+            error_report_err(local_err);
+            exit(1);
+        }
+    }
+
+    if (qemu_opts_foreach(qemu_find_opts("object"),
+                          object_create,
+                          object_create_initial, NULL)) {
         exit(1);
     }


Before: object-initial, chardev, qtest, object-late (not in the patch)

After: chardev, qtest, object-initial, object-late (not in the patch)

Objects must be initialized before chardev (except rng-egd) since in the
future chardev will need to use objects, in particular secret objects.
Was the swap intentional?

Paolo

@@ -4325,15 +4334,6 @@ int main(int argc, char **argv, char **envp)

     configure_accelerator(current_machine);

-    if (qtest_chrdev) {
-        Error *local_err = NULL;
-        qtest_init(qtest_chrdev, qtest_log, &local_err);
-        if (local_err) {
-            error_report_err(local_err);
-            exit(1);
-        }
-    }

> Anyone?
> 
>> ---
>>  exec.c |  5 ++++-
>>  vl.c   | 28 ++++++++++++++--------------
>>  2 files changed, 18 insertions(+), 15 deletions(-)
>>
>> diff --git a/exec.c b/exec.c
>> index b09f18b..acbd4a2 100644
>> --- a/exec.c
>> +++ b/exec.c
>> @@ -51,6 +51,7 @@
>>  #include "qemu/main-loop.h"
>>  #include "translate-all.h"
>>  #include "sysemu/replay.h"
>> +#include "sysemu/qtest.h"
>>  
>>  #include "exec/memory-internal.h"
>>  #include "exec/ram_addr.h"
>> @@ -1196,8 +1197,10 @@ static long gethugepagesize(const char *path, Error **errp)
>>          return 0;
>>      }
>>  
>> -    if (fs.f_type != HUGETLBFS_MAGIC)
>> +    if (!qtest_driver() &&
>> +        fs.f_type != HUGETLBFS_MAGIC) {
>>          fprintf(stderr, "Warning: path not on HugeTLBFS: %s\n", path);
>> +    }
>>  
>>      return fs.f_bsize;
>>  }
>> diff --git a/vl.c b/vl.c
>> index 7d993a5..f9c661a 100644
>> --- a/vl.c
>> +++ b/vl.c
>> @@ -4288,14 +4288,23 @@ int main(int argc, char **argv, char **envp)
>>      page_size_init();
>>      socket_init();
>>  
>> -    if (qemu_opts_foreach(qemu_find_opts("object"),
>> -                          object_create,
>> -                          object_create_initial, NULL)) {
>> +    if (qemu_opts_foreach(qemu_find_opts("chardev"),
>> +                          chardev_init_func, NULL, NULL)) {
>>          exit(1);
>>      }
>>  
>> -    if (qemu_opts_foreach(qemu_find_opts("chardev"),
>> -                          chardev_init_func, NULL, NULL)) {
>> +    if (qtest_chrdev) {
>> +        Error *local_err = NULL;
>> +        qtest_init(qtest_chrdev, qtest_log, &local_err);
>> +        if (local_err) {
>> +            error_report_err(local_err);
>> +            exit(1);
>> +        }
>> +    }
>> +
>> +    if (qemu_opts_foreach(qemu_find_opts("object"),
>> +                          object_create,
>> +                          object_create_initial, NULL)) {
>>          exit(1);
>>      }
>>  
>> @@ -4325,15 +4334,6 @@ int main(int argc, char **argv, char **envp)
>>  
>>      configure_accelerator(current_machine);
>>  
>> -    if (qtest_chrdev) {
>> -        Error *local_err = NULL;
>> -        qtest_init(qtest_chrdev, qtest_log, &local_err);
>> -        if (local_err) {
>> -            error_report_err(local_err);
>> -            exit(1);
>> -        }
>> -    }
>> -
>>      machine_opts = qemu_get_machine_opts();
>>      kernel_filename = qemu_opt_get(machine_opts, "kernel");
>>      initrd_filename = qemu_opt_get(machine_opts, "initrd");
>> -- 
>> 2.5.0

Re: [Qemu-devel] [PATCH] exec: silence hugetlbfs warning under qtest

[Michael S. Tsirkin]

On Wed, Nov 18, 2015 at 10:02:58AM +0100, marcandre.lureau@redhat.com wrote:
> From: Marc-André Lureau <marcandre.lureau@redhat.com>
> 
> vhost-user-test prints a warning. A test should not need to run on
> hugetlbfs, let's silence the warning under qtest. The
> condition can't check on qtest_enabled() since vhost-user-test actually
> doesn't use qtest accel. However, qtest_driver() can be used, if
> qtest_init() is called early enough. For that reason, move chardev and
> qtest initialization early.
> 
> Signed-off-by: Marc-André Lureau <marcandre.lureau@redhat.com>

Seems ok, and let us bring back the 2.4 test that was
dropped in 2.5, but I'd like an ack on this from someone else.

Anyone?

> ---
>  exec.c |  5 ++++-
>  vl.c   | 28 ++++++++++++++--------------
>  2 files changed, 18 insertions(+), 15 deletions(-)
> 
> diff --git a/exec.c b/exec.c
> index b09f18b..acbd4a2 100644
> --- a/exec.c
> +++ b/exec.c
> @@ -51,6 +51,7 @@
>  #include "qemu/main-loop.h"
>  #include "translate-all.h"
>  #include "sysemu/replay.h"
> +#include "sysemu/qtest.h"
>  
>  #include "exec/memory-internal.h"
>  #include "exec/ram_addr.h"
> @@ -1196,8 +1197,10 @@ static long gethugepagesize(const char *path, Error **errp)
>          return 0;
>      }
>  
> -    if (fs.f_type != HUGETLBFS_MAGIC)
> +    if (!qtest_driver() &&
> +        fs.f_type != HUGETLBFS_MAGIC) {
>          fprintf(stderr, "Warning: path not on HugeTLBFS: %s\n", path);
> +    }
>  
>      return fs.f_bsize;
>  }
> diff --git a/vl.c b/vl.c
> index 7d993a5..f9c661a 100644
> --- a/vl.c
> +++ b/vl.c
> @@ -4288,14 +4288,23 @@ int main(int argc, char **argv, char **envp)
>      page_size_init();
>      socket_init();
>  
> -    if (qemu_opts_foreach(qemu_find_opts("object"),
> -                          object_create,
> -                          object_create_initial, NULL)) {
> +    if (qemu_opts_foreach(qemu_find_opts("chardev"),
> +                          chardev_init_func, NULL, NULL)) {
>          exit(1);
>      }
>  
> -    if (qemu_opts_foreach(qemu_find_opts("chardev"),
> -                          chardev_init_func, NULL, NULL)) {
> +    if (qtest_chrdev) {
> +        Error *local_err = NULL;
> +        qtest_init(qtest_chrdev, qtest_log, &local_err);
> +        if (local_err) {
> +            error_report_err(local_err);
> +            exit(1);
> +        }
> +    }
> +
> +    if (qemu_opts_foreach(qemu_find_opts("object"),
> +                          object_create,
> +                          object_create_initial, NULL)) {
>          exit(1);
>      }
>  
> @@ -4325,15 +4334,6 @@ int main(int argc, char **argv, char **envp)
>  
>      configure_accelerator(current_machine);
>  
> -    if (qtest_chrdev) {
> -        Error *local_err = NULL;
> -        qtest_init(qtest_chrdev, qtest_log, &local_err);
> -        if (local_err) {
> -            error_report_err(local_err);
> -            exit(1);
> -        }
> -    }
> -
>      machine_opts = qemu_get_machine_opts();
>      kernel_filename = qemu_opt_get(machine_opts, "kernel");
>      initrd_filename = qemu_opt_get(machine_opts, "initrd");
> -- 
> 2.5.0

[Qemu-devel] [PATCH] exec: silence hugetlbfs warning under qtest

[marcandre.lureau]

From: Marc-André Lureau <marcandre.lureau@redhat.com>

vhost-user-test prints a warning. A test should not need to run on
hugetlbfs, let's silence the warning under qtest. The
condition can't check on qtest_enabled() since vhost-user-test actually
doesn't use qtest accel. However, qtest_driver() can be used, if
qtest_init() is called early enough. For that reason, move chardev and
qtest initialization early.

Signed-off-by: Marc-André Lureau <marcandre.lureau@redhat.com>
---
 exec.c |  5 ++++-
 vl.c   | 28 ++++++++++++++--------------
 2 files changed, 18 insertions(+), 15 deletions(-)

diff --git a/exec.c b/exec.c
index b09f18b..acbd4a2 100644
--- a/exec.c
+++ b/exec.c
@@ -51,6 +51,7 @@
 #include "qemu/main-loop.h"
 #include "translate-all.h"
 #include "sysemu/replay.h"
+#include "sysemu/qtest.h"
 
 #include "exec/memory-internal.h"
 #include "exec/ram_addr.h"
@@ -1196,8 +1197,10 @@ static long gethugepagesize(const char *path, Error **errp)
         return 0;
     }
 
-    if (fs.f_type != HUGETLBFS_MAGIC)
+    if (!qtest_driver() &&
+        fs.f_type != HUGETLBFS_MAGIC) {
         fprintf(stderr, "Warning: path not on HugeTLBFS: %s\n", path);
+    }
 
     return fs.f_bsize;
 }
diff --git a/vl.c b/vl.c
index 7d993a5..f9c661a 100644
--- a/vl.c
+++ b/vl.c
@@ -4288,14 +4288,23 @@ int main(int argc, char **argv, char **envp)
     page_size_init();
     socket_init();
 
-    if (qemu_opts_foreach(qemu_find_opts("object"),
-                          object_create,
-                          object_create_initial, NULL)) {
+    if (qemu_opts_foreach(qemu_find_opts("chardev"),
+                          chardev_init_func, NULL, NULL)) {
         exit(1);
     }
 
-    if (qemu_opts_foreach(qemu_find_opts("chardev"),
-                          chardev_init_func, NULL, NULL)) {
+    if (qtest_chrdev) {
+        Error *local_err = NULL;
+        qtest_init(qtest_chrdev, qtest_log, &local_err);
+        if (local_err) {
+            error_report_err(local_err);
+            exit(1);
+        }
+    }
+
+    if (qemu_opts_foreach(qemu_find_opts("object"),
+                          object_create,
+                          object_create_initial, NULL)) {
         exit(1);
     }
 
@@ -4325,15 +4334,6 @@ int main(int argc, char **argv, char **envp)
 
     configure_accelerator(current_machine);
 
-    if (qtest_chrdev) {
-        Error *local_err = NULL;
-        qtest_init(qtest_chrdev, qtest_log, &local_err);
-        if (local_err) {
-            error_report_err(local_err);
-            exit(1);
-        }
-    }
-
     machine_opts = qemu_get_machine_opts();
     kernel_filename = qemu_opt_get(machine_opts, "kernel");
     initrd_filename = qemu_opt_get(machine_opts, "initrd");
-- 
2.5.0