[PATCH 1/1] vfio: put off the allocation of "minor" in vfio_create_group

[PATCH 1/1] vfio: put off the allocation of "minor" in vfio_create_group

[Zhen Lei]

The next code fragment "list_for_each_entry" is not depend on "minor". With this
patch, the free of "minor" in "list_for_each_entry" can be reduced, and there is
no functional change.

Signed-off-by: Zhen Lei <thunder.leizhen@huawei.com>
---
 drivers/vfio/vfio.c | 13 ++++++-------
 1 file changed, 6 insertions(+), 7 deletions(-)

diff --git a/drivers/vfio/vfio.c b/drivers/vfio/vfio.c
index f018d8d..737eb468 100644
--- a/drivers/vfio/vfio.c
+++ b/drivers/vfio/vfio.c
@@ -225,22 +225,21 @@ static struct vfio_group *vfio_create_group(struct iommu_group *iommu_group)

 	mutex_lock(&vfio.group_lock);

-	minor = vfio_alloc_group_minor(group);
-	if (minor < 0) {
-		vfio_group_unlock_and_free(group);
-		return ERR_PTR(minor);
-	}
-
 	/* Did we race creating this group? */
 	list_for_each_entry(tmp, &vfio.group_list, vfio_next) {
 		if (tmp->iommu_group == iommu_group) {
 			vfio_group_get(tmp);
-			vfio_free_group_minor(minor);
 			vfio_group_unlock_and_free(group);
 			return tmp;
 		}
 	}

+	minor = vfio_alloc_group_minor(group);
+	if (minor < 0) {
+		vfio_group_unlock_and_free(group);
+		return ERR_PTR(minor);
+	}
+
 	dev = device_create(vfio.class, NULL,
 			    MKDEV(MAJOR(vfio.group_devt), minor),
 			    group, "%d", iommu_group_id(iommu_group));
--
1.8.0

Re: [PATCH 1/1] vfio: put off the allocation of "minor" in vfio_create_group

[Alex Williamson]

On Thu, 2014-11-20 at 19:25 +0800, Zhen Lei wrote:
> The next code fragment "list_for_each_entry" is not depend on "minor". With this
> patch, the free of "minor" in "list_for_each_entry" can be reduced, and there is
> no functional change.

A reasonable micro-optimization, but I'm curious if you're actually
seeing some measurable overhead from this.  It seems like we'd need to
have multiple devices, all within the same IOMMU group, all probed by
vfio-pci at the same time to exercise the race condition.  Thanks,

Alex

> Signed-off-by: Zhen Lei <thunder.leizhen@huawei.com>
> ---
>  drivers/vfio/vfio.c | 13 ++++++-------
>  1 file changed, 6 insertions(+), 7 deletions(-)
> 
> diff --git a/drivers/vfio/vfio.c b/drivers/vfio/vfio.c
> index f018d8d..737eb468 100644
> --- a/drivers/vfio/vfio.c
> +++ b/drivers/vfio/vfio.c
> @@ -225,22 +225,21 @@ static struct vfio_group *vfio_create_group(struct iommu_group *iommu_group)
> 
>  	mutex_lock(&vfio.group_lock);
> 
> -	minor = vfio_alloc_group_minor(group);
> -	if (minor < 0) {
> -		vfio_group_unlock_and_free(group);
> -		return ERR_PTR(minor);
> -	}
> -
>  	/* Did we race creating this group? */
>  	list_for_each_entry(tmp, &vfio.group_list, vfio_next) {
>  		if (tmp->iommu_group == iommu_group) {
>  			vfio_group_get(tmp);
> -			vfio_free_group_minor(minor);
>  			vfio_group_unlock_and_free(group);
>  			return tmp;
>  		}
>  	}
> 
> +	minor = vfio_alloc_group_minor(group);
> +	if (minor < 0) {
> +		vfio_group_unlock_and_free(group);
> +		return ERR_PTR(minor);
> +	}
> +
>  	dev = device_create(vfio.class, NULL,
>  			    MKDEV(MAJOR(vfio.group_devt), minor),
>  			    group, "%d", iommu_group_id(iommu_group));
> --
> 1.8.0
> 
> 
> --
> To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
> the body of a message to majordomo@vger.kernel.org
> More majordomo info at  http://vger.kernel.org/majordomo-info.html
> Please read the FAQ at  http://www.tux.org/lkml/

Re: [PATCH 1/1] vfio: put off the allocation of "minor" in vfio_create_group

[leizhen]

On 2014/11/20 23:37, Alex Williamson wrote:
> On Thu, 2014-11-20 at 19:25 +0800, Zhen Lei wrote:
>> The next code fragment "list_for_each_entry" is not depend on "minor". With this
>> patch, the free of "minor" in "list_for_each_entry" can be reduced, and there is
>> no functional change.
> 
> A reasonable micro-optimization, but I'm curious if you're actually
> seeing some measurable overhead from this.  It seems like we'd need to

Oh, I just found this by code review. It's good to put pure check first and
resource allocation later, if they are order independent. Isn't it?
I think it will not improve performance but only save a little code space
and looks feel better.

> have multiple devices, all within the same IOMMU group, all probed by
> vfio-pci at the same time to exercise the race condition.  Thanks,

I think vfio_create_group can not be invoked frequently. So, the performance of this function
is not a major consideration, we don't care slightly performance change.

> 
> Alex
> 
>> Signed-off-by: Zhen Lei <thunder.leizhen@huawei.com>
>> ---
>>  drivers/vfio/vfio.c | 13 ++++++-------
>>  1 file changed, 6 insertions(+), 7 deletions(-)
>>
>> diff --git a/drivers/vfio/vfio.c b/drivers/vfio/vfio.c
>> index f018d8d..737eb468 100644
>> --- a/drivers/vfio/vfio.c
>> +++ b/drivers/vfio/vfio.c
>> @@ -225,22 +225,21 @@ static struct vfio_group *vfio_create_group(struct iommu_group *iommu_group)
>>
>>  	mutex_lock(&vfio.group_lock);
>>
>> -	minor = vfio_alloc_group_minor(group);
>> -	if (minor < 0) {
>> -		vfio_group_unlock_and_free(group);
>> -		return ERR_PTR(minor);
>> -	}
>> -
>>  	/* Did we race creating this group? */
>>  	list_for_each_entry(tmp, &vfio.group_list, vfio_next) {
>>  		if (tmp->iommu_group == iommu_group) {
>>  			vfio_group_get(tmp);
>> -			vfio_free_group_minor(minor);
>>  			vfio_group_unlock_and_free(group);
>>  			return tmp;
>>  		}
>>  	}
>>
>> +	minor = vfio_alloc_group_minor(group);
>> +	if (minor < 0) {
>> +		vfio_group_unlock_and_free(group);
>> +		return ERR_PTR(minor);
>> +	}
>> +
>>  	dev = device_create(vfio.class, NULL,
>>  			    MKDEV(MAJOR(vfio.group_devt), minor),
>>  			    group, "%d", iommu_group_id(iommu_group));
>> --
>> 1.8.0
>>
>>
>> --
>> To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
>> the body of a message to majordomo@vger.kernel.org
>> More majordomo info at  http://vger.kernel.org/majordomo-info.html
>> Please read the FAQ at  http://www.tux.org/lkml/
> 
> 
> 
> 
> .
> 

Re: [PATCH 1/1] vfio: put off the allocation of "minor" in vfio_create_group

[Alex Williamson]

On Thu, 2014-11-20 at 19:25 +0800, Zhen Lei wrote:
> The next code fragment "list_for_each_entry" is not depend on "minor". With this
> patch, the free of "minor" in "list_for_each_entry" can be reduced, and there is
> no functional change.
> 
> Signed-off-by: Zhen Lei <thunder.leizhen@huawei.com>
> ---
>  drivers/vfio/vfio.c | 13 ++++++-------
>  1 file changed, 6 insertions(+), 7 deletions(-)
> 
> diff --git a/drivers/vfio/vfio.c b/drivers/vfio/vfio.c
> index f018d8d..737eb468 100644
> --- a/drivers/vfio/vfio.c
> +++ b/drivers/vfio/vfio.c
> @@ -225,22 +225,21 @@ static struct vfio_group *vfio_create_group(struct iommu_group *iommu_group)
> 
>  	mutex_lock(&vfio.group_lock);
> 
> -	minor = vfio_alloc_group_minor(group);
> -	if (minor < 0) {
> -		vfio_group_unlock_and_free(group);
> -		return ERR_PTR(minor);
> -	}
> -
>  	/* Did we race creating this group? */
>  	list_for_each_entry(tmp, &vfio.group_list, vfio_next) {
>  		if (tmp->iommu_group == iommu_group) {
>  			vfio_group_get(tmp);
> -			vfio_free_group_minor(minor);
>  			vfio_group_unlock_and_free(group);
>  			return tmp;
>  		}
>  	}
> 
> +	minor = vfio_alloc_group_minor(group);
> +	if (minor < 0) {
> +		vfio_group_unlock_and_free(group);
> +		return ERR_PTR(minor);
> +	}
> +
>  	dev = device_create(vfio.class, NULL,
>  			    MKDEV(MAJOR(vfio.group_devt), minor),
>  			    group, "%d", iommu_group_id(iommu_group));

Applied to next for v4.1.  Sorry I didn't put this in earlier, it got
lost in my inbox.  Thanks,

Alex