[dpdk-dev] [PATCH 0/3] Minor rawdev fixes and enhancements

[dpdk-dev] [PATCH 0/3] Minor rawdev fixes and enhancements

[Bruce Richardson]

Improve the rawdev info_get function to return the numa socket id value to
the user, and to allow querying basic info of an unknown device by allowing
generic info to be returned without having to know the device type.

Also add in the missing dump function to the map file.

Bruce Richardson (3):
  rawdev: allow calling info function for unknown rawdevs
  rawdev: return NUMA socket id to the user
  rawdev: fix missing dump function in map file

 lib/librte_rawdev/rte_rawdev.c           | 14 +++++++-------
 lib/librte_rawdev/rte_rawdev.h           |  8 +++++++-
 lib/librte_rawdev/rte_rawdev_version.map |  1 +
 3 files changed, 15 insertions(+), 8 deletions(-)

-- 
2.25.1

[dpdk-dev] [PATCH 1/3] rawdev: allow calling info function for unknown rawdevs

[Bruce Richardson]

To call the rte_rawdev_info_get() function, the user currently has to know
the underlying type of the device in order to pass an appropriate structure
or buffer as the dev_private pointer in the info structure. By allowing a
NULL value for this field, we can skip getting the device-specific info and
just return the generic info - including the device name and driver, which
can be used to determine the device type - to the user.

This ensures that basic info can be get for all rawdevs, without knowing
the type, and even if the info driver API call has not been implemented for
the device.

Cc: stable@dpdk.org
Signed-off-by: Bruce Richardson <bruce.richardson@intel.com>
---
 lib/librte_rawdev/rte_rawdev.c | 6 ++++--
 lib/librte_rawdev/rte_rawdev.h | 8 +++++++-
 2 files changed, 11 insertions(+), 3 deletions(-)

diff --git a/lib/librte_rawdev/rte_rawdev.c b/lib/librte_rawdev/rte_rawdev.c
index aec61f425..b18638435 100644
--- a/lib/librte_rawdev/rte_rawdev.c
+++ b/lib/librte_rawdev/rte_rawdev.c
@@ -90,8 +90,10 @@ rte_rawdev_info_get(uint16_t dev_id, struct rte_rawdev_info *dev_info)
 
 	rawdev = &rte_rawdevs[dev_id];
 
-	RTE_FUNC_PTR_OR_ERR_RET(*rawdev->dev_ops->dev_info_get, -ENOTSUP);
-	(*rawdev->dev_ops->dev_info_get)(rawdev, dev_info->dev_private);
+	if (dev_info->dev_private != NULL) {
+		RTE_FUNC_PTR_OR_ERR_RET(*rawdev->dev_ops->dev_info_get, -ENOTSUP);
+		(*rawdev->dev_ops->dev_info_get)(rawdev, dev_info->dev_private);
+	}
 
 	if (dev_info) {
 
diff --git a/lib/librte_rawdev/rte_rawdev.h b/lib/librte_rawdev/rte_rawdev.h
index ed011ca22..7fb6cb188 100644
--- a/lib/librte_rawdev/rte_rawdev.h
+++ b/lib/librte_rawdev/rte_rawdev.h
@@ -77,7 +77,13 @@ struct rte_rawdev_info;
  *
  * @param[out] dev_info
  *   A pointer to a structure of type *rte_rawdev_info* to be filled with the
- *   contextual information of the device.
+ *   contextual information of the device. The dev_info->dev_private field
+ *   should point to an appropriate buffer space for holding the device-
+ *   specific info for that hardware.
+ *   If the dev_private field is set to NULL, then the device-specific info
+ *   function will not be called and only basic information about the device
+ *   will be returned. This can be used to safely query the type of a rawdev
+ *   instance without needing to know the size of the private data to return.
  *
  * @return
  *   - 0: Success, driver updates the contextual information of the raw device
-- 
2.25.1

[dpdk-dev] [PATCH 2/3] rawdev: return NUMA socket id to the user

[Bruce Richardson]

The rawdev info struct has a socket_id field which was not filled in.

We can also omit the checks for the parameter struct being null, since
that is previously checked in the function.

Cc: stable@dpdk.org
Signed-off-by: Bruce Richardson <bruce.richardson@intel.com>
---
 lib/librte_rawdev/rte_rawdev.c | 8 +++-----
 1 file changed, 3 insertions(+), 5 deletions(-)

diff --git a/lib/librte_rawdev/rte_rawdev.c b/lib/librte_rawdev/rte_rawdev.c
index b18638435..9ee160455 100644
--- a/lib/librte_rawdev/rte_rawdev.c
+++ b/lib/librte_rawdev/rte_rawdev.c
@@ -95,11 +95,9 @@ rte_rawdev_info_get(uint16_t dev_id, struct rte_rawdev_info *dev_info)
 		(*rawdev->dev_ops->dev_info_get)(rawdev, dev_info->dev_private);
 	}
 
-	if (dev_info) {
-
-		dev_info->driver_name = rawdev->driver_name;
-		dev_info->device = rawdev->device;
-	}
+	dev_info->driver_name = rawdev->driver_name;
+	dev_info->device = rawdev->device;
+	dev_info->socket_id = rawdev->socket_id;
 
 	return 0;
 }
-- 
2.25.1

[dpdk-dev] [PATCH 3/3] rawdev: fix missing dump function in map file

[Bruce Richardson]

The rte_rawdev_dump function was missing from the map file, meaning it was
unavailable for use when linking dynamically.

Fixes: c88b3f2558ed ("rawdev: introduce raw device library")
Cc: shreyansh.jain@nxp.com
Cc: stable@dpdk.org

Signed-off-by: Bruce Richardson <bruce.richardson@intel.com>
---
 lib/librte_rawdev/rte_rawdev_version.map | 1 +
 1 file changed, 1 insertion(+)

diff --git a/lib/librte_rawdev/rte_rawdev_version.map b/lib/librte_rawdev/rte_rawdev_version.map
index d847c9e0d..63b54f598 100644
--- a/lib/librte_rawdev/rte_rawdev_version.map
+++ b/lib/librte_rawdev/rte_rawdev_version.map
@@ -5,6 +5,7 @@ DPDK_20.0 {
 	rte_rawdev_configure;
 	rte_rawdev_count;
 	rte_rawdev_dequeue_buffers;
+	rte_rawdev_dump;
 	rte_rawdev_enqueue_buffers;
 	rte_rawdev_firmware_load;
 	rte_rawdev_firmware_status_get;
-- 
2.25.1

Re: [dpdk-dev] [PATCH 3/3] rawdev: fix missing dump function in map file

[Hemant Agrawal]

Acked-by: Hemant Agrawal <hemant.agrawal@nxp.com>

Re: [dpdk-dev] [PATCH 2/3] rawdev: return NUMA socket id to the user

[Hemant Agrawal]

Acked-by: Hemant Agrawal <hemant.agrawal@nxp.com>

Re: [dpdk-dev] [PATCH 1/3] rawdev: allow calling info function for unknown rawdevs

[Hemant Agrawal]

Acked-by: Hemant Agrawal <hemant.agrawal@nxp.com>

Re: [dpdk-dev] [PATCH 1/3] rawdev: allow calling info function for unknown rawdevs

[Thomas Monjalon]

06/07/2020 12:31, Bruce Richardson:
> To call the rte_rawdev_info_get() function, the user currently has to know
> the underlying type of the device in order to pass an appropriate structure
> or buffer as the dev_private pointer in the info structure. By allowing a
> NULL value for this field, we can skip getting the device-specific info and
> just return the generic info - including the device name and driver, which
> can be used to determine the device type - to the user.
> 
> This ensures that basic info can be get for all rawdevs, without knowing
> the type, and even if the info driver API call has not been implemented for
> the device.
> 
> Cc: stable@dpdk.org

This is a new feature. Why should it be backported?

Re: [dpdk-dev] [dpdk-stable] [PATCH 2/3] rawdev: return NUMA socket id to the user

[Thomas Monjalon]

06/07/2020 12:31, Bruce Richardson:
> The rawdev info struct has a socket_id field which was not filled in.
> 
> We can also omit the checks for the parameter struct being null, since
> that is previously checked in the function.
> 
> Cc: stable@dpdk.org

It looks to be a fix.
Which commit introduced this bug? The original one?
Fixes: c88b3f2558ed ("rawdev: introduce raw device library")

Re: [dpdk-dev] [PATCH 1/3] rawdev: allow calling info function for unknown rawdevs

[Bruce Richardson]

On Tue, Jul 07, 2020 at 02:18:31PM +0200, Thomas Monjalon wrote:
> 06/07/2020 12:31, Bruce Richardson:
> > To call the rte_rawdev_info_get() function, the user currently has to
> > know the underlying type of the device in order to pass an appropriate
> > structure or buffer as the dev_private pointer in the info structure.
> > By allowing a NULL value for this field, we can skip getting the
> > device-specific info and just return the generic info - including the
> > device name and driver, which can be used to determine the device type
> > - to the user.
> > 
> > This ensures that basic info can be get for all rawdevs, without
> > knowing the type, and even if the info driver API call has not been
> > implemented for the device.
> > 
> > Cc: stable@dpdk.org
> 
> This is a new feature. Why should it be backported?
> 

I'd largely leave it up to the stable maintainers, but IMHO it should be
backported as its not a big change and I don't see the existing rawdev APIs
as very usable without it. I think the current API leaves the user in a bit
of a catch 22, since info_get() is the API you'd expect to call to find out
the actual type of a rawdev, but unfortunately, to call the info_get API
you need to know the type to pass in the appropriate type-specific
structure parameter. Therefore I see this as fixing a usability bug. :-)

Regards,
/Bruce

Re: [dpdk-dev] [dpdk-stable] [PATCH 2/3] rawdev: return NUMA socket id to the user

[Bruce Richardson]

On Tue, Jul 07, 2020 at 02:23:04PM +0200, Thomas Monjalon wrote:
> 06/07/2020 12:31, Bruce Richardson:
> > The rawdev info struct has a socket_id field which was not filled in.
> > 
> > We can also omit the checks for the parameter struct being null, since
> > that is previously checked in the function.
> > 
> > Cc: stable@dpdk.org
> 
> It looks to be a fix.
> Which commit introduced this bug? The original one?
> Fixes: c88b3f2558ed ("rawdev: introduce raw device library")
>

Yes, it can be viewed as a fix and that is the commit introducing the
issue.

/Bruce

Re: [dpdk-dev] [PATCH 0/3] Minor rawdev fixes and enhancements

[Thomas Monjalon]

06/07/2020 12:31, Bruce Richardson:
> Improve the rawdev info_get function to return the numa socket id value to
> the user, and to allow querying basic info of an unknown device by allowing
> generic info to be returned without having to know the device type.
> 
> Also add in the missing dump function to the map file.
> 
> Bruce Richardson (3):
>   rawdev: allow calling info function for unknown rawdevs
>   rawdev: return NUMA socket id to the user
>   rawdev: fix missing dump function in map file

Applied, thanks