[dpdk-dev,v4,2/9] bus: add device iterator
Checks
Commit Message
From: Jan Blunck <jblunck@infradead.org>
Signed-off-by: Jan Blunck <jblunck@infradead.org>
Signed-off-by: Gaetan Rivet <gaetan.rivet@6wind.com>
---
lib/librte_eal/common/include/rte_bus.h | 7 +++++++
lib/librte_eal/common/include/rte_dev.h | 17 +++++++++++++++++
2 files changed, 24 insertions(+)
Comments
21/06/2017 01:29, Gaetan Rivet:
> +/**
> + * Device comparison function.
> + *
> + * @param dev
> + * Device handle.
> + *
> + * @param data
> + * Data to compare against.
> + *
> + * @return
> + * 0 if the device matches the data.
> + * !0 if the device does not match.
> + * <0 if ordering is possible and the device is lower than the data.
> + * >0 if ordering is possible and the device is greater than the data.
> + */
> +typedef int (*rte_dev_cmp_t)(const struct rte_device *dev, const void *data);
data is really abstract.
Maybe a comment is missing to explain that data is better specified
in bus implementations?
Why not implement it for PCI?
On Wed, Jun 21, 2017 at 01:55:39PM +0200, Thomas Monjalon wrote:
> 21/06/2017 01:29, Gaetan Rivet:
> > +/**
> > + * Device comparison function.
> > + *
> > + * @param dev
> > + * Device handle.
> > + *
> > + * @param data
> > + * Data to compare against.
> > + *
> > + * @return
> > + * 0 if the device matches the data.
> > + * !0 if the device does not match.
> > + * <0 if ordering is possible and the device is lower than the data.
> > + * >0 if ordering is possible and the device is greater than the data.
> > + */
> > +typedef int (*rte_dev_cmp_t)(const struct rte_device *dev, const void *data);
>
> data is really abstract.
> Maybe a comment is missing to explain that data is better specified
> in bus implementations?
>
I'm not sure it is better specified in rte_bus though :).
However, the usage can be understood there, why it exists in the first
place.
I think bus iterators could benefit some more explanation about the why.
> Why not implement it for PCI?
>
I sent this series with only the patches from Jan, initially in the
version he solely developed. Only afterward did I fix a few bugs,
reworked a few APIs.
As such, two other series complete this patchset:
[dpdk-dev] [PATCH v3] pci: implement find_device bus operation
http://dpdk.org/ml/archives/dev/2017-June/067485.html
And
[dpdk-dev] [PATCH v3 0/3] eal: complete attach / detach support
http://dpdk.org/ml/archives/dev/2017-June/067516.html
It might make sense to merge all three series together.
They are conceptually linked very closely. The only reason I did not do
so at first was because I was unsure about who would take responsibility
for the attach / detach patchset, and if it had not be me I did not want to
put undue responsibility of my patches on whomever would.
But that point is moot now.
@@ -82,6 +82,12 @@ typedef int (*rte_bus_scan_t)(void);
typedef int (*rte_bus_probe_t)(void);
/**
+ * Device iterator to find a particular device on a bus.
+ */
+typedef struct rte_device * (*rte_bus_find_device_t)(rte_dev_cmp_t match,
+ const void *data);
+
+/**
* A structure describing a generic bus.
*/
struct rte_bus {
@@ -89,6 +95,7 @@ struct rte_bus {
const char *name; /**< Name of the bus */
rte_bus_scan_t scan; /**< Scan for devices attached to bus */
rte_bus_probe_t probe; /**< Probe devices on bus */
+ rte_bus_find_device_t find_device; /**< Find device on bus */
};
/**
@@ -191,6 +191,23 @@ int rte_eal_dev_attach(const char *name, const char *devargs);
*/
int rte_eal_dev_detach(const char *name);
+/**
+ * Device comparison function.
+ *
+ * @param dev
+ * Device handle.
+ *
+ * @param data
+ * Data to compare against.
+ *
+ * @return
+ * 0 if the device matches the data.
+ * !0 if the device does not match.
+ * <0 if ordering is possible and the device is lower than the data.
+ * >0 if ordering is possible and the device is greater than the data.
+ */
+typedef int (*rte_dev_cmp_t)(const struct rte_device *dev, const void *data);
+
#define RTE_PMD_EXPORT_NAME_ARRAY(n, idx) n##idx[]
#define RTE_PMD_EXPORT_NAME(name, idx) \