diff mbox series

[v4,2/4] drivers/bus: move driver assignment to end of probing

Message ID 20181011210251.7705-3-thomas@monjalon.net (mailing list archive)
State Superseded, archived
Delegated to: Thomas Monjalon
Headers
Series eal: allow hotplug to skip an already probed device |

Checks

Context Check Description
ci/checkpatch success coding style OK
ci/Intel-compilation fail Compilation issues

Commit Message

Thomas Monjalon Oct. 11, 2018, 9:02 p.m. UTC 
  The PCI mapping requires to know the PCI driver to use,
even before the probing is done. That's why the PCI driver is
referenced early inside the PCI device structure. See
commit 1d20a073fa5e ("bus/pci: reference driver structure before mapping")

However the rte_driver does not need to be referenced in rte_device
before the device probing is done.
By moving back this assignment at the end of the device probing,
it becomes possible to make clear the status of a rte_device.

Signed-off-by: Thomas Monjalon <thomas@monjalon.net>
Reviewed-by: Andrew Rybchenko <arybchenko@solarflare.com>
---
 drivers/bus/ifpga/ifpga_bus.c           | 9 ++++-----
 drivers/bus/pci/pci_common.c            | 7 +++----
 drivers/bus/pci/rte_bus_pci.h           | 2 +-
 drivers/bus/vdev/vdev.c                 | 5 ++---
 drivers/bus/vmbus/vmbus_common.c        | 5 +++--
 drivers/net/i40e/i40e_vf_representor.c  | 3 ---
 drivers/net/mlx4/mlx4.c                 | 1 -
 drivers/net/mlx5/mlx5.c                 | 1 -
 lib/librte_eal/common/include/rte_dev.h | 2 +-
 9 files changed, 14 insertions(+), 21 deletions(-)

Comments

Andrew Rybchenko Oct. 12, 2018, 7:44 a.m. UTC | #1 
Hi, Thomas,

On 10/12/18 12:02 AM, Thomas Monjalon wrote:
> The PCI mapping requires to know the PCI driver to use,
> even before the probing is done. That's why the PCI driver is
> referenced early inside the PCI device structure. See
> commit 1d20a073fa5e ("bus/pci: reference driver structure before mapping")
>
> However the rte_driver does not need to be referenced in rte_device
> before the device probing is done.
> By moving back this assignment at the end of the device probing,
> it becomes possible to make clear the status of a rte_device.
>
> Signed-off-by: Thomas Monjalon <thomas@monjalon.net>
> Reviewed-by: Andrew Rybchenko <arybchenko@solarflare.com>

Have you seen driver->name usage in drivers/net/szedata2/rte_eth_szedata2.c
rte_szedata2_eth_dev_init() which is used from probe?

It looks like drivers/net/mlx5/mlx5.c
mlx5_pci_probe()->mlx5_dev_spawn()->mlx5_dev_to_port_id() goes
to device->driver. Is the code OK with the move?

I've added above drivers maintainers in CC.
Jan Remes Oct. 12, 2018, 8:32 a.m. UTC | #2 
Hi Andrew,

On Fri, Oct 12, 2018 at 9:45 AM Andrew Rybchenko
<arybchenko@solarflare.com> wrote:
>
> Hi, Thomas,
>
> On 10/12/18 12:02 AM, Thomas Monjalon wrote:
> > The PCI mapping requires to know the PCI driver to use,
> > even before the probing is done. That's why the PCI driver is
> > referenced early inside the PCI device structure. See
> > commit 1d20a073fa5e ("bus/pci: reference driver structure before mapping")
> >
> > However the rte_driver does not need to be referenced in rte_device
> > before the device probing is done.
> > By moving back this assignment at the end of the device probing,
> > it becomes possible to make clear the status of a rte_device.
> >
> > Signed-off-by: Thomas Monjalon <thomas@monjalon.net>
> > Reviewed-by: Andrew Rybchenko <arybchenko@solarflare.com>
>
> Have you seen driver->name usage in drivers/net/szedata2/rte_eth_szedata2.c
> rte_szedata2_eth_dev_init() which is used from probe?
>

for szedata2, this is easily fixed by using the driver name directly
from the #define above.

diff --git a/drivers/net/szedata2/rte_eth_szedata2.c
b/drivers/net/szedata2/rte_eth_szedata2.c
index 8f92e72..4e5e01c 100644
--- a/drivers/net/szedata2/rte_eth_szedata2.c
+++ b/drivers/net/szedata2/rte_eth_szedata2.c
@@ -1474,7 +1474,7 @@ struct szedata2_tx_queue {
        PMD_INIT_FUNC_TRACE();

        PMD_INIT_LOG(INFO, "Initializing eth_dev %s (driver %s)", data->name,
-                       dev->device->driver->name);
+                       RTE_STR(RTE_SZEDATA2_DRIVER_NAME));

        /* Fill internal private structure. */
        internals->dev = dev;
@@ -1525,7 +1525,7 @@ struct szedata2_tx_queue {
        ether_addr_copy(&eth_addr, data->mac_addrs);

        PMD_INIT_LOG(INFO, "%s device %s successfully initialized",
-                       dev->device->driver->name, data->name);
+                       RTE_STR(RTE_SZEDATA2_DRIVER_NAME), data->name);

        return 0;
 }
@@ -1547,7 +1547,7 @@ struct szedata2_tx_queue {
        rte_free(dev->data->mac_addrs);

        PMD_DRV_LOG(INFO, "%s device %s successfully uninitialized",
-                       dev->device->driver->name, dev->data->name);
+                       RTE_STR(RTE_SZEDATA2_DRIVER_NAME), dev->data->name);

        return 0;
 }

> It looks like drivers/net/mlx5/mlx5.c
> mlx5_pci_probe()->mlx5_dev_spawn()->mlx5_dev_to_port_id() goes
> to device->driver. Is the code OK with the move?
>
> I've added above drivers maintainers in CC.

Jan
Thomas Monjalon Oct. 12, 2018, 10:45 a.m. UTC | #3 
12/10/2018 10:32, Jan Remeš:
> Hi Andrew,
> 
> On Fri, Oct 12, 2018 at 9:45 AM Andrew Rybchenko
> <arybchenko@solarflare.com> wrote:
> >
> > Hi, Thomas,
> >
> > On 10/12/18 12:02 AM, Thomas Monjalon wrote:
> > > The PCI mapping requires to know the PCI driver to use,
> > > even before the probing is done. That's why the PCI driver is
> > > referenced early inside the PCI device structure. See
> > > commit 1d20a073fa5e ("bus/pci: reference driver structure before mapping")
> > >
> > > However the rte_driver does not need to be referenced in rte_device
> > > before the device probing is done.
> > > By moving back this assignment at the end of the device probing,
> > > it becomes possible to make clear the status of a rte_device.
> > >
> > > Signed-off-by: Thomas Monjalon <thomas@monjalon.net>
> > > Reviewed-by: Andrew Rybchenko <arybchenko@solarflare.com>
> >
> > Have you seen driver->name usage in drivers/net/szedata2/rte_eth_szedata2.c
> > rte_szedata2_eth_dev_init() which is used from probe?
> >
> 
> for szedata2, this is easily fixed by using the driver name directly
> from the #define above.
> 
> diff --git a/drivers/net/szedata2/rte_eth_szedata2.c
> b/drivers/net/szedata2/rte_eth_szedata2.c
> index 8f92e72..4e5e01c 100644
> --- a/drivers/net/szedata2/rte_eth_szedata2.c
> +++ b/drivers/net/szedata2/rte_eth_szedata2.c
> @@ -1474,7 +1474,7 @@ struct szedata2_tx_queue {
>         PMD_INIT_FUNC_TRACE();
> 
>         PMD_INIT_LOG(INFO, "Initializing eth_dev %s (driver %s)", data->name,
> -                       dev->device->driver->name);
> +                       RTE_STR(RTE_SZEDATA2_DRIVER_NAME));
> 
>         /* Fill internal private structure. */
>         internals->dev = dev;
> @@ -1525,7 +1525,7 @@ struct szedata2_tx_queue {
>         ether_addr_copy(&eth_addr, data->mac_addrs);
> 
>         PMD_INIT_LOG(INFO, "%s device %s successfully initialized",
> -                       dev->device->driver->name, data->name);
> +                       RTE_STR(RTE_SZEDATA2_DRIVER_NAME), data->name);
> 
>         return 0;
>  }
> @@ -1547,7 +1547,7 @@ struct szedata2_tx_queue {
>         rte_free(dev->data->mac_addrs);
> 
>         PMD_DRV_LOG(INFO, "%s device %s successfully uninitialized",
> -                       dev->device->driver->name, dev->data->name);
> +                       RTE_STR(RTE_SZEDATA2_DRIVER_NAME), dev->data->name);
> 
>         return 0;
>  }

OK, I'll include such change in next version.
Thomas Monjalon Oct. 12, 2018, 3:50 p.m. UTC | #4 
12/10/2018 09:44, Andrew Rybchenko:
> Hi, Thomas,
> 
> On 10/12/18 12:02 AM, Thomas Monjalon wrote:
> > The PCI mapping requires to know the PCI driver to use,
> > even before the probing is done. That's why the PCI driver is
> > referenced early inside the PCI device structure. See
> > commit 1d20a073fa5e ("bus/pci: reference driver structure before mapping")
> >
> > However the rte_driver does not need to be referenced in rte_device
> > before the device probing is done.
> > By moving back this assignment at the end of the device probing,
> > it becomes possible to make clear the status of a rte_device.
> >
> > Signed-off-by: Thomas Monjalon <thomas@monjalon.net>
> > Reviewed-by: Andrew Rybchenko <arybchenko@solarflare.com>
> 
> Have you seen driver->name usage in drivers/net/szedata2/rte_eth_szedata2.c
> rte_szedata2_eth_dev_init() which is used from probe?
> 
> It looks like drivers/net/mlx5/mlx5.c
> mlx5_pci_probe()->mlx5_dev_spawn()->mlx5_dev_to_port_id() goes
> to device->driver. Is the code OK with the move?

This access in mlx5 looks useless.
It is comparing driver names while already comparing rte_device pointers.
If 2 devices are the same, they will have the same driver...
I will remove it.
diff mbox series

Patch

diff --git a/drivers/bus/ifpga/ifpga_bus.c b/drivers/bus/ifpga/ifpga_bus.c
index 80663328a..fca2dbace 100644
--- a/drivers/bus/ifpga/ifpga_bus.c
+++ b/drivers/bus/ifpga/ifpga_bus.c
@@ -280,14 +280,13 @@  ifpga_probe_one_driver(struct rte_afu_driver *drv,
 
 	/* reference driver structure */
 	afu_dev->driver = drv;
-	afu_dev->device.driver = &drv->driver;
 
 	/* call the driver probe() function */
 	ret = drv->probe(afu_dev);
-	if (ret) {
+	if (ret)
 		afu_dev->driver = NULL;
-		afu_dev->device.driver = NULL;
-	}
+	else
+		afu_dev->device.driver = &drv->driver;
 
 	return ret;
 }
@@ -302,7 +301,7 @@  ifpga_probe_all_drivers(struct rte_afu_device *afu_dev)
 		return -1;
 
 	/* Check if a driver is already loaded */
-	if (afu_dev->driver != NULL)
+	if (afu_dev->device.driver != NULL)
 		return 0;
 
 	TAILQ_FOREACH(drv, &ifpga_afu_drv_list, next) {
diff --git a/drivers/bus/pci/pci_common.c b/drivers/bus/pci/pci_common.c
index c7695d108..d63e68045 100644
--- a/drivers/bus/pci/pci_common.c
+++ b/drivers/bus/pci/pci_common.c
@@ -160,14 +160,12 @@  rte_pci_probe_one_driver(struct rte_pci_driver *dr,
 	 * driver flags for adjusting configuration.
 	 */
 	dev->driver = dr;
-	dev->device.driver = &dr->driver;
 
 	if (dr->drv_flags & RTE_PCI_DRV_NEED_MAPPING) {
 		/* map resources for devices that use igb_uio */
 		ret = rte_pci_map_device(dev);
 		if (ret != 0) {
 			dev->driver = NULL;
-			dev->device.driver = NULL;
 			return ret;
 		}
 	}
@@ -176,7 +174,6 @@  rte_pci_probe_one_driver(struct rte_pci_driver *dr,
 	ret = dr->probe(dr, dev);
 	if (ret) {
 		dev->driver = NULL;
-		dev->device.driver = NULL;
 		if ((dr->drv_flags & RTE_PCI_DRV_NEED_MAPPING) &&
 			/* Don't unmap if device is unsupported and
 			 * driver needs mapped resources.
@@ -184,6 +181,8 @@  rte_pci_probe_one_driver(struct rte_pci_driver *dr,
 			!(ret > 0 &&
 				(dr->drv_flags & RTE_PCI_DRV_KEEP_MAPPED_RES)))
 			rte_pci_unmap_device(dev);
+	} else {
+		dev->device.driver = &dr->driver;
 	}
 
 	return ret;
@@ -244,7 +243,7 @@  pci_probe_all_drivers(struct rte_pci_device *dev)
 		return -1;
 
 	/* Check if a driver is already loaded */
-	if (dev->driver != NULL)
+	if (dev->device.driver != NULL)
 		return 0;
 
 	FOREACH_DRIVER_ON_PCIBUS(dr) {
diff --git a/drivers/bus/pci/rte_bus_pci.h b/drivers/bus/pci/rte_bus_pci.h
index 0d1955ffe..984df2b37 100644
--- a/drivers/bus/pci/rte_bus_pci.h
+++ b/drivers/bus/pci/rte_bus_pci.h
@@ -62,7 +62,7 @@  struct rte_pci_device {
 	struct rte_mem_resource mem_resource[PCI_MAX_RESOURCE];
 					    /**< PCI Memory Resource */
 	struct rte_intr_handle intr_handle; /**< Interrupt handle */
-	struct rte_pci_driver *driver;      /**< Associated driver */
+	struct rte_pci_driver *driver;      /**< PCI driver used in probing */
 	uint16_t max_vfs;                   /**< sriov enable if not zero */
 	enum rte_kernel_driver kdrv;        /**< Kernel driver passthrough */
 	char name[PCI_PRI_STR_SIZE+1];      /**< PCI location (ASCII) */
diff --git a/drivers/bus/vdev/vdev.c b/drivers/bus/vdev/vdev.c
index 0142fb2c8..3f27f3510 100644
--- a/drivers/bus/vdev/vdev.c
+++ b/drivers/bus/vdev/vdev.c
@@ -150,10 +150,9 @@  vdev_probe_all_drivers(struct rte_vdev_device *dev)
 
 	if (vdev_parse(name, &driver))
 		return -1;
-	dev->device.driver = &driver->driver;
 	ret = driver->probe(dev);
-	if (ret)
-		dev->device.driver = NULL;
+	if (ret == 0)
+		dev->device.driver = &driver->driver;
 	return ret;
 }
 
diff --git a/drivers/bus/vmbus/vmbus_common.c b/drivers/bus/vmbus/vmbus_common.c
index e3ceb6906..de5548aa4 100644
--- a/drivers/bus/vmbus/vmbus_common.c
+++ b/drivers/bus/vmbus/vmbus_common.c
@@ -111,7 +111,6 @@  vmbus_probe_one_driver(struct rte_vmbus_driver *dr,
 
 	/* reference driver structure */
 	dev->driver = dr;
-	dev->device.driver = &dr->driver;
 
 	if (dev->device.numa_node < 0) {
 		VMBUS_LOG(WARNING, "  Invalid NUMA socket, default to 0");
@@ -124,6 +123,8 @@  vmbus_probe_one_driver(struct rte_vmbus_driver *dr,
 	if (ret) {
 		dev->driver = NULL;
 		rte_vmbus_unmap_device(dev);
+	} else {
+		dev->device.driver = &dr->driver;
 	}
 
 	return ret;
@@ -142,7 +143,7 @@  vmbus_probe_all_drivers(struct rte_vmbus_device *dev)
 	int rc;
 
 	/* Check if a driver is already loaded */
-	if (dev->driver != NULL) {
+	if (dev->device.driver != NULL) {
 		VMBUS_LOG(DEBUG, "VMBUS driver already loaded");
 		return 0;
 	}
diff --git a/drivers/net/i40e/i40e_vf_representor.c b/drivers/net/i40e/i40e_vf_representor.c
index 0bfbb4f60..24751d13c 100644
--- a/drivers/net/i40e/i40e_vf_representor.c
+++ b/drivers/net/i40e/i40e_vf_representor.c
@@ -487,9 +487,6 @@  i40e_vf_representor_init(struct rte_eth_dev *ethdev, void *init_params)
 	if (representor->vf_id >= pf->vf_num)
 		return -ENODEV;
 
-	/** representor shares the same driver as it's PF device */
-	ethdev->device->driver = representor->adapter->eth_dev->device->driver;
-
 	/* Set representor device ops */
 	ethdev->dev_ops = &i40e_representor_dev_ops;
 
diff --git a/drivers/net/mlx4/mlx4.c b/drivers/net/mlx4/mlx4.c
index defc0d4b0..3de7bc53e 100644
--- a/drivers/net/mlx4/mlx4.c
+++ b/drivers/net/mlx4/mlx4.c
@@ -734,7 +734,6 @@  mlx4_pci_probe(struct rte_pci_driver *pci_drv, struct rte_pci_device *pci_dev)
 		eth_dev->data->mac_addrs = priv->mac;
 		eth_dev->device = &pci_dev->device;
 		rte_eth_copy_pci_info(eth_dev, pci_dev);
-		eth_dev->device->driver = &mlx4_driver.driver;
 		/* Initialize local interrupt handle for current port. */
 		priv->intr_handle = (struct rte_intr_handle){
 			.fd = -1,
diff --git a/drivers/net/mlx5/mlx5.c b/drivers/net/mlx5/mlx5.c
index af4a78ce9..f23ef2c26 100644
--- a/drivers/net/mlx5/mlx5.c
+++ b/drivers/net/mlx5/mlx5.c
@@ -1083,7 +1083,6 @@  mlx5_dev_spawn(struct rte_device *dpdk_dev,
 	priv->dev_data = eth_dev->data;
 	eth_dev->data->mac_addrs = priv->mac;
 	eth_dev->device = dpdk_dev;
-	eth_dev->device->driver = &mlx5_driver.driver;
 	err = mlx5_uar_init_primary(eth_dev);
 	if (err) {
 		err = rte_errno;
diff --git a/lib/librte_eal/common/include/rte_dev.h b/lib/librte_eal/common/include/rte_dev.h
index 036180ff3..5084c645b 100644
--- a/lib/librte_eal/common/include/rte_dev.h
+++ b/lib/librte_eal/common/include/rte_dev.h
@@ -156,7 +156,7 @@  struct rte_driver {
 struct rte_device {
 	TAILQ_ENTRY(rte_device) next; /**< Next device */
 	const char *name;             /**< Device name */
-	const struct rte_driver *driver;/**< Associated driver */
+	const struct rte_driver *driver; /**< Driver assigned after probing */
 	const struct rte_bus *bus;    /**< Bus handle assigned on scan */
 	int numa_node;                /**< NUMA node connection */
 	struct rte_devargs *devargs;  /**< Device user arguments */