[v4,3/3] ethdev: allow close function to return an error
Checks
Commit Message
The API function rte_eth_dev_close() was returning void.
The return type is changed to int for notifying of errors.
If an error happens during a close operation,
the status of the port is undefined,
a maximum of resources having been freed.
Signed-off-by: Thomas Monjalon <thomas@monjalon.net>
Reviewed-by: Liron Himi <lironh@marvell.com>
Acked-by: Stephen Hemminger <stephen@networkplumber.org>
---
doc/guides/rel_notes/deprecation.rst | 1 -
doc/guides/rel_notes/release_20_11.rst | 4 +++-
drivers/net/cxgbe/cxgbe_ethdev.c | 5 +++--
drivers/net/cxgbe/cxgbevf_ethdev.c | 5 +++--
drivers/net/failsafe/failsafe_ether.c | 6 +++++-
drivers/net/failsafe/failsafe_ops.c | 25 +++++++++++++++++--------
drivers/net/memif/rte_eth_memif.c | 4 +---
drivers/net/mlx5/mlx5.c | 7 ++++---
drivers/net/mvneta/mvneta_ethdev.c | 5 +++--
drivers/net/mvpp2/mrvl_ethdev.c | 5 +++--
drivers/net/netvsc/hn_ethdev.c | 6 ++++--
drivers/net/netvsc/hn_var.h | 2 +-
drivers/net/netvsc/hn_vf.c | 7 +++++--
drivers/net/virtio/virtio_user_ethdev.c | 4 +---
lib/librte_ethdev/rte_ethdev.c | 16 +++++++++++-----
lib/librte_ethdev/rte_ethdev.h | 5 ++++-
16 files changed, 68 insertions(+), 39 deletions(-)
Comments
On 10/5/2020 6:08 PM, Thomas Monjalon wrote:
> The API function rte_eth_dev_close() was returning void.
> The return type is changed to int for notifying of errors.
>
> If an error happens during a close operation,
> the status of the port is undefined,
> a maximum of resources having been freed.
>
> Signed-off-by: Thomas Monjalon <thomas@monjalon.net>
> Reviewed-by: Liron Himi <lironh@marvell.com>
> Acked-by: Stephen Hemminger <stephen@networkplumber.org>
<...>
> -void
> +int
> rte_eth_dev_close(uint16_t port_id)
> {
> struct rte_eth_dev *dev;
> + int firsterr, binerr;
> + int *lasterr = &firsterr;
>
> - RTE_ETH_VALID_PORTID_OR_RET(port_id);
> + RTE_ETH_VALID_PORTID_OR_ERR_RET(port_id, -EINVAL);
> dev = &rte_eth_devices[port_id];
>
> - RTE_FUNC_PTR_OR_RET(*dev->dev_ops->dev_close);
> - (*dev->dev_ops->dev_close)(dev);
> + RTE_FUNC_PTR_OR_ERR_RET(*dev->dev_ops->dev_close, -ENOTSUP);
> + *lasterr = (*dev->dev_ops->dev_close)(dev);
> + if (*lasterr != 0)
> + lasterr = &binerr;
>
> rte_ethdev_trace_close(port_id);
> - rte_eth_dev_release_port(dev);
> + *lasterr = rte_eth_dev_release_port(dev);
> +
> + return firsterr;
> }
This may be personal taste but above error handling looks like unnecessary
complex, what do you think something like below:
close_err = (*dev->dev_ops->dev_close)(dev);
release_err = rte_eth_dev_release_port(dev);
return close_err ? close_err : release_err;
06/10/2020 11:43, Ferruh Yigit:
> On 10/5/2020 6:08 PM, Thomas Monjalon wrote:
> > The API function rte_eth_dev_close() was returning void.
> > The return type is changed to int for notifying of errors.
> >
> > If an error happens during a close operation,
> > the status of the port is undefined,
> > a maximum of resources having been freed.
> >
> > Signed-off-by: Thomas Monjalon <thomas@monjalon.net>
> > Reviewed-by: Liron Himi <lironh@marvell.com>
> > Acked-by: Stephen Hemminger <stephen@networkplumber.org>
>
> <...>
>
> > -void
> > +int
> > rte_eth_dev_close(uint16_t port_id)
> > {
> > struct rte_eth_dev *dev;
> > + int firsterr, binerr;
> > + int *lasterr = &firsterr;
> >
> > - RTE_ETH_VALID_PORTID_OR_RET(port_id);
> > + RTE_ETH_VALID_PORTID_OR_ERR_RET(port_id, -EINVAL);
> > dev = &rte_eth_devices[port_id];
> >
> > - RTE_FUNC_PTR_OR_RET(*dev->dev_ops->dev_close);
> > - (*dev->dev_ops->dev_close)(dev);
> > + RTE_FUNC_PTR_OR_ERR_RET(*dev->dev_ops->dev_close, -ENOTSUP);
> > + *lasterr = (*dev->dev_ops->dev_close)(dev);
> > + if (*lasterr != 0)
> > + lasterr = &binerr;
> >
> > rte_ethdev_trace_close(port_id);
> > - rte_eth_dev_release_port(dev);
> > + *lasterr = rte_eth_dev_release_port(dev);
> > +
> > + return firsterr;
> > }
>
> This may be personal taste but above error handling looks like unnecessary
> complex, what do you think something like below:
>
> close_err = (*dev->dev_ops->dev_close)(dev);
> release_err = rte_eth_dev_release_port(dev);
> return close_err ? close_err : release_err;
This is what I did first. Then thought about this "elegant" handling.
The pointer solution is just one more line and is more future proof.
I'm fine with any choice. Andrew?
On 10/6/20 1:57 PM, Thomas Monjalon wrote:
> 06/10/2020 11:43, Ferruh Yigit:
>> On 10/5/2020 6:08 PM, Thomas Monjalon wrote:
>>> The API function rte_eth_dev_close() was returning void.
>>> The return type is changed to int for notifying of errors.
>>>
>>> If an error happens during a close operation,
>>> the status of the port is undefined,
>>> a maximum of resources having been freed.
>>>
>>> Signed-off-by: Thomas Monjalon <thomas@monjalon.net>
>>> Reviewed-by: Liron Himi <lironh@marvell.com>
>>> Acked-by: Stephen Hemminger <stephen@networkplumber.org>
>>
>> <...>
>>
>>> -void
>>> +int
>>> rte_eth_dev_close(uint16_t port_id)
>>> {
>>> struct rte_eth_dev *dev;
>>> + int firsterr, binerr;
>>> + int *lasterr = &firsterr;
>>>
>>> - RTE_ETH_VALID_PORTID_OR_RET(port_id);
>>> + RTE_ETH_VALID_PORTID_OR_ERR_RET(port_id, -EINVAL);
Shouldn't it be -ENODEV in fact?
IMHO it would be nice to make it uniform across ethdev API.
>>> dev = &rte_eth_devices[port_id];
>>>
>>> - RTE_FUNC_PTR_OR_RET(*dev->dev_ops->dev_close);
>>> - (*dev->dev_ops->dev_close)(dev);
>>> + RTE_FUNC_PTR_OR_ERR_RET(*dev->dev_ops->dev_close, -ENOTSUP);
>>> + *lasterr = (*dev->dev_ops->dev_close)(dev);
>>> + if (*lasterr != 0)
>>> + lasterr = &binerr;
>>>
>>> rte_ethdev_trace_close(port_id);
>>> - rte_eth_dev_release_port(dev);
>>> + *lasterr = rte_eth_dev_release_port(dev);
>>> +
>>> + return firsterr;
>>> }
>>
>> This may be personal taste but above error handling looks like unnecessary
>> complex, what do you think something like below:
>>
>> close_err = (*dev->dev_ops->dev_close)(dev);
>> release_err = rte_eth_dev_release_port(dev);
>> return close_err ? close_err : release_err;
>
> This is what I did first. Then thought about this "elegant" handling.
> The pointer solution is just one more line and is more future proof.
>
> I'm fine with any choice. Andrew?
>
Hm, really hard to make a choice. I tend to choose Thomas's
version. Yes, I agree that it is a bit over-complicated in
this particular case, but thoughts to be future-proof
overweight for me. Plus it adds a pattern on how to handle
such cases in the future.
13/10/2020 10:40, Andrew Rybchenko:
> On 10/6/20 1:57 PM, Thomas Monjalon wrote:
> > 06/10/2020 11:43, Ferruh Yigit:
> >> On 10/5/2020 6:08 PM, Thomas Monjalon wrote:
> >>> The API function rte_eth_dev_close() was returning void.
> >>> The return type is changed to int for notifying of errors.
> >>>
> >>> If an error happens during a close operation,
> >>> the status of the port is undefined,
> >>> a maximum of resources having been freed.
> >>>
> >>> Signed-off-by: Thomas Monjalon <thomas@monjalon.net>
> >>> Reviewed-by: Liron Himi <lironh@marvell.com>
> >>> Acked-by: Stephen Hemminger <stephen@networkplumber.org>
> >>
> >> <...>
> >>
> >>> -void
> >>> +int
> >>> rte_eth_dev_close(uint16_t port_id)
> >>> {
> >>> struct rte_eth_dev *dev;
> >>> + int firsterr, binerr;
> >>> + int *lasterr = &firsterr;
> >>>
> >>> - RTE_ETH_VALID_PORTID_OR_RET(port_id);
> >>> + RTE_ETH_VALID_PORTID_OR_ERR_RET(port_id, -EINVAL);
>
> Shouldn't it be -ENODEV in fact?
Yes you're right.
> IMHO it would be nice to make it uniform across ethdev API.
Yes, do you want to make the patch to change other occurences to ENODEV?
> >>> dev = &rte_eth_devices[port_id];
> >>>
> >>> - RTE_FUNC_PTR_OR_RET(*dev->dev_ops->dev_close);
> >>> - (*dev->dev_ops->dev_close)(dev);
> >>> + RTE_FUNC_PTR_OR_ERR_RET(*dev->dev_ops->dev_close, -ENOTSUP);
> >>> + *lasterr = (*dev->dev_ops->dev_close)(dev);
> >>> + if (*lasterr != 0)
> >>> + lasterr = &binerr;
> >>>
> >>> rte_ethdev_trace_close(port_id);
> >>> - rte_eth_dev_release_port(dev);
> >>> + *lasterr = rte_eth_dev_release_port(dev);
> >>> +
> >>> + return firsterr;
> >>> }
> >>
> >> This may be personal taste but above error handling looks like unnecessary
> >> complex, what do you think something like below:
> >>
> >> close_err = (*dev->dev_ops->dev_close)(dev);
> >> release_err = rte_eth_dev_release_port(dev);
> >> return close_err ? close_err : release_err;
> >
> > This is what I did first. Then thought about this "elegant" handling.
> > The pointer solution is just one more line and is more future proof.
> >
> > I'm fine with any choice. Andrew?
> >
>
> Hm, really hard to make a choice. I tend to choose Thomas's
> version. Yes, I agree that it is a bit over-complicated in
> this particular case, but thoughts to be future-proof
> overweight for me. Plus it adds a pattern on how to handle
> such cases in the future.
Yes it's an elegant pattern :)
On 10/13/2020 9:55 AM, Thomas Monjalon wrote:
<...>
>>>>> dev = &rte_eth_devices[port_id];
>>>>>
>>>>> - RTE_FUNC_PTR_OR_RET(*dev->dev_ops->dev_close);
>>>>> - (*dev->dev_ops->dev_close)(dev);
>>>>> + RTE_FUNC_PTR_OR_ERR_RET(*dev->dev_ops->dev_close, -ENOTSUP);
>>>>> + *lasterr = (*dev->dev_ops->dev_close)(dev);
>>>>> + if (*lasterr != 0)
>>>>> + lasterr = &binerr;
>>>>>
>>>>> rte_ethdev_trace_close(port_id);
>>>>> - rte_eth_dev_release_port(dev);
>>>>> + *lasterr = rte_eth_dev_release_port(dev);
>>>>> +
>>>>> + return firsterr;
>>>>> }
>>>>
>>>> This may be personal taste but above error handling looks like unnecessary
>>>> complex, what do you think something like below:
>>>>
>>>> close_err = (*dev->dev_ops->dev_close)(dev);
>>>> release_err = rte_eth_dev_release_port(dev);
>>>> return close_err ? close_err : release_err;
>>>
>>> This is what I did first. Then thought about this "elegant" handling.
>>> The pointer solution is just one more line and is more future proof.
>>>
>>> I'm fine with any choice. Andrew?
>>>
>>
>> Hm, really hard to make a choice. I tend to choose Thomas's
>> version. Yes, I agree that it is a bit over-complicated in
>> this particular case, but thoughts to be future-proof
>> overweight for me. Plus it adds a pattern on how to handle
>> such cases in the future.
>
> Yes it's an elegant pattern :)
>
OK :)
@@ -135,7 +135,6 @@ Deprecation Notices
invalid port ID, unsupported operation, failed operation):
- ``rte_eth_dev_stop``
- - ``rte_eth_dev_close``
* ethdev: New offload flags ``DEV_RX_OFFLOAD_FLOW_MARK`` will be added in 19.11.
This will allow application to enable or disable PMDs from updating
@@ -189,7 +189,9 @@ API Changes
* ethdev: ``rte_eth_rx_descriptor_done()`` API has been deprecated.
-* Renamed internal ethdev APIs:
+* ethdev: Added ``int`` return type to ``rte_eth_dev_close()``.
+
+* ethdev: Renamed internal functions:
* ``_rte_eth_dev_callback_process()`` -> ``rte_eth_dev_callback_process()``
* ``_rte_eth_dev_reset`` -> ``rte_eth_dev_internal_reset()``
@@ -1296,12 +1296,13 @@ static int eth_cxgbe_dev_uninit(struct rte_eth_dev *eth_dev)
{
struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(eth_dev);
uint16_t port_id;
+ int err = 0;
/* Free up other ports and all resources */
RTE_ETH_FOREACH_DEV_OF(port_id, &pci_dev->device)
- rte_eth_dev_close(port_id);
+ err |= rte_eth_dev_close(port_id);
- return 0;
+ return err == 0 ? 0 : -EIO;
}
static int eth_cxgbe_pci_probe(struct rte_pci_driver *pci_drv __rte_unused,
@@ -183,12 +183,13 @@ static int eth_cxgbevf_dev_uninit(struct rte_eth_dev *eth_dev)
{
struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(eth_dev);
uint16_t port_id;
+ int err = 0;
/* Free up other ports and all resources */
RTE_ETH_FOREACH_DEV_OF(port_id, &pci_dev->device)
- rte_eth_dev_close(port_id);
+ err |= rte_eth_dev_close(port_id);
- return 0;
+ return err == 0 ? 0 : -EIO;
}
static int eth_cxgbevf_pci_probe(struct rte_pci_driver *pci_drv __rte_unused,
@@ -287,7 +287,11 @@ fs_dev_remove(struct sub_device *sdev)
/* fallthrough */
case DEV_ACTIVE:
failsafe_eth_dev_unregister_callbacks(sdev);
- rte_eth_dev_close(PORT_ID(sdev));
+ ret = rte_eth_dev_close(PORT_ID(sdev));
+ if (ret < 0) {
+ ERROR("Port close failed for sub-device %u",
+ PORT_ID(sdev));
+ }
sdev->state = DEV_PROBED;
/* fallthrough */
case DEV_PROBED:
@@ -638,7 +638,7 @@ failsafe_eth_dev_close(struct rte_eth_dev *dev)
{
struct sub_device *sdev;
uint8_t i;
- int ret;
+ int err, ret = 0;
fs_lock(dev, 0);
failsafe_hotplug_alarm_cancel(dev);
@@ -648,29 +648,38 @@ failsafe_eth_dev_close(struct rte_eth_dev *dev)
FOREACH_SUBDEV_STATE(sdev, i, dev, DEV_ACTIVE) {
DEBUG("Closing sub_device %d", i);
failsafe_eth_dev_unregister_callbacks(sdev);
- rte_eth_dev_close(PORT_ID(sdev));
+ err = rte_eth_dev_close(PORT_ID(sdev));
+ if (err) {
+ ret = ret ? ret : err;
+ ERROR("Error while closing sub-device %u",
+ PORT_ID(sdev));
+ }
sdev->state = DEV_ACTIVE - 1;
}
rte_eth_dev_callback_unregister(RTE_ETH_ALL, RTE_ETH_EVENT_NEW,
failsafe_eth_new_event_callback, dev);
if (rte_eal_process_type() != RTE_PROC_PRIMARY) {
fs_unlock(dev, 0);
- return 0;
+ return ret;
}
fs_dev_free_queues(dev);
- ret = failsafe_eal_uninit(dev);
- if (ret)
+ err = failsafe_eal_uninit(dev);
+ if (err) {
+ ret = ret ? ret : err;
ERROR("Error while uninitializing sub-EAL");
+ }
failsafe_args_free(dev);
rte_free(PRIV(dev)->subs);
rte_free(PRIV(dev)->mcast_addrs);
/* mac_addrs must not be freed alone because part of dev_private */
dev->data->mac_addrs = NULL;
fs_unlock(dev, 0);
- ret = pthread_mutex_destroy(&PRIV(dev)->hotplug_mutex);
- if (ret)
+ err = pthread_mutex_destroy(&PRIV(dev)->hotplug_mutex);
+ if (err) {
+ ret = ret ? ret : err;
ERROR("Error while destroying hotplug mutex");
- return 0;
+ }
+ return ret;
}
static int
@@ -1798,9 +1798,7 @@ rte_pmd_memif_remove(struct rte_vdev_device *vdev)
if (eth_dev == NULL)
return 0;
- rte_eth_dev_close(eth_dev->data->port_id);
-
- return 0;
+ return rte_eth_dev_close(eth_dev->data->port_id);
}
static struct rte_vdev_driver pmd_memif_drv = {
@@ -2010,6 +2010,7 @@ static int
mlx5_pci_remove(struct rte_pci_device *pci_dev)
{
uint16_t port_id;
+ int ret = 0;
RTE_ETH_FOREACH_DEV_OF(port_id, &pci_dev->device) {
/*
@@ -2017,11 +2018,11 @@ mlx5_pci_remove(struct rte_pci_device *pci_dev)
* call the close function explicitly for secondary process.
*/
if (rte_eal_process_type() == RTE_PROC_SECONDARY)
- mlx5_dev_close(&rte_eth_devices[port_id]);
+ ret |= mlx5_dev_close(&rte_eth_devices[port_id]);
else
- rte_eth_dev_close(port_id);
+ ret |= rte_eth_dev_close(port_id);
}
- return 0;
+ return ret == 0 ? 0 : -EIO;
}
static const struct rte_pci_id mlx5_pci_id_map[] = {
@@ -966,14 +966,15 @@ static int
rte_pmd_mvneta_remove(struct rte_vdev_device *vdev)
{
uint16_t port_id;
+ int ret = 0;
RTE_ETH_FOREACH_DEV(port_id) {
if (rte_eth_devices[port_id].device != &vdev->device)
continue;
- rte_eth_dev_close(port_id);
+ ret |= rte_eth_dev_close(port_id);
}
- return 0;
+ return ret == 0 ? 0 : -EIO;
}
static struct rte_vdev_driver pmd_mvneta_drv = {
@@ -3022,14 +3022,15 @@ static int
rte_pmd_mrvl_remove(struct rte_vdev_device *vdev)
{
uint16_t port_id;
+ int ret = 0;
RTE_ETH_FOREACH_DEV(port_id) {
if (rte_eth_devices[port_id].device != &vdev->device)
continue;
- rte_eth_dev_close(port_id);
+ ret |= rte_eth_dev_close(port_id);
}
- return 0;
+ return ret == 0 ? 0 : -EIO;
}
static struct rte_vdev_driver pmd_mrvl_drv = {
@@ -842,14 +842,16 @@ hn_dev_stop(struct rte_eth_dev *dev)
static int
hn_dev_close(struct rte_eth_dev *dev)
{
+ int ret;
+
PMD_INIT_FUNC_TRACE();
if (rte_eal_process_type() != RTE_PROC_PRIMARY)
return 0;
- hn_vf_close(dev);
+ ret = hn_vf_close(dev);
hn_dev_free_queues(dev);
- return 0;
+ return ret;
}
static const struct eth_dev_ops hn_eth_dev_ops = {
@@ -217,7 +217,7 @@ const uint32_t *hn_vf_supported_ptypes(struct rte_eth_dev *dev);
int hn_vf_start(struct rte_eth_dev *dev);
void hn_vf_reset(struct rte_eth_dev *dev);
void hn_vf_stop(struct rte_eth_dev *dev);
-void hn_vf_close(struct rte_eth_dev *dev);
+int hn_vf_close(struct rte_eth_dev *dev);
int hn_vf_allmulticast_enable(struct rte_eth_dev *dev);
int hn_vf_allmulticast_disable(struct rte_eth_dev *dev);
@@ -316,18 +316,21 @@ void hn_vf_reset(struct rte_eth_dev *dev)
VF_ETHDEV_FUNC(dev, rte_eth_dev_reset);
}
-void hn_vf_close(struct rte_eth_dev *dev)
+int hn_vf_close(struct rte_eth_dev *dev)
{
struct hn_data *hv = dev->data->dev_private;
uint16_t vf_port;
+ int ret = 0;
rte_rwlock_read_lock(&hv->vf_lock);
vf_port = hv->vf_port;
if (vf_port != HN_INVALID_PORT)
- rte_eth_dev_close(vf_port);
+ ret = rte_eth_dev_close(vf_port);
hv->vf_port = HN_INVALID_PORT;
rte_rwlock_read_unlock(&hv->vf_lock);
+
+ return ret;
}
int hn_vf_stats_reset(struct rte_eth_dev *dev)
@@ -881,9 +881,7 @@ virtio_user_pmd_remove(struct rte_vdev_device *vdev)
return rte_eth_dev_release_port(eth_dev);
/* make sure the device is stopped, queues freed */
- rte_eth_dev_close(eth_dev->data->port_id);
-
- return 0;
+ return rte_eth_dev_close(eth_dev->data->port_id);
}
static int virtio_user_pmd_dma_map(struct rte_vdev_device *vdev, void *addr,
@@ -1716,19 +1716,25 @@ rte_eth_dev_set_link_down(uint16_t port_id)
return eth_err(port_id, (*dev->dev_ops->dev_set_link_down)(dev));
}
-void
+int
rte_eth_dev_close(uint16_t port_id)
{
struct rte_eth_dev *dev;
+ int firsterr, binerr;
+ int *lasterr = &firsterr;
- RTE_ETH_VALID_PORTID_OR_RET(port_id);
+ RTE_ETH_VALID_PORTID_OR_ERR_RET(port_id, -EINVAL);
dev = &rte_eth_devices[port_id];
- RTE_FUNC_PTR_OR_RET(*dev->dev_ops->dev_close);
- (*dev->dev_ops->dev_close)(dev);
+ RTE_FUNC_PTR_OR_ERR_RET(*dev->dev_ops->dev_close, -ENOTSUP);
+ *lasterr = (*dev->dev_ops->dev_close)(dev);
+ if (*lasterr != 0)
+ lasterr = &binerr;
rte_ethdev_trace_close(port_id);
- rte_eth_dev_release_port(dev);
+ *lasterr = rte_eth_dev_release_port(dev);
+
+ return firsterr;
}
int
@@ -2281,8 +2281,11 @@ int rte_eth_dev_set_link_down(uint16_t port_id);
*
* @param port_id
* The port identifier of the Ethernet device.
+ * @return
+ * - Zero if the port is closed successfully.
+ * - Negative if something went wrong.
*/
-void rte_eth_dev_close(uint16_t port_id);
+int rte_eth_dev_close(uint16_t port_id);
/**
* Reset a Ethernet device and keep its port id.