From: Ferruh Yigit <ferruh.yigit@intel.com>
To: Allain Legacy <allain.legacy@windriver.com>, thomas@monjalon.net
Cc: dev@dpdk.org, Matt Peters <matt.peters@windriver.com>
Subject: Re: [dpdk-dev] [PATCH v2] net/avp: remove resources when port is closed
Date: Wed, 5 Jun 2019 16:17:16 +0100 [thread overview]
Message-ID: <2938bfb6-7215-4098-74b4-d11fcc12dff4@intel.com> (raw)
In-Reply-To: <20190527170255.30853-1-allain.legacy@windriver.com>
On 5/27/2019 6:02 PM, Allain Legacy wrote:
> The rte_eth_dev_close() function now handles freeing resources for
> devices (e.g., mac_addrs). To conform with the new close() behaviour we
> are asserting the RTE_ETH_DEV_CLOSE_REMOVE flag so that
> rte_eth_dev_close() releases all device level dynamic memory.
>
> Second level memory allocated to each individual rx/tx queue is now
> freed as part of the close() operation therefore making it safe for the
> rte_eth_dev_close() function to free the device private data without
> orphaning the rx/tx queue pointers.
>
> Cc: Matt Peters <matt.peters@windriver.com>
> Signed-off-by: Allain Legacy <allain.legacy@windriver.com>
> ---
> drivers/net/avp/avp_ethdev.c | 44 ++++++++++++++++++++++++++++++++++++++++++--
> 1 file changed, 42 insertions(+), 2 deletions(-)
>
> diff --git a/drivers/net/avp/avp_ethdev.c b/drivers/net/avp/avp_ethdev.c
> index 09388d05f..0f7481c86 100644
> --- a/drivers/net/avp/avp_ethdev.c
> +++ b/drivers/net/avp/avp_ethdev.c
> @@ -959,6 +959,8 @@ eth_avp_dev_init(struct rte_eth_dev *eth_dev)
> eth_dev->dev_ops = &avp_eth_dev_ops;
> eth_dev->rx_pkt_burst = &avp_recv_pkts;
> eth_dev->tx_pkt_burst = &avp_xmit_pkts;
> + /* Let rte_eth_dev_close() release the port resources */
> + eth_dev->data->dev_flags |= RTE_ETH_DEV_CLOSE_REMOVE;
>
> if (rte_eal_process_type() != RTE_PROC_PRIMARY) {
> /*
> @@ -1940,8 +1942,25 @@ avp_dev_rx_queue_release(void *rx_queue)
> unsigned int i;
>
> for (i = 0; i < avp->num_rx_queues; i++) {
> - if (data->rx_queues[i] == rxq)
> + if (data->rx_queues[i] == rxq) {
> + rte_free(data->rx_queues[i]);
> data->rx_queues[i] = NULL;
> + }
> + }
> +}
> +
> +static void
> +avp_dev_rx_queue_release_all(struct rte_eth_dev *eth_dev)
> +{
> + struct avp_dev *avp = AVP_DEV_PRIVATE_TO_HW(eth_dev->data->dev_private);
> + struct rte_eth_dev_data *data = avp->dev_data;
> + unsigned int i;
> +
> + for (i = 0; i < avp->num_rx_queues; i++) {
> + if (data->rx_queues[i]) {
> + rte_free(data->rx_queues[i]);
> + data->rx_queues[i] = NULL;
> + }
> }
> }
>
> @@ -1954,8 +1973,25 @@ avp_dev_tx_queue_release(void *tx_queue)
> unsigned int i;
>
> for (i = 0; i < avp->num_tx_queues; i++) {
> - if (data->tx_queues[i] == txq)
> + if (data->tx_queues[i] == txq) {
> + rte_free(data->tx_queues[i]);
> data->tx_queues[i] = NULL;
> + }
> + }
> +}
> +
> +static void
> +avp_dev_tx_queue_release_all(struct rte_eth_dev *eth_dev)
> +{
> + struct avp_dev *avp = AVP_DEV_PRIVATE_TO_HW(eth_dev->data->dev_private);
> + struct rte_eth_dev_data *data = avp->dev_data;
> + unsigned int i;
> +
> + for (i = 0; i < avp->num_tx_queues; i++) {
> + if (data->tx_queues[i]) {
> + rte_free(data->tx_queues[i]);
> + data->tx_queues[i] = NULL;
> + }
> }
> }
>
> @@ -2104,6 +2140,10 @@ avp_dev_close(struct rte_eth_dev *eth_dev)
> /* continue */
> }
>
> + /* release dynamic storage for rx/tx queues */
> + avp_dev_rx_queue_release_all(eth_dev);
> + avp_dev_tx_queue_release_all(eth_dev);
> +
> unlock:
> rte_spinlock_unlock(&avp->lock);
> }
>
Patch looks correct as it is and cover the resource freeing on
'rte_eth_dev_close()' path, but not complete in remove path.
The remove path stack trace is like following:
rte_avp_pmd->.remove [eth_avp_pci_remove]
rte_eth_dev_pci_generic_remove()
eth_avp_dev_uninit()
rte_eth_dev_pci_release()
rte_eth_dev_release_port()
rte_eth_dev_release_port() will free the ethdev allocated resources but not PMD
private ones (like the queues freed above),
it looks like just adding a 'avp_dev_close()' call into the
'eth_avp_dev_uninit()' can solve this, can you please check this option?
And if it make sense, can you please send a new version?
prev parent reply other threads:[~2019-06-05 15:17 UTC|newest]
Thread overview: 4+ messages / expand[flat|nested] mbox.gz Atom feed top
2019-05-27 17:02 Allain Legacy
2019-05-31 14:13 ` Peters, Matt
2019-05-31 15:07 ` Ye Xiaolong
2019-06-05 15:17 ` Ferruh Yigit [this message]
Reply instructions:
You may reply publicly to this message via plain-text email
using any one of the following methods:
* Save the following mbox file, import it into your mail client,
and reply-to-all from there: mbox
Avoid top-posting and favor interleaved quoting:
https://en.wikipedia.org/wiki/Posting_style#Interleaved_style
* Reply using the --to, --cc, and --in-reply-to
switches of git-send-email(1):
git send-email \
--in-reply-to=2938bfb6-7215-4098-74b4-d11fcc12dff4@intel.com \
--to=ferruh.yigit@intel.com \
--cc=allain.legacy@windriver.com \
--cc=dev@dpdk.org \
--cc=matt.peters@windriver.com \
--cc=thomas@monjalon.net \
/path/to/YOUR_REPLY
https://kernel.org/pub/software/scm/git/docs/git-send-email.html
* If your mail client supports setting the In-Reply-To header
via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line
before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for NNTP newsgroup(s).