From: Matan Azrad <matan@mellanox.com>
To: Thomas Monjalon <thomas@monjalon.net>, "dev@dpdk.org" <dev@dpdk.org>
Cc: "ferruh.yigit@intel.com" <ferruh.yigit@intel.com>,
"stable@dpdk.org" <stable@dpdk.org>,
Wenzhuo Lu <wenzhuo.lu@intel.com>,
Jingjing Wu <jingjing.wu@intel.com>,
Bernard Iremonger <bernard.iremonger@intel.com>
Subject: Re: [dpdk-dev] [PATCH 2/2] app/testpmd: fix hot-unplug detaching
Date: Sun, 16 Feb 2020 08:09:03 +0000 [thread overview]
Message-ID: <AM0PR0502MB401990B5DAFA4CDDCEF4479CD2170@AM0PR0502MB4019.eurprd05.prod.outlook.com> (raw)
In-Reply-To: <20200213155226.1024939-3-thomas@monjalon.net>
Hi Thomas
Thanks for the patches, I Saw it just now.
please see small comment below:
From: Thomas Monjalon
> There is a possible race condition in the hotplug path in rmv_port_callback().
> If a port is created between
> close_port(port_id) and detach_port_device(port_id), then the port_id will
> have been reallocated to a different device which will be wrongly detached.
>
> Since a check was added in detach_port_device() for manual detach case,
> the hotplug path was even more broken.
> It became impossible to run because the new check prevented to run
> detach_port_device() after the port is closed.
>
> The solution for both issues is to not rely on the port_id for detaching the
> rte_device.
> The function detach_port_device() is split to allow calling
> detach_device() directly with the rte_device pointer, saved before closing
> the port.
>
> Fixes: 43d0e304980a ("app/testpmd: fix invalid port detaching")
If you fix the race, Don't you think you need to add fixes line for the patch which created the race?
> Cc: stable@dpdk.org
>
> Signed-off-by: Thomas Monjalon <thomas@monjalon.net>
> ---
> app/test-pmd/testpmd.c | 48 ++++++++++++++++++++++++------------------
> 1 file changed, 28 insertions(+), 20 deletions(-)
>
> diff --git a/app/test-pmd/testpmd.c b/app/test-pmd/testpmd.c index
> 11203cb03d..035836adfb 100644
> --- a/app/test-pmd/testpmd.c
> +++ b/app/test-pmd/testpmd.c
> @@ -2633,32 +2633,17 @@ setup_attached_port(portid_t pi)
> printf("Done\n");
> }
>
> -void
> -detach_port_device(portid_t port_id)
> +static void
> +detach_device(struct rte_device *dev)
> {
> - struct rte_device *dev;
> portid_t sibling;
>
> - printf("Removing a device...\n");
> -
> - if (port_id_is_invalid(port_id, ENABLED_WARN))
> - return;
> -
> - dev = rte_eth_devices[port_id].device;
> if (dev == NULL) {
> printf("Device already removed\n");
> return;
> }
>
> - if (ports[port_id].port_status != RTE_PORT_CLOSED) {
> - if (ports[port_id].port_status != RTE_PORT_STOPPED) {
> - printf("Port not stopped\n");
> - return;
> - }
> - printf("Port was not closed\n");
> - if (ports[port_id].flow_list)
> - port_flow_flush(port_id);
> - }
> + printf("Removing a device...\n");
>
> if (rte_dev_remove(dev) < 0) {
> TESTPMD_LOG(ERR, "Failed to detach device %s\n", dev-
> >name); @@ -2676,13 +2661,31 @@ detach_port_device(portid_t port_id)
>
> remove_invalid_ports();
>
> - printf("Device of port %u is detached\n", port_id);
> + printf("Device is detached\n");
> printf("Now total ports is %d\n", nb_ports);
> printf("Done\n");
> return;
> }
>
> void
> +detach_port_device(portid_t port_id)
> +{
> + if (port_id_is_invalid(port_id, ENABLED_WARN))
> + return;
> +
> + if (ports[port_id].port_status != RTE_PORT_CLOSED) {
> + if (ports[port_id].port_status != RTE_PORT_STOPPED) {
> + printf("Port not stopped\n");
> + return;
> + }
> + printf("Port was not closed\n");
> + if (ports[port_id].flow_list)
> + port_flow_flush(port_id);
> + }
> +
> + detach_device(rte_eth_devices[port_id].device);
> +}
> +
> void
> detach_devargs(char *identifier)
> {
> @@ -2873,6 +2876,7 @@ rmv_port_callback(void *arg)
> int need_to_start = 0;
> int org_no_link_check = no_link_check;
> portid_t port_id = (intptr_t)arg;
> + struct rte_device *dev;
>
> RTE_ETH_VALID_PORTID_OR_RET(port_id);
>
> @@ -2883,8 +2887,12 @@ rmv_port_callback(void *arg)
> no_link_check = 1;
> stop_port(port_id);
> no_link_check = org_no_link_check;
> +
> + /* Save rte_device pointer before closing ethdev port */
> + dev = rte_eth_devices[port_id].device;
> close_port(port_id);
> - detach_port_device(port_id);
> + detach_device(dev); /* might be already removed or have more
> ports */
> +
> if (need_to_start)
> start_packet_forwarding(0);
> }
> --
> 2.25.0
next prev parent reply other threads:[~2020-02-16 8:09 UTC|newest]
Thread overview: 8+ messages / expand[flat|nested] mbox.gz Atom feed top
2020-02-13 15:52 [dpdk-dev] [PATCH 0/2] app/testpmd: fix detach Thomas Monjalon
2020-02-13 15:52 ` [dpdk-dev] [PATCH 1/2] app/testpmd: rename function for detaching by devargs Thomas Monjalon
2020-02-13 15:52 ` [dpdk-dev] [PATCH 2/2] app/testpmd: fix hot-unplug detaching Thomas Monjalon
2020-02-16 8:09 ` Matan Azrad [this message]
2020-02-16 9:47 ` Thomas Monjalon
2020-02-17 9:50 ` Kevin Traynor
2020-02-13 18:27 ` [dpdk-dev] [PATCH 0/2] app/testpmd: fix detach Ferruh Yigit
2020-02-13 20:05 ` Thomas Monjalon
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=AM0PR0502MB401990B5DAFA4CDDCEF4479CD2170@AM0PR0502MB4019.eurprd05.prod.outlook.com \
--to=matan@mellanox.com \
--cc=bernard.iremonger@intel.com \
--cc=dev@dpdk.org \
--cc=ferruh.yigit@intel.com \
--cc=jingjing.wu@intel.com \
--cc=stable@dpdk.org \
--cc=thomas@monjalon.net \
--cc=wenzhuo.lu@intel.com \
/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).