From: David Marchand <david.marchand@redhat.com>
To: "Min Hu (Connor)" <humin29@huawei.com>
Cc: dev <dev@dpdk.org>, "Yigit, Ferruh" <ferruh.yigit@intel.com>,
Ciara Power <ciara.power@intel.com>,
"Pattan, Reshma" <reshma.pattan@intel.com>
Subject: Re: [dpdk-dev] [PATCH v2 1/2] telemetry: fix missing check for thread creation
Date: Mon, 12 Apr 2021 09:48:19 +0200 [thread overview]
Message-ID: <CAJFAV8y9uLHnK4OyvX41inzKgtfqC1XfArOKYDMi73V9ZjRdDg@mail.gmail.com> (raw)
In-Reply-To: <1618187540-55614-2-git-send-email-humin29@huawei.com>
On Mon, Apr 12, 2021 at 2:32 AM Min Hu (Connor) <humin29@huawei.com> wrote:
>
> From: Chengwen Feng <fengchengwen@huawei.com>
>
> Add result check and message print out for thread creation after
> failure.
Ah, I was looking at this code too, while looking at your other series :-).
>
> Fixes: b80fe1805eee ("telemetry: introduce backward compatibility")
> Cc: stable@dpdk.org
>
> Signed-off-by: Chengwen Feng <fengchengwen@huawei.com>
> Signed-off-by: Min Hu (Connor) <humin29@huawei.com>
> ---
> lib/librte_telemetry/telemetry.c | 28 +++++++++++++++++++++++++---
> lib/librte_telemetry/telemetry_legacy.c | 10 ++++++++--
> 2 files changed, 33 insertions(+), 5 deletions(-)
>
> diff --git a/lib/librte_telemetry/telemetry.c b/lib/librte_telemetry/telemetry.c
> index 7e08afd..3f1a4c4 100644
> --- a/lib/librte_telemetry/telemetry.c
> +++ b/lib/librte_telemetry/telemetry.c
> @@ -350,6 +350,7 @@ socket_listener(void *socket)
> {
> while (1) {
> pthread_t th;
> + int rc;
> struct socket *s = (struct socket *)socket;
> int s_accepted = accept(s->sock, NULL, NULL);
> if (s_accepted < 0) {
> @@ -366,7 +367,14 @@ socket_listener(void *socket)
> __atomic_add_fetch(s->num_clients, 1,
> __ATOMIC_RELAXED);
> }
> - pthread_create(&th, NULL, s->fn, (void *)(uintptr_t)s_accepted);
> + rc = pthread_create(&th, NULL, s->fn,
> + (void *)(uintptr_t)s_accepted);
> + if (rc) {
Either you use this rc variable (adding strerror(rc) in the log
message below) or you can remove it.
This comment applies to others updates below.
> + TMTY_LOG(ERR, "Error with create thread, telemetry thread quitting\n");
> + close(s_accepted);
> + return NULL;
> + }
> +
> pthread_detach(th);
> }
> return NULL;
> @@ -425,6 +433,7 @@ static int
> telemetry_legacy_init(void)
> {
> pthread_t t_old;
> + int rc;
>
> if (num_legacy_callbacks == 1) {
> TMTY_LOG(WARNING, "No legacy callbacks, legacy socket not created\n");
> @@ -440,7 +449,13 @@ Fixes: 3b3757bda3c3 ("net/ice: get VF hardware index in DCF")
(void)
> v1_socket.sock = create_socket(v1_socket.path);
> if (v1_socket.sock < 0)
> return -1;
> - pthread_create(&t_old, NULL, socket_listener, &v1_socket);
> + rc = pthread_create(&t_old, NULL, socket_listener, &v1_socket);
> + if (rc) {
> + TMTY_LOG(ERR, "Error with create thread\n");
> + unlink_sockets();
Before this patch, v2 mode could be working fine with v1 ko.
I think it was intended since telemetry_legacy_init return value is
not checked and unlink_sockets() checks whether the v1 and v2 sockets
have been created.
Plus, in this error handling, the v1 socket fd is left open.
How about:
+ ret = pthread_create(&t_old, NULL, socket_listener, &v1_socket);
+ if (ret != 0) {
+ TMTY_LOG(DEBUG, "Failed to create handler for legacy
socket: %s\n",
+ strerror(ret));
+ close(v1_socket.sock);
+ v1_socket.sock = -1;
+ if (v1_socket.path[0]) {
+ unlink(v1_socket.path);
+ v1_socket.path[0] = '\0';
+ }
+ return -1;
+ }
> + return -1;
> + }
> +
> pthread_setaffinity_np(t_old, sizeof(*thread_cpuset), thread_cpuset);
>
> TMTY_LOG(DEBUG, "Legacy telemetry socket initialized ok\n");
> @@ -451,6 +466,7 @@ static int
> telemetry_v2_init(void)
> {
> pthread_t t_new;
> + int rc;
>
> v2_socket.num_clients = &v2_clients;
> rte_telemetry_register_cmd("/", list_commands,
> @@ -469,7 +485,13 @@ telemetry_v2_init(void)
> v2_socket.sock = create_socket(v2_socket.path);
> if (v2_socket.sock < 0)
> return -1;
> - pthread_create(&t_new, NULL, socket_listener, &v2_socket);
> + rc = pthread_create(&t_new, NULL, socket_listener, &v2_socket);
> + if (rc) {
> + TMTY_LOG(ERR, "Error with create thread\n");
> + unlink_sockets();
> + return -1;
> + }
> +
Missing a close on v2 socket,
> pthread_setaffinity_np(t_new, sizeof(*thread_cpuset), thread_cpuset);
> atexit(unlink_sockets);
>
> diff --git a/lib/librte_telemetry/telemetry_legacy.c b/lib/librte_telemetry/telemetry_legacy.c
> index 5e9af37..9587bfe 100644
> --- a/lib/librte_telemetry/telemetry_legacy.c
> +++ b/lib/librte_telemetry/telemetry_legacy.c
> @@ -83,6 +83,7 @@ register_client(const char *cmd __rte_unused, const char *params,
> pthread_t th;
> char data[BUF_SIZE];
> int fd;
> + int rc;
> struct sockaddr_un addrs;
> #endif /* !RTE_EXEC_ENV_WINDOWS */
>
> @@ -112,8 +113,13 @@ register_client(const char *cmd __rte_unused, const char *params,
> close(fd);
> return -1;
> }
> - pthread_create(&th, NULL, &legacy_client_handler,
> - (void *)(uintptr_t)fd);
> + rc = pthread_create(&th, NULL, &legacy_client_handler,
> + (void *)(uintptr_t)fd);
> + if (rc) {
> + fprintf(stderr, "Failed to create thread\n");
> + close(fd);
> + return -1;
> + }
> #endif /* !RTE_EXEC_ENV_WINDOWS */
> return 0;
> }
> --
> 2.7.4
>
--
David Marchand
next prev parent reply other threads:[~2021-04-12 7:48 UTC|newest]
Thread overview: 25+ messages / expand[flat|nested] mbox.gz Atom feed top
2021-04-10 10:44 [dpdk-dev] [PATCH 0/2] " Min Hu (Connor)
2021-04-10 10:44 ` [dpdk-dev] [PATCH 1/2] telemetry: " Min Hu (Connor)
2021-04-10 10:44 ` [dpdk-dev] [PATCH 2/2] test: " Min Hu (Connor)
2021-04-12 0:32 ` [dpdk-dev] [PATCH v2 0/2] " Min Hu (Connor)
2021-04-12 0:32 ` [dpdk-dev] [PATCH v2 1/2] telemetry: " Min Hu (Connor)
2021-04-12 7:48 ` David Marchand [this message]
2021-04-15 11:53 ` Min Hu (Connor)
2021-04-12 0:32 ` [dpdk-dev] [PATCH v2 2/2] test: " Min Hu (Connor)
2021-04-15 11:50 ` [dpdk-dev] [PATCH v3 0/2] " Min Hu (Connor)
2021-04-15 11:50 ` [dpdk-dev] [PATCH v3 1/2] telemetry: " Min Hu (Connor)
2021-04-15 14:55 ` David Marchand
2021-04-16 8:21 ` Min Hu (Connor)
2021-04-15 16:11 ` Power, Ciara
2021-04-16 8:19 ` Min Hu (Connor)
2021-04-15 11:50 ` [dpdk-dev] [PATCH v3 2/2] test: " Min Hu (Connor)
2021-04-15 17:05 ` Pattan, Reshma
2021-04-16 8:21 ` Min Hu (Connor)
2021-04-16 8:34 ` Ferruh Yigit
2021-04-16 9:16 ` Min Hu (Connor)
2021-04-16 8:18 ` [dpdk-dev] [PATCH v4 0/2] " Min Hu (Connor)
2021-04-16 8:18 ` [dpdk-dev] [PATCH v4 1/2] telemetry: " Min Hu (Connor)
2021-04-19 8:49 ` Power, Ciara
2021-04-16 8:18 ` [dpdk-dev] [PATCH v4 2/2] test: " Min Hu (Connor)
2021-04-19 9:33 ` Pattan, Reshma
2021-04-21 14:32 ` [dpdk-dev] [PATCH v4 0/2] " 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=CAJFAV8y9uLHnK4OyvX41inzKgtfqC1XfArOKYDMi73V9ZjRdDg@mail.gmail.com \
--to=david.marchand@redhat.com \
--cc=ciara.power@intel.com \
--cc=dev@dpdk.org \
--cc=ferruh.yigit@intel.com \
--cc=humin29@huawei.com \
--cc=reshma.pattan@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).