[v7,3/5] telemetry: use unique socket paths for in-memory mode
Checks
Commit Message
When running in in-memory mode, multiple processes can use the same
runtime dir, leading to conflicts with the telemetry sockets in that
directory. We can resolve this by appending a suffix to each socket
beyond the first, with the suffix being an increasing counter value.
Each process uses the first unused socket counter value.
Signed-off-by: Bruce Richardson <bruce.richardson@intel.com>
---
lib/telemetry/telemetry.c | 45 +++++++++++++++++++++++++--------------
1 file changed, 29 insertions(+), 16 deletions(-)
Comments
Hi Bruce,
>-----Original Message-----
>From: Richardson, Bruce <bruce.richardson@intel.com>
>Sent: Friday 8 October 2021 18:19
>To: dev@dpdk.org
>Cc: Power, Ciara <ciara.power@intel.com>; David Marchand
><david.marchand@redhat.com>; Burakov, Anatoly
><anatoly.burakov@intel.com>; Kevin Traynor <ktraynor@redhat.com>;
>Richardson, Bruce <bruce.richardson@intel.com>
>Subject: [PATCH v7 3/5] telemetry: use unique socket paths for in-memory
>mode
>
>When running in in-memory mode, multiple processes can use the same
>runtime dir, leading to conflicts with the telemetry sockets in that directory.
>We can resolve this by appending a suffix to each socket beyond the first,
>with the suffix being an increasing counter value.
>Each process uses the first unused socket counter value.
>
>Signed-off-by: Bruce Richardson <bruce.richardson@intel.com>
>---
<snip>
Thanks,
Acked-by: Ciara Power <ciara.power@intel.com>
On Tue, Oct 12, 2021 at 04:37:07PM +0100, Power, Ciara wrote:
> Hi Bruce,
>
> >-----Original Message-----
> >From: Richardson, Bruce <bruce.richardson@intel.com>
> >Sent: Friday 8 October 2021 18:19
> >To: dev@dpdk.org
> >Cc: Power, Ciara <ciara.power@intel.com>; David Marchand
> ><david.marchand@redhat.com>; Burakov, Anatoly
> ><anatoly.burakov@intel.com>; Kevin Traynor <ktraynor@redhat.com>;
> >Richardson, Bruce <bruce.richardson@intel.com>
> >Subject: [PATCH v7 3/5] telemetry: use unique socket paths for in-memory
> >mode
> >
> >When running in in-memory mode, multiple processes can use the same
> >runtime dir, leading to conflicts with the telemetry sockets in that directory.
> >We can resolve this by appending a suffix to each socket beyond the first,
> >with the suffix being an increasing counter value.
> >Each process uses the first unused socket counter value.
> >
> >Signed-off-by: Bruce Richardson <bruce.richardson@intel.com>
> >---
> <snip>
>
> Thanks,
>
> Acked-by: Ciara Power <ciara.power@intel.com>
Thanks.
I'm actually in two minds as to whether to this patch should be merged with
the previous one. It was already reviewed and acked as a bugfix, but this
patch also affects a lot of the same areas of code. What do you think?
/Bruce
>-----Original Message-----
>From: Richardson, Bruce <bruce.richardson@intel.com>
>Sent: Tuesday 12 October 2021 16:41
>To: Power, Ciara <ciara.power@intel.com>
>Cc: dev@dpdk.org; David Marchand <david.marchand@redhat.com>;
>Burakov, Anatoly <anatoly.burakov@intel.com>; Kevin Traynor
><ktraynor@redhat.com>
>Subject: Re: [PATCH v7 3/5] telemetry: use unique socket paths for in-
>memory mode
>
>On Tue, Oct 12, 2021 at 04:37:07PM +0100, Power, Ciara wrote:
>> Hi Bruce,
>>
>> >-----Original Message-----
>> >From: Richardson, Bruce <bruce.richardson@intel.com>
>> >Sent: Friday 8 October 2021 18:19
>> >To: dev@dpdk.org
>> >Cc: Power, Ciara <ciara.power@intel.com>; David Marchand
>> ><david.marchand@redhat.com>; Burakov, Anatoly
>> ><anatoly.burakov@intel.com>; Kevin Traynor <ktraynor@redhat.com>;
>> >Richardson, Bruce <bruce.richardson@intel.com>
>> >Subject: [PATCH v7 3/5] telemetry: use unique socket paths for
>> >in-memory mode
>> >
>> >When running in in-memory mode, multiple processes can use the same
>> >runtime dir, leading to conflicts with the telemetry sockets in that
>directory.
>> >We can resolve this by appending a suffix to each socket beyond the
>> >first, with the suffix being an increasing counter value.
>> >Each process uses the first unused socket counter value.
>> >
>> >Signed-off-by: Bruce Richardson <bruce.richardson@intel.com>
>> >---
>> <snip>
>>
>> Thanks,
>>
>> Acked-by: Ciara Power <ciara.power@intel.com>
>
>Thanks.
>
>I'm actually in two minds as to whether to this patch should be merged with
>the previous one. It was already reviewed and acked as a bugfix, but this
>patch also affects a lot of the same areas of code. What do you think?
>
>/Bruce
Yes, I was actually wondering about that myself.
I think it would make more sense to merge them, rather than adding some code in patch 2 that then gets removed in patch 3.
Thanks,
Ciara
@@ -467,14 +467,14 @@ create_socket(char *path)
/* first check if we have a runtime dir */
if (stat(socket_dir, &st) < 0 || !S_ISDIR(st.st_mode)) {
TMTY_LOG(ERR, "Cannot access DPDK runtime directory: %s\n", socket_dir);
- goto error;
+ close(sock);
+ return -ENOENT;
}
/* check if current socket is active */
if (connect(sock, (void *)&sun, sizeof(sun)) == 0) {
- TMTY_LOG(ERR, "Error binding telemetry socket, path already in use\n");
- TMTY_LOG(ERR, "Use '--file-prefix' to select a different socket path, or '--no-telemetry' to disable\n");
- goto error;
+ close(sock);
+ return -EADDRINUSE;
}
/* socket is not active, delete and attempt rebind */
@@ -482,23 +482,20 @@ create_socket(char *path)
unlink(sun.sun_path);
if (bind(sock, (void *) &sun, sizeof(sun)) < 0) {
TMTY_LOG(ERR, "Error binding socket: %s\n", strerror(errno));
- goto error;
+ close(sock);
+ return -errno; /* if unlink failed, this will be -EADDRINUSE as above */
}
}
if (listen(sock, 1) < 0) {
TMTY_LOG(ERR, "Error calling listen for socket: %s\n", strerror(errno));
unlink(sun.sun_path);
- goto error;
+ close(sock);
+ return -errno;
}
TMTY_LOG(DEBUG, "Socket creation and binding ok\n");
return sock;
-
-error:
- close(sock);
- path[0] = 0;
- return -1;
}
static void
@@ -531,8 +528,10 @@ telemetry_legacy_init(void)
return -1;
}
v1_socket.sock = create_socket(v1_socket.path);
- if (v1_socket.sock < 0)
+ if (v1_socket.sock < 0) {
+ v1_socket.path[0] = '\0';
return -1;
+ }
rc = pthread_create(&t_old, NULL, socket_listener, &v1_socket);
if (rc != 0) {
TMTY_LOG(ERR, "Error with create legcay socket thread: %s\n",
@@ -553,7 +552,9 @@ telemetry_legacy_init(void)
static int
telemetry_v2_init(void)
{
+ char spath[sizeof(v2_socket.path)];
pthread_t t_new;
+ short suffix = 0;
int rc;
v2_socket.num_clients = &v2_clients;
@@ -564,15 +565,27 @@ telemetry_v2_init(void)
rte_telemetry_register_cmd("/help", command_help,
"Returns help text for a command. Parameters: string command");
v2_socket.fn = client_handler;
- if (strlcpy(v2_socket.path, get_socket_path(socket_dir, 2),
- sizeof(v2_socket.path)) >= sizeof(v2_socket.path)) {
+ if (strlcpy(spath, get_socket_path(socket_dir, 2), sizeof(spath)) >= sizeof(spath)) {
TMTY_LOG(ERR, "Error with socket binding, path too long\n");
return -1;
}
+ memcpy(v2_socket.path, spath, sizeof(v2_socket.path));
v2_socket.sock = create_socket(v2_socket.path);
- if (v2_socket.sock < 0)
- return -1;
+ while (v2_socket.sock < 0) {
+ /* bail out on unexpected error, or suffix wrap-around */
+ if (v2_socket.sock != -EADDRINUSE || suffix < 0) {
+ v2_socket.path[0] = '\0'; /* clear socket path */
+ return -1;
+ }
+ /* add a suffix to the path if the basic version fails */
+ if (snprintf(v2_socket.path, sizeof(v2_socket.path), "%s:%d",
+ spath, ++suffix) >= (int)sizeof(v2_socket.path)) {
+ TMTY_LOG(ERR, "Error with socket binding, path too long\n");
+ return -1;
+ }
+ v2_socket.sock = create_socket(v2_socket.path);
+ }
rc = pthread_create(&t_new, NULL, socket_listener, &v2_socket);
if (rc != 0) {
TMTY_LOG(ERR, "Error with create socket thread: %s\n",