Received-SPF: pass (google.com: best guess record for domain of linux-kernel-owner@vger.kernel.org designates 209.132.180.67 as permitted sender) client-ip=209.132.180.67;
Date:   Wed, 16 May 2018 17:04:19 +0100
From:   Roman Gushchin <guro@fb.com>
To:     Mike Rapoport <rppt@linux.vnet.ibm.com>
CC:     Shuah Khan <shuah@kernel.org>, Tejun Heo <tj@kernel.org>,
        <linux-kselftest@vger.kernel.org>, <linux-kernel@vger.kernel.org>
Subject: Re: [PATCH] selftests: cgroup: memcontrol: add basic test for socket
 accounting
Message-ID: <20180516160243.GA31445@castle.DHCP.thefacebook.com>
References: <1526470089-27010-1-git-send-email-rppt@linux.vnet.ibm.com>
 <20180516140332.GA19754@castle.DHCP.thefacebook.com>
 <20180516144826.GB23563@rapoport-lnx>
MIME-Version: 1.0
Content-Type: text/plain; charset="us-ascii"
Content-Disposition: inline
In-Reply-To: <20180516144826.GB23563@rapoport-lnx>
User-Agent: Mutt/1.9.2 (2017-12-15)
Received-SPF: None (protection.outlook.com: fb.com does not designate
 permitted sender hosts)
SpamDiagnosticOutput: 1:99
SpamDiagnosticMetadata: NSPM
X-MS-Exchange-CrossTenant-OriginalArrivalTime: 16 May 2018 16:05:08.7351 (UTC)
X-MS-Exchange-CrossTenant-Network-Message-Id: 2d7921ec-b499-4a8d-da6f-08d5bb46ccba
X-MS-Exchange-CrossTenant-FromEntityHeader: Hosted
X-MS-Exchange-CrossTenant-Id: 8ae927fe-1255-47a7-a2af-5f3a069daaa2
X-MS-Exchange-Transport-CrossTenantHeadersStamped: CO1PR15MB1077
X-OriginatorOrg: fb.com
X-Proofpoint-Virus-Version: vendor=fsecure engine=2.50.10434:,, definitions=2018-05-16_08:,,
 signatures=0
X-Proofpoint-Spam-Reason: safe
X-FB-Internal: Safe
Sender: linux-kernel-owner@vger.kernel.org
Precedence: bulk
List-ID: <linux-kernel.vger.kernel.org>
X-Mailing-List: linux-kernel@vger.kernel.org

On Wed, May 16, 2018 at 05:48:26PM +0300, Mike Rapoport wrote:
> On Wed, May 16, 2018 at 03:04:08PM +0100, Roman Gushchin wrote:
> > Hi Mike!
> > 
> > On Wed, May 16, 2018 at 02:28:09PM +0300, Mike Rapoport wrote:
> > > The test verifies that with active TCP traffic memory.current and
> > > memory.stat.sock have similar values.
> > > 
> > > Signed-off-by: Mike Rapoport <rppt@linux.vnet.ibm.com>
> > > ---
> > >  tools/testing/selftests/cgroup/test_memcontrol.c | 184 +++++++++++++++++++++++
> > >  1 file changed, 184 insertions(+)
> > > 
> > > diff --git a/tools/testing/selftests/cgroup/test_memcontrol.c b/tools/testing/selftests/cgroup/test_memcontrol.c
> > > index beae06c9c899..0efdb1009175 100644
> > > --- a/tools/testing/selftests/cgroup/test_memcontrol.c
> > > +++ b/tools/testing/selftests/cgroup/test_memcontrol.c
> > > @@ -9,6 +9,12 @@
> > >  #include <sys/stat.h>
> > >  #include <sys/types.h>
> > >  #include <unistd.h>
> > > +#include <sys/socket.h>
> > > +#include <sys/wait.h>
> > > +#include <arpa/inet.h>
> > > +#include <netinet/in.h>
> > > +#include <netdb.h>
> > > +#include <errno.h>
> > >  
> > >  #include "../kselftest.h"
> > >  #include "cgroup_util.h"
> > > @@ -772,6 +778,183 @@ static int test_memcg_oom_events(const char *root)
> > >  	return ret;
> > >  }
> > >  
> > > +struct tcp_server_args {
> > > +	unsigned short port;
> > > +	int ctl[2];
> > > +};
> > > +
> > > +static int tcp_server(const char *cgroup, void *arg)
> > > +{
> > > +	struct tcp_server_args *srv_args = arg;
> > > +	struct sockaddr_in6 saddr = { 0 };
> > > +	socklen_t slen = sizeof(saddr);
> > > +	int sk, client_sk, ctl_fd, yes = 1, ret = -1;
> > > +
> > > +	close(srv_args->ctl[0]);
> > > +	ctl_fd = srv_args->ctl[1];
> > > +
> > > +	saddr.sin6_family = AF_INET6;
> > > +	saddr.sin6_addr = in6addr_any;
> > > +	saddr.sin6_port = htons(srv_args->port);
> > > +
> > > +	sk = socket(AF_INET6, SOCK_STREAM, 0);
> > > +	if (sk < 0)
> > > +		return ret;
> > > +
> > > +	if (setsockopt(sk, SOL_SOCKET, SO_REUSEADDR, &yes, sizeof(yes)) < 0)
> > > +		goto cleanup;
> > > +
> > > +	if (bind(sk, (struct sockaddr *)&saddr, slen)) {
> > > +		write(ctl_fd, &errno, sizeof(errno));
> > > +		goto cleanup;
> > > +	}
> > > +
> > > +	if (listen(sk, 1))
> > > +		goto cleanup;
> > > +
> > > +	ret = 0;
> > > +	if (write(ctl_fd, &ret, sizeof(ret)) != sizeof(ret)) {
> > > +		ret = -1;
> > > +		goto cleanup;
> > > +	}
> > > +
> > > +	client_sk = accept(sk, NULL, NULL);
> > > +	if (client_sk < 0)
> > > +		goto cleanup;
> > > +
> > > +	ret = -1;
> > > +	for (;;) {
> > > +		uint8_t buf[0x100000];
> > > +
> > > +		if (write(client_sk, buf, sizeof(buf)) <= 0) {
> > > +			if (errno == ECONNRESET)
> > > +				ret = 0;
> > > +			break;
> > > +		}
> > > +	}
> > > +
> > > +	close(client_sk);
> > > +
> > > +cleanup:
> > > +	close(sk);
> > > +	return ret;
> > > +}
> > > +
> > > +static int tcp_client(const char *cgroup, unsigned short port)
> > > +{
> > > +	const char server[] = "localhost";
> > > +	struct addrinfo *ai;
> > > +	char servport[6];
> > > +	int retries = 0x10; /* nice round number */
> > > +	int sk, ret;
> > > +
> > > +	snprintf(servport, sizeof(servport), "%hd", port);
> > > +	ret = getaddrinfo(server, servport, NULL, &ai);
> > > +	if (ret)
> > > +		return ret;
> > > +
> > > +	sk = socket(ai->ai_family, ai->ai_socktype, ai->ai_protocol);
> > > +	if (sk < 0)
> > > +		goto free_ainfo;
> > > +
> > > +	ret = connect(sk, ai->ai_addr, ai->ai_addrlen);
> > > +	if (ret < 0)
> > > +		goto close_sk;
> > > +
> > > +	ret = KSFT_FAIL;
> > > +	while (retries--) {
> > > +		uint8_t buf[0x100000];
> > > +		long current, sock;
> > > +
> > > +		if (read(sk, buf, sizeof(buf)) <= 0)
> > > +			goto close_sk;
> > > +
> > > +		current = cg_read_long(cgroup, "memory.current");
> > > +		if (current < 0)
> > > +			goto close_sk;
> > > +
> > > +		sock = cg_read_key_long(cgroup, "memory.stat", "sock ");
> > > +		if (sock < 0)
> > > +			goto close_sk;
> > > +
> > > +		if (values_close(current, sock, 3)) {
> > > +			ret = KSFT_PASS;
> > > +			break;
> > > +		}
> > 
> > The test is flapping (at least on my dev machine) because of this condition.
> > 
> > I believe it's because of the batching we're using on the page charge path.
> > So, in theory, it should be possible to calculate the maximum difference
> > like num_cpus * PAGE_SIZE * batch_size.
> 
> I afraid it's more complex and timing sensitive
>  
> > Alternatively, just bump allowed error percentage :)
> 
> so I'll bump the error percentage :)

Ok.
Please, add then a check that current >= sock.
It should be always true, right?

>  
> > > +	}
> > > +
> > > +close_sk:
> > > +	close(sk);
> > 
> > It would be great to check that sock and current are getting 0 values
> > after we're closing the socket.
>  
> Hmm, here it's also timing sensitive. I can see that sock is reliably
> getting 0 if I check it after the server exits. But current usually remains
> small but still !0.

Ok, then I'd check that sock == 0 and current >= 0.
That will catch some refcounter underrun issues we had in the past.

Other than that (and assuming the test isn't flapping):
Acked-by: Roman Gushchin <guro@fb.com>

Thanks!