Skip to contentSkip to conversation

recvmsg03: avoid close() and recvmsg() racing

PATCH New Jan Stancek 3 participants 4 messages Jun 14, 2017

lore
Share & tools
lore
#0 lore raw

PATCH 1/1 recvmsg03: avoid close() and recvmsg() racing

From: Sabrina Dubroca <[email protected]>

This testcase rarely hangs on single CPU systems. Problem appears to be between recvmsg() and close():
1. sendmsg(), packet gets queued
2. close(), packet is dropped
3. recvmsg(), nothing to fetch

strace of a buggy run:
[pid 17284] 16:04:41.042617 socket(AF_RDS, SOCK_SEQPACKET, 0) = 3
[pid 17284] 16:04:41.049549 close(3) = 0
strace: Process 17290 attached
[pid 17284] 16:04:41.050765 socket(AF_RDS, SOCK_SEQPACKET, 0 <unfinished ...>
[pid 17290] 16:04:41.050894 futex(0x7f95dfc57014, FUTEX_WAIT, 0, {tv_sec=10, tv_nsec=0} <unfinished ...>
[pid 17284] 16:04:41.051001 <... socket resumed> ) = 3
[pid 17284] 16:04:41.051111 bind(3, {sa_family=AF_INET, sin_port=htons(4000), sin_addr=inet_addr("127.0.0.1")}, 16) = 0
[pid 17284] 16:04:41.051374 futex(0x7f95dfc57014, FUTEX_WAKE, 2147483647) = 1
[pid 17290] 16:04:41.051432 <... futex resumed> ) = 0
[pid 17284] 16:04:41.051442 recvmsg(3, <unfinished ...>
[pid 17290] 16:04:41.051455 socket(AF_RDS, SOCK_SEQPACKET, 0) = 3
[pid 17290] 16:04:41.051666 bind(3, {sa_family=AF_INET, sin_port=htons(4001), sin_addr=inet_addr("127.0.0.1")}, 16) = 0
[pid 17290] 16:04:41.051724 sendmsg(3, {msg_name={sa_family=AF_INET, sin_port=htons(4000), sin_addr=inet_addr("127.0.0.1")}, msg_namelen=16, msg_iov=[{iov_base="hello world\0", iov_len=12}], msg_iovlen=1, msg_controllen=0, msg_flags=0}, 0) = 12
[pid 17290] 16:04:41.051975 close(3) = 0
[pid 17290] 16:04:41.052492 +++ exited with 0 +++
[pid 17284] 16:04:41.052598 <... recvmsg resumed> {msg_namelen=32}, 0) = ? ERESTARTSYS (To be restarted if SA_RESTART is set)
[pid 17284] 16:04:41.052631 --- SIGCHLD {si_signo=SIGCHLD, si_code=CLD_EXITED, si_pid=17290, si_uid=0, si_status=0, si_utime=0, si_stime=0} ---
[pid 17284] 16:04:41.052718 recvmsg(3, // hang

[ wrote changelog message ]
---
 testcases/kernel/syscalls/recvmsg/recvmsg03.c | 4 ++++
 1 file changed, 4 insertions(+)
testcases/kernel/syscalls/recvmsg/recvmsg03.c +4 -0
index 31e7585b66ab..8bbdaceeb464 100644
@@ -90,6 +90,8 @@ static void client(void)
 			"sendmsg() failed to send data to server");
 	}
 
+	TST_CHECKPOINT_WAIT(1);
+
 	SAFE_CLOSE(sock_fd1);
 }
 
@@ -137,6 +139,8 @@ static void server(void)
 			msg.msg_namelen);
 	}
 
+	TST_CHECKPOINT_WAKE(1);
+
 	SAFE_CLOSE(sock_fd2);
 }
 
Signature · 1 line
1.8.3.1
#1 lore raw

PATCH 1/1 recvmsg03: avoid close() and recvmsg() racing

Hi Jan,

Thanks for this fix, it looks good to me. :-)

Thanks,
Xiao Yang.

Quoted 57 lines · Jan Stancek

From: Sabrina Dubroca<[email protected]>

This testcase rarely hangs on single CPU systems. Problem appears to be between recvmsg() and close():
1. sendmsg(), packet gets queued
2. close(), packet is dropped
3. recvmsg(), nothing to fetch

strace of a buggy run:

   [pid 17284] 16:04:41.042617 socket(AF_RDS, SOCK_SEQPACKET, 0) = 3
   [pid 17284] 16:04:41.049549 close(3)    = 0
   strace: Process 17290 attached
   [pid 17284] 16:04:41.050765 socket(AF_RDS, SOCK_SEQPACKET, 0<unfinished ...>
   [pid 17290] 16:04:41.050894 futex(0x7f95dfc57014, FUTEX_WAIT, 0, {tv_sec=10, tv_nsec=0}<unfinished ...>
   [pid 17284] 16:04:41.051001<... socket resumed>  ) = 3
   [pid 17284] 16:04:41.051111 bind(3, {sa_family=AF_INET, sin_port=htons(4000), sin_addr=inet_addr("127.0.0.1")}, 16) = 0
   [pid 17284] 16:04:41.051374 futex(0x7f95dfc57014, FUTEX_WAKE, 2147483647) = 1
   [pid 17290] 16:04:41.051432<... futex resumed>  ) = 0
   [pid 17284] 16:04:41.051442 recvmsg(3,<unfinished ...>
   [pid 17290] 16:04:41.051455 socket(AF_RDS, SOCK_SEQPACKET, 0) = 3
   [pid 17290] 16:04:41.051666 bind(3, {sa_family=AF_INET, sin_port=htons(4001), sin_addr=inet_addr("127.0.0.1")}, 16) = 0
   [pid 17290] 16:04:41.051724 sendmsg(3, {msg_name={sa_family=AF_INET, sin_port=htons(4000), sin_addr=inet_addr("127.0.0.1")}, msg_namelen=16, msg_iov=[{iov_base="hello world\0", iov_len=12}], msg_iovlen=1, msg_controllen=0, msg_flags=0}, 0) = 12
   [pid 17290] 16:04:41.051975 close(3)    = 0
   [pid 17290] 16:04:41.052492 +++ exited with 0 +++
   [pid 17284] 16:04:41.052598<... recvmsg resumed>  {msg_namelen=32}, 0) = ? ERESTARTSYS (To be restarted if SA_RESTART is set)
   [pid 17284] 16:04:41.052631 --- SIGCHLD {si_signo=SIGCHLD, si_code=CLD_EXITED, si_pid=17290, si_uid=0, si_status=0, si_utime=0, si_stime=0} ---
   [pid 17284] 16:04:41.052718 recvmsg(3, // hang
[ wrote changelog message ]
---
  testcases/kernel/syscalls/recvmsg/recvmsg03.c | 4 ++++
  1 file changed, 4 insertions(+)
testcases/kernel/syscalls/recvmsg/recvmsg03.c +4 -0
index 31e7585b66ab..8bbdaceeb464 100644
@@ -90,6 +90,8 @@ static void client(void)
  			"sendmsg() failed to send data to server");
  	}
 
+	TST_CHECKPOINT_WAIT(1);
+
  	SAFE_CLOSE(sock_fd1);
  }
 
@@ -137,6 +139,8 @@ static void server(void)
  			msg.msg_namelen);
  	}
 
+	TST_CHECKPOINT_WAKE(1);
+
  	SAFE_CLOSE(sock_fd2);
  }
 
#3 lore raw

PATCH 1/1 recvmsg03: avoid close() and recvmsg() racing

----- Original Message -----

Quoted 2 lines

Hi!
Looks good.

Pushed.

Regards,
Jan