[LTP] [PATCH] sendmsg02: fix parameter passed to semctl(SETVAL)

[LTP] [PATCH] sendmsg02: fix parameter passed to semctl(SETVAL)

[Jan Stancek]

The fourth has the type union semun.

Signed-off-by: Jan Stancek <jstancek@redhat.com>
---
 testcases/kernel/syscalls/sendmsg/sendmsg02.c |   13 +++++++++++--
 1 files changed, 11 insertions(+), 2 deletions(-)

diff --git a/testcases/kernel/syscalls/sendmsg/sendmsg02.c b/testcases/kernel/syscalls/sendmsg/sendmsg02.c
index 669af20..d96e10c 100644
--- a/testcases/kernel/syscalls/sendmsg/sendmsg02.c
+++ b/testcases/kernel/syscalls/sendmsg/sendmsg02.c
@@ -46,6 +46,12 @@
 #include "usctest.h"
 #include "safe_macros.h"
 
+union semun {
+	int val;
+	struct semid_ds *buf;
+	unsigned short int *array;
+};
+
 char *TCID = "sendmsg02";
 
 static int sem_id;
@@ -126,10 +132,12 @@ static void reproduce(int seconds)
 	int child_count = 0;
 	int *child_pids;
 	int child_pid;
+	union semun u;
 
 	child_pids = SAFE_MALLOC(cleanup, sizeof(int) * child_pairs * 2);
 
-	if (semctl(sem_id, 0, SETVAL, 1) == -1)
+	u.val = 1;
+	if (semctl(sem_id, 0, SETVAL, u) == -1)
 		tst_brkm(TBROK | TERRNO, cleanup, "couldn't set semval to 1");
 
 	/* fork child for each client/server pair */
@@ -168,7 +176,8 @@ static void reproduce(int seconds)
 	if (child_count == child_pairs*2)
 		sleep(seconds);
 
-	if (semctl(sem_id, 0, SETVAL, 0) == -1) {
+	u.val = 0;
+	if (semctl(sem_id, 0, SETVAL, u) == -1) {
 		/* kill children if setting semval failed */
 		for (i = 0; i < child_count; i++)
 			kill(child_pids[i], SIGKILL);
-- 
1.7.1


------------------------------------------------------------------------------
Slashdot TV.  
Video for Nerds.  Stuff that matters.
http://tv.slashdot.org/
_______________________________________________
Ltp-list mailing list
Ltp-list@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/ltp-list

Re: [LTP] [PATCH] sendmsg02: fix parameter passed to semctl(SETVAL)

[chrubis]

Hi!
> The fourth has the type union semun.

The change looks good. I guess that this may have been failing on some
big endian machine where sizeof(void*) != sizeof(int), am I right?

-- 
Cyril Hrubis
chrubis@suse.cz

------------------------------------------------------------------------------
Slashdot TV.  
Video for Nerds.  Stuff that matters.
http://tv.slashdot.org/
_______________________________________________
Ltp-list mailing list
Ltp-list@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/ltp-list

Re: [LTP] [PATCH] sendmsg02: fix parameter passed to semctl(SETVAL)

[Jan Stancek]

----- Original Message -----
> From: "Jan Stancek" <jstancek@redhat.com>
> To: chrubis@suse.cz
> Cc: ltp-list@lists.sourceforge.net
> Sent: Monday, 25 August, 2014 3:12:52 PM
> Subject: Re: [LTP] [PATCH] sendmsg02: fix parameter passed to semctl(SETVAL)
> 
> 
> 
> 
> 
> ----- Original Message -----
> > From: chrubis@suse.cz
> > To: "Jan Stancek" <jstancek@redhat.com>
> > Cc: ltp-list@lists.sourceforge.net
> > Sent: Monday, 25 August, 2014 2:48:38 PM
> > Subject: Re: [LTP] [PATCH] sendmsg02: fix parameter passed to
> > semctl(SETVAL)
> > 
> > Hi!
> > > The fourth has the type union semun.
> > 
> > The change looks good. I guess that this may have been failing on some
> > big endian machine where sizeof(void*) != sizeof(int), am I right?
> 
> You're right it was big endian machine, but with sizeof(void *) == 4 and
> sizeof(int) == 4.
> 
> Fourth parameter is expected to be passed by reference,
> so it was actually crashing on RHEL5.10 ppc:
> 
> void func()
> {
> 	union semun a;
> 	a.val = 0;
> 	semctl(123, 456, SETVAL, a);
> }
> 
> (gdb) disassemble func
> Dump of assembler code for function func:
> 0x1000047c <func+0>:	stwu    r1,-48(r1)
> 0x10000480 <func+4>:	mflr    r0
> 0x10000484 <func+8>:	stw     r31,44(r1)
> 0x10000488 <func+12>:	stw     r0,52(r1)
> 0x1000048c <func+16>:	mr      r31,r1
> 0x10000490 <func+20>:	li      r0,0
> 0x10000494 <func+24>:	stw     r0,8(r31)
> 0x10000498 <func+28>:	lwz     r0,8(r31)
> 0x1000049c <func+32>:	stw     r0,24(r31)
> 0x100004a0 <func+36>:	addi    r0,r31,24
> 0x100004a4 <func+40>:	li      r3,123
> 0x100004a8 <func+44>:	li      r4,456
> 0x100004ac <func+48>:	li      r5,16
> 0x100004b0 <func+52>:	mr      r6,r0
> 0x100004b4 <func+56>:	crclr   4*cr1+eq
> 0x100004b8 <func+60>:	bl      0x10000870 <semctl@plt>
> 0x100004bc <func+64>:	lwz     r11,0(r1)
> 0x100004c0 <func+68>:	lwz     r0,4(r11)
> 0x100004c4 <func+72>:	mtlr    r0
> 0x100004c8 <func+76>:	lwz     r31,-4(r11)
> 0x100004cc <func+80>:	mr      r1,r11
> 0x100004d0 <func+84>:	blr
> End of assembler dump.
> 
> (gdb) b *0x100004b8
> Breakpoint 1 at 0x100004b8: file a.c, line 18.
> 
> (gdb) r
> Starting program:
> /mnt/tests/scratch/jstancek2/ltp-full-20140820/testcases/kernel/syscalls/sendmsg/a.out
> 
> Breakpoint 1, 0x100004b8 in func () at a.c:18
> 18		semctl(123, 456, SETVAL, a);
> 
> (gdb) p &a
> $1 = (union semun *) 0xffacf488
> 
> (gdb) p/x $r6
> $2 = 0xffacf498

Patch pushed.

Regards,
Jan

> 
> ------------------------------------------------------------------------------
> Slashdot TV.
> Video for Nerds.  Stuff that matters.
> http://tv.slashdot.org/
> _______________________________________________
> Ltp-list mailing list
> Ltp-list@lists.sourceforge.net
> https://lists.sourceforge.net/lists/listinfo/ltp-list
> 

------------------------------------------------------------------------------
Slashdot TV.  
Video for Nerds.  Stuff that matters.
http://tv.slashdot.org/
_______________________________________________
Ltp-list mailing list
Ltp-list@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/ltp-list