Understanding Linux kernel semaphore - linux-kernel

Question

My understanding of semaphore is:
count cannot go below 0 : if tried block the task.
count cannot go above the value given during sema_init().
First case is always honoured. However as you can see from below code, I can increase the count of semaphore beyond the given value, I was expecting down_interruptible() get blocked and up() to fail to increment the count.
# uname -r
5.4.0-132-generic
# tree
.
├── Makefile
├── sm_define.c
Makefile:
obj-m := sm_define.o
KVERS := $(shell uname -r)
all:
make -C /lib/modules/$(KVERS)/build M=$(PWD) modules
clean:
make -C /lib/modules/$(KVERS)/build M=$(PWD) clean
sm_define.c
#include <linux/module.h>
#include <linux/init.h>
#include <linux/semaphore.h>
struct semaphore sem;
static int __init
start_module(void)
{
int ret;
sema_init(&sem, 1); /* Binary semaphore */
up(&sem);
up(&sem);
up(&sem);
printk(KERN_INFO "count %d\n", sem.count);
ret = down_interruptible(&sem);
if (ret) {
printk(KERN_INFO "down_interruptible : failed\n");
return -EINVAL;
}
printk(KERN_INFO "count %d\n", sem.count);
ret = down_interruptible(&sem);
if (ret) {
printk(KERN_INFO "down_interruptible : failed\n");
return -EINVAL;
}
printk(KERN_INFO "count %d\n", sem.count);
ret = down_interruptible(&sem);
if (ret) {
printk(KERN_INFO "down_interruptible : failed\n");
return -EINVAL;
}
printk(KERN_INFO "count %d\n", sem.count);
return 0;
}
static void __exit
end_module(void)
{
printk(KERN_INFO "Module exited.\n");
return;
}
module_init(start_module);
module_exit(end_module);
MODULE_LICENSE("GPL");
Output:
# insmod sm_define.ko
# rmmod sm_define
# dmesg
[ 8492.896319] count 4
[ 8492.896320] count 3
[ 8492.896320] count 2
[ 8492.896320] count 1
[ 8500.225030] Module exited.
#
Why this behaviour? and if it is allowed and can't we say notion of binary / counting semaphore is a myth, its just the self followed rule during code, that "I will never increment the value of semaphore before decrementing it?