diff --git a/u04-1/README.md b/u04-1/README.md
index 2c54536..904de75 100644
--- a/u04-1/README.md
+++ b/u04-1/README.md
@@ -1,3 +1,5 @@
+# Problem 4.1: Data exchange point between two threads
+
 Implement a class `Exchanger<T>` that is usable to exchange an object
 of type `T` between exactly two threads:
 
diff --git a/u05-3/README.md b/u05-3/README.md
new file mode 100644
index 0000000..711a103
--- /dev/null
+++ b/u05-3/README.md
@@ -0,0 +1,8 @@
+# Problem 5.3: Implement a semaphore in Java
+
+Implement a Java class `Semaphor` with methods `p()` and `v() `as decribed in the lecture. Your class
+should work for any number of threads using the same semaphore concurrently. Test your class using
+the class `DiningPhilosophersWithForkAccessSema` (on the server in the source code folder
+for chapter 4) by replacing all uses of `java.util.concurrent.Semaphore` by uses of your own
+`Semaphor` class – the program's behavior should still be correct (though not identical, since the
+output of the program is not determinate).
\ No newline at end of file
diff --git a/u05-4/README.md b/u05-4/README.md
new file mode 100644
index 0000000..fd2837a
--- /dev/null
+++ b/u05-4/README.md
@@ -0,0 +1,7 @@
+# Problem 5.4: Work queue of limited length using semaphores
+
+Implement the work queue of limited length from problem 4.2 using semaphores as the only
+synchronization mechanism, i.e. without `synchronized` blocks or methods (and, consequentially,
+not `wait()` or `notify()`). Test your program with your own semaphore class from problem 5.3 and
+with `java.util.concurrent.Semaphore`. Make sure not to use busy waiting: If a thread has
+to wait, use semaphore operations for this.
\ No newline at end of file
diff --git a/u06-1/README.md b/u06-1/README.md
new file mode 100644
index 0000000..0548bca
--- /dev/null
+++ b/u06-1/README.md
@@ -0,0 +1,21 @@
+# Problem 6.1: Barrier without semaphore
+
+Implement a class `Barrier` that realizes a barrier similar to the **`CyclicBarrier`** mentioned on
+slide 4-50 (your barrier does not have to be reusable); barriers are also described in the book by
+Gleim/Schüle in section 3.2.7 (in German). Don't use semaphores, but only `synchronized`, `wait()`
+and `notify/All()`. Your class should look about as follows:
+
+```java
+class Barrier {
+	// …
+	public Barrier(int nThreads){ … }
+		// nThreads: Number of threads using this barrier
+	public int await() throws InterruptedException { … }
+		// Cross the barrier, possibly waiting before doing this
+		// Return value: Position of this thread when arriving at the barrier
+}
+```
+
+Test your barrier with class `BarrierTest` in the source code folder on th server. In the program's
+output, all `a()` invocations must appear before all `b()` invocations `Barrier` if your barrier is
+correct.
\ No newline at end of file
diff --git a/u06-3/README.md b/u06-3/README.md
new file mode 100644
index 0000000..395a51e
--- /dev/null
+++ b/u06-3/README.md
@@ -0,0 +1,9 @@
+# Problem 6.3: Dining philosophers with globally uniform acquisition order
+
+As mentioned in the lecture, one possible strategy to avoid a deadlock is to acquire all resources in the
+same, globally uniform order, so that it cannot happen that e.g. thread A first acquires resource 1 and
+then resource 2, while thread B acquires the same resouces in a different order (first 2 and then 1).
+
+Implement a deadlock-free solution to the "dining philosophers" problem from the lecture using this
+strategy by defining a globally uniform acquisition order for the fork semaphores; an additional
+semaphore as in the lecture will no longer be necessary then.
\ No newline at end of file
diff --git a/u06-4/README.md b/u06-4/README.md
new file mode 100644
index 0000000..496d018
--- /dev/null
+++ b/u06-4/README.md
@@ -0,0 +1,18 @@
+# Problem 6.4: Implement AtomicXXX.updateAndGet
+
+On slide 4-46 two implementations for the modification of an `AtomicXXX` object were presented: A
+loop with a `compareAndSet()` operation and the more abstract and more general
+`updateAndGet()` operation. Implement your own `updateAndGet()` method for the
+`AtomicCounter` class on slide 4-46. Use `compareAndSet()` for this. Use the following interface
+as the parameter type for the lambda expression:
+
+```java
+interface IntUnaryOperator {
+	int applyAsInt(int x);
+}
+```
+
+What condition must hold for the behavior of the code passed to `updateAndGet()`?
+
+Test your class with the class `AtomicCounterTest` to be found in the source code folder on he
+server. Its output must be `100000000`.
\ No newline at end of file
diff --git a/u06-5/README.md b/u06-5/README.md
new file mode 100644
index 0000000..7c4b0ff
--- /dev/null
+++ b/u06-5/README.md
@@ -0,0 +1,20 @@
+# Problem 6.5: Implement a nonblocking spin lock
+
+Spin locks are often used in code close to the hardware (e.g. in operating system kernels) to lock short
+critical sections where the overhead of blocking and later unblocking the thread would cost more time
+than the thread normally has to wait until entering the section. Instead of that, a spin lock contains a
+busy waitig loop that repeats ("spins round and round") until it succeeds to acquire the lock.
+
+Implement a spin lock as a Java class with the following structure and test it using the `SpinLockTest`
+class in the source code folder on the server:
+
+```java
+public class SpinLock {
+	// ...
+	public void lock() {
+		// ...
+	}
+	public boolean unlock() { // returns whether the lock had indeed been acquired by current thread
+		// ...
+	}
+}
\ No newline at end of file
diff --git a/u07-1/README.md b/u07-1/README.md
new file mode 100644
index 0000000..b0605d7
--- /dev/null
+++ b/u07-1/README.md
@@ -0,0 +1,17 @@
+# Problem 7.1: Parallel prime number test using tasks and `Executor`
+
+Implement a simple prime number test, first sequentially and then using concurrent tasks by means of
+the Java `Executor` framework. Your program should test all numbers *n* between 2 and 100.000 for
+primality by simply checking all numbers *d* from 2 to *n* whether *d* divides *n* without a remainder. (Of
+course there are much more efficient primality tests, but the goal here is not efficient primality tests,
+but simply generating some computation load.) As its result, your program should print the number of
+prime numbers found by all tasks (the correct result is 9592).
+
+Divide the work into 100 tasks, each of them testing the next 1000 numbers. To simulate occasional
+blocking due to I/O activity, make each task invoke `Thread.sleep()` after checking half of its
+numbers, with the sleep time chosen randomly between 1 and 100 ms (the sample solution to problem
+4.2 shows an example for this).
+
+Use a `newFixedThreadPool()`, initialized with the number of available core (obtainable using
+`Runtime.getRuntime().availableProcessors()`). Experiment with a higher number of
+threads or with a `newCachedThreadPool`, too.
\ No newline at end of file