Spin locks in linux

Locking lessons¶

Lesson 1: Spin locks¶

The most basic primitive for locking is spinlock:

The above is always safe. It will disable interrupts _locally_, but the spinlock itself will guarantee the global lock, so it will guarantee that there is only one thread-of-control within the region(s) protected by that lock. This works well even under UP also, so the code does _not_ need to worry about UP vs SMP issues: the spinlocks work correctly under both.

NOTE! Implications of spin_locks for memory are further described in:

The above is usually pretty simple (you usually need and want only one spinlock for most things — using more than one spinlock can make things a lot more complex and even slower and is usually worth it only for sequences that you know need to be split up: avoid it at all cost if you aren’t sure).

This is really the only really hard part about spinlocks: once you start using spinlocks they tend to expand to areas you might not have noticed before, because you have to make sure the spinlocks correctly protect the shared data structures everywhere they are used. The spinlocks are most easily added to places that are completely independent of other code (for example, internal driver data structures that nobody else ever touches).

NOTE! The spin-lock is safe only when you also use the lock itself to do locking across CPU’s, which implies that EVERYTHING that touches a shared variable has to agree about the spinlock they want to use.

Lesson 2: reader-writer spinlocks.В¶

If your data accesses have a very natural pattern where you usually tend to mostly read from the shared variables, the reader-writer locks (rw_lock) versions of the spinlocks are sometimes useful. They allow multiple readers to be in the same critical region at once, but if somebody wants to change the variables it has to get an exclusive write lock.

NOTE! reader-writer locks require more atomic memory operations than simple spinlocks. Unless the reader critical section is long, you are better off just using spinlocks.

The routines look the same as above:

The above kind of lock may be useful for complex data structures like linked lists, especially searching for entries without changing the list itself. The read lock allows many concurrent readers. Anything that changes the list will have to get the write lock.

NOTE! RCU is better for list traversal, but requires careful attention to design detail (see Using RCU to Protect Read-Mostly Linked Lists ).

Also, you cannot “upgrade” a read-lock to a write-lock, so if you at _any_ time need to do any changes (even if you don’t do it every time), you have to get the write-lock at the very beginning.

NOTE! We are working hard to remove reader-writer spinlocks in most cases, so please don’t add a new one without consensus. (Instead, see RCU Concepts for complete information.)

Lesson 3: spinlocks revisited.В¶

The single spin-lock primitives above are by no means the only ones. They are the most safe ones, and the ones that work under all circumstances, but partly because they are safe they are also fairly slow. They are slower than they’d need to be, because they do have to disable interrupts (which is just a single instruction on a x86, but it’s an expensive one — and on other architectures it can be worse).

If you have a case where you have to protect a data structure across several CPU’s and you want to use spinlocks you can potentially use cheaper versions of the spinlocks. IFF you know that the spinlocks are never used in interrupt handlers, you can use the non-irq versions:

(and the equivalent read-write versions too, of course). The spinlock will guarantee the same kind of exclusive access, and it will be much faster. This is useful if you know that the data in question is only ever manipulated from a “process context”, ie no interrupts involved.

The reasons you mustn’t use these versions if you have interrupts that play with the spinlock is that you can get deadlocks:

where an interrupt tries to lock an already locked variable. This is ok if the other interrupt happens on another CPU, but it is _not_ ok if the interrupt happens on the same CPU that already holds the lock, because the lock will obviously never be released (because the interrupt is waiting for the lock, and the lock-holder is interrupted by the interrupt and will not continue until the interrupt has been processed).

(This is also the reason why the irq-versions of the spinlocks only need to disable the _local_ interrupts — it’s ok to use spinlocks in interrupts on other CPU’s, because an interrupt on another CPU doesn’t interrupt the CPU that holds the lock, so the lock-holder can continue and eventually releases the lock).

Reference information:В¶

For dynamic initialization, use spin_lock_init() or rwlock_init() as appropriate:

For static initialization, use DEFINE_SPINLOCK() / DEFINE_RWLOCK() or __SPIN_LOCK_UNLOCKED() / __RW_LOCK_UNLOCKED() as appropriate.

© Copyright The kernel development community.

Источник

Synchronization primitives in the Linux kernel. Part 1.

Introduction

This part opens a new chapter in the linux-insides book. Timers and time management related stuff was described in the previous chapter. Now it’s time to move on to the next topic. As you probably recognized from the title, this chapter will describe the synchronization primitives in the Linux kernel.

As always, we will try to know what a synchronization primitive in general is before we deal with any synchronization-related issues. Actually, a synchronization primitive is a software mechanism, that ensures that two or more parallel processes or threads are not running simultaneously on the same code segment. For example, let’s look at the following piece of code:

from the kernel/time/clocksource.c source code file. This code is from the __clocksource_register_scale function which adds the given clocksource to the clock sources list. This function produces different operations on a list with registered clock sources. For example, the clocksource_enqueue function adds the given clock source to the list with registered clocksources — clocksource_list . Note that these lines of code wrapped to two functions: mutex_lock and mutex_unlock which takes one parameter — the clocksource_mutex in our case.

These functions represent locking and unlocking based on mutex synchronization primitive. As mutex_lock will be executed, it allows us to prevent the situation when two or more threads will execute this code while the mutex_unlock will not be executed by process-owner of the mutex. In other words, we prevent parallel operations on a clocksource_list . Why do we need mutex here? What if two parallel processes will try to register a clock source. As we already know, the clocksource_enqueue function adds the given clock source to the clocksource_list list right after a clock source in the list which has the biggest rating (a registered clock source which has the highest frequency in the system):

If two parallel processes will try to do it simultaneously, both process may found the same entry may occur race condition or in other words, the second process which will execute list_add , will overwrite a clock source from the first thread.

Besides this simple example, synchronization primitives are ubiquitous in the Linux kernel. If we will go through the previous chapter or other chapters again or if we will look at the Linux kernel source code in general, we will meet many places like this. We will not consider how mutex is implemented in the Linux kernel. Actually, the Linux kernel provides a set of different synchronization primitives like:

• mutex ;
• semaphores ;
• seqlocks ;
• atomic operations ;
• etc.

We will start this chapter from the spinlock .

Spinlocks in the Linux kernel.

The spinlock is a low-level synchronization mechanism which in simple words, represents a variable which can be in two states:

Each process which wants to acquire a spinlock , must write a value which represents spinlock acquired state to this variable and write spinlock released state to the variable. If a process tries to execute code which is protected by a spinlock , it will be locked while a process which holds this lock will release it. In this case all related operations must be atomic to prevent race conditions state. The spinlock is represented by the spinlock_t type in the Linux kernel. If we will look at the Linux kernel code, we will see that this type is widely used. The spinlock_t is defined as:

and located in the include/linux/spinlock_types.h header file. We may see that its implementation depends on the state of the CONFIG_DEBUG_LOCK_ALLOC kernel configuration option. We will skip this now, because all debugging related stuff will be in the end of this part. So, if the CONFIG_DEBUG_LOCK_ALLOC kernel configuration option is disabled, the spinlock_t contains union with one field which is — raw_spinlock :

The raw_spinlock structure defined in the same header file represents the implementation of normal spinlock. Let’s look how the raw_spinlock structure is defined:

where the arch_spinlock_t represents architecture-specific spinlock implementation. As we mentioned above, we will skip debugging kernel configuration options. As we focus on x86_64 architecture in this book, the arch_spinlock_t that we will consider is defined in the include/asm-generic/qspinlock_types.h header file and looks:

We will not stop on this structures for now. Let’s look at the operations on a spinlock . The Linux kernel provides following main operations on a spinlock :

• spin_lock_init — produces initialization of the given spinlock ;
• spin_lock — acquires given spinlock ;
• spin_lock_bh — disables software interrupts and acquire given spinlock ;
• spin_lock_irqsave and spin_lock_irq — disable interrupts on local processor, preserve/not preserve previous interrupt state in the flags and acquire given spinlock ;
• spin_unlock — releases given spinlock ;
• spin_unlock_bh — releases given spinlock and enables software interrupts;
• spin_is_locked — returns the state of the given spinlock ;
• and etc.

Let’s look on the implementation of the spin_lock_init macro. As I already wrote, this and other macro are defined in the include/linux/spinlock.h header file and the spin_lock_init macro looks:

As we may see, the spin_lock_init macro takes a spinlock and executes two operations: check the given spinlock and execute the raw_spin_lock_init . The implementation of the spinlock_check is pretty easy, this function just returns the raw_spinlock_t of the given spinlock to be sure that we got exactly normal raw spinlock:

The raw_spin_lock_init macro:

assigns the value of the __RAW_SPIN_LOCK_UNLOCKED with the given spinlock to the given raw_spinlock_t . As we may understand from the name of the __RAW_SPIN_LOCK_UNLOCKED macro, this macro does initialization of the given spinlock and set it to released state. This macro is defined in the include/linux/spinlock_types.h header file and expands to the following macros:

As I already wrote above, we will not consider stuff which is related to debugging of synchronization primitives. In this case we will not consider the SPIN_DEBUG_INIT and the SPIN_DEP_MAP_INIT macros. So the __RAW_SPINLOCK_UNLOCKED macro will be expanded to the:

where the __ARCH_SPIN_LOCK_UNLOCKED is:

for the x86_64 architecture. So, after the expansion of the spin_lock_init macro, a given spinlock will be initialized and its state will be — unlocked .

From this moment we know how to initialize a spinlock , now let’s consider API which Linux kernel provides for manipulations of spinlocks . The first is:

function which allows us to acquire a spinlock . The raw_spin_lock macro is defined in the same header file and expands to the call of _raw_spin_lock :

Where _raw_spin_lock is defined depends on whether CONFIG_SMP option is set and CONFIG_INLINE_SPIN_LOCK option is set. If the SMP is disabled, _raw_spin_lock is defined in the include/linux/spinlock_api_up.h header file as a macro and looks like:

If the SMP is enabled and CONFIG_INLINE_SPIN_LOCK is set, it is defined in include/linux/spinlock_api_smp.h header file as the following:

If the SMP is enabled and CONFIG_INLINE_SPIN_LOCK is not set, it is defined in kernel/locking/spinlock.c source code file as the following:

Here we will consider the latter form of _raw_spin_lock . The __raw_spin_lock function looks:

As you may see, first of all we disable preemption by the call of the preempt_disable macro from the include/linux/preempt.h (more about this you may read in the ninth part of the Linux kernel initialization process chapter). When we unlock the given spinlock , preemption will be enabled again:

We need to do this to prevent the process from other processes to preempt it while it is spinning on a lock. The spin_acquire macro which through a chain of other macros expands to the call of the:

The lock_acquire function:

As I wrote above, we will not consider stuff here which is related to debugging or tracing. The main point of the lock_acquire function is to disable hardware interrupts by the call of the raw_local_irq_save macro, because the given spinlock might be acquired with enabled hardware interrupts. In this way the process will not be preempted. Note that in the end of the lock_acquire function we will enable hardware interrupts again with the help of the raw_local_irq_restore macro. As you already may guess, the main work will be in the __lock_acquire function which is defined in the kernel/locking/lockdep.c source code file.

The __lock_acquire function looks big. We will try to understand what this function does, but not in this part. Actually this function is mostly related to the Linux kernel lock validator and it is not topic of this part. If we will return to the definition of the __raw_spin_lock function, we will see that it contains the following definition in the end:

The LOCK_CONTENDED macro is defined in the include/linux/lockdep.h header file and just calls the given function with the given spinlock :

In our case, the lock is do_raw_spin_lock function from the include/linux/spinlock.h header file and the _lock is the given raw_spinlock_t :

The __acquire here is just Sparse related macro and we are not interested in it in this moment. The arch_spin_lock macro is defined in the include/asm-generic/qspinlock.h header file as the following:

We stop here for this part. In the next part, we’ll dive into how queued spinlocks works and related concepts.

Conclusion

This concludes the first part covering synchronization primitives in the Linux kernel. In this part, we met first synchronization primitive spinlock provided by the Linux kernel. In the next part we will continue to dive into this interesting theme and will see other synchronization related stuff.

If you have questions or suggestions, feel free to ping me in twitter 0xAX, drop me email or just create issue.

Please note that English is not my first language and I am really sorry for any inconvenience. If you found any mistakes please send me PR to linux-insides.

Источник