Platform

1. Introduction In a multitasking environment, multiple processes and threads may need to access shared resources concurrently. Without proper synchronization, race conditions, deadlocks, and data corruption can occur. The Linux kernel provides various synchronization primitives to ensure safe concurrent access while maintaining performance. 2. Spinlocks Spinlocks are busy-waiting locks used in scenarios where critical sections are short and must be protected from concurrent access. Key Features: Suitable for short, critical sections. Does not sleep, making it ideal for use in interrupt handlers. If contention occurs, the CPU spins in a loop until the lock is available. Usage: spinlock_t my_lock; spin_lock_init(&my_lock); spin_lock(&my_lock); /* Critical section */ spin_unlock(&my_lock); Types of Spinlocks: ...

Monolithic Kernel All core OS services (memory management, process scheduling, file systems, drivers) reside in kernel space. Example: Linux Kernel. Pros: Fast performance, directaccess to hardware. Cons: Large codebase, difficult debugging, crashes can affect the whole system. Microkernel Minimal core kernel, with most services running in user space. Example: QNX, Minix. Pros: Stability, modularity, better security. Cons: Performance overhead due to inter-process communication (IPC).

Overview Kernel Space: This is where the Linux kernel executes and provides low-level access to hardware, system memory management, process scheduling, and device drivers. Kernel space has privileged access to system resources and is protected from direct user interference. For example, when a user requests data from a hardware sensor, the kernel driver handles communication with the hardware, processes the request, and returns the data to user space through system calls. User Space: This is where applications and system utilities run. User-space processes operate with restricted privileges and interact with the kernel via system calls, libraries, and IPC mechanisms. For example, a user-space daemon may monitor the watchdog status by writing to /dev/watchdog, or a mobile app may read light intensity from /sys/bus/i2c/devices/1-0039/lux. Communication Methods between Kernel and User Space There are several ways to facilitate communication between user space and kernel space in an embedded Linux environment: ...

1. Overview RTOS (Real-Time Operating System): Designed for deterministic, time-critical applications with low-latency response. Why RTOS over Linux? Deterministic Execution: RTOS ensures tasks meet strict timing deadlines, unlike Linux, which has non-deterministic scheduling. Low Overhead: RTOS has minimal context switching overhead and no user/kernel space separation. Resource-Constrained Devices: Ideal for microcontrollers (MCUs) with limited memory and processing power. Fast Boot Times: RTOS boots in milliseconds, while Linux requires a much longer initialization process. Interrupt Handling: More responsive to real-time interrupts, whereas Linux introduces latency due to its complex scheduler. FreeRTOS: A lightweight, open-source RTOS widely used in embedded systems. Linux Kernel: A general-purpose OS with multi-user capabilities, used in complex embedded and desktop/server systems. 2. FreeRTOS vs. Linux Kernel (Key Differences) Kernel vs. User Space FreeRTOS: It doesn’t have the concept of a user space and kernel space like Linux. The whole system is essentially one space, and tasks directly interact with the kernel (RTOS). You can think of FreeRTOS as a single program running with different tasks that can interact with each other or with hardware directly. Linux Kernel: Linux operates with a strict separation between user space and kernel space. User applications cannot directly interact with hardware; they must go through system calls, which are handled by the kernel. Scheduler FreeRTOS: Preemptive, cooperative, or tickless scheduling. Supports priority-based scheduling (fixed priority, round-robin, etc.). Simple task model, each task runs in its own stack but shares memory. Linux Kernel Also has a preemptive scheduler, but it is much more complex, as it must handle multiple users, system calls, different types of scheduling (e.g., real-time, normal tasks), and various priorities. Linux is optimized for fairness CFS (Completely Fair Scheduler) and general-purpose multitasking. The FreeRTOS scheduler, by contrast, is simpler and more deterministic. Processes FreeRTOS: Does not have a “process” model like Linux. Instead, it has tasks. Tasks in FreeRTOS can be thought of as lightweight threads. FreeRTOS doesn’t manage the memory space for each task in the same way Linux does for processes. All tasks share the same address space and run in the same context. Linux Kernel: Linux uses processes, each of which has its own memory space. Processes in Linux can be multi-threaded, and each thread can have different scheduling characteristics. Linux processes are isolated from each other, so one process crashing doesn’t affect others. Memory Management FreeRTOS: Memory management is more manual. FreeRTOS does not have sophisticated memory management like Linux. It provides basic functions for allocating fixed-size blocks or dynamic memory pools (pvPortMalloc, vPortFree). It doesn’t have virtual memory, so all tasks have access to the same memory space, making it much simpler but also more prone to memory corruption if not managed properly. Linux Kernel: Linux includes virtual memory, meaning each process has its own virtual address space. It supports advanced features like paging and memory protection. The Linux kernel has a memory management unit (MMU) and sophisticated memory allocators for heap, stack, and memory mapping. Drivers FreeRTOS: Drivers in FreeRTOS are usually written to interface directly with the hardware. Embedded developers write hardware-specific drivers for devices such as GPIO, UART, SPI, I2C, etc. The drivers are tightly coupled with the hardware and typically run in the same task context as the rest of the application. Interfacing with hardware is done via direct memory-mapped registers and interrupt service routines (ISRs). Linux Kernel: The Linux kernel has a comprehensive set of device drivers for a wide variety of hardware. Drivers in Linux are implemented as kernel modules, which can be dynamically loaded and unloaded. These drivers abstract hardware interactions and often provide a system call interface for user-space applications to interact with hardware. GPIO Management FreeRTOS: Direct register manipulation or vendor-specific HAL libraries. No standard GPIO subsystem like Linux. GPIO interrupts are handled using ISR (Interrupt Service Routines) with FreeRTOS primitives like queues for event notification. Linux Kernel: GPIO Subsystem: Provides an abstraction layer using sysfs, character devices, or device tree bindings. Uses kernel interrupt handling with debounce and polling mechanisms. Interrupt Handling FreeRTOS: Interrupt handling is done through Interrupt Service Routines (ISRs), which are small, time-critical functions that handle hardware interrupts. FreeRTOS provides mechanisms to synchronize tasks with ISRs via semaphores or queues. Linux Kernel: Linux also uses ISRs, but in addition to regular interrupts, it has a more complex mechanism for handling asynchronous events, such as software interrupts, tasklets, work queues, etc. The kernel abstracts much of the interrupt management for portability. Synchronization Mechanisms FreeRTOS: Offers simple synchronization primitives like semaphores, mutexes, queues, and event groups. These are lightweight and highly optimized for small systems with limited resources. Linux Kernel: Linux also provides synchronization mechanisms like semaphores, mutexes, and spinlocks. However, these mechanisms are more complex and support features like priority inversion prevention, as well as various types of locking for different kernel contexts. Filesystem and I/O FreeRTOS: By default, FreeRTOS does not provide any filesystem management or complex I/O subsystem. I/O is typically done through simple APIs provided by the BSP or device driver code. Linux Kernel: Linux supports a full-fledged filesystem with many types (e.g., ext4, NTFS) and includes complex device I/O management, including file descriptors, blocking/non-blocking I/O, and extensive support for network file systems (NFS, CIFS). Conclusion: Feature FreeRTOS Linux Kernel Kernel/User Space Single space Separated Scheduler Priority-based, Preemptive CFS, RT scheduling Driver Model Direct access, HAL-based Kernel module-based GPIO Management Direct register access Standard GPIO subsystem Process Model Tasks only Processes & Threads Memory Management Heap-based, no MMU Virtual memory, MMU support Use Cases Real-time, MCUs High-performance, SBCs, SoCs FreeRTOS and Linux serve different purposes in embedded systems: ...

1. Linux Kernel Internals Importance: Understanding kernel internals is crucial for BSP and driver development as it helps in debugging, optimizing performance, and modifying the kernel to meet hardware-specific requirements. Topics: Kernel Architecture: [Monolithic vs Microkernel]({< ref “/posts/monolithic-vs-microkernel/” >}}), [Kernel Space vs User Space|Kernel and User Space interactions]({< ref “/posts/kernel-space-vs-user-space|kernel-and-user-space-interactions/” >}}). Process Management: Understanding task_struct, process states, scheduling algorithms. Interrupt Handling: SoftIRQs, tasklets, bottom halves, handling IRQs efficiently. Memory Management: Paging, kmalloc/vmalloc, slab allocator, ARM MMU and memory regions. [../../0-Inbox/syscalls|Syscalls]({< ref “/posts/../../0-inbox/syscalls|syscalls/” >}}): How system calls work, writing custom syscalls. [Kernel Synchronization in Linux|Kernel Synchronization]({< ref “/posts/kernel-synchronization-in-linux|kernel-synchronization/” >}}): Spinlocks, mutexes, semaphores, barriers, RCU. Workqueues and Timers: Deferred execution, using timers for scheduling tasks. 2. Linux Device Drivers Importance: Device drivers are the bridge between hardware and the OS. Understanding drivers is crucial for embedded systems and BSP development. ...

Overview The Device Tree (DT) is a data structure used to describe the hardware components of a system in a way that is independent of the operating system and software. It is particularly relevant for systems based on the ARM architecture, where the hardware varies significantly across devices. Instead of hardcoding hardware details in the kernel, the device tree provides a flexible way to inform the kernel about the system’s hardware layout. This simplifies kernel code and enables easier reuse across multiple hardware platforms. ...

Overview Character devices allow byte-by-byte communication between user-space applications and kernel drivers. They are commonly used for devices like serial ports, sensors, and custom hardware interfaces. The Linux kernel provides mechanisms for registering, managing, and interacting with character devices via a device file in /dev. Registering a Character Device To register a character device, the driver needs to: 1. Allocate a Major and Minor Number: Each character device is identified by a major number (device type) and a minor number (specific device). The major number indicates the driver associated with the device, while the minor number is used to differentiate between multiple devices handled by the same driver. If major and minor numbers are repeated, it can cause conflicts and lead to incorrect device identification. To avoid this, the kernel provides alloc_chrdev_region, a function to dynamically allocate major and minor numbers, ensuring uniqueness. These numbers are used in the /dev directory to associate device files with their corresponding drivers. Use alloc_chrdev_region to dynamically allocate a major number. dev_t dev; int result; // kernel/fs/char_dev.c // int alloc_chrdev_region(dev_t *dev, unsigned baseminor, unsigned count, const char *name) result = alloc_chrdev_region(&dev, 0, 1, "my_char_device"); if (result < 0) { pr_err("Failed to allocate major number\n"); return result; } pr_info("Device registered with major %d, minor %d\n", MAJOR(dev), MINOR(dev)); 2. Initialize and Register the Device: Define a cdev structure and initialize it with file operations. Use cdev_add to register the device with the kernel. struct cdev my_cdev; cdev_init(&my_cdev, &my_fops); my_cdev.owner = THIS_MODULE; result = cdev_add(&my_cdev, dev, 1); if (result < 0) { pr_err("Failed to add cdev\n"); unregister_chrdev_region(dev, 1); return result; } 3. Create a Device File (Optional): Creating a device file in /dev is optional because character devices can be accessed directly using their major and minor numbers through system calls or user-space libraries, bypassing the need for a device file. However, creating a file in /dev makes interaction more user-friendly by providing a standard interface. To interact with a character device without creating a device file, you can use system calls like mknod to create a temporary device node or interact with the device directly using its major and minor numbers programmatically. Use class_create and device_create to automatically create a device file in /dev. struct class *my_class; my_class = class_create(THIS_MODULE, "my_device_class"); if (IS_ERR(my_class)) { pr_err("Failed to create class\n"); cdev_del(&my_cdev); unregister_chrdev_region(dev, 1); return PTR_ERR(my_class); } device_create(my_class, NULL, dev, NULL, "my_char_device"); File Operations Character devices are controlled through a set of file operations defined in a struct file_operations. These operations determine how the device responds to system calls like open, read, write, and ioctl. ...

ioctl Implementation in Kernel Device Drivers Overview ioctl (Input/Output Control) is a powerful system call in Linux used to perform device-specific operations that are not covered by standard system calls like read, write, or open. It allows user-space applications to interact with kernel-space drivers for device-specific configurations and data exchanges. How ioctl Works 1. User-Space Interaction: A user-space application invokes ioctl using the following prototype: int ioctl(int fd, unsigned long cmd, void *arg); fd: File descriptor for the device. cmd: Command defining the operation. arg: Pointer to the data or argument passed between user-space and kernel-space. 2. Driver-Side Handling: The ioctl system call is routed to the driver by the kernel. The driver implements a specific unlocked_ioctl or compat_ioctl callback in the file_operations structure. 3. Data Flow: Arguments passed via arg can be pointers to user-space data, requiring the driver to use helper functions like copy_from_user and copy_to_user for secure data transfer. Steps to Implement ioctl in a Kernel Driver 1. Define ioctl Commands: Use macros to define command numbers, typically with the _IO, _IOR, _IOW, and _IOWR macros provided in <linux/ioctl.h>. #define MY_IOCTL_BASE 'M' #define IOCTL_CMD_GET _IOR(MY_IOCTL_BASE, 1, int) #define IOCTL_CMD_SET _IOW(MY_IOCTL_BASE, 2, int) _IOR: Read data from the kernel. _IOW: Write data to the kernel. _IOWR: Read and write data. _IO: Command without data. 2. Implement ioctl Callback: Define the unlocked_ioctl function in the driver. Handle commands appropriately based on cmd. static long my_ioctl(struct file *file, unsigned int cmd, unsigned long arg) { int value; switch (cmd) { case IOCTL_CMD_GET: value = 1234; // Example value if (copy_to_user((int __user *)arg, &value, sizeof(value))) return -EFAULT; break; case IOCTL_CMD_SET: if (copy_from_user(&value, (int __user *)arg, sizeof(value))) return -EFAULT; pr_info("Value set by user: %d\n", value); break; default: return -ENOTTY; // Command not supported } return 0; } 3. Integrate into file_operations: Register the ioctl handler in the file_operations structure. static const struct file_operations my_fops = { .owner = THIS_MODULE, .open = my_open, .release = my_release, .unlocked_ioctl = my_ioctl, }; 4. Test the ioctl Implementation: Write a user-space application to interact with the driver. #include <stdio.h> #include <fcntl.h> #include <unistd.h> #include <sys/ioctl.h> #define MY_IOCTL_BASE 'M' #define IOCTL_CMD_GET _IOR(MY_IOCTL_BASE, 1, int) #define IOCTL_CMD_SET _IOW(MY_IOCTL_BASE, 2, int) int main() { int fd, value = 42; fd = open("/dev/my_device", O_RDWR); if (fd < 0) { perror("Failed to open device"); return -1; } if (ioctl(fd, IOCTL_CMD_SET, &value) < 0) { perror("ioctl SET failed"); } if (ioctl(fd, IOCTL_CMD_GET, &value) < 0) { perror("ioctl GET failed"); } else { printf("Value from device: %d\n", value); } close(fd); return 0; } Best Practices for ioctl Use Explicit Command Definitions: Follow a consistent naming convention for command macros. Secure User-Kernel Data Transfer: Always validate pointers and sizes. Use copy_from_user and copy_to_user for safe data exchange. Error Handling: Return appropriate error codes for unsupported commands or invalid inputs. Limit ioctl Usage: Avoid using ioctl for operations that can be implemented using read or write. Magic Number: Ensure it’s unique (check Documentation/ioctl/ioctl-number.txt in kernel sources). Atomicity: Use locks if hardware operations are not atomic. Cross-Platform: Handle 32/64-bit compatibility with compat_ioctl if needed. Real-World Example: Custom ARM Board For a custom ARM board, you might need an ioctl to configure hardware parameters like GPIO modes or clock frequencies. ...

Problem In bootloader the environment variables are accessible. The values could be accessed and modified perfectly from bootloader cli (hush shell). When booting to the linux kernel and accessing from bash shell (userspace) through fw_printenv the environment variables are not accessible. Getting following logs: Warning: Bad CRC, using default environment Setup: Using MMC storage The /etc/fw_env.config file has following entry and the offset and size are same from uboot’s .config file /dev/mmcblk0p2 0x3f8000 0x8000 Uboot Environment in RK3588 Uboot Config config ENV_OFFSET hex "Environment offset" depends on !ENV_IS_IN_UBI depends on !ENV_IS_NOWHERE || ENVF default 0x0 if ENVF default 0x3f8000 help Offset from the start of the device (or partition) config ENV_SIZE hex "Environment size" default 0x8000 help Size of the environment storage area Parameters CONFIG_ENV_SIZE=0x8000 --> 32K CONFIG_ENV_OFFSET=0x3f8000 --> 4096-32=4064 4096K=4M u-boot/tools/env/fw_env.c Either crc0_ok = (crc0 == *environment.crc); is invalid ...

Number Macro Log Level Description Equivalent 0 pr_emerg Emergency System is unusable. KERN_EMERG 1 pr_alert Alert Action must be taken immediately. KERN_ALERT 2 pr_crit Critical Critical conditions. KERN_CRIT 3 pr_err Error Error conditions. KERN_ERR 4 pr_warn Warning Warning conditions. KERN_WARNING 5 pr_notice Notice Normal but significant condition. KERN_NOTICE 6 pr_info Informational Informational messages. KERN_INFO 7 pr_debug Debug Debugging messages. KERN_DEBUG The number corresponds to the log level used by the Linux kernel, with lower numbers indicating higher severity. For example, if the log level is set to 4 (Warning), only messages from pr_emerg to pr_warn will appear in the system logs. Default log level is generally set to 6.