Lecture 2 - OS Structure¶

System Calls¶

An interface which kernel provides to the user space to interact with the kernel.
Syscall : proviledged instruction

Example Ilustration -- Write()¶

First printf() function is called, then write() function is called. In the write() function, there is a system call to the kernel. [with syscall number $0x1]
Then, go into kernel space, and the kernel will do the actual writing to the file descriptor. 1) kernel_entry code will be called -- Saved all user space registers 2) calls write syscall handler -- Get from syscall_table, which is an array

SYSCALL_DEFINE3(write, unsigned int, fd, const char __user *, buf, size_t, count)
{
    return ksys_write(fd, buf, count);
}

After write finish, call ret_to_user
Restore all saved user space registers
Transfer control flow to user space

System Call Implementation¶

System-call interface maintains a table indexed according to numbers assigned to each system call. Each entry in the table points to the entry point of the system call.

The system call interface invokes the intended system call in OS kernel and returns status of the system call and any return values
The caller needs to know nothing about how the system call is implemented

Example : Copy a file¶

Use strace to trace system calls
strace cp file1 file2
strace -c cp file1 file2 -- Count the number of system calls

strace cp main.c main.copy 2>&1 |wc -l #175 lines

2>&1 -- Redirect stderr to stdout
|wc -l -- Count the number of lines

Time spent in system calls¶

time ap start_kernel -- Time spent in system calls

time count multiple pid and sum them up, so it seems that the time spent in system calls is more than the actual time spent in the system calls.
time grep -Rn start_kernel

System Call Parameter Passing¶

Three general methods used to pass parameters to the OS

Simplest: pass the parameters in registers
In some cases, may be more parameters than registers
Parameters stored in a block, or table, in memory, and address of block passed as a parameter in a register
This approach taken by Linux and Solaris
Parameters placed, or pushed, onto the stack by the program and popped off the stack by the operating system
Block and stack methods do not limit the number or length of parameters being passed

System Service¶

Linkers and Loaders¶

Where does static variable goes? -- .data section
Where does static constant goes? -- .rodata section
Why unintialized in .bss instead of .data? -- To save space, as .bss is not stored in the file
readelf -h main
Entry Address
Magic Number : 0x7f 0x45 0x4c 0x46 -- ELF : Linux defines the format of the file using magic number

Linking¶

Static Linking¶

All needed code is packed in single binary, leading to large binary
可以移植性强

Dynamic linking¶

Reuse libraries to reduce ELF file size.
How to resolve library calls?

readelf -p .interp main

This dump the .interp section of the ELF file, which contains the path of the dynamic linker
It is the loader who resolves lib calls.
lib call: like printf()
loader: ld-linux-aarch64.so.1

Running a Binary¶

While for static linking, the mapping is much less.

Memory layout is in user space ?
User space: stack, heap, data, text
Kernel space: kernel code, kernel data, kernel stack

Questions¶

Who setups ELF file mapping? -- kernel: execve() system call

Who setups stack and heap?
Who setups libraries?

Dynamic linking has to do more system calls

Setup a Binary¶

Static Binary¶

In readelf we see that entry of the main.static is 0x400640
We find that 0x400640 is the address of the start function in the main.static binary
objdump-d a.out

regs->pc = pc here pc is the address of the start function -- elf_entry

Dynamic Binary¶

For dynamic binary, the elf_entry -- interp_elf_ex -> e_entry
ld.so -- Loader resolves the library calls
So loader has to be called first, then the start function

Why Applications are Operating System Specific¶

System calls are different -- name / number

Apps can be multi-operating system * Written in interpreted language like Python, Ruby, and interpreter available on multiple operating systems * App written in language that includes a VM containing the running app (like Java) * Use standard language (like C), compile separately on each operating system to run on each

Application Binary Interface (ABI) is architecture equivalent of API, defines how different components of binary code can interface for a given operating system on a given architecture, CPU, etc

Operating-System Design and Implementation¶

Operating System Structure¶

General-purpose OS is very large program

Various ways to structure ones

Simple structure –MS-DOS
Monolithic –Unix, Linux
Layered –an abstraction
Microkernel –Mach
Good Helper Website

Building and Booting an Operating System¶

Operating System Debugging¶

最后更新: 2024年9月30日 11:27:07
创建日期: 2024年9月30日 11:27:07