So I like to know when I call malloc what exactly happens? Since malloc must be a wrapper so some other function must called inside it may be brk or something, But what happens inside brk, all I know these are memory allocation functions that allocate memory and return non zero value if succeed. But then what is memory management.

So management if memory If I assume is simply keeping sizes of memory against there current value. so for example if I have a two malloc calls both are allocation 10 integer each have the size of allocation will be kept against current values in some table..Whats that table is called. So assume I am implementing a memory management unit is making sense. Is there any standard used in Linux for some struct that can be referred to as this table, Can I have multiple memory management units in one system means there is a parent MMU, and my MMU is just depending on some standard functions

CodePudding user response：

When you call malloc() on Linux, you call a thin wrapper in libc (the C standard library implementation). This thin wrapper will often use the mmap system call to get memory allocated.

On x86-64, system calls are made using syscall in assembly along with some arguments in conventional registers specified in the SysV ABI.

The syscall instruction makes the processor jump in supervisor mode from user mode at the address specified in the STAR64 MSR register of these processors.

On Linux, it will jump at the first instruction of the file entry_64.S. This assembly file will make several checkups to determine what user mode is asking and return a value in some registers.

For mmap, the system call is looking to allocate physical memory and to modify the page tables so that translations from the next portion of virtual memory available will translate to these positions within physical memory. For more info, you can browse my answer at Each program allocates a fixed stack size? Who defines the amount of stack memory for each application running?. The dynamic allocations are done on what's called the heap which can grow as much as there is memory available.

The libc library will actually allocate memory itself to determine how much memory was allocated to your user mode process. This is required because the granularity of allocation on modern x86-64 is one 4kb page. The libc library will thus keep track in what's called "arenas" of the pages that were allocated to try and give you new allocations in those pages instead of requesting new ones.

Also, the Linux kernel keeps track of what memory was allocated to your process in the PCB (the task_struct) in the mm field which is a complex struct in itself.