SO2 Lab 06 - Memory Mapping¶

Lab objectives¶

Understand address space mapping mechanisms
Learn about the most important structures related to memory management

Keywords:

address space
mmap()
struct page
struct vm_area_struct
struct vm_struct
remap_pfn_range
SetPageReserved()
ClearPageReserved()

Overview¶

In the Linux kernel it is possible to map a kernel address space to a user address space. This eliminates the overhead of copying user space information into the kernel space and vice versa. This can be done through a device driver and the user space device interface (/dev).

This feature can be used by implementing the mmap() operation in the device driver's struct file_operations and using the mmap() system call in user space.

The basic unit for virtual memory management is a page, which size is usually 4K, but it can be up to 64K on some platforms. Whenever we work with virtual memory we work with two types of addresses: virtual address and physical address. All CPU access (including from kernel space) uses virtual addresses that are translated by the MMU into physical addresses with the help of page tables.

A physical page of memory is identified by the Page Frame Number (PFN). The PFN can be easily computed from the physical address by dividing it with the size of the page (or by shifting the physical address with PAGE_SHIFT bits to the right).

For efficiency reasons, the virtual address space is divided into user space and kernel space. For the same reason, the kernel space contains a memory mapped zone, called lowmem, which is contiguously mapped in physical memory, starting from the lowest possible physical address (usually 0). The virtual address where lowmem is mapped is defined by PAGE_OFFSET.

On a 32bit system, not all available memory can be mapped in lowmem and because of that there is a separate zone in kernel space called highmem which can be used to arbitrarily map physical memory.

Memory allocated by kmalloc() resides in lowmem and it is physically contiguous. Memory allocated by vmalloc() is not contiguous and does not reside in lowmem (it has a dedicated zone in highmem).

Structures used for memory mapping¶

Before discussing the mechanism of memory-mapping a device, we will present some of the basic structures related to the memory management subsystem of the Linux kernel.

Before discussing about the memory mapping mechanism over a device, we will present some of the basic structures used by the Linux memory management subsystem. Some of the basic structures are: struct page, struct vm_area_struct, struct mm_struct.

`struct page`¶

struct page is used to embed information about all physical pages in the system. The kernel has a struct page structure for all pages in the system.

There are many functions that interact with this structure:

virt_to_page() returns the page associated with a virtual address
pfn_to_page() returns the page associated with a page frame number
page_to_pfn() return the page frame number associated with a struct page
page_address() returns the virtual address of a struct page; this functions can be called only for pages from lowmem
kmap() creates a mapping in kernel for an arbitrary physical page (can be from highmem) and returns a virtual address that can be used to directly reference the page

`struct vm_area_struct`¶

struct vm_area_struct holds information about a contiguous virtual memory area. The memory areas of a process can be viewed by inspecting the maps attribute of the process via procfs:

root@qemux86:~# cat /proc/1/maps
#address          perms offset  device inode     pathname
08048000-08050000 r-xp 00000000 fe:00 761        /sbin/init.sysvinit
08050000-08051000 r--p 00007000 fe:00 761        /sbin/init.sysvinit
08051000-08052000 rw-p 00008000 fe:00 761        /sbin/init.sysvinit
092e1000-09302000 rw-p 00000000 00:00 0          [heap]
4480c000-4482e000 r-xp 00000000 fe:00 576        /lib/ld-2.25.so
4482e000-4482f000 r--p 00021000 fe:00 576        /lib/ld-2.25.so
4482f000-44830000 rw-p 00022000 fe:00 576        /lib/ld-2.25.so
44832000-449a9000 r-xp 00000000 fe:00 581        /lib/libc-2.25.so
449a9000-449ab000 r--p 00176000 fe:00 581        /lib/libc-2.25.so
449ab000-449ac000 rw-p 00178000 fe:00 581        /lib/libc-2.25.so
449ac000-449af000 rw-p 00000000 00:00 0
b7761000-b7763000 rw-p 00000000 00:00 0
b7763000-b7766000 r--p 00000000 00:00 0          [vvar]
b7766000-b7767000 r-xp 00000000 00:00 0          [vdso]
bfa15000-bfa36000 rw-p 00000000 00:00 0          [stack]

A memory area is characterized by a start address, a stop address, length, permissions.

A struct vm_area_struct is created at each mmap() call issued from user space. A driver that supports the mmap() operation must complete and initialize the associated struct vm_area_struct. The most important fields of this structure are:

vm_start, vm_end - the beginning and the end of the memory area, respectively (these fields also appear in /proc/<pid>/maps);
vm_file - the pointer to the associated file structure (if any);
vm_pgoff - the offset of the area within the file;
vm_flags - a set of flags;
vm_ops - a set of working functions for this area
vm_next, vm_prev - the areas of the same process are chained by a list structure

`struct mm_struct`¶

struct mm_struct encompasses all memory areas associated with a process. The mm field of struct task_struct is a pointer to the struct mm_struct of the current process.

Device driver memory mapping¶

Memory mapping is one of the most interesting features of a Unix system. From a driver's point of view, the memory-mapping facility allows direct memory access to a user space device.

To assign a mmap() operation to a driver, the mmap field of the device driver's struct file_operations must be implemented. If that is the case, the user space process can then use the mmap() system call on a file descriptor associated with the device.

The mmap system call takes the following parameters:

void *mmap(caddr_t addr, size_t len, int prot,
           int flags, int fd, off_t offset);

To map memory between a device and user space, the user process must open the device and issue the mmap() system call with the resulting file descriptor.

The device driver mmap() operation has the following signature:

int (*mmap)(struct file *filp, struct vm_area_struct *vma);

The filp field is a pointer to a struct file created when the device is opened from user space. The vma field is used to indicate the virtual address space where the memory should be mapped by the device. A driver should allocate memory (using kmalloc(), vmalloc(), alloc_pages()) and then map it to the user address space as indicated by the vma parameter using helper functions such as remap_pfn_range().

remap_pfn_range() will map a contiguous physical address space into the virtual space represented by vm_area_struct:

int remap_pfn_range (structure vm_area_struct *vma, unsigned long addr,
                     unsigned long pfn, unsigned long size, pgprot_t prot);

remap_pfn_range() expects the following parameters:

vma - the virtual memory space in which mapping is made;
addr - the virtual address space from where remapping begins; page tables for the virtual address space between addr and addr + size will be formed as needed
pfn - the page frame number to which the virtual address should be mapped
size - the size (in bytes) of the memory to be mapped
prot - protection flags for this mapping

Here is an example of using this function that contiguously maps the physical memory starting at page frame number pfn (memory that was previously allocated) to the vma->vm_start virtual address:

struct vm_area_struct *vma;
unsigned long len = vma->vm_end - vma->vm_start;
int ret ;

ret = remap_pfn_range(vma, vma->vm_start, pfn, len, vma->vm_page_prot);
if (ret < 0) {
    pr_err("could not map the address area\n");
    return -EIO;
}

To obtain the page frame number of the physical memory we must consider how the memory allocation was performed. For each kmalloc(), vmalloc(), alloc_pages(), we must used a different approach. For kmalloc() we can use something like:

static char *kmalloc_area;

unsigned long pfn = virt_to_phys((void *)kmalloc_area)>>PAGE_SHIFT;

while for vmalloc():

static char *vmalloc_area;

unsigned long pfn = vmalloc_to_pfn(vmalloc_area);

and finally for alloc_pages():

struct page *page;

unsigned long pfn = page_to_pfn(page);

Attention

Note that memory allocated with vmalloc() is not physically contiguous so if we want to map a range allocated with vmalloc(), we have to map each page individually and compute the physical address for each page.

Since the pages are mapped to user space, they might be swapped out. To avoid this we must set the PG_reserved bit on the page. Enabling is done using SetPageReserved() while reseting it (which must be done before freeing the memory) is done with ClearPageReserved():

void alloc_mmap_pages(int npages)
{
    int i;
    char *mem = kmalloc(PAGE_SIZE * npages);

    if (!mem)
        return mem;

    for(i = 0; i < npages * PAGE_SIZE; i += PAGE_SIZE)
        SetPageReserved(virt_to_page(((unsigned long)mem) + i));

    return mem;
}

void free_mmap_pages(void *mem, int npages)
{
    int i;

    for(i = 0; i < npages * PAGE_SIZE; i += PAGE_SIZE)
        ClearPageReserved(virt_to_page(((unsigned long)mem) + i));

    kfree(mem);
}

Exercises¶

Important

To solve exercises, you need to perform these steps:

prepare skeletons from templates

build modules

copy modules to the VM

start the VM and test the module in the VM.

The current lab name is memory_mapping. See the exercises for the task name.

The skeleton code is generated from full source examples located in tools/labs/templates. To solve the tasks, start by generating the skeleton code for a complete lab:

tools/labs $ make clean
tools/labs $ LABS=<lab name> make skels

You can also generate the skeleton for a single task, using

tools/labs $ LABS=<lab name>/<task name> make skels

Once the skeleton drivers are generated, build the source:

tools/labs $ make build

Then, copy the modules and start the VM:

tools/labs $ make copy
tools/labs $ make boot

The modules are placed in /home/root/skels/memory_mapping/<task_name>.

Alternatively, we can copy files via scp, in order to avoid restarting the VM. For additional details about connecting to the VM via the network, please check Connecting to the Virtual Machine.

Review the Exercises section for more detailed information.

Warning

Before starting the exercises or generating the skeletons, please run git pull inside the Linux repo, to make sure you have the latest version of the exercises.

If you have local changes, the pull command will fail. Check for local changes using git status. If you want to keep them, run git stash before pull and git stash pop after. To discard the changes, run git reset --hard master.

If you already generated the skeleton before git pull you will need to generate it again.

1. Mapping contiguous physical memory to userspace¶

Implement a device driver that maps contiguous physical memory (e.g. obtained via kmalloc()) to userspace.

Review the Device driver memory mapping section, generate the skeleton for the task named kmmap and fill in the areas marked with TODO 1.

Start with allocating a NPAGES+2 memory area page using kmalloc() in the module init function and find the first address in the area that is aligned to a page boundary.

Hint

The size of a page is PAGE_SIZE.

Store the allocated area in kmalloc_ptr and the page aligned address in kmalloc_area:

Use PAGE_ALIGN() to determine kmalloc_area.

Enable the PG_reserved bit of each page with SetPageReserved(). Clear the bit with ClearPageReserved() before freeing the memory.

Hint

Use virt_to_page() to translate virtual pages into physical pages, as required by SetPageReserved() and ClearPageReserved().

For verification purpose (using the test below), fill in the first 4 bytes of each page with the following values: 0xaa, 0xbb, 0xcc, 0xdd.

Implement the mmap() driver function.

Hint

For mapping, use remap_pfn_range(). The third argument for remap_pfn_range() is a page frame number (PFN).

To convert from virtual kernel address to physical address, use virt_to_phys().

To convert a physical address to its PFN, shift the address with PAGE_SHIFT bits to the right.

For testing, load the kernel module and run:

root@qemux86:~# skels/memory_mapping/test/mmap-test 1

If everything goes well, the test will show "matched" messages.

2. Mapping non-contiguous physical memory to userspace¶

Implement a device driver that maps non-contiguous physical memory (e.g. obtained via vmalloc()) to userspace.

Review the Device driver memory mapping section, generate the skeleton for the task named vmmap and fill in the areas marked with TODO 1.

Allocate a memory area of NPAGES with vmalloc().

Hint

The size of a page is PAGE_SIZE. Store the allocated area in vmalloc_area. Memory allocated by vmalloc() is paged aligned.

Enable the PG_reserved bit of each page with SetPageReserved(). Clear the bit with ClearPageReserved() before freeing the memory.

Hint

Use vmalloc_to_page() to translate virtual pages into physical pages used by the functions SetPageReserved() and ClearPageReserved().

For verification purpose (using the test below), fill in the first 4 bytes of each page with the following values: 0xaa, 0xbb, 0xcc, 0xdd.

Implement the mmap driver function.

Hint

To convert from virtual vmalloc address to physical address, use vmalloc_to_pfn() which returns a PFN directly.

Attention

vmalloc pages are not physically contiguous so it is needed to use remap_pfn_range() for each page.

Loop through all virtual pages and for each: * determine the physical address * map it with remap_pfn_range()

Make sure that you determine the physical address each time and that you use a range of one page for mapping.

For testing, load the kernel module and run:

root@qemux86:~# skels/memory_mapping/test/mmap-test 1

If everything goes well, the test will show "matched" messages.

3. Read / write operations in mapped memory¶

Modify one of the previous modules to allow read / write operations on your device. This is a didactic exercise to see that the same space can also be used with the mmap() call and with read() and write() calls.

Fill in areas marked with TODO 2.

Note

The offset parameter sent to the read / write operation can be ignored as all reads / writes from the test program will be done with 0 offsets.

For testing, load the kernel module and run:

root@qemux86:~# skels/memory_mapping/test/mmap-test 2

4. Display memory mapped in procfs¶

Using one of the previous modules, create a procfs file in which you display the total memory mapped by the calling process.

Fill in the areas marked with TODO 3.

Create a new entry in procfs (PROC_ENTRY_NAME, defined in mmap-test.h) that will show the total memory mapped by the process that called the read() on that file.

Hint

Use proc_create(). For the mode parameter, use 0, and for the parent parameter use NULL. Use my_proc_file_ops() for operations.

In the module exit function, delete the PROC_ENTRY_NAME entry using remove_proc_entry().

Note

A (complex) use and description of the struct seq_file interface can be found here in this example .

For this exercise, just a simple use of the interface described here is sufficient. Check the "extra-simple" API described there.

In the my_seq_show() function you will need to:

Obtain the struct mm_struct structure of the current process using the get_task_mm() function.

Hint

The current process is available via the current variable of type struct task_struct*.
Iterate through the entire struct vm_area_struct list associated with the process.

Hint

Use the variable vma_iterator and start from mm->mmap. Use the vm_next field of the struct vm_area_struct to navigate through the list of memory areas. Stop when you reach NULL.
Use vm_start and vm_end for each area to compute the total size.
Use pr_info("%lx %lxn, ...)() to print vm_start and vm_end for each area.
To release struct mm_struct, decrement the reference counter of the structure using mmput().
Use seq_printf() to write to the file. Show only the total count, no other messages. Do not even show newline (n).

In my_seq_open() register the display function (my_seq_show()) using single_open().

Note

single_open() can use NULL as its third argument.

For testing, load the kernel module and run:

root@qemux86:~# skels/memory_mapping/test/mmap-test 3

Note

The test waits for a while (it has an internal sleep instruction). As long as the test waits, use the pmap command in another console to see the mappings of the test and compare those to the test results.

SO2 Lab 06 - Memory Mapping¶

Lab objectives¶

Overview¶

Structures used for memory mapping¶

struct page¶

struct vm_area_struct¶

struct mm_struct¶

Device driver memory mapping¶

Further reading¶

Exercises¶

1. Mapping contiguous physical memory to userspace¶

2. Mapping non-contiguous physical memory to userspace¶

3. Read / write operations in mapped memory¶

4. Display memory mapped in procfs¶

`struct page`¶

`struct vm_area_struct`¶

`struct mm_struct`¶