`Fix-Mapped` addresses are a set of special compile-time addresses whose corresponding physical addresses do not have to be a linear address minus `__START_KERNEL_map`. Each fix-mapped address maps one page frame and the kernel uses them as pointers that never change their address. That is the main point of these addresses. As the comment says: `to have a constant address at compile time, but to set the physical address only in the boot process`. You can remember that in the earliest [part](https://proninyaroslav.gitbooks.io/linux-insides-ru/content/Initialization/linux-initialization-1.html), we already set the `level2_fixmap_pgt`:
As you can see `level2_fixmap_pgt` is right after the `level2_kernel_pgt` which is kernel code+data+bss. Every fix-mapped address is represented by an integer index which is defined in the `fixed_addresses` enum from the [arch/x86/include/asm/fixmap.h](https://github.com/torvalds/linux/blob/16f73eb02d7e1765ccab3d2018e0bd98eb93d973/arch/x86/include/asm/fixmap.h). For example it contains entries for `VSYSCALL_PAGE` - if emulation of legacy vsyscall page is enabled, `FIX_APIC_BASE` for local [apic](http://en.wikipedia.org/wiki/Advanced_Programmable_Interrupt_Controller), etc. In virtual memory fix-mapped area is placed in the modules area:
Here `__end_of_permanent_fixed_addresses` is an element of the `fixed_addresses` enum and as I wrote above: Every fix-mapped address is represented by an integer index which is defined in the `fixed_addresses`. `PAGE_SHIFT` determines the size of a page. For example size of the one page we can get with the `1 << PAGE_SHIFT` expression.
In our case we need to get the size of the fix-mapped area, but not only of one page, that's why we are using `__end_of_permanent_fixed_addresses` for getting the size of the fix-mapped area. The `__end_of_permanent_fixed_addresses` is the last index of the `fixed_addresses` enum or in other words the `__end_of_permanent_fixed_addresses` contains amount of pages in a fixed-mapped area. So if multiply value of the `__end_of_permanent_fixed_addresses` on a page size value we will get size of fix-mapped area. In my case it's a little more than `536` kilobytes. In your case it might be a different number, because the size depends on amount of the fix-mapped addresses which are depends on your kernel's configuration.
The second `FIXADDR_START` macro just subtracts the fix-mapped area size from the last address of the fix-mapped area to get its base virtual address. `FIXADDR_TOP` is a rounded up address from the base address of the [vsyscall](https://lwn.net/Articles/446528/) space:
first of all it checks that the index given for the `fixed_addresses` enum is not greater or equal than `__end_of_fixed_addresses` with the `BUILD_BUG_ON` macro and then returns the result of the `__fix_to_virt` macro:
Here we shift left the given index of a `fix-mapped` area on the `PAGE_SHIFT` which determines size of a page as I wrote above and subtract it from the `FIXADDR_TOP` which is the highest address of the `fix-mapped` area:
```
+-----------------+
| PAGE 1 | FIXADDR_TOP (virt address)
| PAGE 2 |
| PAGE 3 |
| PAGE 4 (idx) | x - 4
| PAGE 5 |
+-----------------+
```
There is an inverse function for getting an index of a fix-mapped area corresponding to the given virtual address:
The `virt_to_fix` takes a virtual address, checks that this address is between `FIXADDR_START` and `FIXADDR_TOP` and calls the `__virt_to_fix` macro which implemented as:
As we may see, the `__virt_to_fix` macro clears the first `12` bits in the given virtual address, subtracts it from the last address the of `fix-mapped` area (`FIXADDR_TOP`) and shifts the result right on `PAGE_SHIFT` which is `12`. Let me explain how it works.
As in previous example (in `__fix_to_virt` macro), we start from the top of the fix-mapped area. We also go back to bottom from the top to search an index of a fix-mapped area corresponding to the given virtual address. As you may see, first of all we will clear the first `12` bits in the given virtual address with `x & PAGE_MASK` expression. This allows us to get base address of page. We need to do this for case when the given virtual address points somewhere in a beginning/middle or end of a page, but not to the base address of it. At the next step subtract this from the `FIXADDR_TOP` and this gives us virtual address of a corresponding page in a fix-mapped area. In the end we just divide value of this address on `PAGE_SHIFT`. This gives us index of a fix-mapped area corresponding to the given virtual address. It may looks hard, but if you will go through this step by step, you will be sure that the `__virt_to_fix` macro is pretty easy.
`Fix-mapped` addresses are used in different [places](http://lxr.free-electrons.com/ident?i=fix_to_virt) in the linux kernel. `IDT` descriptor stored there, [Intel Trusted Execution Technology](http://en.wikipedia.org/wiki/Trusted_Execution_Technology) UUID stored in the `fix-mapped` area started from `FIX_TBOOT_BASE` index, [Xen](http://en.wikipedia.org/wiki/Xen) bootmap and many more... We already saw a little about `fix-mapped` addresses in the fifth [part](https://proninyaroslav.gitbooks.io/linux-insides-ru/content/Initialization/linux-initialization-5.html) about of the linux kernel initialization. We use `fix-mapped` area in the early `ioremap` initialization. Let's look at it more closely and try to understand what `ioremap` is, how it is implemented in the kernel and how it is related to the `fix-mapped` addresses.
The Linux kernel provides many different primitives to manage memory. For this moment we will touch `I/O memory`. Every device is controlled by reading/writing from/to its registers. For example a driver can turn off/on a device by writing to its registers or get the state of a device by reading from its registers. Besides registers, many devices have buffers where a driver can write something or read from there. As we know for this moment there are two ways to access device's registers and data buffers:
In the first case every control register of a device has a number of input and output port. A device driver can read from a port and write to it with two `in` and `out` instructions which we already saw. If you want to know about currently registered port regions, you can learn about them by accessing `/proc/ioports`:
`/proc/ioports` provides information about which driver uses which address of a `I/O` port region. All of these memory regions, for example `0000-0cf7`, were claimed with the `request_region` function from the [include/linux/ioport.h](https://github.com/torvalds/linux/blob/16f73eb02d7e1765ccab3d2018e0bd98eb93d973/include/linux/ioport.h). Actually `request_region` is a macro which is defined as:
`request_region` allocates an `I/O` port region. Very often the `check_region` function is called before the `request_region` to check that the given address range is available and the `release_region` function to release the memory region. `request_region` returns a pointer to the `resource` structure. The `resource` structure represents an abstraction for a tree-like subset of system resources. We already saw the `resource` structure in the fifth part of the kernel [initialization](https://proninyaroslav.gitbooks.io/linux-insides-ru/content/Initialization/linux-initialization-5.html) process and it looks as follows:
and contains start and end addresses of the resource, the name, etc. Every `resource` structure contains pointers to the `parent`, `sibling` and `child` resources. As it has a parent and a child, it means that every subset of resources has root `resource` structure. For example, for `I/O` ports it is the `ioport_resource` structure:
As I have mentioned before, `request_regions` is used to register I/O port regions and this macro is used in many [places](http://lxr.free-electrons.com/ident?i=request_region) in the kernel. For example let's look at [drivers/char/rtc.c](https://github.com/torvalds/linux/blob/16f73eb02d7e1765ccab3d2018e0bd98eb93d973/drivers/char/rtc.c). This source code file provides the [Real Time Clock](http://en.wikipedia.org/wiki/Real-time_clock) interface in the linux kernel. As every kernel module, `rtc` module contains `module_init` definition:
where `rtc_init` is the `rtc` initialization function. This function is defined in the same `rtc.c` source code file. In the `rtc_init` function we can see a couple of calls to the `rtc_request_region` functions, which wrap `request_region` for example:
So with the `request_region(RTC_PORT(0), size, "rtc")` we register a memory region that starts at `0x70` and and has a size of `0x8`. Let's look at `/proc/ioports`:
So, we got it! Ok, that was it for the I/O ports. The second way to communicate with drivers is through the use of `I/O` memory. As I have mentioned above this works by mapping the control registers and the memory of a device to the memory address space. `I/O` memory is a set of contiguous addresses which are provided by a device to the CPU through a bus. None of the memory-mapped I/O addresses are used by the kernel directly. There is a special `ioremap` function which allows us to convert the physical address on a bus to a kernel virtual address. In other words, `ioremap` maps I/O physical memory regions to make them accessible from the kernel. The `ioremap` function takes two parameters:
Part of these addresses are from the call of the `e820_reserve_resources` function. We can find a call to this function in the [arch/x86/kernel/setup.c](https://github.com/torvalds/linux/blob/16f73eb02d7e1765ccab3d2018e0bd98eb93d973/arch/x86/kernel/setup.c) and the function itself is defined in [arch/x86/kernel/e820.c](https://github.com/torvalds/linux/blob/16f73eb02d7e1765ccab3d2018e0bd98eb93d973/arch/x86/kernel/e820.c). `e820_reserve_resources` goes through the [e820](http://en.wikipedia.org/wiki/E820) map and inserts memory regions into the root `iomem` resource structure. All `e820` memory regions which are inserted into the `iomem` resource have the following types:
Now let's try to understand how `ioremap` works. We already know a little about `ioremap`, we saw it in the fifth [part](https://proninyaroslav.gitbooks.io/linux-insides-ru/content/Initialization/linux-initialization-5.html) about linux kernel initialization. If you have read this part, you can remember the call of the `early_ioremap_init` function from the [arch/x86/mm/ioremap.c](https://github.com/torvalds/linux/blob/16f73eb02d7e1765ccab3d2018e0bd98eb93d973/arch/x86/mm/ioremap.c). Initialization of the `ioremap` is split into two parts: there is the early part which we can use before the normal `ioremap` is available and the normal `ioremap` which is available after `vmalloc` initialization and the call of `paging_init`. We do not know anything about `vmalloc` for now, so let's consider early initialization of the `ioremap`. First of all `early_ioremap_init` checks that `fixmap` is aligned on page middle directory boundary:
more about `BUILD_BUG_ON` you can read in the first part about [Linux Kernel initialization](https://proninyaroslav.gitbooks.io/linux-insides-ru/content/Initialization/linux-initialization-1.html). So `BUILD_BUG_ON` macro raises a compilation error if the given expression is true. In the next step after this check, we can see call of the `early_ioremap_setup` function from the [mm/early_ioremap.c](https://github.com/torvalds/linux/blob/16f73eb02d7e1765ccab3d2018e0bd98eb93d973/mm/early_ioremap.c). This function presents generic initialization of the `ioremap`. `early_ioremap_setup` function fills the `slot_virt` array with the virtual addresses of the early fixmaps. All early fixmaps are after `__end_of_permanent_fixed_addresses` in memory. They start at `FIX_BITMAP_BEGIN` (top) and end with `FIX_BITMAP_END` (down). Actually there are `512` temporary boot-time mappings, used by early `ioremap`:
`slot_virt` contains the virtual addresses of the `fix-mapped` areas, `prev_map` array contains addresses of the early ioremap areas. Note that I wrote above: `Actually there are 512 temporary boot-time mappings, used by early ioremap` and you can see that all arrays are defined with the `__initdata` attribute which means that this memory will be released after the kernel initialization process. After `early_ioremap_setup` has finished its work, we're getting page middle directory where early ioremap begins with the `early_ioremap_pmd` function which just gets the base address of the page global directory and calculates the page middle directory for the given address:
*`init_mm` - memory descriptor of the `init` process (you can read about it in the previous [part](https://proninyaroslav.gitbooks.io/linux-insides-ru/content/Initialization/linux-initialization-5.html));
The `pmd_populate_kernel` function is defined in the [arch/x86/include/asm/pgalloc.h](https://github.com/torvalds/linux/blob/16f73eb02d7e1765ccab3d2018e0bd98eb93d973/arch/x86/include/asm/pgalloc.) and populates the page middle directory (`pmd`) provided as an argument with the given page table entries (`bm_pte`):
That's all. Early `ioremap` is ready to use. There are a couple of checks in the `early_ioremap_init` function, but they are not so important, anyway initialization of the `ioremap` is finished.
for mapping/unmapping of I/O physical address to virtual address. Both functions depend on the `CONFIG_MMU` configuration option. [Memory management unit](http://en.wikipedia.org/wiki/Memory_management_unit) is a special block of memory management. The main purpose of this block is the translation of physical addresses to virtual addresses. The memory management unit knows about the high-level page table addresses (`pgd`) from the `cr3` control register. If `CONFIG_MMU` options is set to `n`, `early_ioremap` just returns the given physical address and `early_iounmap` does nothing. If `CONFIG_MMU` option is set to `y`, `early_ioremap` calls `__early_ioremap` which takes three parameters:
First of all in the `__early_ioremap`, we go through all early ioremap fixmap slots and search for the first free one in the `prev_map` array. When we found it we remember its number in the `slot` variable and set up size:
We know that size of a page is 4096 bytes or `1000000000000` in binary. `PAGE_SIZE - 1` will be `111111111111`, but with `~`, we will get `000000000000`, but as we use `~PAGE_MASK` we will get `111111111111` again. On the second line we do the same but clear the first 12 bits and getting page-aligned size of the area on the third line. We getting aligned area and now we need to get the number of pages which are occupied by the new `ioremap` area and calculate the fix-mapped index from `fixed_addresses` in the next steps:
Now we can fill `fix-mapped` area with the given physical addresses. On every iteration in the loop, we call the `__early_set_fixmap` function from the [arch/x86/mm/ioremap.c](https://github.com/torvalds/linux/blob/16f73eb02d7e1765ccab3d2018e0bd98eb93d973/arch/x86/mm/ioremap.c), increase the given physical address by the page size which is `4096` bytes and update the `addresses` index and the number of pages:
In the next step of `early_ioremap_pte` we check the given page flags with the `pgprot_val` macro and call `set_pte` or `pte_clear` depending on the flags given:
flags, so we call `set_pte` function to set the page table entry which works in the same manner as `set_pmd` but for PTEs (read above about it). As we have set all `PTEs` in the loop, we can now take a look at the call of the `__flush_tlb_one` function:
This function is defined in [arch/x86/include/asm/tlbflush.h](https://github.com/torvalds/linux/blob/16f73eb02d7e1765ccab3d2018e0bd98eb93d973) and calls `__flush_tlb_single` or `__flush_tlb` depending on the value of `cpu_has_invlpg`:
The `__flush_tlb_one` function invalidates the given address in the [TLB](http://en.wikipedia.org/wiki/Translation_lookaside_buffer). As you just saw we updated the paging structure, but `TLB` is not informed of the changes, that's why we need to do it manually. There are two ways to do it. The first is to update the `cr3` control register and the `__flush_tlb` function does this:
The second method is to use the `invlpg` instruction to invalidate the `TLB` entry. Let's look at the `__flush_tlb_one` implementation. As you can see, first of all the function checks `cpu_has_invlpg` which is defined as:
or call `__flush_tlb` which just updates the `cr3` register as we have seen. After this step execution of the `__early_set_fixmap` function is finished and we can go back to the `__early_ioremap` implementation. When we have set up the fixmap area for the given address, we need to save the base virtual address of the I/O Re-mapped area in the `prev_map` using the `slot` index:
The second function, `early_iounmap`, unmaps an `I/O` memory region. This function takes two parameters: base address and size of a `I/O` region and generally looks very similar to `early_ioremap`. It also goes through fixmap slots and looks for a slot with the given address. After that, it gets the index of the fixmap slot and calls `__late_clear_fixmap` or `__early_set_fixmap` depending on the `after_paging_init` value. It calls `__early_set_fixmap` with one difference to how `early_ioremap` does it: `early_iounmap` passes `zero` as physical address. And in the end it sets the address of the I/O memory region to `NULL`:
That's all about `fixmaps` and `ioremap`. Of course this part does not cover all features of `ioremap`, only early ioremap but there is also normal ioremap. But we need to know more things before we study that in more detail.
This is the end of the second part about linux kernel memory management. If you have questions or suggestions, ping me on twitter [0xAX](https://twitter.com/0xAX), drop me an [email](anotherworldofworld@gmail.com) or just create an [issue](https://github.com/0xAX/linux-insides/issues/new).
**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 a PR to [linux-insides](https://github.com/0xAX/linux-insides).**
