Suppose to call __get_free_page in order to request the kernel a free page in ZONE_NORMAL zone, so with GFP_KERNEL flag. We know that the returned address is a kernel logical address so there exists a "permanent" mapping in the page table for that logical address and the physical address is calculated with the macro __pa. My question is: if exists a permanent mapping for that page, is the swapping to the disk possible for that page? If yes, i suppose that when it will be swapped in memory again, it will put every time in the same page frame...is this correct?
If you allocate a page in the kernel with __get_free_page in the kernel, it is not swappable. In Linux, a struct page tracks a physical page, which is not swappable. Swapping works on the level of userspace mapping and their contents; so if userspace memory is swapped out, the contents may be swapped back into a different physical page. But that physical page will be mapped at the same userspace virtual address.
Kernel memory is not swappable in Linux.
Related
Virtual address to physical page mapping can be changed during application runtime by swapping or physical page reallocation for memory defragmentation or etc.
What if I want to cache physical page numbers (PPNs) of some virtual address range from /proc/PID/pagemap, since accessing proc/PID/pagemap is extremely expansive overhead to be checked every time, is there a way to be notified if a page has been moved to other physical address or swapped, on the effective address or just any part of memory?
Any kind of method will be ok(not just userspace method, but also those that can be only implemented in kernel space).
Brief Version:
what status are the addresses unpresented in the maps file? Are they belongs to unallocated virtual pages or allocated from anonymous file or others?
Detailed Version
I'm learning about VM. In my book(CS:APP), I learned that all virtual pages can be cut into three sets: unallocated, allocated but not cached, allocated and cached.I have some questions about "what are allocated pages and unallocated pages? When are pages allocated?" And also, is stack and heap belongs to allocated pages or unallocated or only allocate when used?
Trying to solve these problems, I read the /proc/$pid/maps file, while I think I can get anything I want from it. In my mind, the file contains all memory mapping relations. But there isn't information about is it cached(I know maybe it cannot be seen from user mode...), and are the unpresented pages unallocated?
Honestly, I don't really know about the maps file. What I do know is that the information on every page is stored in page structures at all time. I'm gonna take x86-64 as an example.
For x86-64 on Linux you have Page Global Directory (PGD), Page Upper Directory (PUD), Page Middle Directory (PMD) and Page Directory (PD). The address of the bottom of the PGD table is stored in the CR3 register. PGD contains addresses of the PUDs, PUDs contain addresses of the PMDs, PMDs contain addresses of the PDs and PDs contain addresses of the physical pages.
A virtual address, of which only 48 bits are used, is split into 5 parts. The 12 least significant bits are the offset in the physical page. The next chunk of 9 bits is the offset in the PD. The next chunk is the offset in PMD etc. For example let's say you have virtual address 0x0000000000000123. This virtual address will be translated by the MMU in the CPU by looking at entry (offset) 0 of the PGD, entry 0 of the PUD, entry 0 of the PMD, entry 0 of the PD and finally offset 0x123 in the actual physical page in RAM. Every virtual address is 64 bits of which only the 48 least significant bits will be used.
At boot, the kernel makes checks to determine how much memory is available. It then builds its kernel structures accordingly.
When the kernel boots it will mark all pages as unallocated in its own structures (except for kernel needs). The page structure is important for this. The kernel has a page C structure for every page in the system (https://linux-kernel-labs.github.io/refs/heads/master/labs/memory_mapping.html and https://elixir.bootlin.com/linux/v4.6/source/include/linux/mm_types.h). This structure informs the kernel whether the page is allocated or not.
Each physical page in the system has a struct page associated with
it to keep track of whatever it is we are using the page for at the
moment. Note that we have no way to track which tasks are using
a page, though if it is a pagecache page, rmap structures can tell us
who is mapping it.
At first the pages are mostly unallocated. When you start a new process by launching an executable as the user of the system, pages are allocated for your process. On Linux, executables are ELF files. ELF is a conventional format which separates code in loadable segments. Each segment gets a virtual address at which it is going to be loaded in the virtual address space.
Let's say you have an elf file with one segment which should be loaded at virtual address 0x400000. When you launch that ELF executable, the Linux kernel will call certain functions which will look at the size of the code and allocate pages accordingly. The kernel will look at its structures and determine using algorithms where the process will be allocated in RAM. It will then setup the page tables according to where the virtual addresses for that process should land in actual physical memory.
What's important to understand is that each CPU core in the system has only one process running at a time. Each core has it's own set of page tables. When a process switch occurs for one core, the page tables are swapped completely to point to somewhere else in RAM. The same virtual address can point anywhere in RAM depending on how the page tables are set up.
The kernel holds a task_struct for every process running in the system. The task_struct contains a field named pgd which is a pointer to the PGD of the process. Each process has its very own PGD. If you dereference the pointer to the PGD you get the actual value of the first entry of the PGD. This first entry is the address of the PUD. With this only pointer, the kernel can reach every table belonging to the process and modify them at will.
While a process is running, it can ask for more memory. This is called dynamic memory allocation. The kernel has no way to know how much memory the process is going to ask in advance since it is dynamic (done while code is executing). When the process asks for more memory, the kernel determines what page to give to the process depending on an algorithm. It then marks this page as allocated to that process. The task_struct contains a mm field which is of type mm_struct (https://manybutfinite.com/post/how-the-kernel-manages-your-memory/). It is a memory descriptor for that process so that the kernel can know what memory the process is using. In itself the process doesn't need that information since the process should rely only on itself to ask for memory properly to the operating system and to not jump somewhere in RAM where it doesn't belong.
You ask about heap and stack. The stack for a process is allocated at the beginning of the process and I think it has a fixed size. If you overflow the stack, you will throw a CPU exception which will prompt the kernel to kill your process. Each CPU core has a special register called RSP. This is the stack pointer. It points to the top of the stack (the stack grows downward toward low memory). When the kernel allocates a stack for a process you launch, it will set up this register to point at the top of it. The stack pointer contains a virtual address. It will thus be translated using the page tables just like any address.
The heap is allocated and managed completely by the OS. It doesn't have special registers like the stack. It is allocated only when the process asks for more memory during code execution. The kernel knows in advance how much memory a process requires. It is all written in the ELF executable. All static memory is allocated during compilation and thus the kernel knows everything about the size of static memory. The only moment it requires to allocate new memory to a process is when the process actually asks for it. In C++ you use the keyword new to ask for heap memory dynamically. If you don't use this keyword, then the kernel knows in advance where your variables will be allocated (where they will be in memory). Only the stack will be used by static memory.
On the surface, this appears to be a silly question. Some patience please.. :-)
Am structuring this qs into 2 parts:
Part 1:
I fully understand that platform RAM is mapped into the kernel segment; esp on 64-bit systems this will work well. So each kernel virtual address is indeed just an offset from physical memory (DRAM).
Also, it's my understanding that as Linux is a modern virtual memory OS, (pretty much) all addresses are treated as virtual addresses and must "go" via hardware - the TLB/MMU - at runtime and then get translated by the TLB/MMU via kernel paging tables. Again, easy to understand for user-mode processes.
HOWEVER, what about kernel virtual addresses? For efficiency, would it not be simpler to direct-map these (and an identity mapping is indeed setup from PAGE_OFFSET onwards). But still, at runtime, the kernel virtual address must go via the TLB/MMU and get translated right??? Is this actually the case? Or is kernel virtual addr translation just an offset calculation?? (But how can that be, as we must go via hardware TLB/MMU?). As a simple example, lets consider:
char *kptr = kmalloc(1024, GFP_KERNEL);
Now kptr is a kernel virtual address.
I understand that virt_to_phys() can perform the offset calculation and return the physical DRAM address.
But, here's the Actual Question: it can't be done in this manner via software - that would be pathetically slow! So, back to my earlier point: it would have to be translated via hardware (TLB/MMU).
Is this actually the case??
Part 2:
Okay, lets say this is the case, and we do use paging in the kernel to do this, we must of course setup kernel paging tables; I understand it's rooted at swapper_pg_dir.
(I also understand that vmalloc() unlike kmalloc() is a special case- it's a pure virtual region that gets backed by physical frames only on page fault).
If (in Part 1) we do conclude that kernel virtual address translation is done via kernel paging tables, then how exactly does the kernel paging table (swapper_pg_dir) get "attached" or "mapped" to a user-mode process?? This should happen in the context-switch code? How? Where?
Eg.
On an x86_64, 2 processes A and B are alive, 1 cpu.
A is running, so it's higher-canonical addr
0xFFFF8000 00000000 through 0xFFFFFFFF FFFFFFFF "map" to the kernel segment, and it's lower-canonical addr
0x0 through 0x00007FFF FFFFFFFF map to it's private userspace.
Now, if we context-switch A->B, process B's lower-canonical region is unique But
it must "map" to the same kernel of course!
How exactly does this happen? How do we "auto" refer to the kernel paging table when
in kernel mode? Or is that a wrong statement?
Thanks for your patience, would really appreciate a well thought out answer!
First a bit of background.
This is an area where there is a lot of potential variation between
architectures, however the original poster has indicated he is mainly
interested in x86 and ARM, which share several characteristics:
no hardware segments or similar partitioning of the virtual address space (when used by Linux)
hardware page table walk
multiple page sizes
physically tagged caches (at least on modern ARMs)
So if we restrict ourselves to those systems it keeps things simpler.
Once the MMU is enabled, it is never normally turned off. So all CPU
addresses are virtual, and will be translated to physical addresses
using the MMU. The MMU will first look up the virtual address in the
TLB, and only if it doesn't find it in the TLB will it refer to the
page table - the TLB is a cache of the page table - and so we can
ignore the TLB for this discussion.
The page table
describes the entire virtual 32 or 64 bit address space, and includes
information like:
whether the virtual address is valid
which mode(s) the processor must be in for it to be valid
special attributes for things like memory mapped hardware registers
and the physical address to use
Linux divides the virtual address space into two: the lower portion is
used for user processes, and there is a different virtual to physical
mapping for each process. The upper portion is used for the kernel,
and the mapping is the same even when switching between different user
processes. This keep things simple, as an address is unambiguously in
user or kernel space, the page table doesn't need to be changed when
entering or leaving the kernel, and the kernel can simply dereference
pointers into user space for the
current user process. Typically on 32bit processors the split is 3G
user/1G kernel, although this can vary. Pages for the kernel portion
of the address space will be marked as accessible only when the processor
is in kernel mode to prevent them being accessible to user processes.
The portion of the kernel address space which is identity mapped to RAM
(kernel logical addresses) will be mapped using big pages when possible,
which may allow the page table to be smaller but more importantly
reduces the number of TLB misses.
When the kernel starts it creates a single page table for itself
(swapper_pg_dir) which just describes the kernel portion of the
virtual address space and with no mappings for the user portion of the
address space. Then every time a user process is created a new page
table will be generated for that process, the portion which describes
kernel memory will be the same in each of these page tables. This could be
done by copying all of the relevant portion of swapper_pg_dir, but
because page tables are normally a tree structures, the kernel is
frequently able to graft the portion of the tree which describes the
kernel address space from swapper_pg_dir into the page tables for each
user process by just copying a few entries in the upper layer of the
page table structure. As well as being more efficient in memory (and possibly
cache) usage, it makes it easier to keep the mappings consistent. This
is one of the reasons why the split between kernel and user virtual
address spaces can only occur at certain addresses.
To see how this is done for a particular architecture look at the
implementation of pgd_alloc(). For example ARM
(arch/arm/mm/pgd.c) uses:
pgd_t *pgd_alloc(struct mm_struct *mm)
{
...
init_pgd = pgd_offset_k(0);
memcpy(new_pgd + USER_PTRS_PER_PGD, init_pgd + USER_PTRS_PER_PGD,
(PTRS_PER_PGD - USER_PTRS_PER_PGD) * sizeof(pgd_t));
...
}
or
x86 (arch/x86/mm/pgtable.c) pgd_alloc() calls pgd_ctor():
static void pgd_ctor(struct mm_struct *mm, pgd_t *pgd)
{
/* If the pgd points to a shared pagetable level (either the
ptes in non-PAE, or shared PMD in PAE), then just copy the
references from swapper_pg_dir. */
...
clone_pgd_range(pgd + KERNEL_PGD_BOUNDARY,
swapper_pg_dir + KERNEL_PGD_BOUNDARY,
KERNEL_PGD_PTRS);
...
}
So, back to the original questions:
Part 1: Are kernel virtual addresses really translated by the TLB/MMU?
Yes.
Part 2: How is swapper_pg_dir "attached" to a user mode process.
All page tables (whether swapper_pg_dir or those for user processes)
have the same mappings for the portion used for kernel virtual
addresses. So as the kernel context switches between user processes,
changing the current page table, the mappings for the kernel portion
of the address space remain the same.
The kernel address space is mapped to a section of each process for example on 3:1 mapping after address 0xC0000000. If the user code try to access this address space it will generate a page fault and it is guarded by the kernel.
The kernel address space is divided into 2 parts, the logical address space and the virtual address space. It is defined by the constant VMALLOC_START. The CPU is using the MMU all the time, in user space and in kernel space (can't switch on/off).
The kernel virtual address space is mapped the same way as user space mapping. The logical address space is continuous and it is simple to translate it to physical so it can be done on demand using the MMU fault exception. That is the kernel is trying to access an address, the MMU generate fault , the fault handler map the page using macros __pa , __va and change the CPU pc register back to the previous instruction before the fault happened, now everything is ok. This process is actually platform dependent and in some hardware architectures it mapped the same way as user (because the kernel doesn't use a lot of memory).
The linear address beyond 896MB correspond to High memory region ZONE_HIGHMEM.
So the page allocator functions will not work on this region, since they give the linear address of directly mapped page frames in ZONE_NORMAL and ZONE_DMA.
I am confused about these lines specified in Undertanding linux Kernel:
What do they mean when they say "In 64 bit hardware platforms ZONE_HIGHMEM is always empty."
What does this highlighted statement mean: "The allocation of high-memory page frames is done only through alloc_pages() function. These functions do not return linear address since they do not exist. Instead the functions return linear address of the page descriptor of the first allocated page frame. These linear addresses always exist, because all page descriptors are allocated in low memory once and forever during kernel initialization."
What are these Page descriptors and does the 896MB already have all page descriptors of entire RAM.
The x86-32 kernel needs high memory to access more than 1G of physical memory, as it is impossible to permanently map more than 2^{32} addresses within a 32-bit address space and the kernel/user split is 1G/3G.
The x86-64 kernel has no such limitation, as the amount of physically-addressable memory (currently 256T) fits within its 64-bit address space and thus may always be permanently mapped.
High memory is a hack. Ideally you don't need it. Indeed, the point of x86-64 is to be able to directly address all the memory you could possibly want. Taken
from https://www.quora.com/Linux-Kernel/What-is-the-difference-between-high-memory-and-normal-memory
I think page descriptor means struct page. And considering the sizeof struct page. Yes all of them can be stored in ZONE_NORMAL
Ioremap is done in kernel mode.
Does ioremap create an entry in pagetables during
the call of ioremap or when the address is accessed ?
For ioremap, the page tables are updated immediately. Since the function is intended to map physical "I/O" addresses into kernel virtual address space, there would be no point in setting up page table entries to cause a page fault. In essence, a page fault supports the dynamic substitution of one page of physical memory for another. But by definition, I/O space is not substitutable.
For example, one common use of ioremap is to allow a kernel module to access register space on an add-on card or other peripheral. In order to perform operations on the card, the kernel code must obtain a virtual address that refers to the physical bus address corresponding to the card's register area. ioremap causes the allocation of virtual space, and the establishment of a mapping from that virtual range to the card space. It wouldn't make sense to "swap" the underlying physical page: that page isn't a real memory page, it has special functions that can't be duplicated by any other physical memory.