VM_PAGE_ALLOC(9)

✨ NAME ✨

Okay so bestie, welcome to VM_PAGE_ALLOC(9) from the Kernel Developer's Manual — the page that explains vm_page_alloc and its whole friend group: allocate a page of memory. 💖

---

🧚 SYNOPSIS 🧚

#include <sys/param.h>
#include <vm/vm.h>
#include <vm/vm_page.h>

vm_page_t
vm_page_alloc(vm_object_t object, vm_pindex_t pindex, int req);

vm_page_t
vm_page_alloc_after(vm_object_t object, vm_pindex_t pindex, int req,
    vm_page_t mpred);

vm_page_t
vm_page_alloc_contig(vm_object_t object, vm_pindex_t pindex, int req,
    u_long npages, vm_paddr_t low, vm_paddr_t high, u_long alignment,
    vm_paddr_t boundary, vm_memattr_t memattr);

vm_page_t
vm_page_alloc_contig_domain(vm_object_t object, vm_pindex_t pindex,
    int req, u_long npages, vm_paddr_t low, vm_paddr_t high,
    u_long alignment, vm_paddr_t boundary, vm_memattr_t memattr);

vm_page_t
vm_page_alloc_domain(vm_object_t object, vm_pindex_t pindex, int domain,
    int req);

vm_page_t
vm_page_alloc_domain_after(vm_object_t object, vm_pindex_t pindex,
    int domain, int req, vm_page_t mpred);

vm_page_t
vm_page_alloc_noobj(int req);

vm_page_t
vm_page_alloc_noobj_contig(int req, u_long npages, vm_paddr_t low,
    vm_paddr_t high, u_long alignment, vm_paddr_t boundary,
    vm_memattr_t memattr);

vm_page_t
vm_page_alloc_noobj_contig_domain(int domain, int req, u_long npages,
    vm_paddr_t low, vm_paddr_t high, u_long alignment,
    vm_paddr_t boundary, vm_memattr_t memattr);

vm_page_t
vm_page_alloc_noobj_domain(int domain, int req);

---

💖 DESCRIPTION 💖

The vm_page_alloc() family allocates one or more pages of physical memory. Lowkey iconic, but also: most kernel code should not call these functions directly. Instead it should use a kernel memory allocator such as malloc(9) or uma(9), or use a higher-level interface to the page cache, such as vm_page_grab(9). Slay responsibly. 💅

All of these functions take a req parameter that encodes the allocation priority and optional modifier flags — yes, the ones described below. The functions whose names do not include noobj additionally insert the pages starting at index pindex in the VM object object. That object must be write-locked and must not already have a page resident at the specified index. No sneaky duplicates, periodt.

The functions whose names include domain support NUMA-aware allocation by returning pages from the numa(4) domain specified by domain. Fancy, localized, main character memory energy. ✨

vm_page_alloc_after() and vm_page_alloc_domain_after() behave identically to vm_page_alloc() and vm_page_alloc_domain(), respectively, except they take an additional parameter mpred. That mpred must be the page resident in object with the largest index smaller than pindex, or NULL if no such page exists. These functions exist to optimize the common case of loops that allocate multiple pages at successive indices within an object. The way that is such a specific little optimization, I love it. 🫶

vm_page_alloc_contig() and vm_page_alloc_noobj_contig(), plus their NUMA-aware variants, allocate a physically contiguous run of npages pages that satisfies the specified constraints. The low and high parameters specify a physical address range from which the run is to be allocated. The alignment parameter specifies the requested alignment of the first page in the run and must be a power of two. If the boundary parameter is non-zero, the pages constituting the run will not cross a physical address that is a multiple of the parameter value, which must be a power of two. If memattr is not equal to VM_MEMATTR_DEFAULT, then mappings of the returned pages created by, e.g., pmap_enter(9) or pmap_qenter(9), will carry the machine-dependent encoding of the memory attribute. Additionally, the direct mapping of the page, if any, will be updated to reflect the requested memory attribute. It's giving “physically contiguous with boundaries and vibes.” 🦋

---

🔥 REQUEST FLAGS 🔥

All page allocator functions accept a req parameter that governs certain aspects of the function's behavior. This is the part where req is basically the group chat’s settings panel. 💀

The VM_ALLOC_WAITOK, VM_ALLOC_WAITFAIL, and VM_ALLOC_NOWAIT flags specify the behavior of the allocator if free pages could not be immediately allocated. The VM_ALLOC_WAITOK flag can only be used with the noobj variants. If VM_ALLOC_NOWAIT is specified, then the allocator gives up and returns NULL. VM_ALLOC_NOWAIT is specified implicitly if none of the flags are present in the request. If either VM_ALLOC_WAITOK or VM_ALLOC_WAITFAIL is specified, the allocator will put the calling thread to sleep until sufficient free pages become available. At this point, if VM_ALLOC_WAITFAIL is specified the allocator will return NULL, and if VM_ALLOC_WAITOK is specified the allocator will retry the allocation. After a failed VM_ALLOC_WAITFAIL allocation returns, the VM object, if any, will have been unlocked while the thread was sleeping. In this case the VM object write lock will be re-acquired before the function call returns. Locking drama resolved before return — we love a clean exit. 🔒

req also encodes the allocation request priority. By default the page(s) are allocated with no special treatment. If the number of available free pages is below a certain watermark, the allocation will fail or the allocating thread will sleep, depending on the specified wait flag. The watermark is computed at boot time and corresponds to a small (less than one percent) fraction of the system's total physical memory. To allocate memory more aggressively, one of following flags may be specified:

• VM_ALLOC_SYSTEM — The page can be allocated if the free page count is above the interrupt reserved water mark. This flag should be used only when the system really needs the page.
• VM_ALLOC_INTERRUPT — The allocation will fail only if zero free pages are available. This flag should be used only if the consequences of an allocation failure are worse than leaving the system without free memory. For example, this flag is used when allocating kernel page table pages, where allocation failures trigger a kernel panic.

The following optional flags can further modify allocator behavior:

• VM_ALLOC_SBUSY — The returned page will be shared-busy. This flag may only be specified when allocating pages in a VM object.
• VM_ALLOC_NOBUSY — The returned page will not be busy. This flag is implicit when allocating pages without a VM object. When allocating pages in a VM object, and neither VM_ALLOC_SBUSY nor VM_ALLOC_NOBUSY are specified, the returned pages will be exclusively busied.
• VM_ALLOC_NODUMP — The returned page will not be included in any kernel core dumps regardless of whether or not it is mapped in to KVA.
• VM_ALLOC_WIRED — The returned page will be wired.
• VM_ALLOC_ZERO — If this flag is specified, the noobj variants will return zeroed pages. The other allocator interfaces ignore this flag.
• VM_ALLOC_NORECLAIM — If this flag is specified and the request can not be immediately satisfied, the allocator will not attempt to break superpage reservations to satisfy the allocation. This may be useful when the overhead of scanning the reservation queue outweighs the cost of a failed allocation. This flag may be used only with the contig variants, and must not be specified in combination with VM_ALLOC_WAITOK.
• VM_ALLOC_COUNT(n) — Hint that at least n pages will be allocated by the caller in the near future. n must be no larger than 65535. If the system is short of free pages, this hint may cause the kernel to reclaim memory more aggressively than it would otherwise.
• VM_ALLOC_NOFREE — The caller asserts that the returned page will never be released. If this flag is specified, the allocator will try to fetch a page from a special per-domain arena in order to curb long-term physical memory fragmentation.

---

🎀 RETURN VALUES 🎀

If the allocation was successful, a pointer to the struct vm_page corresponding to the allocated page is returned. If the allocation request specified multiple pages, the returned pointer points to an array of struct vm_page constituting the run. Upon failure, NULL is returned. Regardless of whether the allocation succeeds or fails, the VM object object will be write-locked upon return. So bestie, either way, the lock situation is handled. 💖

---

🌈 SEE ALSO 🌈

numa(4), malloc(9), uma(9), vm_page_grab(9), vm_page_sbusy(9)

---

👑 AUTHORS 👑

This manual page was written by Chad David <davidc@acns.ab.ca>.

---

Mwah 💋 this page is from Debian, dated August 4, 2024 — source/version and date noted, and now you're all caught up, bestie. Go forth and allocate responsibly. 🐧💖✨