Skip to content

core: mm: don't walk translation tables for every mapped page #7947

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Merged
jenswikl merged 1 commit into from
Aug 28, 2026
Merged

core: mm: don't walk translation tables for every mapped page #7947

Changes from all commits
Commits
Show all changes
1 commit
Select commit
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
47 changes: 37 additions & 10 deletions core/mm/core_mmu.c
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -2064,11 +2064,27 @@ void core_mmu_map_region(struct mmu_partition *prtn, struct tee_mmap_region *mm)
}
}

/*
* The leaf translation table covering a virtual address only changes every
* CORE_MMU_PGDIR_SIZE. A loop mapping one small page at a time can therefore
* reuse the table found for the previous page instead of walking the
* translation tables from the base table again for every page.
*
* Returns true if @ti still describes the table covering @va. Note that @va
* below @ti->va_base wraps the unsigned subtraction into a large value and
* thus correctly reports the table as not covering @va.
*/
static bool tbl_info_covers_va(struct core_mmu_table_info *ti, vaddr_t va)
{
return ti->table &&
va - ti->va_base < BIT64(ti->shift) * ti->num_entries;
}

TEE_Result core_mmu_map_pages(vaddr_t vstart, paddr_t *pages, size_t num_pages,
enum teecore_memtypes memtype)
{
TEE_Result ret;
struct core_mmu_table_info tbl_info;
struct core_mmu_table_info tbl_info = { };
struct tee_mmap_region *mm;
unsigned int idx;
uint32_t old_attr;
Expand Down Expand Up @@ -2099,21 +2115,25 @@ TEE_Result core_mmu_map_pages(vaddr_t vstart, paddr_t *pages, size_t num_pages,
goto err;
}

while (true) {
while (!tbl_info_covers_va(&tbl_info, vaddr)) {
if (!core_mmu_find_table(NULL, vaddr, UINT_MAX,
&tbl_info))
panic("Can't find pagetable for vaddr ");

idx = core_mmu_va2idx(&tbl_info, vaddr);
if (tbl_info.shift == SMALL_PAGE_SHIFT)
break;

/* This is supertable. Need to divide it. */
idx = core_mmu_va2idx(&tbl_info, vaddr);
if (!core_mmu_entry_to_finer_grained(&tbl_info, idx,
secure))
panic("Failed to spread pgdir on small tables");

/* Force a new walk to reach the divided table */
tbl_info.table = NULL;
}

idx = core_mmu_va2idx(&tbl_info, vaddr);
core_mmu_get_entry(&tbl_info, idx, NULL, &old_attr);
if (old_attr)
panic("Page is already mapped");
Expand Down Expand Up @@ -2172,21 +2192,25 @@ TEE_Result core_mmu_map_contiguous_pages(vaddr_t vstart, paddr_t pstart,
panic("Trying to map into static region");

for (i = 0; i < num_pages; i++) {
while (true) {
while (!tbl_info_covers_va(&tbl_info, vaddr)) {
if (!core_mmu_find_table(NULL, vaddr, UINT_MAX,
&tbl_info))
panic("Can't find pagetable for vaddr ");

idx = core_mmu_va2idx(&tbl_info, vaddr);
if (tbl_info.shift == SMALL_PAGE_SHIFT)
break;

/* This is supertable. Need to divide it. */
idx = core_mmu_va2idx(&tbl_info, vaddr);
if (!core_mmu_entry_to_finer_grained(&tbl_info, idx,
secure))
panic("Failed to spread pgdir on small tables");

/* Force a new walk to reach the divided table */
tbl_info.table = NULL;
}

idx = core_mmu_va2idx(&tbl_info, vaddr);
core_mmu_get_entry(&tbl_info, idx, NULL, &old_attr);
if (old_attr)
panic("Page is already mapped");
Expand Down Expand Up @@ -2274,7 +2298,7 @@ static void maybe_remove_from_mem_map(vaddr_t vstart, size_t num_pages)

void core_mmu_unmap_pages(vaddr_t vstart, size_t num_pages)
{
struct core_mmu_table_info tbl_info;
struct core_mmu_table_info tbl_info = { };
size_t i;
unsigned int idx;
uint32_t exceptions;
Expand All @@ -2284,11 +2308,14 @@ void core_mmu_unmap_pages(vaddr_t vstart, size_t num_pages)
maybe_remove_from_mem_map(vstart, num_pages);

for (i = 0; i < num_pages; i++, vstart += SMALL_PAGE_SIZE) {
if (!core_mmu_find_table(NULL, vstart, UINT_MAX, &tbl_info))
panic("Can't find pagetable");
if (!tbl_info_covers_va(&tbl_info, vstart)) {
if (!core_mmu_find_table(NULL, vstart, UINT_MAX,
&tbl_info))
panic("Can't find pagetable");

if (tbl_info.shift != SMALL_PAGE_SHIFT)
panic("Invalid pagetable level");
if (tbl_info.shift != SMALL_PAGE_SHIFT)
panic("Invalid pagetable level");
}

idx = core_mmu_va2idx(&tbl_info, vstart);
core_mmu_set_entry(&tbl_info, idx, 0, 0);
Expand Down
Loading