michaelh/zephyr
1
0
Fork 0
Code Releases Activity

mmu: backing stores reserve page fault room

Browse source 
If we evict enough pages to completely fill the backing store,
through APIs like k_mem_map(), z_page_frame_evict(), or
z_mem_page_out(), this will produce a crash the next time we
try to handle a page fault.

The backing store now always reserves a free storage location
for actual page faults.

Signed-off-by: Andrew Boie <andrew.p.boie@intel.com>
This commit is contained in:
Andrew Boie 2021-01-15 12:07:45 -08:00 • committed by Anas Nashif
commit c7be5dddda
4 changed files with 62 additions and 11 deletions
Download patch file Download diff file Expand all files Collapse all files

15
subsys/demand_paging/backing_store/ram.c
View file

@ -51,6 +51,7 @@
static char backing_store[CONFIG_MMU_PAGE_SIZE *
CONFIG_BACKING_STORE_RAM_PAGES];
static struct k_mem_slab backing_slabs;
static unsigned int free_slabs;
static void *location_to_slab(uintptr_t location)
{
@ -78,17 +79,21 @@ static uintptr_t slab_to_location(void *slab)
return offset;
}
int z_backing_store_location_get(struct z_page_frame *pf,
uintptr_t *location)
int z_backing_store_location_get(struct z_page_frame *pf, uintptr_t *location,
bool page_fault)
{
int ret;
void *slab;
ret = k_mem_slab_alloc(&backing_slabs, &slab, K_NO_WAIT);
if (ret != 0) {
if ((!page_fault && free_slabs == 1) || free_slabs == 0) {
return -ENOMEM;
}
ret = k_mem_slab_alloc(&backing_slabs, &slab, K_NO_WAIT);
__ASSERT(ret == 0, "slab count mismatch");
(void)ret;
*location = slab_to_location(slab);
free_slabs--;
return 0;
}
@ -98,6 +103,7 @@ void z_backing_store_location_free(uintptr_t location)
void *slab = location_to_slab(location);
k_mem_slab_free(&backing_slabs, &slab);
free_slabs++;
}
void z_backing_store_page_out(uintptr_t location)
@ -121,4 +127,5 @@ void z_backing_store_init(void)
{
k_mem_slab_init(&backing_slabs, backing_store, CONFIG_MMU_PAGE_SIZE,
CONFIG_BACKING_STORE_RAM_PAGES);
free_slabs = CONFIG_BACKING_STORE_RAM_PAGES;
}