[ 02/11] x86-32, mm: Rip out x86_32 NUMA remapping code

From: Greg Kroah-Hartman
Date: Sun Apr 14 2013 - 22:18:23 EST


3.0-stable review patch. If anyone has any objections, please let me know.

------------------

From: Dave Hansen <dave@xxxxxxxxxxxxxxxxxx>

commit f03574f2d5b2d6229dcdf2d322848065f72953c7 upstream.

[was already included in 3.0, but I missed the patch hunk for
arch/x86/mm/numa_32.c - gregkh]

This code was an optimization for 32-bit NUMA systems.

It has probably been the cause of a number of subtle bugs over
the years, although the conditions to excite them would have
been hard to trigger. Essentially, we remap part of the kernel
linear mapping area, and then sometimes part of that area gets
freed back in to the bootmem allocator. If those pages get
used by kernel data structures (say mem_map[] or a dentry),
there's no big deal. But, if anyone ever tried to use the
linear mapping for these pages _and_ cared about their physical
address, bad things happen.

For instance, say you passed __GFP_ZERO to the page allocator
and then happened to get handed one of these pages, it zero the
remapped page, but it would make a pte to the _old_ page.
There are probably a hundred other ways that it could screw
with things.

We don't need to hang on to performance optimizations for
these old boxes any more. All my 32-bit NUMA systems are long
dead and buried, and I probably had access to more than most
people.

This code is causing real things to break today:

https://lkml.org/lkml/2013/1/9/376

I looked in to actually fixing this, but it requires surgery
to way too much brittle code, as well as stuff like
per_cpu_ptr_to_phys().

[ hpa: Cc: this for -stable, since it is a memory corruption issue.
However, an alternative is to simply mark NUMA as depends BROKEN
rather than EXPERIMENTAL in the X86_32 subclause... ]

Link: http://lkml.kernel.org/r/20130131005616.1C79F411@xxxxxxxxxxxxxxxxxxxxxx
Signed-off-by: H. Peter Anvin <hpa@xxxxxxxxxxxxxxx>
Cc: Jiri Slaby <jslaby@xxxxxxx>
Signed-off-by: Greg Kroah-Hartman <gregkh@xxxxxxxxxxxxxxxxxxx>

---
arch/x86/mm/numa_32.c | 161 --------------------------------------------------
1 file changed, 161 deletions(-)

--- a/arch/x86/mm/numa_32.c
+++ b/arch/x86/mm/numa_32.c
@@ -73,167 +73,6 @@ unsigned long node_memmap_size_bytes(int

extern unsigned long highend_pfn, highstart_pfn;

-#define LARGE_PAGE_BYTES (PTRS_PER_PTE * PAGE_SIZE)
-
-static void *node_remap_start_vaddr[MAX_NUMNODES];
-void set_pmd_pfn(unsigned long vaddr, unsigned long pfn, pgprot_t flags);
-
-/*
- * Remap memory allocator
- */
-static unsigned long node_remap_start_pfn[MAX_NUMNODES];
-static void *node_remap_end_vaddr[MAX_NUMNODES];
-static void *node_remap_alloc_vaddr[MAX_NUMNODES];
-
-/**
- * alloc_remap - Allocate remapped memory
- * @nid: NUMA node to allocate memory from
- * @size: The size of allocation
- *
- * Allocate @size bytes from the remap area of NUMA node @nid. The
- * size of the remap area is predetermined by init_alloc_remap() and
- * only the callers considered there should call this function. For
- * more info, please read the comment on top of init_alloc_remap().
- *
- * The caller must be ready to handle allocation failure from this
- * function and fall back to regular memory allocator in such cases.
- *
- * CONTEXT:
- * Single CPU early boot context.
- *
- * RETURNS:
- * Pointer to the allocated memory on success, %NULL on failure.
- */
-void *alloc_remap(int nid, unsigned long size)
-{
- void *allocation = node_remap_alloc_vaddr[nid];
-
- size = ALIGN(size, L1_CACHE_BYTES);
-
- if (!allocation || (allocation + size) > node_remap_end_vaddr[nid])
- return NULL;
-
- node_remap_alloc_vaddr[nid] += size;
- memset(allocation, 0, size);
-
- return allocation;
-}
-
-#ifdef CONFIG_HIBERNATION
-/**
- * resume_map_numa_kva - add KVA mapping to the temporary page tables created
- * during resume from hibernation
- * @pgd_base - temporary resume page directory
- */
-void resume_map_numa_kva(pgd_t *pgd_base)
-{
- int node;
-
- for_each_online_node(node) {
- unsigned long start_va, start_pfn, nr_pages, pfn;
-
- start_va = (unsigned long)node_remap_start_vaddr[node];
- start_pfn = node_remap_start_pfn[node];
- nr_pages = (node_remap_end_vaddr[node] -
- node_remap_start_vaddr[node]) >> PAGE_SHIFT;
-
- printk(KERN_DEBUG "%s: node %d\n", __func__, node);
-
- for (pfn = 0; pfn < nr_pages; pfn += PTRS_PER_PTE) {
- unsigned long vaddr = start_va + (pfn << PAGE_SHIFT);
- pgd_t *pgd = pgd_base + pgd_index(vaddr);
- pud_t *pud = pud_offset(pgd, vaddr);
- pmd_t *pmd = pmd_offset(pud, vaddr);
-
- set_pmd(pmd, pfn_pmd(start_pfn + pfn,
- PAGE_KERNEL_LARGE_EXEC));
-
- printk(KERN_DEBUG "%s: %08lx -> pfn %08lx\n",
- __func__, vaddr, start_pfn + pfn);
- }
- }
-}
-#endif
-
-/**
- * init_alloc_remap - Initialize remap allocator for a NUMA node
- * @nid: NUMA node to initizlie remap allocator for
- *
- * NUMA nodes may end up without any lowmem. As allocating pgdat and
- * memmap on a different node with lowmem is inefficient, a special
- * remap allocator is implemented which can be used by alloc_remap().
- *
- * For each node, the amount of memory which will be necessary for
- * pgdat and memmap is calculated and two memory areas of the size are
- * allocated - one in the node and the other in lowmem; then, the area
- * in the node is remapped to the lowmem area.
- *
- * As pgdat and memmap must be allocated in lowmem anyway, this
- * doesn't waste lowmem address space; however, the actual lowmem
- * which gets remapped over is wasted. The amount shouldn't be
- * problematic on machines this feature will be used.
- *
- * Initialization failure isn't fatal. alloc_remap() is used
- * opportunistically and the callers will fall back to other memory
- * allocation mechanisms on failure.
- */
-void __init init_alloc_remap(int nid, u64 start, u64 end)
-{
- unsigned long start_pfn = start >> PAGE_SHIFT;
- unsigned long end_pfn = end >> PAGE_SHIFT;
- unsigned long size, pfn;
- u64 node_pa, remap_pa;
- void *remap_va;
-
- /*
- * The acpi/srat node info can show hot-add memroy zones where
- * memory could be added but not currently present.
- */
- printk(KERN_DEBUG "node %d pfn: [%lx - %lx]\n",
- nid, start_pfn, end_pfn);
-
- /* calculate the necessary space aligned to large page size */
- size = node_memmap_size_bytes(nid, start_pfn, end_pfn);
- size += ALIGN(sizeof(pg_data_t), PAGE_SIZE);
- size = ALIGN(size, LARGE_PAGE_BYTES);
-
- /* allocate node memory and the lowmem remap area */
- node_pa = memblock_find_in_range(start, end, size, LARGE_PAGE_BYTES);
- if (node_pa == MEMBLOCK_ERROR) {
- pr_warning("remap_alloc: failed to allocate %lu bytes for node %d\n",
- size, nid);
- return;
- }
- memblock_x86_reserve_range(node_pa, node_pa + size, "KVA RAM");
-
- remap_pa = memblock_find_in_range(min_low_pfn << PAGE_SHIFT,
- max_low_pfn << PAGE_SHIFT,
- size, LARGE_PAGE_BYTES);
- if (remap_pa == MEMBLOCK_ERROR) {
- pr_warning("remap_alloc: failed to allocate %lu bytes remap area for node %d\n",
- size, nid);
- memblock_x86_free_range(node_pa, node_pa + size);
- return;
- }
- memblock_x86_reserve_range(remap_pa, remap_pa + size, "KVA PG");
- remap_va = phys_to_virt(remap_pa);
-
- /* perform actual remap */
- for (pfn = 0; pfn < size >> PAGE_SHIFT; pfn += PTRS_PER_PTE)
- set_pmd_pfn((unsigned long)remap_va + (pfn << PAGE_SHIFT),
- (node_pa >> PAGE_SHIFT) + pfn,
- PAGE_KERNEL_LARGE);
-
- /* initialize remap allocator parameters */
- node_remap_start_pfn[nid] = node_pa >> PAGE_SHIFT;
- node_remap_start_vaddr[nid] = remap_va;
- node_remap_end_vaddr[nid] = remap_va + size;
- node_remap_alloc_vaddr[nid] = remap_va;
-
- printk(KERN_DEBUG "remap_alloc: node %d [%08llx-%08llx) -> [%p-%p)\n",
- nid, node_pa, node_pa + size, remap_va, remap_va + size);
-}
-
void __init initmem_init(void)
{
x86_numa_init();


--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at http://vger.kernel.org/majordomo-info.html
Please read the FAQ at http://www.tux.org/lkml/