On Tue, 3 Feb 2015 14:38:02 -0800 Tony Luck <tony.luck@xxxxxxxxx> wrote:
> Try to allocate all boot time kernel data structures from mirrored
> memory. If we run out of mirrored memory print warnings, but fall
> back to using non-mirrored memory to make sure that we still boot.
>
> ...
>
> diff --git a/include/linux/memblock.h b/include/linux/memblock.h
> index 1d448879caae..20bf3dfab564 100644
> --- a/include/linux/memblock.h
> +++ b/include/linux/memblock.h
> @@ -22,6 +22,7 @@
>
> /* Definition of memblock flags. */
> #define MEMBLOCK_HOTPLUG 0x1 /* hotpluggable region */
> +#define MEMBLOCK_MIRROR 0x2 /* mirrored region */
It would be nice to make these an enum. Then all those literal "0"'s
which were added in [1/3] become MEMBLOCK_NONE, which is
self-documenting.
>
> ...
>
> +static inline bool memblock_is_mirror(struct memblock_region *m)
> +{
> + return m->flags & MEMBLOCK_MIRROR;
> +}
> +
>
> ...
>
> +u32 __init_memblock memblock_has_mirror(void)
> +{
> + return memblock_have_mirror ? MEMBLOCK_MIRROR : 0;
> +}
hm, these are very similar. But I guess they're different enough.
Gramatically, a function called "memblock_has_mirror()" should return a
bool. This guy is misnamed. "memblock_mirror_flag()"?
> /* inline so we don't get a warning when pr_debug is compiled out */
> static __init_memblock const char *
> memblock_type_name(struct memblock_type *type)
> @@ -257,8 +263,19 @@ phys_addr_t __init_memblock memblock_find_in_range(phys_addr_t start,
> phys_addr_t end, phys_addr_t size,
> phys_addr_t align)
> {
> - return memblock_find_in_range_node(size, align, start, end,
> + phys_addr_t ret;
> + u32 flag = memblock_has_mirror();
> +
> + ret = memblock_find_in_range_node(size, align, start, end,
> + NUMA_NO_NODE, flag);
> +
> + if (!ret && flag) {
> + pr_warn("Could not allocate %lld bytes of mirrored memory\n", size);
This printk will warn on some configs. Print a phys_addr_t with %pap.
I think. See huge comment over lib/vsprintf.c:pointer(). There are
other instances of this.
> + ret = memblock_find_in_range_node(size, align, start, end,
> NUMA_NO_NODE, 0);
> + }
> +
> + return ret;
> }
>
> ...
>
> phys_addr_t __init memblock_alloc_nid(phys_addr_t size, phys_addr_t align, int nid)
> {
> - return memblock_alloc_base_nid(size, align, MEMBLOCK_ALLOC_ACCESSIBLE, nid, 0);
> + u32 flag = memblock_has_mirror();
> + phys_addr_t ret;
> +
> +again:
> + ret = memblock_alloc_base_nid(size, align, MEMBLOCK_ALLOC_ACCESSIBLE, nid, flag);
> +
> + if (!ret && flag) {
> + flag = 0;
> + goto again;
> + }
What's going on here? This is where we're falling back to
non-mirrored. But it's happening silently? Should it warn, or is that
handled elsewhere?
This function isn't specific to mirrored memory - for any future flags,
falling back to flags==0 may not be the desired behavior. What do we
do then? I guess
if (!ret && (flag & MEMBLOCK_MIRROR)) (
flag &= ~MEMBLOCK_MIRROR;
goto again;
yes?
That can be done later if needed, I suppose.
> + return ret;
> }
>
>
> ...
>
> @@ -1181,13 +1232,13 @@ static void * __init memblock_virt_alloc_internal(
>
> again:
> alloc = memblock_find_in_range_node(size, align, min_addr, max_addr,
> - nid, 0);
> + nid, flag);
> if (alloc)
> goto done;
>
> if (nid != NUMA_NO_NODE) {
> alloc = memblock_find_in_range_node(size, align, min_addr,
> - max_addr, NUMA_NO_NODE, 0);
> + max_addr, NUMA_NO_NODE, flag);
> if (alloc)
> goto done;
> }
> @@ -1195,10 +1246,15 @@ again:
> if (min_addr) {
> min_addr = 0;
> goto again;
> - } else {
> - goto error;
> }
>
> + if (flag) {
> + flag = 0;
> + pr_warn("Could not allocate %lld bytes of mirrored memory\n", size);
printk warning.
Please don't torture people who use 80-col displays!
> + goto again;
> + }
> +
> + return NULL;
>
> ...
>
> @@ -37,11 +37,19 @@ static void * __init __alloc_memory_core_early(int nid, u64 size, u64 align,
> {
> void *ptr;
> u64 addr;
> + u32 flag = memblock_has_mirror();
>
> if (limit > memblock.current_limit)
> limit = memblock.current_limit;
>
> - addr = memblock_find_in_range_node(size, align, goal, limit, nid, 0);
> +again:
> + addr = memblock_find_in_range_node(size, align, goal, limit, nid, flag);
> +
> + if (flag && !addr) {
> + flag = 0;
> + pr_warn("Could not allocate %lld bytes of mirrored memory\n", size);
dittoes.
> + goto again;
> + }
> if (!addr)
> return NULL;
>
--
To unsubscribe, send a message with 'unsubscribe linux-mm' in
the body to majordomo@xxxxxxxxx. For more info on Linux MM,
see: http://www.linux-mm.org/ .
Don't email: <a href=mailto:"dont@xxxxxxxxx";> email@xxxxxxxxx </a>