fast-reboot: parallel memory clearing

Arbitrarily pick 16GB as the unit of parallelism, and
split up clearing memory into jobs and schedule them
node-local to the memory (or on node 0 if we can't
work that out because it's the memory up to SKIBOOT_BASE)

This seems to cut at least ~40% time from memory zeroing on
fast-reboot on a 256GB Boston system.

Signed-off-by: Stewart Smith <stewart@linux.ibm.com>

1 files changed, 4 insertions, 1 deletions

diff --git a/include/device.h b/include/device.h
index fd66df11..a447dcd4 100644
--- a/include/device.h
+++ b/include/device.h
@@ -237,9 +237,12 @@ u32 dt_n_size_cells(const struct dt_node *node);
 u64 dt_get_number(const void *pdata, unsigned int cells);
 
 /* Find an ibm,chip-id property in this node; if not found, walk up the parent
- * nodes. Returns -1 if no chip-id property exists. */
+ * nodes. */
 u32 dt_get_chip_id(const struct dt_node *node);
 
+/* Same as dt_get_chip_id except Returns -1 if no chip-id property exists. */
+u32 __dt_get_chip_id(const struct dt_node *node);
+
 /* Address accessors ("reg" properties parsing). No translation,
  * only support "simple" address forms (1 or 2 cells). Asserts
  * if address doesn't exist