[SLOF] [PATCH v2] rtas-nvram: optimize erase

Nikunj A Dadhania nikunj at linux.vnet.ibm.com
Mon May 9 21:13:37 AEST 2016


As this was done at byte granularity, erasing complete nvram(64K
default) took a lot of time. To reduce the number of rtas call per byte
write which is expensive, the erase is done at one shot using the
nvram_buffer that is initiated during the nvram_init call for
RTAS_NVRAM.

After this patch there is ~450msec improvement during boot. Default qemu
booting does not provide file backed nvram, so every boot there would be
full erase of 64K.

Before this patch:

real	0m2.214s
user	0m0.015s
sys	  0m0.006s

real	0m2.222s
user	0m0.014s
sys	  0m0.005s

real	0m2.201s
user	0m0.010s
sys	  0m0.005s

After this patch:

real	0m1.762s
user	0m0.014s
sys	  0m0.006s

real	0m1.773s
user	0m0.011s
sys   0m0.004s

real	0m1.754s
user	0m0.013s
sys	  0m0.005s

Signed-off-by: Nikunj A Dadhania <nikunj at linux.vnet.ibm.com>
---
 lib/libnvram/nvram.c | 11 +++++++++++
 1 file changed, 11 insertions(+)

diff --git a/lib/libnvram/nvram.c b/lib/libnvram/nvram.c
index 473814e..99deb2a 100644
--- a/lib/libnvram/nvram.c
+++ b/lib/libnvram/nvram.c
@@ -373,6 +373,17 @@ void erase_nvram(int offset, int len)
 {
 	int i;
 
+#ifdef RTAS_NVRAM
+	char *erase_buf = get_nvram_buffer(len);
+	if (erase_buf) {
+		/* Speed up by erasing all memory at once */
+		memset(erase_buf, 0, len);
+		nvram_store(offset, erase_buf, len);
+		free_nvram_buffer(erase_buf);
+		return;
+	}
+	/* If get_nvram_buffer failed, fall through to default code */
+#endif
 	for (i=offset; i<offset+len; i++)
 		nvram_write_byte(i, 0);
 }
-- 
2.5.5



More information about the SLOF mailing list