[coreboot-gerrit] Change in coreboot[master]: cbmem: Add custom aligned memcpy() implementation
Martin Roth (Code Review)
gerrit at coreboot.org
Fri Apr 14 16:33:29 CEST 2017
Martin Roth has submitted this change and it was merged. ( https://review.coreboot.org/18300 )
Change subject: cbmem: Add custom aligned memcpy() implementation
......................................................................
cbmem: Add custom aligned memcpy() implementation
On some architectures (like AArch64), /dev/mem mappings outside of the
area marked as normal RAM use a memory type that does not support
unaligned accesses. The libc memcpy() implementation on these
architectures may not know or expect that and make an unaligned access
for certain source/dest/length alignments. Add a custom memcpy()
implementation that takes these restrictions into account and use it
anywhere we copy straight out of /dev/mem memory.
Change-Id: I03eece380a14a69d4be3805ed72fba640f6f7d9c
Signed-off-by: Julius Werner <jwerner at chromium.org>
Reviewed-on: https://review.coreboot.org/18300
Tested-by: build bot (Jenkins)
Reviewed-by: Aaron Durbin <adurbin at chromium.org>
---
M util/cbmem/cbmem.c
1 file changed, 31 insertions(+), 1 deletion(-)
Approvals:
Aaron Durbin: Looks good to me, approved
build bot (Jenkins): Verified
diff --git a/util/cbmem/cbmem.c b/util/cbmem/cbmem.c
index 7b434d87..e605052 100644
--- a/util/cbmem/cbmem.c
+++ b/util/cbmem/cbmem.c
@@ -61,6 +61,36 @@
static size_t lbtable_size;
/*
+ * Some architectures map /dev/mem memory in a way that doesn't support
+ * unaligned accesses. Most normal libc memcpy()s aren't safe to use in this
+ * case, so build our own which makes sure to never do unaligned accesses on
+ * *src (*dest is fine since we never map /dev/mem for writing).
+ */
+static void *aligned_memcpy(void *dest, const void *src, size_t n)
+{
+ u8 *d = dest;
+ const volatile u8 *s = src; /* volatile to prevent optimization */
+
+ while ((uintptr_t)s & (sizeof(size_t) - 1)) {
+ if (n-- == 0)
+ return dest;
+ *d++ = *s++;
+ }
+
+ while (n >= sizeof(size_t)) {
+ *(size_t *)d = *(const volatile size_t *)s;
+ d += sizeof(size_t);
+ s += sizeof(size_t);
+ n -= sizeof(size_t);
+ }
+
+ while (n-- > 0)
+ *d++ = *s++;
+
+ return dest;
+}
+
+/*
* calculate ip checksum (16 bit quantities) on a passed in buffer. In case
* the buffer length is odd last byte is excluded from the calculation
*/
@@ -608,7 +638,7 @@
console_p = map_memory_size((unsigned long)console.cbmem_addr,
size + sizeof(size) + sizeof(cursor), 1);
- memcpy(console_c, console_p + 8, size);
+ aligned_memcpy(console_c, console_p + 8, size);
printf("%s\n", console_c);
if (size < cursor)
--
To view, visit https://review.coreboot.org/18300
To unsubscribe, visit https://review.coreboot.org/settings
Gerrit-MessageType: merged
Gerrit-Change-Id: I03eece380a14a69d4be3805ed72fba640f6f7d9c
Gerrit-PatchSet: 5
Gerrit-Project: coreboot
Gerrit-Branch: master
Gerrit-Owner: Julius Werner <jwerner at chromium.org>
Gerrit-Reviewer: Aaron Durbin <adurbin at chromium.org>
Gerrit-Reviewer: Furquan Shaikh <furquan at google.com>
Gerrit-Reviewer: Julius Werner <jwerner at chromium.org>
Gerrit-Reviewer: Martin Roth <martinroth at google.com>
Gerrit-Reviewer: Patrick Georgi <pgeorgi at google.com>
Gerrit-Reviewer: Patrick Rudolph <siro at das-labor.org>
Gerrit-Reviewer: Paul Menzel <paulepanter at users.sourceforge.net>
Gerrit-Reviewer: Philippe Mathieu-Daudé <philippe.mathieu.daude at gmail.com>
Gerrit-Reviewer: build bot (Jenkins)
More information about the coreboot-gerrit
mailing list