summaryrefslogtreecommitdiff
path: root/firmware/target/arm/imx233/partitions-imx233.c
diff options
context:
space:
mode:
authorAmaury Pouly <amaury.pouly@gmail.com>2014-01-05 15:21:04 +0100
committerAmaury Pouly <amaury.pouly@gmail.com>2014-01-21 18:52:28 +0100
commit877eb7a112d10d8eb59c4abb50b4193894c32b10 (patch)
tree4ecac92df8f2137ba203f20ce13751cb9ab55557 /firmware/target/arm/imx233/partitions-imx233.c
parent69d0dae55bf3da2f356dbd03dba299c708133902 (diff)
imx233: handle quirk about OF partition address bug/feature
Change-Id: Ib8b854ce7132ac460672f9adf8611f580ca53746
Diffstat (limited to 'firmware/target/arm/imx233/partitions-imx233.c')
-rw-r--r--firmware/target/arm/imx233/partitions-imx233.c26
1 files changed, 17 insertions, 9 deletions
diff --git a/firmware/target/arm/imx233/partitions-imx233.c b/firmware/target/arm/imx233/partitions-imx233.c
index 60c36ebb4d..bb322162c9 100644
--- a/firmware/target/arm/imx233/partitions-imx233.c
+++ b/firmware/target/arm/imx233/partitions-imx233.c
@@ -125,14 +125,22 @@ static int compute_window_freescale(intptr_t user, part_read_fn_t read_fn,
* The first table uses 512-byte sector size and the second one usually uses
* 2048-byte logical sector size.
*
- * We restrict the window to the user partition
- *
* WARNING HACK FIXME BUG
- * Reverse engineering and experiments suggests that the OF ignores the lowest 2 bits
- * of the LBAs in the partition table. There is at least one example
- * (the Creative Zen X-Fi3) where this is important because the LBA of the user partition
- * is not a multiple of 4. The behaviour of the size field is less clear but
- * it seems that it is similarly truncated. */
+ * Reverse engineering suggests that OF has a notion of "internal sector size"
+ * which can either be 512 or 2048 bytes. When applied to the main data drive,
+ * it internally converts LBA from sectors to "internal sectors". Consequently,
+ * if the internal sector size is 2048 and the sector size of 512, it will
+ * drop the lowest 2 bits of the LBA from the partition table. This is the case
+ * of the ZEN X-Fi3 for example. However, if the internal sector size is 512
+ * bytes, then there is no such loss. This is the case of the Zen X-Fi Style
+ * for example.
+ * The behaviour of the size field is less clear but it seems that it is similarly
+ * truncated. */
+#if defined(CREATIVE_ZENXFISTYLE)
+#define DROP_MASK 0
+#else
+#define DROP_MASK 3
+#endif
if(mbr[510] != 0x55 || mbr[511] != 0xAA)
return -101; /* invalid MBR */
if(part == IMX233_PART_DATA)
@@ -141,9 +149,9 @@ static int compute_window_freescale(intptr_t user, part_read_fn_t read_fn,
uint8_t *ent = &mbr[446];
*start = ent[8] | ent[9] << 8 | ent[10] << 16 | ent[11] << 24;
/* ignore two lowest bits(see comment above) */
- *start &= ~3;
+ *start &= ~DROP_MASK;
*end = (ent[12] | ent[13] << 8 | ent[14] << 16 | ent[15] << 24);
- *end &= ~3;
+ *end &= ~DROP_MASK;
/* ignore two lowest bits(order is important, first truncate then add start) */
*end += *start;