Since 2018.11, the pattern of bootcount load and store has changed such that bootcount_inc() and bootcount_error() will load, increment, store, and load *again*. The second load exposes a timing problem whereby the i2c read fails. The root cause _seems_ to relate to performance of the 24c08 EEPROM. Insert a delay after writing to work around this.
Signed-off-by: Ian Ray ian.ray@ge.com --- drivers/bootcount/bootcount_i2c.c | 4 ++++ 1 file changed, 4 insertions(+)
diff --git a/drivers/bootcount/bootcount_i2c.c b/drivers/bootcount/bootcount_i2c.c index ed22389..dd86426 100644 --- a/drivers/bootcount/bootcount_i2c.c +++ b/drivers/bootcount/bootcount_i2c.c @@ -64,6 +64,10 @@ void bootcount_store(ulong a) CONFIG_BOOTCOUNT_ALEN, buf, sizeof(buf)); if (ret != 0) puts("Error writing bootcount\n"); + else + /* Datasheet claims 2 ms typical write cycle time. + * This delay ensures that next read will succeed. */ + mdelay(5);
bootcount_set_bus_back(prev_i2c_bus); }