Allow for controller errors when reading

Controller errors while reading can be caused by corruption caused by a previous powerloss. Because upper level software like JFFS2 treat -EIO as fatal, it is necessary to treat the error as an ECC error.

---
drivers/mtd/onenand/onenand_base.c | 10 ++++++++++

1 files changed, 10 insertions(+), 0 deletions(-)

index 15a62db..daf1965 100644
@@ -329,6 +329,16 @@ static int onenand_wait(struct mtd_info *mtd, int state)

		printk(KERN_ERR "onenand_wait: controller error = 0x%04x\n", ctrl);
		if (ctrl & ONENAND_CTRL_LOCK)
			printk(KERN_ERR "onenand_wait: it's locked error.\n");

+ if (state == FL_READING) {
+ /*
+ * Controller errors while reading can be caused by
+ * corruption caused by a previous powerloss. Because
+ * upper level software like JFFS2 treat -EIO as fatal,
+ * it is necessary to treat the error as an ECC error.
+ */
+ mtd->ecc_stats.failed++;
+ return -EBADMSG;
+ }

		return -EIO;
	}

1.5.2.5

Hi,

Umm interesting. You mean it occurs controller error before ecc error at power loss case.
And it needs to define the terminology the Sudden Power Off Recovery or POR shortly.
I'm not sure a previous powerloss means POR.

Anyway could you show the controller error value? Since it needs to distinguish others e.g., initial bad blocks condition.

Thank you,
Kyungmin Park

No it is not what I meant. I have sent another patch which I hope is clearer.

The controller error value is still printed.

The bbt scan has its own wait function - I guess it is possible to get a controller error while reading oob from an eraseblock that is not actually a bad block. I am not sure how it would be possible to tell the difference.

0x2440

Since it needs to