On 02/10/18 21:23, Daniel Mack wrote:
Hi Miquel,
On 2/10/2018 9:25 AM, Miquel Raynal wrote:
Miquel Raynal miquel.raynal@bootlin.com wrote on Tue, 2 Oct 2018 08:46:01 +0200:
Boris Brezillon boris.brezillon@bootlin.com wrote on Tue, 2 Oct 2018 00:13:28 +0200:
Also, it looks like complete() is not called until the RDDREQ, WRDREQ and WRCMDREQ are cleared in the interrupt handler [1], which is weird. Miquel, do you happen to remember why you had to do that?
The RDDREQ, WRDREQ and WRCMDREQ events might potentially happen while the interrupts are enabled while we only wait for R/B signalling. This check is to avoid calling complete() on these situations.
Actually Boris is right on the fact that while the intention is good, the writing of [1] is not accurate. Daniel, could you please test if the following diff changes something with your setup, without your patch?
diff --git a/drivers/mtd/nand/raw/marvell_nand.c b/drivers/mtd/nand/raw/marvell_nand.c index bc2ef5209783..c7573ccdbacd 100644 --- a/drivers/mtd/nand/raw/marvell_nand.c +++ b/drivers/mtd/nand/raw/marvell_nand.c @@ -686,7 +686,7 @@ static irqreturn_t marvell_nfc_isr(int irq, void *dev_id) marvell_nfc_disable_int(nfc, st & NDCR_ALL_INT);
if (!(st & (NDSR_RDDREQ | NDSR_WRDREQ | NDSR_WRCMDREQ)))
if (st & (NDSR_RDY(0) | NDSR_RDY(1))) complete(&nfc->complete);Yes! That seems to work nicely as a replacement for my patch.
Chris, how is that going on your board?
Looks good to me. I've just re-run stress_{1,2,3} on the custom board and DB-88F6820-AMC.