[PATCH AUTOSEL 5.6 30/38] i2c: tegra: Better handle case where CPU0 is busy for a long time

[PATCH AUTOSEL 5.6 30/38] i2c: tegra: Better handle case where CPU0 is busy for a long time

[Sasha Levin]

From: Dmitry Osipenko <digetx-Re5JQEeQqe8AvxtiuMwx3w@public.gmane.org>

[ Upstream commit a900aeac253729411cf33c6cb598c152e9e4137f ]

Boot CPU0 always handle I2C interrupt and under some rare circumstances
(like running KASAN + NFS root) it may stuck in uninterruptible state for
a significant time. In this case we will get timeout if I2C transfer is
running on a sibling CPU, despite of IRQ being raised. In order to handle
this rare condition, the IRQ status needs to be checked after completion
timeout.

Signed-off-by: Dmitry Osipenko <digetx-Re5JQEeQqe8AvxtiuMwx3w@public.gmane.org>
Signed-off-by: Wolfram Sang <wsa-z923LK4zBo2bacvFa/9K2g@public.gmane.org>
Signed-off-by: Sasha Levin <sashal-DgEjT+Ai2ygdnm+yROfE0A@public.gmane.org>
---
 drivers/i2c/busses/i2c-tegra.c | 27 +++++++++++++++------------
 1 file changed, 15 insertions(+), 12 deletions(-)

diff --git a/drivers/i2c/busses/i2c-tegra.c b/drivers/i2c/busses/i2c-tegra.c
index cbc2ad49043e4..0daa863fb26f2 100644
--- a/drivers/i2c/busses/i2c-tegra.c
+++ b/drivers/i2c/busses/i2c-tegra.c
@@ -1000,14 +1000,13 @@ tegra_i2c_poll_completion_timeout(struct tegra_i2c_dev *i2c_dev,
 	do {
 		u32 status = i2c_readl(i2c_dev, I2C_INT_STATUS);
 
-		if (status) {
+		if (status)
 			tegra_i2c_isr(i2c_dev->irq, i2c_dev);
 
-			if (completion_done(complete)) {
-				s64 delta = ktime_ms_delta(ktimeout, ktime);
+		if (completion_done(complete)) {
+			s64 delta = ktime_ms_delta(ktimeout, ktime);
 
-				return msecs_to_jiffies(delta) ?: 1;
-			}
+			return msecs_to_jiffies(delta) ?: 1;
 		}
 
 		ktime = ktime_get();
@@ -1034,14 +1033,18 @@ tegra_i2c_wait_completion_timeout(struct tegra_i2c_dev *i2c_dev,
 		disable_irq(i2c_dev->irq);
 
 		/*
-		 * There is a chance that completion may happen after IRQ
-		 * synchronization, which is done by disable_irq().
+		 * Under some rare circumstances (like running KASAN +
+		 * NFS root) CPU, which handles interrupt, may stuck in
+		 * uninterruptible state for a significant time.  In this
+		 * case we will get timeout if I2C transfer is running on
+		 * a sibling CPU, despite of IRQ being raised.
+		 *
+		 * In order to handle this rare condition, the IRQ status
+		 * needs to be checked after timeout.
 		 */
-		if (ret == 0 && completion_done(complete)) {
-			dev_warn(i2c_dev->dev,
-				 "completion done after timeout\n");
-			ret = 1;
-		}
+		if (ret == 0)
+			ret = tegra_i2c_poll_completion_timeout(i2c_dev,
+								complete, 0);
 	}
 
 	return ret;
-- 
2.20.1

Re: [PATCH AUTOSEL 5.6 30/38] i2c: tegra: Better handle case where CPU0 is busy for a long time

[Thierry Reding]

[-- Attachment #1: Type: text/plain, Size: 1374 bytes --]

On Fri, Apr 24, 2020 at 08:22:28AM -0400, Sasha Levin wrote:
> From: Dmitry Osipenko <digetx-Re5JQEeQqe8AvxtiuMwx3w@public.gmane.org>
> 
> [ Upstream commit a900aeac253729411cf33c6cb598c152e9e4137f ]
> 
> Boot CPU0 always handle I2C interrupt and under some rare circumstances
> (like running KASAN + NFS root) it may stuck in uninterruptible state for
> a significant time. In this case we will get timeout if I2C transfer is
> running on a sibling CPU, despite of IRQ being raised. In order to handle
> this rare condition, the IRQ status needs to be checked after completion
> timeout.
> 
> Signed-off-by: Dmitry Osipenko <digetx-Re5JQEeQqe8AvxtiuMwx3w@public.gmane.org>
> Signed-off-by: Wolfram Sang <wsa-z923LK4zBo2bacvFa/9K2g@public.gmane.org>
> Signed-off-by: Sasha Levin <sashal-DgEjT+Ai2ygdnm+yROfE0A@public.gmane.org>
> ---
>  drivers/i2c/busses/i2c-tegra.c | 27 +++++++++++++++------------
>  1 file changed, 15 insertions(+), 12 deletions(-)

Hi Sasha,

can you drop this from the v5.6 stable queue please? Jon discovered that
this patch introduces a regression in v5.7, and since we don't have a
good understanding of why this is breaking things I think it'd be best
if we reverted it for v5.7 until we come up with a good fix.

I think the same applies for the other i2c/tegra patch that's 31/38 of
this series.

Thanks,
Thierry

[-- Attachment #2: signature.asc --]
[-- Type: application/pgp-signature, Size: 833 bytes --]

Re: [PATCH AUTOSEL 5.6 30/38] i2c: tegra: Better handle case where CPU0 is busy for a long time

[Sasha Levin]

On Mon, Apr 27, 2020 at 09:22:33AM +0200, Thierry Reding wrote:
>On Fri, Apr 24, 2020 at 08:22:28AM -0400, Sasha Levin wrote:
>> From: Dmitry Osipenko <digetx-Re5JQEeQqe8AvxtiuMwx3w@public.gmane.org>
>>
>> [ Upstream commit a900aeac253729411cf33c6cb598c152e9e4137f ]
>>
>> Boot CPU0 always handle I2C interrupt and under some rare circumstances
>> (like running KASAN + NFS root) it may stuck in uninterruptible state for
>> a significant time. In this case we will get timeout if I2C transfer is
>> running on a sibling CPU, despite of IRQ being raised. In order to handle
>> this rare condition, the IRQ status needs to be checked after completion
>> timeout.
>>
>> Signed-off-by: Dmitry Osipenko <digetx-Re5JQEeQqe8AvxtiuMwx3w@public.gmane.org>
>> Signed-off-by: Wolfram Sang <wsa-z923LK4zBo2bacvFa/9K2g@public.gmane.org>
>> Signed-off-by: Sasha Levin <sashal-DgEjT+Ai2ygdnm+yROfE0A@public.gmane.org>
>> ---
>>  drivers/i2c/busses/i2c-tegra.c | 27 +++++++++++++++------------
>>  1 file changed, 15 insertions(+), 12 deletions(-)
>
>Hi Sasha,
>
>can you drop this from the v5.6 stable queue please? Jon discovered that
>this patch introduces a regression in v5.7, and since we don't have a
>good understanding of why this is breaking things I think it'd be best
>if we reverted it for v5.7 until we come up with a good fix.
>
>I think the same applies for the other i2c/tegra patch that's 31/38 of
>this series.

I've dropped both of these, thanks!

-- 
Thanks,
Sasha

[PATCH AUTOSEL 5.6 31/38] i2c: tegra: Synchronize DMA before termination

[Sasha Levin]

From: Dmitry Osipenko <digetx-Re5JQEeQqe8AvxtiuMwx3w@public.gmane.org>

[ Upstream commit 8814044fe0fa182abc9ff818d3da562de98bc9a7 ]

DMA transfer could be completed, but CPU (which handles DMA interrupt)
may get too busy and can't handle the interrupt in a timely manner,
despite of DMA IRQ being raised. In this case the DMA state needs to
synchronized before terminating DMA transfer in order not to miss the
DMA transfer completion.

Signed-off-by: Dmitry Osipenko <digetx-Re5JQEeQqe8AvxtiuMwx3w@public.gmane.org>
Signed-off-by: Wolfram Sang <wsa-z923LK4zBo2bacvFa/9K2g@public.gmane.org>
Signed-off-by: Sasha Levin <sashal-DgEjT+Ai2ygdnm+yROfE0A@public.gmane.org>
---
 drivers/i2c/busses/i2c-tegra.c | 9 +++++++++
 1 file changed, 9 insertions(+)

diff --git a/drivers/i2c/busses/i2c-tegra.c b/drivers/i2c/busses/i2c-tegra.c
index 0daa863fb26f2..0c6dac770fc3a 100644
--- a/drivers/i2c/busses/i2c-tegra.c
+++ b/drivers/i2c/busses/i2c-tegra.c
@@ -1223,6 +1223,15 @@ static int tegra_i2c_xfer_msg(struct tegra_i2c_dev *i2c_dev,
 		time_left = tegra_i2c_wait_completion_timeout(
 				i2c_dev, &i2c_dev->dma_complete, xfer_time);
 
+		/*
+		 * Synchronize DMA first, since dmaengine_terminate_sync()
+		 * performs synchronization after the transfer's termination
+		 * and we want to get a completion if transfer succeeded.
+		 */
+		dmaengine_synchronize(i2c_dev->msg_read ?
+				      i2c_dev->rx_dma_chan :
+				      i2c_dev->tx_dma_chan);
+
 		dmaengine_terminate_sync(i2c_dev->msg_read ?
 					 i2c_dev->rx_dma_chan :
 					 i2c_dev->tx_dma_chan);
-- 
2.20.1