Received-SPF: pass (google.com: domain of linux-kernel+bounces-181419-linux.lists.archive=gmail.com@vger.kernel.org designates 139.178.88.99 as permitted sender) client-ip=139.178.88.99;
Date: Thu, 16 May 2024 20:46:23 +0300
From: Andy Shevchenko <andriy.shevchenko@linux.intel.com>
To: =?iso-8859-1?Q?N=EDcolas_F=2E_R=2E_A=2E?= Prado <nfraprado@collabora.com>
Cc: Mark Brown <broonie@kernel.org>, linux-spi@vger.kernel.org,
	linux-kernel@vger.kernel.org
Subject: Re: [PATCH v1 1/1] spi: Remove unneded check for orig_nents
Message-ID: <ZkZGbz0RlUHshneC@smile.fi.intel.com>
References: <20240507201028.564630-1-andriy.shevchenko@linux.intel.com>
 <d8930bce-6db6-45f4-8f09-8a00fa48e607@notapiano>
 <ZkXdXO4Xb83270V7@smile.fi.intel.com>
 <ZkYJTxmlM5oWOzFL@smile.fi.intel.com>
 <2fccdd9a-5b97-4dc6-a6b1-ce2d9e0819bd@notapiano>
Precedence: bulk
MIME-Version: 1.0
Content-Type: text/plain; charset=iso-8859-1
Content-Disposition: inline
Content-Transfer-Encoding: 8bit
In-Reply-To: <2fccdd9a-5b97-4dc6-a6b1-ce2d9e0819bd@notapiano>
Organization: Intel Finland Oy - BIC 0357606-4 - Westendinkatu 7, 02160 Espoo

On Thu, May 16, 2024 at 12:25:19PM -0400, N?colas F. R. A. Prado wrote:
> On Thu, May 16, 2024 at 04:25:35PM +0300, Andy Shevchenko wrote:
> > On Thu, May 16, 2024 at 01:18:04PM +0300, Andy Shevchenko wrote:
> > > On Wed, May 15, 2024 at 05:09:33PM -0400, N?colas F. R. A. Prado wrote:
> > > > On Tue, May 07, 2024 at 11:10:27PM +0300, Andy Shevchenko wrote:
> > > > > Both dma_unmap_sgtable() and sg_free_table() in spi_unmap_buf_attrs()
> > > > > have checks for orig_nents against 0. No need to duplicate this.
> > > > > All the same applies to other DMA mapping API calls.
> > > > > 
> > > > > Also note, there is no other user in the kernel that does this kind of
> > > > > checks.
> > > > > 
> > > > > Signed-off-by: Andy Shevchenko <andriy.shevchenko@linux.intel.com>
> > > > 
> > > > this commit caused a regression which I reported here:
> > > > 
> > > > https://lore.kernel.org/all/d3679496-2e4e-4a7c-97ed-f193bd53af1d@notapiano
> > > > 
> > > > along with some thoughts on the cause and a possible solution, though I'm not
> > > > familiar with this code base at all and would really appreciate any feedback you
> > > > may have.
> > > 
> > > Thanks for the report and preliminary analysis!
> > > I'll look at it hopefully sooner than later.
> > > 
> > > But at least what I think now is that my change revealed a problem somewhere
> > > else, because that's how DMA mapping / streaming APIs designed, it's extremely
> > > rare to check orig_nents field.
> > 
> > Can you test the below patch?
> > 
> > diff --git a/drivers/spi/spi.c b/drivers/spi/spi.c
> > index b2efd4964f7c..51811f04e463 100644
> > --- a/drivers/spi/spi.c
> > +++ b/drivers/spi/spi.c
> > @@ -1243,6 +1243,7 @@ static int __spi_map_msg(struct spi_controller *ctlr, struct spi_message *msg)
> >  	else
> >  		rx_dev = ctlr->dev.parent;
> >  
> > +	ret = -ENOMSG;
> >  	list_for_each_entry(xfer, &msg->transfers, transfer_list) {
> >  		/* The sync is done before each transfer. */
> >  		unsigned long attrs = DMA_ATTR_SKIP_CPU_SYNC;
> > @@ -1272,6 +1273,9 @@ static int __spi_map_msg(struct spi_controller *ctlr, struct spi_message *msg)
> >  			}
> >  		}
> >  	}
> > +	/* No transfer has been mapped, bail out with success */
> > +	if (ret)
> > +		return 0;
> >  
> >  	ctlr->cur_rx_dma_dev = rx_dev;
> >  	ctlr->cur_tx_dma_dev = tx_dev;
> 
> Hi Andy,
> 
> thank you for the patch. Unfortunately it didn't completely solve the issue. Now
> the stack trace is slightly different and points at the next line:
> 
> 	dma_sync_sgtable_for_device(rx_dev, &xfer->rx_sg, DMA_FROM_DEVICE);
> 
> So now we're hitting the case where only the tx buffer was DMA mapped, but the
> rx is still uninitialized, though the cur_msg_mapped flag is set to true, since
> it is shared between them. The original code checked for the initialization of
> each scatterlist individually, which is why it worked.

I was kinda expecting that, and already have another patch to try (should
applied on top):

diff --git a/drivers/spi/spi.c b/drivers/spi/spi.c
index 51811f04e463..5c607dd21fe7 100644
--- a/drivers/spi/spi.c
+++ b/drivers/spi/spi.c
@@ -1258,6 +1258,8 @@ static int __spi_map_msg(struct spi_controller *ctlr, struct spi_message *msg)
 						attrs);
 			if (ret != 0)
 				return ret;
+		} else {
+			memset(&xfer->tx_sg, 0, sizeof(xfer->tx_sg));
 		}
 
 		if (xfer->rx_buf != NULL) {
@@ -1271,6 +1273,8 @@ static int __spi_map_msg(struct spi_controller *ctlr, struct spi_message *msg)
 
 				return ret;
 			}
+		} else {
+			memset(&xfer->rx_sg, 0, sizeof(xfer->rx_sg));
 		}
 	}
 	/* No transfer has been mapped, bail out with success */


If my understanding is correct, my patch revealed two issues:
- for non-mapped message at all;
- for unidirect transfers.

But I will wait for your test results to make the final conclusion.

-- 
With Best Regards,
Andy Shevchenko