All of lore.kernel.org
 help / color / mirror / Atom feed
From: Dmitry Lebed <lebed.dmitry@gmail.com>
To: linux-amlogic@lists.infradead.org
Cc: linux-mmc@vger.kernel.org, hkallweit1@gmail.com,
	martin.blumenstingl@googlemail.com,
	Dmitry Lebed <lebed.dmitry@gmail.com>
Subject: [PATCH] mmc: meson-gx: check for scatterlist size alignment in block mode
Date: Thu, 17 Dec 2020 23:53:12 -0800	[thread overview]
Message-ID: <20201218075312.67338-1-lebed.dmitry@gmail.com> (raw)

Enable SGDMA support for SD_IO_RW_EXTENDED and add proper check
for scatterlist size alignment in block mode.

According to documentation, in SDIO block mode meson-gx DMA could
only handle buffers with sizes that are multiples of SDIO block size.

Some SDIO drivers like brcmfmac use scatterlist API, but do not enforce
proper scatterlist buffer size alignemnt, this looks like a root cause
of non-working CMD53.

Some minor style fixes.

Signed-off-by: Dmitry Lebed <lebed.dmitry@gmail.com>
---
 drivers/mmc/host/meson-gx-mmc.c | 37 ++++++++++++++++++++-------------
 1 file changed, 22 insertions(+), 15 deletions(-)

diff --git a/drivers/mmc/host/meson-gx-mmc.c b/drivers/mmc/host/meson-gx-mmc.c
index 13f6a2c0ed04..eb6c02bc4a02 100644
--- a/drivers/mmc/host/meson-gx-mmc.c
+++ b/drivers/mmc/host/meson-gx-mmc.c
@@ -227,7 +227,6 @@ static void meson_mmc_get_transfer_mode(struct mmc_host *mmc,
 	struct mmc_data *data = mrq->data;
 	struct scatterlist *sg;
 	int i;
-	bool use_desc_chain_mode = true;
 
 	/*
 	 * When Controller DMA cannot directly access DDR memory, disable
@@ -237,25 +236,33 @@ static void meson_mmc_get_transfer_mode(struct mmc_host *mmc,
 	if (host->dram_access_quirk)
 		return;
 
-	/*
-	 * Broken SDIO with AP6255-based WiFi on Khadas VIM Pro has been
-	 * reported. For some strange reason this occurs in descriptor
-	 * chain mode only. So let's fall back to bounce buffer mode
-	 * for command SD_IO_RW_EXTENDED.
-	 */
-	if (mrq->cmd->opcode == SD_IO_RW_EXTENDED)
-		return;
+	if (data->blocks > 1) {
+		/*
+		 * In block mode DMA descriptor format, "length" field indicates
+		 * number of blocks and there is no way to pass DMA size that
+		 * is not multiple of SDIO block size, making it impossible to
+		 * tie more than one memory buffer with single SDIO block.
+		 * Block mode sg buffer size should be aligned with SDIO block
+		 * size, otherwise chain mode could not be used.
+		 */
+		for_each_sg(data->sg, sg, data->sg_len, i) {
+			if (sg->length % data->blksz) {
+				WARN_ONCE(1, "unaligned sg len %u blksize %u\n",
+					  sg->length, data->blksz);
+				return;
+			}
+		}
+	}
 
-	for_each_sg(data->sg, sg, data->sg_len, i)
+	for_each_sg(data->sg, sg, data->sg_len, i) {
 		/* check for 8 byte alignment */
-		if (sg->offset & 7) {
+		if (sg->offset % 8) {
 			WARN_ONCE(1, "unaligned scatterlist buffer\n");
-			use_desc_chain_mode = false;
-			break;
+			return;
 		}
+	}
 
-	if (use_desc_chain_mode)
-		data->host_cookie |= SD_EMMC_DESC_CHAIN_MODE;
+	data->host_cookie |= SD_EMMC_DESC_CHAIN_MODE;
 }
 
 static inline bool meson_mmc_desc_chain_mode(const struct mmc_data *data)
-- 
2.24.3 (Apple Git-128)


WARNING: multiple messages have this Message-ID (diff)
From: Dmitry Lebed <lebed.dmitry@gmail.com>
To: linux-amlogic@lists.infradead.org
Cc: martin.blumenstingl@googlemail.com,
	Dmitry Lebed <lebed.dmitry@gmail.com>,
	linux-mmc@vger.kernel.org, hkallweit1@gmail.com
Subject: [PATCH] mmc: meson-gx: check for scatterlist size alignment in block mode
Date: Thu, 17 Dec 2020 23:53:12 -0800	[thread overview]
Message-ID: <20201218075312.67338-1-lebed.dmitry@gmail.com> (raw)

Enable SGDMA support for SD_IO_RW_EXTENDED and add proper check
for scatterlist size alignment in block mode.

According to documentation, in SDIO block mode meson-gx DMA could
only handle buffers with sizes that are multiples of SDIO block size.

Some SDIO drivers like brcmfmac use scatterlist API, but do not enforce
proper scatterlist buffer size alignemnt, this looks like a root cause
of non-working CMD53.

Some minor style fixes.

Signed-off-by: Dmitry Lebed <lebed.dmitry@gmail.com>
---
 drivers/mmc/host/meson-gx-mmc.c | 37 ++++++++++++++++++++-------------
 1 file changed, 22 insertions(+), 15 deletions(-)

diff --git a/drivers/mmc/host/meson-gx-mmc.c b/drivers/mmc/host/meson-gx-mmc.c
index 13f6a2c0ed04..eb6c02bc4a02 100644
--- a/drivers/mmc/host/meson-gx-mmc.c
+++ b/drivers/mmc/host/meson-gx-mmc.c
@@ -227,7 +227,6 @@ static void meson_mmc_get_transfer_mode(struct mmc_host *mmc,
 	struct mmc_data *data = mrq->data;
 	struct scatterlist *sg;
 	int i;
-	bool use_desc_chain_mode = true;
 
 	/*
 	 * When Controller DMA cannot directly access DDR memory, disable
@@ -237,25 +236,33 @@ static void meson_mmc_get_transfer_mode(struct mmc_host *mmc,
 	if (host->dram_access_quirk)
 		return;
 
-	/*
-	 * Broken SDIO with AP6255-based WiFi on Khadas VIM Pro has been
-	 * reported. For some strange reason this occurs in descriptor
-	 * chain mode only. So let's fall back to bounce buffer mode
-	 * for command SD_IO_RW_EXTENDED.
-	 */
-	if (mrq->cmd->opcode == SD_IO_RW_EXTENDED)
-		return;
+	if (data->blocks > 1) {
+		/*
+		 * In block mode DMA descriptor format, "length" field indicates
+		 * number of blocks and there is no way to pass DMA size that
+		 * is not multiple of SDIO block size, making it impossible to
+		 * tie more than one memory buffer with single SDIO block.
+		 * Block mode sg buffer size should be aligned with SDIO block
+		 * size, otherwise chain mode could not be used.
+		 */
+		for_each_sg(data->sg, sg, data->sg_len, i) {
+			if (sg->length % data->blksz) {
+				WARN_ONCE(1, "unaligned sg len %u blksize %u\n",
+					  sg->length, data->blksz);
+				return;
+			}
+		}
+	}
 
-	for_each_sg(data->sg, sg, data->sg_len, i)
+	for_each_sg(data->sg, sg, data->sg_len, i) {
 		/* check for 8 byte alignment */
-		if (sg->offset & 7) {
+		if (sg->offset % 8) {
 			WARN_ONCE(1, "unaligned scatterlist buffer\n");
-			use_desc_chain_mode = false;
-			break;
+			return;
 		}
+	}
 
-	if (use_desc_chain_mode)
-		data->host_cookie |= SD_EMMC_DESC_CHAIN_MODE;
+	data->host_cookie |= SD_EMMC_DESC_CHAIN_MODE;
 }
 
 static inline bool meson_mmc_desc_chain_mode(const struct mmc_data *data)
-- 
2.24.3 (Apple Git-128)


_______________________________________________
linux-amlogic mailing list
linux-amlogic@lists.infradead.org
http://lists.infradead.org/mailman/listinfo/linux-amlogic

             reply	other threads:[~2020-12-18  7:54 UTC|newest]

Thread overview: 8+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2020-12-18  7:53 Dmitry Lebed [this message]
2020-12-18  7:53 ` [PATCH] mmc: meson-gx: check for scatterlist size alignment in block mode Dmitry Lebed
2020-12-18  9:08 ` Heiner Kallweit
2020-12-18  9:08   ` Heiner Kallweit
2021-01-04  9:19   ` Jerome Brunet
2021-01-04  9:19     ` Jerome Brunet
2021-01-13 11:25 ` Ulf Hansson
2021-01-13 11:25   ` Ulf Hansson

Reply instructions:

You may reply publicly to this message via plain-text email
using any one of the following methods:

* Save the following mbox file, import it into your mail client,
  and reply-to-all from there: mbox

  Avoid top-posting and favor interleaved quoting:
  https://en.wikipedia.org/wiki/Posting_style#Interleaved_style

* Reply using the --to, --cc, and --in-reply-to
  switches of git-send-email(1):

  git send-email \
    --in-reply-to=20201218075312.67338-1-lebed.dmitry@gmail.com \
    --to=lebed.dmitry@gmail.com \
    --cc=hkallweit1@gmail.com \
    --cc=linux-amlogic@lists.infradead.org \
    --cc=linux-mmc@vger.kernel.org \
    --cc=martin.blumenstingl@googlemail.com \
    /path/to/YOUR_REPLY

  https://kernel.org/pub/software/scm/git/docs/git-send-email.html

* If your mail client supports setting the In-Reply-To header
  via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line before the message body.
This is an external index of several public inboxes,
see mirroring instructions on how to clone and mirror
all data and code used by this external index.