Re: meson-gx-mmc: 2nd patch set

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

 



On 05.04.2017 20:13, Heiner Kallweit wrote:
Am 05.04.2017 um 14:38 schrieb Helmut Klein:
On 05.04.2017 08:02, Heiner Kallweit wrote:
Am 04.04.2017 um 22:02 schrieb Helmut Klein:
On 04.04.2017 19:37, Heiner Kallweit wrote:
Am 03.04.2017 um 21:50 schrieb Helmut Klein:
On 29.03.2017 16:29, Helmut Klein wrote:
On 29.03.2017 07:46, Heiner Kallweit wrote:
Am 22.03.2017 um 11:09 schrieb Helmut Klein:
On 20.03.2017 20:54, Heiner Kallweit wrote:
Am 20.03.2017 um 14:01 schrieb Helmut Klein:
On 20.03.2017 13:51, Helmut Klein wrote:
On 19.03.2017 23:39, Heiner Kallweit wrote:
Am 19.03.2017 um 21:46 schrieb Helmut Klein:
On 19.03.2017 16:35, Heiner Kallweit wrote:
Am 19.03.2017 um 12:23 schrieb Helmut Klein:
On 18.03.2017 11:24, Heiner Kallweit wrote:
Am 18.03.2017 um 09:11 schrieb Helmut Klein:
On 17.03.2017 20:54, Heiner Kallweit wrote:
Am 17.03.2017 um 19:00 schrieb Helmut Klein:
On 17.03.2017 07:44, Heiner Kallweit wrote:
Am 16.03.2017 um 22:12 schrieb Helmut Klein:
Hallo Heiner,

i applied your 2nd patch sets to linux-amlogic. (and of
course the older set v5 1...10).

The performance of the sd-card and the mmc-chip of my
minimx
is much better now.

sd-card: from 6.5 to 22.5 MBytes/s
mmc: from 14.5 to 28.5 MBytes/s (hs200 compatible device)

But:
the wifi-adapter is no longer operational. The problem
starts
with patch 3 and gets worse with patch 5.

the adapter (ap6330) is attached to the sdio port of the
the
s905. (=mmc1)

Thanks for testing and sharing the feedback. Much
appreciated.
The board I test on supports block devices only.

Could you please share few more details:
- Drivers involved
- All syslog messages related to the device / driver

And please set the relevant mailing lists on cc when
providing
test feedback.
Others might find your feedback helpful too.

Thanks, Heiner

so i think it is important to test your patch set asap
on an
officially supported board with a wifi adapter).

regards
Helmut




The driver for the wifi adapter is brcmfmac. It is
compiled as
a module and loaded via /etc/modules.
the driver depends on the modules brcmutil and cfg80211,
which
are auto loaded.

my user space is debian unstable. The kernel is cross
compiled
on my x86-64 laptop with gcc 5.4.1

i use my own kernel configuration without an initial ramdisk.
mmc, usb & ext4 drivers are compiled into the kernel

kernel & dtb are loaded by u-boot from mmcblk2p1 (sd-card)

the drive for the kernel is sda2 (sd-card inside of a usb
card
reader).


here is the filtered (mmc|brcmfmac) syslog output when only
patches 1 & 2 are applied

[    1.685012] meson-gx-mmc d0074000.mmc: allocated
mmc-pwrseq
[    1.873010] mmc0: new DDR MMC card at address 0001
[    1.873258] mmcblk0: mmc0:0001 NCard  14.5 GiB
[    1.876731] mmcblk0boot0: mmc0:0001 NCard  partition 1
4.00 MiB
[    1.882597] mmcblk0boot1: mmc0:0001 NCard  partition 2
4.00 MiB
[    2.004399] meson-gx-mmc d0070000.mmc: allocated
mmc-pwrseq
[    2.048160] meson-gx-mmc d0072000.mmc: Got CD GPIO
[    2.066962] mmc1: queuing unknown CIS tuple 0x80 (2 bytes)
[    2.068985] mmc1: queuing unknown CIS tuple 0x80 (3 bytes)
[    2.074242] mmc1: queuing unknown CIS tuple 0x80 (3 bytes)
[    2.081275] mmc1: queuing unknown CIS tuple 0x80 (7 bytes)
[    2.124761] mmc2: Skipping voltage switch
[    2.153763] mmc1: new high speed SDIO card at address 0001
[    2.158567] mmc2: new high speed SDHC card at address 59b4
[    2.159219] mmcblk2: mmc2:59b4 USD00 29.5 GiB
[    2.164206]  mmcblk2: p1 p2
[    6.664118] brcmfmac: brcmf_c_preinit_dcmds: Firmware
version = wl0: Jan  6 2014 15:11:29 version 5.90.195.89.13
FWID
01-72f124c5
[    6.736239] brcmfmac: brcmf_cfg80211_reg_notifier: not a
ISO3166 code (0x30 0x30)
[    6.928045] brcmfmac: brcmf_construct_chaninfo: Ignoring
unexpected firmware channel 184
[    6.930475] brcmfmac: brcmf_construct_chaninfo: Ignoring
unexpected firmware channel 188
[    6.938524] brcmfmac: brcmf_construct_chaninfo: Ignoring
unexpected firmware channel 192
[    6.946543] brcmfmac: brcmf_construct_chaninfo: Ignoring
unexpected firmware channel 196
[    6.954566] brcmfmac: brcmf_construct_chaninfo: Ignoring
unexpected firmware channel 200
[    6.954663] brcmfmac: brcmf_cfg80211_reg_notifier: not a
ISO3166 code (0x30 0x30)
[    6.969995] brcmfmac: brcmf_construct_chaninfo: Ignoring
unexpected firmware channel 204
[    6.978023] brcmfmac: brcmf_construct_chaninfo: Ignoring
unexpected firmware channel 208

this is the output after adding patch 3

[    1.684982] meson-gx-mmc d0074000.mmc: allocated
mmc-pwrseq
[    1.873041] mmc0: new DDR MMC card at address 0001
[    1.873273] mmcblk0: mmc0:0001 NCard  14.5 GiB
[    1.876765] mmcblk0boot0: mmc0:0001 NCard  partition 1
4.00 MiB
[    1.882621] mmcblk0boot1: mmc0:0001 NCard  partition 2
4.00 MiB
[    2.004408] meson-gx-mmc d0070000.mmc: allocated
mmc-pwrseq
[    2.048189] meson-gx-mmc d0072000.mmc: Got CD GPIO
[    2.066978] mmc1: queuing unknown CIS tuple 0x80 (2 bytes)
[    2.069001] mmc1: queuing unknown CIS tuple 0x80 (3 bytes)
[    2.074253] mmc1: queuing unknown CIS tuple 0x80 (3 bytes)
[    2.081285] mmc1: queuing unknown CIS tuple 0x80 (7 bytes)
[    2.124874] mmc2: Skipping voltage switch
[    2.153475] mmc1: new high speed SDIO card at address 0001
[    2.158610] mmc2: new high speed SDHC card at address 59b4
[    2.158935] mmcblk2: mmc2:59b4 USD00 29.5 GiB
[    2.163910]  mmcblk2: p1 p2
[    6.586318] brcmfmac: brcmf_sdiod_ramrw: membytes transfer
failed <<<=====
[    6.586810] brcmfmac: brcmf_sdio_download_code_file: error
-84 on writing 239507 membytes at 0x00000000
[    6.596373] brcmfmac: brcmf_sdio_download_firmware: dongle
image file download failed

and this is the output after adding patches 4 & 5

[    1.684941] meson-gx-mmc d0074000.mmc: allocated
mmc-pwrseq
[    1.868910] mmc0: new DDR MMC card at address 0001
[    1.869159] mmcblk0: mmc0:0001 NCard  14.5 GiB
[    1.872639] mmcblk0boot0: mmc0:0001 NCard  partition 1
4.00 MiB
[    1.878497] mmcblk0boot1: mmc0:0001 NCard  partition 2
4.00 MiB
[    1.884356] mmcblk0rpmb: mmc0:0001 NCard  partition 3 128
KiB <<<=====
[    2.000339] meson-gx-mmc d0070000.mmc: allocated
mmc-pwrseq
[    2.044090] meson-gx-mmc d0072000.mmc: Got CD GPIO
[    2.062850] mmc1: queuing unknown CIS tuple 0x80 (2 bytes)
[    2.064869] mmc1: queuing unknown CIS tuple 0x80 (3 bytes)
[    2.070129] mmc1: queuing unknown CIS tuple 0x80 (3 bytes)
[    2.077161] mmc1: queuing unknown CIS tuple 0x80 (7 bytes)
[    2.120693] mmc2: Skipping voltage switch
[    2.148795] mmc1: new high speed SDIO card at address 0001
[    2.154443] mmc2: new high speed SDHC card at address 59b4
[    2.154660] mmcblk2: mmc2:59b4 USD00 29.5 GiB
[    2.159402]  mmcblk2: p1 p2
[    6.550881] brcmfmac: brcmf_chip_recognition: SB chip
is not
supported     <<<=====
[    6.556606] brcmfmac: brcmf_sdio_probe_attach:
brcmf_chip_attach failed!
[    6.563260] brcmfmac: brcmf_sdio_probe:
brcmf_sdio_probe_attach failed
[    6.569796] brcmfmac: brcmf_ops_sdio_probe: F2 error,
probe
failed -19...

regards
Helmut


Thanks a lot for the additional information. The first version
of the patch
set broke byte mode obviously (block mode is working fine).

Before submitting a new version of the patch set:
Could you please apply patches 1, 2, and 4 of the patch set
plus
the
following one and re-test?

Thanks, Heiner

---
 drivers/mmc/host/meson-gx-mmc.c | 222
++++++++++++++++++++++------------------
 1 file changed, 124 insertions(+), 98 deletions(-)

diff --git a/drivers/mmc/host/meson-gx-mmc.c
b/drivers/mmc/host/meson-gx-mmc.c
index 6bfd3da9..ca685902 100644
--- a/drivers/mmc/host/meson-gx-mmc.c
+++ b/drivers/mmc/host/meson-gx-mmc.c
@@ -121,6 +121,13 @@
 #define SD_EMMC_CFG_CMD_GAP 16 /* in clock cycles */
 #define MUX_CLK_NUM_PARENTS 2

+struct sd_emmc_desc {
+    u32 cmd_cfg;
+    u32 cmd_arg;
+    u32 cmd_data;
+    u32 cmd_resp;
+};
+
 struct meson_host {
     struct    device        *dev;
     struct    mmc_host    *mmc;
@@ -136,19 +143,12 @@ struct meson_host {
     struct clk_divider cfg_div;
     struct clk *cfg_div_clk;

-    unsigned int bounce_buf_size;
-    void *bounce_buf;
-    dma_addr_t bounce_dma_addr;
+    struct sd_emmc_desc *descs;
+    dma_addr_t descs_dma_addr;

     bool vqmmc_enabled;
 };

-struct sd_emmc_desc {
-    u32 cmd_cfg;
-    u32 cmd_arg;
-    u32 cmd_data;
-    u32 cmd_resp;
-};
 #define CMD_CFG_LENGTH_SHIFT 0
 #define CMD_CFG_LENGTH_MASK 0x1ff
 #define CMD_CFG_BLOCK_MODE BIT(9)
@@ -185,6 +185,36 @@ static struct mmc_command
*meson_mmc_get_next_command(struct mmc_command *cmd)
         return NULL;
 }

+static enum dma_data_direction meson_mmc_get_data_dir(struct
mmc_data *data)
+{
+    return data->flags & MMC_DATA_WRITE ? DMA_TO_DEVICE :
DMA_FROM_DEVICE;
+}
+
+static void meson_mmc_pre_req(struct mmc_host *mmc, struct
mmc_request *mrq)
+{
+    struct mmc_data *data = mrq->data;
+
+    if (!data)
+        return;
+
+    data->host_cookie = true;
+
+    data->sg_count = dma_map_sg(mmc_dev(mmc), data->sg,
data->sg_len,
+                    meson_mmc_get_data_dir(data));
+    if (!data->sg_count)
+        dev_err(mmc_dev(mmc), "dma_map_sg failed");
+}
+
+static void meson_mmc_post_req(struct mmc_host *mmc, struct
mmc_request *mrq,
+                   int err)
+{
+    struct mmc_data *data = mrq->data;
+
+    if (data && data->sg_count)
+        dma_unmap_sg(mmc_dev(mmc), data->sg, data->sg_len,
+                 meson_mmc_get_data_dir(data));
+}
+
 static int meson_mmc_clk_set(struct meson_host *host,
unsigned
long clk_rate)
 {
     struct mmc_host *mmc = host->mmc;
@@ -434,104 +464,102 @@ static void
meson_mmc_request_done(struct mmc_host *mmc,
 static void meson_mmc_start_cmd(struct mmc_host *mmc, struct
mmc_command *cmd)
 {
     struct meson_host *host = mmc_priv(mmc);
-    struct sd_emmc_desc *desc, desc_tmp;
-    u32 cfg;
-    u8 blk_len, cmd_cfg_timeout;
-    unsigned int xfer_bytes = 0;
+    struct sd_emmc_desc *desc = host->descs;
+    struct mmc_data *data = cmd->data;
+    struct scatterlist *sg;
+    u32 cfg, cmd_cfg = 0;
+    u8 blk_len;
+    int i;

-    /* Setup descriptors */
     dma_rmb();
-    desc = &desc_tmp;
-    memset(desc, 0, sizeof(struct sd_emmc_desc));

-    desc->cmd_cfg |= (cmd->opcode &
CMD_CFG_CMD_INDEX_MASK)    <<
-        CMD_CFG_CMD_INDEX_SHIFT;
-    desc->cmd_cfg |= CMD_CFG_OWNER;  /* owned by CPU */
-    desc->cmd_arg = cmd->arg;
+    cmd_cfg |= (cmd->opcode & CMD_CFG_CMD_INDEX_MASK) <<
+           CMD_CFG_CMD_INDEX_SHIFT;
+    cmd_cfg |= CMD_CFG_OWNER;  /* owned by CPU */

     /* Response */
     if (cmd->flags & MMC_RSP_PRESENT) {
-        desc->cmd_cfg &= ~CMD_CFG_NO_RESP;
         if (cmd->flags & MMC_RSP_136)
-            desc->cmd_cfg |= CMD_CFG_RESP_128;
-        desc->cmd_cfg |= CMD_CFG_RESP_NUM;
-        desc->cmd_resp = 0;
+            cmd_cfg |= CMD_CFG_RESP_128;
+        cmd_cfg |= CMD_CFG_RESP_NUM;

         if (!(cmd->flags & MMC_RSP_CRC))
-            desc->cmd_cfg |= CMD_CFG_RESP_NOCRC;
+            cmd_cfg |= CMD_CFG_RESP_NOCRC;

         if (cmd->flags & MMC_RSP_BUSY)
-            desc->cmd_cfg |= CMD_CFG_R1B;
+            cmd_cfg |= CMD_CFG_R1B;
     } else {
-        desc->cmd_cfg |= CMD_CFG_NO_RESP;
+        cmd_cfg |= CMD_CFG_NO_RESP;
     }

-    /* data? */
-    if (cmd->data) {
-        desc->cmd_cfg |= CMD_CFG_DATA_IO;
-        if (cmd->data->blocks > 1) {
-            desc->cmd_cfg |= CMD_CFG_BLOCK_MODE;
-            desc->cmd_cfg |=
-                (cmd->data->blocks & CMD_CFG_LENGTH_MASK) <<
-                CMD_CFG_LENGTH_SHIFT;
+    if (data) {
+        cmd_cfg |= CMD_CFG_DATA_IO;
+
+        if (data->blocks > 1) {
+            cmd_cfg |= CMD_CFG_BLOCK_MODE;

             /* check if block-size matches, if not update */
             cfg = readl(host->regs + SD_EMMC_CFG);
             blk_len = cfg & (CFG_BLK_LEN_MASK <<
CFG_BLK_LEN_SHIFT);
             blk_len >>= CFG_BLK_LEN_SHIFT;
-            if (blk_len != ilog2(cmd->data->blksz)) {
-                dev_dbg(host->dev, "%s: update blk_len %d ->
%d\n",
-                    __func__, blk_len,
-                    ilog2(cmd->data->blksz));
-                blk_len = ilog2(cmd->data->blksz);
+            if (blk_len != ilog2(data->blksz)) {
+                dev_dbg(host->dev,
+                    "%s: update blk_len %d -> %d\n",
+                    __func__, blk_len, ilog2(data->blksz));
+                blk_len = ilog2(data->blksz);
                 cfg &= ~(CFG_BLK_LEN_MASK <<
CFG_BLK_LEN_SHIFT);
                 cfg |= blk_len << CFG_BLK_LEN_SHIFT;
                 writel(cfg, host->regs + SD_EMMC_CFG);
             }
-        } else {
-            desc->cmd_cfg &= ~CMD_CFG_BLOCK_MODE;
-            desc->cmd_cfg |=
-                (cmd->data->blksz & CMD_CFG_LENGTH_MASK) <<
-                CMD_CFG_LENGTH_SHIFT;
         }

-        cmd->data->bytes_xfered = 0;
-        xfer_bytes = cmd->data->blksz * cmd->data->blocks;
-        if (cmd->data->flags & MMC_DATA_WRITE) {
-            desc->cmd_cfg |= CMD_CFG_DATA_WR;
-            WARN_ON(xfer_bytes > host->bounce_buf_size);
-            sg_copy_to_buffer(cmd->data->sg,
cmd->data->sg_len,
-                      host->bounce_buf, xfer_bytes);
-            cmd->data->bytes_xfered = xfer_bytes;
-            dma_wmb();
-        } else {
-            desc->cmd_cfg &= ~CMD_CFG_DATA_WR;
-        }
+        data->bytes_xfered = 0;
+        if (data->flags & MMC_DATA_WRITE)
+            cmd_cfg |= CMD_CFG_DATA_WR;

-        desc->cmd_data = host->bounce_dma_addr &
CMD_DATA_MASK;
+        cmd_cfg |= ilog2(SD_EMMC_CMD_TIMEOUT_DATA) <<
+               CMD_CFG_TIMEOUT_SHIFT;
+
+        for_each_sg(data->sg, sg, data->sg_count, i) {
+            unsigned int len = sg_dma_len(sg);
+
+            if (data->blocks > 1)
+                len /= data->blksz;
+
+            desc[i].cmd_cfg = cmd_cfg;
+            desc[i].cmd_cfg |= (len & CMD_CFG_LENGTH_MASK) <<
+                       CMD_CFG_LENGTH_SHIFT;
+            if (i > 0)
+                desc[i].cmd_cfg |= CMD_CFG_NO_CMD;
+            desc[i].cmd_arg = cmd->arg;
+            desc[i].cmd_resp = 0;
+            desc[i].cmd_data = sg_dma_address(sg);
+        }
+        desc[data->sg_count - 1].cmd_cfg |=
CMD_CFG_END_OF_CHAIN;

-        cmd_cfg_timeout = ilog2(SD_EMMC_CMD_TIMEOUT_DATA);
     } else {
-        desc->cmd_cfg &= ~CMD_CFG_DATA_IO;
-        cmd_cfg_timeout = ilog2(SD_EMMC_CMD_TIMEOUT);
+        cmd_cfg |= ilog2(SD_EMMC_CMD_TIMEOUT) <<
CMD_CFG_TIMEOUT_SHIFT;
+        cmd_cfg |= CMD_CFG_END_OF_CHAIN;
+        desc[0].cmd_cfg = cmd_cfg;
+        desc[0].cmd_arg = cmd->arg;
+        desc[0].cmd_resp = 0;
+        desc[0].cmd_data = 0;
     }
-    desc->cmd_cfg |= (cmd_cfg_timeout &
CMD_CFG_TIMEOUT_MASK) <<
-        CMD_CFG_TIMEOUT_SHIFT;

     host->cmd = cmd;

-    /* Last descriptor */
-    desc->cmd_cfg |= CMD_CFG_END_OF_CHAIN;
-    writel(desc->cmd_cfg, host->regs + SD_EMMC_CMD_CFG);
-    writel(desc->cmd_data, host->regs + SD_EMMC_CMD_DAT);
-    writel(desc->cmd_resp, host->regs + SD_EMMC_CMD_RSP);
     wmb(); /* ensure descriptor is written before kicked */
-    writel(desc->cmd_arg, host->regs + SD_EMMC_CMD_ARG);
+    cfg = host->descs_dma_addr | START_DESC_BUSY;
+    writel(cfg, host->regs + SD_EMMC_START);
 }

 static void meson_mmc_request(struct mmc_host *mmc, struct
mmc_request *mrq)
 {
     struct meson_host *host = mmc_priv(mmc);
+    bool needs_pre_post_req = mrq->data &&
!mrq->data->host_cookie;
+
+    if (needs_pre_post_req)
+        meson_mmc_pre_req(mmc, mrq);

     /* Stop execution */
     writel(0, host->regs + SD_EMMC_START);
@@ -540,6 +568,9 @@ static void meson_mmc_request(struct
mmc_host *mmc, struct mmc_request *mrq)
         meson_mmc_start_cmd(mmc, mrq->sbc);
     else
         meson_mmc_start_cmd(mmc, mrq->cmd);
+
+    if (needs_pre_post_req)
+        meson_mmc_post_req(mmc, mrq, 0);
 }

 static void meson_mmc_read_resp(struct mmc_host *mmc, struct
mmc_command *cmd)
@@ -560,6 +591,7 @@ static irqreturn_t meson_mmc_irq(int irq,
void *dev_id)
 {
     struct meson_host *host = dev_id;
     struct mmc_command *cmd;
+    struct mmc_data *data;
     u32 irq_en, status, raw_status;
     irqreturn_t ret = IRQ_HANDLED;

@@ -571,6 +603,8 @@ static irqreturn_t meson_mmc_irq(int irq,
void *dev_id)
     if (WARN_ON(!cmd))
         return IRQ_NONE;

+    data = cmd->data;
+
     spin_lock(&host->lock);
     irq_en = readl(host->regs + SD_EMMC_IRQ_EN);
     raw_status = readl(host->regs + SD_EMMC_STATUS);
@@ -608,12 +642,17 @@ static irqreturn_t meson_mmc_irq(int
irq,
void *dev_id)
         dev_dbg(host->dev, "Unhandled IRQ: Descriptor
timeout\n");
         cmd->error = -ETIMEDOUT;
     }
+
+    if (data && !cmd->error)
+        data->bytes_xfered = data->blksz * data->blocks;
+
     if (status & IRQ_SDIO)
         dev_dbg(host->dev, "Unhandled IRQ: SDIO.\n");

-    if (status & (IRQ_END_OF_CHAIN | IRQ_RESP_STATUS))
-        ret = IRQ_WAKE_THREAD;
-    else  {
+    if (status & (IRQ_END_OF_CHAIN | IRQ_RESP_STATUS)) {
+        if (meson_mmc_get_next_command(cmd))
+            ret = IRQ_WAKE_THREAD;
+    } else  {
         dev_warn(host->dev, "Unknown IRQ! status=0x%04x: MMC
CMD%u arg=0x%08x flags=0x%08x stop=%d\n",
              status, cmd->opcode, cmd->arg,
              cmd->flags, cmd->mrq->stop ? 1 : 0);
@@ -642,26 +681,12 @@ static irqreturn_t
meson_mmc_irq_thread(int irq, void *dev_id)
 {
     struct meson_host *host = dev_id;
     struct mmc_command *next_cmd, *cmd = host->cmd;
-    struct mmc_data *data;
-    unsigned int xfer_bytes;

     if (WARN_ON(!cmd))
         return IRQ_NONE;

-    data = cmd->data;
-    if (data && data->flags & MMC_DATA_READ) {
-        xfer_bytes = data->blksz * data->blocks;
-        WARN_ON(xfer_bytes > host->bounce_buf_size);
-        sg_copy_from_buffer(data->sg, data->sg_len,
-                    host->bounce_buf, xfer_bytes);
-        data->bytes_xfered = xfer_bytes;
-    }
-
     next_cmd = meson_mmc_get_next_command(cmd);
-    if (next_cmd)
-        meson_mmc_start_cmd(host->mmc, next_cmd);
-    else
-        meson_mmc_request_done(host->mmc, cmd->mrq);
+    meson_mmc_start_cmd(host->mmc, next_cmd);

     return IRQ_HANDLED;
 }
@@ -695,6 +720,8 @@ static const struct mmc_host_ops
meson_mmc_ops = {
     .request    = meson_mmc_request,
     .set_ios    = meson_mmc_set_ios,
     .get_cd         = meson_mmc_get_cd,
+    .pre_req    = meson_mmc_pre_req,
+    .post_req    = meson_mmc_post_req,
 };

 static int meson_mmc_probe(struct platform_device *pdev)
@@ -774,15 +801,14 @@ static int meson_mmc_probe(struct
platform_device *pdev)

     mmc->caps |= MMC_CAP_CMD23;
     mmc->max_blk_count = CMD_CFG_LENGTH_MASK;
-    mmc->max_req_size = mmc->max_blk_count *
mmc->max_blk_size;
-
-    /* data bounce buffer */
-    host->bounce_buf_size = mmc->max_req_size;
-    host->bounce_buf =
-        dma_alloc_coherent(host->dev, host->bounce_buf_size,
-                   &host->bounce_dma_addr, GFP_KERNEL);
-    if (host->bounce_buf == NULL) {
-        dev_err(host->dev, "Unable to map allocate DMA bounce
buffer.\n");
+    mmc->max_segs = PAGE_SIZE / sizeof(struct sd_emmc_desc);
+    mmc->max_seg_size = mmc->max_blk_count *
mmc->max_blk_size;
+    mmc->max_req_size = mmc->max_seg_size * mmc->max_segs;
+
+    host->descs = dma_alloc_coherent(host->dev, PAGE_SIZE,
+                     &host->descs_dma_addr, GFP_KERNEL);
+    if (!host->descs) {
+        dev_err(host->dev, "Allocating descriptor DMA buffer
failed\n");
         ret = -ENOMEM;
         goto err_div_clk;
     }
@@ -807,8 +833,8 @@ static int meson_mmc_remove(struct
platform_device *pdev)
     /* disable interrupts */
     writel(0, host->regs + SD_EMMC_IRQ_EN);

-    dma_free_coherent(host->dev, host->bounce_buf_size,
-              host->bounce_buf, host->bounce_dma_addr);
+    dma_free_coherent(host->dev, PAGE_SIZE, host->descs,
+              host->descs_dma_addr);

     clk_disable_unprepare(host->cfg_div_clk);
     clk_disable_unprepare(host->core_clk);


compared to the patches 1...5 there is absolutely no change to
patches 1, 2, 4 and the above. Neither syslog nor performance.

Thanks for re-testing. I have a little bit of a hard time to
understand why SD and eMMC mode
are working but SDIO (at least with brcfmac) is not.
Especially as
I don't have HW to test SDIO mode on.

After patches 1, 2, 4 the system is still working normally?
And would it be possible for you to compile a DEBUG kernel and
post the mmc/brcmfmac related output?
In DEBUG mode mode brcmfmac driver logs all SDIO transfers.

Last but not least, could you please post /proc/interrupts ?
This
would give an idea whether any SDIO
or just specific ones fail.

Thanks, Heiner

regards
Helmut




i enabled the debug flags for mmc and brcmfmac in the kernel
configuration.
For the test i also changed meson-gx-mmc from built-in to module.
This resulted in a different order of the mmc devices. The
sdio/wifi is now mmc0 and not mmc1

Without your new patch the wifi adapter works normally.

The debug info was produced with following commad sequence
 dmesg -n 8
 dmesg -D
 logger "loading meson-gx-mmc...."
 modprobe meson-gx-mmc
 sleep 1
 logger "loading brcmfmac"
 modprobe brcmfmac
 logger "modules loaded"
 sleep 1
 cp /var/log/syslog ~/syslog
 cp /proc/interrupts ~/interrupts

and filtered with: egrep "brcmfmac|mmc0|meson-gx-mmc"

because of the length of the files i removed most lines of the
loading of meson-gx-mmc

unhandled interrupts: independent of the number of applied
patches
i've always got 11 of them.
 Only the timings differed with the applied patches

the is the log for applied patches 1, 2, 4 & the new one
  [   47.371608] mmc0: starting CMD52 arg 0020d000 flags 00000195
  [   47.371616] mmc0: req done (CMD52): 0: 00001000 00000000
00000000 00000000
  [   47.371621] mmc0: starting CMD52 arg 0020d200 flags 00000195
  [   47.371629] mmc0: req done (CMD52): 0: 00001000 00000000
00000000 00000000
  [   47.371635] mmc0: starting CMD52 arg 0020d400 flags 00000195
  [   47.371642] mmc0: req done (CMD52): 0: 00001000 00000000
00000000 00000000
  [   47.371648] mmc0: starting CMD52 arg 0020d600 flags 00000195
  [   47.371656] mmc0: req done (CMD52): 0: 00001000 00000000
00000000 00000000
  [   47.371662] mmc0: starting CMD52 arg 0020d800 flags 00000195
  [   47.371670] mmc0: req done (CMD52): 0: 00001000 00000000
00000000 00000000
  [   47.371679] mmc0: starting CMD52 arg 0020da00 flags 00000195
  [   47.371687] mmc0: req done (CMD52): 0: 00001000 00000000
00000000 00000000
  [   47.371694] mmc0: starting CMD52 arg 0020dc00 flags 00000195
  [   47.371702] mmc0: req done (CMD52): 0: 000010ff 00000000
00000000 00000000
  [   47.371716] mmc0: new high speed SDIO card at address 0001
  [   47.372210] meson-gx-mmc d0074000.mmc: Unhandled IRQ:
Response
timeout
  [   47.373142] meson-gx-mmc d0074000.mmc: Unhandled IRQ:
Response
timeout
  [   47.374077] meson-gx-mmc d0074000.mmc: Unhandled IRQ:
Response
timeout
  [   47.375021] meson-gx-mmc d0074000.mmc: Unhandled IRQ:
Response
timeout
  [   47.375950] meson-gx-mmc d0074000.mmc: Unhandled IRQ:
Response
timeout
  [   47.376859] meson-gx-mmc d0074000.mmc: Unhandled IRQ:
Response
timeout
  [   47.377770] meson-gx-mmc d0074000.mmc: Unhandled IRQ:
Response
timeout
  [   47.496509] meson-gx-mmc d0074000.mmc: change clock rate
400000 -> 52000000
  [   47.496541] meson-gx-mmc d0074000.mmc: divider requested
rate
52000000 != actual rate 50000000
  [   47.496602] meson-gx-mmc d0074000.mmc: meson_mmc_set_ios:
SD_EMMC_CFG: 0x00004890 -> 0x00004892
  [   47.497276] meson-gx-mmc d0074000.mmc: meson_mmc_set_ios:
SD_EMMC_CFG: 0x00004892 -> 0x00004896
  loading brcmfmac
  [   48.731368] brcmfmac: brcmfmac_module_init No platform data
available.
  [   48.731451] mmc0: starting CMD52 arg 80022040 flags 00000195
  [   48.731471] mmc0: req done (CMD52): 0: 00001040 00000000
00000000 00000000
  [   48.731489] mmc0: starting CMD52 arg 80022200 flags 00000195
  [   48.731499] mmc0: req done (CMD52): 0: 00001000 00000000
00000000 00000000
  [   48.731507] brcmfmac: brcmf_ops_sdio_probe Enter
  [   48.731510] brcmfmac: brcmf_ops_sdio_probe Class=0
  [   48.731512] brcmfmac: brcmf_ops_sdio_probe sdio vendor
ID: 0x02d0
  [   48.731515] brcmfmac: brcmf_ops_sdio_probe sdio device
ID: 0x4330
  [   48.731517] brcmfmac: brcmf_ops_sdio_probe Function#: 1
  [   48.731538] mmc0: starting CMD52 arg 80042000 flags 00000195
  [   48.731548] mmc0: req done (CMD52): 0: 00001000 00000000
00000000 00000000
  [   48.731555] mmc0: starting CMD52 arg 80042202 flags 00000195
  [   48.731564] mmc0: req done (CMD52): 0: 00001002 00000000
00000000 00000000
  [   48.731580] brcmfmac: brcmf_ops_sdio_probe Enter
  [   48.731582] brcmfmac: brcmf_ops_sdio_probe Class=0
  [   48.731584] brcmfmac: brcmf_ops_sdio_probe sdio vendor
ID: 0x02d0
  [   48.731586] brcmfmac: brcmf_ops_sdio_probe sdio device
ID: 0x4330
  [   48.731588] brcmfmac: brcmf_ops_sdio_probe Function#: 2
  [   48.731593] brcmfmac: brcmf_ops_sdio_probe F2 found, calling
brcmf_sdiod_probe...
  [   48.731596] mmc0: starting CMD52 arg 80022040 flags 00000195
  [   48.731606] mmc0: req done (CMD52): 0: 00001040 00000000
00000000 00000000
  [   48.731613] mmc0: starting CMD52 arg 80022200 flags 00000195
  [   48.731622] mmc0: req done (CMD52): 0: 00001000 00000000
00000000 00000000
  [   48.731629] mmc0: starting CMD52 arg 80042000 flags 00000195
  [   48.731638] mmc0: req done (CMD52): 0: 00001000 00000000
00000000 00000000
  [   48.731644] mmc0: starting CMD52 arg 80042202 flags 00000195
  [   48.731653] mmc0: req done (CMD52): 0: 00001002 00000000
00000000 00000000
  [   48.731661] SDIO: Enabling device mmc0:0001:1...
  [   48.731664] mmc0: starting CMD52 arg 00000400 flags 00000195
  [   48.731673] mmc0: req done (CMD52): 0: 00001000 00000000
00000000 00000000
  [   48.731680] mmc0: starting CMD52 arg 80000402 flags 00000195
  [   48.731689] mmc0: req done (CMD52): 0: 00001002 00000000
00000000 00000000
  [   48.731697] mmc0: starting CMD52 arg 00000600 flags 00000195
  [   48.731706] mmc0: req done (CMD52): 0: 00001002 00000000
00000000 00000000
  [   48.731713] SDIO: Enabled device mmc0:0001:1
  [   48.731717] brcmfmac: brcmf_sdio_probe Enter
  [   48.731795] brcmfmac: brcmf_sdiod_regrl addr:0x18000000
  [   48.731800] brcmfmac: brcmf_sdiod_request_data rw=1, func=1,
addr=0x1000a, nbytes=1
  [   48.731804] mmc0: starting CMD52 arg 92001400 flags 00000195
  [   48.731813] mmc0: req done (CMD52): 0: 00001000 00000000
00000000 00000000
  [   48.731821] brcmfmac: brcmf_sdiod_request_data rw=1, func=1,
addr=0x1000b, nbytes=1
  [   48.731824] mmc0: starting CMD52 arg 92001600 flags 00000195
  [   48.731833] mmc0: req done (CMD52): 0: 00001000 00000000
00000000 00000000
  [   48.731840] brcmfmac: brcmf_sdiod_request_data rw=1, func=1,
addr=0x1000c, nbytes=1
  [   48.731843] mmc0: starting CMD52 arg 92001818 flags 00000195
  [   48.731851] mmc0: req done (CMD52): 0: 00001018 00000000
00000000 00000000
  [   48.731858] brcmfmac: brcmf_sdiod_request_data rw=0, func=1,
addr=0x08000, nbytes=4
  [   48.731863] mmc0: starting CMD53 arg 15000004 flags 000001b5
  [   48.731866] mmc0:     blksz 4 blocks 1 flags 00000200 tsac
1000 ms nsac 0
  [   48.731878] mmc0: req done (CMD53): 0: 00001000 00000000
00000000 00000000
  [   48.731881] mmc0:     4 bytes transferred: 0
  [   48.731887] brcmfmac: brcmf_sdiod_regrl
data:0x00000000             <<<<===== wrong!
  [   48.731890] brcmfmac: F1 signature read @0x18000000=0x   0
  [   48.731892] brcmfmac: brcmf_sdiod_regwb addr:0x0001000e,
data:0x28
  [   48.731895] brcmfmac: brcmf_sdiod_request_data rw=1, func=1,
addr=0x1000e, nbytes=1
  [   48.731898] mmc0: starting CMD52 arg 92001c28 flags 00000195
  [   48.731906] mmc0: req done (CMD52): 0: 00001028 00000000
00000000 00000000
  [   48.731912] brcmfmac: brcmf_sdiod_regrb addr:0x0001000e
  [   48.731915] brcmfmac: brcmf_sdiod_request_data rw=0, func=1,
addr=0x1000e, nbytes=1
  [   48.731918] mmc0: starting CMD52 arg 12001c00 flags 00000195
  [   48.731926] mmc0: req done (CMD52): 0: 00001068 00000000
00000000 00000000
  [   48.731931] brcmfmac: brcmf_sdiod_regrb data:0x68
  [   48.731936] brcmfmac: brcmf_sdiod_regwb addr:0x0001000e,
data:0x28
  [   48.731938] brcmfmac: brcmf_sdiod_request_data rw=1, func=1,
addr=0x1000e, nbytes=1
  [   48.731941] mmc0: starting CMD52 arg 92001c28 flags 00000195
  [   48.731950] mmc0: req done (CMD52): 0: 00001028 00000000
00000000 00000000
  [   48.731957] brcmfmac: brcmf_sdiod_regrb addr:0x0001000e
  [   48.731960] brcmfmac: brcmf_sdiod_request_data rw=0, func=1,
addr=0x1000e, nbytes=1
  [   48.731962] mmc0: starting CMD52 arg 12001c00 flags 00000195
  [   48.731970] mmc0: req done (CMD52): 0: 00001068 00000000
00000000 00000000
  [   48.731985] brcmfmac: brcmf_sdiod_regrb data:0x68
  [   48.731987] brcmfmac: brcmf_sdiod_regrb addr:0x0001000e
  [   48.731990] brcmfmac: brcmf_sdiod_request_data rw=0, func=1,
addr=0x1000e, nbytes=1
  [   48.731993] mmc0: starting CMD52 arg 12001c00 flags 00000195
  [   48.732001] mmc0: req done (CMD52): 0: 00001068 00000000
00000000 00000000
  [   48.732019] brcmfmac: brcmf_sdiod_regrb data:0x68
  [   48.732021] brcmfmac: brcmf_sdiod_regwb addr:0x0001000e,
data:0x21
  [   48.732024] brcmfmac: brcmf_sdiod_request_data rw=1, func=1,
addr=0x1000e, nbytes=1
  [   48.732027] mmc0: starting CMD52 arg 92001c21 flags 00000195
  [   48.732037] mmc0: req done (CMD52): 0: 00001021 00000000
00000000 00000000
  [   48.732151] brcmfmac: brcmf_sdiod_regwb addr:0x0001000f,
data:0x00
  [   48.732160] brcmfmac: brcmf_sdiod_request_data rw=1, func=1,
addr=0x1000f, nbytes=1
  [   48.732167] mmc0: starting CMD52 arg 92001e00 flags 00000195
  [   48.732181] mmc0: req done (CMD52): 0: 00001000 00000000
00000000 00000000
  [   48.732193] brcmfmac: brcmf_sdiod_regrl addr:0x18000000
  [   48.732200] brcmfmac: brcmf_sdiod_request_data rw=0, func=1,
addr=0x08000, nbytes=4
  [   48.732208] mmc0: starting CMD53 arg 15000004 flags 000001b5
  [   48.732214] mmc0:     blksz 4 blocks 1 flags 00000200 tsac
1000 ms nsac 0
  [   48.732228] mmc0: req done (CMD53): 0: 00001000 00000000
00000000 00000000
  [   48.732235] mmc0:     4 bytes transferred: 0
  [   48.732246] brcmfmac: brcmf_sdiod_regrl data:0x00000000
  [   48.732254] brcmfmac: brcmf_chip_recognition found SB chip:
BCM0, rev=0
  [   48.732260] brcmfmac: brcmf_chip_recognition: SB chip is not
supported
  [   48.732266] brcmfmac: brcmf_sdio_probe_attach:
brcmf_chip_attach failed!
  [   48.732273] brcmfmac: brcmf_sdio_probe:
brcmf_sdio_probe_attach failed
  [   48.732280] brcmfmac: brcmf_sdio_remove Enter
  [   48.732285] brcmfmac: brcmf_sdiod_intr_unregister Entering
oob=0 sd=0
  [   48.732290] brcmfmac: brcmf_detach Enter
  [   48.733671] brcmfmac: brcmf_sdio_remove Disconnected
  [   48.733683] SDIO: Disabling device mmc0:0001:2...
  [   48.733689] mmc0: starting CMD52 arg 00000400 flags 00000195
  [   48.733709] mmc0: req done (CMD52): 0: 00001002 00000000
00000000 00000000
  [   48.733717] mmc0: starting CMD52 arg 80000402 flags 00000195
  [   48.733727] mmc0: req done (CMD52): 0: 00001002 00000000
00000000 00000000
  [   48.733732] SDIO: Disabled device mmc0:0001:2
  [   48.733736] SDIO: Disabling device mmc0:0001:1...
  [   48.733739] mmc0: starting CMD52 arg 00000400 flags 00000195
  [   48.733749] mmc0: req done (CMD52): 0: 00001002 00000000
00000000 00000000
  [   48.733760] mmc0: starting CMD52 arg 80000400 flags 00000195
  [   48.733768] mmc0: req done (CMD52): 0: 00001000 00000000
00000000 00000000
  [   48.733773] SDIO: Disabled device mmc0:0001:1
  [   48.733777] brcmfmac: brcmf_ops_sdio_probe: F2 error, probe
failed -19...

this is the debug data for applied patches 1,2 & 4 (left only few
lines after the first CMD53)
  [   52.458082] mmc0: starting CMD52 arg 0020d000 flags 00000195
  [   52.458093] mmc0: req done (CMD52): 0: 00001000 00000000
00000000 00000000
  [   52.458098] mmc0: starting CMD52 arg 0020d200 flags 00000195
  [   52.458108] mmc0: req done (CMD52): 0: 00001000 00000000
00000000 00000000
  [   52.458116] mmc0: starting CMD52 arg 0020d400 flags 00000195
  [   52.458130] mmc0: req done (CMD52): 0: 00001000 00000000
00000000 00000000
  [   52.458135] mmc0: starting CMD52 arg 0020d600 flags 00000195
  [   52.458147] mmc0: req done (CMD52): 0: 00001000 00000000
00000000 00000000
  [   52.458152] mmc0: starting CMD52 arg 0020d800 flags 00000195
  [   52.458165] mmc0: req done (CMD52): 0: 00001000 00000000
00000000 00000000
  [   52.458171] mmc0: starting CMD52 arg 0020da00 flags 00000195
  [   52.458187] mmc0: req done (CMD52): 0: 00001000 00000000
00000000 00000000
  [   52.458193] mmc0: starting CMD52 arg 0020dc00 flags 00000195
  [   52.458204] mmc0: req done (CMD52): 0: 000010ff 00000000
00000000 00000000
  [   52.458217] mmc0: new high speed SDIO card at address 0001
  [   52.560803] meson-gx-mmc d0074000.mmc: change clock rate
400000 -> 52000000
  [   52.560834] meson-gx-mmc d0074000.mmc: divider requested
rate
52000000 != actual rate 50000000
  [   52.560900] meson-gx-mmc d0074000.mmc: meson_mmc_set_ios:
SD_EMMC_CFG: 0x00004890 -> 0x00004892
  [   52.561583] meson-gx-mmc d0074000.mmc: meson_mmc_set_ios:
SD_EMMC_CFG: 0x00004892 -> 0x00004896
  loading brcmfmac
  [   53.796217] brcmfmac: brcmfmac_module_init No platform data
available.
  [   53.796291] mmc0: starting CMD52 arg 80022040 flags 00000195
  [   53.796327] mmc0: req done (CMD52): 0: 00001040 00000000
00000000 00000000
  [   53.796336] mmc0: starting CMD52 arg 80022200 flags 00000195
  [   53.796349] mmc0: req done (CMD52): 0: 00001000 00000000
00000000 00000000
  [   53.796366] brcmfmac: brcmf_ops_sdio_probe Enter
  [   53.796368] brcmfmac: brcmf_ops_sdio_probe Class=0
  [   53.796370] brcmfmac: brcmf_ops_sdio_probe sdio vendor
ID: 0x02d0
  [   53.796372] brcmfmac: brcmf_ops_sdio_probe sdio device
ID: 0x4330
  [   53.796375] brcmfmac: brcmf_ops_sdio_probe Function#: 1
  [   53.796398] mmc0: starting CMD52 arg 80042000 flags 00000195
  [   53.796414] mmc0: req done (CMD52): 0: 00001000 00000000
00000000 00000000
  [   53.796422] mmc0: starting CMD52 arg 80042202 flags 00000195
  [   53.796436] mmc0: req done (CMD52): 0: 00001002 00000000
00000000 00000000
  [   53.796444] brcmfmac: brcmf_ops_sdio_probe Enter
  [   53.796446] brcmfmac: brcmf_ops_sdio_probe Class=0
  [   53.796448] brcmfmac: brcmf_ops_sdio_probe sdio vendor
ID: 0x02d0
  [   53.796450] brcmfmac: brcmf_ops_sdio_probe sdio device
ID: 0x4330
  [   53.796452] brcmfmac: brcmf_ops_sdio_probe Function#: 2
  [   53.796458] brcmfmac: brcmf_ops_sdio_probe F2 found, calling
brcmf_sdiod_probe...
  [   53.796461] mmc0: starting CMD52 arg 80022040 flags 00000195
  [   53.796478] mmc0: req done (CMD52): 0: 00001040 00000000
00000000 00000000
  [   53.796489] mmc0: starting CMD52 arg 80022200 flags 00000195
  [   53.796502] mmc0: req done (CMD52): 0: 00001000 00000000
00000000 00000000
  [   53.796509] mmc0: starting CMD52 arg 80042000 flags 00000195
  [   53.796521] mmc0: req done (CMD52): 0: 00001000 00000000
00000000 00000000
  [   53.796527] mmc0: starting CMD52 arg 80042202 flags 00000195
  [   53.796541] mmc0: req done (CMD52): 0: 00001002 00000000
00000000 00000000
  [   53.796550] SDIO: Enabling device mmc0:0001:1...
  [   53.796552] mmc0: starting CMD52 arg 00000400 flags 00000195
  [   53.796564] mmc0: req done (CMD52): 0: 00001000 00000000
00000000 00000000
  [   53.796570] mmc0: starting CMD52 arg 80000402 flags 00000195
  [   53.796583] mmc0: req done (CMD52): 0: 00001002 00000000
00000000 00000000
  [   53.796589] mmc0: starting CMD52 arg 00000600 flags 00000195
  [   53.796600] mmc0: req done (CMD52): 0: 00001002 00000000
00000000 00000000
  [   53.796605] SDIO: Enabled device mmc0:0001:1
  [   53.796609] brcmfmac: brcmf_sdio_probe Enter
  [   53.796686] brcmfmac: brcmf_sdiod_regrl addr:0x18000000
  [   53.796691] brcmfmac: brcmf_sdiod_request_data rw=1, func=1,
addr=0x1000a, nbytes=1
  [   53.796695] mmc0: starting CMD52 arg 92001400 flags 00000195
  [   53.796707] mmc0: req done (CMD52): 0: 00001000 00000000
00000000 00000000
  [   53.796715] brcmfmac: brcmf_sdiod_request_data rw=1, func=1,
addr=0x1000b, nbytes=1
  [   53.796718] mmc0: starting CMD52 arg 92001600 flags 00000195
  [   53.796729] mmc0: req done (CMD52): 0: 00001000 00000000
00000000 00000000
  [   53.796743] brcmfmac: brcmf_sdiod_request_data rw=1, func=1,
addr=0x1000c, nbytes=1
  [   53.796746] mmc0: starting CMD52 arg 92001818 flags 00000195
  [   53.796758] mmc0: req done (CMD52): 0: 00001018 00000000
00000000 00000000
  [   53.796801] brcmfmac: brcmf_sdiod_request_data rw=0, func=1,
addr=0x08000, nbytes=4
  [   53.796807] mmc0: starting CMD53 arg 15000004 flags 000001b5
  [   53.796810] mmc0:     blksz 4 blocks 1 flags 00000200 tsac
1000 ms nsac 0
  [   53.796826] mmc0: req done (CMD53): 0: 00001000 00000000
00000000 00000000
  [   53.796829] mmc0:     4 bytes transferred: 0
  [   53.796836] brcmfmac: brcmf_sdiod_regrl
data:0x16044330               <<<<<====== ok
  [   53.796839] brcmfmac: F1 signature read
@0x18000000=0x16044330
  [   53.796841] brcmfmac: brcmf_sdiod_regwb addr:0x0001000e,
data:0x28
  [   53.796844] brcmfmac: brcmf_sdiod_request_data rw=1, func=1,
addr=0x1000e, nbytes=1
  [   53.796847] mmc0: starting CMD52 arg 92001c28 flags 00000195
  [   53.796858] mmc0: req done (CMD52): 0: 00001028 00000000
00000000 00000000
  [   53.796872] brcmfmac: brcmf_sdiod_regrb addr:0x0001000e
  [   53.796875] brcmfmac: brcmf_sdiod_request_data rw=0, func=1,
addr=0x1000e, nbytes=1
  [   53.796878] mmc0: starting CMD52 arg 12001c00 flags 00000195
  [   53.796890] mmc0: req done (CMD52): 0: 00001068 00000000
00000000 00000000
  [   53.796897] brcmfmac: brcmf_sdiod_regrb data:0x68
  [   53.796902] brcmfmac: brcmf_sdiod_regwb addr:0x0001000e,
data:0x28
  [   53.796905] brcmfmac: brcmf_sdiod_request_data rw=1, func=1,
addr=0x1000e, nbytes=1

if you need more data: just ask ;-)

Thanks a lot for the logs. So the first read returns 0 instead of
the actual register value.
Could you please apply the following on top of 1, 2, 4, new one
and
send the log?
I hope the two debug messages bring me closer to the root cause of
the issue.

Rgds, Heiner


diff --git a/drivers/mmc/host/meson-gx-mmc.c
b/drivers/mmc/host/meson-gx-mmc.c
index 3b3ddf3e..d6d2d1a1 100644
--- a/drivers/mmc/host/meson-gx-mmc.c
+++ b/drivers/mmc/host/meson-gx-mmc.c
@@ -613,6 +613,9 @@ static void meson_mmc_start_cmd(struct
mmc_host
*mmc, struct mmc_command *cmd)
         for_each_sg(data->sg, sg, data->sg_count, i) {
             unsigned int len = sg_dma_len(sg);

+            if (cmd->opcode == 53)
+                dev_info(host->info, "cmd 53: idx %d len
%u\n", i,
len);
+
             if (data->blocks > 1)
                 len /= data->blksz;

@@ -709,6 +712,9 @@ static irqreturn_t meson_mmc_irq(int irq, void
*dev_id)

     meson_mmc_read_resp(host->mmc, cmd);

+    if (cmd->opcode == 53)
+        dev_info(host->dev, "cmd 53: cmd_data_0: %08x\n",
host->descs[0].cmd_data);
+
     cmd->error = 0;
     if (status & IRQ_RXD_ERR_MASK) {
         dev_dbg(host->dev, "Unhandled IRQ: RXD error\n");


here is the output:
[  107.510508] brcmfmac: brcmf_sdiod_request_data rw=0, func=1,
addr=0x08000, nbytes=4
[  107.510513] mmc0: starting CMD53 arg 15000004 flags 000001b5
[  107.510516] mmc0:     blksz 4 blocks 1 flags 00000200 tsac
1000 ms
nsac 0
[  107.510525] meson-gx-mmc d0070000.mmc: cmd 53: idx 0 len 4
[  107.510535] meson-gx-mmc d0070000.mmc: cmd 53: cmd_data_0:
721d0eb4
[  107.510539] mmc0: req done (CMD53): 0: 00001000 00000000
00000000
00000000
[  107.510541] mmc0:     4 bytes transferred: 0
[  107.510548] brcmfmac: brcmf_sdiod_regrl data:0x00000000

Helmut


Weird .. Everything looks ok so far. I don't see a reason why this
small read DMA returns
zeroed bytes only whilst bigger ones in SD/eMMC work perfectly fine.

OK, one last attempt before I have to think about a better way to
tackle this issue.

Could you please replace the last logging extension patch with
this one?
(I just added a memory barrier to be sure to read the actual value.)

Thanks, Heiner


diff --git a/drivers/mmc/host/meson-gx-mmc.c
b/drivers/mmc/host/meson-gx-mmc.c
index ca685902..5b511944 100644
--- a/drivers/mmc/host/meson-gx-mmc.c
+++ b/drivers/mmc/host/meson-gx-mmc.c
@@ -523,6 +523,9 @@ static void meson_mmc_start_cmd(struct mmc_host
*mmc, struct mmc_command *cmd)
         for_each_sg(data->sg, sg, data->sg_count, i) {
             unsigned int len = sg_dma_len(sg);

+            if (cmd->opcode == 53)
+                pr_info("sg: idx %d len %u\n", i, len);
+
             if (data->blocks > 1)
                 len /= data->blksz;

@@ -619,6 +622,11 @@ static irqreturn_t meson_mmc_irq(int irq, void
*dev_id)

     meson_mmc_read_resp(host->mmc, cmd);

+    if (cmd->opcode == 53) {
+        dma_rmb();
+        pr_info("cmd 53 cmd_data0:  %08x\n",
host->descs[0].cmd_data);
+    }
+
     cmd->error = 0;
     if (status & IRQ_RXD_ERR_MASK) {
         dev_dbg(host->dev, "Unhandled IRQ: RXD error\n");


the output is not really different (for the 2 cmd53)
[   37.113577] sg: idx 0 len 4
[   37.113595] meson-gx-mmc d0070000.mmc: cmd 53 cmd_data0:  72091ab4
[   37.114016] sg: idx 0 len 4
[   37.114034] meson-gx-mmc d0070000.mmc: cmd 53 cmd_data0:  72091ab4

Helmut


I've added my own debugging code to both variations of the driver.
but without a positiv result

here are the patches

for the fully operational version

--- a/drivers/mmc/host/meson-gx-mmc.c    2017-03-18
15:22:31.000000000 +0100
+++ b/drivers/mmc/host/meson-gx-mmc.c    2017-03-20
13:02:05.254762191 +0100
@@ -468,6 +468,9 @@ static void meson_mmc_start_cmd(struct m

     /* data? */
     if (cmd->data) {
+
+        dev_dbg(host->dev, "DBG data");
+
         desc->cmd_cfg |= CMD_CFG_DATA_IO;
         if (cmd->data->blocks > 1) {
             desc->cmd_cfg |= CMD_CFG_BLOCK_MODE;
@@ -526,6 +529,10 @@ static void meson_mmc_start_cmd(struct m
     writel(desc->cmd_data, host->regs + SD_EMMC_CMD_DAT);
     writel(desc->cmd_resp, host->regs + SD_EMMC_CMD_RSP);
     wmb(); /* ensure descriptor is written before kicked */
+
+    dev_dbg(host->dev, "DBG cmd_cfg: 0x%08x, arg: 0x%08x, resp:
0x%08x",
+        desc->cmd_cfg, desc->cmd_arg, desc->cmd_resp);
+
     writel(desc->cmd_arg, host->regs + SD_EMMC_CMD_ARG);
 }

@@ -576,6 +583,8 @@ static irqreturn_t meson_mmc_irq(int irq
     raw_status = readl(host->regs + SD_EMMC_STATUS);
     status = raw_status & irq_en;

+    dev_dbg(host->dev, "DBG status 0x%08x", status);
+
     if (!status) {
         dev_warn(host->dev, "Spurious IRQ! status=0x%08x,
irq_en=0x%08x\n",
              raw_status, irq_en);

and this is for the version with the sdio problem

--- a/drivers/mmc/host/meson-gx-mmc.c    2017-03-20
12:50:32.956474080 +0100
+++ b/drivers/mmc/host/meson-gx-mmc.c    2017-03-20
12:57:27.185837225 +0100
@@ -493,6 +493,9 @@ static void meson_mmc_start_cmd(struct m
     }

     if (data) {
+
+        dev_dbg(host->dev, "DBG data");
+
         cmd_cfg |= CMD_CFG_DATA_IO;

         if (data->blocks > 1) {
@@ -549,6 +552,10 @@ static void meson_mmc_start_cmd(struct m
     host->cmd = cmd;

     wmb(); /* ensure descriptor is written before kicked */
+
+    dev_dbg(host->dev, "DBG cmd_cfg: 0x%08x, arg: 0x%08x, resp:
0x%08x",
+        desc[0].cmd_cfg, desc[0].cmd_arg, desc[0].cmd_resp);
+
     cfg = host->descs_dma_addr | START_DESC_BUSY;
     writel(cfg, host->regs + SD_EMMC_START);
 }
@@ -610,6 +617,8 @@ static irqreturn_t meson_mmc_irq(int irq
     raw_status = readl(host->regs + SD_EMMC_STATUS);
     status = raw_status & irq_en;

+    dev_dbg(host->dev, "DBG status 0x%08x", status);
+
     if (!status) {
         dev_warn(host->dev, "Spurious IRQ! status=0x%08x,
irq_en=0x%08x\n",
              raw_status, irq_en);

the filtered logs are identical up and including the 2nd cmd53.

Helmut


Thanks. Really appreciate your support. Meanwhile I think there must
be a bug
either in the DMA subsystem or there's a hw bug causing issues with
very small
transfers. That's the only difference between SD/eMMC and SDIO on
that level:
SD/eMMC always transfers 512 byte blocks whilst SDIO also uses very
small
transfers.

For up to 4 bytes the chip supports an alternative transfer method.
Let's try this as workaround.

Could you please apply:
patches 1, 2, 4 + the updated bigger patch I sent via mail + the
following:


diff --git a/drivers/mmc/host/meson-gx-mmc.c
b/drivers/mmc/host/meson-gx-mmc.c
index 4f9fafd3..cf9be13f 100644
--- a/drivers/mmc/host/meson-gx-mmc.c
+++ b/drivers/mmc/host/meson-gx-mmc.c
@@ -185,6 +185,12 @@ static struct mmc_command
*meson_mmc_get_next_command(struct mmc_command *cmd)
         return NULL;
 }

+static bool meson_mmc_small_read(const struct mmc_data *data)
+{
+    return data && data->blocks <= 1 && data->blksz <= 4 &&
+           data->flags & MMC_DATA_READ;
+}
+
 static enum dma_data_direction meson_mmc_get_data_dir(struct
mmc_data *data)
 {
     return data->flags & MMC_DATA_WRITE ? DMA_TO_DEVICE :
DMA_FROM_DEVICE;
@@ -208,11 +214,22 @@ static void meson_mmc_pre_req(struct mmc_host
*mmc, struct mmc_request *mrq)
 static void meson_mmc_post_req(struct mmc_host *mmc, struct
mmc_request *mrq,
                    int err)
 {
+    struct meson_host *host = mmc_priv(mmc);
     struct mmc_data *data = mrq->data;

     if (data && data->sg_count)
         dma_unmap_sg(mmc_dev(mmc), data->sg, data->sg_len,
                  meson_mmc_get_data_dir(data));
+
+    if (meson_mmc_small_read(data)) {
+        int cnt;
+
+        dma_rmb();
+        cnt = sg_copy_from_buffer(data->sg, data->sg_len,
+                &host->descs[0].cmd_data, data->blksz);
+        if (cnt != data->blksz)
+            dev_err(host->dev, "error copying to scatterlist\n");
+    }
 }

 static int meson_mmc_clk_set(struct meson_host *host, unsigned long
clk_rate)
@@ -493,6 +510,8 @@ static void meson_mmc_start_cmd(struct mmc_host
*mmc, struct mmc_command *cmd)
     }

     if (data) {
+        bool small_read = meson_mmc_small_read(data);
+
         cmd_cfg |= CMD_CFG_DATA_IO;

         if (data->blocks > 1) {
@@ -513,6 +532,14 @@ static void meson_mmc_start_cmd(struct mmc_host
*mmc, struct mmc_command *cmd)
             }
         }

+        /*
+         * workaround for most likely hw bug
+         * if up to 4 bytes receive them in the descriptor instead
+         * of using dma_map_sg/dma_unmap_sg
+        */
+        if (small_read)
+            cmd_cfg |= CMD_CFG_DATA_NUM;
+
         data->bytes_xfered = 0;
         if (data->flags & MMC_DATA_WRITE)
             cmd_cfg |= CMD_CFG_DATA_WR;
@@ -533,7 +560,7 @@ static void meson_mmc_start_cmd(struct mmc_host
*mmc, struct mmc_command *cmd)
                 desc[i].cmd_cfg |= CMD_CFG_NO_CMD;
             desc[i].cmd_arg = cmd->arg;
             desc[i].cmd_resp = 0;
-            desc[i].cmd_data = sg_dma_address(sg);
+            desc[i].cmd_data = small_read ? 0 : sg_dma_address(sg);
         }
         desc[data->sg_count - 1].cmd_cfg |= CMD_CFG_END_OF_CHAIN;



Hallo Heiner,

here is  the result:
[   44.375027] brcmfmac: brcmf_sdiod_request_data rw=0, func=1,
addr=0x08000, nbytes=4
[   44.375034] mmc0: starting CMD53 arg 15000004 flags 000001b5
[   44.375037] mmc0:     blksz 4 blocks 1 flags 00000200 tsac 1000 ms
nsac 0
[   44.375047] mmc0: req done (CMD53): 0: 00001000 00000000 00000000
00000000
[   44.375050] mmc0:     4 bytes transferred: 0
[   44.375057] brcmfmac: brcmf_sdiod_regrl data:0x16044330
[   44.375059] brcmfmac: F1 signature read @0x18000000=0x16044330

[   44.375296] mmc0: starting CMD53 arg 15000004 flags 000001b5
[   44.375299] mmc0:     blksz 4 blocks 1 flags 00000200 tsac 1000 ms
nsac 0
[   44.375309] mmc0: req done (CMD53): 0: 00001000 00000000 00000000
00000000
[   44.375311] mmc0:     4 bytes transferred: 0
[   44.375323] brcmfmac: brcmf_sdiod_regrl data:0x00000000      <<<
expected 0x16044330
[   44.375327] brcmfmac: brcmf_chip_recognition found SB chip: BCM0,
rev=0
[   44.375330] brcmfmac: brcmf_chip_recognition: SB chip is not
supported
[   44.375333] brcmfmac: brcmf_sdio_probe_attach: brcmf_chip_attach
failed!
[   44.375337] brcmfmac: brcmf_sdio_probe: brcmf_sdio_probe_attach
failed
[   44.375339] brcmfmac: brcmf_sdio_remove Enter

the 1. cmd53 returned the correct data.
the second one failed. 0x0 instead of 0x16044330 (the same value as
in the 1. cmd53)

Helmut


Hallo Helmut,

next attempt .. After some more analysis I suspect a bug in SDIO core
regarding a not
properly aligned buffer (not aligned on 8 byte boundary what is needed
for 64 bit DMA).

I'd appreciate if you could test attached four patches on top of
latest next kernel.
Patches 1 + 2: smaller refactorings which have been submitted already
but are not applied yet
Patch 3: Fix for suspected issue in SDIO core
Patch 4: Switch to descriptor chain mode

By the way, if you go with latest next kernel:
STMMAC ethernet driver is broken currently, I had to go back to
version from March 10th.

Thanks in advance,
Heiner


Hallo Heiner,

i've applied the 4 patches to the current linux-next.

and

i think, you did it!

The adapter type is now detected und the firmware download succeeds.

wlan0 gets an ip-address via dhcp.

but: linux-next itself seems to have a problem with networking. even
without having loaded
meson-gx-mmc and brcmfmac the ethernet communication crashes.
so i cannot do any further tests.

the performance of mmc (28.7 MBytes/s) and sd-card (22.5 MBytes/s) is
much better then with to old driver (none sg-list)

i applied patch #3 to linux-amlogic and copied meson-gx-mmc.c from
linux-next. But failed to compile it

  drivers/mmc/host/meson-gx-mmc.c: In function ‘meson_mmc_pre_req’:
  drivers/mmc/host/meson-gx-mmc.c:199:9: error: implicit declaration of
function ‘mmc_get_dma_dir’ [-Werror=implicit-function-declaration]
         mmc_get_dma_dir(data));
         ^
  ./include/linux/dma-mapping.h:410:58: note: in definition of macro
‘dma_map_sg’
   #define dma_map_sg(d, s, n, r) dma_map_sg_attrs(d, s, n, r, 0)

Helmut



Hallo Heiner,

this evening i made some tests with linux-next. in the current version (2017-04-03)
your first two patches are already included. so i had to appl only 3 and 4.

tcp with the ethernet adpater is ok.

but with the wlan i get warnings
Nov  3 18:18:51 localhost kernel: [  134.461366] WARNING: CPU: 0 PID: 5 at drivers/mmc/host/meson-gx-mmc.c:536 meson_mmc_start_cmd+0x1c8/0x318
Nov  3 18:18:51 localhost kernel: [  134.461368] Modules linked in: brcmfmac cfg80211 brcmutil ir_nec_decoder meson_rng rng_core meson_gxbb_wdt evdev meson_ir rc_core
Nov  3 18:18:51 localhost kernel: [  134.461390] CPU: 0 PID: 5 Comm: kworker/u8:0 Not tainted 4.11.0-rc5-next-20170403+ #6
Nov  3 18:18:51 localhost kernel: [  134.461391] Hardware name: Netxeon MiniMX (DT)
Nov  3 18:18:51 localhost kernel: [  134.461491] Workqueue: brcmf_wq/mmc1:0001:1 brcmf_sdio_dataworker [brcmfmac]
Nov  3 18:18:51 localhost kernel: [  134.461494] task: ffffffc010252600 task.stack: ffffffc010278000
Nov  3 18:18:51 localhost kernel: [  134.461498] PC is at meson_mmc_start_cmd+0x1c8/0x318
Nov  3 18:18:51 localhost kernel: [  134.461502] LR is at meson_mmc_request+0x4c/0x98
Nov  3 18:18:51 localhost kernel: [  134.461504] pc : [<ffffff80083932d0>] lr : [<ffffff800839346c>] pstate: 20000145
Nov  3 18:18:51 localhost kernel: [  134.461505] sp : ffffffc01027b920
Nov  3 18:18:51 localhost kernel: [  134.461507] x29: ffffffc01027b920 x28: 000000000000005c
Nov  3 18:18:51 localhost kernel: [  134.461510] x27: 0000000000000001 x26: ffffffc072540ce0
Nov  3 18:18:51 localhost kernel: [  134.461514] x25: ffffffc074348000 x24: ffffffc01027bab0
Nov  3 18:18:51 localhost kernel: [  134.461518] x23: ffffff800873d000 x22: 00000000b54ca000
Nov  3 18:18:51 localhost kernel: [  134.461521] x21: ffffffc01027baf0 x20: ffffffc074348500
Nov  3 18:18:51 localhost kernel: [  134.461524] x19: 0000000000000000 x18: 0000000000000000
Nov  3 18:18:51 localhost kernel: [  134.461528] x17: 0000000000000000 x16: 0000000000000000
Nov  3 18:18:51 localhost kernel: [  134.461531] x15: ffffffc076f8ac80 x14: 0000000000000000
Nov  3 18:18:51 localhost kernel: [  134.461534] x13: 0000000000000000 x12: 0000000000000000
Nov  3 18:18:51 localhost kernel: [  134.461537] x11: 0000000000000000 x10: ffffff800053e508
Nov  3 18:18:51 localhost kernel: [  134.461541] x9 : ffffffc01027bb98 x8 : ffffffc01027bab0
Nov  3 18:18:51 localhost kernel: [  134.461544] x7 : 0000000000000000 x6 : 0000000000000001
Nov  3 18:18:51 localhost kernel: [  134.461548] x5 : 00000000ffffffff x4 : 0000000000000000
Nov  3 18:18:51 localhost kernel: [  134.461551] x3 : 000000000000003f x2 : ffffffc01027ba90
Nov  3 18:18:51 localhost kernel: [  134.461554] x1 : 0000000000000004 x0 : 000000000000005c
Nov  3 18:18:51 localhost kernel: [  134.461558] ---[ end trace f9fa3cc8e4194a72 ]---
Nov  3 18:18:51 localhost kernel: [  134.461560] Call trace:
Nov  3 18:18:51 localhost kernel: [  134.461564] Exception stack(0xffffffc01027b750 to 0xffffffc01027b880)
Nov  3 18:18:51 localhost kernel: [  134.461566] b740:                 0000000000000000 0000008000000000
Nov  3 18:18:51 localhost kernel: [  134.461570] b760: ffffffc01027b920 ffffff80083932d0 ffffff8008648000 ffffffc076f8ac80
Nov  3 18:18:51 localhost kernel: [  134.461573] b780: ffffffc010252600 ffffffc01027b848 ffffffc01027b7d0 ffffff80080c14b4
Nov  3 18:18:51 localhost kernel: [  134.461576] b7a0: ffffffc076f8ac80 ffffffc076f8ac80 ffffffc01027b7d0 ffffff8008084a3c
Nov  3 18:18:51 localhost kernel: [  134.461579] b7c0: ffffffc01027b7f0 ffffff800844b59c ffffff800864cb80 ffffffc076f8ac80
Nov  3 18:18:51 localhost kernel: [  134.461582] b7e0: ffffffc010252600 ffffffc0743fc000 000000000000005c 0000000000000004
Nov  3 18:18:51 localhost kernel: [  134.461585] b800: ffffffc01027ba90 000000000000003f 0000000000000000 00000000ffffffff
Nov  3 18:18:51 localhost kernel: [  134.461589] b820: 0000000000000001 0000000000000000 ffffffc01027bab0 ffffffc01027bb98
Nov  3 18:18:51 localhost kernel: [  134.461592] b840: ffffff800053e508 0000000000000000 0000000000000000 0000000000000000
Nov  3 18:18:51 localhost kernel: [  134.461595] b860: 0000000000000000 ffffffc076f8ac80 0000000000000000 0000000000000000
Nov  3 18:18:51 localhost kernel: [  134.461598] [<ffffff80083932d0>] meson_mmc_start_cmd+0x1c8/0x318
Nov  3 18:18:51 localhost kernel: [  134.461601] [<ffffff800839346c>] meson_mmc_request+0x4c/0x98
Nov  3 18:18:51 localhost kernel: [  134.461606] [<ffffff800837e540>] __mmc_start_request+0x50/0xf8
Nov  3 18:18:51 localhost kernel: [  134.461609] [<ffffff800837e6e0>] mmc_start_request+0xf8/0x230
Nov  3 18:18:51 localhost kernel: [  134.461612] [<ffffff800837efb4>] mmc_wait_for_req+0x8c/0x100
Nov  3 18:18:51 localhost kernel: [  134.461617] [<ffffff800838acd0>] mmc_io_rw_extended+0x268/0x290
Nov  3 18:18:51 localhost kernel: [  134.461621] [<ffffff800838c470>] sdio_io_rw_ext_helper+0xb8/0x1e8
Nov  3 18:18:51 localhost kernel: [  134.461625] [<ffffff800838c6b8>] sdio_memcpy_toio+0x20/0x28
Nov  3 18:18:51 localhost kernel: [  134.461704] [<ffffff800052dbb8>] brcmf_sdiod_buffrw.isra.4+0x68/0xa8 [brcmfmac]
Nov  3 18:18:51 localhost kernel: [  134.461781] [<ffffff800052f000>] brcmf_sdiod_send_pkt+0xb8/0x108 [brcmfmac]
Nov  3 18:18:51 localhost kernel: [  134.461857] [<ffffff800052a968>] brcmf_sdio_dataworker+0x600/0x2298 [brcmfmac]
Nov  3 18:18:51 localhost kernel: [  134.461863] [<ffffff80080b0478>] process_one_work+0x120/0x338
Nov  3 18:18:51 localhost kernel: [  134.461866] [<ffffff80080b06e0>] worker_thread+0x50/0x440
Nov  3 18:18:51 localhost kernel: [  134.461871] [<ffffff80080b628c>] kthread+0xfc/0x128
Nov  3 18:18:51 localhost kernel: [  134.461875] [<ffffff80080826c0>] ret_from_fork+0x10/0x50

it seems that not every sg buffer is 64 bit aligned.
detection of the wlan adapter, firmware download and dhcp are always successful. just as ping (up to frame sizes of 1500 bytes)
i have the feeling, that only tcp packets trigger the warning

Helmut


Hallo Helmut,

I'd like to understand better whether the unaligned buffer originates in the WiFi driver or in the SDIO core.
I'd appreciate if you could apply the following on top. It should print some (hopefully) helpful info
before the WARN_ON trace.

Thanks, Heiner

diff --git a/drivers/mmc/core/sdio_io.c b/drivers/mmc/core/sdio_io.c
index d40744bb..15cf9820 100644
--- a/drivers/mmc/core/sdio_io.c
+++ b/drivers/mmc/core/sdio_io.c
@@ -14,6 +14,7 @@
 #include <linux/mmc/card.h>
 #include <linux/mmc/sdio.h>
 #include <linux/mmc/sdio_func.h>
+#include <linux/mm.h>

 #include "sdio_ops.h"
 #include "core.h"
@@ -308,10 +309,18 @@ static int sdio_io_rw_ext_helper(struct sdio_func *func, int write,
     unsigned remainder = size;
     unsigned max_blocks;
     int ret;
+    void *buf_orig = buf;
+    unsigned size_orig = size;

     if (!func || (func->num > 7))
         return -EINVAL;

+    if (is_vmalloc_addr(buf))
+        pr_err("sdio: buffer is vmalloc memory\n");
+
+    if (((unsigned long)buf) & 7)
+        pr_err("sdio: buffer %p is not aligned\n", buf);
+
     /* Do the bulk of the transfer using block mode (if supported). */
     if (func->card->cccr.multi_block && (size > sdio_max_byte_size(func))) {
         /* Blocks per command is limited by host count, host transfer
@@ -326,6 +335,10 @@ static int sdio_io_rw_ext_helper(struct sdio_func *func, int write,
                 blocks = max_blocks;
             size = blocks * func->cur_blksize;

+            if (((unsigned long)buf) & 7)
+                pr_err("sdio: buf_orig: %p size_orig: %u buf: %p blocks: %u size: %u\n",
+                       buf_orig, size_orig, buf, blocks, func->cur_blksize);
+
             ret = mmc_io_rw_extended(func->card, write,
                 func->num, addr, incr_addr, buf,
                 blocks, func->cur_blksize);
@@ -343,6 +356,10 @@ static int sdio_io_rw_ext_helper(struct sdio_func *func, int write,
     while (remainder > 0) {
         size = min(remainder, sdio_max_byte_size(func));

+        if (((unsigned long)buf) & 7)
+            pr_err("sdio: buf_orig: %p size_orig: %u buf: %p blocks: %u size: %u blksz: %u\n",
+                   buf_orig, size_orig, buf, 0, size, func->cur_blksize);
+
         /* Indicate byte mode by setting "blocks" = 0 */
         ret = mmc_io_rw_extended(func->card, write, func->num, addr,
              incr_addr, buf, 0, size);


Hallo Heiner,

this is the ouput
[   58.509474] sdio: buffer ffffffc073c4448c is not aligned
[   58.509506] sdio: buf_orig: ffffffc073c4448c size_orig: 84 buf: ffffffc073c4448c blocks: 0 size: 84 blksz: 512
[   58.519115] ------------[ cut here ]------------
[   58.519130] WARNING: CPU: 0 PID: 5 at drivers/mmc/host/meson-gx-mmc.c:536 meson_mmc_start_cmd+0x1c8/0x318
[   58.519132] Modules linked in: brcmfmac cfg80211 ir_nec_decoder brcmutil meson_rng rng_core meson_gxbb_wdt evdev meson_ir rc_core
[   58.519153] CPU: 0 PID: 5 Comm: kworker/u8:0 Not tainted 4.11.0-rc5-next-20170403+ #7
[   58.519154] Hardware name: Netxeon MiniMX (DT)
[   58.519258] Workqueue: brcmf_wq/mmc1:0001:1 brcmf_sdio_dataworker [brcmfmac]
[   58.519269] task: ffffffc010252600 task.stack: ffffffc010278000
[   58.519276] PC is at meson_mmc_start_cmd+0x1c8/0x318
[   58.519280] LR is at meson_mmc_request+0x4c/0x98
[   58.519282] pc : [<ffffff8008393390>] lr : [<ffffff800839352c>] pstate: 20000145
[   58.519283] sp : ffffffc01027b910
[   58.519285] x29: ffffffc01027b910 x28: 0000000000008000
[   58.519288] x27: ffffffc073c444e0 x26: 0000000000000054
[   58.519292] x25: ffffffc0742fd800 x24: ffffffc01027baa0
[   58.519295] x23: ffffff800873d000 x22: 00000000b54ca000
[   58.519298] x21: ffffffc01027bae0 x20: ffffffc0742fdd00
[   58.519302] x19: 0000000000000000 x18: 0000000000000006
[   58.519306] x17: 0000000000000000 x16: 0000000000000000
[   58.519309] x15: ffffff80086f373d x14: 3463333730636666
[   58.519313] x13: 66666666203a6675 x12: 62203438203a6769
[   58.519316] x11: 726f5f657a697320 x10: ffffff80082d56b0
[   58.519337] x9 : ffffffc01027bb88 x8 : ffffffc01027baa0
[   58.519350] x7 : 0000000000000000 x6 : 0000000000000001
[   58.519354] x5 : 00000000ffffffff x4 : 0000000000000000
[   58.519357] x3 : 000000000000003f x2 : ffffffc01027ba80
[   58.519360] x1 : 0000000000000004 x0 : 0000000000000054
[   58.519364] ---[ end trace 22523b85322e1fa4 ]---
[   58.519366] Call trace:
[   58.519370] Exception stack(0xffffffc01027b740 to 0xffffffc01027b870)
[   58.519374] b740: 0000000000000000 0000008000000000 ffffffc01027b910 ffffff8008393390
[   58.519377] b760: ffffff80086f373d ffffffc01027b880 ffffffc01027b780 ffffff8008573945
[   58.519380] b780: ffffffc01027b800 ffffff8008273b38 0000000000000000 ffffff80086f3730
[   58.519383] b7a0: ffffff80086b7a60 ffffff80086f3730 0000000000000062 ffffff80086b7a60
[   58.519386] b7c0: 0000000000000000 0000000000000001 0000000000000400 0000000000000000
[   58.519389] b7e0: 0000000000000054 0000000000000004 ffffffc01027ba80 000000000000003f
[   58.519392] b800: 0000000000000000 00000000ffffffff 0000000000000001 0000000000000000
[   58.519395] b820: ffffffc01027baa0 ffffffc01027bb88 ffffff80082d56b0 726f5f657a697320
[   58.519398] b840: 62203438203a6769 66666666203a6675 3463333730636666 ffffff80086f373d
[   58.519401] b860: 0000000000000000 0000000000000000
[   58.519410] [<ffffff8008393390>] meson_mmc_start_cmd+0x1c8/0x318
[   58.519419] [<ffffff800839352c>] meson_mmc_request+0x4c/0x98
[   58.519428] [<ffffff800837e540>] __mmc_start_request+0x50/0xf8
[   58.519436] [<ffffff800837e6e0>] mmc_start_request+0xf8/0x230
[   58.519443] [<ffffff800837efb4>] mmc_wait_for_req+0x8c/0x100
[   58.519451] [<ffffff800838acd0>] mmc_io_rw_extended+0x268/0x290
[   58.519458] [<ffffff800838c558>] sdio_io_rw_ext_helper+0x1a0/0x2a8
[   58.519464] [<ffffff800838c778>] sdio_memcpy_toio+0x20/0x28
[   58.519552] [<ffffff8000594bb8>] brcmf_sdiod_buffrw.isra.4+0x68/0xa8 [brcmfmac]
[   58.519633] [<ffffff8000596000>] brcmf_sdiod_send_pkt+0xb8/0x108 [brcmfmac]
[   58.519715] [<ffffff8000591968>] brcmf_sdio_dataworker+0x600/0x2298 [brcmfmac]
[   58.519723] [<ffffff80080b0478>] process_one_work+0x120/0x338
[   58.519728] [<ffffff80080b06e0>] worker_thread+0x50/0x440
[   58.519733] [<ffffff80080b628c>] kthread+0xfc/0x128
[   58.519737] [<ffffff80080826c0>] ret_from_fork+0x10/0x50
[   59.517967] sdio: buffer ffffffc073c4448c is not aligned
[   59.517999] sdio: buf_orig: ffffffc073c4448c size_orig: 84 buf: ffffffc073c4448c blocks: 0 size: 84 blksz: 512


[  242.577369] sdio: buffer ffffffc0727bc484 is not aligned
[  242.577400] sdio: buf_orig: ffffffc0727bc484 size_orig: 92 buf: ffffffc0727bc484 blocks: 0 size: 92 blksz: 512
[  242.586996] ------------[ cut here ]------------
[  242.587010] WARNING: CPU: 0 PID: 5 at drivers/mmc/host/meson-gx-mmc.c:536 meson_mmc_start_cmd+0x1c8/0x318
[  242.587012] Modules linked in: brcmfmac cfg80211 ir_nec_decoder brcmutil evdev meson_rng rng_core meson_gxbb_wdt meson_ir rc_core
[  242.587035] CPU: 0 PID: 5 Comm: kworker/u8:0 Not tainted 4.11.0-rc5-next-20170403+ #7
[  242.587036] Hardware name: Netxeon MiniMX (DT)
[  242.587139] Workqueue: brcmf_wq/mmc1:0001:1 brcmf_sdio_dataworker [brcmfmac]
[  242.587151] task: ffffffc010252600 task.stack: ffffffc010278000
[  242.587159] PC is at meson_mmc_start_cmd+0x1c8/0x318
[  242.587162] LR is at meson_mmc_request+0x4c/0x98
[  242.587164] pc : [<ffffff8008393390>] lr : [<ffffff800839352c>] pstate: 20000145
[  242.587166] sp : ffffffc01027b910
[  242.587167] x29: ffffffc01027b910 x28: 0000000000008000
[  242.587171] x27: ffffffc0727bc4e0 x26: 000000000000005c
[  242.587174] x25: ffffffc0742e5000 x24: ffffffc01027baa0
[  242.587177] x23: ffffff800873d000 x22: 00000000b54ca000
[  242.587189] x21: ffffffc01027bae0 x20: ffffffc0742e5500
[  242.587197] x19: 0000000000000000 x18: 0000000000000006
[  242.587200] x17: 0000000000000000 x16: 0000000000000000
[  242.587208] x15: ffffff80086f373d x14: 3438346362373237
[  242.587211] x13: 3063666666666666 x12: 203a667562203239
[  242.587215] x11: 203a6769726f5f65 x10: ffffff80082d56b0
[  242.587218] x9 : ffffffc01027bb88 x8 : ffffffc01027baa0
[  242.587231] x7 : 0000000000000000 x6 : 0000000000000001
[  242.587236] x5 : 00000000ffffffff x4 : 0000000000000000
[  242.587239] x3 : 000000000000003f x2 : ffffffc01027ba80
[  242.587242] x1 : 0000000000000004 x0 : 000000000000005c
[  242.587246] ---[ end trace 7831fd15034513ca ]---
[  242.587249] Call trace:
[  242.587252] Exception stack(0xffffffc01027b740 to 0xffffffc01027b870)
[  242.587257] b740: 0000000000000000 0000008000000000 ffffffc01027b910 ffffff8008393390
[  242.587265] b760: ffffff80086f373d ffffffc01027b880 ffffffc010252600 ffffff8008573945
[  242.587271] b780: ffffffc01027b800 ffffff8008273b38 0000000000000000 ffffff80086f3730
[  242.587276] b7a0: ffffff80086b7b5c ffffff80086f3730 0000000000000062 ffffff80086b7b5c
[  242.587279] b7c0: 0000000000000000 0000000000000001 0000000000000400 0000000000000000
[  242.587282] b7e0: 000000000000005c 0000000000000004 ffffffc01027ba80 000000000000003f
[  242.587284] b800: 0000000000000000 00000000ffffffff 0000000000000001 0000000000000000
[  242.587288] b820: ffffffc01027baa0 ffffffc01027bb88 ffffff80082d56b0 203a6769726f5f65
[  242.587291] b840: 203a667562203239 3063666666666666 3438346362373237 ffffff80086f373d
[  242.587293] b860: 0000000000000000 0000000000000000
[  242.587296] [<ffffff8008393390>] meson_mmc_start_cmd+0x1c8/0x318
[  242.587299] [<ffffff800839352c>] meson_mmc_request+0x4c/0x98
[  242.587306] [<ffffff800837e540>] __mmc_start_request+0x50/0xf8
[  242.587310] [<ffffff800837e6e0>] mmc_start_request+0xf8/0x230
[  242.587313] [<ffffff800837efb4>] mmc_wait_for_req+0x8c/0x100
[  242.587318] [<ffffff800838acd0>] mmc_io_rw_extended+0x268/0x290
[  242.587322] [<ffffff800838c558>] sdio_io_rw_ext_helper+0x1a0/0x2a8
[  242.587325] [<ffffff800838c778>] sdio_memcpy_toio+0x20/0x28
[  242.587416] [<ffffff8000594bb8>] brcmf_sdiod_buffrw.isra.4+0x68/0xa8 [brcmfmac]
[  242.587498] [<ffffff8000596000>] brcmf_sdiod_send_pkt+0xb8/0x108 [brcmfmac]
[  242.587585] [<ffffff8000591968>] brcmf_sdio_dataworker+0x600/0x2298 [brcmfmac]
[  242.587597] [<ffffff80080b0478>] process_one_work+0x120/0x338
[  242.587607] [<ffffff80080b06e0>] worker_thread+0x50/0x440
[  242.587617] [<ffffff80080b628c>] kthread+0xfc/0x128
[  242.587625] [<ffffff80080826c0>] ret_from_fork+0x10/0x50
[  243.584080] sdio: buffer ffffffc0727bc484 is not aligned
[  243.584110] sdio: buf_orig: ffffffc0727bc484 size_orig: 92 buf: ffffffc0727bc484 blocks: 0 size: 92 blksz: 512

i haven't seen any lines with blocks != 0 or "sdio: buffer is vmalloc memory"

i can do udp based data transfers (iperf -s -u) without warning. doing the same with tcp (iperf -s) crashes the wlan interfaces.

Helmut


Thanks, Helmut. At least it's clear now that a not-8-byte-aligned buffer is provided from WiFi driver to SDIO core already.
In the MMC mailig list archive I found some (so far w/o result) discussions regarding SDIO and DMA.
Issue seems to be that buffers provided to SDIO core are used for DMA, but SDIO API doesn't enforce provided buffers
to be DMA-safe.
We can't change all SDIO device drivers, so we have to spend y few more thoughts on how to deal with this in SDIO core
and / or MMC controller drivers.

Heiner


Hallo Heiner,

i found a solution:

diff --git a/drivers/net/wireless/broadcom/brcm80211/brcmfmac/sdio.c b/drivers/net/wireless/broadcom/brcm80211/brcmfmac/sdio.c
index a999f95062c7..b966217f3a6c 100644
--- a/drivers/net/wireless/broadcom/brcm80211/brcmfmac/sdio.c
+++ b/drivers/net/wireless/broadcom/brcm80211/brcmfmac/sdio.c
@@ -540,7 +540,7 @@ static int qcount[NUMPRIO];
 /* Limit on rounding up frames */
 static const uint max_roundup = 512;

-#define ALIGNMENT  4
+#define ALIGNMENT  8

 enum brcmf_sdio_frmtype {
        BRCMF_SDIO_FT_NORMAL,

Just by chance I came across the same setting today and was about to ask you to test it.
Very nice that it solves the issue. However most likely we have to make this chance only
if the platform uses 64 bit DMA.
I will submit a patch with your "Suggested-by".

Regarding your performance figures:
How was it with the mainline driver before the switch to descriptor chain mode?

Heiner



Hallo Heiner,

both driver versions (pure linux-next and the descriptor chain) show the same speeds. when positioning my box directly in front of my router (fritz!box 7390), i get
sending: 17.3 MBit/s
receiving: 34 MBit/s

the amlogic kernel (3.14.73) gets 40.5 MBit/s in both directions.

Helmut

--
To unsubscribe from this list: send the line "unsubscribe linux-mmc" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at  http://vger.kernel.org/majordomo-info.html




[Index of Archives]     [Linux USB Devel]     [Linux Media]     [Video for Linux]     [Linux Audio Users]     [Yosemite News]     [Linux Kernel]     [Linux SCSI]

  Powered by Linux