Re: [PATCH v4 0/3] about data busy

[addy ke <addy.ke@rock-chips.com>]

Hi, Javier and Alim

These days are Spring Festival holiday.
Sorry for late reply.

On 2015/2/15 19:41, Javier Martinez Canillas wrote:
> Hello Addy,
> 
> On Sat, Feb 14, 2015 at 7:17 AM, Addy Ke <addy.ke@rock-chips.com> wrote:
>> patch 1: This patch can fix bug that controller is still data busy after
>>          reset all blocks. After this patch, I still get data busy in
>>          set_ios().
>>
>> patch 2: This patch fix bug 'Timeout sending command'. After patch1 and
>>          patch2, there is no mmc errors after:
>>          cd /sys/bus/platform/drivers/dwmmc_rockchip
>>          for i in $(seq 1 10000); do
>>                 echo "========================" $i
>>                 echo ff0c0000.dwmmc > unbind
>>                 sleep .5
>>                 echo ff0c0000.dwmmc > bind
>>                 sleep 2
>>         done
>>
>> patch3: This patch fix bug that there is data busy before sdio send CMD53.
>>         But This patch is necessary for sd and mmc too.
>>
> 
> I faced the same 'Timeout sending command' error when trying to enable
> support for the SDIO wifi chip attached to mmc@12210000 (mmc1) on an
> Exynos5420 Peach Pit Chromebook. On booting the kernel log shows:
> 
> mmc_host mmc1: Timeout sending command (cmd 0x202000 arg 0x0 status 0x80202000)
> 
> 0x202000 == SDMMC_CMD_UPD_CLK | SDMMC_CMD_PRV_DAT_WAIT so your patch
> #2 dw_mci_setup_bus() avoids the mmc comand to time out. However, it
> has a side effect since with your series the uSD that in mmc@12220000
> (mmc2) fails to be detected and the kernel log shows:
> 
> [    5.466432] Waiting for root device /dev/mmcblk1p4...
> [  240.169436] INFO: task kworker/u16:1:50 blocked for more than 120 seconds.
> [  240.174844]       Not tainted
> 3.19.0-next-20150211-00006-g045d4aba96ce-dirty #476
> [  240.182302] "echo 0 > /proc/sys/kernel/hung_task_timeout_secs"
> disables this message.
> [  240.190109] kworker/u16:1   D c04c2710     0    50      2 0x00000000
> [  240.196446] Workqueue: kmmcd mmc_rescan
> [  240.200249] [<c04c2710>] (__schedule) from [<c04c2ac0>] (schedule+0x34/0x98)
> [  240.207290] [<c04c2ac0>] (schedule) from [<c04c6568>]
> (schedule_timeout+0x120/0x16c)
> [  240.215009] [<c04c6568>] (schedule_timeout) from [<c04c3584>]
> (wait_for_common+0xb0/0x154)
> [  240.223251] [<c04c3584>] (wait_for_common) from [<c038a5ac>]
> (mmc_wait_for_req+0xa0/0x140)
> [  240.231492] [<c038a5ac>] (mmc_wait_for_req) from [<c038a6d4>]
> (mmc_wait_for_cmd+0x88/0xa8)
> [  240.239735] [<c038a6d4>] (mmc_wait_for_cmd) from [<c03905b0>]
> (mmc_go_idle+0x78/0xf8)
> [  240.247540] [<c03905b0>] (mmc_go_idle) from [<c038c578>]
> (mmc_rescan+0x254/0x300)
> [  240.255003] [<c038c578>] (mmc_rescan) from [<c00346e8>]
> (process_one_work+0x120/0x324)
> [  240.262897] [<c00346e8>] (process_one_work) from [<c0034a58>]
> (worker_thread+0x138/0x464)
> [  240.271048] [<c0034a58>] (worker_thread) from [<c0039070>]
> (kthread+0xd8/0xf4)
> [  240.278254] [<c0039070>] (kthread) from [<c000e680>]
> (ret_from_fork+0x14/0x34)
> 
> 
> By enabling debug I see that the card is detected in dw_mci_get_cd() though.
> 
> Alim suggested [0] that dw_mci_wait_busy() should be called in
> mci_send_cmd() instead dw_mci_setup_bus() because the controller hangs
> when when sending update clock cmd in different cases.
> 
> I modified [1] your patch #2 to do what Alim suggested and only with
> that patch on top of linux-next I have neither the the "Timeout
> sending command" error nor the uSD not getting detected errors. Linux
> mounts the rootfs from the uSD and the wifi SDIO device is enumerated
> and listed in /sys/bus/sdio/devices/
> 
> Does that also solve your issue?

After merge Alim patch,and set re_try 8,
it can pass test by:
cd /sys/bus/platform/drivers/dwmmc_rockchip
for i in $(seq 1 10000); do
  echo "========================" $i
  echo ff0c0000.dwmmc > unbind
  sleep .5
  echo ff0c0000.dwmmc > bind
  sleep 2
done

My card is ADATA UHS-1 card(SDR50).
The maximum retry count is 6.

[ 1146.907596] mmc1: card 59b4 removed
[ 1147.421036] dwmmc_rockchip ff0c0000.dwmmc: Using internal DMA controller.
[ 1147.427827] dwmmc_rockchip ff0c0000.dwmmc: Version ID is 270a
[ 1147.433958] dwmmc_rockchip ff0c0000.dwmmc: DW MMC controller at irq 64, 32 bit host data width, 256 deep fifo
[ 1147.444269] dwmmc_rockchip ff0c0000.dwmmc: Got CD GPIO #221.
[ 1147.450381] dwmmc_rockchip ff0c0000.dwmmc: Got WP GPIO #226.
[ 1147.456046] ff0c0000.dwmmc supply card-external-vcc not found, using dummy regulator
[ 1148.519400] dwmmc_rockchip ff0c0000.dwmmc: Data busy (status 0x206)
[ 1149.019451] dwmmc_rockchip ff0c0000.dwmmc: Data busy (status 0x206)
[ 1149.519382] dwmmc_rockchip ff0c0000.dwmmc: Data busy (status 0x206)
[ 1150.019492] dwmmc_rockchip ff0c0000.dwmmc: Data busy (status 0x206)
[ 1150.519442] dwmmc_rockchip ff0c0000.dwmmc: Data busy (status 0x206)
>>>>>>>>>>>>>>>>>>>>>>>>> if re_try is 5, I still get "Timeout sending command".
[ 1150.525711] mmc_host mmc1: Timeout sending command (cmd 0x202000 arg 0x0 status 0x80202000)
[ 1150.534723] rockchip-iodomain io-domains.25: Setting to 3300000 done

So re_try must bigger than 6, but I don't known which value is reansonable.
Do you have any idear about it?

This is the patch Alim suggests:
+       int re_try = 8; /* just random for now, 1 re-try should be ok */

-       mci_writel(host, CMDARG, arg);
-       wmb();
-       mci_writel(host, CMD, SDMMC_CMD_START | cmd);
+       while(re_try--) {
+               mci_writel(host, CMDARG, arg);
+               wmb();
+               mci_writel(host, CMD, SDMMC_CMD_START | cmd);

-       while (time_before(jiffies, timeout)) {
-               cmd_status = mci_readl(host, CMD);
-               if (!(cmd_status & SDMMC_CMD_START))
-                       return;
+               while (time_before(jiffies, timeout)) {
+                       cmd_status = mci_readl(host, CMD);
+                       if (!(cmd_status & SDMMC_CMD_START))
+                               return;
+               }
+
+               dw_mci_wait_busy(slot);
> 
> Best regards,
> Javier
> 
> [0]: https://lkml.org/lkml/2015/2/10/353
> [1]: http://paste.debian.net/plain/148794
> 
> 
>