[U-Boot] [PATCH] mmc: dw_mmc: fix timeout calculate method

[Kever Yang]

Hi Alexey,


On 2019/8/29 下午5:21, Alexey Brodkin wrote:
> Hi Kever,
>
>> -----Original Message-----
>> From: Kever Yang <kever.yang at rock-chips.com>
>> Sent: Thursday, August 29, 2019 10:17 AM
>> To: Peng Fan <peng.fan at nxp.com>; u-boot at lists.denx.de
>> Cc: trini at konsulko.com; Eugeniy.Paltsev at synopsys.com; Alexey Brodkin <abrodkin at synopsys.com>; Simon
>> Glass <sjg at chromium.org>
>> Subject: Re: [PATCH] mmc: dw_mmc: fix timeout calculate method
>>
>> Add ARC maintainers.
> Thanks for adding us in the loop.
>
>> On 2019/8/27 下午5:58, Peng Fan wrote:
>>> Hi Kever,
>>>
>>>> -----Original Message-----
>>>> From: U-Boot <u-boot-bounces at lists.denx.de> On Behalf Of Peng Fan
>>>> Sent: 2019年8月27日 15:40
>>>> To: Kever Yang <kever.yang at rock-chips.com>; u-boot at lists.denx.de
>>>> Cc: trini at konsulko.com
>>>> Subject: Re: [U-Boot] [PATCH] mmc: dw_mmc: fix timeout calculate method
>>>>
>>>>> Subject: [PATCH] mmc: dw_mmc: fix timeout calculate method
>>>>>
>>>>> There are two cases not been considered:
>>>>> - use uint for timeout, it will overflow when size bigger than 512KB for
>>>>>     it *8*1000 at the beginning, but we may use size up to 32MB;
>>>>> - The timeout is using clock speed for data rate, but the device may not
>>>>>     have such high speed, eg. clock is 52MHz while the device write speed
>>>> may
>>>>>     be less than 10MB/s.
>>>>>
>>>>> Fix them in this patch, the max timeout is about 6500 when size is
>>>>> 32MB after fix.
>>>>>
>>>>> Signed-off-by: Kever Yang <kever.yang at rock-chips.com>
>>> This patch breaks build, I need drop it.
>>> See build:
>>> https://travis-ci.org/MrVan/u-boot/jobs/577210464
>> The error happens only in ARC architecture, and seems like the arc
>> compiler not support 64bit div?
> It's not a compiler problem but a subtle issue with libgcc.
> To be independent from toolchain type (and we have quite a few flavors like
> little- and big-endian, ARCv1 or ARCv2 etc) decision was made to move
> libgcc into U-Boot but given full libgcc is quite large only primitives required
> by up-to-date U-Boot were copied from GCC to U-Boot.
>
>> Any idea on how to fix this?
> Add missing primitives like __udivdi3() and their dependencies, see for example:
> https://github.com/gcc-mirror/gcc/blob/master/libgcc/libgcc2.c#L1318.
>
> And I'll happily do it myself but for some reason I cannot reproduce that problem locally.
> I use the same pre-built arc-2018.09 toolchain:
> ---------------------------->8-----------------------
> # arc-linux-gcc -v
> gcc version 8.2.1 20180814 (ARCv2 ISA Linux uClibc toolchain 2018.09)
> ---------------------------->8-----------------------
>
> And may successfully build U-Boot for HSDK from
> https://github.com/MrVan/u-boot/commits/mmc-8-28-2019:
> ---------------------------->8-----------------------
> # git l
> 76c0bf2873 - (HEAD -> mrvan, github-mrvan/mmc-8-28-2019) ARM: dts: ast2500: Add SDHCI nodes (27 hours ago) <Eddie James>
> 07b000cd32 - configs: AST2500 EVB: Enable SD controller (27 hours ago) <Eddie James>
> 5f6a033f10 - mmc: Add Aspeed SD controller driver (27 hours ago) <Eddie James>
> a6503eaa66 - clk: aspeed: Add support for SD clock (27 hours ago) <Eddie James>
> c54e328452 - mmc: Rename timeout parameters for clarification (27 hours ago) <Sam Protsenko>
> be949bb19c - mmc: Fix timeout values passed to mmc_wait_dat0() (27 hours ago) <Sam Protsenko>
> 49842f0edf - dm: mmc: remove unused U_BOOT_DRIVER(mmc) (27 hours ago) <Andy Yan>
> 7b91b47f8f - Prepare v2019.10-rc3 (27 hours ago) <Tom Rini>
> ---------------------------->8-----------------------

I think this tree does not including this patch, Peng drop it because of 
this issue,
so you need to apply this patch in your branch to reproduce the problem.
I have send out V2 patch for this fix with only using 32bit variable 
instead of 64bit.

Thanks,
- Kever
>
> That way:
> ---------------------------->8-----------------------
> ./tools/buildman/buildman hsdk
> boards.cfg is up to date. Nothing to do.
> Building current source for 1 boards (1 thread, 48 jobs per thread)
>         arc:  w+   hsdk
>      0    1    0 /1      hsdk
> ---------------------------->8-----------------------
>
> What do I do wrong?
>
> -Alexey
>
>