[meta-xilinx] Really bad SD/eMMC performance after upgrading 2018.1 to 2018.3

[Mike Looijmans]

On 26-03-19 21:03, Manjukumar Harthikote Matha wrote:
> Hi Mike and all,
> 
> Are there any specific configurations? PMU/Linux ?

What is it you need to know?

If the problem were in Linux, i could easily bisect it.

The problem is caused by the PMU firmware, which isn't as easy to compile and 
deploy, while there are also interdependencies with the ATF causing even more 
headaches.

I saw there were tons of changes in the clocking interface, and I suspect the 
cause is in there.

The clock tree for our board is likely to be completely different from what 
the zcu102 is using.


> 
> We are not able to replicate this issue, we tested it on zcu102 board
> 
> Thanks,
> Manju
> 
>> -----Original Message-----
>> From: meta-xilinx-bounces at yoctoproject.org [mailto:meta-xilinx-
>> bounces at yoctoproject.org] On Behalf Of Mike Looijmans
>> Sent: Thursday, March 21, 2019 10:09 AM
>> To: Peter Smith <salerio at gmail.com>
>> Cc: meta-xilinx at lists.yoctoproject.org
>> Subject: Re: [meta-xilinx] Really bad SD/eMMC performance after upgrading 2018.1
>> to 2018.3
>>
>> Have only tried the MPSoC.
>>
>> I haven't tried the 7k yet with 2018.3, may get to that tomorrow. I
>> suspect it's something with the PMU interface and/or clock dividers.
>>
>>
>> On 21-03-19 17:54, Peter Smith wrote:
>>> Mike, although I have not made any measurements I have noticed in the
>>> last few days (coincidentally I have been doing a lot of work with a
>>> ZCU102) that the SD interface seemed a bit slow. Are you seeing this on
>>> Zynq 7K or US+ MPSoC?
>>> Best Regards
>>> Peter
>>>
>>>
>>> On Thu, 21 Mar 2019 at 16:38, Mike Looijmans <mike.looijmans at topic.nl
>>> <mailto:mike.looijmans at topic.nl>> wrote:
>>>
>>>      The SD and eMMC controller has become terribly slow in 2018.3:
>>>
>>>      # echo 3 > /proc/sys/vm/drop_caches
>>>
>>>      # dd if=/dev/mmcblk0 of=/dev/null bs=1M count=20
>>>      20+0 records in
>>>      20+0 records out
>>>      20971520 bytes (20.0MB) copied, 4.972666 seconds, 4.0MB/s
>>>
>>>      # dd if=/dev/mmcblk1 of=/dev/null bs=1M count=20
>>>      20+0 records in
>>>      20+0 records out
>>>      20971520 bytes (20.0MB) copied, 38.899436 seconds, 526.5KB/s
>>>
>>>
>>>      This used to be over 160MB/s for the eMMC (mmcblk0) and over 20 MB/s
>>>      for the
>>>      SD card (mmcblk0) with the 2018.1 kernel and bootloader.
>>>
>>>      This is only in Linux, the bootloader still reads at about 10MB/s
>>>      from the SD
>>>      card. So apparently the clocks are set up okay.
>>>
>>>      According to the kernel, the clocks and interface are running at
>>>      full speed:
>>>
>>>      # cat /sys/kernel/debug/mmc0/ios
>>>      clock:          200000000 Hz
>>>      actual clock:   200000000 Hz
>>>      vdd:            21 (3.3 ~ 3.4 V)
>>>      bus mode:       2 (push-pull)
>>>      chip select:    0 (don't care)
>>>      power mode:     2 (on)
>>>      bus width:      3 (8 bits)
>>>      timing spec:    9 (mmc HS200)
>>>      signal voltage: 1 (1.80 V)
>>>      driver type:    0 (driver type B)
>>>
>>>      # cat /sys/kernel/debug/mmc1/ios
>>>      clock:          50000000 Hz
>>>      actual clock:   50000000 Hz
>>>      vdd:            21 (3.3 ~ 3.4 V)
>>>      bus mode:       2 (push-pull)
>>>      chip select:    0 (don't care)
>>>      power mode:     2 (on)
>>>      bus width:      2 (4 bits)
>>>      timing spec:    2 (sd high-speed)
>>>      signal voltage: 0 (3.30 V)
>>>      driver type:    0 (driver type B)
>>>
>>>
>>>      Any insights?
>>>      --
>>>      _______________________________________________
>>>      meta-xilinx mailing list
>>>      meta-xilinx at yoctoproject.org <mailto:meta-xilinx at yoctoproject.org>
>>>      https://lists.yoctoproject.org/listinfo/meta-xilinx
>>>
>>
>>
>> --
>> Mike Looijmans
>> --
>> _______________________________________________
>> meta-xilinx mailing list
>> meta-xilinx at yoctoproject.org
>> https://lists.yoctoproject.org/listinfo/meta-xilinx