Journal is not the reason. It's just the first one to hit the bottleneck.

It's not logging anything unusual, just normal trafiic.

But following your suggestion I checked this card in my Lenovo W541 running F25. I've got average write speed of around 7MB/s.

When I sent the same file (2.4 GiB zip) to my PI, and it nearly killed it - 1.5 core constantly occupied by system, rest is blocked by IO waits. But write speed wasn't so much worse: 4.1 MiB/s.

Just for curiosity, I wrote the same file onto USB stick attached to my PI - this time it was slower (3.2 MiB/s), but system seemed to be more responsive.

So writing is CPU intensive and that makes me wonder what is use_spi_crc option responsible for in mmc_core kernel module and how to check it's current value.

After quick look into module's code I see it is supposed to do some data validation with crc which I think may lead to lots of CPU intensive operations in kernel mode ?

How do I check "use_spi_crc" current value? I couldn't find any obvious files in /proc or /sys filesystems.

If it's value is "true" than how do I set it to false? Is proper /etc/modprobe.d file entry sufficient or should I play with kernel parameters at startup since it is responsible for accessing root filesystem ?

regards,

marcinek

2017-05-08 14:55 GMT+02:00 Richard Ryniker <ryniker@alum.mit.edu>:

If the slowdown is due to log writes, what is written into your journal?

It may be awkward to access the journal using the RPi, but move the SD
card to another Fedora system (many laptops have flash card readers built
in, or use a USB device) and a "smoking gun" may be obvious. The journal
format does not depend on machine architecture. Use the --directory=
argument for journalctl.