Drive looks to be in perfect health according to all of the SMART attributes. Your issue is probably due to daemons running on Linux in the background which are doing I/O to your drive. dmesg, smartctl, df, etc. aren't going to help diagnose this. Things like sar or iostat or top -b in a while loop logging to a file might. Furthermore, you should investigate TRIM support on Linux -- it is something you do have to turn on, if I remember right (I'm a FreeBSD guy, not Linux).
00:05.0 IDE interface: nVidia Corporation MCP55 SATA Controller (rev a2)
There's your problem. There are compatibility issues between nvidias fake-AHCI and SSDs. The usual solution is to run the controller in IDE mode and make sure to disable Command Queueing in the driver. How you would do on Linux I have no idea.
Thanks for replies.
I'll check the TRIM.
Two years ago I had a Kingston SNV125-S2/40GB with Intel's hardware inside but without TRIM support and that issue didn't occur... (the same controller and operating system). What do you think now ?
Maybe it's elevator=noop ?
[noop] deadline cfq
I'll check it too.
Beside I forgot to write about the scenerio this issue. It happens (rarely but...) when I unpacking kernel or installing compiled app/package.