Le 19/12/2014 10:08, Maxime Ripard a écrit :
More seriously, what makes you think that the AHCI controller itself
has a flaw? Did you try the same disk on the same kernel configuration
on another ARM SoC? The same kernel on the same SoC with a different
disk? A different kernel with same disk on another machine? Maxime
Hi everybody,
I revive this thread , I also see this odd sata behaviour.
I also have banana pi, but different disks ; here, an SSD (Crucial M500) :
Read is very good :
dd if=/dev/SSD/bench of=/dev/null bs=4M
469762048 octets (470 MB) copiés, 2,73443 s, 172 MB/s
But writing is not that great :
dd if=/dev/zero of=/dev/SSD/bench bs=4M
176160768 octets (176 MB) copiés, 4,57323 s, 38,5 MB/s
I also tried with 7200k RPM disk with globally similar results... Seems
like, in my case, 40 MB/s is hard limit concerning I/O writes.
It doesn't seems to be CPU-bound.
I don't have another ARM SoC with SATA interface for testing, but I can
guarantee this SSD is giving much higher value for writes on X86 boards.
This is with custom-made 3.19.0 kernel, debian jessie, but was observed
with "original banana pi kernels" and older vanilla kernel.
(BTW thanks to all developpers who participated in the mainlining
effort, that's greatly appreciated)
As I understand it, there is lack of documentation of AHCI from
allwinner, so of course it's hard to tell if the problem lies in
software or hardware.
Anyway, just in case, this is a "mee too" report, for what it's worth.
Cheers,
--
You received this message because you are subscribed to the Google Groups
"linux-sunxi" group.
To unsubscribe from this group and stop receiving emails from it, send an email
to [email protected].
For more options, visit https://groups.google.com/d/optout.
