[Bug 244356] Writing to a USB 3.0 stick is very slow

[bugzilla-noreply at freebsd.org]

https://bugs.freebsd.org/bugzilla/show_bug.cgi?id=244356

--- Comment #53 from Olivier Certner <olivier.freebsd at free.fr> ---
(In reply to Sebastien Boisvert from comment #44)

Hi Sébastien,

And first of all, I wanted to say this earlier, maybe now it matters even more:
I did not want to hijack your initial bug report in any way, so I hope you
don't feel that way. Given your initial reports and my experiments, I thought
that we were seeing the same exact problem (at least, the same symptoms). Since
then, more testing has given birth to a much more contrasted picture, at least
on my side. As said in comment #45, SD_64G and SD_128G seem not to work under
`usbtest` (the first at all, the second one when writing), so just may be bad,
and I'm trying to RMA them (initial request sent). The situation for KT_32G
seems more complex. Indeed, it reliably passes `usbtest`, *except* when I/O
size is set to "Random" (the default), in which case no transfer happens and
the stick simply disconnects from the bus. Plus, `iostat` reports a pattern
that I find quite weird. If you're OK, I'm proposing you to do some tests with
your USB 3 sticks (see below), to dig deeper and know to which extent we are
indeed seeing the same behaviors. If you're annoyed by my follow-ups, I will
happily open another PR and put information there.

So, it would be great if you could run the following tests on your USB 3
sticks, and in particular the two models of Kingston.
1. `usbtest` umass: At least, write only, read only, read+write, with I/O size
Random and Increasing.
2. `dd` test (see comment #52), with `iostat -w 1 <stick_device>` running
elsewhere (and reporting both outputs).

Thanks and sorry again if I bothered you.

-- 
You are receiving this mail because:
You are the assignee for the bug.