On Mon, 28 Jan 2013 06:39:07 -0800, Greg KH wrote:

> On Mon, Jan 28, 2013 at 12:42:16PM +0100, Holger Hoffstaette wrote:
>> On Sun, 27 Jan 2013 20:56:58 -0800, Greg KH wrote:
>> 
>> > I'm announcing the release of the 3.7.5 kernel.
>> 
>> Apologies for being late but I'm afraid I just found at least one
>> regression in this release.
>> 
>> I have an external HDD in an USB3 enclosure for the occasional backup. I
>> turn it on, it gets registered as sdX, I mount it/use it/unmount it and
>> turn it off, which worked fine with kernels including 3.7.4. With 3.7.5

FWIW it didn't really work with 3.7.4 either - just didn't notice before.

>> the kernel seems no longer notified of the changed device power status;
>> nothing in dmesg and the device stays in the list of SCSI devices as
>> e.g. shown by lsscsi. This is repeatable. Turning the device back
>> on/using it works fine, it's just that I cannot dynamically remove it
>> any longer.
>> 
>> I cannot run a bisect at the moment, but reading the changelog points to
>> a few USB and libata related changes. I hope this helps.
> 
> If you could run 'git bisect' to determine which patch causes this
> problem, that would be wonderful.

It took a while (original box is a server that's needed for "work"), but I
could reproduce it with a different USB 3.0 card/external device on a free
and sufficiently beefy machine. See below.

> Also, does 3.8-rc5 work for you properly or not on this hardware?

No, current 3.8-rc5 did not work either. I first found 3.7.2 OK and 3.7.3
bad. Bisecting found:

--snip--
f7965c0846d74b270e246c1470ca955d5078eb07 is the first bad commit
commit f7965c0846d74b270e246c1470ca955d5078eb07
Author: Sarah Sharp <[email protected]>
Date:   Wed Nov 14 17:58:04 2012 -0800

    USB: Handle warm reset failure on empty port.
    
    commit 65bdac5effd15d6af619b3b7218627ef4d84ed6a upstream.
    
    An empty port can transition to either Inactive or Compliance Mode if a
    newly connected USB 3.0 device fails to link train.  In that case, we
    issue a warm reset.  Some devices, such as John's Roseweil eusb3
    enclosure, slip back into Compliance Mode after the warm reset.
    [..]
--snip--

Reverting this from 3.7.3 all the way to 3.8-rc5 makes USB 3.0
disconnection/power-off device removal work again. Verified both on the
bisection box and by now also the original machine.

Hope this helps :)

cheers
Holger


--
To unsubscribe from this list: send the line "unsubscribe stable" in
the body of a message to [email protected]
More majordomo info at  http://vger.kernel.org/majordomo-info.html

Reply via email to