Bug 12301

Created attachment 19486 [details]
lsusb output

Sorry, i've a clerical error, i reload the ehci_hcd module

if i skip the '24c0996a6b73e2554104961afcc8659534503e0d' commit, than the fingerprint reader is work

Reply-To: akpm@linux-foundation.org


(switched to email.  Please respond via emailed reply-to-all, not via the
bugzilla web interface).

On Fri, 26 Dec 2008 01:41:57 -0800 (PST) bugme-daemon@bugzilla.kernel.org wrote:

> http://bugzilla.kernel.org/show_bug.cgi?id=12301
> 
>            Summary: Fingerprint reader doesn't respod after the first use
>            Product: Drivers
>            Version: 2.5
>      KernelVersion: >2.6.28-rc8-git4
>           Platform: All
>         OS/Version: Linux
>               Tree: Mainline
>             Status: NEW
>           Severity: normal
>           Priority: P1
>          Component: USB
>         AssignedTo: greg@kroah.com
>         ReportedBy: roka@dawid.hu
> 
> 
> Latest working kernel version: 2.6.28-rc8-00067-g9a9fafb
> Earliest failing kernel version:
> Distribution: Gentoo
> Hardware Environment: HP 2510p, aes2501 fprint reader
> Software Environment:
> Problem Description:
> The fingerprint reader doesn't respod after the first use, until I reload the
> 'ehci-hci' module.
> After some bisecting I think the problem is in the
> '24c0996a6b73e2554104961afcc8659534503e0d' commit.
> Steps to reproduce: I use the fingerprint reader
> 

A regression caused by:

commit 24c0996a6b73e2554104961afcc8659534503e0d
Author:     Alan Stern <stern@rowland.harvard.edu>
AuthorDate: Mon Dec 1 10:24:41 2008 -0500
Commit:     Greg Kroah-Hartman <gregkh@suse.de>
CommitDate: Wed Dec 17 10:49:14 2008 -0800

    USB: skip Set-Interface(0) if already in altsetting 0

Please attach 

On Fri, 26 Dec 2008, Andrew Morton wrote:

> > http://bugzilla.kernel.org/show_bug.cgi?id=12301
> > 
> >            Summary: Fingerprint reader doesn't respod after the first use

> > The fingerprint reader doesn't respod after the first use, until I reload
> the
> > 'ehci-hci' module.
> > After some bisecting I think the problem is in the
> > '24c0996a6b73e2554104961afcc8659534503e0d' commit.

> Author:     Alan Stern <stern@rowland.harvard.edu>
> AuthorDate: Mon Dec 1 10:24:41 2008 -0500
> Commit:     Greg Kroah-Hartman <gregkh@suse.de>
> CommitDate: Wed Dec 17 10:49:14 2008 -0800
> 
>     USB: skip Set-Interface(0) if already in altsetting 0

Please attach to the bug report some usbmon traces showing what happens
when you try to use the fingerprint reader more than once: one trace
with this commit present and one with the commit dropped.  Instructions 
for usbmon are in the kernel source file Documentation/usb/usbmon.txt.

Alan Stern

I have the same problem, with a Thinkpad x06s and 0483:2016 SGS Thomson Microelectronics Fingerprint Reader, and I just located the same commit before I found this bug..

Will do usbmon trace tomorrow.

Created attachment 19505 [details]
usbmon of tf-tool --acquire with stock 2.6.28 kernel

Created attachment 19506 [details]
usbmon of tf-tool --acquire with 2.6.28 kernel with commit 24c099.. reverted

tf-tool is part of the thinkfinger(.sf.net) software.

Actually, my UPEK TouchStrip-reader does not even work once with the stock 2.6.28 kernel. The tf-tool fails before I can scan at all.

Created attachment 19510 [details]
usbmon of attempted login with swipe, stock 2.6.28

Created attachment 19511 [details]
usbmon log of working login swipe, 2.6.28 with commit 24c099 reverted

Created attachment 19550 [details]
Patch to enable interface after driver release

For those who didn't see it in the mailing list, this patch fixes the problem.  It will be submitted and should end up in a 2.6.28.stable kernel.

Works for me.

It works here, too, but thuinkfinger now it is really slow.
After doing "su", it takes up to 10 seconds until 
"Password or swipe finger:" appears, and then it takes up to another 10 seconds, until it accepts finger sweeps.
This is gentoo on a Thinkpad T60,
Bus 005 Device 003: ID 0483:2016 SGS Thomson Microelectronics Fingerprint Reader
latest git kernel, libusb 0.1.12

Oh yeah, my english really rocks today ... sorry for that.
I straced a su command, and there are thousands of

[pid 30406] select(5, NULL, [], NULL, {0, 1000}) = 0 (Timeout)
[pid 30406] ioctl(4, USBDEVFS_REAPURBNDELAY, 0x7fff3d1d0998) = -1 EAGAIN (Resource temporarily unavailable)

lines there. Don't know if thats the problem, but it looks strange to me.

This could be caused by the bad code in libthinkfinger. See http://marc.info/?l=linux-usb&m=123067293718806&w=2

Created attachment 19606 [details]
don't send a direct Set-Config request in _libthinkfinger_usb_hello

Dieter, can you please try the attached patch?  It's yet untested as I don't have the fingerprint reader device at the moment.

Thank you!

Works perfectly. Thanks! 
Should I provide further information, or is a Tested By all you need?

No, that's fine.

I'll do a maintenance release of ThinkFinger in the next few days.

Thanks a lot for testing, Dieter.

If the ThinkFinger package is intended to run under Windows as well as Linux then you should test the change in that environment as well.

Good point.  However I'm not aware of anyone using ThinkFinger on Windows.  I'll post a heads up and call for testing on the ThinkFinger mailing list prior the release.  If anyone is using it on Windows I'll just guard the Set-Config request with some ifdef.

Fixed by commit: 2caf7fcdb8532045680f06b67b9e63f0c9613aaa.  This bug report can be closed now.

Could someone submit this patch to the 2.6.28.y stable queue?

This patch also fixes problems using Altera's USB-Blaster JTAG adapter on 2.6.28 kernels.

It already has been submitted.

(In reply to comment #25)
> It already has been submitted.

Fantastic! Here's hoping it makes it to the linux-2.6.28.y git tree.

For me, this is a serious regression. And the fix is not in 2.6.28.2, i think.

I'm using libusb to upload software to my Lego NXT brick running the LeJOS firmware.

I will try the patch.

(In reply to comment #27)
> I'm using libusb to upload software to my Lego NXT brick running the LeJOS
> firmware.
> 
> I will try the patch.

It seems, the patch is already in 2.6.28.2. Yet, uploading to the NXT brick doesn't work. It used to work with all 2.6.27.x versions.

i had probably the same problem, fingerprint reader was not working (thinkpad T61p/2.6.28.1) and in addition problems reading from usb garmin device as described here: http://bugs.debian.org/cgi-bin/bugreport.cgi?bug=512851.

both issues working for me now with 2.6.28.2.

Created attachment 19999 [details]
Track toggle state in disable-endpoint paths

It turns out that a second patch (attached) is needed in addition to the first one.  I think this will apply correctly to 2.6.28.2 or above (haven't tried it so there might be one or two small mismatches); it should get merged as of 2.6.28.4 or thereabouts.

The patches have been accepted for 2.6.28.stable, commits 2caf7fcdb8532045680f06b67b9e63f0c9613aaa and
ddeac4e75f2527a340f9dc655bde49bb2429b39b.  This bug report can be closed.

thanks for reminding me, closing out.