Bug 140125

I also have the same problem after upgrading to kernel
2.6.9-1.678_FC1smp kernel & udev-039-10.FC3.2 this morning. Palm
Tungsten W had been synching perfectly before today on FC3 by doing
the following.
1. Created a file in /etc/udev/rules.d called 10-local.rules
KERNEL="ttyUSB0", SYSFS{product}="Palm Handheld", NAME="%k",
SYMLINK="pilot"
2. Created a symlink from /dev/ttyUSB0 to /etc/udev/devices/pilot
ln -s /dev/ttyUSB0 /etc/udev/devices/pilot

However after the upgrade when activating pilot-xfer, the same
segmentation fault/error happens as reported above. Here is my desmg
output...

visor 3-2:1.0: Handspring Visor / Palm OS converter detected
usb 3-2: Handspring Visor / Palm OS converter now attached to ttyUSB0
usb 3-2: Handspring Visor / Palm OS converter now attached to ttyUSB1
usbcore: registered new driver visor
drivers/usb/serial/visor.c: USB HandSpring Visor / Palm OS driver v2.1
Unable to handle kernel NULL pointer dereference at virtual address
00000004
 printing eip:
022bafbf
*pde = 00004001
Oops: 0000 [#1]
SMP
Modules linked in: visor usbserial udf parport_pc lp parport autofs4
i2c_dev i2c_core sunrpc ipt_REJECT ipt_state ip_conntrack
iptable_filter ip_tables vfat fat button battery ac md5 ipv6 uhci_hcd
ehci_hcd emu10k1_gp snd_intel8x0 gameport snd_mpu401_uart snd_emu10k1
snd_rawmidi snd_pcm_oss snd_mixer_oss snd_pcm snd_timer snd_seq_device
snd_ac97_codec snd_page_alloc snd_util_mem snd_hwdep snd soundcore
ne2k_pci 8390 tulip floppy dm_snapshot dm_zero dm_mirror ext3 jbd dm_mod
CPU:    1
EIP:    0060:[<022bafbf>]    Not tainted VLI
EFLAGS: 00010002   (2.6.9-1.678_FC3smp)
EIP is at _spin_lock_irqsave+0x7/0x45
eax: 00000000   ebx: 00000202   ecx: 22b59b60   edx: 05672280
esi: 00000000   edi: 0378c700   ebp: 036e2e00   esp: 08684ec8
ds: 007b   es: 007b   ss: 0068
Process jpilot (pid: 3767, threadinfo=08684000 task=05166b10)
Stack: 00000000 0378c700 22b56070 10384000 036e2e00 0378c700 00000000
22b8f2e5
       05672280 ffffffed 05672280 10384000 00000000 021ed74c 00000802
0bc00000
       10384000 00000000 036eb804 08684000 00000000 05db3090 0215b549
05672280
Call Trace:
 [<22b56070>] visor_open+0x70/0x17a [visor]
 [<22b8f2e5>] serial_open+0xa5/0xf9 [usbserial]
 [<021ed74c>] tty_open+0x135/0x24c
 [<0215b549>] chrdev_open+0x14c/0x187
 [<0215355d>] dentry_open+0xcd/0x1a5
 [<0215348a>] filp_open+0x36/0x3c
 [<02153789>] sys_open+0x31/0x7d
Code: <3>Debug: sleeping function called from invalid context at
include/linux/rwsem.h:43
in_atomic():0[expected: 0], irqs_disabled():1
 [<0211dd32>] __might_sleep+0x7d/0x87
 [<0214f041>] rw_vm+0xe5/0x28c
 [<022baf94>] _spin_unlock+0xf/0x27
 [<022baf94>] _spin_unlock+0xf/0x27
 [<0214f4a5>] get_user_size+0x30/0x57
 [<022baf94>] _spin_unlock+0xf/0x27
 [<021061c4>] show_registers+0x115/0x16c
 [<0210635b>] die+0xdb/0x16b
 [<02120424>] vprintk+0x136/0x14a
 [<02119aa7>] do_page_fault+0x421/0x5e7
 [<022bafbf>] _spin_lock_irqsave+0x7/0x45
 [<0211cd57>] __wake_up+0x29/0x3c
 [<0213cf18>] __rmqueue+0xc1/0x10c
 [<0213cfbe>] rmqueue_bulk+0x5b/0x65
 [<0213d34f>] buffered_rmqueue+0x17d/0x1a5
 [<02119686>] do_page_fault+0x0/0x5e7
 [<021e007b>] acpi_ec_read+0x1e/0xdc
 [<022bafbf>] _spin_lock_irqsave+0x7/0x45
 [<22b56070>] visor_open+0x70/0x17a [visor]
 [<22b8f2e5>] serial_open+0xa5/0xf9 [usbserial]
 [<021ed74c>] tty_open+0x135/0x24c
 [<0215b549>] chrdev_open+0x14c/0x187
 [<0215355d>] dentry_open+0xcd/0x1a5
 [<0215348a>] filp_open+0x36/0x3c
 [<02153789>] sys_open+0x31/0x7d
 Bad EIP value.
 <6>usb 3-2: USB disconnect, address 2
visor 3-2:1.0: device disconnected

it lools like a bug in kernel. i assign it to correct component.

*** Bug 139594 has been marked as a duplicate of this bug. ***

Updated to kernel-2.6.9-1.681_FC3. Palm now will connect in a hotsync,
but gpilotd now dies unexpectedly. ttyUSB0 and ttyUSB1 also do not
release properly when the hotsync dies.
 The only way to release USB0 and USB1 seems to be to logoff and then
log back on.

Updated output from dmesg (with comments o what I did) is below:

Started a hotsync:
usb 4-2: new full speed USB device using address 4
visor 4-2:1.0: Handspring Visor / Palm OS converter detected
usb 4-2: Handspring Visor / Palm OS converter now attached to ttyUSB0
usb 4-2: Handspring Visor / Palm OS converter now attached to ttyUSB1
usb 4-2: USB disconnect, address 4

[Palm disconnectes by itself here as if hotsync completed, but in
fact, gpilotd and associated pplets still think hotsync is going]

visor 4-2:1.0: device disconnected

[Attempted second hotsync - Note tht palm now connects to ttyUSB2 and
ttyUSB3 because ttyUSB0 and ttyUSB1 are still tied up...]

usb 4-2: new full speed USB device using address 5
visor 4-2:1.0: Handspring Visor / Palm OS converter detected
usb 4-2: Handspring Visor / Palm OS converter now attached to ttyUSB2
usb 4-2: Handspring Visor / Palm OS converter now attached to ttyUSB3

[Manually cancel hotsync from the palm itself...]

usb 4-2: USB disconnect, address 5
visor ttyUSB2: Handspring Visor / Palm OS converter now disconnected
from ttyUSB2
visor ttyUSB3: Handspring Visor / Palm OS converter now disconnected
from ttyUSB3
visor 4-2:1.0: device disconnected

[Attempt to logoff. Eventually must Ctrl-Alt-Backspace to kill session]
visor ttyUSB0: Handspring Visor / Palm OS converter now disconnected
from ttyUSB0
visor ttyUSB1: Handspring Visor / Palm OS converter now disconnected
from ttyUSB1
usb 1-1: control timeout on ep0in

[Then I log back on to collect output from dmesg]

Created attachment 107191 [details]
Debug log of attempted sync with Treo 90

Happens to me too, on 2.6.9-1.678_FC3smp with a Treo 90.

Attached is a log with the debug flags turned on in the usbserial and visor
modules.

By the way, you don't have to log off when this happens. Do:
$ fuser -k /dev/pilot
and it will kill programs that have it open (if you have permission :)

I'm having a similar issue.  Hitting the hotsync button on my T3
prompts the kernel to create the devices, but then, if I do something
like "pilot-xfer -l", the pilot-xfer lists the first 17 files, then
hangs.  The T3 acts as if the sync was interrupted.  The kernel acts
as if the USB device was disconnected.  If I try syncing within
jpilot, it always hangs on the ToDo list.  This happens both on my
laptop and workstation.

I'm using FC3 and the kernel-smp-2.6.9-1.681_FC3 and udev-039-10.FC3.2
packages.  I keep everything updated at least daily.

This did work before I went away on vacation on 11/12.

I'm having much the same problem, but I've managed to work around it
by setting up netsync over wifi with PPTP and OpenVPN tunneling.  If
you have a Wifi or Bluetooth unit, you may be able to do something
similar.  Devil Linux is a really easy way to set up VPN tunnels.

Still, it'd be nice to have USB sync functionality back, so I can
charge as I sync, rather than drain my batteries really fast as I
sync.  :)

I'm on a Tungsten C with update 3 (english) on it.

(can the bug owner change the desc to something like: "PalmOS USB
Hotsync interrupted by kernel"?)

Yes - it is the same issue. Sequence is:
1 - palmos device is connected and hotsync pressed
2 - usb device and ttyUSB0/1 appear
3 - program connects, traffic happens for a bit
4 - <something goes wrong>
5 - usb device disappears, error appears on palm
6 - user apps hung waiting on orphan ttyUSB devices
7 - user apps are killed, close handles, ttyUSB devices cleaned up.

Check /var/log/messages - you'll see entries similar to:
kernel: usb 2-1: Handspring Visor / Palm OS converter now attached to
ttyUSB0
kernel: usb 2-1: Handspring Visor / Palm OS converter now attached to
ttyUSB1
kernel: usbcore: registered new driver visor
kernel: drivers/usb/serial/visor.c: USB HandSpring Visor / Palm OS
driver v2.1
kernel: usb 2-1: USB disconnect, address 2
kernel: visor 2-1:1.0: device disconnected
kernel: visor ttyUSB0: Handspring Visor / Palm OS converter now
disconnected from ttyUSB0
kernel: visor ttyUSB1: Handspring Visor / Palm OS converter now
disconnected from ttyUSB1

May or may not be the same problem, but this happens very reliably on
my newly-upgraded to FC3 system. 

1. push hotsync button on T3 attached via USB (with correct udev setup)
2. dlpsh at command line
3. ls at dlpsh shell
4. listing of installed files begins,
5. listing stops, and after 30 seconds or so, the Palm disconnects

I am running nvidia kernel module, so if this even remotely points at
issues there, I will gladly retry with untainted kernel.

> (can the bug owner change the desc to something like: "PalmOS USB
> Hotsync interrupted by kernel"?)

I opened the bug - no problem to change it with me if that's where the
issue lies.

Of note, I'm also running the NVidia "tainted" kernel. I rolled back
to a "pure" kernel for a bit and still had the issue, only with slower
video...

Can report the exact same new issues for me when upgrading to
kernel-2.6.9-1.681_FC3smp from 2.6.9-1.678_FC1smp reported earlier by
me. With kernel-2.6.9-1.681_FC3smp the hotsynch at least starts now
with pilot-xfer, j-pilot etc, but dies soon after as the kernel
disconnects the USB device for whatever reason. There isn't enough
time for the hotsync to finish.

Running dmesg I get...

usb 3-2: new full speed USB device using address 4
visor 3-2:1.0: Handspring Visor / Palm OS converter detected
usb 3-2: Handspring Visor / Palm OS converter now attached to ttyUSB0
usb 3-2: Handspring Visor / Palm OS converter now attached to ttyUSB1
usb 3-2: USB disconnect, address 4
visor 3-2:1.0: device disconnected
visor ttyUSB0: Handspring Visor / Palm OS converter now disconnected from
ttyUSB0
visor ttyUSB1: Handspring Visor / Palm OS converter now disconnected from
ttyUSB1

Hope its fixed soon. Am having hotsync withdrawal symptoms! :-)

cheers
Daryn

i have fc3 kernel 2.6.9 and i'm trying to sync my palm pilot.  i've
created the link:

 ln -s /dev/ttyUSB1 /dev/pilot

whenever i hit the sync on the palm pilot i can see the devices:

/dev/ttyUSB0
/dev/ttyUSB1

and i'm trying to setup gnome pilot. i get to the "Initial Sync" where
asks to hit the hotsync, and nothing..
i also tried this:

# pilot-xfer -p /dev/pilot -L
   Listening to port: /dev/pilot

   Please press the HotSync button now.... Connected

Reading list of databases in RAM and ROM...
AddressDB
DatebookDB
ExpenseDB
MailDB
MemoDB
ConnectionMgrDB
NetworkDB
npadDB
PhoneRegistryDB
ToDoDB
Address Book
Calculator
clkp
Card Info
Clipper
Date Book
Dial

and  waits..

This bug is also reproducible on FC2 using the kernel 2.6.9-6 (FC2).
I'm using a Zire 31 and I am syncing it with MultiSync. Bug is also
reproducible in pilot-link command line tools. Trying 'dlpsh' and
doing an 'ls' command, the listing of palm dbs would halt after
listing a few of them.

I can reliably (100%) reproduce this bug on an i686 and a x86_64
system with a Palm Tungsten T3...

I'm experiencing exactly the same on an Athlon-XP 2100 running
kernel-2.6.9-1.678_FC3 attempting to sync with a Sony Clie PEG-SJ22/E.

I reset a Zire71 and try syncing, disabling all conduits and deleting
all files from PC. Syncs. Enabled time conduit. Hangs up. Disabled
time and copy files to PC. Started copying first file and hangs up.
Deleted that file from PC. Reset and try again. The same with all
files. Working on 2.6.9-1.681_FC3smp.

I have the same issues as described by Pete Toscano and Michael Brown,
with a Palm T3. It was all working fine on my system under FC2, and
yesterday I upgraded to FC3 with all the latest updates, and now I
cannot sync :(

With my Samsung i500 I've been able to make it work in kernel version
2.6.9-1.667.  This was a downgrade for me.  Syncing failed in
2.6.9-1.681_FC3 and 2.6.9-1.678_FC3.

The behaviour that I am seeing is the same as in comment #11

Created attachment 107942 [details]
syslog of an interrupted Palm T3 sync

this is kernel 2.6.9-1.681_FC3 is with USB_DEBUG enabled

I can reproduce the interrupted sync on 3 different FC3 machines (all
tested with 2.6.9-1.681_FC3).
On one machine I recompiled the kernel with CONFIG_USB_DEBUG=y, this
machine shows loads of
kernel: drivers/usb/input/hid-core.c: input irq status -84 received
This machine also uses a USB mouse (hence my logs are full of these
errors), mouse works though.

On another machine, I replaced FC3 with FC2 (release kernel), the
problem disappeared.

So it would seem we really broke something in USB between FC2 and FC3.

The Attachment #107942 [details] shows an interrupted sync on the machine that
has USB_Debug enabled

I have had similar problems on 1.681_FC3 (comment #4 is very familiar).

The problem can be fixed by reversing the patchs to
usb/drivers/serial/visor.c in the Fedora RPM and rebuilding. These
changes are patch of the compressed -ac10 patch so cannot be directly
grepped for.

I simply deleted the visor changes from the - ac10 patch, recompressed
it then rebuilt and installed the RPM. This brought pilot-xfer and
gpilotd back to life.

can you try the -698 kernel in updates-testing and see how that fares?

Comment in bug #140005 states that kernel fixes the problem. Will try
it here.

Confirmed, booting with kernel 2.6.9-1.698_FC3, and setting
selinux=disable (because portmap, ntpd, syslog, others would not start
otherwise, and my home is NFS), I have successfully synced with the
gnome conduits, and done a full dlpsh ls listing with no problems,
which both failed previously.

Thanks for the tip, and reenforcing the idea on that other bug.

Confirmed as well, test machine lists pilot contents just fine with
"pilot-xfer -p /dev/ttyUSB1 -l" under 2.6.9-1.698_FC3.

I'll put this kernel on my work machine later and will post results here.

Works on the main machine as well. I still see "kernel:
drivers/usb/input/hid-core.c: input irq status -84 received" but a
sync and a backup of the Palm T3 with jpilot went through fine anyway.

I can confirm as well, that 2.6.901.698_FC3 does fix the USB problem,
but it also breaks ndiswrapper somehow. 

Who should I report that break in the development kernel to?

*** Bug 141956 has been marked as a duplicate of this bug. ***

Just updated FC3 to kernel 2.6.9-1.724. Problem appears to be fixed
with this version.

OK.. this is still a problem with FC4 and 2.6.12-1.1398_FC4.  Is it ever getting
fixed?

This appears to be a recurrence of the 2.6.9 bug with 2.6.12.  It worked right
for 2.6.10 and 11.

Please reopen this bug.

FWIW (and not reopening bug on purpose)

I've had much better results on FC6 in the 3 days I've had it installed now. In
other words 2.6.18-1.2798.fc6 works for me.

With older versions I had to actually reset the Palm just before sync to have
any chance of being able to retrieve my DatebookDB (which at 185K is admittedly
rather big), guess this has to do with the on-the-fly conversion that takes
place in PalmOS 5. By the looks of it, 2.6.18-1.2798.fc6 is simply more patient
on the timeout side.

Ah well, only worked for a while, this morning I had the first instance of the
FC6 box not syncing either until I did a soft reset :-(