Palm Hotsync and usbserial: New Findings

[James Liggett]

Hello all,
I have been working on getting Hotsync to sync to a USB cradle using
usbserial. I did some research, and as it turns out, the functions that
wine uses to poll normal serial ports (like ioctl and tcgetattr) don't
like usbserial devices very much. In fact, when I query my USB cradle
using setserial, it gives me the same "Invalid Arguement" error that
GetCommState does. Additionally, pilot-link doesn't even use these
functions to communicate with a usbserial connected palm. The question
now is, can we get tcgetattr to play nice with usbserial?

Best Regards,
James Liggett

[Uwe Bonnes]

> > > > > "James" == James Liggett <jrliggett(a)cox.net> writes:

James> Hello all, I have been working on getting Hotsync to sync to a
    James> USB cradle using usbserial. I did some research, and as it turns
    James> out, the functions that wine uses to poll normal serial ports
    James> (like ioctl and tcgetattr) don't like usbserial devices very
    James> much. In fact, when I query my USB cradle using setserial, it
    James> gives me the same "Invalid Arguement" error that GetCommState
    James> does. Additionally, pilot-link doesn't even use these functions
    James> to communicate with a usbserial connected palm. The question now
    James> is, can we get tcgetattr to play nice with usbserial?

What USB to serial adapter is used? The usbserial Atmel AVRIce uses a FTDI
FT232 which is well supported. It works quite well, with the serial
related patches I sent to the list( but some of them not yet in CVS).  The
problem you suspect is likely more usb related.

Bye

-- 
Uwe Bonnes                bon(a)elektron.ikp.physik.tu-darmstadt.de

Institut fuer Kernphysik  Schlossgartenstrasse 9  64289 Darmstadt
--------- Tel. [phone] -------- Fax. [phone] ----------

[James Liggett]

I'm not using a "real" adapter. I'm using the visor driver that is used
to sync a USB palm with things like pilot-link. This driver emulates a
serial port in much the same way that a real one does, or so I think. I
was just pointing out that people have similar behavior with real
adapters, so I thought it might be relevant. The visor driver calls
itself a "serial converter", and it does add USB entries in my /dev/tts
folder like a real adapter would. But, my palm cradle is plugged into an
acutal USB port. I do have access to a real USB to serial adapter
though; I'll play around with it. And another thing--I'm also looking
into that WinMM crash that keeps Palm Desktop from running. The problem
is that the function DRIVER_FindFromHDrvr is failing, returning a NULL
value, which wine tries to derefrence in MMDRV_Install, and boom--wine
crashes. 

James

On Thu, 2005-07-07 at 10:51 +0200, Uwe Bonnes wrote:
> > > > > > "James" == James Liggett <jrliggett(a)cox.net> writes:
>
> James> Hello all, I have been working on getting Hotsync to sync to a
>     James> USB cradle using usbserial. I did some research, and as it turns
>     James> out, the functions that wine uses to poll normal serial ports
>     James> (like ioctl and tcgetattr) don't like usbserial devices very
>     James> much. In fact, when I query my USB cradle using setserial, it
>     James> gives me the same "Invalid Arguement" error that GetCommState
>     James> does. Additionally, pilot-link doesn't even use these functions
>     James> to communicate with a usbserial connected palm. The question now
>     James> is, can we get tcgetattr to play nice with usbserial?
>
> What USB to serial adapter is used? The usbserial Atmel AVRIce uses a FTDI
> FT232 which is well supported. It works quite well, with the serial
> related patches I sent to the list( but some of them not yet in CVS).  The
> problem you suspect is likely more usb related.
>
> Bye

[Uwe Bonnes]

> > > > > "James" == James Liggett <jrliggett(a)cox.net> writes:

James> I'm not using a "real" adapter. I'm using the visor driver that
    James> is used to sync a USB palm with things like pilot-link. This

Does the hotplug mechanisme create some /dev/ttyUSBx when you plug in the
device? Or is the USB VID/PID at least know to the linix kernel drivers?

    James> driver emulates a serial port in much the same way that a real
    James> one does, or so I think. I was just pointing out that people have
    James> similar behavior with real adapters, so I thought it might be
    James> relevant. The visor driver calls itself a "serial converter", and
    James> it does add USB entries in my /dev/tts folder like a real adapter
    James> would. But, my palm cradle is plugged into an acutal USB port. I
    James> do have access to a real USB to serial adapter though; I'll play
    James> around with it. And another thing--I'm also looking into that
    James> WinMM crash that keeps Palm Desktop from running. The problem is
    James> that the function DRIVER_FindFromHDrvr is failing, returning a
    James> NULL value, which wine tries to derefrence in MMDRV_Install, and
    James> boom--wine crashes.

Scan the mailing lists. I had another similar problem, Eric pointed me to
some problems in my setup.

-- 
Uwe Bonnes                bon(a)elektron.ikp.physik.tu-darmstadt.de

Institut fuer Kernphysik  Schlossgartenstrasse 9  64289 Darmstadt
--------- Tel. [phone] -------- Fax. [phone] ----------

[James Liggett]

Yes, the hotplug (well, actually it's coldplug and udev) creates two USB
devices, /dev/tts/USB0 and /dev/tts/USB1. With palm devices, these
entries only show up when the palm is actually trying to hotsync, but
palm desktop didn't really seem to care. It tried to use GetCommState
when the devices showed up. If I run hotsync program without trying to
sync, it doesn't complain. The errors only show after I pres the button.
I have USB0 symlinked to /dev/pilot, and I am able to sync with this
device using pilot-link compatible apps (mainly gnome-pilot.) Oh, and
about that crash--does the mailing list archive have a search engine on
it?

Thanks a lot,
James

On Thu, 2005-07-07 at 22:28 +0200, Uwe Bonnes wrote:
> > > > > > "James" == James Liggett <jrliggett(a)cox.net> writes:
>
> James> I'm not using a "real" adapter. I'm using the visor driver that
>     James> is used to sync a USB palm with things like pilot-link. This
>
> Does the hotplug mechanisme create some /dev/ttyUSBx when you plug in the
> device? Or is the USB VID/PID at least know to the linix kernel drivers?
>
> James> driver emulates a serial port in much the same way that a real
>     James> one does, or so I think. I was just pointing out that people have
>     James> similar behavior with real adapters, so I thought it might be
>     James> relevant. The visor driver calls itself a "serial converter", and
>     James> it does add USB entries in my /dev/tts folder like a real adapter
>     James> would. But, my palm cradle is plugged into an acutal USB port. I
>     James> do have access to a real USB to serial adapter though; I'll play
>     James> around with it. And another thing--I'm also looking into that
>     James> WinMM crash that keeps Palm Desktop from running. The problem is
>     James> that the function DRIVER_FindFromHDrvr is failing, returning a
>     James> NULL value, which wine tries to derefrence in MMDRV_Install, and
>     James> boom--wine crashes.
>
> Scan the mailing lists. I had another similar problem, Eric pointed me to
> some problems in my setup.

[Uwe Bonnes]

> > > > > "James" == James Liggett <jrliggett(a)cox.net> writes:

James> device using pilot-link compatible apps (mainly gnome-pilot.) Oh,
    James> and about that crash--does the mailing list archive have a search
    James> engine on it?

Use google. Try "bonnes pouech MMDRV_Install"
-- 
Uwe Bonnes                bon(a)elektron.ikp.physik.tu-darmstadt.de

Institut fuer Kernphysik  Schlossgartenstrasse 9  64289 Darmstadt
--------- Tel. [phone] -------- Fax. [phone] ----------

[James Liggett]

Tried your lead, but no luck. I did find the conversation, but to my
dismay it was a dead end. It ended with you saying that the problem had
gone away out of the blue:
http://www.winehq.com/hypermail/wine-devel/2003/05/0075.html

That's very odd. Btw, I saw that you sent a patch in to check the sanity
of driver pointers returned by DRIVER_FindFromHDrvr. Why wasn't this
committed? And, you were right about that Microquill memory manager bug,
but I realized this after the fact. For a while I thought we weren't on
the same page, but I saw a lot of the stuff about it in the archives
while searching. So that's the cause of this crap? Anyway, thanks for
you help. :)

Regards,
James

On Fri, 2005-07-08 at 00:10 +0200, Uwe Bonnes wrote:
> > > > > > "James" == James Liggett <jrliggett(a)cox.net> writes:
>
> James> device using pilot-link compatible apps (mainly gnome-pilot.) Oh,
>     James> and about that crash--does the mailing list archive have a search
>     James> engine on it?
>
> Use google. Try "bonnes pouech MMDRV_Install"

[Uwe Bonnes]

> > > > > "James" == James Liggett <jrliggett(a)cox.net> writes:

James> Tried your lead, but no luck. I did find the conversation, but to
    James> my dismay it was a dead end. It ended with you saying that the
    James> problem had gone away out of the blue:
    James> http://www.winehq.com/hypermail/wine-devel/2003/05/0075.html

    James> That's very odd. Btw, I saw that you sent a patch in to check the
    James> sanity of driver pointers returned by DRIVER_FindFromHDrvr. Why
    James> wasn't this committed? And, you were right about that Microquill
    James> memory manager bug, but I realized this after the fact. For a
    James> while I thought we weren't on the same page, but I saw a lot of
    James> the stuff about it in the archives while searching. So that's the
    James> cause of this crap? Anyway, thanks for you help. :)

There are two sided of the problem:
- One is microquill not handling HeapValidate
- Second is the loading of the sound driver using this not otherwise used
function 

Reg one: This is a plain bug in shsmp, and I just wrote a letter to
Microquill.
Reg two: Apply appended workaraound. Perhaps Eric can think about a more
valid workaround for this Microquill problem.

Bye
-- 
Uwe Bonnes                bon(a)elektron.ikp.physik.tu-darmstadt.de

Institut fuer Kernphysik  Schlossgartenstrasse 9  64289 Darmstadt
--------- Tel. [phone] -------- Fax. [phone] ----------
Index: wine/dlls/winmm/driver.c
===================================================================
RCS file: /home/wine/wine/dlls/winmm/driver.c,v
retrieving revision 1.32
diff -u -w -r1.32 driver.c
--- wine/dlls/winmm/driver.c	21 Mar 2005 10:28:39 -0000	1.32
+++ wine/dlls/winmm/driver.c	8 Jul 2005 07:47:59 -0000
@@ -75,7 +75,7 @@
 {
     LPWINE_DRIVER	d = (LPWINE_DRIVER)hDrvr;
 
-    if (hDrvr && HeapValidate(GetProcessHeap(), 0, d) && d->dwMagic == WINE_DI_MAGIC) {
+    if (hDrvr /* && HeapValidate(GetProcessHeap(), 0, d) */&& d->dwMagic == WINE_DI_MAGIC) {
 	return d;
     }
     return NULL;

[Robert Shearman]

Uwe Bonnes wrote:
> There are two sided of the problem:
> - One is microquill not handling HeapValidate
> - Second is the loading of the sound driver using this not otherwise used
> function

WinMM should not be using HeapValidate. There are many other ways of 
ensuring the handle that was passed is valid including using exception 
handling with magic values and handle tables.

-- 
Rob Shearman

[James Liggett]

Thanks, Uwe. I really apprecate all of this. I was wondering, why are we
even using this library when there's been bugs in it that we discovered
long ago, but never fixed? And another thing that bothers me: I've
selected ALSA as my sound output, but the debug trace shows wineoss.drv
loading. What's with that?

Jmaes

On Fri, 2005-07-08 at 09:49 +0200, Uwe Bonnes wrote:
> > > > > > "James" == James Liggett <jrliggett(a)cox.net> writes:
>
> James> Tried your lead, but no luck. I did find the conversation, but to
>     James> my dismay it was a dead end. It ended with you saying that the
>     James> problem had gone away out of the blue:
>     James> http://www.winehq.com/hypermail/wine-devel/2003/05/0075.html
>
> James> That's very odd. Btw, I saw that you sent a patch in to check the
>     James> sanity of driver pointers returned by DRIVER_FindFromHDrvr. Why
>     James> wasn't this committed? And, you were right about that Microquill
>     James> memory manager bug, but I realized this after the fact. For a
>     James> while I thought we weren't on the same page, but I saw a lot of
>     James> the stuff about it in the archives while searching. So that's the
>     James> cause of this crap? Anyway, thanks for you help. :)
>
> There are two sided of the problem:
> - One is microquill not handling HeapValidate
> - Second is the loading of the sound driver using this not otherwise used
> function
>
> Reg one: This is a plain bug in shsmp, and I just wrote a letter to
> Microquill.
> Reg two: Apply appended workaraound. Perhaps Eric can think about a more
> valid workaround for this Microquill problem.
>
> Bye

[Uwe Bonnes]

> > > > > "James" == James Liggett <jrliggett(a)cox.net> writes:

James> Thanks, Uwe. I really apprecate all of this. I was wondering, why
    James> are we even using this library when there's been bugs in it that
    James> we discovered long ago, but never fixed? And another thing that
    James> bothers me: I've selected ALSA as my sound output, but the debug
    James> trace shows wineoss.drv loading. What's with that?

It's not we, it is the application (in my case Xilinx ISE) using
Smartheap. But recent mails indicate that the use of HeapValidate in winmm i
dubios...

Bye
-- 
Uwe Bonnes                bon(a)elektron.ikp.physik.tu-darmstadt.de

Institut fuer Kernphysik  Schlossgartenstrasse 9  64289 Darmstadt
--------- Tel. [phone] -------- Fax. [phone] ----------

[James Liggett]

The patch you gave me fixed the crash in winmm, but now I get this:

err:seh:setup_exception stack overflow 496 bytes in thread 0009 eip
7bee164a esp 7b9d0e10 stack 0x7b9d0000-0x7bad0000

What's that?

Thanks,
James

On Fri, 2005-07-08 at 23:17 +0200, Uwe Bonnes wrote:
> > > > > > "James" == James Liggett <jrliggett(a)cox.net> writes:
>
> James> Thanks, Uwe. I really apprecate all of this. I was wondering, why
>     James> are we even using this library when there's been bugs in it that
>     James> we discovered long ago, but never fixed? And another thing that
>     James> bothers me: I've selected ALSA as my sound output, but the debug
>     James> trace shows wineoss.drv loading. What's with that?
>
> It's not we, it is the application (in my case Xilinx ISE) using
> Smartheap. But recent mails indicate that the use of HeapValidate in winmm i
> dubios...
>
> Bye

[Uwe Bonnes]

> > > > > "James" == James Liggett <jrliggett(a)cox.net> writes:

James> The patch you gave me fixed the crash in winmm, but now I get
    James> this: err:seh:setup_exception stack overflow 496 bytes in thread
    James> 0009 eip 7bee164a esp 7b9d0e10 stack 0x7b9d0000-0x7bad0000

    James> What's that?

Another crash :-)

-- 
Uwe Bonnes                bon(a)elektron.ikp.physik.tu-darmstadt.de

Institut fuer Kernphysik  Schlossgartenstrasse 9  64289 Darmstadt
--------- Tel. [phone] -------- Fax. [phone] ----------

[James Liggett]

On Thu, 2005-07-07 at 10:51 +0200, Uwe Bonnes wrote:
> > > > > > "James" == James Liggett <jrliggett(a)cox.net> writes:
>
> James> Hello all, I have been working on getting Hotsync to sync to a
>     James> USB cradle using usbserial. I did some research, and as it turns
>     James> out, the functions that wine uses to poll normal serial ports
>     James> (like ioctl and tcgetattr) don't like usbserial devices very
>     James> much. In fact, when I query my USB cradle using setserial, it
>     James> gives me the same "Invalid Arguement" error that GetCommState
>     James> does. Additionally, pilot-link doesn't even use these functions
>     James> to communicate with a usbserial connected palm. The question now
>     James> is, can we get tcgetattr to play nice with usbserial?
>
> What USB to serial adapter is used? The usbserial Atmel AVRIce uses a FTDI
> FT232 which is well supported. It works quite well, with the serial
> related patches I sent to the list( but some of them not yet in CVS).

Can you send me the patches so I can try them? I tried to use my Dad's
*real* usb to serial converter that he uses to sync his IIIc to his
USB-only laptop, but it doesn't seem to be supported in linux (It uses
some weird chipset, like ATEN or something like that--ever heard of it?)
> The
> problem you suspect is likely more usb related.
>
> Bye

Thanks,
James

[Uwe Bonnes]

> > > > > "James" == James Liggett <jrliggett(a)cox.net> writes:

James> On Thu, 2005-07-07 at 10:51 +0200, Uwe Bonnes wrote:
    >> >>>>> "James" == James Liggett <jrliggett(a)cox.net> writes:
    >> 
    James> Hello all, I have been working on getting Hotsync to sync to a
    James> USB cradle using usbserial. I did some research, and as it turns
    James> out, the functions that wine uses to poll normal serial ports
    James> (like ioctl and tcgetattr) don't like usbserial devices very
    James> much. In fact, when I query my USB cradle using setserial, it
    James> gives me the same "Invalid Arguement" error that GetCommState
    James> does. Additionally, pilot-link doesn't even use these functions
    James> to communicate with a usbserial connected palm. The question now
    James> is, can we get tcgetattr to play nice with usbserial?
    >>  What USB to serial adapter is used? The usbserial Atmel AVRIce uses
    >> a FTDI FT232 which is well supported. It works quite well, with the
    >> serial related patches I sent to the list( but some of them not yet
    >> in CVS).

    James> Can you send me the patches so I can try them? I tried to use my
    James> Dad's *real* usb to serial converter that he uses to sync his
    James> IIIc to his USB-only laptop, but it doesn't seem to be supported
    James> in linux (It uses some weird chipset, like ATEN or something like
    James> that--ever heard of it?)

If the chip is not supported in the linux kernel, no chance to access it
with wine. Does lsusb know about the chip (VID/PID)?
-- 
Uwe Bonnes                bon(a)elektron.ikp.physik.tu-darmstadt.de

Institut fuer Kernphysik  Schlossgartenstrasse 9  64289 Darmstadt
--------- Tel. [phone] -------- Fax. [phone] ----------

[James Liggett]

Yes it does. And I found a driver, so it works now. I guess I spoke a
little too soon...sorry about that.

James

On Tue, 2005-07-12 at 09:33 +0200, Uwe Bonnes wrote:
> > > > > > "James" == James Liggett <jrliggett(a)cox.net> writes:
>
> James> On Thu, 2005-07-07 at 10:51 +0200, Uwe Bonnes wrote:
>     >> >>>>> "James" == James Liggett <jrliggett(a)cox.net> writes:
>     >> 
>     James> Hello all, I have been working on getting Hotsync to sync to a
>     James> USB cradle using usbserial. I did some research, and as it turns
>     James> out, the functions that wine uses to poll normal serial ports
>     James> (like ioctl and tcgetattr) don't like usbserial devices very
>     James> much. In fact, when I query my USB cradle using setserial, it
>     James> gives me the same "Invalid Arguement" error that GetCommState
>     James> does. Additionally, pilot-link doesn't even use these functions
>     James> to communicate with a usbserial connected palm. The question now
>     James> is, can we get tcgetattr to play nice with usbserial?
>     >>  What USB to serial adapter is used? The usbserial Atmel AVRIce uses
>     >> a FTDI FT232 which is well supported. It works quite well, with the
>     >> serial related patches I sent to the list( but some of them not yet
>     >> in CVS).
>
> James> Can you send me the patches so I can try them? I tried to use my
>     James> Dad's *real* usb to serial converter that he uses to sync his
>     James> IIIc to his USB-only laptop, but it doesn't seem to be supported
>     James> in linux (It uses some weird chipset, like ATEN or something like
>     James> that--ever heard of it?)
>
> If the chip is not supported in the linux kernel, no chance to access it
> with wine. Does lsusb know about the chip (VID/PID)?

[Andreas Mohr]

Hi,

On Tue, Jul 12, 2005 at 12:48:36AM -0700, James Liggett wrote:
> Yes it does. And I found a driver, so it works now. I guess I spoke a
> little too soon...sorry about that.
> I knew it. ;) The chances of an unsupported USB-Serial chip in Linux should
> be very slim, since I'd think that such a chip is not the most high-end
> hardware device and should thus be reverse-engineered in a snap for a nice
> Linux driver (maybe using USB interception via VMWare or so).

Andreas Mohr

[Kuba Ober]

> I did some research, and as it turns out, the functions that
> wine uses to poll normal serial ports (like ioctl and tcgetattr) don't
> like usbserial devices very much. In fact, when I query my USB cradle
> using setserial, it gives me the same "Invalid Arguement" error that
> GetCommState does. Additionally, pilot-link doesn't even use these
> functions to communicate with a usbserial connected palm. The question
> now is, can we get tcgetattr to play nice with usbserial?

What kernel is that?

Cheers, Kuba