fails to handle a usb serial port with a specific vendorid

If I run qemu-system-i386 with arguments
-usb -usbdevice serial:vendorid=1221:pty
(this is what the documentation says about how I shoud add a usb device which has a serial port interface and which has a specific vendor id, I used the documentation located here:
http://qemu.weilnetz.de/qemu-doc.html
), it says 
char device redirected to /dev/pts/<something> (label usbserial0)
qemu-system-i386: -usbdevice serial:vendorid=1221:pty: Property '.vendorid' not found
Aborted
and exits. Moreover, if I try to add such a device to a running machine by typing usb_add serial:vendorid=1221:pty in the machine's control terminal (to reach it, I press ctrl-alt-2), qemu also writes 
char device redirected to /dev/pts/<something> (label usbserial0)
Aborted
to the terminal where I run it from and exits. To the quest OS this looks like a power failure which causes all the programs inside the virtual machine to lose their unsaved data.
I have tested this with qemu-1.5.0-rc2, actually, the issue occured in a similar way since 1.0.1, but did not occur in 0.11.1.
The issue is reproducible always, even if I don't specify any hard disk in the command line, i. e.
$ qemu-system-i386 -usb -usbdevice serial:vendorid=1221:pty
, so I believe it is guest OS -independent.

  Hi,

>> (this is what the documentation says about how I shoud add a usb device which has a serial port interface and which has a specific vendor id, I used the documentation located here:
>> http://qemu.weilnetz.de/qemu-doc.html
>> ), it says 
>> char device redirected to /dev/pts/<something> (label usbserial0)
>> qemu-system-i386: -usbdevice serial:vendorid=1221:pty: Property '.vendorid' not found
>> Aborted
> [...]
> 
> Regression; this definitely worked when I wrote docs/qdev-device-use.txt.

> Not a release blocker, since it regressed a long time ago (v0.12).

Guess the docs should be updated, unless someone can come up with a
reasonable use case for the vendorid + deviceid properties.

cheers,
  Gerd


I think the ability to specify a different vendorid + deviceid can be useful. Suppose there is a USB device such that the specifications are open and officially published, but the driver is proprietary. (As far as I know, this is similar to the situation with ATI video cards, but they are not USB devices.) And I suspect that the driver is buggy (i. e. it does not send the data according to the specifications). I want to figure out where exactly it works incorrectly to submit a bug report to the developer of the driver. Or suppose I have a physical device, but it works a bit incorrectly. I want to figure out where exactly the problem is, in the driver or in the device. Since I am not sure that the device is OK, I don't want to write my own driver and interact with the device, maybe I will damage it even more. In both cases, I can emulate the device according to the specifications, install the driver in a guest system, and then see whether the driver sends correct data or where and when exactly the data are incorrect.

Anyway, I think it is more or less ok if qemu crashes right after it starts due to bad command line parameters (nevertherless, the functionality lost this way could be useful as I explained). But I think IT IS NOT OK WHEN A WORKING VM WITH PROGRAMS INSIDE CRASHES after user enters a bad command in the machine's control terminal, unless the user explicitly requests termination (e. g. enters the q command).

Regressed in commit f29783f72ea77dfbd7ea0c993d62d253d4c4e023.

I've just run into this in a similar circumstance: trying to reverse-engineer a driver for a phone to which I can only connect via Bluetooth. No problem, I can just have it pretend to be a USB device. Except that I can't, because the driver won't recognise it.

The crash has now been fixed here:
http://git.qemu.org/?p=qemu.git;a=commitdiff;h=aa612b364ecbe1dc
Please also note that the "-usbdevice serial" syntax is considered as deprecated nowadays - use "-device usb-serial" instead.