From: Gerd Hoffmann <kraxel@redhat.com>
To: Hans de Goede <hdegoede@redhat.com>
Cc: Linqiangmin <linqiangmin@huawei.com>,
"Gonglei (Arei)" <arei.gonglei@huawei.com>,
Luonengjun <luonengjun@huawei.com>,
"qemu-devel@nongnu.org" <qemu-devel@nongnu.org>,
"Huangweidong (Hardware)" <huangweidong@huawei.com>
Subject: Re: [Qemu-devel] usb: a problem of using libusb for usb pass through
Date: Wed, 31 Jul 2013 11:47:16 +0200 [thread overview]
Message-ID: <51F8DD24.5090902@redhat.com> (raw)
In-Reply-To: <51F7F422.2030607@redhat.com>
On 07/30/13 19:13, Hans de Goede wrote:
> Hi,
>
> On 07/22/2013 02:58 PM, Gerd Hoffmann wrote:
>> On 07/16/13 10:45, Gonglei (Arei) wrote:
>>> Hi, Gerd
>>> My Qemu version is 1.5.1, and use libusb for usb pass through.
>>> I pass through a host usb device to the guest by bus number and
>>> physical port, when I unplug the usb device from
>>> the host, and plug in on the same physical port immediately, but the
>>> usb device don't show up in the guest.
>>> The qemu log show:
>>> libusb_release_interface: -4 [NO_DEVICE]
>>> libusbx: error [_get_usbfs_fd] libusbx couldn't open USB device
>>> /dev/bus/usb/002/004: No such file or directory
>>> libusbx: error [_get_usbfs_fd] libusbx couldn't open USB device
>>> /dev/bus/usb/002/004: No such file or directory
>>> libusbx: error [_get_usbfs_fd] libusbx couldn't open USB device
>>> /dev/bus/usb/002/004: No such file or directory
>>>
>>> I find the reason is that the global libusb_context has recorded
>>> the usb devices' devnum, but when the host device was unplugged and
>>> plugged
>>> from the host, the devnum will change, finally cause the problem.
>>> Any ideas ? Thanks!
>>
>> Sounds like libusbx doesn't flush the cache on unplug even though it
>> should. Hans?
>
> This looks like qemu's host-libusb code is re-using the libusb_device
> handle, what it should do after seeing an
> unplug (ie ENODEV error), it should close both the handle, and free any
> devices it may have, then redo
> the libusb_get_device_list() and you should get a libusb_device with the
> new address. This is what the
> usb code in the spice-client does, and it has no problems with sc
host-libusb.c doesn't do that, and I also can't reproduce the issue.
Running libusbx-1.0.14 at the moment.
I get slightly different errors:
libusb_release_interface: -4 [NO_DEVICE]
libusbx: error [_open_sysfs_attr] open
/sys/bus/usb/devices/10-1/bConfigurationValue failed ret=-1 errno=2
Which comes from host-libusb trying to release the (unplugged) device on
closing. They are harmless.
>> [ qemu uses one global libusb_context. libusb_device and
>> libusb_device_handle are allocated and released on open/close ]
>
> That is fine, do you also free the device_list, and re-do the
> libusb_get_device_list() ?
Sure. Each time it looks for new devices it calls libusb_get_device_list().
cheers,
Gerd
next prev parent reply other threads:[~2013-07-31 12:07 UTC|newest]
Thread overview: 7+ messages / expand[flat|nested] mbox.gz Atom feed top
2013-07-16 8:45 [Qemu-devel] usb: a problem of using libusb for usb pass through Gonglei (Arei)
2013-07-22 12:58 ` Gerd Hoffmann
2013-07-25 13:21 ` Gonglei (Arei)
2013-07-30 17:13 ` Hans de Goede
2013-07-30 17:15 ` Hans de Goede
2013-07-31 9:47 ` Gerd Hoffmann [this message]
2013-08-01 1:24 ` Gonglei (Arei)
Reply instructions:
You may reply publicly to this message via plain-text email
using any one of the following methods:
* Save the following mbox file, import it into your mail client,
and reply-to-all from there: mbox
Avoid top-posting and favor interleaved quoting:
https://en.wikipedia.org/wiki/Posting_style#Interleaved_style
* Reply using the --to, --cc, and --in-reply-to
switches of git-send-email(1):
git send-email \
--in-reply-to=51F8DD24.5090902@redhat.com \
--to=kraxel@redhat.com \
--cc=arei.gonglei@huawei.com \
--cc=hdegoede@redhat.com \
--cc=huangweidong@huawei.com \
--cc=linqiangmin@huawei.com \
--cc=luonengjun@huawei.com \
--cc=qemu-devel@nongnu.org \
/path/to/YOUR_REPLY
https://kernel.org/pub/software/scm/git/docs/git-send-email.html
* If your mail client supports setting the In-Reply-To header
via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line
before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for NNTP newsgroup(s).