GNOME Bugzilla – Bug 793885
Crashes in a loop
Last modified: 2018-02-28 17:48:21 UTC
This is an Android device[1] which I plugged in via USB. After browsing through the MTP mount and finding nothing interesting, I launch "adb devices" as root, and gvfsd-mtp started crashing in a loop with this backtrace:
+ Trace 238437
gvfs-mtp-1.34.2-1.fc27.x86_64 Race on umount? [1]: https://www.atgames.us/collections/the-sega-genesis-flashback-collection/products/sega-genesis-flasback
journalctl output: Feb 27 17:17:42 classic kernel: usb 3-9: new high-speed USB device number 51 using xhci_hcd Feb 27 17:17:42 classic kernel: usb 3-9: New USB device found, idVendor=2207, idProduct=0011 Feb 27 17:17:42 classic kernel: usb 3-9: New USB device strings: Mfr=2, Product=3, SerialNumber=4 Feb 27 17:17:42 classic kernel: usb 3-9: Product: rk3036 Feb 27 17:17:42 classic kernel: usb 3-9: Manufacturer: rockchip Feb 27 17:17:42 classic kernel: usb 3-9: SerialNumber: V5NQX36ZJH Feb 27 17:17:42 classic gvfs-mtp-volume[1946]: device (null) has no BUSNUM property, ignoring Feb 27 17:17:42 classic gvfs-mtp-volume[1946]: device (null) has no BUSNUM property, ignoring Feb 27 17:17:42 classic gvfsd[1745]: Device 0 (VID=2207 and PID=0011) is a Various Viewpia DR/bq Kepler Debugging. Feb 27 17:17:42 classic kernel: usb 3-9: reset high-speed USB device number 51 using xhci_hcd Feb 27 17:17:42 classic kernel: usb 3-9: USB disconnect, device number 51 Feb 27 17:17:48 classic gvfsd[1745]: PTP_ERROR_IO: failed to open session, trying again after resetting USB interface Feb 27 17:17:48 classic gvfsd[1745]: LIBMTP libusb: Attempt to reset device Feb 27 17:17:48 classic gvfsd[1745]: inep: usb_get_endpoint_status(): No such device Feb 27 17:17:48 classic gvfsd[1745]: outep: usb_get_endpoint_status(): No such device Feb 27 17:17:48 classic gvfsd[1745]: libusb_open() failed!: No such device Feb 27 17:17:48 classic gvfsd[1745]: LIBMTP PANIC: Could not init USB on second attempt Feb 27 17:17:48 classic gnome-shell[26580]: invalid cast from 'GTask' to 'GSimpleAsyncResult' Feb 27 17:17:48 classic gnome-shell[26580]: g_simple_async_result_propagate_error: assertion 'G_IS_SIMPLE_ASYNC_RESULT (simple)' failed Feb 27 17:17:48 classic gvfsd-mtp[30933]: invalid unclassed pointer in cast to 'GVfsBackendMtp' Feb 27 17:17:48 classic audit[30933]: ANOM_ABEND auid=1000 uid=1000 gid=1000 ses=3 subj=unconfined_u:unconfined_r:unconfined_t:s0-s0:c0.c1023 pid=30933 comm="gvfsd-mtp" exe="/usr/libexec/gvfsd-mtp" sig=11 res=1 Feb 27 17:17:48 classic kernel: gvfsd-mtp[30933]: segfault at 18 ip 00007f9019eaba99 sp 00007ffc673c5dc0 error 4 in libgvfsdaemon.so[7f9019e9e000+25000] Feb 27 17:17:48 classic gvfsd-mtp[30933]: g_object_ref: assertion 'G_IS_OBJECT (object)' failed Feb 27 17:17:48 classic gvfsd-mtp[30933]: g_object_ref: assertion 'G_IS_OBJECT (object)' failed
(The reproducer is a broken version of android-tools and an Android device, see https://bugzilla.redhat.com/show_bug.cgi?id=1470740)
This is just another duplicate of Bug 787992, but maybe finally brings some real reproducer... Can you please provide gvfsd debug log for the segfaults (maybe also with LIBMTP_DEBUG=13)? I will try to reproduce it with the mentioned version of android-tools.
Hmm, "adb devices" works as expected on rawhide and I don't see any repeating resets with one of those broken versions... :-(
(In reply to Ondrej Holy from comment #4) > Hmm, "adb devices" works as expected on rawhide and I don't see any > repeating resets with one of those broken versions... :-( I fixed it in rawhide, you'll need to use one of the broken versions (as mentioned in the bug comments). You'll also need to make sure you killed the old adb daemon that's running from your working version.
Tested gvfs-1.35.91 on my F27 machine (locally built RPM), and even if mtp still throws more errors than I'd like, it doesn't crash anymore. Closing as a dupe. *** This bug has been marked as a duplicate of bug 787992 ***
I've used one from the versions mentioned on that bug, but never mind... but never mind...