Hide Forgot
Description of problem: After a few (1...3) successful runs of "growisofs" to write a DVD/BD image no further accesses are possible bcuase exclusive opens of the device fail: open("/dev/dvdrw", O_RDONLY|O_EXCL|O_NONBLOCK|O_LARGEFILE) = -1 EBUSY (Device or resource busy) However, there are no uses of the drive: # ls -l /dev/dvdrw lrwxrwxrwx 1 root root 3 Jun 28 15:33 /dev/dvdrw -> sr0 # lsof /dev/sr0 lsof: WARNING: can't stat() fuse.gvfs-fuse-daemon file system /home/wd/.gvfs Output information may be incomplete. lsof: WARNING: can't stat() fuse.gvfs-fuse-daemon file system /home/wd/.gvfs Output information may be incomplete. # lsmod shows that "udf" is in use: # lsmod Module Size Used by udf 61737 1 crc_itu_t 1251 1 udf tcp_lp 1867 0 nfs 269067 0 fscache 37475 1 nfs ... It cannot be removed: # rmmod -f udf ERROR: Removing 'udf': Resource temporarily unavailable The only way to recover is a reboot of the system. Version-Release number of selected component (if applicable): kernel-2.6.38.8-32.fc15.i686 How reproducible: Reliably; usually it hangs at the second attempt to write a DVD; so far it never survived a 3rd writing. Steps to Reproduce: 1. Write some files to DVD/BD, for example like this: growisofs -dvd-compat -Z /dev/dvdrw -R -J -udf -allow-limited-size -iso-level=3 Monty_Python.mkv 2. Wait for completion, insert new disk 3. re-run the same command (if it works, go to step 2.) Actual results: growisofs will bail out with an error message: ... :-( unable to O_EXCL /dev/dvdrw: someone was in time to remount? Checking with strace shows that an exclusive open fails: open("/dev/dvdrw", O_RDONLY|O_EXCL|O_NONBLOCK|O_LARGEFILE) = -1 EBUSY (Device or resource busy) As mentioned, in this situation the UDF mdule is locked and cannot be unloaded any more, so I suspect problems in that area. Expected results: It should be possible to write an unlimited number of DVD /BD without need to reboot the system. Additional info:
Sorry, but my description of the steps to trigger the problem is wrong. I understand this only now myself. The problem happens when _mounting_ an UDF file system as greated in the steps described above. Before: # lsmod Module Size Used by tcp_lp 1867 0 sco 13425 2 bnep 12069 2 ... Then run: # mount -o ro /dev/sr0 /mnt/tmp/ The kernel will report: Jun 28 23:21:33 gemini kernel: [22516.221453] UDF-fs: Partition marked readonly; forcing readonly mount Jun 28 23:21:33 gemini kernel: [22516.261759] UDF-fs INFO UDF: Mounting volume 'CDROM', timestamp 2011/06/28 19:07 (1000) and: # lsmod Module Size Used by udf 61737 1 crc_itu_t 1251 1 udf tcp_lp 1867 0 sco 13425 2 bnep 12069 2 ... After an umount I see this: # umount /mnt/tmp/ # lsmod Module Size Used by udf 61737 1 crc_itu_t 1251 1 udf tcp_lp 1867 0 sco 13425 2 bnep 12069 2 ... # rmmod udf ERROR: Module udf is in use ... open("/dev/sr0", O_RDONLY|O_EXCL|O_NONBLOCK|O_LARGEFILE) = -1 EBUSY (Device or resource busy) ... So the Subject: is OK, the actual problem happens when mounting an UDF file system - it blocks the device forever, and the module cannot be unloaded. # uname -a Linux gemini.denx.de 2.6.38.8-32.fc15.i686.PAE #1 SMP Mon Jun 13 19:55:27 UTC 2011 i686 i686 i386 GNU/Linux
Do you get any error messages in the kernel log when you unmount the filesystem?
Not when mounting or unmounting. When mounting, I get this: UDF-fs: Partition marked readonly; forcing readonly mount UDF-fs INFO UDF: Mounting volume 'CDROM', timestamp 2011/06/28 19:07 (1000) There are no messages at all when unmounting. But when I then try to access the drive again (to write a new image), I see: VFS: busy inodes on changed media or resized disk sr0 VFS: busy inodes on changed media or resized disk sr0 VFS: busy inodes on changed media or resized disk sr0
Wolfgang, are you still seeing this with the 2.6.43/3.3 kernels?
(In reply to comment #4) > Wolfgang, are you still seeing this with the 2.6.43/3.3 kernels? I've only tried this 2 or 3 times since, with 3.x kernel versions, and it did not cause problems then.
(In reply to comment #5) > (In reply to comment #4) > > Wolfgang, are you still seeing this with the 2.6.43/3.3 kernels? > > I've only tried this 2 or 3 times since, with 3.x kernel versions, and it > did not cause problems then. OK. We'll close this out as F15 is going EOL. If you see it again, please reopen.