A new File Browser window launches in every suspend cycle when a USB/SDcard is attached.

[mavrothal]

In os850, powerd 26, gnome desktop for both XO-1 and X0-1.5 the autobowse function of nautilus launched a new window of the device (usb/sdcard) in every suspend cycle. With a 15sec timeout setting can be very annoying. Setting "File Browser Preferences -> media -> Browse media when inserted" to false (unckeck box) remedies that. Either powerd or file browser settings should be fixed.

[Quozl]

Reproduced.

[Quozl]

Related #10295. (The filesystem is mounted by the media becoming available, yet is not mounted on suspend).

[Quozl]

Not a regression. Still bad.

[Quozl]

Reproduced day before yesterday, and just now, on an XO-1.5 HS with build os850 and a specific USB device.

Failed to reproduce just now on an XO-1.5 with build os851 using four different devices.

Reproduced just now on XO-1.5 with build os851 using the first USB device.

The symptom follows a specific USB device, which contains a FAT32 partition. At console, after waiting for idle suspend, typing df to wake, repeating it, shows the filesystem is momentarily not mounted and then mounted. dmesg shows a "FAT: bread failed in fat_clusters_flush".

[Quozl]

fail to reproduce dmesg, quozl's blue usb stick

[Quozl]

reproducd dmesg, quozl's white usb stick

[Quozl]

An attempt was made to reproduce with affected and unaffected USB stick, and the dmesg output attached. The dmesg output was trimmed and diff'd, and the only relevant changes were the USB stick SerialNumber and these lines reported after resume:

@@ -604,3 +686,8 @@
 Restarting tasks ... 
 hub 1-0:1.0: over-current change on port 2
 done.
+FAT: bread failed in fat_clusters_flush
+sd 0:0:0:0: [sda] 7856127 512-byte logical blocks: (4.02 GB/3.74 GiB)
+sd 0:0:0:0: [sda] Assuming drive cache: write through
+sd 0:0:0:0: [sda] Assuming drive cache: write through
+ sda: sda1

[Quozl]

Using single-user mode and rsync, but without reformatting the filesystems, I swapped the filesystem content between the affected and unaffected USB sticks. Both USB sticks became affected.

[Quozl]

I'm fairly sure this is not a problem with powerd or nautilus settings, although both things things could be changed to alleviate the symptom; powerd by not suspending when a filesystem is mounted, nautilus by not mounting filesystems automatically. I'm not suggesting either of those changes, though the original poster thinks the solution is with those components.

It seems more likely to be a problem with the kernel handling of FAT filesystems for devices that are momentarily offline. As such, it might occur on directed suspend, such as lid close or power button press.

Tested directed suspend with affected USB devices. Lid close or power button did not cause the symptom of launching a new window for the filesystem. Idle suspend continued to show the symptom.

[martin.langhoff]

This is nasty.

Could it be related to "aggressive/parallelized USB resume"? We need it for fast resume, but it can trip up flaky devices.

It's spooky that your content test led to both devices being affected. Maybe it's the footprint of directory entries? Does a fsck.vfat find any corruption?

[Quozl]

There was no corruption detected by fsck.vfat that I recall. Given the message about FAT clusters, it is probably related to usage or filling of the filesystem, or a certain amount of cluster allocation.

Google for

"bread failed in fat_clusters_flush" lkml suspend resume

... one of those patches might be relevant. It probably doesn't happen on recent kernels. It may have been patched on the Fedora 11 kernel; I gather we are using the upstream Linus kernel. I don't think it has anything to do with OLPC's changes.

[dsd]

I think our idle suspend model is asking for trouble with USB devices, see my comment in #10295.

This bug also mentions SD cards; I'm surprised that this happens there, as I expect SD power to be maintained during suspend so that mounts should not change. I might be missing something. Need to investigate like has been done above for USB.

[mavrothal]

I think this bug should be closed since newer builds/powerd do _not_ have this problem. A new window manager window does not open although "Browse media when inserted" is on in Nautilus preferences

The only remaining issue is that if a USB stick is unmounted but not removed, it auto-remounts and the file manager window opens upon wakeup from idle-suspend
This is _not_ the case for SDcards

[godiard]

Tested in 872, I think can be closed. Open only one window if the device is attached and not mounted, and don not open more windows.

[Quozl]

Agreed, the symptom no longer occurs, when I test for it as I did before; by leaving the filesystem mounted over suspend you no longer get multiple windows containing the same view. If you use the Eject option or icon before the idle suspend, then after the resume a new window is opened.