Comment 10 for bug 529002

Revision history for this message
Heath Hunnicutt (heath-hunnicutt) wrote :

I don't believe the formerly blamed item 509180 is truly the source of these errors. I have read that complete bug thread, and the error messages being posted there are similar to this error message. However, there is a critically important difference between the error messages.

Bug 509180:
    1. Some users are posting error messages which are due to an overly verbose ecryptfs module reporting normal operation when a system call is interrupted: return EINTR.
    2. The root issue being explored is a report of data corruption, and incorrectly reported file lengths, using ecryptfs.

This bug:
    1. The error message indicates a real problem in the ecryptfs operation, not merely forwarding an error condition from a lower-level call.
    2. Nobody has reported the issue (2) from 509180.

My experience with this bug:

uname -a: Linux x 2.6.38-15-generic #61-Ubuntu SMP Tue Jun 12 time 2012 i686 i686 i386 GNU/Linux

I have been using ecryptfs for 2 years on this laptop, multiple Ubuntu upgrades, never gave the module any thought.

Yesterday I rebuilt from source the module udlfb (USB Displaylink Frame Buffer) as found on git Fetch URL http://git.plugable.com/webdav/udlfb

Although I invoked "sudo make install", the kernel module was built into /lib/modules/`uname -r`/updates and I saw output from depmod but modules.dep and modules.order had not been updated and still pointed to the "stock" module.

I did a "sudo modprobe udlfb.ko" in the updates subdirectory, then a depmod, then rebooted.

ecryptfs module failed to load. Of course my home dir was missing. In fact, output from "lsmod" showed that only a couple of dozens modules were loaded, top among them being udlfb, with 0 references (wrong dependencies for it, obvious to me), and i915, but practically nothing else. Wireless (ath9k) was missing. Checking right now, my normal "lsmod | wc -l" is 64.

Figuring dependencies were wrong, I re-ran modprobe and depmod. I powered down, rebooted, and had no improvement. I powered down again, waited a long time to log in, and everything was in working order.

Checking dmesg, I see there was an error from ecryptfs logged 15 minutes after boot:

[ 922.072990] Valid eCryptfs headers not found in file header region or xattr region
[ 922.072994] Either the lower file is not in a valid eCryptfs format, or the key could not be retrieved. Plaintext passthrough mode is not enabled; returning -EIO
[ 926.000041] wlan0: authenticate with 00:26:82:f0:f6:f2 (try 1)

It appears that this occurred while I was logging in. Checking /var/log/kern.log, the error has only been in my log twice, starts only after the episode in which many modules did not load.