Re: Help with a possibly corrupt repo

From: Jonathan Leto <jaleto@gmail.com>
Date: May 20, 2010, 14:57 UTC
Message-ID: <AANLkTikSo8hTPXG8kGgCLOul6wp8THobUW_rSni4D6YC@mail.gmail.com>
In-Reply-To: <20100510212801.GJ14069@dpotapov.dyndns.org>

Howdy,

I ran into this corruption issue using ecryptfs with an ext4 filesystem (on a local SSD drive) on Ubuntu 9.10. The corruption seems to coincide with getting these errors in my dmesg output:

[12773.772426] ecryptfs_decrypt_page: Error attempting to read lower page; rc = [-22] [12773.772431] ecryptfs_readpage: Error decrypting page; rc = [-22] [12796.061574] ecryptfs_read_lower: octets_read = [-4]; expected [4096] [12796.061584] ecryptfs_read_and_validate_header_region: Error reading header region; rc = [-22]

I am lead to believe that ecryptfs+ext4 is not yet production-ready, although it could be a weird failure mode of my drive. I have not had more time to test this setup with other hardware.

Duke

On Mon, May 10, 2010 at 2:28 PM, Dmitry Potapov <dpotapov@gmail.com> wrote:

Show 18 quoted lines

> On Mon, May 10, 2010 at 02:05:09PM -0700, Jonathan Leto wrote:
>> 00020c0 deed 3541 7800 56dd 005c 0000 0000 0000
>> 00020d0 0000 0000 0000 0000 0000 0000 0000 0000
>> *
>> 0003000
>> error: garbage at end of loose object '66fdbaef5ae12a8402f6b0f7c1a73a3ce72e8eba'
>> fatal: object 66fdbaef5ae12a8402f6b0f7c1a73a3ce72e8eba is corrupted
>
> The object is "tree", and it contains a lot of zeros at end.
> I have truncated them, and it appears to fix the problem.
> I have attached the corrected file.
>
> But the important question is how it could happen. What filesystem
> do you use? Do you store it on the local disk or network share?
>
>
> Dmitry
>

-- 
Jonathan "Duke" Leto
jonathan@leto.net
http://leto.net