I have always expected the fall-back behavior of EXT3/2, but recently
have not seen it.  To be specific, the last time I tried to mount an
EXT3 RO, it insisted on trying journal replay.  When I marked the disk
RO in the block driver, still got attempted replay.  (And I/O errors
to match the failing write-back of the replay.)  When I tried to force
mounting as EXT2 (with a "-t ext2"), it still played journal.  That
was last week.  I think the kernel was 2.6.22 which should have a
mature EXT3.

So ... I'm not arguing about what EXT3 is supposed to do, but ... my
experience does not match my expectation.  I therefore recommend EXT2
for any shared RO disk or volume or LUN or partition.  (Unless you
want to go with ISO-9660, which also works.)

-- R;   <><





On Wed, Aug 25, 2010 at 08:26, Christian Paro <[email protected]> wrote:
> Difference is that ext3 is prepared to fall back to ext2 functionality when
> being read from a read-only device - which makes perfect sense, given that
> the journal's purpose is to protect the integrity of writes and (as Mark
> pointed out) it is of no value on a disk which will never again be written
> to.
>
> ReiserFS, not recognizing that the journal is only relevant when it is
> possible to write to the disk, doesn't gracefully fall back to a
> journal-less version of itself the way ext3 does.
>
> As for ext2 vs ext3 for the read-only file system, there is a question of
> whether it's a disk which is to be written to once and then linked to and
> mounted read-only by everyone (in which case ext2 makes the most sense, with
> ext3 adding nothing but unnecessary overhead), or a disk which is regularly
> written to by one system - which you for some reason want be able to link to
> and mount read-only from another (in which case the ext3 journal does still
> serve a purpose while the disk is being accessed in write mode by that first
> system).
>
> On Wed, Aug 25, 2010 at 8:07 AM, Shane <[email protected]> wrote:
>
>> On Tue, 2010-08-24 at 11:00 -0400, Richard Troth wrote:
>>
>> > As others have said, ReiserFS wants to replay the journal.  EXT3 does
>> > too.  (Not sure about EXT4.)  So for RO I recommend EXT2.
>>
>> Whilst I'm prepared to believe anything of reiserfs (not having
>> attempted to use it since before it was named thus), I query the
>> assertion re ext3. Were it to be mounted r/o everywhere, I fail to see
>> how the filesystem could have meta-data that would need replaying.
>> But I admit to not having specifically tested that - maybe someday.
>>
>> Shane ...
>>
>> ----------------------------------------------------------------------
>> For LINUX-390 subscribe / signoff / archive access instructions,
>> send email to [email protected] with the message: INFO LINUX-390 or
>> visit
>> http://www.marist.edu/htbin/wlvindex?LINUX-390
>> ----------------------------------------------------------------------
>> For more information on Linux on System z, visit
>> http://wiki.linuxvm.org/
>>
>
> ----------------------------------------------------------------------
> For LINUX-390 subscribe / signoff / archive access instructions,
> send email to [email protected] with the message: INFO LINUX-390 or visit
> http://www.marist.edu/htbin/wlvindex?LINUX-390
> ----------------------------------------------------------------------
> For more information on Linux on System z, visit
> http://wiki.linuxvm.org/
>

----------------------------------------------------------------------
For LINUX-390 subscribe / signoff / archive access instructions,
send email to [email protected] with the message: INFO LINUX-390 or visit
http://www.marist.edu/htbin/wlvindex?LINUX-390
----------------------------------------------------------------------
For more information on Linux on System z, visit
http://wiki.linuxvm.org/

Reply via email to