now Linux 3.2 can read ext3 16k blocks !

It's still some time until we'll see first linux distributors picking up the new kernel.
Ubuntu is going to release their next LTS version in late april with Kernel 3.2, debian will probably not adopt the new kernel for even longer.
The redhat based distros are usually a little faster, but I'd bet it's march before they release one with the new kernel.

This is good news for all of us Sparc users (I have two NV+ v1 units)

I almost fear it's too early to start the party since I read the news on the new kernel myself.

The H (http://www.h-online.com/open/features/K ... 87311.html) wrote:
The Ext4 filesystem now supports big allocation blocks (e.g 1, 2, 3). The technique, known as bigalloc, bundles the 4K blocks use to store data into clusters of up to 1 MB.

Does this really mean we can now mount filesystems with bigger clustersize or does this only mean that ext4 is now able to combine some internal allocation blocks to bigger ones?
I mean is this bigalloc really what the bigger cluster size on sparc architecture is about or is this some completely different concept?

Looks like someone has to build this kernel and run it to see if it really helps.
Maybe if I find the time I could do that next week, it's not that hard to do, I'm just not sure about the extra fuzz that needs to be built from source like e2fsprogs etc.

Sparc ReadyNAS use EXT3, so if it is only EXT4 then it doesn't apply. From reading the quote you gave, I'm not so sure it is what we want in regards to handling 16k blocks either.

I think you're both right, I'm afraid there's almost no hope to read those disks as I imagined...
- This big blocks feature is mentionned only for ext4
- It is handled by the filesystem itself by clustering block allocation
- The kernel still deals with 4k blocks...

Since ext3 can be read by ext4 without conversion, there is still a small hope if the bigalloc clustering is equivalent to the 16k blocks...

:(

I've tested Linux 3.2 (from Mageia 2 alpha3 distro) : I can't read an ext3 16k made on the ReadNAS Duo...

My tests:
- make an empty 16MB file
- loop with losetup
- make default ext3 fs
=> It gets 1024k block (small fs)
Any Linux with ext3 reads it (loop mount)

- make ext3 with explicit 16k blocks
=> pre 3.2 and 3.2 Linux can't read it... :(
(tried loop mount with automatic fs detection and tried with -t ext4, since ext4 is advertised 16k compatible with 3.2 kernel)

I'll try to ask to kernel or ext4 developers...