Skip to contentSkip to conversation

exofs changes for 2.6.34

GIT PULL Boaz Harrosh 2 participants 3 messages Mar 4, 2010 – Mar 7, 2010 VFS

lore
Share & tools
lore
#0 lore raw

Linus please pull the following changes since commit:

Linux 2.6.33-rc6

They are available in the git repository at:

git://git.open-osd.org/linux-open-osd.git for-linus

Boaz Harrosh (11):
exofs: debug print even less
exofs: Micro-optimize exofs_i_info
exofs: Recover in the case of read-passed-end-of-file
exofs: Move layout related members to a layout structure
exofs: unindent exofs_sbi_read
exofs: Define on-disk per-inode optional layout attribute
exofs: RAID0 support
exofs: convert io_state to use pages array instead of bio at input
exofs: Error recovery if object is missing from storage
exofs: Prepare for groups
exofs: groups support

 fs/exofs/common.h |   39 ++++
 fs/exofs/exofs.h  |   53 ++++-
 fs/exofs/inode.c  |  194 +++++++++++-------
 fs/exofs/ios.c    |  575 +++++++++++++++++++++++++++++++++++++++++++++--------
 fs/exofs/super.c  |  121 +++++++++---
 5 files changed, 779 insertions(+), 203 deletions(-)

What's new:
The main changes are support for data striping and *striping-groups*.
striping-groups is a nice option when there are lots of devices in the system, but the striping of each file is confined to fewer devices, so not to keep all devices busy on each IO. For example a 10 out of 100 striping means each file is striped across 10 devices out of 100 total devices in the full system.
Note that exofs has a unique layout structure that enables Mirrors
orthogonal to any striping/raid arrangement.

Next changes will be support for raid5 and raid6

#1 lore raw
Quoted 1 line · Boaz Harrosh

Next changes will be support for raid5 and raid6

Err, you're not pushing another bloody copy of raid algorithms in, are you?

#2 lore raw
Quoted 6 lines · Christoph Hellwig
Quoted 1 line · Boaz Harrosh

Next changes will be support for raid5 and raid6

Err, you're not pushing another bloody copy of raid algorithms in, are you?

No of course not. You should know me by now. I intend to fully reuse both linux/raid/pq.h and linux/async_tx.h.
[What? lose all these CPU/HW optimizations people have done ?-)]

I'm struggling with re-factoring some of drivers/md/raid* as I'm using bios just as they do, and bio_lists are fine, but currently the code is very md/dm structure specific, there is no abstract bio_list API but I hope I can extract one with minimum risk. I have not yet looked into btrfs, they do raid5 as well right?

But all this is far down the road, off course I'll send RFCs first. The above push is just for that unique objects groups/striping/mirroring, which is very small code, a couple of functions actually, over the basic IO.

Thanks for asking, though.
Boaz