into nuts.ninka.net:/home/davem/src/BK/net-2.5

I don't understand what was on the mind of Procom programmers,
why do all this inverted logic? Its plain confusing, revert it.
Thanks to DaveM for asking if the logic was inverted, I should
have killed this weird stuff a long time ago :-\

With this we make it easier to write correct network families as less
details need to be taken into account, as well in the current state we
make the non-refcounting protocols (the ones still keeping deliver_to_old_ones
in the tree) suck less. 8)

Left a WARN_ON in sk_del_node_init for a while, so that we can catch cases
where we're using __sock_put on a struct sock that has refcnt == 1, which
is not the case for all the ones I tested.

into nuts.ninka.net:/home/davem/src/BK/net-2.5

into nuts.ninka.net:/home/davem/src/BK/net-2.5

into home.transmeta.com:/home/torvalds/v2.5/linux

into kroah.com:/home/greg/linux/BK/i2c-2.5

Boot ROMs have talked TFTP forever.  Some do it over USB now.

  - switch to usb_buffer_alloc

We cannot sensibly support O_DIRECT reads or writes when all writes are
journalled.

This is because the VFS explicitly avoids syncing the file metadata during
O_DIRECT reads and writes.  ext3 with journalled data will leave pending
changes in memory and they will overwrite the results of O_DIRECT writes, and
O_DIRECT reads will not return the latest data.

Setting the a_op to null will cause opens and fcntl(F_SETFL) to return
-EINVAL if O_DIRECT is requested.

Minro tweak: once log_wait_for_space() has created sufficient space in the
journal to start the new handle, hang onto the spinlock as
start_this_handle() loops around to reevaluate the journal's state.

It prevents anyone else from zooming in and stealing the space we just made.

There are various places in which JBD is starting a commit against a
transaction without sufficient locking in place to ensure that that
transaction is still alive.

Change it so that log_start_commit() takes a transaction ID instead.  Make
the caller take a copy of that ID inside the appropriate locks.

log_do_checkpoint is playing around with a transaction pointer without enough
locking to ensure that it is valid.  Fix that up by revalidating the
transaction after acquiring the right locks.

With data=ordered it is often the case that a quick write-and-truncate will
leave large numbers of pages on the page LRU with no ->mapping, and attached
buffers.  Because ext3 was not ready to let the pages go at the time of
truncation.

These pages are trivially reclaimable, but their seeming absence makes the VM
overcommit accounting confused (they don't count as "free", nor as
pagecache).  And they make the /proc/meminfo stats look odd.

So what we do here is to try to strip the buffers from these pages as the
buffers exit the journal commit.

This was a flag which said "the transaction's time is active".
timer_pending() could have told us that, but in fact there is no need to
query it at all.

Fix various problems which cropped up due to MAP_SHARED traffic on
data=journal with blocksize < PAGE_CACHE_SIZE.

All relate to handling the "pending truncate" buffers outside i_size.

start_this_handle() can decide to add this handle to a transaction, but
kjournald then moves the handle into commit phase.

Extend the coverage of j_state_lock so that start_this_transaction()'s
examination of journal->j_state is atomic wrt journal_commit_transaction().

Try to trap some more state when an assertion which cannot happen happens.

Before taking the highly-taken j_list_lock, take a peek to seem if this
buffer is already journalled and in the appropriate state.

ext3 no longer keeps the filesystem-wide free blocks counter and free inodes
counter up to date all the time in the superblock.  Because that requires
fs-wide locking.  These counters are only needed at runtime for the Orlov
allocator heuristics, and we are now using a fuzzy per-cpu coutner for that.

These counters are rather unnecessary: the same info is present in the file
allocation maps and inode tables, the group descriptor blocks and the
bitmaps.

e2fsck will be changed to downgrade the seriousness of this inconsistency.

The filesystem _will_ write these numbers out in the superblock on a clean
unmount, based on the sum of the free block and inode counts in the group
descriptors.

From: Andreas Gruenbacher <agruen@suse.de>

This function can leak a posix_acl on an error path.

The ioctl handler can leave a transaction open on an error path.  That
will wedge up the filesystem.

add a dump_stack() to a can't-happen path which happened during development.

There is a race between transaction commit's attempt to free journal_heads
and journal_try_to_free_buffers' attempt.

Fix that by taking a ref against the journal_head in
journal_try_to_free_buffers().

ext3's fully data-journalled mode has been broken for a year.  This patch
fixes it up.

The prepare_write/commit_write/writepage implementations have been split up.
Instead of having each function handle all three journalling mode we now have
three separate sets of address_space_operations.

The problematic part of data=journal is MAP_SHARED writepage traffic: pages
which don't have buffers.  In 2.4 these were cheatingly treated as
data-ordered buffers and that caused several nasty problems.

Here we do it properly: writepage traffic is fully journalled.  This means
that the various workarounds for the 2.4 scheme can be removed, when I
remember where they all are.

The PG_checked flag has been borrowed: it it set in the atomic set_page_dirty
a_op to tell the subsequent writepage() that this page needs to have buffers
attached, dirtied and journalled.

This rather defines PG_checked as "fs-private info in page->flags" and it
should be renamed sometime.

Avoid holding the journal's j_list_lock while copying the buffer_head's data.
We hold jbd_lock_bh_state() during the copy, which is all that is needed.

In start_this_handle() the caller does not have a handle ref pinning the
transaction open, and so the call to log_start_commit() is racy because some
other CPU could take the transaction into commit state independently.

Fix that by holding j_state_lock (which pins j_running_transaction) across
the log_start_commit() call.

Plug a conceivable race with the freeing up of trasnactions, and add some
more debug checks.

Drop in a few assertions to ensure that the locking rules are being adhered
to.

Add a comment describing why a race isn't there.