JFS: Flush dirty metadata to disk when remounting from read-write
to read-only.  Also fix umount ordering to make sure metadata is
written before journal closed.

With Andrew Morton's recent changes, JFS is no longer writing much
of its dirty metadata when remounting from read-write to read-only.
This causes severe file system corruption.  A JFS root file system
will be corrupted on shutdown.

This patch fixes the problem by explicitly writing the dirty metadata
before the journal is closed.  It also fixes the ordering so that
the dirty metadata is completely written before the journal is closed
for the normal unmount case as well.

into penguin.transmeta.com:/home/penguin/torvalds/repositories/kernel/linux

into penguin.transmeta.com:/home/penguin/torvalds/repositories/kernel/linux

Adding some more flexibility to Config.in (Allow for core built-in but
socket driver modular) cleans up the Makefile as well ;-)

Use the usual list-based approach. Still not too nice.

Old pcmcia-cs versions expect the pcmcia modules in 
/lib/modules/`uname -r`/pcmcia, so we symlinked them for compatibility.
According to the comment, the generation of these links should have
been removed in 2.4, so it's definitely time for it to go now.

Don't shortcut from drivers/Makefile to 
drivers/message/{fusion,i2o}/Makefile, but use intermediate
drivers/message/Makefile.

Cleanup drivers/message/fusion/Config.in and get rid of unnecessary
CONFIG_FUSION_BOOT.

Link the subdirs of drivers/char from drivers/char/Makefile instead the
top-level Makefile. 

Link order changes slightly, shouldn't case any problems, though.

Fix drivers/char/pcmcia/Config.in.

Cleanup drivers/char/pcmcia/Makefile.

Fix rio build rules in drivers/char/Makefile.

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

into tp1.ruhr-uni-bochum.de:/home/kai/kernel/v2.5/linux-2.5.make

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

into penguin.transmeta.com:/home/penguin/torvalds/repositories/kernel/linux

into penguin.transmeta.com:/home/penguin/torvalds/repositories/kernel/linux

There's actually two cases when a new NCCI is created:
outgoing and incoming connections (had missed the latter)

above explaining the {phys,pfn}_base stuff.

into penguin.transmeta.com:/home/penguin/torvalds/repositories/kernel/linux

Applications really ought to take care of maintaing the state themselves,
instead of cheating by listening to controller private messages.

    - Report better errors, and in one case the correct one.
    - Adds strategic wmb() calls
    - Claims the right (scaled) ISO bandwidth
    - Uses non-CVS version ID

This will likely resolve that Archos MP3 Jukebox issue, where "-104"
(-ECONNRESET) was appearing mysteriously.

minor -dj tree updates.

add PMAC changes found in -dj tree

sync up with -dj tree

First step in having the users alloc this as part of the state info
they have to maintain anyway.

change due to moved location of i386's pci.h

    - generalizes/cleans keventd support to also handle
        * rx stalls (and usb 2.0 transaction translator unplug)
        * rx memory shortfalls (latent bug Oliver noticed)
        * cleanup on device disconnect (quiesce first)
    - merges Brad's patch to use the IEEE802 "locally assigned" bit
    - fixes a couple minor bugs on error paths (leak, bogus diagnostic)
    - updates some comments

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

Ah.  That's some left-over code.  Reads will be OK, but writes will be
unexpectedly asynchronous.  Nothing in the kernel uses that function for
writes so it didn't show up.

  Trivial patch to change some instances of suser() and fsuser() to
proper capable() checks.

In Rules.make, a comment says:

# Old makefiles define their own rules for compiling .S files,
# but these standard rules are available for any Makefile that
# wants to use them.  Our plan is to incrementally convert all
# the Makefiles to these standard rules.  -- rmk, mec

This patch does that for the i386 arch Makefiles.

There are missing #includes which will break compilation on some non-x86
platforms. With following patch this compiles and works on alpha.

Summary of changes:
- alpha, arm: code related to PCI-PCI bridges from pcibios_fixup_bus()
  removed - now it's generic;
- pdev_sort_resource: sort resources all together, no matter IO or memory;
- pbus_assign_resources_sorted: ditto;
- pci_bridge_check_ranges, pci_setup_bridge: changed for prefetch support;
- pbus_size_io, pbus_size_mem: core stuff; tested with randomly generated
  sets of resources;
- pbus_size_bridges: pass #2 (pass #1 is PCI probing, common for all archs);
- pbus_assign_resources: pass #3.

Ivan.

This changes PCI resource allocation algorithm to 3 passes vs.
current 2 passes. Extra pass is used for calculation of required
size and alignment of PCI buses behind PCI-PCI bridges. After
that, in the pass #3, these buses get allocated like regular
PCI devices. This gives tighter PCI IO and memory packing -
for instance, this fixes allocation problems on certain alphas
with very small (112Mb) PCI memory range. Also, the new code
- will allow mixed approach to resource allocation:
  architecture can keep BIOS settings for some devices,
  and re-allocate resources for others, including improperly
  initialized bridges;
- makes prefetchable ranges support much simpler;
- allows sizing of IO and memory ranges for the host
  bridges, which might be very useful in some situations.

It was tested on various alphas; I haven't heard any complaints
from rmk and rth, so probably all of this is ok. :-)

Part 1:
- for all archs, 4th argument (align) added to
  pcibios_align_resource (and its callers).
  It's necessary because this function will be called for
  bus resources as well, and in this case size != alignment.
- for several archs, dead/bogus code removed from
  pcibios_fixup_pbus_ranges().

into segfault.osdl.org:/home/mochel/src/kernel/devel/linux-2.5-pci