Minor code cleanup, no functional change.  Instead of remembering
what HIDs & CIDs to add later, just add them immediately.
Signed-off-by: NBjorn Helgaas <bjorn.helgaas@hp.com>
Signed-off-by: NLen Brown <len.brown@intel.com>

Nobody uses acpi_device_uid(), so this patch removes it.
Signed-off-by: NBjorn Helgaas <bjorn.helgaas@hp.com>
Signed-off-by: NLen Brown <len.brown@intel.com>

Every acpi_device has at least one ID (if there's no _HID or _CID, we
give it a synthetic or default ID).  So there's no longer a need to
check whether an ID exists; we can just use it.
Signed-off-by: NBjorn Helgaas <bjorn.helgaas@hp.com>
Signed-off-by: NLen Brown <len.brown@intel.com>

We now keep a single list of IDs that includes both the _HID and any
_CIDs.  We no longer need to keep track of whether the device has a _CID.
Signed-off-by: NBjorn Helgaas <bjorn.helgaas@hp.com>
Signed-off-by: NLen Brown <len.brown@intel.com>

There's no need to treat _HID and _CID differently.  Keeping them in
a single list makes code that uses the IDs a little simpler because it
can just traverse the list rather than checking "do we have a HID?",
"do we have any CIDs?"
Signed-off-by: NBjorn Helgaas <bjorn.helgaas@hp.com>
Reviewed-by: NAlex Chiang <achiang@hp.com>
Signed-off-by: NLen Brown <len.brown@intel.com>

This makes sure every acpi_device has at least one ID.  If we build an
acpi_device for a namespace node with no _HID or _CID, we sometimes
synthesize an ID like "LNXCPU" or "LNXVIDEO".  If we don't even have
that, give it a default "device" ID.

Note that this means things like:
    /sys/devices/LNXSYSTM:00/LNXSYBUS:00/HWP0001:00/HWP0002:04/device:00
(a PCI slot SxFy device) will have "hid" and "modprobe" entries, where
they didn't before.  These aren't very useful (a HID of "device" doesn't
tell you what *kind* of device it is, so it doesn't help find a driver),
but I don't think they're harmful.
Signed-off-by: NBjorn Helgaas <bjorn.helgaas@hp.com>
Signed-off-by: NLen Brown <len.brown@intel.com>

Use acpi_device_hid() rather than accessing acpi_device.pnp.hardware_id
directly.
Signed-off-by: NBjorn Helgaas <bjorn.helgaas@hp.com>
Signed-off-by: NLen Brown <len.brown@intel.com>

This makes \_SB_ show up as /sys/devices/LNXSYSTM:00/LNXSYBUS:00
rather than "device:00".  This has been broken for a loooong time
(at least since 2.6.13) because device->parent is an acpi_device
pointer, not a handle.
Signed-off-by: NBjorn Helgaas <bjorn.helgaas@hp.com>
Signed-off-by: NLen Brown <len.brown@intel.com>

acpi_bus_scan() traverses the namespace to enumerate devices and uses
acpi_add_single_object() to create acpi_devices.  When the platform
notifies us of a hot-plug event, we need to traverse part of the namespace
again to figure out what appeared or disappeared.  (We don't yet call
acpi_bus_scan() during hot-plug, but I plan to do that in the future.)

This patch makes acpi_add_single_object() notice when we already have
an acpi_device, so we don't need to make a new one.
Signed-off-by: NBjorn Helgaas <bjorn.helgaas@hp.com>
Signed-off-by: NLen Brown <len.brown@intel.com>

This patch adds acpi_bus_type_and_status(), which determines the type
of the object and whether we want to build an acpi_device for it.  If
it is acpi_device-worthy, it returns the type and the device's current
status.
Signed-off-by: NBjorn Helgaas <bjorn.helgaas@hp.com>
Signed-off-by: NLen Brown <len.brown@intel.com>

Add acpi_bus_get_status_handle() so we can get the status of a namespace
object before building a struct acpi_device.

This removes a use of "device->flags.dynamic_status", a cached indicator of
whether _STA exists.  It seems simpler and more reliable to just evaluate
_STA and catch AE_NOT_FOUND errors.
Signed-off-by: NBjorn Helgaas <bjorn.helgaas@hp.com>
Signed-off-by: NLen Brown <len.brown@intel.com>

acpi_bus_scan() currently walks the namespace manually.  This patch changes
it to use acpi_walk_namespace() instead.

Besides removing some complicated code, this means we take advantage of the
namespace locking done by acpi_walk_namespace().  The locking isn't so
important at boot-time, but I hope to eventually use this same path to
handle hot-addition of devices, when it will be important.

Note that acpi_walk_namespace() does not actually visit the starting node
first, so we need to do that by hand first.
Signed-off-by: NBjorn Helgaas <bjorn.helgaas@hp.com>
Signed-off-by: NLen Brown <len.brown@intel.com>

We can identify the root of the ACPI device tree by the fact that it
has no parent.  This is simpler than passing around ACPI_BUS_TYPE_SYSTEM
and will help remove special treatment of the device tree root.

Currently, we add the root by hand with ACPI_BUS_TYPE_SYSTEM.  If we
traverse the tree treating the root as just another device and use
acpi_get_type(), the root shows up as ACPI_TYPE_DEVICE.
Signed-off-by: NBjorn Helgaas <bjorn.helgaas@hp.com>
Signed-off-by: NLen Brown <len.brown@intel.com>

This patch changes the order so we enumerate in the "root, namespace,
functional fixed" order instead of the "root, functional fixed, namespace"
order.  When I change acpi_bus_scan() to use acpi_walk_namespace(), it
will use the former order, so this patch isolates the order change for
bisectability.
Signed-off-by: NBjorn Helgaas <bjorn.helgaas@hp.com>
Signed-off-by: NLen Brown <len.brown@intel.com>

This patch changes acpi_bus_scan() to take an acpi_handle rather than an
acpi_device pointer.  I plan to use acpi_bus_scan() in the hotplug path,
and I'd rather not assume that notifications only go to nodes that already
have acpi_devices.

This will also help remove the special case for adding the root node.  We
currently add the root by hand before acpi_bus_scan(), but using a handle
here means we can start the acpi_bus_scan() directly with the root even
though it doesn't have an acpi_device yet.

Note that acpi_bus_scan() currently adds and/or starts the *children* of
its device argument.  It doesn't do anything with the device itself.
Signed-off-by: NBjorn Helgaas <bjorn.helgaas@hp.com>
Signed-off-by: NLen Brown <len.brown@intel.com>

This patch adds acpi_bus_get_parent(), which ascends the namespace until
it finds a parent with an acpi_device.

Then we use acpi_bus_get_parent() in acpi_add_single_object(), so callers
don't have to figure out or keep track of the parent acpi_device.
Signed-off-by: NBjorn Helgaas <bjorn.helgaas@hp.com>
Signed-off-by: NLen Brown <len.brown@intel.com>

acpi_add_single_object() is static, and all callers supply a valid "child"
argument, so we don't need to check it.  This patch also remove some
unnecessary initializations.
Signed-off-by: NBjorn Helgaas <bjorn.helgaas@hp.com>
Signed-off-by: NLen Brown <len.brown@intel.com>

We now save the ACPI bus "device_type" in the acpi_device structure, so
we don't need to pass it around explicitly anymore.
Signed-off-by: NBjorn Helgaas <bjorn.helgaas@hp.com>
Signed-off-by: NLen Brown <len.brown@intel.com>

We only pass the "type" to acpi_device_set_context() so we know whether
the device has a handle to which we can attach the acpi_device pointer.
But it's safer to just check for the handle directly, since it's in the
acpi_device already.
Signed-off-by: NBjorn Helgaas <bjorn.helgaas@hp.com>
Signed-off-by: NLen Brown <len.brown@intel.com>

Check the acpi_device device_type rather than the HID.
Signed-off-by: NBjorn Helgaas <bjorn.helgaas@hp.com>
Signed-off-by: NLen Brown <len.brown@intel.com>

Most uses of the ACPI bus device_type (ACPI_BUS_TYPE_DEVICE,
ACPI_BUS_TYPE_POWER, etc) are during device initialization, but
we do need it later for notify handler installation, since that
is different for fixed hardware devices vs. namespace devices.

This patch saves the device_type in the acpi_device structure,
so we can check that rather than comparing against the _HID string.
Signed-off-by: NBjorn Helgaas <bjorn.helgaas@hp.com>
Signed-off-by: NLen Brown <len.brown@intel.com>

In several cases, functions take handle and parent device pointers in
addition to acpi_device pointers.  But the acpi_device structure contains
both the handle and the parent pointer, so it's pointless and error-prone
to pass them all.  This patch removes the unnecessary "handle" and "parent"
arguments.
Signed-off-by: NBjorn Helgaas <bjorn.helgaas@hp.com>
Signed-off-by: NLen Brown <len.brown@intel.com>

We never use the "root" argument, so just remove it.
Signed-off-by: NBjorn Helgaas <bjorn.helgaas@hp.com>
Signed-off-by: NLen Brown <len.brown@intel.com>

Add debug output for adding an ACPI device.  Enable this with
"acpi.debug_layer=0x00010000" (ACPI_BUS_COMPONENT).
Signed-off-by: NBjorn Helgaas <bjorn.helgaas@hp.com>
Signed-off-by: NLen Brown <len.brown@intel.com>

Commit 15b8dd53 changed info->hardware_id from a static array to
a pointer.  If hardware_id is non-NULL, it points to a NULL-terminated
string, so we don't need to terminate it explicitly.  However, it may
be NULL; in that case, we *can't* add a NULL terminator.

This causes a NULL pointer dereference oops for devices without _HID.
Signed-off-by: NBjorn Helgaas <bjorn.helgaas@hp.com>
CC: Lin Ming <ming.m.lin@intel.com>
CC: Bob Moore <robert.moore@intel.com>
CC: Gary Hade <garyhade@us.ibm.com>
Signed-off-by: NLen Brown <len.brown@intel.com>

* 'writeback' of git://git.kernel.dk/linux-2.6-block:
  writeback: writeback_inodes_sb() should use bdi_start_writeback()
  writeback: don't delay inodes redirtied by a fast dirtier
  writeback: make the super_block pinning more efficient
  writeback: don't resort for a single super_block in move_expired_inodes()
  writeback: move inodes from one super_block together
  writeback: get rid to incorrect references to pdflush in comments
  writeback: improve readability of the wb_writeback() continue/break logic
  writeback: cleanup writeback_single_inode()
  writeback: kupdate writeback shall not stop when more io is possible
  writeback: stop background writeback when below background threshold
  writeback: balance_dirty_pages() shall write more than dirtied pages
  fs: Fix busyloop in wb_writeback()

Pointless to iterate other devices looking for a super, when
we have a bdi mapping.
Signed-off-by: NJens Axboe <jens.axboe@oracle.com>

Debug traces show that in per-bdi writeback, the inode under writeback
almost always get redirtied by a busy dirtier.  We used to call
redirty_tail() in this case, which could delay inode for up to 30s.

This is unacceptable because it now happens so frequently for plain cp/dd,
that the accumulated delays could make writeback of big files very slow.

So let's distinguish between data redirty and metadata only redirty.
The first one is caused by a busy dirtier, while the latter one could
happen in XFS, NFS, etc. when they are doing delalloc or updating isize.

The inode being busy dirtied will now be requeued for next io, while
the inode being redirtied by fs will continue to be delayed to avoid
repeated IO.

CC: Jan Kara <jack@suse.cz>
CC: Theodore Ts'o <tytso@mit.edu>
CC: Dave Chinner <david@fromorbit.com>
CC: Chris Mason <chris.mason@oracle.com>
CC: Christoph Hellwig <hch@infradead.org>
Signed-off-by: NWu Fengguang <fengguang.wu@intel.com>
Signed-off-by: NJens Axboe <jens.axboe@oracle.com>

Currently we pin the inode->i_sb for every single inode. This
increases cache traffic on sb->s_umount sem. Lets instead
cache the inode sb pin state and keep the super_block pinned
for as long as keep writing out inodes from the same
super_block.
Signed-off-by: NJens Axboe <jens.axboe@oracle.com>

If we only moved inodes from a single super_block to the temporary
list, there's no point in doing a resort for multiple super_blocks.
Signed-off-by: NJens Axboe <jens.axboe@oracle.com>

__mark_inode_dirty adds inode to wb dirty list in random order. If a disk has
several partitions, writeback might keep spindle moving between partitions.
To reduce the move, better write big chunk of one partition and then move to
another. Inodes from one fs usually are in one partion, so idealy move indoes
from one fs together should reduce spindle move. This patch tries to address
this. Before per-bdi writeback is added, the behavior is write indoes
from one fs first and then another, so the patch restores previous behavior.
The loop in the patch is a bit ugly, should we add a dirty list for each
superblock in bdi_writeback?

Test in a two partition disk with attached fio script shows about 3% ~ 6%
improvement.
Signed-off-by: NShaohua Li <shaohua.li@intel.com>
Reviewed-by: NWu Fengguang <fengguang.wu@intel.com>
Signed-off-by: NJens Axboe <jens.axboe@oracle.com>

Signed-off-by: NJens Axboe <jens.axboe@oracle.com>

And throw some comments in there, too.
Reviewed-by: NWu Fengguang <fengguang.wu@intel.com>
Signed-off-by: NJens Axboe <jens.axboe@oracle.com>

Make the if-else straight in writeback_single_inode().
No behavior change.

Cc: Jan Kara <jack@suse.cz>
Cc: Michael Rubin <mrubin@google.com>
Cc: Peter Zijlstra <peterz@infradead.org>
Signed-off-by: NFengguang Wu <wfg@mail.ustc.edu.cn>
Signed-off-by: NJens Axboe <jens.axboe@oracle.com>

Fix the kupdate case, which disregards wbc.more_io and stop writeback
prematurely even when there are more inodes to be synced.

wbc.more_io should always be respected.

Also remove the pages_skipped check. It will set when some page(s) of some
inode(s) cannot be written for now. Such inodes will be delayed for a while.
This variable has nothing to do with whether there are other writeable inodes.

CC: Jan Kara <jack@suse.cz>
CC: Dave Chinner <david@fromorbit.com>
CC: Peter Zijlstra <a.p.zijlstra@chello.nl>
Signed-off-by: NWu Fengguang <fengguang.wu@intel.com>
Signed-off-by: NJens Axboe <jens.axboe@oracle.com>

Treat bdi_start_writeback(0) as a special request to do background write,
and stop such work when we are below the background dirty threshold.

Also simplify the (nr_pages <= 0) checks. Since we already pass in
nr_pages=LONG_MAX for WB_SYNC_ALL and background writes, we don't
need to worry about it being decreased to zero.
Reported-by: NRichard Kennedy <richard@rsk.demon.co.uk>
CC: Jan Kara <jack@suse.cz>
Acked-by: NPeter Zijlstra <a.p.zijlstra@chello.nl>
Signed-off-by: NWu Fengguang <fengguang.wu@intel.com>
Signed-off-by: NJens Axboe <jens.axboe@oracle.com>

Some filesystem may choose to write much more than ratelimit_pages
before calling balance_dirty_pages_ratelimited_nr(). So it is safer to
determine number to write based on real number of dirtied pages.

Otherwise it is possible that
  loop {
    btrfs_file_write():     dirty 1024 pages
    balance_dirty_pages():  write up to 48 pages (= ratelimit_pages * 1.5)
  }
in which the writeback rate cannot keep up with dirty rate, and the
dirty pages go all the way beyond dirty_thresh.

The increased write_chunk may make the dirtier more bumpy.
So filesystems shall be take care not to dirty too much at
a time (eg. > 4MB) without checking the ratelimit.
Signed-off-by: NWu Fengguang <fengguang.wu@intel.com>
Acked-by: NPeter Zijlstra <a.p.zijlstra@chello.nl>
Signed-off-by: NJens Axboe <jens.axboe@oracle.com>

If all inodes are under writeback (e.g. in case when there's only one inode
with dirty pages), wb_writeback() with WB_SYNC_NONE work basically degrades
to busylooping until I_SYNC flags of the inode is cleared. Fix the problem by
waiting on I_SYNC flags of an inode on b_more_io list in case we failed to
write anything.
Tested-by: NWu Fengguang <fengguang.wu@intel.com>
Signed-off-by: NJan Kara <jack@suse.cz>
Signed-off-by: NJens Axboe <jens.axboe@oracle.com>