openwrt/staging/blogic.git
17 years ago[XFS] Fix XFS_IOC_FSBULKSTAT{,_SINGLE} & XFS_IOC_FSINUMBERS in compat mode
Michal Marek [Wed, 11 Jul 2007 01:10:19 +0000 (11:10 +1000)]
[XFS] Fix XFS_IOC_FSBULKSTAT{,_SINGLE} & XFS_IOC_FSINUMBERS in compat mode

* 32bit struct xfs_fsop_bulkreq has different size and layout of
members, no matter the alignment. Move the code out of the #else
branch (why was it there in the first place?). Define _32 variants of
the ioctl constants.
* 32bit struct xfs_bstat is different because of time_t and on
i386 because of different padding. Make xfs_bulkstat_one() accept a
custom "output formatter" in the private_data argument which takes care
of the xfs_bulkstat_one_compat() that takes care of the different
layout in the compat case.
* i386 struct xfs_inogrp has different padding.
Add a similar "output formatter" mecanism to xfs_inumbers().

SGI-PV: 967354
SGI-Modid: xfs-linux-melb:xfs-kern:29102a

Signed-off-by: Michal Marek <mmarek@suse.cz>
Signed-off-by: David Chinner <dgc@sgi.com>
Signed-off-by: Tim Shimmin <tes@sgi.com>
17 years ago[XFS] Compat ioctl handler for handle operations
Michal Marek [Wed, 11 Jul 2007 01:10:09 +0000 (11:10 +1000)]
[XFS] Compat ioctl handler for handle operations

32bit struct xfs_fsop_handlereq has different size and offsets (due to
pointers). TODO: case XFS_IOC_{FSSETDM,ATTRLIST,ATTRMULTI}_BY_HANDLE still
not handled.

SGI-PV: 967354
SGI-Modid: xfs-linux-melb:xfs-kern:29101a

Signed-off-by: Michal Marek <mmarek@suse.cz>
Signed-off-by: David Chinner <dgc@sgi.com>
Signed-off-by: Tim Shimmin <tes@sgi.com>
17 years ago[XFS] Compat ioctl handler for XFS_IOC_FSGEOMETRY_V1.
Michal Marek [Wed, 11 Jul 2007 01:09:57 +0000 (11:09 +1000)]
[XFS] Compat ioctl handler for XFS_IOC_FSGEOMETRY_V1.

i386 struct xfs_fsop_geom_v1 has no padding after the last member, so the
size is different.

SGI-PV: 967354
SGI-Modid: xfs-linux-melb:xfs-kern:29100a

Signed-off-by: Michal Marek <mmarek@suse.cz>
Signed-off-by: David Chinner <dgc@sgi.com>
Signed-off-by: Tim Shimmin <tes@sgi.com>
17 years ago[XFS] Clean up function name handling in tracing code
Eric Sandeen [Wed, 11 Jul 2007 01:09:47 +0000 (11:09 +1000)]
[XFS] Clean up function name handling in tracing code

Remove the hardcoded "fnames" for tracing, and just embed them in tracing
macros via __FUNCTION__. Kills a lot of #ifdefs too.

SGI-PV: 967353
SGI-Modid: xfs-linux-melb:xfs-kern:29099a

Signed-off-by: Eric Sandeen <sandeen@sandeen.net>
Signed-off-by: David Chinner <dgc@sgi.com>
Signed-off-by: Tim Shimmin <tes@sgi.com>
17 years ago[XFS] Quota inode has no parent.
David Chinner [Wed, 11 Jul 2007 01:09:33 +0000 (11:09 +1000)]
[XFS] Quota inode has no parent.

Avoid using a special "zero inode" as the parent of the quota inode as
this can confuse the filestreams code into thinking the quota inode has a
parent. We do not want the quota inode to follow filestreams allocation
rules, so pass a NULL as the parent inode and detect this condition when
doing stream associations.

SGI-PV: 964469
SGI-Modid: xfs-linux-melb:xfs-kern:29098a

Signed-off-by: David Chinner <dgc@sgi.com>
Signed-off-by: Tim Shimmin <tes@sgi.com>
17 years ago[XFS] Concurrent Multi-File Data Streams
David Chinner [Wed, 11 Jul 2007 01:09:12 +0000 (11:09 +1000)]
[XFS] Concurrent Multi-File Data Streams

In media spaces, video is often stored in a frame-per-file format. When
dealing with uncompressed realtime HD video streams in this format, it is
crucial that files do not get fragmented and that multiple files a placed
contiguously on disk.

When multiple streams are being ingested and played out at the same time,
it is critical that the filesystem does not cross the streams and
interleave them together as this creates seek and readahead cache miss
latency and prevents both ingest and playout from meeting frame rate
targets.

This patch set creates a "stream of files" concept into the allocator to
place all the data from a single stream contiguously on disk so that RAID
array readahead can be used effectively. Each additional stream gets
placed in different allocation groups within the filesystem, thereby
ensuring that we don't cross any streams. When an AG fills up, we select a
new AG for the stream that is not in use.

The core of the functionality is the stream tracking - each inode that we
create in a directory needs to be associated with the directories' stream.
Hence every time we create a file, we look up the directories' stream
object and associate the new file with that object.

Once we have a stream object for a file, we use the AG that the stream
object point to for allocations. If we can't allocate in that AG (e.g. it
is full) we move the entire stream to another AG. Other inodes in the same
stream are moved to the new AG on their next allocation (i.e. lazy
update).

Stream objects are kept in a cache and hold a reference on the inode.
Hence the inode cannot be reclaimed while there is an outstanding stream
reference. This means that on unlink we need to remove the stream
association and we also need to flush all the associations on certain
events that want to reclaim all unreferenced inodes (e.g. filesystem
freeze).

SGI-PV: 964469
SGI-Modid: xfs-linux-melb:xfs-kern:29096a

Signed-off-by: David Chinner <dgc@sgi.com>
Signed-off-by: Barry Naujok <bnaujok@sgi.com>
Signed-off-by: Donald Douwsma <donaldd@sgi.com>
Signed-off-by: Christoph Hellwig <hch@infradead.org>
Signed-off-by: Tim Shimmin <tes@sgi.com>
Signed-off-by: Vlad Apostolov <vapo@sgi.com>
17 years ago[XFS] Use uninitialized_var macro to stop warning about rtx
Andrew Morton [Thu, 28 Jun 2007 06:46:56 +0000 (16:46 +1000)]
[XFS] Use uninitialized_var macro to stop warning about rtx

Appease gcc in regards to "warning: 'rtx' is used uninitialized in
this function".

SGI-PV: 907752
SGI-Modid: xfs-linux-melb:xfs-kern:29007a

Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Tim Shimmin <tes@sgi.com>
17 years ago[XFS] XFS should not be looking at filp reference counts
Christoph Hellwig [Thu, 28 Jun 2007 06:46:47 +0000 (16:46 +1000)]
[XFS] XFS should not be looking at filp reference counts

A check for file_count is always a bad idea. Linux has the ->release
method to deal with cleanups on last close and ->flush is only for the
very rare case where we want to perform an operation on every drop of a
reference to a file struct.

This patch gets rid of vop_close and surrounding code in favour of simply
doing the page flushing from ->release.

SGI-PV: 966562
SGI-Modid: xfs-linux-melb:xfs-kern:28952a

Signed-off-by: Christoph Hellwig <hch@infradead.org>
Signed-off-by: David Chinner <dgc@sgi.com>
Signed-off-by: Tim Shimmin <tes@sgi.com>
17 years ago[XFS] Use is_power_of_2 instead of open coding checks
Vignesh Babu [Thu, 28 Jun 2007 06:46:37 +0000 (16:46 +1000)]
[XFS] Use is_power_of_2 instead of open coding checks

SGI-PV: 966576
SGI-Modid: xfs-linux-melb:xfs-kern:28950a

Signed-off-by: Vignesh Babu <vignesh.babu@wipro.com>
Signed-off-by: David Chinner <dgc@sgi.com>
Signed-off-by: Tim Shimmin <tes@sgi.com>
17 years ago[XFS] Reduce shouting by removing unnecessary macros from dir2 code.
Christoph Hellwig [Thu, 28 Jun 2007 06:43:50 +0000 (16:43 +1000)]
[XFS] Reduce shouting by removing unnecessary macros from dir2 code.

SGI-PV: 966505
SGI-Modid: xfs-linux-melb:xfs-kern:28947a

Signed-off-by: Christoph Hellwig <hch@infradead.org>
Signed-off-by: David Chinner <dgc@sgi.com>
Signed-off-by: Tim Shimmin <tes@sgi.com>
17 years ago[XFS] Simplify XFS min/max macros.
David Chinner [Thu, 28 Jun 2007 06:43:39 +0000 (16:43 +1000)]
[XFS] Simplify XFS min/max macros.

SGI-PV: 964547
SGI-Modid: xfs-linux-melb:xfs-kern:28945a

Signed-off-by: David Chinner <dgc@sgi.com>
Signed-off-by: Nathan Scott <nscott@aconex.com>
Signed-off-by: Tim Shimmin <tes@sgi.com>
17 years ago[XFS] Kill off xfs_count_bits
Eric Sandeen [Thu, 28 Jun 2007 06:43:30 +0000 (16:43 +1000)]
[XFS] Kill off xfs_count_bits

xfs_count_bits is only called once, and is then compared to 0. IOW, what
it really wants to know is, is the bitmap empty. This can be done more
simply, certainly.

SGI-PV: 966503
SGI-Modid: xfs-linux-melb:xfs-kern:28944a

Signed-off-by: Eric Sandeen <sandeen@sandeen.net>
Signed-off-by: David Chinner <dgc@sgi.com>
Signed-off-by: Tim Shimmin <tes@sgi.com>
17 years ago[XFS] Cancel transactions on xfs_itruncate_start error.
Jesper Juhl [Thu, 28 Jun 2007 06:43:14 +0000 (16:43 +1000)]
[XFS] Cancel transactions on xfs_itruncate_start error.

SGI-PV: 966502
SGI-Modid: xfs-linux-melb:xfs-kern:28943a

Signed-off-by: Jesper Juhl <jesper.juhl@gmail.com>
Signed-off-by: David Chinner <dgc@sgi.com>
Signed-off-by: Tim Shimmin <tes@sgi.com>
17 years ago[XFS] Use do_div() on 64 bit types.
Christoph Hellwig [Mon, 18 Jun 2007 07:57:45 +0000 (17:57 +1000)]
[XFS] Use do_div() on 64 bit types.

SGI-PV: 966145
SGI-Modid: xfs-linux-melb:xfs-kern:28889a

Signed-off-by: Christoph Hellwig <hch@infradead.org>
Signed-off-by: David Chinner <dgc@sgi.com>
Signed-off-by: Tim Shimmin <tes@sgi.com>
17 years ago[XFS] Fix remount,readonly path to flush everything correctly.
David Chinner [Mon, 18 Jun 2007 06:50:48 +0000 (16:50 +1000)]
[XFS] Fix remount,readonly path to flush everything correctly.

The remount readonly path can fail to writeback properly because we still
have active transactions after calling xfs_quiesce_fs(). Further
investigation shows that this path is broken in the same ways that the xfs
freeze path was broken so fix it the same way.

SGI-PV: 964464
SGI-Modid: xfs-linux-melb:xfs-kern:28869a

Signed-off-by: David Chinner <dgc@sgi.com>
Signed-off-by: Christoph Hellwig <hch@infradead.org>
Signed-off-by: Tim Shimmin <tes@sgi.com>
17 years ago[XFS] Cleanup inode extent size hint extraction
David Chinner [Mon, 18 Jun 2007 06:50:37 +0000 (16:50 +1000)]
[XFS] Cleanup inode extent size hint extraction

SGI-PV: 966004
SGI-Modid: xfs-linux-melb:xfs-kern:28866a

Signed-off-by: David Chinner <dgc@sgi.com>
Signed-off-by: Christoph Hellwig <hch@infradead.org>
Signed-off-by: Tim Shimmin <tes@sgi.com>
17 years ago[XFS] Prevent ENOSPC from aborting transactions that need to succeed
David Chinner [Mon, 18 Jun 2007 06:50:27 +0000 (16:50 +1000)]
[XFS] Prevent ENOSPC from aborting transactions that need to succeed

During delayed allocation extent conversion or unwritten extent
conversion, we need to reserve some blocks for transactions reservations.
We need to reserve these blocks in case a btree split occurs and we need
to allocate some blocks.

Unfortunately, we've only ever reserved the number of data blocks we are
allocating, so in both the unwritten and delalloc case we can get ENOSPC
to the transaction reservation. This is bad because in both cases we
cannot report the failure to the writing application.

The fix is two-fold:

1 - leverage the reserved block infrastructure XFS already
has to reserve a small pool of blocks by default to allow
specially marked transactions to dip into when we are at
ENOSPC.
Default setting is min(5%, 1024 blocks).

2 - convert critical transaction reservations to be allowed
to dip into this pool. Spots changed are delalloc
conversion, unwritten extent conversion and growing a
filesystem at ENOSPC.
This also allows growing the filesytsem to succeed at ENOSPC.

SGI-PV: 964468
SGI-Modid: xfs-linux-melb:xfs-kern:28865a

Signed-off-by: David Chinner <dgc@sgi.com>
Signed-off-by: Tim Shimmin <tes@sgi.com>
17 years ago[XFS] Prevent deadlock when flushing inodes on unmount
David Chinner [Mon, 18 Jun 2007 06:50:17 +0000 (16:50 +1000)]
[XFS] Prevent deadlock when flushing inodes on unmount

When we are unmounting the filesystem, we flush all the inodes to disk.
Unfortunately, if we have an inode cluster that has just been freed and
marked stale sitting in an incore log buffer (i.e. hasn't been flushed to
disk), it will be holding all the flush locks on the inodes in that
cluster.

xfs_iflush_all() which is called during unmount walks all the inodes
trying to reclaim them, and it doing so calls xfs_finish_reclaim() on each
inode. If the inode is dirty, if grabs the flush lock and flushes it.
Unfortunately, find dirty inodes that already have their flush lock held
and so we sleep.

At this point in the unmount process, we are running single-threaded.
There is nothing more that can push on the log to force the transaction
holding the inode flush locks to disk and hence we deadlock.

The fix is to issue a log force before flushing the inodes on unmount so
that all the flush locks will be released before we start flushing the
inodes.

SGI-PV: 964538
SGI-Modid: xfs-linux-melb:xfs-kern:28862a

Signed-off-by: David Chinner <dgc@sgi.com>
Signed-off-by: Tim Shimmin <tes@sgi.com>
17 years ago[XFS] Log the agf_length change in xfs_growfs_data_private().
Tim Shimmin [Mon, 18 Jun 2007 06:50:08 +0000 (16:50 +1000)]
[XFS] Log the agf_length change in xfs_growfs_data_private().

SGI-PV: 963528
SGI-Modid: xfs-linux-melb:xfs-kern:28856a

Signed-off-by: Tim Shimmin <tes@sgi.com>
Signed-off-by: David Chinner <dgc@sgi.com>
Signed-off-by: Christoph Hellwig <hch@infradead.org>
17 years ago[XFS] Map unwritten extents correctly for I/o completion processing
David Chinner [Mon, 18 Jun 2007 06:49:58 +0000 (16:49 +1000)]
[XFS] Map unwritten extents correctly for I/o completion processing

If we have multiple unwritten extents within a single page, we fail to
tell the I/o completion construction handlers we need a new handle for the
second and subsequent blocks in the page. While we still issue the I/O
correctly, we do not have the correct ranges recorded in the ioend
structures and hence when we go to convert the unwritten extents we screw
it up.

Make sure we start a new ioend every time the mapping changes so that we
convert the correct ranges on I/O completion.

SGI-PV: 964647
SGI-Modid: xfs-linux-melb:xfs-kern:28797a

Signed-off-by: David Chinner <dgc@sgi.com>
Signed-off-by: Christoph Hellwig <hch@infradead.org>
Signed-off-by: Tim Shimmin <tes@sgi.com>
17 years ago[XFS] Apply transaction delta counts atomically to incore counters
David Chinner [Mon, 18 Jun 2007 06:49:44 +0000 (16:49 +1000)]
[XFS] Apply transaction delta counts atomically to incore counters

With the per-cpu superblock counters, batch updates are no longer atomic
across the entire batch of changes. This is not an issue if each
individual change in the batch is applied atomically. Unfortunately, free
block count changes are not applied atomically, and they are applied in a
manner guaranteed to cause problems.

Essentially, the free block count reservation that the transaction took
initially is returned to the in core counters before a second delta takes
away what is used. because these two operations are not atomic, we can
race with another thread that can use the returned transaction reservation
before the transaction takes the space away again and we can then get
ENOSPC being reported in a spot where we don't have an ENOSPC condition,
nor should we ever see one there.

Fix it up by rolling the two deltas into the one so it can be applied
safely (i.e. atomically) to the incore counters.

SGI-PV: 964465
SGI-Modid: xfs-linux-melb:xfs-kern:28796a

Signed-off-by: David Chinner <dgc@sgi.com>
Signed-off-by: Christoph Hellwig <hch@infradead.org>
Signed-off-by: Tim Shimmin <tes@sgi.com>
17 years ago[XFS] Handle null returned from xfs_vtoi() in xfs_setfilesize().
David Chinner [Tue, 5 Jun 2007 06:24:44 +0000 (16:24 +1000)]
[XFS] Handle null returned from xfs_vtoi() in xfs_setfilesize().

SGI-PV: 965636
SGI-Modid: xfs-linux-melb:xfs-kern:28777a

Signed-off-by: David Chinner <dgc@sgi.com>
Signed-off-by: Olaf Weber <olaf@sgi.com>
Signed-off-by: Tim Shimmin <tes@sgi.com>
17 years ago[XFS] Block on unwritten extent conversion during synchronous direct I/O.
David Chinner [Tue, 5 Jun 2007 06:24:36 +0000 (16:24 +1000)]
[XFS] Block on unwritten extent conversion during synchronous direct I/O.

Currently we do not wait on extent conversion to occur, and hence we can
return to userspace from a synchronous direct I/O write without having
completed all the actions in the write. Hence a read after the write may
see zeroes (unwritten extent) rather than the data that was written.

Block the I/O completion by triggering a synchronous workqueue flush to
ensure that the conversion has occurred before we return to userspace.

SGI-PV: 964092
SGI-Modid: xfs-linux-melb:xfs-kern:28775a

Signed-off-by: David Chinner <dgc@sgi.com>
Signed-off-by: Tim Shimmin <tes@sgi.com>
17 years ago[XFS] Flush the block device before closing it on unmount.
David Chinner [Tue, 5 Jun 2007 06:24:27 +0000 (16:24 +1000)]
[XFS] Flush the block device before closing it on unmount.

SGI-PV: 965630
SGI-Modid: xfs-linux-melb:xfs-kern:28774a

Signed-off-by: David Chinner <dgc@sgi.com>
Signed-off-by: Christoph Hellwig <hch@infradead.org>
Signed-off-by: Tim Shimmin <tes@sgi.com>
17 years ago[XFS] xfs_bmapi fails to update the previous extent pointer
David Chinner [Tue, 5 Jun 2007 06:24:15 +0000 (16:24 +1000)]
[XFS] xfs_bmapi fails to update the previous extent pointer

When processing multiple extent maps, xfs_bmapi needs to keep track of the
extent behind the one it is currently working on to be able to trim extent
ranges correctly. Failing to update the previous pointer can result in
corrupted extent lists in memory and this will result in panics or assert
failures.

Update the previous pointer correctly when we move to the next extent to
process.

SGI-PV: 965631
SGI-Modid: xfs-linux-melb:xfs-kern:28773a

Signed-off-by: David Chinner <dgc@sgi.com>
Signed-off-by: Vlad Apostolov <vapo@sgi.com>
Signed-off-by: Tim Shimmin <tes@sgi.com>
17 years ago[XFS] Fix the transaction flags to make lazy superblock counters work.
David Chinner [Thu, 24 May 2007 05:26:51 +0000 (15:26 +1000)]
[XFS] Fix the transaction flags to make lazy superblock counters work.

SGI-PV: 964999
SGI-Modid: xfs-linux-melb:xfs-kern:28653a

Signed-off-by: David Chinner <dgc@sgi.com>
Signed-off-by: Christoph Hellwig <hch@infradead.org>
Signed-off-by: Tim Shimmin <tes@sgi.com>
17 years ago[XFS] Lazy Superblock Counters
David Chinner [Thu, 24 May 2007 05:26:31 +0000 (15:26 +1000)]
[XFS] Lazy Superblock Counters

When we have a couple of hundred transactions on the fly at once, they all
typically modify the on disk superblock in some way.
create/unclink/mkdir/rmdir modify inode counts, allocation/freeing modify
free block counts.

When these counts are modified in a transaction, they must eventually lock
the superblock buffer and apply the mods. The buffer then remains locked
until the transaction is committed into the incore log buffer. The result
of this is that with enough transactions on the fly the incore superblock
buffer becomes a bottleneck.

The result of contention on the incore superblock buffer is that
transaction rates fall - the more pressure that is put on the superblock
buffer, the slower things go.

The key to removing the contention is to not require the superblock fields
in question to be locked. We do that by not marking the superblock dirty
in the transaction. IOWs, we modify the incore superblock but do not
modify the cached superblock buffer. In short, we do not log superblock
modifications to critical fields in the superblock on every transaction.
In fact we only do it just before we write the superblock to disk every
sync period or just before unmount.

This creates an interesting problem - if we don't log or write out the
fields in every transaction, then how do the values get recovered after a
crash? the answer is simple - we keep enough duplicate, logged information
in other structures that we can reconstruct the correct count after log
recovery has been performed.

It is the AGF and AGI structures that contain the duplicate information;
after recovery, we walk every AGI and AGF and sum their individual
counters to get the correct value, and we do a transaction into the log to
correct them. An optimisation of this is that if we have a clean unmount
record, we know the value in the superblock is correct, so we can avoid
the summation walk under normal conditions and so mount/recovery times do
not change under normal operation.

One wrinkle that was discovered during development was that the blocks
used in the freespace btrees are never accounted for in the AGF counters.
This was once a valid optimisation to make; when the filesystem is full,
the free space btrees are empty and consume no space. Hence when it
matters, the "accounting" is correct. But that means the when we do the
AGF summations, we would not have a correct count and xfs_check would
complain. Hence a new counter was added to track the number of blocks used
by the free space btrees. This is an *on-disk format change*.

As a result of this, lazy superblock counters are a mkfs option and at the
moment on linux there is no way to convert an old filesystem. This is
possible - xfs_db can be used to twiddle the right bits and then
xfs_repair will do the format conversion for you. Similarly, you can
convert backwards as well. At some point we'll add functionality to
xfs_admin to do the bit twiddling easily....

SGI-PV: 964999
SGI-Modid: xfs-linux-melb:xfs-kern:28652a

Signed-off-by: David Chinner <dgc@sgi.com>
Signed-off-by: Christoph Hellwig <hch@infradead.org>
Signed-off-by: Tim Shimmin <tes@sgi.com>
17 years ago[XFS] Use generic shrinker interfaces in XFS.
Andrew Morton [Thu, 24 May 2007 05:25:42 +0000 (15:25 +1000)]
[XFS] Use generic shrinker interfaces in XFS.

SGI-PV: 964986
SGI-Modid: xfs-linux-melb:xfs-kern:28642a

Signed-Off-By: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: David Chinner <dgc@sgi.com>
Signed-off-by: Tim Shimmin <tes@sgi.com>
17 years ago[XFS] Make hole punching at EOF atomic.
David Chinner [Thu, 24 May 2007 05:22:19 +0000 (15:22 +1000)]
[XFS] Make hole punching at EOF atomic.

If hole punching at EOF is done as two steps (i.e. truncate then extend)
the file is in a transient state between the two steps where an
application can see the incorrect file size. Punching a hole to EOF needs
to be treated in teh same way as all other hole punching cases so that the
file size is never seen to change.

SGI-PV: 962012
SGI-Modid: xfs-linux-melb:xfs-kern:28641a

Signed-off-by: David Chinner <dgc@sgi.com>
Signed-off-by: Vlad Apostolov <vapo@sgi.com>
Signed-off-by: Tim Shimmin <tes@sgi.com>
17 years ago[XFS] Fix vmalloc leak on mount/unmount.
David Chinner [Thu, 24 May 2007 05:21:57 +0000 (15:21 +1000)]
[XFS] Fix vmalloc leak on mount/unmount.

When setting the length of the iclogbuf to write out we should just be
changing the desired byte count rather completely reassociating the buffer
memory with the buffer. Reassociating the buffer memory changes the
apparent length of the buffer and hence when we free the buffer, we don't
free all the vmap()d space we originally allocated.

SGI-PV: 964983
SGI-Modid: xfs-linux-melb:xfs-kern:28640a

Signed-off-by: David Chinner <dgc@sgi.com>
Signed-off-by: Christoph Hellwig <hch@infradead.org>
Signed-off-by: Tim Shimmin <tes@sgi.com>
17 years ago[XFS] Fix double free in xfs_buf_get_noaddr error handling path
Christoph Hellwig [Thu, 24 May 2007 05:21:11 +0000 (15:21 +1000)]
[XFS] Fix double free in xfs_buf_get_noaddr error handling path

SGI-PV: 964983
SGI-Modid: xfs-linux-melb:xfs-kern:28639a

Signed-off-by: Christoph Hellwig <hch@lst.de>
Signed-off-by: David Chinner <dgc@sgi.com>
Signed-off-by: Tim Shimmin <tes@sgi.com>
17 years ago[XFS] Fix use-after-free during log unmount.
David Chinner [Mon, 14 May 2007 08:24:16 +0000 (18:24 +1000)]
[XFS] Fix use-after-free during log unmount.

Don't reference the log buffer after running the callbacks as the callback
can trigger the log buffers to be freed during unmount.

SGI-PV: 964545
SGI-Modid: xfs-linux-melb:xfs-kern:28567a

Signed-off-by: David Chinner <dgc@sgi.com>
Signed-off-by: Christoph Hellwig <hch@infradead.org>
Signed-off-by: Tim Shimmin <tes@sgi.com>
17 years ago[XFS] Sleeping with the ilock waiting for I/O completion is Bad.
David Chinner [Mon, 14 May 2007 08:24:09 +0000 (18:24 +1000)]
[XFS] Sleeping with the ilock waiting for I/O completion is Bad.

Recent fixes to the filesystem freezing code introduced a vn_iowait call
in the middle of the sync code. Unfortunately, at the point where this
call was added we are holding the ilock. The ilock is needed by I/O
completion for unwritten extent conversion and now updating the file size.
Hence I/o cannot complete if we hold the ilock while waiting for I/O
completion.

Fix up the bug and clean the code up around it.

SGI-PV: 963674
SGI-Modid: xfs-linux-melb:xfs-kern:28566a

Signed-off-by: David Chinner <dgc@sgi.com>
Signed-off-by: Christoph Hellwig <hch@infradead.org>
Signed-off-by: Tim Shimmin <tes@sgi.com>
17 years ago[XFS] Don't grow filesystems past the size they can index.
Nathan Scott [Mon, 14 May 2007 08:24:02 +0000 (18:24 +1000)]
[XFS] Don't grow filesystems past the size they can index.

When growing a filesystem we don't check to see if the new size overflows
the page cache index range, so we can do silly things like grow a
filesystem page 16TB on a 32bit. Check new filesystem sizes against the
limits the kernel can support.

SGI-PV: 957886
SGI-Modid: xfs-linux-melb:xfs-kern:28563a

Signed-Off-By: Nathan Scott <nscott@aconex.com>
Signed-off-by: David Chinner <dgc@sgi.com>
Signed-off-by: Tim Shimmin <tes@sgi.com>
17 years ago[XFS] Only use refcounted pages for I/O
Christoph Hellwig [Mon, 14 May 2007 08:23:50 +0000 (18:23 +1000)]
[XFS] Only use refcounted pages for I/O

Many block drivers (aoe, iscsi) really want refcountable pages in bios,
which is what almost everyone send down. XFS unfortunately has a few
places where it sends down buffers that may come from kmalloc, which
breaks them.

Fix the places that use kmalloc()d buffers.

SGI-PV: 964546
SGI-Modid: xfs-linux-melb:xfs-kern:28562a

Signed-Off-By: Christoph Hellwig <hch@infradead.org>
Signed-off-by: David Chinner <dgc@sgi.com>
Signed-off-by: Tim Shimmin <tes@sgi.com>
17 years agolots-of-architectures: enable arbitary speed tty support
Alan Cox [Wed, 11 Jul 2007 00:22:27 +0000 (17:22 -0700)]
lots-of-architectures: enable arbitary speed tty support

Add the termios2 structure ready for enabling on most platforms.  One or
two like Sparc are plain weird so have been left alone.  Most can use the
same structure as ktermios for termios2 (ie the newer ioctl uses the
structure matching the current kernel structure)

Signed-off-by: Alan Cox <alan@redhat.com>
Cc: Bryan Wu <bryan.wu@analog.com>
Cc: Ian Molton <spyro@f2s.com>
Cc: Haavard Skinnemoen <hskinnemoen@atmel.com>
Cc: Yoshinori Sato <ysato@users.sourceforge.jp>
Cc: Mikael Starvik <starvik@axis.com>
Cc: David Howells <dhowells@redhat.com>
Cc: "Luck, Tony" <tony.luck@intel.com>
Cc: Hirokazu Takata <takata@linux-m32r.org>
Cc: Geert Uytterhoeven <geert@linux-m68k.org>
Cc: Roman Zippel <zippel@linux-m68k.org>
Cc: Ralf Baechle <ralf@linux-mips.org>
Cc: Kyle McMartin <kyle@mcmartin.ca>
Cc: Matthew Wilcox <willy@debian.org>
Cc: Heiko Carstens <heiko.carstens@de.ibm.com>
Cc: Martin Schwidefsky <schwidefsky@de.ibm.com>
Cc: Paul Mundt <lethal@linux-sh.org>
Cc: Kazumoto Kojima <kkojima@rr.iij4u.or.jp>
Cc: Richard Curnow <rc@rc0.org.uk>
Cc: Miles Bader <uclinux-v850@lsi.nec.co.jp>
Cc: Chris Zankel <chris@zankel.net>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
17 years agoMake common helpers for seq_files that work with list_heads
Pavel Emelianov [Wed, 11 Jul 2007 00:22:26 +0000 (17:22 -0700)]
Make common helpers for seq_files that work with list_heads

Many places in kernel use seq_file API to iterate over a regular list_head.
The code for such iteration is identical in all the places, so it's worth
introducing a common helpers.

This makes code about 300 lines smaller:

The first version of this patch made the helper functions static inline
in the seq_file.h header. This patch moves them to the fs/seq_file.c as
Andrew proposed. The vmlinux .text section sizes are as follows:

2.6.22-rc1-mm1:              0x001794d5
with the previous version:   0x00179505
with this patch:             0x00179135

The config file used was make allnoconfig with the "y" inclusion of all
the possible options to make the files modified by the patch compile plus
drivers I have on the test node.

This patch:

Many places in kernel use seq_file API to iterate over a regular list_head.
The code for such iteration is identical in all the places, so it's worth
introducing a common helpers.

Signed-off-by: Pavel Emelianov <xemul@openvz.org>
Cc: "David S. Miller" <davem@davemloft.net>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
17 years agosx: switch subven and subid values
Jiri Slaby [Wed, 11 Jul 2007 00:22:25 +0000 (17:22 -0700)]
sx: switch subven and subid values

sx.c is failing to locate Graham's card.

Signed-off-by: Jiri Slaby <jirislaby@gmail.com>
Cc: Graham Murray <gmurray@webwayone.co.uk>
Cc: <stable@kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
17 years agoAdd LZO1X algorithm to the kernel
Richard Purdie [Wed, 11 Jul 2007 00:22:24 +0000 (17:22 -0700)]
Add LZO1X algorithm to the kernel

This is a hybrid version of the patch to add the LZO1X compression
algorithm to the kernel.  Nitin and myself have merged the best parts of
the various patches to form this version which we're both happy with (and
are jointly signing off).

The performance of this version is equivalent to the original minilzo code
it was based on.  Bytecode comparisons have also been made on ARM, i386 and
x86_64 with favourable results.

There are several users of LZO lined up including jffs2, crypto and reiser4
since its much faster than zlib.

Signed-off-by: Nitin Gupta <nitingupta910@gmail.com>
Signed-off-by: Richard Purdie <rpurdie@openedhand.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
17 years agoMerge branch 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/drzeus/mmc
Linus Torvalds [Tue, 10 Jul 2007 21:57:52 +0000 (14:57 -0700)]
Merge branch 'for-linus' of git://git./linux/kernel/git/drzeus/mmc

* 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/drzeus/mmc:
  mmc: at91_mci: fix hanging and rework to match flowcharts
  mmc: at91_mci typo
  sdhci: Fix "Unexpected interrupt" handling
  mmc: fix silly copy-and-paste error
  mmc: move layer init and workqueue to core file
  mmc: refactor host class handling
  mmc: refactor bus operations
  sdhci: add ene controller id
  mmc: bounce requests for simple hosts

17 years agoMerge branch 'upstream-linus' of master.kernel.org:/pub/scm/linux/kernel/git/jgarzik...
Linus Torvalds [Tue, 10 Jul 2007 21:56:22 +0000 (14:56 -0700)]
Merge branch 'upstream-linus' of /linux/kernel/git/jgarzik/netdev-2.6

* 'upstream-linus' of master.kernel.org:/pub/scm/linux/kernel/git/jgarzik/netdev-2.6: (40 commits)
  bonding/bond_main.c: make 2 functions static
  ps3: gigabit ethernet driver for PS3, take3
  [netdrvr] Fix dependencies for ax88796 ne2k clone driver
  eHEA: Capability flag for DLPAR support
  Remove sk98lin ethernet driver.
  sunhme.c:quattro_pci_find() must be __devinit
  bonding / ipv6: no addrconf for slaves separately from master
  atl1: remove write-only var in tx handler
  macmace: use "unsigned long flags;"
  Cleanup usbnet_probe() return value handling
  netxen: deinline and sparse fix
  eeprom_93cx6: shorten pulse timing to match spec (bis)
  phylib: Add Marvell 88E1112 phy id
  phylib: cleanup marvell.c a bit
  AX88796 network driver
  IOC3: Switch to pci refcounting safe APIs
  e100: Fix Tyan motherboard e100 not receiving IPMI commands
  QE Ethernet driver writes to wrong register to mask interrupts
  rrunner.c:rr_init() must be __devinit
  tokenring/3c359.c:xl_init() must be __devinit
  ...

17 years agoMerge branch 'upstream-linus' of master.kernel.org:/pub/scm/linux/kernel/git/jgarzik...
Linus Torvalds [Tue, 10 Jul 2007 21:50:16 +0000 (14:50 -0700)]
Merge branch 'upstream-linus' of /linux/kernel/git/jgarzik/libata-dev

* 'upstream-linus' of master.kernel.org:/pub/scm/linux/kernel/git/jgarzik/libata-dev: (32 commits)
  [libata] sata_mv: print out additional chip info during probe
  [libata] Use ATA_UDMAx standard masks when filling driver's udma_mask info
  [libata] AHCI: Add support for Marvell AHCI-like chips (initially 6145)
  [libata] Clean up driver udma_mask initializers
  libata: Support chips with 64K PRD quirk
  Add a PCI ID for santa rosa's PATA controller.
  sata_sil24: sil24_interrupt() micro-optimisation
  Add irq_flags to struct pata_platform_info
  sata_promise: cleanups
  [libata] pata_ixp4xx: kill unused var
  ata_piix: fix pio/mwdma programming
  [libata] ahci: minor internal cleanups
  [ATA] Add named constant for ATAPI command DEVICE RESET
  [libata] sata_sx4, sata_via: minor documentation updates
  [libata] ahci: minor internal cleanups
  [libata] ahci: Factor out SATA port init into a separate function
  [libata] pata_sil680: minor cleanups from benh
  [libata] sata_sx4: named constant cleanup
  [libata] pata_ixp4xx: convert to new EH
  [libata] pdc_adma: Reorder initializers with a couple structs
  ...

17 years agoMerge branch 'upstream' of git://ftp.linux-mips.org/pub/scm/upstream-linus
Linus Torvalds [Tue, 10 Jul 2007 21:48:43 +0000 (14:48 -0700)]
Merge branch 'upstream' of git://ftp.linux-mips.org/upstream-linus

* 'upstream' of git://ftp.linux-mips.org/pub/scm/upstream-linus: (62 commits)
  [MIPS] PNX8550: Cleanup proc code.
  [MIPS] WRPPMC: Fix build.
  [MIPS] Yosemite: Fix modpost warnings.
  [MIPS] Change names of local variables to silence sparse
  [MIPS] SB1: Fix modpost warning.
  [MIPS] PNX: Fix modpost warnings.
  [MIPS] Alchemy: Fix modpost warnings.
  [MIPS] Non-FPAFF: Fix warning.
  [MIPS] DEC: Fix modpost warning.
  [MIPS] MIPSsim: Enable MIPSsim virtual network driver.
  [MIPS] Delete Ocelot 3 support.
  [MIPS] remove LASAT Networks platforms support
  [MIPS] Early check for SMTC kernel on non-MT processor
  [MIPS] Add debugfs files to show fpuemu statistics
  [MIPS] Add some debugfs files to debug unaligned accesses
  [MIPS] rbtx4938: Fix secondary PCIC and glue internal NICs
  [MIPS] tc35815: Load MAC address via platform_device
  [MIPS] Move FPU affinity code into separate file.
  [MIPS] Make ioremap() work on TX39/49 special unmapped segment
  [MIPS] rbtx4938: Update and minimize defconfig
  ...

17 years agoMerge git://git.infradead.org/~dwmw2/battery-2.6
Linus Torvalds [Tue, 10 Jul 2007 21:47:59 +0000 (14:47 -0700)]
Merge git://git.infradead.org/~dwmw2/battery-2.6

* git://git.infradead.org/~dwmw2/battery-2.6:
  [BATTERY] ds2760 W1 slave
  [BATTERY] One Laptop Per Child power/battery driver
  [BATTERY] Apple PMU driver
  [BATTERY] 1-Wire ds2760 chip battery driver
  [BATTERY] APM emulation driver for class batteries
  [BATTERY] pda_power platform driver
  [BATTERY] Universal power supply class (was: battery class)

17 years agoMerge branch 'for-linus' of git://git390.osdl.marist.edu/pub/scm/linux-2.6
Linus Torvalds [Tue, 10 Jul 2007 21:46:09 +0000 (14:46 -0700)]
Merge branch 'for-linus' of git://git390.osdl.marist.edu/linux-2.6

* 'for-linus' of git://git390.osdl.marist.edu/pub/scm/linux-2.6:
  [S390] vmlogrdr function annotation.
  [S390] s390: rename CPU_IDLE to S390_CPU_IDLE
  [S390] cio: Remove prototype for non-existing function cmf_reset().
  [S390] zcrypt: fix request timeout handling
  [S390] system call optimization.
  [S390] dasd: Avoid compile warnings on !CONFIG_DASD_PROFILE
  [S390] Remove volatile from atomic_t
  [S390] Program check in diag 210 under 31 bit
  [S390] Bogomips calculation for 64 bit.
  [S390] smp: Merge smp_count_cpus() and smp_get_save_areas().
  [S390] zcore: Fix __user annotation.
  [S390] fixed cdl-format detection.
  [S390] sclp: Test facility list before executing a service call.
  [S390] sclp: introduce some new interfaces.
  [S390] Fixed comment typo.
  [S390] vmcp cleanup

17 years agoMerge git://git.kernel.org/pub/scm/linux/kernel/git/steve/gfs2-2.6-nmw
Linus Torvalds [Tue, 10 Jul 2007 20:56:13 +0000 (13:56 -0700)]
Merge git://git./linux/kernel/git/steve/gfs2-2.6-nmw

* git://git.kernel.org/pub/scm/linux/kernel/git/steve/gfs2-2.6-nmw: (57 commits)
  [GFS2] Accept old format NFS filehandles
  [GFS2] Small fixes to logging code
  [DLM] dump more lock values
  [GFS2] Remove i_mode passing from NFS File Handle
  [GFS2] Obtaining no_formal_ino from directory entry
  [GFS2] git-gfs2-nmw-build-fix
  [GFS2] System won't suspend with GFS2 file system mounted
  [GFS2] remounting w/o acl option leaves acls enabled
  [GFS2] inode size inconsistency
  [DLM] Telnet to port 21064 can stop all lockspaces
  [GFS2] Fix gfs2_block_truncate_page err return
  [GFS2] Addendum to the journaled file/unmount patch
  [GFS2] Simplify multiple glock aquisition
  [GFS2] assertion failure after writing to journaled file, umount
  [GFS2] Use zero_user_page() in stuffed_readpage()
  [GFS2] Remove bogus '\0' in rgrp.c
  [GFS2] Journaled file write/unstuff bug
  [DLM] don't require FS flag on all nodes
  [GFS2] Fix deallocation issues
  [GFS2] return conflicts for GETLK
  ...

17 years agoMerge branch 'splice-2.6.23' of git://git.kernel.dk/data/git/linux-2.6-block
Linus Torvalds [Tue, 10 Jul 2007 20:51:06 +0000 (13:51 -0700)]
Merge branch 'splice-2.6.23' of git://git.kernel.dk/data/git/linux-2.6-block

* 'splice-2.6.23' of git://git.kernel.dk/data/git/linux-2.6-block:
  pipe: add documentation and comments
  pipe: change the ->pin() operation to ->confirm()
  Remove remnants of sendfile()
  xip sendfile removal
  splice: completely document external interface with kerneldoc
  sendfile: remove bad_sendfile() from bad_file_ops
  shmem: convert to using splice instead of sendfile()
  relay: use splice_to_pipe() instead of open-coding the pipe loop
  pipe: allow passing around of ops private pointer
  splice: divorce the splice structure/function definitions from the pipe header
  splice: relay support
  sendfile: convert nfsd to splice_direct_to_actor()
  sendfile: convert nfs to using splice_read()
  loop: convert to using splice_direct_to_actor() instead of sendfile()
  splice: add void cookie to the actor data
  sendfile: kill generic_file_sendfile()
  sendfile: remove .sendfile from filesystems that use generic_file_sendfile()
  sys_sendfile: switch to using ->splice_read, if available
  vmsplice: add vmsplice-to-user support
  splice: abstract out actor data

17 years agoMerge branch 'trivial-2.6.23' of git://git.kernel.dk/data/git/linux-2.6-block
Linus Torvalds [Tue, 10 Jul 2007 20:49:46 +0000 (13:49 -0700)]
Merge branch 'trivial-2.6.23' of git://git.kernel.dk/data/git/linux-2.6-block

* 'trivial-2.6.23' of git://git.kernel.dk/data/git/linux-2.6-block:
  Documentation/block/barrier.txt is not in sync with the actual code: - blk_queue_ordered() no longer has a gfp_mask parameter - blk_queue_ordered_locked() no longer exists - sd_prepare_flush() looks slightly different
  Use list_for_each_entry() instead of list_for_each() in the block device
  Make a "menuconfig" out of the Kconfig objects "menu, ..., endmenu",
  block/Kconfig already has its own "menuconfig" so remove these
  Use menuconfigs instead of menus, so the whole menu can be disabled at once
  cfq-iosched: fix async queue behaviour
  unexport bio_{,un}map_user
  Remove legacy CDROM drivers
  [PATCH] fix request->cmd == INT cases
  cciss: add new controller support for P700m
  [PATCH] Remove acsi.c
  [BLOCK] drop unnecessary bvec rewinding from flush_dry_bio_endio
  [PATCH] cdrom_sysctl_info fix
  blk_hw_contig_segment(): bad segment size checks
  [TRIVIAL PATCH] Kill blk_congestion_wait() stub for !CONFIG_BLOCK

17 years agobonding/bond_main.c: make 2 functions static
Adrian Bunk [Mon, 9 Jul 2007 18:51:12 +0000 (11:51 -0700)]
bonding/bond_main.c: make 2 functions static

Signed-off-by: Adrian Bunk <bunk@stusta.de>
Cc: Chad Tindel <ctindel@users.sourceforge.net>
Cc: Jay Vosburgh <fubar@us.ibm.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Jeff Garzik <jeff@garzik.org>
17 years agops3: gigabit ethernet driver for PS3, take3
Masakazu Mokuno [Thu, 5 Jul 2007 11:11:16 +0000 (20:11 +0900)]
ps3: gigabit ethernet driver for PS3, take3

Hi,

This is the third submission of the network driver for PS3.
The differences from the previous one are:

  - renamed source file names so that their prefix can match
    with the module name
  - added cbe-oss-dev@ozlabs.org line for MAINTAINER file
  - changed some in copyright comments

If there are no more comments, please apply for 2.6.23.

Thank you

--
Subject: PS3: Ethernet driver

From: Masakazu Mokuno <mokuno@sm.sony.co.jp>

Add Gigabit Ethernet support for the PS3 game console.  The module will
be called ps3_gelic.

CC: Geoff Levand <geoffrey.levand@am.sony.com>
Signed-off-by: Masakazu Mokuno <mokuno@sm.sony.co.jp>
Signed-off-by: Jeff Garzik <jeff@garzik.org>
17 years ago[netdrvr] Fix dependencies for ax88796 ne2k clone driver
Jeff Garzik [Tue, 10 Jul 2007 18:06:48 +0000 (14:06 -0400)]
[netdrvr] Fix dependencies for ax88796 ne2k clone driver

It needs writesb(), not available on all platforms.

Signed-off-by: Jeff Garzik <jeff@garzik.org>
17 years agoeHEA: Capability flag for DLPAR support
Jan-Bernd Themann [Thu, 5 Jul 2007 07:26:25 +0000 (09:26 +0200)]
eHEA: Capability flag for DLPAR support

This patch introduces a capability flag that is used by the DLPAR userspace
tool to check which DLPAR features are supported by the eHEA driver.

Missing goto has been included.

Signed-off-by: Jan-Bernd Themann <themann@de.ibm.com>
Signed-off-by: Jeff Garzik <jeff@garzik.org>
17 years agoRemove sk98lin ethernet driver.
Jeff Garzik [Tue, 10 Jul 2007 16:58:33 +0000 (12:58 -0400)]
Remove sk98lin ethernet driver.

Unmaintained, superceded by skge.

Prodded to deletion by Adrian Bunk.  Acked by Stephen Hemminger.

Signed-off-by: Jeff Garzik <jeff@garzik.org>
17 years agosunhme.c:quattro_pci_find() must be __devinit
Adrian Bunk [Tue, 10 Jul 2007 12:44:49 +0000 (14:44 +0200)]
sunhme.c:quattro_pci_find() must be __devinit

This patch fixes the following section mismatch:

<--  snip  -->

...
  MODPOST vmlinux
WARNING: drivers/built-in.o(.text+0x272f8b): Section mismatch: reference to .init.text:quattro_pci_find (between 'happy_meal_pci_probe' and 'happy_meal_pci_remove')
...

<--  snip  -->

Signed-off-by: Adrian Bunk <bunk@stusta.de>
Signed-off-by: Jeff Garzik <jeff@garzik.org>
17 years agobonding / ipv6: no addrconf for slaves separately from master
Jay Vosburgh [Mon, 9 Jul 2007 17:42:47 +0000 (10:42 -0700)]
bonding / ipv6: no addrconf for slaves separately from master

At present, when a device is enslaved to bonding, if ipv6 is
active then addrconf will be initated on the slave (because it is closed
then opened during the enslavement processing).  This causes DAD and RS
packets to be sent from the slave.  These packets in turn can confuse
switches that perform ipv6 snooping, causing them to incorrectly update
their forwarding tables (if, e.g., the slave being added is an inactve
backup that won't be used right away) and direct traffic away from the
active slave to a backup slave (where the incoming packets will be
dropped).

This patch alters the behavior so that addrconf will only run on
the master device itself.  I believe this is logically correct, as it
prevents slaves from having an IPv6 identity independent from the
master.  This is consistent with the IPv4 behavior for bonding.

This is accomplished by (a) having bonding set IFF_SLAVE sooner
in the enslavement processing than currently occurs (before open, not
after), and (b) having ipv6 addrconf ignore UP and CHANGE events on
slave devices.

The eql driver also uses the IFF_SLAVE flag.  I inspected eql,
and I believe this change is reasonable for its usage of IFF_SLAVE, but
I did not test it.

Signed-off-by: Jay Vosburgh <fubar@us.ibm.com>
Signed-off-by: Jeff Garzik <jeff@garzik.org>
17 years agoatl1: remove write-only var in tx handler
Alexey Dobriyan [Fri, 6 Jul 2007 15:05:25 +0000 (19:05 +0400)]
atl1: remove write-only var in tx handler

Signed-off-by: Alexey Dobriyan <adobriyan@sw.ru>
Signed-off-by: Jeff Garzik <jeff@garzik.org>
17 years agomacmace: use "unsigned long flags;"
Alexey Dobriyan [Fri, 6 Jul 2007 14:57:13 +0000 (18:57 +0400)]
macmace: use "unsigned long flags;"

Code will do local_irq_save() on it.

Signed-off-by: Alexey Dobriyan <adobriyan@sw.ru>
Signed-off-by: Jeff Garzik <jeff@garzik.org>
17 years agoCleanup usbnet_probe() return value handling
Peter Korsgaard [Mon, 2 Jul 2007 22:46:42 +0000 (00:46 +0200)]
Cleanup usbnet_probe() return value handling

usbnet_probe() handles a positive return value from the driver bind()
function as success, but will later only setup the status handler if the
return value was zero, leading to confusion. Patch adjusts this to accept
positive values as success in both checks.

Signed-off-by: Peter Korsgaard <jacmet@sunsite.dk>
Signed-off-by: Jeff Garzik <jeff@garzik.org>
17 years agonetxen: deinline and sparse fix
Stephen Hemminger [Thu, 5 Jul 2007 23:13:55 +0000 (16:13 -0700)]
netxen: deinline and sparse fix

Get rid of dubious casts to (void *) which causes a sparse warning.
And move largeish function from inline to the one file that uses the code,
the compiler can then decide to inline it.

Compile tested only.

Signed-off-by: Stephen Hemminger <shemminger@linux-foundation.org>
Signed-off-by: Jeff Garzik <jeff@garzik.org>
17 years agoeeprom_93cx6: shorten pulse timing to match spec (bis)
Francois Romieu [Tue, 3 Jul 2007 22:31:44 +0000 (00:31 +0200)]
eeprom_93cx6: shorten pulse timing to match spec (bis)

Based on an original idea by John W. Linville.

It is the missing part of 42d45ccd60636c28e35c2016f091783bc14ad99c

Signed-off-by: Francois Romieu <romieu@fr.zoreil.com>
Signed-off-by: Jeff Garzik <jeff@garzik.org>
17 years agophylib: Add Marvell 88E1112 phy id
Olof Johansson [Tue, 3 Jul 2007 21:24:32 +0000 (16:24 -0500)]
phylib: Add Marvell 88E1112 phy id

Add 88E1112 PHY ID to the marvell driver. Seems to do fine with the
88E1111 inits.

Signed-off-by: Olof Johansson <olof@lixom.net>
Signed-off-by: Jeff Garzik <jeff@garzik.org>
17 years agophylib: cleanup marvell.c a bit
Olof Johansson [Tue, 3 Jul 2007 21:23:46 +0000 (16:23 -0500)]
phylib: cleanup marvell.c a bit

Simplify the marvell driver init a bit: Make the supported devices an
array instead of explicitly registering each structure. This makes it
considerably easier to add new devices down the road.

Signed-off-by: Olof Johansson <olof@lixom.net>
Signed-off-by: Jeff Garzik <jeff@garzik.org>
17 years agoAX88796 network driver
Ben Dooks [Tue, 3 Jul 2007 15:53:09 +0000 (16:53 +0100)]
AX88796 network driver

Support for the Asix AX88796 network controller, an
NE2000 compatible 10/100 ethernet device with internal
PHY.

The driver supports PHY settings via either ioctl() or
the ethtool driver ops.

Signed-off-by: Ben Dooks <ben-linux@fluff.org>
Signed-off-by: Jeff Garzik <jeff@garzik.org>
17 years ago[MIPS] PNX8550: Cleanup proc code.
Robert P. J. Day [Tue, 10 Jul 2007 10:37:56 +0000 (06:37 -0400)]
[MIPS] PNX8550: Cleanup proc code.

Here's a slightly cleaner way of creating the /proc structure for the
pnx8850.  mostly, it creates a directory with default mode 555, since the
one you're creating is mode 444, which is somewhat unusual for a directory
under /proc.

Signed-off-by: Ralf Baechle <ralf@linux-mips.org>
17 years ago[MIPS] WRPPMC: Fix build.
Ralf Baechle [Tue, 10 Jul 2007 15:56:33 +0000 (16:56 +0100)]
[MIPS] WRPPMC: Fix build.

Using another systems defines is a safe way to get your code broken by
accident when that system is removed.

Signed-off-by: Ralf Baechle <ralf@linux-mips.org>
17 years ago[MIPS] Yosemite: Fix modpost warnings.
Ralf Baechle [Tue, 10 Jul 2007 15:31:19 +0000 (16:31 +0100)]
[MIPS] Yosemite: Fix modpost warnings.

  MODPOST vmlinux
WARNING: arch/mips/pmc-sierra/yosemite/built-in.o(.text+0x974): Section mismatch: reference to .init.data: (between 'prom_boot_secondary' and 'prom_cpus_done')
WARNING: arch/mips/pmc-sierra/yosemite/built-in.o(.text+0x980): Section mismatch: reference to .init.data: (between 'prom_boot_secondary' and 'prom_cpus_done')
WARNING: arch/mips/pmc-sierra/yosemite/built-in.o(.text+0x978): Section mismatch: reference to .init.data: (between 'prom_boot_secondary' and 'prom_cpus_done')
WARNING: arch/mips/pmc-sierra/yosemite/built-in.o(.text+0x984): Section mismatch: reference to .init.data: (between 'prom_boot_secondary' and 'prom_cpus_done')
WARNING: arch/mips/pmc-sierra/yosemite/built-in.o(.text+0x98c): Section mismatch: reference to .init.data: (between 'prom_boot_secondary' and 'prom_cpus_done')
WARNING: arch/mips/pmc-sierra/yosemite/built-in.o(.text+0x990): Section mismatch: reference to .init.data: (between 'prom_boot_secondary' and 'prom_cpus_done')

Signed-off-by: Ralf Baechle <ralf@linux-mips.org>
17 years ago[MIPS] Change names of local variables to silence sparse
Atsushi Nemoto [Tue, 10 Jul 2007 15:24:14 +0000 (00:24 +0900)]
[MIPS] Change names of local variables to silence sparse

This patch is an workaround for these sparse warnings:

linux/include/linux/calc64.h:25:17: warning: symbol '__quot' shadows an earlier one
linux/include/linux/calc64.h:25:17: originally declared here
linux/include/linux/calc64.h:25:17: warning: symbol '__mod' shadows an earlier one
linux/include/linux/calc64.h:25:17: originally declared here

Signed-off-by: Atsushi Nemoto <anemo@mba.ocn.ne.jp>
Signed-off-by: Ralf Baechle <ralf@linux-mips.org>
17 years ago[MIPS] SB1: Fix modpost warning.
Ralf Baechle [Tue, 10 Jul 2007 15:10:56 +0000 (16:10 +0100)]
[MIPS] SB1: Fix modpost warning.

  MODPOST vmlinux
WARNING: arch/mips/mm/built-in.o(.text+0x1978): Section mismatch: reference to .init.text:except_vec2_sb1 (between 'sb1_cache_init' and 'sb1_flush_cache_sigtramp')
WARNING: arch/mips/mm/built-in.o(.text+0x1988): Section mismatch: reference to .init.text:except_vec2_sb1 (between 'sb1_cache_init' and 'sb1_flush_cache_sigtramp')

Signed-off-by: Ralf Baechle <ralf@linux-mips.org>
17 years ago[MIPS] PNX: Fix modpost warnings.
Ralf Baechle [Tue, 10 Jul 2007 14:04:06 +0000 (15:04 +0100)]
[MIPS] PNX: Fix modpost warnings.

  MODPOST vmlinux
WARNING: arch/mips/philips/pnx8550/common/built-in.o(.text+0xd40): Section mismatch: reference to .init.data: (after 'pnx8550_platform_init')
WARNING: arch/mips/philips/pnx8550/common/built-in.o(.text+0xd44): Section mismatch: reference to .init.data: (after 'pnx8550_platform_init')

Signed-off-by: Ralf Baechle <ralf@linux-mips.org>
17 years ago[MIPS] Alchemy: Fix modpost warnings.
Ralf Baechle [Tue, 10 Jul 2007 08:35:12 +0000 (09:35 +0100)]
[MIPS] Alchemy: Fix modpost warnings.

  MODPOST vmlinux
WARNING: arch/mips/au1000/common/built-in.o(.text+0x1750): Section mismatch: reference to .init.data: (between 'au1xxx_platform_init' and '__fixup_bigphys_addr')
WARNING: arch/mips/au1000/common/built-in.o(.text+0x1754): Section mismatch: reference to .init.data: (between 'au1xxx_platform_init' and '__fixup_bigphys_addr')

Signed-off-by: Ralf Baechle <ralf@linux-mips.org>
17 years ago[MIPS] Non-FPAFF: Fix warning.
Ralf Baechle [Tue, 10 Jul 2007 07:59:17 +0000 (08:59 +0100)]
[MIPS] Non-FPAFF: Fix warning.

Signed-off-by: Ralf Baechle <ralf@linux-mips.org>
17 years ago[MIPS] DEC: Fix modpost warning.
Ralf Baechle [Tue, 10 Jul 2007 07:14:15 +0000 (08:14 +0100)]
[MIPS] DEC: Fix modpost warning.

  LD      vmlinux
  SYSMAP  System.map
  SYSMAP  .tmp_System.map
  MODPOST vmlinux
WARNING: drivers/built-in.o(.data+0x2480): Section mismatch: reference to .init.text: (between 'sercons' and 'ds_parms')

Signed-off-by: Ralf Baechle <ralf@linux-mips.org>
17 years ago[MIPS] MIPSsim: Enable MIPSsim virtual network driver.
Ralf Baechle [Mon, 9 Jul 2007 18:20:20 +0000 (19:20 +0100)]
[MIPS] MIPSsim: Enable MIPSsim virtual network driver.

Signed-off-by: Ralf Baechle <ralf@linux-mips.org>
17 years ago[MIPS] Delete Ocelot 3 support.
Ralf Baechle [Mon, 9 Jul 2007 15:29:16 +0000 (16:29 +0100)]
[MIPS] Delete Ocelot 3 support.

Signed-off-by: Ralf Baechle <ralf@linux-mips.org>
17 years ago[MIPS] remove LASAT Networks platforms support
Yoichi Yuasa [Mon, 9 Jul 2007 04:10:55 +0000 (13:10 +0900)]
[MIPS] remove LASAT Networks platforms support

Signed-off-by: Yoichi Yuasa <yoichi_yuasa@tripeaks.co.jp>
Signed-off-by: Ralf Baechle <ralf@linux-mips.org>
17 years ago[MIPS] Early check for SMTC kernel on non-MT processor
Chris Dearman [Fri, 6 Jul 2007 16:11:20 +0000 (17:11 +0100)]
[MIPS] Early check for SMTC kernel on non-MT processor

Signed-off-by: Chris Dearman <chris@mips.com>
Signed-off-by: Ralf Baechle <ralf@linux-mips.org>
17 years ago[MIPS] Add debugfs files to show fpuemu statistics
Atsushi Nemoto [Sat, 7 Jul 2007 14:21:49 +0000 (23:21 +0900)]
[MIPS] Add debugfs files to show fpuemu statistics

Export contents of struct mips_fpu_emulator_stats via debugfs.

There is no way to read these statistics for now but they (at least
the "emulated" count) might be sometimes useful for performance tuning
on FPU-less CPUs.

Signed-off-by: Atsushi Nemoto <anemo@mba.ocn.ne.jp>
Signed-off-by: Ralf Baechle <ralf@linux-mips.org>
17 years ago[MIPS] Add some debugfs files to debug unaligned accesses
Atsushi Nemoto [Fri, 29 Jun 2007 15:55:48 +0000 (00:55 +0900)]
[MIPS] Add some debugfs files to debug unaligned accesses

Currently a number of unaligned instructions is counted but not used.
Add /debug/mips/unaligned_instructions file to show the value.

And add /debug/mips/unaligned_action to control behavior upon an
unaligned access.  Possible actions are:

0: silently fixup the unaligned access.
1: send SIGBUS.
2: dump registers, process name, etc. and fixup.

Signed-off-by: Atsushi Nemoto <anemo@mba.ocn.ne.jp>
Signed-off-by: Ralf Baechle <ralf@linux-mips.org>
17 years ago[MIPS] rbtx4938: Fix secondary PCIC and glue internal NICs
Atsushi Nemoto [Mon, 2 Jul 2007 13:43:06 +0000 (22:43 +0900)]
[MIPS] rbtx4938: Fix secondary PCIC and glue internal NICs

* Fix pci ops for secondary PCIC
* Do not reserve 1MB for PCI MEM region (leave PCIBIOS_MIN_MEM zero)
* Use platform_device to provide ethernet addresses for internal NICs.
  (background: TX49XX SoCs include PCI NIC (TC35815 compatible)
  connected via its internal PCI bus, but the NIC's PROM interface is
  not connected to SEEPROM.  So we must provide its ethernet address
  by another way.)
* Check return value of early_read_config_word()

Signed-off-by: Atsushi Nemoto <anemo@mba.ocn.ne.jp>
Signed-off-by: Ralf Baechle <ralf@linux-mips.org>
17 years ago[MIPS] tc35815: Load MAC address via platform_device
Atsushi Nemoto [Fri, 29 Jun 2007 13:34:53 +0000 (22:34 +0900)]
[MIPS] tc35815: Load MAC address via platform_device

TX49XX SoCs include PCI NIC (TC35815 compatible) connected via its
internal PCI bus, but the NIC's PROM interface is not connected to
SEEPROM.  So we must provide its ethernet address by another way.

Signed-off-by: Atsushi Nemoto <anemo@mba.ocn.ne.jp>
Acked-by: Jeff Garzik <jeff@garzik.org>
Signed-off-by: Ralf Baechle <ralf@linux-mips.org>
17 years ago[MIPS] Move FPU affinity code into separate file.
Ralf Baechle [Tue, 3 Jul 2007 13:37:43 +0000 (14:37 +0100)]
[MIPS] Move FPU affinity code into separate file.

Gets rid of some of the CONFIG_MIPS_MT_FPAFF #ifdefery.

Signed-off-by: Ralf Baechle <ralf@linux-mips.org>
17 years ago[MIPS] Make ioremap() work on TX39/49 special unmapped segment
Atsushi Nemoto [Mon, 25 Jun 2007 16:14:01 +0000 (01:14 +0900)]
[MIPS] Make ioremap() work on TX39/49 special unmapped segment

TX39XX and TX49XX have "reserved" segment in CKSEG3 area.
0xff000000-0xff3fffff on TX49XX and 0xff000000-0xfffeffff on TX39XX
are reserved (unmapped, uncached).  Controllers on these SoCs are
placed in this segment.

This patch add plat_ioremap() and plat_iounmap() to override default
behavior and implement these hooks for TX39/TX49.

Signed-off-by: Atsushi Nemoto <anemo@mba.ocn.ne.jp>
Signed-off-by: Ralf Baechle <ralf@linux-mips.org>
17 years ago[MIPS] rbtx4938: Update and minimize defconfig
Atsushi Nemoto [Fri, 22 Jun 2007 14:22:29 +0000 (23:22 +0900)]
[MIPS] rbtx4938: Update and minimize defconfig

Signed-off-by: Atsushi Nemoto <anemo@mba.ocn.ne.jp>
Signed-off-by: Ralf Baechle <ralf@linux-mips.org>
17 years ago[MIPS] rbtx4938: Convert SPI codes to use generic SPI drivers
Atsushi Nemoto [Fri, 22 Jun 2007 14:22:06 +0000 (23:22 +0900)]
[MIPS] rbtx4938: Convert SPI codes to use generic SPI drivers

Use rtc-rs5c348 and at25 spi protocol driver and spi_txx9 spi
controller driver instead of platform dependent codes.

This patch also removes dependencies to old RTC interfaces such as
rtc_mips_get_time, etc.

Signed-off-by: Atsushi Nemoto <anemo@mba.ocn.ne.jp>
Signed-off-by: Ralf Baechle <ralf@linux-mips.org>
17 years ago[MIPS] rbtx4938: Add generic GPIO support
Atsushi Nemoto [Fri, 22 Jun 2007 14:21:55 +0000 (23:21 +0900)]
[MIPS] rbtx4938: Add generic GPIO support

GPIO 0..15 are for TX4938 PIO pins, GPIO 16..18 are for FPGA-driven
chipselect signals for SPI devices.

Signed-off-by: Atsushi Nemoto <anemo@mba.ocn.ne.jp>
Signed-off-by: Ralf Baechle <ralf@linux-mips.org>
17 years ago[MIPS] SNI RM updates
Thomas Bogendoerfer [Wed, 20 Jun 2007 21:36:47 +0000 (23:36 +0200)]
[MIPS] SNI RM updates

- use RTC_CLASS instead of GEN_RTC
- get rid of ds1216 in favour of a RTC_CLASS driver
- use correct console device for older RM400
- use physical addresses for 82596 device
- use 128 byte L1 cache line size (this is needed because most of the
  SNI caches are using 128 L2 cache lines)

Signed-off-by: Thomas Bogendoerfer <tsbogend@alpha.franken.de>
Signed-off-by: Ralf Baechle <ralf@linux-mips.org>
17 years ago[MIPS] PMC MSP71xx default configuration
Marc St-Jean [Thu, 14 Jun 2007 21:57:11 +0000 (15:57 -0600)]
[MIPS] PMC MSP71xx default configuration

Patch to add default configuration for the PMC-Sierra
MSP71xx devices.

Signed-off-by: Marc St-Jean <Marc_St-Jean@pmc-sierra.com>
Signed-off-by: Ralf Baechle <ralf@linux-mips.org>
17 years ago[MIPS] PMC MSP71xx PCI support
Marc St-Jean [Thu, 14 Jun 2007 21:56:23 +0000 (15:56 -0600)]
[MIPS] PMC MSP71xx PCI support

Patch to add PCI support for the PMC-Sierra MSP71xx devices.

Signed-off-by: Marc St-Jean <Marc_St-Jean@pmc-sierra.com>
Signed-off-by: Ralf Baechle <ralf@linux-mips.org>
17 years ago[MIPS] PMC MSP71xx mips common
Marc St-Jean [Thu, 14 Jun 2007 21:55:31 +0000 (15:55 -0600)]
[MIPS] PMC MSP71xx mips common

Patch to add mips common support for the PMC-Sierra MSP71xx devices.

Signed-off-by: Marc St-Jean <Marc_St-Jean@pmc-sierra.com>
Signed-off-by: Ralf Baechle <ralf@linux-mips.org>
17 years ago[MIPS] PMC MSP71xx core platform
Marc St-Jean [Thu, 14 Jun 2007 21:54:47 +0000 (15:54 -0600)]
[MIPS] PMC MSP71xx core platform

Patch to add core platform support for the PMC-Sierra MSP71xx devices.

Signed-off-by: Marc St-Jean <Marc_St-Jean@pmc-sierra.com>
Signed-off-by: Ralf Baechle <ralf@linux-mips.org>
17 years ago[MIPS] Removes the few leftovers of the MOMENCO_JAGUAR_ATX removal.
Adrian Bunk [Tue, 19 Jun 2007 20:27:04 +0000 (22:27 +0200)]
[MIPS] Removes the few leftovers of the MOMENCO_JAGUAR_ATX removal.

Signed-off-by: Adrian Bunk <bunk@stusta.de>
Signed-off-by: Ralf Baechle <ralf@linux-mips.org>
17 years ago[MIPS] New files for lemote fulong mini-PC support
Songmao Tian [Wed, 6 Jun 2007 06:52:38 +0000 (14:52 +0800)]
[MIPS] New files for lemote fulong mini-PC support

Signed-off-by: Fuxin Zhang <zhangfx@lemote.com>
Signed-off-by: Songmao Tian <tiansm@lemote.com>
Signed-off-by: Ralf Baechle <ralf@linux-mips.org>
17 years ago[MIPS] define Hit_Invalidate_I to Index_Invalidate_I for loongson2
Fuxin Zhang [Wed, 6 Jun 2007 06:52:43 +0000 (14:52 +0800)]
[MIPS] define Hit_Invalidate_I to Index_Invalidate_I for loongson2

Signed-off-by: Fuxin Zhang <zhangfx@lemote.com>
Signed-off-by: Ralf Baechle <ralf@linux-mips.org>
17 years ago[MIPS] Convert init_thread initialization to ISO C initializers.
Ralf Baechle [Tue, 10 Jul 2007 16:33:02 +0000 (17:33 +0100)]
[MIPS] Convert init_thread initialization to ISO C initializers.

Signed-off-by: Ralf Baechle <ralf@linux-mips.org>
17 years ago[MIPS] Enable support for the userlocal hardware register
Ralf Baechle [Tue, 10 Jul 2007 16:33:02 +0000 (17:33 +0100)]
[MIPS] Enable support for the userlocal hardware register

Which will cut down the cost of RDHWR $29 which is used to obtain the
TLS pointer and so far being emulated in software down to a single cycle
operation.

Signed-off-by: Ralf Baechle <ralf@linux-mips.org>
17 years ago[MIPS] FP affinity: Coding style cleanups
Ralf Baechle [Tue, 10 Jul 2007 16:33:02 +0000 (17:33 +0100)]
[MIPS] FP affinity: Coding style cleanups

Signed-off-by: Ralf Baechle <ralf@linux-mips.org>
17 years ago[MIPS] Put an end to <asm/serial.h>'s long and annyoing existence
Ralf Baechle [Tue, 10 Jul 2007 16:33:01 +0000 (17:33 +0100)]
[MIPS] Put an end to <asm/serial.h>'s long and annyoing existence

Signed-off-by: Ralf Baechle <ralf@linux-mips.org>
17 years ago[MIPS] Remove Momenco Ocelot C support
Franck Bui-Huu [Mon, 11 Jun 2007 13:08:55 +0000 (15:08 +0200)]
[MIPS] Remove Momenco Ocelot C support

Signed-off-by: Franck Bui-Huu <fbuihuu@gmail.com>
Signed-off-by: Ralf Baechle <ralf@linux-mips.org>
 delete mode 100644 arch/mips/configs/ocelot_c_defconfig
 delete mode 100644 arch/mips/momentum/ocelot_c/Makefile
 delete mode 100644 arch/mips/momentum/ocelot_c/cpci-irq.c
 delete mode 100644 arch/mips/momentum/ocelot_c/dbg_io.c
 delete mode 100644 arch/mips/momentum/ocelot_c/irq.c
 delete mode 100644 arch/mips/momentum/ocelot_c/ocelot_c_fpga.h
 delete mode 100644 arch/mips/momentum/ocelot_c/platform.c
 delete mode 100644 arch/mips/momentum/ocelot_c/prom.c
 delete mode 100644 arch/mips/momentum/ocelot_c/reset.c
 delete mode 100644 arch/mips/momentum/ocelot_c/setup.c
 delete mode 100644 arch/mips/momentum/ocelot_c/uart-irq.c
 delete mode 100644 arch/mips/pci/fixup-ocelot-c.c
 delete mode 100644 arch/mips/pci/pci-ocelot-c.c

17 years ago[MIPS] IP32: Remove experimental tag from kconfig.
Ralf Baechle [Tue, 10 Jul 2007 16:33:00 +0000 (17:33 +0100)]
[MIPS] IP32: Remove experimental tag from kconfig.

Signed-off-by: Ralf Baechle <ralf@linux-mips.org>
17 years ago[MIPS] PCI: Make dev pointer argument of pcibios_map_irq const.
Ralf Baechle [Tue, 10 Jul 2007 16:33:00 +0000 (17:33 +0100)]
[MIPS] PCI: Make dev pointer argument of pcibios_map_irq const.

This is to break the code of people who think they are supposed to scribble
into the pci device structure - it's off limits.

Signed-off-by: Ralf Baechle <ralf@linux-mips.org>