Commit 42cb14b1 ("mm: migrate dirty page without
clear_page_dirty_for_io etc") simplified the migration of a PageDirty
pagecache page: one stat needs moving from zone to zone and that's about
all.

It's convenient and safest for it to shift the PageDirty bit from old
page to new, just before updating the zone stats: before copying data
and marking the new PageUptodate.  This is all done while both pages are
isolated and locked, just as before; and just as before, there's a
moment when the new page is visible in the radix_tree, but not yet
PageUptodate.  What's new is that it may now be briefly visible as
PageDirty before it is PageUptodate.

When I scoured the tree to see if this could cause a problem anywhere,
the only places I found were in two similar functions __r4w_get_page():
which look up a page with find_get_page() (not using page lock), then
claim it's uptodate if it's PageDirty or PageWriteback or PageUptodate.

I'm not sure whether that was right before, but now it might be wrong
(on rare occasions): only claim the page is uptodate if PageUptodate.
Or perhaps the page in question could never be migratable anyway?
Signed-off-by: NHugh Dickins <hughd@google.com>
Tested-by: NBoaz Harrosh <ooo@electrozaur.com>
Cc: Benny Halevy <bhalevy@panasas.com>
Cc: Trond Myklebust <trond.myklebust@primarydata.com>
Cc: Christoph Lameter <cl@linux.com>
Signed-off-by: NAndrew Morton <akpm@linux-foundation.org>
Signed-off-by: NLinus Torvalds <torvalds@linux-foundation.org>

IS_ERR(_OR_NULL) already contain an 'unlikely' compiler flag and there
is no need to do that again from its callers. Drop it.
Signed-off-by: NViresh Kumar <viresh.kumar@linaro.org>
Reviewed-by: NJeff Layton <jlayton@poochiereds.net>
Reviewed-by: NDavid Howells <dhowells@redhat.com>
Reviewed-by: NSteve French <smfrench@gmail.com>
Signed-off-by: NJiri Kosina <jkosina@suse.cz>

The LAYOUTCOMMIT operation means different things to different layout types.
For blocks and objects, it is both a data and metadata consistency operation.
For files and flexfiles, it is only a metadata consistency operation.

This patch separates out the 2 cases, allowing the files/flexfiles layout
drivers to optimise away the data consistency calls to layoutcommit.
Signed-off-by: NTrond Myklebust <trond.myklebust@primarydata.com>

Use of synchronize_rcu() when unmounting and potentially freeing a lot
of deviceids is problematic. There really is no reason why we can't just
use kfree_rcu() here.
Signed-off-by: NTrond Myklebust <trond.myklebust@primarydata.com>

Let it return current nfs_pgio_mirror in use depending on pg_mirror_count.
For read, we always use pg_mirrors[0], so this effectively gives us freedom
to use pg_mirror_idx to track the actual mirror to read from through out the
IO stack.
Signed-off-by: NPeng Tao <tao.peng@primarydata.com>
Signed-off-by: NTom Haynes <loghyr@primarydata.com>

This patch adds mirrored write support to the pgio layer. The default
is to use one mirror, but pgio callers may define callbacks to change
this to any value up to the (arbitrarily selected) limit of 16.

The basic idea is to break out members of nfs_pageio_descriptor that cannot
be shared between mirrored DSes and put them in a new structure.
Signed-off-by: NWeston Andros Adamson <dros@primarydata.com>

This is needed to support mirrored writes - the first write can't just
trash the lseg, we need to keep it around until all mirrors have
written.
Signed-off-by: NWeston Andros Adamson <dros@primarydata.com>

I no longer have access to the Panasas email.
So change to an email that can always reach me.
Signed-off-by: NBoaz Harrosh <ooo@electrozaur.com>

The kbuild test robot complained about a new sparse warning in
objio_alloc_deviceid_node, but it turns out that this was just a moved
reference to an existing variable.  Fix it to have the right big endian
annotated type.

Note that there are some other endianess issues in this file that I didn't
bother to sort out as they involve global headers.
Reported-by: Nkbuild test robot <fengguang.wu@intel.com>
Signed-off-by: NChristoph Hellwig <hch@lst.de>
Signed-off-by: NTrond Myklebust <trond.myklebust@primarydata.com>

Add support to the common pNFS core to issue GETDEVICEINFO calls on
a device ID cache miss.  The code is taken from the well debugged
file layout implementation and calls out to the layoutdriver through
a new alloc_deviceid_node method.  The calling conventions for
nfs4_find_get_deviceid are changed so that all information needed to
send a GETDEVICEINFO request is passed to the common code.
Signed-off-by: NChristoph Hellwig <hch@lst.de>
Signed-off-by: NTrond Myklebust <trond.myklebust@primarydata.com>

struct nfs_pgio_data only exists as a member of nfs_pgio_header, but is
passed around everywhere, because there used to be multiple _data structs
per _header. Many of these functions then use the _data to find a pointer
to the _header.  This patch cleans this up by merging the nfs_pgio_data
structure into nfs_pgio_header and passing nfs_pgio_header around instead.
Reviewed-by: NChristoph Hellwig <hch@lst.de>
Signed-off-by: NWeston Andros Adamson <dros@primarydata.com>
Signed-off-by: NTrond Myklebust <trond.myklebust@primarydata.com>

Now that pg_test can change the size of the request (by returning a non-zero
size smaller than the request), pg_test functions that call other
pg_test functions must return the minimum of the result - or 0 if any fail.

Also clean up the logic of some pg_test functions so that all checks are
for contitions where coalescing is not possible.
Signed-off-by: NWeston Andros Adamson <dros@primarydata.com>
Signed-off-by: NTrond Myklebust <trond.myklebust@primarydata.com>

This is a step toward allowing pg_test to inform the the
coalescing code to reduce the size of requests so they may fit in
whatever scheme the pg_test callback wants to define.

For now, just return the size of the request if there is space, or 0
if there is not.  This shouldn't change any behavior as it acts
the same as when the pg_test functions returned bool.
Signed-off-by: NWeston Andros Adamson <dros@primarydata.com>
Signed-off-by: NTrond Myklebust <trond.myklebust@primarydata.com>

At this point, the only difference between nfs_read_data and
nfs_write_data is the write verifier.
Signed-off-by: NAnna Schumaker <Anna.Schumaker@Netapp.com>
Signed-off-by: NTrond Myklebust <trond.myklebust@primarydata.com>

Correct spelling typos in printk and comments.
Signed-off-by: NMasanari Iida <standby24x7@gmail.com>
Acked-by: NRandy Dunlap <rdunlap@infradead.org>
Signed-off-by: NJiri Kosina <jkosina@suse.cz>

now pnfs client uses block layout, maybe we can remove
blocklayoutdriver first. if we umount later,
it can cause oops in unset_pnfs_layoutdriver.
because nfss->pnfs_curr_ld->clear_layoutdriver is invalid.

reproduce it:
 modprobe  blocklayoutdriver
 mount -t nfs4 -o minorversion=1 pnfsip:/ /mnt/
 rmmod blocklayoutdriver
 umount /mnt

then you can see following

CPU 0
Pid: 17023, comm: umount.nfs4 Tainted: GF          O 3.7.0-rc6-pnfs #1 VMware, Inc. VMware Virtual Platform/440BX Desktop Reference Platform
RIP: 0010:[<ffffffffa04cfe6d>]  [<ffffffffa04cfe6d>] unset_pnfs_layoutdriver+0x1d/0x70 [nfsv4]
RSP: 0018:ffff8800022d9e48  EFLAGS: 00010286
RAX: ffffffffa04a1b00 RBX: ffff88000b013800 RCX: 0000000000000001
RDX: ffffffff81ae8ee0 RSI: ffff880001ee94b8 RDI: ffff88000b013800
RBP: ffff8800022d9e58 R08: 0000000000000001 R09: 0000000000000000
R10: 0000000000000000 R11: 0000000000000000 R12: ffff880001ee9400
R13: ffff8800105978c0 R14: 00007fff25846c08 R15: 0000000001bba550
FS:  00007f45ae7f0700(0000) GS:ffff880012c00000(0000) knlGS:0000000000000000
CS:  0010 DS: 0000 ES: 0000 CR0: 000000008005003b
CR2: ffffffffa04a1b38 CR3: 0000000002c0c000 CR4: 00000000000006f0
DR0: 0000000000000000 DR1: 0000000000000000 DR2: 0000000000000000
DR3: 0000000000000000 DR6: 00000000ffff0ff0 DR7: 0000000000000400
Process umount.nfs4 (pid: 17023, threadinfo ffff8800022d8000, task ffff880006e48aa0)
Stack:
ffff8800105978c0 ffff88000b013800 ffff8800022d9e78 ffffffffa04cd0ce
ffff8800022d9e78 ffff88000b013800 ffff8800022d9ea8 ffffffffa04755a7
ffff8800022d9ea8 ffff880002f96400 ffff88000b013800 ffff880002f96400
Call Trace:
[<ffffffffa04cd0ce>] nfs4_destroy_server+0x1e/0x30 [nfsv4]
[<ffffffffa04755a7>] nfs_free_server+0xb7/0x150 [nfs]
[<ffffffffa047d4d5>] nfs_kill_super+0x35/0x40 [nfs]
[<ffffffff81178d35>] deactivate_locked_super+0x45/0x70
[<ffffffff8117986a>] deactivate_super+0x4a/0x70
[<ffffffff81193ee2>] mntput_no_expire+0xd2/0x130
[<ffffffff81194d62>] sys_umount+0x72/0xe0
[<ffffffff8154af59>] system_call_fastpath+0x16/0x1b
Code: 06 e1 b8 ea ff ff ff eb 9e 0f 1f 44 00 00 55 48 89 e5 53 48 83 ec 08 66 66 66 66 90 48 8b 87 80 03 00 00 48 89 fb 48 85 c0 74 29 <48> 8b 40 38 48 85 c0 74 02 ff d0 48 8b 03 3e ff 48 04 0f 94 c2
RIP  [<ffffffffa04cfe6d>] unset_pnfs_layoutdriver+0x1d/0x70 [nfsv4]
RSP <ffff8800022d9e48>
CR2: ffffffffa04a1b38
---[ end trace 29f75aaedda058bf ]---

Signed-off-by: fanchaoting<fanchaoting@cn.fujitsu.com>
Signed-off-by: NTrond Myklebust <Trond.Myklebust@netapp.com>
Cc: stable@vger.kernel.org

Signed-off-by: NTrond Myklebust <Trond.Myklebust@netapp.com>

For buffer write, block layout client scan inode mapping to find
next hole and use offset-to-hole as layoutget length. Object
layout client uses offset-to-isize as layoutget length.

For direct write, both block layout and object layout use dreq->bytes_left.
Signed-off-by: NPeng Tao <tao.peng@emc.com>
Signed-off-by: NTrond Myklebust <Trond.Myklebust@netapp.com>

Depending on layout and ARCH, ORE has some limits on max IO sizes
which is communicated on (what else) ore_layout->max_io_length,
which is always stripe aligned.
This was considered as the pg_test boundary for splitting and starting
a new IO.

But in the case of a long IO where the start offset is not aligned
what would happen is that both end of IO[N] and start of IO[N+1]
would be unaligned, causing each IO boundary parity unit to be
calculated and written twice.

So what we do in this patch is split the very start of an unaligned
IO, up to a stripe boundary, and then next IO's can continue fully
aligned til the end.

We might be sacrificing the case where the full unaligned IO would
fit within a single max_io_length, but the sacrifice is well worth
the elimination of double calculation and parity units IO.
Actually the sacrificing is marginal and is almost unmeasurable.

TODO:
	If we know the total expected linear segment that will
	be received, at pg_init, we could use that information
	in many places:
	1. blocks-layout get_layout write segment size
	2. Better mds-threshold
	3. In above situation for a better clean split

	I will do this in future submission.
Signed-off-by: NBoaz Harrosh <bharrosh@panasas.com>
Signed-off-by: NTrond Myklebust <Trond.Myklebust@netapp.com>

It is very common for the end of the file to be unaligned on
stripe size. But since we know it's beyond file's end then
the XOR should be preformed with all zeros.

Old code used to just read zeros out of the OSD devices, which is a great
waist. But what scares me more about this situation is that, we now have
pages attached to the file's mapping that are beyond i_size. I don't
like the kind of bugs this calls for.

Fix both birds, by returning a global zero_page, if offset is beyond
i_size.

TODO:
	Change the API to ->__r4w_get_page() so a NULL can be
	returned without being considered as error, since XOR API
	treats NULL entries as zero_pages.

[Bug since 3.2. Should apply the same way to all Kernels since]
CC: Stable Tree <stable@kernel.org>
Signed-off-by: NBoaz Harrosh <bharrosh@panasas.com>

[Bug since 3.2 Kernel]
CC: Stable Tree <stable@kernel.org>
Signed-off-by: NBoaz Harrosh <bharrosh@panasas.com>

Fix the following sparse warnings:

fs/nfs/direct.c:221:6: warning: symbol 'nfs_direct_readpage_release' was
not declared. Should it be static?
fs/nfs/read.c:38:43: warning: non-ANSI function declaration of function
'nfs_readhdr_alloc'
fs/nfs/objlayout/objio_osd.c:214:5: warning: symbol '__alloc_objio_seg'
was not declared. Should it be static?
Reported-by: NDan Carpenter <dan.carpenter@oracle.com>
Signed-off-by: NTrond Myklebust <Trond.Myklebust@netapp.com>
Cc: Fred Isaman <iisaman@netapp.com>
Cc: Boaz Harrosh <bharrosh@panasas.com>

In order to avoid duplicating all the data in nfs_read_data whenever we
split it up into multiple RPC calls (either due to a short read result
or due to rsize < PAGE_SIZE), we split out the bits that are the same
per RPC call into a separate "header" structure.

The goal this patch moves towards is to have a single header
refcounted by several rpc_data structures.  Thus, want to always refer
from rpc_data to the header, and not the other way.  This patch comes
close to that ideal, but the directio code currently needs some
special casing, isolated in the nfs_direct_[read_write]hdr_release()
functions.  This will be dealt with in a future patch.
Signed-off-by: NFred Isaman <iisaman@netapp.com>
Signed-off-by: NTrond Myklebust <Trond.Myklebust@netapp.com>

The pnfs-objects protocol mandates that we autologin into devices not
present in the system, according to information specified in the
get_device_info returned from the server.

The Protocol specifies two login hints.
1. An IP address:port combination
2. A string URI which is constructed as a URL with a protocol prefix
   followed by :// and a string as address. For each  protocol prefix
   the string-address format might be different.

We only support the second option. The first option is just redundant
to the second one.
NOTE: The Kernel part of autologin does not parse the URI string. It
just channels it to a user-mode script. So any new login protocols should
only update the user-mode script which is a part of the nfs-utils package,
but the Kernel need not change.

We implement the autologin by using the call_usermodehelper() API.
(Thanks to Steve Dickson <steved@redhat.com> for pointing it out)
So there is no running daemon needed, and/or special setup.

We Add the osd_login_prog Kernel module parameters which defaults to:
	/sbin/osd_login

Kernel try's to upcall the program specified in osd_login_prog. If the file is
not found or the execution fails Kernel will disable any farther upcalls, by
zeroing out  osd_login_prog, Until Admin re-enables it by setting the
osd_login_prog parameter to a proper program.

Also add text about the osd_login program command line API to:
	Documentation/filesystems/nfs/pnfs.txt
and documentation of the new  osd_login_prog  module parameter to:
	Documentation/kernel-parameters.txt

TODO: Add timeout option in the case osd_login program gets
              stuck
Signed-off-by: NSachin Bhamare <sbhamare@panasas.com>
Signed-off-by: NBoaz Harrosh <bharrosh@panasas.com>
Signed-off-by: NTrond Myklebust <Trond.Myklebust@netapp.com>

At some past instance Linus Trovalds wrote:
> From: Linus Torvalds <torvalds@linux-foundation.org>
> commit a84a79e4 upstream.
>
> The size is always valid, but variable-length arrays generate worse code
> for no good reason (unless the function happens to be inlined and the
> compiler sees the length for the simple constant it is).
>
> Also, there seems to be some code generation problem on POWER, where
> Henrik Bakken reports that register r28 can get corrupted under some
> subtle circumstances (interrupt happening at the wrong time?).  That all
> indicates some seriously broken compiler issues, but since variable
> length arrays are bad regardless, there's little point in trying to
> chase it down.
>
> "Just don't do that, then".

Since then any use of "variable length arrays" has become blasphemous.
Even in perfectly good, beautiful, perfectly safe code like the one
below where the variable length arrays are only used as a sizeof()
parameter, for type-safe dynamic structure allocations. GCC is not
executing any stack allocation code.

I have produced a small file which defines two functions main1(unsigned numdevs)
and main2(unsigned numdevs). main1 uses code as before with call to malloc
and main2 uses code as of after this patch. I compiled it as:
	gcc -O2 -S see_asm.c
and here is what I get:

<see_asm.s>
main1:
.LFB7:
	.cfi_startproc
	mov	%edi, %edi
	leaq	4(%rdi,%rdi), %rdi
	salq	$3, %rdi
	jmp	malloc
	.cfi_endproc
.LFE7:
	.size	main1, .-main1
	.p2align 4,,15
	.globl	main2
	.type	main2, @function
main2:
.LFB8:
	.cfi_startproc
	mov	%edi, %edi
	addq	$2, %rdi
	salq	$4, %rdi
	jmp	malloc
	.cfi_endproc
.LFE8:
	.size	main2, .-main2
	.section	.text.startup,"ax",@progbits
	.p2align 4,,15
</see_asm.s>

*Exact* same code !!!

So please seriously consider not accepting this patch and leave the
perfectly good code intact.

CC: Linus Torvalds <torvalds@linux-foundation.org>
Signed-off-by: NBoaz Harrosh <bharrosh@panasas.com>
Signed-off-by: NTrond Myklebust <Trond.Myklebust@netapp.com>

This patch addresses printks that have some context to show that they are
from fs/nfs/, but for the sake of consistency now start with NFS:
Signed-off-by: NWeston Andros Adamson <dros@netapp.com>
Signed-off-by: NTrond Myklebust <Trond.Myklebust@netapp.com>

As mandated by the standard. In case of an IO error, a pNFS
objects layout driver must return it's layout. This is because
all device errors are reported to the server as part of the
layout return buffer.

This is implemented the same way PNFS_LAYOUTRET_ON_SETATTR
is done, through a bit flag on the pnfs_layoutdriver_type->flags
member. The flag is set by the layout driver that wants a
layout_return preformed at pnfs_ld_{write,read}_done in case
of an error.
(Though I have not defined a wrapper like pnfs_ld_layoutret_on_setattr
 because this code is never called outside of pnfs.c and pnfs IO
 paths)

Without this patch 3.[0-2] Kernels leak memory and have an annoying
WARN_ON after every IO error utilizing the pnfs-obj driver.

[This patch is for 3.2 Kernel. 3.1/0 Kernels need a different patch]
CC: Stable Tree <stable@kernel.org>
Signed-off-by: NBoaz Harrosh <bharrosh@panasas.com>
Signed-off-by: NTrond Myklebust <Trond.Myklebust@netapp.com>

The ore need suplied a r4w_get_page/r4w_put_page API
from Filesystem so it can get cache pages to read-into when
writing parial stripes.
Signed-off-by: NBoaz Harrosh <bharrosh@panasas.com>
Signed-off-by: NTrond Myklebust <Trond.Myklebust@netapp.com>

Finally remove all the old raid engine, which is by now
dead code.
Signed-off-by: NBoaz Harrosh <bharrosh@panasas.com>
Signed-off-by: NTrond Myklebust <Trond.Myklebust@netapp.com>

In this patch we are actually moving to the ORE.
(Object Raid Engine).

objio_state holds a pointer to an ore_io_state. Once
we have an ore_io_state at hand we can call the ore
for reading/writing. We register on the done path
to kick off the nfs io_done mechanism.

Again for Ease of reviewing the old code is "#if 0"
but is not removed so the diff command works better.
The old code will be removed in the next patch.

fs/exofs/Kconfig::ORE is modified to also be auto-included
if PNFS_OBJLAYOUT is set. Since we now depend on ORE.
(See comments in fs/exofs/Kconfig)
Signed-off-by: NBoaz Harrosh <bharrosh@panasas.com>
Signed-off-by: NTrond Myklebust <Trond.Myklebust@netapp.com>

For Ease of reviewing I split the move to ore into 3 parts
	move to ore 01: ore_layout & ore_components
	move to ore 02: move to ORE
	move to ore 03: Remove old raid engine

This patch modifies the objio_lseg, layout-segment level
and devices and components arrays to use the ORE types.

Though it will be removed soon, also the raid engine
is modified to actually compile, possibly run, with
the new types. So it is the same old raid engine but
with some new ORE types.

For Ease of reviewing, some of the old code is
"#if 0" but is not removed so the diff command works
better. The old code will be removed in the 3rd patch.
Signed-off-by: NBoaz Harrosh <bharrosh@panasas.com>
Signed-off-by: NTrond Myklebust <Trond.Myklebust@netapp.com>

* All instances of objlayout_io_state => objlayout_io_res
* All instances of state => oir;
* All instances of ol_state => oir;

Big but nothing to it
Signed-off-by: NBoaz Harrosh <bharrosh@panasas.com>
Signed-off-by: NTrond Myklebust <Trond.Myklebust@netapp.com>

This is part of moving objio_osd to use the ORE.

objlayout_io_state had two functions:
1. It was used in the error reporting mechanism at layout_return.
   This function is kept intact.
   (Later patch will rename objlayout_io_state => objlayout_io_res)
2. Carrier of rw io members into the objio_read/write_paglist API.
   This is removed in this patch.

The {r,w}data received from NFS are passed directly to the
objio_{read,write}_paglist API. The io_engine is now allocating
it's own IO state as part of the read/write. The minimal
functionality that was part of the generic allocation is passed
to the io_engine.

So part of this patch is rename of:
	ios->ol_state.foo => ios->foo

At objlayout_{read,write}_done an objlayout_io_state is passed that
denotes the result of the IO. (Hence the later name change).
If the IO is successful objlayout calls an objio_free_result() API
immediately (Which for objio_osd causes the release of the io_state).
If the IO ended in an error it is hanged onto until reported in
layout_return and is released later through the objio_free_result()
API. (All this is not new just renamed and cleaned)
Signed-off-by: NBoaz Harrosh <bharrosh@panasas.com>
Signed-off-by: NTrond Myklebust <Trond.Myklebust@netapp.com>

objlayout driver was always returning PNFS_ATTEMPTED from it's
read/write_pagelist operations. Even on error. Fix that.

Start by establishing an error return API from io-engine, by
not returning ssize_t (length-or-error) but returning "int"
0=OK, 0>Error. And clean up all return types in io-engine.

Then if io-engine returned error return PNFS_NOT_ATTEMPTED
to generic layer. (With a dprint)
Signed-off-by: NBoaz Harrosh <bharrosh@panasas.com>
Signed-off-by: NTrond Myklebust <Trond.Myklebust@netapp.com>

There were bugs in the case of partial layout where olo_comp_index
is not zero. This used to work and was tested but one of the later
cleanup SQUASHMEs broke it and was not tested since.

Also add a dprint that specify those received layout parameters.
Everything else was already printed.

[Needed in v3.0]
CC: Stable Tree <stable@kernel.org>
Signed-off-by: NBoaz Harrosh <bharrosh@panasas.com>
Signed-off-by: NTrond Myklebust <Trond.Myklebust@netapp.com>

When we have a situation that the number of pages we want
to encode is bigger then the size of the bio. (Which can
currently happen only when all IO is going to a single device
.e.g group_width==1) then the IO is submitted short and we
report back only the amount of bytes we actually wrote/read
and all is fine. BUT ...

There was a bug that the current length counter was advanced
before the fail to add the extra page, and we come to a situation
that the CDB length was one-page longer then the actual bio size,
which is of course rejected by the osd-target.

While here also fix the bio size calculation, in the case
that we received more then one group of devices.

CC: Stable Tree <stable@kernel.org>
Signed-off-by: NBoaz Harrosh <bharrosh@panasas.com>
Signed-off-by: NTrond Myklebust <Trond.Myklebust@netapp.com>

Embed the necessary alias into the module rather than waiting for
someone to add it to /etc/modprobe.conf
Signed-off-by: NJ. Bruce Fields <bfields@redhat.com>
Signed-off-by: NTrond Myklebust <Trond.Myklebust@netapp.com>

...and ensure that we recoalese to take into account differences in
differences in block sizes when falling back to write through the MDS.
Signed-off-by: NTrond Myklebust <Trond.Myklebust@netapp.com>

...and ensure that we recoalese to take into account differences in
block sizes when falling back to read through the MDS.
Signed-off-by: NTrond Myklebust <Trond.Myklebust@netapp.com>

Signed-off-by: NTrond Myklebust <Trond.Myklebust@netapp.com>