Constify all instances of blk_mq_ops, as they are never modified.
Signed-off-by: NEric Biggers <ebiggers@google.com>
Signed-off-by: NJens Axboe <axboe@fb.com>

Add a system call to make extended file information available, including
file creation and some attribute flags where available through the
underlying filesystem.

The getattr inode operation is altered to take two additional arguments: a
u32 request_mask and an unsigned int flags that indicate the
synchronisation mode.  This change is propagated to the vfs_getattr*()
function.

Functions like vfs_stat() are now inline wrappers around new functions
vfs_statx() and vfs_statx_fd() to reduce stack usage.

========
OVERVIEW
========

The idea was initially proposed as a set of xattrs that could be retrieved
with getxattr(), but the general preference proved to be for a new syscall
with an extended stat structure.

A number of requests were gathered for features to be included.  The
following have been included:

 (1) Make the fields a consistent size on all arches and make them large.

 (2) Spare space, request flags and information flags are provided for
     future expansion.

 (3) Better support for the y2038 problem [Arnd Bergmann] (tv_sec is an
     __s64).

 (4) Creation time: The SMB protocol carries the creation time, which could
     be exported by Samba, which will in turn help CIFS make use of
     FS-Cache as that can be used for coherency data (stx_btime).

     This is also specified in NFSv4 as a recommended attribute and could
     be exported by NFSD [Steve French].

 (5) Lightweight stat: Ask for just those details of interest, and allow a
     netfs (such as NFS) to approximate anything not of interest, possibly
     without going to the server [Trond Myklebust, Ulrich Drepper, Andreas
     Dilger] (AT_STATX_DONT_SYNC).

 (6) Heavyweight stat: Force a netfs to go to the server, even if it thinks
     its cached attributes are up to date [Trond Myklebust]
     (AT_STATX_FORCE_SYNC).

And the following have been left out for future extension:

 (7) Data version number: Could be used by userspace NFS servers [Aneesh
     Kumar].

     Can also be used to modify fill_post_wcc() in NFSD which retrieves
     i_version directly, but has just called vfs_getattr().  It could get
     it from the kstat struct if it used vfs_xgetattr() instead.

     (There's disagreement on the exact semantics of a single field, since
     not all filesystems do this the same way).

 (8) BSD stat compatibility: Including more fields from the BSD stat such
     as creation time (st_btime) and inode generation number (st_gen)
     [Jeremy Allison, Bernd Schubert].

 (9) Inode generation number: Useful for FUSE and userspace NFS servers
     [Bernd Schubert].

     (This was asked for but later deemed unnecessary with the
     open-by-handle capability available and caused disagreement as to
     whether it's a security hole or not).

(10) Extra coherency data may be useful in making backups [Andreas Dilger].

     (No particular data were offered, but things like last backup
     timestamp, the data version number and the DOS archive bit would come
     into this category).

(11) Allow the filesystem to indicate what it can/cannot provide: A
     filesystem can now say it doesn't support a standard stat feature if
     that isn't available, so if, for instance, inode numbers or UIDs don't
     exist or are fabricated locally...

     (This requires a separate system call - I have an fsinfo() call idea
     for this).

(12) Store a 16-byte volume ID in the superblock that can be returned in
     struct xstat [Steve French].

     (Deferred to fsinfo).

(13) Include granularity fields in the time data to indicate the
     granularity of each of the times (NFSv4 time_delta) [Steve French].

     (Deferred to fsinfo).

(14) FS_IOC_GETFLAGS value.  These could be translated to BSD's st_flags.
     Note that the Linux IOC flags are a mess and filesystems such as Ext4
     define flags that aren't in linux/fs.h, so translation in the kernel
     may be a necessity (or, possibly, we provide the filesystem type too).

     (Some attributes are made available in stx_attributes, but the general
     feeling was that the IOC flags were to ext[234]-specific and shouldn't
     be exposed through statx this way).

(15) Mask of features available on file (eg: ACLs, seclabel) [Brad Boyer,
     Michael Kerrisk].

     (Deferred, probably to fsinfo.  Finding out if there's an ACL or
     seclabal might require extra filesystem operations).

(16) Femtosecond-resolution timestamps [Dave Chinner].

     (A __reserved field has been left in the statx_timestamp struct for
     this - if there proves to be a need).

(17) A set multiple attributes syscall to go with this.

===============
NEW SYSTEM CALL
===============

The new system call is:

	int ret = statx(int dfd,
			const char *filename,
			unsigned int flags,
			unsigned int mask,
			struct statx *buffer);

The dfd, filename and flags parameters indicate the file to query, in a
similar way to fstatat().  There is no equivalent of lstat() as that can be
emulated with statx() by passing AT_SYMLINK_NOFOLLOW in flags.  There is
also no equivalent of fstat() as that can be emulated by passing a NULL
filename to statx() with the fd of interest in dfd.

Whether or not statx() synchronises the attributes with the backing store
can be controlled by OR'ing a value into the flags argument (this typically
only affects network filesystems):

 (1) AT_STATX_SYNC_AS_STAT tells statx() to behave as stat() does in this
     respect.

 (2) AT_STATX_FORCE_SYNC will require a network filesystem to synchronise
     its attributes with the server - which might require data writeback to
     occur to get the timestamps correct.

 (3) AT_STATX_DONT_SYNC will suppress synchronisation with the server in a
     network filesystem.  The resulting values should be considered
     approximate.

mask is a bitmask indicating the fields in struct statx that are of
interest to the caller.  The user should set this to STATX_BASIC_STATS to
get the basic set returned by stat().  It should be noted that asking for
more information may entail extra I/O operations.

buffer points to the destination for the data.  This must be 256 bytes in
size.

======================
MAIN ATTRIBUTES RECORD
======================

The following structures are defined in which to return the main attribute
set:

	struct statx_timestamp {
		__s64	tv_sec;
		__s32	tv_nsec;
		__s32	__reserved;
	};

	struct statx {
		__u32	stx_mask;
		__u32	stx_blksize;
		__u64	stx_attributes;
		__u32	stx_nlink;
		__u32	stx_uid;
		__u32	stx_gid;
		__u16	stx_mode;
		__u16	__spare0[1];
		__u64	stx_ino;
		__u64	stx_size;
		__u64	stx_blocks;
		__u64	__spare1[1];
		struct statx_timestamp	stx_atime;
		struct statx_timestamp	stx_btime;
		struct statx_timestamp	stx_ctime;
		struct statx_timestamp	stx_mtime;
		__u32	stx_rdev_major;
		__u32	stx_rdev_minor;
		__u32	stx_dev_major;
		__u32	stx_dev_minor;
		__u64	__spare2[14];
	};

The defined bits in request_mask and stx_mask are:

	STATX_TYPE		Want/got stx_mode & S_IFMT
	STATX_MODE		Want/got stx_mode & ~S_IFMT
	STATX_NLINK		Want/got stx_nlink
	STATX_UID		Want/got stx_uid
	STATX_GID		Want/got stx_gid
	STATX_ATIME		Want/got stx_atime{,_ns}
	STATX_MTIME		Want/got stx_mtime{,_ns}
	STATX_CTIME		Want/got stx_ctime{,_ns}
	STATX_INO		Want/got stx_ino
	STATX_SIZE		Want/got stx_size
	STATX_BLOCKS		Want/got stx_blocks
	STATX_BASIC_STATS	[The stuff in the normal stat struct]
	STATX_BTIME		Want/got stx_btime{,_ns}
	STATX_ALL		[All currently available stuff]

stx_btime is the file creation time, stx_mask is a bitmask indicating the
data provided and __spares*[] are where as-yet undefined fields can be
placed.

Time fields are structures with separate seconds and nanoseconds fields
plus a reserved field in case we want to add even finer resolution.  Note
that times will be negative if before 1970; in such a case, the nanosecond
fields will also be negative if not zero.

The bits defined in the stx_attributes field convey information about a
file, how it is accessed, where it is and what it does.  The following
attributes map to FS_*_FL flags and are the same numerical value:

	STATX_ATTR_COMPRESSED		File is compressed by the fs
	STATX_ATTR_IMMUTABLE		File is marked immutable
	STATX_ATTR_APPEND		File is append-only
	STATX_ATTR_NODUMP		File is not to be dumped
	STATX_ATTR_ENCRYPTED		File requires key to decrypt in fs

Within the kernel, the supported flags are listed by:

	KSTAT_ATTR_FS_IOC_FLAGS

[Are any other IOC flags of sufficient general interest to be exposed
through this interface?]

New flags include:

	STATX_ATTR_AUTOMOUNT		Object is an automount trigger

These are for the use of GUI tools that might want to mark files specially,
depending on what they are.

Fields in struct statx come in a number of classes:

 (0) stx_dev_*, stx_blksize.

     These are local system information and are always available.

 (1) stx_mode, stx_nlinks, stx_uid, stx_gid, stx_[amc]time, stx_ino,
     stx_size, stx_blocks.

     These will be returned whether the caller asks for them or not.  The
     corresponding bits in stx_mask will be set to indicate whether they
     actually have valid values.

     If the caller didn't ask for them, then they may be approximated.  For
     example, NFS won't waste any time updating them from the server,
     unless as a byproduct of updating something requested.

     If the values don't actually exist for the underlying object (such as
     UID or GID on a DOS file), then the bit won't be set in the stx_mask,
     even if the caller asked for the value.  In such a case, the returned
     value will be a fabrication.

     Note that there are instances where the type might not be valid, for
     instance Windows reparse points.

 (2) stx_rdev_*.

     This will be set only if stx_mode indicates we're looking at a
     blockdev or a chardev, otherwise will be 0.

 (3) stx_btime.

     Similar to (1), except this will be set to 0 if it doesn't exist.

=======
TESTING
=======

The following test program can be used to test the statx system call:

	samples/statx/test-statx.c

Just compile and run, passing it paths to the files you want to examine.
The file is built automatically if CONFIG_SAMPLES is enabled.

Here's some example output.  Firstly, an NFS directory that crosses to
another FSID.  Note that the AUTOMOUNT attribute is set because transiting
this directory will cause d_automount to be invoked by the VFS.

	[root@andromeda ~]# /tmp/test-statx -A /warthog/data
	statx(/warthog/data) = 0
	results=7ff
	  Size: 4096            Blocks: 8          IO Block: 1048576  directory
	Device: 00:26           Inode: 1703937     Links: 125
	Access: (3777/drwxrwxrwx)  Uid:     0   Gid:  4041
	Access: 2016-11-24 09:02:12.219699527+0000
	Modify: 2016-11-17 10:44:36.225653653+0000
	Change: 2016-11-17 10:44:36.225653653+0000
	Attributes: 0000000000001000 (-------- -------- -------- -------- -------- -------- ---m---- --------)

Secondly, the result of automounting on that directory.

	[root@andromeda ~]# /tmp/test-statx /warthog/data
	statx(/warthog/data) = 0
	results=7ff
	  Size: 4096            Blocks: 8          IO Block: 1048576  directory
	Device: 00:27           Inode: 2           Links: 125
	Access: (3777/drwxrwxrwx)  Uid:     0   Gid:  4041
	Access: 2016-11-24 09:02:12.219699527+0000
	Modify: 2016-11-17 10:44:36.225653653+0000
	Change: 2016-11-17 10:44:36.225653653+0000
Signed-off-by: NDavid Howells <dhowells@redhat.com>
Signed-off-by: NAl Viro <viro@zeniv.linux.org.uk>

loop_reread_partitions() needs to do I/O, but we just froze the queue,
so we end up waiting forever. This can easily be reproduced with losetup
-P. Fix it by moving the reread to after we unfreeze the queue.

Fixes: ecdd0959 ("block/loop: fix race between I/O and set_status")
Reported-by: NTejun Heo <tj@kernel.org>
Cc: stable@vger.kernel.org
Signed-off-by: NOmar Sandoval <osandov@fb.com>
Reviewed-by: NMing Lei <tom.leiming@gmail.com>
Signed-off-by: NJens Axboe <axboe@fb.com>

Fix typos and add the following to the scripts/spelling.txt:

  algined||aligned

While we are here, fix the "appplication" in the touched line in
drivers/block/loop.c.  Also, fix the "may not naturally ..." to
"may not be naturally ..." in the touched line in mm/page_alloc.

Link: http://lkml.kernel.org/r/1481573103-11329-9-git-send-email-yamada.masahiro@socionext.comSigned-off-by: NMasahiro Yamada <yamada.masahiro@socionext.com>
Signed-off-by: NAndrew Morton <akpm@linux-foundation.org>
Signed-off-by: NLinus Torvalds <torvalds@linux-foundation.org>

Signed-off-by: NMiklos Szeredi <mszeredi@redhat.com>

Inside set_status, transfer need to setup again, so
we have to drain IO before the transition, otherwise
oops may be triggered like the following:

	divide error: 0000 [#1] SMP KASAN
	CPU: 0 PID: 2935 Comm: loop7 Not tainted 4.10.0-rc7+ #213
	Hardware name: QEMU Standard PC (i440FX + PIIX, 1996), BIOS Bochs
	01/01/2011
	task: ffff88006ba1e840 task.stack: ffff880067338000
	RIP: 0010:transfer_xor+0x1d1/0x440 drivers/block/loop.c:110
	RSP: 0018:ffff88006733f108 EFLAGS: 00010246
	RAX: 0000000000000000 RBX: ffff8800688d7000 RCX: 0000000000000059
	RDX: 0000000000000000 RSI: 1ffff1000d743f43 RDI: ffff880068891c08
	RBP: ffff88006733f160 R08: ffff8800688d7001 R09: 0000000000000000
	R10: 0000000000000000 R11: 0000000000000000 R12: ffff8800688d7000
	R13: ffff880067b7d000 R14: dffffc0000000000 R15: 0000000000000000
	FS:  0000000000000000(0000) GS:ffff88006d000000(0000)
	knlGS:0000000000000000
	CS:  0010 DS: 0000 ES: 0000 CR0: 0000000080050033
	CR2: 00000000006c17e0 CR3: 0000000066e3b000 CR4: 00000000001406f0
	Call Trace:
	 lo_do_transfer drivers/block/loop.c:251 [inline]
	 lo_read_transfer drivers/block/loop.c:392 [inline]
	 do_req_filebacked drivers/block/loop.c:541 [inline]
	 loop_handle_cmd drivers/block/loop.c:1677 [inline]
	 loop_queue_work+0xda0/0x49b0 drivers/block/loop.c:1689
	 kthread_worker_fn+0x4c3/0xa30 kernel/kthread.c:630
	 kthread+0x326/0x3f0 kernel/kthread.c:227
	 ret_from_fork+0x31/0x40 arch/x86/entry/entry_64.S:430
	Code: 03 83 e2 07 41 29 df 42 0f b6 04 30 4d 8d 44 24 01 38 d0 7f 08
	84 c0 0f 85 62 02 00 00 44 89 f8 41 0f b6 48 ff 25 ff 01 00 00 99 <f7>
	7d c8 48 63 d2 48 03 55 d0 48 89 d0 48 89 d7 48 c1 e8 03 83
	RIP: transfer_xor+0x1d1/0x440 drivers/block/loop.c:110 RSP:
	ffff88006733f108
	---[ end trace 0166f7bd3b0c0933 ]---
Reported-by: NDmitry Vyukov <dvyukov@google.com>
Cc: stable@vger.kernel.org
Signed-off-by: NMing Lei <tom.leiming@gmail.com>
Tested-by: NDmitry Vyukov <dvyukov@google.com>
Signed-off-by: NJens Axboe <axboe@fb.com>

This was entirely automated, using the script by Al:

  PATT='^[[:blank:]]*#[[:blank:]]*include[[:blank:]]*<asm/uaccess.h>'
  sed -i -e "s!$PATT!#include <linux/uaccess.h>!" \
        $(git grep -l "$PATT"|grep -v ^include/linux/uaccess.h)

to do the replacement at the end of the merge window.
Requested-by: NAl Viro <viro@zeniv.linux.org.uk>
Signed-off-by: NLinus Torvalds <torvalds@linux-foundation.org>

->queue_rq() should return one of the BLK_MQ_RQ_QUEUE_* constants, not
an errno.

f4aa4c7b ("block: loop: convert to per-device workqueue")
Signed-off-by: NOmar Sandoval <osandov@fb.com>
Cc: stable@vger.kernel.org
Signed-off-by: NJens Axboe <axboe@fb.com>

A good practice is to prefix the names of functions by the name
of the subsystem.

The kthread worker API is a mix of classic kthreads and workqueues.  Each
worker has a dedicated kthread.  It runs a generic function that process
queued works.  It is implemented as part of the kthread subsystem.

This patch renames the existing kthread worker API to use
the corresponding name from the workqueues API prefixed by
kthread_:

__init_kthread_worker()		-> __kthread_init_worker()
init_kthread_worker()		-> kthread_init_worker()
init_kthread_work()		-> kthread_init_work()
insert_kthread_work()		-> kthread_insert_work()
queue_kthread_work()		-> kthread_queue_work()
flush_kthread_work()		-> kthread_flush_work()
flush_kthread_worker()		-> kthread_flush_worker()

Note that the names of DEFINE_KTHREAD_WORK*() macros stay
as they are. It is common that the "DEFINE_" prefix has
precedence over the subsystem names.

Note that INIT() macros and init() functions use different
naming scheme. There is no good solution. There are several
reasons for this solution:

  + "init" in the function names stands for the verb "initialize"
    aka "initialize worker". While "INIT" in the macro names
    stands for the noun "INITIALIZER" aka "worker initializer".

  + INIT() macros are used only in DEFINE() macros

  + init() functions are used close to the other kthread()
    functions. It looks much better if all the functions
    use the same scheme.

  + There will be also kthread_destroy_worker() that will
    be used close to kthread_cancel_work(). It is related
    to the init() function. Again it looks better if all
    functions use the same naming scheme.

  + there are several precedents for such init() function
    names, e.g. amd_iommu_init_device(), free_area_init_node(),
    jump_label_init_type(),  regmap_init_mmio_clk(),

  + It is not an argument but it was inconsistent even before.

[arnd@arndb.de: fix linux-next merge conflict]
 Link: http://lkml.kernel.org/r/20160908135724.1311726-1-arnd@arndb.de
Link: http://lkml.kernel.org/r/1470754545-17632-3-git-send-email-pmladek@suse.comSuggested-by: NAndrew Morton <akpm@linux-foundation.org>
Signed-off-by: NPetr Mladek <pmladek@suse.com>
Cc: Oleg Nesterov <oleg@redhat.com>
Cc: Tejun Heo <tj@kernel.org>
Cc: Ingo Molnar <mingo@redhat.com>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Steven Rostedt <rostedt@goodmis.org>
Cc: "Paul E. McKenney" <paulmck@linux.vnet.ibm.com>
Cc: Josh Triplett <josh@joshtriplett.org>
Cc: Thomas Gleixner <tglx@linutronix.de>
Cc: Jiri Kosina <jkosina@suse.cz>
Cc: Borislav Petkov <bp@suse.de>
Cc: Michal Hocko <mhocko@suse.cz>
Cc: Vlastimil Babka <vbabka@suse.cz>
Signed-off-by: NArnd Bergmann <arnd@arndb.de>
Signed-off-by: NAndrew Morton <akpm@linux-foundation.org>
Signed-off-by: NLinus Torvalds <torvalds@linux-foundation.org>

All drivers use the default, so provide an inline version of it.  If we
ever need other queue mapping we can add an optional method back,
although supporting will also require major changes to the queue setup
code.

This provides better code generation, and better debugability as well.
Signed-off-by: NChristoph Hellwig <hch@lst.de>
Reviewed-by: NKeith Busch <keith.busch@intel.com>
Signed-off-by: NJens Axboe <axboe@fb.com>

Use a switch statement to iterate over the possible operations and
error out if it's an incorrect one.
Signed-off-by: NJens Axboe <axboe@fb.com>

Fix a fat-fingered conversion to the req_op accessors, and also
use a switch statement to make it more obvious what is being checked.
Signed-off-by: NChristoph Hellwig <hch@lst.de>
Reported-by: NDave Chinner <david@fromorbit.com>
Fixes: c2df40 ("drivers: use req op accessor");
Reviewed-by: NMing Lei <ming.lei@canonical.com>
Signed-off-by: NJens Axboe <axboe@fb.com>

There is no error number returned if loop driver fails in function
alloc_disk to add new loop device. Add a correct error number to make
user notify in this case.
Signed-off-by: NMinfei Huang <mnghuan@gmail.com>
Reviewed-by: NMing Lei <tom.leiming@gmail.com>
Reviewed-by: NJeff Moyer <jmoyer@redhat.com>
Signed-off-by: NJens Axboe <axboe@fb.com>

This adds a REQ_OP_FLUSH operation that is sent to request_fn
based drivers by the block layer's flush code, instead of
sending requests with the request->cmd_flags REQ_FLUSH bit set.
Signed-off-by: NMike Christie <mchristi@redhat.com>
Reviewed-by: NChristoph Hellwig <hch@lst.de>
Reviewed-by: NHannes Reinecke <hare@suse.com>
Signed-off-by: NJens Axboe <axboe@fb.com>

The req operation REQ_OP is separated from the rq_flag_bits
definition. This converts the block layer drivers to
use req_op to get the op from the request struct.
Signed-off-by: NMike Christie <mchristi@redhat.com>
Reviewed-by: NHannes Reinecke <hare@suse.com>
Signed-off-by: NJens Axboe <axboe@fb.com>

We currently set REQ_WRITE/WRITE for all non READ IOs
like discard, flush, writesame, etc. In the next patches where we
no longer set up the op as a bitmap, we will not be able to
detect a operation direction like writesame by testing if REQ_WRITE is
set.

This patch converts the drivers and cgroup to use the
op_is_write helper. This should just cover the simple
cases. I did dm, md and bcache in their own patches
because they were more involved.
Signed-off-by: NMike Christie <mchristi@redhat.com>
Reviewed-by: NHannes Reinecke <hare@suse.com>
Signed-off-by: NJens Axboe <axboe@fb.com>

Starting from commit e36f6204(block: split bios to max possible length),
block core starts to split bio in the middle of bvec.

Unfortunately loop dio/aio doesn't consider this situation, and
always treat 'iter.iov_offset' as zero. Then filesystem corruption
is observed.

This patch figures out the offset of the base bvevc via
'bio->bi_iter.bi_bvec_done' and fixes the issue by passing the offset
to iov iterator.

Fixes: e36f6204 (block: split bios to max possible length)
Cc: Keith Busch <keith.busch@intel.com>
Cc: Al Viro <viro@zeniv.linux.org.uk>
Cc: stable@vger.kernel.org (4.5)
Signed-off-by: NMing Lei <ming.lei@canonical.com>
Signed-off-by: NJens Axboe <axboe@fb.com>

Signed-off-by: NJens Axboe <axboe@fb.com>
Reviewed-by: NChristoph Hellwig <hch@lst.de>

blk_mq_complete_request may be a no-op if the request has already
been completed by others means (e.g. a timeout or cancellation), but
currently drivers have to set rq->errors before calling
blk_mq_complete_request, which might leave us with the wrong error value.

Add an error parameter to blk_mq_complete_request so that we can
defer setting rq->errors until we known we won the race to complete the
request.
Signed-off-by: NChristoph Hellwig <hch@lst.de>
Reviewed-by: NSagi Grimberg <sagig@mellanox.com>
Signed-off-by: NJens Axboe <axboe@fb.com>

There are at least 3 advantages to use direct I/O and AIO on
read/write loop's backing file:

1) double cache can be avoided, then memory usage gets
decreased a lot

2) not like user space direct I/O, there isn't cost of
pinning pages

3) avoid context switch for obtaining good throughput
- in buffered file read, random I/O top throughput is often obtained
only if they are submitted concurrently from lots of tasks; but for
sequential I/O, most of times they can be hit from page cache, so
concurrent submissions often introduce unnecessary context switch
and can't improve throughput much. There was such discussion[1]
to use non-blocking I/O to improve the problem for application.
- with direct I/O and AIO, concurrent submissions can be
avoided and random read throughput can't be affected meantime

xfstests(-g auto, ext4) is basically passed when running with
direct I/O(aio), one exception is generic/232, but it failed in
loop buffered I/O(4.2-rc6-next-20150814) too.

Follows the fio test result for performance purpose:
	4 jobs fio test inside ext4 file system over loop block

1) How to run
	- KVM: 4 VCPUs, 2G RAM
	- linux kernel: 4.2-rc6-next-20150814(base) with the patchset
	- the loop block is over one image on SSD.
	- linux psync, 4 jobs, size 1500M, ext4 over loop block
	- test result: IOPS from fio output

2) Throughput(IOPS) becomes a bit better with direct I/O(aio)
        -------------------------------------------------------------
        test cases          |randread   |read   |randwrite  |write  |
        -------------------------------------------------------------
        base                |8015       |113811 |67442      |106978
        -------------------------------------------------------------
        base+loop aio       |8136       |125040 |67811      |111376
        -------------------------------------------------------------

- somehow, it should be caused by more page cache avaiable for
application or one extra page copy is avoided in case of direct I/O

3) context switch
        - context switch decreased by ~50% with loop direct I/O(aio)
	compared with loop buffered I/O(4.2-rc6-next-20150814)

4) memory usage from /proc/meminfo
        -------------------------------------------------------------
                                   | Buffers       | Cached
        -------------------------------------------------------------
        base                       | > 760MB       | ~950MB
        -------------------------------------------------------------
        base+loop direct I/O(aio)  | < 5MB         | ~1.6GB
        -------------------------------------------------------------

- so there are much more page caches available for application with
direct I/O

[1] https://lwn.net/Articles/612483/Signed-off-by: NMing Lei <ming.lei@canonical.com>
Reviewed-by: NChristoph Hellwig <hch@lst.de>
Signed-off-by: NJens Axboe <axboe@fb.com>

If loop block is mounted via 'mount -o loop', it isn't easy
to pass file descriptor opened as O_DIRECT, so this patch
introduces a new command to support direct IO for this case.

Cc: linux-api@vger.kernel.org
Signed-off-by: NMing Lei <ming.lei@canonical.com>
Reviewed-by: NChristoph Hellwig <hch@lst.de>
Signed-off-by: NJens Axboe <axboe@fb.com>

This patches provides one interface for enabling direct IO
from user space:

	- userspace(such as losetup) can pass 'file' which is
	opened/fcntl as O_DIRECT

Also __loop_update_dio() is introduced to check if direct I/O
can be used on current loop setting.

The last big change is to introduce LO_FLAGS_DIRECT_IO flag
for userspace to know if direct IO is used to access backing
file.

Cc: linux-api@vger.kernel.org
Signed-off-by: NMing Lei <ming.lei@canonical.com>
Reviewed-by: NChristoph Hellwig <hch@lst.de>
Signed-off-by: NJens Axboe <axboe@fb.com>

The following patch will use dio/aio to submit IO to backing file,
then it needn't to schedule IO concurrently from work, so
use kthread_work for decreasing context switch cost a lot.

For non-AIO case, single thread has been used for long long time,
and it was just converted to work in v4.0, which has caused performance
regression for fedora live booting already. In discussion[1], even
though submitting I/O via work concurrently can improve random read IO
throughput, meantime it might hurt sequential read IO performance, so
better to restore to single thread behaviour.

For the following AIO support, it is better to use multi hw-queue
with per-hwq kthread than current work approach suppose there is so
high performance requirement for loop.

[1] http://marc.info/?t=143082678400002&r=1&w=2Signed-off-by: NMing Lei <ming.lei@canonical.com>
Reviewed-by: NChristoph Hellwig <hch@lst.de>
Signed-off-by: NJens Axboe <axboe@fb.com>

It doesn't make sense to enable merge because the I/O
submitted to backing file is handled page by page.
Signed-off-by: NMing Lei <ming.lei@canonical.com>
Reviewed-by: NChristoph Hellwig <hch@lst.de>
Signed-off-by: NJens Axboe <axboe@fb.com>

Some drivers use it now, others just set the limits field manually.
But in preparation for splitting this into a hard and soft limit,
ensure that they all call the proper function for setting the hw
limit for discards.
Reviewed-by: NJeff Moyer <jmoyer@redhat.com>
Signed-off-by: NJens Axboe <axboe@fb.com>

Turn
	d_path(&file->f_path, ...);
into
	file_path(file, ...);
Signed-off-by: NMiklos Szeredi <mszeredi@suse.cz>
Signed-off-by: NAl Viro <viro@zeniv.linux.org.uk>

gcc, righfully, complains:

drivers/block/loop.c:1369:1: warning: label 'out' defined but not used [-Wunused-label]

Kill it.
Signed-off-by: NJens Axboe <axboe@fb.com>

loop_clr_fd() can be run piggyback with lo_release(), and
under this situation, reread partition may always fail because
bd_mutex has been held already.

This patch detects the situation by the reference count, and
call __blkdev_reread_part() to avoid acquiring the lock again.

In the meantime, this patch switches to new kernel APIs
of blkdev_reread_part() and __blkdev_reread_part().
Reviewed-by: NChristoph Hellwig <hch@lst.de>
Tested-by: NJarod Wilson <jarod@redhat.com>
Acked-by: NJarod Wilson <jarod@redhat.com>
Signed-off-by: NJarod Wilson <jarod@redhat.com>
Signed-off-by: NMing Lei <ming.lei@canonical.com>
Signed-off-by: NJens Axboe <axboe@fb.com>

The lo_ctl_mutex is held for running all ioctl handlers, and
in some ioctl handlers, ioctl_by_bdev(BLKRRPART) is called for
rereading partitions, which requires bd_mutex.

So it is easy to cause failure because trylock(bd_mutex) may
fail inside blkdev_reread_part(), and follows the lock context:

blkid or other application:
	->open()
		->mutex_lock(bd_mutex)
		->lo_open()
			->mutex_lock(lo_ctl_mutex)

losetup(set fd ioctl):
	->mutex_lock(lo_ctl_mutex)
	->ioctl_by_bdev(BLKRRPART)
		->trylock(bd_mutex)

This patch trys to eliminate the ABBA lock dependency by removing
lo_ctl_mutext in lo_open() with the following approach:

1) make lo_refcnt as atomic_t and avoid acquiring lo_ctl_mutex in lo_open():
	- for open vs. add/del loop, no any problem because of loop_index_mutex
	- freeze request queue during clr_fd, so I/O can't come until
	  clearing fd is completed, like the effect of holding lo_ctl_mutex
	  in lo_open
	- both open() and release() have been serialized by bd_mutex already

2) don't hold lo_ctl_mutex for decreasing/checking lo_refcnt in
lo_release(), then lo_ctl_mutex is only required for the last release.
Reviewed-by: NChristoph Hellwig <hch@lst.de>
Tested-by: NJarod Wilson <jarod@redhat.com>
Acked-by: NJarod Wilson <jarod@redhat.com>
Signed-off-by: NMing Lei <ming.lei@canonical.com>
Signed-off-by: NJens Axboe <axboe@fb.com>

If there are too many pending per work I/O, too many
high priority work thread can be generated so that
system performance can be effected.

This patch limits the max_active parameter of workqueue as 16.

This patch fixes Fedora 22 live booting performance
regression when it is booted from squashfs over dm
based on loop, and looks the following reasons are
related with the problem:

- not like other filesyststems(such as ext4), squashfs
is a bit special, and I observed that increasing I/O jobs
to access file in squashfs only improve I/O performance a
little, but it can make big difference for ext4

- nested loop: both squashfs.img and ext3fs.img are mounted
as loop block, and ext3fs.img is inside the squashfs

- during booting, lots of tasks may run concurrently

Fixes: b5dd2f60
Cc: stable@vger.kernel.org (v4.0)
Cc: Justin M. Forbes <jforbes@fedoraproject.org>
Signed-off-by: NMing Lei <ming.lei@canonical.com>
Acked-by: NTejun Heo <tj@kernel.org>
Signed-off-by: NJens Axboe <axboe@fb.com>

Documentation/workqueue.txt:
	If there is dependency among multiple work items used
	during memory reclaim, they should be queued to separate
	wq each with WQ_MEM_RECLAIM.

Loop devices can be stacked, so we have to convert to per-device
workqueue. One example is Fedora live CD.

Fixes: b5dd2f60
Cc: stable@vger.kernel.org (v4.0)
Cc: Justin M. Forbes <jforbes@fedoraproject.org>
Signed-off-by: NMing Lei <ming.lei@canonical.com>
Acked-by: NTejun Heo <tj@kernel.org>
Signed-off-by: NJens Axboe <axboe@fb.com>

Because of the peculiar way that md devices are created (automatically
when the device node is opened), a new device can be created and
registered immediately after the
	blk_unregister_region(disk_devt(disk), disk->minors);
call in del_gendisk().

Therefore it is important that all visible artifacts of the previous
device are removed before this call.  In particular, the 'bdi'.

Since:
commit c4db59d3
Author: Christoph Hellwig <hch@lst.de>
    fs: don't reassign dirty inodes to default_backing_dev_info

moved the
   device_unregister(bdi->dev);
call from bdi_unregister() to bdi_destroy() it has been quite easy to
lose a race and have a new (e.g.) "md127" be created after the
blk_unregister_region() call and before bdi_destroy() is ultimately
called by the final 'put_disk', which must come after del_gendisk().

The new device finds that the bdi name is already registered in sysfs
and complains

> [ 9627.630029] WARNING: CPU: 18 PID: 3330 at fs/sysfs/dir.c:31 sysfs_warn_dup+0x5a/0x70()
> [ 9627.630032] sysfs: cannot create duplicate filename '/devices/virtual/bdi/9:127'

We can fix this by moving the bdi_destroy() call out of
blk_release_queue() (which can happen very late when a refcount
reaches zero) and into blk_cleanup_queue() - which happens exactly when the md
device driver calls it.

Then it is only necessary for md to call blk_cleanup_queue() before
del_gendisk().  As loop.c devices are also created on demand by
opening the device node, we make the same change there.

Fixes: c4db59d3Reported-by: NAzat Khuzhin <a3at.mail@gmail.com>
Cc: Christoph Hellwig <hch@lst.de>
Cc: stable@vger.kernel.org (v4.0)
Signed-off-by: NNeilBrown <neilb@suse.de>
Reviewed-by: NChristoph Hellwig <hch@lst.de>
Signed-off-by: NJens Axboe <axboe@fb.com>

Signed-off-by: NChristoph Hellwig <hch@lst.de>
Signed-off-by: NAl Viro <viro@zeniv.linux.org.uk>

all writable files that might be used as backing store for /dev/loop
already support ->write_iter()
Signed-off-by: NAl Viro <viro@zeniv.linux.org.uk>

Looks like we pull it in through other ways on x86, but we fail
on sparc:

In file included from drivers/block/cryptoloop.c:30:0:
drivers/block/loop.h:63:24: error: field 'tag_set' has incomplete type
struct blk_mq_tag_set tag_set;

Add the include to loop.h, kill it from loop.c.
Signed-off-by: NJens Axboe <axboe@fb.com>

block core handles REQ_FUA by its flush state machine, so
won't do it in loop explicitly.
Signed-off-by: NMing Lei <ming.lei@canonical.com>
Signed-off-by: NJens Axboe <axboe@fb.com>

No behaviour change, just move the handling for REQ_DISCARD
and REQ_FLUSH in these two functions.
Signed-off-by: NMing Lei <ming.lei@canonical.com>
Signed-off-by: NJens Axboe <axboe@fb.com>

Switch to block request completely.
Signed-off-by: NMing Lei <ming.lei@canonical.com>
Signed-off-by: NJens Axboe <axboe@fb.com>

The conversion is a bit straightforward, and use work queue to
dispatch requests of loop block, and one big change is that requests
is submitted to backend file/device concurrently with work queue,
so throughput may get improved much. Given write requests over same
file are often run exclusively, so don't handle them concurrently for
avoiding extra context switch cost, possible lock contention and work
schedule cost. Also with blk-mq, there is opportunity to get loop I/O
merged before submitting to backend file/device.

In the following test:
	- base: v3.19-rc2-2041231
	- loop over file in ext4 file system on SSD disk
	- bs: 4k, libaio, io depth: 64, O_DIRECT, num of jobs: 1
	- throughput: IOPS

	------------------------------------------------------
	|            | base      | base with loop-mq | delta |
	------------------------------------------------------
	| randread   | 1740      | 25318             | +1355%|
	------------------------------------------------------
	| read       | 42196     | 51771             | +22.6%|
	-----------------------------------------------------
	| randwrite  | 35709     | 34624             | -3%   |
	-----------------------------------------------------
	| write      | 39137     | 40326             | +3%   |
	-----------------------------------------------------

So loop-mq can improve throughput for both read and randread, meantime,
performance of write and randwrite isn't hurted basically.

Another benefit is that loop driver code gets simplified
much after blk-mq conversion, and the patch can be thought as
cleanup too.
Signed-off-by: NMing Lei <ming.lei@canonical.com>
Signed-off-by: NJens Axboe <axboe@fb.com>

Replace various -20/+19 hardcoded nice values with MIN_NICE/MAX_NICE.
Signed-off-by: NDongsheng Yang <yangds.fnst@cn.fujitsu.com>
Acked-by: NTejun Heo <tj@kernel.org>
Signed-off-by: NPeter Zijlstra <peterz@infradead.org>
Link: http://lkml.kernel.org/r/ff13819fd09b7a5dba5ab5ae797f2e7019bdfa17.1394532288.git.yangds.fnst@cn.fujitsu.com
Cc: devel@driverdev.osuosl.org
Cc: devicetree@vger.kernel.org
Cc: fcoe-devel@open-fcoe.org
Cc: linux390@de.ibm.com
Cc: linux-kernel@vger.kernel.org
Cc: linux-mm@kvack.org
Cc: linux-s390@vger.kernel.org
Cc: linux-scsi@vger.kernel.org
Cc: nbd-general@lists.sourceforge.net
Cc: ocfs2-devel@oss.oracle.com
Cc: openipmi-developer@lists.sourceforge.net
Cc: qla2xxx-upstream@qlogic.com
Cc: linux-arch@vger.kernel.org
[ Consolidated the patches, twiddled the changelog. ]
Signed-off-by: NIngo Molnar <mingo@kernel.org>