J.R. Okajima reports that the call to sync_inode() in nfs_wb_page() can
deadlock with other writeback flush calls. It boils down to the fact
that we cannot ever call writeback_single_inode() while holding a page
lock (even if we do set nr_to_write to zero) since another process may
already be waiting in the call to do_writepages(), and so will deny us
the I_SYNC lock.
Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
If we exit from nfs_commit_inode() without ensuring that the COMMIT rpc
call has been completed, we must re-mark the inode as dirty. Otherwise,
future calls to sync_inode() with the WB_SYNC_ALL flag set will fail to
ensure that the data is on the disk.
Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
Commit 9c7e7e2337 (NFS: Don't call iput() in
nfs_access_cache_shrinker) unintentionally removed the spin unlock for the
inode->i_lock.
Reported-by: David Howells <dhowells@redhat.com>
Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
This changes O_DIRECT write code to mark extents as delalloc
while it is processing them. Yan Zheng has reworked the
enospc accounting based on tracking delalloc extents and
this makes it much easier to track enospc in the O_DIRECT code.
There are a few space cases with the O_DIRECT code though,
it only sets the EXTENT_DELALLOC bits, instead of doing
EXTENT_DELALLOC | EXTENT_DIRTY | EXTENT_UPTODATE, because
we don't want to mess with clearing the dirty and uptodate
bits when things go wrong. This is important because there
are no pages in the page cache, so any extent state structs
that we put in the tree won't get freed by releasepage. We have
to clear them ourselves as the DIO ends.
With this commit, we reserve space at in btrfs_file_aio_write,
and then as each btrfs_direct_IO call progresses it sets
EXTENT_DELALLOC on the range.
btrfs_get_blocks_direct is responsible for clearing the delalloc
at the same time it drops the extent lock.
Signed-off-by: Chris Mason <chris.mason@oracle.com>
This adds:
alias: devname:<name>
to some common kernel modules, which will allow the on-demand loading
of the kernel module when the device node is accessed.
Ideally all these modules would be compiled-in, but distros seems too
much in love with their modularization that we need to cover the common
cases with this new facility. It will allow us to remove a bunch of pretty
useless init scripts and modprobes from init scripts.
The static device node aliases will be carried in the module itself. The
program depmod will extract this information to a file in the module directory:
$ cat /lib/modules/2.6.34-00650-g537b60d-dirty/modules.devname
# Device nodes to trigger on-demand module loading.
microcode cpu/microcode c10:184
fuse fuse c10:229
ppp_generic ppp c108:0
tun net/tun c10:200
dm_mod mapper/control c10:235
Udev will pick up the depmod created file on startup and create all the
static device nodes which the kernel modules specify, so that these modules
get automatically loaded when the device node is accessed:
$ /sbin/udevd --debug
...
static_dev_create_from_modules: mknod '/dev/cpu/microcode' c10:184
static_dev_create_from_modules: mknod '/dev/fuse' c10:229
static_dev_create_from_modules: mknod '/dev/ppp' c108:0
static_dev_create_from_modules: mknod '/dev/net/tun' c10:200
static_dev_create_from_modules: mknod '/dev/mapper/control' c10:235
udev_rules_apply_static_dev_perms: chmod '/dev/net/tun' 0666
udev_rules_apply_static_dev_perms: chmod '/dev/fuse' 0666
A few device nodes are switched to statically allocated numbers, to allow
the static nodes to work. This might also useful for systems which still run
a plain static /dev, which is completely unsafe to use with any dynamic minor
numbers.
Note:
The devname aliases must be limited to the *common* and *single*instance*
device nodes, like the misc devices, and never be used for conceptually limited
systems like the loop devices, which should rather get fixed properly and get a
control node for losetup to talk to, instead of creating a random number of
device nodes in advance, regardless if they are ever used.
This facility is to hide the mess distros are creating with too modualized
kernels, and just to hide that these modules are not compiled-in, and not to
paper-over broken concepts. Thanks! :)
Cc: Greg Kroah-Hartman <gregkh@suse.de>
Cc: David S. Miller <davem@davemloft.net>
Cc: Miklos Szeredi <miklos@szeredi.hu>
Cc: Chris Mason <chris.mason@oracle.com>
Cc: Alasdair G Kergon <agk@redhat.com>
Cc: Tigran Aivazian <tigran@aivazian.fsnet.co.uk>
Cc: Ian Kent <raven@themaw.net>
Signed-Off-By: Kay Sievers <kay.sievers@vrfy.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
* git://git.kernel.org/pub/scm/linux/kernel/git/steve/gfs2-2.6-fixes:
GFS2: Fix permissions checking for setflags ioctl()
GFS2: Don't "get" xattrs for ACLs when ACLs are turned off
GFS2: Rework reclaiming unlinked dinodes
* 'for-linus' of git://oss.sgi.com/xfs/xfs:
xfs: Ensure inode allocation buffers are fully replayed
xfs: enable background pushing of the CIL
xfs: forced unmounts need to push the CIL
xfs: Introduce delayed logging core code
xfs: Delayed logging design documentation
xfs: Improve scalability of busy extent tracking
xfs: make the log ticket ID available outside the log infrastructure
xfs: clean up log ticket overrun debug output
xfs: Clean up XFS_BLI_* flag namespace
xfs: modify buffer item reference counting
xfs: allow log ticket allocation to take allocation flags
xfs: Don't reuse the same transaction ID for duplicated transactions.
Quote from Nick piggin's about btrfs patch
- http://www.mail-archive.com/linux-btrfs@vger.kernel.org/msg04472.html.
"add_to_page_cache_lru is exported, so it should be used. Benefits over
using a private pagevec: neater code, 128 bytes fewer stack used, percpu
lru ordering is preserved, and finally don't need to flush pagevec
before returning so batching may be shared with other LRU insertions."
Let's use it instead of private pagevec in ntfs, too.
Signed-off-by: Minchan Kim <minchan.kim@gmail.com>
Acked-by: Anton Altaparmakov <aia21@cantab.net>
Cc: Nick Piggin <nickpiggin@yahoo.com.au>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
cached_page and lru_pvec have not been used. Let's remove the arguments.
Signed-off-by: Minchan Kim <minchan.kim@gmail.com>
Acked-by: Anton Altaparmakov <aia21@cantab.net>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
- C99 knows about USHRT_MAX/SHRT_MAX/SHRT_MIN, not
USHORT_MAX/SHORT_MAX/SHORT_MIN.
- Make SHRT_MIN of type s16, not int, for consistency.
[akpm@linux-foundation.org: fix drivers/dma/timb_dma.c]
[akpm@linux-foundation.org: fix security/keys/keyring.c]
Signed-off-by: Alexey Dobriyan <adobriyan@gmail.com>
Acked-by: WANG Cong <xiyou.wangcong@gmail.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
When wb_writeback() hasn't written anything it will re-acquire the inode
lock before calling inode_wait_for_writeback.
This change tests the sync bit first so that is doesn't need to drop &
re-acquire the lock if the inode became available while wb_writeback() was
waiting to get the lock.
Signed-off-by: Richard Kennedy <richard@rsk.demon.co.uk>
Cc: Alexander Viro <viro@zeniv.linux.org.uk>
Cc: Jens Axboe <jens.axboe@oracle.com>
Cc: Wu Fengguang <fengguang.wu@intel.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Page migration requires rmap to be able to find all ptes mapping a page
at all times, otherwise the migration entry can be instantiated, but it
is possible to leave one behind if the second rmap_walk fails to find
the page. If this page is later faulted, migration_entry_to_page() will
call BUG because the page is locked indicating the page was migrated by
the migration PTE not cleaned up. For example
kernel BUG at include/linux/swapops.h:105!
invalid opcode: 0000 [#1] PREEMPT SMP
...
Call Trace:
[<ffffffff810e951a>] handle_mm_fault+0x3f8/0x76a
[<ffffffff8130c7a2>] do_page_fault+0x44a/0x46e
[<ffffffff813099b5>] page_fault+0x25/0x30
[<ffffffff8114de33>] load_elf_binary+0x152a/0x192b
[<ffffffff8111329b>] search_binary_handler+0x173/0x313
[<ffffffff81114896>] do_execve+0x219/0x30a
[<ffffffff8100a5c6>] sys_execve+0x43/0x5e
[<ffffffff8100320a>] stub_execve+0x6a/0xc0
RIP [<ffffffff811094ff>] migration_entry_wait+0xc1/0x129
There is a race between shift_arg_pages and migration that triggers this
bug. A temporary stack is setup during exec and later moved. If
migration moves a page in the temporary stack and the VMA is then removed
before migration completes, the migration PTE may not be found leading to
a BUG when the stack is faulted.
This patch causes pages within the temporary stack during exec to be
skipped by migration. It does this by marking the VMA covering the
temporary stack with an otherwise impossible combination of VMA flags.
These flags are cleared when the temporary stack is moved to its final
location.
[kamezawa.hiroyu@jp.fujitsu.com: idea for having migration skip temporary stacks]
Signed-off-by: Mel Gorman <mel@csn.ul.ie>
Reviewed-by: KAMEZAWA Hiroyuki <kamezawa.hiroyu@jp.fujitsu.com>
Reviewed-by: Rik van Riel <riel@redhat.com>
Acked-by: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Minchan Kim <minchan.kim@gmail.com>
Cc: Christoph Lameter <cl@linux.com>
Cc: Andrea Arcangeli <aarcange@redhat.com>
Cc: Rik van Riel <riel@redhat.com>
Cc: Peter Zijlstra <peterz@infradead.org>
Reviewed-by: KOSAKI Motohiro <kosaki.motohiro@jp.fujitsu.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
If !CONFIG_HUGETLB_PAGE, pagemap_hugetlb_range() is never called. So put
it (and its calling function) into #ifdef block.
Signed-off-by: Naoya Horiguchi <n-horiguchi@ah.jp.nec.com>
Acked-by: Matt Mackall <mpm@selenic.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
The async helper threads offload crc work onto all the
CPUs, and make streaming writes much faster. This
changes the O_DIRECT write code to use them. The only
small complication was that we need to pass in the
logical offset in the file for each bio, because we can't
find it in the bio's pages.
Signed-off-by: Chris Mason <chris.mason@oracle.com>
Yan Zheng noticed two places we were doing a lot of work
without task->state set to TASK_RUNNING. This sets the state
properly after we get ready to sleep but decide not to.
Signed-off-by: Chris Mason <chris.mason@oracle.com>
In order for AIO to work, we need to implement aio_write. This patch converts
our btrfs_file_write to btrfs_aio_write. I've tested this with xfstests and
nothing broke, and the AIO stuff magically started working. Thanks,
Signed-off-by: Josef Bacik <josef@redhat.com>
Signed-off-by: Chris Mason <chris.mason@oracle.com>
This provides basic DIO support for reading and writing. It does not do the
work to recover from mismatching checksums, that will come later. A few design
changes have been made from Jim's code (sorry Jim!)
1) Use the generic direct-io code. Jim originally re-wrote all the generic DIO
code in order to account for all of BTRFS's oddities, but thanks to that work it
seems like the best bet is to just ignore compression and such and just opt to
fallback on buffered IO.
2) Fallback on buffered IO for compressed or inline extents. Jim's code did
it's own buffering to make dio with compressed extents work. Now we just
fallback onto normal buffered IO.
3) Use ordered extents for the writes so that all of the
lock_extent()
lookup_ordered()
type checks continue to work.
4) Do the lock_extent() lookup_ordered() loop in readpage so we don't race with
DIO writes.
I've tested this with fsx and everything works great. This patch depends on my
dio and filemap.c patches to work. Thanks,
Signed-off-by: Josef Bacik <josef@redhat.com>
Signed-off-by: Chris Mason <chris.mason@oracle.com>
Btrfs cannot handle having logically non-contiguous requests submitted. For
example if you have
Logical: [0-4095][HOLE][8192-12287]
Physical: [0-4095] [4096-8191]
Normally the DIO code would put these into the same BIO's. The problem is we
need to know exactly what offset is associated with what BIO so we can do our
checksumming and unlocking properly, so putting them in the same BIO doesn't
work. So add another check where we submit the current BIO if the physical
blocks are not contigous OR the logical blocks are not contiguous.
Signed-off-by: Josef Bacik <josef@redhat.com>
Signed-off-by: Chris Mason <chris.mason@oracle.com>
Because BTRFS can do RAID and such, we need our own submit hook so we can setup
the bio's in the correct fashion, and handle checksum errors properly. So there
are a few changes here
1) The submit_io hook. This is straightforward, just call this instead of
submit_bio.
2) Allow the fs to return -ENOTBLK for reads. Usually this has only worked for
writes, since writes can fallback onto buffered IO. But BTRFS needs the option
of falling back on buffered IO if it encounters a compressed extent, since we
need to read the entire extent in and decompress it. So if we get -ENOTBLK back
from get_block we'll return back and fallback on buffered just like the write
case.
I've tested these changes with fsx and everything seems to work. Thanks,
Signed-off-by: Josef Bacik <josef@redhat.com>
Signed-off-by: Chris Mason <chris.mason@oracle.com>
This patch adds metadata ENOSPC handling for the balance code.
It is consisted by following major changes:
1. Avoid COW tree leave in the phrase of merging tree.
2. Handle interaction with snapshot creation.
3. make the backref cache can live across transactions.
Signed-off-by: Yan Zheng <zheng.yan@oracle.com>
Signed-off-by: Chris Mason <chris.mason@oracle.com>
Pre-allocate space for data relocation. This can detect ENOPSC
condition caused by fragmentation of free space.
Signed-off-by: Yan Zheng <zheng.yan@oracle.com>
Signed-off-by: Chris Mason <chris.mason@oracle.com>
Previous patches make the allocater return -ENOSPC if there is no
unreserved free metadata space. This patch updates tree log code
and various other places to propagate/handle the ENOSPC error.
Signed-off-by: Yan Zheng <zheng.yan@oracle.com>
Signed-off-by: Chris Mason <chris.mason@oracle.com>
Reserve metadata space for extent tree, checksum tree and root tree
Signed-off-by: Yan Zheng <zheng.yan@oracle.com>
Signed-off-by: Chris Mason <chris.mason@oracle.com>
Introduce metadata reservation context for delayed allocation
and update various related functions.
This patch also introduces EXTENT_FIRST_DELALLOC control bit for
set/clear_extent_bit. It tells set/clear_bit_hook whether they
are processing the first extent_state with EXTENT_DELALLOC bit
set. This change is important if set/clear_extent_bit involves
multiple extent_state.
Signed-off-by: Yan Zheng <zheng.yan@oracle.com>
Signed-off-by: Chris Mason <chris.mason@oracle.com>
Besides simplify the code, this change makes sure all metadata
reservation for normal metadata operations are released after
committing transaction.
Changes since V1:
Add code that check if unlink and rmdir will free space.
Add ENOSPC handling for clone ioctl.
Signed-off-by: Yan Zheng <zheng.yan@oracle.com>
Signed-off-by: Chris Mason <chris.mason@oracle.com>
Introducing metadata reseravtion contexts has two major advantages.
First, it makes metadata reseravtion more traceable. Second, it can
reclaim freed space and re-add them to the itself after transaction
committed.
Besides add btrfs_block_rsv structure and related helper functions,
This patch contains following changes:
Move code that decides if freed tree block should be pinned into
btrfs_free_tree_block().
Make space accounting more accurate, mainly for handling read only
block groups.
Signed-off-by: Chris Mason <chris.mason@oracle.com>
All code in init_btrfs_i can be moved into btrfs_alloc_inode()
Signed-off-by: Yan Zheng <zheng.yan@oracle.com>
Signed-off-by: Chris Mason <chris.mason@oracle.com>
Shrink delayed allocation space in a synchronized manner is more
controllable than flushing all delay allocated space in an async
thread.
Signed-off-by: Yan Zheng <zheng.yan@oracle.com>
Signed-off-by: Chris Mason <chris.mason@oracle.com>
We already have fs_info->chunk_mutex to avoid concurrent
chunk creation.
Signed-off-by: Yan Zheng <zheng.yan@oracle.com>
Signed-off-by: Chris Mason <chris.mason@oracle.com>
The size of reserved space is stored in space_info. If block groups
of different raid types are linked to separate space_info, changing
allocation profile will corrupt reserved space accounting.
Signed-off-by: Yan Zheng <zheng.yan@oracle.com>
Signed-off-by: Chris Mason <chris.mason@oracle.com>
Allow userspace filesystem implementation to use splice() to read from
the fuse device.
The userspace filesystem can now transfer data coming from a WRITE
request to an arbitrary file descriptor (regular file, block device or
socket) without having to go through a userspace buffer.
The semantics of using splice() to read messages are:
1) with a single splice() call move the whole message from the fuse
device to a temporary pipe
2) read the header from the pipe and determine the message type
3a) if message is a WRITE then splice data from pipe to destination
3b) else read rest of message to userspace buffer
Signed-off-by: Miklos Szeredi <mszeredi@suse.cz>
When splicing buffers to the fuse device with SPLICE_F_MOVE, try to
move pages from the pipe buffer into the page cache. This allows
populating the fuse filesystem's cache without ever touching the page
contents, i.e. zero copy read capability.
The following steps are performed when trying to move a page into the
page cache:
- buf->ops->confirm() to make sure the new page is uptodate
- buf->ops->steal() to try to remove the new page from it's previous place
- remove_from_page_cache() on the old page
- add_to_page_cache_locked() on the new page
If any of the above steps fail (non fatally) then the code falls back
to copying the page. In particular ->steal() will fail if there are
external references (other than the page cache and the pipe buffer) to
the page.
Also since the remove_from_page_cache() + add_to_page_cache_locked()
are non-atomic it is possible that the page cache is repopulated in
between the two and add_to_page_cache_locked() will fail. This could
be fixed by creating a new atomic replace_page_cache_page() function.
fuse_readpages_end() needed to be reworked so it works even if
page->mapping is NULL for some or all pages which can happen if the
add_to_page_cache_locked() failed.
A number of sanity checks were added to make sure the stolen pages
don't have weird flags set, etc... These could be moved into generic
splice/steal code.
Signed-off-by: Miklos Szeredi <mszeredi@suse.cz>
Allow userspace filesystem implementation to use splice() to write to
the fuse device. The semantics of using splice() are:
1) buffer the message header and data in a temporary pipe
2) with a *single* splice() call move the message from the temporary pipe
to the fuse device
The READ reply message has the most interesting use for this, since
now the data from an arbitrary file descriptor (which could be a
regular file, a block device or a socket) can be tranferred into the
fuse device without having to go through a userspace buffer. It will
also allow zero copy moving of pages.
One caveat is that the protocol on the fuse device requires the length
of the whole message to be written into the header. But the length of
the data transferred into the temporary pipe may not be known in
advance. The current library implementation works around this by
using vmplice to write the header and modifying the header after
splicing the data into the pipe (error handling omitted):
struct fuse_out_header out;
iov.iov_base = &out;
iov.iov_len = sizeof(struct fuse_out_header);
vmsplice(pip[1], &iov, 1, 0);
len = splice(input_fd, input_offset, pip[1], NULL, len, 0);
/* retrospectively modify the header: */
out.len = len + sizeof(struct fuse_out_header);
splice(pip[0], NULL, fuse_chan_fd(req->ch), NULL, out.len, flags);
This works since vmsplice only saves a pointer to the data, it does
not copy the data itself.
Since pipes are currently limited to 16 pages and messages need to be
spliced atomically, the length of the data is limited to 15 pages (or
60kB for 4k pages).
Signed-off-by: Miklos Szeredi <mszeredi@suse.cz>
Acquire a page ref on pages in ->readpages() and release them when the
read has finished. Not acquiring a reference didn't seem to cause any
trouble since the page is locked and will not be kicked out of the
page cache during the read.
However the following patches will want to remove the page from the
cache so a separate ref is needed. Making the reference in req->pages
explicit also makes the code easier to understand.
Signed-off-by: Miklos Szeredi <mszeredi@suse.cz>
Replace uses of get_user_pages() with get_user_pages_fast(). It looks
nicer and should be faster in most cases.
Signed-off-by: Miklos Szeredi <mszeredi@suse.cz>
"map" isn't needed any more after: 0bd87182d3 "fuse: fix kunmap in
fuse_ioctl_copy_user"
Signed-off-by: Dan Carpenter <error27@gmail.com>
Signed-off-by: Miklos Szeredi <mszeredi@suse.cz>
mapping_gfp_mask() is not supposed to store allocation contex details,
only page location details. So mapping_gfp_mask should be applied to the
pagecache page allocation, wheras normal (kernel mapped) memory should be
used for surrounding allocations such as radix-tree nodes allocated by
add_to_page_cache. Context modifiers should be applied on a per-callsite
basis.
So change splice to follow this convention (which is followed in similar
code patterns in core code).
Signed-off-by: Nick Piggin <npiggin@suse.de>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Jens Axboe <jens.axboe@oracle.com>
Instead of requiring an exact number of pages as the argument and
return value, change the API to deal with number of bytes instead.
This also relaxes the requirement that the passed in size must
result in a power-of-2 page array size. Round up to the nearest
power-of-2 automatically and return the resulting size of the pipe
on success.
Signed-off-by: Jens Axboe <jens.axboe@oracle.com>
If the passed in size is larger than what has been set as the
system wide limit and the user is not root, we want to return
permission denied (not invalid value).
Signed-off-by: Jens Axboe <jens.axboe@oracle.com>
With delayed logging, we can get inode allocation buffers in the
same transaction inode unlink buffers. We don't currently mark inode
allocation buffers in the log, so inode unlink buffers take
precedence over allocation buffers.
The result is that when they are combined into the same checkpoint,
only the unlinked inode chain fields are replayed, resulting in
uninitialised inode buffers being detected when the next inode
modification is replayed.
To fix this, we need to ensure that we do not set the inode buffer
flag in the buffer log item format flags if the inode allocation has
not already hit the log. To avoid requiring a change to log
recovery, we really need to make this a modification that relies
only on in-memory sate.
We can do this by checking during buffer log formatting (while the
CIL cannot be flushed) if we are still in the same sequence when we
commit the unlink transaction as the inode allocation transaction.
If we are, then we do not add the inode buffer flag to the buffer
log format item flags. This means the entire buffer will be
replayed, not just the unlinked fields. We do this while
CIL flusheѕ are locked out to ensure that we don't race with the
sequence numbers changing and hence fail to put the inode buffer
flag in the buffer format flags when we really need to.
Signed-off-by: Dave Chinner <dchinner@redhat.com>
Reviewed-by: Christoph Hellwig <hch@lst.de>
Signed-off-by: Alex Elder <aelder@sgi.com>
If we let the CIL grow without bound, it will grow large enough to violate
recovery constraints (must be at least one complete transaction in the log at
all times) or take forever to write out through the log buffers. Hence we need
a check during asynchronous transactions as to whether the CIL needs to be
pushed.
We track the amount of log space the CIL consumes, so it is relatively simple
to limit it on a pure size basis. Make the limit the minimum of just under half
the log size (recovery constraint) or 8MB of log space (which is an awful lot
of metadata).
Signed-off-by: Dave Chinner <dchinner@redhat.com>
Reviewed-by: Christoph Hellwig <hch@lst.de>
Signed-off-by: Alex Elder <aelder@sgi.com>
If the filesystem is being shut down and the there is no log error,
the current code forces out the current log buffers. This code now needs
to push the CIL before it forces out the log buffers to acheive the same
result.
Signed-off-by: Dave Chinner <dchinner@redhat.com>
Reviewed-by: Christoph Hellwig <hch@lst.de>
Signed-off-by: Alex Elder <aelder@sgi.com>
The delayed logging code only changes in-memory structures and as
such can be enabled and disabled with a mount option. Add the mount
option and emit a warning that this is an experimental feature that
should not be used in production yet.
We also need infrastructure to track committed items that have not
yet been written to the log. This is what the Committed Item List
(CIL) is for.
The log item also needs to be extended to track the current log
vector, the associated memory buffer and it's location in the Commit
Item List. Extend the log item and log vector structures to enable
this tracking.
To maintain the current log format for transactions with delayed
logging, we need to introduce a checkpoint transaction and a context
for tracking each checkpoint from initiation to transaction
completion. This includes adding a log ticket for tracking space
log required/used by the context checkpoint.
To track all the changes we need an io vector array per log item,
rather than a single array for the entire transaction. Using the new
log vector structure for this requires two passes - the first to
allocate the log vector structures and chain them together, and the
second to fill them out. This log vector chain can then be passed
to the CIL for formatting, pinning and insertion into the CIL.
Formatting of the log vector chain is relatively simple - it's just
a loop over the iovecs on each log vector, but it is made slightly
more complex because we re-write the iovec after the copy to point
back at the memory buffer we just copied into.
This code also needs to pin log items. If the log item is not
already tracked in this checkpoint context, then it needs to be
pinned. Otherwise it is already pinned and we don't need to pin it
again.
The only other complexity is calculating the amount of new log space
the formatting has consumed. This needs to be accounted to the
transaction in progress, and the accounting is made more complex
becase we need also to steal space from it for log metadata in the
checkpoint transaction. Calculate all this at insert time and update
all the tickets, counters, etc correctly.
Once we've formatted all the log items in the transaction, attach
the busy extents to the checkpoint context so the busy extents live
until checkpoint completion and can be processed at that point in
time. Transactions can then be freed at this point in time.
Now we need to issue checkpoints - we are tracking the amount of log space
used by the items in the CIL, so we can trigger background checkpoints when the
space usage gets to a certain threshold. Otherwise, checkpoints need ot be
triggered when a log synchronisation point is reached - a log force event.
Because the log write code already handles chained log vectors, writing the
transaction is trivial, too. Construct a transaction header, add it
to the head of the chain and write it into the log, then issue a
commit record write. Then we can release the checkpoint log ticket
and attach the context to the log buffer so it can be called during
Io completion to complete the checkpoint.
We also need to allow for synchronising multiple in-flight
checkpoints. This is needed for two things - the first is to ensure
that checkpoint commit records appear in the log in the correct
sequence order (so they are replayed in the correct order). The
second is so that xfs_log_force_lsn() operates correctly and only
flushes and/or waits for the specific sequence it was provided with.
To do this we need a wait variable and a list tracking the
checkpoint commits in progress. We can walk this list and wait for
the checkpoints to change state or complete easily, an this provides
the necessary synchronisation for correct operation in both cases.
Signed-off-by: Dave Chinner <dchinner@redhat.com>
Reviewed-by: Christoph Hellwig <hch@lst.de>
Signed-off-by: Alex Elder <aelder@sgi.com>
When we free a metadata extent, we record it in the per-AG busy
extent array so that it is not re-used before the freeing
transaction hits the disk. This array is fixed size, so when it
overflows we make further allocation transactions synchronous
because we cannot track more freed extents until those transactions
hit the disk and are completed. Under heavy mixed allocation and
freeing workloads with large log buffers, we can overflow this array
quite easily.
Further, the array is sparsely populated, which means that inserts
need to search for a free slot, and array searches often have to
search many more slots that are actually used to check all the
busy extents. Quite inefficient, really.
To enable this aspect of extent freeing to scale better, we need
a structure that can grow dynamically. While in other areas of
XFS we have used radix trees, the extents being freed are at random
locations on disk so are better suited to being indexed by an rbtree.
So, use a per-AG rbtree indexed by block number to track busy
extents. This incures a memory allocation when marking an extent
busy, but should not occur too often in low memory situations. This
should scale to an arbitrary number of extents so should not be a
limitation for features such as in-memory aggregation of
transactions.
However, there are still situations where we can't avoid allocating
busy extents (such as allocation from the AGFL). To minimise the
overhead of such occurences, we need to avoid doing a synchronous
log force while holding the AGF locked to ensure that the previous
transactions are safely on disk before we use the extent. We can do
this by marking the transaction doing the allocation as synchronous
rather issuing a log force.
Because of the locking involved and the ordering of transactions,
the synchronous transaction provides the same guarantees as a
synchronous log force because it ensures that all the prior
transactions are already on disk when the synchronous transaction
hits the disk. i.e. it preserves the free->allocate order of the
extent correctly in recovery.
By doing this, we avoid holding the AGF locked while log writes are
in progress, hence reducing the length of time the lock is held and
therefore we increase the rate at which we can allocate and free
from the allocation group, thereby increasing overall throughput.
The only problem with this approach is that when a metadata buffer is
marked stale (e.g. a directory block is removed), then buffer remains
pinned and locked until the log goes to disk. The issue here is that
if that stale buffer is reallocated in a subsequent transaction, the
attempt to lock that buffer in the transaction will hang waiting
the log to go to disk to unlock and unpin the buffer. Hence if
someone tries to lock a pinned, stale, locked buffer we need to
push on the log to get it unlocked ASAP. Effectively we are trading
off a guaranteed log force for a much less common trigger for log
force to occur.
Ideally we should not reallocate busy extents. That is a much more
complex fix to the problem as it involves direct intervention in the
allocation btree searches in many places. This is left to a future
set of modifications.
Finally, now that we track busy extents in allocated memory, we
don't need the descriptors in the transaction structure to point to
them. We can replace the complex busy chunk infrastructure with a
simple linked list of busy extents. This allows us to remove a large
chunk of code, making the overall change a net reduction in code
size.
Signed-off-by: Dave Chinner <david@fromorbit.com>
Reviewed-by: Christoph Hellwig <hch@lst.de>
Signed-off-by: Alex Elder <aelder@sgi.com>
