Record the byte count for an osd request rather than the page count.
The number of pages can always be derived from the byte count (and
alignment/offset) but the reverse is not true.
Signed-off-by: Alex Elder <elder@inktank.com>
Reviewed-by: Josh Durgin <josh.durgin@inktank.com>
An osd request defines information about where data to be read
should be placed as well as where data to write comes from.
Currently these are represented by common fields.
Keep information about data for writing separate from data to be
read by splitting these into data_in and data_out fields.
This is the key patch in this whole series, in that it actually
identifies which osd requests generate outgoing data and which
generate incoming data. It's less obvious (currently) that an osd
CALL op generates both outgoing and incoming data; that's the focus
of some upcoming work.
This resolves:
http://tracker.ceph.com/issues/4127
Signed-off-by: Alex Elder <elder@inktank.com>
Reviewed-by: Josh Durgin <josh.durgin@inktank.com>
An osd request uses either pages or a bio list for its data. Use a
union to record information about the two, and add a data type
tag to select between them.
Signed-off-by: Alex Elder <elder@inktank.com>
Reviewed-by: Josh Durgin <josh.durgin@inktank.com>
Pull the fields in an osd request structure that define the data for
the request out into a separate structure.
Signed-off-by: Alex Elder <elder@inktank.com>
Reviewed-by: Josh Durgin <josh.durgin@inktank.com>
Currently ceph_osdc_new_request() assigns an osd request's
r_num_pages and r_alignment fields. The only thing it does
after that is call ceph_osdc_build_request(), and that doesn't
need those fields to be assigned.
Move the assignment of those fields out of ceph_osdc_new_request()
and into its caller. As a result, the page_align parameter is no
longer used, so get rid of it.
Note that in ceph_sync_write(), the value for req->r_num_pages had
already been calculated earlier (as num_pages, and fortunately
it was computed the same way). So don't bother recomputing it,
but because it's not needed earlier, move that calculation after the
call to ceph_osdc_new_request(). Hold off making the assignment to
r_alignment, doing it instead r_pages and r_num_pages are
getting set.
Similarly, in start_read(), nr_pages already holds the number of
pages in the array (and is calculated the same way), so there's no
need to recompute it. Move the assignment of the page alignment
down with the others there as well.
This and the next few patches are preparation work for:
http://tracker.ceph.com/issues/4127
Signed-off-by: Alex Elder <elder@inktank.com>
Reviewed-by: Josh Durgin <josh.durgin@inktank.com>
(This is being reposted. The first one had a problem because it
erroneously added a similar change elsewhere; that change has been
dropped.)
The next patch in this series points out that the calculation for
the number of pages in an osd request is getting done twice. It
is not obvious, but the result of both calculations is identical.
This patch simplifies one of them--as a separate step--to make
it clear that the transformation in the next patch is valid.
In ceph_sync_write() there is some magic that computes page_align
for an osd request. But a little analysis shows it can be
simplified.
First, we have:
io_align = pos & ~PAGE_MASK;
which is used here:
page_align = (pos - io_align + buf_align) & ~PAGE_MASK;
Note (pos - io_align) simply rounds "pos" down to the nearest multiple
of the page size.
We also have:
buf_align = (unsigned long)data & ~PAGE_MASK;
Adding buf_align to that rounded-down "pos" value will stay within
the same page; the result will just be offset by the page offset for
the "data" pointer. The final mask therefore leaves just the value
of "buf_align".
One more simplification. Note that the result of calc_pages_for()
is invariant of which page the offset starts in--the only thing that
matters is the offset within the starting page. We will have
put the proper page offset to use into "page_align", so just use
that in calculating num_pages.
This resolves:
http://tracker.ceph.com/issues/4166
Signed-off-by: Alex Elder <elder@inktank.com>
Reviewed-by: Josh Durgin <josh.durgin@inktank.com>
make __ceph_do_pending_vmtruncate() acquire the i_mutex if the caller
does not hold the i_mutex, so ceph_aio_read() can call safely.
Signed-off-by: Yan, Zheng <zheng.z.yan@intel.com>
Reviewed-by: Greg Farnum <greg@inktank.com>
ceph_aio_write() has an optimization that marks CEPH_CAP_FILE_WR
cap dirty before data is copied to page cache and inode size is
updated. The optimization avoids slow cap revocation caused by
balance_dirty_pages(), but introduces inode size update race. If
ceph_check_caps() flushes the dirty cap before the inode size is
updated, MDS can miss the new inode size. So just remove the
optimization.
Signed-off-by: Yan, Zheng <zheng.z.yan@intel.com>
Reviewed-by: Greg Farnum <greg@inktank.com>
commit 22cddde104 breaks the atomicity of write operation, it also
introduces a deadlock between write and truncate.
Signed-off-by: Yan, Zheng <zheng.z.yan@intel.com>
Reviewed-by: Greg Farnum <greg@inktank.com>
Conflicts:
fs/ceph/addr.c
Pull Ceph updates from Sage Weil:
"A few groups of patches here. Alex has been hard at work improving
the RBD code, layout groundwork for understanding the new formats and
doing layering. Most of the infrastructure is now in place for the
final bits that will come with the next window.
There are a few changes to the data layout. Jim Schutt's patch fixes
some non-ideal CRUSH behavior, and a set of patches from me updates
the client to speak a newer version of the protocol and implement an
improved hashing strategy across storage nodes (when the server side
supports it too).
A pair of patches from Sam Lang fix the atomicity of open+create
operations. Several patches from Yan, Zheng fix various mds/client
issues that turned up during multi-mds torture tests.
A final set of patches expose file layouts via virtual xattrs, and
allow the policies to be set on directories via xattrs as well
(avoiding the awkward ioctl interface and providing a consistent
interface for both kernel mount and ceph-fuse users)."
* 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/sage/ceph-client: (143 commits)
libceph: add support for HASHPSPOOL pool flag
libceph: update osd request/reply encoding
libceph: calculate placement based on the internal data types
ceph: update support for PGID64, PGPOOL3, OSDENC protocol features
ceph: update "ceph_features.h"
libceph: decode into cpu-native ceph_pg type
libceph: rename ceph_pg -> ceph_pg_v1
rbd: pass length, not op for osd completions
rbd: move rbd_osd_trivial_callback()
libceph: use a do..while loop in con_work()
libceph: use a flag to indicate a fault has occurred
libceph: separate non-locked fault handling
libceph: encapsulate connection backoff
libceph: eliminate sparse warnings
ceph: eliminate sparse warnings in fs code
rbd: eliminate sparse warnings
libceph: define connection flag helpers
rbd: normalize dout() calls
rbd: barriers are hard
rbd: ignore zero-length requests
...
The "num_reply" parameter to ceph_osdc_new_request() is never
used inside that function, so get rid of it.
Note that ceph_sync_write() passes 2 for that argument, while all
other callers pass 1. It doesn't matter, but perhaps someone should
verify this doesn't indicate a problem.
Signed-off-by: Alex Elder <elder@inktank.com>
Reviewed-by: Josh Durgin <josh.durgin@inktank.com>
The mds now sends back a created inode if the create request
performed the create. If the file already existed, no inode is
returned in the reply. This allows ceph to set the created flag
in atomic_open so that permissions are properly checked in the case
that the file wasn't created by the create call to the mds.
To ensure compability with previous kernels, a feature for sending
back the inode in the create reply was added, so that the mds will
only send back the inode if the client indicates it supports the
feature.
Signed-off-by: Sam Lang <sam.lang@inktank.com>
Reviewed-by: Sage Weil <sage@inktank.com>
The error returned by ceph_mdsc_do_request includes errors sending the
request, errors on timeout, or any errors coming from the mds. If
ceph_mdsc_do_request returns an error, the reply struct will most likely
be bogus. We need to bail out and propogate the error instead of
overwriting it.
Signed-off-by: Sam Lang <sam.lang@inktank.com>
Reviewed-by: Sage Weil <sage@inktank.com>
Pull Ceph update from Sage Weil:
"There are a few different groups of commits here. The largest is
Alex's ongoing work to enable the coming RBD features (cloning,
striping). There is some cleanup in libceph that goes along with it.
Cyril and David have fixed some problems with NFS reexport (leaking
dentries and page locks), and there is a batch of patches from Yan
fixing problems with the fs client when running against a clustered
MDS. There are a few bug fixes mixed in for good measure, many of
which will be going to the stable trees once they're upstream.
My apologies for the late pull. There is still a gremlin in the rbd
map/unmap code and I was hoping to include the fix for that as well,
but we haven't been able to confirm the fix is correct yet; I'll send
that in a separate pull once it's nailed down."
* 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/sage/ceph-client: (68 commits)
rbd: get rid of rbd_{get,put}_dev()
libceph: register request before unregister linger
libceph: don't use rb_init_node() in ceph_osdc_alloc_request()
libceph: init event->node in ceph_osdc_create_event()
libceph: init osd->o_node in create_osd()
libceph: report connection fault with warning
libceph: socket can close in any connection state
rbd: don't use ENOTSUPP
rbd: remove linger unconditionally
rbd: get rid of RBD_MAX_SEG_NAME_LEN
libceph: avoid using freed osd in __kick_osd_requests()
ceph: don't reference req after put
rbd: do not allow remove of mounted-on image
libceph: Unlock unprocessed pages in start_read() error path
ceph: call handle_cap_grant() for cap import message
ceph: Fix __ceph_do_pending_vmtruncate
ceph: Don't add dirty inode to dirty list if caps is in migration
ceph: Fix infinite loop in __wake_requests
ceph: Don't update i_max_size when handling non-auth cap
bdi_register: add __printf verification, fix arg mismatch
...
But the kernel decided to call it "origin" instead. Fix most of the
sites.
Acked-by: Hugh Dickins <hughd@google.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
ceph_aio_write() has an optimization that marks cap EPH_CAP_FILE_WR
dirty before data is copied to page cache and inode size is updated.
If ceph_check_caps() flushes the dirty cap before the inode size is
updated, MDS can miss the new inode size. The fix is move
ceph_{get,put}_cap_refs() into ceph_write_{begin,end}() and call
__ceph_mark_dirty_caps() after inode size is updated.
Signed-off-by: Yan, Zheng <zheng.z.yan@intel.com>
Signed-off-by: Sage Weil <sage@inktank.com>
If we are creating an osd request and get an invalid layout, return
an EINVAL to the caller. We switch up the return to have an error
code instead of NULL implying -ENOMEM.
Signed-off-by: Sage Weil <sage@inktank.com>
Reviewed-by: Alex Elder <elder@inktank.com>
The initial ->atomic_open op was carried over from the old intent code,
which was incomplete and didn't really work. Replace it with a fresh
method. In particular:
* always attempt to do an atomic open+lookup, both for the create case
and for lookups of existing files.
* fix symlink handling by returning 1 to the VFS so that we can follow
the link to its destination. This fixes a longstanding ceph bug (#2392).
Signed-off-by: Sage Weil <sage@inktank.com>
Just pass struct file *. Methods are happier that way...
There's no need to return struct file * from finish_open() now,
so let it return int. Next: saner prototypes for parts in
namei.c
Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
Change of calling conventions:
old new
NULL 1
file 0
ERR_PTR(-ve) -ve
Caller *knows* that struct file *; no need to return it.
Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
... and let finish_open() report having opened the file via that sucker.
Next step: don't modify od->filp at all.
[AV: FILE_CREATE was already used by cifs; Miklos' fix folded]
Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
Add an ->atomic_open implementation which replaces the atomic lookup+open+create
operation implemented via ->lookup and ->create operations.
Signed-off-by: Miklos Szeredi <mszeredi@suse.cz>
CC: Sage Weil <sage@newdream.net>
Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
What was the purpose of this?
Signed-off-by: Miklos Szeredi <mszeredi@suse.cz>
CC: Sage Weil <sage@newdream.net>
Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
This was an ill-conceived feature that has been removed from Ceph. Do
this gracefully:
- reject attempts to specify a preferred_osd via the ioctl
- stop exposing this information via virtual xattrs
- always fill in -1 for requests, in case we talk to an older server
- don't calculate preferred_osd placements/pgids
Reviewed-by: Alex Elder <elder@inktank.com>
Signed-off-by: Sage Weil <sage@inktank.com>
Commit 06222e491e got the if wrong so that
it always evaluates as true. This is semantically harmless, but makes
SEEK_CUR and SEEK_SET needlessly query the server.
Rewrite the if to explicitly enumerate the cases we DO need a valid i_size
to make this code less fragile.
Reported-by: Roel Kluin <roel.kluin@gmail.com>
Signed-off-by: Sage Weil <sage@newdream.net>
We have been using i_lock to protect all kinds of data structures in the
ceph_inode_info struct, including lists of inodes that we need to iterate
over while avoiding races with inode destruction. That requires grabbing
a reference to the inode with the list lock protected, but igrab() now
takes i_lock to check the inode flags.
Changing the list lock ordering would be a painful process.
However, using a ceph-specific i_ceph_lock in the ceph inode instead of
i_lock is a simple mechanical change and avoids the ordering constraints
imposed by igrab().
Reported-by: Amon Ott <a.ott@m-privacy.de>
Signed-off-by: Sage Weil <sage@newdream.net>
* 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/sage/ceph-client: (23 commits)
ceph: document unlocked d_parent accesses
ceph: explicitly reference rename old_dentry parent dir in request
ceph: document locking for ceph_set_dentry_offset
ceph: avoid d_parent in ceph_dentry_hash; fix ceph_encode_fh() hashing bug
ceph: protect d_parent access in ceph_d_revalidate
ceph: protect access to d_parent
ceph: handle racing calls to ceph_init_dentry
ceph: set dir complete frag after adding capability
rbd: set blk_queue request sizes to object size
ceph: set up readahead size when rsize is not passed
rbd: cancel watch request when releasing the device
ceph: ignore lease mask
ceph: fix ceph_lookup_open intent usage
ceph: only link open operations to directory unsafe list if O_CREAT|O_TRUNC
ceph: fix bad parent_inode calc in ceph_lookup_open
ceph: avoid carrying Fw cap during write into page cache
libceph: don't time out osd requests that haven't been received
ceph: report f_bfree based on kb_avail rather than diffing.
ceph: only queue capsnap if caps are dirty
ceph: fix snap writeback when racing with writes
...
d_parent is protected by d_lock: use it when looking up a dentry's parent
directory inode. Also take a reference and drop it in the caller to avoid
a use-after-free.
Reported-by: Al Viro <viro@ZenIV.linux.org.uk>
Reviewed-by: Yehuda Sadeh <yehuda@hq.newdream.net>
Signed-off-by: Sage Weil <sage@newdream.net>
We weren't properly calling lookup_instantiate_filp when setting up the
lookup intent, which could lead to file leakage on errors. So:
- use separate helper for the hidden snapdir translation, immediately
following the mds request
- use ceph_finish_lookup for the final dentry/return value dance in the
exit path
- lookup_instantiate_filp on success
Reported-by: Al Viro <viro@ZenIV.linux.org.uk>
Reviewed-by: Yehuda Sadeh <yehuda@hq.newdream.net>
Signed-off-by: Sage Weil <sage@newdream.net>
We only need to put these on the directory unsafe list if they have
side effects that fsync(2) should flush out.
Reviewed-by: Yehuda Sadeh <yehuda@hq.newdream.net>
Signed-off-by: Sage Weil <sage@newdream.net>
We were always getting NULL here because the intent file f_dentry is always
NULL at this point, which means we were always passing NULL to
ceph_mdsc_do_request. In reality, this was fine, since this isn't
currently ever a write operation that needs to get strung on the dir's
unsafe list.
Use the dir explicitly, and only pass it if this open has side-effects that
a dir fsync should flush.
Reviewed-by: Yehuda Sadeh <yehuda@hq.newdream.net>
Signed-off-by: Sage Weil <sage@newdream.net>
The generic_file_aio_write call may block on balance_dirty_pages while we
flush data to the OSDs. If we hold a reference to the FILE_WR cap during
that interval revocation by the MDS (e.g., to do a stat(2)) may be very
slow.
Reviewed-by: Yehuda Sadeh <yehuda@hq.newdream.net>
Signed-off-by: Sage Weil <sage@newdream.net>
This allows us to force IO through the sync path which you normally only
get when multiple clients are reading/writing to the same file or by
mounting with -o sync. Among other things, this lets test programs verify
correctness with a single mount.
Reviewed-by: Yehuda Sadeh <yehuda@hq.newdream.net>
Signed-off-by: Sage Weil <sage@newdream.net>
This converts everybody to handle SEEK_HOLE/SEEK_DATA properly. In some cases
we just return -EINVAL, in others we do the normal generic thing, and in others
we're simply making sure that the properly due-dilligence is done. For example
in NFS/CIFS we need to make sure the file size is update properly for the
SEEK_HOLE and SEEK_DATA case, but since it calls the generic llseek stuff itself
that is all we have to do. Thanks,
Signed-off-by: Josef Bacik <josef@redhat.com>
Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
We were iterating across stripe boundaries properly, but not moving the
write buffer pointer forward. This caused us to rewrite the same data
after the break. Fix by adjusting the data pointer forward, and
recalculating the io and buffer alignment after the break.
Signed-off-by: Sage Weil <sage@newdream.net>
Getting ENOENT is equivalent to reading 0 bytes. Make that correction
before setting up the hit_stripe and was_short flags.
Fixes the following case:
dd if=/dev/zero of=/mnt/fs_depot/dd3 bs=1 seek=1048576 count=0
dd if=/mnt/fs_depot/dd3 of=/root/ddout1 skip=8 bs=500 count=2 iflag=direct
Reported-by: Henry C Chang <henry.cy.chang@gmail.com>
Signed-off-by: Sage Weil <sage@newdream.net>
If we get a short read from the OSD because the object is small, we need to
zero the remainder of the buffer. For O_DIRECT reads, the attempted range
is not trimmed to i_size by the VFS, so we were actually looping
indefinitely.
Fix by trimming by i_size, and the unconditionally zeroing the trailing
range.
Reported-by: Jeff Wu <cpwu@tnsoft.com.cn>
Signed-off-by: Sage Weil <sage@newdream.net>
We should use ihold whenever we already have a stable inode ref, even
when we aren't holding i_lock. This avoids adding new and unnecessary
locking dependencies.
Signed-off-by: Sage Weil <sage@newdream.net>
The __mark_dirty_inode helper now takes i_lock as of 250df6ed. Fix the
one ceph callers that held i_lock (__ceph_mark_dirty_caps) to return the
flags value so that the callers can do it outside of i_lock.
Signed-off-by: Sage Weil <sage@newdream.net>
In sync_write_wait(), we assume that the newest request is at the
tail of unsafe write list. We should maintain the semantics here.
Signed-off-by: Henry C Chang <henry_c_chang@tcloudcomputing.com>
Signed-off-by: Sage Weil <sage@newdream.net>
For read operation, we have to set the argument _write_ of get_user_pages
to 1 since we will write data to pages. Also, we need to SetPageDirty before
releasing these pages.
Signed-off-by: Henry C Chang <henry_c_chang@tcloudcomputing.com>
Signed-off-by: Sage Weil <sage@newdream.net>
The user buffer may be 512-byte aligned, not page-aligned. We were
assuming the buffer was page-aligned and only accounting for
non-page-aligned io offsets.
Signed-off-by: Henry C Chang <henry_c_chang@tcloudcomputing.com>
Signed-off-by: Sage Weil <sage@newdream.net>
We used to infer alignment of IOs within a page based on the file offset,
which assumed they matched. This broke with direct IO that was not aligned
to pages (e.g., 512-byte aligned IO). We were also trusting the alignment
specified in the OSD reply, which could have been adjusted by the server.
Explicitly specify the page alignment when setting up OSD IO requests.
Signed-off-by: Sage Weil <sage@newdream.net>
Normally when we open a file we already have a cap, and simply update the
wanted set. However, if we open a file for write, but don't have an auth
cap, that doesn't work; we need to open a new cap with the auth MDS. Only
reuse existing caps if we are opening for read or the existing cap is auth.
Signed-off-by: Sage Weil <sage@newdream.net>
