forge/linux-pinenote
1
0
Fork 0
Code Issues Pull requests Projects Releases Packages Wiki Activity Actions
linux-pinenote/fs/ceph/file.c

810 lines
22 KiB
C
Raw Normal View History

ceph: factor out libceph from Ceph file system This factors out protocol and low-level storage parts of ceph into a separate libceph module living in net/ceph and include/linux/ceph. This is mostly a matter of moving files around. However, a few key pieces of the interface change as well: - ceph_client becomes ceph_fs_client and ceph_client, where the latter captures the mon and osd clients, and the fs_client gets the mds client and file system specific pieces. - Mount option parsing and debugfs setup is correspondingly broken into two pieces. - The mon client gets a generic handler callback for otherwise unknown messages (mds map, in this case). - The basic supported/required feature bits can be expanded (and are by ceph_fs_client). No functional change, aside from some subtle error handling cases that got cleaned up in the refactoring process. Signed-off-by: Sage Weil <sage@newdream.net>
2010-04-06 15:14:15 -07:00
#include <linux/ceph/ceph_debug.h>
ceph: file operations File open and close operations, and read and write methods that ensure we have obtained the proper capabilities from the MDS cluster before performing IO on a file. We take references on held capabilities for the duration of the read/write to avoid prematurely releasing them back to the MDS. We implement two main paths for read and write: one that is buffered (and uses generic_aio_{read,write}), and one that is fully synchronous and blocking (operating either on a __user pointer or, if O_DIRECT, directly on user pages). Signed-off-by: Sage Weil <sage@newdream.net>
2009-10-06 11:31:08 -07:00
ceph: factor out libceph from Ceph file system This factors out protocol and low-level storage parts of ceph into a separate libceph module living in net/ceph and include/linux/ceph. This is mostly a matter of moving files around. However, a few key pieces of the interface change as well: - ceph_client becomes ceph_fs_client and ceph_client, where the latter captures the mon and osd clients, and the fs_client gets the mds client and file system specific pieces. - Mount option parsing and debugfs setup is correspondingly broken into two pieces. - The mon client gets a generic handler callback for otherwise unknown messages (mds map, in this case). - The basic supported/required feature bits can be expanded (and are by ceph_fs_client). No functional change, aside from some subtle error handling cases that got cleaned up in the refactoring process. Signed-off-by: Sage Weil <sage@newdream.net>
2010-04-06 15:14:15 -07:00
#include <linux/module.h>
ceph: file operations File open and close operations, and read and write methods that ensure we have obtained the proper capabilities from the MDS cluster before performing IO on a file. We take references on held capabilities for the duration of the read/write to avoid prematurely releasing them back to the MDS. We implement two main paths for read and write: one that is buffered (and uses generic_aio_{read,write}), and one that is fully synchronous and blocking (operating either on a __user pointer or, if O_DIRECT, directly on user pages). Signed-off-by: Sage Weil <sage@newdream.net>
2009-10-06 11:31:08 -07:00
#include <linux/sched.h>
include cleanup: Update gfp.h and slab.h includes to prepare for breaking implicit slab.h inclusion from percpu.h percpu.h is included by sched.h and module.h and thus ends up being included when building most .c files. percpu.h includes slab.h which in turn includes gfp.h making everything defined by the two files universally available and complicating inclusion dependencies. percpu.h -> slab.h dependency is about to be removed. Prepare for this change by updating users of gfp and slab facilities include those headers directly instead of assuming availability. As this conversion needs to touch large number of source files, the following script is used as the basis of conversion. http://userweb.kernel.org/~tj/misc/slabh-sweep.py The script does the followings. * Scan files for gfp and slab usages and update includes such that only the necessary includes are there. ie. if only gfp is used, gfp.h, if slab is used, slab.h. * When the script inserts a new include, it looks at the include blocks and try to put the new include such that its order conforms to its surrounding. It's put in the include block which contains core kernel includes, in the same order that the rest are ordered - alphabetical, Christmas tree, rev-Xmas-tree or at the end if there doesn't seem to be any matching order. * If the script can't find a place to put a new include (mostly because the file doesn't have fitting include block), it prints out an error message indicating which .h file needs to be added to the file. The conversion was done in the following steps. 1. The initial automatic conversion of all .c files updated slightly over 4000 files, deleting around 700 includes and adding ~480 gfp.h and ~3000 slab.h inclusions. The script emitted errors for ~400 files. 2. Each error was manually checked. Some didn't need the inclusion, some needed manual addition while adding it to implementation .h or embedding .c file was more appropriate for others. This step added inclusions to around 150 files. 3. The script was run again and the output was compared to the edits from #2 to make sure no file was left behind. 4. Several build tests were done and a couple of problems were fixed. e.g. lib/decompress_*.c used malloc/free() wrappers around slab APIs requiring slab.h to be added manually. 5. The script was run on all .h files but without automatically editing them as sprinkling gfp.h and slab.h inclusions around .h files could easily lead to inclusion dependency hell. Most gfp.h inclusion directives were ignored as stuff from gfp.h was usually wildly available and often used in preprocessor macros. Each slab.h inclusion directive was examined and added manually as necessary. 6. percpu.h was updated not to include slab.h. 7. Build test were done on the following configurations and failures were fixed. CONFIG_GCOV_KERNEL was turned off for all tests (as my distributed build env didn't work with gcov compiles) and a few more options had to be turned off depending on archs to make things build (like ipr on powerpc/64 which failed due to missing writeq). * x86 and x86_64 UP and SMP allmodconfig and a custom test config. * powerpc and powerpc64 SMP allmodconfig * sparc and sparc64 SMP allmodconfig * ia64 SMP allmodconfig * s390 SMP allmodconfig * alpha SMP allmodconfig * um on x86_64 SMP allmodconfig 8. percpu.h modifications were reverted so that it could be applied as a separate patch and serve as bisection point. Given the fact that I had only a couple of failures from tests on step 6, I'm fairly confident about the coverage of this conversion patch. If there is a breakage, it's likely to be something in one of the arch headers which should be easily discoverable easily on most builds of the specific arch. Signed-off-by: Tejun Heo <tj@kernel.org> Guess-its-ok-by: Christoph Lameter <cl@linux-foundation.org> Cc: Ingo Molnar <mingo@redhat.com> Cc: Lee Schermerhorn <Lee.Schermerhorn@hp.com>
2010-03-24 17:04:11 +09:00
#include <linux/slab.h>
ceph: file operations File open and close operations, and read and write methods that ensure we have obtained the proper capabilities from the MDS cluster before performing IO on a file. We take references on held capabilities for the duration of the read/write to avoid prematurely releasing them back to the MDS. We implement two main paths for read and write: one that is buffered (and uses generic_aio_{read,write}), and one that is fully synchronous and blocking (operating either on a __user pointer or, if O_DIRECT, directly on user pages). Signed-off-by: Sage Weil <sage@newdream.net>
2009-10-06 11:31:08 -07:00
#include <linux/file.h>
#include <linux/namei.h>
#include <linux/writeback.h>
#include "super.h"
#include "mds_client.h"
/*
* Ceph file operations
*
* Implement basic open/close functionality, and implement
* read/write.
*
* We implement three modes of file I/O:
* - buffered uses the generic_file_aio_{read,write} helpers
*
* - synchronous is used when there is multi-client read/write
* sharing, avoids the page cache, and synchronously waits for an
* ack from the OSD.
*
* - direct io takes the variant of the sync path that references
* user pages directly.
*
* fsync() flushes and waits on dirty pages, but just queues metadata
* for writeback: since the MDS can recover size and mtime there is no
* need to wait for MDS acknowledgement.
*/
/*
* Prepare an open request. Preallocate ceph_cap to avoid an
* inopportune ENOMEM later.
*/
static struct ceph_mds_request *
prepare_open_request(struct super_block *sb, int flags, int create_mode)
{
ceph: factor out libceph from Ceph file system This factors out protocol and low-level storage parts of ceph into a separate libceph module living in net/ceph and include/linux/ceph. This is mostly a matter of moving files around. However, a few key pieces of the interface change as well: - ceph_client becomes ceph_fs_client and ceph_client, where the latter captures the mon and osd clients, and the fs_client gets the mds client and file system specific pieces. - Mount option parsing and debugfs setup is correspondingly broken into two pieces. - The mon client gets a generic handler callback for otherwise unknown messages (mds map, in this case). - The basic supported/required feature bits can be expanded (and are by ceph_fs_client). No functional change, aside from some subtle error handling cases that got cleaned up in the refactoring process. Signed-off-by: Sage Weil <sage@newdream.net>
2010-04-06 15:14:15 -07:00
struct ceph_fs_client *fsc = ceph_sb_to_client(sb);
struct ceph_mds_client *mdsc = fsc->mdsc;
ceph: file operations File open and close operations, and read and write methods that ensure we have obtained the proper capabilities from the MDS cluster before performing IO on a file. We take references on held capabilities for the duration of the read/write to avoid prematurely releasing them back to the MDS. We implement two main paths for read and write: one that is buffered (and uses generic_aio_{read,write}), and one that is fully synchronous and blocking (operating either on a __user pointer or, if O_DIRECT, directly on user pages). Signed-off-by: Sage Weil <sage@newdream.net>
2009-10-06 11:31:08 -07:00
struct ceph_mds_request *req;
int want_auth = USE_ANY_MDS;
int op = (flags & O_CREAT) ? CEPH_MDS_OP_CREATE : CEPH_MDS_OP_OPEN;
if (flags & (O_WRONLY|O_RDWR|O_CREAT|O_TRUNC))
want_auth = USE_AUTH_MDS;
req = ceph_mdsc_create_request(mdsc, op, want_auth);
if (IS_ERR(req))
goto out;
req->r_fmode = ceph_flags_to_mode(flags);
req->r_args.open.flags = cpu_to_le32(flags);
req->r_args.open.mode = cpu_to_le32(create_mode);
ceph: fix sparse endian warning Use the __le macro, even though for -1 it doesn't matter. Signed-off-by: Sage Weil <sage@newdream.net>
2009-11-04 11:40:05 -08:00
req->r_args.open.preferred = cpu_to_le32(-1);
ceph: file operations File open and close operations, and read and write methods that ensure we have obtained the proper capabilities from the MDS cluster before performing IO on a file. We take references on held capabilities for the duration of the read/write to avoid prematurely releasing them back to the MDS. We implement two main paths for read and write: one that is buffered (and uses generic_aio_{read,write}), and one that is fully synchronous and blocking (operating either on a __user pointer or, if O_DIRECT, directly on user pages). Signed-off-by: Sage Weil <sage@newdream.net>
2009-10-06 11:31:08 -07:00
out:
return req;
}
/*
* initialize private struct file data.
* if we fail, clean up by dropping fmode reference on the ceph_inode
*/
static int ceph_init_file(struct inode *inode, struct file *file, int fmode)
{
struct ceph_file_info *cf;
int ret = 0;
switch (inode->i_mode & S_IFMT) {
case S_IFREG:
case S_IFDIR:
dout("init_file %p %p 0%o (regular)\n", inode, file,
inode->i_mode);
cf = kmem_cache_alloc(ceph_file_cachep, GFP_NOFS | __GFP_ZERO);
if (cf == NULL) {
ceph_put_fmode(ceph_inode(inode), fmode); /* clean up */
return -ENOMEM;
}
cf->fmode = fmode;
cf->next_offset = 2;
file->private_data = cf;
BUG_ON(inode->i_fop->release != ceph_release);
break;
case S_IFLNK:
dout("init_file %p %p 0%o (symlink)\n", inode, file,
inode->i_mode);
ceph_put_fmode(ceph_inode(inode), fmode); /* clean up */
break;
default:
dout("init_file %p %p 0%o (special)\n", inode, file,
inode->i_mode);
/*
* we need to drop the open ref now, since we don't
* have .release set to ceph_release.
*/
ceph_put_fmode(ceph_inode(inode), fmode); /* clean up */
BUG_ON(inode->i_fop->release == ceph_release);
/* call the proper open fop */
ret = inode->i_fop->open(inode, file);
}
return ret;
}
/*
* If the filp already has private_data, that means the file was
* already opened by intent during lookup, and we do nothing.
*
* If we already have the requisite capabilities, we can satisfy
* the open request locally (no need to request new caps from the
* MDS). We do, however, need to inform the MDS (asynchronously)
* if our wanted caps set expands.
*/
int ceph_open(struct inode *inode, struct file *file)
{
struct ceph_inode_info *ci = ceph_inode(inode);
ceph: factor out libceph from Ceph file system This factors out protocol and low-level storage parts of ceph into a separate libceph module living in net/ceph and include/linux/ceph. This is mostly a matter of moving files around. However, a few key pieces of the interface change as well: - ceph_client becomes ceph_fs_client and ceph_client, where the latter captures the mon and osd clients, and the fs_client gets the mds client and file system specific pieces. - Mount option parsing and debugfs setup is correspondingly broken into two pieces. - The mon client gets a generic handler callback for otherwise unknown messages (mds map, in this case). - The basic supported/required feature bits can be expanded (and are by ceph_fs_client). No functional change, aside from some subtle error handling cases that got cleaned up in the refactoring process. Signed-off-by: Sage Weil <sage@newdream.net>
2010-04-06 15:14:15 -07:00
struct ceph_fs_client *fsc = ceph_sb_to_client(inode->i_sb);
struct ceph_mds_client *mdsc = fsc->mdsc;
ceph: file operations File open and close operations, and read and write methods that ensure we have obtained the proper capabilities from the MDS cluster before performing IO on a file. We take references on held capabilities for the duration of the read/write to avoid prematurely releasing them back to the MDS. We implement two main paths for read and write: one that is buffered (and uses generic_aio_{read,write}), and one that is fully synchronous and blocking (operating either on a __user pointer or, if O_DIRECT, directly on user pages). Signed-off-by: Sage Weil <sage@newdream.net>
2009-10-06 11:31:08 -07:00
struct ceph_mds_request *req;
struct ceph_file_info *cf = file->private_data;
struct inode *parent_inode = file->f_dentry->d_parent->d_inode;
int err;
int flags, fmode, wanted;
if (cf) {
dout("open file %p is already opened\n", file);
return 0;
}
/* filter out O_CREAT|O_EXCL; vfs did that already. yuck. */
flags = file->f_flags & ~(O_CREAT|O_EXCL);
if (S_ISDIR(inode->i_mode))
flags = O_DIRECTORY; /* mds likes to know */
dout("open inode %p ino %llx.%llx file %p flags %d (%d)\n", inode,
ceph_vinop(inode), file, flags, file->f_flags);
fmode = ceph_flags_to_mode(flags);
wanted = ceph_caps_for_mode(fmode);
/* snapped files are read-only */
if (ceph_snap(inode) != CEPH_NOSNAP && (file->f_mode & FMODE_WRITE))
return -EROFS;
/* trivially open snapdir */
if (ceph_snap(inode) == CEPH_SNAPDIR) {
spin_lock(&inode->i_lock);
__ceph_get_fmode(ci, fmode);
spin_unlock(&inode->i_lock);
return ceph_init_file(inode, file, fmode);
}
/*
ceph: fix open for write on clustered mds 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>
2010-11-07 09:07:15 -08:00
* No need to block if we have caps on the auth MDS (for
* write) or any MDS (for read). Update wanted set
ceph: file operations File open and close operations, and read and write methods that ensure we have obtained the proper capabilities from the MDS cluster before performing IO on a file. We take references on held capabilities for the duration of the read/write to avoid prematurely releasing them back to the MDS. We implement two main paths for read and write: one that is buffered (and uses generic_aio_{read,write}), and one that is fully synchronous and blocking (operating either on a __user pointer or, if O_DIRECT, directly on user pages). Signed-off-by: Sage Weil <sage@newdream.net>
2009-10-06 11:31:08 -07:00
* asynchronously.
*/
spin_lock(&inode->i_lock);
ceph: fix open for write on clustered mds 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>
2010-11-07 09:07:15 -08:00
if (__ceph_is_any_real_caps(ci) &&
(((fmode & CEPH_FILE_MODE_WR) == 0) || ci->i_auth_cap)) {
ceph: file operations File open and close operations, and read and write methods that ensure we have obtained the proper capabilities from the MDS cluster before performing IO on a file. We take references on held capabilities for the duration of the read/write to avoid prematurely releasing them back to the MDS. We implement two main paths for read and write: one that is buffered (and uses generic_aio_{read,write}), and one that is fully synchronous and blocking (operating either on a __user pointer or, if O_DIRECT, directly on user pages). Signed-off-by: Sage Weil <sage@newdream.net>
2009-10-06 11:31:08 -07:00
int mds_wanted = __ceph_caps_mds_wanted(ci);
int issued = __ceph_caps_issued(ci, NULL);
dout("open %p fmode %d want %s issued %s using existing\n",
inode, fmode, ceph_cap_string(wanted),
ceph_cap_string(issued));
__ceph_get_fmode(ci, fmode);
spin_unlock(&inode->i_lock);
/* adjust wanted? */
if ((issued & wanted) != wanted &&
(mds_wanted & wanted) != wanted &&
ceph_snap(inode) != CEPH_SNAPDIR)
ceph_check_caps(ci, 0, NULL);
return ceph_init_file(inode, file, fmode);
} else if (ceph_snap(inode) != CEPH_NOSNAP &&
(ci->i_snap_caps & wanted) == wanted) {
__ceph_get_fmode(ci, fmode);
spin_unlock(&inode->i_lock);
return ceph_init_file(inode, file, fmode);
}
spin_unlock(&inode->i_lock);
dout("open fmode %d wants %s\n", fmode, ceph_cap_string(wanted));
req = prepare_open_request(inode->i_sb, flags, 0);
if (IS_ERR(req)) {
err = PTR_ERR(req);
goto out;
}
req->r_inode = igrab(inode);
req->r_num_caps = 1;
err = ceph_mdsc_do_request(mdsc, parent_inode, req);
if (!err)
err = ceph_init_file(inode, file, req->r_fmode);
ceph_mdsc_put_request(req);
dout("open result=%d on %llx.%llx\n", err, ceph_vinop(inode));
out:
return err;
}
/*
* Do a lookup + open with a single request.
*
* If this succeeds, but some subsequent check in the vfs
* may_open() fails, the struct *file gets cleaned up (i.e.
* ceph_release gets called). So fear not!
*/
/*
* flags
* path_lookup_open -> LOOKUP_OPEN
* path_lookup_create -> LOOKUP_OPEN|LOOKUP_CREATE
*/
struct dentry *ceph_lookup_open(struct inode *dir, struct dentry *dentry,
struct nameidata *nd, int mode,
int locked_dir)
{
ceph: factor out libceph from Ceph file system This factors out protocol and low-level storage parts of ceph into a separate libceph module living in net/ceph and include/linux/ceph. This is mostly a matter of moving files around. However, a few key pieces of the interface change as well: - ceph_client becomes ceph_fs_client and ceph_client, where the latter captures the mon and osd clients, and the fs_client gets the mds client and file system specific pieces. - Mount option parsing and debugfs setup is correspondingly broken into two pieces. - The mon client gets a generic handler callback for otherwise unknown messages (mds map, in this case). - The basic supported/required feature bits can be expanded (and are by ceph_fs_client). No functional change, aside from some subtle error handling cases that got cleaned up in the refactoring process. Signed-off-by: Sage Weil <sage@newdream.net>
2010-04-06 15:14:15 -07:00
struct ceph_fs_client *fsc = ceph_sb_to_client(dir->i_sb);
struct ceph_mds_client *mdsc = fsc->mdsc;
ceph: file operations File open and close operations, and read and write methods that ensure we have obtained the proper capabilities from the MDS cluster before performing IO on a file. We take references on held capabilities for the duration of the read/write to avoid prematurely releasing them back to the MDS. We implement two main paths for read and write: one that is buffered (and uses generic_aio_{read,write}), and one that is fully synchronous and blocking (operating either on a __user pointer or, if O_DIRECT, directly on user pages). Signed-off-by: Sage Weil <sage@newdream.net>
2009-10-06 11:31:08 -07:00
struct file *file = nd->intent.open.file;
struct inode *parent_inode = get_dentry_parent_inode(file->f_dentry);
struct ceph_mds_request *req;
int err;
int flags = nd->intent.open.flags - 1; /* silly vfs! */
dout("ceph_lookup_open dentry %p '%.*s' flags %d mode 0%o\n",
dentry, dentry->d_name.len, dentry->d_name.name, flags, mode);
/* do the open */
req = prepare_open_request(dir->i_sb, flags, mode);
if (IS_ERR(req))
fs/ceph: Use ERR_CAST Use ERR_CAST(x) rather than ERR_PTR(PTR_ERR(x)). The former makes more clear what is the purpose of the operation, which otherwise looks like a no-op. In the case of fs/ceph/inode.c, ERR_CAST is not needed, because the type of the returned value is the same as the type of the enclosing function. The semantic patch that makes this change is as follows: (http://coccinelle.lip6.fr/) // <smpl> @@ type T; T x; identifier f; @@ T f (...) { <+... - ERR_PTR(PTR_ERR(x)) + x ...+> } @@ expression x; @@ - ERR_PTR(PTR_ERR(x)) + ERR_CAST(x) // </smpl> Signed-off-by: Julia Lawall <julia@diku.dk> Signed-off-by: Sage Weil <sage@newdream.net>
2010-05-22 12:01:14 +02:00
return ERR_CAST(req);
ceph: file operations File open and close operations, and read and write methods that ensure we have obtained the proper capabilities from the MDS cluster before performing IO on a file. We take references on held capabilities for the duration of the read/write to avoid prematurely releasing them back to the MDS. We implement two main paths for read and write: one that is buffered (and uses generic_aio_{read,write}), and one that is fully synchronous and blocking (operating either on a __user pointer or, if O_DIRECT, directly on user pages). Signed-off-by: Sage Weil <sage@newdream.net>
2009-10-06 11:31:08 -07:00
req->r_dentry = dget(dentry);
req->r_num_caps = 2;
if (flags & O_CREAT) {
req->r_dentry_drop = CEPH_CAP_FILE_SHARED;
req->r_dentry_unless = CEPH_CAP_FILE_EXCL;
}
req->r_locked_dir = dir; /* caller holds dir->i_mutex */
err = ceph_mdsc_do_request(mdsc, parent_inode, req);
dentry = ceph_finish_lookup(req, dentry, err);
if (!err && (flags & O_CREAT) && !req->r_reply_info.head->is_dentry)
err = ceph_handle_notrace_create(dir, dentry);
if (!err)
err = ceph_init_file(req->r_dentry->d_inode, file,
req->r_fmode);
ceph_mdsc_put_request(req);
dout("ceph_lookup_open result=%p\n", dentry);
return dentry;
}
int ceph_release(struct inode *inode, struct file *file)
{
struct ceph_inode_info *ci = ceph_inode(inode);
struct ceph_file_info *cf = file->private_data;
dout("release inode %p file %p\n", inode, file);
ceph_put_fmode(ci, cf->fmode);
if (cf->last_readdir)
ceph_mdsc_put_request(cf->last_readdir);
kfree(cf->last_name);
kfree(cf->dir_info);
dput(cf->dentry);
kmem_cache_free(ceph_file_cachep, cf);
ceph: return EBADF if waiting for caps on closed file Verify the file is actually open for the given caps when we are waiting for caps. This ensures we will wake up and return EBADF if another thread closes the file out from under us. Note that EBADF is also the correct return code from write(2) when called on a file handle opened for reading (although the vfs should catch that). Signed-off-by: Sage Weil <sage@newdream.net>
2010-03-01 09:57:54 -08:00
/* wake up anyone waiting for caps on this inode */
ceph: use complete_all and wake_up_all This fixes an issue triggered by running concurrent syncs. One of the syncs would go through while the other would just hang indefinitely. In any case, we never actually want to wake a single waiter, so the *_all functions should be used. Signed-off-by: Yehuda Sadeh <yehuda@hq.newdream.net> Signed-off-by: Sage Weil <sage@newdream.net>
2010-07-27 13:11:08 -07:00
wake_up_all(&ci->i_cap_wq);
ceph: file operations File open and close operations, and read and write methods that ensure we have obtained the proper capabilities from the MDS cluster before performing IO on a file. We take references on held capabilities for the duration of the read/write to avoid prematurely releasing them back to the MDS. We implement two main paths for read and write: one that is buffered (and uses generic_aio_{read,write}), and one that is fully synchronous and blocking (operating either on a __user pointer or, if O_DIRECT, directly on user pages). Signed-off-by: Sage Weil <sage@newdream.net>
2009-10-06 11:31:08 -07:00
return 0;
}
/*
* Read a range of bytes striped over one or more objects. Iterate over
* objects we stripe over. (That's not atomic, but good enough for now.)
*
* If we get a short result from the OSD, check against i_size; we need to
* only return a short read to the caller if we hit EOF.
*/
static int striped_read(struct inode *inode,
u64 off, u64 len,
ceph: fix sync read eof check deadlock If a sync read gets a short result from the OSD, it may need to do a getattr to see if it is short due to reaching end-of-file. The getattr was being done while holding a reference to FILE_RD, which can lead to a deadlock if the MDS is revoking that capability bit and can't process the getattr until it does. We fix this by setting a flag if EOF size validation is needed, and doing the getattr in ceph_aio_read, after the RD cap ref is dropped. If the read needs to be continued, we loop and continue traversing the file. Signed-off-by: Sage Weil <sage@newdream.net>
2010-02-09 14:04:02 -08:00
struct page **pages, int num_pages,
ceph: make page alignment explicit in osd interface 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>
2010-11-09 12:43:12 -08:00
int *checkeof, bool align_to_pages)
ceph: file operations File open and close operations, and read and write methods that ensure we have obtained the proper capabilities from the MDS cluster before performing IO on a file. We take references on held capabilities for the duration of the read/write to avoid prematurely releasing them back to the MDS. We implement two main paths for read and write: one that is buffered (and uses generic_aio_{read,write}), and one that is fully synchronous and blocking (operating either on a __user pointer or, if O_DIRECT, directly on user pages). Signed-off-by: Sage Weil <sage@newdream.net>
2009-10-06 11:31:08 -07:00
{
ceph: factor out libceph from Ceph file system This factors out protocol and low-level storage parts of ceph into a separate libceph module living in net/ceph and include/linux/ceph. This is mostly a matter of moving files around. However, a few key pieces of the interface change as well: - ceph_client becomes ceph_fs_client and ceph_client, where the latter captures the mon and osd clients, and the fs_client gets the mds client and file system specific pieces. - Mount option parsing and debugfs setup is correspondingly broken into two pieces. - The mon client gets a generic handler callback for otherwise unknown messages (mds map, in this case). - The basic supported/required feature bits can be expanded (and are by ceph_fs_client). No functional change, aside from some subtle error handling cases that got cleaned up in the refactoring process. Signed-off-by: Sage Weil <sage@newdream.net>
2010-04-06 15:14:15 -07:00
struct ceph_fs_client *fsc = ceph_inode_to_client(inode);
ceph: file operations File open and close operations, and read and write methods that ensure we have obtained the proper capabilities from the MDS cluster before performing IO on a file. We take references on held capabilities for the duration of the read/write to avoid prematurely releasing them back to the MDS. We implement two main paths for read and write: one that is buffered (and uses generic_aio_{read,write}), and one that is fully synchronous and blocking (operating either on a __user pointer or, if O_DIRECT, directly on user pages). Signed-off-by: Sage Weil <sage@newdream.net>
2009-10-06 11:31:08 -07:00
struct ceph_inode_info *ci = ceph_inode(inode);
u64 pos, this_len;
ceph: make page alignment explicit in osd interface 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>
2010-11-09 12:43:12 -08:00
int io_align, page_align;
ceph: fix short synchronous reads Zeroing of holes was not done correctly: page_off was miscalculated and zeroing the tail didn't not adjust the 'read' value to include the zeroed portion. Signed-off-by: Yehuda Sadeh <yehuda@hq.newdream.net> Signed-off-by: Sage Weil <sage@newdream.net>
2010-02-04 13:41:41 -08:00
int page_off = off & ~PAGE_CACHE_MASK; /* first byte's offset in page */
ceph: file operations File open and close operations, and read and write methods that ensure we have obtained the proper capabilities from the MDS cluster before performing IO on a file. We take references on held capabilities for the duration of the read/write to avoid prematurely releasing them back to the MDS. We implement two main paths for read and write: one that is buffered (and uses generic_aio_{read,write}), and one that is fully synchronous and blocking (operating either on a __user pointer or, if O_DIRECT, directly on user pages). Signed-off-by: Sage Weil <sage@newdream.net>
2009-10-06 11:31:08 -07:00
int left, pages_left;
int read;
struct page **page_pos;
int ret;
bool hit_stripe, was_short;
/*
* we may need to do multiple reads. not atomic, unfortunately.
*/
pos = off;
left = len;
page_pos = pages;
pages_left = num_pages;
read = 0;
ceph: make page alignment explicit in osd interface 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>
2010-11-09 12:43:12 -08:00
io_align = off & ~PAGE_MASK;
ceph: file operations File open and close operations, and read and write methods that ensure we have obtained the proper capabilities from the MDS cluster before performing IO on a file. We take references on held capabilities for the duration of the read/write to avoid prematurely releasing them back to the MDS. We implement two main paths for read and write: one that is buffered (and uses generic_aio_{read,write}), and one that is fully synchronous and blocking (operating either on a __user pointer or, if O_DIRECT, directly on user pages). Signed-off-by: Sage Weil <sage@newdream.net>
2009-10-06 11:31:08 -07:00
more:
ceph: make page alignment explicit in osd interface 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>
2010-11-09 12:43:12 -08:00
if (align_to_pages)
page_align = (pos - io_align) & ~PAGE_MASK;
else
page_align = pos & ~PAGE_MASK;
ceph: file operations File open and close operations, and read and write methods that ensure we have obtained the proper capabilities from the MDS cluster before performing IO on a file. We take references on held capabilities for the duration of the read/write to avoid prematurely releasing them back to the MDS. We implement two main paths for read and write: one that is buffered (and uses generic_aio_{read,write}), and one that is fully synchronous and blocking (operating either on a __user pointer or, if O_DIRECT, directly on user pages). Signed-off-by: Sage Weil <sage@newdream.net>
2009-10-06 11:31:08 -07:00
this_len = left;
ceph: factor out libceph from Ceph file system This factors out protocol and low-level storage parts of ceph into a separate libceph module living in net/ceph and include/linux/ceph. This is mostly a matter of moving files around. However, a few key pieces of the interface change as well: - ceph_client becomes ceph_fs_client and ceph_client, where the latter captures the mon and osd clients, and the fs_client gets the mds client and file system specific pieces. - Mount option parsing and debugfs setup is correspondingly broken into two pieces. - The mon client gets a generic handler callback for otherwise unknown messages (mds map, in this case). - The basic supported/required feature bits can be expanded (and are by ceph_fs_client). No functional change, aside from some subtle error handling cases that got cleaned up in the refactoring process. Signed-off-by: Sage Weil <sage@newdream.net>
2010-04-06 15:14:15 -07:00
ret = ceph_osdc_readpages(&fsc->client->osdc, ceph_vino(inode),
ceph: file operations File open and close operations, and read and write methods that ensure we have obtained the proper capabilities from the MDS cluster before performing IO on a file. We take references on held capabilities for the duration of the read/write to avoid prematurely releasing them back to the MDS. We implement two main paths for read and write: one that is buffered (and uses generic_aio_{read,write}), and one that is fully synchronous and blocking (operating either on a __user pointer or, if O_DIRECT, directly on user pages). Signed-off-by: Sage Weil <sage@newdream.net>
2009-10-06 11:31:08 -07:00
&ci->i_layout, pos, &this_len,
ci->i_truncate_seq,
ci->i_truncate_size,
ceph: make page alignment explicit in osd interface 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>
2010-11-09 12:43:12 -08:00
page_pos, pages_left, page_align);
ceph: file operations File open and close operations, and read and write methods that ensure we have obtained the proper capabilities from the MDS cluster before performing IO on a file. We take references on held capabilities for the duration of the read/write to avoid prematurely releasing them back to the MDS. We implement two main paths for read and write: one that is buffered (and uses generic_aio_{read,write}), and one that is fully synchronous and blocking (operating either on a __user pointer or, if O_DIRECT, directly on user pages). Signed-off-by: Sage Weil <sage@newdream.net>
2009-10-06 11:31:08 -07:00
hit_stripe = this_len < left;
was_short = ret >= 0 && ret < this_len;
if (ret == -ENOENT)
ret = 0;
dout("striped_read %llu~%u (read %u) got %d%s%s\n", pos, left, read,
ret, hit_stripe ? " HITSTRIPE" : "", was_short ? " SHORT" : "");
if (ret > 0) {
int didpages =
((pos & ~PAGE_CACHE_MASK) + ret) >> PAGE_CACHE_SHIFT;
if (read < pos - off) {
dout(" zero gap %llu to %llu\n", off + read, pos);
ceph: factor out libceph from Ceph file system This factors out protocol and low-level storage parts of ceph into a separate libceph module living in net/ceph and include/linux/ceph. This is mostly a matter of moving files around. However, a few key pieces of the interface change as well: - ceph_client becomes ceph_fs_client and ceph_client, where the latter captures the mon and osd clients, and the fs_client gets the mds client and file system specific pieces. - Mount option parsing and debugfs setup is correspondingly broken into two pieces. - The mon client gets a generic handler callback for otherwise unknown messages (mds map, in this case). - The basic supported/required feature bits can be expanded (and are by ceph_fs_client). No functional change, aside from some subtle error handling cases that got cleaned up in the refactoring process. Signed-off-by: Sage Weil <sage@newdream.net>
2010-04-06 15:14:15 -07:00
ceph_zero_page_vector_range(page_off + read,
pos - off - read, pages);
ceph: file operations File open and close operations, and read and write methods that ensure we have obtained the proper capabilities from the MDS cluster before performing IO on a file. We take references on held capabilities for the duration of the read/write to avoid prematurely releasing them back to the MDS. We implement two main paths for read and write: one that is buffered (and uses generic_aio_{read,write}), and one that is fully synchronous and blocking (operating either on a __user pointer or, if O_DIRECT, directly on user pages). Signed-off-by: Sage Weil <sage@newdream.net>
2009-10-06 11:31:08 -07:00
}
pos += ret;
read = pos - off;
left -= ret;
page_pos += didpages;
pages_left -= didpages;
/* hit stripe? */
if (left && hit_stripe)
goto more;
}
if (was_short) {
/* was original extent fully inside i_size? */
if (pos + left <= inode->i_size) {
dout("zero tail\n");
ceph: factor out libceph from Ceph file system This factors out protocol and low-level storage parts of ceph into a separate libceph module living in net/ceph and include/linux/ceph. This is mostly a matter of moving files around. However, a few key pieces of the interface change as well: - ceph_client becomes ceph_fs_client and ceph_client, where the latter captures the mon and osd clients, and the fs_client gets the mds client and file system specific pieces. - Mount option parsing and debugfs setup is correspondingly broken into two pieces. - The mon client gets a generic handler callback for otherwise unknown messages (mds map, in this case). - The basic supported/required feature bits can be expanded (and are by ceph_fs_client). No functional change, aside from some subtle error handling cases that got cleaned up in the refactoring process. Signed-off-by: Sage Weil <sage@newdream.net>
2010-04-06 15:14:15 -07:00
ceph_zero_page_vector_range(page_off + read, len - read,
pages);
ceph: fix short synchronous reads Zeroing of holes was not done correctly: page_off was miscalculated and zeroing the tail didn't not adjust the 'read' value to include the zeroed portion. Signed-off-by: Yehuda Sadeh <yehuda@hq.newdream.net> Signed-off-by: Sage Weil <sage@newdream.net>
2010-02-04 13:41:41 -08:00
read = len;
ceph: file operations File open and close operations, and read and write methods that ensure we have obtained the proper capabilities from the MDS cluster before performing IO on a file. We take references on held capabilities for the duration of the read/write to avoid prematurely releasing them back to the MDS. We implement two main paths for read and write: one that is buffered (and uses generic_aio_{read,write}), and one that is fully synchronous and blocking (operating either on a __user pointer or, if O_DIRECT, directly on user pages). Signed-off-by: Sage Weil <sage@newdream.net>
2009-10-06 11:31:08 -07:00
goto out;
}
/* check i_size */
ceph: fix sync read eof check deadlock If a sync read gets a short result from the OSD, it may need to do a getattr to see if it is short due to reaching end-of-file. The getattr was being done while holding a reference to FILE_RD, which can lead to a deadlock if the MDS is revoking that capability bit and can't process the getattr until it does. We fix this by setting a flag if EOF size validation is needed, and doing the getattr in ceph_aio_read, after the RD cap ref is dropped. If the read needs to be continued, we loop and continue traversing the file. Signed-off-by: Sage Weil <sage@newdream.net>
2010-02-09 14:04:02 -08:00
*checkeof = 1;
ceph: file operations File open and close operations, and read and write methods that ensure we have obtained the proper capabilities from the MDS cluster before performing IO on a file. We take references on held capabilities for the duration of the read/write to avoid prematurely releasing them back to the MDS. We implement two main paths for read and write: one that is buffered (and uses generic_aio_{read,write}), and one that is fully synchronous and blocking (operating either on a __user pointer or, if O_DIRECT, directly on user pages). Signed-off-by: Sage Weil <sage@newdream.net>
2009-10-06 11:31:08 -07:00
}
out:
if (ret >= 0)
ret = read;
dout("striped_read returns %d\n", ret);
return ret;
}
/*
* Completely synchronous read and write methods. Direct from __user
* buffer to osd, or directly to user pages (if O_DIRECT).
*
* If the read spans object boundary, just do multiple reads.
*/
static ssize_t ceph_sync_read(struct file *file, char __user *data,
ceph: fix sync read eof check deadlock If a sync read gets a short result from the OSD, it may need to do a getattr to see if it is short due to reaching end-of-file. The getattr was being done while holding a reference to FILE_RD, which can lead to a deadlock if the MDS is revoking that capability bit and can't process the getattr until it does. We fix this by setting a flag if EOF size validation is needed, and doing the getattr in ceph_aio_read, after the RD cap ref is dropped. If the read needs to be continued, we loop and continue traversing the file. Signed-off-by: Sage Weil <sage@newdream.net>
2010-02-09 14:04:02 -08:00
unsigned len, loff_t *poff, int *checkeof)
ceph: file operations File open and close operations, and read and write methods that ensure we have obtained the proper capabilities from the MDS cluster before performing IO on a file. We take references on held capabilities for the duration of the read/write to avoid prematurely releasing them back to the MDS. We implement two main paths for read and write: one that is buffered (and uses generic_aio_{read,write}), and one that is fully synchronous and blocking (operating either on a __user pointer or, if O_DIRECT, directly on user pages). Signed-off-by: Sage Weil <sage@newdream.net>
2009-10-06 11:31:08 -07:00
{
struct inode *inode = file->f_dentry->d_inode;
struct page **pages;
u64 off = *poff;
int num_pages = calc_pages_for(off, len);
int ret;
dout("sync_read on file %p %llu~%u %s\n", file, off, len,
(file->f_flags & O_DIRECT) ? "O_DIRECT" : "");
ceph: fix comment, remove extraneous args The offset/length arguments aren't used. Signed-off-by: Sage Weil <sage@newdream.net>
2010-11-09 12:24:53 -08:00
if (file->f_flags & O_DIRECT)
pages = ceph_get_direct_page_vector(data, num_pages);
else
ceph: all allocation functions should get gfp_mask This is essential, as for the rados block device we'll need to run in different contexts that would need flags that are other than GFP_NOFS. Signed-off-by: Yehuda Sadeh <yehuda@hq.newdream.net> Signed-off-by: Sage Weil <sage@newdream.net>
2010-04-06 14:33:58 -07:00
pages = ceph_alloc_page_vector(num_pages, GFP_NOFS);
ceph: file operations File open and close operations, and read and write methods that ensure we have obtained the proper capabilities from the MDS cluster before performing IO on a file. We take references on held capabilities for the duration of the read/write to avoid prematurely releasing them back to the MDS. We implement two main paths for read and write: one that is buffered (and uses generic_aio_{read,write}), and one that is fully synchronous and blocking (operating either on a __user pointer or, if O_DIRECT, directly on user pages). Signed-off-by: Sage Weil <sage@newdream.net>
2009-10-06 11:31:08 -07:00
if (IS_ERR(pages))
return PTR_ERR(pages);
ceph: fix comment, remove extraneous args The offset/length arguments aren't used. Signed-off-by: Sage Weil <sage@newdream.net>
2010-11-09 12:24:53 -08:00
/*
* flush any page cache pages in this range. this
* will make concurrent normal and sync io slow,
* but it will at least behave sensibly when they are
* in sequence.
*/
ceph: sync read/write considers page cache In the cases where we either do a sync read or a write, we need to make sure that everything in the page cache is flushed. In the case of a sync write we invalidate the relevant pages, so that subsequent read/write reflects the new data written. Signed-off-by: Yehuda Sadeh <yehuda@hq.newdream.net> Signed-off-by: Sage Weil <sage@newdream.net>
2010-02-09 11:14:41 -08:00
ret = filemap_write_and_wait(inode->i_mapping);
if (ret < 0)
goto done;
ceph: make page alignment explicit in osd interface 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>
2010-11-09 12:43:12 -08:00
ret = striped_read(inode, off, len, pages, num_pages, checkeof,
file->f_flags & O_DIRECT);
ceph: file operations File open and close operations, and read and write methods that ensure we have obtained the proper capabilities from the MDS cluster before performing IO on a file. We take references on held capabilities for the duration of the read/write to avoid prematurely releasing them back to the MDS. We implement two main paths for read and write: one that is buffered (and uses generic_aio_{read,write}), and one that is fully synchronous and blocking (operating either on a __user pointer or, if O_DIRECT, directly on user pages). Signed-off-by: Sage Weil <sage@newdream.net>
2009-10-06 11:31:08 -07:00
if (ret >= 0 && (file->f_flags & O_DIRECT) == 0)
ceph: factor out libceph from Ceph file system This factors out protocol and low-level storage parts of ceph into a separate libceph module living in net/ceph and include/linux/ceph. This is mostly a matter of moving files around. However, a few key pieces of the interface change as well: - ceph_client becomes ceph_fs_client and ceph_client, where the latter captures the mon and osd clients, and the fs_client gets the mds client and file system specific pieces. - Mount option parsing and debugfs setup is correspondingly broken into two pieces. - The mon client gets a generic handler callback for otherwise unknown messages (mds map, in this case). - The basic supported/required feature bits can be expanded (and are by ceph_fs_client). No functional change, aside from some subtle error handling cases that got cleaned up in the refactoring process. Signed-off-by: Sage Weil <sage@newdream.net>
2010-04-06 15:14:15 -07:00
ret = ceph_copy_page_vector_to_user(pages, data, off, ret);
ceph: file operations File open and close operations, and read and write methods that ensure we have obtained the proper capabilities from the MDS cluster before performing IO on a file. We take references on held capabilities for the duration of the read/write to avoid prematurely releasing them back to the MDS. We implement two main paths for read and write: one that is buffered (and uses generic_aio_{read,write}), and one that is fully synchronous and blocking (operating either on a __user pointer or, if O_DIRECT, directly on user pages). Signed-off-by: Sage Weil <sage@newdream.net>
2009-10-06 11:31:08 -07:00
if (ret >= 0)
*poff = off + ret;
ceph: sync read/write considers page cache In the cases where we either do a sync read or a write, we need to make sure that everything in the page cache is flushed. In the case of a sync write we invalidate the relevant pages, so that subsequent read/write reflects the new data written. Signed-off-by: Yehuda Sadeh <yehuda@hq.newdream.net> Signed-off-by: Sage Weil <sage@newdream.net>
2010-02-09 11:14:41 -08:00
done:
ceph: file operations File open and close operations, and read and write methods that ensure we have obtained the proper capabilities from the MDS cluster before performing IO on a file. We take references on held capabilities for the duration of the read/write to avoid prematurely releasing them back to the MDS. We implement two main paths for read and write: one that is buffered (and uses generic_aio_{read,write}), and one that is fully synchronous and blocking (operating either on a __user pointer or, if O_DIRECT, directly on user pages). Signed-off-by: Sage Weil <sage@newdream.net>
2009-10-06 11:31:08 -07:00
if (file->f_flags & O_DIRECT)
ceph: factor out libceph from Ceph file system This factors out protocol and low-level storage parts of ceph into a separate libceph module living in net/ceph and include/linux/ceph. This is mostly a matter of moving files around. However, a few key pieces of the interface change as well: - ceph_client becomes ceph_fs_client and ceph_client, where the latter captures the mon and osd clients, and the fs_client gets the mds client and file system specific pieces. - Mount option parsing and debugfs setup is correspondingly broken into two pieces. - The mon client gets a generic handler callback for otherwise unknown messages (mds map, in this case). - The basic supported/required feature bits can be expanded (and are by ceph_fs_client). No functional change, aside from some subtle error handling cases that got cleaned up in the refactoring process. Signed-off-by: Sage Weil <sage@newdream.net>
2010-04-06 15:14:15 -07:00
ceph_put_page_vector(pages, num_pages);
ceph: file operations File open and close operations, and read and write methods that ensure we have obtained the proper capabilities from the MDS cluster before performing IO on a file. We take references on held capabilities for the duration of the read/write to avoid prematurely releasing them back to the MDS. We implement two main paths for read and write: one that is buffered (and uses generic_aio_{read,write}), and one that is fully synchronous and blocking (operating either on a __user pointer or, if O_DIRECT, directly on user pages). Signed-off-by: Sage Weil <sage@newdream.net>
2009-10-06 11:31:08 -07:00
else
ceph_release_page_vector(pages, num_pages);
dout("sync_read result %d\n", ret);
return ret;
}
/*
* Write commit callback, called if we requested both an ACK and
* ONDISK commit reply from the OSD.
*/
static void sync_write_commit(struct ceph_osd_request *req,
struct ceph_msg *msg)
{
struct ceph_inode_info *ci = ceph_inode(req->r_inode);
dout("sync_write_commit %p tid %llu\n", req, req->r_tid);
spin_lock(&ci->i_unsafe_lock);
list_del_init(&req->r_unsafe_item);
spin_unlock(&ci->i_unsafe_lock);
ceph_put_cap_refs(ci, CEPH_CAP_FILE_WR);
}
/*
* Synchronous write, straight from __user pointer or user pages (if
* O_DIRECT).
*
* If write spans object boundary, just do multiple writes. (For a
* correct atomic write, we should e.g. take write locks on all
* objects, rollback on failure, etc.)
*/
static ssize_t ceph_sync_write(struct file *file, const char __user *data,
size_t left, loff_t *offset)
{
struct inode *inode = file->f_dentry->d_inode;
struct ceph_inode_info *ci = ceph_inode(inode);
ceph: factor out libceph from Ceph file system This factors out protocol and low-level storage parts of ceph into a separate libceph module living in net/ceph and include/linux/ceph. This is mostly a matter of moving files around. However, a few key pieces of the interface change as well: - ceph_client becomes ceph_fs_client and ceph_client, where the latter captures the mon and osd clients, and the fs_client gets the mds client and file system specific pieces. - Mount option parsing and debugfs setup is correspondingly broken into two pieces. - The mon client gets a generic handler callback for otherwise unknown messages (mds map, in this case). - The basic supported/required feature bits can be expanded (and are by ceph_fs_client). No functional change, aside from some subtle error handling cases that got cleaned up in the refactoring process. Signed-off-by: Sage Weil <sage@newdream.net>
2010-04-06 15:14:15 -07:00
struct ceph_fs_client *fsc = ceph_inode_to_client(inode);
ceph: file operations File open and close operations, and read and write methods that ensure we have obtained the proper capabilities from the MDS cluster before performing IO on a file. We take references on held capabilities for the duration of the read/write to avoid prematurely releasing them back to the MDS. We implement two main paths for read and write: one that is buffered (and uses generic_aio_{read,write}), and one that is fully synchronous and blocking (operating either on a __user pointer or, if O_DIRECT, directly on user pages). Signed-off-by: Sage Weil <sage@newdream.net>
2009-10-06 11:31:08 -07:00
struct ceph_osd_request *req;
struct page **pages;
int num_pages;
long long unsigned pos;
u64 len;
int written = 0;
int flags;
int do_sync = 0;
int check_caps = 0;
ceph: make page alignment explicit in osd interface 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>
2010-11-09 12:43:12 -08:00
int page_align, io_align;
ceph: file operations File open and close operations, and read and write methods that ensure we have obtained the proper capabilities from the MDS cluster before performing IO on a file. We take references on held capabilities for the duration of the read/write to avoid prematurely releasing them back to the MDS. We implement two main paths for read and write: one that is buffered (and uses generic_aio_{read,write}), and one that is fully synchronous and blocking (operating either on a __user pointer or, if O_DIRECT, directly on user pages). Signed-off-by: Sage Weil <sage@newdream.net>
2009-10-06 11:31:08 -07:00
int ret;
struct timespec mtime = CURRENT_TIME;
if (ceph_snap(file->f_dentry->d_inode) != CEPH_NOSNAP)
return -EROFS;
dout("sync_write on file %p %lld~%u %s\n", file, *offset,
(unsigned)left, (file->f_flags & O_DIRECT) ? "O_DIRECT" : "");
if (file->f_flags & O_APPEND)
pos = i_size_read(inode);
else
pos = *offset;
ceph: make page alignment explicit in osd interface 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>
2010-11-09 12:43:12 -08:00
io_align = pos & ~PAGE_MASK;
ceph: sync read/write considers page cache In the cases where we either do a sync read or a write, we need to make sure that everything in the page cache is flushed. In the case of a sync write we invalidate the relevant pages, so that subsequent read/write reflects the new data written. Signed-off-by: Yehuda Sadeh <yehuda@hq.newdream.net> Signed-off-by: Sage Weil <sage@newdream.net>
2010-02-09 11:14:41 -08:00
ret = filemap_write_and_wait_range(inode->i_mapping, pos, pos + left);
if (ret < 0)
return ret;
ret = invalidate_inode_pages2_range(inode->i_mapping,
pos >> PAGE_CACHE_SHIFT,
(pos + left) >> PAGE_CACHE_SHIFT);
if (ret < 0)
dout("invalidate_inode_pages2_range returned %d\n", ret);
ceph: file operations File open and close operations, and read and write methods that ensure we have obtained the proper capabilities from the MDS cluster before performing IO on a file. We take references on held capabilities for the duration of the read/write to avoid prematurely releasing them back to the MDS. We implement two main paths for read and write: one that is buffered (and uses generic_aio_{read,write}), and one that is fully synchronous and blocking (operating either on a __user pointer or, if O_DIRECT, directly on user pages). Signed-off-by: Sage Weil <sage@newdream.net>
2009-10-06 11:31:08 -07:00
flags = CEPH_OSD_FLAG_ORDERSNAP |
CEPH_OSD_FLAG_ONDISK |
CEPH_OSD_FLAG_WRITE;
if ((file->f_flags & (O_SYNC|O_DIRECT)) == 0)
flags |= CEPH_OSD_FLAG_ACK;
else
do_sync = 1;
/*
* we may need to do multiple writes here if we span an object
* boundary. this isn't atomic, unfortunately. :(
*/
more:
len = left;
ceph: make page alignment explicit in osd interface 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>
2010-11-09 12:43:12 -08:00
if (file->f_flags & O_DIRECT)
/* write from beginning of first page, regardless of
io alignment */
page_align = (pos - io_align) & ~PAGE_MASK;
else
page_align = pos & ~PAGE_MASK;
ceph: factor out libceph from Ceph file system This factors out protocol and low-level storage parts of ceph into a separate libceph module living in net/ceph and include/linux/ceph. This is mostly a matter of moving files around. However, a few key pieces of the interface change as well: - ceph_client becomes ceph_fs_client and ceph_client, where the latter captures the mon and osd clients, and the fs_client gets the mds client and file system specific pieces. - Mount option parsing and debugfs setup is correspondingly broken into two pieces. - The mon client gets a generic handler callback for otherwise unknown messages (mds map, in this case). - The basic supported/required feature bits can be expanded (and are by ceph_fs_client). No functional change, aside from some subtle error handling cases that got cleaned up in the refactoring process. Signed-off-by: Sage Weil <sage@newdream.net>
2010-04-06 15:14:15 -07:00
req = ceph_osdc_new_request(&fsc->client->osdc, &ci->i_layout,
ceph: file operations File open and close operations, and read and write methods that ensure we have obtained the proper capabilities from the MDS cluster before performing IO on a file. We take references on held capabilities for the duration of the read/write to avoid prematurely releasing them back to the MDS. We implement two main paths for read and write: one that is buffered (and uses generic_aio_{read,write}), and one that is fully synchronous and blocking (operating either on a __user pointer or, if O_DIRECT, directly on user pages). Signed-off-by: Sage Weil <sage@newdream.net>
2009-10-06 11:31:08 -07:00
ceph_vino(inode), pos, &len,
CEPH_OSD_OP_WRITE, flags,
ci->i_snap_realm->cached_context,
do_sync,
ci->i_truncate_seq, ci->i_truncate_size,
ceph: make page alignment explicit in osd interface 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>
2010-11-09 12:43:12 -08:00
&mtime, false, 2, page_align);
ceph: make ceph_msg_new return NULL on failure; clean up, fix callers Returning ERR_PTR(-ENOMEM) is useless extra work. Return NULL on failure instead, and fix up the callers (about half of which were wrong anyway). Signed-off-by: Sage Weil <sage@newdream.net>
2010-04-01 16:06:19 -07:00
if (!req)
return -ENOMEM;
ceph: file operations File open and close operations, and read and write methods that ensure we have obtained the proper capabilities from the MDS cluster before performing IO on a file. We take references on held capabilities for the duration of the read/write to avoid prematurely releasing them back to the MDS. We implement two main paths for read and write: one that is buffered (and uses generic_aio_{read,write}), and one that is fully synchronous and blocking (operating either on a __user pointer or, if O_DIRECT, directly on user pages). Signed-off-by: Sage Weil <sage@newdream.net>
2009-10-06 11:31:08 -07:00
num_pages = calc_pages_for(pos, len);
if (file->f_flags & O_DIRECT) {
ceph: make page alignment explicit in osd interface 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>
2010-11-09 12:43:12 -08:00
pages = ceph_get_direct_page_vector(data, num_pages);
ceph: file operations File open and close operations, and read and write methods that ensure we have obtained the proper capabilities from the MDS cluster before performing IO on a file. We take references on held capabilities for the duration of the read/write to avoid prematurely releasing them back to the MDS. We implement two main paths for read and write: one that is buffered (and uses generic_aio_{read,write}), and one that is fully synchronous and blocking (operating either on a __user pointer or, if O_DIRECT, directly on user pages). Signed-off-by: Sage Weil <sage@newdream.net>
2009-10-06 11:31:08 -07:00
if (IS_ERR(pages)) {
ret = PTR_ERR(pages);
goto out;
}
/*
* throw out any page cache pages in this range. this
* may block.
*/
ceph: whitespace cleanup Signed-off-by: Sage Weil <sage@newdream.net>
2010-08-03 10:25:11 -07:00
truncate_inode_pages_range(inode->i_mapping, pos,
ceph: fix direct io truncate offset truncate_inode_pages_range wants the end offset to align with the last byte in a page. Signed-off-by: Sage Weil <sage@newdream.net>
2010-04-22 13:48:59 -07:00
(pos+len) | (PAGE_CACHE_SIZE-1));
ceph: file operations File open and close operations, and read and write methods that ensure we have obtained the proper capabilities from the MDS cluster before performing IO on a file. We take references on held capabilities for the duration of the read/write to avoid prematurely releasing them back to the MDS. We implement two main paths for read and write: one that is buffered (and uses generic_aio_{read,write}), and one that is fully synchronous and blocking (operating either on a __user pointer or, if O_DIRECT, directly on user pages). Signed-off-by: Sage Weil <sage@newdream.net>
2009-10-06 11:31:08 -07:00
} else {
ceph: all allocation functions should get gfp_mask This is essential, as for the rados block device we'll need to run in different contexts that would need flags that are other than GFP_NOFS. Signed-off-by: Yehuda Sadeh <yehuda@hq.newdream.net> Signed-off-by: Sage Weil <sage@newdream.net>
2010-04-06 14:33:58 -07:00
pages = ceph_alloc_page_vector(num_pages, GFP_NOFS);
ceph: file operations File open and close operations, and read and write methods that ensure we have obtained the proper capabilities from the MDS cluster before performing IO on a file. We take references on held capabilities for the duration of the read/write to avoid prematurely releasing them back to the MDS. We implement two main paths for read and write: one that is buffered (and uses generic_aio_{read,write}), and one that is fully synchronous and blocking (operating either on a __user pointer or, if O_DIRECT, directly on user pages). Signed-off-by: Sage Weil <sage@newdream.net>
2009-10-06 11:31:08 -07:00
if (IS_ERR(pages)) {
ret = PTR_ERR(pages);
goto out;
}
ceph: factor out libceph from Ceph file system This factors out protocol and low-level storage parts of ceph into a separate libceph module living in net/ceph and include/linux/ceph. This is mostly a matter of moving files around. However, a few key pieces of the interface change as well: - ceph_client becomes ceph_fs_client and ceph_client, where the latter captures the mon and osd clients, and the fs_client gets the mds client and file system specific pieces. - Mount option parsing and debugfs setup is correspondingly broken into two pieces. - The mon client gets a generic handler callback for otherwise unknown messages (mds map, in this case). - The basic supported/required feature bits can be expanded (and are by ceph_fs_client). No functional change, aside from some subtle error handling cases that got cleaned up in the refactoring process. Signed-off-by: Sage Weil <sage@newdream.net>
2010-04-06 15:14:15 -07:00
ret = ceph_copy_user_to_page_vector(pages, data, pos, len);
ceph: file operations File open and close operations, and read and write methods that ensure we have obtained the proper capabilities from the MDS cluster before performing IO on a file. We take references on held capabilities for the duration of the read/write to avoid prematurely releasing them back to the MDS. We implement two main paths for read and write: one that is buffered (and uses generic_aio_{read,write}), and one that is fully synchronous and blocking (operating either on a __user pointer or, if O_DIRECT, directly on user pages). Signed-off-by: Sage Weil <sage@newdream.net>
2009-10-06 11:31:08 -07:00
if (ret < 0) {
ceph_release_page_vector(pages, num_pages);
goto out;
}
if ((file->f_flags & O_SYNC) == 0) {
/* get a second commit callback */
req->r_safe_callback = sync_write_commit;
req->r_own_pages = 1;
}
}
req->r_pages = pages;
req->r_num_pages = num_pages;
req->r_inode = inode;
ceph: factor out libceph from Ceph file system This factors out protocol and low-level storage parts of ceph into a separate libceph module living in net/ceph and include/linux/ceph. This is mostly a matter of moving files around. However, a few key pieces of the interface change as well: - ceph_client becomes ceph_fs_client and ceph_client, where the latter captures the mon and osd clients, and the fs_client gets the mds client and file system specific pieces. - Mount option parsing and debugfs setup is correspondingly broken into two pieces. - The mon client gets a generic handler callback for otherwise unknown messages (mds map, in this case). - The basic supported/required feature bits can be expanded (and are by ceph_fs_client). No functional change, aside from some subtle error handling cases that got cleaned up in the refactoring process. Signed-off-by: Sage Weil <sage@newdream.net>
2010-04-06 15:14:15 -07:00
ret = ceph_osdc_start_request(&fsc->client->osdc, req, false);
ceph: file operations File open and close operations, and read and write methods that ensure we have obtained the proper capabilities from the MDS cluster before performing IO on a file. We take references on held capabilities for the duration of the read/write to avoid prematurely releasing them back to the MDS. We implement two main paths for read and write: one that is buffered (and uses generic_aio_{read,write}), and one that is fully synchronous and blocking (operating either on a __user pointer or, if O_DIRECT, directly on user pages). Signed-off-by: Sage Weil <sage@newdream.net>
2009-10-06 11:31:08 -07:00
if (!ret) {
if (req->r_safe_callback) {
/*
* Add to inode unsafe list only after we
* start_request so that a tid has been assigned.
*/
spin_lock(&ci->i_unsafe_lock);
ceph: fix list_add usage on unsafe_writes list Fix argument order. Signed-off-by: Henry C Chang <henry_c_chang@tcloudcomputing.com> Signed-off-by: Sage Weil <sage@newdream.net>
2010-09-22 20:21:17 -07:00
list_add(&req->r_unsafe_item, &ci->i_unsafe_writes);
ceph: file operations File open and close operations, and read and write methods that ensure we have obtained the proper capabilities from the MDS cluster before performing IO on a file. We take references on held capabilities for the duration of the read/write to avoid prematurely releasing them back to the MDS. We implement two main paths for read and write: one that is buffered (and uses generic_aio_{read,write}), and one that is fully synchronous and blocking (operating either on a __user pointer or, if O_DIRECT, directly on user pages). Signed-off-by: Sage Weil <sage@newdream.net>
2009-10-06 11:31:08 -07:00
spin_unlock(&ci->i_unsafe_lock);
ceph_get_cap_refs(ci, CEPH_CAP_FILE_WR);
}
ceph: factor out libceph from Ceph file system This factors out protocol and low-level storage parts of ceph into a separate libceph module living in net/ceph and include/linux/ceph. This is mostly a matter of moving files around. However, a few key pieces of the interface change as well: - ceph_client becomes ceph_fs_client and ceph_client, where the latter captures the mon and osd clients, and the fs_client gets the mds client and file system specific pieces. - Mount option parsing and debugfs setup is correspondingly broken into two pieces. - The mon client gets a generic handler callback for otherwise unknown messages (mds map, in this case). - The basic supported/required feature bits can be expanded (and are by ceph_fs_client). No functional change, aside from some subtle error handling cases that got cleaned up in the refactoring process. Signed-off-by: Sage Weil <sage@newdream.net>
2010-04-06 15:14:15 -07:00
ret = ceph_osdc_wait_request(&fsc->client->osdc, req);
ceph: file operations File open and close operations, and read and write methods that ensure we have obtained the proper capabilities from the MDS cluster before performing IO on a file. We take references on held capabilities for the duration of the read/write to avoid prematurely releasing them back to the MDS. We implement two main paths for read and write: one that is buffered (and uses generic_aio_{read,write}), and one that is fully synchronous and blocking (operating either on a __user pointer or, if O_DIRECT, directly on user pages). Signed-off-by: Sage Weil <sage@newdream.net>
2009-10-06 11:31:08 -07:00
}
if (file->f_flags & O_DIRECT)
ceph: factor out libceph from Ceph file system This factors out protocol and low-level storage parts of ceph into a separate libceph module living in net/ceph and include/linux/ceph. This is mostly a matter of moving files around. However, a few key pieces of the interface change as well: - ceph_client becomes ceph_fs_client and ceph_client, where the latter captures the mon and osd clients, and the fs_client gets the mds client and file system specific pieces. - Mount option parsing and debugfs setup is correspondingly broken into two pieces. - The mon client gets a generic handler callback for otherwise unknown messages (mds map, in this case). - The basic supported/required feature bits can be expanded (and are by ceph_fs_client). No functional change, aside from some subtle error handling cases that got cleaned up in the refactoring process. Signed-off-by: Sage Weil <sage@newdream.net>
2010-04-06 15:14:15 -07:00
ceph_put_page_vector(pages, num_pages);
ceph: file operations File open and close operations, and read and write methods that ensure we have obtained the proper capabilities from the MDS cluster before performing IO on a file. We take references on held capabilities for the duration of the read/write to avoid prematurely releasing them back to the MDS. We implement two main paths for read and write: one that is buffered (and uses generic_aio_{read,write}), and one that is fully synchronous and blocking (operating either on a __user pointer or, if O_DIRECT, directly on user pages). Signed-off-by: Sage Weil <sage@newdream.net>
2009-10-06 11:31:08 -07:00
else if (file->f_flags & O_SYNC)
ceph_release_page_vector(pages, num_pages);
out:
ceph_osdc_put_request(req);
if (ret == 0) {
pos += len;
written += len;
left -= len;
if (left)
goto more;
ret = written;
*offset = pos;
if (pos > i_size_read(inode))
check_caps = ceph_inode_set_size(inode, pos);
if (check_caps)
ceph_check_caps(ceph_inode(inode), CHECK_CAPS_AUTHONLY,
NULL);
}
return ret;
}
/*
* Wrap generic_file_aio_read with checks for cap bits on the inode.
* Atomically grab references, so that those bits are not released
* back to the MDS mid-read.
*
* Hmm, the sync read case isn't actually async... should it be?
*/
static ssize_t ceph_aio_read(struct kiocb *iocb, const struct iovec *iov,
unsigned long nr_segs, loff_t pos)
{
struct file *filp = iocb->ki_filp;
ceph: perform lazy reads when file mode and caps permit If the file mode is marked as "lazy," perform cached/buffered reads when the caps permit it. Adjust the rdcache_gen and invalidation logic accordingly so that we manage our cache based on the FILE_CACHE -or- FILE_LAZYIO cap bits. Signed-off-by: Sage Weil <sage@newdream.net>
2010-05-27 10:40:43 -07:00
struct ceph_file_info *fi = filp->private_data;
ceph: file operations File open and close operations, and read and write methods that ensure we have obtained the proper capabilities from the MDS cluster before performing IO on a file. We take references on held capabilities for the duration of the read/write to avoid prematurely releasing them back to the MDS. We implement two main paths for read and write: one that is buffered (and uses generic_aio_{read,write}), and one that is fully synchronous and blocking (operating either on a __user pointer or, if O_DIRECT, directly on user pages). Signed-off-by: Sage Weil <sage@newdream.net>
2009-10-06 11:31:08 -07:00
loff_t *ppos = &iocb->ki_pos;
size_t len = iov->iov_len;
struct inode *inode = filp->f_dentry->d_inode;
struct ceph_inode_info *ci = ceph_inode(inode);
ceph: code cleanup Mainly fixing minor issues reported by sparse. Signed-off-by: Yehuda Sadeh <yehuda@hq.newdream.net> Signed-off-by: Sage Weil <sage@newdream.net>
2010-06-11 16:58:48 -07:00
void __user *base = iov->iov_base;
ceph: file operations File open and close operations, and read and write methods that ensure we have obtained the proper capabilities from the MDS cluster before performing IO on a file. We take references on held capabilities for the duration of the read/write to avoid prematurely releasing them back to the MDS. We implement two main paths for read and write: one that is buffered (and uses generic_aio_{read,write}), and one that is fully synchronous and blocking (operating either on a __user pointer or, if O_DIRECT, directly on user pages). Signed-off-by: Sage Weil <sage@newdream.net>
2009-10-06 11:31:08 -07:00
ssize_t ret;
ceph: perform lazy reads when file mode and caps permit If the file mode is marked as "lazy," perform cached/buffered reads when the caps permit it. Adjust the rdcache_gen and invalidation logic accordingly so that we manage our cache based on the FILE_CACHE -or- FILE_LAZYIO cap bits. Signed-off-by: Sage Weil <sage@newdream.net>
2010-05-27 10:40:43 -07:00
int want, got = 0;
ceph: fix sync read eof check deadlock If a sync read gets a short result from the OSD, it may need to do a getattr to see if it is short due to reaching end-of-file. The getattr was being done while holding a reference to FILE_RD, which can lead to a deadlock if the MDS is revoking that capability bit and can't process the getattr until it does. We fix this by setting a flag if EOF size validation is needed, and doing the getattr in ceph_aio_read, after the RD cap ref is dropped. If the read needs to be continued, we loop and continue traversing the file. Signed-off-by: Sage Weil <sage@newdream.net>
2010-02-09 14:04:02 -08:00
int checkeof = 0, read = 0;
ceph: file operations File open and close operations, and read and write methods that ensure we have obtained the proper capabilities from the MDS cluster before performing IO on a file. We take references on held capabilities for the duration of the read/write to avoid prematurely releasing them back to the MDS. We implement two main paths for read and write: one that is buffered (and uses generic_aio_{read,write}), and one that is fully synchronous and blocking (operating either on a __user pointer or, if O_DIRECT, directly on user pages). Signed-off-by: Sage Weil <sage@newdream.net>
2009-10-06 11:31:08 -07:00
dout("aio_read %p %llx.%llx %llu~%u trying to get caps on %p\n",
inode, ceph_vinop(inode), pos, (unsigned)len, inode);
ceph: fix sync read eof check deadlock If a sync read gets a short result from the OSD, it may need to do a getattr to see if it is short due to reaching end-of-file. The getattr was being done while holding a reference to FILE_RD, which can lead to a deadlock if the MDS is revoking that capability bit and can't process the getattr until it does. We fix this by setting a flag if EOF size validation is needed, and doing the getattr in ceph_aio_read, after the RD cap ref is dropped. If the read needs to be continued, we loop and continue traversing the file. Signed-off-by: Sage Weil <sage@newdream.net>
2010-02-09 14:04:02 -08:00
again:
ceph: file operations File open and close operations, and read and write methods that ensure we have obtained the proper capabilities from the MDS cluster before performing IO on a file. We take references on held capabilities for the duration of the read/write to avoid prematurely releasing them back to the MDS. We implement two main paths for read and write: one that is buffered (and uses generic_aio_{read,write}), and one that is fully synchronous and blocking (operating either on a __user pointer or, if O_DIRECT, directly on user pages). Signed-off-by: Sage Weil <sage@newdream.net>
2009-10-06 11:31:08 -07:00
__ceph_do_pending_vmtruncate(inode);
ceph: perform lazy reads when file mode and caps permit If the file mode is marked as "lazy," perform cached/buffered reads when the caps permit it. Adjust the rdcache_gen and invalidation logic accordingly so that we manage our cache based on the FILE_CACHE -or- FILE_LAZYIO cap bits. Signed-off-by: Sage Weil <sage@newdream.net>
2010-05-27 10:40:43 -07:00
if (fi->fmode & CEPH_FILE_MODE_LAZY)
want = CEPH_CAP_FILE_CACHE | CEPH_CAP_FILE_LAZYIO;
else
want = CEPH_CAP_FILE_CACHE;
ret = ceph_get_caps(ci, CEPH_CAP_FILE_RD, want, &got, -1);
ceph: file operations File open and close operations, and read and write methods that ensure we have obtained the proper capabilities from the MDS cluster before performing IO on a file. We take references on held capabilities for the duration of the read/write to avoid prematurely releasing them back to the MDS. We implement two main paths for read and write: one that is buffered (and uses generic_aio_{read,write}), and one that is fully synchronous and blocking (operating either on a __user pointer or, if O_DIRECT, directly on user pages). Signed-off-by: Sage Weil <sage@newdream.net>
2009-10-06 11:31:08 -07:00
if (ret < 0)
goto out;
dout("aio_read %p %llx.%llx %llu~%u got cap refs on %s\n",
inode, ceph_vinop(inode), pos, (unsigned)len,
ceph_cap_string(got));
ceph: perform lazy reads when file mode and caps permit If the file mode is marked as "lazy," perform cached/buffered reads when the caps permit it. Adjust the rdcache_gen and invalidation logic accordingly so that we manage our cache based on the FILE_CACHE -or- FILE_LAZYIO cap bits. Signed-off-by: Sage Weil <sage@newdream.net>
2010-05-27 10:40:43 -07:00
if ((got & (CEPH_CAP_FILE_CACHE|CEPH_CAP_FILE_LAZYIO)) == 0 ||
ceph: file operations File open and close operations, and read and write methods that ensure we have obtained the proper capabilities from the MDS cluster before performing IO on a file. We take references on held capabilities for the duration of the read/write to avoid prematurely releasing them back to the MDS. We implement two main paths for read and write: one that is buffered (and uses generic_aio_{read,write}), and one that is fully synchronous and blocking (operating either on a __user pointer or, if O_DIRECT, directly on user pages). Signed-off-by: Sage Weil <sage@newdream.net>
2009-10-06 11:31:08 -07:00
(iocb->ki_filp->f_flags & O_DIRECT) ||
(inode->i_sb->s_flags & MS_SYNCHRONOUS))
/* hmm, this isn't really async... */
ceph: fix sync read eof check deadlock If a sync read gets a short result from the OSD, it may need to do a getattr to see if it is short due to reaching end-of-file. The getattr was being done while holding a reference to FILE_RD, which can lead to a deadlock if the MDS is revoking that capability bit and can't process the getattr until it does. We fix this by setting a flag if EOF size validation is needed, and doing the getattr in ceph_aio_read, after the RD cap ref is dropped. If the read needs to be continued, we loop and continue traversing the file. Signed-off-by: Sage Weil <sage@newdream.net>
2010-02-09 14:04:02 -08:00
ret = ceph_sync_read(filp, base, len, ppos, &checkeof);
ceph: file operations File open and close operations, and read and write methods that ensure we have obtained the proper capabilities from the MDS cluster before performing IO on a file. We take references on held capabilities for the duration of the read/write to avoid prematurely releasing them back to the MDS. We implement two main paths for read and write: one that is buffered (and uses generic_aio_{read,write}), and one that is fully synchronous and blocking (operating either on a __user pointer or, if O_DIRECT, directly on user pages). Signed-off-by: Sage Weil <sage@newdream.net>
2009-10-06 11:31:08 -07:00
else
ret = generic_file_aio_read(iocb, iov, nr_segs, pos);
out:
dout("aio_read %p %llx.%llx dropping cap refs on %s = %d\n",
inode, ceph_vinop(inode), ceph_cap_string(got), (int)ret);
ceph_put_cap_refs(ci, got);
ceph: fix sync read eof check deadlock If a sync read gets a short result from the OSD, it may need to do a getattr to see if it is short due to reaching end-of-file. The getattr was being done while holding a reference to FILE_RD, which can lead to a deadlock if the MDS is revoking that capability bit and can't process the getattr until it does. We fix this by setting a flag if EOF size validation is needed, and doing the getattr in ceph_aio_read, after the RD cap ref is dropped. If the read needs to be continued, we loop and continue traversing the file. Signed-off-by: Sage Weil <sage@newdream.net>
2010-02-09 14:04:02 -08:00
if (checkeof && ret >= 0) {
int statret = ceph_do_getattr(inode, CEPH_STAT_CAP_SIZE);
/* hit EOF or hole? */
if (statret == 0 && *ppos < inode->i_size) {
dout("aio_read sync_read hit hole, reading more\n");
read += ret;
base += ret;
len -= ret;
checkeof = 0;
goto again;
}
}
if (ret >= 0)
ret += read;
ceph: file operations File open and close operations, and read and write methods that ensure we have obtained the proper capabilities from the MDS cluster before performing IO on a file. We take references on held capabilities for the duration of the read/write to avoid prematurely releasing them back to the MDS. We implement two main paths for read and write: one that is buffered (and uses generic_aio_{read,write}), and one that is fully synchronous and blocking (operating either on a __user pointer or, if O_DIRECT, directly on user pages). Signed-off-by: Sage Weil <sage@newdream.net>
2009-10-06 11:31:08 -07:00
return ret;
}
/*
* Take cap references to avoid releasing caps to MDS mid-write.
*
* If we are synchronous, and write with an old snap context, the OSD
* may return EOLDSNAPC. In that case, retry the write.. _after_
* dropping our cap refs and allowing the pending snap to logically
* complete _before_ this write occurs.
*
* If we are near ENOSPC, write synchronously.
*/
static ssize_t ceph_aio_write(struct kiocb *iocb, const struct iovec *iov,
unsigned long nr_segs, loff_t pos)
{
struct file *file = iocb->ki_filp;
ceph: perform lazy writes when file mode and caps permit If we have marked a file as "lazy" (using the ceph ioctl), perform buffered writes when the MDS caps allow it. Signed-off-by: Sage Weil <sage@newdream.net>
2010-05-26 14:31:27 -07:00
struct ceph_file_info *fi = file->private_data;
ceph: file operations File open and close operations, and read and write methods that ensure we have obtained the proper capabilities from the MDS cluster before performing IO on a file. We take references on held capabilities for the duration of the read/write to avoid prematurely releasing them back to the MDS. We implement two main paths for read and write: one that is buffered (and uses generic_aio_{read,write}), and one that is fully synchronous and blocking (operating either on a __user pointer or, if O_DIRECT, directly on user pages). Signed-off-by: Sage Weil <sage@newdream.net>
2009-10-06 11:31:08 -07:00
struct inode *inode = file->f_dentry->d_inode;
struct ceph_inode_info *ci = ceph_inode(inode);
ceph: factor out libceph from Ceph file system This factors out protocol and low-level storage parts of ceph into a separate libceph module living in net/ceph and include/linux/ceph. This is mostly a matter of moving files around. However, a few key pieces of the interface change as well: - ceph_client becomes ceph_fs_client and ceph_client, where the latter captures the mon and osd clients, and the fs_client gets the mds client and file system specific pieces. - Mount option parsing and debugfs setup is correspondingly broken into two pieces. - The mon client gets a generic handler callback for otherwise unknown messages (mds map, in this case). - The basic supported/required feature bits can be expanded (and are by ceph_fs_client). No functional change, aside from some subtle error handling cases that got cleaned up in the refactoring process. Signed-off-by: Sage Weil <sage@newdream.net>
2010-04-06 15:14:15 -07:00
struct ceph_osd_client *osdc =
&ceph_sb_to_client(inode->i_sb)->client->osdc;
ceph: file operations File open and close operations, and read and write methods that ensure we have obtained the proper capabilities from the MDS cluster before performing IO on a file. We take references on held capabilities for the duration of the read/write to avoid prematurely releasing them back to the MDS. We implement two main paths for read and write: one that is buffered (and uses generic_aio_{read,write}), and one that is fully synchronous and blocking (operating either on a __user pointer or, if O_DIRECT, directly on user pages). Signed-off-by: Sage Weil <sage@newdream.net>
2009-10-06 11:31:08 -07:00
loff_t endoff = pos + iov->iov_len;
ceph: perform lazy writes when file mode and caps permit If we have marked a file as "lazy" (using the ceph ioctl), perform buffered writes when the MDS caps allow it. Signed-off-by: Sage Weil <sage@newdream.net>
2010-05-26 14:31:27 -07:00
int want, got = 0;
ceph: don't clobber write return value when using O_SYNC Signed-off-by: Yehuda Sadeh <yehuda@hq.newdream.net> Signed-off-by: Sage Weil <sage@newdream.net>
2010-02-23 18:16:23 +00:00
int ret, err;
ceph: file operations File open and close operations, and read and write methods that ensure we have obtained the proper capabilities from the MDS cluster before performing IO on a file. We take references on held capabilities for the duration of the read/write to avoid prematurely releasing them back to the MDS. We implement two main paths for read and write: one that is buffered (and uses generic_aio_{read,write}), and one that is fully synchronous and blocking (operating either on a __user pointer or, if O_DIRECT, directly on user pages). Signed-off-by: Sage Weil <sage@newdream.net>
2009-10-06 11:31:08 -07:00
if (ceph_snap(inode) != CEPH_NOSNAP)
return -EROFS;
retry_snap:
if (ceph_osdmap_flag(osdc->osdmap, CEPH_OSDMAP_FULL))
return -ENOSPC;
__ceph_do_pending_vmtruncate(inode);
dout("aio_write %p %llx.%llx %llu~%u getting caps. i_size %llu\n",
inode, ceph_vinop(inode), pos, (unsigned)iov->iov_len,
inode->i_size);
ceph: perform lazy writes when file mode and caps permit If we have marked a file as "lazy" (using the ceph ioctl), perform buffered writes when the MDS caps allow it. Signed-off-by: Sage Weil <sage@newdream.net>
2010-05-26 14:31:27 -07:00
if (fi->fmode & CEPH_FILE_MODE_LAZY)
want = CEPH_CAP_FILE_BUFFER | CEPH_CAP_FILE_LAZYIO;
else
want = CEPH_CAP_FILE_BUFFER;
ret = ceph_get_caps(ci, CEPH_CAP_FILE_WR, want, &got, endoff);
ceph: file operations File open and close operations, and read and write methods that ensure we have obtained the proper capabilities from the MDS cluster before performing IO on a file. We take references on held capabilities for the duration of the read/write to avoid prematurely releasing them back to the MDS. We implement two main paths for read and write: one that is buffered (and uses generic_aio_{read,write}), and one that is fully synchronous and blocking (operating either on a __user pointer or, if O_DIRECT, directly on user pages). Signed-off-by: Sage Weil <sage@newdream.net>
2009-10-06 11:31:08 -07:00
if (ret < 0)
goto out;
dout("aio_write %p %llx.%llx %llu~%u got cap refs on %s\n",
inode, ceph_vinop(inode), pos, (unsigned)iov->iov_len,
ceph_cap_string(got));
ceph: perform lazy writes when file mode and caps permit If we have marked a file as "lazy" (using the ceph ioctl), perform buffered writes when the MDS caps allow it. Signed-off-by: Sage Weil <sage@newdream.net>
2010-05-26 14:31:27 -07:00
if ((got & (CEPH_CAP_FILE_BUFFER|CEPH_CAP_FILE_LAZYIO)) == 0 ||
ceph: file operations File open and close operations, and read and write methods that ensure we have obtained the proper capabilities from the MDS cluster before performing IO on a file. We take references on held capabilities for the duration of the read/write to avoid prematurely releasing them back to the MDS. We implement two main paths for read and write: one that is buffered (and uses generic_aio_{read,write}), and one that is fully synchronous and blocking (operating either on a __user pointer or, if O_DIRECT, directly on user pages). Signed-off-by: Sage Weil <sage@newdream.net>
2009-10-06 11:31:08 -07:00
(iocb->ki_filp->f_flags & O_DIRECT) ||
(inode->i_sb->s_flags & MS_SYNCHRONOUS)) {
ret = ceph_sync_write(file, iov->iov_base, iov->iov_len,
&iocb->ki_pos);
} else {
ret = generic_file_aio_write(iocb, iov, nr_segs, pos);
if ((ret >= 0 || ret == -EIOCBQUEUED) &&
((file->f_flags & O_SYNC) || IS_SYNC(file->f_mapping->host)
ceph: don't clobber write return value when using O_SYNC Signed-off-by: Yehuda Sadeh <yehuda@hq.newdream.net> Signed-off-by: Sage Weil <sage@newdream.net>
2010-02-23 18:16:23 +00:00
|| ceph_osdmap_flag(osdc->osdmap, CEPH_OSDMAP_NEARFULL))) {
sanitize vfs_fsync calling conventions Now that the last user passing a NULL file pointer is gone we can remove the redundant dentry argument and associated hacks inside vfs_fsynmc_range. The next step will be removig the dentry argument from ->fsync, but given the luck with the last round of method prototype changes I'd rather defer this until after the main merge window. Signed-off-by: Christoph Hellwig <hch@lst.de> Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
2010-03-22 17:32:25 +01:00
err = vfs_fsync_range(file, pos, pos + ret - 1, 1);
ceph: don't clobber write return value when using O_SYNC Signed-off-by: Yehuda Sadeh <yehuda@hq.newdream.net> Signed-off-by: Sage Weil <sage@newdream.net>
2010-02-23 18:16:23 +00:00
if (err < 0)
ret = err;
}
ceph: file operations File open and close operations, and read and write methods that ensure we have obtained the proper capabilities from the MDS cluster before performing IO on a file. We take references on held capabilities for the duration of the read/write to avoid prematurely releasing them back to the MDS. We implement two main paths for read and write: one that is buffered (and uses generic_aio_{read,write}), and one that is fully synchronous and blocking (operating either on a __user pointer or, if O_DIRECT, directly on user pages). Signed-off-by: Sage Weil <sage@newdream.net>
2009-10-06 11:31:08 -07:00
}
if (ret >= 0) {
spin_lock(&inode->i_lock);
__ceph_mark_dirty_caps(ci, CEPH_CAP_FILE_WR);
spin_unlock(&inode->i_lock);
}
out:
dout("aio_write %p %llx.%llx %llu~%u dropping cap refs on %s\n",
inode, ceph_vinop(inode), pos, (unsigned)iov->iov_len,
ceph_cap_string(got));
ceph_put_cap_refs(ci, got);
if (ret == -EOLDSNAPC) {
dout("aio_write %p %llx.%llx %llu~%u got EOLDSNAPC, retrying\n",
inode, ceph_vinop(inode), pos, (unsigned)iov->iov_len);
goto retry_snap;
}
return ret;
}
/*
* llseek. be sure to verify file size on SEEK_END.
*/
static loff_t ceph_llseek(struct file *file, loff_t offset, int origin)
{
struct inode *inode = file->f_mapping->host;
int ret;
mutex_lock(&inode->i_mutex);
__ceph_do_pending_vmtruncate(inode);
switch (origin) {
case SEEK_END:
ret = ceph_do_getattr(inode, CEPH_STAT_CAP_SIZE);
if (ret < 0) {
offset = ret;
goto out;
}
offset += inode->i_size;
break;
case SEEK_CUR:
/*
* Here we special-case the lseek(fd, 0, SEEK_CUR)
* position-querying operation. Avoid rewriting the "same"
* f_pos value back to the file because a concurrent read(),
* write() or lseek() might have altered it
*/
if (offset == 0) {
offset = file->f_pos;
goto out;
}
offset += file->f_pos;
break;
}
if (offset < 0 || offset > inode->i_sb->s_maxbytes) {
offset = -EINVAL;
goto out;
}
/* Special lock needed here? */
if (offset != file->f_pos) {
file->f_pos = offset;
file->f_version = 0;
}
out:
mutex_unlock(&inode->i_mutex);
return offset;
}
const struct file_operations ceph_file_fops = {
.open = ceph_open,
.release = ceph_release,
.llseek = ceph_llseek,
.read = do_sync_read,
.write = do_sync_write,
.aio_read = ceph_aio_read,
.aio_write = ceph_aio_write,
.mmap = ceph_mmap,
.fsync = ceph_fsync,
ceph: add flock/fcntl lock support Implement flock inode operation to support advisory file locking. All lock/unlock operations are synchronous with the MDS. Lock state is sent when reconnecting to a recovering MDS to restore the shared lock state. Signed-off-by: Greg Farnum <gregf@hq.newdream.net> Signed-off-by: Sage Weil <sage@newdream.net>
2010-08-02 15:34:23 -07:00
.lock = ceph_lock,
.flock = ceph_flock,
ceph: file operations File open and close operations, and read and write methods that ensure we have obtained the proper capabilities from the MDS cluster before performing IO on a file. We take references on held capabilities for the duration of the read/write to avoid prematurely releasing them back to the MDS. We implement two main paths for read and write: one that is buffered (and uses generic_aio_{read,write}), and one that is fully synchronous and blocking (operating either on a __user pointer or, if O_DIRECT, directly on user pages). Signed-off-by: Sage Weil <sage@newdream.net>
2009-10-06 11:31:08 -07:00
.splice_read = generic_file_splice_read,
.splice_write = generic_file_splice_write,
.unlocked_ioctl = ceph_ioctl,
.compat_ioctl = ceph_ioctl,
};
Reference in a new issue Copy permalink