d/freebsd-dev
1
0
Fork 0
Code Issues Pull Requests Packages Projects Releases Wiki Activity
7dc71949f2
freebsd-dev/module/zfs/zpl_xattr.c

1200 lines
27 KiB
C
Raw Normal View History

Add Hooks for Linux Xattr Operations The Linux specific xattr operations have all been located in the file zpl_xattr.c. These functions primarily rely on the reworked zfs_* functions to do their job. They are also responsible for converting the possible Solaris style error codes to negative Linux errors.
2011-01-26 20:10:01 +00:00
/*
* CDDL HEADER START
*
* The contents of this file are subject to the terms of the
* Common Development and Distribution License (the "License").
* You may not use this file except in compliance with the License.
*
* You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE
* or http://www.opensolaris.org/os/licensing.
* See the License for the specific language governing permissions
* and limitations under the License.
*
* When distributing Covered Code, include this CDDL HEADER in each
* file and include the License file at usr/src/OPENSOLARIS.LICENSE.
* If applicable, add the following below this CDDL HEADER, with the
* fields enclosed by brackets "[]" replaced with your own identifying
* information: Portions Copyright [yyyy] [name of copyright owner]
*
* CDDL HEADER END
*/
/*
* Copyright (c) 2011, Lawrence Livermore National Security, LLC.
*
* Extended attributes (xattr) on Solaris are implemented as files
* which exist in a hidden xattr directory. These extended attributes
* can be accessed using the attropen() system call which opens
* the extended attribute. It can then be manipulated just like
* a standard file descriptor. This has a couple advantages such
* as practically no size limit on the file, and the extended
* attributes permissions may differ from those of the parent file.
* This interface is really quite clever, but it's also completely
Implement SA based xattrs The current ZFS implementation stores xattrs on disk using a hidden directory. In this directory a file name represents the xattr name and the file contexts are the xattr binary data. This approach is very flexible and allows for arbitrarily large xattrs. However, it also suffers from a significant performance penalty. Accessing a single xattr can requires up to three disk seeks. 1) Lookup the dnode object. 2) Lookup the dnodes's xattr directory object. 3) Lookup the xattr object in the directory. To avoid this performance penalty Linux filesystems such as ext3 and xfs try to store the xattr as part of the inode on disk. When the xattr is to large to store in the inode then a single external block is allocated for them. In practice most xattrs are small and this approach works well. The addition of System Attributes (SA) to zfs provides us a clean way to make this optimization. When the dataset property 'xattr=sa' is set then xattrs will be preferentially stored as System Attributes. This allows tiny xattrs (~100 bytes) to be stored with the dnode and up to 64k of xattrs to be stored in the spill block. If additional xattr space is required, which is unlikely under Linux, they will be stored using the traditional directory approach. This optimization results in roughly a 3x performance improvement when accessing xattrs which brings zfs roughly to parity with ext4 and xfs (see table below). When multiple xattrs are stored per-file the performance improvements are even greater because all of the xattrs stored in the spill block will be cached. However, by default SA based xattrs are disabled in the Linux port to maximize compatibility with other implementations. If you do enable SA based xattrs then they will not be visible on platforms which do not support this feature. ---------------------------------------------------------------------- Time in seconds to get/set one xattr of N bytes on 100,000 files ------+--------------------------------+------------------------------ | setxattr | getxattr bytes | ext4 xfs zfs-dir zfs-sa | ext4 xfs zfs-dir zfs-sa ------+--------------------------------+------------------------------ 1 | 2.33 31.88 21.50 4.57 | 2.35 2.64 6.29 2.43 32 | 2.79 30.68 21.98 4.60 | 2.44 2.59 6.78 2.48 256 | 3.25 31.99 21.36 5.92 | 2.32 2.71 6.22 3.14 1024 | 3.30 32.61 22.83 8.45 | 2.40 2.79 6.24 3.27 4096 | 3.57 317.46 22.52 10.73 | 2.78 28.62 6.90 3.94 16384 | n/a 2342.39 34.30 19.20 | n/a 45.44 145.90 7.55 65536 | n/a 2941.39 128.15 131.32* | n/a 141.92 256.85 262.12* Legend: * ext4 - Stock RHEL6.1 ext4 mounted with '-o user_xattr'. * xfs - Stock RHEL6.1 xfs mounted with default options. * zfs-dir - Directory based xattrs only. * zfs-sa - Prefer SAs but spill in to directories as needed, a trailing * indicates overflow in to directories occured. NOTE: Ext4 supports 4096 bytes of xattr name/value pairs per file. NOTE: XFS and ZFS have no limit on xattr name/value pairs per file. NOTE: Linux limits individual name/value pairs to 65536 bytes. NOTE: All setattr/getattr's were done after dropping the cache. NOTE: All tests were run against a single hard drive. Signed-off-by: Brian Behlendorf <behlendorf1@llnl.gov> Issue #443
2011-10-24 23:55:20 +00:00
* different than what is supported on Linux. It also comes with a
* steep performance penalty when accessing small xattrs because they
* are not stored with the parent file.
Add Hooks for Linux Xattr Operations The Linux specific xattr operations have all been located in the file zpl_xattr.c. These functions primarily rely on the reworked zfs_* functions to do their job. They are also responsible for converting the possible Solaris style error codes to negative Linux errors.
2011-01-26 20:10:01 +00:00
*
* Under Linux extended attributes are manipulated by the system
* calls getxattr(2), setxattr(2), and listxattr(2). They consider
* extended attributes to be name/value pairs where the name is a
* NULL terminated string. The name must also include one of the
Implement SA based xattrs The current ZFS implementation stores xattrs on disk using a hidden directory. In this directory a file name represents the xattr name and the file contexts are the xattr binary data. This approach is very flexible and allows for arbitrarily large xattrs. However, it also suffers from a significant performance penalty. Accessing a single xattr can requires up to three disk seeks. 1) Lookup the dnode object. 2) Lookup the dnodes's xattr directory object. 3) Lookup the xattr object in the directory. To avoid this performance penalty Linux filesystems such as ext3 and xfs try to store the xattr as part of the inode on disk. When the xattr is to large to store in the inode then a single external block is allocated for them. In practice most xattrs are small and this approach works well. The addition of System Attributes (SA) to zfs provides us a clean way to make this optimization. When the dataset property 'xattr=sa' is set then xattrs will be preferentially stored as System Attributes. This allows tiny xattrs (~100 bytes) to be stored with the dnode and up to 64k of xattrs to be stored in the spill block. If additional xattr space is required, which is unlikely under Linux, they will be stored using the traditional directory approach. This optimization results in roughly a 3x performance improvement when accessing xattrs which brings zfs roughly to parity with ext4 and xfs (see table below). When multiple xattrs are stored per-file the performance improvements are even greater because all of the xattrs stored in the spill block will be cached. However, by default SA based xattrs are disabled in the Linux port to maximize compatibility with other implementations. If you do enable SA based xattrs then they will not be visible on platforms which do not support this feature. ---------------------------------------------------------------------- Time in seconds to get/set one xattr of N bytes on 100,000 files ------+--------------------------------+------------------------------ | setxattr | getxattr bytes | ext4 xfs zfs-dir zfs-sa | ext4 xfs zfs-dir zfs-sa ------+--------------------------------+------------------------------ 1 | 2.33 31.88 21.50 4.57 | 2.35 2.64 6.29 2.43 32 | 2.79 30.68 21.98 4.60 | 2.44 2.59 6.78 2.48 256 | 3.25 31.99 21.36 5.92 | 2.32 2.71 6.22 3.14 1024 | 3.30 32.61 22.83 8.45 | 2.40 2.79 6.24 3.27 4096 | 3.57 317.46 22.52 10.73 | 2.78 28.62 6.90 3.94 16384 | n/a 2342.39 34.30 19.20 | n/a 45.44 145.90 7.55 65536 | n/a 2941.39 128.15 131.32* | n/a 141.92 256.85 262.12* Legend: * ext4 - Stock RHEL6.1 ext4 mounted with '-o user_xattr'. * xfs - Stock RHEL6.1 xfs mounted with default options. * zfs-dir - Directory based xattrs only. * zfs-sa - Prefer SAs but spill in to directories as needed, a trailing * indicates overflow in to directories occured. NOTE: Ext4 supports 4096 bytes of xattr name/value pairs per file. NOTE: XFS and ZFS have no limit on xattr name/value pairs per file. NOTE: Linux limits individual name/value pairs to 65536 bytes. NOTE: All setattr/getattr's were done after dropping the cache. NOTE: All tests were run against a single hard drive. Signed-off-by: Brian Behlendorf <behlendorf1@llnl.gov> Issue #443
2011-10-24 23:55:20 +00:00
* following namespace prefixes:
Add Hooks for Linux Xattr Operations The Linux specific xattr operations have all been located in the file zpl_xattr.c. These functions primarily rely on the reworked zfs_* functions to do their job. They are also responsible for converting the possible Solaris style error codes to negative Linux errors.
2011-01-26 20:10:01 +00:00
*
* user - No restrictions and is available to user applications.
* trusted - Restricted to kernel and root (CAP_SYS_ADMIN) use.
* system - Used for access control lists (system.nfs4_acl, etc).
* security - Used by SELinux to store a files security context.
*
Implement SA based xattrs The current ZFS implementation stores xattrs on disk using a hidden directory. In this directory a file name represents the xattr name and the file contexts are the xattr binary data. This approach is very flexible and allows for arbitrarily large xattrs. However, it also suffers from a significant performance penalty. Accessing a single xattr can requires up to three disk seeks. 1) Lookup the dnode object. 2) Lookup the dnodes's xattr directory object. 3) Lookup the xattr object in the directory. To avoid this performance penalty Linux filesystems such as ext3 and xfs try to store the xattr as part of the inode on disk. When the xattr is to large to store in the inode then a single external block is allocated for them. In practice most xattrs are small and this approach works well. The addition of System Attributes (SA) to zfs provides us a clean way to make this optimization. When the dataset property 'xattr=sa' is set then xattrs will be preferentially stored as System Attributes. This allows tiny xattrs (~100 bytes) to be stored with the dnode and up to 64k of xattrs to be stored in the spill block. If additional xattr space is required, which is unlikely under Linux, they will be stored using the traditional directory approach. This optimization results in roughly a 3x performance improvement when accessing xattrs which brings zfs roughly to parity with ext4 and xfs (see table below). When multiple xattrs are stored per-file the performance improvements are even greater because all of the xattrs stored in the spill block will be cached. However, by default SA based xattrs are disabled in the Linux port to maximize compatibility with other implementations. If you do enable SA based xattrs then they will not be visible on platforms which do not support this feature. ---------------------------------------------------------------------- Time in seconds to get/set one xattr of N bytes on 100,000 files ------+--------------------------------+------------------------------ | setxattr | getxattr bytes | ext4 xfs zfs-dir zfs-sa | ext4 xfs zfs-dir zfs-sa ------+--------------------------------+------------------------------ 1 | 2.33 31.88 21.50 4.57 | 2.35 2.64 6.29 2.43 32 | 2.79 30.68 21.98 4.60 | 2.44 2.59 6.78 2.48 256 | 3.25 31.99 21.36 5.92 | 2.32 2.71 6.22 3.14 1024 | 3.30 32.61 22.83 8.45 | 2.40 2.79 6.24 3.27 4096 | 3.57 317.46 22.52 10.73 | 2.78 28.62 6.90 3.94 16384 | n/a 2342.39 34.30 19.20 | n/a 45.44 145.90 7.55 65536 | n/a 2941.39 128.15 131.32* | n/a 141.92 256.85 262.12* Legend: * ext4 - Stock RHEL6.1 ext4 mounted with '-o user_xattr'. * xfs - Stock RHEL6.1 xfs mounted with default options. * zfs-dir - Directory based xattrs only. * zfs-sa - Prefer SAs but spill in to directories as needed, a trailing * indicates overflow in to directories occured. NOTE: Ext4 supports 4096 bytes of xattr name/value pairs per file. NOTE: XFS and ZFS have no limit on xattr name/value pairs per file. NOTE: Linux limits individual name/value pairs to 65536 bytes. NOTE: All setattr/getattr's were done after dropping the cache. NOTE: All tests were run against a single hard drive. Signed-off-by: Brian Behlendorf <behlendorf1@llnl.gov> Issue #443
2011-10-24 23:55:20 +00:00
* The value under Linux to limited to 65536 bytes of binary data.
* In practice, individual xattrs tend to be much smaller than this
* and are typically less than 100 bytes. A good example of this
* are the security.selinux xattrs which are less than 100 bytes and
* exist for every file when xattr labeling is enabled.
Add Hooks for Linux Xattr Operations The Linux specific xattr operations have all been located in the file zpl_xattr.c. These functions primarily rely on the reworked zfs_* functions to do their job. They are also responsible for converting the possible Solaris style error codes to negative Linux errors.
2011-01-26 20:10:01 +00:00
*
Implement SA based xattrs The current ZFS implementation stores xattrs on disk using a hidden directory. In this directory a file name represents the xattr name and the file contexts are the xattr binary data. This approach is very flexible and allows for arbitrarily large xattrs. However, it also suffers from a significant performance penalty. Accessing a single xattr can requires up to three disk seeks. 1) Lookup the dnode object. 2) Lookup the dnodes's xattr directory object. 3) Lookup the xattr object in the directory. To avoid this performance penalty Linux filesystems such as ext3 and xfs try to store the xattr as part of the inode on disk. When the xattr is to large to store in the inode then a single external block is allocated for them. In practice most xattrs are small and this approach works well. The addition of System Attributes (SA) to zfs provides us a clean way to make this optimization. When the dataset property 'xattr=sa' is set then xattrs will be preferentially stored as System Attributes. This allows tiny xattrs (~100 bytes) to be stored with the dnode and up to 64k of xattrs to be stored in the spill block. If additional xattr space is required, which is unlikely under Linux, they will be stored using the traditional directory approach. This optimization results in roughly a 3x performance improvement when accessing xattrs which brings zfs roughly to parity with ext4 and xfs (see table below). When multiple xattrs are stored per-file the performance improvements are even greater because all of the xattrs stored in the spill block will be cached. However, by default SA based xattrs are disabled in the Linux port to maximize compatibility with other implementations. If you do enable SA based xattrs then they will not be visible on platforms which do not support this feature. ---------------------------------------------------------------------- Time in seconds to get/set one xattr of N bytes on 100,000 files ------+--------------------------------+------------------------------ | setxattr | getxattr bytes | ext4 xfs zfs-dir zfs-sa | ext4 xfs zfs-dir zfs-sa ------+--------------------------------+------------------------------ 1 | 2.33 31.88 21.50 4.57 | 2.35 2.64 6.29 2.43 32 | 2.79 30.68 21.98 4.60 | 2.44 2.59 6.78 2.48 256 | 3.25 31.99 21.36 5.92 | 2.32 2.71 6.22 3.14 1024 | 3.30 32.61 22.83 8.45 | 2.40 2.79 6.24 3.27 4096 | 3.57 317.46 22.52 10.73 | 2.78 28.62 6.90 3.94 16384 | n/a 2342.39 34.30 19.20 | n/a 45.44 145.90 7.55 65536 | n/a 2941.39 128.15 131.32* | n/a 141.92 256.85 262.12* Legend: * ext4 - Stock RHEL6.1 ext4 mounted with '-o user_xattr'. * xfs - Stock RHEL6.1 xfs mounted with default options. * zfs-dir - Directory based xattrs only. * zfs-sa - Prefer SAs but spill in to directories as needed, a trailing * indicates overflow in to directories occured. NOTE: Ext4 supports 4096 bytes of xattr name/value pairs per file. NOTE: XFS and ZFS have no limit on xattr name/value pairs per file. NOTE: Linux limits individual name/value pairs to 65536 bytes. NOTE: All setattr/getattr's were done after dropping the cache. NOTE: All tests were run against a single hard drive. Signed-off-by: Brian Behlendorf <behlendorf1@llnl.gov> Issue #443
2011-10-24 23:55:20 +00:00
* The Linux xattr implemenation has been written to take advantage of
* this typical usage. When the dataset property 'xattr=sa' is set,
* then xattrs will be preferentially stored as System Attributes (SA).
* This allows tiny xattrs (~100 bytes) to be stored with the dnode and
* up to 64k of xattrs to be stored in the spill block. If additional
* xattr space is required, which is unlikely under Linux, they will
* be stored using the traditional directory approach.
Add Hooks for Linux Xattr Operations The Linux specific xattr operations have all been located in the file zpl_xattr.c. These functions primarily rely on the reworked zfs_* functions to do their job. They are also responsible for converting the possible Solaris style error codes to negative Linux errors.
2011-01-26 20:10:01 +00:00
*
Implement SA based xattrs The current ZFS implementation stores xattrs on disk using a hidden directory. In this directory a file name represents the xattr name and the file contexts are the xattr binary data. This approach is very flexible and allows for arbitrarily large xattrs. However, it also suffers from a significant performance penalty. Accessing a single xattr can requires up to three disk seeks. 1) Lookup the dnode object. 2) Lookup the dnodes's xattr directory object. 3) Lookup the xattr object in the directory. To avoid this performance penalty Linux filesystems such as ext3 and xfs try to store the xattr as part of the inode on disk. When the xattr is to large to store in the inode then a single external block is allocated for them. In practice most xattrs are small and this approach works well. The addition of System Attributes (SA) to zfs provides us a clean way to make this optimization. When the dataset property 'xattr=sa' is set then xattrs will be preferentially stored as System Attributes. This allows tiny xattrs (~100 bytes) to be stored with the dnode and up to 64k of xattrs to be stored in the spill block. If additional xattr space is required, which is unlikely under Linux, they will be stored using the traditional directory approach. This optimization results in roughly a 3x performance improvement when accessing xattrs which brings zfs roughly to parity with ext4 and xfs (see table below). When multiple xattrs are stored per-file the performance improvements are even greater because all of the xattrs stored in the spill block will be cached. However, by default SA based xattrs are disabled in the Linux port to maximize compatibility with other implementations. If you do enable SA based xattrs then they will not be visible on platforms which do not support this feature. ---------------------------------------------------------------------- Time in seconds to get/set one xattr of N bytes on 100,000 files ------+--------------------------------+------------------------------ | setxattr | getxattr bytes | ext4 xfs zfs-dir zfs-sa | ext4 xfs zfs-dir zfs-sa ------+--------------------------------+------------------------------ 1 | 2.33 31.88 21.50 4.57 | 2.35 2.64 6.29 2.43 32 | 2.79 30.68 21.98 4.60 | 2.44 2.59 6.78 2.48 256 | 3.25 31.99 21.36 5.92 | 2.32 2.71 6.22 3.14 1024 | 3.30 32.61 22.83 8.45 | 2.40 2.79 6.24 3.27 4096 | 3.57 317.46 22.52 10.73 | 2.78 28.62 6.90 3.94 16384 | n/a 2342.39 34.30 19.20 | n/a 45.44 145.90 7.55 65536 | n/a 2941.39 128.15 131.32* | n/a 141.92 256.85 262.12* Legend: * ext4 - Stock RHEL6.1 ext4 mounted with '-o user_xattr'. * xfs - Stock RHEL6.1 xfs mounted with default options. * zfs-dir - Directory based xattrs only. * zfs-sa - Prefer SAs but spill in to directories as needed, a trailing * indicates overflow in to directories occured. NOTE: Ext4 supports 4096 bytes of xattr name/value pairs per file. NOTE: XFS and ZFS have no limit on xattr name/value pairs per file. NOTE: Linux limits individual name/value pairs to 65536 bytes. NOTE: All setattr/getattr's were done after dropping the cache. NOTE: All tests were run against a single hard drive. Signed-off-by: Brian Behlendorf <behlendorf1@llnl.gov> Issue #443
2011-10-24 23:55:20 +00:00
* This optimization results in roughly a 3x performance improvement
* when accessing xattrs because it avoids the need to perform a seek
* for every xattr value. When multiple xattrs are stored per-file
* the performance improvements are even greater because all of the
* xattrs stored in the spill block will be cached.
*
* However, by default SA based xattrs are disabled in the Linux port
* to maximize compatibility with other implementations. If you do
* enable SA based xattrs then they will not be visible on platforms
* which do not support this feature.
*
* NOTE: One additional consequence of the xattr directory implementation
* is that when an extended attribute is manipulated an inode is created.
* This inode will exist in the Linux inode cache but there will be no
* associated entry in the dentry cache which references it. This is
* safe but it may result in some confusion. Enabling SA based xattrs
* largely avoids the issue except in the overflow case.
Add Hooks for Linux Xattr Operations The Linux specific xattr operations have all been located in the file zpl_xattr.c. These functions primarily rely on the reworked zfs_* functions to do their job. They are also responsible for converting the possible Solaris style error codes to negative Linux errors.
2011-01-26 20:10:01 +00:00
*/
#include <sys/zfs_vfsops.h>
#include <sys/zfs_vnops.h>
#include <sys/zfs_znode.h>
Linux 3.11 compat: fops->iterate() Commit torvalds/linux@2233f31aade393641f0eaed43a71110e629bb900 replaced ->readdir() with ->iterate() in struct file_operations. All filesystems must now use the new ->iterate method. To handle this the code was reworked to use the new ->iterate interface. Care was taken to keep the majority of changes confined to the ZPL layer which is already Linux specific. However, minor changes were required to the common zfs_readdir() function. Compatibility with older kernels was accomplished by adding versions of the trivial dir_emit* helper functions. Also the various *_readdir() functions were reworked in to wrappers which create a dir_context structure to pass to the new *_iterate() functions. Unfortunately, the new dir_emit* functions prevent us from passing a private pointer to the filldir function. The xattr directory code leveraged this ability through zfs_readdir() to generate the list of xattr names. Since we can no longer use zfs_readdir() a simplified zpl_xattr_readdir() function was added to perform the same task. Signed-off-by: Richard Yao <ryao@cs.stonybrook.edu> Signed-off-by: Brian Behlendorf <behlendorf1@llnl.gov> Closes #1653 Issue #1591
2013-08-07 12:53:45 +00:00
#include <sys/zap.h>
Add Hooks for Linux Xattr Operations The Linux specific xattr operations have all been located in the file zpl_xattr.c. These functions primarily rely on the reworked zfs_* functions to do their job. They are also responsible for converting the possible Solaris style error codes to negative Linux errors.
2011-01-26 20:10:01 +00:00
#include <sys/vfs.h>
#include <sys/zpl.h>
typedef struct xattr_filldir {
size_t size;
size_t offset;
char *buf;
struct inode *inode;
} xattr_filldir_t;
static int
Linux 3.11 compat: fops->iterate() Commit torvalds/linux@2233f31aade393641f0eaed43a71110e629bb900 replaced ->readdir() with ->iterate() in struct file_operations. All filesystems must now use the new ->iterate method. To handle this the code was reworked to use the new ->iterate interface. Care was taken to keep the majority of changes confined to the ZPL layer which is already Linux specific. However, minor changes were required to the common zfs_readdir() function. Compatibility with older kernels was accomplished by adding versions of the trivial dir_emit* helper functions. Also the various *_readdir() functions were reworked in to wrappers which create a dir_context structure to pass to the new *_iterate() functions. Unfortunately, the new dir_emit* functions prevent us from passing a private pointer to the filldir function. The xattr directory code leveraged this ability through zfs_readdir() to generate the list of xattr names. Since we can no longer use zfs_readdir() a simplified zpl_xattr_readdir() function was added to perform the same task. Signed-off-by: Richard Yao <ryao@cs.stonybrook.edu> Signed-off-by: Brian Behlendorf <behlendorf1@llnl.gov> Closes #1653 Issue #1591
2013-08-07 12:53:45 +00:00
zpl_xattr_filldir(xattr_filldir_t *xf, const char *name, int name_len)
Add Hooks for Linux Xattr Operations The Linux specific xattr operations have all been located in the file zpl_xattr.c. These functions primarily rely on the reworked zfs_* functions to do their job. They are also responsible for converting the possible Solaris style error codes to negative Linux errors.
2011-01-26 20:10:01 +00:00
{
if (!strncmp(name, XATTR_USER_PREFIX, XATTR_USER_PREFIX_LEN))
Linux compat 2.6.39: mount_nodev() The .get_sb callback has been replaced by a .mount callback in the file_system_type structure. When using the new interface the caller must now use the mount_nodev() helper. Unfortunately, the new interface no longer passes the vfsmount down to the zfs layers. This poses a problem for the existing implementation because we currently save this pointer in the super block for latter use. It provides our only entry point in to the namespace layer for manipulating certain mount options. This needed to be done originally to allow commands like 'zfs set atime=off tank' to work properly. It also allowed me to keep more of the original Solaris code unmodified. Under Solaris there is a 1-to-1 mapping between a mount point and a file system so this is a fairly natural thing to do. However, under Linux they many be multiple entries in the namespace which reference the same filesystem. Thus keeping a back reference from the filesystem to the namespace is complicated. Rather than introduce some ugly hack to get the vfsmount and continue as before. I'm leveraging this API change to update the ZFS code to do things in a more natural way for Linux. This has the upside that is resolves the compatibility issue for the long term and fixes several other minor bugs which have been reported. This commit updates the code to remove this vfsmount back reference entirely. All modifications to filesystem mount options are now passed in to the kernel via a '-o remount'. This is the expected Linux mechanism and allows the namespace to properly handle any options which apply to it before passing them on to the file system itself. Aside from fixing the compatibility issue, removing the vfsmount has had the benefit of simplifying the code. This change which fairly involved has turned out nicely. Closes #246 Closes #217 Closes #187 Closes #248 Closes #231
2011-05-19 18:44:07 +00:00
if (!(ITOZSB(xf->inode)->z_flags & ZSB_XATTR))
Add Hooks for Linux Xattr Operations The Linux specific xattr operations have all been located in the file zpl_xattr.c. These functions primarily rely on the reworked zfs_* functions to do their job. They are also responsible for converting the possible Solaris style error codes to negative Linux errors.
2011-01-26 20:10:01 +00:00
return (0);
if (!strncmp(name, XATTR_TRUSTED_PREFIX, XATTR_TRUSTED_PREFIX_LEN))
if (!capable(CAP_SYS_ADMIN))
return (0);
/* When xf->buf is NULL only calculate the required size. */
if (xf->buf) {
if (xf->offset + name_len + 1 > xf->size)
return (-ERANGE);
memcpy(xf->buf + xf->offset, name, name_len);
xf->buf[xf->offset + name_len] = '\0';
}
xf->offset += (name_len + 1);
return (0);
}
Linux 3.11 compat: fops->iterate() Commit torvalds/linux@2233f31aade393641f0eaed43a71110e629bb900 replaced ->readdir() with ->iterate() in struct file_operations. All filesystems must now use the new ->iterate method. To handle this the code was reworked to use the new ->iterate interface. Care was taken to keep the majority of changes confined to the ZPL layer which is already Linux specific. However, minor changes were required to the common zfs_readdir() function. Compatibility with older kernels was accomplished by adding versions of the trivial dir_emit* helper functions. Also the various *_readdir() functions were reworked in to wrappers which create a dir_context structure to pass to the new *_iterate() functions. Unfortunately, the new dir_emit* functions prevent us from passing a private pointer to the filldir function. The xattr directory code leveraged this ability through zfs_readdir() to generate the list of xattr names. Since we can no longer use zfs_readdir() a simplified zpl_xattr_readdir() function was added to perform the same task. Signed-off-by: Richard Yao <ryao@cs.stonybrook.edu> Signed-off-by: Brian Behlendorf <behlendorf1@llnl.gov> Closes #1653 Issue #1591
2013-08-07 12:53:45 +00:00
/*
* Read as many directory entry names as will fit in to the provided buffer,
* or when no buffer is provided calculate the required buffer size.
*/
int
zpl_xattr_readdir(struct inode *dxip, xattr_filldir_t *xf)
{
zap_cursor_t zc;
zap_attribute_t zap;
int error;
zap_cursor_init(&zc, ITOZSB(dxip)->z_os, ITOZ(dxip)->z_id);
while ((error = -zap_cursor_retrieve(&zc, &zap)) == 0) {
if (zap.za_integer_length != 8 || zap.za_num_integers != 1) {
error = -ENXIO;
break;
}
error = zpl_xattr_filldir(xf, zap.za_name, strlen(zap.za_name));
if (error)
break;
zap_cursor_advance(&zc);
}
zap_cursor_fini(&zc);
if (error == -ENOENT)
error = 0;
return (error);
}
Implement SA based xattrs The current ZFS implementation stores xattrs on disk using a hidden directory. In this directory a file name represents the xattr name and the file contexts are the xattr binary data. This approach is very flexible and allows for arbitrarily large xattrs. However, it also suffers from a significant performance penalty. Accessing a single xattr can requires up to three disk seeks. 1) Lookup the dnode object. 2) Lookup the dnodes's xattr directory object. 3) Lookup the xattr object in the directory. To avoid this performance penalty Linux filesystems such as ext3 and xfs try to store the xattr as part of the inode on disk. When the xattr is to large to store in the inode then a single external block is allocated for them. In practice most xattrs are small and this approach works well. The addition of System Attributes (SA) to zfs provides us a clean way to make this optimization. When the dataset property 'xattr=sa' is set then xattrs will be preferentially stored as System Attributes. This allows tiny xattrs (~100 bytes) to be stored with the dnode and up to 64k of xattrs to be stored in the spill block. If additional xattr space is required, which is unlikely under Linux, they will be stored using the traditional directory approach. This optimization results in roughly a 3x performance improvement when accessing xattrs which brings zfs roughly to parity with ext4 and xfs (see table below). When multiple xattrs are stored per-file the performance improvements are even greater because all of the xattrs stored in the spill block will be cached. However, by default SA based xattrs are disabled in the Linux port to maximize compatibility with other implementations. If you do enable SA based xattrs then they will not be visible on platforms which do not support this feature. ---------------------------------------------------------------------- Time in seconds to get/set one xattr of N bytes on 100,000 files ------+--------------------------------+------------------------------ | setxattr | getxattr bytes | ext4 xfs zfs-dir zfs-sa | ext4 xfs zfs-dir zfs-sa ------+--------------------------------+------------------------------ 1 | 2.33 31.88 21.50 4.57 | 2.35 2.64 6.29 2.43 32 | 2.79 30.68 21.98 4.60 | 2.44 2.59 6.78 2.48 256 | 3.25 31.99 21.36 5.92 | 2.32 2.71 6.22 3.14 1024 | 3.30 32.61 22.83 8.45 | 2.40 2.79 6.24 3.27 4096 | 3.57 317.46 22.52 10.73 | 2.78 28.62 6.90 3.94 16384 | n/a 2342.39 34.30 19.20 | n/a 45.44 145.90 7.55 65536 | n/a 2941.39 128.15 131.32* | n/a 141.92 256.85 262.12* Legend: * ext4 - Stock RHEL6.1 ext4 mounted with '-o user_xattr'. * xfs - Stock RHEL6.1 xfs mounted with default options. * zfs-dir - Directory based xattrs only. * zfs-sa - Prefer SAs but spill in to directories as needed, a trailing * indicates overflow in to directories occured. NOTE: Ext4 supports 4096 bytes of xattr name/value pairs per file. NOTE: XFS and ZFS have no limit on xattr name/value pairs per file. NOTE: Linux limits individual name/value pairs to 65536 bytes. NOTE: All setattr/getattr's were done after dropping the cache. NOTE: All tests were run against a single hard drive. Signed-off-by: Brian Behlendorf <behlendorf1@llnl.gov> Issue #443
2011-10-24 23:55:20 +00:00
static ssize_t
zpl_xattr_list_dir(xattr_filldir_t *xf, cred_t *cr)
Add Hooks for Linux Xattr Operations The Linux specific xattr operations have all been located in the file zpl_xattr.c. These functions primarily rely on the reworked zfs_* functions to do their job. They are also responsible for converting the possible Solaris style error codes to negative Linux errors.
2011-01-26 20:10:01 +00:00
{
Implement SA based xattrs The current ZFS implementation stores xattrs on disk using a hidden directory. In this directory a file name represents the xattr name and the file contexts are the xattr binary data. This approach is very flexible and allows for arbitrarily large xattrs. However, it also suffers from a significant performance penalty. Accessing a single xattr can requires up to three disk seeks. 1) Lookup the dnode object. 2) Lookup the dnodes's xattr directory object. 3) Lookup the xattr object in the directory. To avoid this performance penalty Linux filesystems such as ext3 and xfs try to store the xattr as part of the inode on disk. When the xattr is to large to store in the inode then a single external block is allocated for them. In practice most xattrs are small and this approach works well. The addition of System Attributes (SA) to zfs provides us a clean way to make this optimization. When the dataset property 'xattr=sa' is set then xattrs will be preferentially stored as System Attributes. This allows tiny xattrs (~100 bytes) to be stored with the dnode and up to 64k of xattrs to be stored in the spill block. If additional xattr space is required, which is unlikely under Linux, they will be stored using the traditional directory approach. This optimization results in roughly a 3x performance improvement when accessing xattrs which brings zfs roughly to parity with ext4 and xfs (see table below). When multiple xattrs are stored per-file the performance improvements are even greater because all of the xattrs stored in the spill block will be cached. However, by default SA based xattrs are disabled in the Linux port to maximize compatibility with other implementations. If you do enable SA based xattrs then they will not be visible on platforms which do not support this feature. ---------------------------------------------------------------------- Time in seconds to get/set one xattr of N bytes on 100,000 files ------+--------------------------------+------------------------------ | setxattr | getxattr bytes | ext4 xfs zfs-dir zfs-sa | ext4 xfs zfs-dir zfs-sa ------+--------------------------------+------------------------------ 1 | 2.33 31.88 21.50 4.57 | 2.35 2.64 6.29 2.43 32 | 2.79 30.68 21.98 4.60 | 2.44 2.59 6.78 2.48 256 | 3.25 31.99 21.36 5.92 | 2.32 2.71 6.22 3.14 1024 | 3.30 32.61 22.83 8.45 | 2.40 2.79 6.24 3.27 4096 | 3.57 317.46 22.52 10.73 | 2.78 28.62 6.90 3.94 16384 | n/a 2342.39 34.30 19.20 | n/a 45.44 145.90 7.55 65536 | n/a 2941.39 128.15 131.32* | n/a 141.92 256.85 262.12* Legend: * ext4 - Stock RHEL6.1 ext4 mounted with '-o user_xattr'. * xfs - Stock RHEL6.1 xfs mounted with default options. * zfs-dir - Directory based xattrs only. * zfs-sa - Prefer SAs but spill in to directories as needed, a trailing * indicates overflow in to directories occured. NOTE: Ext4 supports 4096 bytes of xattr name/value pairs per file. NOTE: XFS and ZFS have no limit on xattr name/value pairs per file. NOTE: Linux limits individual name/value pairs to 65536 bytes. NOTE: All setattr/getattr's were done after dropping the cache. NOTE: All tests were run against a single hard drive. Signed-off-by: Brian Behlendorf <behlendorf1@llnl.gov> Issue #443
2011-10-24 23:55:20 +00:00
struct inode *ip = xf->inode;
Add Hooks for Linux Xattr Operations The Linux specific xattr operations have all been located in the file zpl_xattr.c. These functions primarily rely on the reworked zfs_* functions to do their job. They are also responsible for converting the possible Solaris style error codes to negative Linux errors.
2011-01-26 20:10:01 +00:00
struct inode *dxip = NULL;
int error;
/* Lookup the xattr directory */
error = -zfs_lookup(ip, NULL, &dxip, LOOKUP_XATTR, cr, NULL, NULL);
if (error) {
if (error == -ENOENT)
error = 0;
Implement SA based xattrs The current ZFS implementation stores xattrs on disk using a hidden directory. In this directory a file name represents the xattr name and the file contexts are the xattr binary data. This approach is very flexible and allows for arbitrarily large xattrs. However, it also suffers from a significant performance penalty. Accessing a single xattr can requires up to three disk seeks. 1) Lookup the dnode object. 2) Lookup the dnodes's xattr directory object. 3) Lookup the xattr object in the directory. To avoid this performance penalty Linux filesystems such as ext3 and xfs try to store the xattr as part of the inode on disk. When the xattr is to large to store in the inode then a single external block is allocated for them. In practice most xattrs are small and this approach works well. The addition of System Attributes (SA) to zfs provides us a clean way to make this optimization. When the dataset property 'xattr=sa' is set then xattrs will be preferentially stored as System Attributes. This allows tiny xattrs (~100 bytes) to be stored with the dnode and up to 64k of xattrs to be stored in the spill block. If additional xattr space is required, which is unlikely under Linux, they will be stored using the traditional directory approach. This optimization results in roughly a 3x performance improvement when accessing xattrs which brings zfs roughly to parity with ext4 and xfs (see table below). When multiple xattrs are stored per-file the performance improvements are even greater because all of the xattrs stored in the spill block will be cached. However, by default SA based xattrs are disabled in the Linux port to maximize compatibility with other implementations. If you do enable SA based xattrs then they will not be visible on platforms which do not support this feature. ---------------------------------------------------------------------- Time in seconds to get/set one xattr of N bytes on 100,000 files ------+--------------------------------+------------------------------ | setxattr | getxattr bytes | ext4 xfs zfs-dir zfs-sa | ext4 xfs zfs-dir zfs-sa ------+--------------------------------+------------------------------ 1 | 2.33 31.88 21.50 4.57 | 2.35 2.64 6.29 2.43 32 | 2.79 30.68 21.98 4.60 | 2.44 2.59 6.78 2.48 256 | 3.25 31.99 21.36 5.92 | 2.32 2.71 6.22 3.14 1024 | 3.30 32.61 22.83 8.45 | 2.40 2.79 6.24 3.27 4096 | 3.57 317.46 22.52 10.73 | 2.78 28.62 6.90 3.94 16384 | n/a 2342.39 34.30 19.20 | n/a 45.44 145.90 7.55 65536 | n/a 2941.39 128.15 131.32* | n/a 141.92 256.85 262.12* Legend: * ext4 - Stock RHEL6.1 ext4 mounted with '-o user_xattr'. * xfs - Stock RHEL6.1 xfs mounted with default options. * zfs-dir - Directory based xattrs only. * zfs-sa - Prefer SAs but spill in to directories as needed, a trailing * indicates overflow in to directories occured. NOTE: Ext4 supports 4096 bytes of xattr name/value pairs per file. NOTE: XFS and ZFS have no limit on xattr name/value pairs per file. NOTE: Linux limits individual name/value pairs to 65536 bytes. NOTE: All setattr/getattr's were done after dropping the cache. NOTE: All tests were run against a single hard drive. Signed-off-by: Brian Behlendorf <behlendorf1@llnl.gov> Issue #443
2011-10-24 23:55:20 +00:00
return (error);
Add Hooks for Linux Xattr Operations The Linux specific xattr operations have all been located in the file zpl_xattr.c. These functions primarily rely on the reworked zfs_* functions to do their job. They are also responsible for converting the possible Solaris style error codes to negative Linux errors.
2011-01-26 20:10:01 +00:00
}
Linux 3.11 compat: fops->iterate() Commit torvalds/linux@2233f31aade393641f0eaed43a71110e629bb900 replaced ->readdir() with ->iterate() in struct file_operations. All filesystems must now use the new ->iterate method. To handle this the code was reworked to use the new ->iterate interface. Care was taken to keep the majority of changes confined to the ZPL layer which is already Linux specific. However, minor changes were required to the common zfs_readdir() function. Compatibility with older kernels was accomplished by adding versions of the trivial dir_emit* helper functions. Also the various *_readdir() functions were reworked in to wrappers which create a dir_context structure to pass to the new *_iterate() functions. Unfortunately, the new dir_emit* functions prevent us from passing a private pointer to the filldir function. The xattr directory code leveraged this ability through zfs_readdir() to generate the list of xattr names. Since we can no longer use zfs_readdir() a simplified zpl_xattr_readdir() function was added to perform the same task. Signed-off-by: Richard Yao <ryao@cs.stonybrook.edu> Signed-off-by: Brian Behlendorf <behlendorf1@llnl.gov> Closes #1653 Issue #1591
2013-08-07 12:53:45 +00:00
error = zpl_xattr_readdir(dxip, xf);
Implement SA based xattrs The current ZFS implementation stores xattrs on disk using a hidden directory. In this directory a file name represents the xattr name and the file contexts are the xattr binary data. This approach is very flexible and allows for arbitrarily large xattrs. However, it also suffers from a significant performance penalty. Accessing a single xattr can requires up to three disk seeks. 1) Lookup the dnode object. 2) Lookup the dnodes's xattr directory object. 3) Lookup the xattr object in the directory. To avoid this performance penalty Linux filesystems such as ext3 and xfs try to store the xattr as part of the inode on disk. When the xattr is to large to store in the inode then a single external block is allocated for them. In practice most xattrs are small and this approach works well. The addition of System Attributes (SA) to zfs provides us a clean way to make this optimization. When the dataset property 'xattr=sa' is set then xattrs will be preferentially stored as System Attributes. This allows tiny xattrs (~100 bytes) to be stored with the dnode and up to 64k of xattrs to be stored in the spill block. If additional xattr space is required, which is unlikely under Linux, they will be stored using the traditional directory approach. This optimization results in roughly a 3x performance improvement when accessing xattrs which brings zfs roughly to parity with ext4 and xfs (see table below). When multiple xattrs are stored per-file the performance improvements are even greater because all of the xattrs stored in the spill block will be cached. However, by default SA based xattrs are disabled in the Linux port to maximize compatibility with other implementations. If you do enable SA based xattrs then they will not be visible on platforms which do not support this feature. ---------------------------------------------------------------------- Time in seconds to get/set one xattr of N bytes on 100,000 files ------+--------------------------------+------------------------------ | setxattr | getxattr bytes | ext4 xfs zfs-dir zfs-sa | ext4 xfs zfs-dir zfs-sa ------+--------------------------------+------------------------------ 1 | 2.33 31.88 21.50 4.57 | 2.35 2.64 6.29 2.43 32 | 2.79 30.68 21.98 4.60 | 2.44 2.59 6.78 2.48 256 | 3.25 31.99 21.36 5.92 | 2.32 2.71 6.22 3.14 1024 | 3.30 32.61 22.83 8.45 | 2.40 2.79 6.24 3.27 4096 | 3.57 317.46 22.52 10.73 | 2.78 28.62 6.90 3.94 16384 | n/a 2342.39 34.30 19.20 | n/a 45.44 145.90 7.55 65536 | n/a 2941.39 128.15 131.32* | n/a 141.92 256.85 262.12* Legend: * ext4 - Stock RHEL6.1 ext4 mounted with '-o user_xattr'. * xfs - Stock RHEL6.1 xfs mounted with default options. * zfs-dir - Directory based xattrs only. * zfs-sa - Prefer SAs but spill in to directories as needed, a trailing * indicates overflow in to directories occured. NOTE: Ext4 supports 4096 bytes of xattr name/value pairs per file. NOTE: XFS and ZFS have no limit on xattr name/value pairs per file. NOTE: Linux limits individual name/value pairs to 65536 bytes. NOTE: All setattr/getattr's were done after dropping the cache. NOTE: All tests were run against a single hard drive. Signed-off-by: Brian Behlendorf <behlendorf1@llnl.gov> Issue #443
2011-10-24 23:55:20 +00:00
iput(dxip);
return (error);
}
static ssize_t
zpl_xattr_list_sa(xattr_filldir_t *xf)
{
znode_t *zp = ITOZ(xf->inode);
nvpair_t *nvp = NULL;
int error = 0;
mutex_enter(&zp->z_lock);
if (zp->z_xattr_cached == NULL)
error = -zfs_sa_get_xattr(zp);
mutex_exit(&zp->z_lock);
if (error)
return (error);
ASSERT(zp->z_xattr_cached);
while ((nvp = nvlist_next_nvpair(zp->z_xattr_cached, nvp)) != NULL) {
ASSERT3U(nvpair_type(nvp), ==, DATA_TYPE_BYTE_ARRAY);
Linux 3.11 compat: fops->iterate() Commit torvalds/linux@2233f31aade393641f0eaed43a71110e629bb900 replaced ->readdir() with ->iterate() in struct file_operations. All filesystems must now use the new ->iterate method. To handle this the code was reworked to use the new ->iterate interface. Care was taken to keep the majority of changes confined to the ZPL layer which is already Linux specific. However, minor changes were required to the common zfs_readdir() function. Compatibility with older kernels was accomplished by adding versions of the trivial dir_emit* helper functions. Also the various *_readdir() functions were reworked in to wrappers which create a dir_context structure to pass to the new *_iterate() functions. Unfortunately, the new dir_emit* functions prevent us from passing a private pointer to the filldir function. The xattr directory code leveraged this ability through zfs_readdir() to generate the list of xattr names. Since we can no longer use zfs_readdir() a simplified zpl_xattr_readdir() function was added to perform the same task. Signed-off-by: Richard Yao <ryao@cs.stonybrook.edu> Signed-off-by: Brian Behlendorf <behlendorf1@llnl.gov> Closes #1653 Issue #1591
2013-08-07 12:53:45 +00:00
error = zpl_xattr_filldir(xf, nvpair_name(nvp),
strlen(nvpair_name(nvp)));
Implement SA based xattrs The current ZFS implementation stores xattrs on disk using a hidden directory. In this directory a file name represents the xattr name and the file contexts are the xattr binary data. This approach is very flexible and allows for arbitrarily large xattrs. However, it also suffers from a significant performance penalty. Accessing a single xattr can requires up to three disk seeks. 1) Lookup the dnode object. 2) Lookup the dnodes's xattr directory object. 3) Lookup the xattr object in the directory. To avoid this performance penalty Linux filesystems such as ext3 and xfs try to store the xattr as part of the inode on disk. When the xattr is to large to store in the inode then a single external block is allocated for them. In practice most xattrs are small and this approach works well. The addition of System Attributes (SA) to zfs provides us a clean way to make this optimization. When the dataset property 'xattr=sa' is set then xattrs will be preferentially stored as System Attributes. This allows tiny xattrs (~100 bytes) to be stored with the dnode and up to 64k of xattrs to be stored in the spill block. If additional xattr space is required, which is unlikely under Linux, they will be stored using the traditional directory approach. This optimization results in roughly a 3x performance improvement when accessing xattrs which brings zfs roughly to parity with ext4 and xfs (see table below). When multiple xattrs are stored per-file the performance improvements are even greater because all of the xattrs stored in the spill block will be cached. However, by default SA based xattrs are disabled in the Linux port to maximize compatibility with other implementations. If you do enable SA based xattrs then they will not be visible on platforms which do not support this feature. ---------------------------------------------------------------------- Time in seconds to get/set one xattr of N bytes on 100,000 files ------+--------------------------------+------------------------------ | setxattr | getxattr bytes | ext4 xfs zfs-dir zfs-sa | ext4 xfs zfs-dir zfs-sa ------+--------------------------------+------------------------------ 1 | 2.33 31.88 21.50 4.57 | 2.35 2.64 6.29 2.43 32 | 2.79 30.68 21.98 4.60 | 2.44 2.59 6.78 2.48 256 | 3.25 31.99 21.36 5.92 | 2.32 2.71 6.22 3.14 1024 | 3.30 32.61 22.83 8.45 | 2.40 2.79 6.24 3.27 4096 | 3.57 317.46 22.52 10.73 | 2.78 28.62 6.90 3.94 16384 | n/a 2342.39 34.30 19.20 | n/a 45.44 145.90 7.55 65536 | n/a 2941.39 128.15 131.32* | n/a 141.92 256.85 262.12* Legend: * ext4 - Stock RHEL6.1 ext4 mounted with '-o user_xattr'. * xfs - Stock RHEL6.1 xfs mounted with default options. * zfs-dir - Directory based xattrs only. * zfs-sa - Prefer SAs but spill in to directories as needed, a trailing * indicates overflow in to directories occured. NOTE: Ext4 supports 4096 bytes of xattr name/value pairs per file. NOTE: XFS and ZFS have no limit on xattr name/value pairs per file. NOTE: Linux limits individual name/value pairs to 65536 bytes. NOTE: All setattr/getattr's were done after dropping the cache. NOTE: All tests were run against a single hard drive. Signed-off-by: Brian Behlendorf <behlendorf1@llnl.gov> Issue #443
2011-10-24 23:55:20 +00:00
if (error)
return (error);
}
return (0);
}
ssize_t
zpl_xattr_list(struct dentry *dentry, char *buffer, size_t buffer_size)
{
znode_t *zp = ITOZ(dentry->d_inode);
zfs_sb_t *zsb = ZTOZSB(zp);
xattr_filldir_t xf = { buffer_size, 0, buffer, dentry->d_inode };
cred_t *cr = CRED();
int error = 0;
crhold(cr);
rw_enter(&zp->z_xattr_lock, RW_READER);
if (zsb->z_use_sa && zp->z_is_sa) {
error = zpl_xattr_list_sa(&xf);
if (error)
goto out;
}
error = zpl_xattr_list_dir(&xf, cr);
Add Hooks for Linux Xattr Operations The Linux specific xattr operations have all been located in the file zpl_xattr.c. These functions primarily rely on the reworked zfs_* functions to do their job. They are also responsible for converting the possible Solaris style error codes to negative Linux errors.
2011-01-26 20:10:01 +00:00
if (error)
goto out;
error = xf.offset;
out:
Implement SA based xattrs The current ZFS implementation stores xattrs on disk using a hidden directory. In this directory a file name represents the xattr name and the file contexts are the xattr binary data. This approach is very flexible and allows for arbitrarily large xattrs. However, it also suffers from a significant performance penalty. Accessing a single xattr can requires up to three disk seeks. 1) Lookup the dnode object. 2) Lookup the dnodes's xattr directory object. 3) Lookup the xattr object in the directory. To avoid this performance penalty Linux filesystems such as ext3 and xfs try to store the xattr as part of the inode on disk. When the xattr is to large to store in the inode then a single external block is allocated for them. In practice most xattrs are small and this approach works well. The addition of System Attributes (SA) to zfs provides us a clean way to make this optimization. When the dataset property 'xattr=sa' is set then xattrs will be preferentially stored as System Attributes. This allows tiny xattrs (~100 bytes) to be stored with the dnode and up to 64k of xattrs to be stored in the spill block. If additional xattr space is required, which is unlikely under Linux, they will be stored using the traditional directory approach. This optimization results in roughly a 3x performance improvement when accessing xattrs which brings zfs roughly to parity with ext4 and xfs (see table below). When multiple xattrs are stored per-file the performance improvements are even greater because all of the xattrs stored in the spill block will be cached. However, by default SA based xattrs are disabled in the Linux port to maximize compatibility with other implementations. If you do enable SA based xattrs then they will not be visible on platforms which do not support this feature. ---------------------------------------------------------------------- Time in seconds to get/set one xattr of N bytes on 100,000 files ------+--------------------------------+------------------------------ | setxattr | getxattr bytes | ext4 xfs zfs-dir zfs-sa | ext4 xfs zfs-dir zfs-sa ------+--------------------------------+------------------------------ 1 | 2.33 31.88 21.50 4.57 | 2.35 2.64 6.29 2.43 32 | 2.79 30.68 21.98 4.60 | 2.44 2.59 6.78 2.48 256 | 3.25 31.99 21.36 5.92 | 2.32 2.71 6.22 3.14 1024 | 3.30 32.61 22.83 8.45 | 2.40 2.79 6.24 3.27 4096 | 3.57 317.46 22.52 10.73 | 2.78 28.62 6.90 3.94 16384 | n/a 2342.39 34.30 19.20 | n/a 45.44 145.90 7.55 65536 | n/a 2941.39 128.15 131.32* | n/a 141.92 256.85 262.12* Legend: * ext4 - Stock RHEL6.1 ext4 mounted with '-o user_xattr'. * xfs - Stock RHEL6.1 xfs mounted with default options. * zfs-dir - Directory based xattrs only. * zfs-sa - Prefer SAs but spill in to directories as needed, a trailing * indicates overflow in to directories occured. NOTE: Ext4 supports 4096 bytes of xattr name/value pairs per file. NOTE: XFS and ZFS have no limit on xattr name/value pairs per file. NOTE: Linux limits individual name/value pairs to 65536 bytes. NOTE: All setattr/getattr's were done after dropping the cache. NOTE: All tests were run against a single hard drive. Signed-off-by: Brian Behlendorf <behlendorf1@llnl.gov> Issue #443
2011-10-24 23:55:20 +00:00
rw_exit(&zp->z_xattr_lock);
Linux 2.6.29 compat, credentials As of Linux 2.6.29 a clean credential API was added to the Linux kernel. Previously the credential was embedded in the task_struct. Because the SPL already has considerable support for handling this API change the ZPL code has been updated to use the Solaris credential API.
2011-03-22 18:13:41 +00:00
crfree(cr);
Add Hooks for Linux Xattr Operations The Linux specific xattr operations have all been located in the file zpl_xattr.c. These functions primarily rely on the reworked zfs_* functions to do their job. They are also responsible for converting the possible Solaris style error codes to negative Linux errors.
2011-01-26 20:10:01 +00:00
return (error);
}
static int
Implement SA based xattrs The current ZFS implementation stores xattrs on disk using a hidden directory. In this directory a file name represents the xattr name and the file contexts are the xattr binary data. This approach is very flexible and allows for arbitrarily large xattrs. However, it also suffers from a significant performance penalty. Accessing a single xattr can requires up to three disk seeks. 1) Lookup the dnode object. 2) Lookup the dnodes's xattr directory object. 3) Lookup the xattr object in the directory. To avoid this performance penalty Linux filesystems such as ext3 and xfs try to store the xattr as part of the inode on disk. When the xattr is to large to store in the inode then a single external block is allocated for them. In practice most xattrs are small and this approach works well. The addition of System Attributes (SA) to zfs provides us a clean way to make this optimization. When the dataset property 'xattr=sa' is set then xattrs will be preferentially stored as System Attributes. This allows tiny xattrs (~100 bytes) to be stored with the dnode and up to 64k of xattrs to be stored in the spill block. If additional xattr space is required, which is unlikely under Linux, they will be stored using the traditional directory approach. This optimization results in roughly a 3x performance improvement when accessing xattrs which brings zfs roughly to parity with ext4 and xfs (see table below). When multiple xattrs are stored per-file the performance improvements are even greater because all of the xattrs stored in the spill block will be cached. However, by default SA based xattrs are disabled in the Linux port to maximize compatibility with other implementations. If you do enable SA based xattrs then they will not be visible on platforms which do not support this feature. ---------------------------------------------------------------------- Time in seconds to get/set one xattr of N bytes on 100,000 files ------+--------------------------------+------------------------------ | setxattr | getxattr bytes | ext4 xfs zfs-dir zfs-sa | ext4 xfs zfs-dir zfs-sa ------+--------------------------------+------------------------------ 1 | 2.33 31.88 21.50 4.57 | 2.35 2.64 6.29 2.43 32 | 2.79 30.68 21.98 4.60 | 2.44 2.59 6.78 2.48 256 | 3.25 31.99 21.36 5.92 | 2.32 2.71 6.22 3.14 1024 | 3.30 32.61 22.83 8.45 | 2.40 2.79 6.24 3.27 4096 | 3.57 317.46 22.52 10.73 | 2.78 28.62 6.90 3.94 16384 | n/a 2342.39 34.30 19.20 | n/a 45.44 145.90 7.55 65536 | n/a 2941.39 128.15 131.32* | n/a 141.92 256.85 262.12* Legend: * ext4 - Stock RHEL6.1 ext4 mounted with '-o user_xattr'. * xfs - Stock RHEL6.1 xfs mounted with default options. * zfs-dir - Directory based xattrs only. * zfs-sa - Prefer SAs but spill in to directories as needed, a trailing * indicates overflow in to directories occured. NOTE: Ext4 supports 4096 bytes of xattr name/value pairs per file. NOTE: XFS and ZFS have no limit on xattr name/value pairs per file. NOTE: Linux limits individual name/value pairs to 65536 bytes. NOTE: All setattr/getattr's were done after dropping the cache. NOTE: All tests were run against a single hard drive. Signed-off-by: Brian Behlendorf <behlendorf1@llnl.gov> Issue #443
2011-10-24 23:55:20 +00:00
zpl_xattr_get_dir(struct inode *ip, const char *name, void *value,
size_t size, cred_t *cr)
Add Hooks for Linux Xattr Operations The Linux specific xattr operations have all been located in the file zpl_xattr.c. These functions primarily rely on the reworked zfs_* functions to do their job. They are also responsible for converting the possible Solaris style error codes to negative Linux errors.
2011-01-26 20:10:01 +00:00
{
struct inode *dxip = NULL;
struct inode *xip = NULL;
int error;
/* Lookup the xattr directory */
error = -zfs_lookup(ip, NULL, &dxip, LOOKUP_XATTR, cr, NULL, NULL);
if (error)
goto out;
/* Lookup a specific xattr name in the directory */
error = -zfs_lookup(dxip, (char *)name, &xip, 0, cr, NULL, NULL);
if (error)
goto out;
if (!size) {
error = i_size_read(xip);
goto out;
}
Correctly return ERANGE in getxattr(2) According to the getxattr(2) man page the ERANGE errno should be returned when the size of the value buffer is to small to hold the result. Prior to this patch the implementation would just truncate the value to size bytes. Signed-off-by: Brian Behlendorf <behlendorf1@llnl.gov> Closes #1408
2013-04-17 20:07:36 +00:00
if (size < i_size_read(xip)) {
error = -ERANGE;
goto out;
}
Implement SA based xattrs The current ZFS implementation stores xattrs on disk using a hidden directory. In this directory a file name represents the xattr name and the file contexts are the xattr binary data. This approach is very flexible and allows for arbitrarily large xattrs. However, it also suffers from a significant performance penalty. Accessing a single xattr can requires up to three disk seeks. 1) Lookup the dnode object. 2) Lookup the dnodes's xattr directory object. 3) Lookup the xattr object in the directory. To avoid this performance penalty Linux filesystems such as ext3 and xfs try to store the xattr as part of the inode on disk. When the xattr is to large to store in the inode then a single external block is allocated for them. In practice most xattrs are small and this approach works well. The addition of System Attributes (SA) to zfs provides us a clean way to make this optimization. When the dataset property 'xattr=sa' is set then xattrs will be preferentially stored as System Attributes. This allows tiny xattrs (~100 bytes) to be stored with the dnode and up to 64k of xattrs to be stored in the spill block. If additional xattr space is required, which is unlikely under Linux, they will be stored using the traditional directory approach. This optimization results in roughly a 3x performance improvement when accessing xattrs which brings zfs roughly to parity with ext4 and xfs (see table below). When multiple xattrs are stored per-file the performance improvements are even greater because all of the xattrs stored in the spill block will be cached. However, by default SA based xattrs are disabled in the Linux port to maximize compatibility with other implementations. If you do enable SA based xattrs then they will not be visible on platforms which do not support this feature. ---------------------------------------------------------------------- Time in seconds to get/set one xattr of N bytes on 100,000 files ------+--------------------------------+------------------------------ | setxattr | getxattr bytes | ext4 xfs zfs-dir zfs-sa | ext4 xfs zfs-dir zfs-sa ------+--------------------------------+------------------------------ 1 | 2.33 31.88 21.50 4.57 | 2.35 2.64 6.29 2.43 32 | 2.79 30.68 21.98 4.60 | 2.44 2.59 6.78 2.48 256 | 3.25 31.99 21.36 5.92 | 2.32 2.71 6.22 3.14 1024 | 3.30 32.61 22.83 8.45 | 2.40 2.79 6.24 3.27 4096 | 3.57 317.46 22.52 10.73 | 2.78 28.62 6.90 3.94 16384 | n/a 2342.39 34.30 19.20 | n/a 45.44 145.90 7.55 65536 | n/a 2941.39 128.15 131.32* | n/a 141.92 256.85 262.12* Legend: * ext4 - Stock RHEL6.1 ext4 mounted with '-o user_xattr'. * xfs - Stock RHEL6.1 xfs mounted with default options. * zfs-dir - Directory based xattrs only. * zfs-sa - Prefer SAs but spill in to directories as needed, a trailing * indicates overflow in to directories occured. NOTE: Ext4 supports 4096 bytes of xattr name/value pairs per file. NOTE: XFS and ZFS have no limit on xattr name/value pairs per file. NOTE: Linux limits individual name/value pairs to 65536 bytes. NOTE: All setattr/getattr's were done after dropping the cache. NOTE: All tests were run against a single hard drive. Signed-off-by: Brian Behlendorf <behlendorf1@llnl.gov> Issue #443
2011-10-24 23:55:20 +00:00
error = zpl_read_common(xip, value, size, 0, UIO_SYSSPACE, 0, cr);
Add Hooks for Linux Xattr Operations The Linux specific xattr operations have all been located in the file zpl_xattr.c. These functions primarily rely on the reworked zfs_* functions to do their job. They are also responsible for converting the possible Solaris style error codes to negative Linux errors.
2011-01-26 20:10:01 +00:00
out:
if (xip)
iput(xip);
if (dxip)
iput(dxip);
Implement SA based xattrs The current ZFS implementation stores xattrs on disk using a hidden directory. In this directory a file name represents the xattr name and the file contexts are the xattr binary data. This approach is very flexible and allows for arbitrarily large xattrs. However, it also suffers from a significant performance penalty. Accessing a single xattr can requires up to three disk seeks. 1) Lookup the dnode object. 2) Lookup the dnodes's xattr directory object. 3) Lookup the xattr object in the directory. To avoid this performance penalty Linux filesystems such as ext3 and xfs try to store the xattr as part of the inode on disk. When the xattr is to large to store in the inode then a single external block is allocated for them. In practice most xattrs are small and this approach works well. The addition of System Attributes (SA) to zfs provides us a clean way to make this optimization. When the dataset property 'xattr=sa' is set then xattrs will be preferentially stored as System Attributes. This allows tiny xattrs (~100 bytes) to be stored with the dnode and up to 64k of xattrs to be stored in the spill block. If additional xattr space is required, which is unlikely under Linux, they will be stored using the traditional directory approach. This optimization results in roughly a 3x performance improvement when accessing xattrs which brings zfs roughly to parity with ext4 and xfs (see table below). When multiple xattrs are stored per-file the performance improvements are even greater because all of the xattrs stored in the spill block will be cached. However, by default SA based xattrs are disabled in the Linux port to maximize compatibility with other implementations. If you do enable SA based xattrs then they will not be visible on platforms which do not support this feature. ---------------------------------------------------------------------- Time in seconds to get/set one xattr of N bytes on 100,000 files ------+--------------------------------+------------------------------ | setxattr | getxattr bytes | ext4 xfs zfs-dir zfs-sa | ext4 xfs zfs-dir zfs-sa ------+--------------------------------+------------------------------ 1 | 2.33 31.88 21.50 4.57 | 2.35 2.64 6.29 2.43 32 | 2.79 30.68 21.98 4.60 | 2.44 2.59 6.78 2.48 256 | 3.25 31.99 21.36 5.92 | 2.32 2.71 6.22 3.14 1024 | 3.30 32.61 22.83 8.45 | 2.40 2.79 6.24 3.27 4096 | 3.57 317.46 22.52 10.73 | 2.78 28.62 6.90 3.94 16384 | n/a 2342.39 34.30 19.20 | n/a 45.44 145.90 7.55 65536 | n/a 2941.39 128.15 131.32* | n/a 141.92 256.85 262.12* Legend: * ext4 - Stock RHEL6.1 ext4 mounted with '-o user_xattr'. * xfs - Stock RHEL6.1 xfs mounted with default options. * zfs-dir - Directory based xattrs only. * zfs-sa - Prefer SAs but spill in to directories as needed, a trailing * indicates overflow in to directories occured. NOTE: Ext4 supports 4096 bytes of xattr name/value pairs per file. NOTE: XFS and ZFS have no limit on xattr name/value pairs per file. NOTE: Linux limits individual name/value pairs to 65536 bytes. NOTE: All setattr/getattr's were done after dropping the cache. NOTE: All tests were run against a single hard drive. Signed-off-by: Brian Behlendorf <behlendorf1@llnl.gov> Issue #443
2011-10-24 23:55:20 +00:00
return (error);
}
static int
zpl_xattr_get_sa(struct inode *ip, const char *name, void *value, size_t size)
{
znode_t *zp = ITOZ(ip);
uchar_t *nv_value;
uint_t nv_size;
int error = 0;
Add Hooks for Linux Xattr Operations The Linux specific xattr operations have all been located in the file zpl_xattr.c. These functions primarily rely on the reworked zfs_* functions to do their job. They are also responsible for converting the possible Solaris style error codes to negative Linux errors.
2011-01-26 20:10:01 +00:00
Implement SA based xattrs The current ZFS implementation stores xattrs on disk using a hidden directory. In this directory a file name represents the xattr name and the file contexts are the xattr binary data. This approach is very flexible and allows for arbitrarily large xattrs. However, it also suffers from a significant performance penalty. Accessing a single xattr can requires up to three disk seeks. 1) Lookup the dnode object. 2) Lookup the dnodes's xattr directory object. 3) Lookup the xattr object in the directory. To avoid this performance penalty Linux filesystems such as ext3 and xfs try to store the xattr as part of the inode on disk. When the xattr is to large to store in the inode then a single external block is allocated for them. In practice most xattrs are small and this approach works well. The addition of System Attributes (SA) to zfs provides us a clean way to make this optimization. When the dataset property 'xattr=sa' is set then xattrs will be preferentially stored as System Attributes. This allows tiny xattrs (~100 bytes) to be stored with the dnode and up to 64k of xattrs to be stored in the spill block. If additional xattr space is required, which is unlikely under Linux, they will be stored using the traditional directory approach. This optimization results in roughly a 3x performance improvement when accessing xattrs which brings zfs roughly to parity with ext4 and xfs (see table below). When multiple xattrs are stored per-file the performance improvements are even greater because all of the xattrs stored in the spill block will be cached. However, by default SA based xattrs are disabled in the Linux port to maximize compatibility with other implementations. If you do enable SA based xattrs then they will not be visible on platforms which do not support this feature. ---------------------------------------------------------------------- Time in seconds to get/set one xattr of N bytes on 100,000 files ------+--------------------------------+------------------------------ | setxattr | getxattr bytes | ext4 xfs zfs-dir zfs-sa | ext4 xfs zfs-dir zfs-sa ------+--------------------------------+------------------------------ 1 | 2.33 31.88 21.50 4.57 | 2.35 2.64 6.29 2.43 32 | 2.79 30.68 21.98 4.60 | 2.44 2.59 6.78 2.48 256 | 3.25 31.99 21.36 5.92 | 2.32 2.71 6.22 3.14 1024 | 3.30 32.61 22.83 8.45 | 2.40 2.79 6.24 3.27 4096 | 3.57 317.46 22.52 10.73 | 2.78 28.62 6.90 3.94 16384 | n/a 2342.39 34.30 19.20 | n/a 45.44 145.90 7.55 65536 | n/a 2941.39 128.15 131.32* | n/a 141.92 256.85 262.12* Legend: * ext4 - Stock RHEL6.1 ext4 mounted with '-o user_xattr'. * xfs - Stock RHEL6.1 xfs mounted with default options. * zfs-dir - Directory based xattrs only. * zfs-sa - Prefer SAs but spill in to directories as needed, a trailing * indicates overflow in to directories occured. NOTE: Ext4 supports 4096 bytes of xattr name/value pairs per file. NOTE: XFS and ZFS have no limit on xattr name/value pairs per file. NOTE: Linux limits individual name/value pairs to 65536 bytes. NOTE: All setattr/getattr's were done after dropping the cache. NOTE: All tests were run against a single hard drive. Signed-off-by: Brian Behlendorf <behlendorf1@llnl.gov> Issue #443
2011-10-24 23:55:20 +00:00
ASSERT(RW_LOCK_HELD(&zp->z_xattr_lock));
mutex_enter(&zp->z_lock);
if (zp->z_xattr_cached == NULL)
error = -zfs_sa_get_xattr(zp);
mutex_exit(&zp->z_lock);
if (error)
return (error);
ASSERT(zp->z_xattr_cached);
error = -nvlist_lookup_byte_array(zp->z_xattr_cached, name,
&nv_value, &nv_size);
if (error)
return (error);
if (!size)
return (nv_size);
Correctly return ERANGE in getxattr(2) According to the getxattr(2) man page the ERANGE errno should be returned when the size of the value buffer is to small to hold the result. Prior to this patch the implementation would just truncate the value to size bytes. Signed-off-by: Brian Behlendorf <behlendorf1@llnl.gov> Closes #1408
2013-04-17 20:07:36 +00:00
if (size < nv_size)
return (-ERANGE);
memcpy(value, nv_value, nv_size);
Implement SA based xattrs The current ZFS implementation stores xattrs on disk using a hidden directory. In this directory a file name represents the xattr name and the file contexts are the xattr binary data. This approach is very flexible and allows for arbitrarily large xattrs. However, it also suffers from a significant performance penalty. Accessing a single xattr can requires up to three disk seeks. 1) Lookup the dnode object. 2) Lookup the dnodes's xattr directory object. 3) Lookup the xattr object in the directory. To avoid this performance penalty Linux filesystems such as ext3 and xfs try to store the xattr as part of the inode on disk. When the xattr is to large to store in the inode then a single external block is allocated for them. In practice most xattrs are small and this approach works well. The addition of System Attributes (SA) to zfs provides us a clean way to make this optimization. When the dataset property 'xattr=sa' is set then xattrs will be preferentially stored as System Attributes. This allows tiny xattrs (~100 bytes) to be stored with the dnode and up to 64k of xattrs to be stored in the spill block. If additional xattr space is required, which is unlikely under Linux, they will be stored using the traditional directory approach. This optimization results in roughly a 3x performance improvement when accessing xattrs which brings zfs roughly to parity with ext4 and xfs (see table below). When multiple xattrs are stored per-file the performance improvements are even greater because all of the xattrs stored in the spill block will be cached. However, by default SA based xattrs are disabled in the Linux port to maximize compatibility with other implementations. If you do enable SA based xattrs then they will not be visible on platforms which do not support this feature. ---------------------------------------------------------------------- Time in seconds to get/set one xattr of N bytes on 100,000 files ------+--------------------------------+------------------------------ | setxattr | getxattr bytes | ext4 xfs zfs-dir zfs-sa | ext4 xfs zfs-dir zfs-sa ------+--------------------------------+------------------------------ 1 | 2.33 31.88 21.50 4.57 | 2.35 2.64 6.29 2.43 32 | 2.79 30.68 21.98 4.60 | 2.44 2.59 6.78 2.48 256 | 3.25 31.99 21.36 5.92 | 2.32 2.71 6.22 3.14 1024 | 3.30 32.61 22.83 8.45 | 2.40 2.79 6.24 3.27 4096 | 3.57 317.46 22.52 10.73 | 2.78 28.62 6.90 3.94 16384 | n/a 2342.39 34.30 19.20 | n/a 45.44 145.90 7.55 65536 | n/a 2941.39 128.15 131.32* | n/a 141.92 256.85 262.12* Legend: * ext4 - Stock RHEL6.1 ext4 mounted with '-o user_xattr'. * xfs - Stock RHEL6.1 xfs mounted with default options. * zfs-dir - Directory based xattrs only. * zfs-sa - Prefer SAs but spill in to directories as needed, a trailing * indicates overflow in to directories occured. NOTE: Ext4 supports 4096 bytes of xattr name/value pairs per file. NOTE: XFS and ZFS have no limit on xattr name/value pairs per file. NOTE: Linux limits individual name/value pairs to 65536 bytes. NOTE: All setattr/getattr's were done after dropping the cache. NOTE: All tests were run against a single hard drive. Signed-off-by: Brian Behlendorf <behlendorf1@llnl.gov> Issue #443
2011-10-24 23:55:20 +00:00
Correctly return ERANGE in getxattr(2) According to the getxattr(2) man page the ERANGE errno should be returned when the size of the value buffer is to small to hold the result. Prior to this patch the implementation would just truncate the value to size bytes. Signed-off-by: Brian Behlendorf <behlendorf1@llnl.gov> Closes #1408
2013-04-17 20:07:36 +00:00
return (nv_size);
Implement SA based xattrs The current ZFS implementation stores xattrs on disk using a hidden directory. In this directory a file name represents the xattr name and the file contexts are the xattr binary data. This approach is very flexible and allows for arbitrarily large xattrs. However, it also suffers from a significant performance penalty. Accessing a single xattr can requires up to three disk seeks. 1) Lookup the dnode object. 2) Lookup the dnodes's xattr directory object. 3) Lookup the xattr object in the directory. To avoid this performance penalty Linux filesystems such as ext3 and xfs try to store the xattr as part of the inode on disk. When the xattr is to large to store in the inode then a single external block is allocated for them. In practice most xattrs are small and this approach works well. The addition of System Attributes (SA) to zfs provides us a clean way to make this optimization. When the dataset property 'xattr=sa' is set then xattrs will be preferentially stored as System Attributes. This allows tiny xattrs (~100 bytes) to be stored with the dnode and up to 64k of xattrs to be stored in the spill block. If additional xattr space is required, which is unlikely under Linux, they will be stored using the traditional directory approach. This optimization results in roughly a 3x performance improvement when accessing xattrs which brings zfs roughly to parity with ext4 and xfs (see table below). When multiple xattrs are stored per-file the performance improvements are even greater because all of the xattrs stored in the spill block will be cached. However, by default SA based xattrs are disabled in the Linux port to maximize compatibility with other implementations. If you do enable SA based xattrs then they will not be visible on platforms which do not support this feature. ---------------------------------------------------------------------- Time in seconds to get/set one xattr of N bytes on 100,000 files ------+--------------------------------+------------------------------ | setxattr | getxattr bytes | ext4 xfs zfs-dir zfs-sa | ext4 xfs zfs-dir zfs-sa ------+--------------------------------+------------------------------ 1 | 2.33 31.88 21.50 4.57 | 2.35 2.64 6.29 2.43 32 | 2.79 30.68 21.98 4.60 | 2.44 2.59 6.78 2.48 256 | 3.25 31.99 21.36 5.92 | 2.32 2.71 6.22 3.14 1024 | 3.30 32.61 22.83 8.45 | 2.40 2.79 6.24 3.27 4096 | 3.57 317.46 22.52 10.73 | 2.78 28.62 6.90 3.94 16384 | n/a 2342.39 34.30 19.20 | n/a 45.44 145.90 7.55 65536 | n/a 2941.39 128.15 131.32* | n/a 141.92 256.85 262.12* Legend: * ext4 - Stock RHEL6.1 ext4 mounted with '-o user_xattr'. * xfs - Stock RHEL6.1 xfs mounted with default options. * zfs-dir - Directory based xattrs only. * zfs-sa - Prefer SAs but spill in to directories as needed, a trailing * indicates overflow in to directories occured. NOTE: Ext4 supports 4096 bytes of xattr name/value pairs per file. NOTE: XFS and ZFS have no limit on xattr name/value pairs per file. NOTE: Linux limits individual name/value pairs to 65536 bytes. NOTE: All setattr/getattr's were done after dropping the cache. NOTE: All tests were run against a single hard drive. Signed-off-by: Brian Behlendorf <behlendorf1@llnl.gov> Issue #443
2011-10-24 23:55:20 +00:00
}
static int
__zpl_xattr_get(struct inode *ip, const char *name, void *value, size_t size,
cred_t *cr)
{
znode_t *zp = ITOZ(ip);
zfs_sb_t *zsb = ZTOZSB(zp);
int error;
ASSERT(RW_LOCK_HELD(&zp->z_xattr_lock));
if (zsb->z_use_sa && zp->z_is_sa) {
error = zpl_xattr_get_sa(ip, name, value, size);
Only check directory xattr on ENOENT When SA xattrs are enabled only fallback to checking the directory xattrs when the name is not found as a SA xattr. Otherwise, the SA error which should be returned to the caller is overwritten by the directory xattr errors. Positive return values indicating success will also be immediately returned. In the case of #1437 the ERANGE error was being correctly returned by zpl_xattr_get_sa() only to be overridden with ENOENT which was returned by the subsequent unnessisary call to zpl_xattr_get_dir(). Signed-off-by: Brian Behlendorf <behlendorf1@llnl.gov> Closes #1437
2013-05-08 16:20:04 +00:00
if (error != -ENOENT)
Implement SA based xattrs The current ZFS implementation stores xattrs on disk using a hidden directory. In this directory a file name represents the xattr name and the file contexts are the xattr binary data. This approach is very flexible and allows for arbitrarily large xattrs. However, it also suffers from a significant performance penalty. Accessing a single xattr can requires up to three disk seeks. 1) Lookup the dnode object. 2) Lookup the dnodes's xattr directory object. 3) Lookup the xattr object in the directory. To avoid this performance penalty Linux filesystems such as ext3 and xfs try to store the xattr as part of the inode on disk. When the xattr is to large to store in the inode then a single external block is allocated for them. In practice most xattrs are small and this approach works well. The addition of System Attributes (SA) to zfs provides us a clean way to make this optimization. When the dataset property 'xattr=sa' is set then xattrs will be preferentially stored as System Attributes. This allows tiny xattrs (~100 bytes) to be stored with the dnode and up to 64k of xattrs to be stored in the spill block. If additional xattr space is required, which is unlikely under Linux, they will be stored using the traditional directory approach. This optimization results in roughly a 3x performance improvement when accessing xattrs which brings zfs roughly to parity with ext4 and xfs (see table below). When multiple xattrs are stored per-file the performance improvements are even greater because all of the xattrs stored in the spill block will be cached. However, by default SA based xattrs are disabled in the Linux port to maximize compatibility with other implementations. If you do enable SA based xattrs then they will not be visible on platforms which do not support this feature. ---------------------------------------------------------------------- Time in seconds to get/set one xattr of N bytes on 100,000 files ------+--------------------------------+------------------------------ | setxattr | getxattr bytes | ext4 xfs zfs-dir zfs-sa | ext4 xfs zfs-dir zfs-sa ------+--------------------------------+------------------------------ 1 | 2.33 31.88 21.50 4.57 | 2.35 2.64 6.29 2.43 32 | 2.79 30.68 21.98 4.60 | 2.44 2.59 6.78 2.48 256 | 3.25 31.99 21.36 5.92 | 2.32 2.71 6.22 3.14 1024 | 3.30 32.61 22.83 8.45 | 2.40 2.79 6.24 3.27 4096 | 3.57 317.46 22.52 10.73 | 2.78 28.62 6.90 3.94 16384 | n/a 2342.39 34.30 19.20 | n/a 45.44 145.90 7.55 65536 | n/a 2941.39 128.15 131.32* | n/a 141.92 256.85 262.12* Legend: * ext4 - Stock RHEL6.1 ext4 mounted with '-o user_xattr'. * xfs - Stock RHEL6.1 xfs mounted with default options. * zfs-dir - Directory based xattrs only. * zfs-sa - Prefer SAs but spill in to directories as needed, a trailing * indicates overflow in to directories occured. NOTE: Ext4 supports 4096 bytes of xattr name/value pairs per file. NOTE: XFS and ZFS have no limit on xattr name/value pairs per file. NOTE: Linux limits individual name/value pairs to 65536 bytes. NOTE: All setattr/getattr's were done after dropping the cache. NOTE: All tests were run against a single hard drive. Signed-off-by: Brian Behlendorf <behlendorf1@llnl.gov> Issue #443
2011-10-24 23:55:20 +00:00
goto out;
}
error = zpl_xattr_get_dir(ip, name, value, size, cr);
out:
Add Hooks for Linux Xattr Operations The Linux specific xattr operations have all been located in the file zpl_xattr.c. These functions primarily rely on the reworked zfs_* functions to do their job. They are also responsible for converting the possible Solaris style error codes to negative Linux errors.
2011-01-26 20:10:01 +00:00
if (error == -ENOENT)
error = -ENODATA;
return (error);
}
static int
Implement SA based xattrs The current ZFS implementation stores xattrs on disk using a hidden directory. In this directory a file name represents the xattr name and the file contexts are the xattr binary data. This approach is very flexible and allows for arbitrarily large xattrs. However, it also suffers from a significant performance penalty. Accessing a single xattr can requires up to three disk seeks. 1) Lookup the dnode object. 2) Lookup the dnodes's xattr directory object. 3) Lookup the xattr object in the directory. To avoid this performance penalty Linux filesystems such as ext3 and xfs try to store the xattr as part of the inode on disk. When the xattr is to large to store in the inode then a single external block is allocated for them. In practice most xattrs are small and this approach works well. The addition of System Attributes (SA) to zfs provides us a clean way to make this optimization. When the dataset property 'xattr=sa' is set then xattrs will be preferentially stored as System Attributes. This allows tiny xattrs (~100 bytes) to be stored with the dnode and up to 64k of xattrs to be stored in the spill block. If additional xattr space is required, which is unlikely under Linux, they will be stored using the traditional directory approach. This optimization results in roughly a 3x performance improvement when accessing xattrs which brings zfs roughly to parity with ext4 and xfs (see table below). When multiple xattrs are stored per-file the performance improvements are even greater because all of the xattrs stored in the spill block will be cached. However, by default SA based xattrs are disabled in the Linux port to maximize compatibility with other implementations. If you do enable SA based xattrs then they will not be visible on platforms which do not support this feature. ---------------------------------------------------------------------- Time in seconds to get/set one xattr of N bytes on 100,000 files ------+--------------------------------+------------------------------ | setxattr | getxattr bytes | ext4 xfs zfs-dir zfs-sa | ext4 xfs zfs-dir zfs-sa ------+--------------------------------+------------------------------ 1 | 2.33 31.88 21.50 4.57 | 2.35 2.64 6.29 2.43 32 | 2.79 30.68 21.98 4.60 | 2.44 2.59 6.78 2.48 256 | 3.25 31.99 21.36 5.92 | 2.32 2.71 6.22 3.14 1024 | 3.30 32.61 22.83 8.45 | 2.40 2.79 6.24 3.27 4096 | 3.57 317.46 22.52 10.73 | 2.78 28.62 6.90 3.94 16384 | n/a 2342.39 34.30 19.20 | n/a 45.44 145.90 7.55 65536 | n/a 2941.39 128.15 131.32* | n/a 141.92 256.85 262.12* Legend: * ext4 - Stock RHEL6.1 ext4 mounted with '-o user_xattr'. * xfs - Stock RHEL6.1 xfs mounted with default options. * zfs-dir - Directory based xattrs only. * zfs-sa - Prefer SAs but spill in to directories as needed, a trailing * indicates overflow in to directories occured. NOTE: Ext4 supports 4096 bytes of xattr name/value pairs per file. NOTE: XFS and ZFS have no limit on xattr name/value pairs per file. NOTE: Linux limits individual name/value pairs to 65536 bytes. NOTE: All setattr/getattr's were done after dropping the cache. NOTE: All tests were run against a single hard drive. Signed-off-by: Brian Behlendorf <behlendorf1@llnl.gov> Issue #443
2011-10-24 23:55:20 +00:00
zpl_xattr_get(struct inode *ip, const char *name, void *value, size_t size)
{
znode_t *zp = ITOZ(ip);
cred_t *cr = CRED();
int error;
crhold(cr);
rw_enter(&zp->z_xattr_lock, RW_READER);
error = __zpl_xattr_get(ip, name, value, size, cr);
rw_exit(&zp->z_xattr_lock);
crfree(cr);
return (error);
}
static int
zpl_xattr_set_dir(struct inode *ip, const char *name, const void *value,
size_t size, int flags, cred_t *cr)
Add Hooks for Linux Xattr Operations The Linux specific xattr operations have all been located in the file zpl_xattr.c. These functions primarily rely on the reworked zfs_* functions to do their job. They are also responsible for converting the possible Solaris style error codes to negative Linux errors.
2011-01-26 20:10:01 +00:00
{
struct inode *dxip = NULL;
struct inode *xip = NULL;
vattr_t *vap = NULL;
ssize_t wrote;
Prevent xattr remove from creating xattr directory Attempting to remove an xattr from a file which does not contain any directory based xattrs would result in the xattr directory being created. This behavior is non-optimal because it results in write operations to the pool in addition to the expected error being returned. To prevent this the CREATE_XATTR_DIR flag is only passed in zpl_xattr_set_dir() when setting a non-NULL xattr value. In addition, zpl_xattr_set() is updated similarly such that it will return immediately if passed an xattr name which doesn't exist and a NULL value. Signed-off-by: Massimo Maggi <me@massimo-maggi.eu> Signed-off-by: Richard Yao <ryao@gentoo.org> Signed-off-by: Brian Behlendorf <behlendorf1@llnl.gov> Issue #170
2013-10-28 16:07:00 +00:00
int lookup_flags, error;
Truncate the xattr znode when updating existing attributes. If the attribute's new value was shorter than the old one the old code would leave parts of the old value in the xattr znode. Signed-off-by: Brian Behlendorf <behlendorf1@llnl.gov> Closes #203
2011-04-17 18:31:33 +00:00
const int xattr_mode = S_IFREG | 0644;
Add Hooks for Linux Xattr Operations The Linux specific xattr operations have all been located in the file zpl_xattr.c. These functions primarily rely on the reworked zfs_* functions to do their job. They are also responsible for converting the possible Solaris style error codes to negative Linux errors.
2011-01-26 20:10:01 +00:00
Prevent xattr remove from creating xattr directory Attempting to remove an xattr from a file which does not contain any directory based xattrs would result in the xattr directory being created. This behavior is non-optimal because it results in write operations to the pool in addition to the expected error being returned. To prevent this the CREATE_XATTR_DIR flag is only passed in zpl_xattr_set_dir() when setting a non-NULL xattr value. In addition, zpl_xattr_set() is updated similarly such that it will return immediately if passed an xattr name which doesn't exist and a NULL value. Signed-off-by: Massimo Maggi <me@massimo-maggi.eu> Signed-off-by: Richard Yao <ryao@gentoo.org> Signed-off-by: Brian Behlendorf <behlendorf1@llnl.gov> Issue #170
2013-10-28 16:07:00 +00:00
/*
* Lookup the xattr directory. When we're adding an entry pass
* CREATE_XATTR_DIR to ensure the xattr directory is created.
* When removing an entry this flag is not passed to avoid
* unnecessarily creating a new xattr directory.
*/
lookup_flags = LOOKUP_XATTR;
if (value != NULL)
lookup_flags |= CREATE_XATTR_DIR;
error = -zfs_lookup(ip, NULL, &dxip, lookup_flags, cr, NULL, NULL);
Add Hooks for Linux Xattr Operations The Linux specific xattr operations have all been located in the file zpl_xattr.c. These functions primarily rely on the reworked zfs_* functions to do their job. They are also responsible for converting the possible Solaris style error codes to negative Linux errors.
2011-01-26 20:10:01 +00:00
if (error)
goto out;
Implement SA based xattrs The current ZFS implementation stores xattrs on disk using a hidden directory. In this directory a file name represents the xattr name and the file contexts are the xattr binary data. This approach is very flexible and allows for arbitrarily large xattrs. However, it also suffers from a significant performance penalty. Accessing a single xattr can requires up to three disk seeks. 1) Lookup the dnode object. 2) Lookup the dnodes's xattr directory object. 3) Lookup the xattr object in the directory. To avoid this performance penalty Linux filesystems such as ext3 and xfs try to store the xattr as part of the inode on disk. When the xattr is to large to store in the inode then a single external block is allocated for them. In practice most xattrs are small and this approach works well. The addition of System Attributes (SA) to zfs provides us a clean way to make this optimization. When the dataset property 'xattr=sa' is set then xattrs will be preferentially stored as System Attributes. This allows tiny xattrs (~100 bytes) to be stored with the dnode and up to 64k of xattrs to be stored in the spill block. If additional xattr space is required, which is unlikely under Linux, they will be stored using the traditional directory approach. This optimization results in roughly a 3x performance improvement when accessing xattrs which brings zfs roughly to parity with ext4 and xfs (see table below). When multiple xattrs are stored per-file the performance improvements are even greater because all of the xattrs stored in the spill block will be cached. However, by default SA based xattrs are disabled in the Linux port to maximize compatibility with other implementations. If you do enable SA based xattrs then they will not be visible on platforms which do not support this feature. ---------------------------------------------------------------------- Time in seconds to get/set one xattr of N bytes on 100,000 files ------+--------------------------------+------------------------------ | setxattr | getxattr bytes | ext4 xfs zfs-dir zfs-sa | ext4 xfs zfs-dir zfs-sa ------+--------------------------------+------------------------------ 1 | 2.33 31.88 21.50 4.57 | 2.35 2.64 6.29 2.43 32 | 2.79 30.68 21.98 4.60 | 2.44 2.59 6.78 2.48 256 | 3.25 31.99 21.36 5.92 | 2.32 2.71 6.22 3.14 1024 | 3.30 32.61 22.83 8.45 | 2.40 2.79 6.24 3.27 4096 | 3.57 317.46 22.52 10.73 | 2.78 28.62 6.90 3.94 16384 | n/a 2342.39 34.30 19.20 | n/a 45.44 145.90 7.55 65536 | n/a 2941.39 128.15 131.32* | n/a 141.92 256.85 262.12* Legend: * ext4 - Stock RHEL6.1 ext4 mounted with '-o user_xattr'. * xfs - Stock RHEL6.1 xfs mounted with default options. * zfs-dir - Directory based xattrs only. * zfs-sa - Prefer SAs but spill in to directories as needed, a trailing * indicates overflow in to directories occured. NOTE: Ext4 supports 4096 bytes of xattr name/value pairs per file. NOTE: XFS and ZFS have no limit on xattr name/value pairs per file. NOTE: Linux limits individual name/value pairs to 65536 bytes. NOTE: All setattr/getattr's were done after dropping the cache. NOTE: All tests were run against a single hard drive. Signed-off-by: Brian Behlendorf <behlendorf1@llnl.gov> Issue #443
2011-10-24 23:55:20 +00:00
/* Lookup a specific xattr name in the directory */
Add Hooks for Linux Xattr Operations The Linux specific xattr operations have all been located in the file zpl_xattr.c. These functions primarily rely on the reworked zfs_* functions to do their job. They are also responsible for converting the possible Solaris style error codes to negative Linux errors.
2011-01-26 20:10:01 +00:00
error = -zfs_lookup(dxip, (char *)name, &xip, 0, cr, NULL, NULL);
Implement SA based xattrs The current ZFS implementation stores xattrs on disk using a hidden directory. In this directory a file name represents the xattr name and the file contexts are the xattr binary data. This approach is very flexible and allows for arbitrarily large xattrs. However, it also suffers from a significant performance penalty. Accessing a single xattr can requires up to three disk seeks. 1) Lookup the dnode object. 2) Lookup the dnodes's xattr directory object. 3) Lookup the xattr object in the directory. To avoid this performance penalty Linux filesystems such as ext3 and xfs try to store the xattr as part of the inode on disk. When the xattr is to large to store in the inode then a single external block is allocated for them. In practice most xattrs are small and this approach works well. The addition of System Attributes (SA) to zfs provides us a clean way to make this optimization. When the dataset property 'xattr=sa' is set then xattrs will be preferentially stored as System Attributes. This allows tiny xattrs (~100 bytes) to be stored with the dnode and up to 64k of xattrs to be stored in the spill block. If additional xattr space is required, which is unlikely under Linux, they will be stored using the traditional directory approach. This optimization results in roughly a 3x performance improvement when accessing xattrs which brings zfs roughly to parity with ext4 and xfs (see table below). When multiple xattrs are stored per-file the performance improvements are even greater because all of the xattrs stored in the spill block will be cached. However, by default SA based xattrs are disabled in the Linux port to maximize compatibility with other implementations. If you do enable SA based xattrs then they will not be visible on platforms which do not support this feature. ---------------------------------------------------------------------- Time in seconds to get/set one xattr of N bytes on 100,000 files ------+--------------------------------+------------------------------ | setxattr | getxattr bytes | ext4 xfs zfs-dir zfs-sa | ext4 xfs zfs-dir zfs-sa ------+--------------------------------+------------------------------ 1 | 2.33 31.88 21.50 4.57 | 2.35 2.64 6.29 2.43 32 | 2.79 30.68 21.98 4.60 | 2.44 2.59 6.78 2.48 256 | 3.25 31.99 21.36 5.92 | 2.32 2.71 6.22 3.14 1024 | 3.30 32.61 22.83 8.45 | 2.40 2.79 6.24 3.27 4096 | 3.57 317.46 22.52 10.73 | 2.78 28.62 6.90 3.94 16384 | n/a 2342.39 34.30 19.20 | n/a 45.44 145.90 7.55 65536 | n/a 2941.39 128.15 131.32* | n/a 141.92 256.85 262.12* Legend: * ext4 - Stock RHEL6.1 ext4 mounted with '-o user_xattr'. * xfs - Stock RHEL6.1 xfs mounted with default options. * zfs-dir - Directory based xattrs only. * zfs-sa - Prefer SAs but spill in to directories as needed, a trailing * indicates overflow in to directories occured. NOTE: Ext4 supports 4096 bytes of xattr name/value pairs per file. NOTE: XFS and ZFS have no limit on xattr name/value pairs per file. NOTE: Linux limits individual name/value pairs to 65536 bytes. NOTE: All setattr/getattr's were done after dropping the cache. NOTE: All tests were run against a single hard drive. Signed-off-by: Brian Behlendorf <behlendorf1@llnl.gov> Issue #443
2011-10-24 23:55:20 +00:00
if (error && (error != -ENOENT))
goto out;
Add Hooks for Linux Xattr Operations The Linux specific xattr operations have all been located in the file zpl_xattr.c. These functions primarily rely on the reworked zfs_* functions to do their job. They are also responsible for converting the possible Solaris style error codes to negative Linux errors.
2011-01-26 20:10:01 +00:00
error = 0;
/* Remove a specific name xattr when value is set to NULL. */
if (value == NULL) {
if (xip)
error = -zfs_remove(dxip, (char *)name, cr);
goto out;
}
/* Lookup failed create a new xattr. */
if (xip == NULL) {
vap = kmem_zalloc(sizeof(vattr_t), KM_SLEEP);
Truncate the xattr znode when updating existing attributes. If the attribute's new value was shorter than the old one the old code would leave parts of the old value in the xattr znode. Signed-off-by: Brian Behlendorf <behlendorf1@llnl.gov> Closes #203
2011-04-17 18:31:33 +00:00
vap->va_mode = xattr_mode;
Add Hooks for Linux Xattr Operations The Linux specific xattr operations have all been located in the file zpl_xattr.c. These functions primarily rely on the reworked zfs_* functions to do their job. They are also responsible for converting the possible Solaris style error codes to negative Linux errors.
2011-01-26 20:10:01 +00:00
vap->va_mask = ATTR_MODE;
Linux 2.6.29 compat, credentials As of Linux 2.6.29 a clean credential API was added to the Linux kernel. Previously the credential was embedded in the task_struct. Because the SPL already has considerable support for handling this API change the ZPL code has been updated to use the Solaris credential API.
2011-03-22 18:13:41 +00:00
vap->va_uid = crgetfsuid(cr);
vap->va_gid = crgetfsgid(cr);
Add Hooks for Linux Xattr Operations The Linux specific xattr operations have all been located in the file zpl_xattr.c. These functions primarily rely on the reworked zfs_* functions to do their job. They are also responsible for converting the possible Solaris style error codes to negative Linux errors.
2011-01-26 20:10:01 +00:00
error = -zfs_create(dxip, (char *)name, vap, 0, 0644, &xip,
cr, 0, NULL);
if (error)
goto out;
}
ASSERT(xip != NULL);
Truncate the xattr znode when updating existing attributes. If the attribute's new value was shorter than the old one the old code would leave parts of the old value in the xattr znode. Signed-off-by: Brian Behlendorf <behlendorf1@llnl.gov> Closes #203
2011-04-17 18:31:33 +00:00
error = -zfs_freesp(ITOZ(xip), 0, 0, xattr_mode, TRUE);
if (error)
goto out;
Add Hooks for Linux Xattr Operations The Linux specific xattr operations have all been located in the file zpl_xattr.c. These functions primarily rely on the reworked zfs_* functions to do their job. They are also responsible for converting the possible Solaris style error codes to negative Linux errors.
2011-01-26 20:10:01 +00:00
wrote = zpl_write_common(xip, value, size, 0, UIO_SYSSPACE, 0, cr);
if (wrote < 0)
error = wrote;
out:
if (vap)
kmem_free(vap, sizeof(vattr_t));
if (xip)
iput(xip);
if (dxip)
iput(dxip);
if (error == -ENOENT)
error = -ENODATA;
ASSERT3S(error, <=, 0);
return (error);
}
Implement SA based xattrs The current ZFS implementation stores xattrs on disk using a hidden directory. In this directory a file name represents the xattr name and the file contexts are the xattr binary data. This approach is very flexible and allows for arbitrarily large xattrs. However, it also suffers from a significant performance penalty. Accessing a single xattr can requires up to three disk seeks. 1) Lookup the dnode object. 2) Lookup the dnodes's xattr directory object. 3) Lookup the xattr object in the directory. To avoid this performance penalty Linux filesystems such as ext3 and xfs try to store the xattr as part of the inode on disk. When the xattr is to large to store in the inode then a single external block is allocated for them. In practice most xattrs are small and this approach works well. The addition of System Attributes (SA) to zfs provides us a clean way to make this optimization. When the dataset property 'xattr=sa' is set then xattrs will be preferentially stored as System Attributes. This allows tiny xattrs (~100 bytes) to be stored with the dnode and up to 64k of xattrs to be stored in the spill block. If additional xattr space is required, which is unlikely under Linux, they will be stored using the traditional directory approach. This optimization results in roughly a 3x performance improvement when accessing xattrs which brings zfs roughly to parity with ext4 and xfs (see table below). When multiple xattrs are stored per-file the performance improvements are even greater because all of the xattrs stored in the spill block will be cached. However, by default SA based xattrs are disabled in the Linux port to maximize compatibility with other implementations. If you do enable SA based xattrs then they will not be visible on platforms which do not support this feature. ---------------------------------------------------------------------- Time in seconds to get/set one xattr of N bytes on 100,000 files ------+--------------------------------+------------------------------ | setxattr | getxattr bytes | ext4 xfs zfs-dir zfs-sa | ext4 xfs zfs-dir zfs-sa ------+--------------------------------+------------------------------ 1 | 2.33 31.88 21.50 4.57 | 2.35 2.64 6.29 2.43 32 | 2.79 30.68 21.98 4.60 | 2.44 2.59 6.78 2.48 256 | 3.25 31.99 21.36 5.92 | 2.32 2.71 6.22 3.14 1024 | 3.30 32.61 22.83 8.45 | 2.40 2.79 6.24 3.27 4096 | 3.57 317.46 22.52 10.73 | 2.78 28.62 6.90 3.94 16384 | n/a 2342.39 34.30 19.20 | n/a 45.44 145.90 7.55 65536 | n/a 2941.39 128.15 131.32* | n/a 141.92 256.85 262.12* Legend: * ext4 - Stock RHEL6.1 ext4 mounted with '-o user_xattr'. * xfs - Stock RHEL6.1 xfs mounted with default options. * zfs-dir - Directory based xattrs only. * zfs-sa - Prefer SAs but spill in to directories as needed, a trailing * indicates overflow in to directories occured. NOTE: Ext4 supports 4096 bytes of xattr name/value pairs per file. NOTE: XFS and ZFS have no limit on xattr name/value pairs per file. NOTE: Linux limits individual name/value pairs to 65536 bytes. NOTE: All setattr/getattr's were done after dropping the cache. NOTE: All tests were run against a single hard drive. Signed-off-by: Brian Behlendorf <behlendorf1@llnl.gov> Issue #443
2011-10-24 23:55:20 +00:00
static int
zpl_xattr_set_sa(struct inode *ip, const char *name, const void *value,
size_t size, int flags, cred_t *cr)
{
znode_t *zp = ITOZ(ip);
nvlist_t *nvl;
size_t sa_size;
int error;
ASSERT(zp->z_xattr_cached);
nvl = zp->z_xattr_cached;
if (value == NULL) {
error = -nvlist_remove(nvl, name, DATA_TYPE_BYTE_ARRAY);
if (error == -ENOENT)
error = zpl_xattr_set_dir(ip, name, NULL, 0, flags, cr);
} else {
/* Limited to 32k to keep nvpair memory allocations small */
if (size > DXATTR_MAX_ENTRY_SIZE)
return (-EFBIG);
/* Prevent the DXATTR SA from consuming the entire SA region */
error = -nvlist_size(nvl, &sa_size, NV_ENCODE_XDR);
if (error)
return (error);
if (sa_size > DXATTR_MAX_SA_SIZE)
return (-EFBIG);
error = -nvlist_add_byte_array(nvl, name,
(uchar_t *)value, size);
if (error)
return (error);
}
/* Update the SA for additions, modifications, and removals. */
if (!error)
error = -zfs_sa_set_xattr(zp);
ASSERT3S(error, <=, 0);
return (error);
}
static int
zpl_xattr_set(struct inode *ip, const char *name, const void *value,
size_t size, int flags)
{
znode_t *zp = ITOZ(ip);
zfs_sb_t *zsb = ZTOZSB(zp);
cred_t *cr = CRED();
int error;
crhold(cr);
rw_enter(&ITOZ(ip)->z_xattr_lock, RW_WRITER);
/*
* Before setting the xattr check to see if it already exists.
* This is done to ensure the following optional flags are honored.
*
* XATTR_CREATE: fail if xattr already exists
* XATTR_REPLACE: fail if xattr does not exist
*/
error = __zpl_xattr_get(ip, name, NULL, 0, cr);
if (error < 0) {
if (error != -ENODATA)
goto out;
Prevent xattr remove from creating xattr directory Attempting to remove an xattr from a file which does not contain any directory based xattrs would result in the xattr directory being created. This behavior is non-optimal because it results in write operations to the pool in addition to the expected error being returned. To prevent this the CREATE_XATTR_DIR flag is only passed in zpl_xattr_set_dir() when setting a non-NULL xattr value. In addition, zpl_xattr_set() is updated similarly such that it will return immediately if passed an xattr name which doesn't exist and a NULL value. Signed-off-by: Massimo Maggi <me@massimo-maggi.eu> Signed-off-by: Richard Yao <ryao@gentoo.org> Signed-off-by: Brian Behlendorf <behlendorf1@llnl.gov> Issue #170
2013-10-28 16:07:00 +00:00
if (flags & XATTR_REPLACE)
goto out;
/* The xattr to be removed already doesn't exist */
error = 0;
if (value == NULL)
Implement SA based xattrs The current ZFS implementation stores xattrs on disk using a hidden directory. In this directory a file name represents the xattr name and the file contexts are the xattr binary data. This approach is very flexible and allows for arbitrarily large xattrs. However, it also suffers from a significant performance penalty. Accessing a single xattr can requires up to three disk seeks. 1) Lookup the dnode object. 2) Lookup the dnodes's xattr directory object. 3) Lookup the xattr object in the directory. To avoid this performance penalty Linux filesystems such as ext3 and xfs try to store the xattr as part of the inode on disk. When the xattr is to large to store in the inode then a single external block is allocated for them. In practice most xattrs are small and this approach works well. The addition of System Attributes (SA) to zfs provides us a clean way to make this optimization. When the dataset property 'xattr=sa' is set then xattrs will be preferentially stored as System Attributes. This allows tiny xattrs (~100 bytes) to be stored with the dnode and up to 64k of xattrs to be stored in the spill block. If additional xattr space is required, which is unlikely under Linux, they will be stored using the traditional directory approach. This optimization results in roughly a 3x performance improvement when accessing xattrs which brings zfs roughly to parity with ext4 and xfs (see table below). When multiple xattrs are stored per-file the performance improvements are even greater because all of the xattrs stored in the spill block will be cached. However, by default SA based xattrs are disabled in the Linux port to maximize compatibility with other implementations. If you do enable SA based xattrs then they will not be visible on platforms which do not support this feature. ---------------------------------------------------------------------- Time in seconds to get/set one xattr of N bytes on 100,000 files ------+--------------------------------+------------------------------ | setxattr | getxattr bytes | ext4 xfs zfs-dir zfs-sa | ext4 xfs zfs-dir zfs-sa ------+--------------------------------+------------------------------ 1 | 2.33 31.88 21.50 4.57 | 2.35 2.64 6.29 2.43 32 | 2.79 30.68 21.98 4.60 | 2.44 2.59 6.78 2.48 256 | 3.25 31.99 21.36 5.92 | 2.32 2.71 6.22 3.14 1024 | 3.30 32.61 22.83 8.45 | 2.40 2.79 6.24 3.27 4096 | 3.57 317.46 22.52 10.73 | 2.78 28.62 6.90 3.94 16384 | n/a 2342.39 34.30 19.20 | n/a 45.44 145.90 7.55 65536 | n/a 2941.39 128.15 131.32* | n/a 141.92 256.85 262.12* Legend: * ext4 - Stock RHEL6.1 ext4 mounted with '-o user_xattr'. * xfs - Stock RHEL6.1 xfs mounted with default options. * zfs-dir - Directory based xattrs only. * zfs-sa - Prefer SAs but spill in to directories as needed, a trailing * indicates overflow in to directories occured. NOTE: Ext4 supports 4096 bytes of xattr name/value pairs per file. NOTE: XFS and ZFS have no limit on xattr name/value pairs per file. NOTE: Linux limits individual name/value pairs to 65536 bytes. NOTE: All setattr/getattr's were done after dropping the cache. NOTE: All tests were run against a single hard drive. Signed-off-by: Brian Behlendorf <behlendorf1@llnl.gov> Issue #443
2011-10-24 23:55:20 +00:00
goto out;
} else {
error = -EEXIST;
if (flags & XATTR_CREATE)
goto out;
}
/* Preferentially store the xattr as a SA for better performance */
if (zsb->z_use_sa && zsb->z_xattr_sa && zp->z_is_sa) {
error = zpl_xattr_set_sa(ip, name, value, size, flags, cr);
if (error == 0)
goto out;
}
error = zpl_xattr_set_dir(ip, name, value, size, flags, cr);
out:
rw_exit(&ITOZ(ip)->z_xattr_lock);
crfree(cr);
ASSERT3S(error, <=, 0);
return (error);
}
Add Hooks for Linux Xattr Operations The Linux specific xattr operations have all been located in the file zpl_xattr.c. These functions primarily rely on the reworked zfs_* functions to do their job. They are also responsible for converting the possible Solaris style error codes to negative Linux errors.
2011-01-26 20:10:01 +00:00
static int
Linux 2.6.33 compat, get/set xattr callbacks The xattr handler prototypes were sanitized with the idea being that the same handlers could be used for multiple methods. The result of this was the inode type was changes to a dentry, and both the get() and set() hooks had a handler_flags argument added. The list() callback was similiarly effected but no autoconf check was added because we do not use the list() callback.
2011-02-11 18:33:01 +00:00
__zpl_xattr_user_get(struct inode *ip, const char *name,
Implement SA based xattrs The current ZFS implementation stores xattrs on disk using a hidden directory. In this directory a file name represents the xattr name and the file contexts are the xattr binary data. This approach is very flexible and allows for arbitrarily large xattrs. However, it also suffers from a significant performance penalty. Accessing a single xattr can requires up to three disk seeks. 1) Lookup the dnode object. 2) Lookup the dnodes's xattr directory object. 3) Lookup the xattr object in the directory. To avoid this performance penalty Linux filesystems such as ext3 and xfs try to store the xattr as part of the inode on disk. When the xattr is to large to store in the inode then a single external block is allocated for them. In practice most xattrs are small and this approach works well. The addition of System Attributes (SA) to zfs provides us a clean way to make this optimization. When the dataset property 'xattr=sa' is set then xattrs will be preferentially stored as System Attributes. This allows tiny xattrs (~100 bytes) to be stored with the dnode and up to 64k of xattrs to be stored in the spill block. If additional xattr space is required, which is unlikely under Linux, they will be stored using the traditional directory approach. This optimization results in roughly a 3x performance improvement when accessing xattrs which brings zfs roughly to parity with ext4 and xfs (see table below). When multiple xattrs are stored per-file the performance improvements are even greater because all of the xattrs stored in the spill block will be cached. However, by default SA based xattrs are disabled in the Linux port to maximize compatibility with other implementations. If you do enable SA based xattrs then they will not be visible on platforms which do not support this feature. ---------------------------------------------------------------------- Time in seconds to get/set one xattr of N bytes on 100,000 files ------+--------------------------------+------------------------------ | setxattr | getxattr bytes | ext4 xfs zfs-dir zfs-sa | ext4 xfs zfs-dir zfs-sa ------+--------------------------------+------------------------------ 1 | 2.33 31.88 21.50 4.57 | 2.35 2.64 6.29 2.43 32 | 2.79 30.68 21.98 4.60 | 2.44 2.59 6.78 2.48 256 | 3.25 31.99 21.36 5.92 | 2.32 2.71 6.22 3.14 1024 | 3.30 32.61 22.83 8.45 | 2.40 2.79 6.24 3.27 4096 | 3.57 317.46 22.52 10.73 | 2.78 28.62 6.90 3.94 16384 | n/a 2342.39 34.30 19.20 | n/a 45.44 145.90 7.55 65536 | n/a 2941.39 128.15 131.32* | n/a 141.92 256.85 262.12* Legend: * ext4 - Stock RHEL6.1 ext4 mounted with '-o user_xattr'. * xfs - Stock RHEL6.1 xfs mounted with default options. * zfs-dir - Directory based xattrs only. * zfs-sa - Prefer SAs but spill in to directories as needed, a trailing * indicates overflow in to directories occured. NOTE: Ext4 supports 4096 bytes of xattr name/value pairs per file. NOTE: XFS and ZFS have no limit on xattr name/value pairs per file. NOTE: Linux limits individual name/value pairs to 65536 bytes. NOTE: All setattr/getattr's were done after dropping the cache. NOTE: All tests were run against a single hard drive. Signed-off-by: Brian Behlendorf <behlendorf1@llnl.gov> Issue #443
2011-10-24 23:55:20 +00:00
void *value, size_t size)
Add Hooks for Linux Xattr Operations The Linux specific xattr operations have all been located in the file zpl_xattr.c. These functions primarily rely on the reworked zfs_* functions to do their job. They are also responsible for converting the possible Solaris style error codes to negative Linux errors.
2011-01-26 20:10:01 +00:00
{
char *xattr_name;
int error;
if (strcmp(name, "") == 0)
return -EINVAL;
Linux compat 2.6.39: mount_nodev() The .get_sb callback has been replaced by a .mount callback in the file_system_type structure. When using the new interface the caller must now use the mount_nodev() helper. Unfortunately, the new interface no longer passes the vfsmount down to the zfs layers. This poses a problem for the existing implementation because we currently save this pointer in the super block for latter use. It provides our only entry point in to the namespace layer for manipulating certain mount options. This needed to be done originally to allow commands like 'zfs set atime=off tank' to work properly. It also allowed me to keep more of the original Solaris code unmodified. Under Solaris there is a 1-to-1 mapping between a mount point and a file system so this is a fairly natural thing to do. However, under Linux they many be multiple entries in the namespace which reference the same filesystem. Thus keeping a back reference from the filesystem to the namespace is complicated. Rather than introduce some ugly hack to get the vfsmount and continue as before. I'm leveraging this API change to update the ZFS code to do things in a more natural way for Linux. This has the upside that is resolves the compatibility issue for the long term and fixes several other minor bugs which have been reported. This commit updates the code to remove this vfsmount back reference entirely. All modifications to filesystem mount options are now passed in to the kernel via a '-o remount'. This is the expected Linux mechanism and allows the namespace to properly handle any options which apply to it before passing them on to the file system itself. Aside from fixing the compatibility issue, removing the vfsmount has had the benefit of simplifying the code. This change which fairly involved has turned out nicely. Closes #246 Closes #217 Closes #187 Closes #248 Closes #231
2011-05-19 18:44:07 +00:00
if (!(ITOZSB(ip)->z_flags & ZSB_XATTR))
Add Hooks for Linux Xattr Operations The Linux specific xattr operations have all been located in the file zpl_xattr.c. These functions primarily rely on the reworked zfs_* functions to do their job. They are also responsible for converting the possible Solaris style error codes to negative Linux errors.
2011-01-26 20:10:01 +00:00
return -EOPNOTSUPP;
xattr_name = kmem_asprintf("%s%s", XATTR_USER_PREFIX, name);
Implement SA based xattrs The current ZFS implementation stores xattrs on disk using a hidden directory. In this directory a file name represents the xattr name and the file contexts are the xattr binary data. This approach is very flexible and allows for arbitrarily large xattrs. However, it also suffers from a significant performance penalty. Accessing a single xattr can requires up to three disk seeks. 1) Lookup the dnode object. 2) Lookup the dnodes's xattr directory object. 3) Lookup the xattr object in the directory. To avoid this performance penalty Linux filesystems such as ext3 and xfs try to store the xattr as part of the inode on disk. When the xattr is to large to store in the inode then a single external block is allocated for them. In practice most xattrs are small and this approach works well. The addition of System Attributes (SA) to zfs provides us a clean way to make this optimization. When the dataset property 'xattr=sa' is set then xattrs will be preferentially stored as System Attributes. This allows tiny xattrs (~100 bytes) to be stored with the dnode and up to 64k of xattrs to be stored in the spill block. If additional xattr space is required, which is unlikely under Linux, they will be stored using the traditional directory approach. This optimization results in roughly a 3x performance improvement when accessing xattrs which brings zfs roughly to parity with ext4 and xfs (see table below). When multiple xattrs are stored per-file the performance improvements are even greater because all of the xattrs stored in the spill block will be cached. However, by default SA based xattrs are disabled in the Linux port to maximize compatibility with other implementations. If you do enable SA based xattrs then they will not be visible on platforms which do not support this feature. ---------------------------------------------------------------------- Time in seconds to get/set one xattr of N bytes on 100,000 files ------+--------------------------------+------------------------------ | setxattr | getxattr bytes | ext4 xfs zfs-dir zfs-sa | ext4 xfs zfs-dir zfs-sa ------+--------------------------------+------------------------------ 1 | 2.33 31.88 21.50 4.57 | 2.35 2.64 6.29 2.43 32 | 2.79 30.68 21.98 4.60 | 2.44 2.59 6.78 2.48 256 | 3.25 31.99 21.36 5.92 | 2.32 2.71 6.22 3.14 1024 | 3.30 32.61 22.83 8.45 | 2.40 2.79 6.24 3.27 4096 | 3.57 317.46 22.52 10.73 | 2.78 28.62 6.90 3.94 16384 | n/a 2342.39 34.30 19.20 | n/a 45.44 145.90 7.55 65536 | n/a 2941.39 128.15 131.32* | n/a 141.92 256.85 262.12* Legend: * ext4 - Stock RHEL6.1 ext4 mounted with '-o user_xattr'. * xfs - Stock RHEL6.1 xfs mounted with default options. * zfs-dir - Directory based xattrs only. * zfs-sa - Prefer SAs but spill in to directories as needed, a trailing * indicates overflow in to directories occured. NOTE: Ext4 supports 4096 bytes of xattr name/value pairs per file. NOTE: XFS and ZFS have no limit on xattr name/value pairs per file. NOTE: Linux limits individual name/value pairs to 65536 bytes. NOTE: All setattr/getattr's were done after dropping the cache. NOTE: All tests were run against a single hard drive. Signed-off-by: Brian Behlendorf <behlendorf1@llnl.gov> Issue #443
2011-10-24 23:55:20 +00:00
error = zpl_xattr_get(ip, xattr_name, value, size);
Add Hooks for Linux Xattr Operations The Linux specific xattr operations have all been located in the file zpl_xattr.c. These functions primarily rely on the reworked zfs_* functions to do their job. They are also responsible for converting the possible Solaris style error codes to negative Linux errors.
2011-01-26 20:10:01 +00:00
strfree(xattr_name);
return (error);
}
Linux 2.6.33 compat, get/set xattr callbacks The xattr handler prototypes were sanitized with the idea being that the same handlers could be used for multiple methods. The result of this was the inode type was changes to a dentry, and both the get() and set() hooks had a handler_flags argument added. The list() callback was similiarly effected but no autoconf check was added because we do not use the list() callback.
2011-02-11 18:33:01 +00:00
ZPL_XATTR_GET_WRAPPER(zpl_xattr_user_get);
Add Hooks for Linux Xattr Operations The Linux specific xattr operations have all been located in the file zpl_xattr.c. These functions primarily rely on the reworked zfs_* functions to do their job. They are also responsible for converting the possible Solaris style error codes to negative Linux errors.
2011-01-26 20:10:01 +00:00
static int
Linux 2.6.33 compat, get/set xattr callbacks The xattr handler prototypes were sanitized with the idea being that the same handlers could be used for multiple methods. The result of this was the inode type was changes to a dentry, and both the get() and set() hooks had a handler_flags argument added. The list() callback was similiarly effected but no autoconf check was added because we do not use the list() callback.
2011-02-11 18:33:01 +00:00
__zpl_xattr_user_set(struct inode *ip, const char *name,
Add Hooks for Linux Xattr Operations The Linux specific xattr operations have all been located in the file zpl_xattr.c. These functions primarily rely on the reworked zfs_* functions to do their job. They are also responsible for converting the possible Solaris style error codes to negative Linux errors.
2011-01-26 20:10:01 +00:00
const void *value, size_t size, int flags)
{
char *xattr_name;
int error;
if (strcmp(name, "") == 0)
return -EINVAL;
Linux compat 2.6.39: mount_nodev() The .get_sb callback has been replaced by a .mount callback in the file_system_type structure. When using the new interface the caller must now use the mount_nodev() helper. Unfortunately, the new interface no longer passes the vfsmount down to the zfs layers. This poses a problem for the existing implementation because we currently save this pointer in the super block for latter use. It provides our only entry point in to the namespace layer for manipulating certain mount options. This needed to be done originally to allow commands like 'zfs set atime=off tank' to work properly. It also allowed me to keep more of the original Solaris code unmodified. Under Solaris there is a 1-to-1 mapping between a mount point and a file system so this is a fairly natural thing to do. However, under Linux they many be multiple entries in the namespace which reference the same filesystem. Thus keeping a back reference from the filesystem to the namespace is complicated. Rather than introduce some ugly hack to get the vfsmount and continue as before. I'm leveraging this API change to update the ZFS code to do things in a more natural way for Linux. This has the upside that is resolves the compatibility issue for the long term and fixes several other minor bugs which have been reported. This commit updates the code to remove this vfsmount back reference entirely. All modifications to filesystem mount options are now passed in to the kernel via a '-o remount'. This is the expected Linux mechanism and allows the namespace to properly handle any options which apply to it before passing them on to the file system itself. Aside from fixing the compatibility issue, removing the vfsmount has had the benefit of simplifying the code. This change which fairly involved has turned out nicely. Closes #246 Closes #217 Closes #187 Closes #248 Closes #231
2011-05-19 18:44:07 +00:00
if (!(ITOZSB(ip)->z_flags & ZSB_XATTR))
Add Hooks for Linux Xattr Operations The Linux specific xattr operations have all been located in the file zpl_xattr.c. These functions primarily rely on the reworked zfs_* functions to do their job. They are also responsible for converting the possible Solaris style error codes to negative Linux errors.
2011-01-26 20:10:01 +00:00
return -EOPNOTSUPP;
xattr_name = kmem_asprintf("%s%s", XATTR_USER_PREFIX, name);
error = zpl_xattr_set(ip, xattr_name, value, size, flags);
strfree(xattr_name);
return (error);
}
Linux 2.6.33 compat, get/set xattr callbacks The xattr handler prototypes were sanitized with the idea being that the same handlers could be used for multiple methods. The result of this was the inode type was changes to a dentry, and both the get() and set() hooks had a handler_flags argument added. The list() callback was similiarly effected but no autoconf check was added because we do not use the list() callback.
2011-02-11 18:33:01 +00:00
ZPL_XATTR_SET_WRAPPER(zpl_xattr_user_set);
Add Hooks for Linux Xattr Operations The Linux specific xattr operations have all been located in the file zpl_xattr.c. These functions primarily rely on the reworked zfs_* functions to do their job. They are also responsible for converting the possible Solaris style error codes to negative Linux errors.
2011-01-26 20:10:01 +00:00
Linux 2.6.35 compat, const struct xattr_handler The const keyword was added to the 'struct xattr_handler' in the generic Linux super_block structure. To handle this we define an appropriate xattr_handler_t typedef which can be used. This was the preferred solution because it keeps the code clean and readable.
2011-02-11 00:16:52 +00:00
xattr_handler_t zpl_xattr_user_handler = {
Add Hooks for Linux Xattr Operations The Linux specific xattr operations have all been located in the file zpl_xattr.c. These functions primarily rely on the reworked zfs_* functions to do their job. They are also responsible for converting the possible Solaris style error codes to negative Linux errors.
2011-01-26 20:10:01 +00:00
.prefix = XATTR_USER_PREFIX,
.get = zpl_xattr_user_get,
.set = zpl_xattr_user_set,
};
static int
Linux 2.6.33 compat, get/set xattr callbacks The xattr handler prototypes were sanitized with the idea being that the same handlers could be used for multiple methods. The result of this was the inode type was changes to a dentry, and both the get() and set() hooks had a handler_flags argument added. The list() callback was similiarly effected but no autoconf check was added because we do not use the list() callback.
2011-02-11 18:33:01 +00:00
__zpl_xattr_trusted_get(struct inode *ip, const char *name,
Implement SA based xattrs The current ZFS implementation stores xattrs on disk using a hidden directory. In this directory a file name represents the xattr name and the file contexts are the xattr binary data. This approach is very flexible and allows for arbitrarily large xattrs. However, it also suffers from a significant performance penalty. Accessing a single xattr can requires up to three disk seeks. 1) Lookup the dnode object. 2) Lookup the dnodes's xattr directory object. 3) Lookup the xattr object in the directory. To avoid this performance penalty Linux filesystems such as ext3 and xfs try to store the xattr as part of the inode on disk. When the xattr is to large to store in the inode then a single external block is allocated for them. In practice most xattrs are small and this approach works well. The addition of System Attributes (SA) to zfs provides us a clean way to make this optimization. When the dataset property 'xattr=sa' is set then xattrs will be preferentially stored as System Attributes. This allows tiny xattrs (~100 bytes) to be stored with the dnode and up to 64k of xattrs to be stored in the spill block. If additional xattr space is required, which is unlikely under Linux, they will be stored using the traditional directory approach. This optimization results in roughly a 3x performance improvement when accessing xattrs which brings zfs roughly to parity with ext4 and xfs (see table below). When multiple xattrs are stored per-file the performance improvements are even greater because all of the xattrs stored in the spill block will be cached. However, by default SA based xattrs are disabled in the Linux port to maximize compatibility with other implementations. If you do enable SA based xattrs then they will not be visible on platforms which do not support this feature. ---------------------------------------------------------------------- Time in seconds to get/set one xattr of N bytes on 100,000 files ------+--------------------------------+------------------------------ | setxattr | getxattr bytes | ext4 xfs zfs-dir zfs-sa | ext4 xfs zfs-dir zfs-sa ------+--------------------------------+------------------------------ 1 | 2.33 31.88 21.50 4.57 | 2.35 2.64 6.29 2.43 32 | 2.79 30.68 21.98 4.60 | 2.44 2.59 6.78 2.48 256 | 3.25 31.99 21.36 5.92 | 2.32 2.71 6.22 3.14 1024 | 3.30 32.61 22.83 8.45 | 2.40 2.79 6.24 3.27 4096 | 3.57 317.46 22.52 10.73 | 2.78 28.62 6.90 3.94 16384 | n/a 2342.39 34.30 19.20 | n/a 45.44 145.90 7.55 65536 | n/a 2941.39 128.15 131.32* | n/a 141.92 256.85 262.12* Legend: * ext4 - Stock RHEL6.1 ext4 mounted with '-o user_xattr'. * xfs - Stock RHEL6.1 xfs mounted with default options. * zfs-dir - Directory based xattrs only. * zfs-sa - Prefer SAs but spill in to directories as needed, a trailing * indicates overflow in to directories occured. NOTE: Ext4 supports 4096 bytes of xattr name/value pairs per file. NOTE: XFS and ZFS have no limit on xattr name/value pairs per file. NOTE: Linux limits individual name/value pairs to 65536 bytes. NOTE: All setattr/getattr's were done after dropping the cache. NOTE: All tests were run against a single hard drive. Signed-off-by: Brian Behlendorf <behlendorf1@llnl.gov> Issue #443
2011-10-24 23:55:20 +00:00
void *value, size_t size)
Add Hooks for Linux Xattr Operations The Linux specific xattr operations have all been located in the file zpl_xattr.c. These functions primarily rely on the reworked zfs_* functions to do their job. They are also responsible for converting the possible Solaris style error codes to negative Linux errors.
2011-01-26 20:10:01 +00:00
{
char *xattr_name;
int error;
if (!capable(CAP_SYS_ADMIN))
return -EACCES;
if (strcmp(name, "") == 0)
return -EINVAL;
xattr_name = kmem_asprintf("%s%s", XATTR_TRUSTED_PREFIX, name);
Implement SA based xattrs The current ZFS implementation stores xattrs on disk using a hidden directory. In this directory a file name represents the xattr name and the file contexts are the xattr binary data. This approach is very flexible and allows for arbitrarily large xattrs. However, it also suffers from a significant performance penalty. Accessing a single xattr can requires up to three disk seeks. 1) Lookup the dnode object. 2) Lookup the dnodes's xattr directory object. 3) Lookup the xattr object in the directory. To avoid this performance penalty Linux filesystems such as ext3 and xfs try to store the xattr as part of the inode on disk. When the xattr is to large to store in the inode then a single external block is allocated for them. In practice most xattrs are small and this approach works well. The addition of System Attributes (SA) to zfs provides us a clean way to make this optimization. When the dataset property 'xattr=sa' is set then xattrs will be preferentially stored as System Attributes. This allows tiny xattrs (~100 bytes) to be stored with the dnode and up to 64k of xattrs to be stored in the spill block. If additional xattr space is required, which is unlikely under Linux, they will be stored using the traditional directory approach. This optimization results in roughly a 3x performance improvement when accessing xattrs which brings zfs roughly to parity with ext4 and xfs (see table below). When multiple xattrs are stored per-file the performance improvements are even greater because all of the xattrs stored in the spill block will be cached. However, by default SA based xattrs are disabled in the Linux port to maximize compatibility with other implementations. If you do enable SA based xattrs then they will not be visible on platforms which do not support this feature. ---------------------------------------------------------------------- Time in seconds to get/set one xattr of N bytes on 100,000 files ------+--------------------------------+------------------------------ | setxattr | getxattr bytes | ext4 xfs zfs-dir zfs-sa | ext4 xfs zfs-dir zfs-sa ------+--------------------------------+------------------------------ 1 | 2.33 31.88 21.50 4.57 | 2.35 2.64 6.29 2.43 32 | 2.79 30.68 21.98 4.60 | 2.44 2.59 6.78 2.48 256 | 3.25 31.99 21.36 5.92 | 2.32 2.71 6.22 3.14 1024 | 3.30 32.61 22.83 8.45 | 2.40 2.79 6.24 3.27 4096 | 3.57 317.46 22.52 10.73 | 2.78 28.62 6.90 3.94 16384 | n/a 2342.39 34.30 19.20 | n/a 45.44 145.90 7.55 65536 | n/a 2941.39 128.15 131.32* | n/a 141.92 256.85 262.12* Legend: * ext4 - Stock RHEL6.1 ext4 mounted with '-o user_xattr'. * xfs - Stock RHEL6.1 xfs mounted with default options. * zfs-dir - Directory based xattrs only. * zfs-sa - Prefer SAs but spill in to directories as needed, a trailing * indicates overflow in to directories occured. NOTE: Ext4 supports 4096 bytes of xattr name/value pairs per file. NOTE: XFS and ZFS have no limit on xattr name/value pairs per file. NOTE: Linux limits individual name/value pairs to 65536 bytes. NOTE: All setattr/getattr's were done after dropping the cache. NOTE: All tests were run against a single hard drive. Signed-off-by: Brian Behlendorf <behlendorf1@llnl.gov> Issue #443
2011-10-24 23:55:20 +00:00
error = zpl_xattr_get(ip, xattr_name, value, size);
Add Hooks for Linux Xattr Operations The Linux specific xattr operations have all been located in the file zpl_xattr.c. These functions primarily rely on the reworked zfs_* functions to do their job. They are also responsible for converting the possible Solaris style error codes to negative Linux errors.
2011-01-26 20:10:01 +00:00
strfree(xattr_name);
return (error);
}
Linux 2.6.33 compat, get/set xattr callbacks The xattr handler prototypes were sanitized with the idea being that the same handlers could be used for multiple methods. The result of this was the inode type was changes to a dentry, and both the get() and set() hooks had a handler_flags argument added. The list() callback was similiarly effected but no autoconf check was added because we do not use the list() callback.
2011-02-11 18:33:01 +00:00
ZPL_XATTR_GET_WRAPPER(zpl_xattr_trusted_get);
Add Hooks for Linux Xattr Operations The Linux specific xattr operations have all been located in the file zpl_xattr.c. These functions primarily rely on the reworked zfs_* functions to do their job. They are also responsible for converting the possible Solaris style error codes to negative Linux errors.
2011-01-26 20:10:01 +00:00
static int
Linux 2.6.33 compat, get/set xattr callbacks The xattr handler prototypes were sanitized with the idea being that the same handlers could be used for multiple methods. The result of this was the inode type was changes to a dentry, and both the get() and set() hooks had a handler_flags argument added. The list() callback was similiarly effected but no autoconf check was added because we do not use the list() callback.
2011-02-11 18:33:01 +00:00
__zpl_xattr_trusted_set(struct inode *ip, const char *name,
Add Hooks for Linux Xattr Operations The Linux specific xattr operations have all been located in the file zpl_xattr.c. These functions primarily rely on the reworked zfs_* functions to do their job. They are also responsible for converting the possible Solaris style error codes to negative Linux errors.
2011-01-26 20:10:01 +00:00
const void *value, size_t size, int flags)
{
char *xattr_name;
int error;
if (!capable(CAP_SYS_ADMIN))
return -EACCES;
if (strcmp(name, "") == 0)
return -EINVAL;
xattr_name = kmem_asprintf("%s%s", XATTR_TRUSTED_PREFIX, name);
error = zpl_xattr_set(ip, xattr_name, value, size, flags);
strfree(xattr_name);
return (error);
}
Linux 2.6.33 compat, get/set xattr callbacks The xattr handler prototypes were sanitized with the idea being that the same handlers could be used for multiple methods. The result of this was the inode type was changes to a dentry, and both the get() and set() hooks had a handler_flags argument added. The list() callback was similiarly effected but no autoconf check was added because we do not use the list() callback.
2011-02-11 18:33:01 +00:00
ZPL_XATTR_SET_WRAPPER(zpl_xattr_trusted_set);
Add Hooks for Linux Xattr Operations The Linux specific xattr operations have all been located in the file zpl_xattr.c. These functions primarily rely on the reworked zfs_* functions to do their job. They are also responsible for converting the possible Solaris style error codes to negative Linux errors.
2011-01-26 20:10:01 +00:00
Linux 2.6.35 compat, const struct xattr_handler The const keyword was added to the 'struct xattr_handler' in the generic Linux super_block structure. To handle this we define an appropriate xattr_handler_t typedef which can be used. This was the preferred solution because it keeps the code clean and readable.
2011-02-11 00:16:52 +00:00
xattr_handler_t zpl_xattr_trusted_handler = {
Add Hooks for Linux Xattr Operations The Linux specific xattr operations have all been located in the file zpl_xattr.c. These functions primarily rely on the reworked zfs_* functions to do their job. They are also responsible for converting the possible Solaris style error codes to negative Linux errors.
2011-01-26 20:10:01 +00:00
.prefix = XATTR_TRUSTED_PREFIX,
.get = zpl_xattr_trusted_get,
.set = zpl_xattr_trusted_set,
};
static int
Linux 2.6.33 compat, get/set xattr callbacks The xattr handler prototypes were sanitized with the idea being that the same handlers could be used for multiple methods. The result of this was the inode type was changes to a dentry, and both the get() and set() hooks had a handler_flags argument added. The list() callback was similiarly effected but no autoconf check was added because we do not use the list() callback.
2011-02-11 18:33:01 +00:00
__zpl_xattr_security_get(struct inode *ip, const char *name,
Implement SA based xattrs The current ZFS implementation stores xattrs on disk using a hidden directory. In this directory a file name represents the xattr name and the file contexts are the xattr binary data. This approach is very flexible and allows for arbitrarily large xattrs. However, it also suffers from a significant performance penalty. Accessing a single xattr can requires up to three disk seeks. 1) Lookup the dnode object. 2) Lookup the dnodes's xattr directory object. 3) Lookup the xattr object in the directory. To avoid this performance penalty Linux filesystems such as ext3 and xfs try to store the xattr as part of the inode on disk. When the xattr is to large to store in the inode then a single external block is allocated for them. In practice most xattrs are small and this approach works well. The addition of System Attributes (SA) to zfs provides us a clean way to make this optimization. When the dataset property 'xattr=sa' is set then xattrs will be preferentially stored as System Attributes. This allows tiny xattrs (~100 bytes) to be stored with the dnode and up to 64k of xattrs to be stored in the spill block. If additional xattr space is required, which is unlikely under Linux, they will be stored using the traditional directory approach. This optimization results in roughly a 3x performance improvement when accessing xattrs which brings zfs roughly to parity with ext4 and xfs (see table below). When multiple xattrs are stored per-file the performance improvements are even greater because all of the xattrs stored in the spill block will be cached. However, by default SA based xattrs are disabled in the Linux port to maximize compatibility with other implementations. If you do enable SA based xattrs then they will not be visible on platforms which do not support this feature. ---------------------------------------------------------------------- Time in seconds to get/set one xattr of N bytes on 100,000 files ------+--------------------------------+------------------------------ | setxattr | getxattr bytes | ext4 xfs zfs-dir zfs-sa | ext4 xfs zfs-dir zfs-sa ------+--------------------------------+------------------------------ 1 | 2.33 31.88 21.50 4.57 | 2.35 2.64 6.29 2.43 32 | 2.79 30.68 21.98 4.60 | 2.44 2.59 6.78 2.48 256 | 3.25 31.99 21.36 5.92 | 2.32 2.71 6.22 3.14 1024 | 3.30 32.61 22.83 8.45 | 2.40 2.79 6.24 3.27 4096 | 3.57 317.46 22.52 10.73 | 2.78 28.62 6.90 3.94 16384 | n/a 2342.39 34.30 19.20 | n/a 45.44 145.90 7.55 65536 | n/a 2941.39 128.15 131.32* | n/a 141.92 256.85 262.12* Legend: * ext4 - Stock RHEL6.1 ext4 mounted with '-o user_xattr'. * xfs - Stock RHEL6.1 xfs mounted with default options. * zfs-dir - Directory based xattrs only. * zfs-sa - Prefer SAs but spill in to directories as needed, a trailing * indicates overflow in to directories occured. NOTE: Ext4 supports 4096 bytes of xattr name/value pairs per file. NOTE: XFS and ZFS have no limit on xattr name/value pairs per file. NOTE: Linux limits individual name/value pairs to 65536 bytes. NOTE: All setattr/getattr's were done after dropping the cache. NOTE: All tests were run against a single hard drive. Signed-off-by: Brian Behlendorf <behlendorf1@llnl.gov> Issue #443
2011-10-24 23:55:20 +00:00
void *value, size_t size)
Add Hooks for Linux Xattr Operations The Linux specific xattr operations have all been located in the file zpl_xattr.c. These functions primarily rely on the reworked zfs_* functions to do their job. They are also responsible for converting the possible Solaris style error codes to negative Linux errors.
2011-01-26 20:10:01 +00:00
{
char *xattr_name;
int error;
if (strcmp(name, "") == 0)
return -EINVAL;
xattr_name = kmem_asprintf("%s%s", XATTR_SECURITY_PREFIX, name);
Implement SA based xattrs The current ZFS implementation stores xattrs on disk using a hidden directory. In this directory a file name represents the xattr name and the file contexts are the xattr binary data. This approach is very flexible and allows for arbitrarily large xattrs. However, it also suffers from a significant performance penalty. Accessing a single xattr can requires up to three disk seeks. 1) Lookup the dnode object. 2) Lookup the dnodes's xattr directory object. 3) Lookup the xattr object in the directory. To avoid this performance penalty Linux filesystems such as ext3 and xfs try to store the xattr as part of the inode on disk. When the xattr is to large to store in the inode then a single external block is allocated for them. In practice most xattrs are small and this approach works well. The addition of System Attributes (SA) to zfs provides us a clean way to make this optimization. When the dataset property 'xattr=sa' is set then xattrs will be preferentially stored as System Attributes. This allows tiny xattrs (~100 bytes) to be stored with the dnode and up to 64k of xattrs to be stored in the spill block. If additional xattr space is required, which is unlikely under Linux, they will be stored using the traditional directory approach. This optimization results in roughly a 3x performance improvement when accessing xattrs which brings zfs roughly to parity with ext4 and xfs (see table below). When multiple xattrs are stored per-file the performance improvements are even greater because all of the xattrs stored in the spill block will be cached. However, by default SA based xattrs are disabled in the Linux port to maximize compatibility with other implementations. If you do enable SA based xattrs then they will not be visible on platforms which do not support this feature. ---------------------------------------------------------------------- Time in seconds to get/set one xattr of N bytes on 100,000 files ------+--------------------------------+------------------------------ | setxattr | getxattr bytes | ext4 xfs zfs-dir zfs-sa | ext4 xfs zfs-dir zfs-sa ------+--------------------------------+------------------------------ 1 | 2.33 31.88 21.50 4.57 | 2.35 2.64 6.29 2.43 32 | 2.79 30.68 21.98 4.60 | 2.44 2.59 6.78 2.48 256 | 3.25 31.99 21.36 5.92 | 2.32 2.71 6.22 3.14 1024 | 3.30 32.61 22.83 8.45 | 2.40 2.79 6.24 3.27 4096 | 3.57 317.46 22.52 10.73 | 2.78 28.62 6.90 3.94 16384 | n/a 2342.39 34.30 19.20 | n/a 45.44 145.90 7.55 65536 | n/a 2941.39 128.15 131.32* | n/a 141.92 256.85 262.12* Legend: * ext4 - Stock RHEL6.1 ext4 mounted with '-o user_xattr'. * xfs - Stock RHEL6.1 xfs mounted with default options. * zfs-dir - Directory based xattrs only. * zfs-sa - Prefer SAs but spill in to directories as needed, a trailing * indicates overflow in to directories occured. NOTE: Ext4 supports 4096 bytes of xattr name/value pairs per file. NOTE: XFS and ZFS have no limit on xattr name/value pairs per file. NOTE: Linux limits individual name/value pairs to 65536 bytes. NOTE: All setattr/getattr's were done after dropping the cache. NOTE: All tests were run against a single hard drive. Signed-off-by: Brian Behlendorf <behlendorf1@llnl.gov> Issue #443
2011-10-24 23:55:20 +00:00
error = zpl_xattr_get(ip, xattr_name, value, size);
Add Hooks for Linux Xattr Operations The Linux specific xattr operations have all been located in the file zpl_xattr.c. These functions primarily rely on the reworked zfs_* functions to do their job. They are also responsible for converting the possible Solaris style error codes to negative Linux errors.
2011-01-26 20:10:01 +00:00
strfree(xattr_name);
return (error);
}
Linux 2.6.33 compat, get/set xattr callbacks The xattr handler prototypes were sanitized with the idea being that the same handlers could be used for multiple methods. The result of this was the inode type was changes to a dentry, and both the get() and set() hooks had a handler_flags argument added. The list() callback was similiarly effected but no autoconf check was added because we do not use the list() callback.
2011-02-11 18:33:01 +00:00
ZPL_XATTR_GET_WRAPPER(zpl_xattr_security_get);
Add Hooks for Linux Xattr Operations The Linux specific xattr operations have all been located in the file zpl_xattr.c. These functions primarily rely on the reworked zfs_* functions to do their job. They are also responsible for converting the possible Solaris style error codes to negative Linux errors.
2011-01-26 20:10:01 +00:00
static int
Linux 2.6.33 compat, get/set xattr callbacks The xattr handler prototypes were sanitized with the idea being that the same handlers could be used for multiple methods. The result of this was the inode type was changes to a dentry, and both the get() and set() hooks had a handler_flags argument added. The list() callback was similiarly effected but no autoconf check was added because we do not use the list() callback.
2011-02-11 18:33:01 +00:00
__zpl_xattr_security_set(struct inode *ip, const char *name,
Add Hooks for Linux Xattr Operations The Linux specific xattr operations have all been located in the file zpl_xattr.c. These functions primarily rely on the reworked zfs_* functions to do their job. They are also responsible for converting the possible Solaris style error codes to negative Linux errors.
2011-01-26 20:10:01 +00:00
const void *value, size_t size, int flags)
{
char *xattr_name;
int error;
if (strcmp(name, "") == 0)
return -EINVAL;
xattr_name = kmem_asprintf("%s%s", XATTR_SECURITY_PREFIX, name);
error = zpl_xattr_set(ip, xattr_name, value, size, flags);
strfree(xattr_name);
return (error);
}
Linux 2.6.33 compat, get/set xattr callbacks The xattr handler prototypes were sanitized with the idea being that the same handlers could be used for multiple methods. The result of this was the inode type was changes to a dentry, and both the get() and set() hooks had a handler_flags argument added. The list() callback was similiarly effected but no autoconf check was added because we do not use the list() callback.
2011-02-11 18:33:01 +00:00
ZPL_XATTR_SET_WRAPPER(zpl_xattr_security_set);
Add Hooks for Linux Xattr Operations The Linux specific xattr operations have all been located in the file zpl_xattr.c. These functions primarily rely on the reworked zfs_* functions to do their job. They are also responsible for converting the possible Solaris style error codes to negative Linux errors.
2011-01-26 20:10:01 +00:00
Linux 3.2 compat, security_inode_init_security() The security_inode_init_security() API has been changed to include a filesystem specific callback to write security extended attributes. This was done to support the initialization of multiple LSM xattrs and the EVM xattr. This change updates the code to use the new API when it's available. Otherwise it falls back to the previous implementation. In addition, the ZFS_AC_KERNEL_6ARGS_SECURITY_INODE_INIT_SECURITY autoconf test has been made more rigerous by passing the expected types. This is done to ensure we always properly the detect the correct form for the security_inode_init_security() API. Signed-off-by: Brian Behlendorf <behlendorf1@llnl.gov> Closes #516
2012-01-12 21:59:44 +00:00
#ifdef HAVE_CALLBACK_SECURITY_INODE_INIT_SECURITY
static int
__zpl_xattr_security_init(struct inode *ip, const struct xattr *xattrs,
void *fs_info)
{
const struct xattr *xattr;
int error = 0;
for (xattr = xattrs; xattr->name != NULL; xattr++) {
error = __zpl_xattr_security_set(ip,
xattr->name, xattr->value, xattr->value_len, 0);
if (error < 0)
break;
}
Apply the ZoL coding standard to zpl_xattr.c Make the indenting in the zpl_xattr.c file consistent with the Sun coding standard by removing soft tabs. Signed-off-by: Brian Behlendorf <behlendorf1@llnl.gov>
2012-01-05 20:27:36 +00:00
return (error);
Linux 3.2 compat, security_inode_init_security() The security_inode_init_security() API has been changed to include a filesystem specific callback to write security extended attributes. This was done to support the initialization of multiple LSM xattrs and the EVM xattr. This change updates the code to use the new API when it's available. Otherwise it falls back to the previous implementation. In addition, the ZFS_AC_KERNEL_6ARGS_SECURITY_INODE_INIT_SECURITY autoconf test has been made more rigerous by passing the expected types. This is done to ensure we always properly the detect the correct form for the security_inode_init_security() API. Signed-off-by: Brian Behlendorf <behlendorf1@llnl.gov> Closes #516
2012-01-12 21:59:44 +00:00
}
int
zpl_xattr_security_init(struct inode *ip, struct inode *dip,
const struct qstr *qstr)
{
return security_inode_init_security(ip, dip, qstr,
&__zpl_xattr_security_init, NULL);
}
#else
Add Hooks for Linux Xattr Operations The Linux specific xattr operations have all been located in the file zpl_xattr.c. These functions primarily rely on the reworked zfs_* functions to do their job. They are also responsible for converting the possible Solaris style error codes to negative Linux errors.
2011-01-26 20:10:01 +00:00
int
Linux compat 2.6.39: security_inode_init_security() The security_inode_init_security() function now takes an additional qstr argument which must be passed in from the dentry if available. Passing a NULL is safe when no qstr is available the relevant security checks will just be skipped. Closes #246 Closes #217 Closes #187
2011-05-19 19:47:32 +00:00
zpl_xattr_security_init(struct inode *ip, struct inode *dip,
const struct qstr *qstr)
Add Hooks for Linux Xattr Operations The Linux specific xattr operations have all been located in the file zpl_xattr.c. These functions primarily rely on the reworked zfs_* functions to do their job. They are also responsible for converting the possible Solaris style error codes to negative Linux errors.
2011-01-26 20:10:01 +00:00
{
Apply the ZoL coding standard to zpl_xattr.c Make the indenting in the zpl_xattr.c file consistent with the Sun coding standard by removing soft tabs. Signed-off-by: Brian Behlendorf <behlendorf1@llnl.gov>
2012-01-05 20:27:36 +00:00
int error;
size_t len;
void *value;
char *name;
Add Hooks for Linux Xattr Operations The Linux specific xattr operations have all been located in the file zpl_xattr.c. These functions primarily rely on the reworked zfs_* functions to do their job. They are also responsible for converting the possible Solaris style error codes to negative Linux errors.
2011-01-26 20:10:01 +00:00
Apply the ZoL coding standard to zpl_xattr.c Make the indenting in the zpl_xattr.c file consistent with the Sun coding standard by removing soft tabs. Signed-off-by: Brian Behlendorf <behlendorf1@llnl.gov>
2012-01-05 20:27:36 +00:00
error = zpl_security_inode_init_security(ip, dip, qstr,
&name, &value, &len);
if (error) {
if (error == -EOPNOTSUPP)
return 0;
return (error);
}
Add Hooks for Linux Xattr Operations The Linux specific xattr operations have all been located in the file zpl_xattr.c. These functions primarily rely on the reworked zfs_* functions to do their job. They are also responsible for converting the possible Solaris style error codes to negative Linux errors.
2011-01-26 20:10:01 +00:00
Linux 2.6.33 compat, get/set xattr callbacks The xattr handler prototypes were sanitized with the idea being that the same handlers could be used for multiple methods. The result of this was the inode type was changes to a dentry, and both the get() and set() hooks had a handler_flags argument added. The list() callback was similiarly effected but no autoconf check was added because we do not use the list() callback.
2011-02-11 18:33:01 +00:00
error = __zpl_xattr_security_set(ip, name, value, len, 0);
Add Hooks for Linux Xattr Operations The Linux specific xattr operations have all been located in the file zpl_xattr.c. These functions primarily rely on the reworked zfs_* functions to do their job. They are also responsible for converting the possible Solaris style error codes to negative Linux errors.
2011-01-26 20:10:01 +00:00
Apply the ZoL coding standard to zpl_xattr.c Make the indenting in the zpl_xattr.c file consistent with the Sun coding standard by removing soft tabs. Signed-off-by: Brian Behlendorf <behlendorf1@llnl.gov>
2012-01-05 20:27:36 +00:00
kfree(name);
kfree(value);
Add Hooks for Linux Xattr Operations The Linux specific xattr operations have all been located in the file zpl_xattr.c. These functions primarily rely on the reworked zfs_* functions to do their job. They are also responsible for converting the possible Solaris style error codes to negative Linux errors.
2011-01-26 20:10:01 +00:00
Apply the ZoL coding standard to zpl_xattr.c Make the indenting in the zpl_xattr.c file consistent with the Sun coding standard by removing soft tabs. Signed-off-by: Brian Behlendorf <behlendorf1@llnl.gov>
2012-01-05 20:27:36 +00:00
return (error);
Add Hooks for Linux Xattr Operations The Linux specific xattr operations have all been located in the file zpl_xattr.c. These functions primarily rely on the reworked zfs_* functions to do their job. They are also responsible for converting the possible Solaris style error codes to negative Linux errors.
2011-01-26 20:10:01 +00:00
}
Linux 3.2 compat, security_inode_init_security() The security_inode_init_security() API has been changed to include a filesystem specific callback to write security extended attributes. This was done to support the initialization of multiple LSM xattrs and the EVM xattr. This change updates the code to use the new API when it's available. Otherwise it falls back to the previous implementation. In addition, the ZFS_AC_KERNEL_6ARGS_SECURITY_INODE_INIT_SECURITY autoconf test has been made more rigerous by passing the expected types. This is done to ensure we always properly the detect the correct form for the security_inode_init_security() API. Signed-off-by: Brian Behlendorf <behlendorf1@llnl.gov> Closes #516
2012-01-12 21:59:44 +00:00
#endif /* HAVE_CALLBACK_SECURITY_INODE_INIT_SECURITY */
Add Hooks for Linux Xattr Operations The Linux specific xattr operations have all been located in the file zpl_xattr.c. These functions primarily rely on the reworked zfs_* functions to do their job. They are also responsible for converting the possible Solaris style error codes to negative Linux errors.
2011-01-26 20:10:01 +00:00
Linux 2.6.35 compat, const struct xattr_handler The const keyword was added to the 'struct xattr_handler' in the generic Linux super_block structure. To handle this we define an appropriate xattr_handler_t typedef which can be used. This was the preferred solution because it keeps the code clean and readable.
2011-02-11 00:16:52 +00:00
xattr_handler_t zpl_xattr_security_handler = {
Add Hooks for Linux Xattr Operations The Linux specific xattr operations have all been located in the file zpl_xattr.c. These functions primarily rely on the reworked zfs_* functions to do their job. They are also responsible for converting the possible Solaris style error codes to negative Linux errors.
2011-01-26 20:10:01 +00:00
.prefix = XATTR_SECURITY_PREFIX,
.get = zpl_xattr_security_get,
.set = zpl_xattr_security_set,
};
Honor CONFIG_FS_POSIX_ACL kernel option The required Posix ACL interfaces are only available for kernels with CONFIG_FS_POSIX_ACL defined. Therefore, only enable Posix ACL support for these kernels. All major distribution kernels enable CONFIG_FS_POSIX_ACL by default. If your kernel does not support Posix ACLs the following warning will be printed at ZFS module load time. "ZFS: Posix ACLs disabled by kernel" Signed-off-by: Massimo Maggi <me@massimo-maggi.eu> Signed-off-by: Brian Behlendorf <behlendorf1@llnl.gov> Closes #1825
2013-11-02 23:40:26 +00:00
#ifdef CONFIG_FS_POSIX_ACL
Posix ACL Support This change adds support for Posix ACLs by storing them as an xattr which is common practice for many Linux file systems. Since the Posix ACL is stored as an xattr it will not overwrite any existing ZFS/NFSv4 ACLs which may have been set. The Posix ACL will also be non-functional on other platforms although it may be visible as an xattr if that platform understands SA based xattrs. By default Posix ACLs are disabled but they may be enabled with the new 'aclmode=noacl|posixacl' property. Set the property to 'posixacl' to enable them. If ZFS/NFSv4 ACL support is ever added an appropriate acltype will be added. This change passes the POSIX Test Suite cleanly with the exception of xacl/00.t test 45 which is incorrect for Linux (Ext4 fails too). http://www.tuxera.com/community/posix-test-suite/ Signed-off-by: Massimo Maggi <me@massimo-maggi.eu> Signed-off-by: Richard Yao <ryao@gentoo.org> Signed-off-by: Brian Behlendorf <behlendorf1@llnl.gov> Closes #170
2013-10-28 16:22:15 +00:00
int
zpl_set_acl(struct inode *ip, int type, struct posix_acl *acl)
{
struct super_block *sb = ITOZSB(ip)->z_sb;
char *name, *value = NULL;
int error = 0;
size_t size = 0;
if (S_ISLNK(ip->i_mode))
return (-EOPNOTSUPP);
switch(type) {
case ACL_TYPE_ACCESS:
name = POSIX_ACL_XATTR_ACCESS;
if (acl) {
zpl_equivmode_t mode = ip->i_mode;
error = posix_acl_equiv_mode(acl, &mode);
if (error < 0) {
return (error);
} else {
/*
* The mode bits will have been set by
* ->zfs_setattr()->zfs_acl_chmod_setattr()
* using the ZFS ACL conversion. If they
* differ from the Posix ACL conversion dirty
* the inode to write the Posix mode bits.
*/
if (ip->i_mode != mode) {
ip->i_mode = mode;
ip->i_ctime = current_fs_time(sb);
mark_inode_dirty(ip);
}
if (error == 0)
acl = NULL;
}
}
break;
case ACL_TYPE_DEFAULT:
name = POSIX_ACL_XATTR_DEFAULT;
if (!S_ISDIR(ip->i_mode))
return (acl ? -EACCES : 0);
break;
default:
return (-EINVAL);
}
if (acl) {
size = posix_acl_xattr_size(acl->a_count);
value = kmem_alloc(size, KM_SLEEP);
error = zpl_acl_to_xattr(acl, value, size);
if (error < 0) {
kmem_free(value, size);
return (error);
}
}
error = zpl_xattr_set(ip, name, value, size, 0);
if (value)
kmem_free(value, size);
if (!error) {
if (acl)
zpl_set_cached_acl(ip, type, acl);
else
zpl_forget_cached_acl(ip, type);
}
return (error);
}
struct posix_acl *
zpl_get_acl(struct inode *ip, int type)
{
struct posix_acl *acl;
void *value = NULL;
char *name;
int size;
#ifdef HAVE_POSIX_ACL_CACHING
acl = get_cached_acl(ip, type);
if (acl != ACL_NOT_CACHED)
return (acl);
#endif /* HAVE_POSIX_ACL_CACHING */
switch (type) {
case ACL_TYPE_ACCESS:
name = POSIX_ACL_XATTR_ACCESS;
break;
case ACL_TYPE_DEFAULT:
name = POSIX_ACL_XATTR_DEFAULT;
break;
default:
return ERR_PTR(-EINVAL);
}
size = zpl_xattr_get(ip, name, NULL, 0);
if (size > 0) {
value = kmem_alloc(size, KM_PUSHPAGE);
size = zpl_xattr_get(ip, name, value, size);
}
if (size > 0) {
acl = zpl_acl_from_xattr(value, size);
} else if (size == -ENODATA || size == -ENOSYS) {
acl = NULL;
} else {
acl = ERR_PTR(-EIO);
}
if (size > 0)
kmem_free(value, size);
if (!IS_ERR(acl))
zpl_set_cached_acl(ip, type, acl);
return (acl);
}
#if !defined(HAVE_GET_ACL)
static int
__zpl_check_acl(struct inode *ip, int mask)
{
struct posix_acl *acl;
int error;
acl = zpl_get_acl(ip, ACL_TYPE_ACCESS);
if (IS_ERR(acl))
return (PTR_ERR(acl));
if (acl) {
error = posix_acl_permission(ip, acl, mask);
zpl_posix_acl_release(acl);
return (error);
}
return (-EAGAIN);
}
#if defined(HAVE_CHECK_ACL_WITH_FLAGS)
int
zpl_check_acl(struct inode *ip, int mask, unsigned int flags)
{
return __zpl_check_acl(ip, mask);
}
#elif defined(HAVE_CHECK_ACL)
int
zpl_check_acl(struct inode *ip, int mask)
{
return __zpl_check_acl(ip , mask);
}
#elif defined(HAVE_PERMISSION_WITH_NAMEIDATA)
int
zpl_permission(struct inode *ip, int mask, struct nameidata *nd)
{
return generic_permission(ip, mask, __zpl_check_acl);
}
#elif defined(HAVE_PERMISSION)
int
zpl_permission(struct inode *ip, int mask)
{
return generic_permission(ip, mask, __zpl_check_acl);
}
#endif /* HAVE_CHECK_ACL | HAVE_PERMISSION */
#endif /* !HAVE_GET_ACL */
int
zpl_init_acl(struct inode *ip, struct inode *dir)
{
struct posix_acl *acl = NULL;
int error = 0;
if (ITOZSB(ip)->z_acl_type != ZFS_ACLTYPE_POSIXACL)
return (0);
if (!S_ISLNK(ip->i_mode)) {
if (ITOZSB(ip)->z_acl_type == ZFS_ACLTYPE_POSIXACL) {
acl = zpl_get_acl(dir, ACL_TYPE_DEFAULT);
if (IS_ERR(acl))
return (PTR_ERR(acl));
}
if (!acl) {
ip->i_mode &= ~current_umask();
ip->i_ctime = current_fs_time(ITOZSB(ip)->z_sb);
mark_inode_dirty(ip);
return (0);
}
}
if ((ITOZSB(ip)->z_acl_type == ZFS_ACLTYPE_POSIXACL) && acl) {
umode_t mode;
if (S_ISDIR(ip->i_mode)) {
error = zpl_set_acl(ip, ACL_TYPE_DEFAULT, acl);
if (error)
goto out;
}
mode = ip->i_mode;
error = posix_acl_create(&acl,GFP_KERNEL, &mode);
if (error >= 0) {
ip->i_mode = mode;
mark_inode_dirty(ip);
if (error > 0)
error = zpl_set_acl(ip, ACL_TYPE_ACCESS, acl);
}
}
out:
zpl_posix_acl_release(acl);
return (error);
}
int
zpl_chmod_acl(struct inode *ip)
{
struct posix_acl *acl;
int error;
if (ITOZSB(ip)->z_acl_type != ZFS_ACLTYPE_POSIXACL)
return (0);
if (S_ISLNK(ip->i_mode))
return (-EOPNOTSUPP);
acl = zpl_get_acl(ip, ACL_TYPE_ACCESS);
if (IS_ERR(acl) || !acl)
return (PTR_ERR(acl));
error = posix_acl_chmod(&acl,GFP_KERNEL, ip->i_mode);
if (!error)
error = zpl_set_acl(ip,ACL_TYPE_ACCESS, acl);
zpl_posix_acl_release(acl);
return (error);
}
static size_t
zpl_xattr_acl_list(struct inode *ip, char *list, size_t list_size,
const char *name, size_t name_len, int type)
{
char *xattr_name;
size_t xattr_size;
if (ITOZSB(ip)->z_acl_type != ZFS_ACLTYPE_POSIXACL)
return (0);
switch (type) {
case ACL_TYPE_ACCESS:
xattr_name = POSIX_ACL_XATTR_ACCESS;
xattr_size = sizeof(xattr_name);
break;
case ACL_TYPE_DEFAULT:
xattr_name = POSIX_ACL_XATTR_DEFAULT;
xattr_size = sizeof(xattr_name);
break;
default:
return (0);
}
if (list && xattr_size <= list_size)
memcpy(list, xattr_name, xattr_size);
return (xattr_size);
}
#ifdef HAVE_DENTRY_XATTR_LIST
static size_t
zpl_xattr_acl_list_access(struct dentry *dentry, char *list,
size_t list_size, const char *name, size_t name_len, int type)
{
ASSERT3S(type, ==, ACL_TYPE_ACCESS);
return zpl_xattr_acl_list(dentry->d_inode,
list, list_size, name, name_len, type);
}
static size_t
zpl_xattr_acl_list_default(struct dentry *dentry, char *list,
size_t list_size, const char *name, size_t name_len, int type)
{
ASSERT3S(type, ==, ACL_TYPE_DEFAULT);
return zpl_xattr_acl_list(dentry->d_inode,
list, list_size, name, name_len, type);
}
#else
static size_t
zpl_xattr_acl_list_access(struct inode *ip, char *list, size_t list_size,
const char *name, size_t name_len)
{
return zpl_xattr_acl_list(ip,
list, list_size, name, name_len, ACL_TYPE_ACCESS);
}
static size_t
zpl_xattr_acl_list_default(struct inode *ip, char *list, size_t list_size,
const char *name, size_t name_len)
{
return zpl_xattr_acl_list(ip,
list, list_size, name, name_len, ACL_TYPE_DEFAULT);
}
#endif /* HAVE_DENTRY_XATTR_LIST */
static int
zpl_xattr_acl_get(struct inode *ip, const char *name,
void *buffer, size_t size, int type)
{
struct posix_acl *acl;
int error;
if (strcmp(name, "") != 0)
return (-EINVAL);
if (ITOZSB(ip)->z_acl_type != ZFS_ACLTYPE_POSIXACL)
return (-EOPNOTSUPP);
acl = zpl_get_acl(ip, type);
if (IS_ERR(acl))
return (PTR_ERR(acl));
if (acl == NULL)
return (-ENODATA);
error = zpl_acl_to_xattr(acl, buffer, size);
zpl_posix_acl_release(acl);
return (error);
}
#ifdef HAVE_DENTRY_XATTR_GET
static int
zpl_xattr_acl_get_access(struct dentry *dentry, const char *name,
void *buffer, size_t size, int type)
{
ASSERT3S(type, ==, ACL_TYPE_ACCESS);
return zpl_xattr_acl_get(dentry->d_inode, name, buffer, size, type);
}
static int
zpl_xattr_acl_get_default(struct dentry *dentry, const char *name,
void *buffer, size_t size, int type)
{
ASSERT3S(type, ==, ACL_TYPE_DEFAULT);
return zpl_xattr_acl_get(dentry->d_inode, name, buffer, size, type);
}
#else
static int
zpl_xattr_acl_get_access(struct inode *ip, const char *name,
void *buffer, size_t size)
{
return zpl_xattr_acl_get(ip, name, buffer, size, ACL_TYPE_ACCESS);
}
static int
zpl_xattr_acl_get_default(struct inode *ip, const char *name,
void *buffer, size_t size)
{
return zpl_xattr_acl_get(ip, name, buffer, size, ACL_TYPE_DEFAULT);
}
#endif /* HAVE_DENTRY_XATTR_GET */
static int
zpl_xattr_acl_set(struct inode *ip, const char *name,
const void *value, size_t size, int flags, int type)
{
struct posix_acl *acl;
int error = 0;
if (strcmp(name, "") != 0)
return (-EINVAL);
if (ITOZSB(ip)->z_acl_type != ZFS_ACLTYPE_POSIXACL)
return (-EOPNOTSUPP);
if (!zpl_inode_owner_or_capable(ip))
return (-EPERM);
if (value) {
acl = zpl_acl_from_xattr(value, size);
if (IS_ERR(acl))
return (PTR_ERR(acl));
else if (acl) {
error = posix_acl_valid(acl);
if (error) {
zpl_posix_acl_release(acl);
return (error);
}
}
} else {
acl = NULL;
}
error = zpl_set_acl(ip, type, acl);
zpl_posix_acl_release(acl);
return (error);
}
#ifdef HAVE_DENTRY_XATTR_SET
static int
zpl_xattr_acl_set_access(struct dentry *dentry, const char *name,
const void *value, size_t size, int flags, int type)
{
ASSERT3S(type, ==, ACL_TYPE_ACCESS);
return zpl_xattr_acl_set(dentry->d_inode,
name, value, size, flags, type);
}
static int
zpl_xattr_acl_set_default(struct dentry *dentry, const char *name,
const void *value, size_t size,int flags, int type)
{
ASSERT3S(type, ==, ACL_TYPE_DEFAULT);
return zpl_xattr_acl_set(dentry->d_inode,
name, value, size, flags, type);
}
#else
static int
zpl_xattr_acl_set_access(struct inode *ip, const char *name,
const void *value, size_t size, int flags)
{
return zpl_xattr_acl_set(ip,
name, value, size, flags, ACL_TYPE_ACCESS);
}
static int
zpl_xattr_acl_set_default(struct inode *ip, const char *name,
const void *value, size_t size, int flags)
{
return zpl_xattr_acl_set(ip,
name, value, size, flags, ACL_TYPE_DEFAULT);
}
#endif /* HAVE_DENTRY_XATTR_SET */
struct xattr_handler zpl_xattr_acl_access_handler =
{
.prefix = POSIX_ACL_XATTR_ACCESS,
.list = zpl_xattr_acl_list_access,
.get = zpl_xattr_acl_get_access,
.set = zpl_xattr_acl_set_access,
#ifdef HAVE_DENTRY_XATTR_LIST
.flags = ACL_TYPE_ACCESS,
#endif /* HAVE_DENTRY_XATTR_LIST */
};
struct xattr_handler zpl_xattr_acl_default_handler =
{
.prefix = POSIX_ACL_XATTR_DEFAULT,
.list = zpl_xattr_acl_list_default,
.get = zpl_xattr_acl_get_default,
.set = zpl_xattr_acl_set_default,
#ifdef HAVE_DENTRY_XATTR_LIST
.flags = ACL_TYPE_DEFAULT,
#endif /* HAVE_DENTRY_XATTR_LIST */
};
Honor CONFIG_FS_POSIX_ACL kernel option The required Posix ACL interfaces are only available for kernels with CONFIG_FS_POSIX_ACL defined. Therefore, only enable Posix ACL support for these kernels. All major distribution kernels enable CONFIG_FS_POSIX_ACL by default. If your kernel does not support Posix ACLs the following warning will be printed at ZFS module load time. "ZFS: Posix ACLs disabled by kernel" Signed-off-by: Massimo Maggi <me@massimo-maggi.eu> Signed-off-by: Brian Behlendorf <behlendorf1@llnl.gov> Closes #1825
2013-11-02 23:40:26 +00:00
#endif /* CONFIG_FS_POSIX_ACL */
Linux 2.6.35 compat, const struct xattr_handler The const keyword was added to the 'struct xattr_handler' in the generic Linux super_block structure. To handle this we define an appropriate xattr_handler_t typedef which can be used. This was the preferred solution because it keeps the code clean and readable.
2011-02-11 00:16:52 +00:00
xattr_handler_t *zpl_xattr_handlers[] = {
Add Hooks for Linux Xattr Operations The Linux specific xattr operations have all been located in the file zpl_xattr.c. These functions primarily rely on the reworked zfs_* functions to do their job. They are also responsible for converting the possible Solaris style error codes to negative Linux errors.
2011-01-26 20:10:01 +00:00
&zpl_xattr_security_handler,
&zpl_xattr_trusted_handler,
&zpl_xattr_user_handler,
Honor CONFIG_FS_POSIX_ACL kernel option The required Posix ACL interfaces are only available for kernels with CONFIG_FS_POSIX_ACL defined. Therefore, only enable Posix ACL support for these kernels. All major distribution kernels enable CONFIG_FS_POSIX_ACL by default. If your kernel does not support Posix ACLs the following warning will be printed at ZFS module load time. "ZFS: Posix ACLs disabled by kernel" Signed-off-by: Massimo Maggi <me@massimo-maggi.eu> Signed-off-by: Brian Behlendorf <behlendorf1@llnl.gov> Closes #1825
2013-11-02 23:40:26 +00:00
#ifdef CONFIG_FS_POSIX_ACL
Add Hooks for Linux Xattr Operations The Linux specific xattr operations have all been located in the file zpl_xattr.c. These functions primarily rely on the reworked zfs_* functions to do their job. They are also responsible for converting the possible Solaris style error codes to negative Linux errors.
2011-01-26 20:10:01 +00:00
&zpl_xattr_acl_access_handler,
Apply the ZoL coding standard to zpl_xattr.c Make the indenting in the zpl_xattr.c file consistent with the Sun coding standard by removing soft tabs. Signed-off-by: Brian Behlendorf <behlendorf1@llnl.gov>
2012-01-05 20:27:36 +00:00
&zpl_xattr_acl_default_handler,
Honor CONFIG_FS_POSIX_ACL kernel option The required Posix ACL interfaces are only available for kernels with CONFIG_FS_POSIX_ACL defined. Therefore, only enable Posix ACL support for these kernels. All major distribution kernels enable CONFIG_FS_POSIX_ACL by default. If your kernel does not support Posix ACLs the following warning will be printed at ZFS module load time. "ZFS: Posix ACLs disabled by kernel" Signed-off-by: Massimo Maggi <me@massimo-maggi.eu> Signed-off-by: Brian Behlendorf <behlendorf1@llnl.gov> Closes #1825
2013-11-02 23:40:26 +00:00
#endif /* CONFIG_FS_POSIX_ACL */
Add missing NULL in zpl_xattr_handlers The xattr_resolve_name() helper function expects the registered list of xattr handlers to be NULL terminated. This NULL was accidentally missing which could result in a NULL dereference. Interestingly this issue only manifested itself on certain 32-bit systems. Presumably on 64-bit kernels we just always happen to get lucky and the memory following the structure is zeroed. Signed-off-by: Brian Behlendorf <behlendorf1@llnl.gov> Issue #594
2012-03-14 19:36:49 +00:00
NULL
Add Hooks for Linux Xattr Operations The Linux specific xattr operations have all been located in the file zpl_xattr.c. These functions primarily rely on the reworked zfs_* functions to do their job. They are also responsible for converting the possible Solaris style error codes to negative Linux errors.
2011-01-26 20:10:01 +00:00
};
Reference in New Issue Copy Permalink