| Message ID | 20230607180958.645115-1-aleksandr.mikhalitsyn@canonical.com (mailing list archive) |
|---|---|
| Headers | show |
| Series | ceph: support idmapped mounts | expand |
Hi Alexander, As I mentioned in V2 thread https://www.spinics.net/lists/kernel/msg4810994.html, we should use the 'idmap' for all the requests below, because MDS will do the 'check_access()' for all the requests by using the caller uid/gid, please see https://github.com/ceph/ceph/blob/main/src/mds/Server.cc#L3294-L3310. Cscope tag: ceph_mdsc_do_request # line filename / context / line 1 321 fs/ceph/addr.c <<ceph_netfs_issue_op_inline>> err = ceph_mdsc_do_request(mdsc, NULL, req); 2 443 fs/ceph/dir.c <<ceph_readdir>> err = ceph_mdsc_do_request(mdsc, NULL, req); 3 838 fs/ceph/dir.c <<ceph_lookup>> err = ceph_mdsc_do_request(mdsc, NULL, req); 4 933 fs/ceph/dir.c <<ceph_mknod>> err = ceph_mdsc_do_request(mdsc, dir, req); 5 1045 fs/ceph/dir.c <<ceph_symlink>> err = ceph_mdsc_do_request(mdsc, dir, req); 6 1120 fs/ceph/dir.c <<ceph_mkdir>> err = ceph_mdsc_do_request(mdsc, dir, req); 7 1180 fs/ceph/dir.c <<ceph_link>> err = ceph_mdsc_do_request(mdsc, dir, req); 8 1365 fs/ceph/dir.c <<ceph_unlink>> err = ceph_mdsc_do_request(mdsc, dir, req); 9 1431 fs/ceph/dir.c <<ceph_rename>> err = ceph_mdsc_do_request(mdsc, old_dir, req); 10 1927 fs/ceph/dir.c <<ceph_d_revalidate>> err = ceph_mdsc_do_request(mdsc, NULL, req); 11 154 fs/ceph/export.c <<__lookup_inode>> err = ceph_mdsc_do_request(mdsc, NULL, req); 12 262 fs/ceph/export.c <<__snapfh_to_dentry>> err = ceph_mdsc_do_request(mdsc, NULL, req); 13 347 fs/ceph/export.c <<__get_parent>> err = ceph_mdsc_do_request(mdsc, NULL, req); 14 490 fs/ceph/export.c <<__get_snap_name>> err = ceph_mdsc_do_request(fsc->mdsc, NULL, req); 15 561 fs/ceph/export.c <<ceph_get_name>> err = ceph_mdsc_do_request(mdsc, NULL, req); 16 339 fs/ceph/file.c <<ceph_renew_caps>> err = ceph_mdsc_do_request(mdsc, NULL, req); 17 434 fs/ceph/file.c <<ceph_open>> err = ceph_mdsc_do_request(mdsc, NULL, req); 18 855 fs/ceph/file.c <<ceph_atomic_open>> err = ceph_mdsc_do_request(mdsc, (flags & O_CREAT) ? dir : NULL, req); 19 2715 fs/ceph/inode.c <<__ceph_setattr>> err = ceph_mdsc_do_request(mdsc, NULL, req); 20 2839 fs/ceph/inode.c <<__ceph_do_getattr>> err = ceph_mdsc_do_request(mdsc, NULL, req); 21 2883 fs/ceph/inode.c <<ceph_do_getvxattr>> err = ceph_mdsc_do_request(mdsc, NULL, req); 22 126 fs/ceph/ioctl.c <<ceph_ioctl_set_layout>> err = ceph_mdsc_do_request(mdsc, NULL, req); 23 171 fs/ceph/ioctl.c <<ceph_ioctl_set_layout_policy>> err = ceph_mdsc_do_request(mdsc, inode, req); 24 216 fs/ceph/locks.c <<ceph_lock_wait_for_completion>> err = ceph_mdsc_do_request(mdsc, inode, intr_req); 25 1091 fs/ceph/super.c <<open_root_dentry>> err = ceph_mdsc_do_request(mdsc, NULL, req); 26 1151 fs/ceph/xattr.c <<ceph_sync_setxattr>> err = ceph_mdsc_do_request(mdsc, NULL, req); And also could you squash the similar commit into one ? Thanks - Xiubo On 6/8/23 02:09, Alexander Mikhalitsyn wrote: > Dear friends, > > This patchset was originally developed by Christian Brauner but I'll continue > to push it forward. Christian allowed me to do that :) > > This feature is already actively used/tested with LXD/LXC project. > > Git tree (based on https://github.com/ceph/ceph-client.git master): > https://github.com/mihalicyn/linux/tree/fs.idmapped.ceph > > In the version 3 I've changed only two commits: > - fs: export mnt_idmap_get/mnt_idmap_put > - ceph: allow idmapped setattr inode op > and added a new one: > - ceph: pass idmap to __ceph_setattr > > In the version 4 I've reworked the ("ceph: stash idmapping in mdsc request") > commit. Now we take idmap refcounter just in place where req->r_mnt_idmap > is filled. It's more safer approach and prevents possible refcounter underflow > on error paths where __register_request wasn't called but ceph_mdsc_release_request is > called. > > I can confirm that this version passes xfstests. > > Links to previous versions: > v1: https://lore.kernel.org/all/20220104140414.155198-1-brauner@kernel.org/ > v2: https://lore.kernel.org/lkml/20230524153316.476973-1-aleksandr.mikhalitsyn@canonical.com/ > v3: https://lore.kernel.org/lkml/20230607152038.469739-1-aleksandr.mikhalitsyn@canonical.com/#t > > Kind regards, > Alex > > Original description from Christian: > ======================================================================== > This patch series enables cephfs to support idmapped mounts, i.e. the > ability to alter ownership information on a per-mount basis. > > Container managers such as LXD support sharaing data via cephfs between > the host and unprivileged containers and between unprivileged containers. > They may all use different idmappings. Idmapped mounts can be used to > create mounts with the idmapping used for the container (or a different > one specific to the use-case). > > There are in fact more use-cases such as remapping ownership for > mountpoints on the host itself to grant or restrict access to different > users or to make it possible to enforce that programs running as root > will write with a non-zero {g,u}id to disk. > > The patch series is simple overall and few changes are needed to cephfs. > There is one cephfs specific issue that I would like to discuss and > solve which I explain in detail in: > > [PATCH 02/12] ceph: handle idmapped mounts in create_request_message() > > It has to do with how to handle mds serves which have id-based access > restrictions configured. I would ask you to please take a look at the > explanation in the aforementioned patch. > > The patch series passes the vfs and idmapped mount testsuite as part of > xfstests. To run it you will need a config like: > > [ceph] > export FSTYP=ceph > export TEST_DIR=/mnt/test > export TEST_DEV=10.103.182.10:6789:/ > export TEST_FS_MOUNT_OPTS="-o name=admin,secret=$password > > and then simply call > > sudo ./check -g idmapped > > ======================================================================== > > Alexander Mikhalitsyn (2): > fs: export mnt_idmap_get/mnt_idmap_put > ceph: pass idmap to __ceph_setattr > > Christian Brauner (12): > ceph: stash idmapping in mdsc request > ceph: handle idmapped mounts in create_request_message() > ceph: allow idmapped mknod inode op > ceph: allow idmapped symlink inode op > ceph: allow idmapped mkdir inode op > ceph: allow idmapped rename inode op > ceph: allow idmapped getattr inode op > ceph: allow idmapped permission inode op > ceph: allow idmapped setattr inode op > ceph/acl: allow idmapped set_acl inode op > ceph/file: allow idmapped atomic_open inode op > ceph: allow idmapped mounts > > fs/ceph/acl.c | 6 +++--- > fs/ceph/dir.c | 4 ++++ > fs/ceph/file.c | 10 ++++++++-- > fs/ceph/inode.c | 29 +++++++++++++++++------------ > fs/ceph/mds_client.c | 27 +++++++++++++++++++++++---- > fs/ceph/mds_client.h | 1 + > fs/ceph/super.c | 2 +- > fs/ceph/super.h | 3 ++- > fs/mnt_idmapping.c | 2 ++ > include/linux/mnt_idmapping.h | 3 +++ > 10 files changed, 64 insertions(+), 23 deletions(-) >
On Thu, Jun 8, 2023 at 5:01 AM Xiubo Li <xiubli@redhat.com> wrote: > > Hi Alexander, Dear Xiubo, > > As I mentioned in V2 thread > https://www.spinics.net/lists/kernel/msg4810994.html, we should use the > 'idmap' for all the requests below, because MDS will do the > 'check_access()' for all the requests by using the caller uid/gid, > please see > https://github.com/ceph/ceph/blob/main/src/mds/Server.cc#L3294-L3310. > > > Cscope tag: ceph_mdsc_do_request > # line filename / context / line > 1 321 fs/ceph/addr.c <<ceph_netfs_issue_op_inline>> > err = ceph_mdsc_do_request(mdsc, NULL, req); > 2 443 fs/ceph/dir.c <<ceph_readdir>> > err = ceph_mdsc_do_request(mdsc, NULL, req); > 3 838 fs/ceph/dir.c <<ceph_lookup>> > err = ceph_mdsc_do_request(mdsc, NULL, req); > 4 933 fs/ceph/dir.c <<ceph_mknod>> > err = ceph_mdsc_do_request(mdsc, dir, req); > 5 1045 fs/ceph/dir.c <<ceph_symlink>> > err = ceph_mdsc_do_request(mdsc, dir, req); > 6 1120 fs/ceph/dir.c <<ceph_mkdir>> > err = ceph_mdsc_do_request(mdsc, dir, req); > 7 1180 fs/ceph/dir.c <<ceph_link>> > err = ceph_mdsc_do_request(mdsc, dir, req); > 8 1365 fs/ceph/dir.c <<ceph_unlink>> > err = ceph_mdsc_do_request(mdsc, dir, req); > 9 1431 fs/ceph/dir.c <<ceph_rename>> > err = ceph_mdsc_do_request(mdsc, old_dir, req); > 10 1927 fs/ceph/dir.c <<ceph_d_revalidate>> > err = ceph_mdsc_do_request(mdsc, NULL, req); > 11 154 fs/ceph/export.c <<__lookup_inode>> > err = ceph_mdsc_do_request(mdsc, NULL, req); > 12 262 fs/ceph/export.c <<__snapfh_to_dentry>> > err = ceph_mdsc_do_request(mdsc, NULL, req); > 13 347 fs/ceph/export.c <<__get_parent>> > err = ceph_mdsc_do_request(mdsc, NULL, req); > 14 490 fs/ceph/export.c <<__get_snap_name>> > err = ceph_mdsc_do_request(fsc->mdsc, NULL, req); > 15 561 fs/ceph/export.c <<ceph_get_name>> > err = ceph_mdsc_do_request(mdsc, NULL, req); > 16 339 fs/ceph/file.c <<ceph_renew_caps>> > err = ceph_mdsc_do_request(mdsc, NULL, req); > 17 434 fs/ceph/file.c <<ceph_open>> > err = ceph_mdsc_do_request(mdsc, NULL, req); > 18 855 fs/ceph/file.c <<ceph_atomic_open>> > err = ceph_mdsc_do_request(mdsc, (flags & O_CREAT) ? dir : > NULL, req); > 19 2715 fs/ceph/inode.c <<__ceph_setattr>> > err = ceph_mdsc_do_request(mdsc, NULL, req); > 20 2839 fs/ceph/inode.c <<__ceph_do_getattr>> > err = ceph_mdsc_do_request(mdsc, NULL, req); > 21 2883 fs/ceph/inode.c <<ceph_do_getvxattr>> > err = ceph_mdsc_do_request(mdsc, NULL, req); > 22 126 fs/ceph/ioctl.c <<ceph_ioctl_set_layout>> > err = ceph_mdsc_do_request(mdsc, NULL, req); > 23 171 fs/ceph/ioctl.c <<ceph_ioctl_set_layout_policy>> > err = ceph_mdsc_do_request(mdsc, inode, req); > 24 216 fs/ceph/locks.c <<ceph_lock_wait_for_completion>> > err = ceph_mdsc_do_request(mdsc, inode, intr_req); > 25 1091 fs/ceph/super.c <<open_root_dentry>> > err = ceph_mdsc_do_request(mdsc, NULL, req); > 26 1151 fs/ceph/xattr.c <<ceph_sync_setxattr>> > err = ceph_mdsc_do_request(mdsc, NULL, req); Sure, I remember about this point and as far as I mentioned earlier https://lore.kernel.org/all/20230519134420.2d04e5f70aad15679ab566fc@canonical.com/ It is a discussional thing, because not all the inode_operations allows us to get a mount idmapping. For instance, lookup, get_link, get_inode_acl, readlink, link, unlink, rmdir, listxattr, fiemap, update_time, fileattr_get inode_operations are not provided with an idmapping. atomic_open also lacks the mnt_idmap argument, but we have a struct file so we can get an idmapping through it. As far as I can see from the code: https://raw.githubusercontent.com/ceph/ceph/main/src/mds/Server.cc We have Server::check_access calls for all inode_operations, including the lookup. It means that with the current VFS we are not able to support MDS UID/GID-based path restriction with idmapped mounts. But we can return to it later if someone really wants it. If I understand your idea right, you want me to set req->r_mnt_idmap to an actual idmapping everywhere, where it is possible, and ignore inode_operations where we have no idmapping passed? Christian's approach was more conservative here, his idea was to pass an idmapping only to the operations that are creating some nodes on the filesystem, but pass a "nop_mnt_idmap" to everyone else. So, I'll try to set up MDS UID/GID-based path restriction on my local environment and reproduce the issue with it, but as I mentioned earlier we can't support it right now anyway. But as we already have an idmappings supported for most of existing filesystems, having it supported for cephfs would be great (even with this limitation about MDS UID/GID-based path restriction), because we already have a real world use cases for cephfs idmapped mounts and this particular patchset is used by LXD/LXC for more that a year. We can extend this later, if someone really wants to use this combination and once we extend the VFS layer. > > > And also could you squash the similar commit into one ? Sure, you mean commits that do `req->r_mnt_idmap = mnt_idmap_get(idmap)`? Will do. Big thanks for the fast reaction/review on this series, Xiubo! > Kind regards, Alex > > Thanks > > - Xiubo > > > On 6/8/23 02:09, Alexander Mikhalitsyn wrote: > > Dear friends, > > > > This patchset was originally developed by Christian Brauner but I'll continue > > to push it forward. Christian allowed me to do that :) > > > > This feature is already actively used/tested with LXD/LXC project. > > > > Git tree (based on https://github.com/ceph/ceph-client.git master): > > https://github.com/mihalicyn/linux/tree/fs.idmapped.ceph > > > > In the version 3 I've changed only two commits: > > - fs: export mnt_idmap_get/mnt_idmap_put > > - ceph: allow idmapped setattr inode op > > and added a new one: > > - ceph: pass idmap to __ceph_setattr > > > > In the version 4 I've reworked the ("ceph: stash idmapping in mdsc request") > > commit. Now we take idmap refcounter just in place where req->r_mnt_idmap > > is filled. It's more safer approach and prevents possible refcounter underflow > > on error paths where __register_request wasn't called but ceph_mdsc_release_request is > > called. > > > > I can confirm that this version passes xfstests. > > > > Links to previous versions: > > v1: https://lore.kernel.org/all/20220104140414.155198-1-brauner@kernel.org/ > > v2: https://lore.kernel.org/lkml/20230524153316.476973-1-aleksandr.mikhalitsyn@canonical.com/ > > v3: https://lore.kernel.org/lkml/20230607152038.469739-1-aleksandr.mikhalitsyn@canonical.com/#t > > > > Kind regards, > > Alex > > > > Original description from Christian: > > ======================================================================== > > This patch series enables cephfs to support idmapped mounts, i.e. the > > ability to alter ownership information on a per-mount basis. > > > > Container managers such as LXD support sharaing data via cephfs between > > the host and unprivileged containers and between unprivileged containers. > > They may all use different idmappings. Idmapped mounts can be used to > > create mounts with the idmapping used for the container (or a different > > one specific to the use-case). > > > > There are in fact more use-cases such as remapping ownership for > > mountpoints on the host itself to grant or restrict access to different > > users or to make it possible to enforce that programs running as root > > will write with a non-zero {g,u}id to disk. > > > > The patch series is simple overall and few changes are needed to cephfs. > > There is one cephfs specific issue that I would like to discuss and > > solve which I explain in detail in: > > > > [PATCH 02/12] ceph: handle idmapped mounts in create_request_message() > > > > It has to do with how to handle mds serves which have id-based access > > restrictions configured. I would ask you to please take a look at the > > explanation in the aforementioned patch. > > > > The patch series passes the vfs and idmapped mount testsuite as part of > > xfstests. To run it you will need a config like: > > > > [ceph] > > export FSTYP=ceph > > export TEST_DIR=/mnt/test > > export TEST_DEV=10.103.182.10:6789:/ > > export TEST_FS_MOUNT_OPTS="-o name=admin,secret=$password > > > > and then simply call > > > > sudo ./check -g idmapped > > > > ======================================================================== > > > > Alexander Mikhalitsyn (2): > > fs: export mnt_idmap_get/mnt_idmap_put > > ceph: pass idmap to __ceph_setattr > > > > Christian Brauner (12): > > ceph: stash idmapping in mdsc request > > ceph: handle idmapped mounts in create_request_message() > > ceph: allow idmapped mknod inode op > > ceph: allow idmapped symlink inode op > > ceph: allow idmapped mkdir inode op > > ceph: allow idmapped rename inode op > > ceph: allow idmapped getattr inode op > > ceph: allow idmapped permission inode op > > ceph: allow idmapped setattr inode op > > ceph/acl: allow idmapped set_acl inode op > > ceph/file: allow idmapped atomic_open inode op > > ceph: allow idmapped mounts > > > > fs/ceph/acl.c | 6 +++--- > > fs/ceph/dir.c | 4 ++++ > > fs/ceph/file.c | 10 ++++++++-- > > fs/ceph/inode.c | 29 +++++++++++++++++------------ > > fs/ceph/mds_client.c | 27 +++++++++++++++++++++++---- > > fs/ceph/mds_client.h | 1 + > > fs/ceph/super.c | 2 +- > > fs/ceph/super.h | 3 ++- > > fs/mnt_idmapping.c | 2 ++ > > include/linux/mnt_idmapping.h | 3 +++ > > 10 files changed, 64 insertions(+), 23 deletions(-) > > >
Dear friends, This patchset was originally developed by Christian Brauner but I'll continue to push it forward. Christian allowed me to do that :) This feature is already actively used/tested with LXD/LXC project. Git tree (based on https://github.com/ceph/ceph-client.git master): https://github.com/mihalicyn/linux/tree/fs.idmapped.ceph In the version 3 I've changed only two commits: - fs: export mnt_idmap_get/mnt_idmap_put - ceph: allow idmapped setattr inode op and added a new one: - ceph: pass idmap to __ceph_setattr In the version 4 I've reworked the ("ceph: stash idmapping in mdsc request") commit. Now we take idmap refcounter just in place where req->r_mnt_idmap is filled. It's more safer approach and prevents possible refcounter underflow on error paths where __register_request wasn't called but ceph_mdsc_release_request is called. I can confirm that this version passes xfstests. Links to previous versions: v1: https://lore.kernel.org/all/20220104140414.155198-1-brauner@kernel.org/ v2: https://lore.kernel.org/lkml/20230524153316.476973-1-aleksandr.mikhalitsyn@canonical.com/ v3: https://lore.kernel.org/lkml/20230607152038.469739-1-aleksandr.mikhalitsyn@canonical.com/#t Kind regards, Alex Original description from Christian: ======================================================================== This patch series enables cephfs to support idmapped mounts, i.e. the ability to alter ownership information on a per-mount basis. Container managers such as LXD support sharaing data via cephfs between the host and unprivileged containers and between unprivileged containers. They may all use different idmappings. Idmapped mounts can be used to create mounts with the idmapping used for the container (or a different one specific to the use-case). There are in fact more use-cases such as remapping ownership for mountpoints on the host itself to grant or restrict access to different users or to make it possible to enforce that programs running as root will write with a non-zero {g,u}id to disk. The patch series is simple overall and few changes are needed to cephfs. There is one cephfs specific issue that I would like to discuss and solve which I explain in detail in: [PATCH 02/12] ceph: handle idmapped mounts in create_request_message() It has to do with how to handle mds serves which have id-based access restrictions configured. I would ask you to please take a look at the explanation in the aforementioned patch. The patch series passes the vfs and idmapped mount testsuite as part of xfstests. To run it you will need a config like: [ceph] export FSTYP=ceph export TEST_DIR=/mnt/test export TEST_DEV=10.103.182.10:6789:/ export TEST_FS_MOUNT_OPTS="-o name=admin,secret=$password and then simply call sudo ./check -g idmapped ======================================================================== Alexander Mikhalitsyn (2): fs: export mnt_idmap_get/mnt_idmap_put ceph: pass idmap to __ceph_setattr Christian Brauner (12): ceph: stash idmapping in mdsc request ceph: handle idmapped mounts in create_request_message() ceph: allow idmapped mknod inode op ceph: allow idmapped symlink inode op ceph: allow idmapped mkdir inode op ceph: allow idmapped rename inode op ceph: allow idmapped getattr inode op ceph: allow idmapped permission inode op ceph: allow idmapped setattr inode op ceph/acl: allow idmapped set_acl inode op ceph/file: allow idmapped atomic_open inode op ceph: allow idmapped mounts fs/ceph/acl.c | 6 +++--- fs/ceph/dir.c | 4 ++++ fs/ceph/file.c | 10 ++++++++-- fs/ceph/inode.c | 29 +++++++++++++++++------------ fs/ceph/mds_client.c | 27 +++++++++++++++++++++++---- fs/ceph/mds_client.h | 1 + fs/ceph/super.c | 2 +- fs/ceph/super.h | 3 ++- fs/mnt_idmapping.c | 2 ++ include/linux/mnt_idmapping.h | 3 +++ 10 files changed, 64 insertions(+), 23 deletions(-)