On Thu, Jan 27, 2022 at 4:53 PM Jeff Layton <jlayton@xxxxxxxxxx> wrote: > > On Thu, 2022-01-27 at 08:26 +0000, Milind Changire wrote: > > Problem: > > Directory vxattrs like ceph.dir.pin* and ceph.dir.layout* may not be > > propagated to the client as frequently to keep them updated. This > > creates vxattr availability problems. > > > > Solution: > > Adds new getvxattr op to fetch ceph.dir.pin*, ceph.dir.layout* and > > ceph.file.layout* vxattrs. > > If the entire layout for a dir or a file is being set, then it is > > expected that the layout be set in standard JSON format. Individual > > field value retrieval is not wrapped in JSON. The JSON format also > > applies while setting the vxattr if the entire layout is being set in > > one go. > > As a temporary measure, setting a vxattr can also be done in the old > > format. The old format will be deprecated in the future. > > > > URL: https://tracker.ceph.com/issues/51062 > > Signed-off-by: Milind Changire <mchangir@xxxxxxxxxx> > > --- > > fs/ceph/inode.c | 51 ++++++++++++++++++++++++++++++++++++ > > fs/ceph/mds_client.c | 27 ++++++++++++++++++- > > fs/ceph/mds_client.h | 12 ++++++++- > > fs/ceph/strings.c | 1 + > > fs/ceph/super.h | 1 + > > fs/ceph/xattr.c | 17 ++++++++++++ > > include/linux/ceph/ceph_fs.h | 1 + > > 7 files changed, 108 insertions(+), 2 deletions(-) > > > > diff --git a/fs/ceph/inode.c b/fs/ceph/inode.c > > index e3322fcb2e8d..efdce049b7f0 100644 > > --- a/fs/ceph/inode.c > > +++ b/fs/ceph/inode.c > > @@ -2291,6 +2291,57 @@ int __ceph_do_getattr(struct inode *inode, struct page *locked_page, > > return err; > > } > > > > +int ceph_do_getvxattr(struct inode *inode, const char *name, void *value, > > + size_t size) > > +{ > > + struct ceph_fs_client *fsc = ceph_sb_to_client(inode->i_sb); > > + struct ceph_mds_client *mdsc = fsc->mdsc; > > + struct ceph_mds_request *req; > > + int mode = USE_AUTH_MDS; > > + int err; > > + char *xattr_value; > > + size_t xattr_value_len; > > + > > + req = ceph_mdsc_create_request(mdsc, CEPH_MDS_OP_GETVXATTR, mode); > > + if (IS_ERR(req)) { > > + err = -ENOMEM; > > + goto out; > > + } > > + > > + req->r_path2 = kstrdup(name, GFP_NOFS); > > + if (!req->r_path2) { > > + err = -ENOMEM; > > + goto put; > > + } > > + > > + ihold(inode); > > + req->r_inode = inode; > > + err = ceph_mdsc_do_request(mdsc, NULL, req); > > + if (err < 0) > > + goto put; > > + > > + xattr_value = req->r_reply_info.xattr_info.xattr_value; > > + xattr_value_len = req->r_reply_info.xattr_info.xattr_value_len; > > + > > + dout("do_getvxattr xattr_value_len:%zu, size:%zu\n", xattr_value_len, size); > > + > > + err = (int)xattr_value_len; > > + if (size == 0) > > + goto put; > > + > > + if (xattr_value_len > size) { > > + err = -ERANGE; > > + goto put; > > + } > > + > > + memcpy(value, xattr_value, xattr_value_len); > > +put: > > + ceph_mdsc_put_request(req); > > +out: > > + dout("do_getvxattr result=%d\n", err); > > + return err; > > +} > > + > > > > /* > > * Check inode permissions. We verify we have a valid value for > > diff --git a/fs/ceph/mds_client.c b/fs/ceph/mds_client.c > > index c30eefc0ac19..a5eafc71d976 100644 > > --- a/fs/ceph/mds_client.c > > +++ b/fs/ceph/mds_client.c > > @@ -555,6 +555,29 @@ static int parse_reply_info_create(void **p, void *end, > > return -EIO; > > } > > > > +static int parse_reply_info_getvxattr(void **p, void *end, > > + struct ceph_mds_reply_info_parsed *info, > > + u64 features) > > +{ > > + u8 struct_v, struct_compat; > > + u32 struct_len; > > + u32 value_len; > > + > > + ceph_decode_8_safe(p, end, struct_v, bad); > > + ceph_decode_8_safe(p, end, struct_compat, bad); > > + ceph_decode_32_safe(p, end, struct_len, bad); > > + ceph_decode_32_safe(p, end, value_len, bad); > > + > > + if (value_len == end - *p) { > > + info->xattr_info.xattr_value = *p; > > + info->xattr_info.xattr_value_len = end - *p; > > + *p = end; > > + return info->xattr_info.xattr_value_len; > > + } > > +bad: > > + return -EIO; > > +} > > + > > /* > > * parse extra results > > */ > > @@ -570,6 +593,8 @@ static int parse_reply_info_extra(void **p, void *end, > > return parse_reply_info_readdir(p, end, info, features); > > else if (op == CEPH_MDS_OP_CREATE) > > return parse_reply_info_create(p, end, info, features, s); > > + else if (op == CEPH_MDS_OP_GETVXATTR) > > + return parse_reply_info_getvxattr(p, end, info, features); > > else > > return -EIO; > > } > > @@ -615,7 +640,7 @@ static int parse_reply_info(struct ceph_mds_session *s, struct ceph_msg *msg, > > > > if (p != end) > > goto bad; > > - return 0; > > + return err; > > > > Why this delta? It looks like this function could now return err > uninitialized with it. delta undone this was something left behind while testing to understand the code > > > bad: > > err = -EIO; > > diff --git a/fs/ceph/mds_client.h b/fs/ceph/mds_client.h > > index 97c7f7bfa55f..f2a8e5af3c2e 100644 > > --- a/fs/ceph/mds_client.h > > +++ b/fs/ceph/mds_client.h > > @@ -29,8 +29,10 @@ enum ceph_feature_type { > > CEPHFS_FEATURE_MULTI_RECONNECT, > > CEPHFS_FEATURE_DELEG_INO, > > CEPHFS_FEATURE_METRIC_COLLECT, > > + CEPHFS_FEATURE_ALTERNATE_NAME, > > + CEPHFS_FEATURE_GETVXATTR, > > > > - CEPHFS_FEATURE_MAX = CEPHFS_FEATURE_METRIC_COLLECT, > > + CEPHFS_FEATURE_MAX = CEPHFS_FEATURE_GETVXATTR, > > }; > > > > /* > > @@ -45,6 +47,8 @@ enum ceph_feature_type { > > CEPHFS_FEATURE_MULTI_RECONNECT, \ > > CEPHFS_FEATURE_DELEG_INO, \ > > CEPHFS_FEATURE_METRIC_COLLECT, \ > > + CEPHFS_FEATURE_ALTERNATE_NAME, \ > > + CEPHFS_FEATURE_GETVXATTR, \ > > \ > > CEPHFS_FEATURE_MAX, \ > > } > > @@ -100,6 +104,11 @@ struct ceph_mds_reply_dir_entry { > > loff_t offset; > > }; > > > > +struct ceph_mds_reply_xattr { > > + char *xattr_value; > > + size_t xattr_value_len; > > +}; > > + > > /* > > * parsed info about an mds reply, including information about > > * either: 1) the target inode and/or its parent directory and dentry, > > @@ -115,6 +124,7 @@ struct ceph_mds_reply_info_parsed { > > char *dname; > > u32 dname_len; > > struct ceph_mds_reply_lease *dlease; > > + struct ceph_mds_reply_xattr xattr_info; > > > > /* extra */ > > union { > > diff --git a/fs/ceph/strings.c b/fs/ceph/strings.c > > index 573bb9556fb5..e36e8948e728 100644 > > --- a/fs/ceph/strings.c > > +++ b/fs/ceph/strings.c > > @@ -60,6 +60,7 @@ const char *ceph_mds_op_name(int op) > > case CEPH_MDS_OP_LOOKUPINO: return "lookupino"; > > case CEPH_MDS_OP_LOOKUPNAME: return "lookupname"; > > case CEPH_MDS_OP_GETATTR: return "getattr"; > > + case CEPH_MDS_OP_GETVXATTR: return "getvxattr"; > > case CEPH_MDS_OP_SETXATTR: return "setxattr"; > > case CEPH_MDS_OP_SETATTR: return "setattr"; > > case CEPH_MDS_OP_RMXATTR: return "rmxattr"; > > diff --git a/fs/ceph/super.h b/fs/ceph/super.h > > index ac331aa07cfa..a627fa69668e 100644 > > --- a/fs/ceph/super.h > > +++ b/fs/ceph/super.h > > @@ -1043,6 +1043,7 @@ static inline bool ceph_inode_is_shutdown(struct inode *inode) > > > > /* xattr.c */ > > int __ceph_setxattr(struct inode *, const char *, const void *, size_t, int); > > +int ceph_do_getvxattr(struct inode *inode, const char *name, void *value, size_t size); > > ssize_t __ceph_getxattr(struct inode *, const char *, void *, size_t); > > extern ssize_t ceph_listxattr(struct dentry *, char *, size_t); > > extern struct ceph_buffer *__ceph_build_xattrs_blob(struct ceph_inode_info *ci); > > diff --git a/fs/ceph/xattr.c b/fs/ceph/xattr.c > > index fcf7dfdecf96..9a4fbe48963f 100644 > > --- a/fs/ceph/xattr.c > > +++ b/fs/ceph/xattr.c > > @@ -924,6 +924,7 @@ ssize_t __ceph_getxattr(struct inode *inode, const char *name, void *value, > > struct ceph_inode_info *ci = ceph_inode(inode); > > struct ceph_inode_xattr *xattr; > > struct ceph_vxattr *vxattr = NULL; > > + struct ceph_mds_session *session = NULL; > > int req_mask; > > ssize_t err; > > > > @@ -945,6 +946,22 @@ ssize_t __ceph_getxattr(struct inode *inode, const char *name, void *value, > > err = -ERANGE; > > } > > return err; > > + } else { > > + err = -ENODATA; > > + spin_lock(&ci->i_ceph_lock); > > + if (strncmp(name, XATTR_CEPH_PREFIX, XATTR_CEPH_PREFIX_LEN)) > > + goto out; > > This looks much better. One thing I'd probably change is to make this do > this strncmp first, before calling ceph_match_vxattr. That way we can > avoid walking the vxattr list unless the thing starts with "ceph.". done ... check v7 patch > > > + /* check if the auth mds supports the getvxattr feature */ > > + session = ci->i_auth_cap->session; > > > i_auth_cap can be NULL, so this may end up being a NULL pointer > dereference. You probably also need to be holding the i_ceph_lock when > you check it, since that's what protects it. > > But, beyond that, the logic here is problematic. This is going to just > fail with -ENODATA if we don't happen to have auth caps. That seems > wrong. I think if you don't have auth caps, then you should probably > walk the sessions array and see if at least one MDS supports GETVXATTR. > If it does, then go ahead and do the call. done ... check v7 patch > > > + if (!session) > > + goto out; > > + > > + if (test_bit(CEPHFS_FEATURE_GETVXATTR, &session->s_features)) { > > + spin_unlock(&ci->i_ceph_lock); > > + err = ceph_do_getvxattr(inode, name, value, size); > > + spin_lock(&ci->i_ceph_lock); > > + } > > + goto out; > > } > > > > req_mask = __get_request_mask(inode); > > diff --git a/include/linux/ceph/ceph_fs.h b/include/linux/ceph/ceph_fs.h > > index 7ad6c3d0db7d..66db21ac5f0c 100644 > > --- a/include/linux/ceph/ceph_fs.h > > +++ b/include/linux/ceph/ceph_fs.h > > @@ -328,6 +328,7 @@ enum { > > CEPH_MDS_OP_LOOKUPPARENT = 0x00103, > > CEPH_MDS_OP_LOOKUPINO = 0x00104, > > CEPH_MDS_OP_LOOKUPNAME = 0x00105, > > + CEPH_MDS_OP_GETVXATTR = 0x00106, > > > > CEPH_MDS_OP_SETXATTR = 0x01105, > > CEPH_MDS_OP_RMXATTR = 0x01106, > > -- > Jeff Layton <jlayton@xxxxxxxxxx>
