1 #include <linux/in.h> 2 3 #include "super.h" 4 #include "mds_client.h" 5 #include <linux/ceph/ceph_debug.h> 6 7 #include "ioctl.h" 8 9 10 /* 11 * ioctls 12 */ 13 14 /* 15 * get and set the file layout 16 */ 17 static long ceph_ioctl_get_layout(struct file *file, void __user *arg) 18 { 19 struct ceph_inode_info *ci = ceph_inode(file->f_dentry->d_inode); 20 struct ceph_ioctl_layout l; 21 int err; 22 23 err = ceph_do_getattr(file->f_dentry->d_inode, CEPH_STAT_CAP_LAYOUT); 24 if (!err) { 25 l.stripe_unit = ceph_file_layout_su(ci->i_layout); 26 l.stripe_count = ceph_file_layout_stripe_count(ci->i_layout); 27 l.object_size = ceph_file_layout_object_size(ci->i_layout); 28 l.data_pool = le32_to_cpu(ci->i_layout.fl_pg_pool); 29 l.preferred_osd = (s32)-1; 30 if (copy_to_user(arg, &l, sizeof(l))) 31 return -EFAULT; 32 } 33 34 return err; 35 } 36 37 static long __validate_layout(struct ceph_mds_client *mdsc, 38 struct ceph_ioctl_layout *l) 39 { 40 int i, err; 41 42 /* validate striping parameters */ 43 if ((l->object_size & ~PAGE_MASK) || 44 (l->stripe_unit & ~PAGE_MASK) || 45 (l->stripe_unit != 0 && 46 ((unsigned)l->object_size % (unsigned)l->stripe_unit))) 47 return -EINVAL; 48 49 /* make sure it's a valid data pool */ 50 mutex_lock(&mdsc->mutex); 51 err = -EINVAL; 52 for (i = 0; i < mdsc->mdsmap->m_num_data_pg_pools; i++) 53 if (mdsc->mdsmap->m_data_pg_pools[i] == l->data_pool) { 54 err = 0; 55 break; 56 } 57 mutex_unlock(&mdsc->mutex); 58 if (err) 59 return err; 60 61 return 0; 62 } 63 64 static long ceph_ioctl_set_layout(struct file *file, void __user *arg) 65 { 66 struct inode *inode = file->f_dentry->d_inode; 67 struct inode *parent_inode; 68 struct ceph_mds_client *mdsc = ceph_sb_to_client(inode->i_sb)->mdsc; 69 struct ceph_mds_request *req; 70 struct ceph_ioctl_layout l; 71 struct ceph_inode_info *ci = ceph_inode(file->f_dentry->d_inode); 72 struct ceph_ioctl_layout nl; 73 int err; 74 75 if (copy_from_user(&l, arg, sizeof(l))) 76 return -EFAULT; 77 78 /* validate changed params against current layout */ 79 err = ceph_do_getattr(file->f_dentry->d_inode, CEPH_STAT_CAP_LAYOUT); 80 if (err) 81 return err; 82 83 memset(&nl, 0, sizeof(nl)); 84 if (l.stripe_count) 85 nl.stripe_count = l.stripe_count; 86 else 87 nl.stripe_count = ceph_file_layout_stripe_count(ci->i_layout); 88 if (l.stripe_unit) 89 nl.stripe_unit = l.stripe_unit; 90 else 91 nl.stripe_unit = ceph_file_layout_su(ci->i_layout); 92 if (l.object_size) 93 nl.object_size = l.object_size; 94 else 95 nl.object_size = ceph_file_layout_object_size(ci->i_layout); 96 if (l.data_pool) 97 nl.data_pool = l.data_pool; 98 else 99 nl.data_pool = ceph_file_layout_pg_pool(ci->i_layout); 100 101 /* this is obsolete, and always -1 */ 102 nl.preferred_osd = le64_to_cpu(-1); 103 104 err = __validate_layout(mdsc, &nl); 105 if (err) 106 return err; 107 108 req = ceph_mdsc_create_request(mdsc, CEPH_MDS_OP_SETLAYOUT, 109 USE_AUTH_MDS); 110 if (IS_ERR(req)) 111 return PTR_ERR(req); 112 req->r_inode = inode; 113 ihold(inode); 114 req->r_inode_drop = CEPH_CAP_FILE_SHARED | CEPH_CAP_FILE_EXCL; 115 116 req->r_args.setlayout.layout.fl_stripe_unit = 117 cpu_to_le32(l.stripe_unit); 118 req->r_args.setlayout.layout.fl_stripe_count = 119 cpu_to_le32(l.stripe_count); 120 req->r_args.setlayout.layout.fl_object_size = 121 cpu_to_le32(l.object_size); 122 req->r_args.setlayout.layout.fl_pg_pool = cpu_to_le32(l.data_pool); 123 124 parent_inode = ceph_get_dentry_parent_inode(file->f_dentry); 125 err = ceph_mdsc_do_request(mdsc, parent_inode, req); 126 iput(parent_inode); 127 ceph_mdsc_put_request(req); 128 return err; 129 } 130 131 /* 132 * Set a layout policy on a directory inode. All items in the tree 133 * rooted at this inode will inherit this layout on creation, 134 * (It doesn't apply retroactively ) 135 * unless a subdirectory has its own layout policy. 136 */ 137 static long ceph_ioctl_set_layout_policy (struct file *file, void __user *arg) 138 { 139 struct inode *inode = file->f_dentry->d_inode; 140 struct ceph_mds_request *req; 141 struct ceph_ioctl_layout l; 142 int err; 143 struct ceph_mds_client *mdsc = ceph_sb_to_client(inode->i_sb)->mdsc; 144 145 /* copy and validate */ 146 if (copy_from_user(&l, arg, sizeof(l))) 147 return -EFAULT; 148 149 err = __validate_layout(mdsc, &l); 150 if (err) 151 return err; 152 153 req = ceph_mdsc_create_request(mdsc, CEPH_MDS_OP_SETDIRLAYOUT, 154 USE_AUTH_MDS); 155 156 if (IS_ERR(req)) 157 return PTR_ERR(req); 158 req->r_inode = inode; 159 ihold(inode); 160 161 req->r_args.setlayout.layout.fl_stripe_unit = 162 cpu_to_le32(l.stripe_unit); 163 req->r_args.setlayout.layout.fl_stripe_count = 164 cpu_to_le32(l.stripe_count); 165 req->r_args.setlayout.layout.fl_object_size = 166 cpu_to_le32(l.object_size); 167 req->r_args.setlayout.layout.fl_pg_pool = 168 cpu_to_le32(l.data_pool); 169 170 err = ceph_mdsc_do_request(mdsc, inode, req); 171 ceph_mdsc_put_request(req); 172 return err; 173 } 174 175 /* 176 * Return object name, size/offset information, and location (OSD 177 * number, network address) for a given file offset. 178 */ 179 static long ceph_ioctl_get_dataloc(struct file *file, void __user *arg) 180 { 181 struct ceph_ioctl_dataloc dl; 182 struct inode *inode = file->f_dentry->d_inode; 183 struct ceph_inode_info *ci = ceph_inode(inode); 184 struct ceph_osd_client *osdc = 185 &ceph_sb_to_client(inode->i_sb)->client->osdc; 186 u64 len = 1, olen; 187 u64 tmp; 188 struct ceph_object_layout ol; 189 struct ceph_pg pgid; 190 191 /* copy and validate */ 192 if (copy_from_user(&dl, arg, sizeof(dl))) 193 return -EFAULT; 194 195 down_read(&osdc->map_sem); 196 ceph_calc_file_object_mapping(&ci->i_layout, dl.file_offset, &len, 197 &dl.object_no, &dl.object_offset, &olen); 198 dl.file_offset -= dl.object_offset; 199 dl.object_size = ceph_file_layout_object_size(ci->i_layout); 200 dl.block_size = ceph_file_layout_su(ci->i_layout); 201 202 /* block_offset = object_offset % block_size */ 203 tmp = dl.object_offset; 204 dl.block_offset = do_div(tmp, dl.block_size); 205 206 snprintf(dl.object_name, sizeof(dl.object_name), "%llx.%08llx", 207 ceph_ino(inode), dl.object_no); 208 ceph_calc_object_layout(&ol, dl.object_name, &ci->i_layout, 209 osdc->osdmap); 210 211 pgid = ol.ol_pgid; 212 dl.osd = ceph_calc_pg_primary(osdc->osdmap, pgid); 213 if (dl.osd >= 0) { 214 struct ceph_entity_addr *a = 215 ceph_osd_addr(osdc->osdmap, dl.osd); 216 if (a) 217 memcpy(&dl.osd_addr, &a->in_addr, sizeof(dl.osd_addr)); 218 } else { 219 memset(&dl.osd_addr, 0, sizeof(dl.osd_addr)); 220 } 221 up_read(&osdc->map_sem); 222 223 /* send result back to user */ 224 if (copy_to_user(arg, &dl, sizeof(dl))) 225 return -EFAULT; 226 227 return 0; 228 } 229 230 static long ceph_ioctl_lazyio(struct file *file) 231 { 232 struct ceph_file_info *fi = file->private_data; 233 struct inode *inode = file->f_dentry->d_inode; 234 struct ceph_inode_info *ci = ceph_inode(inode); 235 236 if ((fi->fmode & CEPH_FILE_MODE_LAZY) == 0) { 237 spin_lock(&ci->i_ceph_lock); 238 ci->i_nr_by_mode[fi->fmode]--; 239 fi->fmode |= CEPH_FILE_MODE_LAZY; 240 ci->i_nr_by_mode[fi->fmode]++; 241 spin_unlock(&ci->i_ceph_lock); 242 dout("ioctl_layzio: file %p marked lazy\n", file); 243 244 ceph_check_caps(ci, 0, NULL); 245 } else { 246 dout("ioctl_layzio: file %p already lazy\n", file); 247 } 248 return 0; 249 } 250 251 static long ceph_ioctl_syncio(struct file *file) 252 { 253 struct ceph_file_info *fi = file->private_data; 254 255 fi->flags |= CEPH_F_SYNC; 256 return 0; 257 } 258 259 long ceph_ioctl(struct file *file, unsigned int cmd, unsigned long arg) 260 { 261 dout("ioctl file %p cmd %u arg %lu\n", file, cmd, arg); 262 switch (cmd) { 263 case CEPH_IOC_GET_LAYOUT: 264 return ceph_ioctl_get_layout(file, (void __user *)arg); 265 266 case CEPH_IOC_SET_LAYOUT: 267 return ceph_ioctl_set_layout(file, (void __user *)arg); 268 269 case CEPH_IOC_SET_LAYOUT_POLICY: 270 return ceph_ioctl_set_layout_policy(file, (void __user *)arg); 271 272 case CEPH_IOC_GET_DATALOC: 273 return ceph_ioctl_get_dataloc(file, (void __user *)arg); 274 275 case CEPH_IOC_LAZYIO: 276 return ceph_ioctl_lazyio(file); 277 278 case CEPH_IOC_SYNCIO: 279 return ceph_ioctl_syncio(file); 280 } 281 282 return -ENOTTY; 283 } 284