Commit ddee5cdb70e6f87de2fc696b87bd7bd184a51eb8

Authored by Tristan Ye
Committed by Joel Becker
1 parent 152831be91

Ocfs2: Add new OCFS2_IOC_INFO ioctl for ocfs2 v8.

The reason why we need this ioctl is to offer the none-privileged
end-user a possibility to get filesys info gathering.

We use OCFS2_IOC_INFO to manipulate the new ioctl, userspace passes a
structure to kernel containing an array of request pointers and request
count, such as,

* From userspace:

struct ocfs2_info_blocksize oib = {
        .ib_req = {
                .ir_magic = OCFS2_INFO_MAGIC,
                .ir_code = OCFS2_INFO_BLOCKSIZE,
                ...
        }
        ...
}

struct ocfs2_info_clustersize oic = {
        ...
}

uint64_t reqs[2] = {(unsigned long)&oib,
                    (unsigned long)&oic};

struct ocfs2_info info = {
        .oi_requests = reqs,
        .oi_count = 2,
}

ret = ioctl(fd, OCFS2_IOC_INFO, &info);

* In kernel:

Get the request pointers from *info*, then handle each request one bye one.

Idea here is to make the spearated request small enough to guarantee
a better backward&forward compatibility since a small piece of request
would be less likely to be broken if filesys on raw disk get changed.

Currently, the following 7 requests are supported per the requirement from
userspace tool o2info, and I believe it will grow over time:-)

        OCFS2_INFO_CLUSTERSIZE
        OCFS2_INFO_BLOCKSIZE
        OCFS2_INFO_MAXSLOTS
        OCFS2_INFO_LABEL
        OCFS2_INFO_UUID
        OCFS2_INFO_FS_FEATURES
        OCFS2_INFO_JOURNAL_SIZE

This ioctl is only specific to OCFS2.

Signed-off-by: Tristan Ye <tristan.ye@oracle.com>
Signed-off-by: Joel Becker <joel.becker@oracle.com>

Showing 2 changed files with 451 additions and 0 deletions Side-by-side Diff

... ... @@ -26,6 +26,26 @@
26 26  
27 27 #include <linux/ext2_fs.h>
28 28  
  29 +#define o2info_from_user(a, b) \
  30 + copy_from_user(&(a), (b), sizeof(a))
  31 +#define o2info_to_user(a, b) \
  32 + copy_to_user((typeof(a) __user *)b, &(a), sizeof(a))
  33 +
  34 +/*
  35 + * This call is void because we are already reporting an error that may
  36 + * be -EFAULT. The error will be returned from the ioctl(2) call. It's
  37 + * just a best-effort to tell userspace that this request caused the error.
  38 + */
  39 +static inline void __o2info_set_request_error(struct ocfs2_info_request *kreq,
  40 + struct ocfs2_info_request __user *req)
  41 +{
  42 + kreq->ir_flags |= OCFS2_INFO_FL_ERROR;
  43 + (void)put_user(kreq->ir_flags, (__u32 __user *)&(req->ir_flags));
  44 +}
  45 +
  46 +#define o2info_set_request_error(a, b) \
  47 + __o2info_set_request_error((struct ocfs2_info_request *)&(a), b)
  48 +
29 49 static int ocfs2_get_inode_attr(struct inode *inode, unsigned *flags)
30 50 {
31 51 int status;
... ... @@ -109,6 +129,328 @@
109 129 return status;
110 130 }
111 131  
  132 +int ocfs2_info_handle_blocksize(struct inode *inode,
  133 + struct ocfs2_info_request __user *req)
  134 +{
  135 + int status = -EFAULT;
  136 + struct ocfs2_info_blocksize oib;
  137 +
  138 + if (o2info_from_user(oib, req))
  139 + goto bail;
  140 +
  141 + oib.ib_blocksize = inode->i_sb->s_blocksize;
  142 + oib.ib_req.ir_flags |= OCFS2_INFO_FL_FILLED;
  143 +
  144 + if (o2info_to_user(oib, req))
  145 + goto bail;
  146 +
  147 + status = 0;
  148 +bail:
  149 + if (status)
  150 + o2info_set_request_error(oib, req);
  151 +
  152 + return status;
  153 +}
  154 +
  155 +int ocfs2_info_handle_clustersize(struct inode *inode,
  156 + struct ocfs2_info_request __user *req)
  157 +{
  158 + int status = -EFAULT;
  159 + struct ocfs2_info_clustersize oic;
  160 + struct ocfs2_super *osb = OCFS2_SB(inode->i_sb);
  161 +
  162 + if (o2info_from_user(oic, req))
  163 + goto bail;
  164 +
  165 + oic.ic_clustersize = osb->s_clustersize;
  166 + oic.ic_req.ir_flags |= OCFS2_INFO_FL_FILLED;
  167 +
  168 + if (o2info_to_user(oic, req))
  169 + goto bail;
  170 +
  171 + status = 0;
  172 +bail:
  173 + if (status)
  174 + o2info_set_request_error(oic, req);
  175 +
  176 + return status;
  177 +}
  178 +
  179 +int ocfs2_info_handle_maxslots(struct inode *inode,
  180 + struct ocfs2_info_request __user *req)
  181 +{
  182 + int status = -EFAULT;
  183 + struct ocfs2_info_maxslots oim;
  184 + struct ocfs2_super *osb = OCFS2_SB(inode->i_sb);
  185 +
  186 + if (o2info_from_user(oim, req))
  187 + goto bail;
  188 +
  189 + oim.im_max_slots = osb->max_slots;
  190 + oim.im_req.ir_flags |= OCFS2_INFO_FL_FILLED;
  191 +
  192 + if (o2info_to_user(oim, req))
  193 + goto bail;
  194 +
  195 + status = 0;
  196 +bail:
  197 + if (status)
  198 + o2info_set_request_error(oim, req);
  199 +
  200 + return status;
  201 +}
  202 +
  203 +int ocfs2_info_handle_label(struct inode *inode,
  204 + struct ocfs2_info_request __user *req)
  205 +{
  206 + int status = -EFAULT;
  207 + struct ocfs2_info_label oil;
  208 + struct ocfs2_super *osb = OCFS2_SB(inode->i_sb);
  209 +
  210 + if (o2info_from_user(oil, req))
  211 + goto bail;
  212 +
  213 + memcpy(oil.il_label, osb->vol_label, OCFS2_MAX_VOL_LABEL_LEN);
  214 + oil.il_req.ir_flags |= OCFS2_INFO_FL_FILLED;
  215 +
  216 + if (o2info_to_user(oil, req))
  217 + goto bail;
  218 +
  219 + status = 0;
  220 +bail:
  221 + if (status)
  222 + o2info_set_request_error(oil, req);
  223 +
  224 + return status;
  225 +}
  226 +
  227 +int ocfs2_info_handle_uuid(struct inode *inode,
  228 + struct ocfs2_info_request __user *req)
  229 +{
  230 + int status = -EFAULT;
  231 + struct ocfs2_info_uuid oiu;
  232 + struct ocfs2_super *osb = OCFS2_SB(inode->i_sb);
  233 +
  234 + if (o2info_from_user(oiu, req))
  235 + goto bail;
  236 +
  237 + memcpy(oiu.iu_uuid_str, osb->uuid_str, OCFS2_TEXT_UUID_LEN + 1);
  238 + oiu.iu_req.ir_flags |= OCFS2_INFO_FL_FILLED;
  239 +
  240 + if (o2info_to_user(oiu, req))
  241 + goto bail;
  242 +
  243 + status = 0;
  244 +bail:
  245 + if (status)
  246 + o2info_set_request_error(oiu, req);
  247 +
  248 + return status;
  249 +}
  250 +
  251 +int ocfs2_info_handle_fs_features(struct inode *inode,
  252 + struct ocfs2_info_request __user *req)
  253 +{
  254 + int status = -EFAULT;
  255 + struct ocfs2_info_fs_features oif;
  256 + struct ocfs2_super *osb = OCFS2_SB(inode->i_sb);
  257 +
  258 + if (o2info_from_user(oif, req))
  259 + goto bail;
  260 +
  261 + oif.if_compat_features = osb->s_feature_compat;
  262 + oif.if_incompat_features = osb->s_feature_incompat;
  263 + oif.if_ro_compat_features = osb->s_feature_ro_compat;
  264 + oif.if_req.ir_flags |= OCFS2_INFO_FL_FILLED;
  265 +
  266 + if (o2info_to_user(oif, req))
  267 + goto bail;
  268 +
  269 + status = 0;
  270 +bail:
  271 + if (status)
  272 + o2info_set_request_error(oif, req);
  273 +
  274 + return status;
  275 +}
  276 +
  277 +int ocfs2_info_handle_journal_size(struct inode *inode,
  278 + struct ocfs2_info_request __user *req)
  279 +{
  280 + int status = -EFAULT;
  281 + struct ocfs2_info_journal_size oij;
  282 + struct ocfs2_super *osb = OCFS2_SB(inode->i_sb);
  283 +
  284 + if (o2info_from_user(oij, req))
  285 + goto bail;
  286 +
  287 + oij.ij_journal_size = osb->journal->j_inode->i_size;
  288 +
  289 + oij.ij_req.ir_flags |= OCFS2_INFO_FL_FILLED;
  290 +
  291 + if (o2info_to_user(oij, req))
  292 + goto bail;
  293 +
  294 + status = 0;
  295 +bail:
  296 + if (status)
  297 + o2info_set_request_error(oij, req);
  298 +
  299 + return status;
  300 +}
  301 +
  302 +int ocfs2_info_handle_unknown(struct inode *inode,
  303 + struct ocfs2_info_request __user *req)
  304 +{
  305 + int status = -EFAULT;
  306 + struct ocfs2_info_request oir;
  307 +
  308 + if (o2info_from_user(oir, req))
  309 + goto bail;
  310 +
  311 + oir.ir_flags &= ~OCFS2_INFO_FL_FILLED;
  312 +
  313 + if (o2info_to_user(oir, req))
  314 + goto bail;
  315 +
  316 + status = 0;
  317 +bail:
  318 + if (status)
  319 + o2info_set_request_error(oir, req);
  320 +
  321 + return status;
  322 +}
  323 +
  324 +/*
  325 + * Validate and distinguish OCFS2_IOC_INFO requests.
  326 + *
  327 + * - validate the magic number.
  328 + * - distinguish different requests.
  329 + * - validate size of different requests.
  330 + */
  331 +int ocfs2_info_handle_request(struct inode *inode,
  332 + struct ocfs2_info_request __user *req)
  333 +{
  334 + int status = -EFAULT;
  335 + struct ocfs2_info_request oir;
  336 +
  337 + if (o2info_from_user(oir, req))
  338 + goto bail;
  339 +
  340 + status = -EINVAL;
  341 + if (oir.ir_magic != OCFS2_INFO_MAGIC)
  342 + goto bail;
  343 +
  344 + switch (oir.ir_code) {
  345 + case OCFS2_INFO_BLOCKSIZE:
  346 + if (oir.ir_size == sizeof(struct ocfs2_info_blocksize))
  347 + status = ocfs2_info_handle_blocksize(inode, req);
  348 + break;
  349 + case OCFS2_INFO_CLUSTERSIZE:
  350 + if (oir.ir_size == sizeof(struct ocfs2_info_clustersize))
  351 + status = ocfs2_info_handle_clustersize(inode, req);
  352 + break;
  353 + case OCFS2_INFO_MAXSLOTS:
  354 + if (oir.ir_size == sizeof(struct ocfs2_info_maxslots))
  355 + status = ocfs2_info_handle_maxslots(inode, req);
  356 + break;
  357 + case OCFS2_INFO_LABEL:
  358 + if (oir.ir_size == sizeof(struct ocfs2_info_label))
  359 + status = ocfs2_info_handle_label(inode, req);
  360 + break;
  361 + case OCFS2_INFO_UUID:
  362 + if (oir.ir_size == sizeof(struct ocfs2_info_uuid))
  363 + status = ocfs2_info_handle_uuid(inode, req);
  364 + break;
  365 + case OCFS2_INFO_FS_FEATURES:
  366 + if (oir.ir_size == sizeof(struct ocfs2_info_fs_features))
  367 + status = ocfs2_info_handle_fs_features(inode, req);
  368 + break;
  369 + case OCFS2_INFO_JOURNAL_SIZE:
  370 + if (oir.ir_size == sizeof(struct ocfs2_info_journal_size))
  371 + status = ocfs2_info_handle_journal_size(inode, req);
  372 + break;
  373 + default:
  374 + status = ocfs2_info_handle_unknown(inode, req);
  375 + break;
  376 + }
  377 +
  378 +bail:
  379 + return status;
  380 +}
  381 +
  382 +int ocfs2_get_request_ptr(struct ocfs2_info *info, int idx,
  383 + u64 *req_addr, int compat_flag)
  384 +{
  385 + int status = -EFAULT;
  386 + u64 __user *bp = NULL;
  387 +
  388 + if (compat_flag) {
  389 +#ifdef CONFIG_COMPAT
  390 + /*
  391 + * pointer bp stores the base address of a pointers array,
  392 + * which collects all addresses of separate request.
  393 + */
  394 + bp = (u64 __user *)(unsigned long)compat_ptr(info->oi_requests);
  395 +#else
  396 + BUG();
  397 +#endif
  398 + } else
  399 + bp = (u64 __user *)(unsigned long)(info->oi_requests);
  400 +
  401 + if (o2info_from_user(*req_addr, bp + idx))
  402 + goto bail;
  403 +
  404 + status = 0;
  405 +bail:
  406 + return status;
  407 +}
  408 +
  409 +/*
  410 + * OCFS2_IOC_INFO handles an array of requests passed from userspace.
  411 + *
  412 + * ocfs2_info_handle() recevies a large info aggregation, grab and
  413 + * validate the request count from header, then break it into small
  414 + * pieces, later specific handlers can handle them one by one.
  415 + *
  416 + * Idea here is to make each separate request small enough to ensure
  417 + * a better backward&forward compatibility, since a small piece of
  418 + * request will be less likely to be broken if disk layout get changed.
  419 + */
  420 +int ocfs2_info_handle(struct inode *inode, struct ocfs2_info *info,
  421 + int compat_flag)
  422 +{
  423 + int i, status = 0;
  424 + u64 req_addr;
  425 + struct ocfs2_info_request __user *reqp;
  426 +
  427 + if ((info->oi_count > OCFS2_INFO_MAX_REQUEST) ||
  428 + (!info->oi_requests)) {
  429 + status = -EINVAL;
  430 + goto bail;
  431 + }
  432 +
  433 + for (i = 0; i < info->oi_count; i++) {
  434 +
  435 + status = ocfs2_get_request_ptr(info, i, &req_addr, compat_flag);
  436 + if (status)
  437 + break;
  438 +
  439 + reqp = (struct ocfs2_info_request *)(unsigned long)req_addr;
  440 + if (!reqp) {
  441 + status = -EINVAL;
  442 + goto bail;
  443 + }
  444 +
  445 + status = ocfs2_info_handle_request(inode, reqp);
  446 + if (status)
  447 + break;
  448 + }
  449 +
  450 +bail:
  451 + return status;
  452 +}
  453 +
112 454 long ocfs2_ioctl(struct file *filp, unsigned int cmd, unsigned long arg)
113 455 {
114 456 struct inode *inode = filp->f_path.dentry->d_inode;
... ... @@ -120,6 +462,7 @@
120 462 struct reflink_arguments args;
121 463 const char *old_path, *new_path;
122 464 bool preserve;
  465 + struct ocfs2_info info;
123 466  
124 467 switch (cmd) {
125 468 case OCFS2_IOC_GETFLAGS:
... ... @@ -174,6 +517,12 @@
174 517 preserve = (args.preserve != 0);
175 518  
176 519 return ocfs2_reflink_ioctl(inode, old_path, new_path, preserve);
  520 + case OCFS2_IOC_INFO:
  521 + if (copy_from_user(&info, (struct ocfs2_info __user *)arg,
  522 + sizeof(struct ocfs2_info)))
  523 + return -EFAULT;
  524 +
  525 + return ocfs2_info_handle(inode, &info, 0);
177 526 default:
178 527 return -ENOTTY;
179 528 }
... ... @@ -185,6 +534,7 @@
185 534 bool preserve;
186 535 struct reflink_arguments args;
187 536 struct inode *inode = file->f_path.dentry->d_inode;
  537 + struct ocfs2_info info;
188 538  
189 539 switch (cmd) {
190 540 case OCFS2_IOC32_GETFLAGS:
... ... @@ -209,6 +559,12 @@
209 559  
210 560 return ocfs2_reflink_ioctl(inode, compat_ptr(args.old_path),
211 561 compat_ptr(args.new_path), preserve);
  562 + case OCFS2_IOC_INFO:
  563 + if (copy_from_user(&info, (struct ocfs2_info __user *)arg,
  564 + sizeof(struct ocfs2_info)))
  565 + return -EFAULT;
  566 +
  567 + return ocfs2_info_handle(inode, &info, 1);
212 568 default:
213 569 return -ENOIOCTLCMD;
214 570 }
fs/ocfs2/ocfs2_ioctl.h
... ... @@ -76,5 +76,100 @@
76 76 };
77 77 #define OCFS2_IOC_REFLINK _IOW('o', 4, struct reflink_arguments)
78 78  
  79 +/* Following definitions dedicated for ocfs2_info_request ioctls. */
  80 +#define OCFS2_INFO_MAX_REQUEST (50)
  81 +#define OCFS2_TEXT_UUID_LEN (OCFS2_VOL_UUID_LEN * 2)
  82 +
  83 +/* Magic number of all requests */
  84 +#define OCFS2_INFO_MAGIC (0x4F32494E)
  85 +
  86 +/*
  87 + * Always try to separate info request into small pieces to
  88 + * guarantee the backward&forward compatibility.
  89 + */
  90 +struct ocfs2_info {
  91 + __u64 oi_requests; /* Array of __u64 pointers to requests */
  92 + __u32 oi_count; /* Number of requests in info_requests */
  93 + __u32 oi_pad;
  94 +};
  95 +
  96 +struct ocfs2_info_request {
  97 +/*00*/ __u32 ir_magic; /* Magic number */
  98 + __u32 ir_code; /* Info request code */
  99 + __u32 ir_size; /* Size of request */
  100 + __u32 ir_flags; /* Request flags */
  101 +/*10*/ /* Request specific fields */
  102 +};
  103 +
  104 +struct ocfs2_info_clustersize {
  105 + struct ocfs2_info_request ic_req;
  106 + __u32 ic_clustersize;
  107 + __u32 ic_pad;
  108 +};
  109 +
  110 +struct ocfs2_info_blocksize {
  111 + struct ocfs2_info_request ib_req;
  112 + __u32 ib_blocksize;
  113 + __u32 ib_pad;
  114 +};
  115 +
  116 +struct ocfs2_info_maxslots {
  117 + struct ocfs2_info_request im_req;
  118 + __u32 im_max_slots;
  119 + __u32 im_pad;
  120 +};
  121 +
  122 +struct ocfs2_info_label {
  123 + struct ocfs2_info_request il_req;
  124 + __u8 il_label[OCFS2_MAX_VOL_LABEL_LEN];
  125 +} __attribute__ ((packed));
  126 +
  127 +struct ocfs2_info_uuid {
  128 + struct ocfs2_info_request iu_req;
  129 + __u8 iu_uuid_str[OCFS2_TEXT_UUID_LEN + 1];
  130 +} __attribute__ ((packed));
  131 +
  132 +struct ocfs2_info_fs_features {
  133 + struct ocfs2_info_request if_req;
  134 + __u32 if_compat_features;
  135 + __u32 if_incompat_features;
  136 + __u32 if_ro_compat_features;
  137 + __u32 if_pad;
  138 +};
  139 +
  140 +struct ocfs2_info_journal_size {
  141 + struct ocfs2_info_request ij_req;
  142 + __u64 ij_journal_size;
  143 +};
  144 +
  145 +/* Codes for ocfs2_info_request */
  146 +enum ocfs2_info_type {
  147 + OCFS2_INFO_CLUSTERSIZE = 1,
  148 + OCFS2_INFO_BLOCKSIZE,
  149 + OCFS2_INFO_MAXSLOTS,
  150 + OCFS2_INFO_LABEL,
  151 + OCFS2_INFO_UUID,
  152 + OCFS2_INFO_FS_FEATURES,
  153 + OCFS2_INFO_JOURNAL_SIZE,
  154 + OCFS2_INFO_NUM_TYPES
  155 +};
  156 +
  157 +/* Flags for struct ocfs2_info_request */
  158 +/* Filled by the caller */
  159 +#define OCFS2_INFO_FL_NON_COHERENT (0x00000001) /* Cluster coherency not
  160 + required. This is a hint.
  161 + It is up to ocfs2 whether
  162 + the request can be fulfilled
  163 + without locking. */
  164 +/* Filled by ocfs2 */
  165 +#define OCFS2_INFO_FL_FILLED (0x40000000) /* Filesystem understood
  166 + this request and
  167 + filled in the answer */
  168 +
  169 +#define OCFS2_INFO_FL_ERROR (0x80000000) /* Error happened during
  170 + request handling. */
  171 +
  172 +#define OCFS2_IOC_INFO _IOR('o', 5, struct ocfs2_info)
  173 +
79 174 #endif /* OCFS2_IOCTL_H */