提交 db2196a5 编写于 作者: I Ilya Dryomov

libceph: eliminate overflows in ceph_calc_file_object_mapping()

bl, stripeno and objsetno should be u64 -- otherwise large enough files
get corrupted.  How large depends on file layout:

- 4M-objects layout (default): any file over 16P
- 64K-objects layout (smallest possible object size): any file over 512T

Only CephFS is affected, rbd doesn't use ceph_calc_file_object_mapping()
yet.  Fortunately, CephFS has a max_file_size configurable, the default
for which is way below both of the above numbers.

Reimplement the logic from scratch with no layout validation -- it's
done on the MDS side.
Signed-off-by: NIlya Dryomov <idryomov@gmail.com>
Reviewed-by: NAlex Elder <elder@linaro.org>
上级 24f1df60
...@@ -2141,72 +2141,40 @@ bool ceph_osds_changed(const struct ceph_osds *old_acting, ...@@ -2141,72 +2141,40 @@ bool ceph_osds_changed(const struct ceph_osds *old_acting,
} }
/* /*
* calculate file layout from given offset, length. * Map a file extent to a stripe unit within an object.
* fill in correct oid, logical length, and object extent * Fill in objno, offset into object, and object extent length (i.e. the
* offset, length. * number of bytes mapped, less than or equal to @l->stripe_unit).
* *
* for now, we write only a single su, until we can * Example for stripe_count = 3, stripes_per_object = 4:
* pass a stride back to the caller. *
* blockno | 0 3 6 9 | 1 4 7 10 | 2 5 8 11 | 12 15 18 21 | 13 16 19
* stripeno | 0 1 2 3 | 0 1 2 3 | 0 1 2 3 | 4 5 6 7 | 4 5 6
* stripepos | 0 | 1 | 2 | 0 | 1
* objno | 0 | 1 | 2 | 3 | 4
* objsetno | 0 | 1
*/ */
int ceph_calc_file_object_mapping(struct ceph_file_layout *layout, int ceph_calc_file_object_mapping(struct ceph_file_layout *l,
u64 off, u64 len, u64 off, u64 len,
u64 *ono, u64 *objno, u64 *objoff, u64 *xlen)
u64 *oxoff, u64 *oxlen) {
{ u32 stripes_per_object = l->object_size / l->stripe_unit;
u32 osize = layout->object_size; u64 blockno; /* which su in the file (i.e. globally) */
u32 su = layout->stripe_unit; u32 blockoff; /* offset into su */
u32 sc = layout->stripe_count; u64 stripeno; /* which stripe */
u32 bl, stripeno, stripepos, objsetno; u32 stripepos; /* which su in the stripe,
u32 su_per_object; which object in the object set */
u64 t, su_offset; u64 objsetno; /* which object set */
u32 objsetpos; /* which stripe in the object set */
dout("mapping %llu~%llu osize %u fl_su %u\n", off, len,
osize, su);
if (su == 0 || sc == 0)
goto invalid;
su_per_object = osize / su;
if (su_per_object == 0)
goto invalid;
dout("osize %u / su %u = su_per_object %u\n", osize, su,
su_per_object);
if ((su & ~PAGE_MASK) != 0)
goto invalid;
/* bl = *off / su; */
t = off;
do_div(t, su);
bl = t;
dout("off %llu / su %u = bl %u\n", off, su, bl);
stripeno = bl / sc;
stripepos = bl % sc;
objsetno = stripeno / su_per_object;
*ono = objsetno * sc + stripepos;
dout("objset %u * sc %u = ono %u\n", objsetno, sc, (unsigned int)*ono);
/* *oxoff = *off % layout->fl_stripe_unit; # offset in su */
t = off;
su_offset = do_div(t, su);
*oxoff = su_offset + (stripeno % su_per_object) * su;
/* blockno = div_u64_rem(off, l->stripe_unit, &blockoff);
* Calculate the length of the extent being written to the selected stripeno = div_u64_rem(blockno, l->stripe_count, &stripepos);
* object. This is the minimum of the full length requested (len) or objsetno = div_u64_rem(stripeno, stripes_per_object, &objsetpos);
* the remainder of the current stripe being written to.
*/
*oxlen = min_t(u64, len, su - su_offset);
dout(" obj extent %llu~%llu\n", *oxoff, *oxlen); *objno = objsetno * l->stripe_count + stripepos;
return 0; *objoff = objsetpos * l->stripe_unit + blockoff;
*xlen = min_t(u64, len, l->stripe_unit - blockoff);
invalid: return 0;
dout(" invalid layout\n");
*ono = 0;
*oxoff = 0;
*oxlen = 0;
return -EINVAL;
} }
EXPORT_SYMBOL(ceph_calc_file_object_mapping); EXPORT_SYMBOL(ceph_calc_file_object_mapping);
......
Markdown is supported
0% .
You are about to add 0 people to the discussion. Proceed with caution.
先完成此消息的编辑!
想要评论请 注册