[LU-3162] lov_getstripe() makes incorrect assumptions about layout of lov_mds_md_v1 Created: 12/Apr/13 Updated: 16/Oct/13 Resolved: 16/Oct/13 |
|
| Status: | Resolved |
| Project: | Lustre |
| Component/s: | None |
| Affects Version/s: | Lustre 2.4.0 |
| Fix Version/s: | Lustre 2.5.0 |
| Type: | Bug | Priority: | Minor |
| Reporter: | John Hammond | Assignee: | WC Triage |
| Resolution: | Fixed | Votes: | 0 |
| Labels: | lov | ||
| Severity: | 3 |
| Rank (Obsolete): | 7705 |
| Description |
|
The rarely tested lov_getstripe() assumes that (char*)(&lmmk->lmm_stripe_count) + sizeof(lmmk->lmm_stripe_count) is the address of lmm_objects when lmmk is cast to lov_mds_md_v1. (This was written before lmm_layout_gen was introduced. But this not the way to do that in any case.) |
| Comments |
| Comment by Swapnil Pimpale (Inactive) [ 07/Aug/13 ] |
|
The introduction of lmm_layout_gen in lov_mds_md_v1 makes the assumption that lmm_objects is present after lmm_stripe_count incorrect. |
| Comment by Jodi Levi (Inactive) [ 16/Oct/13 ] |
|
Closing ticket as patch landed to Master. Please let me know if more work is needed in this ticket and I will reopen. |