[LU-507] Lustre sometimes return EPROTO instead of ENOSPC

• Type: Bug
• Resolution: Fixed
• Priority: Minor
• Story Points: 1
• Severity: 3
• Project: Orion
• Assignee: Johann Lombardi (Inactive)
• Reporter: Johann Lombardi (Inactive)

Description

Seen while running acc-sm:

== sanity test 64b: check out-of-space detection on client =========================================== 03:50:40 (1311072640)
STRIPECOUNT=2 ORIGFREE=292008 MAXFREE=800000
BEFORE dd started
lustre-OST0000 avl=146676 grnt=38592 diff=108084
lustre-OST0001 avl=145332 grnt=9804 diff=135528
ERROR: dd not return ENOSPC
LOG: dd: writing `/mnt/lustre/oosfile': Protocol error
LOG: 291009+0 records in
LOG: 291008+0 records out
LOG: 297992192 bytes (298 MB) copied, 38.9363 seconds, 7.7 MB/s
AFTER dd
lustre-OST0000 avl=1580 grnt=1332 diff=248 FULL
lustre-OST0001 avl=32 grnt=0 diff=32 FULL
LOG file
dd: writing `/mnt/lustre/oosfile': Protocol error

The problem lies in check_write_rcs():

static int check_write_rcs(struct ptlrpc_request *req,
                           int requested_nob, int niocount,
                           obd_count page_count, struct brw_page **pga)
{
        int     i;   
        __u32   *remote_rcs;
...
        if (remote_rcs[i] < 0) 
                return(remote_rcs[i]);
        if (remote_rcs[i] != 0) { 
                CDEBUG(D_INFO, "rc[%d] invalid (%d) req %p\n",
                        i, remote_rcs[i], req);
                return(-EPROTO);
        } 

The check remote_rcs[i] < 0 is always false since remote_rcs[i] is unsigned.