lkml.org 
[lkml]   [2017]   [Apr]   [7]   [last100]   RSS Feed
Views: [wrap][no wrap]   [headers]  [forward] 
 
Messages in this thread
/
Date
From
SubjectRe: [PATCH] RDS: IB: ensure an initialized ret is printed in pr_warn message
Setting it to zero doesn't seem like the right thing, it should be an
error code. Oh, heh... Smatch parses this one correctly. "ret" is
always initialized but the code is probably buggy in a different way:

net/rds/ib_rdma.c
539 void *rds_ib_get_mr(struct scatterlist *sg, unsigned long nents,
540 struct rds_sock *rs, u32 *key_ret)
541 {
542 struct rds_ib_device *rds_ibdev;
543 struct rds_ib_mr *ibmr = NULL;
544 struct rds_ib_connection *ic = rs->rs_conn->c_transport_data;
545 int ret;
546
547 rds_ibdev = rds_ib_get_device(rs->rs_bound_addr);
548 if (!rds_ibdev) {
549 ret = -ENODEV;
550 goto out;
551 }
552
553 if (!rds_ibdev->mr_8k_pool || !rds_ibdev->mr_1m_pool) {
554 ret = -ENODEV;
555 goto out;
556 }
557
558 if (rds_ibdev->use_fastreg)
559 ibmr = rds_ib_reg_frmr(rds_ibdev, ic, sg, nents, key_ret);
560 else
561 ibmr = rds_ib_reg_fmr(rds_ibdev, sg, nents, key_ret);
562 if (ibmr)
^^^^
This condition is always true because those functions return ERR_PTRs
not NULLs.

563 rds_ibdev = NULL;
564
565 out:
566 if (!ibmr)
^^^^^
This condition implies that "ret" is set to an error code.

567 pr_warn("RDS/IB: rds_ib_get_mr failed (errno=%d)\n", ret);
568
569 if (rds_ibdev)
570 rds_ib_dev_put(rds_ibdev);
571
572 return ibmr;
573 }

regards,
dan carpenter

\
 
 \ /
  Last update: 2017-04-07 16:27    [W:0.056 / U:0.352 seconds]
©2003-2020 Jasper Spaans|hosted at Digital Ocean and TransIP|Read the blog|Advertise on this site