RDMA/mlx5: Fix prefetch memory leak if get_prefetchable_mr fails

destroy_prefetch_work() must always be called if the work is not going
to be queued. The num_sge also should have been set to i, not i-1
which avoids the condition where it shouldn't have been called in the
first place.

Cc: stable@vger.kernel.org
Fixes: fb985e278a ("RDMA/mlx5: Use SRCU properly in ODP prefetch")
Link: https://lore.kernel.org/r/20200727095712.495652-1-leon@kernel.org
Signed-off-by: Leon Romanovsky <leonro@mellanox.com>
Signed-off-by: Jason Gunthorpe <jgg@nvidia.com>
This commit is contained in:
Jason Gunthorpe 2020-07-27 12:57:12 +03:00
parent 31142a4ba6
commit 5351a56b1a

View File

@ -1797,9 +1797,7 @@ static bool init_prefetch_work(struct ib_pd *pd,
work->frags[i].mr =
get_prefetchable_mr(pd, advice, sg_list[i].lkey);
if (!work->frags[i].mr) {
work->num_sge = i - 1;
if (i)
destroy_prefetch_work(work);
work->num_sge = i;
return false;
}
@ -1865,6 +1863,7 @@ int mlx5_ib_advise_mr_prefetch(struct ib_pd *pd,
srcu_key = srcu_read_lock(&dev->odp_srcu);
if (!init_prefetch_work(pd, advice, pf_flags, work, sg_list, num_sge)) {
srcu_read_unlock(&dev->odp_srcu, srcu_key);
destroy_prefetch_work(work);
return -EINVAL;
}
queue_work(system_unbound_wq, &work->work);