| Message ID | 55BF9F11.6050708@sandisk.com (mailing list archive) |
|---|---|
| State | Accepted |
| Headers | show |
On 08/03/2015 01:04 PM, Bart Van Assche wrote: > Fix the memory leaks in update_gids_task() and reset_gids_task(). > kmemleak reported one of these leaks as follows: > > unreferenced object 0xffff88044576d668 (size 4096): > comm "softirq", pid 0, jiffies 4294942043 (age 976.560s) > hex dump (first 32 bytes): > 00 02 00 00 00 00 00 00 70 d6 76 45 04 88 ff ff ........p.vE.... > 70 d6 76 45 04 88 ff ff 60 d6 2d a0 ff ff ff ff p.vE....`.-..... > backtrace: > [<ffffffff814f2f1e>] kmemleak_alloc+0x4e/0xb0 > [<ffffffff811b0678>] kmem_cache_alloc_trace+0xc8/0x2d0 > [<ffffffffa02e061e>] reset_gid_table+0x2e/0xe0 [mlx4_ib] > [<ffffffffa02e1c59>] mlx4_ib_netdev_event+0x449/0x450 [mlx4_ib] > [<ffffffff814011ca>] register_netdevice_notifier+0x8a/0x1c0 > [<ffffffffa02e173b>] mlx4_ib_add+0x106b/0x1140 [mlx4_ib] > [<ffffffffa00dc3d1>] mlx4_add_device+0x41/0xa0 [mlx4_core] > [<ffffffffa00dc4bc>] mlx4_register_interface+0x8c/0xf0 [mlx4_core] > [<ffffffffa02fc056>] ext4_has_free_clusters+0x6/0x170 [ext4] > [<ffffffff8100211d>] do_one_initcall+0xad/0x1e0 > [<ffffffff814f7238>] do_init_module+0x61/0x1ec > [<ffffffff810f95f2>] load_module+0x1d42/0x26b0 > [<ffffffff810fa150>] SyS_finit_module+0x80/0xb0 > [<ffffffff81502332>] entry_SYSCALL_64_fastpath+0x16/0x7a > [<ffffffffffffffff>] 0xffffffffffffffff > > Fixes: fa417f7b520e ("IB/mlx4: Add support for IBoE") > Signed-off-by: Bart Van Assche <bart.vanassche@sandisk.com> > Cc: Eli Cohen <eli@mellanox.co.il> > Cc: stable@vger.kernel.org # v2.6.37 Thanks, I've picked this up.
diff --git a/drivers/infiniband/hw/mlx4/main.c b/drivers/infiniband/hw/mlx4/main.c index 8d88760..d79346e 100644 --- a/drivers/infiniband/hw/mlx4/main.c +++ b/drivers/infiniband/hw/mlx4/main.c @@ -1572,12 +1572,12 @@ static void update_gids_task(struct work_struct *work) int is_bonded = mlx4_is_bonded(dev); if (!gw->dev->ib_active) - return; + goto free; mailbox = mlx4_alloc_cmd_mailbox(dev); if (IS_ERR(mailbox)) { pr_warn("update gid table failed %ld\n", PTR_ERR(mailbox)); - return; + goto free; } gids = mailbox->buf; @@ -1595,6 +1595,8 @@ static void update_gids_task(struct work_struct *work) IB_EVENT_GID_CHANGE); mlx4_free_cmd_mailbox(dev, mailbox); + +free: kfree(gw); } @@ -1608,7 +1610,7 @@ static void reset_gids_task(struct work_struct *work) struct mlx4_dev *dev = gw->dev->dev; if (!gw->dev->ib_active) - return; + goto free; mailbox = mlx4_alloc_cmd_mailbox(dev); if (IS_ERR(mailbox)) {
Fix the memory leaks in update_gids_task() and reset_gids_task(). kmemleak reported one of these leaks as follows: unreferenced object 0xffff88044576d668 (size 4096): comm "softirq", pid 0, jiffies 4294942043 (age 976.560s) hex dump (first 32 bytes): 00 02 00 00 00 00 00 00 70 d6 76 45 04 88 ff ff ........p.vE.... 70 d6 76 45 04 88 ff ff 60 d6 2d a0 ff ff ff ff p.vE....`.-..... backtrace: [<ffffffff814f2f1e>] kmemleak_alloc+0x4e/0xb0 [<ffffffff811b0678>] kmem_cache_alloc_trace+0xc8/0x2d0 [<ffffffffa02e061e>] reset_gid_table+0x2e/0xe0 [mlx4_ib] [<ffffffffa02e1c59>] mlx4_ib_netdev_event+0x449/0x450 [mlx4_ib] [<ffffffff814011ca>] register_netdevice_notifier+0x8a/0x1c0 [<ffffffffa02e173b>] mlx4_ib_add+0x106b/0x1140 [mlx4_ib] [<ffffffffa00dc3d1>] mlx4_add_device+0x41/0xa0 [mlx4_core] [<ffffffffa00dc4bc>] mlx4_register_interface+0x8c/0xf0 [mlx4_core] [<ffffffffa02fc056>] ext4_has_free_clusters+0x6/0x170 [ext4] [<ffffffff8100211d>] do_one_initcall+0xad/0x1e0 [<ffffffff814f7238>] do_init_module+0x61/0x1ec [<ffffffff810f95f2>] load_module+0x1d42/0x26b0 [<ffffffff810fa150>] SyS_finit_module+0x80/0xb0 [<ffffffff81502332>] entry_SYSCALL_64_fastpath+0x16/0x7a [<ffffffffffffffff>] 0xffffffffffffffff Fixes: fa417f7b520e ("IB/mlx4: Add support for IBoE") Signed-off-by: Bart Van Assche <bart.vanassche@sandisk.com> Cc: Eli Cohen <eli@mellanox.co.il> Cc: stable@vger.kernel.org # v2.6.37 --- drivers/infiniband/hw/mlx4/main.c | 8 +++++--- 1 file changed, 5 insertions(+), 3 deletions(-)