[<prev] [next>] [day] [month] [year] [list]
Message-ID: <4bf45476-ad99-042b-d0ea-3c97bb429a98@web.de>
Date: Mon, 25 May 2020 14:54:47 +0200
From: Markus Elfring <Markus.Elfring@....de>
To: Qiushi Wu <wu000273@....edu>, linux-rdma@...r.kernel.org
Cc: kernel-janitors@...r.kernel.org, linux-kernel@...r.kernel.org,
Doug Ledford <dledford@...hat.com>,
Jason Gunthorpe <jgg@...pe.ca>, Kangjie Lu <kjlu@....edu>,
Leon Romanovsky <leon@...nel.org>
Subject: Re: [PATCH] RDMA/core: Complete exception handling in add_port()
> In function add_port(), pointer p is not released in error paths.
1. I would prefer to describe that an ib_port data structure was not released
in some error cases.
How relevant can its size be here?
> Fix this issue by adding a kfree(p) into the end of error path.
2. I suggest to improve also this change description.
3. I find an other subject more appropriate.
…
> +++ b/drivers/infiniband/core/sysfs.c
> @@ -1202,6 +1202,7 @@ static int add_port(struct ib_core_device *coredev, int port_num)
>
> err_put:
> kobject_put(&p->kobj);
> + kfree(p);
> return ret;
> }
4. I recommend to add also the label “free_port” before the missed function call.
Another source code place should be accordingly adjusted then.
ret = kobject_init_and_add(&p->kobj, &port_type,
coredev->ports_kobj,
"%d", port_num);
- if (ret) {
- kfree(p);
- return ret;
- }
+ if (ret)
+ goto free_port;
5. Will a similar adjustment be needed for the data structure member “gid_attr_group”
according to the desired complete exception handling?
Regards,
Markus
Powered by blists - more mailing lists