2015-11-26 13:46:26

by Minfei Huang

[permalink] [raw]
Subject: [PATCH] null_blk: Fix error path in module initialization

From: Minfei Huang <[email protected]>

Module couldn't release resource properly during the initialization. To
fix this issue, we will clean up the proper resource before returning.

Signed-off-by: Minfei Huang <[email protected]>
---
drivers/block/null_blk.c | 24 +++++++++++++++++-------
1 file changed, 17 insertions(+), 7 deletions(-)

diff --git a/drivers/block/null_blk.c b/drivers/block/null_blk.c
index 5c8ba54..ec99568 100644
--- a/drivers/block/null_blk.c
+++ b/drivers/block/null_blk.c
@@ -780,7 +780,9 @@ out:

static int __init null_init(void)
{
+ int ret = 0;
unsigned int i;
+ struct nullb *nullb;

if (bs > PAGE_SIZE) {
pr_warn("null_blk: invalid block size\n");
@@ -835,22 +837,30 @@ static int __init null_init(void)
0, 0, NULL);
if (!ppa_cache) {
pr_err("null_blk: unable to create ppa cache\n");
- return -ENOMEM;
+ ret = -ENOMEM;
+ goto err_ppa;
}
}

for (i = 0; i < nr_devices; i++) {
- if (null_add_dev()) {
- unregister_blkdev(null_major, "nullb");
- goto err_ppa;
- }
+ ret = null_add_dev();
+ if (ret)
+ goto err_dev;
}

pr_info("null: module loaded\n");
return 0;
+
+err_dev:
+ while (!list_empty(&nullb_list)) {
+ nullb = list_entry(nullb_list.next, struct nullb, list);
+ null_del_dev(nullb);
+ }
+ if (use_lightnvm)
+ kmem_cache_destroy(ppa_cache);
err_ppa:
- kmem_cache_destroy(ppa_cache);
- return -EINVAL;
+ unregister_blkdev(null_major, "nullb");
+ return ret;
}

static void __exit null_exit(void)
--
1.8.3.1


2015-12-08 12:35:46

by Minfei Huang

[permalink] [raw]
Subject: Re: [PATCH] null_blk: Fix error path in module initialization

Ping.

Any comment is appreciate.

Thanks
Minfei

On 11/26/15 at 09:48P, Minfei Huang wrote:
> From: Minfei Huang <[email protected]>
>
> Module couldn't release resource properly during the initialization. To
> fix this issue, we will clean up the proper resource before returning.
>
> Signed-off-by: Minfei Huang <[email protected]>
> ---
> drivers/block/null_blk.c | 24 +++++++++++++++++-------
> 1 file changed, 17 insertions(+), 7 deletions(-)
>
> diff --git a/drivers/block/null_blk.c b/drivers/block/null_blk.c
> index 5c8ba54..ec99568 100644
> --- a/drivers/block/null_blk.c
> +++ b/drivers/block/null_blk.c
> @@ -780,7 +780,9 @@ out:
>
> static int __init null_init(void)
> {
> + int ret = 0;
> unsigned int i;
> + struct nullb *nullb;
>
> if (bs > PAGE_SIZE) {
> pr_warn("null_blk: invalid block size\n");
> @@ -835,22 +837,30 @@ static int __init null_init(void)
> 0, 0, NULL);
> if (!ppa_cache) {
> pr_err("null_blk: unable to create ppa cache\n");
> - return -ENOMEM;
> + ret = -ENOMEM;
> + goto err_ppa;
> }
> }
>
> for (i = 0; i < nr_devices; i++) {
> - if (null_add_dev()) {
> - unregister_blkdev(null_major, "nullb");
> - goto err_ppa;
> - }
> + ret = null_add_dev();
> + if (ret)
> + goto err_dev;
> }
>
> pr_info("null: module loaded\n");
> return 0;
> +
> +err_dev:
> + while (!list_empty(&nullb_list)) {
> + nullb = list_entry(nullb_list.next, struct nullb, list);
> + null_del_dev(nullb);
> + }
> + if (use_lightnvm)
> + kmem_cache_destroy(ppa_cache);
> err_ppa:
> - kmem_cache_destroy(ppa_cache);
> - return -EINVAL;
> + unregister_blkdev(null_major, "nullb");
> + return ret;
> }
>
> static void __exit null_exit(void)
> --
> 1.8.3.1
>

2015-12-08 20:49:07

by Jens Axboe

[permalink] [raw]
Subject: Re: [PATCH] null_blk: Fix error path in module initialization

On 12/08/2015 05:35 AM, Minfei Huang wrote:
> Ping.
>
> Any comment is appreciate.

Applied, it is indeed pretty broken after the lightnvm update.
I killed the use_lightnvm check before kmem_cache_destroy, that's not
needed.

--
Jens Axboe

2015-12-09 04:18:22

by Minfei Huang

[permalink] [raw]
Subject: Re: [PATCH] null_blk: Fix error path in module initialization

On 12/08/15 at 01:48pm, Jens Axboe wrote:
> On 12/08/2015 05:35 AM, Minfei Huang wrote:
> >Ping.
> >
> >Any comment is appreciate.
>
> Applied, it is indeed pretty broken after the lightnvm update.
> I killed the use_lightnvm check before kmem_cache_destroy, that's
> not needed.

Agreed.

Thanks
Minfei