diff mbox series

[v2] vdpa_sim: fix possible memory leak in vdpasim_net_init() and vdpasim_blk_init()

Message ID 20221110050446.3932031-1-ruanjinjie@huawei.com
State New
Headers
Series [v2] vdpa_sim: fix possible memory leak in vdpasim_net_init() and vdpasim_blk_init() |

Commit Message

Jinjie Ruan Nov. 10, 2022, 5:04 a.m. UTC 
  Inject fault while probing module, if device_register() fails in
vdpasim_net_init() or vdpasim_blk_init(), but the refcount of kobject is
not decreased to 0, the name allocated in dev_set_name() is leaked.
Fix this by calling put_device(), so that name can be freed in
callback function kobject_cleanup().

(vdpa_sim_net)
unreferenced object 0xffff88807eebc370 (size 16):
  comm "modprobe", pid 3848, jiffies 4362982860 (age 18.153s)
  hex dump (first 16 bytes):
    76 64 70 61 73 69 6d 5f 6e 65 74 00 6b 6b 6b a5  vdpasim_net.kkk.
  backtrace:
    [<ffffffff8174f19e>] __kmalloc_node_track_caller+0x4e/0x150
    [<ffffffff81731d53>] kstrdup+0x33/0x60
    [<ffffffff83a5d421>] kobject_set_name_vargs+0x41/0x110
    [<ffffffff82d87aab>] dev_set_name+0xab/0xe0
    [<ffffffff82d91a23>] device_add+0xe3/0x1a80
    [<ffffffffa0270013>] 0xffffffffa0270013
    [<ffffffff81001c27>] do_one_initcall+0x87/0x2e0
    [<ffffffff813739cb>] do_init_module+0x1ab/0x640
    [<ffffffff81379d20>] load_module+0x5d00/0x77f0
    [<ffffffff8137bc40>] __do_sys_finit_module+0x110/0x1b0
    [<ffffffff83c4d505>] do_syscall_64+0x35/0x80
    [<ffffffff83e0006a>] entry_SYSCALL_64_after_hwframe+0x46/0xb0

(vdpa_sim_blk)
unreferenced object 0xffff8881070c1250 (size 16):
  comm "modprobe", pid 6844, jiffies 4364069319 (age 17.572s)
  hex dump (first 16 bytes):
    76 64 70 61 73 69 6d 5f 62 6c 6b 00 6b 6b 6b a5  vdpasim_blk.kkk.
  backtrace:
    [<ffffffff8174f19e>] __kmalloc_node_track_caller+0x4e/0x150
    [<ffffffff81731d53>] kstrdup+0x33/0x60
    [<ffffffff83a5d421>] kobject_set_name_vargs+0x41/0x110
    [<ffffffff82d87aab>] dev_set_name+0xab/0xe0
    [<ffffffff82d91a23>] device_add+0xe3/0x1a80
    [<ffffffffa0220013>] 0xffffffffa0220013
    [<ffffffff81001c27>] do_one_initcall+0x87/0x2e0
    [<ffffffff813739cb>] do_init_module+0x1ab/0x640
    [<ffffffff81379d20>] load_module+0x5d00/0x77f0
    [<ffffffff8137bc40>] __do_sys_finit_module+0x110/0x1b0
    [<ffffffff83c4d505>] do_syscall_64+0x35/0x80
    [<ffffffff83e0006a>] entry_SYSCALL_64_after_hwframe+0x46/0xb0

Signed-off-by: ruanjinjie <ruanjinjie@huawei.com>
---
v2:
- add fault inject message
---
 drivers/vdpa/vdpa_sim/vdpa_sim_blk.c | 4 +++-
 drivers/vdpa/vdpa_sim/vdpa_sim_net.c | 4 +++-
 2 files changed, 6 insertions(+), 2 deletions(-)

Comments

Stefano Garzarella Nov. 10, 2022, 8:01 a.m. UTC | #1 
Hi,

On Thu, Nov 10, 2022 at 01:04:46PM +0800, ruanjinjie wrote:
>Inject fault while probing module, if device_register() fails in
>vdpasim_net_init() or vdpasim_blk_init(), but the refcount of kobject is
>not decreased to 0, the name allocated in dev_set_name() is leaked.
>Fix this by calling put_device(), so that name can be freed in
>callback function kobject_cleanup().
>
>(vdpa_sim_net)
>unreferenced object 0xffff88807eebc370 (size 16):
>  comm "modprobe", pid 3848, jiffies 4362982860 (age 18.153s)
>  hex dump (first 16 bytes):
>    76 64 70 61 73 69 6d 5f 6e 65 74 00 6b 6b 6b a5  vdpasim_net.kkk.
>  backtrace:
>    [<ffffffff8174f19e>] __kmalloc_node_track_caller+0x4e/0x150
>    [<ffffffff81731d53>] kstrdup+0x33/0x60
>    [<ffffffff83a5d421>] kobject_set_name_vargs+0x41/0x110
>    [<ffffffff82d87aab>] dev_set_name+0xab/0xe0
>    [<ffffffff82d91a23>] device_add+0xe3/0x1a80
>    [<ffffffffa0270013>] 0xffffffffa0270013
>    [<ffffffff81001c27>] do_one_initcall+0x87/0x2e0
>    [<ffffffff813739cb>] do_init_module+0x1ab/0x640
>    [<ffffffff81379d20>] load_module+0x5d00/0x77f0
>    [<ffffffff8137bc40>] __do_sys_finit_module+0x110/0x1b0
>    [<ffffffff83c4d505>] do_syscall_64+0x35/0x80
>    [<ffffffff83e0006a>] entry_SYSCALL_64_after_hwframe+0x46/0xb0
>
>(vdpa_sim_blk)
>unreferenced object 0xffff8881070c1250 (size 16):
>  comm "modprobe", pid 6844, jiffies 4364069319 (age 17.572s)
>  hex dump (first 16 bytes):
>    76 64 70 61 73 69 6d 5f 62 6c 6b 00 6b 6b 6b a5  vdpasim_blk.kkk.
>  backtrace:
>    [<ffffffff8174f19e>] __kmalloc_node_track_caller+0x4e/0x150
>    [<ffffffff81731d53>] kstrdup+0x33/0x60
>    [<ffffffff83a5d421>] kobject_set_name_vargs+0x41/0x110
>    [<ffffffff82d87aab>] dev_set_name+0xab/0xe0
>    [<ffffffff82d91a23>] device_add+0xe3/0x1a80
>    [<ffffffffa0220013>] 0xffffffffa0220013
>    [<ffffffff81001c27>] do_one_initcall+0x87/0x2e0
>    [<ffffffff813739cb>] do_init_module+0x1ab/0x640
>    [<ffffffff81379d20>] load_module+0x5d00/0x77f0
>    [<ffffffff8137bc40>] __do_sys_finit_module+0x110/0x1b0
>    [<ffffffff83c4d505>] do_syscall_64+0x35/0x80
>    [<ffffffff83e0006a>] entry_SYSCALL_64_after_hwframe+0x46/0xb0
>
>Signed-off-by: ruanjinjie <ruanjinjie@huawei.com>
>---
>v2:
>- add fault inject message
>---
> drivers/vdpa/vdpa_sim/vdpa_sim_blk.c | 4 +++-
> drivers/vdpa/vdpa_sim/vdpa_sim_net.c | 4 +++-
> 2 files changed, 6 insertions(+), 2 deletions(-)

Thanks for the v2 of this patch!
It LGTM, just a couple of comments about the submit process.

Usually when you get A-b R-b's and the patch doesn't change, it's better 
to bring them in later versions.

Also, we had suggested reporting the Fixes tag, because usually when we 
fix a bug it's good to identify which patch introduced the problem, so 
it's easier to backport this fix into stable versions of the kernel.
https://www.kernel.org/doc/html/v4.17/process/submitting-patches.html#describe-your-changes

In this case we should use the following tags:

Fixes: 899c4d187f6a ("vdpa_sim_blk: add support for vdpa management tool")
Fixes: a3c06ae158dd ("vdpa_sim_net: Add support for user supported devices")

With them:

Reviewed-by: Stefano Garzarella <sgarzare@redhat.com>

Thanks,
Stefano
Jinjie Ruan Nov. 10, 2022, 8:59 a.m. UTC | #2 
Thank you very much!

On 2022/11/10 16:01, Stefano Garzarella wrote:
> Hi,
> 
> On Thu, Nov 10, 2022 at 01:04:46PM +0800, ruanjinjie wrote:
>> Inject fault while probing module, if device_register() fails in
>> vdpasim_net_init() or vdpasim_blk_init(), but the refcount of kobject is
>> not decreased to 0, the name allocated in dev_set_name() is leaked.
>> Fix this by calling put_device(), so that name can be freed in
>> callback function kobject_cleanup().
>>
>> (vdpa_sim_net)
>> unreferenced object 0xffff88807eebc370 (size 16):
>>  comm "modprobe", pid 3848, jiffies 4362982860 (age 18.153s)
>>  hex dump (first 16 bytes):
>>    76 64 70 61 73 69 6d 5f 6e 65 74 00 6b 6b 6b a5  vdpasim_net.kkk.
>>  backtrace:
>>    [<ffffffff8174f19e>] __kmalloc_node_track_caller+0x4e/0x150
>>    [<ffffffff81731d53>] kstrdup+0x33/0x60
>>    [<ffffffff83a5d421>] kobject_set_name_vargs+0x41/0x110
>>    [<ffffffff82d87aab>] dev_set_name+0xab/0xe0
>>    [<ffffffff82d91a23>] device_add+0xe3/0x1a80
>>    [<ffffffffa0270013>] 0xffffffffa0270013
>>    [<ffffffff81001c27>] do_one_initcall+0x87/0x2e0
>>    [<ffffffff813739cb>] do_init_module+0x1ab/0x640
>>    [<ffffffff81379d20>] load_module+0x5d00/0x77f0
>>    [<ffffffff8137bc40>] __do_sys_finit_module+0x110/0x1b0
>>    [<ffffffff83c4d505>] do_syscall_64+0x35/0x80
>>    [<ffffffff83e0006a>] entry_SYSCALL_64_after_hwframe+0x46/0xb0
>>
>> (vdpa_sim_blk)
>> unreferenced object 0xffff8881070c1250 (size 16):
>>  comm "modprobe", pid 6844, jiffies 4364069319 (age 17.572s)
>>  hex dump (first 16 bytes):
>>    76 64 70 61 73 69 6d 5f 62 6c 6b 00 6b 6b 6b a5  vdpasim_blk.kkk.
>>  backtrace:
>>    [<ffffffff8174f19e>] __kmalloc_node_track_caller+0x4e/0x150
>>    [<ffffffff81731d53>] kstrdup+0x33/0x60
>>    [<ffffffff83a5d421>] kobject_set_name_vargs+0x41/0x110
>>    [<ffffffff82d87aab>] dev_set_name+0xab/0xe0
>>    [<ffffffff82d91a23>] device_add+0xe3/0x1a80
>>    [<ffffffffa0220013>] 0xffffffffa0220013
>>    [<ffffffff81001c27>] do_one_initcall+0x87/0x2e0
>>    [<ffffffff813739cb>] do_init_module+0x1ab/0x640
>>    [<ffffffff81379d20>] load_module+0x5d00/0x77f0
>>    [<ffffffff8137bc40>] __do_sys_finit_module+0x110/0x1b0
>>    [<ffffffff83c4d505>] do_syscall_64+0x35/0x80
>>    [<ffffffff83e0006a>] entry_SYSCALL_64_after_hwframe+0x46/0xb0
>>
>> Signed-off-by: ruanjinjie <ruanjinjie@huawei.com>
>> ---
>> v2:
>> - add fault inject message
>> ---
>> drivers/vdpa/vdpa_sim/vdpa_sim_blk.c | 4 +++-
>> drivers/vdpa/vdpa_sim/vdpa_sim_net.c | 4 +++-
>> 2 files changed, 6 insertions(+), 2 deletions(-)
> 
> Thanks for the v2 of this patch!
> It LGTM, just a couple of comments about the submit process.
> 
> Usually when you get A-b R-b's and the patch doesn't change, it's better
> to bring them in later versions.
> 
> Also, we had suggested reporting the Fixes tag, because usually when we
> fix a bug it's good to identify which patch introduced the problem, so
> it's easier to backport this fix into stable versions of the kernel.
> https://www.kernel.org/doc/html/v4.17/process/submitting-patches.html#describe-your-changes
> 
> In this case we should use the following tags:
> 
> Fixes: 899c4d187f6a ("vdpa_sim_blk: add support for vdpa management tool")
> Fixes: a3c06ae158dd ("vdpa_sim_net: Add support for user supported
> devices")
> 
> With them:
> 
> Reviewed-by: Stefano Garzarella <sgarzare@redhat.com>
> 
> Thanks,
> Stefano
>
diff mbox series

Patch

diff --git a/drivers/vdpa/vdpa_sim/vdpa_sim_blk.c b/drivers/vdpa/vdpa_sim/vdpa_sim_blk.c
index c6db1a1baf76..f745926237a8 100644
--- a/drivers/vdpa/vdpa_sim/vdpa_sim_blk.c
+++ b/drivers/vdpa/vdpa_sim/vdpa_sim_blk.c
@@ -427,8 +427,10 @@  static int __init vdpasim_blk_init(void)
 	int ret;
 
 	ret = device_register(&vdpasim_blk_mgmtdev);
-	if (ret)
+	if (ret) {
+		put_device(&vdpasim_blk_mgmtdev);
 		return ret;
+	}
 
 	ret = vdpa_mgmtdev_register(&mgmt_dev);
 	if (ret)
diff --git a/drivers/vdpa/vdpa_sim/vdpa_sim_net.c b/drivers/vdpa/vdpa_sim/vdpa_sim_net.c
index c3cb225ea469..11f5a121df24 100644
--- a/drivers/vdpa/vdpa_sim/vdpa_sim_net.c
+++ b/drivers/vdpa/vdpa_sim/vdpa_sim_net.c
@@ -305,8 +305,10 @@  static int __init vdpasim_net_init(void)
 	int ret;
 
 	ret = device_register(&vdpasim_net_mgmtdev);
-	if (ret)
+	if (ret) {
+		put_device(&vdpasim_net_mgmtdev);
 		return ret;
+	}
 
 	ret = vdpa_mgmtdev_register(&mgmt_dev);
 	if (ret)