[PATCH] stm class: Fix double add issue when store source_link
Jinlong Mao
quic_jinlmao at quicinc.com
Wed Nov 2 02:34:16 PDT 2022
Hi all,
Please help to review this patch.
Thanks
Jinlong Mao
On 5/16/2022 3:14 PM, Jinlong Mao wrote:
> Hi Reviewers,
>
> Could you please help to review this patch ?
>
> Thanks
>
> Jinlong Mao
>
> On 4/18/2022 7:46 PM, Mao Jinlong wrote:
>> If two threads store the same stm device to stm_source_link
>> at the same time when stm->link_list is empty, it is possible
>> that stm_source_link_add will be called for both of these two
>> threads. Then double add issue below will happen. Add mutex
>> lock for stm_source_link drop and stm_source_link add to avoid
>> this race condition.
>>
>> [ 12.386579][ T1024] list_add double add: new=ffffff87b73ebd90,
>> prev=ffffff87b73ebd90, next=ffffffc012737700.
>> [ 12.386657][ T1024] -----------[ cut here ]-----------
>> [ 12.386671][ T1024] kernel BUG at lib/list_debug.c:31!
>> [ 12.388845][ T1024] CPU: 2 PID: 1024 Comm: sh
>> [ 12.389162][ T1024] Call trace:
>> [ 12.389174][ T1024] __list_add_valid+0x68/0x98
>> [ 12.389199][ T1024] stm_source_link_store+0xcc/0x314 [stm_core]
>> [ 12.389213][ T1024] dev_attr_store+0x38/0x8c
>> [ 12.389228][ T1024] sysfs_kf_write+0xa0/0x100
>> [ 12.389239][ T1024] kernfs_fop_write_iter+0x1b0/0x2f8
>> [ 12.389253][ T1024] vfs_write+0x300/0x37c
>> [ 12.389264][ T1024] ksys_write+0x84/0x12c
>>
>> Signed-off-by: Yuanfang Zhang <quic_yuanfang at quicinc.com>
>> Signed-off-by: Mao Jinlong <quic_jinlmao at quicinc.com>
>> ---
>> drivers/hwtracing/stm/core.c | 7 ++++++-
>> drivers/hwtracing/stm/stm.h | 1 +
>> 2 files changed, 7 insertions(+), 1 deletion(-)
>>
>> diff --git a/drivers/hwtracing/stm/core.c b/drivers/hwtracing/stm/core.c
>> index 2712e699ba08..e73ac961acb2 100644
>> --- a/drivers/hwtracing/stm/core.c
>> +++ b/drivers/hwtracing/stm/core.c
>> @@ -1171,11 +1171,14 @@ static ssize_t stm_source_link_store(struct
>> device *dev,
>> struct stm_device *link;
>> int err;
>> + mutex_lock(&src->link_mutex);
>> stm_source_link_drop(src);
>> link = stm_find_device(buf);
>> - if (!link)
>> + if (!link) {
>> + mutex_lock(&src->link_mutex);
>> return -EINVAL;
>> + }
>> pm_runtime_get(&link->dev);
>> @@ -1185,6 +1188,7 @@ static ssize_t stm_source_link_store(struct
>> device *dev,
>> /* matches the stm_find_device() above */
>> stm_put_device(link);
>> }
>> + mutex_unlock(&src->link_mutex);
>> return err ? : count;
>> }
>> @@ -1251,6 +1255,7 @@ int stm_source_register_device(struct device
>> *parent,
>> stm_output_init(&src->output);
>> spin_lock_init(&src->link_lock);
>> + mutex_init(&src->link_mutex);
>> INIT_LIST_HEAD(&src->link_entry);
>> src->data = data;
>> data->src = src;
>> diff --git a/drivers/hwtracing/stm/stm.h b/drivers/hwtracing/stm/stm.h
>> index a9be49fc7a6b..60b814cc00e0 100644
>> --- a/drivers/hwtracing/stm/stm.h
>> +++ b/drivers/hwtracing/stm/stm.h
>> @@ -79,6 +79,7 @@ void stm_put_device(struct stm_device *stm);
>> struct stm_source_device {
>> struct device dev;
>> struct stm_source_data *data;
>> + struct mutex link_mutex;
>> spinlock_t link_lock;
>> struct stm_device __rcu *link;
>> struct list_head link_entry;
More information about the linux-arm-kernel
mailing list