On Mon, 4 Jun 2012 22:49:02 -0700, neil wrote:
>On Tue, 5 Jun 2012 13:37:15 +0800 majianpeng <majianpeng@xxxxxxxxx> wrote:
>
>> When raid takeovered, it used old personality to set task->comm[].
>> So we add judgement and set corrected task->comm[].
>>
>> Signed-off-by: majianpeng <majianpeng@xxxxxxxxx>
>> ---
>> drivers/md/md.c | 14 ++++++++++++++
>> 1 files changed, 14 insertions(+), 0 deletions(-)
>>
>> diff --git a/drivers/md/md.c b/drivers/md/md.c
>> index 1c2f904..6b3ce6e 100644
>> --- a/drivers/md/md.c
>> +++ b/drivers/md/md.c
>> @@ -6738,11 +6738,25 @@ struct md_thread *md_register_thread(void (*run) (struct mddev *), struct mddev
>> const char *name)
>> {
>> struct md_thread *thread;
>> + struct md_personality *pers = NULL;
>>
>> thread = kzalloc(sizeof(struct md_thread), GFP_KERNEL);
>> if (!thread)
>> return NULL;
>>
>> + /*
>> + *When raid takeovered,the mddev->pers was old,so the
>> + *thread_name is wrong.
>> + */
>> + if (name == NULL && mddev->level != mddev->new_level) {
>> + spin_lock(&pers_lock);
>> + list_for_each_entry(pers, &pers_list, list)
>> + if (pers->level == mddev->new_level) {
>> + name = pers->name;
>> + break;
>> + }
>> + spin_unlock(&pers_lock);
>> + }
>> init_waitqueue_head(&thread->wqueue);
>>
>> thread->run = run;
>
>No, I really don't like that. It is to focussed on fixing the particular
>symptom rather that making the code generally cleaner.
>
>Was there a reason you didn't just change the last arg of md_register_thread
>to be the level name instead of NULL??
>
No.I first wanted to corrected in which call md_register_thread.But I found it will take some codes.
So i think corrected this function may better.
>NeilBrown
>
>
--------------
majianpeng?韬{.n?????%??檩??w?{.n???{炳盯w???塄}?财??j:+v?? ? ???2??璀??摺?囤??z夸z罐?+?????w棹f
