Re: [Ocfs2-devel] [Resend PATCH v3] ocfs2: fix kernel crash after mounting when journal doesn't ready

[Joseph Qi via Ocfs2-devel <ocfs2-devel@oss.oracle.com>]

On 3/28/22 10:44 AM, heming.zhao--- via Ocfs2-devel wrote:
> On 3/28/22 10:08, Joseph Qi wrote:
>>
>>
>> On 3/28/22 12:56 AM, heming.zhao@suse.com wrote:
>>> On 3/28/22 00:04, Joseph Qi wrote:
>>>>
>>>>
>>>> On 3/27/22 6:37 PM, heming.zhao@suse.com wrote:
>>>>> Hello Joseph,
>>>>>
>>>>> Could we speed up this crash fixing? I could trigger this kind of crash at least two ways.
>>>>> Do you have another alternative patch?
>>>>>
>>>> As I've described in previous mail, I don't want to fix it in the way
>>>> you proposed.
>>>
>>> I am OK with your mind, and only hope to fix this problem ASAP.
>>>
>>>> And I am afraid we need do a thoroughly check since I've found there
>>>> are other issues in ocfs2_fill_super(), e.g.
>>>>
>>>> ocfs2_fill_super
>>>>     ocfs2_initialize_super // fails in check max_slots
>>>>
>>>> Then osb->osb_mount_event is not properly initialized, and then will
>>>> crash at wake_up(&osb->osb_mount_event).
>>>>
>>>
>>> The error handling in ocfs2_fill_super() is too simple. IMO, there needs to add more
>>> kinds of "goto lable" or replace some "goto" with "return" for directly breaking
>>> init routine.
>>
>> Agree, we need handle error cases more gracefully.
>> And we can't simply call ocfs2_dismount_volume() in error handling
>> since it involves too much actions.
>> So do you have interest to take this?
>>
> 
> NO problem, Let me try.
> I will start a new mail thread for the mounting error handling story.
> 
Glad to here that. You may split these into a series.
And I will try my best to review ASAP.

Thanks,
Joseph

_______________________________________________
Ocfs2-devel mailing list
Ocfs2-devel@oss.oracle.com
https://oss.oracle.com/mailman/listinfo/ocfs2-devel