On 5/8/20 6:36 AM, Matthew Wilcox wrote:
> On Thu, May 07, 2020 at 05:02:42PM -0700, Andrew Morton wrote:
>> Here's how I resolved things. Please check?
>>
>> static struct kern_ipc_perm *sysvipc_find_ipc(struct ipc_ids *ids, loff_t
>> pos,
>> loff_t *new_pos)
>> {
>> unsigned long index = pos;
>> struct kern_ipc_perm *ipc;
>>
>> rcu_read_lock();
>> ipc = xa_find(&ids->ipcs, &index, ULONG_MAX, XA_PRESENT);
>> if (ipc)
>> ipc_lock_object(ipc);
>> else
>> rcu_read_unlock();
>> *new_pos = pos + 1;
>> return ipc;
>> }
>
> Surely that should be '*new_pos = index + 1'? Or did I misunderstand
> the reasoning behind the other patch?
I'm not sure however it looks like xa_find() can return index < pos
xa_find in our case will call xas_find_marked() that have following description
* If no marked entry is found and the array is smaller than @max, @xas is
* set to the bounds state and xas->xa_index is set to the smallest index
* not yet in the array. This allows @xas to be immediately passed to
* xas_store().
Matthew, could you please clarify this question?
Thank you,
Vasily Averin
