Re: [Qemu-devel] [PATCH v5 1/1] Introduce "xen-load-devices-state"

[Changlong Xie <xiecl.fnst@cn.fujitsu.com>]

On 06/03/2016 10:13 AM, Changlong Xie wrote:
> On 06/03/2016 09:45 AM, Eric Blake wrote:
>> On 06/02/2016 07:26 PM, Changlong Xie wrote:
>>
>>>>> +
>>>>> +    ioc = qio_channel_file_new_path(filename, O_WRONLY | O_CREAT,
>>>>> 0660, errp);
>>>>
>>>> This does not look right, it looks like it's going to open the file
>>>> to write to it. You probably want O_RDONLY, also I don't think the
>>>> O_CREAT flag is needed. (and without O_WRONLY, mode can be 0 instead of
>>>> 0660.)
>>>>
>>>
>>> Yes, as you said. We should use 0_RDONLY for open(2), so mode should
>>> be 0.
>>
>> Huh?  mode doesn't affect the current fd, but DOES affect the next
>> person to open the file.  If you are truly creating the file, then a
>> mode of 0 means you won't be able to reopen it without chmod.  And if
>> you are doing O_RDONLY | O_CREAT, all you will be able to create is an
>> empty file, which is a pretty boring read.  So drop the O_CREAT, and
>> then you don't need a mode argument at all.
>>
>
> Yes, i just mean qio_channel_file_new_path(filename, O_RDONLY, 0, errp)

I just notice that, qemu specifies flag 'O_BINARY' to allow system to 
differentiate between a text file and a binary file( I guess so?). For 
backward compatibility, refer to function test_io_channel_file(), i will 
use

qio_channel_file_new_path(filename, O_RDONLY | O_BINARY, 0, errp)

here.

> here. Maybe my poor english make you confused :(
>
> Thanks
>      -Xie
>
>
>
>