Patchwork chardev: Get filename for new qapi backend

login
register
mail settings
Submitter Lei Li
Date May 15, 2013, 3:52 a.m.
Message ID <1368589953-22455-1-git-send-email-lilei@linux.vnet.ibm.com>
Download mbox | patch
Permalink /patch/243896/
State New
Headers show

Comments

Lei Li - May 15, 2013, 3:52 a.m.
This patch add the filename when the new qapi backend init from opts.

Commit 2c5f488293c7d0cd095635c74157c2526e2c4947 add support for
qapi-based chardev initialization, but miss the filename of the
char device as below:

(qemu) info chardev
parallel0: filename=(null)
serial0: filename=(null)
compat_monitor0: filename=(null)


Signed-off-by: Lei Li <lilei@linux.vnet.ibm.com>
---
 qemu-char.c |    2 ++
 1 files changed, 2 insertions(+), 0 deletions(-)
Gerd Hoffmann - May 15, 2013, 5:39 a.m.
On 05/15/13 05:52, Lei Li wrote:
> This patch add the filename when the new qapi backend init from opts.
> 
> Commit 2c5f488293c7d0cd095635c74157c2526e2c4947 add support for
> qapi-based chardev initialization, but miss the filename of the
> char device as below:
> 
> (qemu) info chardev
> parallel0: filename=(null)
> serial0: filename=(null)
> compat_monitor0: filename=(null)

> @@ -3276,6 +3276,7 @@ CharDriverState *qemu_chr_new_from_opts(QemuOpts *opts,
>          ChardevReturn *ret = NULL;
>          const char *id = qemu_opts_id(opts);
>          const char *bid = NULL;
> +        char *filename = g_strdup(qemu_opt_get(opts, "backend"));
>  
>          if (qemu_opt_get_bool(opts, "mux", 0)) {
>              bid = g_strdup_printf("%s-base", id);
> @@ -3308,6 +3309,7 @@ CharDriverState *qemu_chr_new_from_opts(QemuOpts *opts,
>          }
>  
>          chr = qemu_chr_find(id);
> +        chr->filename = filename;

That should happen in qmp_chardev_add(), so filename is set consistently
for every chardev no matter how it gets created.

Take care to not overwrite filename, some functions
(qmp_chardev_open_socket for example) do set chr->filename already.

Maybe it's better to put that into the individual qmp_chardev_open_*
functions anyway.

cheers,
  Gerd
Lei Li - May 15, 2013, 7:03 a.m.
On 05/15/2013 01:39 PM, Gerd Hoffmann wrote:
> On 05/15/13 05:52, Lei Li wrote:
>> This patch add the filename when the new qapi backend init from opts.
>>
>> Commit 2c5f488293c7d0cd095635c74157c2526e2c4947 add support for
>> qapi-based chardev initialization, but miss the filename of the
>> char device as below:
>>
>> (qemu) info chardev
>> parallel0: filename=(null)
>> serial0: filename=(null)
>> compat_monitor0: filename=(null)
>> @@ -3276,6 +3276,7 @@ CharDriverState *qemu_chr_new_from_opts(QemuOpts *opts,
>>           ChardevReturn *ret = NULL;
>>           const char *id = qemu_opts_id(opts);
>>           const char *bid = NULL;
>> +        char *filename = g_strdup(qemu_opt_get(opts, "backend"));
>>   
>>           if (qemu_opt_get_bool(opts, "mux", 0)) {
>>               bid = g_strdup_printf("%s-base", id);
>> @@ -3308,6 +3309,7 @@ CharDriverState *qemu_chr_new_from_opts(QemuOpts *opts,
>>           }
>>   
>>           chr = qemu_chr_find(id);
>> +        chr->filename = filename;
> That should happen in qmp_chardev_add(), so filename is set consistently
> for every chardev no matter how it gets created.

Yeah, I was hesitant to add it here for that it does not looks very logical/...
/But I am not sure about the best place for it..
//  <http://www.google.com.hk/search?newwindow=1&safe=strict&biw=1280&bih=646&q=i+am+a+little+Hesitant&spell=1&sa=X&ei=OiaTUaGALpPI4AOOoIDoCQ&ved=0CCkQvwUoAA>

>
> Take care to not overwrite filename, some functions
> (qmp_chardev_open_socket for example) do set chr->filename already.

Yes, for the char device pipe, pty, socket, they set chr->filename in their
own open functions. But I think the overwrite will not happen because
it will not get "backend" by qemu_opt_get() for these chardevs.

>
> Maybe it's better to put that into the individual qmp_chardev_open_*
> functions anyway.

OK, I'll have a try and see if everyone like it, I don't have better idea..

>
> cheers,
>    Gerd
>
>
Lei Li - May 15, 2013, 7:26 a.m.
On 05/15/2013 01:39 PM, Gerd Hoffmann wrote:
> On 05/15/13 05:52, Lei Li wrote:
>> This patch add the filename when the new qapi backend init from opts.
>>
>> Commit 2c5f488293c7d0cd095635c74157c2526e2c4947 add support for
>> qapi-based chardev initialization, but miss the filename of the
>> char device as below:
>>
>> (qemu) info chardev
>> parallel0: filename=(null)
>> serial0: filename=(null)
>> compat_monitor0: filename=(null)
>> @@ -3276,6 +3276,7 @@ CharDriverState *qemu_chr_new_from_opts(QemuOpts *opts,
>>           ChardevReturn *ret = NULL;
>>           const char *id = qemu_opts_id(opts);
>>           const char *bid = NULL;
>> +        char *filename = g_strdup(qemu_opt_get(opts, "backend"));
>>   
>>           if (qemu_opt_get_bool(opts, "mux", 0)) {
>>               bid = g_strdup_printf("%s-base", id);
>> @@ -3308,6 +3309,7 @@ CharDriverState *qemu_chr_new_from_opts(QemuOpts *opts,
>>           }
>>   
>>           chr = qemu_chr_find(id);
>> +        chr->filename = filename;
> That should happen in qmp_chardev_add(), so filename is set consistently
> for every chardev no matter how it gets created.

Yeah, I was hesitant to add it here for that it does not looks very logical...
But I am not sure about the best place for it..

> Take care to not overwrite filename, some functions
> (qmp_chardev_open_socket for example) do set chr->filename already.

Yes, for the char device pipe, pty, socket, they set chr->filename in their
own open functions. But I think the overwrite will not happen because
it will not get "backend" by qemu_opt_get() for these chardevs.

>
> Maybe it's better to put that into the individual qmp_chardev_open_*
> functions anyway.

OK, I'll have a try and see if everyone like it, I don't have better idea.

>
> cheers,
>    Gerd
>
>
>
Lei Li - May 17, 2013, 8:35 a.m.
On 05/15/2013 03:26 PM, Lei Li wrote:
> On 05/15/2013 01:39 PM, Gerd Hoffmann wrote:
>> On 05/15/13 05:52, Lei Li wrote:
>>> This patch add the filename when the new qapi backend init from opts.
>>>
>>> Commit 2c5f488293c7d0cd095635c74157c2526e2c4947 add support for
>>> qapi-based chardev initialization, but miss the filename of the
>>> char device as below:
>>>
>>> (qemu) info chardev
>>> parallel0: filename=(null)
>>> serial0: filename=(null)
>>> compat_monitor0: filename=(null)
>>> @@ -3276,6 +3276,7 @@ CharDriverState 
>>> *qemu_chr_new_from_opts(QemuOpts *opts,
>>>           ChardevReturn *ret = NULL;
>>>           const char *id = qemu_opts_id(opts);
>>>           const char *bid = NULL;
>>> +        char *filename = g_strdup(qemu_opt_get(opts, "backend"));
>>>             if (qemu_opt_get_bool(opts, "mux", 0)) {
>>>               bid = g_strdup_printf("%s-base", id);
>>> @@ -3308,6 +3309,7 @@ CharDriverState 
>>> *qemu_chr_new_from_opts(QemuOpts *opts,
>>>           }
>>>             chr = qemu_chr_find(id);
>>> +        chr->filename = filename;
>> That should happen in qmp_chardev_add(), so filename is set consistently
>> for every chardev no matter how it gets created.
>
> Yeah, I was hesitant to add it here for that it does not looks very 
> logical...
> But I am not sure about the best place for it..
>
>> Take care to not overwrite filename, some functions
>> (qmp_chardev_open_socket for example) do set chr->filename already.
>
> Yes, for the char device pipe, pty, socket, they set chr->filename in 
> their
> own open functions. But I think the overwrite will not happen because
> it will not get "backend" by qemu_opt_get() for these chardevs.
>
>>
>> Maybe it's better to put that into the individual qmp_chardev_open_*
>> functions anyway.
>
> OK, I'll have a try and see if everyone like it, I don't have better 
> idea.
>
Hi Gerd,

As I tried to add the filename into individual open functions, I think it might
not be a good idea. For example,

chardev backend file and console have the same open function qemu_chr_open_win_file().
Then we may need to add flag to distinguish them.

And we already parse every backend into QemuOpts, if we want to put the filename setting
job into open functions (by hardcoding) seems a little overdone?

So why not just set it in qemu_chr_new_from_opts(), by qemu_opt_get() directly?

>>
>> cheers,
>>    Gerd
>>
>>
>>
>
>
Anthony Liguori - May 22, 2013, 10:59 p.m.
Applied.  Thanks.

Regards,

Anthony Liguori

Patch

diff --git a/qemu-char.c b/qemu-char.c
index 30a2ddf..a9f618a 100644
--- a/qemu-char.c
+++ b/qemu-char.c
@@ -3276,6 +3276,7 @@  CharDriverState *qemu_chr_new_from_opts(QemuOpts *opts,
         ChardevReturn *ret = NULL;
         const char *id = qemu_opts_id(opts);
         const char *bid = NULL;
+        char *filename = g_strdup(qemu_opt_get(opts, "backend"));
 
         if (qemu_opt_get_bool(opts, "mux", 0)) {
             bid = g_strdup_printf("%s-base", id);
@@ -3308,6 +3309,7 @@  CharDriverState *qemu_chr_new_from_opts(QemuOpts *opts,
         }
 
         chr = qemu_chr_find(id);
+        chr->filename = filename;
 
     qapi_out:
         qapi_free_ChardevBackend(backend);