[PATCH] Un-export vg_read_internal.

[PATCH] Un-export vg_read_internal.

[Petr Rockai]

Hi,

this was waiting a fairly long time for a round tuit, but here it is. This
should properly banish vg_read_internal into confinement in metadata.c. Should
be fairly obvious, I am just being paranoid here. (I made sure clvmd still
compiles this time, too.)

Yours,
   Petr.

-------------- next part --------------
A non-text attachment was scrubbed...
Name: unexport-vg_read_internal.diff
Type: text/x-diff
Size: 5412 bytes
Desc: not available
URL: <http://listman.redhat.com/archives/lvm-devel/attachments/20091102/217e6e45/attachment.bin>

[PATCH] Un-export vg_read_internal.

[Milan Broz]

On 11/02/2009 04:43 PM, Petr Rockai wrote:
> this was waiting a fairly long time for a round tuit, but here it is. This
> should properly banish vg_read_internal into confinement in metadata.c. Should
> be fairly obvious, I am just being paranoid here. (I made sure clvmd still
> compiles this time, too.)

> @@ -780,9 +780,9 @@ void lvm_do_backup(const char *vgname)
>  
> 	pthread_mutex_lock(&lvm_lock);
>  
> -	vg = vg_read_internal(cmd, vgname, NULL /*vgid*/, &consistent);
> +	vg = vg_read(cmd, vgname, NULL /*vgid*/, 0 /*flags*/);
>  
> -	if (vg && consistent)
> +	if (!vg_read_error(vg))
>  		check_current_backup(vg);

Cannot this slightly change when remote meatadata backup is run? (missing PVs?)
But it is not problem IMHO.

ACK.

Milan
--
mbroz at redhat.com

[PATCH] Un-export vg_read_internal.

[Milan Broz]

On 11/19/2009 01:11 PM, Milan Broz wrote:
> On 11/02/2009 04:43 PM, Petr Rockai wrote:
>> this was waiting a fairly long time for a round tuit, but here it is. This
>> should properly banish vg_read_internal into confinement in metadata.c. Should
>> be fairly obvious, I am just being paranoid here. (I made sure clvmd still
>> compiles this time, too.)
> 
>> @@ -780,9 +780,9 @@ void lvm_do_backup(const char *vgname)
>>  
>> 	pthread_mutex_lock(&lvm_lock);
>>  
>> -	vg = vg_read_internal(cmd, vgname, NULL /*vgid*/, &consistent);
>> +	vg = vg_read(cmd, vgname, NULL /*vgid*/, 0 /*flags*/);

post-mortem NACK

we cannot take lock inside vg_read here. it kills clvmd.

Milan