[PATCH] fs: make i_generation a u64

[PATCH] fs: make i_generation a u64

[Josef Bacik]

Btrfs stores generation numbers as 64bit numbers, which means we have to
carry around a u64 in our incore inode in addition to setting i_generation.
So convert to a u64 so btrfs can kill it's incore generation.  Thanks,

Signed-off-by: Josef Bacik <josef@redhat.com>
---
 include/linux/fs.h |    2 +-
 1 files changed, 1 insertions(+), 1 deletions(-)

diff --git a/include/linux/fs.h b/include/linux/fs.h
index 9be896d..40564e0 100644
--- a/include/linux/fs.h
+++ b/include/linux/fs.h
@@ -831,7 +831,7 @@ struct inode {
 		struct cdev		*i_cdev;
 	};
 
-	__u32			i_generation;
+	u64			i_generation;
 
 #ifdef CONFIG_FSNOTIFY
 	__u32			i_fsnotify_mask; /* all events this inode cares about */
-- 
1.7.7.6

Re: [PATCH] fs: make i_generation a u64

[Chris Mason]

On Wed, Apr 11, 2012 at 04:42:48PM -0400, Josef Bacik wrote:
> Btrfs stores generation numbers as 64bit numbers, which means we have to
> carry around a u64 in our incore inode in addition to setting i_generation.
> So convert to a u64 so btrfs can kill it's incore generation.  Thanks,
> 
> Signed-off-by: Josef Bacik <josef@redhat.com>
> ---
>  include/linux/fs.h |    2 +-
>  1 files changed, 1 insertions(+), 1 deletions(-)
> 
> diff --git a/include/linux/fs.h b/include/linux/fs.h
> index 9be896d..40564e0 100644
> --- a/include/linux/fs.h
> +++ b/include/linux/fs.h
> @@ -831,7 +831,7 @@ struct inode {
>  		struct cdev		*i_cdev;
>  	};
>  
> -	__u32			i_generation;
> +	u64			i_generation;

Why was this using __u32 instead of u32?

-chris

Re: [PATCH] fs: make i_generation a u64

[Marco Stornelli]

2012/4/11 Josef Bacik <josef@redhat.com>:
> Btrfs stores generation numbers as 64bit numbers, which means we have to
> carry around a u64 in our incore inode in addition to setting i_generation.
> So convert to a u64 so btrfs can kill it's incore generation.  Thanks,
>
> Signed-off-by: Josef Bacik <josef@redhat.com>
> ---
>  include/linux/fs.h |    2 +-
>  1 files changed, 1 insertions(+), 1 deletions(-)
>
> diff --git a/include/linux/fs.h b/include/linux/fs.h
> index 9be896d..40564e0 100644
> --- a/include/linux/fs.h
> +++ b/include/linux/fs.h
> @@ -831,7 +831,7 @@ struct inode {
>                struct cdev             *i_cdev;
>        };
>
> -       __u32                   i_generation;
> +       u64                     i_generation;
>
>  #ifdef CONFIG_FSNOTIFY
>        __u32                   i_fsnotify_mask; /* all events this inode cares about */
> --
> 1.7.7.6

This patch can have several impact on other fs. Only to do an example
you can see the code in ioctl of ext4. I haven't got study the code
but the ioctl returns a long, but on 32bit system, long means 4 bytes,
so how we can return i_generation with ioctl?

Marco
--
To unsubscribe from this list: send the line "unsubscribe linux-btrfs" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html

Re: [PATCH] fs: make i_generation a u64

[Josef Bacik]

On Thu, Apr 12, 2012 at 08:46:14AM +0200, Marco Stornelli wrote:
> 2012/4/11 Josef Bacik <josef@redhat.com>:
> > Btrfs stores generation numbers as 64bit numbers, which means we have to
> > carry around a u64 in our incore inode in addition to setting i_generation.
> > So convert to a u64 so btrfs can kill it's incore generation.  Thanks,
> >
> > Signed-off-by: Josef Bacik <josef@redhat.com>
> > ---
> >  include/linux/fs.h |    2 +-
> >  1 files changed, 1 insertions(+), 1 deletions(-)
> >
> > diff --git a/include/linux/fs.h b/include/linux/fs.h
> > index 9be896d..40564e0 100644
> > --- a/include/linux/fs.h
> > +++ b/include/linux/fs.h
> > @@ -831,7 +831,7 @@ struct inode {
> >                struct cdev             *i_cdev;
> >        };
> >
> > -       __u32                   i_generation;
> > +       u64                     i_generation;
> >
> >  #ifdef CONFIG_FSNOTIFY
> >        __u32                   i_fsnotify_mask; /* all events this inode cares about */
> > --
> > 1.7.7.6
> 
> This patch can have several impact on other fs. Only to do an example
> you can see the code in ioctl of ext4. I haven't got study the code
> but the ioctl returns a long, but on 32bit system, long means 4 bytes,
> so how we can return i_generation with ioctl?
> 

Thanks for pointing this out, I will fix ext4 up and look to see if anybody else
does something similar,

Josef
--
To unsubscribe from this list: send the line "unsubscribe linux-btrfs" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html

Re: [PATCH] fs: make i_generation a u64

[Josef Bacik]

On Thu, Apr 12, 2012 at 08:46:14AM +0200, Marco Stornelli wrote:
> 2012/4/11 Josef Bacik <josef@redhat.com>:
> > Btrfs stores generation numbers as 64bit numbers, which means we have to
> > carry around a u64 in our incore inode in addition to setting i_generation.
> > So convert to a u64 so btrfs can kill it's incore generation.  Thanks,
> >
> > Signed-off-by: Josef Bacik <josef@redhat.com>
> > ---
> >  include/linux/fs.h |    2 +-
> >  1 files changed, 1 insertions(+), 1 deletions(-)
> >
> > diff --git a/include/linux/fs.h b/include/linux/fs.h
> > index 9be896d..40564e0 100644
> > --- a/include/linux/fs.h
> > +++ b/include/linux/fs.h
> > @@ -831,7 +831,7 @@ struct inode {
> >                struct cdev             *i_cdev;
> >        };
> >
> > -       __u32                   i_generation;
> > +       u64                     i_generation;
> >
> >  #ifdef CONFIG_FSNOTIFY
> >        __u32                   i_fsnotify_mask; /* all events this inode cares about */
> > --
> > 1.7.7.6
> 
> This patch can have several impact on other fs. Only to do an example
> you can see the code in ioctl of ext4. I haven't got study the code
> but the ioctl returns a long, but on 32bit system, long means 4 bytes,
> so how we can return i_generation with ioctl?

So looking through everybody I'd have to convert a bunch of cpu_to_le32 to
(u32)cpu_to_le64 and some other such crap which I'm not terribly interested in
doing, I don't care _that_ much about saving 8 bytes, so I'll just drop this for
now.  Thanks,

Josef
--
To unsubscribe from this list: send the line "unsubscribe linux-btrfs" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html

Re: [PATCH] fs: make i_generation a u64

[Ted Ts'o]

On Wed, Apr 11, 2012 at 04:42:48PM -0400, Josef Bacik wrote:
> Btrfs stores generation numbers as 64bit numbers, which means we have to
> carry around a u64 in our incore inode in addition to setting i_generation.
> So convert to a u64 so btrfs can kill it's incore generation.  Thanks,
> 
> Signed-off-by: Josef Bacik <josef@redhat.com>

Why is btrfs using a 64-bit generation number, out of curiosity?  The
only user of the inode generation number as far as I can tell is NFS,
and even NFSv4 is using a 32-bit generation number....

						- Ted

Re: [PATCH] fs: make i_generation a u64

[Josef Bacik]

On Thu, Apr 12, 2012 at 03:42:17PM -0400, Ted Ts'o wrote:
> On Wed, Apr 11, 2012 at 04:42:48PM -0400, Josef Bacik wrote:
> > Btrfs stores generation numbers as 64bit numbers, which means we have to
> > carry around a u64 in our incore inode in addition to setting i_generation.
> > So convert to a u64 so btrfs can kill it's incore generation.  Thanks,
> > 
> > Signed-off-by: Josef Bacik <josef@redhat.com>
> 
> Why is btrfs using a 64-bit generation number, out of curiosity?  The
> only user of the inode generation number as far as I can tell is NFS,
> and even NFSv4 is using a 32-bit generation number....
> 

It's just tied to our transaction id #'s which are 64bit, no super awesome
reason or anything.  Thanks,

Josef