Home » Mailing list archives » tux3 » 2008 » December » 12

[Tux3] [PATCH] fix mtime support by adding tux3_setattr

Previous thread: Re: [Tux3] Feature interaction between multiple volumes and atomic update by Matthew Dillon on Friday, August 29, 2008 - 11:31 pm. (6 messages)

Next thread: [Tux3] Tux3 command interpreter by Daniel Phillips on Monday, September 1, 2008 - 4:15 pm. (4 messages)
[view original message]
From: Shapor Naghibzadeh <shapor@...>
To: <tux3@...>
Subject: [Tux3] [PATCH] fix mtime support by adding tux3_setattr
Date: Thursday, December 11, 2008 - 10:13 pm

This patch implements setattr fileop for Tux3.  Tux3 now stores the

mtime properly and persistently, which I tested under UML.


I'm updating the MTIME_BIT in tuxnode->present every time, before

calling inode_setattr, which marks the inode dirty.  I considered

twiddling the MTIME_BIT afterwards, but I wasn't sure if that was safe

since the inode was already set dirty (and could possibly get

flushed?), so I was forced to check if the mtime/ctime was changing.


Thanks to Daniel for the excellent instructions:

http://mailman.tux3.org/pipermail/tux3/2008-November/000380.html

Now we just need a git repo that works :)


diff -rup linux-2.6.26.5.orig/fs/tux3/inode.c linux-2.6.26.5/fs/tux3/inode.c

--- linux-2.6.26.5.orig/fs/tux3/inode.c 2008-12-11 17:42:16.000000000 -0800

+++ linux-2.6.26.5/fs/tux3/inode.c      2008-12-11 17:50:29.000000000 -0800

@@ -218,7 +218,7 @@ static const struct file_operations tux_

 static const struct inode_operations tux_file_iops = {

 //     .truncate       = ext4_truncate,

 //     .permission     = ext4_permission,

-//     .setattr        = ext4_setattr,

+       .setattr        = tux3_setattr,

 //     .getattr        = ext4_getattr

 #ifdef CONFIG_EXT4DEV_FS_XATTR

 //     .setxattr       = generic_setxattr,

@@ -324,4 +324,22 @@ struct inode *tux3_iget(struct super_blo

        unlock_new_inode(inode);

        return inode;

 }

+

+int tux3_setattr(struct dentry *dentry, struct iattr *iattr)

+{

+       struct inode *inode = dentry->d_inode;

+       tuxnode_t *tuxnode = tux_inode(inode);

+       int error;

+

+       error = inode_change_ok(inode, iattr);

+       if (error)

+               return error;

+       if (timespec_equal(iattr->ia_valid & ATTR_MTIME ?

&iattr->ia_mtime : &inode->i_mtime,

+                          iattr->ia_valid & ATTR_CTIME ?

&iattr->ia_ctime : &inode->i_ctime))

+               tuxnode->present &= ~MTIME_BIT;

+       else

+               tuxnode->present |= MTIME_BIT;

+

+...
[ message continues ]
[view original message]
From: Michael Keulkeul <kriptomik@...>
To: <Tux3@...>
Subject: [Tux3] Hi all, I'm a weak coder but you've my full support
Date: Saturday, August 30, 2008 - 8:13 pm

Hi all !


I'm a modest Linux lover which has a lot of interest in file systems

and storage. Nobody is perfect, You must know I'm working at netapp,

but I hope this is going to change soon (Paris suxx, I must leave this

town)


I've seen a query for easy NVRAM not limited with write cycles, I may

have a stupid trick that you could use : Use a cheap laptop (battery

backed of course, 1G ethernet give better latencies) and export an

iSCSI or AoE Lun using a ramdrive as backend. It's not as fast as

memory, but it's as fast as flash and does not die that quickly on

massive overwrites. Nice other thing is that multiple initiators can

use it, so it's not as expensive as it appears.


Let me know if I can help anywhere (I can translate into french for

exemple) I'll do my best to support a project like this.


Michael B.


_______________________________________________

Tux3 mailing list

Tux3@tux3.org

http://tux3.org/cgi-bin/mailman/listinfo/tux3
[view original message]
From: Daniel Phillips <phillips@...>
To: <tux3@...>
Subject: Re: [Tux3] [PATCH] fix mtime support by adding tux3_setattr
Date: Thursday, December 11, 2008 - 11:49 pm

Pulled from your hg repo.  A proper git tree on kernel.org is coming

soon.  Checking in to hg always works.  There was a one liner bash

script floating around to symlink the user/kernel files from the

kernel tree.  Hard links would be even better.


I only tested that it builds, not that it works :-)


Anybody?


Regards,


Daniel


_______________________________________________

Tux3 mailing list

Tux3@tux3.org

http://mailman.tux3.org/cgi-bin/mailman/listinfo/tux3
[view original message]
From: Daniel Phillips <phillips@...>
To: <tux3@...>
Cc: Michael Keulkeul <kriptomik@...>
Subject: Re: [Tux3] Hi all, I'm a weak coder but you've my full support
Date: Sunday, August 31, 2008 - 2:30 am

Hi Michael,


A weak coder becomes a strong coder by reading code and writing code -

every day, for fun.


As it happened, my initial demonstration of Ramback (see "faster than

a speeding bullet") was done on a laptop.  I did not fail to notice

that the battery essentially turns a laptop into NVRAM, as long as you

have software like Ramback that can get the memory onto disk before

the battery runs out.


There is always a way to help with a project, just keep your eyes open

and watch what is going on.  Try cloning the mercurial reposititory

and see if that works for you, then run "make" and "make tests", and

see if the tests run ok.  Check out the output from the tests and see

if you can make sense of what is going on.  If not, come back here and

ask.


Try reading the code and figuring out how it works.  If you don't see

why it does something a particular way, ask.  That means we probably

should have a comment in the code there.  That kind of feedback is

helpful.


Regards,


Daniel


_______________________________________________

Tux3 mailing list

Tux3@tux3.org

http://tux3.org/cgi-bin/mailman/listinfo/tux3
[view original message]
From: Michael Keulkeul <kriptomik@...>
To: Daniel Phillips <phillips@...>
Cc: <tux3@...>
Subject: Re: [Tux3] Hi all, I'm a weak coder but you've my full support
Date: Sunday, August 31, 2008 - 6:24 am

Hi !


Ok so I'll start with cloning the mercurial repository, and alway keep

an eye on this project !


I started digging a bit into the code but I must confess I feel like a

newborn just poping into reality... Bah it makes my brain work a lot

which is essentially a good thing !


Good luck, anyway ! I'll to catch up things, starting with

understanding your notes.


Best reagards


Michael


_______________________________________________

Tux3 mailing list

Tux3@tux3.org

http://tux3.org/cgi-bin/mailman/listinfo/tux3
Previous thread: Re: [Tux3] Feature interaction between multiple volumes and atomic update by Matthew Dillon on Friday, August 29, 2008 - 11:31 pm. (6 messages)

Next thread: [Tux3] Tux3 command interpreter by Daniel Phillips on Monday, September 1, 2008 - 4:15 pm. (4 messages)