On Thu, Feb 21, 2013 at 03:56:51PM +0100, Lukáš Czerner wrote:
> On Thu, 21 Feb 2013, Zheng Liu wrote:
>
> ..snip..
>
> > > > > > /*
> > > > > > * All of the blocks before first_data_block are overhead
> > > > > > */
> > > > > > - overhead = EXT4_B2C(sbi, le32_to_cpu(es->s_first_data_block));
> > > > > > + overhead = EXT4_NUM_B2C(sbi, le32_to_cpu(es->s_first_data_block));
> > > >
> > > > ...except this. I do not think this is right because we do not skip
> > > > the first cluster right ? We're still using it, but we can never use
> > > > the block before es->s_first_data_block. Please correct me if I am
> > > > wrong.
> >
> > Yes, I think you are right.
> >
> > >
> > > moreover we do not allow bigalloc file system with block size < 4k.
> >
> > No, we allow user to use bigalloc with block size < 4k, such as:
> >
> > mkfs.ext4 -b 1024 -C 4096 -O bigalloc ${dev}
> >
> > This command formats a bigalloc filesystem with blocksize = 1k and
> > clustersize = 4k, at least in e2fsprogs 1.42.7 it works well.
> >
>
> Ok, i was pretty sure that we do not allow that, it's good to know.
> Also, does it make any sense ? I do not think so, and I would really
> consider the fact that we allow that as a bug. We should not allow
> that otherwise it unnecessarily extending the test matrix.
>
> What people think about restricting bigalloc _only_ for 4k block
> size file systems ?
I agree with you that we should forbid user to use bigalloc feature with
block size = 1k or 2k because I guess no one really use it, at least in
Taobao we always use bigalloc feature with block size = 4k.
FWIW, recently I am considering whether we could remove 'data=journal'
and 'data=writeback' mode. 'data=journal' mode hurts performance
dramatically. Further, 'data=writeback' seems also useless, especially
after we have 'no journal' feature. TBH, these modes are lack of tests.
Maybe this is a crazy idea in my mind.
Regards,
- Zheng
--
To unsubscribe from this list: send the line "unsubscribe linux-ext4" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at http://vger.kernel.org/majordomo-info.html
