different block sizes and badblocks

different block sizes and badblocks

[Newsmail]

I have two questions: I see that in reiserfsprogs 3.x.1c-pre5 we have the 
option to make filesystem with different block sizes than the default 4096 
bytes. what does this mean? I can choose whatever blocksize I want? or it 
has to be a multiple of 4096? actually as I heard if I have a filesystem 
that has big files it is better to have big blocksizes, so the HDs have to 
seek less. actually I have a fileserver having mostly 7-10 meg files, that 
means that we can consider these as large files, and using lets say 256k 
block sizes would be more efficient than 4k block sizes right? and as 
reiserfs has the 'tail' option, I wont loose a single byte because of this 
big blocksize right? could somebody clear my mind about this?

the other question is about badblocks. we have badblock support in 
reiserfsprogs 3.x.1c-pre5, but its disabled by default. Oleg, could you 
give me some infos about it, how can I reenable it, or how does it work 
actually? it would be really nice :)

regards,
greg

Re: different block sizes and badblocks

[Oleg Drokin]

Hello!

On Thu, Jun 20, 2002 at 02:05:11PM +0100, Newsmail wrote:
> I have two questions: I see that in reiserfsprogs 3.x.1c-pre5 we have the 
> option to make filesystem with different block sizes than the default 4096 
> bytes. what does this mean? I can choose whatever blocksize I want? or it 
> has to be a multiple of 4096? actually as I heard if I have a filesystem 

Right now kernel lacks support for blocksize other than 4096 bytes.

> the other question is about badblocks. we have badblock support in 
> reiserfsprogs 3.x.1c-pre5, but its disabled by default. Oleg, could you 
> give me some infos about it, how can I reenable it, or how does it work 
> actually? it would be really nice :)

Right now badblocks support is disabled on purpose.
It conflicts with savelinks and we have not decided how to solve that yet.
But we will announce badblocks support once we think it is ready.

Bye,
    Oleg

RE: different block sizes and badblocks

[berthiaume_wayne]

	How does reiserfs normally handle a bad block encountered on a read
at this time?

-----Original Message-----
From: Oleg Drokin [mailto:green@namesys.com]
Sent: Thursday, June 20, 2002 12:22 PM
To: Newsmail
Cc: reiserfs-list@namesys.com
Subject: Re: [reiserfs-list] different block sizes and badblocks


Hello!

On Thu, Jun 20, 2002 at 02:05:11PM +0100, Newsmail wrote:
> I have two questions: I see that in reiserfsprogs 3.x.1c-pre5 we have the 
> option to make filesystem with different block sizes than the default 4096

> bytes. what does this mean? I can choose whatever blocksize I want? or it 
> has to be a multiple of 4096? actually as I heard if I have a filesystem 

Right now kernel lacks support for blocksize other than 4096 bytes.

> the other question is about badblocks. we have badblock support in 
> reiserfsprogs 3.x.1c-pre5, but its disabled by default. Oleg, could you 
> give me some infos about it, how can I reenable it, or how does it work 
> actually? it would be really nice :)

Right now badblocks support is disabled on purpose.
It conflicts with savelinks and we have not decided how to solve that yet.
But we will announce badblocks support once we think it is ready.

Bye,
    Oleg

Re: different block sizes and badblocks

[Oleg Drokin]

Hello!

On Fri, Jun 21, 2002 at 10:35:25AM -0400, berthiaume_wayne@emc.com wrote:

> 	How does reiserfs normally handle a bad block encountered on a read
> at this time?

I/O errro on journal read or superblock read or root node read are threatened
like mount failures.
I/O error on read in is usually handled to userspace as -EIO.

Bye,
    Oleg 

Re: different block sizes and badblocks

[Kuba Ober]

> > the other question is about badblocks. we have badblock support in
> > reiserfsprogs 3.x.1c-pre5, but its disabled by default. Oleg, could you
> > give me some infos about it, how can I reenable it, or how does it work
> > actually? it would be really nice :)
>
> Right now badblocks support is disabled on purpose.
> It conflicts with savelinks and we have not decided how to solve that yet.
> But we will announce badblocks support once we think it is ready.

With many modern IDE drives, the bad blocks list can potentially vary between 
disk power-ups, and usually it dynamically changes while you access the disk. 
Most modern drives should be considered dead when you get bad sectors. For 
your own good, throw them away ASAP, or get to know smartctl if your drive 
supports S.M.A.R.T.

Again, for your own good, either low-level reformat the drive using a decent 
tool (smartctl uses drive's own SMART programs, or buy SpinRite 
http://www.grc.com).

Cheers, Kuba