lkml.org 
[lkml]   [2011]   [Nov]   [4]   [last100]   RSS Feed
Views: [wrap][no wrap]   [headers]  [forward] 
 
Messages in this thread
/
Date
From
SubjectRe: [GIT PULL] Squashfs updates for 3.2
NamJae Jeon wrote:

>
> I already posted this patch before ([PATCH] squashfs : devblksize set
> to 4KB intead of BLOCK_SIZE(1KB).).

No you didn't. You posted a patch that simply unconditionally changed
the block size from 1K -> 4K.

https://lkml.org/lkml/2011/9/18/66

This is an unacceptable change, Squashfs is used on many devices not only
NAND, and the default value of 1K is optimal for these other devices, and
should not be changed.

Second, if you are going to change long-term existing behaviour you should
always allow users to "buy-in" to the change, rather than surprising them
with new unexpected behaviour.

> It is similar with my patch except option.

The option *is* the patch.

> Have you ever seen this patch ? I didn't response about this patch from you.

Since 2008 (and probably before) I have had reports that a 1K block size was
causing performance issues on NAND

http://old.nabble.com/Default-FS-block-size-td15423970.html

However, I chose to do nothing at that time because the results were
inconclusive.

The impetus for moving to a 1K block on NAND was due to the development
of the UBIBLK driver for NAND earlier this year

http://lists.infradead.org/pipermail/linux-mtd/2011-June/036595.html

where the 1K dev block behaviour of Squashfs was discovered to be the
reason (in the early V1 driver referenced above) why Squashfs filesystems
worked, but ext2/3 and vfat filesystems did not.

Your patch was merely the 3rd or 4th unacceptable patch I have
received changing the block size unconditionally.

The month before your patch I received this truly horrible patch, which
though it is extremely long, does nothing more than change the max dev
block size to 4K. I dropped that patch too.

Phillip


\
 
 \ /
  Last update: 2011-11-04 17:27    [W:0.050 / U:0.432 seconds]
©2003-2020 Jasper Spaans|hosted at Digital Ocean and TransIP|Read the blog|Advertise on this site