Minimum/optimal sector_size for jffs2

Minimum/optimal sector_size for jffs2

[Peter Menzebach]

Hi,
since my grep through the source code was not successful:
Is there a minimum possible jffs2_sb_info.sector_size and somewhere a 
definition?

Is there somewhat like an optimal sector_size?

I have here a device (dataflash), which has a very small
erase/write page size (1056 bytes).

Best regards
Peter
-- 
Peter Menzebach
Menzebach und Wolff IT-Consulting GbR
Phone +49 751 355 387 1

Re: Minimum/optimal sector_size for jffs2

[Artem B. Bityutskiy]

Peter Menzebach wrote:
> Hi,
> since my grep through the source code was not successful:
> Is there a minimum possible jffs2_sb_info.sector_size and somewhere a 
> definition?
There is probably no minimum value, but not because it doesn't matter, 
just because nobody cared adding a check.

> Is there somewhat like an optimal sector_size?
Not sure about optimal, probably yes. Consider the following aspects:

1. There is an array (c->blocks[]) with one element per eraseblock. The 
smaller is yur eraseblock - the larger is the array.

2. The eraseblock size cannot be less then PAGE_SIZE + sizeof(struct 
jffs2_raw_inode). PAGE_SIZE is mostly 4096 bytes.

3. The space at the end of eraseblock is not used if there are less then 
JFFS2_MIN_DATA_LEN bytes, or if the node beinf written does not fit that 
space. So, the smaller is the eraseblock, the more space is wasted.

> I have here a device (dataflash), which has a very small
> erase/write page size (1056 bytes).
Err, AFAIR, I explained why you cannot use 1056 bytes eraseblocks... It 
is simply too small.

-- 
Best Regards,
Artem B. Bityuckiy,
St.-Petersburg, Russia.

Re: Minimum/optimal sector_size for jffs2

[Peter Menzebach]

Artem B. Bityutskiy wrote:
> Peter Menzebach wrote:
> 
>> Hi,
>> since my grep through the source code was not successful:
>> Is there a minimum possible jffs2_sb_info.sector_size and somewhere a 
>> definition?
> 
> There is probably no minimum value, but not because it doesn't matter, 
> just because nobody cared adding a check.
> 
>> Is there somewhat like an optimal sector_size?
> 
> Not sure about optimal, probably yes. Consider the following aspects:
> 
> 1. There is an array (c->blocks[]) with one element per eraseblock. The 
> smaller is yur eraseblock - the larger is the array.
> 
> 2. The eraseblock size cannot be less then PAGE_SIZE + sizeof(struct 
> jffs2_raw_inode). PAGE_SIZE is mostly 4096 bytes.
> 
Sorry,
I was not precise in terms. When I talked about erasesize, I mean 
mtd_info.erasesize.

As far as I see in the code, for jffs2 the erase vloack size (used by 
jeb) is jffs2_sb_info.sector_size.

So, in these terms, the jffs2 sector_size has a minimum:
jffs2 sector_size >= PAGE_SIZE + sizeof(struct jffs2_raw_inode)

That's the information, I wanted to confirm.

>> I have here a device (dataflash), which has a very small
>> erase/write page size (1056 bytes).
> 
> Err, AFAIR, I explained why you cannot use 1056 bytes eraseblocks... It 
> is simply too small.
See explanation above, sorry for the confusion ;) .


Best regards
Peter

-- 
Peter Menzebach
Menzebach und Wolff IT-Consulting GbR
Phone +49 751 355 387 1

Re: Minimum/optimal sector_size for jffs2

[Jörn Engel]

On Fri, 23 September 2005 13:30:33 +0200, Peter Menzebach wrote:
>
> I was not precise in terms. When I talked about erasesize, I mean 
> mtd_info.erasesize.
> 
> As far as I see in the code, for jffs2 the erase vloack size (used by 
> jeb) is jffs2_sb_info.sector_size.

Hmm.  "sector_size" is a confusing name.  It reminds people of a disk
sector, which is semantically similar to the writesize (or pagesize,
if NAND people prefer).  But JFFS2 uses sector_size as erasesize.

Oh well, just another one.

Jörn

-- 
Don't patch bad code, rewrite it.
-- Kernigham and Pike, according to Rusty

Re: Minimum/optimal sector_size for jffs2

[Artem B. Bityutskiy]

Jörn Engel wrote:
> Hmm.  "sector_size" is a confusing name.  It reminds people of a disk
> sector, which is semantically similar to the writesize (or pagesize,
> if NAND people prefer).  But JFFS2 uses sector_size as erasesize.
> 
I would not mind if one rename that field to erasesize.

-- 
Best Regards,
Artem B. Bityuckiy,
St.-Petersburg, Russia.

Re: Minimum/optimal sector_size for jffs2

[Artem B. Bityutskiy]

Hi,

I wonder, did you finally heal your JFFS2/Dataflash ? If yes, send your 
patch, I assume we should commit it before JFFS2 has gone to mainstream

-- 
Best Regards,
Artem B. Bityuckiy,
St.-Petersburg, Russia.

Re: Minimum/optimal sector_size for jffs2

[Peter Menzebach]

Artem B. Bityutskiy wrote:
> Hi,
> 
> I wonder, did you finally heal your JFFS2/Dataflash ? If yes, send your 
> patch, I assume we should commit it before JFFS2 has gone to mainstream
> 
Yes,
dataflash is successfully healed. Thanks for your help.
I am on the road this week. I will send it friday this week...

Best regards
Peter

-- 
Peter Menzebach
Menzebach und Wolff IT-Consulting GbR
Phone +49 751 355 387 1