LFS/non-LFS error handling

LFS/non-LFS error handling

[David Chow]

Dear all,

If my fs don't support LFS (files larger than 4GB), what should I set in 
my sb->s_maxbytes? In kernel 2.4.2 , I have experieced if not set the 
s_maxbytes generic_file_write() will fail, but for later version of 
kernel I simply leave it unset, so far I don't have any problem. If it 
is wrong, what should s_maxbytes=? . Since this behaviour changes across 
kernel version, is there a standard implemenation?

Since it still work for me for not setting the s_maxbytes, it seems to 
me there is a potential risk of data corruption at limit points, because 
don't know who to handle the over flow situation, does this handle by 
VFS? or I have to take care myself? Thanks.

regards,

David

Re: LFS/non-LFS error handling

[Ion Badulescu]

On Sat, 13 Apr 2002 00:46:14 +0800, David Chow <davidchow@shaolinmicro.com> wrote:
> Dear all,
> 
> If my fs don't support LFS (files larger than 4GB), what should I set in 
> my sb->s_maxbytes? 

MAX_NON_LFS.

> In kernel 2.4.2 , I have experieced if not set the 
> s_maxbytes generic_file_write() will fail, but for later version of 
> kernel I simply leave it unset, so far I don't have any problem. If it 
> is wrong, what should s_maxbytes=? . Since this behaviour changes across 
> kernel version, is there a standard implemenation?
> 
> Since it still work for me for not setting the s_maxbytes, it seems to 
> me there is a potential risk of data corruption at limit points, because 
> don't know who to handle the over flow situation, does this handle by 
> VFS? or I have to take care myself? Thanks.

The VFS will initialize sb->s_maxbytes to MAX_NON_LFS if you don't.

There was one kernel version (2.4.2, I think, or maybe 2.4.3) when you
_had_ to set it, otherwise writes would fail -- as you discovered yourself.

Ion

-- 
  It is better to keep your mouth shut and be thought a fool,
            than to open it and remove all doubt.

Re: LFS/non-LFS error handling

[David Chow]

Actually, I tested long ago with RedHat 7.1 kernel 2.4.2-2 which is 
properly a 2.4.3 (I don't know but it seems to be) .  I am trying to 
automate the compiling process for my fs and discovering kernel version 
issues, it seems there are so many kernel version issues as well. We try 
to distribute a generic binary kernel module with patching kernel 
version symbols with /proc/ksysms during installation so that users 
don't have to custom build using kernel source when they do 
installation,  that's why we are now studying kernel version 
dependencies. Also some of the calls seems have different storage sizes 
among different versions even we stick with 2.4.x only.


I really want to find my way on solving dependency issues among 
different kernel version but not recompiling from source.

regards,

David

Ion Badulescu wrote:

>On Sat, 13 Apr 2002 00:46:14 +0800, David Chow <davidchow@shaolinmicro.com> wrote:
>
>>Dear all,
>>
>>If my fs don't support LFS (files larger than 4GB), what should I set in 
>>my sb->s_maxbytes? 
>>
>
>MAX_NON_LFS.
>
>>In kernel 2.4.2 , I have experieced if not set the 
>>s_maxbytes generic_file_write() will fail, but for later version of 
>>kernel I simply leave it unset, so far I don't have any problem. If it 
>>is wrong, what should s_maxbytes=? . Since this behaviour changes across 
>>kernel version, is there a standard implemenation?
>>
>>Since it still work for me for not setting the s_maxbytes, it seems to 
>>me there is a potential risk of data corruption at limit points, because 
>>don't know who to handle the over flow situation, does this handle by 
>>VFS? or I have to take care myself? Thanks.
>>
>
>The VFS will initialize sb->s_maxbytes to MAX_NON_LFS if you don't.
>
>There was one kernel version (2.4.2, I think, or maybe 2.4.3) when you
>_had_ to set it, otherwise writes would fail -- as you discovered yourself.
>
>Ion
>