RFC : fat : default errors use FAT_ERRORS_CONT instead of FAT_ERRORS_RO.

RFC : fat : default errors use FAT_ERRORS_CONT instead of FAT_ERRORS_RO.

[NamJae Jeon]

Hi ogawa.

I found different error exception between linux fat and windows fat.

When fat is currputed by sudden unplug usb device formatted fat, fat
does never change RO partition on windows.

It said windows fat provide a chance that continuously access
partition except corrupted file.

Currently when access corrupted file, linux fat change RO parition
from RW. So user no longer do open/write on linux fat.

I suggest that the default errors of linux fat use FAT_ERRORS_CONT
instead of FAT_ERRORS_RO.

I want to know your opinion.

Thanks.

Re: RFC : fat : default errors use FAT_ERRORS_CONT instead of FAT_ERRORS_RO.

[Valdis.Kletnieks]

[-- Attachment #1: Type: text/plain, Size: 722 bytes --]

On Thu, 13 Oct 2011 09:24:52 +0900, NamJae Jeon said:

> It said windows fat provide a chance that continuously access
> partition except corrupted file.

It also gives you a chance to make things worse by continuing to write to the
device.  If you force it RO at the first sign of trouble, maybe you lose a few
blocks of one file.  If you keep writing to it, creating new files, erasing
files, and so on, pretty soon your corrupted filesystem is even more corrupted.

> I suggest that the default errors of linux fat use FAT_ERRORS_CONT
> instead of FAT_ERRORS_RO.

And what user benefit does that give?  If you keep going, they won't
realize their device is corrupted, and won't have any reason to *fix* the
filesystem.

[-- Attachment #2: Type: application/pgp-signature, Size: 227 bytes --]

Re: RFC : fat : default errors use FAT_ERRORS_CONT instead of FAT_ERRORS_RO.

[NamJae Jeon]

2011/10/13  <Valdis.Kletnieks@vt.edu>:
> On Thu, 13 Oct 2011 09:24:52 +0900, NamJae Jeon said:
>
>> It said windows fat provide a chance that continuously access
>> partition except corrupted file.
>
> It also gives you a chance to make things worse by continuing to write to the
> device.  If you force it RO at the first sign of trouble, maybe you lose a few
> blocks of one file.  If you keep writing to it, creating new files, erasing
> files, and so on, pretty soon your corrupted filesystem is even more corrupted.
There is no journal feature in FAT. it means FAT is corrupted well in
unexpected status.
Can you explain specifically what error will happen in case of keep writing ?
I think that MS consider also about this. maybe they think that
keeping access is better than changing RO type by one file is
corrupted.
>
>> I suggest that the default errors of linux fat use FAT_ERRORS_CONT
>> instead of FAT_ERRORS_RO.
>
> And what user benefit does that give?  If you keep going, they won't
> realize their device is corrupted, and won't have any reason to *fix* the
> filesystem.
for example, FAT is RW filesystem. and it is corrupted well by
unexpected status.
If so, user can not write FAT filesystem anytime while using linux
FAT.  there is no mehod except they should directly remount fat
parittion. well known linux user can do it only.
>

Re: RFC : fat : default errors use FAT_ERRORS_CONT instead of FAT_ERRORS_RO.

[NamJae Jeon]

2011/10/13 NamJae Jeon <linkinjeon@gmail.com>:
> 2011/10/13  <Valdis.Kletnieks@vt.edu>:
>> On Thu, 13 Oct 2011 09:24:52 +0900, NamJae Jeon said:
>>
>>> It said windows fat provide a chance that continuously access
>>> partition except corrupted file.
>>
>> It also gives you a chance to make things worse by continuing to write to the
>> device.  If you force it RO at the first sign of trouble, maybe you lose a few
>> blocks of one file.  If you keep writing to it, creating new files, erasing
>> files, and so on, pretty soon your corrupted filesystem is even more corrupted.
> There is no journal feature in FAT. it means FAT is corrupted well in
> unexpected status.
> Can you explain specifically what error will happen in case of keep writing ?
> I think that MS consider also about this. maybe they think that
> keeping access is better than changing RO type by one file is
> corrupted.
>>
>>> I suggest that the default errors of linux fat use FAT_ERRORS_CONT
>>> instead of FAT_ERRORS_RO.
>>
>> And what user benefit does that give?  If you keep going, they won't
>> realize their device is corrupted, and won't have any reason to *fix* the
>> filesystem.
> for example, FAT is RW filesystem. and it is corrupted well by
> unexpected status.
> If so, user can not write FAT filesystem anytime while using linux
> FAT.  there is no mehod except they should directly remount fat
> parittion. well known linux user can do it only.
And user can know file is corrupted or not by corrupted message and
return value when accessing this file.
>>
>

Re: RFC : fat : default errors use FAT_ERRORS_CONT instead of FAT_ERRORS_RO.

[NamJae Jeon]

2011/10/13 NamJae Jeon <linkinjeon@gmail.com>:
> 2011/10/13 NamJae Jeon <linkinjeon@gmail.com>:
>> 2011/10/13  <Valdis.Kletnieks@vt.edu>:
>>> On Thu, 13 Oct 2011 09:24:52 +0900, NamJae Jeon said:
>>>
>>>> It said windows fat provide a chance that continuously access
>>>> partition except corrupted file.
>>>
>>> It also gives you a chance to make things worse by continuing to write to the
>>> device.  If you force it RO at the first sign of trouble, maybe you lose a few
>>> blocks of one file.  If you keep writing to it, creating new files, erasing
>>> files, and so on, pretty soon your corrupted filesystem is even more corrupted.
>> There is no journal feature in FAT. it means FAT is corrupted well in
>> unexpected status.
>> Can you explain specifically what error will happen in case of keep writing ?
>> I think that MS consider also about this. maybe they think that
>> keeping access is better than changing RO type by one file is
>> corrupted.
And currenly we still can write on FAT if corrupted file is not accessed.
>>>
>>>> I suggest that the default errors of linux fat use FAT_ERRORS_CONT
>>>> instead of FAT_ERRORS_RO.
>>>
>>> And what user benefit does that give?  If you keep going, they won't
>>> realize their device is corrupted, and won't have any reason to *fix* the
>>> filesystem.
>> for example, FAT is RW filesystem. and it is corrupted well by
>> unexpected status.
>> If so, user can not write FAT filesystem anytime while using linux
>> FAT.  there is no mehod except they should directly remount fat
>> parittion. well known linux user can do it only.
> And user can know file is corrupted or not by corrupted message and
> return value when accessing this file.
>>>
>>
>

Re: RFC : fat : default errors use FAT_ERRORS_CONT instead of FAT_ERRORS_RO.

[OGAWA Hirofumi]

NamJae Jeon <linkinjeon@gmail.com> writes:

> Hi ogawa.
>
> I found different error exception between linux fat and windows fat.
>
> When fat is currputed by sudden unplug usb device formatted fat, fat
> does never change RO partition on windows.
>
> It said windows fat provide a chance that continuously access
> partition except corrupted file.
>
> Currently when access corrupted file, linux fat change RO parition
> from RW. So user no longer do open/write on linux fat.
>
> I suggest that the default errors of linux fat use FAT_ERRORS_CONT
> instead of FAT_ERRORS_RO.
>
> I want to know your opinion.

I think it is dangerous, and the default should be safer side.

If driver noticed the corruption, it means FS is not unknown state
anymore.  Any files/directories are possibly corrupted.

Even if you change the file that seems to be uncorrupted, it may make FS
unrecoverable more. (e.g. lose all data in worst case)

Thanks.
-- 
OGAWA Hirofumi <hirofumi@mail.parknet.co.jp>

Re: RFC : fat : default errors use FAT_ERRORS_CONT instead of FAT_ERRORS_RO.

[NamJae Jeon]

2011/10/13 OGAWA Hirofumi <hirofumi@mail.parknet.co.jp>:
> NamJae Jeon <linkinjeon@gmail.com> writes:
>
>> Hi ogawa.
>>
>> I found different error exception between linux fat and windows fat.
>>
>> When fat is currputed by sudden unplug usb device formatted fat, fat
>> does never change RO partition on windows.
>>
>> It said windows fat provide a chance that continuously access
>> partition except corrupted file.
>>
>> Currently when access corrupted file, linux fat change RO parition
>> from RW. So user no longer do open/write on linux fat.
>>
>> I suggest that the default errors of linux fat use FAT_ERRORS_CONT
>> instead of FAT_ERRORS_RO.
>>
>> I want to know your opinion.
>
> I think it is dangerous, and the default should be safer side.
>
> If driver noticed the corruption, it means FS is not unknown state
> anymore.  Any files/directories are possibly corrupted.
>
> Even if you change the file that seems to be uncorrupted, it may make FS
> unrecoverable more. (e.g. lose all data in worst case)
Hi ogawa.
Thanks for your reply. I follow your opinion.

>
> Thanks.
> --
> OGAWA Hirofumi <hirofumi@mail.parknet.co.jp>
>

Re: RFC : fat : default errors use FAT_ERRORS_CONT instead of FAT_ERRORS_RO.

[Alan Cox]

> If so, user can not write FAT filesystem anytime while using linux
> FAT.  there is no mehod except they should directly remount fat
> parittion. well known linux user can do it only.

That is a user space tools problem. Nothing stops a desktop doing things
like allowing you to click on a device and have a 'repair' menu entry
when appropriate. We have a FAT fsck (although neither it nor the
Microsoft one are brilliant at rescuing FAT disks) and the OS supports
mount/unmount/remount.

Alan