[PATCH] guess_fstype applet

[Tanguy Pruvot]

Thanks, i think it will be another divergence, the main purpose of this
patch is for the android recovery which should be able to mount badly set
fstype (exemple: for a backup)

Even if in CM the recovery can be installed on each nightly, we have the
ability to keep a custom one (to handle different backup formats etc...)
and the fstab can be different of the recovery one.

Now we have another problem about that, its the ability to use encrypted
file systems...

2015-02-11 20:37 GMT+01:00 Tanguy Pruvot <tanguy.pruvot at gmail.com>:

> Thanks for the review... was merged too fast
>
> 2015-02-11 20:32 GMT+01:00 Isaac Dunham <ibid.ag at gmail.com>:
>
>> On Wed, Feb 11, 2015 at 07:50:49PM +0100, Tanguy Pruvot wrote:
>> > 2015-02-11 19:48 GMT+01:00 Denys Vlasenko <vda.linux at googlemail.com>:
>> >
>> > > On Fri, Feb 6, 2015 at 5:35 PM, Tanguy Pruvot <
>> tanguy.pruvot at gmail.com>
>> > > wrote:
>> > > > From CyanogenMod :
>> > > >
>> > > > http://review.cyanogenmod.org/#/c/87995/
>> > > >
>> > > > 2013-08-23 6:20 GMT+02:00 James B <jamesbond3142 at gmail.com>:
>> > > >>
>> > > >> If the interface can be simplifed to detect filesystem only
>> (instead
>> > > >> of also looking for labels and UUID) I think I think it would get
>> an
>> > > >> order of magnitude speed-up (taking the cues from Puppy Linux's
>> > > >> original guess_fstype which basically does just that). I'm not so
>> sure
>> > > >> about code-size reduction, though.
>> > > >>
>> > > >> It is called guess_fstype for historical reason (because that's
>> how it
>> > > >> was used and called in Puppy Linux), I'm happy to change it to
>> > > >> "fstype" if that helps everybody else :)
>> > >
>> > > I'm confused.
>> > > Is here something proposed for inclusion into busybox,
>> > > or you are now talking about hacks for a particular project only?
>> >
>> > its a feature to enhance the auto mount with blkid code... we are about
>> to
>> > merge this commit soon in CM12
>>
>> Comments:
>> - do NOT let code override a user specification.
>> If I specify -t (ntfs|ntfs-3g) or (msdos|vfat) manually, it should force
>> mount to try mounting using that specific driver and then *fail* if
>> it's impossible.
>> If fstype overrides user specifications, you may end up breaking future
>> filesystems.
>> I would think that this would be better:
>>         // Treat fstype "auto" as unspecified
>>         if (mp->mnt_type && strcmp(mp->mnt_type, "auto") == 0)
>>                 mp->mnt_type = NULL;
>> +#if ENABLE_FEATURE_BLKID_TYPE
>> +       // try to to autodetect type
>> +       if (!mp->mnt_type)
>> +               mp->mnt_type = get_fstype_from_devname(mp->mnt_fsname);
>> #endif
>>
>> OTOH, if I specify -t auto and userspace code selects the wrong fs type
>> (as it surely will, because bugs and unexpected inputs happen...),
>> it would be proper for mount to keep retrying, as would be done if
>> mp->mnt_type is unset after the new code.
>>
>> - I'm guessing that the use of != to compare two strings is a mistake?
>> +       || (detected_fstype != NULL && detected_fstype != mp->mnt_type))
>>
>> - guess_fstype, fstype, blkid commands:
>>  guess_fstype is the name that Puppy Linux has used for a command that
>>  outputs filesystem type without label/uuid.
>>  fstype is the name that it goes by in Toybox, which is now in Android.
>>  However, upstream Android turns it off and just uses blkid.
>>
>> HTH,
>> Isaac Dunham
>>
>
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.busybox.net/pipermail/busybox/attachments/20150212/ecfa8d52/attachment-0001.html>