Userspace write to UBI partitions

Userspace write to UBI partitions

[Ricard Wanderlof]

Is it possible to do a userspace write to a UBI volume as if it were plain 
flash memory, i.e. should I be able to do "cat somefile > /dev/ubi0_0" ? I 
get a 'Write Error: Operation not permitted' when I do this (Linux 2.6.35) 
but I'm pretty sure that I've used UBI volumes that way in the past, may 
have been with older kernels then.

Since there are ioctls for volume update and atomic logical ereasblock 
change, it would seem natural that writing to UBI volumes must be preceded 
by these ioctls, still I can't seem to see anywhere that doing a plain 
write would not be allowed, and as I say I'm pretty sure I've done so in 
the past.

/Ricard
-- 
Ricard Wolf Wanderlöf                           ricardw(at)axis.com
Axis Communications AB, Lund, Sweden            www.axis.com
Phone +46 46 272 2016                           Fax +46 46 13 61 30

Re: Userspace write to UBI partitions

[Shmulik Ladkani]

On Tue, 7 Feb 2012 16:20:58 +0100 (CET) Ricard Wanderlof <ricard.wanderlof@axis.com> wrote:
> 
> Is it possible to do a userspace write to a UBI volume as if it were plain 
> flash memory, i.e. should I be able to do "cat somefile > /dev/ubi0_0" ? I 
> get a 'Write Error: Operation not permitted' when I do this (Linux 2.6.35) 
> but I'm pretty sure that I've used UBI volumes that way in the past, may 
> have been with older kernels then.

Is your volume dynamic or static?

If dynamic, have you tried setting the UBI_VOL_PROP_DIRECT_WRITE
property by issuing a UBI_IOCSETVOLPROP volume ioctl?
(formerly named UBI_PROP_DIRECT_WRITE and UBI_IOCSETPROP)

Regards,
Shmulik

Re: Userspace write to UBI partitions

[Ricard Wanderlof]

On Tue, 7 Feb 2012, Shmulik Ladkani wrote:

> On Tue, 7 Feb 2012 16:20:58 +0100 (CET) Ricard Wanderlof <ricard.wanderlof@axis.com> wrote:
>>
>> Is it possible to do a userspace write to a UBI volume as if it were plain
>> flash memory, i.e. should I be able to do "cat somefile > /dev/ubi0_0" ? I
>> get a 'Write Error: Operation not permitted' when I do this (Linux 2.6.35)
>> but I'm pretty sure that I've used UBI volumes that way in the past, may
>> have been with older kernels then.
>
> Is your volume dynamic or static?

I tried both, but given the nature of the data we want to keep in the 
volume it should really be dynamic.

> If dynamic, have you tried setting the UBI_VOL_PROP_DIRECT_WRITE
> property by issuing a UBI_IOCSETVOLPROP volume ioctl?
> (formerly named UBI_PROP_DIRECT_WRITE and UBI_IOCSETPROP)

Thanks for the pointer, I've never noticed this ioctl before.

I'll try this out. In the meantime, there doesn't seem to be an ioctl for 
erasing blocks?

/Ricard
-- 
Ricard Wolf Wanderlöf                           ricardw(at)axis.com
Axis Communications AB, Lund, Sweden            www.axis.com
Phone +46 46 272 2016                           Fax +46 46 13 61 30

Re: Userspace write to UBI partitions

[Shmulik Ladkani]

On Tue, 7 Feb 2012 17:36:25 +0100 (CET) Ricard Wanderlof <ricard.wanderlof@axis.com> wrote:
> 
> On Tue, 7 Feb 2012, Shmulik Ladkani wrote:
> 
> > On Tue, 7 Feb 2012 16:20:58 +0100 (CET) Ricard Wanderlof <ricard.wanderlof@axis.com> wrote:
> >>
> >> Is it possible to do a userspace write to a UBI volume as if it were plain
> >> flash memory, i.e. should I be able to do "cat somefile > /dev/ubi0_0" ? I
> >> get a 'Write Error: Operation not permitted' when I do this (Linux 2.6.35)
> >> but I'm pretty sure that I've used UBI volumes that way in the past, may
> >> have been with older kernels then.
> >
> > Is your volume dynamic or static?
> 
> I tried both, but given the nature of the data we want to keep in the 
> volume it should really be dynamic.
> 
> > If dynamic, have you tried setting the UBI_VOL_PROP_DIRECT_WRITE
> > property by issuing a UBI_IOCSETVOLPROP volume ioctl?
> > (formerly named UBI_PROP_DIRECT_WRITE and UBI_IOCSETPROP)
> 
> Thanks for the pointer, I've never noticed this ioctl before.
> 
> I'll try this out. In the meantime, there doesn't seem to be an ioctl for 
> erasing blocks?
> 

Well, another option would be using Gluebi - (CONFIG_MTD_UBI_GLUEBI).
It exposes a "regular" mtd for every existing volume.

You'd be able to work using standard mtd interfaces on it (mtdchar
ioctls, e.g. MEMERASE, or even the mtdblock abstraction).

Regards,
Shmulik

Re: Userspace write to UBI partitions

[Ricard Wanderlof]

On Tue, 7 Feb 2012, Shmulik Ladkani wrote:

>> I'll try this out. In the meantime, there doesn't seem to be an ioctl for
>> erasing blocks?
>>
> Well, another option would be using Gluebi - (CONFIG_MTD_UBI_GLUEBI).
> It exposes a "regular" mtd for every existing volume.
>
> You'd be able to work using standard mtd interfaces on it (mtdchar
> ioctls, e.g. MEMERASE, or even the mtdblock abstraction).

For other reasons, gluebi is not a viable alternative for us. Mainly 
because it puts an mtd device on _every_ ubi volume, meaning that volumes 
can't be removed or resized while gluebi is enabled.

/Ricard
-- 
Ricard Wolf Wanderlöf                           ricardw(at)axis.com
Axis Communications AB, Lund, Sweden            www.axis.com
Phone +46 46 272 2016                           Fax +46 46 13 61 30

Re: Userspace write to UBI partitions

[Artem Bityutskiy]

[-- Attachment #1: Type: text/plain, Size: 883 bytes --]

On Wed, 2012-02-08 at 08:35 +0100, Ricard Wanderlof wrote:
> On Tue, 7 Feb 2012, Shmulik Ladkani wrote:
> 
> >> I'll try this out. In the meantime, there doesn't seem to be an ioctl for
> >> erasing blocks?
> >>
> > Well, another option would be using Gluebi - (CONFIG_MTD_UBI_GLUEBI).
> > It exposes a "regular" mtd for every existing volume.
> >
> > You'd be able to work using standard mtd interfaces on it (mtdchar
> > ioctls, e.g. MEMERASE, or even the mtdblock abstraction).
> 
> For other reasons, gluebi is not a viable alternative for us. Mainly 
> because it puts an mtd device on _every_ ubi volume, meaning that volumes 
> can't be removed or resized while gluebi is enabled.

I do not recommend to use gluebi in general - it is just a hack, but the
limitation you brought up is fixable if you really need to :-)

-- 
Best Regards,
Artem Bityutskiy

[-- Attachment #2: This is a digitally signed message part --]
[-- Type: application/pgp-signature, Size: 836 bytes --]

Re: Userspace write to UBI partitions

[Ricard Wanderlof]

> > For other reasons, gluebi is not a viable alternative for us. Mainly
> > because it puts an mtd device on _every_ ubi volume, meaning that
> > volumes can't be removed or resized while gluebi is enabled.
> 
> I do not recommend to use gluebi in general - it is just a hack, but the 
> limitation you brought up is fixable if you really need to :-)

Yes, I think there was a discussion here on the list a while ago about it.

It seemed to me that the work that might be put into making gluebi more 
flexible would be better spent in making applications we have that rely on 
mtd use ubi instead.

/Ricard
-- 
Ricard Wolf Wanderlöf                           ricardw(at)axis.com
Axis Communications AB, Lund, Sweden            www.axis.com
Phone +46 46 272 2016                           Fax +46 46 13 61 30

Re: Userspace write to UBI partitions

[Artem Bityutskiy]

[-- Attachment #1: Type: text/plain, Size: 597 bytes --]

On Tue, 2012-02-07 at 17:36 +0100, Ricard Wanderlof wrote:
> > If dynamic, have you tried setting the UBI_VOL_PROP_DIRECT_WRITE
> > property by issuing a UBI_IOCSETVOLPROP volume ioctl?
> > (formerly named UBI_PROP_DIRECT_WRITE and UBI_IOCSETPROP)
> 
> Thanks for the pointer, I've never noticed this ioctl before.
> 
> I'll try this out.

It should help.

>  In the meantime, there doesn't seem to be an ioctl for 
> erasing blocks?

In UBI it is called "unmap" instead of "erase":

http://www.linux-mtd.infradead.org/doc/ubi.html#L_unmap

-- 
Best Regards,
Artem Bityutskiy

[-- Attachment #2: This is a digitally signed message part --]
[-- Type: application/pgp-signature, Size: 836 bytes --]

Re: Userspace write to UBI partitions

[Ricard Wanderlof]

> >  In the meantime, there doesn't seem to be an ioctl for erasing
> > blocks?
> 
> In UBI it is called "unmap" instead of "erase":
> 
> http://www.linux-mtd.infradead.org/doc/ubi.html#L_unmap

Thanks. I've been staring at the UBI page for too long now I guess, didn't 
realize that was what I should have been looking for...

/Ricard
-- 
Ricard Wolf Wanderlöf                           ricardw(at)axis.com
Axis Communications AB, Lund, Sweden            www.axis.com
Phone +46 46 272 2016                           Fax +46 46 13 61 30

RE: Userspace write to UBI partitions

[Atlant Schmidt]

All:

  But there are times when you really want a block erased
  and erased NOW; in that case, "unmap" isn't what you
  want as it only schedules the PEB for erasure at some
  future time. And if a power cut intervenes, that
  scheduling for erasure may be lost and the ultimate
  erasure may not take place for a long, long time.

  So if you really intended to destroy the data, "unmap"
  is not the correct function to call.

                              Atlant

-----Original Message-----
From: linux-mtd-bounces@lists.infradead.org [mailto:linux-mtd-bounces@lists.infradead.org] On Behalf Of Ricard Wanderlof
Sent: Wednesday, February 08, 2012 07:19
To: Artem Bityutskiy
Cc: Ricard Wanderlof; Linux mtd; Shmulik Ladkani
Subject: Re: Userspace write to UBI partitions


> >  In the meantime, there doesn't seem to be an ioctl for erasing
> > blocks?
>
> In UBI it is called "unmap" instead of "erase":
>
> http://www.linux-mtd.infradead.org/doc/ubi.html#L_unmap

Thanks. I've been staring at the UBI page for too long now I guess, didn't
realize that was what I should have been looking for...

/Ricard
--
Ricard Wolf Wanderlöf                           ricardw(at)axis.com
Axis Communications AB, Lund, Sweden            www.axis.com
Phone +46 46 272 2016                           Fax +46 46 13 61 30

______________________________________________________
Linux MTD discussion mailing list
http://lists.infradead.org/mailman/listinfo/linux-mtd/


 Click https://www.mailcontrol.com/sr/iNq3RabdjW!TndxI!oX7UskfEqpntbSQBh5+L1x62eAEpsHd6qIu4!3pOM!83jtSy7QxM5RExv!1lqOECv!Swg==  to report this email as spam.

This e-mail and the information, including any attachments, it contains are intended to be a confidential communication only to the person or entity to whom it is addressed and may contain information that is privileged. If the reader of this message is not the intended recipient, you are hereby notified that any dissemination, distribution or copying of this communication is strictly prohibited. If you have received this communication in error, please immediately notify the sender and destroy the original message.

Thank you.

Please consider the environment before printing this email.

RE: Userspace write to UBI partitions

[Artem Bityutskiy]

[-- Attachment #1: Type: text/plain, Size: 670 bytes --]

On Wed, 2012-02-08 at 08:49 -0500, Atlant Schmidt wrote:
> All:
> 
>   But there are times when you really want a block erased
>   and erased NOW; in that case, "unmap" isn't what you
>   want as it only schedules the PEB for erasure at some
>   future time. And if a power cut intervenes, that
>   scheduling for erasure may be lost and the ultimate
>   erasure may not take place for a long, long time.
> 
>   So if you really intended to destroy the data, "unmap"
>   is not the correct function to call.

In this case you should first unmap and than map:

http://www.linux-mtd.infradead.org/doc/ubi.html#L_map

-- 
Best Regards,
Artem Bityutskiy

[-- Attachment #2: This is a digitally signed message part --]
[-- Type: application/pgp-signature, Size: 836 bytes --]