32-bit user suspend program on 64-bit kernel?

32-bit user suspend program on 64-bit kernel?

[Johannes Berg]

[-- Attachment #1.1: Type: text/plain, Size: 182 bytes --]

Is that configuration possible? It seems not since I can't find any
compat code for the ioctl which shares pointers and structs, is there
any way to make it possible?

johannes

[-- Attachment #1.2: This is a digitally signed message part --]
[-- Type: application/pgp-signature, Size: 190 bytes --]

[-- Attachment #2: Type: text/plain, Size: 0 bytes --]

Re: 32-bit user suspend program on 64-bit kernel?

[Pavel Machek]

Hi!

> Is that configuration possible? It seems not since I can't find any

Yes.

> compat code for the ioctl which shares pointers and structs, is there
> any way to make it possible?

That means that either ioctl is very hw-specific, or you've found a
bug.

								Pavel

-- 
(english) http://www.livejournal.com/~pavelmachek
(cesky, pictures) http://atrey.karlin.mff.cuni.cz/~pavel/picture/horses/blog.html

Re: 32-bit user suspend program on 64-bit kernel?

[Johannes Berg]

[-- Attachment #1.1: Type: text/plain, Size: 306 bytes --]

On Wed, 2007-03-21 at 13:28 +0100, Pavel Machek wrote:

> > compat code for the ioctl which shares pointers and structs, is there
> > any way to make it possible?
> 
> That means that either ioctl is very hw-specific, or you've found a
> bug.

I think I'll just have to give it a try.

johannes

[-- Attachment #1.2: This is a digitally signed message part --]
[-- Type: application/pgp-signature, Size: 190 bytes --]

[-- Attachment #2: Type: text/plain, Size: 0 bytes --]

Re: 32-bit user suspend program on 64-bit kernel?

[Rafael J. Wysocki]

On Wednesday, 21 March 2007 13:31, Johannes Berg wrote:
> On Wed, 2007-03-21 at 13:28 +0100, Pavel Machek wrote:
> 
> > > compat code for the ioctl which shares pointers and structs, is there
> > > any way to make it possible?
> > 
> > That means that either ioctl is very hw-specific, or you've found a
> > bug.
> 
> I think I'll just have to give it a try.

Unfortunately the numbers of ioctls are not portable due to my rookie mistake
in the beginning (using void * in _IOW() and friends which passed with no
objections).

I'm ashamed of that, but it's probably too late to fix it. :-(

Greetings,
Rafael

Re: 32-bit user suspend program on 64-bit kernel?

[Johannes Berg]

[-- Attachment #1.1: Type: text/plain, Size: 675 bytes --]

On Wed, 2007-03-21 at 15:05 +0100, Rafael J. Wysocki wrote:

> Unfortunately the numbers of ioctls are not portable due to my rookie mistake
> in the beginning (using void * in _IOW() and friends which passed with no
> objections).
> 
> I'm ashamed of that, but it's probably too late to fix it. :-(

Bugger. What would be required to fix it?

Thing is, this is going to make powerpc distro's lives difficult since
part of uswsusp needs to live in the initrd which usually just contains
kernel modules + normal programs, and no powerpc distro I know of ships
64-bit userspace on a 32-bit kernel since, well, it doesn't really have
many advantages.

johannes

[-- Attachment #1.2: This is a digitally signed message part --]
[-- Type: application/pgp-signature, Size: 190 bytes --]

[-- Attachment #2: Type: text/plain, Size: 0 bytes --]

Re: 32-bit user suspend program on 64-bit kernel?

[Rafael J. Wysocki]

On Wednesday, 21 March 2007 15:04, Johannes Berg wrote:
> On Wed, 2007-03-21 at 15:05 +0100, Rafael J. Wysocki wrote:
> 
> > Unfortunately the numbers of ioctls are not portable due to my rookie mistake
> > in the beginning (using void * in _IOW() and friends which passed with no
> > objections).
> > 
> > I'm ashamed of that, but it's probably too late to fix it. :-(
> 
> Bugger. What would be required to fix it?

Well, the change itself wouldn't be big (three #defines or so), but it would
break all of the existing setups.

Alternatively, we can add some ioctls that duplicate those with nonportable
numbers and tell people to use the new ones in the future (documentation
changes would be needed).

Greetings,
Rafael

Re: 32-bit user suspend program on 64-bit kernel?

[Johannes Berg]

[-- Attachment #1.1: Type: text/plain, Size: 625 bytes --]

On Wed, 2007-03-21 at 15:17 +0100, Rafael J. Wysocki wrote:

> Well, the change itself wouldn't be big (three #defines or so), but it would
> break all of the existing setups.
> 
> Alternatively, we can add some ioctls that duplicate those with nonportable
> numbers and tell people to use the new ones in the future (documentation
> changes would be needed).

Sounds like a plan then, no? Make them print a big warning that people
actually see and put the other numbers on the feature deprecation
schedule :)

I do have a 64-bit powerpc that runs 32-bit userspace exclusively,
that's why I asked.

johannes

[-- Attachment #1.2: This is a digitally signed message part --]
[-- Type: application/pgp-signature, Size: 190 bytes --]

[-- Attachment #2: Type: text/plain, Size: 0 bytes --]