Re: [PATCH] dl: Use "adr" assembler command to get proper load address

[Lukasz Majewski <lukma@denx.de>]

[-- Attachment #1: Type: text/plain, Size: 4992 bytes --]

Dear Community,

> On Thu, 9 Sep 2021 11:49:36 +0200
> Lukasz Majewski <lukma@denx.de> wrote:
> 
> > On Thu, 9 Sep 2021 09:18:06 +0200
> > Lukasz Majewski <lukma@denx.de> wrote:
> >   
> > > On Wed, 8 Sep 2021 22:34:21 +0200
> > > Lukasz Majewski <lukma@denx.de> wrote:
> > >     
> > > > Hi Adhemerval,
> > > >       
> > > > > On 08/09/2021 12:05, Lukasz Majewski wrote:        
> > > > > > 
> > > > > > The r1 gets the 0xffffee80 (negative offset) value. It is
> > > > > > then added to pc and used to calculate r2.
> > > > > > 
> > > > > > For working code (with this patch applied) - there are NO
> > > > > > such large values (like aforementioned 0xffffee80). The
> > > > > > arithmetic is done on 
> > > > > > 
> > > > > >    1690:       00000020        .word   0x00000020
> > > > > >    1694:       0002be7e        .word   0x0002be7e
> > > > > > 
> > > > > > which seems to work.
> > > > > > 
> > > > > > This shouldn't be a problem as with U2 the arithmetic shall
> > > > > > work. However, I've noticed (with passing LD_DEBUG=all) that
> > > > > > the ld-linux-armhf.so.3 (and init) are relocated twice
> > > > > > before execution.
> > > > > > 
> > > > > > Why do we need to relocate it?
> > > > > > 
> > > > > > Another question is why on this particular case the large
> > > > > > (i.e. negative) offset matters?          
> > > > > 
> > > > > I think it is highly unlikely the negative offset plays any
> > > > > role here. Do you have a working example to trigger this
> > > > > issue? 
> > > > 
> > > > I'm just using QEMU (runqemu -d y2038-image-devel nographic)
> > > > from meta-y2038 build.
> > > > 
> > > > Recent versions: 
> > > > poky: SHA1: 1e2e9a84d6dd81d7f6dd69c0d119d0149d10ade1
> > > > qemu_6.0.0
> > > > binutils_2.37
> > > > gcc_11.2
> > > > 
> > > > The QEMU board uses Cortex-A9 core.      
> > > 
> > > I've updated the poky to be the newest -master:
> > > master:50293eec9a7d0602b748220ab100b070b8d3432a
> > > 
> > > No change.
> > > 
> > > I will try the same setup with Cortex-A5 core - maybe there is
> > > some kind of subtle handling of such emulation in qemu?    
> > 
> > No change. On the qemuarm yocto/OE MACHINE the same situation is
> > observed.
> >   
> 
> Do you have any more ideas where to look for solution of this problem?

Can we make a decision regarding this fix?

The only finding, which I get, is the large negative offset added to the
pc in the glibc broken (current master) code, which causes QEMU system
to fail. 

I've shared this problem with the QEMU community [1], but no reply
received till today.


Shall we:

1. Apply this patch or

2. Revert the conversion (whole) patch [2] for ARM.

3. Wait till QEMU response for the problem?



Links:

[1] -
https://lists.nongnu.org/archive/html/qemu-devel/2021-09/msg03332.html

[2] - SHA1: bca0f5cbc9257c13322b99e55235c4f21ba0bd82

> 
> > After applying this patch the board works without issues.
> >   
> > >     
> > > >       
> > > > > I am currently
> > > > > testing arm for different compilers (gcc 6.2, gcc 10, gcc 11)
> > > > > and with different configurations (armv5, armv6, armv7, with
> > > > > and without thumb) and I haven't see any issue so far.
> > > > > 
> > > > > It might binutils related, which version are you using?
> > > > >  
> > > > 
> > > > 
> > > > 
> > > > 
> > > > Best regards,
> > > > 
> > > > Lukasz Majewski
> > > > 
> > > > --
> > > > 
> > > > DENX Software Engineering GmbH,      Managing Director: Wolfgang
> > > > Denk HRB 165235 Munich, Office: Kirchenstr.5, D-82194
> > > > Groebenzell, Germany Phone: (+49)-8142-66989-59 Fax:
> > > > (+49)-8142-66989-80 Email: lukma@denx.de      
> > > 
> > > 
> > > 
> > > 
> > > Best regards,
> > > 
> > > Lukasz Majewski
> > > 
> > > --
> > > 
> > > DENX Software Engineering GmbH,      Managing Director: Wolfgang
> > > Denk HRB 165235 Munich, Office: Kirchenstr.5, D-82194 Groebenzell,
> > > Germany Phone: (+49)-8142-66989-59 Fax: (+49)-8142-66989-80 Email:
> > > lukma@denx.de    
> > 
> > 
> > 
> > 
> > Best regards,
> > 
> > Lukasz Majewski
> > 
> > --
> > 
> > DENX Software Engineering GmbH,      Managing Director: Wolfgang
> > Denk HRB 165235 Munich, Office: Kirchenstr.5, D-82194 Groebenzell,
> > Germany Phone: (+49)-8142-66989-59 Fax: (+49)-8142-66989-80 Email:
> > lukma@denx.de  
> 
> 
> 
> 
> Best regards,
> 
> Lukasz Majewski
> 
> --
> 
> DENX Software Engineering GmbH,      Managing Director: Wolfgang Denk
> HRB 165235 Munich, Office: Kirchenstr.5, D-82194 Groebenzell, Germany
> Phone: (+49)-8142-66989-59 Fax: (+49)-8142-66989-80 Email:
> lukma@denx.de




Best regards,

Lukasz Majewski

--

DENX Software Engineering GmbH,      Managing Director: Wolfgang Denk
HRB 165235 Munich, Office: Kirchenstr.5, D-82194 Groebenzell, Germany
Phone: (+49)-8142-66989-59 Fax: (+49)-8142-66989-80 Email: lukma@denx.de

[-- Attachment #2: OpenPGP digital signature --]
[-- Type: application/pgp-signature, Size: 488 bytes --]