build failure of mips decstation_r4k_defconfig with binutils-2_37

build failure of mips decstation_r4k_defconfig with binutils-2_37

[Sudip Mukherjee]

Hi All,

While I was testing v5.4.142-rc2 I noticed mips build of
decstation_r4k_defconfig fails with binutils-2_37. The error is:

arch/mips/dec/prom/locore.S: Assembler messages:
arch/mips/dec/prom/locore.S:29: Error: opcode not supported on this
processor: r4600 (mips3) `rfe'

I have also reported this at https://sourceware.org/bugzilla/show_bug.cgi?id=28241

--
Regards
Sudip

Re: build failure of mips decstation_r4k_defconfig with binutils-2_37

[Ralf Baechle]

On Tue, Aug 17, 2021 at 12:54:32PM +0100, Sudip Mukherjee wrote:

> While I was testing v5.4.142-rc2 I noticed mips build of
> decstation_r4k_defconfig fails with binutils-2_37. The error is:
> 
> arch/mips/dec/prom/locore.S: Assembler messages:
> arch/mips/dec/prom/locore.S:29: Error: opcode not supported on this
> processor: r4600 (mips3) `rfe'
> 
> I have also reported this at https://sourceware.org/bugzilla/show_bug.cgi?id=28241

It would appear gas got more anal about ISA checking for the RFE instructions
which did only exist in MIPS I and II; MIPS III and later use ERET for
returning from an exception.

The older gas I've got installed here happily accepts RFE in MIPS III/R4000
mode:

$ cat s.s
	rfe
	eret
$ mips-linux-as -o s.o s.s
s.s: Assembler messages:
s.s:2: Error: opcode not supported on this processor: mips1 (mips1) `eret'
$ mips-linux-as -march=r4000 -o s.o s.s
$ mips-linux-objdump -d s.o 

s.o:     file format elf32-tradbigmips


Disassembly of section .text:

00000000 <.text>:
   0:	42000010 	c0	0x10		# <- RFE
   4:	42000018 	eret
	...
$

It's easy to find arguments for why this gas change is the right thing to
do - and not the right thing to do.

It should be fixable by simply putting gas into mips1 mode.  Can you test
below patch?

  Ralf

Signed-off-by: Ralf Baechle <ralf@linux-mips.org>

diff --git a/arch/mips/dec/prom/locore.S b/arch/mips/dec/prom/locore.S
index 0eb8fab62ab0..8d00ca8872f9 100644
--- a/arch/mips/dec/prom/locore.S
+++ b/arch/mips/dec/prom/locore.S
@@ -16,6 +16,7 @@
 NESTED(genexcept_early, 0, sp)
 	.set	noat
 	.set	noreorder
+	.set	mips1
 
 	mfc0	k0, CP0_STATUS
 	la	k1, mem_err

Re: build failure of mips decstation_r4k_defconfig with binutils-2_37

[Sudip Mukherjee]

Hi Ralf,

On Tue, Aug 17, 2021 at 1:55 PM Ralf Baechle <ralf@linux-mips.org> wrote:
>
> On Tue, Aug 17, 2021 at 12:54:32PM +0100, Sudip Mukherjee wrote:
>
> > While I was testing v5.4.142-rc2 I noticed mips build of
> > decstation_r4k_defconfig fails with binutils-2_37. The error is:
> >
> > arch/mips/dec/prom/locore.S: Assembler messages:
> > arch/mips/dec/prom/locore.S:29: Error: opcode not supported on this
> > processor: r4600 (mips3) `rfe'
> >
> > I have also reported this at https://sourceware.org/bugzilla/show_bug.cgi?id=28241
>
> It would appear gas got more anal about ISA checking for the RFE instructions
> which did only exist in MIPS I and II; MIPS III and later use ERET for
> returning from an exception.
>
> The older gas I've got installed here happily accepts RFE in MIPS III/R4000
> mode:
>

<snip>

>
> It's easy to find arguments for why this gas change is the right thing to
> do - and not the right thing to do.
>
> It should be fixable by simply putting gas into mips1 mode.  Can you test
> below patch?

Tested on top of v5.4.142-rc2 and it builds again. Thanks for the
quick reply and the patch.


-- 
Regards
Sudip

Re: build failure of mips decstation_r4k_defconfig with binutils-2_37

[Maciej W. Rozycki]

On Tue, 17 Aug 2021, Ralf Baechle wrote:

> > While I was testing v5.4.142-rc2 I noticed mips build of
> > decstation_r4k_defconfig fails with binutils-2_37. The error is:
> > 
> > arch/mips/dec/prom/locore.S: Assembler messages:
> > arch/mips/dec/prom/locore.S:29: Error: opcode not supported on this
> > processor: r4600 (mips3) `rfe'
> > 
> > I have also reported this at https://sourceware.org/bugzilla/show_bug.cgi?id=28241
> 
> It would appear gas got more anal about ISA checking for the RFE instructions
> which did only exist in MIPS I and II; MIPS III and later use ERET for
> returning from an exception.

 Yes, I made such a change when I discovered the mess around coprocessor 
instructions in binutils earlier this year.  Eighteen patches total to 
straighten it out.  Some instructions were misassembled even, and no 
proper subsetting was made, so you could come with a nonsensical mixture 
when disassembling.  And RFE was disassembled as `c0 0x10' regardless of 
the ISA chosen.

> It should be fixable by simply putting gas into mips1 mode.  Can you test
> below patch?

 But it's missing the point, as I noted in the other message.  I've been 
too busy with higher priority stuff to get to a proper fix right away (I 
thought I was the last one using the DECstation, and especially with the 
top of the tree binutils).

  Maciej

Re: build failure of mips decstation_r4k_defconfig with binutils-2_37

[Maciej W. Rozycki]

On Tue, 17 Aug 2021, Sudip Mukherjee wrote:

> While I was testing v5.4.142-rc2 I noticed mips build of
> decstation_r4k_defconfig fails with binutils-2_37. The error is:
> 
> arch/mips/dec/prom/locore.S: Assembler messages:
> arch/mips/dec/prom/locore.S:29: Error: opcode not supported on this
> processor: r4600 (mips3) `rfe'
> 
> I have also reported this at https://sourceware.org/bugzilla/show_bug.cgi?id=28241

 Thanks.  I have known it for a while thanks to an earlier report from 
Jan-Benedict Glaw, but didn't get to fixing it.  I mean to address it 
sometime later this month or early Sep the latest.

 This file shouldn't be built for R4k configurations in the first place 
because they use the REX firmware exclusively, which this source has 
nothing to do with.  A trivial fix would be to override the ISA level 
temporarily across RFE, but that's missing the point as it's dead code 
anyway with R4k.  OTOH a proper fix requires proper verification.

 As a workaround use older binutils for the time being.  Apologies for the 
inconvenience.

  Maciej