fix DECSTATION depends
PATCH 1/1 fix DECSTATION depends
Hi Ralf,
it seems the following is required in Linus' tree to get correct depends for DECSTATION:
arch/mips/Kconfig renamed from arch/mips/Kconfig.old +1 -1
@@ -51,7 +51,7 @@ config DECSTATION bool "Support for DECstations" - depends on MIPS32 || EXPERIMENTAL + depends on MIPS32 && EXPERIMENTAL ---help--- This enables support for DEC's MIPS based workstations. For details see the Linux/MIPS FAQ on <http://oss.sgi.com/mips/> and the
cu
Adrian
Signature · 4 lines
"Is there not promise of rain?" Ling Tan asked suddenly out
of the darkness. There had been need of rain for many days.
"Only a promise," Lao Er said.
Pearl S. Buck - Dragon SeedQuoted 2 lines · Adrian Bunk
it seems the following is required in Linus' tree to get correct depends for DECSTATION:
Thanks, applied.
Ralf
Quoted 4 lines · Ralf Baechle
Quoted 2 lines
it seems the following is required in Linus' tree to get correct depends for DECSTATION:
Thanks, applied.
The dependency was intentional: stable for 32-bit, experimental for
64-bit. I'm reverting the change immediately. Please always contact me before applying non-obvious changes for the DECstation.
If there's anything wrong with the depends, it should be fixed elsewhere.
Details, please.
Maciej
Signature · 3 lines
+ Maciej W. Rozycki, Technical University of Gdansk, Poland + +--------------------------------------------------------------+ + e-mail: [email protected], PGP key available +
Quoted 9 lines · Maciej W. Rozycki
Quoted 5 lines
Quoted 3 lines
it seems the following is required in Linus' tree to get correct depends for DECSTATION:
Thanks, applied.
The dependency was intentional: stable for 32-bit, experimental for
64-bit. I'm reverting the change immediately. Please always contact me before applying non-obvious changes for the DECstation.
Well, this one seemed to be obvious but sometimes things are not what they seem to be ...
Ralf
Quoted 6 lines · Ralf Baechle
Quoted 3 lines
The dependency was intentional: stable for 32-bit, experimental for
64-bit. I'm reverting the change immediately. Please always contact me before applying non-obvious changes for the DECstation.Well, this one seemed to be obvious but sometimes things are not what they seem to be ...
Hmm, a change of semantics wouldn't qualify as obvious for me...
I admit this construct may seem surprising at first sight when compared
to many other uses of EXPERIMENTAL -- the equivalent more verbose syntax used for 2.4 leaves no doubt about the intent, though.
Maciej
Signature · 3 lines
+ Maciej W. Rozycki, Technical University of Gdansk, Poland + +--------------------------------------------------------------+ + e-mail: [email protected], PGP key available +
Quoted 5 lines · Maciej W. Rozycki
Quoted 3 lines · Ralf Baechle
Quoted 2 lines
it seems the following is required in Linus' tree to get correct depends for DECSTATION:
Quoted 8 lines
Quoted 1 line
Thanks, applied.
The dependency was intentional: stable for 32-bit, experimental for
64-bit. I'm reverting the change immediately. Please always contact me before applying non-obvious changes for the DECstation.If there's anything wrong with the depends, it should be fixed elsewhere.
Details, please.
Does -mabi=64 really work on any DECstation?
AFAIK none of the R2000, R3x00 and the R4x00 do support the 64bit ABI.
Quoted 1 line
Maciej
cu
Adrian
Signature · 4 lines
"Is there not promise of rain?" Ling Tan asked suddenly out
of the darkness. There had been need of rain for many days.
"Only a promise," Lao Er said.
Pearl S. Buck - Dragon SeedQuoted 3 lines · Adrian Bunk
Does -mabi=64 really work on any DECstation?
AFAIK none of the R2000, R3x00 and the R4x00 do support the 64bit ABI.
Well, the R4000 and the R4400 implement the MIPS3 ISA (the R4000 was
actually first to implement it) and are thus obviously 64-bit processors. And I'm actually using a 64-bit kernel on my R4k DECstations routinely -- currently I'm torturing it on my 5000/260 with gcc 3.4 bootstraps and it seems quite happy -- it appears rock-solid.
The problem is the official kernel would work fine for a 64-bit
DECstation if it had an R4400 rev.2.0 or later. I haven't heard of any having such a processor -- the ones I have and the others reported by poeple have either an R4000 rev.3.0 or an R4400 rev.1.0. These processors have errata that lead to erroneous behavior in a few common 64-bit operations (according to the errata sheet, the R4000 actually has a serious 32-bit erratum as well, but I haven't been able to trigger it yet). I have implemented appropriate workarounds (available upon request), but they require changes not only to Linux, but to gcc and gas as well. I'm preparing to merge the changes to the tools -- hence my current gcc 3.4 effort -- but until then the 64-bit port has to be marked as experimental (marking R4000 and R4400 processor selections as such for 64-bit operation would be more accurate, but currently we don't have a separate setting for them).
See also arch/mips/dec/prom/call_o32.S for the only chunk of explicit
support code for 64-bit operation for the DECstation -- everything else just works as is (modulo possible protability bugs in drivers).
Going back to the subject -- what's the problem with dependencies?
Maciej
Signature · 3 lines
+ Maciej W. Rozycki, Technical University of Gdansk, Poland + +--------------------------------------------------------------+ + e-mail: [email protected], PGP key available +
Quoted 15 lines · Maciej W. Rozycki
The problem is the official kernel would work fine for a 64-bit
DECstation if it had an R4400 rev.2.0 or later. I haven't heard of any having such a processor -- the ones I have and the others reported by poeple have either an R4000 rev.3.0 or an R4400 rev.1.0. These processors have errata that lead to erroneous behavior in a few common 64-bit operations (according to the errata sheet, the R4000 actually has a serious 32-bit erratum as well, but I haven't been able to trigger it yet). I have implemented appropriate workarounds (available upon request), but they require changes not only to Linux, but to gcc and gas as well. I'm preparing to merge the changes to the tools -- hence my current gcc 3.4 effort -- but until then the 64-bit port has to be marked as experimental (marking R4000 and R4400 processor selections as such for 64-bit operation would be more accurate, but currently we don't have a separate setting for them).
Quoted 5 lines
See also
arch/mips/dec/prom/call_o32.Sfor the only chunk of explicit
support code for 64-bit operation for the DECstation -- everything else just works as is (modulo possible protability bugs in drivers).Going back to the subject -- what's the problem with dependencies?
Nothing. It was looking like you meant something else and me and Adrian got trapped by that. Feel free to change it back to what it was but maybe "depends on MIPS32 || (MIPS64 && EXPERIMENTAL)" is less ambigous?
Ralf
Quoted 5 lines · Ralf Baechle
Quoted 1 line
Going back to the subject -- what's the problem with dependencies?
Nothing. It was looking like you meant something else and me and Adrian got trapped by that. Feel free to change it back to what it was but maybe "depends on MIPS32 || (MIPS64 && EXPERIMENTAL)" is less ambigous?
I thought the construct triggered a problem elsewhere and the change was
supposed to work it around. I'm pretty surprised you've forgotten I ported the DECstation code to 64 bits -- it's already over a year and a half since I did the first boot.
I feel a bit hesitant about changing the expression, but perhaps I may
add a comment to the help text, on why it's experimental for 64-bit. I suspect hardly anyone will notice, though -- if built with unpatched tools and run on a problematic processor the kernel complains and refuses to run asking to contact <[email protected]> and nobody did that yet...
Maciej
Signature · 3 lines
+ Maciej W. Rozycki, Technical University of Gdansk, Poland + +--------------------------------------------------------------+ + e-mail: [email protected], PGP key available +