[PATCH] KVM: 'asm' operand has impossible constraints

[PATCH] KVM: 'asm' operand has impossible constraints

[S.Çağlar Onur]

[-- Attachment #1: Type: text/plain, Size: 1142 bytes --]

Hi;

-rc6 fails with latest gcc 4.2 snapshot as following;

CC [M]  drivers/kvm/svm.o
drivers/kvm/svm.c:206: warning: 'inject_db' defined but not used
drivers/kvm/svm.c: In function 'svm_vcpu_run':
drivers/kvm/kvm.h:560: error: 'asm' operand has impossible constraints
make[2]: *** [drivers/kvm/svm.o] Error 1
make[1]: *** [drivers/kvm] Error 2
make: *** [drivers] Error 2

And according to this thread http://lkml.org/lkml/2006/11/11/129, solution is 
converting g to rm, patch follows

Signed-off-by: S.Çağlar Onur <caglar@pardus.org.tr>

Index: linux-2.6/drivers/kvm/kvm.h
===================================================================
--- linux-2.6.orig/drivers/kvm/kvm.h	2007-01-26 01:38:35.000000000 +0200
+++ linux-2.6/drivers/kvm/kvm.h	2007-01-26 01:37:48.000000000 +0200
@@ -557,7 +557,7 @@
 #ifndef load_ldt
 static inline void load_ldt(u16 sel)
 {
-	asm ("lldt %0" : : "g"(sel));
+	asm ("lldt %0" : : "rm"(sel));
 }
 #endif


-- 
S.Çağlar Onur <caglar@pardus.org.tr>
http://cekirdek.pardus.org.tr/~caglar/

Linux is like living in a teepee. No Windows, no Gates and an Apache in house!

[-- Attachment #2: Type: application/pgp-signature, Size: 189 bytes --]

Re: [PATCH] KVM: 'asm' operand has impossible constraints

[Avi Kivity]

S.Çağlar Onur wrote:
> Hi;
>
> -rc6 fails with latest gcc 4.2 snapshot as following;
>
> CC [M]  drivers/kvm/svm.o
> drivers/kvm/svm.c:206: warning: 'inject_db' defined but not used
> drivers/kvm/svm.c: In function 'svm_vcpu_run':
> drivers/kvm/kvm.h:560: error: 'asm' operand has impossible constraints
> make[2]: *** [drivers/kvm/svm.o] Error 1
> make[1]: *** [drivers/kvm] Error 2
> make: *** [drivers] Error 2
>
> And according to this thread http://lkml.org/lkml/2006/11/11/129, solution is 
> converting g to rm, patch follows
>
>   

The patch looks correct, but I don't understand the gcc error message.  
Are we sure this isn't a gcc 4.2 bug?

"g" appears to be equivalent to "rmi", if "i" is impossible, gcc is free 
to use "r" or "m", no?


> Signed-off-by: S.Çağlar Onur <caglar@pardus.org.tr>
>
> Index: linux-2.6/drivers/kvm/kvm.h
> ===================================================================
> --- linux-2.6.orig/drivers/kvm/kvm.h	2007-01-26 01:38:35.000000000 +0200
> +++ linux-2.6/drivers/kvm/kvm.h	2007-01-26 01:37:48.000000000 +0200
> @@ -557,7 +557,7 @@
>  #ifndef load_ldt
>  static inline void load_ldt(u16 sel)
>  {
> -	asm ("lldt %0" : : "g"(sel));
> +	asm ("lldt %0" : : "rm"(sel));
>  }
>  #endif
>
>
>   


-- 
Do not meddle in the internals of kernels, for they are subtle and quick to panic.

Re: [PATCH] KVM: 'asm' operand has impossible constraints

[Paweł Sikora]

On Saturday 27 of January 2007 10:05:53 Avi Kivity wrote:

> "g" appears to be equivalent to "rmi", if "i" is impossible, gcc is free
> to use "r" or "m", no?

`r'
     A register operand is allowed provided that it is in a general
     register.
`g'
     Any register, memory or immediate integer operand is allowed,
     except for registers that are not general registers.

so, it looks like g == !r for registers ( not general vs. general regs ).

Re: [PATCH] KVM: 'asm' operand has impossible constraints

[S.Çağlar Onur]

[-- Attachment #1: Type: text/plain, Size: 717 bytes --]

27 Oca 2007 Cts tarihinde, Avi Kivity şunları yazmıştı: 
> The patch looks correct, but I don't understand the gcc error message.
> Are we sure this isn't a gcc 4.2 bug?
>
> "g" appears to be equivalent to "rmi", if "i" is impossible, gcc is free
> to use "r" or "m", no?

Accorgind to GCC devs. its not a bug 
(http://gcc.gnu.org/bugzilla/show_bug.cgi?id=29808), on comment #5 the 
problem described like;

"g" means "r"+"i" so the register allocator in the -O0 case is selecting "r" 
while in the optimize case is selecting "i"

-- 
S.Çağlar Onur <caglar@pardus.org.tr>
http://cekirdek.pardus.org.tr/~caglar/

Linux is like living in a teepee. No Windows, no Gates and an Apache in house!

[-- Attachment #2: Type: application/pgp-signature, Size: 189 bytes --]

Re: [PATCH] KVM: 'asm' operand has impossible constraints

[D. Hazelton]

On Saturday 27 January 2007 16:28, S.Çağlar Onur wrote:
> 27 Oca 2007 Cts tarihinde, Avi Kivity şunları yazmıştı:
> > The patch looks correct, but I don't understand the gcc error message.
> > Are we sure this isn't a gcc 4.2 bug?
> >
> > "g" appears to be equivalent to "rmi", if "i" is impossible, gcc is free
> > to use "r" or "m", no?
>
> Accorgind to GCC devs. its not a bug
> (http://gcc.gnu.org/bugzilla/show_bug.cgi?id=29808), on comment #5 the
> problem described like;
>
> "g" means "r"+"i" so the register allocator in the -O0 case is selecting
> "r" while in the optimize case is selecting "i"

Sounds like a bug to me! After all, shouldn't the different sections of code 
be selecting the *same* bits ?

DRH

Re: [PATCH] KVM: 'asm' operand has impossible constraints

[H. Peter Anvin]

S.Çağlar Onur wrote:
> 
> "g" means "r"+"i" so the register allocator in the -O0 case is selecting "r" 
> while in the optimize case is selecting "i"
> 

"g" means "rmi", not "ri".

	-hpa

Re: [PATCH] KVM: 'asm' operand has impossible constraints

[Avi Kivity]

Paweł Sikora wrote:
> On Saturday 27 of January 2007 10:05:53 Avi Kivity wrote:
>
>   
>> "g" appears to be equivalent to "rmi", if "i" is impossible, gcc is free
>> to use "r" or "m", no?
>>     
>
> `r'
>      A register operand is allowed provided that it is in a general
>      register.
> `g'
>      Any register, memory or immediate integer operand is allowed,
>      except for registers that are not general registers.
>
> so, it looks like g == !r for registers ( not general vs. general regs ).
>   

I read it as g == !!r for registers ("except" and "not" constitute a 
double negative).


-- 
error compiling committee.c: too many arguments to function

Re: [PATCH] KVM: 'asm' operand has impossible constraints

[Avi Kivity]

S.Çağlar Onur wrote:
> 27 Oca 2007 Cts tarihinde, Avi Kivity şunları yazmıştı: 
>   
>> The patch looks correct, but I don't understand the gcc error message.
>> Are we sure this isn't a gcc 4.2 bug?
>>
>> "g" appears to be equivalent to "rmi", if "i" is impossible, gcc is free
>> to use "r" or "m", no?
>>     
>
> Accorgind to GCC devs. its not a bug 
> (http://gcc.gnu.org/bugzilla/show_bug.cgi?id=29808), on comment #5 the 
> problem described like;
>
> "g" means "r"+"i" so the register allocator in the -O0 case is selecting "r" 
> while in the optimize case is selecting "i"
>
>   

That's a different bug. The gcc PR has an assembler error message, as 
expected, whereas the kvm miscompile has a compiler error, which I don't 
understand.

Anyhow, your patch is correct (because selecting "i" is indeed 
erroneous) so I'll apply it, but I'm worried that there's a gcc bug in 
there that we ought to report.


-- 
error compiling committee.c: too many arguments to function