Re: [uml-devel] tls incrementals for 2.6.15 breaking build ... still?

["D. Bahi" <dbahi@enterasys.com>]

[-- Attachment #1.1: Type: text/plain, Size: 3212 bytes --]

so instead of just my usual whining attached is a suggested fix ;)

D. Bahi wrote:
> i'm trying to build an 'uptodate' UML guest and see exactly this
> 'undefined reference to indirect_set_thread_area' (reported in November)
> with 2.6.15 and all the incrementals from the patch tarball for
> 2006-01-08 on user-mode-linux.sf.net.
>
> the comment related to this patch says CONFIG_MODE_ TT 'off' ... but
> this email says try it 'on' and that just leads to:
>
>   CC      arch/um/os-Linux/tls.o
> arch/um/os-Linux/tls.c:48: error: syntax error before "get_thread_area"
> arch/um/os-Linux/tls.c:48: warning: type defaults to `int' in
> declaration of `_syscall1'
> arch/um/os-Linux/tls.c:48: warning: function declaration isn't a prototype
> arch/um/os-Linux/tls.c:48: warning: data definition has no type or
> storage class
> arch/um/os-Linux/tls.c:49: error: syntax error before "set_thread_area"
> arch/um/os-Linux/tls.c:49: warning: type defaults to `int' in
> declaration of `_syscall1'
> arch/um/os-Linux/tls.c:49: warning: function declaration isn't a prototype
> arch/um/os-Linux/tls.c:49: warning: data definition has no type or
> storage class
> arch/um/os-Linux/tls.c:51: warning: "struct user_desc" declared inside
> parameter list
> arch/um/os-Linux/tls.c:51: warning: its scope is only this definition or
> declaration, which is probably not what you want
> arch/um/os-Linux/tls.c: In function `do_set_thread_area_tt':
> arch/um/os-Linux/tls.c:55: warning: implicit declaration of function
> `set_thread_area'
> arch/um/os-Linux/tls.c: At top level:
> arch/um/os-Linux/tls.c:62: warning: "struct user_desc" declared inside
> parameter list
> arch/um/os-Linux/tls.c: In function `do_get_thread_area_tt':
> arch/um/os-Linux/tls.c:66: warning: implicit declaration of function
> `get_thread_area'
> make[1]: *** [arch/um/os-Linux/tls.o] Error 1
>
>
> Blaisorblade wrote:
>   
>> On Friday 18 November 2005 09:59, Rob Landley wrote:
>>   
>>     
>>> On Friday 18 November 2005 01:58, Blaisorblade wrote:
>>>     
>>>       
>>>>> arch/um/sys-i386/built-in.o(.text+0x3b60): In function `load_TLS':
>>>>> : undefined reference to `indirect_set_thread_area'
>>>>>
>>>>> collect2: ld returned 1 exit status
>>>>>   KSYM    .tmp_kallsyms1.S
>>>>> nm: '.tmp_vmlinux1': No such file
>>>>> No valid symbol.
>>>>> make[2]: *** [.tmp_kallsyms1.S] Error 1
>>>>> make[1]: *** [_all] Error 2
>>>>> make: *** [all] Error 2
>>>>>
>>>>> This is progress!
>>>>>         
>>>>>           
>>>> Disable TT mode or play with such things IIRC
>>>>       
>>>>         
>>> Nope, tt mode already is disabled.
>>>     
>>>       
>> Be creative - enable it then... yep, indirect_... is defined only when TT is 
>> active, but is also used otherwise.
>>
>>   
>>     
>>>> - actually there's a proper
>>>> fix along lines of "move #ifdef's inside sys-i386/tls.c to properly cater
>>>> to dependencies" if you want.
>>>>
>>>> This comes from the TLS patches btw...
>>>>       
>>>>         
>>   
>>     
>>> I'll be banging on this more in a bit...
>>>     
>>>       
>>   
>>     
>>> Rob
>>>     
>>>       
>>   
>>     
>
>   

[-- Warning: decoded text below may be mangled, UTF-8 assumed --]
[-- Attachment #1.2: tls_no_tt.patch --]
[-- Type: text/x-patch; name="tls_no_tt.patch", Size: 502 bytes --]

--- linux-2.6.15/arch/um/sys-i386/tls.c~	2006-01-11 20:04:35.528386774 -0500
+++ linux-2.6.15/arch/um/sys-i386/tls.c	2006-01-11 20:05:45.600027039 -0500
@@ -134,7 +134,7 @@ int load_TLS(int flags, struct task_stru
 		if (!(flags & O_FORCE) && curr->flushed)
 			continue;
 
-		ret = (flags & O_INDIRECT) ? indirect_set_thread_area(&curr->tls): do_set_thread_area(&curr->tls);
+		ret = CHOOSE_MODE_PROC( indirect_set_thread_area, do_set_thread_area, &curr->tls );
 		if (ret)
 			goto out;
 

[-- Attachment #2: OpenPGP digital signature --]
[-- Type: application/pgp-signature, Size: 187 bytes --]