pvops dom0 compile issues

pvops dom0 compile issues

[Michael D Labriola]

[-- Attachment #1: Type: text/plain, Size: 1449 bytes --]

Jeremy,

I've been playing with your pvops tree for a few weeks pretty 
successfully.
My first attempt had some strange linking errors that I wrote off as a
misconfiguration on my part.  I just pulled fresh from your xen/master
branch and snagged the example 32bit .config from the pvops Wiki and tried
to do a fresh compile.

The first problem I had was some inlining errors in mptsas.c.  I'm 
assuming
that since that driver is enabled in the example .config that it compiles
for you... which means it's related to compiler differences perhaps?
Anyway, I attached a patch that fixed this.  Basically just moved the
offending function to before it's first invocation.

Now that that compile error is fixed, I get all the way to linking and get
this:

drivers/built-in.o: In function `kmalloc':
include/trace/events/kmem.h:47: undefined reference to `.L1445'

I also saw some warnings suggesting I do a 'make
CONFIG_DEBUG_SECTION_MISMATCH=y', so I've done that and attached the last
page or so of output as well.

I got these errors to go away last time I compiled by gutting ALL the
debug stuff out of my .config... but I don't want to do that.

Any ideas?  I'm trying to get up to speed on this stuff, but am definitely
feeling a bit over my head.  ;-)

-Mike

PS - I'm using gcc 3.4.6 on a custom linux distro forked off of fc5 a long
time ago...

 
---
Michael D Labriola
Electric Boat
mlabriol@gdeb.com
401-848-8871 (office)
401-316-9844 (cell)

[-- Attachment #2: mptsas_inline.patch --]
[-- Type: application/octet-stream, Size: 1632 bytes --]

diff --git a/drivers/message/fusion/mptsas.c b/drivers/message/fusion/mptsas.c
index 55ff252..8d69958 100644
--- a/drivers/message/fusion/mptsas.c
+++ b/drivers/message/fusion/mptsas.c
@@ -441,6 +441,23 @@ mptsas_is_end_device(struct mptsas_devinfo * attached)
 		return 0;
 }
 
+static inline void
+mptsas_set_rphy(MPT_ADAPTER *ioc, struct mptsas_phyinfo *phy_info, struct sas_rphy *rphy)
+{
+	if (phy_info->port_details) {
+		phy_info->port_details->rphy = rphy;
+		dsaswideprintk(ioc, printk(MYIOC_s_DEBUG_FMT "sas_rphy_add: rphy=%p\n",
+		    ioc->name, rphy));
+	}
+
+	if (rphy) {
+		dsaswideprintk(ioc, dev_printk(KERN_DEBUG,
+		    &rphy->dev, MYIOC_s_FMT "add:", ioc->name));
+		dsaswideprintk(ioc, printk(MYIOC_s_DEBUG_FMT "rphy=%p release=%p\n",
+		    ioc->name, rphy, rphy->dev.release));
+	}
+}
+
 /* no mutex */
 static void
 mptsas_port_delete(MPT_ADAPTER *ioc, struct mptsas_portinfo_details * port_details)
@@ -479,23 +496,6 @@ mptsas_get_rphy(struct mptsas_phyinfo *phy_info)
 		return NULL;
 }
 
-static inline void
-mptsas_set_rphy(MPT_ADAPTER *ioc, struct mptsas_phyinfo *phy_info, struct sas_rphy *rphy)
-{
-	if (phy_info->port_details) {
-		phy_info->port_details->rphy = rphy;
-		dsaswideprintk(ioc, printk(MYIOC_s_DEBUG_FMT "sas_rphy_add: rphy=%p\n",
-		    ioc->name, rphy));
-	}
-
-	if (rphy) {
-		dsaswideprintk(ioc, dev_printk(KERN_DEBUG,
-		    &rphy->dev, MYIOC_s_FMT "add:", ioc->name));
-		dsaswideprintk(ioc, printk(MYIOC_s_DEBUG_FMT "rphy=%p release=%p\n",
-		    ioc->name, rphy, rphy->dev.release));
-	}
-}
-
 static inline struct sas_port *
 mptsas_get_port(struct mptsas_phyinfo *phy_info)
 {

[-- Attachment #3: compile.log --]
[-- Type: application/octet-stream, Size: 2230 bytes --]

  CC      arch/x86/lib/memcpy_32.o
  AS      arch/x86/lib/putuser.o
  CC      arch/x86/lib/string_32.o
  CC      arch/x86/lib/strstr_32.o
  CC      arch/x86/lib/usercopy_32.o
  AR      arch/x86/lib/lib.a
  LD      vmlinux.o
  MODPOST vmlinux.o
WARNING: vmlinux.o(.init.text+0x2b417): Section mismatch in reference from the function netchan2_init() to the function .exit.text:netfront2_exit()
The function __init netchan2_init() references
a function __exit netfront2_exit().
This is often seen when error handling in the init function
uses functionality in the exit path.
The fix is often to remove the __exit annotation of
netfront2_exit() so it may be used outside an exit section.

WARNING: vmlinux.o(.data+0x1a950): Section mismatch in reference from the variable cpu_evtchn_mask_p to the variable .init.data:init_evtchn_mask
The variable cpu_evtchn_mask_p references
the variable __initdata init_evtchn_mask
If the reference is valid then annotate the
variable with __init* or __refdata (see linux/init.h) or name the variable:
*driver, *_template, *_timer, *_sht, *_ops, *_probe, *_probe_one, *_console, 

WARNING: vmlinux.o(.data+0x2710c): Section mismatch in reference from the variable netfront2 to the function .devinit.text:netfront_probe()
The variable netfront2 references
the function __devinit netfront_probe()
If the reference is valid then annotate the
variable with __init* or __refdata (see linux/init.h) or name the variable:
*driver, *_template, *_timer, *_sht, *_ops, *_probe, *_probe_one, *_console, 

WARNING: vmlinux.o(.data+0x27114): Section mismatch in reference from the variable netfront2 to the function .devexit.text:netfront_remove()
The variable netfront2 references
the function __devexit netfront_remove()
If the reference is valid then annotate the
variable with __exit* (see linux/init.h) or name the variable:
*driver, *_template, *_timer, *_sht, *_ops, *_probe, *_probe_one, *_console, 

  GEN     .version
  CHK     include/linux/compile.h
  UPD     include/linux/compile.h
  CC      init/version.o
  LD      init/built-in.o
  LD      .tmp_vmlinux1
drivers/built-in.o: In function `kmalloc':
include/trace/events/kmem.h:47: undefined reference to `.L1445'
make: *** [.tmp_vmlinux1] Error 1

[-- Attachment #4: Type: text/plain, Size: 138 bytes --]

_______________________________________________
Xen-devel mailing list
Xen-devel@lists.xensource.com
http://lists.xensource.com/xen-devel

Re: pvops dom0 compile issues

[Konrad Rzeszutek Wilk]

> Any ideas?  I'm trying to get up to speed on this stuff, but am definitely
> feeling a bit over my head.  ;-)

Besides the 'CONFIG_DEBUG_SECTION_MISMATCH' I am not seeing any of the errors
you saw. The compiler I am using is 4.1.2 (SLES10 SP2) and 4.4.0(FC11), so nothing
as ancient as 3.4.6.


Does a vanialla Linux 2.6 link with the .config file? Or does it fail
at the same spot?

Re: pvops dom0 compile issues

[Michael D Labriola]

xen-devel-bounces@lists.xensource.com wrote on 12/04/2009 09:17:25 AM:

> > Any ideas?  I'm trying to get up to speed on this stuff, but am 
definitely
> > feeling a bit over my head.  ;-)
> 
> Besides the 'CONFIG_DEBUG_SECTION_MISMATCH' I am not seeing any of the 
errors
> you saw. The compiler I am using is 4.1.2 (SLES10 SP2) and 4.4.
> 0(FC11), so nothing
> as ancient as 3.4.6.
> 
> 
> Does a vanialla Linux 2.6 link with the .config file? Or does it fail
> at the same spot?

I've been compiling 2.6.28 with a similar .config w/out problems.  I'll go 
try 2.6.31.4 kernel.org via make oldconfig and see what happens.

-Mike

Re: pvops dom0 compile issues

[Michael D Labriola]

xen-devel-bounces@lists.xensource.com wrote on 12/04/2009 09:44:42 AM:

> xen-devel-bounces@lists.xensource.com wrote on 12/04/2009 09:17:25 AM:
> 
> > > Any ideas?  I'm trying to get up to speed on this stuff, but am 
> definitely
> > > feeling a bit over my head.  ;-)
> > 
> > Besides the 'CONFIG_DEBUG_SECTION_MISMATCH' I am not seeing any of the 

> errors
> > you saw. The compiler I am using is 4.1.2 (SLES10 SP2) and 4.4.
> > 0(FC11), so nothing
> > as ancient as 3.4.6.
> > 
> > 
> > Does a vanialla Linux 2.6 link with the .config file? Or does it fail
> > at the same spot?
> 
> I've been compiling 2.6.28 with a similar .config w/out problems.  I'll 
go 
> try 2.6.31.4 kernel.org via make oldconfig and see what happens.
> 

Just got the same errors on v2.6.31.4 with that .config.

-Mike

Re: pvops dom0 compile issues

[Michael D Labriola]

xen-devel-bounces@lists.xensource.com wrote on 12/04/2009 11:16:19 AM:

> xen-devel-bounces@lists.xensource.com wrote on 12/04/2009 09:44:42 AM:
> 
> > xen-devel-bounces@lists.xensource.com wrote on 12/04/2009 09:17:25 AM:
> > 
> > > > Any ideas?  I'm trying to get up to speed on this stuff, but am 
> > definitely
> > > > feeling a bit over my head.  ;-)
> > > 
> > > Besides the 'CONFIG_DEBUG_SECTION_MISMATCH' I am not seeing any of 
the 
> 
> > errors
> > > you saw. The compiler I am using is 4.1.2 (SLES10 SP2) and 4.4.
> > > 0(FC11), so nothing
> > > as ancient as 3.4.6.
> > > 
> > > 
> > > Does a vanialla Linux 2.6 link with the .config file? Or does it 
fail
> > > at the same spot?
> > 
> > I've been compiling 2.6.28 with a similar .config w/out problems. I'll 

> go 
> > try 2.6.31.4 kernel.org via make oldconfig and see what happens.
> > 
> 
> Just got the same errors on v2.6.31.4 with that .config.
> 

I just double checked that v2.6.28 compiles fine.  I had to disable some 
mac laptop backlight drivers, but no link error.  So, somewhere between 
2.6.28 and 2.6.31.4 we lost the ability to link with gcc 3.4.6?

Suppose I'm gonna have to post this on the kernel-devel list, huh?  Wonder 
if they'll all just laugh at me for using such an "ancient" version of 
gcc.  ;-)

-Mike

Re: pvops dom0 compile issues

[Jeremy Fitzhardinge]

On 12/04/09 04:49, Michael D Labriola wrote:
> Jeremy,
>
> I've been playing with your pvops tree for a few weeks pretty
> successfully.
> My first attempt had some strange linking errors that I wrote off as a
> misconfiguration on my part.  I just pulled fresh from your xen/master
> branch and snagged the example 32bit .config from the pvops Wiki and tried
> to do a fresh compile.
>
> The first problem I had was some inlining errors in mptsas.c.  I'm
> assuming
> that since that driver is enabled in the example .config that it compiles
> for you... which means it's related to compiler differences perhaps?
> Anyway, I attached a patch that fixed this.  Basically just moved the
> offending function to before it's first invocation.
>
> Now that that compile error is fixed, I get all the way to linking and get
> this:
>
> drivers/built-in.o: In function `kmalloc':
> include/trace/events/kmem.h:47: undefined reference to `.L1445'
>
> I also saw some warnings suggesting I do a 'make
> CONFIG_DEBUG_SECTION_MISMATCH=y', so I've done that and attached the last
> page or so of output as well.
>
> I got these errors to go away last time I compiled by gutting ALL the
> debug stuff out of my .config... but I don't want to do that.
>
> Any ideas?  I'm trying to get up to speed on this stuff, but am definitely
> feeling a bit over my head.  ;-)
>
> -Mike
>
> PS - I'm using gcc 3.4.6 on a custom linux distro forked off of fc5 a long
> time ago...
>    

gcc 3.4.6 is known to have compilation problems with pvops kernels, 
apparently because of 42854dc0a6320.  Bug 
http://bugzilla.kernel.org/show_bug.cgi?id=14729 is tracking it, but its 
not clear what the right fix is.  It's interesting you're seeing a 
different failure mode.

I was wondering if anyone is still using 3.4.6, but I guess you've 
answered that question. Nevertheless, I'd recommend upgrading if you can.

     J

Re: pvops dom0 compile issues

[Michael D Labriola]

xen-devel-bounces@lists.xensource.com wrote on 12/04/2009 02:26:00 PM:

> On 12/04/09 04:49, Michael D Labriola wrote:
> > Jeremy,
> >
> > I've been playing with your pvops tree for a few weeks pretty
> > successfully.
> > My first attempt had some strange linking errors that I wrote off as a
> > misconfiguration on my part.  I just pulled fresh from your xen/master
> > branch and snagged the example 32bit .config from the pvops Wiki and 
tried
> > to do a fresh compile.
> >
> > The first problem I had was some inlining errors in mptsas.c.  I'm
> > assuming
> > that since that driver is enabled in the example .config that it 
compiles
> > for you... which means it's related to compiler differences perhaps?
> > Anyway, I attached a patch that fixed this.  Basically just moved the
> > offending function to before it's first invocation.
> >
> > Now that that compile error is fixed, I get all the way to linking and 
get
> > this:
> >
> > drivers/built-in.o: In function `kmalloc':
> > include/trace/events/kmem.h:47: undefined reference to `.L1445'
> >
> > I also saw some warnings suggesting I do a 'make
> > CONFIG_DEBUG_SECTION_MISMATCH=y', so I've done that and attached the 
last
> > page or so of output as well.
> >
> > I got these errors to go away last time I compiled by gutting ALL the
> > debug stuff out of my .config... but I don't want to do that.
> >
> > Any ideas?  I'm trying to get up to speed on this stuff, but am 
definitely
> > feeling a bit over my head.  ;-)
> >
> > -Mike
> >
> > PS - I'm using gcc 3.4.6 on a custom linux distro forked off of fc5 a 
long
> > time ago...
> > 
> 
> gcc 3.4.6 is known to have compilation problems with pvops kernels, 
> apparently because of 42854dc0a6320.  Bug 
> http://bugzilla.kernel.org/show_bug.cgi?id=14729 is tracking it, but its 

> not clear what the right fix is.  It's interesting you're seeing a 
> different failure mode.
> 

I just tried upgrading to the latest binutils from gnu.org, since I'm 
failing on an ld call, but that didn't help at all.

I took a look at bug 14729 and 13503 (which involve paravirt.h compilation 
errors) and bug 14153 (which sounds more like ld problem I'm having). Very 
odd that I don't see the asm/paravirt.h errors...  I'm trying out the 
patch you stuck in 14153 right now.

> I was wondering if anyone is still using 3.4.6, but I guess you've 
> answered that question. Nevertheless, I'd recommend upgrading if you 
can.
> 

I have some very unfortunate dependancies on stuff that doesn't compile 
with gcc >= 4.0.  I'll have to figure out which is worse to work through.

Re: pvops dom0 compile issues

[Michael D Labriola]

xen-devel-bounces@lists.xensource.com wrote on 12/04/2009 03:52:50 PM:

> xen-devel-bounces@lists.xensource.com wrote on 12/04/2009 02:26:00 PM:
> 
> > On 12/04/09 04:49, Michael D Labriola wrote:
> > > Jeremy,
> > >
> > > I've been playing with your pvops tree for a few weeks pretty
> > > successfully.
> > > My first attempt had some strange linking errors that I wrote off as 
a
> > > misconfiguration on my part.  I just pulled fresh from your 
xen/master
> > > branch and snagged the example 32bit .config from the pvops Wiki and 

> tried
> > > to do a fresh compile.
> > >
> > > The first problem I had was some inlining errors in mptsas.c.  I'm
> > > assuming
> > > that since that driver is enabled in the example .config that it 
> compiles
> > > for you... which means it's related to compiler differences perhaps?
> > > Anyway, I attached a patch that fixed this.  Basically just moved 
the
> > > offending function to before it's first invocation.
> > >
> > > Now that that compile error is fixed, I get all the way to linking 
and 
> get
> > > this:
> > >
> > > drivers/built-in.o: In function `kmalloc':
> > > include/trace/events/kmem.h:47: undefined reference to `.L1445'
> > >
> > > I also saw some warnings suggesting I do a 'make
> > > CONFIG_DEBUG_SECTION_MISMATCH=y', so I've done that and attached the 

> last
> > > page or so of output as well.
> > >
> > > I got these errors to go away last time I compiled by gutting ALL 
the
> > > debug stuff out of my .config... but I don't want to do that.
> > >
> > > Any ideas?  I'm trying to get up to speed on this stuff, but am 
> definitely
> > > feeling a bit over my head.  ;-)
> > >
> > > -Mike
> > >
> > > PS - I'm using gcc 3.4.6 on a custom linux distro forked off of fc5 
a 
> long
> > > time ago...
> > > 
> > 
> > gcc 3.4.6 is known to have compilation problems with pvops kernels, 
> > apparently because of 42854dc0a6320.  Bug 
> > http://bugzilla.kernel.org/show_bug.cgi?id=14729 is tracking it, but 
its 
> 
> > not clear what the right fix is.  It's interesting you're seeing a 
> > different failure mode.
> > 
> 
> I just tried upgrading to the latest binutils from gnu.org, since I'm 
> failing on an ld call, but that didn't help at all.
> 
> I took a look at bug 14729 and 13503 (which involve paravirt.h 
compilation 
> errors) and bug 14153 (which sounds more like ld problem I'm having). 
Very 
> odd that I don't see the asm/paravirt.h errors...  I'm trying out the 
> patch you stuck in 14153 right now.
> 

That didn't help either.  I didn't really think it would.

> > I was wondering if anyone is still using 3.4.6, but I guess you've 
> > answered that question. Nevertheless, I'd recommend upgrading if you 
> can.
> > 
> 
> I have some very unfortunate dependancies on stuff that doesn't compile 
> with gcc >= 4.0.  I'll have to figure out which is worse to work 
through.
> 
> _______________________________________________
> Xen-devel mailing list
> Xen-devel@lists.xensource.com
> http://lists.xensource.com/xen-devel

Re: pvops dom0 compile issues

[Konrad Rzeszutek Wilk]

> I have some very unfortunate dependancies on stuff that doesn't compile 
> with gcc >= 4.0.  I'll have to figure out which is worse to work through.

Oooh that complicates things.
> 
You could compile the kernel using a different compiler than your applications.
You just need to make sure that the modules were compiled using the same
compiler as the kernel. 

Re: pvops dom0 compile issues

[Jeremy Fitzhardinge]

On 12/07/09 06:37, Konrad Rzeszutek Wilk wrote:
>> I have some very unfortunate dependancies on stuff that doesn't compile
>> with gcc>= 4.0.  I'll have to figure out which is worse to work through.
>>      
> Oooh that complicates things.
>    
>>      
> You could compile the kernel using a different compiler than your applications.
> You just need to make sure that the modules were compiled using the same
> compiler as the kernel.
>    

It's possible to install multiple versions of gcc.  Fedora, for example, 
offers the "compat-gcc-34*" set of packages which can be installed in 
parallel with the current gcc versions.

     J