lpfc: problems with -Werror

lpfc: problems with -Werror

[Fengguang Wu]

Hi Alex,

In commit d438bd3aef ("Add -Werror compilation flag") you added this:

+ccflags-y += -Werror

Now I run into a number of gcc errors when running randconfig tests.
They smell like false warnings. I don't quite understand why you made
that "-Werror" conditional and only fail on CONFIG_SCSI_LPFC=y.  But
if you prefer to keep that -Werror, I'll send a (perhaps silly) patch
to quiet these error messages.

drivers/scsi/lpfc/lpfc_sli.c:3654:3: error: 'resp_data' may be used uninitialized in this function [-Werror=maybe-uninitialized]
drivers/scsi/lpfc/lpfc_sli.c:3637:2: error: 'ha_copy' may be used uninitialized in this function [-Werror=maybe-uninitialized]
arch/x86/include/asm/io.h:63:1: error: 'hc_copy' may be used uninitialized in this function [-Werror=maybe-uninitialized]
drivers/scsi/lpfc/lpfc_sli.c:3510:18: error: 'status' may be used uninitialized in this function [-Werror=maybe-uninitialized]
drivers/scsi/lpfc/lpfc_sli.c:4068:17: error: 'status' may be used uninitialized in this function [-Werror=maybe-uninitialized]
drivers/scsi/lpfc/lpfc_sli.c:3777:23: error: 'ha_copy' may be used uninitialized in this function [-Werror=maybe-uninitialized]
drivers/scsi/lpfc/lpfc_sli.c:3742:9: error: 'status' may be used uninitialized in this function [-Werror=maybe-uninitialized]
drivers/scsi/lpfc/lpfc_sli.c:6694:21: error: 'ha_copy' may be used uninitialized in this function [-Werror=maybe-uninitialized]
drivers/scsi/lpfc/lpfc_sli.c:6693:18: error: 'word0' may be used uninitialized in this function [-Werror=maybe-uninitialized]
drivers/scsi/lpfc/lpfc_sli.c:9718:12: error: 'creg_val' may be used uninitialized in this function [-Werror=maybe-uninitialized]
drivers/scsi/lpfc/lpfc_sli.c:9872:2: error: 'ha_copy' may be used uninitialized in this function [-Werror=maybe-uninitialized]
drivers/scsi/lpfc/lpfc_sli.c:10235:14: error: 'control' may be used uninitialized in this function [-Werror=maybe-uninitialized]
drivers/scsi/lpfc/lpfc_sli.c:10176:9: error: 'hc_copy' may be used uninitialized in this function [-Werror=maybe-uninitialized]
drivers/scsi/lpfc/lpfc_sli.c:10152:12: error: 'ha_copy' may be used uninitialized in this function [-Werror=maybe-uninitialized]
drivers/scsi/lpfc/lpfc_sli.c:10447:9: error: 'ha_copy' may be used uninitialized in this function [-Werror=maybe-uninitialized]
drivers/scsi/lpfc/lpfc_sli.c:10551:8: error: 'hc_copy' may be used uninitialized in this function [-Werror=maybe-uninitialized]
drivers/scsi/lpfc/lpfc_els.c:95:2: error: 'ha_copy' may be used uninitialized in this function [-Werror=maybe-uninitialized]
drivers/scsi/lpfc/lpfc_els.c:4731:11: error: 'cmd' may be used uninitialized in this function [-Werror=maybe-uninitialized]
drivers/scsi/lpfc/lpfc_els.c:4731:11: error: 'length' may be used uninitialized in this function [-Werror=maybe-uninitialized]
drivers/scsi/lpfc/lpfc_init.c:522:9: error: 'status' may be used uninitialized in this function [-Werror=maybe-uninitialized]
drivers/scsi/lpfc/lpfc_attr.c:1491:12: error: 'creg_val' may be used uninitialized in this function [-Werror=maybe-uninitialized]
drivers/scsi/lpfc/lpfc_bsg.c:358:12: error: 'creg_val' may be used uninitialized in this function [-Werror=maybe-uninitialized]
drivers/scsi/lpfc/lpfc_bsg.c:615:12: error: 'creg_val' may be used uninitialized in this function [-Werror=maybe-uninitialized]
drivers/scsi/lpfc/lpfc_bsg.c:1385:12: error: 'creg_val' may be used uninitialized in this function [-Werror=maybe-uninitialized]

Thanks,
Fengguang

Re: lpfc: problems with -Werror

[James Bottomley]

On Thu, 2012-06-21 at 09:30 +0800, Fengguang Wu wrote:
> Hi Alex,
> 
> In commit d438bd3aef ("Add -Werror compilation flag") you added this:
> 
> +ccflags-y += -Werror
> 
> Now I run into a number of gcc errors when running randconfig tests.
> They smell like false warnings. I don't quite understand why you made
> that "-Werror" conditional and only fail on CONFIG_SCSI_LPFC=y.  But
> if you prefer to keep that -Werror, I'll send a (perhaps silly) patch
> to quiet these error messages.

They're all bogus, but we can't really have the build break because gcc
is fragile.

James and Alex, how about this: It's the aic7xxx solution: it allows you
to build the kernel with make WARNINGS_BECOME_ERRORS=1 to get -Werror
but it's not on by default?

James

---

diff --git a/drivers/scsi/lpfc/Makefile b/drivers/scsi/lpfc/Makefile
index fe5d396..e2516ba 100644
--- a/drivers/scsi/lpfc/Makefile
+++ b/drivers/scsi/lpfc/Makefile
@@ -22,7 +22,9 @@
 ccflags-$(GCOV) := -fprofile-arcs -ftest-coverage
 ccflags-$(GCOV) += -O0
 
+ifdef WARNINGS_BECOME_ERRORS
 ccflags-y += -Werror
+endif
 
 obj-$(CONFIG_SCSI_LPFC) := lpfc.o
 

Re: lpfc: problems with -Werror

[Matthew Wilcox]

On Thu, Jun 21, 2012 at 07:50:02AM +0000, James Bottomley wrote:
> @@ -22,7 +22,9 @@
>  ccflags-$(GCOV) := -fprofile-arcs -ftest-coverage
>  ccflags-$(GCOV) += -O0
>  
> +ifdef WARNINGS_BECOME_ERRORS
>  ccflags-y += -Werror
> +endif
>  
>  obj-$(CONFIG_SCSI_LPFC) := lpfc.o

Why not more simply:

ccflags-$(WARNINGS_BECOME_ERRORS) += -Werror

> N?????r??y????b?X????v?^?)??{.n?+????{???"?{ay?\x1d
?????,j\a??f???h???z?\x1e
?w???\f
???j:+v???w?j?m????\a????zZ+???????j"??!

That's some nasty line noise on your modem there, James.

-- 
Matthew Wilcox				Intel Open Source Technology Centre
"Bill, look, we understand that you're interested in selling us this
operating system, but compare it to ours.  We can't possibly take such
a retrograde step."

Re: lpfc: problems with -Werror

[James Bottomley]

On Fri, 2012-06-22 at 12:46 -0600, Matthew Wilcox wrote:
> On Thu, Jun 21, 2012 at 07:50:02AM +0000, James Bottomley wrote:
> > @@ -22,7 +22,9 @@
> >  ccflags-$(GCOV) := -fprofile-arcs -ftest-coverage
> >  ccflags-$(GCOV) += -O0
> >  
> > +ifdef WARNINGS_BECOME_ERRORS
> >  ccflags-y += -Werror
> > +endif
> >  
> >  obj-$(CONFIG_SCSI_LPFC) := lpfc.o
> 
> Why not more simply:
> 
> ccflags-$(WARNINGS_BECOME_ERRORS) += -Werror

Because it's not a config option, it's a makefile option.  If we go with
what you propose, the format of make has to be

make WARNINGS_BECOME_ERRORS=y

which is a bit non intuitive.

> > N?????r??y????b?X????v?^?)??{.n?+????{???"?{ay?\x1d
?????,j\a??f???h???z?\x1e
?w???\f
???j:+v???w?j?m????\a????zZ+???????j"??!
> 
> That's some nasty line noise on your modem there, James.

No, it's because majordomo has some issue doing base64.

James