Re: linux-next: scsi tree build failure

Re: linux-next: scsi tree build failure

[James Bottomley]

On Mon, 2009-03-16 at 21:39 +1100, Stephen Rothwell wrote:
> Hi James,
> 
> Today's linux-next build (powerpc allyesconfig) failed like this:
> 
> drivers/message/built-in.o:(.data.rel+0x88): multiple definition of `ioc_list'
> drivers/scsi/built-in.o:(.data.rel+0x2760): first defined here
> 
> Caused by commit 635374e7eb110e80d9918b8611198edd56a32975 ("[SCSI]
> mpt2sas v00.100.11.15").  ioc_list is not a good choice for a global
> name.  The other definition is in drivers/message/fusion/mptbase.c.
> 
> I have reverted that commit for today.

Eric,

Vould we get this one sorted out before the merge window opens? ... the
simple fix seems to be to rename ioc_list in mpt2sas ... and track down
any more global symbols with unfortunate names.

James

Re: linux-next: scsi tree build failure

[James Bottomley]

cc: linux-scsi added

On Thu, 2009-04-02 at 16:49 +1100, Stephen Rothwell wrote:
> Hi James,
> 
> Today's linux-next build (powerpc allyesconfig) failed like this:
> 
> drivers/scsi/fcoe/fcoe.c:982: error: __ksymtab_fcoe_fc_crc causes a section type conflict
> drivers/scsi/fcoe/fcoe.c:982: error: __ksymtab_fcoe_fc_crc causes a section type conflict
> 
> Reverting commit 0c3cb56ec3321b23e9826cccf822470717be6e08 ("[SCSI] fcoe:
> cleans up libfcoe.h and adds fcoe.h for fcoe module") makes the error go
> away.  I have no idea why - the only clue is that the only global
> declaration of fcoe_fc_crc() is removed by that commit.
> 
> Using a different compiler version (4.1.2) also fixed the problem.

Yes, that would be why I didn't see the issue.

> I also had to revert commits 60390544fcc212e0657cf1eea639691d09651943
> ("[SCSI] fcoe, libfc: fix double fcoe_softc memory alloc") and
> ba070644bff4390c965e99b1997f2cbec4157b46 ("[SCSI] fcoe, libfcoe: Add
> support for FIP. FCoE discovery and keep-alive") since they followed the
> one above.
> 
> I wonder if that function even needs to be exported as it is only used
> withing the one file.

I think it's because the above commit removes the external declaration
of fcoe_fc_crc ... so now we have an export symbol with no header
defining it.  I think the correct patch is just not to export the
symbol.

James

---

diff --git a/drivers/scsi/fcoe/fcoe.c b/drivers/scsi/fcoe/fcoe.c
index 5cdda6e..94e1e31 100644
--- a/drivers/scsi/fcoe/fcoe.c
+++ b/drivers/scsi/fcoe/fcoe.c
@@ -979,7 +979,6 @@ u32 fcoe_fc_crc(struct fc_frame *fp)
 	}
 	return crc;
 }
-EXPORT_SYMBOL_GPL(fcoe_fc_crc);
 
 /**
  * fcoe_xmit() - FCoE frame transmit function

Re: linux-next: scsi tree build failure

[Robert Love]

On Thu, 2009-04-02 at 07:22 -0700, James Bottomley wrote:
> cc: linux-scsi added
> 
> On Thu, 2009-04-02 at 16:49 +1100, Stephen Rothwell wrote:
> > Hi James,
> > 
> > Today's linux-next build (powerpc allyesconfig) failed like this:
> > 
> > drivers/scsi/fcoe/fcoe.c:982: error: __ksymtab_fcoe_fc_crc causes a section type conflict
> > drivers/scsi/fcoe/fcoe.c:982: error: __ksymtab_fcoe_fc_crc causes a section type conflict
> > 
> > Reverting commit 0c3cb56ec3321b23e9826cccf822470717be6e08 ("[SCSI] fcoe:
> > cleans up libfcoe.h and adds fcoe.h for fcoe module") makes the error go
> > away.  I have no idea why - the only clue is that the only global
> > declaration of fcoe_fc_crc() is removed by that commit.
> > 
> > Using a different compiler version (4.1.2) also fixed the problem.
> 
> Yes, that would be why I didn't see the issue.
> 
I built using gcc version 4.3.2 (on x86_64) and didn't see this either.

> > I also had to revert commits 60390544fcc212e0657cf1eea639691d09651943
> > ("[SCSI] fcoe, libfc: fix double fcoe_softc memory alloc") and
> > ba070644bff4390c965e99b1997f2cbec4157b46 ("[SCSI] fcoe, libfcoe: Add
> > support for FIP. FCoE discovery and keep-alive") since they followed the
> > one above.
> > 
> > I wonder if that function even needs to be exported as it is only used
> > withing the one file.
> 
> I think it's because the above commit removes the external declaration
> of fcoe_fc_crc ... so now we have an export symbol with no header
> defining it.  I think the correct patch is just not to export the
> symbol.
> 
> James
> 
> ---
> 
> diff --git a/drivers/scsi/fcoe/fcoe.c b/drivers/scsi/fcoe/fcoe.c
> index 5cdda6e..94e1e31 100644
> --- a/drivers/scsi/fcoe/fcoe.c
> +++ b/drivers/scsi/fcoe/fcoe.c
> @@ -979,7 +979,6 @@ u32 fcoe_fc_crc(struct fc_frame *fp)
>  	}
>  	return crc;
>  }
> -EXPORT_SYMBOL_GPL(fcoe_fc_crc);
>  
>  /**
>   * fcoe_xmit() - FCoE frame transmit function
> 
> 
How can we resolve this? Should we provide an updated patch without this
line or a new patch to remove the line?

Re: linux-next: scsi tree build failure

[James Bottomley]

On Thu, 2009-04-02 at 10:19 -0700, Robert Love wrote:
> On Thu, 2009-04-02 at 07:22 -0700, James Bottomley wrote:
> > cc: linux-scsi added
> > 
> > On Thu, 2009-04-02 at 16:49 +1100, Stephen Rothwell wrote:
> > > Hi James,
> > > 
> > > Today's linux-next build (powerpc allyesconfig) failed like this:
> > > 
> > > drivers/scsi/fcoe/fcoe.c:982: error: __ksymtab_fcoe_fc_crc causes a section type conflict
> > > drivers/scsi/fcoe/fcoe.c:982: error: __ksymtab_fcoe_fc_crc causes a section type conflict
> > > 
> > > Reverting commit 0c3cb56ec3321b23e9826cccf822470717be6e08 ("[SCSI] fcoe:
> > > cleans up libfcoe.h and adds fcoe.h for fcoe module") makes the error go
> > > away.  I have no idea why - the only clue is that the only global
> > > declaration of fcoe_fc_crc() is removed by that commit.
> > > 
> > > Using a different compiler version (4.1.2) also fixed the problem.
> > 
> > Yes, that would be why I didn't see the issue.
> > 
> I built using gcc version 4.3.2 (on x86_64) and didn't see this either.
> 
> > > I also had to revert commits 60390544fcc212e0657cf1eea639691d09651943
> > > ("[SCSI] fcoe, libfc: fix double fcoe_softc memory alloc") and
> > > ba070644bff4390c965e99b1997f2cbec4157b46 ("[SCSI] fcoe, libfcoe: Add
> > > support for FIP. FCoE discovery and keep-alive") since they followed the
> > > one above.
> > > 
> > > I wonder if that function even needs to be exported as it is only used
> > > withing the one file.
> > 
> > I think it's because the above commit removes the external declaration
> > of fcoe_fc_crc ... so now we have an export symbol with no header
> > defining it.  I think the correct patch is just not to export the
> > symbol.
> > 
> > James
> > 
> > ---
> > 
> > diff --git a/drivers/scsi/fcoe/fcoe.c b/drivers/scsi/fcoe/fcoe.c
> > index 5cdda6e..94e1e31 100644
> > --- a/drivers/scsi/fcoe/fcoe.c
> > +++ b/drivers/scsi/fcoe/fcoe.c
> > @@ -979,7 +979,6 @@ u32 fcoe_fc_crc(struct fc_frame *fp)
> >  	}
> >  	return crc;
> >  }
> > -EXPORT_SYMBOL_GPL(fcoe_fc_crc);
> >  
> >  /**
> >   * fcoe_xmit() - FCoE frame transmit function
> > 
> > 
> How can we resolve this? Should we provide an updated patch without this
> line or a new patch to remove the line?

Right at the moment, I need someone to verify that this is the fix,
since I don't have a complaining compiler myself.  If it's this simple,
I can adjust the patch accordingly.

James

Re: linux-next: scsi tree build failure

[Stephen Rothwell]

[-- Attachment #1: Type: text/plain, Size: 1129 bytes --]

Hi James,

On Thu, 02 Apr 2009 18:23:42 +0000 James Bottomley <James.Bottomley@HansenPartnership.com> wrote:
>
> On Thu, 2009-04-02 at 10:19 -0700, Robert Love wrote:
> >
> > I built using gcc version 4.3.2 (on x86_64) and didn't see this either.

It may be powerpc specific, I guess.

> > > diff --git a/drivers/scsi/fcoe/fcoe.c b/drivers/scsi/fcoe/fcoe.c
> > > index 5cdda6e..94e1e31 100644
> > > --- a/drivers/scsi/fcoe/fcoe.c
> > > +++ b/drivers/scsi/fcoe/fcoe.c
> > > @@ -979,7 +979,6 @@ u32 fcoe_fc_crc(struct fc_frame *fp)
> > >  	}
> > >  	return crc;
> > >  }
> > > -EXPORT_SYMBOL_GPL(fcoe_fc_crc);
> > >  
> > >  /**
> > >   * fcoe_xmit() - FCoE frame transmit function
> > > 
> > > 
> Right at the moment, I need someone to verify that this is the fix,
> since I don't have a complaining compiler myself.  If it's this simple,
> I can adjust the patch accordingly.

This fixes it for me.  I will apply that patch to linux-next for today
(instead of the reverts I did yesterday).

-- 
Cheers,
Stephen Rothwell                    sfr@canb.auug.org.au
http://www.canb.auug.org.au/~sfr/

[-- Attachment #2: Type: application/pgp-signature, Size: 197 bytes --]

Re: linux-next: scsi tree build failure

[James Bottomley]

On Fri, 2009-04-03 at 11:26 +1100, Stephen Rothwell wrote:
> Hi James,
> 
> On Thu, 02 Apr 2009 18:23:42 +0000 James Bottomley <James.Bottomley@HansenPartnership.com> wrote:
> >
> > On Thu, 2009-04-02 at 10:19 -0700, Robert Love wrote:
> > >
> > > I built using gcc version 4.3.2 (on x86_64) and didn't see this either.
> 
> It may be powerpc specific, I guess.
> 
> > > > diff --git a/drivers/scsi/fcoe/fcoe.c b/drivers/scsi/fcoe/fcoe.c
> > > > index 5cdda6e..94e1e31 100644
> > > > --- a/drivers/scsi/fcoe/fcoe.c
> > > > +++ b/drivers/scsi/fcoe/fcoe.c
> > > > @@ -979,7 +979,6 @@ u32 fcoe_fc_crc(struct fc_frame *fp)
> > > >  	}
> > > >  	return crc;
> > > >  }
> > > > -EXPORT_SYMBOL_GPL(fcoe_fc_crc);
> > > >  
> > > >  /**
> > > >   * fcoe_xmit() - FCoE frame transmit function
> > > > 
> > > > 
> > Right at the moment, I need someone to verify that this is the fix,
> > since I don't have a complaining compiler myself.  If it's this simple,
> > I can adjust the patch accordingly.
> 
> This fixes it for me.  I will apply that patch to linux-next for today
> (instead of the reverts I did yesterday).

Perfect, thanks!

I fixed up the patch in my tree, so you should not need this next time
around.

James

Re: linux-next: scsi tree build failure

[Stephen Rothwell]

[-- Attachment #1: Type: text/plain, Size: 327 bytes --]

Hi James,

On Fri, 03 Apr 2009 01:16:40 +0000 James Bottomley <James.Bottomley@HansenPartnership.com> wrote:
>
> I fixed up the patch in my tree, so you should not need this next time
> around.

Great, thanks.

-- 
Cheers,
Stephen Rothwell                    sfr@canb.auug.org.au
http://www.canb.auug.org.au/~sfr/

[-- Attachment #2: Type: application/pgp-signature, Size: 197 bytes --]

Re: linux-next: scsi tree build failure

[James Bottomley]

On Wed, 2008-12-31 at 03:16 +1100, Stephen Rothwell wrote:
> Hi James,
> 
> Today's linux-next build (sparc defconfig) failed like this:
> 
> drivers/scsi/qlogicpti.c: In function 'qlogicpti_load_firmware':
> drivers/scsi/qlogicpti.c:480: error: 'struct qlogicpti' has no member named 'sdev'
> 
> An interaction between commit 68175399c8b5ee90960e78c1bb8e3353e71fbf33
> ("[SCSI] qlogicpti: use request_firmware") from the scsi tree and commit
> 5dc2536bb8d8adb5fdfbe76cd6fdcdf9de3f40f8 ("qlogicpti: Convert to pure OF
> driver") which is in Linus' tree.
> 
> The version of the firmware commit that has been in the firmware tree
> since Nov 19 has this conflict corrected.
> 
> I have left it for today (as it is getting way to late).  Please fix it
> up for tomorrow.

I'm slightly annoyed this hasn't been building.  I accepted it into the
SCSI tree because I was under the impression it had been well tested in
linux-next.

I can't actually build this driver, but I think this is the fix, if
someone can verify, I'll fold it into the patch.

Looking at the firmware patches, they all seem to have a kconfig
deficiency in that the modified drivers should now be selecting
FW_LOADER ... it this the case?

James

---

diff --git a/drivers/scsi/qlogicpti.c b/drivers/scsi/qlogicpti.c
index e17a4d2..a342426 100644
--- a/drivers/scsi/qlogicpti.c
+++ b/drivers/scsi/qlogicpti.c
@@ -477,7 +477,7 @@ static int __devinit qlogicpti_load_firmware(struct qlogicpti *qpti)
 	unsigned long flags;
 	int i, timeout;
 
-	err = request_firmware(&fw, FIRMWARE_1000, &qpti->sdev->ofdev.dev);
+	err = request_firmware(&fw, FIRMWARE_1000, &qpti->op->dev);
 	if (err) {
 		printk(KERN_ERR "Failed to load image \"%s\" err %d\n",
 		       FIRMWARE_1000, err);

Re: linux-next: scsi tree build failure

[James Bottomley]

Correct cc's added

On Mon, 2008-07-07 at 22:25 +1000, Stephen Rothwell wrote:
> Hi James,
> 
> Today's linux-next build (powerpc ppc64_defconfig) failed like this:
> 
> drivers/scsi/ibmvscsi/ibmvscsi.c: In function 'map_sg_data':
> drivers/scsi/ibmvscsi/ibmvscsi.c:431: error: 'FW_FEATURE_CMO' undeclared (first use in this function)
> drivers/scsi/ibmvscsi/ibmvscsi.c:431: error: (Each undeclared identifier is reported only once
> drivers/scsi/ibmvscsi/ibmvscsi.c:431: error: for each function it appears in.)
> drivers/scsi/ibmvscsi/ibmvscsi.c: In function 'ibmvscsi_queuecommand':
> drivers/scsi/ibmvscsi/ibmvscsi.c:750: error: 'FW_FEATURE_CMO' undeclared (first use in this function)
> drivers/scsi/ibmvscsi/ibmvscsi.c: In function 'send_mad_adapter_info':
> drivers/scsi/ibmvscsi/ibmvscsi.c:864: error: 'FW_FEATURE_CMO' undeclared (first use in this function)
> drivers/scsi/ibmvscsi/ibmvscsi.c: In function 'ibmvscsi_do_host_config':
> drivers/scsi/ibmvscsi/ibmvscsi.c:1412: error: 'FW_FEATURE_CMO' undeclared (first use in this function)
> drivers/scsi/ibmvscsi/ibmvscsi.c: At top level:
> drivers/scsi/ibmvscsi/ibmvscsi.c:1769: error: unknown field 'get_io_entitlement' specified in initializer
> drivers/scsi/ibmvscsi/ibmvscsi.c:1769: warning: missing braces around initializer
> drivers/scsi/ibmvscsi/ibmvscsi.c:1769: warning: (near initialization for 'ibmvscsi_driver.driver')
> drivers/scsi/ibmvscsi/ibmvscsi.c:1769: warning: initialization from incompatible pointer type
> 
> Caused because commit 341b56db6804040aa9559e913865108424e3b18b ("[SCSI]
> ibmvscsi: driver enablement for CMO"), which was 15/16 in a series, has
> been merged before any of the other patches in the series.  I have
> reverted that commit.

Do I detect the fact that IBM sent a patch for a SCSI driver for which
the core features weren't yet enabled (and which I couldn't check, not
having a ppc build system)?

James

Re: linux-next: scsi tree build failure

[Robert Jennings]

* James Bottomley (James.Bottomley@HansenPartnership.com) wrote:
> Correct cc's added
> 
> On Mon, 2008-07-07 at 22:25 +1000, Stephen Rothwell wrote:
> > Hi James,
> > 
> > Today's linux-next build (powerpc ppc64_defconfig) failed like this:
> > 
> > drivers/scsi/ibmvscsi/ibmvscsi.c: In function 'map_sg_data':
> > drivers/scsi/ibmvscsi/ibmvscsi.c:431: error: 'FW_FEATURE_CMO' undeclared (first use in this function)
> > drivers/scsi/ibmvscsi/ibmvscsi.c:431: error: (Each undeclared identifier is reported only once
> > drivers/scsi/ibmvscsi/ibmvscsi.c:431: error: for each function it appears in.)
> > drivers/scsi/ibmvscsi/ibmvscsi.c: In function 'ibmvscsi_queuecommand':
> > drivers/scsi/ibmvscsi/ibmvscsi.c:750: error: 'FW_FEATURE_CMO' undeclared (first use in this function)
> > drivers/scsi/ibmvscsi/ibmvscsi.c: In function 'send_mad_adapter_info':
> > drivers/scsi/ibmvscsi/ibmvscsi.c:864: error: 'FW_FEATURE_CMO' undeclared (first use in this function)
> > drivers/scsi/ibmvscsi/ibmvscsi.c: In function 'ibmvscsi_do_host_config':
> > drivers/scsi/ibmvscsi/ibmvscsi.c:1412: error: 'FW_FEATURE_CMO' undeclared (first use in this function)
> > drivers/scsi/ibmvscsi/ibmvscsi.c: At top level:
> > drivers/scsi/ibmvscsi/ibmvscsi.c:1769: error: unknown field 'get_io_entitlement' specified in initializer
> > drivers/scsi/ibmvscsi/ibmvscsi.c:1769: warning: missing braces around initializer
> > drivers/scsi/ibmvscsi/ibmvscsi.c:1769: warning: (near initialization for 'ibmvscsi_driver.driver')
> > drivers/scsi/ibmvscsi/ibmvscsi.c:1769: warning: initialization from incompatible pointer type
> > 
> > Caused because commit 341b56db6804040aa9559e913865108424e3b18b ("[SCSI]
> > ibmvscsi: driver enablement for CMO"), which was 15/16 in a series, has
> > been merged before any of the other patches in the series.  I have
> > reverted that commit.
> 
> Do I detect the fact that IBM sent a patch for a SCSI driver for which
> the core features weren't yet enabled (and which I couldn't check, not
> having a ppc build system)?
> 
> James

I didn't not properly note in the patch description my intent to have
this merged in the linuxppc-dev tree.  The patch is part of a larger set
enabling the feature.  Linux-scsi was CC'ed for this driver for review.

Robert Jennings

Re: linux-next: scsi tree build failure

[Benjamin Herrenschmidt]

On Mon, 2008-07-07 at 09:39 -0500, James Bottomley wrote:
> > Caused because commit 341b56db6804040aa9559e913865108424e3b18b
> ("[SCSI]
> > ibmvscsi: driver enablement for CMO"), which was 15/16 in a series,
> has
> > been merged before any of the other patches in the series.  I have
> > reverted that commit.
> 
> Do I detect the fact that IBM sent a patch for a SCSI driver for which
> the core features weren't yet enabled (and which I couldn't check, not
> having a ppc build system)?

I suspect the sender (Robert) was asking for comments/review :-) In any
case, Robert, next time make it explicit on a patch CCed to a separate
list from the rest of the serie that it isn't to be merged without
dependencies.

James, if you ack it, I'll put it in powerpc.git.

Cheers,
Ben.

Re: linux-next: scsi tree build failure

[James Bottomley]

On Tue, 2008-07-08 at 08:05 +1000, Benjamin Herrenschmidt wrote:
> On Mon, 2008-07-07 at 09:39 -0500, James Bottomley wrote:
> > > Caused because commit 341b56db6804040aa9559e913865108424e3b18b
> > ("[SCSI]
> > > ibmvscsi: driver enablement for CMO"), which was 15/16 in a series,
> > has
> > > been merged before any of the other patches in the series.  I have
> > > reverted that commit.
> > 
> > Do I detect the fact that IBM sent a patch for a SCSI driver for which
> > the core features weren't yet enabled (and which I couldn't check, not
> > having a ppc build system)?
> 
> I suspect the sender (Robert) was asking for comments/review :-) In any
> case, Robert, next time make it explicit on a patch CCed to a separate
> list from the rest of the serie that it isn't to be merged without
> dependencies.
> 
> James, if you ack it, I'll put it in powerpc.git.

I thought it was all done and dusted (which is why I put it in
scsi-misc).  However, now Brian King seems to have some late arriving
comments.

James

Re: linux-next: scsi tree build failure

[Benjamin Herrenschmidt]

On Mon, 2008-07-07 at 17:07 -0500, James Bottomley wrote:

> I thought it was all done and dusted (which is why I put it in
> scsi-misc).  However, now Brian King seems to have some late arriving
> comments.

There are comments on the whole serie yup, I haven't put it in the
tree yet. 

Cheers,
Ben.