qla1280 compile problems

qla1280 compile problems

[Christoph Hellwig]

The qla1280 update still has some issues, mostly because it backs
out all the 2.5 changes applied to the driver in the last months,
also it doesn't work on ia64 because the IA64 generic build references
sn2-specific symbols that don't exist in that build.

Jes, can you look at the 2.5 changes that went in after your last merge
and make the SN2 vchan workaround a config option?

For reference, this is the compile errors on ppc:

In file included from drivers/scsi/qla1280.c:299:
include/linux/blk.h:1:2: warning: #warning this file is obsolete, please use <linux/blkdev.h> instead
drivers/scsi/qla1280.c: In function `qla1280_done':
drivers/scsi/qla1280.c:1833: warning: implicit declaration of function `qla1280_get_target_options'
drivers/scsi/qla1280.c: At top level:
drivers/scsi/qla1280.c:4506: warning: type mismatch with previous implicit declaration
drivers/scsi/qla1280.c:1833: warning: previous implicit declaration of `qla1280_get_target_options'
drivers/scsi/qla1280.c:4506: warning: `qla1280_get_target_options' was previously implicitly declared to return `int'
drivers/scsi/qla1280.c:4830: warning: initialization from incompatible pointer type
drivers/scsi/qla1280.c:4830: error: unknown field `command' specified in initializer

qla1280 compile problems

[Christoph Hellwig]

[and now with jes's _full_ email address for an actually working
 delivery.  did I mention I hate email today?]

The qla1280 update still has some issues, mostly because it backs
out all the 2.5 changes applied to the driver in the last months,
also it doesn't work on ia64 because the IA64 generic build references
sn2-specific symbols that don't exist in that build.

Jes, can you look at the 2.5 changes that went in after your last merge
and make the SN2 vchan workaround a config option?

For reference, this is the compile errors on ppc:

In file included from drivers/scsi/qla1280.c:299:
include/linux/blk.h:1:2: warning: #warning this file is obsolete, please use <linux/blkdev.h> instead
drivers/scsi/qla1280.c: In function `qla1280_done':
drivers/scsi/qla1280.c:1833: warning: implicit declaration of function `qla1280_get_target_options'
drivers/scsi/qla1280.c: At top level:
drivers/scsi/qla1280.c:4506: warning: type mismatch with previous implicit declaration
drivers/scsi/qla1280.c:1833: warning: previous implicit declaration of `qla1280_get_target_options'
drivers/scsi/qla1280.c:4506: warning: `qla1280_get_target_options' was previously implicitly declared to return `int'
drivers/scsi/qla1280.c:4830: warning: initialization from incompatible pointer type
drivers/scsi/qla1280.c:4830: error: unknown field `command' specified in initializer

Re: qla1280 compile problems

[Jes Sorensen]

>>>>> "Christoph" == Christoph Hellwig <hch@lst.de> writes:

Christoph> The qla1280 update still has some issues, mostly because it
Christoph> backs out all the 2.5 changes applied to the driver in the
Christoph> last months, also it doesn't work on ia64 because the IA64
Christoph> generic build references sn2-specific symbols that don't
Christoph> exist in that build.

Christoph,

Which version of the driver are you testing and what kernel? I tried
it out fairly recently on 2.5.x (2.5.59 or so) and it compiled without
any problems.

Christoph> Jes, can you look at the 2.5 changes that went in after
Christoph> your last merge and make the SN2 vchan workaround a config
Christoph> option?

I will take a look at building it against the latest 2.6.0-test
kernel. However, I am not going to make the vchan workaround a new
config option, it doesn't make any sense as it's already protected by
CONFIG_IA64_SGI_SN2 || CONFIG_IA64_GENERIC. The appropriate patches
for that have already been pushed for 2.4 and Jesse is about to push
them out for 2.5.

Cheers,
Jes

Re: qla1280 compile problems

[Christoph Hellwig]

On Tue, Aug 05, 2003 at 08:40:03AM -0400, Jes Sorensen wrote:
> >>>>> "Christoph" == Christoph Hellwig <hch@lst.de> writes:
> 
> Christoph> The qla1280 update still has some issues, mostly because it
> Christoph> backs out all the 2.5 changes applied to the driver in the
> Christoph> last months, also it doesn't work on ia64 because the IA64
> Christoph> generic build references sn2-specific symbols that don't
> Christoph> exist in that build.
> 
> Christoph,
> 
> Which version of the driver are you testing and what kernel? I tried
> it out fairly recently on 2.5.x (2.5.59 or so) and it compiled without
> any problems.

That's Jame's latest scsi-misc-2.5 tree, i.e. Linus' tree from four days
ago plus lots of scsi-related chanhges including qla 3.23.34.

> I will take a look at building it against the latest 2.6.0-test
> kernel. However, I am not going to make the vchan workaround a new
> config option, it doesn't make any sense as it's already protected by
> CONFIG_IA64_SGI_SN2 || CONFIG_IA64_GENERIC. The appropriate patches
> for that have already been pushed for 2.4 and Jesse is about to push
> them out for 2.5.

It does make sense.  Your punishing all ia64 users for SN2 breakage
although 95% (or probably more) of the IA64 boxens won't ever need it.

And no, so far SN2 isn't part of the generic IA64 build in either 2.4
or 2.5 although it's hopefully going to happen soon for 2.5.

Re: qla1280 compile problems

[Jes Sorensen]

>>>>> "Christoph" == Christoph Hellwig <hch@lst.de> writes:

Christoph> On Tue, Aug 05, 2003 at 08:40:03AM -0400, Jes Sorensen
Christoph> wrote:
>> Which version of the driver are you testing and what kernel? I
>> tried it out fairly recently on 2.5.x (2.5.59 or so) and it
>> compiled without any problems.

Christoph> That's Jame's latest scsi-misc-2.5 tree, i.e. Linus' tree
Christoph> from four days ago plus lots of scsi-related chanhges
Christoph> including qla 3.23.34.

Ok, will take a look at this.

Christoph> It does make sense.  Your punishing all ia64 users for SN2
Christoph> breakage although 95% (or probably more) of the IA64 boxens
Christoph> won't ever need it.

Punishing? thats why you have the option to compile the kernel as
CONFIG_IA64_HP_ZX1 or CONFIG_IA64_DIG. The point of the generic kernel
is that it will boot on all boxes, thats why this feature is only
enabled for generic kernels or sn2 specific. Second, the 'cost' of
this is a test and a branch and even on sn2 it just flips a
bit. Hardly comparable to the cost of loading the address into the
device.

Christoph> And no, so far SN2 isn't part of the generic IA64 build in
Christoph> either 2.4 or 2.5 although it's hopefully going to happen
Christoph> soon for 2.5.

I'll check what the status is, if it prevents current kernels from
building I'll of course take it out for now.

Jes

Re: qla1280 compile problems

[James Bottomley]

On Tue, 2003-08-05 at 05:18, Christoph Hellwig wrote:
> The qla1280 update still has some issues, mostly because it backs
> out all the 2.5 changes applied to the driver in the last months,
> also it doesn't work on ia64 because the IA64 generic build references
> sn2-specific symbols that don't exist in that build.

Could you be more explicit on this?  I thought I'd re-added all of the
backed out patches (although I might have missed some).

> In file included from drivers/scsi/qla1280.c:299:
> include/linux/blk.h:1:2: warning: #warning this file is obsolete, please use <linux/blkdev.h> instead
> drivers/scsi/qla1280.c: In function `qla1280_done':
> drivers/scsi/qla1280.c:1833: warning: implicit declaration of function `qla1280_get_target_options'
> drivers/scsi/qla1280.c: At top level:
> drivers/scsi/qla1280.c:4506: warning: type mismatch with previous implicit declaration
> drivers/scsi/qla1280.c:1833: warning: previous implicit declaration of `qla1280_get_target_options'
> drivers/scsi/qla1280.c:4506: warning: `qla1280_get_target_options' was previously implicitly declared to return `int'
> drivers/scsi/qla1280.c:4830: warning: initialization from incompatible pointer type
> drivers/scsi/qla1280.c:4830: error: unknown field `command' specified in initializer

For what it's worth, this compiled for me without warnings on parisc64
(although I haven't got around to actually testing it).

James

Re: qla1280 compile problems

[Christoph Hellwig]

On Tue, Aug 05, 2003 at 09:17:38AM -0700, James Bottomley wrote:
> On Tue, 2003-08-05 at 05:18, Christoph Hellwig wrote:
> > The qla1280 update still has some issues, mostly because it backs
> > out all the 2.5 changes applied to the driver in the last months,
> > also it doesn't work on ia64 because the IA64 generic build references
> > sn2-specific symbols that don't exist in that build.
> 
> Could you be more explicit on this?  I thought I'd re-added all of the
> backed out patches (although I might have missed some).


These seem to be missing at least

 - remove all #include <linux/blk.h>'s (Adrian Bunk)
 - give ->proc_info a struct Scsi_Host * paramter (me)
 - move all host templates into .c files (me)
 - Spellig fixes for consistent, dependent and persistent
 - Spellig fixes
 - [patch, 2.5] scsi_qla1280.c free on error path
 

> For what it's worth, this compiled for me without warnings on parisc64
> (although I haven't got around to actually testing it).

strange, I can's see how it could compile at least with the superflous
initializer for sht.command.