RE: I have modified the megaraid driver v2.00.5 from kernel 2.6 t o 2 .4.22-prex-ac1

RE: I have modified the megaraid driver v2.00.5 from kernel 2.6 t o 2 .4.22-prex-ac1

[Tomita, Haruo]

Hello Matthew,

     thanks for the reply!

Haruo> I have modified the megaraid driver  v2.00.5 from kernel 2.6 
Haruo> to 2.4.22-pre3-ac1 or 2.4.22-pre6-ac1.
 
Matthew> You mean you backported the 2.6 driver to 2.4?  If so, it seems the
Matthew> driver is in better state in 2.4 than 2.6 as many of these changes
Matthew> are outright wrong.

Yes, I  backported the 2.6 driver to 2.4.
Thank you for pointing out the affair of compile string literals.
I will correct a mistake. 

Matthew> PCI subsystem vendor IDs should use the defined constants in
Matthew> include/linux/pci_ids.h

I think that your opinion is right. My patch is wrong. 
I will correct a code. It will become the code of the driver of 2.6.

Matthew> if we want linux to be PCI200X compliant (a dubious point of view
since
Matthew> it means we'd have to stop supporting hundreds of hardware devices
designed
Matthew> before this became mandatory), this should be done in the PCI
subsystem,
Matthew> not in individual drivers.

Please explain in detail to a slight degree.
Although it corresponds to PCI2001 by 2.6, 
is it having stopped corresponding with my patch?

Matthew> some of the other code around here needs cleaning up too; for
example
Matthew> you should be using pci_name() rather than printing the slot/func
yourself.

Cleaning up of a code is not considered yet. 
I think it priority to make it operate by 2.4.

Matthew> btw, any plans to convert this driver to be hotplug capable?
 
I do not understand.

> 
> > @@ -1799,12 +1809,15 @@
> >  		}
> >  		set_irq_state(adapter, byte);
> >  
> > +		nstatus = 0xFF;
> >  		while((nstatus = (volatile u8)adapter->mbox->numstatus)
> >  				== 0xFF)

Matthew> this initialisation is not necessary; the assignment in the while
loop
Matthew> occurs before the test. (other occurrences of the same problem
deleted).

Thank you for a careful check.

Thanks again.
Regards
--
Haruo

Re: I have modified the megaraid driver v2.00.5 from kernel 2.6 t o 2 .4.22-prex-ac1

[Matthew Wilcox]

On Thu, Jul 24, 2003 at 02:05:24PM +0900, Tomita, Haruo wrote:
>      thanks for the reply!

no problem.  just trying to make the code better.

> Yes, I  backported the 2.6 driver to 2.4.
> Thank you for pointing out the affair of compile string literals.
> I will correct a mistake. 

Great.

> Matthew> if we want linux to be PCI200X compliant (a dubious point of view
> since
> Matthew> it means we'd have to stop supporting hundreds of hardware devices
> designed
> Matthew> before this became mandatory), this should be done in the PCI
> subsystem,
> Matthew> not in individual drivers.
> 
> Please explain in detail to a slight degree.
> Although it corresponds to PCI2001 by 2.6, 
> is it having stopped corresponding with my patch?

No, not at all.  I'm criticising the original code here, not your patch.
In my opinion, we should simply delete these lines:

                subsysvid = pdev->subsystem_vendor;
                subsysid = pdev->subsystem_device;

                /*
                 * If we do not find the valid subsys vendor id, refuse to
                 * load the driver. This is part of PCI200X compliance
                 * We load the driver if subsysvid is 0.
                 */
                if( subsysvid && (subsysvid != AMI_SUBSYS_VID) &&
                                (subsysvid != DELL_SUBSYS_VID) &&
                                (subsysvid != HP_SUBSYS_VID) &&
                                (subsysvid != LSI_SUBSYS_VID) ) continue;


> Matthew> some of the other code around here needs cleaning up too; for
> example
> Matthew> you should be using pci_name() rather than printing the slot/func
> yourself.
> 
> Cleaning up of a code is not considered yet. 
> I think it priority to make it operate by 2.4.

Is the driver in 2.4 not functional for you?

> Matthew> btw, any plans to convert this driver to be hotplug capable?
>  
> I do not understand.

In 2.5, it is finally possible to write a scsi driver that can cope with
the pci card being hotplugged.  The changes are actually not too hard, but
they do have some consequences we have to accept.

For example, converting the sym8xx driver to this means the controllers will
be detected in a different order.  This is going to suck for some people,
but it's really unavoidable.  It doesn't look like megaraid would have
the same problem, though this comment scared me:

                /*
                 * re-order hosts so that one with bootable logical drive
                 * comes first
                 */

but there's no code associated with it, so it doesn't matter.

-- 
"It's not Hollywood.  War is real, war is primarily not about defeat or
victory, it is about death.  I've seen thousands and thousands of dead bodies.
Do you think I want to have an academic debate on this subject?" -- Robert Fisk

Re: I have modified the megaraid driver v2.00.5 from kernel 2.6 t o 2 .4.22-prex-ac1

[Alan Cox]

> No, not at all.  I'm criticising the original code here, not your patch.
> In my opinion, we should simply delete these lines:

That would be a bad idea. Some of the devices share the same non subsys id
with non megaraid cards (INTEL I960) and the result is that megaraid cards
fall out with some other stuff.

Re: I have modified the megaraid driver v2.00.5 from kernel 2.6 t o 2 .4.22-prex-ac1

[Matthew Wilcox]

On Thu, Jul 24, 2003 at 09:09:39AM -0400, Alan Cox wrote:
> > No, not at all.  I'm criticising the original code here, not your patch.
> > In my opinion, we should simply delete these lines:
> 
> That would be a bad idea. Some of the devices share the same non subsys id
> with non megaraid cards (INTEL I960) and the result is that megaraid cards
> fall out with some other stuff.

Then the comment needs to be fixed -- we shouldn't be rejecting cards
for non-PCI-200X compliance; we should be rejecting them because another
driver wants them.  Yuck.

-- 
"It's not Hollywood.  War is real, war is primarily not about defeat or
victory, it is about death.  I've seen thousands and thousands of dead bodies.
Do you think I want to have an academic debate on this subject?" -- Robert Fisk

RE: I have modified the megaraid driver v2.00.5 from kernel 2.6 t o 2 .4.22-prex-ac1

[Tomita, Haruo]

Hi Matthew,

Haruo> Please explain in detail to a slight degree.
Haruo> Although it corresponds to PCI2001 by 2.6, 
Haruo> is it having stopped corresponding with my patch?

Matthew> No, not at all.  I'm criticising the original code here, not 
Matthew> your patch.In my opinion, we should simply delete these lines:

OK, I will leave an original code. 

Haruo> I think it priority to make it operate by 2.4.
Matthew> Is the driver in 2.4 not functional for you?
 
No, I should also operate the driver of v2.00.x by the kernel 2.4.

Matthew> In 2.5, it is finally possible to write a scsi driver that can cope
with
Matthew> the pci card being hotplugged.  The changes are actually not too
hard, but
Matthew> they do have some consequences we have to accept.
Matthew> For example, converting the sym8xx driver to this means the
controllers will
Matthew> be detected in a different order.  This is going to suck for some
people,
Matthew> but it's really unavoidable.  It doesn't look like megaraid would
have
Matthew> the same problem, though this comment scared me:
Matthew> but there's no code associated with it, so it doesn't matter.

Thank you for intelligible description. I understood in general.

Thanks!!
Haruo