two sym53c8xx.o modules

two sym53c8xx.o modules

[Xose Vazquez Perez]

hi,

kernel 2.4 has two modules with *same name*:
/lib/modules/2.4.XX/kernel/drivers/scsi/sym53c8xx_2/sym53c8xx.o
/lib/modules/2.4.XX/kernel/drivers/scsi/sym53c8xx.o

"# modprobe sym53c8xx" tries to load sym53c8xx.o first and then sym53c8xx_2/sym53c8xx.o
bug or feature?

should be sym53c8xx_2/sym53c8xx.o renamed to sym53c8xx_2/sym53c8xx_2.o ?
                                                                  ^^
-thanks-

Re: two sym53c8xx.o modules

[Marcelo Tosatti]

On Thu, 9 Oct 2003, Xose Vazquez Perez wrote:

> hi,
> 
> kernel 2.4 has two modules with *same name*:
> /lib/modules/2.4.XX/kernel/drivers/scsi/sym53c8xx_2/sym53c8xx.o
> /lib/modules/2.4.XX/kernel/drivers/scsi/sym53c8xx.o
> 
> "# modprobe sym53c8xx" tries to load sym53c8xx.o first and then sym53c8xx_2/sym53c8xx.o
> bug or feature?
> 
> should be sym53c8xx_2/sym53c8xx.o renamed to sym53c8xx_2/sym53c8xx_2.o ?

Xose,

One should not have both drivers loaded at the same time, so I think this
is alright.

Re: two sym53c8xx.o modules

[Erik Mouw]

On Thu, Oct 09, 2003 at 08:41:49AM -0300, Marcelo Tosatti wrote:
> On Thu, 9 Oct 2003, Xose Vazquez Perez wrote:
> > kernel 2.4 has two modules with *same name*:
> > /lib/modules/2.4.XX/kernel/drivers/scsi/sym53c8xx_2/sym53c8xx.o
> > /lib/modules/2.4.XX/kernel/drivers/scsi/sym53c8xx.o
> > 
> > "# modprobe sym53c8xx" tries to load sym53c8xx.o first and then sym53c8xx_2/sym53c8xx.o
> > bug or feature?
> > 
> > should be sym53c8xx_2/sym53c8xx.o renamed to sym53c8xx_2/sym53c8xx_2.o ?
> 
> One should not have both drivers loaded at the same time, so I think this
> is alright.

No, it's not allright. Modprobe can't distinguish between
sym53c8xx_2/sym53c8xx.o and sym53c8xx.o, you have to figure out the
full path and insmod one of them manually. Xose is right in that
sym53c8xx_2/sym53c8xx.o should be renamed in sym53c8xx_2/sym53c8xx_2.o.
Compare with aic7xxx and aic7xxx_old.


Erik

-- 
+-- Erik Mouw -- www.harddisk-recovery.com -- +31 70 370 12 90 --
| Lab address: Delftechpark 26, 2628 XH, Delft, The Netherlands
| Data lost? Stay calm and contact Harddisk-recovery.com

Re: two sym53c8xx.o modules

[Matthew Wilcox]

On Thu, Oct 09, 2003 at 02:24:28PM +0200, Erik Mouw wrote:
> No, it's not allright. Modprobe can't distinguish between
> sym53c8xx_2/sym53c8xx.o and sym53c8xx.o, you have to figure out the
> full path and insmod one of them manually. Xose is right in that
> sym53c8xx_2/sym53c8xx.o should be renamed in sym53c8xx_2/sym53c8xx_2.o.
> Compare with aic7xxx and aic7xxx_old.

Is it really important to change this at this stage of 2.4?  For
reference, sym1 has already gone away in 2.6, so it'll be incovenient
for users no matter what you change.  Let's stick with the devil we know.

-- 
"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: two sym53c8xx.o modules

[Marcelo Tosatti]

On Thu, 9 Oct 2003, Erik Mouw wrote:

> On Thu, Oct 09, 2003 at 08:41:49AM -0300, Marcelo Tosatti wrote:
> > On Thu, 9 Oct 2003, Xose Vazquez Perez wrote:
> > > kernel 2.4 has two modules with *same name*:
> > > /lib/modules/2.4.XX/kernel/drivers/scsi/sym53c8xx_2/sym53c8xx.o
> > > /lib/modules/2.4.XX/kernel/drivers/scsi/sym53c8xx.o
> > > 
> > > "# modprobe sym53c8xx" tries to load sym53c8xx.o first and then sym53c8xx_2/sym53c8xx.o
> > > bug or feature?
> > > 
> > > should be sym53c8xx_2/sym53c8xx.o renamed to sym53c8xx_2/sym53c8xx_2.o ?
> > 
> > One should not have both drivers loaded at the same time, so I think this
> > is alright.
> 
> No, it's not allright. Modprobe can't distinguish between
> sym53c8xx_2/sym53c8xx.o and sym53c8xx.o, you have to figure out the
> full path and insmod one of them manually. Xose is right in that
> sym53c8xx_2/sym53c8xx.o should be renamed in sym53c8xx_2/sym53c8xx_2.o.
> Compare with aic7xxx and aic7xxx_old.

True. Mind sending me a patch? 

Re: two sym53c8xx.o modules

[Erik Mouw]

On Thu, Oct 09, 2003 at 01:38:57PM +0100, Matthew Wilcox wrote:
> Is it really important to change this at this stage of 2.4?  For
> reference, sym1 has already gone away in 2.6, so it'll be incovenient
> for users no matter what you change.  Let's stick with the devil we know.

Yes it is. The _2 driver supports newer hardware (or better: can do
faster speeds on newer hardware), while the old driver doesn't. The old
driver, however, is more reliable in case of hardware errors. It's very
nice to be able to change drivers on the fly without having to figure
out the full path to the correct module.

Hmm, just tested the _2 driver and now I remember why I hardly use it
with flakey drives:

  root@dagobert:~ # scsiinfo -eXR /dev/sda 0 0 0 0 0 0 0 0 0 0 0 0 0 1
  <4>sym0:4:0:extraneous data discarded.
  <4>sym0:4:0:extraneous data discarded.
  <4>sym0:4:0:extraneous data discarded.
  <4>sym0:4:0:extraneous data discarded.
  <4>sym0:4:0:extraneous data discarded.
  Unable to store Read-Write Error Recovery Page 01h
  18 00 00 00 18 00 00 00 70 00 05 00 00 00 00 18 
  00 00 00 00 26 00 00 80 02 00 01 0a 00 00 00 00 
  00 00 00 00 00 01 00 00 00 00 00 00 00 00 00 00 

I haven't figured out what it could be yet, scsiinfo works perfectly
well with the old sym58xxx driver (and the aic7xxx driver, FWIW).

Anyway, here's a patch to rename sym53c8xx_2/sym53c8xx.o to
sym53c8xx_2/sym53c8xx_2.o. It's against 2.4.22, but should apply to all
2.4 kernels.


Erik

--- linux/drivers/scsi/Makefile.old	Tue Aug 26 12:16:09 2003
+++ linux/drivers/scsi/Makefile	Thu Oct  9 14:48:40 2003
@@ -97,7 +97,7 @@
 obj-$(CONFIG_SCSI_NCR53C7xx)	+= 53c7,8xx.o 
 subdir-$(CONFIG_SCSI_SYM53C8XX_2)	+= sym53c8xx_2
 ifeq ($(CONFIG_SCSI_SYM53C8XX_2),y)
-  obj-$(CONFIG_SCSI_SYM53C8XX_2)	+= sym53c8xx_2/sym53c8xx.o
+  obj-$(CONFIG_SCSI_SYM53C8XX_2)	+= sym53c8xx_2/sym53c8xx_2.o
 endif
 obj-$(CONFIG_SCSI_SYM53C8XX)	+= sym53c8xx.o 
 obj-$(CONFIG_SCSI_NCR53C8XX)	+= ncr53c8xx.o 
--- linux/drivers/scsi/sym53c8xx_2/Makefile.old	Sat Nov 10 00:22:54 2001
+++ linux/drivers/scsi/sym53c8xx_2/Makefile	Thu Oct  9 14:49:43 2003
@@ -1,14 +1,14 @@
-# File: drivers/sym53c8xx/Makefile
+# File: drivers/scsi/sym53c8xx_2/Makefile
 # Makefile for the NCR/SYMBIOS/LSI 53C8XX PCI SCSI controllers driver.
 
-list-multi := sym53c8xx.o
-sym53c8xx-objs := sym_fw.o sym_glue.o sym_hipd.o sym_malloc.o sym_misc.o sym_nvram.o
-obj-$(CONFIG_SCSI_SYM53C8XX_2) := sym53c8xx.o
+list-multi := sym53c8xx_2.o
+sym53c8xx_2-objs := sym_fw.o sym_glue.o sym_hipd.o sym_malloc.o sym_misc.o sym_nvram.o
+obj-$(CONFIG_SCSI_SYM53C8XX_2) := sym53c8xx_2.o
 
 EXTRA_CFLAGS += -I.
 
-sym53c8xx.o: $(sym53c8xx-objs)
-	$(LD) -r -o $@ $(sym53c8xx-objs)
+sym53c8xx_2.o: $(sym53c8xx_2-objs)
+	$(LD) -r -o $@ $(sym53c8xx_2-objs)
 
 include $(TOPDIR)/Rules.make
 


-- 
+-- Erik Mouw -- www.harddisk-recovery.nl -- 0800 220 20 20 --
| Eigen lab: Delftechpark 26, 2628 XH, Delft, Nederland
| Files foetsie, bestanden kwijt, alle data weg?!
| Blijf kalm en neem contact op met Harddisk-recovery.nl!

Re: two sym53c8xx.o modules

[Matthew Wilcox]

On Thu, Oct 09, 2003 at 03:11:52PM +0200, Erik Mouw wrote:
> On Thu, Oct 09, 2003 at 01:38:57PM +0100, Matthew Wilcox wrote:
> > Is it really important to change this at this stage of 2.4?  For
> > reference, sym1 has already gone away in 2.6, so it'll be incovenient
> > for users no matter what you change.  Let's stick with the devil we know.
> 
> Yes it is. The _2 driver supports newer hardware (or better: can do
> faster speeds on newer hardware), while the old driver doesn't. The old
> driver, however, is more reliable in case of hardware errors. It's very
> nice to be able to change drivers on the fly without having to figure
> out the full path to the correct module.

No, that's not the point.  For people who are already using either driver,
there's no need to change anything between 2.4 and 2.6.  It's only the
foolish people who compile both as modules that have a problem.

If you change the sym2 driver to have a different module name, now people
have to change when they switch between 2.4 and 2.6.  That's bad.

We've had this problem all through 2.4 and it's never been an issue
until now.  I really think this is one problem which shouldn't be fixed.

> Hmm, just tested the _2 driver and now I remember why I hardly use it
> with flakey drives:
> 
>   root@dagobert:~ # scsiinfo -eXR /dev/sda 0 0 0 0 0 0 0 0 0 0 0 0 0 1
>   <4>sym0:4:0:extraneous data discarded.
>   <4>sym0:4:0:extraneous data discarded.
>   <4>sym0:4:0:extraneous data discarded.
>   <4>sym0:4:0:extraneous data discarded.
>   <4>sym0:4:0:extraneous data discarded.
>   Unable to store Read-Write Error Recovery Page 01h
>   18 00 00 00 18 00 00 00 70 00 05 00 00 00 00 18 
>   00 00 00 00 26 00 00 80 02 00 01 0a 00 00 00 00 
>   00 00 00 00 00 01 00 00 00 00 00 00 00 00 00 00 
> 
> I haven't figured out what it could be yet, scsiinfo works perfectly
> well with the old sym58xxx driver (and the aic7xxx driver, FWIW).

It works perfectly fine with the drives I have access to under 2.6 ...

> Anyway, here's a patch to rename sym53c8xx_2/sym53c8xx.o to
> sym53c8xx_2/sym53c8xx_2.o. It's against 2.4.22, but should apply to all
> 2.4 kernels.

Please don't apply this patch.

-- 
"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: two sym53c8xx.o modules

[Erik Mouw]

On Thu, Oct 09, 2003 at 02:21:41PM +0100, Matthew Wilcox wrote:
> On Thu, Oct 09, 2003 at 03:11:52PM +0200, Erik Mouw wrote:
> > Yes it is. The _2 driver supports newer hardware (or better: can do
> > faster speeds on newer hardware), while the old driver doesn't. The old
> > driver, however, is more reliable in case of hardware errors. It's very
> > nice to be able to change drivers on the fly without having to figure
> > out the full path to the correct module.
> 
> No, that's not the point.  For people who are already using either driver,
> there's no need to change anything between 2.4 and 2.6.  It's only the
> foolish people who compile both as modules that have a problem.

It's not foolish to compile both modules. The 53c895 works faster with
the sym538cxxx_2 driver, the 53c810 can do with the sym53cxxx driver.
For otherwise the same machines, it's nice to compile a single kernel
for all of them and have the sym58cxxx drivers as modules, so you can
specify in modules.conf which driver to use. (also think about fully
modular distribution kernels).

> If you change the sym2 driver to have a different module name, now people
> have to change when they switch between 2.4 and 2.6.  That's bad.

Upgrading major kernel versions usually breaks things. That has never
been a problem in the past, so I don't see why it should be one in this
particular case. There were other modules with completely different
names, like usb-uhci --> uhci-hcd.

> We've had this problem all through 2.4 and it's never been an issue
> until now.  I really think this is one problem which shouldn't be fixed.

It hasn't been an issue until now simply because nobody has actually
seen it going wrong. If you load sym53cxxx, you get the old driver, and
hey, it works well with all sym53cxxx SCSI cards. The old driver simply
masks the new driver and I think the new driver didn't get as much
real-life testing as it should have had.

Selecting everything as a module never was a problem in Linux, you can
load any driver just by name. The sym53cxxx is an exception to this:
the old driver masks the new one. It's a namespace collision. That's
bad and it should be fixed.

> > I haven't figured out what it could be yet, scsiinfo works perfectly
> > well with the old sym58xxx driver (and the aic7xxx driver, FWIW).
> 
> It works perfectly fine with the drives I have access to under 2.6 ...

I know (I think you even told me). The fact that I only recently
figured out about this bug proves my point: in linux-2.4 the sym53cxx_2
driver didn't get as much testing as it should have had. Apply my
patch, and the driver might even get real use.


Erik

-- 
+-- Erik Mouw -- www.harddisk-recovery.com -- +31 70 370 12 90 --
| Lab address: Delftechpark 26, 2628 XH, Delft, The Netherlands
| Data lost? Stay calm and contact Harddisk-recovery.com

Re: two sym53c8xx.o modules

[Xose Vazquez Perez]

Erik Mouw wrote:

> I know (I think you even told me). The fact that I only recently
> figured out about this bug proves my point: in linux-2.4 the sym53cxx_2
> driver didn't get as much testing as it should have had. Apply my
> patch, and the driver might even get real use.

What's best approach ? to replace sym53c8xx_2/sym53c8xx.o with
sym53c8xx_2/sym53c8xx_2.o or sym53c8xx.o with sym53c8xx_old.o

:-?

Re: two sym53c8xx.o modules

[Xose Vazquez Perez]

Marcelo Tosatti wrote:

> One should not have both drivers loaded at the same time, so I think this
> is alright.

problem is "modprobe sym53c8xx" command loads at *same time* both.
I sent you a patch long time ago http://marc.theaimsgroup.com/?l=linux-kernel&m=105262722022952&w=2
;-)

-thanks-