This patch is intended to fix the "pdc20265 secondary controller
misidentified as primary" problem that occurs on motherboards including the
20265 as an onboard raid controller. The patch is based on one suggested
by Nick Orlov in August, but adds an additional printk to make the
alternatives more palatable.
The fix in the patch is to switch the PDC20265 from ON_BOARD to OFF_BOARD
when CONFIG_PDC202XX_FORCE is not set. Since this was proposed before, I
will try to address the two main objections raised and explain why I think
the new printk solves them.
1: This fixes some MBs, but break others:
First, more MBs seem to use the pdc20265 as a secondary controller than a
primary
Second, of those that do use the pdc20265 as the primary controller
(others have asserted this is a non-empty set, I have no knowledge of one),
if the BIOS sets it up to use the standard I/O port 0x01f0 for ide0, then
the first loop of ide-pci.c:ide-match-hwif() will notice the ioport and set
it as primary regardless of the ON_BOARD or OFF_BOARD setting, so these MBs
should work fine.
So the only case where there is a problem is a MB with a pdc20265 as
primary that doesn't set it up at the normal 0x01f0/0x0170 ioports. The
question in my mind is how many MBs are we talking about here. My
understanding is this set must be extremely small, if it is not non-empty.
So I believe that such a small percentage of MBs is not worth making the
default very painful for the vast majority of properly configured MBs. But
for those that are misconfigured, we still need a solution, which motivates
the new printk:
The right way, IMHO, to select the appropriate controller when it is a
problem is with ideX=base. The problem with suggesting this is that there
is no easy way to identify which value to use for base---the 20265, for
instance, gets 8 entries in /proc/ioports, and the VIA primary on the
A7V266-E gets 6. Unless you pick the right one, ide_match_hwif won't put
the controller where you request it. The solution I propose is to add a
printk to the bootup messages that will identify the base associated with
each interface on the controller, as seen by ide-match-hwif. This way a
quick look at the bootup screen will give the information needed to
rearrange the controllers.
This solution requires a kernel boot parameter to be added for the
(hopefully very small set of) MBs that have a 20265 as the primary
controller, but not at the standard address. Most MBs will continue to
work without a problem.
2: Won't fly with the non-sysadmin crowd
Right now, linux up to 2.4.18 would DTRT on most MBs and have the 20265 as
secondary. 2.5.41 also has the 20265 as OFF_BOARD, so it also identifies
it as secondary. If we want to make life easiest for the
non-kernel-hacker, the 20265 goes back to being OFF_BOARD. If this isn't
fixed, non-sysadmins (and sysadmins who have to support the machines) will
be very unhappy as their hard disks move from hda to hde and back to hda.
Sysadmins will be able to figure out the appropriate ideX=base parameter
and it should (hopefully) continue to work. Or controllers will continue
to be recognized on ide2, but at least it will be the same. This change
should be made just to ensure version-to-version compatibility.
FTR, I suspect doing this right (perfectly) would require a major reworking
of the ide probe code to do something like pre-scan all IDE devices to
check their ioports and then assigning them to ideX devices according to
ioport, known on-board controller, and pci order. But I don't think that
kind of effort is appropriate in a 2.4 kernel. We have a good boot
parameter solution of ideX=base, but it's hard to find the right number to
plug in for base. I think the best way to solve this problem is to add
additional information to the bootup messages. Then people who have a real
problem with IDE device ordering can solve it easily with ideX=base, rather
than relying on ide=reverse, which really seems like a hack (what if I want
the middle controller?).
In the case of my A7N266-E MBs, I have verified that the boot flags "ro
root=/dev/hde1 ide0=0xb400 ide1=0xa800 ide2=0x1f0 ide3=0x170" allows me to
boot with the 20265 on ide0 and ide1, should I want to.
There are two patches below. They both have the change ON_BOARD to
OFF_BOARD fix, but with different printk solutions. The first adds a line
like the one below (marked with *). Unfortunately, this printk occurs
before the interfaces are named, and even before the code has checked that
the second interface is enabled. But it puts all the information on one
line.
PDC20265: chipset revision 2
PDC20265: not 100% native mode: will probe irqs later
PDC20265: (U)DMA Burst Bit ENABLED Primary PCI Mode Secondary PCI Mode.
*PDC20265: interface bases at 0xb400 & 0xa800
ide2: BM-DMA at 0xa000-0xa007, BIOS settings: hde:pio, hdf:pio
ide3: BM-DMA at 0xa008-0xa00f, BIOS settings: hdg:pio, hdh:pio
The second adds lines like this, which solve the probem of naming and
making sure they're enabled, but I think it's somewhat uglier:
PDC20265: chipset revision 2
PDC20265: not 100% native mode: will probe irqs later
PDC20265: (U)DMA Burst Bit ENABLED Primary PCI Mode Secondary PCI Mode.
* ide2: located at 0xb400
ide2: BM-DMA at 0xa000-0xa007, BIOS settings: hde:pio, hdf:pio
* ide3: located at 0xa800
ide3: BM-DMA at 0xa008-0xa00f, BIOS settings: hdg:pio, hdh:pio
I think the best fix might be something like changing those last two lines
to:
ide2@0xb400: BM-DMA at 0xa000-0xa007, BIOS settings: hde:pio, hdf:pio
ide3@0xa800: BM-DMA at 0xa008-0xa00f, BIOS settings: hdg:pio, hdh:pio
but that involves changing all of the calls to ide_setup_dma in about 10
source files in the ide directory. I am willing to come up with that
patch, but only if there is positive feedback on this that makes it seem
likely the work would be worthwhile.
Comments are appreciated. Getting this solved in a reasonable manner is
important for those of us trying to maintain several machines with these
MBs. If there is a reason to believe that there is a significant number of
MBs with a primary onboard 20265 that doesn't live at 0x01f0, is there a
better solution?
Bruce Lowekamp
first patch that produces just one more line:
diff -u linux-2.4.20-pre9/drivers/ide/ide-pci.c.orig
linux-2.4.20-pre9/drivers/ide/ide-pci.c
--- linux-2.4.20-pre9/drivers/ide/ide-pci.c.orig Mon Oct 7 21:41:54
2002+++ linux-2.4.20-pre9/drivers/ide/ide-pci.c Tue Oct 8 16:20:57 2002
@@ -405,7 +405,7 @@
#ifndef CONFIG_PDC202XX_FORCE
{DEVID_PDC20246,"PDC20246", PCI_PDC202XX, NULL, INIT_PDC202XX,
NULL, {{0x00,0x00,0x00}, {0x00,0x00,0x00}}, OFF_BOARD,
16 },
{DEVID_PDC20262,"PDC20262", PCI_PDC202XX, ATA66_PDC202XX,
INIT_PDC202XX, NULL, {{0x00,0x00,0x00}, {0x00,0x00,0x00}}, OFF_BOARD,
48 },
- {DEVID_PDC20265,"PDC20265", PCI_PDC202XX, ATA66_PDC202XX,
INIT_PDC202XX, NULL, {{0x00,0x00,0x00}, {0x00,0x00,0x00}}, ON_BOARD,
48 },
+ {DEVID_PDC20265,"PDC20265", PCI_PDC202XX, ATA66_PDC202XX,
INIT_PDC202XX, NULL, {{0x00,0x00,0x00}, {0x00,0x00,0x00}}, OFF_BOARD,
48 },
{DEVID_PDC20267,"PDC20267", PCI_PDC202XX, ATA66_PDC202XX,
INIT_PDC202XX, NULL, {{0x00,0x00,0x00}, {0x00,0x00,0x00}}, OFF_BOARD,
48 },
#else /* !CONFIG_PDC202XX_FORCE */
{DEVID_PDC20246,"PDC20246", PCI_PDC202XX, NULL, INIT_PDC202XX, NULL,
{{0x50,0x02,0x02}, {0x50,0x04,0x04}}, OFF_BOARD,
16 },
@@ -744,6 +744,10 @@
continue;
#endif
}
+ if(port==0)
+ printk("%s: first interface base=0x%04lx, second interface
base=0x%04lx\n",
+ d->name, dev->resource[2*0].start,
+ dev->resource[2*1].start);
}
if ((ctl && !base) || (base && !ctl)) {
printk("%s: inconsistent baseregs (BIOS) for port %d, skipping\n",
d->name, port);
@@ -751,8 +755,11 @@
}
if (!ctl)
ctl = port ? 0x374 : 0x3f4; /* use default value */
- if (!base)
+ if (!base){
base = port ? 0x170 : 0x1f0; /* use default value */
+ if(port==0)
+ printk("%s: default first interface base=0x01f0, second interface
base=0x170\n");
+ }
if ((hwif = ide_match_hwif(base, d->bootable, d->name)) == NULL)
continue; /* no room in ide_hwifs[] */
if (hwif->io_ports[IDE_DATA_OFFSET] != base) {
second option that produces multiple lines, but more accurate:
diff -u linux-2.4.20-pre9/drivers/ide/ide-pci.c.orig
linux-2.4.20-pre9/drivers/ide/ide-pci.c
--- linux-2.4.20-pre9/drivers/ide/ide-pci.c.orig Mon Oct 7 21:41:54
2002+++ linux-2.4.20-pre9/drivers/ide/ide-pci.c Tue Oct 8 16:23:24 2002
@@ -405,7 +405,7 @@
#ifndef CONFIG_PDC202XX_FORCE
{DEVID_PDC20246,"PDC20246", PCI_PDC202XX, NULL, INIT_PDC202XX,
NULL, {{0x00,0x00,0x00}, {0x00,0x00,0x00}}, OFF_BOARD,
16 },
{DEVID_PDC20262,"PDC20262", PCI_PDC202XX, ATA66_PDC202XX,
INIT_PDC202XX, NULL, {{0x00,0x00,0x00}, {0x00,0x00,0x00}}, OFF_BOARD,
48 },
- {DEVID_PDC20265,"PDC20265", PCI_PDC202XX, ATA66_PDC202XX,
INIT_PDC202XX, NULL, {{0x00,0x00,0x00}, {0x00,0x00,0x00}}, ON_BOARD,
48 },
+ {DEVID_PDC20265,"PDC20265", PCI_PDC202XX, ATA66_PDC202XX,
INIT_PDC202XX, NULL, {{0x00,0x00,0x00}, {0x00,0x00,0x00}}, OFF_BOARD,
48 },
{DEVID_PDC20267,"PDC20267", PCI_PDC202XX, ATA66_PDC202XX,
INIT_PDC202XX, NULL, {{0x00,0x00,0x00}, {0x00,0x00,0x00}}, OFF_BOARD,
48 },
#else /* !CONFIG_PDC202XX_FORCE */
{DEVID_PDC20246,"PDC20246", PCI_PDC202XX, NULL, INIT_PDC202XX, NULL,
{{0x50,0x02,0x02}, {0x50,0x04,0x04}}, OFF_BOARD,
16 },
@@ -755,6 +755,7 @@
base = port ? 0x170 : 0x1f0; /* use default value */
if ((hwif = ide_match_hwif(base, d->bootable, d->name)) == NULL)
continue; /* no room in ide_hwifs[] */
+ printk(" %s: located at 0x%04lx\n", hwif->name, base);
if (hwif->io_ports[IDE_DATA_OFFSET] != base) {
ide_init_hwif_ports(&hwif->hw, base, (ctl | 2), NULL);
memcpy(hwif->io_ports, hwif->hw.io_ports, sizeof(hwif->io_ports));
I made a mistake when producing the two different patches. Here is a
replacement for the first patch that has all of the parameters it should.
this version does:
PDC20265: chipset revision 2
PDC20265: not 100% native mode: will probe irqs later
PDC20265: (U)DMA Burst Bit ENABLED Primary PCI Mode Secondary PCI Mode.
*PDC20265: interface bases at 0xb400 & 0xa800
ide2: BM-DMA at 0xa000-0xa007, BIOS settings: hde:pio, hdf:pio
ide3: BM-DMA at 0xa008-0xa00f, BIOS settings: hdg:pio, hdh:pio
diff -u linux-2.4.20-pre9/drivers/ide/ide-pci.c.orig
linux-2.4.20-pre9/drivers/ide/ide-pci.c
--- linux-2.4.20-pre9/drivers/ide/ide-pci.c.orig Mon Oct 7 21:41:54
2002+++ linux-2.4.20-pre9/drivers/ide/ide-pci.c Tue Oct 8 16:44:59 2002
@@ -405,7 +405,7 @@
#ifndef CONFIG_PDC202XX_FORCE
{DEVID_PDC20246,"PDC20246", PCI_PDC202XX, NULL, INIT_PDC202XX,
NULL, {{0x00,0x00,0x00}, {0x00,0x00,0x00}}, OFF_BOARD,
16 },
{DEVID_PDC20262,"PDC20262", PCI_PDC202XX, ATA66_PDC202XX,
INIT_PDC202XX, NULL, {{0x00,0x00,0x00}, {0x00,0x00,0x00}}, OFF_BOARD,
48 },
- {DEVID_PDC20265,"PDC20265", PCI_PDC202XX, ATA66_PDC202XX,
INIT_PDC202XX, NULL, {{0x00,0x00,0x00}, {0x00,0x00,0x00}}, ON_BOARD,
48 },
+ {DEVID_PDC20265,"PDC20265", PCI_PDC202XX, ATA66_PDC202XX,
INIT_PDC202XX, NULL, {{0x00,0x00,0x00}, {0x00,0x00,0x00}}, OFF_BOARD,
48 },
{DEVID_PDC20267,"PDC20267", PCI_PDC202XX, ATA66_PDC202XX,
INIT_PDC202XX, NULL, {{0x00,0x00,0x00}, {0x00,0x00,0x00}}, OFF_BOARD,
48 },
#else /* !CONFIG_PDC202XX_FORCE */
{DEVID_PDC20246,"PDC20246", PCI_PDC202XX, NULL, INIT_PDC202XX, NULL,
{{0x50,0x02,0x02}, {0x50,0x04,0x04}}, OFF_BOARD,
16 },
@@ -744,6 +744,10 @@
continue;
#endif
}
+ if(port==0)
+ printk("%s: first interface base=0x%04lx, second interface
base=0x%04lx\n",
+ d->name, dev->resource[2*0].start,
+ dev->resource[2*1].start);
}
if ((ctl && !base) || (base && !ctl)) {
printk("%s: inconsistent baseregs (BIOS) for port %d, skipping\n",
d->name, port);
@@ -751,8 +755,11 @@
}
if (!ctl)
ctl = port ? 0x374 : 0x3f4; /* use default value */
- if (!base)
+ if (!base){
base = port ? 0x170 : 0x1f0; /* use default value */
+ if(port==0)
+ printk("%s: default first interface base=0x01f0, second interface
base=0x170\n", d->name);
+ }
if ((hwif = ide_match_hwif(base, d->bootable, d->name)) == NULL)
continue; /* no room in ide_hwifs[] */
if (hwif->io_ports[IDE_DATA_OFFSET] != base) {
> This patch is intended to fix the "pdc20265 secondary controller
> misidentified as primary" problem that occurs on motherboards including
> the 20265 as an onboard raid controller. The patch is based on one
> suggested by Nick Orlov in August, but adds an additional printk to make
> the alternatives more palatable.
I was going to post revised patches, and go to a different mail client that
didn't munge them so badly, but...
Since 2.4.20-pre10-ac1 merges the IDE from 2.5 into 2.4, it resolves this
problem.
Bruce