The Intel 82378ZB System I/O (SIO) and 82379AB System I/O APIC (SIO.A)
ISA bridges implement PCI interrupt steering with a PIRQ router[1][2]
that is exactly the same as that of the PIIX and ICH southbridges (or
actually the other way round, given that the SIO ASIC was there first).
An earlier version of the SIO, the 82378IB[3][4], does not implement PCI
interrupt steering however, so we need to exclude it by checking the low
nibble of the PCI Revision Identification Register[5][6] for being at
least 3.
There is a note in the 82379AB specification update[7] saying that the
device ID for that chip is 0x7, rather than 0x484 as stated in the
datasheet[8]. It looks like a red herring however, for no report has
been ever seen with that value quoted and it matches the documented
default value of the PCI Command Register, which comes next after the
PCI Device Identification Register, so it looks like a copy-&-paste
editorial mistake.
NB the 82378ZB has been commonly used with smaller DEC Alpha systems
with the contents of the Revision Identification Register reported as
one of 0x3, 0x43, or 0x84, so the masking of the high nibble seems
indeed right by empirical observation. The value in the high nibble
might be either random, or depend on the batch, or correspond to some
other state such as reset straps.
References:
[1] "82378 System I/O (SIO)", Intel Corporation, Order Number:
290473-004, December 1994, Section 4.1.26 "PIRQ[3:0]#--PIRQ Route
Control Registers"
[2] "82378ZB System I/O (SIO) and 82379AB System I/O APIC (SIO.A)",
Intel Corporation, Order Number: 290571-001, March 1996, Section
3.1.25. "PIRQ[3:0]#--PIRQ Route Control Registers", p. 48
[3] "82378IB System I/O (SIO)", Intel Corporation, Order Number:
290473-002, April 1993, Section 5.8.7.7 "Edge and Level Triggered
Modes"
[4] "82378IB to 82378ZB Errata Fix and Feature Enhancement Conversion
FOL933002-01",
<https://web.archive.org/web/19990421045433/http://support.intel.com/support/chipsets/420/8511.htm>
[5] "82378 System I/O (SIO)", Intel Corporation, Order Number:
290473-004, December 1994, Section 4.1.5. "RID--Revision
Identification Register"
[6] "82378ZB System I/O (SIO) and 82379AB System I/O APIC (SIO.A)",
Intel Corporation, Order Number: 290571-001, March 1996, Section
3.1.5. "RID--Revision Identification Register", p. 34
[7] "Intel 82379AB (SIO.A) System I/O Component Specification Update",
Intel Corporation, Order Number: 297734-001, May, 1996, "Component
Identification via Programming Interface", p. 5
[8] "82378ZB System I/O (SIO) and 82379AB System I/O APIC (SIO.A)",
Intel Corporation, Order Number: 290571-001, March 1996, Section
3.1.2. "DID--Device Identification Register", p. 33
Signed-off-by: Maciej W. Rozycki <[email protected]>
---
Hi,
This patch was dropped from x86/irq due to a bug in a follow-up patch and
when resent it was not re-picked up along with the other patches for some
reason. It applies unchanged to 6.2.0-rc3.
Please apply.
Maciej
Changes from v2:
- Regenerate for a merge conflict.
Changes from v1:
- Add [PATCH] annotation (umm...).
- Fix RID values listed to include 0x84 rather than 0x83 (braino).
---
arch/x86/pci/irq.c | 11 +++++++++--
1 file changed, 9 insertions(+), 2 deletions(-)
linux-x86-pirq-router-sio.diff
Index: linux-macro/arch/x86/pci/irq.c
===================================================================
--- linux-macro.orig/arch/x86/pci/irq.c
+++ linux-macro/arch/x86/pci/irq.c
@@ -974,11 +974,18 @@ static __init int intel_router_probe(str
return 0;
switch (device) {
+ u8 rid;
case PCI_DEVICE_ID_INTEL_82375:
r->name = "PCEB/ESC";
r->get = pirq_esc_get;
r->set = pirq_esc_set;
return 1;
+ case PCI_DEVICE_ID_INTEL_82378:
+ pci_read_config_byte(router, PCI_REVISION_ID, &rid);
+ /* Tell 82378IB (rev < 3) and 82378ZB/82379AB apart. */
+ if ((rid & 0xfu) < 3)
+ break;
+ fallthrough;
case PCI_DEVICE_ID_INTEL_82371FB_0:
case PCI_DEVICE_ID_INTEL_82371SB_0:
case PCI_DEVICE_ID_INTEL_82371AB_0:
@@ -1020,7 +1027,7 @@ static __init int intel_router_probe(str
case PCI_DEVICE_ID_INTEL_ICH10_3:
case PCI_DEVICE_ID_INTEL_PATSBURG_LPC_0:
case PCI_DEVICE_ID_INTEL_PATSBURG_LPC_1:
- r->name = "PIIX/ICH";
+ r->name = "SIO/PIIX/ICH";
r->get = pirq_piix_get;
r->set = pirq_piix_set;
return 1;
@@ -1039,7 +1046,7 @@ static __init int intel_router_probe(str
device <= PCI_DEVICE_ID_INTEL_DH89XXCC_LPC_MAX)
|| (device >= PCI_DEVICE_ID_INTEL_PANTHERPOINT_LPC_MIN &&
device <= PCI_DEVICE_ID_INTEL_PANTHERPOINT_LPC_MAX)) {
- r->name = "PIIX/ICH";
+ r->name = "SIO/PIIX/ICH";
r->get = pirq_piix_get;
r->set = pirq_piix_set;
return 1;
On Sun, 8 Jan 2023, Maciej W. Rozycki wrote:
> The Intel 82378ZB System I/O (SIO) and 82379AB System I/O APIC (SIO.A)
> ISA bridges implement PCI interrupt steering with a PIRQ router[1][2]
> that is exactly the same as that of the PIIX and ICH southbridges (or
> actually the other way round, given that the SIO ASIC was there first).
Ping for:
<https://lore.kernel.org/lkml/[email protected]/>.
Maciej
On Sun, 8 Jan 2023, Maciej W. Rozycki wrote:
> The Intel 82378ZB System I/O (SIO) and 82379AB System I/O APIC (SIO.A)
> ISA bridges implement PCI interrupt steering with a PIRQ router[1][2]
> that is exactly the same as that of the PIIX and ICH southbridges (or
> actually the other way round, given that the SIO ASIC was there first).
Ping for:
<https://lore.kernel.org/lkml/[email protected]/>.
Maciej
On Sun, 8 Jan 2023, Maciej W. Rozycki wrote:
> The Intel 82378ZB System I/O (SIO) and 82379AB System I/O APIC (SIO.A)
> ISA bridges implement PCI interrupt steering with a PIRQ router[1][2]
> that is exactly the same as that of the PIIX and ICH southbridges (or
> actually the other way round, given that the SIO ASIC was there first).
Ping for:
<https://lore.kernel.org/lkml/[email protected]/>.
I think the patch is fairly obvious. Are there any outstanding concerns
that prevent it from being applied?
Maciej
On March 1, 2023 5:14:59 AM PST, "Maciej W. Rozycki" <[email protected]> wrote:
>On Sun, 8 Jan 2023, Maciej W. Rozycki wrote:
>
>> The Intel 82378ZB System I/O (SIO) and 82379AB System I/O APIC (SIO.A)
>> ISA bridges implement PCI interrupt steering with a PIRQ router[1][2]
>> that is exactly the same as that of the PIIX and ICH southbridges (or
>> actually the other way round, given that the SIO ASIC was there first).
>
> Ping for:
><https://lore.kernel.org/lkml/[email protected]/>.
>
> I think the patch is fairly obvious. Are there any outstanding concerns
>that prevent it from being applied?
>
> Maciej
>
Has this patch been actually tested on a real machine, or is it purely theoretical?
On Wed, 1 Mar 2023, H. Peter Anvin wrote:
> >> The Intel 82378ZB System I/O (SIO) and 82379AB System I/O APIC (SIO.A)
> >> ISA bridges implement PCI interrupt steering with a PIRQ router[1][2]
> >> that is exactly the same as that of the PIIX and ICH southbridges (or
> >> actually the other way round, given that the SIO ASIC was there first).
> >
> > Ping for:
> ><https://lore.kernel.org/lkml/[email protected]/>.
> >
> > I think the patch is fairly obvious. Are there any outstanding concerns
> >that prevent it from being applied?
>
> Has this patch been actually tested on a real machine, or is it purely
> theoretical?
I have no way to verify it on real x86 hw, my only SIO southbridge is in
a DEC Alpha machine, so not relevant.
This is I believe the final Intel device we're missing PIRQ support for,
and this work was prompted by a user having issues with his network card,
which ultimately and with a lot of confusion around I was able to narrow
down to missing PIRQ support rather than any actual issue with hardware.
At least Nikolai was patient enough to go through all this and I was
vigilant enough to actually catch the ongoing discussion on netdev in the
flood. See: <https://lore.kernel.org/netdev/[email protected]/>.
I'd rather we did not frustrate someone else with something as trivial
again, but if you think it's not enough for justification to merge this
change, then I'll accept it. I'd like such a decision to be explicitly
stated though rather than assumed by the loss of the patch in mailing list
noise (I know we're all overloaded with such stuff).
Thank your for your input.
Maciej