blob: 30b4c714fbe127cd088f6511fb19af74a41e388f [file] [log] [blame]
Linus Torvalds1da177e2005-04-16 15:20:36 -07001Most (all) Intel-MP compliant SMP boards have the so-called 'IO-APIC',
Nick Andrew248fb892008-02-17 18:01:42 +11002which is an enhanced interrupt controller. It enables us to route
3hardware interrupts to multiple CPUs, or to CPU groups. Without an
4IO-APIC, interrupts from hardware will be delivered only to the
5CPU which boots the operating system (usually CPU#0).
Linus Torvalds1da177e2005-04-16 15:20:36 -07006
7Linux supports all variants of compliant SMP boards, including ones with
Nick Andrew248fb892008-02-17 18:01:42 +11008multiple IO-APICs. Multiple IO-APICs are used in high-end servers to
9distribute IRQ load further.
Linus Torvalds1da177e2005-04-16 15:20:36 -070010
Nick Andrew248fb892008-02-17 18:01:42 +110011There are (a few) known breakages in certain older boards, such bugs are
Linus Torvalds1da177e2005-04-16 15:20:36 -070012usually worked around by the kernel. If your MP-compliant SMP board does
13not boot Linux, then consult the linux-smp mailing list archives first.
14
15If your box boots fine with enabled IO-APIC IRQs, then your
16/proc/interrupts will look like this one:
17
18 ---------------------------->
19 hell:~> cat /proc/interrupts
20 CPU0
21 0: 1360293 IO-APIC-edge timer
22 1: 4 IO-APIC-edge keyboard
23 2: 0 XT-PIC cascade
24 13: 1 XT-PIC fpu
25 14: 1448 IO-APIC-edge ide0
26 16: 28232 IO-APIC-level Intel EtherExpress Pro 10/100 Ethernet
27 17: 51304 IO-APIC-level eth0
28 NMI: 0
29 ERR: 0
30 hell:~>
31 <----------------------------
32
Nick Andrew248fb892008-02-17 18:01:42 +110033Some interrupts are still listed as 'XT PIC', but this is not a problem;
Linus Torvalds1da177e2005-04-16 15:20:36 -070034none of those IRQ sources is performance-critical.
35
36
Nick Andrew248fb892008-02-17 18:01:42 +110037In the unlikely case that your board does not create a working mp-table,
Linus Torvalds1da177e2005-04-16 15:20:36 -070038you can use the pirq= boot parameter to 'hand-construct' IRQ entries. This
Nick Andrew248fb892008-02-17 18:01:42 +110039is non-trivial though and cannot be automated. One sample /etc/lilo.conf
Linus Torvalds1da177e2005-04-16 15:20:36 -070040entry:
41
42 append="pirq=15,11,10"
43
Nick Andrew248fb892008-02-17 18:01:42 +110044The actual numbers depend on your system, on your PCI cards and on their
Linus Torvalds1da177e2005-04-16 15:20:36 -070045PCI slot position. Usually PCI slots are 'daisy chained' before they are
46connected to the PCI chipset IRQ routing facility (the incoming PIRQ1-4
47lines):
48
49 ,-. ,-. ,-. ,-. ,-.
50 PIRQ4 ----| |-. ,-| |-. ,-| |-. ,-| |--------| |
51 |S| \ / |S| \ / |S| \ / |S| |S|
52 PIRQ3 ----|l|-. `/---|l|-. `/---|l|-. `/---|l|--------|l|
53 |o| \/ |o| \/ |o| \/ |o| |o|
54 PIRQ2 ----|t|-./`----|t|-./`----|t|-./`----|t|--------|t|
55 |1| /\ |2| /\ |3| /\ |4| |5|
56 PIRQ1 ----| |- `----| |- `----| |- `----| |--------| |
57 `-' `-' `-' `-' `-'
58
Nick Andrew248fb892008-02-17 18:01:42 +110059Every PCI card emits a PCI IRQ, which can be INTA, INTB, INTC or INTD:
Linus Torvalds1da177e2005-04-16 15:20:36 -070060
61 ,-.
62 INTD--| |
63 |S|
64 INTC--|l|
65 |o|
66 INTB--|t|
67 |x|
68 INTA--| |
69 `-'
70
71These INTA-D PCI IRQs are always 'local to the card', their real meaning
72depends on which slot they are in. If you look at the daisy chaining diagram,
Nick Andrewc0c20fb2008-03-04 15:05:40 -080073a card in slot4, issuing INTA IRQ, it will end up as a signal on PIRQ4 of
Linus Torvalds1da177e2005-04-16 15:20:36 -070074the PCI chipset. Most cards issue INTA, this creates optimal distribution
75between the PIRQ lines. (distributing IRQ sources properly is not a
76necessity, PCI IRQs can be shared at will, but it's a good for performance
77to have non shared interrupts). Slot5 should be used for videocards, they
78do not use interrupts normally, thus they are not daisy chained either.
79
80so if you have your SCSI card (IRQ11) in Slot1, Tulip card (IRQ9) in
81Slot2, then you'll have to specify this pirq= line:
82
83 append="pirq=11,9"
84
85the following script tries to figure out such a default pirq= line from
86your PCI configuration:
87
88 echo -n pirq=; echo `scanpci | grep T_L | cut -c56-` | sed 's/ /,/g'
89
90note that this script wont work if you have skipped a few slots or if your
91board does not do default daisy-chaining. (or the IO-APIC has the PIRQ pins
92connected in some strange way). E.g. if in the above case you have your SCSI
93card (IRQ11) in Slot3, and have Slot1 empty:
94
95 append="pirq=0,9,11"
96
97[value '0' is a generic 'placeholder', reserved for empty (or non-IRQ emitting)
98slots.]
99
Nick Andrew248fb892008-02-17 18:01:42 +1100100Generally, it's always possible to find out the correct pirq= settings, just
Linus Torvalds1da177e2005-04-16 15:20:36 -0700101permute all IRQ numbers properly ... it will take some time though. An
102'incorrect' pirq line will cause the booting process to hang, or a device
Nick Andrew248fb892008-02-17 18:01:42 +1100103won't function properly (e.g. if it's inserted as a module).
Linus Torvalds1da177e2005-04-16 15:20:36 -0700104
Nick Andrew248fb892008-02-17 18:01:42 +1100105If you have 2 PCI buses, then you can use up to 8 pirq values, although such
Linus Torvalds1da177e2005-04-16 15:20:36 -0700106boards tend to have a good configuration.
107
108Be prepared that it might happen that you need some strange pirq line:
109
110 append="pirq=0,0,0,0,0,0,9,11"
111
Nick Andrew248fb892008-02-17 18:01:42 +1100112Use smart trial-and-error techniques to find out the correct pirq line ...
Linus Torvalds1da177e2005-04-16 15:20:36 -0700113
Nick Andrew248fb892008-02-17 18:01:42 +1100114Good luck and mail to linux-smp@vger.kernel.org or
Linus Torvalds1da177e2005-04-16 15:20:36 -0700115linux-kernel@vger.kernel.org if you have any problems that are not covered
116by this document.
117
118-- mingo
119