QUESTION: Network hangs with BP6 and 2.4.x kernels, hardware related?

How?
You send an EOI by writing 0 to the EOI register of the local apic, and then the local apic automagically checks it's ISR bitfield. It takes the highest set bit and clears it. Then it checks that bit in the TMR, and it if's also set in the TMR then it sends an EOI to the IO apic.

The magic seems to be tamper proof: all bits are read only.

The bit on the IO apic is also read only.
Perhaps with brute force? Switch the interrupt to edge triggered on the io apic, wait 1 usec, switch it back to level triggered. The IRR bit is undefined for edge triggered interrupts, perhaps that clears the IRR bit.

I would first concentrate on the differences between 2.2 and 2.4:

Frank, could you try what happens with the NMI oopser disabled?

The second major difference I'm immediately aware of is the number of the reschedule/tlb flush/etc interrupt: 2.2 uses the lowest priority, 2.4 the highest priority.

--
Manfred
-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in the body of a message to [email protected]
Please read the FAQ at http://www.tux.org/lkml/

Here's the results with nmi_watchdog=0

Before network hang (nmi_watchdog=0)
====================================

Jan 12 18:24:43 behemoth kernel: SysRq:
Jan 12 18:24:43 behemoth kernel: print_PIC()
Jan 12 18:24:43 behemoth kernel:
Jan 12 18:24:43 behemoth kernel: printing PIC contents
Jan 12 18:24:43 behemoth kernel: ... PIC  IMR: fffa
Jan 12 18:24:43 behemoth kernel: ... PIC  IRR: 0000
Jan 12 18:24:43 behemoth kernel: ... PIC  ISR: 0000
Jan 12 18:24:43 behemoth kernel: ... PIC ELCR: 1e00
Jan 12 18:24:43 behemoth kernel: print_IO_APIC()
Jan 12 18:24:43 behemoth kernel: number of MP IRQ sources: 23.
Jan 12 18:24:43 behemoth kernel: number of IO-APIC #2 registers: 24.
Jan 12 18:24:43 behemoth kernel: testing the IO APIC.......................
Jan 12 18:24:43 behemoth kernel:
Jan 12 18:24:43 behemoth kernel: IO APIC #2......
Jan 12 18:24:43 behemoth kernel: .... register #00: 02000000
Jan 12 18:24:43 behemoth kernel: .......    : physical APIC id: 02
Jan 12 18:24:43 behemoth kernel: .... register #01: 00170011
Jan 12 18:24:43 behemoth kernel: .......     : max redirection entries: 0017
Jan 12 18:24:43 behemoth kernel: .......     : IO APIC version: 0011
Jan 12 18:24:43 behemoth kernel: .... register #02: 00000000
Jan 12 18:24:43 behemoth kernel: .......     : arbitration: 00
Jan 12 18:24:43 behemoth kernel: .... IRQ redirection table:
Jan 12 18:24:43 behemoth kernel:  NR Log Phy Mask Trig IRR Pol Stat Dest Deli Vect:
Jan 12 18:24:43 behemoth kernel:  00 000 00  1    0    0   0   0    0    0    00
Jan 12 18:24:43 behemoth kernel:  01 003 03  0    0    0   0   0    1    1    39
Jan 12 18:24:43 behemoth kernel:  02 003 03  0    0    0   0   0    1    1    31
Jan 12 18:24:43 behemoth kernel:  03 003 03  0    0    0   0   0    1    1    41
Jan 12 18:24:43 behemoth kernel:  04 003 03  0    0    0   0   0    1    1    49
Jan 12 18:24:43 behemoth kernel:  05 003 03  0    0    0   0   0    1    1    51
Jan 12 18:24:43 behemoth kernel:  06 003 03  0    0    0   0   0    1    1    59
Jan 12 18:24:43 behemoth kernel:  07 003 03  0    0    0   0   0    1    1    61
Jan 12 18:24:43 behemoth kernel:  08 003 03  0    0    0   0   0    1    1    69
Jan 12 18:24:43 behemoth kernel:  09 000 00  1    0    0   0   0    0    0    00
Jan 12 18:24:43 behemoth kernel:  0a 000 00  1    0    0   0   0    0    0    00
Jan 12 18:24:43 behemoth kernel:  0b 000 00  1    0    0   0   0    0    0    00
Jan 12 18:24:43 behemoth kernel:  0c 000 00  1    0    0   0   0    0    0    00
Jan 12 18:24:43 behemoth kernel:  0d 000 00  1    0    0   0   0    0    0    00
Jan 12 18:24:43 behemoth kernel:  0e 003 03  0    0    0   0   0    1    1    71
Jan 12 18:24:43 behemoth kernel:  0f 003 03  0    0    0   0   0    1    1    79
Jan 12 18:24:43 behemoth kernel:  10 003 03  0    1    0   1   0    1    1    81
Jan 12 18:24:43 behemoth kernel:  11 003 03  0    1    0   1   0    1    1    89
Jan 12 18:24:43 behemoth kernel:  12 003 03  0    1    0   1   0    1    1    91
Jan 12 18:24:43 behemoth kernel:  13 003 03  0    1    0   1   0    1    1    99
Jan 12 18:24:43 behemoth kernel:  14 000 00  1    0    0   0   0    0    0    00
Jan 12 18:24:43 behemoth kernel:  15 000 00  1    0    0   0   0    0    0    00
Jan 12 18:24:43 behemoth kernel:  16 000 00  1    0    0   0   0    0    0    00
Jan 12 18:24:43 behemoth kernel:  17 000 00  1    0    0   0   0    0    0    00
Jan 12 18:24:43 behemoth kernel: IRQ to pin mappings:
Jan 12 18:24:43 behemoth kernel: IRQ0 -> 2
Jan 12 18:24:43 behemoth kernel: IRQ1 -> 1
Jan 12 18:24:43 behemoth kernel: IRQ3 -> 3
Jan 12 18:24:43 behemoth kernel: IRQ4 -> 4
Jan 12 18:24:43 behemoth kernel: IRQ5 -> 5
Jan 12 18:24:43 behemoth kernel: IRQ6 -> 6
Jan 12 18:24:43 behemoth kernel: IRQ7 -> 7
Jan 12 18:24:43 behemoth kernel: IRQ8 -> 8
Jan 12 18:24:43 behemoth kernel: IRQ13 -> 13
Jan 12 18:24:43 behemoth kernel: IRQ14 -> 14
Jan 12 18:24:43 behemoth kernel: IRQ15 -> 15
Jan 12 18:24:43 behemoth kernel: IRQ16 -> 16
Jan 12 18:24:43 behemoth kernel: IRQ17 -> 17
Jan 12 18:24:43 behemoth kernel: IRQ18 -> 18
Jan 12 18:24:43 behemoth kernel: IRQ19 -> 19
Jan 12 18:24:43 behemoth kernel: .................................... done.
Jan 12 18:24:43 behemoth kernel: print_all_local_APICs()
Jan 12 18:24:43 behemoth kernel:
Jan 12 18:24:43 behemoth kernel: printing local APIC contents on CPU#1/1:
Jan 12 18:24:43 behemoth kernel: ... APIC ID:      01000000 (1)
Jan 12 18:24:43 behemoth kernel: ... APIC VERSION: 00040011
Jan 12 18:24:43 behemoth kernel: ... APIC TASKPRI: 00000000 (00)
Jan 12 18:24:43 behemoth kernel: ... APIC ARBPRI: 00000000 (00)
Jan 12 18:24:43 behemoth kernel: ... APIC PROCPRI: 00000000
Jan 12 18:24:43 behemoth kernel: ... APIC EOI: 00000000
Jan 12 18:24:43 behemoth kernel: ... APIC LDR: 02000000
Jan 12 18:24:43 behemoth kernel: ... APIC DFR: ffffffff
Jan 12 18:24:43 behemoth kernel: ... APIC SPIV: 000003ff
Jan 12 18:24:43 behemoth kernel: ... APIC ISR field:
Jan 12 18:24:43 behemoth kernel: 0123456789abcdef0123456789abcdef
Jan 12 18:24:43 behemoth kernel: 00000000000000000000000000000000
Jan 12 18:24:43 behemoth last message repeated 7 times
Jan 12 18:24:43 behemoth kernel: ... APIC TMR field:
Jan 12 18:24:43 behemoth kernel: 0123456789abcdef0123456789abcdef
Jan 12 18:24:43 behemoth kernel: 00000000000000000000000000000000
Jan 12 18:24:43 behemoth last message repeated 3 times
Jan 12 18:24:43 behemoth kernel: 00000000010000000000000001000000
Jan 12 18:24:43 behemoth kernel: 00000000000000000000000000000000
Jan 12 18:24:43 behemoth last message repeated 2 times
Jan 12 18:24:43 behemoth kernel: ... APIC IRR field:
Jan 12 18:24:43 behemoth kernel: 0123456789abcdef0123456789abcdef
Jan 12 18:24:43 behemoth kernel: 00000000000000000000000000000000
Jan 12 18:24:43 behemoth last message repeated 7 times
Jan 12 18:24:43 behemoth kernel: ... APIC ESR: 00000000
Jan 12 18:24:43 behemoth kernel: ... APIC ICR: 000008fc
Jan 12 18:24:43 behemoth kernel: ... APIC ICR2: 01000000
Jan 12 18:24:43 behemoth kernel: ... APIC LVTT: 000200ef
Jan 12 18:24:43 behemoth kernel: ... APIC LVTPC: 00010000
Jan 12 18:24:43 behemoth kernel: ... APIC LVT0: 00010700
Jan 12 18:24:43 behemoth kernel: ... APIC LVT1: 00010400
Jan 12 18:24:43 behemoth kernel: ... APIC LVTERR: 000000fe
Jan 12 18:24:43 behemoth kernel: ... APIC TMICT: 0000a322
Jan 12 18:24:43 behemoth kernel: ... APIC TMCCT: 00000be6
Jan 12 18:24:43 behemoth kernel: ... APIC TDCR: 00000003
Jan 12 18:24:43 behemoth kernel:
Jan 12 18:24:43 behemoth kernel:
Jan 12 18:24:43 behemoth kernel: printing local APIC contents on CPU#0/0:
Jan 12 18:24:43 behemoth kernel: ... APIC ID:      00000000 (0)
Jan 12 18:24:43 behemoth kernel: ... APIC VERSION: 00040011
Jan 12 18:24:43 behemoth kernel: ... APIC TASKPRI: 00000000 (00)
Jan 12 18:24:43 behemoth kernel: ... APIC ARBPRI: 000000e0 (e0)
Jan 12 18:24:43 behemoth kernel: ... APIC PROCPRI: 00000000
Jan 12 18:24:43 behemoth kernel: ... APIC EOI: 00000000
Jan 12 18:24:43 behemoth kernel: ... APIC LDR: 01000000
Jan 12 18:24:43 behemoth kernel: ... APIC DFR: ffffffff
Jan 12 18:24:43 behemoth kernel: ... APIC SPIV: 000003ff
Jan 12 18:24:43 behemoth kernel: ... APIC ISR field:
Jan 12 18:24:43 behemoth kernel: 0123456789abcdef0123456789abcdef
Jan 12 18:24:43 behemoth kernel: 00000000000000000000000000000000
Jan 12 18:24:43 behemoth last message repeated 7 times
Jan 12 18:24:43 behemoth kernel: ... APIC TMR field:
Jan 12 18:24:43 behemoth kernel: 0123456789abcdef0123456789abcdef
Jan 12 18:24:43 behemoth kernel: 00000000000000000000000000000000
Jan 12 18:24:43 behemoth last message repeated 3 times
Jan 12 18:24:43 behemoth kernel: 00000000010000000000000001000000
Jan 12 18:24:43 behemoth kernel: 00000000000000000000000000000000
Jan 12 18:24:43 behemoth last message repeated 2 times
Jan 12 18:24:43 behemoth kernel: ... APIC IRR field:
Jan 12 18:24:43 behemoth kernel: 0123456789abcdef0123456789abcdef
Jan 12 18:24:43 behemoth kernel: 00000000000000000000000000000000
Jan 12 18:24:43 behemoth kernel: 00000000000000000100000000000000
Jan 12 18:24:43 behemoth kernel: 00000000000000000000000000000000
Jan 12 18:24:43 behemoth last message repeated 4 times
Jan 12 18:24:43 behemoth kernel: 00000000000000010000000000000000
Jan 12 18:24:43 behemoth kernel: ... APIC ESR: 00000000
Jan 12 18:24:43 behemoth kernel: ... APIC ICR: 000c08fb
Jan 12 18:24:43 behemoth kernel: ... APIC ICR2: 02000000
Jan 12 18:24:43 behemoth kernel: ... APIC LVTT: 000200ef
Jan 12 18:24:43 behemoth kernel: ... APIC LVTPC: 00010000
Jan 12 18:24:43 behemoth kernel: ... APIC LVT0: 00010700
Jan 12 18:24:43 behemoth kernel: ... APIC LVT1: 00000400
Jan 12 18:24:43 behemoth kernel: ... APIC LVTERR: 000000fe
Jan 12 18:24:43 behemoth kernel: ... APIC TMICT: 0000a322
Jan 12 18:24:43 behemoth kernel: ... APIC TMCCT: 000041e1
Jan 12 18:24:43 behemoth kernel: ... APIC TDCR: 00000003

After network hang (nmi_watchdog=0)
===================================

Jan 12 18:26:21 behemoth kernel: SysRq:
Jan 12 18:26:21 behemoth kernel: print_PIC()
Jan 12 18:26:21 behemoth kernel:
Jan 12 18:26:21 behemoth kernel: printing PIC contents
Jan 12 18:26:21 behemoth kernel: ... PIC  IMR: fffa
Jan 12 18:26:21 behemoth kernel: ... PIC  IRR: 0000
Jan 12 18:26:21 behemoth kernel: ... PIC  ISR: 0000
Jan 12 18:26:21 behemoth kernel: ... PIC ELCR: 1e00
Jan 12 18:26:21 behemoth kernel: print_IO_APIC()
Jan 12 18:26:21 behemoth kernel: number of MP IRQ sources: 23.
Jan 12 18:26:21 behemoth kernel: number of IO-APIC #2 registers: 24.
Jan 12 18:26:21 behemoth kernel: testing the IO APIC.......................
Jan 12 18:26:21 behemoth kernel:
Jan 12 18:26:21 behemoth kernel: IO APIC #2......
Jan 12 18:26:21 behemoth kernel: .... register #00: 02000000
Jan 12 18:26:21 behemoth kernel: .......    : physical APIC id: 02
Jan 12 18:26:21 behemoth kernel: .... register #01: 00170011
Jan 12 18:26:21 behemoth kernel: .......     : max redirection entries: 0017
Jan 12 18:26:21 behemoth kernel: .......     : IO APIC version: 0011
Jan 12 18:26:21 behemoth kernel: .... register #02: 00000000
Jan 12 18:26:21 behemoth kernel: .......     : arbitration: 00
Jan 12 18:26:21 behemoth kernel: .... IRQ redirection table:
Jan 12 18:26:21 behemoth kernel:  NR Log Phy Mask Trig IRR Pol Stat Dest Deli Vect:
Jan 12 18:26:21 behemoth kernel:  00 000 00  1    0    0   0   0    0    0    00
Jan 12 18:26:21 behemoth kernel:  01 003 03  0    0    0   0   0    1    1    39
Jan 12 18:26:21 behemoth kernel:  02 003 03  0    0    0   0   0    1    1    31
Jan 12 18:26:21 behemoth kernel:  03 003 03  0    0    0   0   0    1    1    41
Jan 12 18:26:21 behemoth kernel:  04 003 03  0    0    0   0   0    1    1    49
Jan 12 18:26:21 behemoth kernel:  05 003 03  0    0    0   0   0    1    1    51
Jan 12 18:26:21 behemoth kernel:  06 003 03  0    0    0   0   0    1    1    59
Jan 12 18:26:21 behemoth kernel:  07 003 03  0    0    0   0   0    1    1    61
Jan 12 18:26:21 behemoth kernel:  08 003 03  0    0    0   0   0    1    1    69
Jan 12 18:26:21 behemoth kernel:  09 000 00  1    0    0   0   0    0    0    00
Jan 12 18:26:21 behemoth kernel:  0a 000 00  1    0    0   0   0    0    0    00
Jan 12 18:26:21 behemoth kernel:  0b 000 00  1    0    0   0   0    0    0    00
Jan 12 18:26:21 behemoth kernel:  0c 000 00  1    0    0   0   0    0    0    00
Jan 12 18:26:21 behemoth kernel:  0d 000 00  1    0    0   0   0    0    0    00
Jan 12 18:26:21 behemoth kernel:  0e 003 03  0    0    0   0   0    1    1    71
Jan 12 18:26:21 behemoth kernel:  0f 003 03  0    0    0   0   0    1    1    79
Jan 12 18:26:21 behemoth kernel:  10 003 03  0    1    0   1   0    1    1    81
Jan 12 18:26:21 behemoth kernel:  11 003 03  0    1    0   1   0    1    1    89
Jan 12 18:26:21 behemoth kernel:  12 003 03  0    1    0   1   0    1    1    91
Jan 12 18:26:21 behemoth kernel:  13 003 03  0    1    1   1   0    1    1    99
Jan 12 18:26:21 behemoth kernel:  14 000 00  1    0    0   0   0    0    0    00
Jan 12 18:26:21 behemoth kernel:  15 000 00  1    0    0   0   0    0    0    00
Jan 12 18:26:21 behemoth kernel:  16 000 00  1    0    0   0   0    0    0    00
Jan 12 18:26:21 behemoth kernel:  17 000 00  1    0    0   0   0    0    0    00
Jan 12 18:26:21 behemoth kernel: IRQ to pin mappings:
Jan 12 18:26:21 behemoth kernel: IRQ0 -> 2
Jan 12 18:26:21 behemoth kernel: IRQ1 -> 1
Jan 12 18:26:21 behemoth kernel: IRQ3 -> 3
Jan 12 18:26:21 behemoth kernel: IRQ4 -> 4
Jan 12 18:26:21 behemoth kernel: IRQ5 -> 5
Jan 12 18:26:21 behemoth kernel: IRQ6 -> 6
Jan 12 18:26:21 behemoth kernel: IRQ7 -> 7
Jan 12 18:26:21 behemoth kernel: IRQ8 -> 8
Jan 12 18:26:21 behemoth kernel: IRQ13 -> 13
Jan 12 18:26:21 behemoth kernel: IRQ14 -> 14
Jan 12 18:26:21 behemoth kernel: IRQ15 -> 15
Jan 12 18:26:21 behemoth kernel: IRQ16 -> 16
Jan 12 18:26:21 behemoth kernel: IRQ17 -> 17
Jan 12 18:26:21 behemoth kernel: IRQ18 -> 18
Jan 12 18:26:21 behemoth kernel: IRQ19 -> 19
Jan 12 18:26:21 behemoth kernel: .................................... done.
Jan 12 18:26:21 behemoth kernel: print_all_local_APICs()
Jan 12 18:26:21 behemoth kernel:
Jan 12 18:26:21 behemoth kernel: printing local APIC contents on CPU#1/1:
Jan 12 18:26:21 behemoth kernel: ... APIC ID:      01000000 (1)
Jan 12 18:26:21 behemoth kernel: ... APIC VERSION: 00040011
Jan 12 18:26:21 behemoth kernel: ... APIC TASKPRI: 00000000 (00)
Jan 12 18:26:21 behemoth kernel: ... APIC ARBPRI: 00000000 (00)
Jan 12 18:26:21 behemoth kernel: ... APIC PROCPRI: 00000000
Jan 12 18:26:21 behemoth kernel: ... APIC EOI: 00000000
Jan 12 18:26:21 behemoth kernel: ... APIC LDR: 02000000
Jan 12 18:26:21 behemoth kernel: ... APIC DFR: ffffffff
Jan 12 18:26:21 behemoth kernel: ... APIC SPIV: 000003ff
Jan 12 18:26:21 behemoth kernel: ... APIC ISR field:
Jan 12 18:26:21 behemoth kernel: 0123456789abcdef0123456789abcdef
Jan 12 18:26:21 behemoth kernel: 00000000000000000000000000000000
Jan 12 18:26:21 behemoth last message repeated 7 times
Jan 12 18:26:21 behemoth kernel: ... APIC TMR field:
Jan 12 18:26:21 behemoth kernel: 0123456789abcdef0123456789abcdef
Jan 12 18:26:21 behemoth kernel: 00000000000000000000000000000000
Jan 12 18:26:21 behemoth last message repeated 3 times
Jan 12 18:26:21 behemoth kernel: 00000000010000000000000000000000
Jan 12 18:26:21 behemoth kernel: 00000000000000000000000000000000
Jan 12 18:26:21 behemoth last message repeated 2 times
Jan 12 18:26:21 behemoth kernel: ... APIC IRR field:
Jan 12 18:26:21 behemoth kernel: 0123456789abcdef0123456789abcdef
Jan 12 18:26:21 behemoth kernel: 00000000000000000000000000000000
Jan 12 18:26:21 behemoth last message repeated 7 times
Jan 12 18:26:21 behemoth kernel: ... APIC ESR: 00000000
Jan 12 18:26:21 behemoth kernel: ... APIC ICR: 000008fc
Jan 12 18:26:21 behemoth kernel: ... APIC ICR2: 01000000
Jan 12 18:26:21 behemoth kernel: ... APIC LVTT: 000200ef
Jan 12 18:26:21 behemoth kernel: ... APIC LVTPC: 00010000
Jan 12 18:26:21 behemoth kernel: ... APIC LVT0: 00010700
Jan 12 18:26:21 behemoth kernel: ... APIC LVT1: 00010400
Jan 12 18:26:21 behemoth kernel: ... APIC LVTERR: 000000fe
Jan 12 18:26:21 behemoth kernel: ... APIC TMICT: 0000a322
Jan 12 18:26:21 behemoth kernel: ... APIC TMCCT: 00001803
Jan 12 18:26:21 behemoth kernel: ... APIC TDCR: 00000003
Jan 12 18:26:21 behemoth kernel:
Jan 12 18:26:21 behemoth kernel:
Jan 12 18:26:21 behemoth kernel: printing local APIC contents on CPU#0/0:
Jan 12 18:26:21 behemoth kernel: ... APIC ID:      00000000 (0)
Jan 12 18:26:21 behemoth kernel: ... APIC VERSION: 00040011
Jan 12 18:26:21 behemoth kernel: ... APIC TASKPRI: 00000000 (00)
Jan 12 18:26:21 behemoth kernel: ... APIC ARBPRI: 000000e0 (e0)
Jan 12 18:26:21 behemoth kernel: ... APIC PROCPRI: 00000000
Jan 12 18:26:21 behemoth kernel: ... APIC EOI: 00000000
Jan 12 18:26:21 behemoth kernel: ... APIC LDR: 01000000
Jan 12 18:26:21 behemoth kernel: ... APIC DFR: ffffffff
Jan 12 18:26:21 behemoth kernel: ... APIC SPIV: 000003ff
Jan 12 18:26:21 behemoth kernel: ... APIC ISR field:
Jan 12 18:26:21 behemoth kernel: 0123456789abcdef0123456789abcdef
Jan 12 18:26:21 behemoth kernel: 00000000000000000000000000000000
Jan 12 18:26:21 behemoth last message repeated 7 times
Jan 12 18:26:21 behemoth kernel: ... APIC TMR field:
Jan 12 18:26:21 behemoth kernel: 0123456789abcdef0123456789abcdef
Jan 12 18:26:21 behemoth kernel: 00000000000000000000000000000000
Jan 12 18:26:21 behemoth last message repeated 3 times
Jan 12 18:26:21 behemoth kernel: 00000000010000000000000001000000
Jan 12 18:26:21 behemoth kernel: 00000000000000000000000000000000
Jan 12 18:26:21 behemoth last message repeated 2 times
Jan 12 18:26:21 behemoth kernel: ... APIC IRR field:
Jan 12 18:26:21 behemoth kernel: 0123456789abcdef0123456789abcdef
Jan 12 18:26:21 behemoth kernel: 00000000000000000000000000000000
Jan 12 18:26:21 behemoth last message repeated 6 times
Jan 12 18:26:21 behemoth kernel: 00000000000000010000000000000000
Jan 12 18:26:21 behemoth kernel: ... APIC ESR: 00000000
Jan 12 18:26:21 behemoth kernel: ... APIC ICR: 000c08fb
Jan 12 18:26:21 behemoth kernel: ... APIC ICR2: 02000000
Jan 12 18:26:21 behemoth kernel: ... APIC LVTT: 000200ef
Jan 12 18:26:21 behemoth kernel: ... APIC LVTPC: 00010000
Jan 12 18:26:21 behemoth kernel: ... APIC LVT0: 00010700
Jan 12 18:26:21 behemoth kernel: ... APIC LVT1: 00000400
Jan 12 18:26:21 behemoth kernel: ... APIC LVTERR: 000000fe
Jan 12 18:26:21 behemoth kernel: ... APIC TMICT: 0000a322
Jan 12 18:26:21 behemoth kernel: ... APIC TMCCT: 00004e26
Jan 12 18:26:21 behemoth kernel: ... APIC TDCR: 00000003

  WWWWW      _______________________
 ## o o\    /     Frank de Lange     \
 }#   \|   /                          \
  ##---# _/     <Hacker for Hire>      \
   ####   \      +31-320-252965        /
           \    [email protected]    /
            -------------------------
 [ "Omnis enim res, quae dando non deficit, dum habetur
    et non datur, nondum habetur, quomodo habenda est."  ]
-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to [email protected]
Please read the FAQ at http://www.tux.org/lkml/

Im trying to remember what they were, but some APIC versions do have errata and someting about 3 irqs at the same priority level rings a bell.

-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in the body of a message to [email protected]
Please read the FAQ at http://www.tux.org/lkml/

The PPro local apic documentation says:
<<<<<<<
The processor's local APIC includes an in-service entry and a holding entry for each priority level. To avoid losing interrupts, software should allocate no more than 2 interrupt vectors per priority.

Ok, we must reorder the vector numbers for our own interrupts (0xfb-0xff), but that doesn't explain our problems: we don't loose reschedule interrupts, we have problems with normal interrupts - and there we only use 2 irq at the same priority level.

Btw, the kick patch I sent a few minutes ago revives my io apic.

--
Manfred
-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in the body of a message to [email protected]
Please read the FAQ at http://www.tux.org/lkml/

we *already* reorder vector numbers and spread them out as much as possible. We do this in 2.2 as well. We did this almost from day 1 of IO-APIC support. If any manually allocated IRQ vector creates a '3 vectors in the same 16-vector region' situation then thats a bug in hw_irq.h..

the 'loss of interrupts' above does not include external interrupts, only local interrupts (such as the APIC timer interrupt) can get lost in such a situation.

(nevertheless there is something going on.)

Ingo

-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in the body of a message to [email protected]
Please read the FAQ at http://www.tux.org/lkml/

I reread the 2.2 and 2.4 code:

2.2 spreads all vectors, even the internal interrupts (tlb flush, reschedule, timer, call function) are spread.
Hmm. AFAICS there is one bug: 3 interrupts are in priority 5:

CALL_FUNCTION_VECTOR 0x50
IRQ0_TRAP_VECTOR 0x51
irq1: irq0+8 0x59

check arch/i386/kernel/irq.h and assign_irq_vector (io_apic.c)

2.4 spreads the vectors for the external (hardware, from io apic) interrupts, but 5 ipi vectors have the same priority: reschedule, call function, tlb invalidate, apic error, spurious interrupt.

But that doesn't explain what happens with ne2k cards: neither 2.2 nor 2.4 have more than 2 interrupts in class for the hardware interrupt 16/19.
With 2.2 I don't have any problems, 2.4 hangs.

--
Manfred
-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in the body of a message to [email protected]
Please read the FAQ at http://www.tux.org/lkml/

my reading of the errata is that the lost APIC timer IRQ happens only if the APIC timer IRQ vector's priority level has more than 2 active vectors. It's a very limited case, which does not happen in recent CPUs anyway (such as the PIII).

yep.

Ingo

-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in the body of a message to [email protected]
Please read the FAQ at http://www.tux.org/lkml/

In article <[email protected]>,

Does this seem to happen mainly with drivers that use "disable_irq()" and "enable_irq()"? I know the ne drivers do (through the 8390 module), and some others do too (3c59x).

"disable_irq()"/"enable_irq()" has always tended to be slightly problematic. It's not a set of semantics that maps well onto all interrupt controllers (io-apic definitely included). Drivers would generally be better off if they disabled their own chip from sending interrupts, rather than disabling the interrupt line the chip is on.

(Of course, most drivers would be even _better_ off if they didn't play games with irq disabling at all, but I think the 8390 driver does it because otherwise it would suck too badly for words).

If you are seeing this with a 8390 core, try to see if the problem goes away if you remove the "disable_irq_nosync(dev->irq);" and "enable_irq()" thing (which means that you need to change the spinlocking at the same place to use irq-safe versions - this _will_ make for bad interrupt latency especially with ISA ne2000 cards, but it would be interesting to hear if it makes the problem less likely to happen).

Linus
-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in the body of a message to [email protected]
Please read the FAQ at http://www.tux.org/lkml/

That doesn't work very well because the device irq can arrive a measurable number of clocks after you disable it on the source and there is no way to say 'and be sure the stupid thing has propogated the apic bus'

Alan

-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in the body of a message to [email protected]
Please read the FAQ at http://www.tux.org/lkml/

I agree. Asynchronous buses are nasty that way. However, if your "locking" is based on device state, you can still do it: you just have to have an internal device protocol. The simplest example of this is:

	interrupt_handler()
	{
		status = readl(dev->status);
		if (status & MY_IRQ_DISABLE)
			return;

	}


	{
		status = readl(dev->status);
		writel(dev->status, status | MY_IRQ_DISABLE);
		spin_lock();
		.. critical region ..
		spin_unlock();
		writel(dev->status, status);
	}

Notice how the above does NOT guarantee that the physical interrupt might not be "in flight". It does, however, synchronize other ways, simply by virtue of using the _synchronous_ bus (PCI, in this case) to figure out when the asynchronous bus (interrupts) has a bogus event.

Linus

-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in the body of a message to [email protected]
Please read the FAQ at http://www.tux.org/lkml/

Unfortunately on the 8390 the IRQ statud register is on page 0. The code on the other CPU might not be on page 0. That means we can't even safely check if there is an irq pending or clear it down (bad news on ne2k-pci) without getting that lock.

That means we have to be able to just block that one irq source to avoid horrible SMP latency problems.

Alan

-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in the body of a message to [email protected]
Please read the FAQ at http://www.tux.org/lkml/

Alan.

THINK.

The "synchronous channel" can be _memory_. I just used the above as an example of using a synchronous channel to make sure that the asynchronous buses aren't screwing us.

You only need one bit of synchronous information. The example used the very same bit that the irq flag on the device is anyway - which works on a lot of devices simply because they need to read the status flags anyway in order to handle the interrupt.

But if you have problems with that, just use an atomic flag off "dev->private" instead. Big deal.

Linus

-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in the body of a message to [email protected]
Please read the FAQ at http://www.tux.org/lkml/

As per Linus' suggestion, I removed the disable_irq/enable_irq statements from the 8390 core driver, and replace the spinlocks with irq-safe versions. This seems to solve the network hangs, as I am currently running a heavy network load (which would have killed a non-patched driver within seconds). Network latency seems a bit higher, and there are some hiccups in the streaming audio (part of the network load, easy indicator of performance...), but no hangs. Here's the patch:

@@ -242,15 +242,15 @@
 
 	/* Ugly but a reset can be slow, yet must be protected */
 		
-	disable_irq_nosync(dev->irq);
-	spin_lock(&ei_local->page_lock);
+	/* disable_irq_nosync(dev->irq); */
+	spin_lock_irq(&ei_local->page_lock);
 		
 	/* Try to restart the card.  Perhaps the user has fixed something. */
 	ei_reset_8390(dev);
 	NS8390_init(dev, 1);
 		
-	spin_unlock(&ei_local->page_lock);
-	enable_irq(dev->irq);
+	spin_unlock_irq(&ei_local->page_lock);
+	/* enable_irq(dev->irq); */
 	netif_wake_queue(dev);
 }
     

@@ -285,9 +285,9 @@
 	 *	Slow phase with lock held.
 	 */
 	 
-	disable_irq_nosync(dev->irq);
+	/* disable_irq_nosync(dev->irq); */
 	
-	spin_lock(&ei_local->page_lock);
+	spin_lock_irq(&ei_local->page_lock);
 	
 	ei_local->irqlock = 1;
 

@@ -383,8 +383,8 @@
 	ei_local->irqlock = 0;
 	outb_p(ENISR_ALL, e8390_base + EN0_IMR);
 	
-	spin_unlock(&ei_local->page_lock);
-	enable_irq(dev->irq);
+	spin_unlock_irq(&ei_local->page_lock);
+	/* enable_irq(dev->irq); */
 
 	dev_kfree_skb (skb);
 	ei_local->stat.tx_bytes += send_length;

  WWWWW      _______________________
 ## o o\    /     Frank de Lange     \
 }#   \|   /                          \
  ##---# _/     <Hacker for Hire>      \
   ####   \      +31-320-252965        /
           \    [email protected]    /
            -------------------------
 [ "Omnis enim res, quae dando non deficit, dum habetur
    et non datur, nondum habetur, quomodo habenda est."  ]
-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to [email protected]
Please read the FAQ at http://www.tux.org/lkml/

In article <[email protected]>,

Ok, so it's tentatively the IOAPIC disable/enable code. But it could obviously be something that just interacts with it, including just a timing issue (ie the _real_ bug might just be bad behaviour when changing IO-APIC state at the same time as an interrupt happens, and disable/enable-irq just happen to be the only things that do it at a high enough frequency that you can see the problem).

Remind me: what polarity are your io-apic irq's? Level, edge, sideways? Anything else that might be relevant?

Linus
-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in the body of a message to [email protected]
Please read the FAQ at http://www.tux.org/lkml/

Hi Alan,

Will you incorporate Andre IDE patches into the 2.2.19preXX kernel? If not, do you have any plan to do so in the very near future? I know this issue has been discussed before. But there has not been any progress.

I have been following the kernel development for years. I have seen many drivers added to the kernel. Some weren't working properly and had to be removed quickly. Strangely, the IDE patches drag on and on after all the kernel releases.

Are you still concern about the stability of the driver?

Regards,

Jeff

ASL Inc.

-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in the body of a message to [email protected]
Please read the FAQ at http://www.tux.org/lkml/

my BP6 with the patch frank sent me and the apic code at line 273 (or so) defined as '1' and a flood ping :

Jan 14 19:56:19 grobbebol kernel: APIC error on CPU1: 02(02)
Jan 14 19:56:25 grobbebol kernel: APIC error on CPU1: 02(02)
Jan 14 19:58:10 grobbebol last message repeated 2 times
Jan 14 20:00:01 grobbebol kernel: APIC error on CPU1: 02(02)
Jan 14 20:01:11 grobbebol last message repeated 2 times
Jan 14 20:01:48 grobbebol kernel: APIC error on CPU1: 02(02)
Jan 14 20:01:59 grobbebol kernel: APIC error on CPU1: 02(08)
Jan 14 20:02:10 grobbebol kernel: APIC error on CPU1: 08(08)
Jan 14 20:02:39 grobbebol kernel: APIC error on CPU1: 08(02)
Jan 14 20:02:39 grobbebol kernel: unexpected IRQ trap at vector 8d
Jan 14 20:15:32 grobbebol kernel: APIC error on CPU1: 02(08)

[....]

ad the network is dead. however, no crashes seen during this.

Grobbebol's Home                   |  Don't give in to spammers.   -o)
http://www.xs4all.nl/~bengel       | Use your real e-mail address   /\
Linux 2.2.16 SMP 2x466MHz / 256 MB |        on Usenet.             _\_v  
-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to [email protected]
Please read the FAQ at http://www.tux.org/lkml/

It still hangs.

Frank, I've attached a proposed kick_IOAPIC pin. Could you try it? I'm rebooting with that patch right now.
--
Manfred

1) add to the end of io_apic.c:

static void print_line(struct IO_APIC_route_entry* entry)

{
	printk(KERN_EMERG " %02x %03X %02X  ",
			0,
			entry->dest.logical.logical_dest,
			entry->dest.physical.physical_dest
		);

	printk("%1d    %1d    %1d   %1d   %1d    %1d    %1d    %02X\n",
			entry->mask,
			entry->trigger,
			entry->irr,
			entry->polarity,
			entry->delivery_status,
			entry->dest_mode,
			entry->delivery_mode,
			entry->vector
		);
}

void kick_IOAPIC_pin(int pin)

{
    	unsigned long flags;
	struct IO_APIC_route_entry entry;

	local_irq_save(flags);

	*(((int *)&entry) + 1) = io_apic_read(0, 0x11 + 2 * pin);
	*(((int *)&entry) + 0) = io_apic_read(0, 0x10 + 2 * pin);

	printk(KERN_EMERG " NR Log Phy Mask Trig IRR Pol"
			  " Stat Dest Deli Vect:   \n");
	printk(KERN_EMERG "Before:\n");
	print_line(&entry);

	entry.trigger = 0;
	io_apic_write(0, 0x11 + 2 * pin, *(((int *)&entry) + 1));
	io_apic_write(0, 0x10 + 2 * pin, *(((int *)&entry) + 0));
	udelay(10);
	printk(KERN_EMERG "After switching to edge:\n");
	print_line(&entry);

	entry.trigger = 1;
	io_apic_write(0, 0x11 + 2 * pin, *(((int *)&entry) + 1));
	io_apic_write(0, 0x10 + 2 * pin, *(((int *)&entry) + 0));
	udelay(10);
	printk(KERN_EMERG "After switch back:\n");
	print_line(&entry);

	local_irq_restore(flags);
}

2) add to sysrq.c:

@@ -137,6 +137,9 @@
 		send_sig_all(SIGKILL, 1);
 		orig_log_level = 8;
 		break;
+	case 'q':
+		kick_IOAPIC_pin(19);
+
 	default:					    /* Unknown: help */
 		if (kbd)
 			printk("unRaw ");

I added the patch, and tried it out. When the network hangs, I am able to revive it with ALT-SYSRQ-Q. The debug log shows these entries:

Jan 12 19:22:57 behemoth kernel: SysRq: <0> NR Log Phy Mask Trig IRR Pol Stat Dest Deli Vect:
Jan 12 19:22:57 behemoth kernel: Before:
Jan 12 19:22:57 behemoth kernel:  00 003 03  0    1    1   1   1    1    1    99
Jan 12 19:22:57 behemoth kernel: After switching to edge:
Jan 12 19:22:57 behemoth kernel:  00 003 03  0    0    1   1   1    1    1    99
Jan 12 19:22:57 behemoth kernel: After switch back:
Jan 12 19:22:57 behemoth kernel:  00 003 03  0    1    1   1   1    1    1    99

  WWWWW      _______________________
 ## o o\    /     Frank de Lange     \
 }#   \|   /                          \
  ##---# _/     <Hacker for Hire>      \
   ####   \      +31-320-252965        /
           \    [email protected]    /
            -------------------------
 [ "Omnis enim res, quae dando non deficit, dum habetur
    et non datur, nondum habetur, quomodo habenda est."  ]
-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to [email protected]
Please read the FAQ at http://www.tux.org/lkml/

I removed the disable_irq lines from 8390.c, and that fixed the problem:
no hang within 2 minutes - the test is still running.

Frank, could you double check it?

--
Manfred

// $Header$
// Kernel Version:

//  VERSION = 2
//  PATCHLEVEL = 4
//  SUBLEVEL = 0
//  EXTRAVERSION =

@@ -242,15 +242,15 @@
 
 	/* Ugly but a reset can be slow, yet must be protected */
 		
-	disable_irq_nosync(dev->irq);
-	spin_lock(&ei_local->page_lock);
+/*	disable_irq_nosync(dev->irq);*/
+	spin_lock_irqsave(&ei_local->page_lock, flags);
 		
 	/* Try to restart the card.  Perhaps the user has fixed something. */
 	ei_reset_8390(dev);
 	NS8390_init(dev, 1);
 		
-	spin_unlock(&ei_local->page_lock);
-	enable_irq(dev->irq);
+	spin_unlock_irqrestore(&ei_local->page_lock, flags);
+/*	enable_irq(dev->irq); */
 	netif_wake_queue(dev);
 }
     

@@ -285,9 +285,9 @@
 	 *	Slow phase with lock held.
 	 */
 	 
-	disable_irq_nosync(dev->irq);
+/*	disable_irq_nosync(dev->irq);*/
 	
-	spin_lock(&ei_local->page_lock);
+	spin_lock_irqsave(&ei_local->page_lock, flags);
 	
 	ei_local->irqlock = 1;
 

@@ -327,8 +327,8 @@
 		ei_local->irqlock = 0;
 		netif_stop_queue(dev);
 		outb_p(ENISR_ALL, e8390_base + EN0_IMR);
-		spin_unlock(&ei_local->page_lock);
-		enable_irq(dev->irq);
+		spin_unlock_irqrestore(&ei_local->page_lock, flags);
+/*		enable_irq(dev->irq);*/
 		ei_local->stat.tx_errors++;
 		return 1;
 	}

@@ -383,8 +383,8 @@
 	ei_local->irqlock = 0;
 	outb_p(ENISR_ALL, e8390_base + EN0_IMR);
 	
-	spin_unlock(&ei_local->page_lock);
-	enable_irq(dev->irq);
+	spin_unlock_irqrestore(&ei_local->page_lock, flags);
+/*	enable_irq(dev->irq); */
 
 	dev_kfree_skb (skb);
 	ei_local->stat.tx_bytes += send_length;

Hm, I also sent in a (somewhat different) patch on my own... :-)]

Anyway, still running under heavy load...

Cheers//Frank

  WWWWW      _______________________
 ## o o\    /     Frank de Lange     \
 }#   \|   /                          \
  ##---# _/     <Hacker for Hire>      \
   ####   \      +31-320-252965        /
           \    [email protected]    /
            -------------------------
 [ "Omnis enim res, quae dando non deficit, dum habetur
    et non datur, nondum habetur, quomodo habenda est."  ]
-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to [email protected]
Please read the FAQ at http://www.tux.org/lkml/

I'm currently running my own patched version, which uses spin_lock_irq/spin_unlock_irq instead of spin_lock_irqsave/spin_unlock_irqrestore like you patch uses. Looking at spinlock.h, spin_lock_irq does a local irq disable, which seems to be closer to the original intent (disable_irq) than spin_lock_irqsave. Anyone want to comment on this?

Anyway, still running under load, also got USB (which uses the same irq) to produce some interrupts by scanning some stuff. No problems so far...

Cheers//Frank

  WWWWW      _______________________
 ## o o\    /     Frank de Lange     \
 }#   \|   /                          \
  ##---# _/     <Hacker for Hire>      \
   ####   \      +31-320-252965        /
           \    [email protected]    /
            -------------------------
 [ "Omnis enim res, quae dando non deficit, dum habetur
    et non datur, nondum habetur, quomodo habenda est."  ]
-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to [email protected]
Please read the FAQ at http://www.tux.org/lkml/

It's a bit dangerous: _if_ one of the function is called with disabled local interrupts, then spin_unlock_irq would enable these interrupts. That could cause other problems, but I haven't checked if these function are actually called with disabled interrupts - e.g. the transmit function is called with enabled interrupts.

Frank, the 2.4.0 contains 2 band aids that were added for ne2k smp:

* From Ingo: focus cpu disabled, in arch/i386/kernel/apic.c
* From myself: TARGET_CPU = cpu_online_mask, was 0xFF.

Could you disable both bandaids? I disabled them, no problems so far.

Now back to the disable_irq_nosync().
--
Manfred
-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in the body of a message to [email protected]
Please read the FAQ at http://www.tux.org/lkml/

Ok so it looks like the disable_irq code is buggy. Unfortunately its not just used for these drivers they are just the heaviest users.

Given that we can see the IRQ is still set on the APIC can we fake an IRQ in that condition ?

-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in the body of a message to [email protected]
Please read the FAQ at http://www.tux.org/lkml/

Switch to edge triggered, wait , switch back to level triggered - I've added that to my sysrq commands, and it clears the IRR bit in the IO APIC. Sometimes 2 or 3 attempts are required.

I tried several changes to the ioapic code, but without success.

I found:
* the io_apic_sync() in __mask_IO_APIC_irq() is probably wrong, I'd say it should be _after_ the io_apic_modify, but currently it is before the io_apic_modify(). 2.2 uses the same code, that doesn't explain the problem.

* this comment in
http://oss.sgi.com/www.linux.sgi.com/intel/visws/visws.2210.28jul99.patch

+/* XXX ouch... is this really our only choice for masking this interrupt? */
+/* XXX not fully safe for 2 reasons:
+ * 1) should not touch an apic entry while (whole) apic is enabled
+ * 2) careful about storing to IRR bit (unless we know this intr is
idle)
+ */
+static void disable_cobalt_irq(unsigned int irq) /* disable_irq()
*/
+{
+ int ent = is_co_apic(irq);
+ if (ent == -1) {
+ return; /* could actually be a panic */
+ }
+
+ /* Note: h/w nada like read-mod-write! Vec saved in
IRQ_VECTOR() */
+ co_apic_write(CO_APIC_LO(ent), CO_APIC_MASK);
+ (void)co_apic_read(CO_APIC_LO(ent)); /* sync cpu to cobalt apic
*/
+}
+
It's possible that the comment is about an cobalt specific problem, the IRR bit is - according the Intel documentation - read only.

--
Manfred
-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in the body of a message to [email protected]
Please read the FAQ at http://www.tux.org/lkml/

It may well not be disable_irq() that is buggy. In fact, there's good reason to believe that it's a hardware problem.

Linus

-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in the body of a message to [email protected]
Please read the FAQ at http://www.tux.org/lkml/

Perhaps a problem with the 82093AA external IO APIC used for 440BX board? I haven't seen any reports from newer Intel boards (the ICH2 includes an IO APIC) or from Via boards.

The problem is not SMP specific: Frank wrote that it also occured when he booted with "max_cpus=1".

--
Manfred
-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in the body of a message to [email protected]
Please read the FAQ at http://www.tux.org/lkml/

I am inclined to believe it IS a hardware problem... If disable_irq were buggy, wouldn't the problem occur more frequently in other irq-heavy areas? A quick count shows that disable_irq* is used in 84 sourcefiles in the driver/* directory. This includes drivers which generate many interrupts in a short timeframe (like ide).

Frank

  WWWWW      _______________________
 ## o o\    /     Frank de Lange     \
 }#   \|   /                          \
  ##---# _/     <Hacker for Hire>      \
   ####   \      +31-320-252965        /
           \    [email protected]    /
            -------------------------
 [ "Omnis enim res, quae dando non deficit, dum habetur
    et non datur, nondum habetur, quomodo habenda est."  ]
-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to [email protected]
Please read the FAQ at http://www.tux.org/lkml/

IDE is not my favourite example of a "known stable driver". Also, in many cases IDE is for historical reasons connected to an EDGE io-apic pin (ie it's still considered an ISA interrupt). Which probably wouldn't show this problem anyway.

Also, IDE doesn't generate all that many interrupts. You can make a network driver do a _lot_ more interrupts than just about any disk driver by simply sending/receiving a lot of packets. With disks it is very hard to get the same kind of irq load - Linux will merge the requests and do at least 1kB worth of transfer per interrupt etc. On a ne2k 100Mbps PCI card, you can probably _easily_ generate a much higher stream of interrupts.

Linus

-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in the body of a message to [email protected]
Please read the FAQ at http://www.tux.org/lkml/

I've been having similar problem with my network card 3c59x. My BP6 was hanging with
50% probablity while running netscape from nfs mounted /usr partition).

For 2.2 kernel I've created this patch which has solved this problem (so I don't have any problems until recentely I've started to use UDMA4 with ATA66
- but this seems to be related to BX chipset problem and as far as I know Andre
is working on this issue)

While using this patch - I've got many messages about irq_enter in my kernel log,
but the system was stable and running quite happy.

With the latest 2.4 kernels I do not have same problems (at least it looks so far)
(So I assume the problem was hidden somewhere inside NFS). However with 2.4.0 & ac patches I'm seeing some problem when one CPU is locked
and the computer becomes unusable without any reason - there is no network trafick
no ide intensive actions - all I can do is to move mouse within Xfree (so maybe
its xfree bug - but its hard to recognize this - all I could say is that I'm not seeing
such problems with -test12 kernels)

Here the patch for 2.2 I've been using:

--- linux.orig/arch/i386/kernel/irq.h Wed May 31 11:21:04
2000
+++ linux/arch/i386/kernel/irq.h Wed May 31 11:21:47
2000
@@ -138,8 +138,22
@@
static inline void irq_enter(int cpu, unsigned int
irq)

{

hardirq_enter(cpu);
- while (test_bit(0,&global_irq_lock))
{
- /* nothing
*/;
+ if (global_irq_holder == cpu && test_bit(0, &global_irq_lock))
{
+ printk(KERN_WARNING "irq_enter - CPU:%d already holder
(count:%
+ cpu, global_irq_count,
local_irq_count[cpu]);
+ /* avoid deadlock bellow
*/
+
clear_bit(0,&global_irq_lock);
+ } else
{
+ unsigned long i =
1000000;
+ while (test_bit(0,&global_irq_lock) && i)
{
+
i--;
+ /* nothing
*/;
+
}
+ if (!i)
{
+
clear_bit(0,&global_irq_lock);
+ printk(KERN_WARNING "irq_enter - loop timeout CPU:%d
Holder:%d!
+ cpu,
global_irq_holder);
+
}

}

}
printk(KERN_WARNING "irq_enter - loop timeout CPU:%d
Holder:%d!

             There are three types of people in the world:
               those who can count, and those who can't.
  Zdenek Kabelac  http://i.am/kabi/ [email protected] {debian.org; fi.muni.cz}

-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to [email protected]
Please read the FAQ at http://www.tux.org/lkml/

They (ide interrupts) are indeed EDGE-triggered on my box. I have not enabled the HPT366 (ATA66) controller on this board, so I can not tell if that controller is EDGE-triggered as well.

There's sound... The msnd.c (Turtle Beach MultiSound) driver (and its derivatives, like msnd_pinnacle) uses disable_irq. Running esd (esound daemon), sound can easily generate > 1000 interrupts/second, since esd uses small dma transfers. This can be seen quite clearly from /proc/interrupts on my soundserver:

CPU0

  0:  276867328          XT-PIC  timer
  1:          2          XT-PIC  keyboard
  2:          0          XT-PIC  cascade
  3:    7631519          XT-PIC  eth1
  4:    2751419          XT-PIC  serial
  5: 1907346678          XT-PIC  soundblaster
  8:          1          XT-PIC  rtc
  9:   45022986          XT-PIC  eth0
 13:          1          XT-PIC  fpu
 14:    4320643          XT-PIC  ide0
 15:    4409193          XT-PIC  ide1

NMI: 0

OK, this is an ageing P166, and it uses a different driver, etc. I have not found any problems with hanging sound drivers in Google query for 'linux msnd bp6' or 'linux multisound bp6'. Of course, this is no conclusive evidence, far from it... It could be that people using those cards are not the ones who tend to go for the (somewhat tricky) BP6 board...

Cheers//Frank

  WWWWW      _______________________
 ## o o\    /     Frank de Lange     \
 }#   \|   /                          \
  ##---# _/     <Hacker for Hire>      \
   ####   \      +31-320-252965        /
           \    [email protected]    /
            -------------------------
 [ "Omnis enim res, quae dando non deficit, dum habetur
    et non datur, nondum habetur, quomodo habenda est."  ]
-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to [email protected]
Please read the FAQ at http://www.tux.org/lkml/

Actually, without mult count you will do only 512b of I/O per interrupt on IDE. Regardless of merging etc. Still doesn't reach nic levels, but it's _bad_ anyway :-)

* Jens Axboe <[email protected]>
* SuSE Labs
-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to [email protected]
Please read the FAQ at http://www.tux.org/lkml/

I doubt that it's BP6 specific: I have the problem with a Gigabyte BXD board and I doubt that Ingo used an BP6. Perhaps 82093AA specific (the IO APIC chip used for SMP 440BX board)

I can't find any spec updates for that chip: either it's the first perfect chip Intel ever produced, or ...

--
Manfred
-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in the body of a message to [email protected]
Please read the FAQ at http://www.tux.org/lkml/

3c59x calls disable_irq() once per minute, and seems to be
one of the most-affected drivers.
-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in the body of a message to [email protected]
Please read the FAQ at http://www.tux.org/lkml/

The ne2k thing seems to be the _most_ affected one, as far as I can tell.

However, it could easily be a matter of timing - for example, if the driver does something to trigger an interrupt _just_ before (or after, considering the asynchronous nature of writing to the IO-APIC) doing the enable_irq(), then..

I'm also nervous about the complete lack of locking in vortex_timer():
disabling interrupts doesn't mean that transmits couldn't be pending. But maybe the hardware is ok with changing status concurrently.

Linus

-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in the body of a message to [email protected]
Please read the FAQ at http://www.tux.org/lkml/

mm.. It's a little racy wrt vortex_ioctl(), but otherwise OK. del_timer_sync() in vortex_ioctl() seems to be needed.

disable_irq() is very useful in functions such as this. It would be a shame to have to stop using it.
-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in the body of a message to [email protected]
Please read the FAQ at http://www.tux.org/lkml/

Doesn't the NCR53C9x SCSI drivers use disable_irq() a lot? Do they have any problems?
_____

  |_____| ------------------------------------------------- ---+---+-
  |   |         Russell King        [email protected]      --- ---
  | | | | http://www.arm.linux.org.uk/personal/aboutme.html   /  /  |
  | +-+-+                                                     --- -+-
  /   |               THE developer of ARM Linux              |+| /|\
 /  | | |                                                     ---  |
    +-+-+ -------------------------------------------------  /\\\  |

-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in the body of a message to [email protected]
Please read the FAQ at http://www.tux.org/lkml/

It seems that a certain timing is necessary: one flood ping or a single ncp usually doesn't trigger any problems, but 2 concurrent flood pings hang the network after 5-10 seconds. It's not multi processor specific, both I and Frank can trigger it when we boot with one cpu.

So far it seems that only the io apic for the BX chipset is affected (only SMP BX boards contain an io apic).

Any volunteers with ne2k-pci cards and other motherboards that include an io apic (e.g. all Intel motherboards that use an IO Controller Hub, Via Apollo Pro133, Pro133A, KX133)?

Please:
* apply the attached patch.
* compile the kernel for SMP, or at least enable uniprocessor io apic support.
* reboot.
* flood ping the computer with 2 concurrent flood pings from a second computer.
* wait one minute.

According to the ICH2 documentation the IRR bit on the IO APIC is writable - that's either a docu error, or could cause further problems.

--
Manfred

@@ -270,7 +270,7 @@
 	 *   PCI Ne2000 networking cards and PII/PIII processors, dual
 	 *   BX chipset. ]
 	 */
-#if 0
+#if 1
 	/* Enable focus processor (bit==0) */
 	value &= ~(1<<9);
 #else

It seems that noone uses a Ne2000 compatible pci NIC with a newer motherboard (every K7 board, Intel 8xx boards, via apollo pro 133), but I've set up a tiny web site that describes my problem:

colorfullife.com/~manfred/io_apic

--
Manfred
-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in the body of a message to [email protected]
Please read the FAQ at http://www.tux.org/lkml/

Volunteer. Mine is a Realtek 8029:
00:0d.0 Ethernet controller: Realtek Semiconductor Co., Ltd. RTL-8029(AS)
Subsystem: Realtek Semiconductor Co., Ltd. RT8029(AS)
Flags: medium devsel, IRQ 11
I/O ports at ef40 [size=32]

Board: SuperMicro P6DGU (440GX,PIIX4), 256Mb, 2xPII@400(Deschutes,512Kb). The only problem is that I just have a cable connection. Is it enough a 'loopback' connection (ie, ping myself) ? Does it goes at least until the card and generates interrupts or stops at the kernel soft level ? Anyways, if a 128K connection can generate enough flood for you, i'll send you my results. I am going to test it anyways...

dmesg selected info:
I/O APIC #2 Version 17 at 0xFEC00000.
Int: type 3, pol 0, trig 0, bus 2, IRQ 00, APIC ID 2, APIC INT 00
Int: type 0, pol 0, trig 0, bus 2, IRQ 01, APIC ID 2, APIC INT 01
Int: type 0, pol 0, trig 0, bus 2, IRQ 00, APIC ID 2, APIC INT 02
Int: type 0, pol 0, trig 0, bus 2, IRQ 03, APIC ID 2, APIC INT 03
Int: type 0, pol 0, trig 0, bus 2, IRQ 04, APIC ID 2, APIC INT 04
Int: type 0, pol 0, trig 0, bus 2, IRQ 06, APIC ID 2, APIC INT 06
Int: type 0, pol 0, trig 0, bus 2, IRQ 07, APIC ID 2, APIC INT 07
Int: type 0, pol 1, trig 1, bus 2, IRQ 08, APIC ID 2, APIC INT 08
Int: type 0, pol 0, trig 0, bus 2, IRQ 09, APIC ID 2, APIC INT 09
Int: type 0, pol 0, trig 0, bus 2, IRQ 0c, APIC ID 2, APIC INT 0c
Int: type 0, pol 0, trig 0, bus 2, IRQ 0d, APIC ID 2, APIC INT 0d
Int: type 0, pol 0, trig 0, bus 2, IRQ 0e, APIC ID 2, APIC INT 0e
Int: type 0, pol 0, trig 0, bus 2, IRQ 0f, APIC ID 2, APIC INT 0f
Int: type 0, pol 3, trig 3, bus 2, IRQ 0a, APIC ID 2, APIC INT 10
Int: type 0, pol 3, trig 3, bus 2, IRQ 0b, APIC ID 2, APIC INT 11
Int: type 0, pol 3, trig 3, bus 2, IRQ 0b, APIC ID 2, APIC INT 12
Int: type 0, pol 3, trig 3, bus 2, IRQ 05, APIC ID 2, APIC INT 13
Int: type 2, pol 0, trig 0, bus 2, IRQ 00, APIC ID 2, APIC INT 17
Lint: type 3, pol 0, trig 0, bus 0, IRQ 00, APIC ID ff, APIC LINT 00
Lint: type 1, pol 0, trig 0, bus 0, IRQ 00, APIC ID ff, APIC LINT 01
Processors: 2
mapped APIC to ffffe000 (fee00000)
mapped IOAPIC to ffffd000 (fec00000)
..
ENABLING IO-APIC IRQs
..changing IO-APIC physical APIC ID to 2 ... ok.
Synchronizing Arb IDs.
init IO_APIC IRQs
IO-APIC (apicid-pin) 2-0, 2-5, 2-10, 2-11, 2-20, 2-21, 2-22, 2-23 not connected
.
.TIMER: vector=49 pin1=2 pin2=0
activating NMI Watchdog ... done.
testing NMI watchdog ... OK.
number of MP IRQ sources: 18.
number of IO-APIC #2 registers: 24.
testing the IO APIC.......................

IO APIC #2......
... register #00: 02000000
...... : physical APIC id: 02
... register #01: 00170011
...... : max redirection entries: 0017
...... : IO APIC version: 0011
... register #02: 00000000
...... : arbitration: 00
... IRQ redirection table:
NR Log Phy Mask Trig IRR Pol Stat Dest Deli Vect:

 00 000 00  1    0    0   0   0    0    0    00
 01 003 03  0    0    0   0   0    1    1    39
 02 003 03  0    0    0   0   0    1    1    31
 03 003 03  0    0    0   0   0    1    1    41
 04 003 03  0    0    0   0   0    1    1    49
 05 000 00  1    0    0   0   0    0    0    00
 06 003 03  0    0    0   0   0    1    1    51
 07 003 03  0    0    0   0   0    1    1    59
 08 003 03  0    0    0   0   0    1    1    61
 09 003 03  0    0    0   0   0    1    1    69
 0a 000 00  1    0    0   0   0    0    0    00
 0b 000 00  1    0    0   0   0    0    0    00
 0c 003 03  0    0    0   0   0    1    1    71
 0d 000 00  1    0    0   0   0    0    0    00
 0e 003 03  0    0    0   0   0    1    1    79
 0f 003 03  0    0    0   0   0    1    1    81
 10 003 03  1    1    0   1   0    1    1    89
 11 003 03  1    1    0   1   0    1    1    91
 12 003 03  1    1    0   1   0    1    1    91
 13 003 03  1    1    0   1   0    1    1    99
 14 000 00  1    0    0   0   0    0    0    00
 15 000 00  1    0    0   0   0    0    0    00
 16 000 00  1    0    0   0   0    0    0    00
 17 000 00  1    0    0   0   0    0    0    00

IRQ to pin mappings:
IRQ0 -> 2
IRQ1 -> 1
IRQ3 -> 3
IRQ4 -> 4
IRQ5 -> 19
IRQ6 -> 6
IRQ7 -> 7
IRQ8 -> 8
IRQ9 -> 9
IRQ10 -> 16
IRQ11 -> 17-> 18
IRQ12 -> 12
IRQ13 -> 13
IRQ14 -> 14
IRQ15 -> 15
................................... done.
calibrating APIC timer ...
.... CPU clock speed is 400.9147 MHz.
.... host bus clock speed is 100.2284 MHz.
cpu: 0, clocks: 1002284, slice: 334094
CPU0<T0:1002272,T1:668176,D:2,S:334094,C:1002284>
cpu: 1, clocks: 1002284, slice: 334094
CPU1<T0:1002272,T1:334080,D:4,S:334094,C:1002284>
checking TSC synchronization across CPUs: passed.
Setting commenced=1, go go go
..
Board Vendor: Supermicro Inc..
Board Name: Intel 440BX/440GX.

J.A. Magallon                                                      $> cd pub
mailto:[email protected]                                          $> more beer

Linux werewolf 2.4.0-ac8 #1 SMP Fri Jan 12 18:02:50 CET 2001 i686

-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to [email protected]
Please read the FAQ at http://www.tux.org/lkml/

In my case, (440GX/BX, PIIX4), network goes off (both with a ping-flood or some web browsing) with a message:

Jan 14 03:01:25 werewolf kernel: NETDEV WATCHDOG: eth0: transmit timed out
Jan 14 03:01:57 werewolf last message repeated 19 times

J.A. Magallon                                                      $> cd pub
mailto:[email protected]                                          $> more beer

Linux werewolf 2.4.0-ac9 #2 SMP Sun Jan 14 01:46:07 CET 2001 i686

-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to [email protected]
Please read the FAQ at http://www.tux.org/lkml/

It isn't. But I just meant to indicate that the mere fact that I could not find any problem-report for that combination does not indicate that there ARE no problems...

:-)

Well, the BX chipset is one of their better attempts I think...

Frank

  WWWWW      _______________________
 ## o o\    /     Frank de Lange     \
 }#   \|   /                          \
  ##---# _/     <Hacker for Hire>      \
   ####   \      +31-320-252965        /
           \    [email protected]    /
            -------------------------
 [ "Omnis enim res, quae dando non deficit, dum habetur
    et non datur, nondum habetur, quomodo habenda est."  ]
-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to [email protected]
Please read the FAQ at http://www.tux.org/lkml/

As you think about this problem, please bear in mind that this loss of APIC interrupts also occurs in 2.2 kernels.

Donald may have more details.

-
-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in the body of a message to [email protected]
Please read the FAQ at http://www.tux.org/lkml/

I disabled both (I guess you meant the 'define TARGET_CPUS cpu_online' in io_apic.c?), and reverted my own patch, added your patch... Now running with the usual heavy network load, no problems so far... Also made USB produce interrupts (shares irq with network), no problems...

Could this really be the solution?

Cheers//Frank

  WWWWW      _______________________
 ## o o\    /     Frank de Lange     \
 }#   \|   /                          \
  ##---# _/     <Hacker for Hire>      \
   ####   \      +31-320-252965        /
           \    [email protected]    /
            -------------------------
 [ "Omnis enim res, quae dando non deficit, dum habetur
    et non datur, nondum habetur, quomodo habenda est."  ]
-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to [email protected]
Please read the FAQ at http://www.tux.org/lkml/

I'd like to know _which_ of the two makes a difference (or does it only trigger with both of them enabled)? And even then I'm not sure that it is "the" solution - both changes to io-apic handling had some reason for them. Ingo, what was the focus-cpu thing?

Linus

-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in the body of a message to [email protected]
Please read the FAQ at http://www.tux.org/lkml/

well, some time ago i had an ne2k card in an SMP system as well, and found this very problem. Disabling/enabling focus-cpu appeared to make a difference, but later on i made experiments that show that in both cases the hang happens. I spent a good deal of time trying to fix this problem, but failed - so any fresh ideas are more than welcome.

Ingo

-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in the body of a message to [email protected]
Please read the FAQ at http://www.tux.org/lkml/

Frank, please clarify:
you still run without disable_irq_nosync() in 8390.c?

I have a first idea: we send an EOI to an interrupt that is masked on the IO apic, perhaps that causes the problems.

I'm right now typing a patch.

--
Manfred
-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in the body of a message to [email protected]
Please read the FAQ at http://www.tux.org/lkml/

I am running with your patched version of 8390.c (so WITHOUT disable_irq_nosync()).

In addition, I patched apic.c (focus cpu enabled) In addition, I patched io_apic ((TARGET_CPUS 0xff)

Sound plausible...

I'll await yours instead of making my own patch this time... :-)

Cheers//Frank

  WWWWW      _______________________
 ## o o\    /     Frank de Lange     \
 }#   \|   /                          \
  ##---# _/     <Hacker for Hire>      \
   ####   \      +31-320-252965        /
           \    [email protected]    /
            -------------------------
 [ "Omnis enim res, quae dando non deficit, dum habetur
    et non datur, nondum habetur, quomodo habenda est."  ]
-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to [email protected]
Please read the FAQ at http://www.tux.org/lkml/

please try it with the focus CPU enabling change (we want to enable that feature, i only disabled it due to the stuck-ne2k bug), but with TARGET_CPUS set to cpu_online_mask. (this later is needed for certain crappy BIOSes.)

i believe the ne2k driver change is the key.

does not help. I've tried it (and many other combinations). I did not find any direct workaround for this problem. (i tried very hard.)

Ingo

-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in the body of a message to [email protected]
Please read the FAQ at http://www.tux.org/lkml/

WITH or WITHOUT the changed 8390 driver? I can already give you the results for running WITH the changed driver: it works. I have not yet tried it WITHOUT the changed 8390 driver (so that would be stock 8390, patched apic.c, stock io_apic.c). Please let me know which you want...

Frank

  WWWWW      _______________________
 ## o o\    /     Frank de Lange     \
 }#   \|   /                          \
  ##---# _/     <Hacker for Hire>      \
   ####   \      +31-320-252965        /
           \    [email protected]    /
            -------------------------
 [ "Omnis enim res, quae dando non deficit, dum habetur
    et non datur, nondum habetur, quomodo habenda est."  ]
-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to [email protected]
Please read the FAQ at http://www.tux.org/lkml/

WITH. patched 8390.c, patched apic.c, sock io_apic.c. My very strong feeling is that this will be a stable combination, and that this is what we want as a final solution.

Ingo

-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in the body of a message to [email protected]
Please read the FAQ at http://www.tux.org/lkml/

If you do that please #ifdef SMP all the changes. Its impossible to use a modem and an ne2K together on a typical PC otherwise. The copy from the NE2K with irq disabled is just _SO_ slow you drop bytes continually.

I did all the horrible magic in the ne2k driver for a reason. The other alternative is to provide a way to force the system back out of apic mode so the ne2K driver can do a

goodbye_apic_crap()

type call

-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in the body of a message to [email protected]
Please read the FAQ at http://www.tux.org/lkml/

It is. As I already mentioned in other messages, I already tested with JUST the patched 8390.c driver, no other patches. It was stable. I then patched apic.c AND io_apic.c, which did not introduce new instabilities. Unless you think that reverting back to a stock io_apic.c would cause instabilities (which would be weird, since I had no instabilities running only a patched 8390.c), I think the patch to 8390.c DOES remove the symptoms all by itself. No other patches seem necessary to get a stable box.

But I'll patch the mess again just fox kicks :-)

Cheers//Frank

  WWWWW      _______________________
 ## o o\    /     Frank de Lange     \
 }#   \|   /                          \
  ##---# _/     <Hacker for Hire>      \
   ####   \      +31-320-252965        /
           \    [email protected]    /
            -------------------------
 [ "Omnis enim res, quae dando non deficit, dum habetur
    et non datur, nondum habetur, quomodo habenda est."  ]
-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to [email protected]
Please read the FAQ at http://www.tux.org/lkml/

okay - i just wanted to hear a definitive word from you that this fixes your problem, because this is what we'll have to do as a final solution. (barring any other solution.)

Ingo

-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in the body of a message to [email protected]
Please read the FAQ at http://www.tux.org/lkml/

Patching 8390.c won't fix this for me. The only thing on IRQ19 when I saw interrupts die was usb-uhci, and that doesn't appear to use disable_irq.

But then again, I've only ever seen this happen once. It's not repeatable.

dwmw2


-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to [email protected]
Please read the FAQ at http://www.tux.org/lkml/

Ingo, is that possible?

The current fix is "disable_irq_nosync() and enable_irq() cause deadlocks with level triggered ioapic irqs, do not use them" - I'm sure ne2k-pci isn't the only driver that uses these function.

I have found one combination that doesn't hang with the unpatched 8390.c, but network throughput is down to 1/2. I hope that's due to the debugging changes.

I'll restart now from a fresh 2.4.0 tree:
Changes:

1) enable focus cpu.
2) apply the attached patch.

I'm not sure if it's a real fix or if it just hides the problem: my sysrq patch has shown that clearing and setting the "level trigger" bit in the io apic reanimates the IO APIC.

--
Manfred

@@ -134,6 +134,30 @@
 	spin_unlock_irqrestore(&ioapic_lock, flags);
 }
 
+DO_ACTION( __trigger_level,    0, |= 0x00008000, io_apic_sync(entry->apic))/* mask = 1 */
+DO_ACTION( __trigger_edge,  0, &= 0xffff7fff, )				/* mask = 0 */
+
+
+static void unmask_level_IO_APIC_irq (unsigned int irq)
+{
+	unsigned long flags;
+
+	spin_lock_irqsave(&ioapic_lock, flags);
+	__trigger_level_IO_APIC_irq(irq);
+	__unmask_IO_APIC_irq(irq);
+	spin_unlock_irqrestore(&ioapic_lock, flags);
+}
+
+static void mask_level_IO_APIC_irq (unsigned int irq)
+{
+	unsigned long flags;
+
+	spin_lock_irqsave(&ioapic_lock, flags);
+	__mask_IO_APIC_irq(irq);
+	__trigger_edge_IO_APIC_irq(irq);
+	spin_unlock_irqrestore(&ioapic_lock, flags);
+}
+
 static void unmask_IO_APIC_irq (unsigned int irq)
 {
 	unsigned long flags;

@@ -143,6 +167,7 @@
 	spin_unlock_irqrestore(&ioapic_lock, flags);
 }
 
+
 void clear_IO_APIC_pin(unsigned int apic, unsigned int pin)
 {
 	struct IO_APIC_route_entry entry;

@@ -1181,14 +1206,14 @@
  */
 static unsigned int startup_level_ioapic_irq (unsigned int irq)
 {
-	unmask_IO_APIC_irq(irq);
+	unmask_level_IO_APIC_irq(irq);
 
 	return 0; /* don't check for pending */
 }
 
-#define shutdown_level_ioapic_irq	mask_IO_APIC_irq
-#define enable_level_ioapic_irq		unmask_IO_APIC_irq
-#define disable_level_ioapic_irq	mask_IO_APIC_irq
+#define shutdown_level_ioapic_irq	mask_level_IO_APIC_irq
+#define enable_level_ioapic_irq		unmask_level_IO_APIC_irq
+#define disable_level_ioapic_irq	mask_level_IO_APIC_irq
 
 static void end_level_ioapic_irq (unsigned int i)
 {

Hm, could it be that the fact that network throughput is halved causes the problem not to appear? Remember, it only appears under HEAVY network load. A single nfs cp -rd <big_dir> was not enough to hang my network, I needed to add at least another cp -rd or some streaming audio or something else...

Cheers//Frank

  WWWWW      _______________________
 ## o o\    /     Frank de Lange     \
 }#   \|   /                          \
  ##---# _/     <Hacker for Hire>      \
   ####   \      +31-320-252965        /
           \    [email protected]    /
            -------------------------
 [ "Omnis enim res, quae dando non deficit, dum habetur
    et non datur, nondum habetur, quomodo habenda est."  ]
-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to [email protected]
Please read the FAQ at http://www.tux.org/lkml/

No. The problem is still there. But now lots of losts packets instead of a total hang.

Due to the modification of mask_irq now disable_irq_nosync and enable_irq act as if I would press SysRQ+q every millisecond, and thus the io apic is immediatly reset when it got stuck.

Btw, my initial assumption about EOI to masked interrupt must be wrong:
2.2 always first masks the irq, then it sends the EOI, and 2.2 doesn't hang.

--
Manfred
-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in the body of a message to [email protected]
Please read the FAQ at http://www.tux.org/lkml/

Now running with this config:

PATCHED 8390.c (using irq_safe spinlocks instead of disable_irq) PATCHED apic.c (focus cpu ENABLED)
STOCK io_apic.c

No problems under heavy network load.

Gentleman, this (the patch to 8390.c) seems to fix the problem.

Cheers//Frank

  WWWWW      _______________________
 ## o o\    /     Frank de Lange     \
 }#   \|   /                          \
  ##---# _/     <Hacker for Hire>      \
   ####   \      +31-320-252965        /
           \    [email protected]    /
            -------------------------
 [ "Omnis enim res, quae dando non deficit, dum habetur
    et non datur, nondum habetur, quomodo habenda est."  ]
-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to [email protected]
Please read the FAQ at http://www.tux.org/lkml/

great. Back when i had the same problem, flood pinging another host (on the local network) was the quickest way to reproduce the hang:

ping -f -s 10 otherhost

this produced an IOAPIC-hang within seconds.

Ingo

-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in the body of a message to [email protected]
Please read the FAQ at http://www.tux.org/lkml/

Apart from killing streaming audio and interactive network use, nothing hangs. As soon as the ping flood is stopped, audio streams on and ssh sessions are useable again. So, it seems to fix it...

Frank

  WWWWW      _______________________
 ## o o\    /     Frank de Lange     \
 }#   \|   /                          \
  ##---# _/     <Hacker for Hire>      \
   ####   \      +31-320-252965        /
           \    [email protected]    /
            -------------------------
 [ "Omnis enim res, quae dando non deficit, dum habetur
    et non datur, nondum habetur, quomodo habenda est."  ]
-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to [email protected]
Please read the FAQ at http://www.tux.org/lkml/

The problem with this patch is that anybody with a slow ISA ne2000 clone will basically have absolutely _horrible_ interrupt latency because we hold the irq lock over some quite expensive operations.

The spin_lock_irqsave() is absolutely my preferred fix, and if I remember correctly this is in fact how some early 2.1.x code fixed the ne2000 driver when the original irq scalability stuff happened (for some time during development we did not have a working "disable_irq()" AT ALL because the irq-disabling counters etc logic hadn't been done).

The spinlock was changed to "disable_irq()" by a patch from Alan, if I remember correctly, exactly because people couldn't access serial lines at any kind of high speeds otherwise - even on "reasonable" hardware.

Alan may remember details better. The fact is that as a general design principle we should _not_ be using "disable_irq/enable_irq" anyway. BUT that there are some real-world concerns that make the "better" spinlock handling have some problems too.

So yes, I bet the spinlock fixes it. But..

Linus

-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in the body of a message to [email protected]
Please read the FAQ at http://www.tux.org/lkml/

And that's the patch I meant... Manfred's spin_lock_irqsave/spin_unlock_irqrestore based one, not my (spin_lock_irq/spin_unlock_irq) based patch. That is also the one I'm running now.

Frank

  WWWWW      _______________________
 ## o o\    /     Frank de Lange     \
 }#   \|   /                          \
  ##---# _/     <Hacker for Hire>      \
   ####   \      +31-320-252965        /
           \    [email protected]    /
            -------------------------
 [ "Omnis enim res, quae dando non deficit, dum habetur
    et non datur, nondum habetur, quomodo habenda est."  ]
-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to [email protected]
Please read the FAQ at http://www.tux.org/lkml/

The old code did it with #ifdef __SMP__ tests so it only screwed up SMP boxes, which at the time was quite acceptable because real people didnt have them and certainly at the price didnt put ne2000's in them 8)

The basic problem is that you cannot allow

	-	set multicast list
	-	open/close
	-	irq
	-	transmit

to occur except when serialized because of the indirection and register gunge on the chip. The copies are slow and long enough that they prevent 28.8 modem sessions being usable.

I'd have to check the chip manual to be sure you even disable the irqs without corrupting an in progress transfer

-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in the body of a message to [email protected]
Please read the FAQ at http://www.tux.org/lkml/

I do have a 3c503 and a ne2k-pci both of them use the 8390, I can hang the ne2k-pci easily by doing a ping -f, bigger packet size => early the hang. But I cannot hang the 3c503 by doing this.

Now with 2.4.0 the ne2k-pci behaviour is that: doing a ping -f works for some amount of time, then stops for a BIG amount of time (various minutes), and then it works again (it seems), but at a much slower speed, and if you test it with normal ping (ping host) you don't get replies.

The packets really go down to the wire and I even got replies. but I don't receive it.

Previous versions of 2.4.0-testX caused ne2k-pci to hang and remain hanged until reboot.

System: Mb Gigabyte 586dx, 2x200MMX, 96Mb RAM,

Jorge Nerin
<[email protected]>
-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to [email protected]
Please read the FAQ at http://www.tux.org/lkml/

for the record. my BP6, non OC, apic smp system with ne2k fails within 24 hours here too. if I can be of any help..... (2.4.0. kernel. no vmware or opensound)

Grobbebol's Home                   |  Don't give in to spammers.   -o)
http://www.xs4all.nl/~bengel       | Use your real e-mail address   /\
Linux 2.2.16 SMP 2x466MHz / 256 MB |        on Usenet.             _\_v  
-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to [email protected]
Please read the FAQ at http://www.tux.org/lkml/

You can help yourself by applying Manfred's patch to 8390.c (in preference to my own patch to the same file). This will sove the hanging-network problem. If your entire box hangs, that's another story which will probably not be fixed by that patch. You can find the patch in Manfred's posting to the list from Fri Jan 12 2001 - 14:04:24 EST.

I've been running a patched driver for more than a day now, under heavy network load, without problems.

Frank

  WWWWW      _______________________
 ## o o\    /     Frank de Lange     \
 }#   \|   /                          \
  ##---# _/     <Hacker for Hire>      \
   ####   \      +31-320-252965        /
           \    [email protected]    /
            -------------------------
 [ "Omnis enim res, quae dando non deficit, dum habetur
    et non datur, nondum habetur, quomodo habenda est."  ]
-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to [email protected]
Please read the FAQ at http://www.tux.org/lkml/

Well, with 'this' (in 'could THIS be') I really meant the move from disable_irq to the irq_safe spinlocks. I'm currently running with the patched 8390.c driver, patched io_apic (TARGET_CPUS 0xff) and patched apic.c (focus cpu enabled), and have had no problems yet... even though I'm running several simulatnsous nfs cp -rd <big_dir>, streaming network audio, scanning with an USB scanner, etc.

So far, it seems that the patch to 8390.c removed the symptoms. The changes to apic.c and io_apic.c did not make the network hang come back.

Cheers//Frank

  WWWWW      _______________________
 ## o o\    /     Frank de Lange     \
 }#   \|   /                          \
  ##---# _/     <Hacker for Hire>      \
   ####   \      +31-320-252965        /
           \    [email protected]    /
            -------------------------
 [ "Omnis enim res, quae dando non deficit, dum habetur
    et non datur, nondum habetur, quomodo habenda est."  ]
-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to [email protected]
Please read the FAQ at http://www.tux.org/lkml/

Well, sideways ofcourse! :-)

here's a cat /proc/interrupts from the (BP6) box:

CPU0 CPU1

  0:     104936     105433    IO-APIC-edge  timer
  1:       4444       4384    IO-APIC-edge  keyboard
  2:          0          0          XT-PIC  cascade
  3:         79         59    IO-APIC-edge  serial
  4:      12743      12850    IO-APIC-edge  serial
 14:       7855       7885    IO-APIC-edge  ide0
 15:       1990       1703    IO-APIC-edge  ide1
 16:          0          0   IO-APIC-level  es1371, mga@PCI:1:0:0
 17:         24         28   IO-APIC-level  sym53c8xx
 18:          0          0   IO-APIC-level  bttv
 19:     460435     460402   IO-APIC-level  eth0, eth1, usb-uhci
NMI:     210303     210303 
LOC:     210285     210284 
ERR:          0

The interrupt which caused problems was 19 (with both network cards and USB on it). It shows a high number of interrupts because I've been load-testing the network. The mere fact that it shows this hig number of interrupts shows the fix works...

As this is a BP6, I'm now supposed to go on about the dead chickens, dedicated air conditioners, nuclear powersupplies and other magic you're supposed to buy to get these boards running. Well, nothing of that sort, it is running on a simple (but high quality) 235W PSU with heatgreased coolers on the CPUs and the BX xhipset. Nothing is overclocked. CPU and chipset tmeperatures are 24.C and 32.C, respectively.

In short, nothing remarkable. All PCI slots are used, as you can see from my first posting in this thread (which contains more info on the hardware).

//Frank

  WWWWW      _______________________
 ## o o\    /     Frank de Lange     \
 }#   \|   /                          \
  ##---# _/     <Hacker for Hire>      \
   ####   \      +31-320-252965        /
           \    [email protected]    /
            -------------------------
 [ "Omnis enim res, quae dando non deficit, dum habetur
    et non datur, nondum habetur, quomodo habenda est."  ]
-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to [email protected]
Please read the FAQ at http://www.tux.org/lkml/