Every semaphore call results in "uninterruptable sleep"

Every semaphore function (semget(),..) , "ipcs -s" ... results in "uninterruptable sleep".
The processes cannot be killed by kill -9.

I've two boxes with Dual Xeons 1.8 GHz and HT option enabled in BIOS. When I boot 2.4.20-rc2 with default arguments the kernel detects 4 processors and also reports 4 processors in /proc/cpuinfo wiht the "ht" feature.

However, if I boot with the "noht" option (wich I believed turned off HT and therefore only two processors should be available), the kernel still detects 4 processors, _but_ now the "ht" feature in /proc/cpuinfo is not there. Is this the intention of the "noht" option ?

If so, are there any
options available to turn off HT support in the kernel completely, so that I don't have to go into the BIOS to turn it off ?

If you want to I can provide the dmesg output and .config.

I think this issue is not just related to 2.4.20-rc2, but earlier kernels aswell.

Quoted 9 lines · Steffen Persvold

I've two boxes with Dual Xeons 1.8 GHz and HT option enabled in BIOS. When I boot 2.4.20-rc2 with default arguments the kernel detects 4 processors and also reports 4 processors in /proc/cpuinfo wiht the "ht" feature.

However, if I boot with the "noht" option (wich I believed turned off HT and therefore only two processors should be available), the kernel still detects 4 processors, _but_ now the "ht" feature in /proc/cpuinfo is not there. Is this the intention of the "noht" option ?

No, the intention of the "noht" option is as you believed, and it used to work that way. I looked at latest sourcec, didn't see anything obviously wrong. Sorry, but I don't have a suitable HT machine to check at present.

Quoted 4 lines

If so, are there any
options available to turn off HT support in the kernel completely, so that I don't have to go into the BIOS to turn it off ?

Quoted 4 lines

If you want to I can provide the dmesg output and .config.

I think this issue is not just related to 2.4.20-rc2, but earlier kernels aswell.

Could others check and report if 2.4.20-rc2 "noht" works for them or not?

Quoted 19 lines · Hugh Dickins

Quoted 9 lines · Steffen Persvold

I've two boxes with Dual Xeons 1.8 GHz and HT option enabled in BIOS. When I boot 2.4.20-rc2 with default arguments the kernel detects 4 processors and also reports 4 processors in /proc/cpuinfo wiht the "ht" feature.

However, if I boot with the "noht" option (wich I believed turned off HT and therefore only two processors should be available), the kernel still detects 4 processors, _but_ now the "ht" feature in /proc/cpuinfo is not there. Is this the intention of the "noht" option ?

No, the intention of the "noht" option is as you believed, and it used to work that way. I looked at latest sourcec, didn't see anything obviously wrong. Sorry, but I don't have a suitable HT machine to check at present.

Quoted 4 lines

If so, are there any
options available to turn off HT support in the kernel completely, so that I don't have to go into the BIOS to turn it off ?

Quoted 7 lines

Quoted 4 lines

If you want to I can provide the dmesg output and .config.

I think this issue is not just related to 2.4.20-rc2, but earlier kernels aswell.

Could others check and report if 2.4.20-rc2 "noht" works for them or not?

Thanks for your reply Hugh,

After digging into it a bit, I _think_ I know what happens. When the machine is bootet with default parameters, the processors are detected in the ACPI table, and the MP table search is skipped :

Linux version 2.4.20-rc2 ([email protected]) (gcc version 3.2 20020903 (Red Hat Linux 8.0 3.2-7)) #4 SMP Sat Nov 16 14:06:13 CET 2002 BIOS-provided physical RAM map:
BIOS-e820: 0000000000000000 - 000000000009f400 (usable)
BIOS-e820: 000000000009f400 - 00000000000a0000 (reserved)
BIOS-e820: 00000000000d8000 - 00000000000e0000 (reserved)
BIOS-e820: 00000000000e4000 - 0000000000100000 (reserved)
BIOS-e820: 0000000000100000 - 000000003fef0000 (usable)
BIOS-e820: 000000003fef0000 - 000000003fefc000 (ACPI data)
BIOS-e820: 000000003fefc000 - 000000003ff00000 (ACPI NVS)
BIOS-e820: 000000003ff00000 - 000000003ff80000 (usable)
BIOS-e820: 000000003ff80000 - 0000000040000000 (reserved)
BIOS-e820: 00000000fec00000 - 00000000fec10000 (reserved)
BIOS-e820: 00000000fee00000 - 00000000fee01000 (reserved)
BIOS-e820: 00000000ff800000 - 00000000ffc00000 (reserved)
BIOS-e820: 00000000fff00000 - 0000000100000000 (reserved)
127MB HIGHMEM available.
896MB LOWMEM available.
found SMP MP-table at 000f7030
hm, page 000f7000 reserved twice.
hm, page 000f8000 reserved twice.
hm, page 0009f000 reserved twice.
hm, page 000a0000 reserved twice.
On node 0 totalpages: 262016
zone(0): 4096 pages.
zone(1): 225280 pages.
zone(2): 32640 pages.
ACPI: Searched entire block, no RSDP was found.
ACPI: RSDP located at physical address c00f7090
RSD PTR v0 [PTLTD ]
__va_range(0x3fef84e0, 0x68): idx=8 mapped at ffff6000
ACPI table found: RSDT v1 [PTLTD RSDT 1540.0]
__va_range(0x3fefbe78, 0x24): idx=8 mapped at ffff6000
__va_range(0x3fefbe78, 0x74): idx=8 mapped at ffff6000
ACPI table found: FACP v1 [INTEL K_CANYON 1540.0]
__va_range(0x3fefbeec, 0x24): idx=8 mapped at ffff6000
__va_range(0x3fefbeec, 0x9c): idx=8 mapped at ffff6000
ACPI table found: APIC v1 [PTLTD APIC 1540.0]
__va_range(0x3fefbeec, 0x9c): idx=8 mapped at ffff6000
LAPIC (acpi_id[0x0000] id[0x0] enabled[1])
CPU 0 (0x0000) enabledProcessor #0 Pentium 4(tm) XEON(tm) APIC version 16

LAPIC (acpi_id[0x0001] id[0x6] enabled[1])
CPU 1 (0x0600) enabledProcessor #6 Pentium 4(tm) XEON(tm) APIC version 16

LAPIC (acpi_id[0x0002] id[0x1] enabled[1])
CPU 2 (0x0100) enabledProcessor #1 Pentium 4(tm) XEON(tm) APIC version 16

LAPIC (acpi_id[0x0003] id[0x7] enabled[1])
CPU 3 (0x0700) enabledProcessor #7 Pentium 4(tm) XEON(tm) APIC version 16

IOAPIC (id[0x2] address[0xfec00000] global_irq_base[0x0]) IOAPIC (id[0x3] address[0xfec80000] global_irq_base[0x18]) IOAPIC (id[0x4] address[0xfec80400] global_irq_base[0x30]) INT_SRC_OVR (bus[0] irq[0x0] global_irq[0x2] polarity[0x1] trigger[0x1]) INT_SRC_OVR (bus[0] irq[0x9] global_irq[0x9] polarity[0x1] trigger[0x3]) LAPIC_NMI (acpi_id[0x0000] polarity[0x1] trigger[0x1] lint[0x1]) LAPIC_NMI (acpi_id[0x0001] polarity[0x1] trigger[0x1] lint[0x1]) LAPIC_NMI (acpi_id[0x0002] polarity[0x1] trigger[0x1] lint[0x1]) LAPIC_NMI (acpi_id[0x0003] polarity[0x1] trigger[0x1] lint[0x1]) 4 CPUs total
Local APIC address fee00000
__va_range(0x3fefbf88, 0x24): idx=8 mapped at ffff6000 __va_range(0x3fefbf88, 0x28): idx=8 mapped at ffff6000
ACPI table found: BOOT v1 [PTLTD $SBFTBL$ 1540.0] __va_range(0x3fefbfb0, 0x24): idx=8 mapped at ffff6000 __va_range(0x3fefbfb0, 0x50): idx=8 mapped at ffff6000
ACPI table found: SPCR v1 [PTLTD $UCRTBL$ 1540.0]
Enabling the CPU's according to the ACPI table
Intel MultiProcessor Specification v1.4
Virtual Wire compatibility mode.
OEM ID: Product ID: Kings Canyon APIC at: 0xFEE00000
I/O APIC #2 Version 32 at 0xFEC00000.
I/O APIC #3 Version 32 at 0xFEC80000.
I/O APIC #4 Version 32 at 0xFEC80400.
Processors: 4
Kernel command line: ro root=LABEL=/ console=ttyS0,9600n8 nmi_watchdog=1 [snip]

However, when I boot the machine with the "noht" option the ACPI table search is skipped, _but_ it still finds 4 processors in the MP table :

Linux version 2.4.20-rc2 ([email protected]) (gcc version 3.2 20020903 (Red Hat Linux 8.0 3.2-7)) #4 SMP Sat Nov 16 14:06:13 CET 2002 BIOS-provided physical RAM map:
BIOS-e820: 0000000000000000 - 000000000009f400 (usable)
BIOS-e820: 000000000009f400 - 00000000000a0000 (reserved)
BIOS-e820: 00000000000d8000 - 00000000000e0000 (reserved)
BIOS-e820: 00000000000e4000 - 0000000000100000 (reserved)
BIOS-e820: 0000000000100000 - 000000003fef0000 (usable)
BIOS-e820: 000000003fef0000 - 000000003fefc000 (ACPI data)
BIOS-e820: 000000003fefc000 - 000000003ff00000 (ACPI NVS)
BIOS-e820: 000000003ff00000 - 000000003ff80000 (usable)
BIOS-e820: 000000003ff80000 - 0000000040000000 (reserved)
BIOS-e820: 00000000fec00000 - 00000000fec10000 (reserved)
BIOS-e820: 00000000fee00000 - 00000000fee01000 (reserved)
BIOS-e820: 00000000ff800000 - 00000000ffc00000 (reserved)
BIOS-e820: 00000000fff00000 - 0000000100000000 (reserved)
127MB HIGHMEM available.
896MB LOWMEM available.
found SMP MP-table at 000f7030
hm, page 000f7000 reserved twice.
hm, page 000f8000 reserved twice.
hm, page 0009f000 reserved twice.
hm, page 000a0000 reserved twice.
On node 0 totalpages: 262016
zone(0): 4096 pages.
zone(1): 225280 pages.
zone(2): 32640 pages.
Intel MultiProcessor Specification v1.4
Virtual Wire compatibility mode.
OEM ID: Product ID: Kings Canyon APIC at: 0xFEE00000
Processor #0 Pentium 4(tm) XEON(tm) APIC version 20
Processor #6 Pentium 4(tm) XEON(tm) APIC version 20
Processor #1 Pentium 4(tm) XEON(tm) APIC version 20
Processor #7 Pentium 4(tm) XEON(tm) APIC version 20
I/O APIC #2 Version 32 at 0xFEC00000.
I/O APIC #3 Version 32 at 0xFEC80000.
I/O APIC #4 Version 32 at 0xFEC80400.
Processors: 4
Kernel command line: ro root=LABEL=/ console=ttyS0,9600n8 nmi_watchdog=1 noht

Since the HT feature is enabled in the BIOS it seems normal to me that 4 processors are reported both in the MP table and in the ACPI table. Shouldn't the 'noht' option handle this ?

I have two of these machines and can assist in testing out patches etc.

I know too little to comment definitively, but it's my understanding that a dual HT machine should only show 2 processors in its MP table, their siblings only appearing through analysis of the ACPI tables.

Whether it's that your MP table has been wrongly set up, or that you've really been given 4 processors when you only asked for 2 (sue your supplier!), I cannot say. I've copied Jun at Intel and Arjan at RedHat, and hope they can shed more light on this.

Quoted 9 lines · Hugh Dickins

I know too little to comment definitively, but it's my understanding that a dual HT machine should only show 2 processors in its MP table, their siblings only appearing through analysis of the ACPI tables.

Whether it's that your MP table has been wrongly set up, or that you've really been given 4 processors when you only asked for 2 (sue your supplier!), I cannot say. I've copied Jun at Intel and Arjan at RedHat, and hope they can shed more light on this.

Linux has zero problem with a sane MP table that lists all CPU's. Intel normally seems to recommend against it (maybe N3.51 doesn't like it or so) but it's all fair as far as I'm concerned. The bios is supposed to offer you a choice
to disable hyperthreading, use that ;)

Sure, the bios has this option (and it works). I just believed the 'noht' option would disable it from a kernel perspective. I understand that if the MP table lists 4 processors, the kernel must think it is 4 processors and enable them. But what is the purpose of the 'noht' option ? If it is to avoid scanning the ACPI table for CPUs, wouldn't it be less confusing to call it something like 'acpismp=disable', since you apparently can't disable the siblings anyway (when they are also listed in the MP table) ?