Le 16/03/13 09:01, Richard PALO a écrit :
Le 14/03/13 17:42, Richard PALO a écrit :
richard@x3200:~# echo ::interrupts -d |mdb -k
IRQ Vect IPL Bus Trg Type CPU Share APIC/INT# Driver Name(s)
1 0x41 5 ISA Edg Fixed 3 1 0x0/0x1 i8042#0
9 0x80 9 PCI Lvl Fixed 1 1 0x0/0x9 acpi_wrapper_isr
12 0x42 5 ISA Edg Fixed 0 1 0x0/0xc i8042#0
14 0x40 5 ISA Edg Fixed 0 1 0x0/0xe ata#0
20 0x60 6 PCI Lvl Fixed 2 3 0x0/0x14 nge#0, pci-ide#1,
pci-ide#1
21 0x81 9 PCI Lvl Fixed 1 1 0x0/0x15 nvidia#0
22 0x82 9 PCI Lvl Fixed 2 1 0x0/0x16 ehci#0
23 0x83 9 PCI Lvl Fixed 3 1 0x0/0x17 ohci#0
160 0xa0 0 Edg IPI all 0 - poke_cpu
208 0xd0 14 Edg IPI all 1 -
kcpc_hw_overflow_intr
209 0xd1 14 Edg IPI all 1 - cbe_fire
210 0xd3 14 Edg IPI all 1 - cbe_fire
240 0xe0 15 Edg IPI all 1 - xc_serv
241 0xe1 15 Edg IPI all 1 - apic_error_intr
after scratching my ear a bit, I realised that my sata disks were being
served by the ide driver (as indicated above).
The chipset is MCP61 (nforce 630a), so I did the following and am
relieved that nv_sata took over.
# update_drv -a -i "pci10de,3f6" nv_sata
( https://www.illumos.org/issues/1023 )
I'm trying to see with ASROCK support if I can prevent the shared IRQs
since there's quite a bit of room:
echo ::interrupts -d |mdb -k
IRQ Vect IPL Bus Trg Type CPU Share APIC/INT# Driver Name(s)
1 0x43 5 ISA Edg Fixed 2 1 0x0/0x1 i8042#0
9 0x80 9 PCI Lvl Fixed 1 1 0x0/0x9 acpi_wrapper_isr
12 0x44 5 ISA Edg Fixed 3 1 0x0/0xc i8042#0
14 0x41 5 ISA Edg Fixed 3 1 0x0/0xe ata#0
20 0x83 9 PCI Lvl Fixed 2 2 0x0/0x14 ohci#0, nv_sata#0
21 0x60 6 PCI Lvl Fixed 1 2 0x0/0x15 nge#0, nv_sata#1
22 0x81 9 PCI Lvl Fixed 0 1 0x0/0x16 nvidia#0
23 0x82 9 PCI Lvl Fixed 1 1 0x0/0x17 ehci#0
160 0xa0 0 Edg IPI all 0 - poke_cpu
208 0xd0 14 Edg IPI all 1 - kcpc_hw_overflow_intr
209 0xd1 14 Edg IPI all 1 - cbe_fire
210 0xd3 14 Edg IPI all 1 - cbe_fire
240 0xe0 15 Edg IPI all 1 - xc_serv
241 0xe1 15 Edg IPI all 1 - apic_error_intr
I'm curious whether this has a bit to do with the hangs (heavy disk
banging during builds)... because since using nv_sata, I have not
noticed any more real hangs (outside of 100% cpu). In this event, good
news on that front.
testing in parallele acpica-unix-20130214, if successful, I'll post my
updates in illumos.devel
-------------------------------------------
The good news is that since about a week I've been running
acpica-20130823 (see illumos dev for webrev) and I haven't had a hang since.
Good thing as I was ready to toss this ASROCK, now I'm rather happy
(except still can't complete an omnios build which is still on the old
acpica bits, I'm booted on hipster which is greatly improved as of late)
I notice a bit better response in general... and the nv_sata devices
align on the same IRQ now as well which may have had at least something
to do with the problem.
Before I could rarely finish a full illumos-gate build or a serious
pkgsrc compile, I've done many of both since without problem.
It would be extremely useful to know if other mainboards can benefit
from the updates to get a push committed after the upcoming september
release once tested out.
In the meanwhile, I'll try to finish the acpidump update (which is now
included in acpica (will be easy to integrate with Mr Pankovs
restructuring work) in order to check out what's up with dsdt-hpet
tables as Mr Saini indicates.
I might try the MSI trick again as well, for grins.
Also, I noticed the following in freebsd's code:
#if defined(__amd64__) || defined(__i386__)
/*
* XXX: Certain BIOSes have buggy AML that specify an IRQ that is
* edge-sensitive and active-lo. However, edge-sensitive IRQs
* should be active-hi. Force IRQs with an ISA IRQ value to be
* active-hi instead.
*/
if (irq < 16 && trig == ACPI_EDGE_SENSITIVE && pol == ACPI_ACTIVE_LOW)
pol = ACPI_ACTIVE_HIGH;
#endif
BUS_CONFIG_INTR(dev, irq, (trig == ACPI_EDGE_SENSITIVE) ?
INTR_TRIGGER_EDGE : INTR_TRIGGER_LEVEL, (pol == ACPI_ACTIVE_HIGH) ?
INTR_POLARITY_HIGH : INTR_POLARITY_LOW);
}
I will add a debug statement in psm_common.c to see what arrives in case
this is also a corner case given I see edge-sensitive IRQs (kbd/mouse).
-------------------------------------------
illumos-discuss
Archives: https://www.listbox.com/member/archive/182180/=now
RSS Feed: https://www.listbox.com/member/archive/rss/182180/21175430-2e6923be
Modify Your Subscription:
https://www.listbox.com/member/?member_id=21175430&id_secret=21175430-6a77cda4
Powered by Listbox: http://www.listbox.com
