FreeBSD 9.0-RELEASE amd64 strange interrupts behavior
Александр В. Шутко
AVShutko at mail.khstu.ru
Tue May 29 02:49:55 UTC 2012
Hi, All. I can't get working Adaptec 6805 driver on 9.0 for unknown reason.
I got Supermicro X8DTH-iF based server with two Xeon 5645 (6 cores x 2 threads),
24Gb RAM, Adaptec sas 6805 raid. I downloaded 6805 driver code for FreeBSD 8.3
from adaptec site and compiled it under fresh-installed FreeBSD 9.0-RELEASE (amd64),
kldloaded it and got hung and timeouts:
aacu0: <Adaptec RAID Controller> mem 0xf8c00000-0xf8ffffff,0xf8bbc000-0xf8bbc7ff,
0xf8bbe000-0xf8bbe0ff irq 50 at device 0.0 on pci132
aacu0: Enable Raw I/O
aacu0: Enable 64-bit array
aacu0: New comm. interface type1 enabled
ioapic2: routing intpin 2 (PCI IRQ 50) to lapic 0 vector 50
aacu0: Adaptec 6805, aac driver 2.4.2-18668
aacu0: Supported Options=141f1d7c<WCACHE,DATA64,HOSTTIME,RAID50,WINDOW4GB,
SOFTERR,SGMAP64,ALARM,NONDASD,ADPTINFO,NEWCOMM,ARRAY64BIT,HEATSENSOR>
aacd0: <RAID 6> on aacu0
aacd0: 2860022MB (5857325056 sectors)
GEOM: new disk aacd0
pci133: driver added
pci134: driver added
pci254: driver added
pci255: driver added
aacp0: <SCSI Passthrough Bus> on aacu0
aacp1: <SCSI Passthrough Bus> on aacu0
aacp2: <SCSI Passthrough Bus> on aacu0
pci0: driver added
aacu0: COMMAND 0xffffff8001fd08e8 TIMEOUT AFTER 40 SECONDS
aacu0: COMMAND 0xffffff8001fd0ac8 TIMEOUT AFTER 40 SECONDS
aacu0: COMMAND 0xffffff8001fd0ca8 TIMEOUT AFTER 40 SECONDS
aacu0: COMMAND 0xffffff8001fd0e88 TIMEOUT AFTER 40 SECONDS
aacu0: COMMAND 0xffffff8001fd1068 TIMEOUT AFTER 40 SECONDS
aacu0: COMMAND 0xffffff8001fd1248 TIMEOUT AFTER 40 SECONDS
aacu0: COMMAND 0xffffff8001fd1428 TIMEOUT AFTER 40 SECONDS
aacu0: COMMAND 0xffffff8001fd1608 TIMEOUT AFTER 40 SECONDS
aacu0: COMMAND 0xffffff8001fd17e8 TIMEOUT AFTER 40 SECONDS
.....
Then I installed 8.3, compiled driver and it started without any problem.
After that I cvsup-ed FreeBSD10 code (head, 28.05.2012), compiled it,
booted new kernel (set module_path/boot clang), compiled driver and it
kldloaded again without any error (raid6 2.8Tb disk read-write speed good,
no errors).
Here is kernel 10 aacu driver dmesg lines:
aacu0: <Adaptec RAID Controller> mem 0xf8c00000-0xf8ffffff,
0xf8bbc000-0xf8bbc7ff,0xf8bbe000-0xf8bbe0ff irq 50 at device 0.0 on pci132
aacu0: Enable Raw I/O
aacu0: Enable 64-bit array
aacu0: New comm. interface type1 enabled
ioapic2: routing intpin 2 (PCI IRQ 50) to lapic 33 vector 49
aacu0: Adaptec 6805, aac driver 2.4.2-18668
aacu0: Supported Options=141f1d7c<WCACHE,DATA64,HOSTTIME,RAID50,WINDOW4GB,SOFTERR,SGMAP64,
ALARM,NONDASD,ADPTINFO,NEWCOMM,ARRAY64BIT,HEATSENSOR>
aacd0 on aacu0
aacd0: 2860022MB (5857325056 sectors)
GEOM: new disk aacd0
pci133: driver added
pci134: driver added
......
The problem exists only under 9.0-RELEASE. Interrupts doesn't work for this
device at all. Driver makes device initialization using sync requests and
then hungs on first async io waiting interrupt.
Driver interrupts install code looks good:
sc->aac_irq_rid = 0;
if ((sc->aac_irq = bus_alloc_resource_any(sc->aac_dev, SYS_RES_IRQ,
&sc->aac_irq_rid,
RF_SHAREABLE | RF_ACTIVE)) == NULL) {
device_printf(sc->aac_dev, "can't allocate interrupt\n");
return (EINVAL);
}
if (sc->flags & AAC_FLAGS_NEW_COMM_TYPE1) {
if (bus_setup_intr(sc->aac_dev, sc->aac_irq,
INTR_MPSAFE|INTR_TYPE_BIO, NULL,.
aac_new_intr_type1, sc, &sc->aac_intr)) {
device_printf(sc->aac_dev, "can't set up interrupt\n");
return (EINVAL);
}
I added "panic" call at the beginning of interrupt handler
"aac_new_intr_type1()" and found it never called during driver
initialization on 9.0. Is this 9.0-RELEASE bug or future ?
--
Александр В. Шутко <AVShutko at mail.khstu.ru>
More information about the freebsd-stable
mailing list