ATAng lockups (Re: geom<->ata forever cycle)

 Sat, Nov 22, 2003 at 13:08:10, netch (Valentin Nechayev) wrote about "geom<->ata forever cycle": 

VN> 5.1-current of 2003-11-20 hangs during kernel initialization with forever
VN> cycle around geom partitioning detection and ATA read failure.
VN> Message rate is too fast to record, no stopping is available, no serial
VN> console.
VN> Previous 5-current on this machine was of 2003-05-23.
VN> How to diagnose?

Well, I restricted time interval to between 2003-08-18 and 2003-08-25, so it is
ATAng. Also tested 2003-11-20 and 2003-10-06 with the same result.
Different poses were seen:
- forever cycle around GEOM detecting
- simple lockup
- panic("initiate_write_inodeblock_ufs1: already started")
- another kinds of panic
and all have common sign - impossibility of driver to dig itself out of
lockup (switch to PIO, reset bus, etc.)
Dirty fix works to set hw.ata.ata_dma=0 in loader; moreover, `atacontrol
mode 1 udma100 -' doesn't cause problems, but `atacontrol mode 0 udma33 -'
drops system to broken state in a few minutes.

Please say how to dig this problem.

Full verbose dmesg.boot (from 2003-08-18 kernel) and config (unchanged)
was put in previous letters and can be resent personally.
Entries, bound with ATA, in first problematic (2003-08-25) dmesg.boot:

pcib0: <Intel 82815 (i815 GMCH) Host To Hub bridge> at pcibus 0 on motherboard
atapci0: <Intel ICH2 UDMA100 controller> port 0xf000-0xf00f at device 31.1 on pci0
ata0: pre reset mask=03 ostat0=50 ostat2=00
ata0-master: ATAPI 00 00
ata0-slave: ATAPI 14 eb
ata0: after reset mask=03 stat0=50 stat1=00
ata0-master: ATA 01 a5
ata0: devices=09
ata0: at 0x1f0 irq 14 on atapci0
ata0-slave: pio=0x0c wdma=0x22 udma=0x42 cable=40pin
ata0-master: pio=0x0c wdma=0x22 udma=0x44 cable=40pin
ad0: setting PIO4 on Intel ICH2 chip
ad0: <IBM-DJNA-351520/J56OA30K> ATA-4 disk at ata0-master
ad0: 14664MB (30033360 sectors), 29795 C, 16 H, 63 S, 512 B
ad0: 16 secs/int, 1 depth queue, PIO4
acd0: setting PIO4 on Intel ICH2 chip
acd0: <CD-540E/1.0A> CDROM drive at ata0 as slave
acd0: read 6890KB/s (6890KB/s), 128KB buffer, PIO4
(remind that it is with hw.ata.ata_dma=0)


-netch-