Messages in this thread | | | Date | Wed, 22 Mar 2000 23:24:17 +0100 | From | Ove Ewerlid <> | Subject | Re: Linux 2.3.x and Promise IDE card |
| |
Andre Hedrick wrote: > > yOn Mon, 20 Mar 2000, Ove Ewerlid wrote: > > > Andre Hedrick wrote: > > > > > > That is because one could enable the chipset with out AUTODMA and then > > > try, but crash the system, to retune the drive but the chipset is still > > > set at PIO 0. Thus attempting to do DMA transfers from the drive to a > > > PIO 0 configured host adapter will CRASH........ > > > > > > This will be fixed or is pixed in 2.3.99pre2 > > > > Just tried it with 2.3.99pre3 and the DMA crash is still there (20262 > > chipset). > > Still works fine with no DMA (at reduced speed). > > (This time I saw a crash error message saying that DMA was attempted > > while > > not supported by the chipset, consistent with what you say above.) > > Mar 20 02:20:43 p6dnf kernel: > PDC20262: ROM enabled at 0xfebb0000 > PDC20262: (U)DMA Burst Bit ENABLED Primary PCI Mode Secondary PCI Mode. > PDC20262: FORCING PRIMARY MODE BIT 0x00 -> 0x01 MASTER > PDC20262: FORCING SECONDARY MODE BIT 0x00 -> 0x01 MASTER > ide0: BM-DMA at 0xef00-0xef07, BIOS settings: hda:DMA, hdb:DMA > ide1: BM-DMA at 0xef08-0xef0f, BIOS settings: hdc:DMA, hdd:pio > > hda: QUANTUM FIREBALL ST6.4A, 6149MB w/81kB Cache, CHS=13328/15/63, UDMA(33) > hdb: QUANTUM FIREBALL ST3.2A, 3079MB w/81kB Cache, CHS=782/128/63, UDMA(33) > hdc: QUANTUM FIREBALLP KA9.1, 8809MB w/371kB Cache, CHS=17898/16/63, UDMA(66) > > I went and dug-out this card see if it was broken and it is not.
Turns out that the problem was an unnecessary 'hdparm -d 1 /dev/XXX'. The error message is:
Mar 22 18:30:13 labridge kernel: hdh: timeout waiting for DMA Mar 22 18:30:13 labridge kernel: ide_dmaproc: choppiest supported ide_dma_timeout func only: 12 Mar 22 18:30:13 labridge kernel: hdh: irq timeout: status=0xd0 { Busy } Mar 22 18:30:13 labridge kernel: hdh: DMA disabled Mar 22 18:30:13 labridge kernel: ide3: reset: success
However, running some stress test scripts on dual PII/400 I observed this: 1) 2.3.XX is much faster on sequential IO (ala dd if=/dev/hde of=/dev/null) then 2.2.x 2) Alas, 2.3.XX locks up under this load:
dd if=/dev/hde of=/dev/null & dd if=/dev/hdf of=/dev/null & dd if=/dev/hdg of=/dev/null & dd if=/dev/hdh of=/dev/null &
This runs OK for hours using DMA, but adding:
rsh server "cat /dev/zero" > /dev/null
(Gives close to 100 Mbit/s) Gives errors like this within 10 seconds or so:
Mar 22 18:54:17 labridge kernel: ide_dmaproc: choppiest supported ide_dma_lostirq func only: 11 Mar 22 18:54:17 labridge kernel: hdf: lost interrupt Mar 22 18:54:17 labridge kernel: ide_dmaproc: choppiest supported ide_dma_lostirq func only: 11 Mar 22 18:54:17 labridge kernel: hdh: lost interrupt Mar 22 18:54:27 labridge kernel: ide_dmaproc: choppiest supported ide_dma_lostirq func only: 11 Mar 22 18:54:27 labridge kernel: hde: lost interrupt Mar 22 18:54:27 labridge kernel: ide_dmaproc: choppiest supported ide_dma_lostirq func only: 11 Mar 22 18:54:27 labridge kernel: hdg: lost interrupt Mar 22 18:54:37 labridge kernel: ide_dmaproc: choppiest supported ide_dma_lostirq func only: 11 Processes accessing the devices locks and eventually dies after kill -9. All devices attached to the 20262 are effectively dead after this. (Have tried both 3c90x and tulip network cards) Same problem if maxcpus=1 or if DMA is not set.
This problem does not happen in 2.2.x.
Ove
-- Ove Ewerlid Email: Ove.Ewerlid@[syscon|signal|material].uu.se Phone: +46 70 666 23 63, Fax: +46 18 503 611, +46 18 555 096
- To unsubscribe from this list: send the line "unsubscribe linux-kernel" in the body of a message to majordomo@vger.rutgers.edu Please read the FAQ at http://www.tux.org/lkml/
| |