All, I'm running 4.9-RC3 on a new installation with three new hard drives that are all Ultra100-compatible. BIOS is set to Auto for IDE drives and correctly reports sizes with UDMA 100 LBA modes. The OS sees: ad0: 76319MB <WDC WD800JB-00ETA0> [155061/16/63] at ata0-master UDMA100 ad1: 238475MB <WDC WD2500JB-32FUA0> [484521/16/63] at ata0-slave UDMA100 ad2: 238475MB <WDC WD2500JB-32FUA0> [484521/16/63] at ata1-master UDMA100 When I connect the drives to the onboard IDE, only the first IDE channel (ad0 and ad1) works right at UDMA100. On the second channel, with ad2 only, I get errors when writing (reading is not a problem): ad2s1e: UDMA ICRC error writing fsbn 643383313 of 321036265-321036392 (ad2s1 bn 643383313; cn 40048 tn 193 sn 34) retrying ad2s1e: UDMA ICRC error writing fsbn 643383313 of 321036265-321036392 (ad2s1 bn 643383313; cn 40048 tn 193 sn 34) retrying ad2s1e: UDMA ICRC error writing fsbn 643383313 of 321036265-321036392 (ad2s1 bn 643383313; cn 40048 tn 193 sn 34) retrying Occasionally, the retries succeed. But eventually, the OS sets the ATA mode to PIO: ad2s1e: UDMA ICRC error writing fsbn 643383313 of 321036265-321036392 (ad2s1 bn 643383313; cn 40048 tn 193 sn 34) falling back to PIO mode Of course, PIO mode works (albeit slowly). If I manually set the ATA mode with atacontrol to UDMA66, it works fine. Moving the drives to a Promise Ultra100 PCI card works, too, but is not what I want. I've tried two UDMA100 cables with no difference (and they work on the Promise card as a control test, so they're fine). BIOS is set to auto for everything -- CPU, bus, SDRAM frequency -- no over-clocking or any other aggressive settings. I'm accessing a vinum striped and mirrored file system across ad1 and ad2, if that's significant. (But if it were, why would the errors appear only on the ad2 device?). I've run out of ideas. Am I'm missing something? Thanks for your help. --Serious P.S. Please include my address in your reply.
Serious Signal wrote:> All, > > I'm running 4.9-RC3 on a new installation with three new hard drives that > are all Ultra100-compatible. BIOS is set to Auto for IDE drives and > correctly reports sizes with UDMA 100 LBA modes. The OS sees: > > ad0: 76319MB <WDC WD800JB-00ETA0> [155061/16/63] at ata0-master UDMA100 > ad1: 238475MB <WDC WD2500JB-32FUA0> [484521/16/63] at ata0-slave UDMA100 > ad2: 238475MB <WDC WD2500JB-32FUA0> [484521/16/63] at ata1-master UDMA100 > > When I connect the drives to the onboard IDE, only the first IDE channel > (ad0 and ad1) works right at UDMA100. On the second channel, with ad2 only, > I get errors when writing (reading is not a problem): > > ad2s1e: UDMA ICRC error writing fsbn 643383313 of 321036265-321036392 (ad2s1 > bn 643383313; cn 40048 tn 193 sn 34) retrying > ad2s1e: UDMA ICRC error writing fsbn 643383313 of 321036265-321036392 (ad2s1 > bn 643383313; cn 40048 tn 193 sn 34) retrying > ad2s1e: UDMA ICRC error writing fsbn 643383313 of 321036265-321036392 (ad2s1 > bn 643383313; cn 40048 tn 193 sn 34) retrying > > Occasionally, the retries succeed. But eventually, the OS sets the ATA mode > to PIO: > > ad2s1e: UDMA ICRC error writing fsbn 643383313 of 321036265-321036392 (ad2s1 > bn 643383313; cn 40048 tn 193 sn 34) falling back to PIO modeInteresting that I get a similar situation, be it with a regular hard disk (no special setups such as vinum). 'dmesg' says: [...] ata1-slave: ATA identify retries exceeded ad0: 76319MB <ST380021A> [155061/16/63] at ata0-master UDMA66 acd0: CDROM <GCR-8521B> at ata1-master PIO4 Mounting root from ufs:/dev/ad0s1a ad0s1a: UDMA ICRC error reading fsbn 137023 of 68480-68607 (ad0s1 bn 137023; cn 8 tn 134 sn 61) retrying ad0s1a: UDMA ICRC error reading fsbn 137023 of 68480-68607 (ad0s1 bn 137023; cn 8 tn 134 sn 61) retrying ad0s1a: UDMA ICRC error reading fsbn 137023 of 68480-68607 (ad0s1 bn 137023; cn 8 tn 134 sn 61) retrying ad0s1a: UDMA ICRC error reading fsbn 137023 of 68480-68607 (ad0s1 bn 137023; cn 8 tn 134 sn 61) falling back to PIO mode When I reported this to the STABLE list, my PC was blamed of having some broken communication to my harddisk. Now I wonder if this is indeed the case; or will you also be blamed for having a bad PC :). Regards, Rob.