Problems with PCI SATA Controller
Hi, I've had a number of problems with this card. Please note I'm not using this as a RAID card. [EMAIL PROTECTED]:3:0: class=0x010400 card=0x61121095 chip=0x31121095 rev=0x02 hdr=0x00 vendor = 'Silicon Image Inc (Was: CMD Technology Inc)' device = 'SiI 3112 SATALink/SATARaid Controller' class= mass storage subclass = RAID Under FreeBSD 5, it'd continually generate timeout - WRITE_DMA errors which would make the disks operate really slow... I found many others to be having the same issue and they recommended dropping back to PIO mode... seems I cannot do that on this card. Under FreeBSD 6-BETA5, I never managed to get it installed... was getting an error DANGER WILL ROBINSON Under FreeBSD 6-RELEASE, I wasn't able to install with multiple disk slices which I have attempted several times for the machine to lock up when it gets to 28% copied each time. I ended up partitioning as a single slice and strangely this worked. However, it seems whenever there is a bit of disk activity the machine locks up just after dumping the error ata2: DISCONNECT requested Strangely this seems to happen everytime if I begin accessing both disks I have attached to this controller at once. ___ freebsd-stable@freebsd.org mailing list http://lists.freebsd.org/mailman/listinfo/freebsd-stable To unsubscribe, send any mail to [EMAIL PROTECTED]
Re: Problems with PCI SATA Controller
Peter Hoskin wrote: Hi, I've had a number of problems with this card. Please note I'm not using this as a RAID card. [EMAIL PROTECTED]:3:0: class=0x010400 card=0x61121095 chip=0x31121095 rev=0x02 hdr=0x00 vendor = 'Silicon Image Inc (Was: CMD Technology Inc)' device = 'SiI 3112 SATALink/SATARaid Controller' class= mass storage subclass = RAID Under FreeBSD 5, it'd continually generate timeout - WRITE_DMA errors which would make the disks operate really slow... I found many others to be having the same issue and they recommended dropping back to PIO mode... seems I cannot do that on this card. Under FreeBSD 6-BETA5, I never managed to get it installed... was getting an error DANGER WILL ROBINSON Under FreeBSD 6-RELEASE, I wasn't able to install with multiple disk slices which I have attempted several times for the machine to lock up when it gets to 28% copied each time. I ended up partitioning as a single slice and strangely this worked. However, it seems whenever there is a bit of disk activity the machine locks up just after dumping the error ata2: DISCONNECT requested Strangely this seems to happen everytime if I begin accessing both disks I have attached to this controller at once. Unfortunately the sii 3112 is a bit of a horrornumerous people have experienced issues with it (web search on sii 3112 data corruption makes quite interesting reading). I seem to recall a posting suggesting that some success might be had with just 1 SATA channel (i.e 1 disk) attached, however I can't find it offhand. Cheers Mark ___ freebsd-stable@freebsd.org mailing list http://lists.freebsd.org/mailman/listinfo/freebsd-stable To unsubscribe, send any mail to [EMAIL PROTECTED]
Re: Problems with PCI SATA Controller
Mark Kirkwood wrote: Unfortunately the sii 3112 is a bit of a horrornumerous people have experienced issues with it (web search on sii 3112 data corruption makes quite interesting reading). I seem to recall a posting suggesting that some success might be had with just 1 SATA channel (i.e 1 disk) attached, however I can't find it offhand. Cheers Mark *sigh* I guess I'll be taking the card back and getting some PATA - SATA adaptors. I need two drives as I wish to do a mirrored RAID, which with this card seems to be out of the question. Regards, Peter Hoskin ___ freebsd-stable@freebsd.org mailing list http://lists.freebsd.org/mailman/listinfo/freebsd-stable To unsubscribe, send any mail to [EMAIL PROTECTED]
Re: Problems with PCI SATA Controller
Peter Hoskin wrote: Mark Kirkwood wrote: Unfortunately the sii 3112 is a bit of a horrornumerous people have experienced issues with it (web search on sii 3112 data corruption makes quite interesting reading). I seem to recall a posting suggesting that some success might be had with just 1 SATA channel (i.e 1 disk) attached, however I can't find it offhand. Cheers Mark *sigh* I guess I'll be taking the card back and getting some PATA - SATA adaptors. I need two drives as I wish to do a mirrored RAID, which with this card seems to be out of the question. Regards, Peter Hoskin The Promise 2 channel cards (TX2200, TX2300) seem to work quite well (and are reasonably cheap). regards Mark ___ freebsd-stable@freebsd.org mailing list http://lists.freebsd.org/mailman/listinfo/freebsd-stable To unsubscribe, send any mail to [EMAIL PROTECTED]
Re: Problems with PCI SATA controller (bug in ATA driver? both ATAng and ATAmkIII)
* Brooks Davis ([EMAIL PROTECTED]) wrote: The SII3112 is a piece of crap that won't work reliably. Order something better (Soren recommends Promise cards). Are all Promise cards currently supported? Even their 8/16+ port SATA-II RAID6 cards? PCI-Express? The list of supported controllers suggests it maxes out at some of their more basic 4 port cards, but maybe the bigger cards just have multiple controller chips? The ata manpage could use a little more detail on this; if nothing else, it should have an asterisk next to the chipsets that aren't expected to work reliably, like the entire SiI range :) -- Thomas 'Freaky' Hurst http://hur.st/ ___ freebsd-stable@freebsd.org mailing list http://lists.freebsd.org/mailman/listinfo/freebsd-stable To unsubscribe, send any mail to [EMAIL PROTECTED]
Problems with PCI SATA controller (bug in ATA driver? both ATAng and ATAmkIII)
(Please CC me in replies as I'm not a subscriber of this list) Hello, I'm having difficulties getting a system with an PCI Serial ATA controller stable under stress. Sorry for the long email, but I'm pretty confused about the problem, so my description won't be as structured as I'd have liked. First some background about the system. It's a Intel D915GAV mainboard with 1GB RAM and a P4 3GHz (Hyperthreading turned on, both in BIOS and in FreeBSD). It has four SATA channels via the ICH6 SATA controller and one PATA channel. It has two SATA HD's (mirrored via geom_mirror) for the OS, and three SATA HD's (RAID 5 via gvinum) for data. The system disks work flawlessly, and the system is stable under stress (eg. a make world). However, after some time of stress on the data RAID array, eg. copying 35G via rsync over the fxp0 network interface, I get errors about READ_DMA timeouts (exact errors below). The drive that gives the errors eventually disappears, and gvinum takes down the RAID array if two disks do this. This only happens to the disks that are connected to the PCI SATA controller. I tried both a Promise SATAII 150 TX4 controller (PDC 40518) and an Adaptec SATA Connect SII3112A based controller. The SII3112A even gives errors (ad6: TIMEOUT - READ_DMA retrying (2 retries left) LBA=21679174) when I run newfs. I planned to run FreeBSD 5.4 on it. When copying lots of data to the data RAID array, I eventually (sometimes after 10 minutes, sometimes after 2 hours) get this errors: ad6: TIMEOUT - READ_DMA retrying (2 retries left) LBA=21679174 ad6: FAILURE - ATA_IDENTIFY timed out ad6: FAILURE - ATA_IDENTIFY timed out agvinum: lost drive 'backup2' d6: WARNING - removed from configuration aGEOM_VINUM: subdisk backup.p0.s1 state change: up - down GEOM_VINUM: plex backup.p0 state change: up - degraded ta3-master: FAILURE - READ_DMA timed out Sometimes it spontaneously rebooted, or crashed, in other cases, all processes that tried to access the disks stopped. This is one of the crashes: GEOM_VINUM: plex backup.p0 state change: degraded - down ARNING - removed from configuration GEOM_VINUM: plex backup.p0 state change: degraded - down ARNING - removed from configuration a Fatal trap 12: page fault while in kernel mode cpuid = 1; apic id = 01 fault virtual address = 0x98 fault code= supervisor write, page not present instruction pointer = 0x8:0xc04edb6a stack pointer = 0x10:0xe336bcc8 frame pointer = 0x10:0xe336bcc8 code segment = base 0x0, limit 0xf, type 0x1b = DPL 0, pres 1, def32 1, gran 1 processor eflags = interrupt enabled, resume, IOPL = 0 current process = 3 (g_up) trap number = 12 panic: page fault cpuid=1 ta5-master: FAILURE - READ_DMA timed out boot() called on cpu#0 Uptime: 12m57s GEOM_MIRROR: Device gm0: provider mirror/gm0 destroyed. ad14: WARNING - WRITE_DMA interrupt was seen but timeout fired LBA=69625709 ad14: WARNING - WRITE_DMA interrupt was seen but timeout fired LBA=69625709 (sorry, no backtrace or dump, don't have that kernel anymore) ad6 and ad14 were connected to the Promise TX4. I've never seen this error for a disk connected to the ICH6 controller. I also ran the HD manufacturer's test utility (advanced test), and it didn't find any errors, so I assume the HD's are fine. I also replaced the SATA cables. I tried disabling ACPI and hyperthreading. I don't think a SATA disk can run without UDMA (PIO mode), at least atacontrol won't do it. It doesn't many how many disks are on the PCI controller, so it doesn't seem to have anything todo with concurrent access to multiple channels. It doesn't surprise me that I get this errors with the SII3112A controller, since it's supposed to be pretty crappy, but I expected the promise controller to be well supported. I also tried the ATA-mkIIIn patch from people.freebsd.org/~sos/ATA, the only difference was a different error message: ad4: WARNING - SETFEATURES SET TRANSFER MODE interrupt was seen but timeout fired ad4: req=0xc3cdfa28 SETFEATURES SET TRANSFER MODE semaphore timeout !! DANGER Will Robinsion !! (lots of these, and all processes that try to access the FS hang) Same with FreeBSD 6.0-RC1. I've yet to try a different mainboard, but I want to have this server working soon, so maybe I'll just put four disks on the ICH6 controller and forget about the fifth disk until I figure this out. Please let me know if you've any ideas, this certainly looks like a bug in the ATA driver to me. Alson A verbose dmesg from 5.4-RELEASE (most recent RELENG_5_4, -p7 or so) is below (I believe the top part is missing, probably because it didn't fit in the ring buffer, let me know if it's important and I'll try to obtain it). bus=0, slot=28, func=1 class=06-04-00, hdrtype=0x01, mfdev=1 cmdreg=0x0106, statreg=0x0010, cachelnsz=16 (dwords) lattimer=0x00 (0 ns), mingnt=0x06 (1500 ns), maxlat=0x00 (0 ns) intpin=b, irq=255 found-
Re: Problems with PCI SATA controller (bug in ATA driver? both ATAng and ATAmkIII)
(Please CC me in replies as I'm not a subscriber of this list) Update: I tested with an i810-based mainboard (Celeron 1GHz, RTL8139 ethernet, Promise SATAII 150 TX4 controller, 3 SATA disks in RAID 5, FreeBSD 6.0-RC1). It remained stable for two hours. I suspect this is because it has far less bandwith (iostat showed only about 3MB/s to the disks, as opposed to 12MB/s with the i915 mainboard). After I added a dd if=/dev/zero of=foo bs=128k (this increased the bandwidth usage to the disk to about 9MB/s according to iostat), it crashed in about 40 minutes. This suggests that it crashes because of the large amount of I/O. However, it's only about 10MB/s per disk (for three disks), so it doesn't seem to be that exotic to me. Since this is a completely different mainboard, it seems clearly a software issue to me. The built-in ICH6 controller works fine however, so it may be PDC*0518/SII311* specific (which basically means any PCI SATA controller available locally). Alson ___ freebsd-stable@freebsd.org mailing list http://lists.freebsd.org/mailman/listinfo/freebsd-stable To unsubscribe, send any mail to [EMAIL PROTECTED]
Re: Problems with PCI SATA controller (bug in ATA driver? both ATAng and ATAmkIII)
On Thu, Oct 20, 2005 at 04:30:48AM +0200, Alson van der Meulen wrote: (Please CC me in replies as I'm not a subscriber of this list) Update: I tested with an i810-based mainboard (Celeron 1GHz, RTL8139 ethernet, Promise SATAII 150 TX4 controller, 3 SATA disks in RAID 5, FreeBSD 6.0-RC1). It remained stable for two hours. I suspect this is because it has far less bandwith (iostat showed only about 3MB/s to the disks, as opposed to 12MB/s with the i915 mainboard). After I added a dd if=/dev/zero of=foo bs=128k (this increased the bandwidth usage to the disk to about 9MB/s according to iostat), it crashed in about 40 minutes. This suggests that it crashes because of the large amount of I/O. However, it's only about 10MB/s per disk (for three disks), so it doesn't seem to be that exotic to me. Since this is a completely different mainboard, it seems clearly a software issue to me. The built-in ICH6 controller works fine however, so it may be PDC*0518/SII311* specific (which basically means any PCI SATA controller available locally). The SII3112 is a piece of crap that won't work reliably. Order something better (Soren recommends Promise cards). -- Brooks ___ freebsd-stable@freebsd.org mailing list http://lists.freebsd.org/mailman/listinfo/freebsd-stable To unsubscribe, send any mail to [EMAIL PROTECTED]
Re: Problems with PCI SATA controller (bug in ATA driver? both ATAng and ATAmkIII)
(Please CC me in replies as I'm not a subscriber of this list) * Brooks Davis [EMAIL PROTECTED] [2005-10-20 06:15]: On Thu, Oct 20, 2005 at 04:30:48AM +0200, Alson van der Meulen wrote: Since this is a completely different mainboard, it seems clearly a software issue to me. The built-in ICH6 controller works fine however, so it may be PDC*0518/SII311* specific (which basically means any PCI SATA controller available locally). The SII3112 is a piece of crap that won't work reliably. Order something better (Soren recommends Promise cards). That's why I ordered a Promise SATAII 150 TX4. Only when I encountered the problems described in my emails, I got a controller from the only different brand I could find: SII (which costs about 125 euro because it has an Adaptec sticker on it, this thing is definitely going back), to rule out a broken controller. Most of my testing is done with the PDC40518, I occasionally use the SII3112A for comparison purposes (just an extra data point). Because the Promise cards are supposed to be pretty well supported, I'm suprised to encounter these issues in both ATAng and ATAmkIII. BTW: The ICH6 is now being stress tested for 5 hours, and still working fine, so the rest of the system is definitely stable. Alson ___ freebsd-stable@freebsd.org mailing list http://lists.freebsd.org/mailman/listinfo/freebsd-stable To unsubscribe, send any mail to [EMAIL PROTECTED]
Re: Problems with PCI SATA controller (bug in ATA driver? both ATAng and ATAmkIII)
On Wednesday 19 October 2005 21:15, Brooks Davis wrote: On Thu, Oct 20, 2005 at 04:30:48AM +0200, Alson van der Meulen wrote: (Please CC me in replies as I'm not a subscriber of this list) Update: I tested with an i810-based mainboard (Celeron 1GHz, RTL8139 ethernet, Promise SATAII 150 TX4 controller, 3 SATA disks in RAID 5, FreeBSD 6.0-RC1). It remained stable for two hours. I suspect this is because it has far less bandwith (iostat showed only about 3MB/s to the disks, as opposed to 12MB/s with the i915 mainboard). After I added a dd if=/dev/zero of=foo bs=128k (this increased the bandwidth usage to the disk to about 9MB/s according to iostat), it crashed in about 40 minutes. This suggests that it crashes because of the large amount of I/O. However, it's only about 10MB/s per disk (for three disks), so it doesn't seem to be that exotic to me. Since this is a completely different mainboard, it seems clearly a software issue to me. The built-in ICH6 controller works fine however, so it may be PDC*0518/SII311* specific (which basically means any PCI SATA controller available locally). The SII3112 is a piece of crap that won't work reliably. Order something better (Soren recommends Promise cards). -- Brooks Why does the SII3112 work better in FreeBSD 5.4 than in 6.0? I have a Highpoint 1820 on order but it still bugs me having to toss hardware that worked in 5.4 inorder to keep current. -Mike ___ freebsd-stable@freebsd.org mailing list http://lists.freebsd.org/mailman/listinfo/freebsd-stable To unsubscribe, send any mail to [EMAIL PROTECTED]