Bug 128180
Summary: | sata_sil hangs harddrives when loaded (IO-APIC related) | ||||||
---|---|---|---|---|---|---|---|
Product: | [Fedora] Fedora | Reporter: | Hubert Bahr <hab> | ||||
Component: | kernel | Assignee: | Jeff Garzik <jgarzik> | ||||
Status: | CLOSED INSUFFICIENT_DATA | QA Contact: | |||||
Severity: | high | Docs Contact: | |||||
Priority: | medium | ||||||
Version: | 4 | CC: | alexandr.kara, allord_3, davej, etai-ab, jgarzik, k9srb, kevin, kf, mdhouse, nitind, peterm, pmorry, timw, trevor, wtogami | ||||
Target Milestone: | --- | ||||||
Target Release: | --- | ||||||
Hardware: | i686 | ||||||
OS: | Linux | ||||||
Whiteboard: | |||||||
Fixed In Version: | Doc Type: | Bug Fix | |||||
Doc Text: | Story Points: | --- | |||||
Clone Of: | Environment: | ||||||
Last Closed: | 2006-05-05 14:51:01 UTC | Type: | --- | ||||
Regression: | --- | Mount Type: | --- | ||||
Documentation: | --- | CRM: | |||||
Verified Versions: | Category: | --- | |||||
oVirt Team: | --- | RHEL 7.3 requirements from Atomic Host: | |||||
Cloudforms Team: | --- | Target Upstream Version: | |||||
Embargoed: | |||||||
Attachments: |
|
Description
Hubert Bahr
2004-07-19 19:19:27 UTC
Confirm same problem here, identical motherboard. kernel-2.6.6-1.435.2.3 boots fine, kernel-2.6.7-1.494.2.2 fails. It happens with a custom build of kernel 2.6.8-rc4 too. Same problem here on an A7N8X-Deluxe 2.0 mb. System is installed RAID 1 on 2 160GB SATA Seagate drives. The FC2 2.6.5 kernel using the old driver works fine. The 2.6.8 update and the FC3 kernel are unusable. The system locks hard when the driver loads. Starting 2.6.8 with acpi=off I see irq 5: nobody cared! (screaming interrupt?) irq 5: Please try booting with acpi=off and report a bug Stack pointer is garbage, not printing trace handlers: [<4285e2a5>] (ata_interrupt+0x0/0x1ee [libata]) Disabling IRQ #5 Something is still vey broken here. The old cruddy driver works fine. I forgot to add, if anybody (Jeff?) wants to give me a pointer to debugging this, I am at your disposal :-) any improvement with the 2.6.9 based kernel update ? Not with the 2.6.9-1.6_FC@ kernel no. Booting with or without acpi=off yields: irq 5: nobody cared! (screaming interrupt?) irq 5: Please try boot with acpi=off and report a bug handlers: [<42864ab9>] (ata_interrupt+0x0/0x1ee [libata]) Disabling IRQ #5 I have some things I am going to try. The interrupt is shared with the USB2 controller and apparently people have had success turning that off in the BIOS. It will at least be a data point. Otherwise, I'm going to turn the libata tracing on. The old siimage driver still works fine on the same hardware (as does Windows XP), so it's definitely something broken in sata_sil/libata. Same problem here on FC3 using kernel-2.6.9-1.681_FC3, also happened with original FC3 and FC3test3 kernels. The system is an Abit NF7-S v1 nForce2 mb w/ SIL3112A. When a Hitachi SATA drive is attached the boot halts with: irq 11: nobody cared! (screaming interrupt?) irq 11: Please try boot with acpi=off and report a bug handlers: [<42876897>] (ata_interrupt+0x0/0x212 [libata]) Disabling IRQ #11 If the SATA drive is disconnected all loads fine: libata version 1.02 loaded. sata_sil version 0.54 I have found some other bug reports that seems to express the same problem and might provide additional info: 133893, 135479, http://bugzilla.kernel.org/show_bug.cgi?id=3282 I don't know why this bug is classed under FC2. I'm having this problem trying to install FC3 on an ASUS A7N8X deluxe motherboard with 2 SATA drives. I get loading SATA_SIL driver and then an ugly white on black chunk of text that says "Disabling IRQ #5" and then hang city. Any work-around available, such as creating a new install with the new kernel but having the previous and working SATA_SIL driver? Additional information: I tried to load the SiI6512 driver during install (dd hde=noprobe hdg=noprobe). This driver is provided by ASUS for the A7N8X deluxe for systems that need it. It's listed as a driver suitable for RedHat 9 - so I assumed it might be safe to try in the Fedora Core 3 install. The installer asked for the driver disk and after some grinding told me that no suitable drivers were found. So I selected the SiI6512 driver from the list and things seemed to go along nicely - until a dialog came up saying no hard disks were found. That's strange - since I can see them when booting in the SATA raid utility bios information. So it asked if I wanted to load a driver and I said yes, and again chose the SiI6512. Things proceeded until I got to partitioning. Then Anaconda reported that it could (again) detect no drives and rebooted. So my workaround didn't work... or i did something wrong in the process perhaps? OK, there's something very odd going on with the SiL 3112 on the ASUS boards. I currently have 2.6.9-1.6FC2 up and running but only via some fairly grotesque hackery: I patched drivers/scsi/libata-core.c in ata_interrupt to stop check (!(qc->tf.ctl & ATA_NIEN)) to force calling of ata_host_intr even if we thought interrupts were disabled, and turned on ATA_IRQ_TRAP in include/linux/libata.h, and fixing the code in ata_host_intr that is enabled by this define to actually work again. The system now "works", but the very instant the ATA IDENTIFY command is sent, we get an interrupt back, eventually fielded by that ATA_IRQ_TRAP fixup: ata1: irq trap and ata2: irq trap I subsequently see occasional errors of the form: ata1: irq trap ata1: command 0x25 timeout, stat 0x50, host_stat 0x60 or command 0x35. The system is however, stable and usable apart from these hiccups. It looks like the chipset generates interrupts for other reasons that the driver doesn't handle. This is the chipset has a BIOS with the pseudo raid support. Any ideas if that changes things?? This may be redundant, however, when booting kernel-2.6.10-1.741_FC3 with the irqpoll option sata works on an Abit NF7-S mb. If this is fixed in a recent kernel (741), does anyone know how this could be applied to installing with stock FC3 anaconda? FC3 anaconda hangs when loading sata_sil with similar errors to what is listed above (see bug 135479). It isn't fixed in -741, or not for me. I still had to rebuild the kernel disabling sata_sil and re-enabling the siimage support in the ide driver. It's not an ACPI issue. The siimage driver works flawlessly. Something is different with sata_sil, maybe in the register setup, that causes all these extraneous interrupts that it fails to handle. As I don't have any chipset documentation, I'm running out of ideas. I am willing to test code, but for the moment, I am forced to rebuild the kernel each time. It's still happening on the -766 kernel. However, now, irqpoll does actually allow the kernel to boot and run. It seems to me there are two possible things happening: 1) siimage and sata_sil program the chipset differently causing the additional unhandled interrupts in sata_sil, or 2) siimage is also receiving the extra interrupts but is happily fielding them (i.e. handling the interrupt and doing nothing), preventing the problem. Any ideas as to how I might find which scenario is true? Can someone help me? I want to do a fresh FC3 install on a system with no PATA drive -- just a SATA. I don't care what (not too expensive) SATA card I have to buy, I just need to know what chipset will actually boot from FC3 anaconda! I've tried the sata_sil, and that fails. You guys mention using the siimage driver, but that doesn't appear to be in the stock FC3 kernel? If it's not in the stock kernel, how can I boot it from anaconda? I looked up Promise and Adaptec SATA controllers and they too appear to have serious boot issues (mostly the same interrupt problem). Does anyone know of a SATA chipest/vendor out there that I can actually install FC3 with? Any other ideas? Perhaps booting an older FC1 or FC2 first then upgrading? FC4test? I've yet to see any indication anywhere that anyone has *installed* FC3 on a SATA-only machine at all! Loading sata_sil driver ... Disabling IRQ #11 Anyone else get this message and a concurrent hard freeze of the system while installing Fedora Core 3 ? A7N8X deluxe asus mobo Seagate 80 GB Barracuda Sata with NCQ IMPORTANT NOTE!!! I installed same Fedora Core 3 on same system but (( with a NON NCQ Sata barracuda. )) And the install went perfectly. This makes me suspect that Fedora Core 3 is having issues with my brand new Seagate NCQ. Anyone else think it is the NCQ technology on the new SATA's that might be causing me problems ? I also had issues installing Ubuntu and Debian on same drive. Everyone's getting some weird error like that. I get "disabling IRQ 12" or whatever, depending on where the card is installed. Yes, my drives are Seagate w/NCQ. I don't have any non-NCQ drives to try. I wonder if the Seagates have a way to disable the NCQ? Jumper? Bam! John has it nailed. My drives are Seagate 160GB with NCQ (ST3160827AS). Looks like this is the problem combo. The siimage driver still works. But something about NCQ and the SiI controller causes these extra interrupts. Sounds like the driver doesn't handle the NCQ features properly. Does anyone know if it's even supposed to (yet)? I didn't even know the SiI card supported NCQ! Darn, now I regret getting the "better" NCQ drives. Yep. According to the Seagate site, the SiI controller is in the list of those that support NCQ, but, as you mention, the driver doesn't know about it. Sounds like we might be inadvertantly enabling it? Or failing to disable it? Jeff, is there any publicly available programming information for the chipset, or is it only under NDA? Silicon Image kindly made the hardware documentation publicly available: http://gkernel.sourceforge.net/specs/sii/3112A_SiI-DS-0095-B2.pdf.bz2 It's a bit strange that NCQ drives are triggering bad behavior, because the 311x chips do not support NCQ. I wonder if NCQ drives send a FIS (a.k.a. a single SATA data packet) that the 311x controller does not recognize. Argh! I just spent all night fighting to build a custom "linux dd" driver disk with the old (working) siimage driver on it, only to realize at the end that it appears you can't do modular ide drivers at all! If I can't fudge siimage to work in FC3 anaconda, then what else can I try to get a fresh FC3 install to go on this SiI3112 card w/NCQ drives? Does anyone know if a newer SiI driver is in the works somewhere in kernel-land? At least now I know how to build them into a custom driver floppy. I'm getting desperate to get this system loaded. I would use another SATA PCI card, but all the ones I can buy all have a bunch of open bugzillas on them with similar problems. The situation is dire. OK, I dug into the source and the kernel bugzilla and it looks like there's some fudging required for many Seagate disks to work with the sata_sil driver. Look at function sil_drivelist and there are lots of Segate models "blacklisted" to use "SIL_QUIRK_MOD15WRITE". I got to thinking that adding in my Seagate model (ST380817AS) and recompiling might at least let me install FC3. If this works for me, then it may help others if they add their Seagates to the blacklist. I got it compile but for the life of me I can't get the custom "linux dd driver disk" I am trying to create to recognize on the target system. I'm pretty sure I have all the structures correct, but it refuses to load the modules. Since I've checked the dd disk structure so carefully, I'm thinking maybe I'm not compiling valid modules. Is the following sufficient to build a module (without a complete rebuild of the kernel)?: edit Makefile to tweak EXTRAVERSION = -1.667 and -1.667BOOT cd /usr/src/redhat/BUILD/kernel-2.6.9/linux-2.6.9/drivers/scsi make -C /usr/src/redhat/BUILD/kernel-2.6.9/linux-2.6.9 include/linux/version.h make -C /usr/src/redhat/BUILD/kernel-2.6.9/linux-2.6.9 SUBDIRS=$PWD modules I'm trying to build this in the stock FC3 BOOT kernel so I can use the stock boot CD iso. I *finally* got a custom hacked sata_sil module to load from a "linux dd" driver floppy! Man, that has got to be the hardest thing in the world. Anyways, even with my ST380817AS added to the blacklist, the bug still hits. I get the same old interrupts error on the ALT-F1 screen. However, ALT-F4 is interesting. It shows that indeed my hacked module is blacklisting the drives. However, the interrupt error occurs *before* the blacklist takes place?! Maybe the blacklist *does* work but it doesn't have a chance to before it's interrupted? See the screenshot of ALT-F4 I'm attaching next. Is there anyway we can get the sata_sil author in on this bug? I'm completely fresh out of ideas to try. PS: if anyone wants a custom hacked FC3 installer dd floppy, let me know and I can try to provide some instructions. It really is a nightmare to get this right. Created attachment 111970 [details]
screeshot of the driver hang on ALT-F4 console, sorry for the blurriness
It doesn't appear to be the Seagate problem. There is at least one other person reporting the issue with a different drive manufacturer. What seems to be common is that the problematic drives are SATA-II or at least support NCQ. That seems to cause these unexplained interrupts from the Silicon Image controller. I wonder if there's information missing from the documentation. (In reply to comment #24) > I *finally* got a custom hacked sata_sil module to load from a "linux dd" driver > floppy! Man, that has got to be the hardest thing in the world. Anyways, even > with my ST380817AS added to the blacklist, the bug still hits. I get the same > old interrupts error on the ALT-F1 screen. > > However, ALT-F4 is interesting. It shows that indeed my hacked module is > blacklisting the drives. However, the interrupt error occurs *before* the > blacklist takes place?! Maybe the blacklist *does* work but it doesn't have a > chance to before it's interrupted? > > See the screenshot of ALT-F4 I'm attaching next. > > Is there anyway we can get the sata_sil author in on this bug? I'm completely > fresh out of ideas to try. > > PS: if anyone wants a custom hacked FC3 installer dd floppy, let me know and I > can try to provide some instructions. It really is a nightmare to get this right. > Hello, I'm looking for a solution to install FC3 or FC4t1 with the Sata Sil3112A on an Asus A7N8X-E. If you have the driver disk that make the installation working, i'm interresting. Regards. I've got an ABIT IC7-G with the same silicon image controller and 2 WD80GB sata hds, I wanted to install the FC3 and the FC4 test1 and it hang prior to the anaconda. BTW the Mandrake 10.1 and rhe FC2 had worked perfectly, can I extract a driver from there, in order to put it on a floppy disk so I can use the Linux dd option??? I'm under the impression (and my testing further verifies this) that you can't take the old siimage driver and put it on a dd disk because siimage is an IDE driver and must be compiled into the kernel and can't be used as a module. I did get a driver disk going with a hacked sata_sil module by trying to blacklist my drive, but that didn't help. If someone were to give me an updated sata_sil that tried to address this bug, I could create a linux dd driver disk for it for stock FC3 install boot, and I could post the img of that here for everyone to use/test. However, there has yet to be any word from any developer that anyone has even started to look at this problem, even in FC4t1 and the new kernels. From the looks of it, this patch would probably not help my situation (Seagate NCQ interrupts), right? Or is it worth a shot still? can this help anyone (and by that I mean can help someone help me) http://www.linuxmafia.com/faq/Hardware/sata.html#siliconimage I had similar problems when I added an sata_sil card to a FC3 system which was working beautifully before. Strange this is that it will boot just fine when I have the card but no drives, but when I attach a 120GB Western Digital drive I get the 'Disabling IRQ #11' and the system hangs immediately. Attempted noacpi, nousb (as usb was the only thing claiming IRQ 11) -- no joy Then went in and made sure the module was loaded in my initrd image (with a --preload sata_sil on the mkinitrd command line), and the system again hung, but gave a useful pointer in suggesting that I add irqpoll to the kernel boot line. Now it boots, but still says its disabling irq11. System comes up, recognises drive, allows me to format it and then mount it. The warning messages about disabling #11 still occur, but occur earlier in the boot sequence and seem to be handled more gracefully. At any rate, I'm using kernel 2.6.10-1.770_FC3 on a ASUS A7V333, and can supply more info if that is helpful, but wanted to get out the word that it's not just the Segate drives, or the motherboards mentioned so far Thought I'd add the following from dmesg (from the booted system) in the hopes this might shed some light. Note, this is from the system that is RUNNING: ACPI wakeup devices: PCI0 PCI1 UAR1 USB0 USB1 ACPI: (supports S0 S1 S4 S5) Freeing unused kernel memory: 160k freed SCSI subsystem initialized libata version 1.10 loaded. sata_sil version 0.8 ACPI: PCI Interrupt Link [LNKD] enabled at IRQ 11 PCI: setting IRQ 11 as level-triggered ACPI: PCI interrupt 0000:00:10.0[A] -> GSI 11 (level, low) -> IRQ 11 ata1: SATA max UDMA/100 cmd 0xF883C080 ctl 0xF883C08A bmdma 0xF883C000 irq 11 ata2: SATA max UDMA/100 cmd 0xF883C0C0 ctl 0xF883C0CA bmdma 0xF883C008 irq 11 irq 11: nobody cared (try booting with the "irqpoll" option. [<c013e0a0>] __report_bad_irq+0x2b/0x68 [<c013e169>] note_interrupt+0x73/0x96 [<c013d6cc>] __do_IRQ+0x1bd/0x249 [<c0104e04>] do_IRQ+0x5e/0x7a ======================= [<c01035b2>] common_interrupt+0x1a/0x20 [<c0120b50>] __do_softirq+0x2c/0x79 [<c0104edc>] do_softirq+0x38/0x3f ======================= [<c0104e16>] do_IRQ+0x70/0x7a [<c01035b2>] common_interrupt+0x1a/0x20 [<c010103b>] default_idle+0x23/0x29 [<c010108f>] cpu_idle+0x1f/0x34 [<c03a5665>] start_kernel+0x16b/0x16d handlers: [<f8870be7>] (ata_interrupt+0x0/0x210 [libata]) Disabling IRQ #11 ata1: dev 0 cfg 49:2f00 82:346b 83:7f61 84:4003 85:3469 86:3c41 87:4003 88:207f ata1: dev 0 ATA, max UDMA/133, 234441648 sectors: lba48 ata1: dev 0 configured for UDMA/100 scsi0 : sata_sil ata2: no device found (phy stat 00000000) scsi1 : sata_sil Vendor: ATA Model: WDC WD1200JD-22H Rev: 08.0 Type: Direct-Access ANSI SCSI revision: 05 SCSI device sda: 234441648 512-byte hdwr sectors (120034 MB) SCSI device sda: drive cache: write back SCSI device sda: 234441648 512-byte hdwr sectors (120034 MB) SCSI device sda: drive cache: write back sda: unknown partition table Attached scsi disk sda at scsi0, channel 0, id 0, lun 0 device-mapper: 4.3.0-ioctl (2004-09-30) initialised: dm-devel That's interesting because I'm pretty sure the WD drives do NOT have NCQ yet, which would mean it's NOT isolated to NCQ drives? I'm working with 2 WD 80GB attached via Abits Serillel 2 ( IDE to Serial ATA Adapter ) connected to the Silicon Image 3112 and want to install the FC3 to a Fujitsu HD 8GB via IDE and it hangs up, so it's not limited to Seagate HDs. I'm actually want the SATA HD just to be mounted (and the FC3 will be on the IDE drive) but I can't install because it hang on the installation. I even tried to install when the SATA was disabled and it worked fine until I tried booting when the SATA was enanbled and it hangs again. Same problem here of the one listed above (see bug # 135479 : https://bugzilla.redhat.com/bugzilla/show_bug.cgi?id=135479). For this purpose, I'd like to create a custom kernel (vmlinuz) to fit in a custom FC3/4 installation disk. After many tests (compiling the kernel and burning a new boot CD with it), i realized that it might boot correctly and recognize my sata disk (hde,hdg), but it unfortunately stopped asking my the installation media (like it was not able to recognize my cd-rom any longer). So the question is. How can I obtain a .config of the kernel present in the installation disk (vmlinuz) so that I can only modify the entry "CONFIG_BLK_DEV_IDE_SATA=y" and recompiling a new vmlinuz ? May be I need to build a kernel with more parameter build into kernel instead of loading them as modules. Thus I would like to see the original .config of that kernel and starting from there. Tnx a lot for any suggestions . How about this patch? http://www.mail-archive.com/linux-kernel@vger.kernel.org/msg63401.html Thanks for the link! That looks promising, if a bit kludgy. I will whip up a linux dd disk with that when I have time and test it out. Fedora Core 2 has now reached end of life, and no further updates will be provided by Red Hat. The Fedora legacy project will be producing further kernel updates for security problems only. If this bug has not been fixed in the latest Fedora Core 2 update kernel, please try to reproduce it under Fedora Core 3, and reopen if necessary, changing the product version accordingly. Thank you. reopening against FC3 on request from user in private mail. FYI, the patch mentioned above at http://www.mail-archive.com/linux-kernel@vger.kernel.org/msg63401.html does prevent hanging, but causes a large number of "ata trap" messages (seemingly one or two per I/O completion {I get more as I am running RAID-1}). So it's not really a viable fix. Whatever is causing the chipset to want to send interrupts is not a one initialization time thing, but something that apparently needs acknowledging and handling. Kernel kernel-2.6.11-1.14_FC3 still exhibits the same issues. I'm back to running with the pci-ide SATA support, but have not yet got to the point of determining if it is programming the chipset differently and thus not getting the extra interrupts, or if it is simply getting them and silently acknowledging them and getting on with life :-) (In reply to comment #37) > Same problem here of the one listed above (see bug # 135479 : > https://bugzilla.redhat.com/bugzilla/show_bug.cgi?id=135479). > > For this purpose, I'd like to create a custom kernel (vmlinuz) to fit in a > custom FC3/4 installation disk. After many tests (compiling the kernel and > burning a new boot CD with it), i realized that it might boot correctly and > recognize my sata disk (hde,hdg), but it unfortunately stopped asking my the > installation media (like it was not able to recognize my cd-rom any longer). > So the question is. How can I obtain a .config of the kernel present in the > installation disk (vmlinuz) so that I can only modify the entry > "CONFIG_BLK_DEV_IDE_SATA=y" and recompiling a new vmlinuz ? > May be I need to build a kernel with more parameter build into kernel instead of > loading them as modules. Thus I would like to see the original .config of that > kernel and starting from there. > > Tnx a lot for any suggestions . I made a new .rpm kernel with only "CONFIG_BLK_DEV_IDE_SATA=y" changed from the original FC3 kernel (for i686 ). But not yet tested ! Do you want to try it ? You can mail me directly if you want ! my email : deletepmorryatliemans.comdelete I just went to try the patch in comment #38 and I think it can't be used on a linux dd driver floppy since it applies to libata-core.c and not sata_sil.c it doesn't produce a .ko module, just a .o? Does that mean libata-core is not a module but integrated directly in the monolithic kernel? Which means that I can't make a dd driver disk for it to use during an install of stock FC3, can I? There goes that idea. Also, I brought in a new WD 80G drive to try out to see if it has the same problem as my Seagate NCQ's and indeed the WD does NOT work. This is interesting because I'm nearly positive that WD does NOT have NCQ support yet. This means the problem is not isolated to NCQ drives. That makes this a BIG BIG problem as the SIL chipset seems to break on nearly ALL recent SATA drives now. As before, if someone links me to a patch to try, I'll be able to make a linux dd floppy for installers to use to load FC3 and I'll post it here... BUT it would seem I can only do this if the patch applies to sata_sil.c and not the supporting libata code. I'm willing to try your patch but I can't understand what do you mean by .rpm? shouldn't it be a dd? I'm a newbee so you might want to tell me what to do? An RPM kernel will not help you install a new OS, but you can use it as an upgrade if you've already installed an OS, say onto a PATA drive. This doesn't help you (or me) if you have no OS installed and your only hard drive is SATA off of the SIL. For that we need a newer sata_sil driver on a dd disk, which so far no one has provided a patch from which we can create one. I want to install my linux on a PATA drive but it hangs whenever the SATA HDs are connected (even after the installation), can it solve my problem? can I install the linux on a PATA hd (with the SATA HDs disconnected), and install the RPM and then connect the SATA HDs? I still have the same motherboard but I installed FC3 from the distribution boot disk. Upon rebooting same old problem. But recognizing that the installation disk uses the i586 version of the kernel I replaced the i686 version of the kernel installed by default for my Athlon with the i586 version. This boots and runs from my SATA controller. This is the libata version not ide version. on FC2 I had just removed extraneous drivers until the libata version no longer had conflicts. the libata version works except it has conflicts with something else in the default i686 version of the kernel. I want to install my linux on a PATA drive but it hangs whenever the SATA HDs are connected (even after the installation), can it solve my problem? can I install the linux on a PATA hd (with the SATA HDs disconnected), and install the RPM and then connect the SATA HDs? (In reply to comment #49) > I want to install my linux on a PATA drive but it hangs whenever the SATA HDs > are connected (even after the installation), can it solve my problem? > can I install the linux on a PATA hd (with the SATA HDs disconnected), and > install the RPM and then connect the SATA HDs? > You have various options to prepare a kernel that doesn't adversely react to the SATA HDs. Install the rpm and make sure that it becomes the default load. After you verify that this new kernel is loaded on reboot, then reconect the SATA drives. If it still doesn't work we can find another solution. Once you have a compatible kernel you can even intall it on a sata drive if you eventually want to remove the pata drives. OK and how do I do that? If I forgot to mention I'm a rookie at LINUX but a novice in WINDOWS :-) Is there any change in our situation? Did anyone manage to pass the hurdle? FYI, I try another distrib. (Ubuntu a debian's like !). And my Sata sil 3112 was working ! may be a patch from this distrib could help us ? Comment #52 : I have yet to get a working solution that lets me install with a stock FC3 install CD (ie: something that lets me just remake a patched sata_sil dd driver floppy). Comment #53 : In the distro that's working, can you lsmod |grep sata and tell us what drivers are being used? If the distro uses the ide sata stuff then that could be why it's working. If they have it working with sata_sil then that would be news! Also, give us the kernel version. Has anybody else tried the i586 rpm of the kernel. That works for me and provides the sata-sil driver. the i686 kernel still gives me the same problem. I haven't investigated the differences, but I think it is still just a conflict between the default drivers in the stock i686 rpm The stock FC3 install CD/DVD is an i586 2.6.9 kernel and it fails almost immediately with the interrupt error. I think it is a function of what driver is being used, not what version or architecture of kernel you are using. reply to #54 : In the Ubuntu distrib, the module used is sata_sil and libata, so the one we need ! The kernel is 2.6.10-5-386 ! reply to #55: Please could you tell exactly which kernel for i586 you tried ? What's the URL for me to get the source code for the Ubuntu kernel? I'd like to go in and see what patches they might be running. If there was a patch that addressed this issue, I'm sure the driver maintainer would have mentioned it by now? Ok, that's good ! I try with the next kernel parameters with the CD/DVD of the FC3 (just write this): linux irqpoll and it's working, the sata disk are seeing by the kernel ! For the last kernel of FC2 : 2.6.10-1.772_FC2 (i586, didn't test the i686), just modify the grub.conf like this kernel /vmlinuz-2.6.10-1.771_FC2 ro root=LABEL=/ irqpoll instead of : kernel /vmlinuz-2.6.10-1.771_FC2 ro root=LABEL=/ rhgb quiet Now, the sata disk are seeing like sdx drive ! irqpoll works around the problem, but it is not a fix. It has other bad side-effects (can't use my cdrom and dvd on the motherboard when it is in effect). Currently, the only reliable solution is to turn off sata_sil and build the PCI IDE sata driver back in (which still works flwalessly on all these newer drives). I think, you will find all the source here : But I m not sure it's the good way ! http://fr.archive.ubuntu.com/ubuntu/pool/main/l/linux-source-2.6.10 BTW when I installed Mandrake 10.1 a month and a half ago it worked perfectly, it recognized all the sata drives etc. no problem at all. But I needed Fedora form my Homework in the University :-( I see no one has managed to crack the code, ha? Not for FC3 and not for FC4? nothing? I guess they're smarted than us :-) As per comment #57 ... if the bug is not in Ubuntu, yet Ubuntu uses sata_sil and libata, then where does that leave us? Is there a chance the bug is outside of these modules and in something else in the kernel? I don't know enough about the inner-workings of the kernel to guess. It's interesting how comment #57 says it works with Ubuntu using kernel 2.6.10-5-386. Note the -386 part. That's interesting along with comment #55 which says that -586 kernels work. However, in my tests, trying to load 2.6.9-586 which is the kernel on the FC3 install CD, always fails with the interrupt error. Hab, what distro (and version) and kernel version were you testing when you said -586 worked in comment #55? As for poring over the code, I haven't had time, but will try to get Ubuntu source today to take a look. However, I'm not a kernel-source nor a C guy. We really need some bones thrown from the higher-up developers for this. Does anyone know where the heck one can download source code for the kernels used in Ubuntu or Mandrake? Both are reported to not have this bug yet both seem to make it exceeding difficult to find or download sources. That URL in comment #61 appears to contain nothing but binary rpms in .deb format. (In reply to comment #64) > I was using the dvd iso for FC3. It installed correctly, however when I rebooted It failed. It had installed the i686 kernel. Once I found out that it used i586 for the installation, I then installed the i586 version, and that is what it uses now. Some additional info to my original posting. The drives on my sata controller are standard parallel drives talking through a pata to sata bridge. that mounts directly to the individual drives. One drive is on a 1 meter cable. the other on a .5 meter cable. As I mentioned in my earlier trials, I could build a custom kernel that would work with sata-sil. It looks to me like there is another driver that is interfereing with sata-sil and it is not sata-sil that is the only problem. hubert I've experienced the problem with a sil3112A w/ a Seagate 120GB SATA w/ NCQ. Brian Kuschak's hack works fine for me. I'm running Debian's modified 2.6.11. Comment #66: Your case is interesting as you're using non-NCQ parallel drives through adapters. The fact it works with i586 for you but not i686 is also interesting. I think the more common behaviour is for installation to fail while in i586. It is starting to look like you may be correct in that it's not just sata_sil's fault, but I'd say it's still a 50/50 chance. Also interesting how this error is now not just a NCQ issue, as myself and others have reproduced this bug even with non-NCQ drives. Same problem occurring here during install on an Asus A7N8X Deluxe v1.04 using the onboard SiI 3112A with a message that IRQ #11 has been disabled, then system hangs using Fedora Core 4. Using 2 Maxtor drives. One is SATA, the other is PATA with an Abit SATA adapter. Have been able to install on PATA drive by removing SATA adapter and unplugging the other drive from SATA port. IRQPOLL, acpi=off, etc. options to the kernel did me no good when trying to install. Just learned how to rebuild the kernel, and have been trying for 2 days now to get it to work with various configs. May try Brian Kuschak's hack first, then going to try the old IDE drivers with sata_sil removed from the kernel. Any more insights appreciated, or if people want me to try various builds of the kernel, since I'm in a somewhat unique situation (a working install of FC4 where I can rebuild the kernel and try things). Yesterday, I upgrade my kernel of my FC3 with the last one (2.6.11-1-27 ). I try the smp kernel, and it was working without the sata kernel bug. The same kernel without smp, didn't work... Still the sata problem ! Can someone could try the last kernel with SMP and confirm this ? If SMP kernel works, but uniprocessor does not, that may be an interrupt routing / IOAPIC issue separate from sata_sil driver. (note to others, that does not diminish your reported problems) Having the same problem with FC4 installation with ASUS A7N8X-E deluxe motherboard (the same as some of the people from previous posts had). It also prints information about disabling INT 11 and hangs. I "fixed" it before in FC2 by using a self-built kernel, but I don't remember which setting made it work. I think I had to enable APIC, local APIC, IO APIC, ACPI IRQ routing - just enable everything around. The next best option (didn't hang but didn't work too well either) was to disable all of the above... I will try more after installing FC4. Same problem seems to occur with Promise SATA card, see bug #139667. I can now confirm that after upgrading to 2.6.12.2 kernel and changing: # CONFIG_X86_UP_APIC is not set to CONFIG_X86_UP_APIC=y CONFIG_X86_UP_IOAPIC=y CONFIG_X86_LOCAL_APIC=y CONFIG_X86_IO_APIC=y the SATA disc is working OK (without any error). However before a GNOME session starts, there is about 2 minute delay probably related to dbus (pstree shows some dbus processes attached to gnome-session). I don't know whether it is relevant or not. An update has been released for Fedora Core 3 (kernel-2.6.12-1.1372_FC3) which may contain a fix for your problem. Please update to this new kernel, and report whether or not it fixes your problem. If you have updated to Fedora Core 4 since this bug was opened, and the problem still occurs with the latest updates for that release, please change the version field of this bug to 'fc4'. Thank you. Someone else who has the ability to boot from PATA will have to test this. In my situation I'm trying to do it all SATA-only through the sil card and I can't even get an OS installed because neither FC3 nor FC4 install ISO's will work. AFAIK, there's no way for me to use an updated kernel to help me install with anaconda. Forgot to mention: I will be able to test this when a FC5t1 (or similar) comes out where the anaconda install kernel has this patch. (Of course, it might be a while before we see a FC5t1.) You can actually grab the boot.iso from http://download.fedora.redhat.com/pub/fedora/linux/core/development/i386/images/ which gets rebuilt daily using the latest kernel. This way, you won't need to wait for test1 to appear. Hey, thanks! I did not know that. I will download it and try it asap! Just booted from the latest July 15 2005 boot.iso and immediately on loading sata_sil driver it hangs on error "disabling IRQ #5". Exact same problem as with earlier kernels. Drive used for this test was my Seagate NCQ as detailed above. Ding ding ding, we have a winner! Alexandr.Kara has the solution at least for me. I rebuilt the kernel changing CPU type to K7 and enabling the local APIC *AND* the local IO/APIC. Everything works flawlessly. libata has landed on IRQ 18 and all is sweetness and light. I'm in the middle of an FC4 upgrade via yum and am running a rebuilt FC4 kernel currently. Here's /proc/interrupts on the now happy system: kryten:~$ cat /proc/interrupts CPU0 0: 2605259 IO-APIC-edge timer 1: 7404 IO-APIC-edge i8042 7: 5 IO-APIC-edge parport0 8: 1 IO-APIC-edge rtc 9: 0 IO-APIC-level acpi 12: 79412 IO-APIC-edge i8042 15: 67837 IO-APIC-edge ide1 18: 97822 IO-APIC-level libata 20: 0 IO-APIC-level ohci_hcd:usb2 21: 93207 IO-APIC-level ehci_hcd:usb1, eth0 22: 2826 IO-APIC-level ohci1394, NVidia nForce2, ohci_hcd:usb3 NMI: 0 LOC: 2592574 ERR: 0 MIS: 0 The problem we have hear is historically, enabling UP APIC broke some boxes. Things may have improved, so I'll contemplate turning this on for FC5test1 to see how things go. I'm puzzled why legacy irq routing doesn't work correctly though, that may be something else that needs fixing. Same story here. Rebuilt 2.6.12 kernel with Athlon CPU type, and did as Tim Wright suggested with enabling the local APIC and the local IO/APIC. First time that FC3 has been able to boot with a kernel higher than 2.6.9 for me. Also was able to remove the 'irqpoll' line from my kernel parameters without issue. (In reply to comment #78) > You can actually grab the boot.iso from > http://download.fedora.redhat.com/pub/fedora/linux/core/development/i386/images/ > > which gets rebuilt daily using the latest kernel. This way, you won't need to > wait for test1 to appear. > I grabbed todays boot.iso, and it seems to work, including getting sata_sil loaded successfully where FC4 had not. But it won't then let me install from my FC4 discs - apparently it notices the version mismatch between the boot cd and the install media. Is there any way around this? I don't want to install a rawhide system. That is interesting. Because it doesn't have any changes to the apic config options. More likely is that someone fixed something in the IRQ routing code. I'll go over the diffs and see if I can root something out. Making an updated ISO with an updated kernel is aparently a bit of a pain to do. I've got "write howto" on my todo, but it keeps getting preempted by more important things.. I am seeing the same failure as many other people when booting from the fc4 CD's: sata_sil driver loading message, then Interrupt 12 Disabled message, then silence. F3/F4 shows the same error messages as other folks (unclaimed interrupt, some stuff, then interrupt disabled). Ctl-Alt-Del does exit the install (unless booted with NOPROBE; I have a USB keyboard). I've tried booting with NOPROBE, irqpoll, and other workarounds suggested for this bug and #135479, none of which worked. I am using an Intel D845PEBT2 motherboard (no firewire), with a 2.4GHz P4 processor, and both a PATA (WinXT) and SATA (intended to be Linux) HD. The SATA HD is a Maxtor 250GB. I put a small FAT32 partition on the SATA with Maxell's utility under WinXP, and Windows XP mounts it and works fine. The Debis Sarge network install CD hangs with the same error as the FC4 install disk. The MEPIS (small Debis-based) distribution boots and runs fine, recognizing the SATA drive, but has no sata_sil driver, so presumably has sata support compiled into the kernel IDE driver. (If anyone would like further information or have me run a test, please let me know). I plan to unplug my SATA drive, repartion the PATA drive to generate enough space for an FC4 install, install it, build a kernel with Alexandr Kara's fix from #74 reply in it, and plug in the SATA drive and see if the driver loads and runs. (In reply to comment #86) Try to boot with SMP kernel, Sata works with this kernel for me ! (sata sil3112). Regards, > I am seeing the same failure as many other people when booting from the fc4 > CD's: sata_sil driver loading message, then Interrupt 12 Disabled message, > then silence. F3/F4 shows the same error messages as other folks (unclaimed > interrupt, some stuff, then interrupt disabled). Ctl-Alt-Del does exit the > install (unless booted with NOPROBE; I have a USB keyboard). I've tried > booting with NOPROBE, irqpoll, and other workarounds suggested for this bug > and #135479, none of which worked. I am using an Intel D845PEBT2 motherboard > (no firewire), with a 2.4GHz P4 processor, and both a PATA (WinXT) and SATA > (intended to be Linux) HD. The SATA HD is a Maxtor 250GB. I put a small > FAT32 partition on the SATA with Maxell's utility under WinXP, and Windows XP > mounts it and works fine. The Debis Sarge network install CD hangs with the > same error as the FC4 install disk. The MEPIS (small Debis-based) > distribution boots and runs fine, recognizing the SATA drive, but has no > sata_sil driver, so presumably has sata support compiled into the kernel IDE > driver. (If anyone would like further information or have me run a test, > please let me know). > > I plan to unplug my SATA drive, repartion the PATA drive to generate enough > space for an FC4 install, install it, build a kernel with Alexandr Kara's fix > from #74 reply in it, and plug in the SATA drive and see if the driver loads > and runs. (In reply to comment #86) Try to boot with SMP kernel, Sata works with this kernel for me ! (sata sil3112). Regards, > I am seeing the same failure as many other people when booting from the fc4 > CD's: sata_sil driver loading message, then Interrupt 12 Disabled message, > then silence. F3/F4 shows the same error messages as other folks (unclaimed > interrupt, some stuff, then interrupt disabled). Ctl-Alt-Del does exit the > install (unless booted with NOPROBE; I have a USB keyboard). I've tried > booting with NOPROBE, irqpoll, and other workarounds suggested for this bug > and #135479, none of which worked. I am using an Intel D845PEBT2 motherboard > (no firewire), with a 2.4GHz P4 processor, and both a PATA (WinXT) and SATA > (intended to be Linux) HD. The SATA HD is a Maxtor 250GB. I put a small > FAT32 partition on the SATA with Maxell's utility under WinXP, and Windows XP > mounts it and works fine. The Debis Sarge network install CD hangs with the > same error as the FC4 install disk. The MEPIS (small Debis-based) > distribution boots and runs fine, recognizing the SATA drive, but has no > sata_sil driver, so presumably has sata support compiled into the kernel IDE > driver. (If anyone would like further information or have me run a test, > please let me know). > > I plan to unplug my SATA drive, repartion the PATA drive to generate enough > space for an FC4 install, install it, build a kernel with Alexandr Kara's fix > from #74 reply in it, and plug in the SATA drive and see if the driver loads > and runs. k9srb: in comment #86 you didn't say what SATA card you were using. Please tell us the card, or if onboard please confirm the chipset. I didn't think Intel-brand (as distinct from Intel-chipset!) boards used onboard SIL. The motherboard is an Intel D845PEBT2, http://www.intel.com/design/motherbd/bt2/, which uses the 845PE chipset, and has an Sil 3112A chip on the motherboard. My motherboard does not have the Firewire option. Just came across some interesting info. There are lots of reports of 3112-based cards and onboards having similar problems in Windows! I assumed the card would work great in Windows. If the card doesn't work in Windows or Linux then how the heck can they be selling this card? Argh! See: http://www.sysopt.com/forum/printthread.php?threadid=141781 http://www.pcreview.co.uk/forums/thread-1943729.php http://www.pcreview.co.uk/forums/thread-1514517.php http://www.computerbanter.com/showthread.php?t=65700 The posts indicate it *may* be card brand specific, but there certainly are lots of brands that are known to have issues and no definitive reports of brands working. Also, SiI has put updated drivers + bios on their site: http://www.siliconimage.com/support/downloadresults.aspx?pid=63&bios=0&drivers=0&sataraid=0& I installed the updated BIOS and it doesn't seem to help anything. (In reply to comment #92) > Just came across some interesting info. There are lots of reports of 3112-based > cards and onboards having similar problems in Windows! I assumed the card would > work great in Windows. If the card doesn't work in Windows or Linux then how > the heck can they be selling this card? Argh! > > See: > http://www.sysopt.com/forum/printthread.php?threadid=141781 > http://www.pcreview.co.uk/forums/thread-1943729.php > http://www.pcreview.co.uk/forums/thread-1514517.php > http://www.computerbanter.com/showthread.php?t=65700 > > The posts indicate it *may* be card brand specific, but there certainly are lots > of brands that are known to have issues and no definitive reports of brands working. > > Also, SiI has put updated drivers + bios on their site: > http://www.siliconimage.com/support/downloadresults.aspx?pid=63&bios=0&drivers=0&sataraid=0& > > I installed the updated BIOS and it doesn't seem to help anything. > The Card SIL3112 Work on Redhat Linux with SMP Kernel.... But not without SMP ! I tink, it'll be good to look in that way, make a diff between SMP kernel and the monoprocessor kernel (I'm not a developer, so I can't do it !). For information, The Ubuntu distribution work perfectly with sata SIL3112. This is a mass-update to all currently open Fedora Core 3 kernel bugs. Fedora Core 3 support has transitioned to the Fedora Legacy project. Due to the limited resources of this project, typically only updates for new security issues are released. As this bug isn't security related, it has been migrated to a Fedora Core 4 bug. Please upgrade to this newer release, and test if this bug is still present there. This bug has been placed in NEEDINFO_REPORTER state. Due to the large volume of inactive bugs in bugzilla, if this bug is still in this state in two weeks time, it will be closed. Should this bug still be relevant after this period, the reporter can reopen the bug at any time. Any other users on the Cc: list of this bug can request that the bug be reopened by adding a comment to the bug. Thank you. I can confirm the bug is present in the lastest FC4 kernel. I don't have access to the machine right now, but I updated the kernel from updates repository last week and the computer froze again after printing "Disabling IRQ 11". I can post more info next week. It can be fixed by enabling both local APIC and local IO/APIC. Probably also works with SMP kernels as it turns APIC on (if I recall correctly). This is a mass-update to all currently open kernel bugs. A new kernel update has been released (Version: 2.6.15-1.1830_FC4) based upon a new upstream kernel release. Please retest against this new kernel, as a large number of patches go into each upstream release, possibly including changes that may address this problem. This bug has been placed in NEEDINFO_REPORTER state. Due to the large volume of inactive bugs in bugzilla, if this bug is still in this state in two weeks time, it will be closed. Should this bug still be relevant after this period, the reporter can reopen the bug at any time. Any other users on the Cc: list of this bug can request that the bug be reopened by adding a comment to the bug. If this bug is a problem preventing you from installing the release this version is filed against, please see bug 169613. Thank you. Closing per last comment. |