I've got a machine here that I'm trying to install FreeBSD 5.4 on. It's identical (except for the hard drive) to another machine that I ran the 4.x series on for some time with no issues. I've got several other Pentium 3-based machines running 5.4-RELEASE-p3 with a GENERIC kernel, and I have a 5.3 installer disk, so my strategy was to do a minimal install of 5.3, then NFS mount /usr/src and /usr/obj from my organizational build server and upgrade to 5.4 from there. The 5.3 install went smoothly, make installkernel went smoothly, but when I tried to reboot I got an error saying that it couldn't find the root filesystem. Further investigation found that it wasn't able to find the ATA HDD (master on ata0) at all, but could find the ATAPI CDROM drive (master on ata0). I verified that the drive had not gone bad by again booting off the 5.3 installer disk, which found ad0 with no difficulty. I then downloaded the 5.4 mini-CD, which was unable to find ad0. Finally, I swapped in another hard drive with identical results. Neither booting with ACPI disabled (didn't bother to check whether this machine even has it) nor booting in safe mode helped at all. I did happen to notice that when the 5.3 CD boots, there is a delay of several seconds when probing ata0. Could this be a timing issue with regard to the controller? The machine is a Gateway ATXSTF FED Pro M1000 with a 1GHz Pentium III and 256MB of RAM. It has an onboard Intel ICH2 UDMA100 controller. The first HDD I tried was a Western Digital Caviar 32500 2559.8MB drive, and the second was a Caviar 24300 4311.9MB drive. If necessary, I'm willing to build custom kernels for testing purposes or to add debugging statements etc, if anyone can give me pointers in the right direction. Thanks in advance. -- Tim Howe Software Engineer Celebrity Resorts, Inc.
On Fri, 2005-Sep-23 22:52:09 -0400, Tim Howe wrote:>I've got several other Pentium 3-based machines running 5.4-RELEASE-p3 >with a GENERIC kernel, and I have a 5.3 installer disk, so my strategy >was to do a minimal install of 5.3, then NFS mount /usr/src and /usr/obj >from my organizational build server and upgrade to 5.4 from there.Did you reboot after the 5.3 install or do the upgrade whilst booted from that install disk?>root filesystem. Further investigation found that it wasn't able to >find the ATA HDD (master on ata0) at all, but could find the ATAPI CDROM >drive (master on ata0).You shouldn't have two masters on ata0. I hope that's a typo. How far through the boot process do you get? I gather the loader runs successfully and loads the kernel but the kernel can't find ad0. Does it find the controller (ata0 or whatever)? -- Peter Jeremy
I'm no expert on this, but I'd let that machine have it's own build environment instead of NFS importing it from a server. It _does_ require more disk space, but in the long run, it might be for the better. If you _must_ export from server, then export only /usr/src and not /usr/obj Owe J?rgensen On Fri, 23 Sep 2005 22:52:09 -0400, Tim Howe <tim.howe@celebrityresorts.com> wrote:> I've got a machine here that I'm trying to install FreeBSD 5.4 on. It's > identical (except for the hard drive) to another machine that I ran the > 4.x series on for some time with no issues. > > I've got several other Pentium 3-based machines running 5.4-RELEASE-p3 > with a GENERIC kernel, and I have a 5.3 installer disk, so my strategy > was to do a minimal install of 5.3, then NFS mount /usr/src and /usr/obj > from my organizational build server and upgrade to 5.4 from there. > > The 5.3 install went smoothly, make installkernel went smoothly, but > when I tried to reboot I got an error saying that it couldn't find the > root filesystem. Further investigation found that it wasn't able to > find the ATA HDD (master on ata0) at all, but could find the ATAPI CDROM > drive (master on ata0). > > I verified that the drive had not gone bad by again booting off the 5.3 > installer disk, which found ad0 with no difficulty. I then downloaded > the 5.4 mini-CD, which was unable to find ad0. Finally, I swapped in > another hard drive with identical results. > > Neither booting with ACPI disabled (didn't bother to check whether this > machine even has it) nor booting in safe mode helped at all. > > I did happen to notice that when the 5.3 CD boots, there is a delay of > several seconds when probing ata0. Could this be a timing issue with > regard to the controller? > > The machine is a Gateway ATXSTF FED Pro M1000 with a 1GHz Pentium III > and 256MB of RAM. It has an onboard Intel ICH2 UDMA100 controller. The > first HDD I tried was a Western Digital Caviar 32500 2559.8MB drive, and > the second was a Caviar 24300 4311.9MB drive. > > If necessary, I'm willing to build custom kernels for testing purposes > or to add debugging statements etc, if anyone can give me pointers in > the right direction. > > Thanks in advance. >-- mvh, Owe Andr? J?rgensen
Okay, Tim, I just reported the same bug in a previous post. So did Julian C. Dunn. I think this is an issue with some recent code changes in CVS... See these threads: critical BOOT failure updating to latest 5-Stable (5.4) 9/23/2005 12:07 AM ATA lockup with 5.4-STABLE 9/21/2005 10:19 PM I have privately emailed Mr. Dunn, and his problem seems a little different perhaps, but I see some similarities. His system is an IBM (so is mine) and his uses the Intel PIIX4 ata chipset. Mine uses the PIIX3. Mine is by far the oldest, and yours is the newest system. Of note maybe is that I was also using an older drive as you are, in my case an 8 or 9 GB Quantum Fireball. Anyway, I count three systems that are normally boting with 5.3 or 5.4, and then the upgrade to the latest 5-Stable breaks the boot. Perhaps you could test one of those identical machines with a simple hard disk swap? Just put the bad booting drive in a working system and see if the error is still there? Has anyone checked the CVS repository for a possible change to the ATA detection or boot code? Can you post your dmesg for 5.3 perhaps? BTW, you can boot that machine to the old kernel using the procedure unload load /boot/kernel.old/kernel boot (or boot -v) Do all of that with option 6 on the Boot Loader menu. Thanks, Billy Tim Howe wrote:> I've got a machine here that I'm trying to install FreeBSD 5.4 on. It's > identical (except for the hard drive) to another machine that I ran the > 4.x series on for some time with no issues. > > I've got several other Pentium 3-based machines running 5.4-RELEASE-p3 > with a GENERIC kernel, and I have a 5.3 installer disk, so my strategy > was to do a minimal install of 5.3, then NFS mount /usr/src and /usr/obj > from my organizational build server and upgrade to 5.4 from there. > > The 5.3 install went smoothly, make installkernel went smoothly, but > when I tried to reboot I got an error saying that it couldn't find the > root filesystem. Further investigation found that it wasn't able to > find the ATA HDD (master on ata0) at all, but could find the ATAPI CDROM > drive (master on ata0). > > I verified that the drive had not gone bad by again booting off the 5.3 > installer disk, which found ad0 with no difficulty. I then downloaded > the 5.4 mini-CD, which was unable to find ad0. Finally, I swapped in > another hard drive with identical results. > > Neither booting with ACPI disabled (didn't bother to check whether this > machine even has it) nor booting in safe mode helped at all. > > I did happen to notice that when the 5.3 CD boots, there is a delay of > several seconds when probing ata0. Could this be a timing issue with > regard to the controller? > > The machine is a Gateway ATXSTF FED Pro M1000 with a 1GHz Pentium III > and 256MB of RAM. It has an onboard Intel ICH2 UDMA100 controller. The > first HDD I tried was a Western Digital Caviar 32500 2559.8MB drive, and > the second was a Caviar 24300 4311.9MB drive. > > If necessary, I'm willing to build custom kernels for testing purposes > or to add debugging statements etc, if anyone can give me pointers in > the right direction.
On Sat, 24 Sep 2005, Billy Newsom wrote:> Okay, Tim, I just reported the same bug in a previous post. So did Julian C. > Dunn. I think this is an issue with some recent code changes in CVS... > > See these threads: > > critical BOOT failure updating to latest 5-Stable (5.4) > 9/23/2005 12:07 AM > > ATA lockup with 5.4-STABLE > 9/21/2005 10:19 PM >I have a similar problem (previously posted to this mailing list) with a Toshiba 8100 which locks up when atapci0 is called during the boot using recent kernels. I can boot off an old 5.4-RELEASE kernel (May 2005) and can boot off the more recent kernels if I copy the old version of acpi.ko over the new version, which leads me to believe that it is changes in the ACPI code which are the source of the problem. I've posted to the ACPI mailing list (I'm not aware of any responses). Mike Squires