I sync with 7-STABLE almost every day. I build everything on a Toshiba U205 Satellite. Things are fine for months on end. I did this on June 8th. Everything was fine. I did this on June 10th. The machine no longer booted. The entire root partition got clobbered. I reinstalled a snapshot of 7-STABLE from May 28th that I had put on a DVD. Everything was once again fine. I then sync'd again this morning June 11th with 7-STABLE, did a full build, and reproducibly, BOOM - the entire root partition got clobbered. Gone. Again. After the reboot it just comes up with: BTX loader 1.00 BTX version is 1.02 Consoles: internal video/keyboard BIOS drive A: is disk0 BIOS drive C: is disk1 and stops. Nothing else is printed. There is no choice of how to boot. There is no files for me to send, no log files to inspect, no remnants. I unfortunately did not see where in the build this happened. My build does the canonical steps exactly as outlined in / usr/src/Makefile and then does a reboot. It builds userland and the kernel. When I inspect it from the bootable DVD the partition that my root filesystem was in has no association with it having the root any more. The partition is listed, but it looks like it had been freshly partitioned. Something is very, very wrong. Ideas? Dan Allen
On 6/11/09, Dan Allen <danallen46@airwired.net> wrote:> I sync with 7-STABLE almost every day. I build everything on a > Toshiba U205 Satellite. Things are fine for months on end. > > I did this on June 8th. Everything was fine. > > I did this on June 10th. The machine no longer booted. The entire > root partition got clobbered. I reinstalled a snapshot of 7-STABLE > from May 28th that I had put on a DVD. Everything was once again fine. > > I then sync'd again this morning June 11th with 7-STABLE, did a full > build, and reproducibly, BOOM - the entire root partition got > clobbered. Gone. Again. After the reboot it just comes up with: > > BTX loader 1.00 BTX version is 1.02 > Consoles: internal video/keyboard > BIOS drive A: is disk0 > BIOS drive C: is disk1 > > and stops. Nothing else is printed. There is no choice of how to > boot. There is no files for me to send, no log files to inspect, no > remnants. I unfortunately did not see where in the build this > happened. My build does the canonical steps exactly as outlined in / > usr/src/Makefile and then does a reboot. It builds userland and the > kernel. > > When I inspect it from the bootable DVD the partition that my root > filesystem was in has no association with it having the root any > more. The partition is listed, but it looks like it had been freshly > partitioned. > > Something is very, very wrong. > > Ideas?Are you using ZFS on root partition? -- Paul
On 11 Jun 2009, at 3:40 PM, Paul B. Mahol wrote:> Are you using ZFS on root partition?No. The disk is the default (UFS2 I believe). So I just reinstalled BSD again and this time I did not reinitialize the file system and after a brief disk integrity check it reinstalled and files I had added were still there! So apparently the file system did not get munged as much as the main disk partition map got nailed. So, what has changed recently that could mess with the disk partition map? I have Windows on the first disk partition and it has not been harmed with these problems. Dan
In trying to figure this out, I rebuilt a GENERIC kernel after sync'ing to today's RELENG_7 sources. I then installed it, held my breath, and rebooted. It works! So I am now looking at userland (make buildworld && make installworld) to see if it is the culprit. Another possibility is my custom kernel build. Stay tuned... Dan
Okay. I did a make buildkernel && installkernel and rebooted, no problems. I then did a make buildworld and rebooted, no problems. I then did a make installworld which completed normally, rebooted, and BINGO - my disk partition table has been zapped. The problem appears to be something that runs during this 'make installworld'! There are no problems with the build itself that I can tell, but some program is munging the disk partition table. In a zany sort of way this is progress. Of course now I have to reinstall the OS, again... Dan Allen
On 13 Jun 2009, at 2:41 PM, Dan Allen wrote:> > On 13 Jun 2009, at 12:50 PM, Paul B. Mahol wrote: > >> I doubt it is loader fault, from your description it appears that >> loader is never started. >> >> Could you try to remove -DLOADER_ZFS_SUPPORT from Makefile? >> >>> /usr/src/sys/boot/i386/loader/Makefile > > BINGO! LOADER_ZFS_SUPPORT is the culprit. > > I rebuilt the good world, and brought in the loader changes which > have caused previously caused death to my drive, then deleted the > ZFS_SUPPORT path in /usr/src/sys/boot/i386/loader/Makefile as you > recommended, and rebuilt things and everything works fine.My wording may not have been clear: I simply replaced the errant June 10th version of /usr/src/sys/boot/i386/loader/Makefile with the June 8th version of the same file which simply defaults to no LOADER_ZFS_SUPPORT. Dan