On all of our post 2.4.9 era kernels, running parted on disks which have mounted partitions even just for reading results in *many* messages of invalidate: busy buffer. In some cases, you just get the spewage to syslog for 5-10 seconds and nothing more (no biggie other than confusing people). In others, though, I've had my laptop sit hard locked for a minute or more giving no response while spewing several thousand instances of the message.
2 parts in this bug: 1) partitioning while live fs mounted: not supported 2) the busy buffers in case of no longer live.... do you know if it helps if you "sync" inbetween unmounting and the invalidate ?
But I'm not really repartitioning. Parted is just doing the reading of the disks to get this to happen (run parted /dev/hda; print). It's the parted equivalent of fdisk -l, and if that's not supported, it's not going to be exactly nice :( I'll take a look at seeing what interaction changes I can get by doing some syncing as well as fsck'ing my disk to see if that helps since it seems a lot worse on my laptop than my desktop.
beta2, Linux pc17 2.4.18-0.1 #1 Wed Feb 27 07:42:03 EST 2002 i686 unknown http://www.beta.redhat.com/betahardware.cgi?action=form&id=6 I get many of these, always after something load parport automatically (also note "try to free nonexistent resource": Mar 13 17:08:40 pc17 gconfd (root-1247): Unable to open saved state file '/root/ .gconfd/saved_state': No such file or directory Mar 13 17:08:53 pc17 xinetd[1260]: warning: can't get client address: Transport endpoint is not connected Mar 14 15:13:27 pc17 kernel: parport0: PC-style at 0x278 (0x678) [PCSPP,TRISTATE ,EPP] Mar 14 15:13:27 pc17 kernel: parport0: irq 5 detected Mar 14 15:13:30 pc17 kernel: invalidate: busy buffer Mar 14 15:13:39 pc17 last message repeated 1008 times Mar 14 15:13:39 pc17 kernel: invalidate: dirty buffer Mar 14 15:13:39 pc17 kernel: invalidate: busy buffer Mar 14 15:13:39 pc17 kernel: invalidate: dirty buffer Mar 14 15:13:39 pc17 kernel: invalidate: busy buffer Mar 14 15:13:39 pc17 kernel: invalidate: dirty buffer Mar 14 15:13:39 pc17 kernel: invalidate: busy buffer Mar 14 15:13:46 pc17 last message repeated 1213 times .... Mar 15 17:07:53 pc17 kdm[994]: Rereading access file /etc/X11/xdm/Xaccess Mar 15 17:08:19 pc17 kde(pam_unix)[1430]: session opened for user root by (uid=0 ) Mar 15 17:08:20 pc17 xinetd[1569]: warning: can't get client address: Transport endpoint is not connected Mar 15 17:08:25 pc17 modprobe: modprobe: Can't locate module sound-slot-0 Mar 15 17:08:25 pc17 modprobe: modprobe: Can't locate module sound-service-0-3 Mar 15 17:08:25 pc17 modprobe: modprobe: Can't locate module sound-slot-0 Mar 15 17:08:25 pc17 modprobe: modprobe: Can't locate module sound-service-0-0 Mar 15 17:08:25 pc17 modprobe: modprobe: Can't locate module sound-slot-0 Mar 15 17:08:25 pc17 modprobe: modprobe: Can't locate module sound-service-0-0 Mar 15 17:08:25 pc17 modprobe: modprobe: Can't locate module sound-slot-1 Mar 15 17:08:25 pc17 modprobe: modprobe: Can't locate module sound-service-1-0 Mar 15 17:08:26 pc17 modprobe: modprobe: Can't locate module sound-slot-1 Mar 15 17:08:26 pc17 modprobe: modprobe: Can't locate module sound-service-1-0 Mar 15 17:09:34 pc17 kernel: parport0: PC-style at 0x278 (0x678) [PCSPP,TRISTATE ,EPP] Mar 15 17:09:34 pc17 kernel: parport0: irq 5 detected Mar 15 17:09:35 pc17 kernel: invalidate: busy buffer Mar 15 17:09:37 pc17 last message repeated 15982 times Mar 15 17:12:40 pc17 kernel: hdc: unknown partition table Mar 15 17:12:40 pc17 modprobe: modprobe: Can't locate module block-major-33 Mar 15 17:12:40 pc17 modprobe: modprobe: Can't locate module block-major-33 Mar 15 17:12:40 pc17 modprobe: modprobe: Can't locate module block-major-34 Mar 15 17:12:40 pc17 modprobe: modprobe: Can't locate module block-major-34 Mar 15 17:12:41 pc17 modprobe: modprobe: Can't locate module block-major-56 Mar 15 17:12:41 pc17 modprobe: modprobe: Can't locate module block-major-56 Mar 15 17:12:41 pc17 modprobe: modprobe: Can't locate module block-major-57 Mar 15 17:12:41 pc17 modprobe: modprobe: Can't locate module block-major-57 Mar 15 17:12:41 pc17 modprobe: modprobe: Can't locate module block-major-88 Mar 15 17:12:41 pc17 modprobe: modprobe: Can't locate module block-major-88 Mar 15 17:12:41 pc17 modprobe: modprobe: Can't locate module block-major-89 Mar 15 17:12:41 pc17 modprobe: modprobe: Can't locate module block-major-89 Mar 15 17:12:41 pc17 modprobe: modprobe: Can't locate module block-major-90 Mar 15 17:12:41 pc17 modprobe: modprobe: Can't locate module block-major-90 Mar 15 17:12:41 pc17 modprobe: modprobe: Can't locate module block-major-91 Mar 15 17:12:41 pc17 modprobe: modprobe: Can't locate module block-major-91 Mar 15 17:12:41 pc17 modprobe: modprobe: Can't locate module block-major-80 Mar 15 17:12:42 pc17 modprobe: modprobe: Can't locate module block-major-45 Mar 15 17:12:42 pc17 modprobe: modprobe: Can't locate module block-major-72 Mar 15 17:12:42 pc17 modprobe: modprobe: Can't locate module block-major-48 Mar 15 17:12:42 pc17 kernel: XD: Loaded as a module. Mar 15 17:12:42 pc17 kernel: Trying to free nonexistent resource <00000320-00000 323> Mar 15 17:12:42 pc17 insmod: /lib/modules/2.4.18-0.1/kernel/drivers/block/xd.o: init_module: Operation not permitted Mar 15 17:12:42 pc17 insmod: Hint: insmod errors can be caused by incorrect modu le parameters, including invalid IO or IRQ parameters Mar 15 17:12:42 pc17 insmod: /lib/modules/2.4.18-0.1/kernel/drivers/block/xd.o: insmod block-major-13 failed Mar 15 17:12:43 pc17 kernel: hdc: unknown partition table Mar 15 17:12:43 pc17 kernel: ide-scsi: hdd: unsupported comman
I'm using 2.4.18-0.13 BETA kernel and get tons of "invalidate: busy buffer" errors when shutting down, not during startup. It's very repeatable. Happens every time I shut down. This box has 1G of memory and I'm not doing any "work" on the box, so it's possible that no dirty pages are being written to disk until shutdown. I have software RAID configured with two drives hda, hdc. Looking at the code in fs/buffer.c, we could get a lot more information that might be helpful. kdev, bh, bh->count. I can recompile and print some additional info. What should I collect? Or is this a different problem?
Should be fixed in our internal tree now: look for it in a subsequent rawhide kernel build.