On Tue, May 12, 2009 at 7:00 PM, Andy Isaacson <adi@xxxxxxxxxxxxx> wrote: > On Tue, May 12, 2009 at 05:23:57PM -0600, Robert Hancock wrote: >> >Looking at hald/linux/addons/addon-storage.c it does appear to >> >open(O_EXCL) as well, so perhaps the O_EXCL multiple-open logic broke >> >recently? >> > >> >Complete dmesg, kconfig, and so on at >> > >> >http://web.hexapodia.org/~adi/sysinfo/1242164411_cvpe4300_2.6.30-rc5-dmar-00096-ga4d7749/ >> >> I'm not sure how this is supposed to work. The only place I can see any >> O_EXCL handling is in the sg driver. But if you access the sr device >> directly, it doesn't seem like that would have any effect. >> >> CCing linux-scsi. Could be there's a bad user-space assumption here, or >> something. > > Perhaps, I haven't tracked down exactly what the difference in behavior > between working 2.6.29 and broken 2.6.30-rc2 is. It seems like the disk > spins up more often (in response to hal background probes presumably) > with 30-rc2 than it did before, and running "udevadm monitor" I do see > the following repeated "change" events happening, even on an idle > system: > > KERNEL[1242175547.793029] change /devices/pci0000:00/0000:00:1f.2/host1/target1:0:0/1:0:0:0 (scsi) > KERNEL[1242175547.794669] change /devices/pci0000:00/0000:00:1f.2/host1/target1:0:0/1:0:0:0/block/sr0 (block) > UDEV [1242175547.794912] change /devices/pci0000:00/0000:00:1f.2/host1/target1:0:0/1:0:0:0 (scsi) > UDEV [1242175547.888539] change /devices/pci0000:00/0000:00:1f.2/host1/target1:0:0/1:0:0:0/block/sr0 (block) > KERNEL[1242175673.309129] change /devices/pci0000:00/0000:00:1f.2/host1/target1:0:0/1:0:0:0 (scsi) > KERNEL[1242175673.310775] change /devices/pci0000:00/0000:00:1f.2/host1/target1:0:0/1:0:0:0/block/sr0 (block) > UDEV [1242175673.311013] change /devices/pci0000:00/0000:00:1f.2/host1/target1:0:0/1:0:0:0 (scsi) > UDEV [1242175674.658718] change /devices/pci0000:00/0000:00:1f.2/host1/target1:0:0/1:0:0:0/block/sr0 (block) > KERNEL[1242175690.227604] change /devices/pci0000:00/0000:00:1f.2/host1/target1:0:0/1:0:0:0 (scsi) > KERNEL[1242175690.229730] change /devices/pci0000:00/0000:00:1f.2/host1/target1:0:0/1:0:0:0/block/sr0 (block) > UDEV [1242175690.229960] change /devices/pci0000:00/0000:00:1f.2/host1/target1:0:0/1:0:0:0 (scsi) > UDEV [1242175714.060440] change /devices/pci0000:00/0000:00:1f.2/host1/target1:0:0/1:0:0:0/block/sr0 (block) > KERNEL[1242175727.909819] change /devices/pci0000:00/0000:00:1f.2/host1/target1:0:0/1:0:0:0 (scsi) > KERNEL[1242175727.911356] change /devices/pci0000:00/0000:00:1f.2/host1/target1:0:0/1:0:0:0/block/sr0 (block) > UDEV [1242175727.911690] change /devices/pci0000:00/0000:00:1f.2/host1/target1:0:0/1:0:0:0 (scsi) > UDEV [1242175727.987747] change /devices/pci0000:00/0000:00:1f.2/host1/target1:0:0/1:0:0:0/block/sr0 (block) > KERNEL[1242175731.591853] change /devices/pci0000:00/0000:00:1f.2/host1/target1:0:0/1:0:0:0 (scsi) > KERNEL[1242175731.593541] change /devices/pci0000:00/0000:00:1f.2/host1/target1:0:0/1:0:0:0/block/sr0 (block) > > I've verified that O_EXCL works at least with a trivial test program, so > it's not simply that EXCL broke. > > Also I've verified with strace that hald-addon-storage and wodim are > both using O_EXCL, and that hald-addon-storage gets EBUSY while wodim is > writing the disk. (The straced run didn't fail, unfortunately.) Those events seem rather suspicious. Hmm: [ 2.770959] ata2.00: ATAPI: TSSTcorp DVD+/-RW TS-U633A, D200, max UDMA/100, ATAPI AN Did libata report it was using ATAPI AN in 2.6.29? You could try bisecting to see if you can find which patch started triggering those spurious events.. -- To unsubscribe from this list: send the line "unsubscribe linux-ide" in the body of a message to majordomo@xxxxxxxxxxxxxxx More majordomo info at http://vger.kernel.org/majordomo-info.html