On 09:11, Phillip Susi wrote: > Umm... it seems to me that if you open your disk O_RDWR, for instance, > to update your boot loader, while you have a partition on the disk > mounted, and the kernel destroys an actively mounted partition device, > then the kernel is very, very broken...
Strangely, the GRUB install step doesn't seem to trigger this... I can't remember if I was using GPT. A mere dd if=/dev/null bs=512 of=/dev/ada triggers it. (No actual blocks written). But to open that O_RDWR when there are mounted filesystems, requires kern.geom.debugflags to enable "foot shooting" mode first. I've only confirmed that it affects ZFS so far, with kfreebsd 10.1, but I can see other partitions being detached / re-attached on MEDIACHANGE. > As for the performance implications, yes, parted should not be using > O_RDWR when it do doesn't need it and does take care not to on linux. OK I'll look into this, as we probably should be doing the same. Regards, -- Steven Chamberlain ste...@pyro.eu.org -- To UNSUBSCRIBE, email to debian-bugs-dist-requ...@lists.debian.org with a subject of "unsubscribe". Trouble? Contact listmas...@lists.debian.org
