Ah, no, forget about it, sorry. It's a bug in libparted and that's why I observed the wrong argument in strace which made it into my code as well. The length should be of course 512 and not 1 due to the different units. Patch would be adding the following line to libparted/arch/linux.c:2551: +++ linux_part.length *= disk->dev->sector_size; (function _blkpg_resize_partition) Well, thanks for reading it anyway. I will submit a patch. Am 09.04.19 um 18:41 schrieb Kai Lüke: > Hello, > > I write to you because I believe that the BLKPG ioctl misbehaves when an > existing extended MBR partition is changed and a logical partition added > in it. > Should I report like this or should I sign up for bugzilla.kernel.org? > > The /sys/block/sdx/sdxY/size entry is set to 0 and even opening the > extended partition fails. > This leads to e.g. udevd failing to have access. > https://github.com/storaged-project/udisks/issues/425 > > To fix this situation from userspace, a BLKRRPART ioctl is needed. > > To reproduce it, I attached a small program, which you can run on this > prepared image or on any drive with a single extended partition on it. > > $ wget http://kailueke.spline.de/test.img > $ udisksctl loop-setup -f test.img > > $ sudo ./sim loop0 > new size in /sys/block/loop0/loop0p1/size: 0 > issuing BLKRRPART > new size in /sys/block/loop0/loop0p1/size: 2 > > > Do you think that this should be fixed in the kernel? > Regards, > Kai Lüke > -- dm-devel mailing list dm-devel@xxxxxxxxxx https://www.redhat.com/mailman/listinfo/dm-devel
