Fix kpart -d regression where matching devices not found

The Lorax project found a regression in kpartx -d behavior introduced by the following commit: 9bdfa3eb8e24b668e6c2bb882cddb0ccfe23ed5b

This patch should fix the problem. It's easy to reproduce this problem (I did it on Fedora 28) with the following commands:

        # kpartx -av disk.img
        ...
        # kpartx -dv disk.img

Notice that the loop devices and partition mappings created by the first command are not cleaned up by the second command.

Stef Walter (1):
kpartx: Use absolute paths to create mappings

 kpartx/kpartx.c | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

2.17.1

kpartx -d now uses absolute paths to delete mappings, since the commit 9bdfa3eb8e24b668e6c2bb882cddb0ccfe23ed5b. We should use those same absolute paths to create the mappings.

Without this patch, the following workflow (as seen in the manual page for kpartx) fails to actually remove the devices:

    # kpartx -av disk.img
    ...
    # kpartx -d disk.img
---
 kpartx/kpartx.c | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

index 442b6bd9..d4fb53be 100644
@@ -341,7 +341,7 @@ main(int argc, char **argv){
 		if (!loopdev) {
 			loopdev = find_unused_loop_device();
 
-			if (set_loop(loopdev, device, 0, &ro)) {
+			if (set_loop(loopdev, rpath, 0, &ro)) {
 				fprintf(stderr, "can't set up loop\n");
 				exit (1);
 			}

2.17.1