[Gptfdisk-general] [PATCH] Fix problem resizing RAID disks

[Frediano Z. <fre...@ci...>]

This patch fix a nasty problem using -Z option in some cases if raid type
is change. The reason is that usually when you change RAID type only first
part get cleaned so you end up with GPT backup copy (at end) still
present but at different location.

Imaging you have a RAID0 disk with 2 disk so you have clusters
alternated with last cluster ending at disk #2 containing the backup
GPT. You destroy the RAID and create another RAID0 with 4 disks. If disk
#2 now is disk #4 (last) of new RAID you end up with last cluster
containing the old GPT. However the LBA information still point to old
location which now point at the middle of the disk while protective MBR
and main GPT cleared. In this situation doing a --zap-all command gdisk
try to load GPTs headers finding the backup. Then it try to zero them
using LBA informations just read so it clear some sectors in the middle
of the disk. The result if that disk is not cleared correctly.

The patch fix LBA addresses before writing zeroes clearing correct
locations.

Signed-off-by: Frediano Ziglio <fre...@ci...>
---
 gpt.cc |    1 +
 1 file changed, 1 insertion(+)

diff --git a/gpt.cc b/gpt.cc
index fa7b661..3244cc9 100644
--- a/gpt.cc
+++ b/gpt.cc
@@ -1314,6 +1314,7 @@ int GPTData::DestroyGPT(void) {
    uint8_t* emptyTable;
 
    memset(blankSector, 0, sizeof(blankSector));
+   ClearGPTData();
 
    if (myDisk.OpenForWrite()) {
       if (!myDisk.Seek(mainHeader.currentLBA))
-- 
1.7.10.4