Windows eats Fedora installation sticks

Old news, I know. But I've been burned again. A Fedora installation USB stick, prepared by dd-ing the .iso file onto the raw stick works fine. But if you have that stick in the computer, and Windows sees it, it will drop a few bytes on it. The result is that the stick cannot be used to boot Fedora (until you rewrite it). Sheesh. Why did it happen to me, when I know of the problem? To run the stick, during power on, I have to hit a key to get into the choose-a-device-to-boot-from menu. It's a race, and sometimes I lose.

On Wed, Sep 04, 2024 at 11:43:40AM -0400, D. Hugh Redelmeier via talk wrote:
Old news, I know. But I've been burned again.
A Fedora installation USB stick, prepared by dd-ing the .iso file onto the raw stick works fine.
But if you have that stick in the computer, and Windows sees it, it will drop a few bytes on it. The result is that the stick cannot be used to boot Fedora (until you rewrite it). Sheesh.
Why did it happen to me, when I know of the problem? To run the stick, during power on, I have to hit a key to get into the choose-a-device-to-boot-from menu. It's a race, and sometimes I lose.
It appears windows takes it upon itself to "fix" the GPT. It changes the offset to the end of disk and where the backup partition table should be, and the checksum. I suspect a USB key without a GPT would not be affected. 000001e0: 0000 0000 0000 0000 0000 0000 0000 0000 ................ 000001f0: 0000 0000 0000 0000 0000 0000 0000 55aa ..............U. 00000200: 4546 4920 5041 5254 0000 0100 5c00 0000 EFI PART....\... -00000210: 00f9 743c 0000 0000 0100 0000 0000 0000 ................ <- checksum first 4 bytes +00000210: 5acf 1fd8 0000 0000 0100 0000 0000 0000 ................ -00000220: f76b 4400 0000 0000 4000 0000 0000 0000 ........@....... <- location of backup GPT first 8 bytes +00000220: ff7f 3707 0000 0000 4000 0000 0000 0000 ........@....... -00000230: b86b 4400 0000 0000 7de2 0cbd e0a6 374c ........}.....7L <- last usable sector of device first 8 bytes +00000230: c07f 3707 0000 0000 7de2 0cbd e0a6 374c ........}.....7L 00000240: beec ab4b 6f9b 6a73 0200 0000 0000 0000 ...Ko.js........ 00000250: f800 0000 8000 0000 8bbd c3e0 0000 0000 ................ 00000260: 0000 0000 0000 0000 0000 0000 0000 0000 ................ I must admit I always found the media check in redhat weird. Like how unreliable do you think media is these days? And don't your packages have checksums on them so it would fail if there was a media error? Wasting time checking the media before installing compared to the minute chance of it failing during the install? I don't get it. I don't recall if Debian has a media check option, although if it does I have certainly never used it. -- Len Sorensen

On Tue, Sep 10, 2024 at 12:31:26PM -0400, wrote:
It appears windows takes it upon itself to "fix" the GPT. It changes the offset to the end of disk and where the backup partition table should be, and the checksum.
I suspect a USB key without a GPT would not be affected.
000001e0: 0000 0000 0000 0000 0000 0000 0000 0000 ................ 000001f0: 0000 0000 0000 0000 0000 0000 0000 55aa ..............U. 00000200: 4546 4920 5041 5254 0000 0100 5c00 0000 EFI PART....\...
-00000210: 00f9 743c 0000 0000 0100 0000 0000 0000 ................ <- checksum first 4 bytes +00000210: 5acf 1fd8 0000 0000 0100 0000 0000 0000 ................
-00000220: f76b 4400 0000 0000 4000 0000 0000 0000 ........@....... <- location of backup GPT first 8 bytes +00000220: ff7f 3707 0000 0000 4000 0000 0000 0000 ........@.......
-00000230: b86b 4400 0000 0000 7de2 0cbd e0a6 374c ........}.....7L <- last usable sector of device first 8 bytes +00000230: c07f 3707 0000 0000 7de2 0cbd e0a6 374c ........}.....7L
00000240: beec ab4b 6f9b 6a73 0200 0000 0000 0000 ...Ko.js........ 00000250: f800 0000 8000 0000 8bbd c3e0 0000 0000 ................ 00000260: 0000 0000 0000 0000 0000 0000 0000 0000 ................
I must admit I always found the media check in redhat weird. Like how unreliable do you think media is these days? And don't your packages have checksums on them so it would fail if there was a media error? Wasting time checking the media before installing compared to the minute chance of it failing during the install? I don't get it. I don't recall if Debian has a media check option, although if it does I have certainly never used it.
I just checked debian's installer and they only use a DOS MBR partition table, not a GPT like fedora. So that avoids getting corrupted I guess, and they still have an efi boot partition of course, so it will boot just fine on modern machines. Not sure why fedora decided to make their installer image more complicated than it had to be. I don't see where they would have any benefit from it. -- Len Sorensen
participants (2)
-
D. Hugh Redelmeier
-
Lennart Sorensen