Linux thinks there are partitions on disks that were just tested

[antifuchs]

I believe some people are getting very unlucky with the random data being written to their disks: They end up with partitions on devices that just finished testing.

One of these disks has these first 512 bytes:

:;    hexdump first-512.bin
0000000 f6ad 79c8 e65c 9ec9 a586 1e92 8d4a de49
0000010 10ee 8cc9 c9fc 2f35 598a 92c8 7f41 5e33
0000020 d5d3 d27a ab35 a843 a6b8 5111 c16d 7782
0000030 6cae ab2d fc9e cad8 b435 809b be12 c64d
0000040 dc72 f6c1 db6b 7acf 28bd dfb0 5088 74c1
0000050 a921 45e4 63e0 d854 ff15 c46a e43c 2787
0000060 08e9 b5d3 6d42 babd 86ea 34e2 01c5 9b05
0000070 b604 ac69 e890 b90c bc6f 4796 1247 6c60
0000080 3f5a a52e 8744 e0cc 06ff f2b2 4e25 8fda
0000090 4722 69e7 c7a0 1b79 732d 402d 599e cd36
00000a0 90a0 6dad aeeb bec2 f217 f4d4 b516 8dca
00000b0 f2de 31f5 ab45 76cf c50e dac4 daf5 5d82
00000c0 5c2c 9cd6 3321 c960 d39c cb0e 132c e5c9
00000d0 dd1f 66f8 9cb3 1185 743a 1032 21c4 7a80
00000e0 c0de 42c2 d8b1 5b18 ed3c 8e19 22a1 42f1
00000f0 44f7 502e 5266 b3a3 5d94 c797 4588 1151
0000100 b4a9 596f fba2 8c00 779a 7556 d111 77fa
0000110 5199 fcc3 8467 88ea 9185 0452 644c 9f16
0000120 1d64 5913 cfca 500d 453e 1fab f320 4996
0000130 3fc9 85da 5b83 37c8 72a0 ab5c e98d ab22
0000140 68a8 4c75 fc32 d320 c9ab 7bda 19a7 f367
0000150 f4a6 b2cc c844 e23c b6eb 9438 2747 9136
0000160 baa1 3d9c 578a d9ae 6275 a381 a97f 0b40
0000170 bafa b2f2 5fd6 7819 28b5 ab03 a65b c41e
0000180 671e d48b 8d36 bafa 91b4 af14 603f 6e10
0000190 e2a4 20c8 2949 5f42 5614 6bd6 0cf4 4482
00001a0 0f55 063b db11 6450 88ff aff2 d94a 512b
00001b0 7d59 c925 23d0 5e05 13a1 31e7 ee8d 0ce6
00001c0 d9cb 8b17 e889 e6e1 fd76 b694 978a f64f
00001d0 9212 0afb 9a96 1e6c f820 0c32 67ca adfb
00001e0 f776 6816 a1fb 06c2 d18c 5386 d0e6 adba
00001f0 101b 1a9b 7e3e 60e1 6cbf 1408 64e4 a350
0000200

[richo]

I got this experience, although for completeness sake I did not need to wait for it to finish testing for Linux to decide there are partitions.

It also doesn't trip disk-spinners sane drive detection, so it's not a huge issue but it is a strange remnant leftover on my system.

[antifuchs]

This is so very confusing: I would expect an MBR that confuses the partition detector to end in 55FF, as documented here. Maybe Linux thinks these are GPT partitions, even? But there, LBA1 (the 512 bytes after an opaque MBR) would have to start with "EFI PART", even more absurd.

If you see this again, can you post the output from p on the fdisk prompt for the affected drive?

[richo]

Sure can I am about to head off on a trip for a week, but when I return I'll shove more disks in there. Unless there's some other spooky action going on I think it should be pretty easy to reproduce.

[antifuchs]

That sounds great! Thanks in advance. I should add that I haven’t seen this once in my HBA/disk configuration, so I wonder if this could be a hardware thing in addition to everything.