Hello, I am testing block device corruption with fio, and i believe fio lacks a certain helpful feature. This is the config i am using: ``` [global] ioengine=libaio direct=1 group_reporting exitall bs=8k rw=write iodepth=64 size=671088640 numjobs=1 verify=pattern loops=1000 verify_dump=1 verify_fatal=1 filename=/dev/sdad [job0x0] verify_pattern=0x0102030405060708090a0b0c0d0e0f210908163826%o offset=0 [job0x1] verify_pattern=0x1112131415161718191a1b1c1d1e1f210908163826%o offset=671088640 [job0x2] verify_pattern=0x2122232425262728292a2b2c2d2e2f210908163826%o offset=1342177280 [job0x3] verify_pattern=0x3132333435363738393a3b3c3d3e3f210908163826%o offset=2013265920 [job0x4] verify_pattern=0x4142434445464748494a4b4c4d4e4f210908163826%o offset=2684354560 [job0x5] verify_pattern=0x5152535455565758595a5b5c5d5e5f210908163826%o offset=3355443200 [job0x6] verify_pattern=0x6162636465666768696a6b6c6d6e6f210908163826%o offset=4026531840 [job0x7] verify_pattern=0x7172737475767778797a7b7c7d7e7f210908163826%o offset=4697620480 [job0x8] verify_pattern=0x8182838485868788898a8b8c8d8e8f210908163826%o offset=5368709120 [job0x9] verify_pattern=0x9192939495969798999a9b9c9d9e9f210908163826%o offset=6039797760 [job0xa] verify_pattern=0xa1a2a3a4a5a6a7a8a9aaabacadaeaf210908163826%o offset=6710886400 [job0xb] verify_pattern=0xb1b2b3b4b5b6b7b8b9babbbcbdbebf210908163826%o offset=7381975040 [job0xc] verify_pattern=0xc1c2c3c4c5c6c7c8c9cacbcccdcecf210908163826%o offset=8053063680 [job0xd] verify_pattern=0xd1d2d3d4d5d6d7d8d9dadbdcdddedf210908163826%o offset=8724152320 [job0xe] verify_pattern=0xe1e2e3e4e5e6e7e8e9eaebecedeeef210908163826%o offset=9395240960 [job0xf] verify_pattern=0xf1f2f3f4f5f6f7f8f9fafbfcfdfeff210908163826%o offset=10066329600 ``` By changing the timestamp in verify_pattern i can tell if the .received block i get is left over from the last run or is a newly written one. Unfortunately, i have to use the `loops` option to have a better chance of getting corruption in a single run. This leaves an ambiguity in the results i get: i am not able to tell if a corruption has arisen from the driver accidentally duplicating a write request, misplacing it or loosing it completely. I would like to be able to add e.g. a `%i` format option that would update with each iteration, i.e. one of the patterns above would be rewritten as verify_pattern=0x0102030405060708090a0b0c0d0e0f210908163826%o%i and would expand to 0x0102030405060708090a0b0c0d0e0f210908163826 0020130000000000 0000000000000000 on the very first (0th) iteration of the loop, and would expand to 0x0102030405060708090a0b0c0d0e0f210908163826 0020130000000000 0f27000000000000 on the very last (9999th, i.e 0x270Fth) iteration of the loop (assuming pattern for the block #153 of size 8k, i.e. offset of 1253376 bytes), whitespace inserted for clarity. This feature would give very valuable insight into the verification fail i am investigating. Thank you, Abrakhin Andrei
