I dont know if this is still of interest: As it seemed that all my truncation errors appeared for DLEN=64 packets, I Limited to DLEN=32. After 38 h I got an "increase error". Seems logical that this can happen if the DLEN is read wrong. 1: idx 8 exp:2A5#00 08 27 05 A8 9F 48 02 94 30 F4 DA AD 3D 0C 53 8C 7F 91 7E 53 84 9B EB 1: rx:2A5#00 08 27 05 A8 9F 48 02 94 30 F4 DA AD 3D 0C 53 8C 7F 91 7E 53 84 9B EB C7 1D 92 24 D1 B4 72 29 Now the interesting part: 1: RX72 (010/009) 2A5#00 09 D2 91 35 52 03 A5 41 DA 2A 68 DF D3 07 27 D5 9B 57 C9 76 04 06 82 1: RX72 (010/008) 2A5#00 08 27 05 A8 9F 48 02 94 30 F4 DA AD 3D 0C 53 8C 7F 91 7E 53 84 9B EB C7 1D 92 24 D1 B4 72 29 ******** 1: RX72 (010/007) 2A5#00 07 7F 72 70 EF D7 85 1: RX72 (010/006) 2A5#00 06 E7 F2 46 7D 49 32 A6 4E 72 3D F2 D4 F3 1D 98 4E F5 4A 1: RX72 (010/005) 2A5#00 05 6F 65 15 38 2E 23 78 43 59 F5 1: RX72 (009/004) 2A5#00 04 95 57 EC 29 05 40 97 A3 62 B5 2B C3 5B D8 B2 31 7C 26 1: RX72 (008/003) 2A5#00 03 1E 88 61 A6 AC EE D5 A4 74 3E E8 EF C8 C9 0E 40 15 36 7D A4 98 50 1: RX72 (005/002) 2A5#00 02 E9 40 87 24 8D 2F F8 85 69 44 7B 8A E0 1F 77 4C 27 3D 8E 62 D3 F1 BF 9B 94 AD 54 6D 9E 21 1: RX72 (004/001) 2A5#00 01 63 5C C8 A2 EE B9 D9 F1 CD B1 1: RX72 (004/000) 2A5#00 00 F8 39 4C F5 21 9C A4 ED 6C 08 C6 42 3B 95 1: RX72 (010/009) 2A5#00 09 B2 54 EC AB 1B 3A A0 D9 01 AD BB 3E 12 2F ED B7 50 41 C1 9F 31 76 1: RX72 (010/008) 2A5#00 08 0C E2 3D 65 74 AF A4 21 53 0A 4E F0 B5 8D AB 75 91 F5 73 AE A4 BC 1: RX72 (010/007) 2A5#00 07 F8 C7 4D F5 BE C7 1: RX72 (010/006) 2A5#00 06 AD A4 66 2D 9D AA 3E BA C0 04 4A FE 1D AF 48 04 AF 30 1: RX72 (010/005) 2A5#00 05 FD F7 79 17 24 76 59 92 72 3F 1: RX72 (010/004) 2A5#00 04 7F F9 44 11 1D 15 C6 90 3E 18 64 4A E7 00 E8 F4 17 1D 1: RX72 (009/003) 2A5#00 03 52 D4 0B CE 9C 9E 0D 17 35 08 DF 5C 1A 2D FC 42 6E FC E6 1B 31 27 1: RX72 (006/002) 2A5#00 02 62 E2 7C C7 92 3F 7B 97 FB C8 27 11 4D A0 28 41 00 A4 AD 35 41 64 C7 1D 92 24 D1 B4 72 29 ********* 1: RX72 (005/001) 2A5#00 01 B6 92 DC 2E 5D B3 22 4E 58 9E 1: RX72 (004/000) 2A5#00 00 04 BC 4D 90 16 85 78 16 F3 60 7C 6C 35 9D 1: RX72 (010/009) 2A5#00 09 9E 6A E2 7E 4F 81 1F BE AD 84 15 E5 50 D1 80 03 3A 46 63 85 F2 CB 1: RX72 (010/008) 2A5#00 08 65 56 38 CC BC 9B B2 69 C5 60 4B AC 68 94 C7 97 5D EA 3A 92 BA EF 1: RX72 (010/007) 2A5#00 07 E0 E5 9E A0 6F 2C 1: RX72 (010/006) 2A5#00 06 AC 2A AD E2 E6 08 49 1B 2E C8 23 A3 35 D0 B5 0C B0 F3 The on-top data has been received before (see marked lines). However at a queue length of 20 (according to ethtool) I would not have expected that the distance to this frame Would be 16, but 20. Doesn't the chip/driver use the queue elements one-by-one? -- Stefan
