Characterizing error correlation based on error logging for computer buses
Systems and devices can include forward error correction (FEC) logic to identify a correctable error in the first flit, and correct the correctable error using three error correcting code (ECC) groups. System and devices can also include an error log, the correctable error log to log a symbol number in the first flit corrected by each ECC group, and to log a magnitude of the correctable error corrected by each ECC group in the first flit; and a configuration register to log link error correlation, the link error correlation comprising a indication of one or more bits in error in the first flit.
1. An apparatus comprising:
a port to receive flow control units (flits) across a serial interconnect, the flits comprising a first flit and subsequent flits; and
a protocol stack comprising:
forward error correction (FEC) logic to identify a correctable error in the first flit using each of three error correcting code (ECC) groups, the FEC logic to correct the error using each of the ECC group; and
an error log, the error log to log:
a symbol number in the first flit corrected by each ECC group, and
a magnitude of the error corrected by each ECC group in the first flit.
2. The apparatus of claim 1 , wherein the error log comprises information to indicate that each ECC group corrected an error in the first flit and that the first flit did not include an uncorrectable error.
3. The apparatus of claim 1 , wherein the magnitude of the error corrected by each ECC group comprises an amount of error correction with a symbol indicated by the symbol number.
4. The apparatus of claim 1 , wherein the error log comprises information indicating that at least one subsequent flit included an error, the error in the at least one subsequent flit comprising one of a correctable error or an uncorrectable error.
5. The apparatus of claim 4 , wherein the error log is a first error log, the apparatus comprising a second error log to log:
a symbol number in at least one subsequent flit corrected by each ECC group, and
a magnitude of the error corrected by each ECC group in the at least one subsequent flit.
6. The apparatus of claim 1 , wherein apparatus comprises a configuration register to log link error correlation.
7. The apparatus of claim 1 , further comprising cyclic redundancy check (CRC) logic to identify a correctable error in the first flit and to correct the error in the first flit, the FEC error log to log the correctable error corrected by the CRC logic.
8. The apparatus of claim 1 , wherein the FEC logic is to detect an uncorrectable error in the first flit, the apparatus further comprising an uncorrectable error log to log uncorrectable errors detected in the first flit.
9. The apparatus of claim 8 , wherein the uncorrectable error log is to log a type of error in the first flit, the type of error comprising no error, a correctable error, or an uncorrectable error.
10. The apparatus of claim 8 , further comprising:
a replay buffer to store an outgoing flit, the outgoing flit comprising the uncorrectable error detected by the FEC logic; and
error detection logic to:
compare a retried flit that does not include the uncorrectable error with the outgoing flit in the replay buffer;
determine a bit position of the uncorrectable error based on the comparison; and
log the bit position of the uncorrectable error in the error log.
11. The apparatus of claim 10 , further comprising exclusive OR logic to compare the retried flit with the outgoing flit.
12. The apparatus of claim 8 , further comprising error detection logic to:
determine by a SKP OS that a next flit will contain all zeros (all zero flit); and
compare the next flit with all zeros to identify a bit location of an error in the next flit.
13. A method comprising:
receiving a first flow control unit (flit) at a port across a serial interconnect, the first flit a first flit of a series of flits to be received by the port;
determining a correctable error in the first flit;
correcting the correctable error in the first flit by each of three of error correcting code (ECC) groups;
recording a symbol number in the first flit corrected by each ECC group in an error log; and
recording a magnitude of the error corrected by each ECC group in the first flit in the error log.
14. The method of claim 13 , further comprising recording in the error log information indicating that each ECC group corrected an error in the first flit and that no uncorrectable errors were detected in the first flit.
15. The method of claim 13 , wherein the magnitude of the error corrected by each ECC group comprises an amount of error correction with a symbol indicated by the symbol number.
16. The method of claim 13 , further comprising:
detecting an uncorrectable error in the first flit; and
recording the uncorrectable error in an uncorrectable errors log.
17. The method of claim 16 , further comprising recording in the uncorrectable errors log a type of error in the first flit, the type of error comprising no error, a correctable error, or an uncorrectable error.
18. The method of claim 16 , further comprising:
comparing a retried flit that does not include the uncorrectable error with an outgoing flit in a replay buffer;
determine a bit position of the uncorrectable error based on the comparison; and
log the bit position of the uncorrectable error in the error log.
19. The method of claim 16 , further comprising:
receiving a SKiP ordered set (SKP OS);
determining by a SKP OS that a next flit will contain all zeros (all zero flit);
comparing the next flit with all zero flit; and
identifying a bit location of the uncorrectable error in the next flit based on the comparison.
20. A system comprising:
a hardware processor;
a transmitter port for transmitting flow control units (flits) across a serial interconnect;
a receiver port to receive flits from across a serial interconnect;
a protocol stack comprising:
forward error correction (FEC) logic to:
identify a correctable error in a first flit, and
correct the correctable error using three error correcting code (ECC) groups; and
an error log, the correctable error log to log a symbol number in the first flit corrected by each ECC group, and to log a magnitude of the correctable error corrected by each ECC group in the first flit; and
a configuration register to log link error correlation, the link error correlation comprising a indication of one or more bits in error in the first flit.
21. The system of claim 20 , wherein the magnitude of the error corrected by each ECC group comprises an amount of error correction with a symbol indicated by the symbol number.
22. The system of claim 20 , wherein the system comprises an uncorrectable error log; and wherein the FEC logic is to detect an uncorrectable error in the first flit and to log the uncorrectable error detected in the first flit.
23. The system of claim 22 , wherein the uncorrectable error log is to log a type of error in the first flit, the type of error comprising no error, a correctable error, or an uncorrectable error.
24. The system of claim 22 , further comprising:
a replay buffer to store an outgoing flit, the outgoing flit comprising the uncorrectable error detected by the FEC logic; and
error detection logic to:
compare a retried flit that does not include the uncorrectable error with the outgoing flit in the replay buffer;
determine a bit position of the uncorrectable error based on the comparison; and
log the bit position of the uncorrectable error in the error log.
25. The system of claim 24 , further comprising exclusive OR logic to compare the retried flit with the outgoing flit.
26. The system of claim 22 , further comprising error detection logic to:
determine by a SKP OS that a next flit will contain all zeros (all zero flit); and
compare the next flit with all zeros to identify a bit location of an error in the next flit.