erasure code - erasure_code_perf.c only checks for erasure corrections in data rows, not parity rows.

[jeffareid]

The code should check for erasures in the parity rows, since that is needed if rebuilding an array after a drive replacement.

[pablodelara]

Hi @jeffareid. Could you share the command line to reproduce this issue? I am running the following command line and it works fine:
./erasure_code/erasure_code_perf -k 4 -p 10 -e 1

[pablodelara]

Right, that option doesn't work because of the condition that "e <= k". I'll look into it, but there is no issue with any infinite loop, it just prints:
Number of errors (5) cannot be higher than number of data buffers (4)

[jeffareid]

I forgot I had modified it to test an alternate faster encoding scheme (LFSR with self-reciprocal polynomial), which is how it got stuck in a loop. I edited the original issues, and deleted no longer needed posts.

Suggested changes: the loop to find random error index numbers < k (number of data rows) should be changed to find random numbers < m (total number of rows, data+parity), to check for erasures in both data and parity rows (needed if erasure code used to rebuild an array). With this change, then the number of errors just has to be less than or equal to p (number of parity rows).

[pablodelara]

Hi @jeffareid. I pushed an update on this application. Feel free to test it ;)

[jeffareid]

@pablodelara - erasure_code_update_perf.c should get the same change (%k -> %m).

[pablodelara]

@pablodelara - erasure_code_update_perf.c should get the same change (%k -> %m).
I will make the change after the release.