Comment 197 for bug 453579

I believe I must withdraw my bug report. I have a test case that reproduces the problem, but it does not seems to be related to ext4, as turned out today.

I read about 2.6.31-18 here and updated yesterday. But also with the new kernel I could reproduce the problem. Wondering about that I created an XFS partition and repeated the test on it ... also positive. So my problem is somewhere else.

The test is copying 4 big files totaling 11gb via nfs from an old Dapper box to a local partition. One of the files always ended up with a mismatching sha1 sum. The "cmp -l" output is always a single contiguous 128 byte block at some random offset. The values don't seem to be affected by single bit flips (110010 -> 101010, 11111101 -> 11010101, 10100 -> 101010, 1110 -> 11010111, 11110001 -> 11010010). Memtest86 also ran fine, at least on this box, I did not yet test the Dapper one.