Comment 15 for bug 444915

Ok after much debugging it appears there is a bug in the elevator affecting the merging of read-ahead requests. This is exposed horribly by the sequential scan of the FAT as required to calculate the current free information for a FAT filesystem on first mount. I think we have it nailed and have produced some test kernels with the fix. If you could test those and report back here that would be very useful. Kernels will be at the URL below shortly:

    http://people.canonical.com/~apw/lp444915-karmic/