I was just wondering how hard it would be to modify the kernel to write
to two swap partitions at the same time?
If the swap-in read call failed on one swap partition it would try the
other partition before returning a swap-in failed.
Its easy to setup multiple swap partitions over different partitions
currently but if one partition fails and the kernel had some important
memory items in their it's core dump time:-)
Unlike RAID just telling the kernel to write to two partitions at the
same time shouldn't speedup or slowdown operations that much. If RAID-1
was fully used it should speed up read & write operations but that would
increase the kernel size and complexity too much I expect.
If both swap partitions were exactly equal size then the kernel wouldn't
really need to do anything flash except have two file descriptors open
and do two write calls with the same data on different file descriptors,
then on error (-1) from a read call do a read on the second file
descriptor before failing.
Anyone know the kernel memory swap C code?
Regards,
David Robinson
[EMAIL PROTECTED]
