On Fri, 24 Oct 2003, Mark Scott wrote:Just checking... you mean the difference between -18mdk and -19mdk is a newer libata, in which only sata_promise changed? (i.e. no change in sata_via)
Ah, yes, doing a "modprobe sata_via" makes the drive work. So I guess the kernel doesn't need patching, but maybe updating (the last libata patch was 17th Oct). I don't know if this is reliable or not... I'll find out over time :-)
AFAICS, this only brings up sata_promise 0.83, which is now in 2.4.22-19mdk.
I am interested to know the performance of your plain IDE drive. Strangely, it does appear that if sata_via is loaded, kernel doesn't recognize normal disk as UDMA133 capable...I'm afraid I don't have that problem, the PATA drive isn't UDMA133 capable, it's a Seagate ST36561A (6.3Gb). It gets a buffered disk read speed (from hdparm -t) of 9.46MB/sec. The SATA disk achieves 50.67MB/sec.
I also did the following to create a 512 Mb file: # time dd if=/dev/zero of=testfile count=1000000 1000000+0 records in 1000000+0 records out 0.61user 7.10system 0:14.15elapsed 54%CPU (0avgtext+0avgdata 0maxresident)k 0inputs+0outputs (146major+42minor)pagefaults 0swaps
No idea how good a metric it is, but writing 512Mb in 14.15s is 36.18Mb/s (ext3 fs)
BTW, ia32 installer won't know about those SATA drivers unless it is rebuilt with latest -BOOT kernel & ldetect-lst.
Uhuh... I was thinking of doing 'urpmi --root ...' to install a basic bootable system on the SATA disk (and putting sata_via in its initrd), bypassing the need to work out how to rebuild the installer. Unless that's more beneficial to everyone...
-- Mark
