I have investigated more on the effect of decreased speed with EIDE drives
at least with Triton I and II boards on the recent kernel.
For that purpose, I did a binary search on the older kernels I still have
available. Between 2.0.20 and 2.0.23 I can observe the drop in speed. I
have a P133 32MByte Triton II (Gigabyte HX) board. This drop is best visible
with a command like
"time dd if=/dev/hda of=/dev/null bs=1024k count=64", as there is no change
that parts of the read process get chaches.
Looking at the patch files
patch-2.0.21.gz
patch-2.0.22.gz
patch-2.0.23.gz
I only find the "SUPPORT_SLOW_DATA_PORTS" and the "slow"-field in the
ide_drive_s structure introduced in the area of ide.h, ide.c and triton.c.
As I was in doubt, how the slow field is intitalized, I explicitly set
drive->slow = 0;
in the initializations at the end of init_hwif_data() in
(2.0.27/)../ide.c. After recompiling and rebooting with this new kernel, I
still observed the slow IDE-speed. The next try was to
#define SUPPORT_SLOW_DATA_PORTS 0 /* 0 to reduce kernel size */
in (2.0.27/)../ide.h. Recompiling and rebooting again didn't get back high
IDE-speed.
So I guess some other mechanisme outside of ide.[ch] and triton.c must have
changed and does have a bad influence on the IDE-read speed.
Has anyone any clue?
Best regards
-- Uwe Bonnes bon@elektron.ikp.physik.th-darmstadt.deInstitut fuer Kernphysik Schlossgartenstrasse 9 64289 Darmstadt --------- Tel. 06151 162516 -------- Fax. 06151 164321 ----------