notfixed 479217 2.6.25-2
reopen 479217
thanks
-=| Ben Armstrong, Wed, May 28, 2008 at 06:35:06AM -0300 |=-
> Package: linux-latest-2.6
> Version: 2.6.25-4
> Followup-For: Bug #479217
>
> synrg@dove:~$ dmesg | egrep 'Linux version|cable'
> [ 0.000000] Linux version 2.6.25-2-686 (Debian 2.6.25-4) (maks@debian.org) (gcc version 4.1.3 20080420 (prerelease) (Debian 4.1.2-22)) #1 SMP Tue May 27 15:38:35 UTC 2008
> [ 5.060487] ata2.00: limited to UDMA/33 due to 40-wire cable
Same here.
This is the test case in the Git commit:
> Before patch:
> $ dd if=/dev/sda of=/dev/null bs=2M count=64 iflag=direct
> 134217728 bytes (134 MB) copied, 5.29612 s, 25.3 MB/s
>
> After patch:
> $ dd if=/dev/sda of=/dev/null bs=2M count=64 iflag=direct
> 134217728 bytes (134 MB) copied, 3.94955 s, 34.0 MB/s
This is what I observe:
134217728 bytes (134MB) copied, 4.85051 s, 27.7 MB/s
A bit higher than 25.3 MB/s, yes, but still far from 34.0 MB/s. And I
see the "limited UDMA/33 due to 40-wire cable" too.
lspci -n shows the same IDs:
00:1f.2 IDE interface [0101]: Intel Corporation 82801FBM (ICH6M)
SATA Controller [8086:2653] (rev 04) (prog-if 80 [Master])
Subsystem: ASUSTeK Computer Inc. Device [1043:82d8]
Flags: bus master, 66MHz, medium devsel, latency 0, IRQ 19
I did a naive attempt to trace the problem, please don't laugh :)
The "limited to UDMA/33 due to 40-wire cable" message seems to come
from drivers/ata/libata-core.c:4141, but it is only shown if
cable_is_40wire() returns true. cable_is_40wire() should return false
in EeePC case as the cable is ATA_CBL_PATA40_SHORT (or at least the
patch makes ich_pata_cable_detect() return that).
So I am puzzled (no surprises here). Either ich_pata_cable_detect is
not used, or something else is going on. Hopefully people with hands
in the kernel code will have some clues.
--
dam JabberID: dam@jabber.minus273.org
Attachment:
signature.asc
Description: Digital signature