[ltp] Problems with disc-performance and libata
Oli
linux-thinkpad@linux-thinkpad.org
Wed, 22 Aug 2007 11:48:22 +0200
Hi,
I'm using a Lenovo Thinkpad with TOSHIBA MK1234GS HD and MATSHITA
DVD-RAM UJ-850 on 82801GBM/GHM (ICH7 Family) Serial ATA Storage
Controller IDE. The Kernel is a 2.6.22.1 from Kernel.org. I'm using
Ubuntu FF.
sudo hdparm -t /dev/scd0 gives
Timing buffered disk reads: 2 MB in 4.47 seconds = 457.75 kB/sec
sudo hdparm -t /dev/sda gives
Timing buffered disk reads: 102 MB in 3.04 seconds = 33.51 MB/sec
which is a quite poor performance and makes any streaming impossible.
Obviously UDMA is working correctly as dmesg | grep UDMA says:
[ 43.636590] ata1: SATA max UDMA/133 cmd 0xffffc20004612500 ctl
0x0000000000000000 bmdma 0x0000000000000000 irq 314
[ 43.636595] ata2: SATA max UDMA/133 cmd 0xffffc20004612580 ctl
0x0000000000000000 bmdma 0x0000000000000000 irq 314
[ 43.636599] ata3: SATA max UDMA/133 cmd 0xffffc20004612600 ctl
0x0000000000000000 bmdma 0x0000000000000000 irq 314
[ 43.636603] ata4: SATA max UDMA/133 cmd 0xffffc20004612680 ctl
0x0000000000000000 bmdma 0x0000000000000000 irq 314
[ 44.177980] ata1.00: ATA-7: TOSHIBA MK1234GSX, AH002E, max UDMA/100
[ 44.179192] ata1.00: configured for UDMA/100
[ 44.326918] ata5: PATA max UDMA/133 cmd 0x00000000000101f0 ctl
0x00000000000103f6 bmdma 0x0000000000011880 irq 14
[ 44.326922] ata6: PATA max UDMA/133 cmd 0x0000000000010170 ctl
0x0000000000010376 bmdma 0x0000000000011888 irq 15
[ 44.571970] ata5.00: ATAPI: MATSHITADVD-RAM UJ-850, RB01, max UDMA/33
[ 44.738552] ata5.00: configured for UDMA/33
[ 0.610332] ata5.00: configured for UDMA/33
[ 0.750296] ata1.00: configured for UDMA/100
Is there any solution for that problem or at least a workaround?
Ubuntu-maintainers won't fix that soon:
https://bugs.launchpad.net/ubuntu/+source/linux-source-2.6.22/+bug/110636
I also tried to recompile the kernel as recommended in the official
libata-faq ( http://linux-ata.org/faq.html ) but the system stuck when
booting the root-partition. Booting with combined_mode=libata or
combined_mode=ide didn't affect the performance, too.
Thanks in advance,
Oli