Harddisk Performance

Hi!

Running 2.5.2-pre3, hdparm shows up very poor harddisk performance on my system compared to 2.4.x:

[Kernel 2.5.2-pre3]

elfie:~ # hdparm -tT /dev/hda

/dev/hda:

Timing buffer-cache reads:   128 MB in  1.77 seconds = 72.32 MB/sec
Timing buffered disk reads:  64 MB in 22.09 seconds =  2.90 MB/sec
                                                       ^^^^

[Kernel 2.4.17]

elfie:~ # hdparm -tT /dev/hda

/dev/hda:

Timing buffer-cache reads:   128 MB in  1.81 seconds = 70.72 MB/sec
Timing buffered disk reads:  64 MB in  6.61 seconds =  9.68 MB/sec
                                                       ^^^^

The tests were made under the same circumstances, of course, so no other process can be the cause.

Here come the harddisk and system informations:

elfie:~ # hdparm -i /dev/hda

/dev/hda:

Model=IBM-DHEA-36481, FwRev=HP6OA20C, SerialNo=SG0SGFJ0623 Config={ HardSect NotMFM HdSw>15uSec Fixed DTR>10Mbs } RawCHS=12592/16/63, TrkSize=0, SectSize=0, ECCbytes=28 BuffType=DualPortCache, BuffSize=472kB, MaxMultSect=16, MultSect=16 CurCHS=12592/16/63, CurSects=12692736, LBA=yes, LBAsects=12692736 IORDY=on/off, tPIO={min:240,w/IORDY:120}, tDMA={min:120,rec:120} PIO modes: pio0 pio1 pio2 pio3 pio4
DMA modes: sdma0 sdma1 sdma2 mdma0 mdma1 mdma2 udma0 udma1 *udma2

Linux version 2.4.17 (root@elfie) (gcc version 2.95.3 20010315 (release)) #1 Sat Dec 29 12:05:30 CET 2001
BIOS-provided physical RAM map:
BIOS-e820: 0000000000000000 - 00000000000a0000 (usable)
BIOS-e820: 00000000000f0000 - 0000000000100000 (reserved)
BIOS-e820: 0000000000100000 - 0000000010000000 (usable)
BIOS-e820: 00000000ffff0000 - 0000000100000000 (reserved) On node 0 totalpages: 65536
zone(0): 4096 pages.
zone(1): 61440 pages.
zone(2): 0 pages.
No local APIC present or hardware disabled
Kernel command line: auto BOOT_IMAGE=Linux ro root=301 hdc=ide-scsi hdd=ide-scsi idebus=33
ide_setup: hdc=ide-scsi
ide_setup: hdd=ide-scsi
ide_setup: idebus=33
Initializing CPU#0
Detected 400.914 MHz processor.
Console: colour VGA+ 80x25
Calibrating delay loop... 799.53 BogoMIPS
Memory: 256696k/262144k available (841k kernel code, 5064k reserved, 216k data, 204k init, 0k highmem)

Dentry-cache hash table entries: 32768 (order: 6, 262144 bytes)
Inode-cache hash table entries: 16384 (order: 5, 131072 bytes)
Mount-cache hash table entries: 4096 (order: 3, 32768 bytes)
Buffer-cache hash table entries: 16384 (order: 4, 65536 bytes)
CPU: Before vendor init, caps: 008021bf 808029bf 00000000, vendor = 2
CPU: L1 I Cache: 32K (32 bytes/line), D cache 32K (32 bytes/line)
CPU: L2 Cache: 256K (32 bytes/line)
CPU: After vendor init, caps: 008021bf 808029bf 00000000 00000002
CPU:     After generic, caps: 008021bf 808029bf 00000000 00000002
CPU:             Common caps: 008021bf 808029bf 00000000 00000002
CPU: AMD-K6(tm) 3D+ Processor stepping 01

Checking 'hlt' instruction... OK.
POSIX conformance testing by UNIFIX
mtrr: v1.40 (20010327) Richard Gooch ([email protected])
mtrr: detected mtrr type: AMD K6
PCI: PCI BIOS revision 2.10 entry at 0xfb3c0, last bus=1
PCI: Using configuration type 1
PCI: Probing PCI hardware
Unknown bridge resource 1: assuming transparent
Unknown bridge resource 2: assuming transparent
PCI: Using IRQ router VIA [1106/0586] at 00:07.0 Activating ISA DMA hang workarounds.
Linux NET4.0 for Linux 2.4
Based upon Swansea University Computer Society NET3.039 Initializing RT netlink socket
Starting kswapd
Real Time Clock Driver v1.10e
block: 128 slots per queue, batch=32
Uniform Multi-Platform E-IDE driver Revision: 6.31
ide: Assuming 33MHz system bus speed for PIO modes
VP_IDE: IDE controller on PCI bus 00 dev 39
VP_IDE: chipset revision 6
VP_IDE: not 100% native mode: will probe irqs later
VP_IDE: VIA vt82c586b (rev 41) IDE UDMA33 controller on pci00:07.1
ide0: BM-DMA at 0xe000-0xe007, BIOS settings: hda:DMA, hdb:DMA
ide1: BM-DMA at 0xe008-0xe00f, BIOS settings: hdc:DMA, hdd:DMA
hda: IBM-DHEA-36481, ATA DISK drive
hdb: Conner Peripherals 1275MB - CFS1275A, ATA DISK drive
hdc: CD-540E, ATAPI CD/DVD-ROM drive
hdd: CD-W54E, ATAPI CD/DVD-ROM drive

     ide0 at 0x1f0-0x1f7,0x3f6 on irq 14
     ide1 at 0x170-0x177,0x376 on irq 15

hda: 12692736 sectors (6499 MB) w/472KiB Cache, CHS=790/255/63, UDMA(33)
hdb: 2496876 sectors (1278 MB) w/64KiB Cache, CHS=619/64/63, DMA Partition check:
hda: hda1 hda2 hda3 < hda5 hda6 >
hdb: hdb1 < hdb5 >
Floppy drive(s): fd0 is 1.44M
FDC 0 is a post-1991 82077

SCSI subsystem driver Revision: 1.00
scsi0 : SCSI host adapter emulation for IDE ATAPI devices
Vendor: TEAC      Model: CD-540E           Rev: 3.0A
Type:   CD-ROM                             ANSI SCSI revision: 02
Vendor: TEAC      Model: CD-W54E           Rev: 1.1B
Type:   CD-ROM                             ANSI SCSI revision: 02

Attached scsi CD-ROM sr0 at scsi0, channel 0, id 0, lun 0 Attached scsi CD-ROM sr1 at scsi0, channel 0, id 1, lun 0
sr0: scsi3-mmc drive: 40x/40x cd/rw xa/form2 cdda tray Uniform CD-ROM driver Revision: 3.12
sr1: scsi3-mmc drive: 32x/32x writer cd/rw xa/form2 cdda tray
NET4: Linux TCP/IP 1.0 for NET4.0
IP Protocols: ICMP, UDP, TCP
IP: routing cache hash table of 2048 buckets, 16Kbytes
TCP: Hash tables configured (established 16384 bind 16384)
NET4: Unix domain sockets 1.0/SMP for Linux NET4.0.
VFS: Mounted root (ext2 filesystem) readonly. Freeing unused kernel memory: 204k freed
Adding Swap: 130748k swap-space (priority -1) [....]

elfie:~ # lspci
00:00.0 Host bridge: VIA Technologies, Inc. VT82C598 [Apollo MVP3] (rev 04) 00:01.0 PCI bridge: VIA Technologies, Inc. VT82C598 [Apollo MVP3 AGP] 00:07.0 ISA bridge: VIA Technologies, Inc. VT82C586/A/B PCI-to-ISA [Apollo VP] (rev 41)
00:07.1 IDE interface: VIA Technologies, Inc. VT82C586 IDE [Apollo] (rev 06) 00:07.3 Host bridge: VIA Technologies, Inc. VT82C586B ACPI (rev 10) 00:11.0 VGA compatible controller: S3 Inc. ViRGE/DX or /GX (rev 01)

# Heinz Diehl, 68259 Mannheim, Germany

Yes I just noticed that too (someone else reported it) -- seems to be due to missed merges, I'm investigating. The list corruption bug has higher prio rigth now, though.

Jens Axboe

I have found that if the elevator says BACK_MERGE or FRONT_MERGE, but the request queue doesn't allow the merge, the request will be put last in the queue instead of next to the request where the merge would have been done if allowed.

I first noticed this when playing with the pktcdvd.o module, where the result was duplicated writes to the same packet, but it does happen also for normal IDE hard disks.

I think this is a problem in 2.4 too, so it probably doesn't explain why 2.5 is slower than 2.4. Maybe this is a problem with the "deadline I/O scheduler" too. I haven't tested it yet, but it doesn't seem to touch this part of ll_rw_blk.c.

This patch has been running for some time now on my system without problem.

@@ -1121,8 +1121,10 @@
 	switch (el_ret) {
 		case ELEVATOR_BACK_MERGE:
 			BUG_ON(!rq_mergeable(req));
-			if (!q->back_merge_fn(q, req, bio))
+			if (!q->back_merge_fn(q, req, bio)) {
+				insert_here = &req->queuelist;
 				break;
+			}
 
 			elv_merge_cleanup(q, req, nr_sectors);
 

@@ -1135,8 +1137,10 @@
 
 		case ELEVATOR_FRONT_MERGE:
 			BUG_ON(!rq_mergeable(req));
-			if (!q->front_merge_fn(q, req, bio))
+			if (!q->front_merge_fn(q, req, bio)) {
+				insert_here = req->queuelist.prev;
 				break;
+			}
 
 			elv_merge_cleanup(q, req, nr_sectors);
 

Peter Osterlund - [email protected]
http://w1.894.telia.com/~u89404340

Excellent spotting! Applied.

Jens Axboe